See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/317369476

Understanding the Control of Metabolism

Book · January 1997

CITATIONS READS

804 11,634

1 author:

David Andrew Fell

Oxford Brookes University
171 PUBLICATIONS   11,064 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Cell cycle simulation View project

Metabolic control analysis View project

All content following this page was uploaded by David Andrew Fell on 06 June 2017.

The user has requested enhancement of the downloaded file.

Understanding the Control of
Metabolism

David Fell
School of Biological & Molecular Sciences,
Oxford Brookes University

Version: May 1, 2004

Dedication: For my parents, Sidney and Rachel Fell.

Contents

Preface v

1 Introduction: regulation and control 1

1.1 Regulation and control . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Approaches to metabolic regulation . . . . . . . . . . . . . . . . 4
1.2.1 Molecular or reductionist explanations . . . . . . . . . . 5
1.2.2 Qualitative systemic approach . . . . . . . . . . . . . . 6
1.2.3 Quantitative systemic theory . . . . . . . . . . . . . . . 7
1.3 An overview of mechanisms of regulation and control . . . . . . 8
1.4 Basic theory of metabolism . . . . . . . . . . . . . . . . . . . . 10
1.4.1 Metabolic steady states . . . . . . . . . . . . . . . . . . 10
1.4.2 Thermodynamics of metabolic pathways . . . . . . . . . 13
1.4.3 Principles of regulation and control . . . . . . . . . . . . 18
1.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.6 Appendix: free energy changes . . . . . . . . . . . . . . . . . . 21

2 Methods for studying metabolism and its regulation 23

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Experimental systems . . . . . . . . . . . . . . . . . . . . . . . 23
2.2.1 Whole multicellular organisms . . . . . . . . . . . . . . 24
2.2.2 Isolated tissues and organs . . . . . . . . . . . . . . . . 25
2.2.3 Tissue slices . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2.4 Isolated cells . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2.5 Permeabilized cells . . . . . . . . . . . . . . . . . . . . . 29
2.2.6 Cell–free systems . . . . . . . . . . . . . . . . . . . . . . 29
2.2.7 Isolated organelles . . . . . . . . . . . . . . . . . . . . . 29
2.2.8 Isolated enzymes . . . . . . . . . . . . . . . . . . . . . . 30
2.3 Measurement of metabolites . . . . . . . . . . . . . . . . . . . . 30
2.3.1 Fixing concentrations . . . . . . . . . . . . . . . . . . . 30
2.3.2 Low concentrations in complex mixtures . . . . . . . . . 32
2.3.3 Compartmentation . . . . . . . . . . . . . . . . . . . . . 34
2.3.4 Free and bound metabolites . . . . . . . . . . . . . . . . 36
2.4 Measurement of enzyme activity . . . . . . . . . . . . . . . . . 37
2.4.1 Assay conditions . . . . . . . . . . . . . . . . . . . . . . 37

i
Contents

2.4.2 What to measure . . . . . . . . . . . . . . . . . . . . . . 38

2.4.3 Control of other factors affecting in vivo activity . . . . 38
2.5 Measurement of metabolic fluxes . . . . . . . . . . . . . . . . . 39
2.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3 Enzyme activity: the molecular basis for its regulation 47

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 Michaelis–Menten kinetics . . . . . . . . . . . . . . . . . . . . . 47
3.2.1 The measurement of Km and V . . . . . . . . . . . . . . 50
3.2.2 Product inhibition . . . . . . . . . . . . . . . . . . . . . 55
3.3 Two–substrate enzymes . . . . . . . . . . . . . . . . . . . . . . 56
3.3.1 Experimental investigation . . . . . . . . . . . . . . . . 60
3.3.2 Compulsory order mechanism . . . . . . . . . . . . . . . 61
3.3.3 Random order mechanism . . . . . . . . . . . . . . . . . 63
3.3.4 Double displacement mechanism . . . . . . . . . . . . . 64
3.4 Binding characteristics of enzyme sites . . . . . . . . . . . . . . 65
3.4.1 Identical independent binding sites . . . . . . . . . . . . 66
3.4.2 Non–identical independent sites . . . . . . . . . . . . . . 67
3.4.3 Identical interacting sites . . . . . . . . . . . . . . . . . 68
3.5 Allosteric enzymes . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.5.1 The concerted model . . . . . . . . . . . . . . . . . . . . 73
3.5.2 The sequential model . . . . . . . . . . . . . . . . . . . 76
3.5.3 Specific examples . . . . . . . . . . . . . . . . . . . . . . 79
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

4 Traditional approaches to metabolic regulation 87

4.1 The teleological approach . . . . . . . . . . . . . . . . . . . . . 87
4.2 Non–equilibrium enzymes . . . . . . . . . . . . . . . . . . . . . 88
4.3 Isotopic measurement of flux . . . . . . . . . . . . . . . . . . . 90
4.4 Maximal enzyme activities . . . . . . . . . . . . . . . . . . . . . 92
4.5 Addition of intermediates . . . . . . . . . . . . . . . . . . . . . 95
4.6 The cross–over theorem . . . . . . . . . . . . . . . . . . . . . . 95
4.7 Enzyme properties . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.8 Metabolic mutants . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

5 Metabolic control analysis 103

5.1 The problems of the traditional approaches . . . . . . . . . . . 103
5.2 Flux control coefficients . . . . . . . . . . . . . . . . . . . . . . 105
5.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.2.2 Interpretation . . . . . . . . . . . . . . . . . . . . . . . . 107
5.2.3 The summation theorem . . . . . . . . . . . . . . . . . . 111
5.3 Elasticities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.3.1 Definition and examples . . . . . . . . . . . . . . . . . . 114
5.3.2 Use and interpretation . . . . . . . . . . . . . . . . . . . 121

ii
 Contents

5.3.3 The connectivity theorem . . . . . . . . . . . . . . . . . 121

5.4 Response coefficients . . . . . . . . . . . . . . . . . . . . . . . . 124
5.5 Control Analysis and traditional approaches . . . . . . . . . . . 128
5.5.1 Displacement from equilibrium . . . . . . . . . . . . . . 128
5.5.2 Maximal enzyme activities . . . . . . . . . . . . . . . . 129
5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.7 Appendix 1: More about flux control coefficients . . . . . . . . 130
5.8 Appendix 2: Concentration control coefficients . . . . . . . . . 132
5.8.1 Theorems for concentration control coefficients. . . . . . 133
5.9 Appendix 3: More about elasticities . . . . . . . . . . . . . . . 134
5.9.1 Elasticities and enzyme kinetics . . . . . . . . . . . . . . 134
5.9.2 Algebraic evaluation of elasticities . . . . . . . . . . . . 135

6 Measuring control coefficients 139

6.1 Manipulation of enzyme activity . . . . . . . . . . . . . . . . . 140
6.1.1 Altering enzyme activity by genetic means. . . . . . . . 140
6.1.2 Natural alteration of expressed activity . . . . . . . . . 151
6.1.3 Titration with purified enzyme . . . . . . . . . . . . . . 154
6.1.4 Titration of enzymes by specific inhibitors . . . . . . . . 155
6.2 Control coefficients from computer models . . . . . . . . . . . . 168
6.3 Control coefficients from elasticities . . . . . . . . . . . . . . . . 170
6.3.1 An experimental example . . . . . . . . . . . . . . . . . 173
6.3.2 Experimental measurement in vivo by modulation. . . . 176
6.3.3 Experimental measurement in vivo: the ‘top–down’ ap-
proach. . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
6.3.4 Calculation of elasticities. . . . . . . . . . . . . . . . . . 191
6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

7 Control structures in metabolism 201

7.1 Supply and demand . . . . . . . . . . . . . . . . . . . . . . . . 201
7.2 Feedback inhibition . . . . . . . . . . . . . . . . . . . . . . . . . 205
7.2.1 Discovery and relationship to allosteric enzymes . . . . 205
7.2.2 Feedback inhibition and control analysis . . . . . . . . . 206
7.2.3 Patterns of feedback inhibition in branched pathways . 210
7.3 Substrate or ‘futile’ cycles . . . . . . . . . . . . . . . . . . . . . 217
7.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 217
7.3.2 Evidence for substrate cycling . . . . . . . . . . . . . . . 220
7.3.3 Suggested functions . . . . . . . . . . . . . . . . . . . . 223
7.3.4 Thermogenesis . . . . . . . . . . . . . . . . . . . . . . . 224
7.3.5 Sensitivity of control . . . . . . . . . . . . . . . . . . . . 225
7.3.6 Switching the direction of flux . . . . . . . . . . . . . . 227
7.3.7 Buffering metabolite concentrations . . . . . . . . . . . 227
7.4 Regulation by covalent modification of enzymes . . . . . . . . . 229
7.4.1 Irreversible and cyclic cascades . . . . . . . . . . . . . . 230
7.4.2 Types of reversible modification . . . . . . . . . . . . . . 231
7.4.3 Phosphorylation . . . . . . . . . . . . . . . . . . . . . . 232

iii
Contents

7.4.4 Nucleotidylation . . . . . . . . . . . . . . . . . . . . . . 245

7.4.5 Properties of cyclic modification systems . . . . . . . . . 248
7.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
7.6 Appendix 1: Control Analysis of substrate cycles . . . . . . . . 258

8 Conclusion 261
8.1 A new view of the physiological control of flux . . . . . . . . . 263
8.1.1 Evidence for multisite modulation in cells . . . . . . . . 265
8.1.2 Implications of multisite modulation . . . . . . . . . . . 274
8.2 Limitations of Control Analysis . . . . . . . . . . . . . . . . . . 274
8.2.1 Objections to Control Analysis . . . . . . . . . . . . . . 275
8.2.2 Extensions of Control Analysis . . . . . . . . . . . . . . 277
8.3 Applications of Metabolic Control Analysis . . . . . . . . . . . 280
8.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282

References 285

iv
Preface

There is little doubt that the central areas of traditional biochemistry (such as
metabolism, metabolic regulation and enzymology) are currently unfashion-
able and generally regarded as problems long solved. They have been eclipsed
by molecular genetics, which is not only able to identify the genes for protein
components and their regulatory elements, but which also promises the means
to change the genes and their control elements at will, to the point where it
could truly be called biological engineering. In the press and on television,
there are already debates about how we should use, and not abuse, the power
we are expected to posess tomorrow to manipulate even complex multigenic,
and environmentally–influenced, characteristics such as human intelligence
and resistance to disease. How simple it should be, then, to engineer a trait
such as metabolism, where there are such clear and well–understood links
from the genes to the enzymes to the metabolic pathways.
Why then, have I written a book now about metabolism and its regula-
tion which returns to these central, unfashionable concerns of biochemistry
and which only mentions molecular genetics in a subordinate role? The an-
swer is that even 20 years ago, a small number of researchers were convinced
that there were fundamental flaws in the explanations that biochemists gave
for the regulation and control of metabolism. Together, these critics created
Metabolic Control Analysis as an alternative means to understand and ex-
plain these problems, and they have gradually been persuading more and
more biological scientists to design and interpret their experiments in a differ-
ent way. However, their criticisms and reworkings of traditional biochemical
explanations are still not widely known, and have as yet had little influence on
the contents of the standard biochemistry textbooks, which still cling to the
rejected concepts, usually without even mentioning the doubts and problems.
The purpose of my book is therefore to present an introduction to Metabolic
Control Analysis that places it in the context of the biochemical results and
phenomena to which it must be applied. My hope is that this will provide a
more accessible account of the major features of Metabolic Control Analysis
than has previously been available, since most of the existing literature con-
sists of reviews and original articles in research journals, or brief outlines of a
few selected aspects.
My full case for why Metabolic Control Analysis deserves to be more widely
known is contained in the rest of the book, but a representative illustration

v
Preface

comes from genetic engineering itself. There have already been attempts to
manipulate metabolism, and these have involved successfully changing the
amounts or kinetic properties of certain target enzymes. These enzymes have
been selected because traditional biochemical theory, and the textbooks, iden-
tify them as the key steps in the control of a metabolic process. Virtually none
of these experiments has had the expected outcome: the technology exists,
but traditional metabolic biochemistry does not provide the understanding
needed to use it purposefully. Nor, it should be said, is Metabolic Control
Analysis certain to be able to predict the outcome, though it can provide
a framework in which the possibility of achieving a specific modification of
metabolism by a particular genetic intervention can be assessed.
One factor that has undeniably worked against the wide acceptance of
Metabolic Control Analysis is that it inevitably involves more mathematical
concepts and numerical computation than is usual in metabolic biochem-
istry. This is inevitable because it is the essence of the Control Analysis case
against the traditional concepts that they cannot be rigorously justified and
are not adapted to the quantitative testing that would be necessary to discrim-
inate between alternative explanations. Nevertheless, I have tried to limit the
amount of mathematics that is used in this book, and to explain the meaning
and interpretation of the essential equations. Sometimes I have placed addi-
tional mathematical material in boxes or appendices if it is not essential to
the main lines of the argument. I have not, as a result, presented Metabolic
Control Analysis in its fullest or most mathematically rigorous form, nor have
I provided derivations of all its theorems; these can be found elsewhere. How-
ever, it would be a delusion for metabolic biochemists to imagine that the
detailed behaviour of systems as complex as metabolic pathways can be pre-
dicted with the aid of a few qualitative, verbal principles. Other biological
scientists studying comparably complex phenomena, such as population dy-
namics and genetics, or physiology, have long known this, and biochemistry
is remarkable for its limited use of mathematical and theoretical analysis.
The ideas I present in this book draw on the work of a great many re-
searchers and are only partly my own. Even the ideas I would claim as my
own owe much to the influences of other scientists and colleagues, some of
whom I wish specifically to identify.
Eric Newsholme, first stimulated my interest in the control of metabolism,
with his excellent lectures and tutorials at the University of Oxford, even
though I have not followed his approach nor, in the end, agreed with his
conclusions.
Henrik Kacser’s article The control of flux 119 (with Jim Burns) made me
realize that a more rational approach to metabolic regulation was possible.
Later, when I first stated to publish my results in Metabolic Control Analysis,
Henrik’s encouragement gave me confidence to continue. His death in March
1995 whilst still actively involved in research on Metabolic Control Analysis,
even though officially in retirement, was a great loss to the field of Metabolic
Control Analysis of which he was a cofounder.
Herbert Sauro’s unsuppressible curiosity when he was my research student

vi
 Preface

led us to make our first discoveries in the theory of Metabolic Control Analysis.
He also then began the development of the metabolic computer simulation
package SCAMP, which has continued to be a useful tool in my research
group ever since and which has contributed some of the simulations reported
in this book.
My next research student, Rankin Small, helped continue the momentum
of the work begun with Herbert and began the development of the Control
Analysis of covalent modification systems described in Chapter 7. Some of
his subsequent work with Henrik Kacser influenced the ideas I put forward in
the final chapter.
More recently, Simon Thomas has worked with me, first as a research
student and then as a Leverhulme Special Research Fellow, on some of the
ideas presented in the final two chapters and he drew my attention to some
of the examples I have used there.
The elasticity calculus presented in the Appendix to Chapter 5 has been
developed further and applied in computer programs by Simon Thomas, John
Woods and Herbert Sauro.
A number of colleagues helped me by supplying various materials for the
book. Kevin Brindle provided me with an original copy of Fig. 2.3, which
appeared first as Fig. 1 in Biochim. Biophys. Acta, 847, 285–292 (1985) and
which is reproduced with permission of the publishers. Jean–Pierre Mazat
and Thierry Letellier provided me with the data and fitting functions for their
inhibitor titrations in Fig. 6.16, but also, more importantly, much–appreciated
hospitality in Bordeaux on several occasions for discussions about Control
Analysis. Christoph Giersch sent me a manuscript before publication and the
data that appears in Figs. 6.18 and 6.19; we also had useful discussions about
the related text.
I am grateful to all those who have commented on various parts of the
book. Keith Snell, as editor of this series, read the whole text and made
useful suggestions about presentation. Athel Cornish–Bowden corrected my
errors in Chapter 3 on enzyme kinetics and also had helpful ideas about the
presentation. Simon Thomas read much of the draft and helped to improve
the text in several places. Others who have commented on the text include
Jofao Pedro Moniz Barreto, Stefan Schuster, John Woods and my wife, Mary.
Many other friends and colleagues have made an indirect contribution to
this book through conversations, discussions and even critical scrutiny of my
ideas, particularly at conferences. I cannot list them all, but three who have
had a direct influence on material in the book are Athel Cornish–Bowden,
Jannie Hofmeyr and Mark Stitt. Athel and Jannie have provided me with
advance copies of several of their papers that relate particularly to material
in the first chapter and the last two. Mark drew my attention to some of the
illustrative material I have used from plant biochemistry.
I am also grateful to the undergraduates of Oxford Brookes University who
were the unwitting guinea pigs whilst I rehearsed the presentation of parts of
this book.
In addition to those mentioned previously, the following authors and pub-

vii
Preface

lishers have kindly given me permission to use their diagrams in this book:

• J. Katz et al. , J. Biol. Chem. 253, 4530–4536, (1977) for Fig. 2.4;

• E. Chance and the Journal of Biological Chemistry 36 for Fig. 2.5;

• A. B. Tulp, Trends in Biochemical Sciences, 11, 13 (1986) for Fig. 4.6,

and

• Bert Groen for Fig. 6.14 from his PhD thesis.89

I am also greatly indebted to the authors of the excellent public domain

software that I have found so useful in preparing this book. Firstly, this
manuscript was prepared using the LATEXdocument preparation system by L.
Lamport, itself a macro package for TEX by D. E. Knuth. Many other authors
have contributed the parts that make this a working system, from the various
TEXdistributions on the different computers I have used and the auxiliary
programs for making the bibliography, indexing, drawing diagrams and view-
ing the output (BIBTEX, MakeIndex, TeXcad and the DVI processors) to the
additional packages for citation, drawing (PICTEX) and inclusion of Postscript
figures. Many of the graphs of data and functions were plotted with GnuPlot
from the Free Software Foundation and viewed with GhostScript. Most of the
text editting was done with various versions of MicroEmacs. I have received
much help with finding, implementing and using this software from Herbert
Sauro, John Woods and Jof ao Pedro Moniz Barreto.

David Fell
Oxford
January 1996

viii
 1

Introduction: regulation and

control

1.1 Regulation and control

Much of the detail of metabolism, represented by the interconversions of
metabolites catalyzed by enzymes, is known, and can be found in biochem-
istry texts and on metabolic maps. But biochemistry is a relatively young
science and this detail has been discovered recently, mostly within the last
seventy years. Our knowledge of this and other aspects of biochemistry and
molecular biology reflects the success of the reductionist approach: the study
of a whole system by detailed examination of the properties of its constituent
parts, in the expectation that the behaviour of the system can be predicted
once the properties of the parts are understood. In metabolism, the reduc-
tionist strategy has started with the discovery of the component pathways
followed by that of the separate reactions in the pathways, and then passed
to characterization and detailed study of the enzymes responsible for the re-
actions, and ultimately even to the identification of the genes that encode the
enzymes. Knowing the routes and the molecules involved does not by itself,
however, lead to an understanding of what determines the material flows in
different pathways, nor of how the production and utilization of metabolites
are kept in balance. The focus of metabolic research has therefore shifted to
tackle these problems of regulation and control, though a number of different
approaches have been adopted , as will be shown later.
Firstly though, why is the study of regulation and control of metabolism
important? The reasons include:

• Gaining a more complete understanding of metabolism, under both nor-

mal and abnormal conditions. The former might be represented by
healthy organisms in favourable conditions, and the latter by organisms
affected by mutation, disease or hostile conditions. I will argue later

1
2 1 Introduction: regulation and control

that knowing the molecular mechanisms and interactions involved in

regulation and control is not enough; we must be able to show in a con-
vincing way how these lead to the observed metabolic responses of the
organism.

• Understanding the actions of hormones and drugs or other chemical

agents that affect metabolism, for which it is necessary to understand
how metabolism responds to various types of perturbation of enzyme
activities.

• Aiding the search for drugs that affect metabolism by identifying ap-
propriate target sites for bringing about the required change.

• Enabling biotechnologists to increase formation of the desired metabolic

products, and decrease formation of unwanted secondary metabolites,
in industrial processes for producing biological materials such as drugs,
antibodies, enzymes or nutrients.

Indeed, our lack of understanding of metabolic regulation has been revealed

by poor results from attempts to increase the rates of selected metabolic path-
ways. Genetic engineering techniques have advanced to an amazing extent,
so it is now possible to change the activities of enzymes at will, but when
this has been applied to the enzymes stated by biochemistry textbooks to
be rate–controlling, there has often been no significant corresponding change
in metabolism. Our ability to interfere with an organism’s genetics has far
outstripped our ability to predict the effects on its metabolism.
One of the difficulties in discussing regulation and control is that, in biol-
ogy, there often appears to be no consistent distinction between, the terms.
Very often they are used interchangeably. If we try to be more specific about
the terms, and to make our usage consistent with other areas of science and
technology, such as engineering, we find a fundamental difficulty in drawing
analogies between control systems in engineering and biology: in engineer-
ing we know the purpose the mechanism serves, but in biology we can only
presume the function as a hypothesis, and we cannot be certain that there
is not a range of other requirements that are be met by it. For example, we
know why there is a thermostat in a refrigerator, whereas we do not know
why the regulation of glycogen metabolism has to involve all the molecular
components that it does (see Chapter 7.4.3.3, p. 243). Nevertheless, it is
important to discriminate between the terms regulation and control.
Regulation, both in technological and biological systems, is occurring when
a system maintains some variable (e.g. temperature or concentration) con-
stant over time, in spite of fluctuations in external conditions. (See, for ex-
ample, the McGraw–Hill Encyclopaedia of Science & Technology (1982) and
the Longman Dictionary of Scientific Usage (1979).) Regulation, in the sense
of minimizing the effects of disturbances, is therefore linked to homoeostasis,
the maintenance of a relatively constant internal state. For example, Jannie
Hofmeyr and Athel Cornish–Bowden106 recently defined metabolic regulation
1.2 Approaches to metabolic regulation 3

in terms of the performance of the metabolic system rather than the existence
of particular mechanisms in the system. They stated that the effectiveness
of metabolic regulation should be assessed from the way a metabolic system
responds to environmental changes. For example, metabolic regulation could
be a sensitive response of the rate of a metabolic pathway to environmental
conditions whilst metabolite concentrations are kept near constant. In other
circumstances, the notable feature of the performance might be the mainte-
nance of near-constant metabolic rates (for example, of ATP synthesis) in the
face of fluctuations in the environment. This underlines the problem that it
might not be completely clear what aspect of a metabolic system is being
regulated; this might be assumed or presumed, whereas in a technological
system, this aspect of function would be an explicit feature of the design.

Control is used in technology to refer to adjusting the output of a system

with time, perhaps to match some required profile (e.g. a time–varying input
signal) or to obtain a desired response. (See Chambers’ Science & Technology
Dictionary (1974) and Longman’s Dictionary of Scientific Usage (1979).) In
common usage, control as a verb implies the ability to start, stop, direct or
adjust something. Again these meanings can often not be applied literally
to biological systems, because to do so would imply there is an identifiable
and unique characteristic of the system under consideration, whereas we know
that many metabolic pathways are multifunctional and that the ultimate re-
quirement is that they are controlled in such a way as to ensure the fitness
(in an evolutionary sense) of a species in its ecological niche. However, we
can regard metabolic control as the power to change the state of metabolism
in response to an external signal. Hofmeyr & Cornish–Bowden106 similarly
regard control as being measurable in terms of the degree of influence that an
external factor has on the state of a metabolic system. The advantage of this
definition is that metabolic control can be assessed in terms of the strength
of any of the responses to the external factor without making any assumption
about the function or purpose of that response. It also means control is a
simpler, lower–level concept than regulation, because it does not require a
judgement about the function of the system.

To complicate matters further, whether regulation or control is the ap-

propriate term depends as much on the context and level of study as on the
system under consideration. For example, vertebrate blood glucose is regu-
lated by insulin and glucagon (homoeostasis), but these hormones achieve this
by controlling the metabolism of various body tissues. In this example, we
could say that the concentration of a metabolite is kept constant (regulated)
by varying (controlling) the rates of metabolism, if we wanted to emphasize
the homoeostatic and active aspects respectively of the two terms. Hofmeyr
and Cornish–Bowden’s definitions imply that we refer to the regulation of the
blood glucose because it is the behaviour of a complete system, underlying
which is the control that the hormones exert on specific components of the
system, particularly the rates of metabolic pathways in various tissues.
4 1 Introduction: regulation and control

1.2 Approaches to metabolic regulation

The study of regulation and control draws on a range of different approaches;
the challenge is to combine these into a logically consistent whole. My aim
in this book is to present the basics of these approaches and to suggest that
Metabolic Control Analysis offers one way of achieving a consistent overview.
What are the different strands that have to be combined? They are the
views that result from studying metabolic biochemistry either at relatively
high levels of organization and integration, the system level, or at much lower
levels of organization, effectively at the molecular level.
Study at the system level typically involves whole pathways, organelles,
cells or even organs. The topics belonging to the system level within this book
are:

• Methods of studying metabolism.

– Measuring metabolites
– Measuring metabolic fluxes
– Measuring enzyme contents

• Methods of identifying supposed ‘rate–limiting steps’.

• Metabolic Control Analysis: replacing the erroneous concept of the rate–

limiting step.

(Note that I have used metabolic fluxes rather than metabolic rates. Although
the two expressions mean essentially the same, it is conventional in Metabolic
Control Analysis to use flux to refer to a rate in a multicomponent system
such as a metabolic pathway and to reserve rate for individual component
steps, such as enzymes. I will maintain this distinction in the rest of this
book.)
At the molecular level, the properties of individual molecular components
of metabolic systems are characterized. Examples from later in the book
include:

• Enzyme kinetics.

– single substrate enzymes

– two–substrate enzymes
– allosteric enzymes

• Feedback inhibition.

• Covalent modification of enzymes.

– irreversible (degradation or activation)

– reversible (phosphorylation etc.)
1.2 Approaches to metabolic regulation 5

Researchers in the areas I have just crudely allocated into these two groups
may resent my implication that, by studying a topic from one of them, they
are distanced from biochemistry at the other level. They might point out
that system–level studies have established that particular pathways have some
means of responding to certain physiological or environmental stimuli, and
that molecular–level studies have identified molecules that relay the signals
and the enzymes on which they act. However, I still believe that to build
a bridge between these two levels, to ensure that the knowledge and under-
standing we have at each can be combined as a consistent whole, is extremely
difficult and has not received sufficient attention until recently. Even though
the discovery of control molecules and signals has required work at the two
levels (e.g. Chapter 7.4), I do not think I am proved wrong because there have
been many unresolved disagreements about the physiological significance and
role of many of the molecular events that have been discovered to have the
potential for regulation and control. The relative lack of attention given to
relating the two levels, and to developing an adequate theory able to resolve
disputes, probably reflects the optimistic tendency in science to exaggerate
the extent of our islands of knowledge and to stress the importance of work at
their margins, whilst minimizing the significance of the relatively uncharted
seas of ignorance that separate them. To explain how this applies to metabolic
regulation, I will consider in the following sections three types of approach
to explanation of metabolic control and regulation. Although the aim of the
book is to promote the advantages of the third type of approach, examples
of the other two will be found, sometimes to show how current thinking has
evolved from previous ideas, sometimes to criticize ideas that are still current,
and sometimes because the third type of approach is incomplete and does not
yet have a view to offer.

1.2.1 Molecular or reductionist explanations

The justification for the reductionist approach is that, once the properties of
the component parts are known, it will be possible to explain the behaviour
of the whole system. For the biochemist, the system might be a particular
pathway, or the whole metabolism of a particular cell type, and what must
be explained includes why there is a certain rate of flow through a pathway
under some defined set of external conditions. Now, a metabolic map does
not supply this sort of answer, any more than a town map reveals how much
traffic can flow through the streets. Even if the information available includes
the catalytic activity of the enzymes in the cell, and how the activities respond
to signals from the concentrations of metabolites, the answer is still elusive,
in the same way that the traffic flow problem is still a problem even if the
map shows the relative widths of the streets and where the traffic signals are.
How feasible is it to construct an explanation of metabolism in terms of
the properties of the parts? There can be thousands of enzymes and metabo-
lites in a cell; if every enzyme interacted with all the metabolites, the number
of separate interactions to be investigated would be in the millions. Fortu-
6 1 Introduction: regulation and control

nately, most of these interactions do not occur, and the most important ones
are obvious: first of all, an enzyme must interact with its substrates and
products, and then perhaps with certain other metabolites from its own path-
way. Even so, for an enzyme such as glutamine synthase, which has three
substrates, three products and nine significant effectors, the problem of ob-
taining an accurate description of its behaviour for all concentrations of these
fifteen metabolites is huge. Michael Savageau217 pointed out that it would be
necessary to make 415 or ≈ 109 measurements in order to investigate all the
possible interactions, even if only four different concentrations of each of the
fifteen metabolites would be needed to characterize its response. Obviously,
it is unlikely that complete descriptions of complex enzymes will ever exist,
though the principal characteristics of many of the common enzymes, such
as those in glycolysis and the tricarboxylic acid cycle, are known for several
different cell types. Indeed, there have been attempts to build this infor-
mation into computer models to simulate metabolism, particularly by David
Garfinkel and his group working on the carbohydrate catabolism of mam-
malian heart muscle,78 and by Barbara Wright and her co–workers on the
carbohydrate metabolism of slime moulds2, 278 (Chapter 6.2). Whilst these
models have had some success, in order to make their behaviour realistic it
is usually necessary to change some of the values for some of the enzymes’
properties, since the experimental values seem unsuitable, often for no very
clear reason.279
In terms of my previous metaphor, reductionists are not worried about a
sea of ignorance between the molecular level and system level of explanation,
because they believe that in principle it will eventually be crossed with the
boat of mathematical modelling and computer simulation, even though the
record of successful crossings is not very good. It seems we are still a long way
from being able to simulate metabolism easily and routinely on the basis of
our present biochemical knowledge. If we cannot do this, can we be justified in
claiming that our knowledge of the molecular properties explains the system–
level behaviour in biochemistry? Up to a point we probably can, because
the types of behaviour observed with real metabolic pathways are seen in
computer simulations of metabolic systems composed of model enzymes with
properties typical of real enzymes, even though we cannot easily reproduce
the exact behaviour of a specific metabolic pathway.

1.2.2 Qualitative systemic approach

In this approach, the information about the underlying molecular structure
(e.g. existence of a feedback inhibition on an enzyme) is used in general terms
only. That is, there is a qualitative, or at best semi–quantitative decision, as
to whether a particular interaction is large enough to have a significant effect
at the system level or not. A verbal description is then built up of how the
enzymes and metabolites interact to give the system properties of the whole
pathway, perhaps drawing on analogies with other disciplines such as engi-
neering. The verbal model might be used to predict responses of the system
1.2 Approaches to metabolic regulation 7

to external events, but exact comparison with experimental observations is

difficult because these predictions are qualitative rather than precise. When
different groups have produced different verbal models for the same phenom-
ena, it has proved difficult to resolve the dispute in favour of one or the other
because model performance cannot easily be objectively assessed. Neverthe-
less, such qualitative verbal models have been dominant in biochemistry as
a whole and in metabolic regulation in particular. This is understandable in
a young science studying complex little–known systems; it offers a starting
point for the generation of models of the system behaviour that can be re-
fined and tested. What is disturbing is the apparent lack of any ambition
to progress to more rigorous specifications of the models so that quantitative
predictions emerge that can be compared with experimental results.
In terms of my metaphor, the qualitative systems approach tries to avoid
falling into the sea of ignorance by never walking close enough to the water’s
edge to be able to see it clearly, and therefore does not know how far the sea
extends. For each little island of knowledge, there are lots of travel guides
that tell you what you might see there, but little in the way of maps to show
how the different features are connected together.

1.2.3 Quantitative systemic theory

A quantitative systems approach involves using information about the general
features of the underlying molecular structure to build a formal (mathemati-
cal) description of the system properties, without attempting to incorporate
a fully detailed model of all the behaviour at a molecular level. Instead, one
of several theoretical schemes is used to build in a general description of the
behaviour of metabolic systems. If this is combined with the measurement or
calculation of some selected system parameters, the description can be made
quantitative and compared with experimental observations. Generally, the
quantitative description is best close to a measured experimental state, and
the larger the movement away from this state, the less exact the predictions
become because the model is not sufficiently detailed to give an exact repre-
sentation over a large range of conditions. The advantage of forgoing a precise
model of the system is that the amount of information needed for these sys-
tems level models is much smaller than for complete molecular models, so it
is more feasible to collect it experimentally. Even though the models are not
highly precise, they are quantitative so it ought to be easier to decide whether
different models give significantly different predictions and whether they are
consistent with experimental observations.
Thus it is accepted that there will be a degree of ignorance about whether
the molecular properties account for the system behaviour, but that does not
mean that there can be no link between the two levels of description and it
is also possible to have a quantitative model. In this case, we are saying that
we do not expect to have the islands of knowledge that correspond to our
system components mapped internally in very great detail; however, more
importance is attached to determining the position of each of these islands on
8 1 Introduction: regulation and control

the world map, so that we have a good impression of the global geography
even though we accept that there is a lack of very specific detail.

1.3 An overview of mechanisms of regulation and

control
One notable feature of metabolism is the very wide range of time scales over
which changes can occur. In some cases, adjustments can happen within sec-
onds or less, whereas in others they may be linked to seasonal changes or
the life cycle of the organism. Underlying these events can be mechanistic
steps, such as the binding of a hormone by a receptor, that take tiny fractions
of a second, or the breakdown of unwanted enzyme occurring over hours or
days. Not surprisingly, it is difficult to consider such disparate events simul-
taneously. So both theoretically and experimentally, it is usual to choose a
time span to work on, and to try to arrange matters so that changes requir-
ing longer periods of adjustment can be ignored and changes taking place
on shorter time scales can be regarded as having reached completion almost
immediately. For example, the laboratory measurement of an enzyme activ-
ity is usually performed so that the initial events of substrate binding to the
enzyme on mixing are essentially instantaneous and the measurement is com-
pleted before the inevitable long–term depletion of substrate occurs to any
significant extent.
In the regulation and control of metabolism, it is possible to group the
operative mechanisms into three major classes on the basis of the relative
lengths of time they take to bring about adjustments. These are:

• Long time scales, i.e. of the order of hours or days in eukaryotes,

though only a few minutes in prokaryotes. The mechanisms on these
time scales typically involve the control of amounts of enzyme protein
by:

– the rate and degree of gene expression.

– the rate of enzyme degradation.

Mechanisms for the first of these are known in great molecular detail in a
number of cases, beginning with the work of Jacob and Monod on the control
of the lac operon of E. coli. Much less is known about the mechanisms of the
second, though it is known that in mammals, for example, some enzymes have
lifetimes of only a few minutes, whilst other enzymes in the same cells can
last for days. Proteolytic breakdown of enzymes can be selectively targeted,
for example with the ubiquitin system,102 and can be initiated in response to
specific stimuli. The dynamic range of these mechanisms can be very large;
that is, the amount of enzyme protein in a cell can be varied between nothing
and the level of the major metabolic enzymes such as those of glycolysis. This
book will not deal with the molecular details of these mechanisms, examples
1.3 An overview of mechanisms of regulation and control 9

of which are commonly given in many textbooks. However, the effects of

the variations in enzyme amount that are caused in such ways have not been
studied so systematically, and this will be considered later in the book. For
example, it was over 20 years after Jacob and Monod’s proposal of the lac
operon before the contributions of the enzymes of the operon to the growth
of the bacteria on galactose were assessed in detail (see Chapter 6.1.1.2 and
Chapter 6.1.2.).

• Medium time scales, i.e. of the order of minutes or a few seconds. A

major mechanism for changes on this scale is the cyclic activation and
deactivation of existing enzymes by covalent modification of the protein.

The Nobel Laureates E G Krebs and E Fischer first demonstrated this in mam-
malian muscle for the activation of glycogen phosphorylase by phosphoryla-
tion (Chapter 7.4.1); since then, many eukaryotic enzymes have been found
to be affected by phosphorylation and dephosphorylation, and more recently
the process has been discovered in bacteria. There are also other types of
covalent modification apart from phosphorylation (see Chapter 7.4.2, p. 231).
In animals, the modification reactions may respond to hormonal or nervous
signals, and are in fact the means whereby these signals can have relatively
rapid effects on metabolism. The potential degree of change from the lowest
enzymic activity to the highest (the dynamic range) is very large with this
mechanism. This is because the inactive form of the enzyme can be almost
completely devoid of activity, and virtually all the enzyme can potentially be
converted to this state; on the other hand, when activation occurs, there can
be essentially complete conversion of all the enzyme present in the cell to the
active form. Examples of cyclic covalent modification schemes are considered
later in this book (Chapter 7.4.3 & 7.4.4), together with an examination of
the types of behaviour they can exhibit (Chapter 7.4.5).

• Short time scales i.e. seconds or less. The mechanisms on this time
scale are presumed to be the reversible binding of metabolites to en-
zymes, causing effects such as allosteric inhibition and activation (Chap-
ter 3.5, p. 71).

The forces holding metabolites to enzymes are usually non–covalent and there-
fore relatively weak, but as the links can easily break and reform, the equilibria
are established extremely rapidly. On the other hand, the dynamic range of
these effects on enzyme activity is limited, since inhibition and activation
involving non–covalent binding are rarely 100% effective. The maximum acti-
vation is also limited by the amount of enzyme available, but it is not possible
to have a large reserve of enzyme to compensate for this partial activation
because then the inhibition effects would not be strong enough to slow the
enzyme when it is not needed. These effects are considered in several places
in this book; in the chapter on enzyme kinetics (Chapter 3.5) the molecular
mechanisms of the effects are considered, whereas in Chapter 7.2 examples
are given of the ways in which they implement feedback control systems.
10 1 Introduction: regulation and control

1- 2- 3- -
Glucose Glc6P Fru6P Ethanol + CO2

Fig. 1.1: A metabolic pathway.

This is part of the catabolic pathway from the source, glucose, available from the medium
to the sink, the ethanol plus CO2 that are released to the surroundings. At steady state,
the rate of formation of Glc6P in the cell by reaction 1 is equal to its consumption
in reaction 2 (assuming there are no other significant uses of Glc6P). Thus the rates
of reactions 1 and 2 are the same. Similarly, since the rate of formation of Fru6P by
reaction 2 is the same as its consumption by reaction 3, the latter also works at the same
rate as reaction 1. The steady state in metabolite concentrations is therefore equivalent
to a constant rate through the whole pathway.

The basis on which it is legitimate to concentrate on one of these time

scales and to ignore the events and mechanisms on longer time scales is ex-
plored in the next section.

1.4 Basic theory of metabolism

1.4.1 Metabolic steady states
One of the characteristic features of living organisms is their ability to main-
tain a relatively constant composition whilst continually taking in nutrients
from the environment and returning excretory products. These two aspects
are necessarily linked: organisms can only keep their internal state constant
by this flux of matter and energy through the metabolic pathways of their
cells. This is termed dynamic equilibrium to distinguish it from chemical
equilibrium, in which a constant state is reached when the net reaction fluxes
die away. The concept of the steady state corresponds to a perfect dynamic
equilibrium; we suppose that a metabolic pathway starts with a source of
material that is derived at constant concentration from the environment (di-
rectly or indirectly) and finishes with an end product, the sink, that is kept
at constant concentration by direct or indirect excretion into to the environ-
ment. In most cases, this will lead to the development of a steady state, where
the concentrations of the intermediates remain constant because their rates
of formation have come to be in exact balance with their rates of degrada-
1.4 Basic theory of metabolism 11

tion (Fig. 1.1). This also requires that the flow through the pathway remains
constant for the reasons explained in the figure legend. It is not inevitable
that any sequence of successive enzymic reactions will always be able to reach
a steady state in these circumstances, but the consequence of not reaching
a steady state would be that one or more intermediate metabolites would
continue to accumulate in ever increasing amounts. Such an increase would
start to give a cell severe osmotic problems, and would be at odds with the
general impression of dynamic equilibrium. Indeed, biochemists must have an
expectation that a metabolic system tends to a reproducible state, since sim-
ilar experimental results could only be obtained from day to day in this case.
It would not generally be possible with non–steady states, since the amounts
of metabolites would depend very sensitively on the previous history of the
sample to such an extent that it would be difficult to repeat any observation.
(An exception to this would be the metabolic oscillations that are sometimes
observed reproducibly; however, although these show regular fluctuations in
concentrations and fluxes, their average values are constant.)
Nevertheless, an exact steady state is a mathematical abstraction. One
reason for this is that as a pathway comes closer and closer to steady state, its
rate of approach becomes ever slower, so that in theory it will only arrive there
after an infinite amount of time. In practice this does not matter, because the
limited accuracy of methods of biochemical measurement mean that a system
that is more than 95% of the way to steady state is virtually indistinguishable
from the steady state itself, and 95% attainment of the steady state could
well occur in the preincubation phase of a metabolic experiment.
Even so, there is another limitation to the concept of a steady state: the
environment, whether inside some laboratory glassware or in nature itself,
is never completely constant, not least because of the action of metabolism.
It is inevitable that the source materials of metabolic pathways will slowly
deplete, and that excretory products will accumulate. Also, as mentioned in
the previous section, there are other events that come into effect on longer
time scales, such as induction or repression of enzyme synthesis. Another
possibility is that minor metabolic pathways, such as those for the synthesis
of adenine or nicotinamide nucleotides, cause slow changes in the amounts
of coenzymes like ATP and NAD+ . If there are very slow changes in the
concentrations of metabolites or the pathway flux caused by any of these
factors, the pathway may still be regarded as being in quasi steady state
provided the time scale of the changes is very much longer than the time
taken by the pathway to approach steady state. In other words, if there
are slow changes in one or more metabolites, this must mean that the rates
of synthesis and degradation are not quite in balance, but if the degree of
imbalance is only a small fraction of the rate of synthesis, then it may still be
a good approximation to regard it as being at steady state. Fig. 1.2 illustrates
the way in which the quasi steady state is relative to the time scale of the
experimental observations. In the chapter on enzyme kinetics (Chapter 3),
it is shown that exactly the same concept of quasi steady state is routinely
accepted as a basis for interpreting experiments.
12 1 Introduction: regulation and control

.....Enzyme
.........................synthesis
............................

Enzyme binds Metabolite Covalent enzyme

ligand ..
....................................... ..........crosses
......................cell
..............................modification
...............................................

Enzyme turns Metabolic pathway

reaches steady state
......over
. ................
...............once . .. ..
.........................................................

10-4 10-3 10-2 10-1 1 101 102 103 104

seconds

A B C D

..................................................................
....................... . . . . . . . . . ........................
... . . . . . . . ..... . .. . .. . .. . .. . .. . .. . .. . .. . .. . .. . .. .
. . . Fast processes. . . . . . . . Slow processes . . . .
... ..... Processes ...
. . . at dynamic . . . . . under study . . . . . .ignored ....
. . . equilibrium . . . . . . . . . . . . . . . . . . . . . .. . ..
....................... . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . .. . .. . .. . .. . .. . .. . .. . .. . .. . .. . ..
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Experimental viewing frame

Fig. 1.2: Metabolic time scales and the quasi steady state.
For different types of experiment, the experimental viewing frame is placed over the
time axis in different positions, masking off the regions that are not being observed. For
example, a fast reaction enzyme kineticist would centre the window of the frame above
the letter A, a metabolic biochemist working on isolated cells or tissues might center
it in the range B to C, and a biochemist working on whole animals, or a nutritionist,
might centre it above D. The time scales of various biochemical events are indicated as
very approximate ranges.
1.4 Basic theory of metabolism 13

1.4.2 Thermodynamics of metabolic pathways

The study of the regulation and control of metabolic pathways is concerned
with the factors that affect the rates of the pathways, so at first sight it might
seem that thermodynamics can offer little insight. After all, in spite of its
name, classical thermodynamics gives no information about the rates of pro-
cesses; even worse, many of its formulae are only precise about systems that
are at equilibrium or changing infinitely slowly. Nevertheless, thermodynam-
ics as the science of energy relationships reveals the energy constraints to
which pathways are subject and which, independently of the details of the
enzyme kinetics, place some limits on possible steady states of a pathway.
A more recent branch of thermodynamics, nonequilibrium thermodynamics
(developed by the Belgian Nobel Laureate I. Prigogine in particular), does
make explicit links between the rates of processes and their energetics. Fur-
thermore, it is not enough for mechanisms of regulation and control just to
bring about the necessary steady state fluxes through the pathways. Other
constraints have to be satisfied as well, and these can be directly linked to
the energetics of the processes. For example, Dan E. Atkinson6, 7 pointed out
that the concentrations of most intermediary metabolites must be kept low,
for otherwise, with thousands of different metabolites in a cell, the total vol-
ume they would occupy and their demand for water of solvation would be far
greater than space available in the cell. In addition, he argued that the low
concentrations of metabolites ensure that non–enzymic side reactions between
metabolites are minimized.
Classical thermodynamics deals principally with closed systems, defined
as systems that can exchange energy, but not matter with their surround-
ings. The only steady state that can develop under these circumstances is
chemical equilibrium, when all net chemical change will have ceased. Living
organisms, in contrast, are open systems that exchange energy and matter
with their surroundings. This crucial difference allows the development of
a dynamic equilibrium as a stable displacement from chemical equilibrium.
(Non–equilibrium stable states are explained by nonequilibrium thermody-
namics.)
According to classical thermodynamics, the change in free energy ∆G for
a reaction at equilibrium is zero, but is non–zero if the reaction is displaced
from equilibrium. (The Appendix to this chapter explains why the symbol
∆G is inappropriate for the concept of free energy, but it is retained here to
be consistent with general usage.) If the reaction is taking place of its own
accord (i.e. spontaneously, with no input of work energy such as an electric
current) then it will have a negative free energy change and will be going in the
direction towards equilibrium. Incidentally, nonequilibrium thermodynamics
can show that metabolic pathways at dynamic equilibrium are indeed systems
of spontaneous reactions. Where there is a net flow through a metabolic
pathway, there must be a negative free energy change at each reaction step,
every one of which must therefore be displaced from equilibrium. However,
whereas successive reactions in an unbranched metabolic pathway must be
14 1 Introduction: regulation and control

going at the same rate in a steady state, there is no requirement that they
all have the same free energy change, since this has no direct link with the
kinetics. In fact, it is a matter of observation (see Chapter 4.2, p. 88) that
the degrees of displacement from equilibrium of the reactions of a pathway
are extremely variable. Before we consider the significance of this, let us first
define the degree of displacement more precisely.
Consider the reaction:

A + B*
)C + D
The actual free energy of this reaction for a given set of concentrations of the
substrates and products is ∆G0 (where the prime indicates the (biological)
standard condition of pH 7), related to the standard free energy ∆G◦0 by the
expression:
CD
∆G0 = ∆G◦0 + RT ln (1.1)
AB
where R is the gas constant, T the absolute temperature and ln is the natural
logarithm. The ratio of product concentrations to substrate concentrations is
defined as the mass action ratio, Γ, i.e. :

CD
Γ= (1.2)
AB

At equilibrium, Γ = Keq , and ∆G0 = 0. Hence:

0 = ∆G◦0 + RT ln Keq

or
∆G◦0 = −RT ln Keq (1.3)

Substituting eqns. [1.2] and [1.3] into Eqn. [1.1] gives:

∆G0 = −RT ln Keq + RT ln Γ

which rearranges as:

µ ¶
Γ
∆G0 = RT ln (1.4)
Keq

The disequilibrium ratio, ρ, defined as Γ/Keq gives a measure of the displace-

ment of a reaction from equilibrium. As Eqn. [1.4] shows, it is 1 at equilibrium
when ∆G0 = 0, and is always less than 1 when ∆G0 is negative, becoming
small for large negative free energy changes. (This is because for logarithms
in any base, ln 1 = 0, and the logarithm of any number between 1 and 0 is
negative, approaching −∞ as the numbers approach closer to 0.)
For reasons to be considered below and in the next section, the value of
ρ, the disequilibrium ratio for a reaction is thought to determine whether it
is a potential site for regulation and control. Accordingly, it is common to
1.4 Basic theory of metabolism 15

find that reactions are classified as either non–equilibrium or near–equilibrium

reactions. The former are reactions for which ρ is very small, and the latter
are those for which it is close to 1. Unfortunately there is no strong theoretical
basis for deciding the demarcation point between these two classes, as will be
discussed later (Chapter 4.2, p. 88, and Chapter 5.5.1, p. 128). However, there
are many reactions where the classification is not disputed, and generally the
non–equilibrium reactions are those which have a large negative standard free
energy change, corresponding to a large equilibrium constant.
On the other hand, the undisputed near–equilibrium reactions tend to
be those that have a small standard free energy change, corresponding to
an equilibrium constant close to 1. This is not a surprising finding, for a
reason that has little to do with regulation and control. Most metabolite
concentrations are in the micromolar to millimolar (10−6 M to 10−3 M) range.
If concentrations were above this range, cells would have problems with the
osmotic strength of the cytoplasm and the water requirement for solvating all
the metabolites. In addition, higher metabolite levels would slow down the
response rates of metabolism, because the large pools of metabolites would
buffer the impact of sudden changes in enzymic activity caused by a control
signal (an effect analysed in more detail by John Easterby in his studies of the
transition times of metabolic pathways62 ). In contrast, if the concentrations
were very low, enzymes acting on the metabolites would have to have a very
high affinity for them, and whilst enzymes could no doubt evolve to have
much higher affinities for their substrates, there are theoretical grounds for
believing that this is incompatible with the development of high catalytic
efficiency. Therefore it is likely that metabolites are on the whole constrained
to be in the micromolar to millimolar region.
If this is so, then reactions with large equilibrium constants will have
to be non–equilibrium reactions. For example, the first three reactions of
glycolysis (Fig. 1.1) have equilibrium constants of approximately 4000, 0.4
and 1000 respectively, giving an overall equilibrium constant for the three
reactions of 1.6 × 106 . Given the concentrations of glucose, ATP and ADP in
the human red blood cell, the fructose–1,6–bisphosphate concentration would
have to be more than a million molar for these three reactions to approach
near to equilibrium. In the case of reactions with small standard free energy
changes, corresponding to equilibrium constants close to 1, a similar argument
suggests that they have to be near to equilibrium. If such a reaction were non–
equilibrium, this would mean that its products would be present in much
lower concentrations than the substrates, and this effect would multiply if
there were successive reactions with small standard free energy changes, each
displaced far from equilibrium so that the metabolite concentrations would
become progressively smaller. In one of the founding papers in Metabolic
Control Analysis,99 Reinhart Heinrich and Tom Rapoport in Berlin in 1974
were able to show by a mathematical proof that there is a natural tendency
in a pathway at steady state for the reactions with the largest equilibrium
constants to be the furthest displaced from equilibrium.
In conclusion, the simplest way that metabolite concentrations can be
16 1 Introduction: regulation and control

kept within reasonable bounds is for reactions with large negative standard
free energy changes to be far from equilibrium, and for reactions with small
standard free energy changes to be near to equilibrium. In addition, as Atkin-
son7 pointed out, chemical energy can only be conserved by maintaining a
reaction such as ATP hydrolysis far from equilibrium; free energy coupling in
metabolism requires the existence of non–equilibrium reactions.
If these thermodynamic considerations are now combined with chemical
kinetics, it is possible to show why non–equilibrium reactions are more likely
than near–equilibrium reactions to have a determining influence on the rate of
a metabolic pathway, using arguments developed by a number of biochemists
in the 1960s.
Consider first the reversible reaction:
f v
A −→
←− B
vr

If we consider simple chemical kinetics, where the rate vi of reaction i is a rate

constant, ki , times the reactant concentration, then vf = kf A and vr = kr B.
At equilibrium, there is no net change, so vf = vr , so:

kf A = k r B

and since B/A = Keq :

B kf
= = Keq (1.5)
A kr
If we now consider this reaction when it is not at equilibrium, we can sub-
stitute this equation (Eqn. 1.5), and the definition of the mass action ratio
(Eqn. 1.2) into the expression for vf /vr as follows:
vr kr B
=
vf kf A
kr
= Γ
kf
Γ
= =ρ (1.6)
Keq

Now let us consider that this reversible reaction occurs in a metabolic

pathway, which has reached a steady state in which there is a flow vp along
the pathway:
vp f vvp
−→ A −→
←− B −→
vr

For the pathway to be at steady state, vf − vr = vp , or vf − vp = vr , so using

Eqn. [1.6] gives:
vr vf − v p
= =ρ
vf vf
1.4 Basic theory of metabolism 17

Alternatively,
vp
=1−ρ
vf
With vp > 0, this shows that the reaction can only be near–equilibrium (i.e.
with ρ ' 1) if vf À vp so that vf ≈ vr . A non–equilibrium reaction has
a small ρ, and therefore the pathway rate, vp is comparable in size to the
forward reaction rate vf , and the rate of the reverse reaction, vr , is relatively
very small. (Although these relationships between ρ and the ratio of the
reverse and forward reaction rates have been derived here using first–order
chemical kinetics, the same expression is obtained for the ratio of the reaction
rates if proper enzyme kinetic equations are used.201 )
Such calculations convinced biochemists that a near–equilibrium reaction
is catalyzed by an enzyme that is present in a great excess over the amount
required to deliver the overall flux of the pathway, and that therefore it is
not limiting the rate of the pathway. The non–equilibrium reaction, on the
other hand, is a potentially limiting factor on the overall pathway flux, be-
cause the catalyzed reaction rate is only slightly greater. Furthermore, many
biochemists gained the impression from these equations for the ratios of rates
that the reverse and net rates of a non–equilibrium reaction would not be
significantly affected by changes in the product concentration. Unfortunately
this is not true for enzyme–catalyzed reactions because the values of the net
and reverse rates do in general depend on product concentration, though as
the proportional effect is the same on both, the ratio does not show this. As
a result, product inhibition effects have been wrongly discounted unless they
are unusually strong.
A second argument (originally put forward by another Nobel Laureate,
Hans Krebs140 ) reinforced the view that non–equilibrium enzymes can be
rate–limiting, but near–equilibrium enzymes cannot. Suppose more enzyme
is added to catalyze a reaction that is already near–equilibrium; this will tend
to increase the amount of product and decrease the amount of substrate to
bring the reaction closer to equilibrium (as Eqn. [1.6] shows for an increase
in vf ). However, since the reaction is already close to equilibrium, if it is
brought closer to equilibrium, there will not be much change in the metabolite
concentrations, nor will the reaction using the product be stimulated to go
much faster so there is no reason to expect a great change in the pathway rate.
On the other hand, for a non–equilibrium reaction, increasing the amount
of enzyme could bring the reaction much closer to equilibrium, causing the
product concentration to rise significantly, and stimulating the next reaction
to go faster. Thus the change could propagate through the pathway to cause
a significant change in pathway flux.
These arguments may seem persuasive and in many cases they probably
have some validity. A potential weakness though is that they focus too closely
on a single reaction in the pathway, and give insufficient weight to how the
metabolite concentrations will change as the whole system responds to an
attempt to change the rate at one step. Later we will see why these fears are
justified (Chapter 5.5.1).
18 1 Introduction: regulation and control

1.4.3 Principles of regulation and control

Until recently, what I termed the qualitative systems approach has been dom-
inant in the attempt to formulate principles of regulation and control. That
is, the theories were developed verbally, and little in the way of quantitative
tests was undertaken to check their adequacy. Unfortunately, there is one
seriously inadequate concept that has held a dominant position throughout
the history of the study of metabolic regulation until relatively recently. Fur-
thermore, because it has been so dominant, it is impossible to explain the
motivation behind many of the experiments in metabolic regulation without
referring to it. This troublesome concept is that of the rate–limiting step,
originally proposed in the form:

When a process is conditioned as to its rapidity by a number of

separate factors, the rate of the process is limited by the pace of
the slowest factor. (Blackman 18 )

This concept has been described as a ‘truism’ and ‘self–evident’, and versions
of it continue to be stated in biochemistry textbooks up to the present day.
In many ways this is odd, because it is difficult to give a literal interpretation
of the concept. As explained in the earlier section on the steady state (p. 10),
a requirement of successive steps along a linear metabolic pathway is that
they are operating at the same rate in the steady state, so none of them
could be termed the ‘slowest’ step in the normal sense of the word. What was
meant by the many advocates of the ‘rate–limiting step’ was that the rate of
a pathway could be altered only by changing the activity of one particular
enzyme in the pathway. On this basis, the study of the control of a pathway
entailed the identification of the rate–limiting step. From time to time, doubts
were expressed: why should the fact that the alteration of the activity of one
enzyme can change the rate of a pathway mean that this cannot be true of
another enzyme in the pathway? Attempts to analyse the kinetics of multi–
reaction systems did not support the rate–limiting step; it was possible to
show that the overall rate depended in principle on the kinetics of every
reaction in the system. Until recently, these criticisms had little impact, but
this is changing with the growing acceptance of Metabolic Control Analysis.
From Chapter 5 on, the theoretical and experimental grounds for rejecting
the rate–limiting step concept will be presented. In summary, it has been
found that more than one enzyme can affect the rate of a pathway; although
it is possible to imagine conditions where only one step affects the rate of a
pathway, experimental studies show that this is not the usual case.
In studying the control of metabolism, we therefore have a problem. Much
of the work that has been done has been interpreted in terms of the concept
of the rate–limiting enzyme. This work cannot be disregarded, because it
has given us methodologies for studying metabolic pathways, has provided
evidence about what does happen in pathways as their rates change and has
led to the discovery of many mechanisms that appear to have a role in the
control of pathway rates. Since it is necessary to look back at this work, as
1.4 Basic theory of metabolism 19

I shall do particularly in Chapter 4, it is necessary to recognize that it was

often motivated by the search for a rate–limiting enzyme. This search was
conducted by looking for enzymes that had the characteristics of potential
regulatory enzymes, and then attempting to gather evidence that one of these
was the rate–limiting step.
The key characteristics of regulatory enzymes were variously given as:

• an enzyme that catalyses a non–equilibrium reaction (e.g. Newsholme

& Start, 1973 171 ).

• an enzyme that catalyses a non–equilibrium reaction and whose activ-

ity is controlled by factors other than substrate concentration. (e.g.
Newsholme & Start, 1973; Denton & Pogson 1976 60 ).

• . . . the enzyme which responds to the original metabolic signal and

thereby initiates subsequent changes in the activity of the remaining
enzymes . . . (after Newsholme & Start, 1973).

The reasons for concentrating on non–equilibrium enzymes were explained

previously. However, pathways often contain several non–equilibrium en-
zymes, so it was thought necessary to look for further features that distinguish
between them. All enzymes will change their rate of reaction if their substrate
concentrations are changed, unless they are already saturated (which is not
usually the case in a pathway). The second definition of a regulatory enzyme
therefore attempts to eliminate enzymes that can be regarded as responding
passively to a requirement to change rate because some other change in a
pathway has altered the rate of supply of substrate, but that lack any mech-
anism to initiate a change in the pathway rate. The third definition goes
further by stating that there might be a number of enzymes that satisfy the
second definition, but that one amongst these must be the one that first re-
acted to some signal to change the pathway rate. Chapter 4 describes the
ways in which these characteristics were identified in the various enzymes of
a pathway.
The difference between the second and third definitions of regulatory en-
zymes given above implies that some of them could have functions other than
flux control. Bücher and Rüssmann24 noted in 1963 that one of the remark-
able things about muscle glycolysis was that the pathway flux could change by
a factor of more than a hundred between the resting and working state with
only small changes in the relative concentrations of the metabolites. They
linked this homoeostasis of cellular metabolite ratios (as they termed it) to
the requirement for functional readiness of the muscle on the basis that the
time taken to establish a new steady state when the flux is changed is shorter if
the associated changes in metabolite concentrations are small. This is because
the time taken to reach a new steady state includes the time taken to syn-
thesize or remove the metabolic intermediates. (This concept has since been
verified in more detailed analyses, first by John Easterby 62, 64 and later by
others.157 ) Bücher & Rüssmann therefore proposed that regulatory enzymes
20 1 Introduction: regulation and control

were involved as much in homoeostasis of metabolite levels as in flux control,

and for this reason, multiple sites of control were needed. Biochemists gener-
ally seemed to pay little attention to this original suggestion of a requirement
for multiple control sites, but towards the end of this book, I will raise the
topic again and look at the reasons for believing Bücher & Rüssmann were
right (Chapter 8.1, p. 263). Their observations about homoeostasis of concen-
trations were noted though, as reflected in Newsholme & Gevers’ definition 169
of a regulatory enzyme as one:
. . . the activity of which is controlled by factors other than sub-
strate availability, and the activity of which controls the rate of
flux or the concentrations of metabolic intermediates.
Some biochemists, therefore, had recognized that there could be two types of
role for regulatory enzymes:
• The stabilization of the pathway to adapt its production or consumption
of a particular metabolite to the rest of metabolism. This is metabolic
homoeostasis, and often involves internal signals acting on the regula-
tory enzymes.
• The response to external signals that change the state of a pathway or
metabolism in general (e.g. hormones, second messengers etc). This is
control, again acting on regulatory enzymes.
Metabolic Control Analysis introduces coefficients that measure the potential
of enzymes to fulfill the second of these roles. One of the reasons that some
biochemists have been reluctant to give up the traditional characterization of
regulatory enzymes is that these measures from Metabolic Control Analysis
do show that some enzymes that clearly possess regulatory properties cannot
control pathway flux. My reason for mentioning this is to underline my final
point: these two aspects of regulation and control are distinct, and there is
no reason to assume that they are both performed by the same regulatory
enzymes. In fact, Metabolic Control Analysis gives grounds for believing
that there is a basic incompatibility between these two roles. The traditional
approaches to the search for rate–limiting enzymes, on the other hand, offer
no clear criteria for distinguishing between them.

1.5 Summary
1. Regulation is homoeostasis, or the maintenance of constant conditions
in the face of external perturbations.
2. Control is the ability to make changes as necessary.
3. Regulation and control are properties of metabolic systems of great com-
plexity and the unsolved challenge is to link our knowledge of molecular
details to system–level explanations in a convincing yet feasible manner.
1.6 Appendix: free energy changes 21

4. In metabolism, different mechanisms for regulation and control exist on

different time–scales:

• On long time scales, amounts of enzyme are changed by enzyme

synthesis and degradation.
• On medium time scales, the activities of ready–formed enzyme are
altered by covalent modification.
• On short time scales, the activities of enzymes alter in response to
changes in metabolite concentrations.

5. Metabolic pathways generally tend to a steady state, a dynamic equi-

librium where rates of formation of intermediates equal their rates of
breakdown and concentrations remain constant. Individual reactions
are displaced from chemical equilibrium, but reactions with small stan-
dard free energy changes tend to be close to equilibrium and reactions
with large standard free energy changes tend to be far from equilibrium.

6. Enzymes that catalyze reactions that are far from equilibrium are thought
to be more important in regulation and control than those whose reac-
tions are near to equilibrium.

7. In the past, analysis of regulation and control has concentrated on iden-

tifying the supposed rate–limiting enzymes for each pathway, but this
concept is unsatisfactory in many respects.

8. Regulatory enzymes can function to regulate intermediate concentra-

tions or to control flux.

1.6 Appendix: free energy changes

There are serious objections to the use of ∆G to represent the free energy
change in a biochemical reaction for reasons that have been given by Rick
Welch in New Orleans.271 When ∆H is used to represent the enthalpy change
of a reaction, it indicates the difference in enthalpy between the reactants and
products, measured by the heat change on complete conversion. The free
energy value given for a (bio)chemical reaction in solution is not a difference
between initial and final states after complete conversion, but the rate at
which the free energy is changing during the conversion of reactants to the
products at the current values of their concentrations. It should really be
represented as ∂G/∂ξ, where ξ is a measure of the extent of the reaction in
terms of moles converted. This is also equal to −A, where A is the chemical
affinity used in nonequilibrium thermodynamics to characterize the driving
force on a reaction. Unfortunately, this more appropriate terminology is not
familiar from standard biochemical texts, so the usual, misleading symbol will
be used here.
22 1 Introduction: regulation and control

Further reading
1. Atkinson, D. E. Cellular Energy Metabolism and its Regulation. Aca-
demic Press, New York (1977).
2. van Dam, K. Biochemistry is a Quantitative Science. Trends Biochem.
Sci. 11, 13–14 (1986).
3. Kacser, H. Control of Metabolism, Chap. 2 (sections I & IIA–C, pp. 39–
44) in The Biochemistry of Plants, Vol. 11, D. D. Davies (ed.) Academic
Press, New York (1987).

Problems
1. Phosphofructokinase catalyses the reaction:
Fru-6-P + ATP *
) Fru-1,6-bisP + ADP
The following substrate and product concentrations were measured in
the perfused rat heart: Rat heart contains 5.67 cm3 of H2 O/g dry

Metabolite Tissue content

(µmol/gdry tissue)
Fru-6-P 0.154
Fru-1,6-bisP 0.042
ATP 21.70
ADP 2.49

weight; 50% of this water is intracellular and 50% extracellular. Cal-

culate the intracellular concentrations of the metabolites, and then the
mass action ratio, Γ. The equilibrium constant Keq is 1.0 × 103 . Calcu-
late the ratio Γ/Keq and thus conclude the degree of displacement from
equilibrium.
2. Phosphoglucoseisomerase catalyses the reaction:

Glu-6-P *
) Fru-6-P

for which the equilibrium constant is believed to be in the range 0.36 -

0.47. Under control conditions, the contents of the two intermediates in
perfused rat hearts were determined as 290 and 63 nmol/g wet weight
respectively. Given that rat heart contains 0.43 cm3 of intracellular
water per g wet weight, what are the concentrations of these metabolites
in the cytoplasm? What is the mass action ratio, Γ, for the reaction,
and what can be inferred about its displacement from equilibrium?
 2

Methods for studying

metabolism and its regulation

2.1 Introduction
Although my aim in this book is to describe a new approach to the inter-
pretation of control and regulation in metabolism, the experimental methods
that are needed are exactly the same as those traditionally used in biochem-
istry. This chapter therefore introduces the experimental measurements that
have been used in the past to develop explanations of control and regulation
and that continue to be used in the growing body of experimental studies in
Metabolic Control Analysis.
There are four main issues that have to be addressed before carrying out
most experiments in metabolism:

• the choice of the particular experimental system in which the metabol-

ism is to be studied;

• the problems of measuring metabolite concentrations in vivo;

• the problems of making meaningful measurements of the activities of

the relevant enzymes, and

• deciding how to measure the metabolic fluxes.

Each of these issues is considered in turn in the following four sections.

2.2 Experimental systems

In principle, the choice of experimental system could start with the organism
in which to study the metabolic activity of interest, but in practice, the range
of this element of the choice can be very limited from the beginning. The

23
24 2 Methods for studying metabolism and its regulation

reasons for this, though, are varied. Some will relate to the extrinsic motiva-
tions for studying a particular metabolic pathway in the first instance. For
example, if the aim is to explore the differences between an aspect of human
metabolism in health and disease, and if suitable human cell or tissue samples
cannot readily be obtained ethically, then it will almost certainly be necessary
to use animal material. The animal would need to be no larger than necessary
to supply sufficient material for study (since larger animals are more expen-
sive to feed and house humanely), have a metabolism sufficiently similar to
humans for the results to be relevant, and have been sufficiently studied in
the past that the results can be placed in context. It is also an advantage if
inbred strains are available that have a restricted range of individual varia-
tion. In most cases, this points to the small rodents. Another aim might be
to study the metabolism of an organism of economic importance, for example
an animal or plant used as food, or a microorganism that produces a valuable
chemical such as a drug, and in this case there is no real choice. What I
am implying, of course, is that the extrinsic motivation to study a particular
organism arises from the existence of a organization prepared to fund bio-
chemists to carry out the experiments. (Unfortunately, there are few sources
of funding for carrying out basic research propelled by an intrinsic motiva-
tion to understand metabolic regulation better, in spite of the evidence that
the funded applied research is handicapped by the limitations in our basic
understanding.)
Given, then, that a choice of organism has been made, or forced, there
still remains the choice of how it, or parts of it, are to be used. Here there
are a range of options (not all of which are appropriate to all organisms):

2.2.1 Whole multicellular organisms

At this level of study, examples of the types of experiment that can be done
include the effects of diet or environmental conditions on the tissue contents
of enzymes and measurement of the overall patterns of conversion of nutri-
ents into various end products. While the organism is alive and actively me-
tabolizing, the methods of investigation must cause little perturbation. For
example, with animals, non–invasive methods such as analysis of respiratory
gases, urine and faeces can be used, or else minimally invasive methods such
as taking blood samples or small tissue samples with a biopsy needle. Other
possibilities can include the tracking of the metabolism of compounds that
have been labelled isotopically, for example by substituting hydrogen with
its radioactive isotope tritium ( 3 H) or carbon by its radioactive isotope 14 C
and monitoring the distribution of radioactivity. Recently, the technique of
nuclear magnetic resonance (NMR), which produces characteristic spectra of
chemicals based on the magnetic properties of certain atomic nuclei (including
hydrogen, phosphorous and the carbon isotope 13 C) has been scaled up to
allow non–invasive measurements of metabolites in vivo on human limbs or
whole small animals. (More details of NMR spectrometry are given later, in
section 2.3.2.)
2.2 Experimental systems 25

After death of the organism, the full range of analytical biochemical tech-
niques can be applied to tissue samples, but this obviously terminates the
experiment with that individual, and different individuals must be analysed
to determine the effects of different experimental conditions.
The advantage of whole organisms studies is that metabolism is taking
place under physiologically normal conditions.
The disadvantages are that:
• physiologically normal conditions for multicellular organisms means that
there are metabolic interactions between tissues, which usually have dif-
ferent metabolic specializations. Observations will show the net result of
these interactions, as modulated by the effects of hormonal and nervous
stimuli (which may be difficult to control).
• metabolizable substrates must reach the site of their metabolism, and
the products escape, by crossing the natural permeability barriers. Since
these are selective, this reduces the scope for introducing artifical sub-
strates or changing the amounts of intermediates that are normally con-
fined within the cells.
• the natural biological variability of the organisms on which the exper-
iments are performed means that it is necessary to carry out replicate
experiments to determine the mean behaviour and to perform statisti-
cal tests to establish the significance of any observed differences between
control and experimental conditions. This increases the number of ex-
perimental subjects that have to be used, which increases the time taken
and the cost, and, in the case of animal experiments raises the ethical
dilemma of minimizing the number of individuals used whilst adequately
verifying the findings.

2.2.2 Isolated tissues and organs

The metabolism of a specific part of a multicellular organism can be studied
in isolation if it can be maintained in aviable state.
The advantages that this offers include:
• the metabolic responses of the tissue can be studied in the absence of
interactions with other organs;
• it is easier to control the concentrations of nutrients, hormones and so
on to which the tissue is exposed, and yet
• the functional integrity of the tissue is retained.
The disadvantages of studying whole organs include:
• the supply of nutrients and removal of excretory products can be tech-
nically difficult, particularly with animal tissues in the absence of a
functioning blood circulation. The simplest solution is to use naturally
26 2 Methods for studying metabolism and its regulation

Fig. 2.1: Apparatus for perfusion of isolated rat heart.

The components are not to scale.

thin tissues and to rely on exchange of materials with the surroundings

by passive diffusion. Examples of suitable thin tissues include plant
leaves and the diaphragm muscle and epididymal fat (the fat around
the testes) of rats. With larger animal organs it is necessary to replace
the blood circulation by perfusion of an oxygenated nutrient solution
through the blood vessels. This is more technically difficult and re-
quires more equipment (see Fig. 2.1). Liver and heart are two organs
often studied in this way.
• tissues and organs themselves are often heterogenous with different cell
types exhibiting different metabolic functions and responses;
• most of the permeability barriers that restrict experiments in whole
organisms are still intact in the isolated organs;
• biological variability will still result in the need for replicate experiments
and statistical analysis. If, as with animal liver and heart, there is only
one such organ per animal, the number of experimental subjects needed
is no lower than with whole animal studies. Where there is more than
2.2 Experimental systems 27

one organ per organism, then the numbers needed are reduced because
control and test experiments can be performed on matched organs from
the same organism.

2.2.3 Tissue slices

With tissues that are too thick for diffusion to be an efficient means of sup-
plying nutrients and removing waste products, another option is to cut from
them uniform slices of the order of 1 mm thick. This works well with potato
tubers and the kidney cortex of rats for example, but is less successful with
liver as the slices prove to be less viable.
The advantages are:

• the procedure is technically simpler than organ perfusion, since the slices
just have to be incubated in oxygenated nutrient solution as the nutrient
supply is by diffusion, and

• fewer individuals are used in animal studies because many slices can
be taken from one organ and this also reduces the effects of biological
variability since control and test treatments can be compared on slices
from the same organ.

There are some disadvantages in that:

• there is inevitably cell damage at the cut surfaces, and substances re-
leased from the damaged cells may affect the undamaged ones, and

• the permeability barrier formed by the cell membranes is still in place.

2.2.4 Isolated cells

If the initial choice of organism was a microorganism, this will be the first
relevant option. However, it is increasingly an option in plant and animal
studies. Of course, there are few types of animal cell that exist naturally in a
free state; red blood cells and sperm are examples, but they have extremely
specialized metabolism. Certain types of cancer cell have been established as
cultures that can be grown readily in laboratory conditions and are available
from cell culture collections. However, cancer cells show significant metabolic
differences from the normal cells from which they are derived and whilst this
is important in its own right for the study of cancer as a disease process, they
cannot be used as reliable guide to the metabolism of normal cells. Most
differentiated animal cells will not grow indefinitely in culture conditions, al-
though some relatively undifferentiated precursor cells may, most readily as
cell monolayers attached to a surface. In general, animal cells die after about
20 to 40 cell divisions (a limitation on the growth of cultures often termed
the Hayflick limit). Undifferentiated cells persist in plants throughout their
28 2 Methods for studying metabolism and its regulation

lives and can often be grown as cell suspensions. In plants, even differenti-
ated cells may dedifferentiate and grow well, whereas differentiation is largely
irreversible in animal cells.
Techniques have also been developed for isolating cells from animal organs.
For example, in 1969 Berry and Friend16 in Australia devised the method
commonly used these days for obtaining individual cells, hepatocytes, from
mammalian liver in high yield. Their innovation was to perfuse the liver
with an oxygenated solution containing proteolytic enzymes that attack the
intercellular ground substance and the adhesion factors on the cell surfaces.
After a short while, the tissue loses its integrity and can be dispersed to yield
free cells. Similar techniques have been applied to other tissues, for example,
giving fat cells (adipocytes) from fat tissue. In most cases, these isolated
cells will not grow in culture and can only be maintained for relatively short
periods.
Compared to the other systems examined so far, the advantages of using
isolated cells include:

• diffusion is entirely adequate for nutrient supply provided the cell sus-
pension is kept adequately mixed and there is sufficient surface area for
gas exchange (where necessary). The equipment required is therefore
relatively simple.

• many samples can be taken from the same batch of cells, and since
each sample might typically contain 105 –107 cells, inter–sample varia-
tion should be minimal. This should reduce the number of experiments
needed to establish statistically significant findings. Variability between
different experiments, and between different laboratories, can be reduced
if clones of cells are used.

• in studies of animal metabolism, many fewer experimental animals will

be needed. This may be just because of the previous reason, but in the
case of cell culture, the material from the original animal is multiplied
up many times. However, it should be noted that most animal cell
culture media need the inclusion of some foetal calf serum to supply the
growth factors without which most isolated animal cells die. (In plant
cell cultures, coconut milk supplies the needed factors.)

The disadvantages of studies on isolated cells are that:

• the amounts of cellular material are small, particularly in relation to

the amount of suspension medium, so it is necessary to have efficient
techniques for separating the cells and sensitive analytical techniques
for assaying their contents, and

• the cell membranes still pose permeability barriers.

2.2 Experimental systems 29

2.2.5 Permeabilized cells

Cells can be made permeable to small molecules by partial removal of lipid
from their plasma membranes by treatment with aqueous solutions of organic
solvents or nonionic detergents and still remain metabolically active. The
treatment has to be brief and carefully controlled so that the holes made in
the membrane are too small to allow the enzymes to leak out. In an early
experiment of this type, Serrano, Gancedo & Gancedo permeabilized yeast
cells and showed that it was possible to measure the kinetics of hexokinase
retained within.225
The advantages of using permeabilized calls are the same as those of
isolated cells plus the greater access to the cell interior, which allows the
addition of metabolites that would not normally cross the membrane.
The disadvantages, apart from those associated with isolated cells gen-
erally include:

• the dilution or loss of key metabolites of low molecular mass, and

• possible uncertainty about the extent of damage to the cells and the
degree of loss of enzymes etc.

2.2.6 Cell–free systems

One of the first experiments in metabolic biochemistry was when the Buch-
ner brothers ground up a paste of yeast cells with sand and demonstrated
that ethanol formation from sugar could take place in a cell extract. Hans
Krebs’ experiments that led to his discovery of the tricarboxylic acid cycle
used homogenates of pigeon breast muscle, so metabolically competent, cell–
free systems obtained by mechanically breaking tissues and cells have a long
history. A more recent variant is to take isolated, purified components, such as
individual enzymes, and to combine them to form a working, though limited,
replica of a particular segment of metabolism.
The advantages are the simplicity and reproducibility of the procedure
and the lack of permeability barriers. The disadvantages are:

• the variable degrees of loss of (eukaryotic) cell structure, such as the

separate compartments formed by intracellular organelles and any pos-
sible cytoskeletal structure. Whether this is significant for the pathway
under study would have to be a matter for experiment.

• the possible dilution of key metabolites and enzymes if the cells are lysed
or homogenized (as is usual) in a liquid medium.

2.2.7 Isolated organelles

Just as there is metabolic specialization amongst the tissues of multicellular
organisms, so there is in the organelles of eukaryotic cells. Therefore, it is
30 2 Methods for studying metabolism and its regulation

often desirable to obtain a particular organelle in a pure state by subcellu-

lar fractionation (usually by differential or zonal centrifugation) in order to
determine its metabolic capabilities in the absence of interactions with the
cytoplasm and other organelles. For example, oxidative phosphorylation is
most usually studied on preparations of isolated mitochondria by the tech-
niques described in more detail later in the book (Chapter 6.1.4.2, p. 161).
The advantages, apart from that of isolating a particular metabolic sub-
system of the cell, are similar to those of studying isolated cells (indeed,
mitochondria and chloroplasts are comparable in size to bacterial cells).
The disadvantages are that:
• the degree of damage during isolation and the purity of the samples
are difficult to assess. Some types of damage are inevitable and re-
producible. For example, the biochemist’s microsomes are extensively
disrupted fragments of endoplasmic reticulum. Again, the mitochondria
of many cell types are large extended structures present in small num-
bers (unlike those from the animal liver cells that are used as their most
frequent source) and presumably fragment into smaller pieces during
isolation.
• the cellular environment is replaced by an artificial one, although this
does offer greater opportunity for experimental manipulation, and
• some organelles’ membranes act as permeability barriers restricting ac-
cess.

2.2.8 Isolated enzymes

The next lower level of study is the enzymes themselves. Techniques for
purification of enzymes have been developed throughout this century, and
many are now prepared on industrial scale and marketed. Obviously, an
individual enzyme is hardly a metabolic system in its own right. However,
studies of the properties of isolated enzymes have an important role in the
study of metabolic regulation and will be introduced in Chapter 3.

2.3 Measurement of metabolites

In any approach to understanding the inner workings of metabolism it is
essential to know the concentrations of the intermediary metabolites whilst
the pathway is in operation and to know how they change in circumstances
that affect the rate of the pathway. In making these measurements there are
several distinct problems that I will describe in turn.

2.3.1 Fixing concentrations

Only a small number of analytical techniques allow the estimation of in vivo
concentrations of metabolites. Examples include the use of fluorescence spec-
2.3 Measurement of metabolites 31

trophotometry to measure NADH concentrations, dual beam spectrophotom-

etry to measure the concentrations of redox components of the mitochondrial
and chloroplast electron transport chains, and nuclear magnetic resonance
spectrophotometry (which unfortunately can take up to 45 minutes to take
a single spectrum, and even then can only measure the more concentrated
metabolites with an error of about 10%). Most other methods of analysis
therefore require release of metabolites from cells and take some time to com-
plete. For these methods, the problem they all have in common is how to
prevent changes in metabolite concentrations between the time a sample is
taken for analysis and the results recorded. The goal is to stop all the en-
zymic reactions rapidly and simultaneously to get an instantaneous snapshot
of the concentrations. How rapidly this needs to be depends on the amount of
metabolite in the cell relative to the rate at which it is produced or consumed;
the amount divided by the rate gives the turnover time. Typical values range
from several seconds to milliseconds in the worst cases (such as oxaloacetate
in the tricarboxylic acid cycle — for which the situation is virtually hopeless).
Methods that have been used to quench metabolism rapidly include:

• sudden pH change, e.g. by mixing with perchloric acid (HClO4 ), which

is a powerful protein denaturant. This is particularly suitable for cell
free systems that can be rapidly injected into a well–mixed solution of
the acid. It can be used with cells, but will be slower because of the
lag time whilst the acid penetrates the cells. One of the motives for
using perchloric acid is that the resulting solution can be neutralized
with KOH, and the perchlorate removed because potassium perchlorate
precipitates from cold solutions.

• rapid freezing at liquid N2 temperatures. This relies on the rapid drop in

temperature first slowing enzymic reactions and then stopping them as
the system solidifies. Freezing rates are very rapid if cell–free solutions
or cell suspensions are sprayed into the liquid nitrogen. Larger objects
such as isolated organs cannot be frozen sufficiently rapidly unless they
are very thin. This is because the nitrogen boils and forms an insulating
layer of bubbles over their surface. For isolated organs, such as heart
muscle, the technique of freeze–clamping is used instead. In this, two
blocks of metal (aluminium or copper) with a high heat conductivity and
mounted on long scissor–like handles (a Wollenberger clamp) are cooled
in liquid nitrogen. The organ is then squashed between the two metal
blocks. The temperature of a guinea pig kidney has been measured to
fall from 38◦ to 0◦ in 90 ms and to -160◦ in 0.5 s when frozen in this
way. Whatever the way in which the samples have been frozen, they
are ground to a powder whilst still frozen, and then thawed by rapid
mixing with perchloric acid so that enzymic activity does not recover as
the metabolites are taken into solution.
• mixing with organic solvents. Again this relies on the denaturing effect
of the solvent to stop enzyme activity. For example, Melvin Calvin used
32 2 Methods for studying metabolism and its regulation

hot methanol to stop metabolism in Chlorella algae in his experiments

to determine the photosynthetic carbon pathway by determining the
rate of appearance of 14 C in intracellular metabolites from CO2 , which
was one of the early studies to make use of radioisotope labelling.

2.3.2 Low concentrations in complex mixtures

ADP ATP NADH NAD+

... ... .
... .. ... ..
... . . .. .
... . .
...
.... ... ... .
.... ..... .. .....
..... .... .... ...
....... ...... ..... ..... .....
pep .............................. ......... .................................. pyruvate ................................. .... .........................................
.
...
. .
...
.... ...
... .
...
..
lactate

pyruvate kinase lactate dehydrogenase

Fig. 2.2: A coupled enzyme assay.

Lactate dehydrogenase alone can be used to assay pyruvate in a sample by the fall in
NADH concentration. If pyruvate kinase and ADP are then added, the coupled assay
can be used to measure phosphoenol pyruvate (pep). Alternatively, if pyruvate kinase
and pep are added, ADP can be assayed.

Most intermediary metabolites occur in cells at concentrations in the mi-

cromolar to millimolar range amongst a mixture of thousands of other sub-
stances, some of which will be chemically similar. Measurement of their con-
centrations therefore requires sensitive methods that are either intrinsically
selective or that include a separation step. Examples of techniques that have
proved useful include:

• enzymic analysis, often using spectrophotometric or fluorimetric detec-

tion. Here the selectivity of the analysis is provided by choosing an
enzyme with a narrow specificity for the metabolite to be assayed. Of
course, this only helps if the products of the reaction are easier to de-
tect than the metabolite itself. A common strategy is to use dehydro-
genases so that the reaction consumes or produces NADH or NADPH,
both of which can be detected spectrophotometrically and, with about
a thousand times greater sensitivity, fluorimetrically. If the metabolite
of interest is not acted on by a dehydrogenase, coupled enzyme assays
can be used, where a sequence of two or three enzymes ends with a
dehydrogenase reaction. An example is shown in Fig. 2.2. Many other
2.3 Measurement of metabolites 33

detection techniques can be linked to the enzyme assay. One that pro-
vides particularly high sensitivity for measuring metabolites in very low
concentrations is luminescence. This involves the detection of the weak
pulse of light produced by the light–emitting enzymes, the luciferases.
Luciferase derived from fireflies acts on ATP, so detects it and reactions
that produce it whilst bacterial luciferase reacts with NADH and so can
be used with coupled enzyme assays involving NAD–linked dehydroge-
nases.

• chromatographic separation and quantitation. The requirements for

separation of complex biological mixtures have long been a driving force
in the development of chromatographic methods, and most techniques
have had some application in the separation of metabolic intermedi-
ates. The most useful methods are those which are linked to some form
of quantitative detection and these days the most versatile method is
HPLC (high performance liquid chromatography). This is based on ap-
paratus designed to force solvents at high pressure through strong, finely
divided support particles that can carry one of a variety of selective sta-
tionary phases that are responsible for the separations. Relatively high
flow rates allow separations to be achieved in around ten minutes, and
a variety of highly sensitive detectors have been developed to monitor
substances in the column eluate. Most metabolic intermediates carry
ionic charges and can be separated by ion exchange or reverse phase
columns.

• immunoassay methods. These exploit the specificity and high affinity of

binding by antibodies, but first it is necessary that the metabolite can
act as antigens and elicit an immune response, which is not straight-
forward since most metabolites are small molecules that are not in-
trinsically immunogenic. If they are attached to a larger immunogenic
molecule, then a response to them can sometimes be stimulated. Im-
munoassay methods are capable of extremely high sensitivity and have
been used for metabolites that are present at particularly low levels
and that are difficult to assay by other techniques, such as the second
messenger, cyclic AMP.

• nuclear magnetic resonance (NMR) spectrometry. Most spectrophoto-

metric techniques, for example visible/ultraviolet spectrophotometry,
are not sufficiently selective to be applied to a complex mixture of
metabolites in order to identify and assay specific components. The ex-
ception is nuclear magnetic resonance spectrophotometry. This detects
certain atomic nuclei that have a magnetic moment by means of their
absorbance and emission of high frequency radio waves when they are
placed in very strong magnetic fields generated with super–conducting
electromagnets. The magnetic field and radio frequency combination
that leads to absorption by an atomic nucleus is subtly influenced by
the valency electrons surrounding the nucleus, so that nuclei in different
34 2 Methods for studying metabolism and its regulation

Fig. 2.3: 31 P NMR spectrum of glycolysing yeast.

This spectrum was recorded in vivo by Kevin Brindle & Susan Krikler (Biochim. Biophys.
Acta, 847, 285–292, 1985, Fig.1.) MDP, reference signal; Pin i , cytosolic phosphate;
Pout
i , extracellular phosphate; SP, sugar phosphates; NTPγ, γ–phosphate of nucleotide
triphosphates, etc; PP1 terminal phosphate of polyphosphates; PP2 and PP3 , penulti-
mate phosphates of polyphosphate. The result is the average of 512 spectra collected
over 40 minutes.

chemical compounds absorb in different parts of the spectrum. Help-

fully for biologists, the nuclei of two common elements, hydrogen ( 1 H)
and phosphorous ( 31 P) are magnetic and give a signal. The rare isotope
of carbon, 13 C, also gives a signal, though this is extremely weak at the
levels of this isotope found naturally in organic material. However, by
using compounds enriched in their content of this stable isotope, it is
possible to monitor its spread through a metabolic pathway. NMR is
so selective that it can be used to detect and measure metabolites in
vivo without any need to quench metabolism and separate the metabo-
lites (Fig. 2.3). The disadvantage is that the sensitivity of the technique
is not quite as high as is needed. It can measure down to fractions of
millimolar, but this is only enough to detect the more abundant interme-
diary metabolites (though this does include important compounds like
ATP). In addition, the signals from phosphorous and carbon are very
weak, and may have to be collected for half an hour or more. Finally,
the equipment needed to make these measurements is very expensive to
buy, and costly to use because the super–conducting magnets have to
be kept near absolute zero all the time with a supply of liquid helium
(in a vacuum flask itself surrounded by liquid nitrogen).

2.3.3 Compartmentation
Tissues and cells contain a number of liquid compartments separated by se-
lectively permeable membranes. Most metabolites will not be present at a
uniform concentration in every compartment, and whilst average values may
show significant changes with experimental conditions, they cannot accurately
2.3 Measurement of metabolites 35

represent the concentrations to which a particular enzyme is exposed within

the compartment in which it is found.
The first distinction to be made is between intracellular and extracellu-
lar fluid compartments. Obviously these are both present in tissues, but
less obviously, they are also present in samples of packed cells obtained from
cell suspensions by centrifugation or filtration, since these techniques always
leave some medium around and between the cells. We can assume that most
metabolites are present only in the intracellular compartment, but that when
substances are present in the extracellular compartment, their concentrations
there can be determined by assay of the bulk medium in which the cells or
tissues were incubated. The key to the problem of determining intracellu-
lar concentrations is then to determine the relative volumes of water these
two compartments. A common approach is to determine the total volume
of water in the sample and then subtract the volume of extracellular water.
The total water content can be measured by the loss of weight of a sample
on gentle drying, though this is not compatible with sample preparation for
metabolite assays. Another possibility is to add tritiated water, [ 3 H]H2 O,
to the medium. A measurement of the radioactivity from the tritium in a
sample will then indicate its total water content, since the tritiated water will
distribute uniformly and rapidly through the system. The extracellular water
content can be determined by adding a substance to the incubation medium
that can be readily measured, is not naturally present in the cells and is un-
able to enter them across the cell membrane. Large polysaccharides, such as
inulin, labelled with radioactive 14 C, are often used.
A more difficult problem is to determine distribution of metabolites within
cells. In most cells, the cytoplasm will be the largest aqueous compartment,
so if a metabolite is reasonably uniformly distributed within the cell, the
average concentration obtained by dividing its content in the sample by the
intracellular water content will be close to the cytoplasmic concentration.
(Plant cells are more of a problem as the vacuole is the largest intracellular
volume.)
Few techniques allow the direct observation of metabolite distributions
in eukaryotic cells. NMR may be able resolve and measure cytoplasmic and
mitochondrial ATP because the 31 P spectra are pH–dependent and the pH
difference between these two compartments can be sufficient to separate the
peaks. Photometric measurements can be made on microscope images, but
either the cells must be fixed and stained for the metabolite in such a way that
it does not redistribute (for example by rapid freezing) or the measurements
must be made on living cells (for example, using the natural fluorescence
of compounds such as NADH, or fluorescent indicators and labels, such as
aquorein, which is sensitive to calcium concentrations).
Rapid animal cell fractionation techniques have been developed to measure
mitochondrial metabolite levels, which are the ones most likely to show signif-
icant differences from cytoplasmic concentrations in animal cells. One of these
was developed for hepatocytes by Zuurendonk & Tager283, 284 in Amsterdam
in the laboratories from which some of the major experimental advances in
36 2 Methods for studying metabolism and its regulation

Metabolic Control Analysis were to emerge a few years later. In their method,
a tube for a small centrifuge that accelerates very rapidly is filled with a small
volume of perchloric acid solution overlaid by a layer of inert silicone oil. On
top of this is laid a medium containing the metal–chelating agent EDTA and
the detergent digitonin. Digitonin attacks cell membranes according to their
cholesterol content, which is much higher in plasma membranes than in mi-
tochondrial membranes. In a short period of exposure, therefore, the outer
membranes of the cell lyse, but the mitochondria remain intact. The whole
tube is cooled on ice before the cell sample is added to the top layer and
mixed. 10–15 seconds later, the centrifuge is started and the mitochondria
are rapidly spun down through the oil into the perchloric acid. The centrifuge
is stopped and the top layer containing the cytoplasm is acidified. The lower
layer is assayed to give the metabolite concentrations in the mitochondria and
the top layer those in the cytoplasm. The separation of the mitochondrial and
cytoplasmic compartments is 80–90% efficient, and changes in the cytosolic
adenine nucleotide levels by continuing metabolism are minimized by the low
temperature and the EDTA, which removes essential Mg2+ ions.

2.3.4 Free and bound metabolites

Some of the metabolites in cells will, at any one time, be bound to proteins, in
particular to the enzymes that act on them. The fraction that is bound varies
depending on the concentrations of the metabolite and the binding protein
and the affinity of the binding reaction. Only the free metabolite in solution
contributes to the concentration that is ‘seen’ by enzymes, but the majority
of analysis methods measure the total amount of metabolite. Whether there
is a significant difference between the free and total concentrations varies on a
case by case basis. In muscle cells, for example, there is a measurable amount
of total ADP, but the evidence is that most of this is bound to the myosin
ATPase. Probably around half of the cellular content of the second messenger
cyclic AMP is bound to cyclic AMP–dependent protein kinase.
The sorts of experimental evidence that support these statements include:

• NMR measurements. For technical reasons, the NMR signals from

metabolites that are bound to large molecules are not usually visible
in the spectra and only molecules in free solution are detected. NMR
showed that ADP was undetectable in muscle cells, whereas the mea-
sured total concentrations were high enough to have been visible. Un-
fortunately, the limited sensitivity of NMR measurements means that
it cannot provide information on the metabolites at low concentration
that might be most affected by binding to proteins.

• determination of the binding characteristics of e.g. a cell homogenate.

This can be undertaken in the same way as binding experiments on
enzymes (see Chapter 3.4), except that there is a possibility that the
metabolite might be converted by enzymes in the homogenate. (This
2.4 Measurement of enzyme activity 37

could be averted by removing essential coenzymes and cofactors.) This

was one of the ways that the binding of cyclic AMP was assessed.

• estimation using measured concentrations of the known major binding

proteins and the dissociation constant. All the necessary information
may not be easy to collect with sufficient accuracy to make a reliable
calculation, but it is often possible to assess whether sufficient binding is
likely to occur for further investigation to be warranted. In both specific
cases I have cited, binding of cyclic AMP by protein kinase and muscle
ADP by myosin, such calculations indicated the likelihood of significant
binding.

• calculation of the concentration assuming the metabolite is in equilib-

rium with other metabolites that are unbound. The assumption that
the enzyme catalysing the reaction involved is sufficiently active to
bring the reaction near to equilibrium is obviously difficult to test when
the metabolite concentration measurements are themselves in question.
However, ADP concentrations calculated from the equilibrium constant
of the creatine kinase reaction and the measured concentrations of ATP,
creatine and creatine phosphate are much lower than measured total
ADP concentrations.164, 258 Thus discrepancies in such calculations, or
equivalently, the observation of consistently unusual disequilibrium ra-
tios involving a particular metabolite, can offer evidence that there is a
problem.

2.4 Measurement of enzyme activity

Measurements of enzyme activity are invariably involved in the experimental
investigation of metabolic regulation and control, yet it is often not clear how
these measurements should be made in order to provide relevant evidence. I
will not describe basic techniques of enzyme assay here; let us assume that
assay methods exist for the enzymes of interest and consider some of the
remaining problemmatical issues.

2.4.1 Assay conditions

Many enzyme assays were developed and optimized to aid detection of the
enzyme and to carry out kinetic investigations for identifying the catalytic
mechanism (see Chapter 3). The resulting assay conditions were not neces-
sarily intended to reproduce in vivo conditions. For example, the pH of an
assay may have been chosen because it is the optimum pH for enzyme activity,
or because it is the pH at which the change in light absorption used for the
assay is at a maximum, or because the equilibrium constant of the reaction is
more favourable, or because it is the value at which the enzyme is most stable.
Resemblance to intracellular pH will have come low down the list of priorities.
Similar considerations may have swayed the choice of the assay temperature.
38 2 Methods for studying metabolism and its regulation

Does this matter? If the aim of the measurements is to detect relative

changes in the amount of active enzyme in different physiological states, then
probably not. If the aim is to compare the measured enzymic activity with the
rate of metabolism (one of the traditional stages in the search for regulatory
enzymes — see Chapter 4.4), then it would be better if pH, temperature and
the ionic environment were as close as possible to in vivo values.

2.4.2 What to measure

Similar difficulties surround the choice of substrate concentrations at which
to measure the activity. If high concentrations (relative to the Km values)
are used, the measurement approximates to the limiting rate (or maximum
velocity), V . Again, this may be a useful measurement if the aim is to detect
variations in the amount of active enzyme. It is of limited use otherwise, since
few enzymes are exposed to such high substrate concentrations in vivo, and
act in the presence of the reaction products, which are rarely included in assay
mixtures. (See Fig. 3.5 to see how much difference the products can make.)
It is doubtful whether many enzymes in vivo are acting at more than a few
per cent of their limiting rates.

2.4.3 Control of other factors affecting in vivo activity

Finally there are the miscellaneous factors that are the stuff of biochemists’
nightmares because, if they affect the enzyme in question, but have escaped
detection or not been adequately controlled, they could render the results
meaningless. Some of the possibilities are:

• existence of unknown effectors. If the enzyme samples contained variable

amounts of such an effector, then the results would be uninterpretable.
This was found to have happened in 1980 in a metabolic pathway that
has been known for most of this century: glycolysis. Hers, Hue and
Van Schaftingen in Belgium discovered that the most powerful acti-
vator of phosphofructokinase was the previously unknown compound
fructose–2,6–bis phosphate.253, 254 It acts as a messenger of hormone ac-
tion. Ignorance of this effector has almost certainly resulted in erroneous
conclusions about the regulation of glycolysis, since much has been made
of the apparent (and unusual) product activation of mammalian liver
phosphofructokinase by fructose–1,6–bis phosphate, but this is probably
an artefact caused by binding of the product at the site for the effector.
When fructose–2,6–bis phosphate is present at in vivo levels, activation
by fructose–1,6–bis phosphate is much less significant and is often out-
weighed by its action as a normal product inhibitor,255 as is the case
for yeast phosphofructokinase.11, 186 Strangely, one of the criticisms of
Metabolic Control Analysis has been that it would lead to the wrong
conclusions if there were undiscovered effectors. Apart from the obvi-
ous point that all previous explanation of whatever form would be in
2.5 Measurement of metabolic fluxes 39

danger of being wrong if there were undiscovered effects, the criticism

seems to overlook that it is a key characteristic of science to discard old
explanations and embrace new ones when fresh discoveries are made.

• degree of activation by cyclic covalent modfication reactions. One of

the ways that enzyme activity is altered in vivo is by enzyme–catalyzed
modification reactions (see Chapter 7.4). These can either intercon-
vert the enzyme between active and inactive forms, or change its ki-
netic properties more subtly. The difficulty is ensuring that the degree
of modification does not change from its in vivo level when the sam-
ple is taken for enzyme assay. For obvious reasons, it is not possible
to use protein denaturation to inhibit the modification reactions, as is
done for metabolite assays. Instead, the modification reactions have to
be stopped by removal of necessary cofactors or the use of specific in-
hibitors. For example, glycogen synthase and glycogen phosphorylase
have been extracted in an unchanged state by homogenizing freeze–
clamped tissue with a buffer containing EDTA and KF125 as part of the
study mentioned in Chapter 4.4, Fig. 4.4.

• possible effects from loss of intracellular structure by homogenization

and dilution. Enzymes are often assayed at concentrations thousands of
times lower than their concentration in the cell. The assay environment
is a simple aqueous solution. In the cell, enzymes are in a protein–rich
environment, and may bind to other enzymes, membranes, glycogen
particles or cytoskeletal elements. Investigations are still continuing
into how common such interactions are and whether they have major
implications for enzyme activity.

2.5 Measurement of metabolic fluxes

Measuring fluxes is a problem that can be either very easy or very hard.
The easy part is measuring the fluxes of uptake of materials by the metabolic
system or of output of end products, since these can be determined by the rates
of disappearance or appearance of the relevant compounds. Of course, it may
be necessary to look carefully to find all the products. For example, if yeast
cells are metabolizing glucose, measurement of ethanol production and either
oxygen consumption or CO2 production will define the rates of anaerobic
fermentation to ethanol and aerobic catabolism. These will probably account
for a large part of the glucose consumption, but other minor end products
could include glycerol, storage compounds such as trehalose and glycogen,
and yeast cell biomass if the cells are growing, all of which could be measured
to provide a balance sheet for the fate of glucose.
The more complicated problem is determining metabolic fluxes in the in-
ternal parts of metabolism, where the branches and cycles create multiple
routes between an input and output metabolite. Sometimes it is possible to
40 2 Methods for studying metabolism and its regulation

Fig. 2.4: Ion exchange HPLC of 14 C–labelled metabolites in rat hepatocytes.

Chromatograms of extracts from cells incubated with [U– 14 C]glucose. The broken lines
show the positions of 3 H–labelled markers. LACT, lactate; G1P, glucose–1–P; αGP,
3–phosphoglycerol; F6P, fructose–6–P, PGA, 3–phosphoglycerate; FDP, fructose–1,6–
bisphosphate, and UDPG, UDP–glucose. From Fig. 3 of Katz et al. , J. Biol. Chem.
253, 4530–4536, (1977).

infer the pathway usage by calculating a pattern of pathway fluxes that ac-
counts for the mass balances between the inputs and outputs and ensures that
all the internal metabolites (including coenzymes such as NAD+ ) have rates
of formation equal to their rates of consumption. However, this does require
very extensive and accurate measurements of all the inputs and outputs of the
system, as well as full knowledge of the pathways that are potentially active.
The most common method of determining internal fluxes experimentally
is by isotopic labelling. The system under investigation is incubated for a
while to allow it to reach a metabolic steady state. One or more of the input
2.5 Measurement of metabolic fluxes 41

metabolites is then replaced by material in which one or more atoms have

been labelled by substitution with a rare isotope of the element (e.g. 2 H or
3
H for 1 H, 13 C or 14 C for 12 C, 15 N for 14 N, 32 P for 31 P). Isotopically labelled
compounds are metabolized at essentially the same rate as the natural com-
pounds. (Although there can be slight differences, they will be too small to
be significant in these experiments, except perhaps in the case of hydrogen
isotopes.) If the pathway is at steady state before the a small amount of label
is added, the spread of label follows simple linear kinetics, even though the
enzyme kinetics of the steps involved are generally non–linear. Fluxes can be
calculated from measurements of the degree of labelling of particular metabo-
lites (their specific activities) against time. In the past, the calculations tended
to use simple approximations, based on assumptions about the relative rates
of diferent parts of metabolism. Nowadays, best practice is to use a computer
to find the best–fit solution to the results.19 For example, Jacob Blum’s group
at Duke University, North Carolina have determined networks of component
metabolic fluxes in the catabolic pathways of the ciliate Tetrahymena pyri-
formis 19 and the gluconeogenic pathways of rat hepatocytes.10, 58, 190 Barbara
Wright’s group in Montana have built up an extensive picture of the carbo-
hydrate and amino acid metabolism during differentiation of the slime mold
Dictyostelium discoideum by computer modelling of their isotopic tracer stud-
ies.130, 280
More information about fluxes, especially of internal, cyclic pathways, can
be obtained from the experiments if the rates of spread of the isotope into
individual positions in the compound are measured, and not just its overall
activity.
The simplest and cheapest method of measurement is of the radioactivity
associated with unstable isotopes such as 3 H and 14 C. The metabolites to be
analyzed must first be separated from other compounds carrying the isotope,
and the total amount of labelled and unlabelled compound must be measured
by conventional methods so that the degree of labelling (the specific activity)
can be determined. An example of the separation by ion exchange HPLC and
radioactive counting of glycolytic metabolites in an experiment128 to mea-
sure fluxes in glucose metabolism in rat hepatocytes is shown in Fig. 2.4. If
the position of the isotopic label in the compound is to be determined, the
compound must be broken up by enzymic or chemical means and the radioac-
tivity of the fragments measured. For example, Blum’s group measured the
radioactive labelling in the C–1 position of glucose in their studies on gluco-
neogenesis in rat hepatocytes by enzymically cleaving this carbon atom from
a sample of the glucose using the enzymes hexokinase, glucose–6–phosphate
dehydrogenase and 6–phosphogluconate dehydrogenase. The carbon atom is
released as CO2 that is collected and has its radioactivity counted.58
Isotopic tracer studies can be carried out with stable (i.e. non–radioactive)
isotopes if a mass spectrometer is used to make the measurements. A major
advantage of using stable isotopes rather than radioactive ones is that they al-
low metabolic and nutritional investigations that would otherwise be unethical
to be carried out safely on humans. Mass spectrometry separates molecules
42 2 Methods for studying metabolism and its regulation

and molecular fragments on the basis of molecular mass with sufficient sensi-
tivity to distinguish molecules containing the rare isotope from those contain-
ing the natural one, so that the relative proportions of labelled and unlabelled
molecules can be measured in the same spectrum. Often the mass spectrom-
eter is used as a detector after a chromatographic separation of the sample
by gas liquid chromatography or HPLC, though it is not essential to purify
the target compound completely since the mass spectrum will separate the
impurities. Furthermore, since compounds break down in a reproducible way
into fragments in the mass spectrometer, the extent of labelling at particular
positions in the molecule can be determined from the mass spectrum of the
fragments. For example, the relative fluxes of gluconeogenesis and glucose
phosphorylation in rats after 17 hours fasting were estimated by the relative
rates of incorporation of deuterium (2 H) from heavy water ([2 H]H2 O) onto the
C–6 and C–2 carbon atoms of glucose respectively.227 Mass spectrometry, like
radioactivity measurements, has much higher sensitivity and precision than
NMR measurements of isotopic labelling. The major disadvantages compared
with radioisotope studies are the much higher cost and greater complexity of
the apparatus.
Nuclear magnetic resonance can be used to detect the spread of 13 C
through a pathway. Since individual carbon atoms give separate lines in
the spectrum, the position of labelling in the molecule is determined at the
same time (Fig. 2.5). As mentioned previously, the method can be applied to
live, metabolizing specimens, though the low sensitivity of detection of 13 C
means that it takes a long time to collect a measurement — a clear disadvan-
tage when the degree of labelling is changing. For these purposes, it is better
to collect samples as for other types of metabolic experiment and then take
the spectra of the extracts. Then the measurement time is unimportant, and
better accuracy can be achieved. The results are best interpreted by fitting
to a computer model of the metabolism. For example, citric acid cycle fluxes
in perfused rat hearts were measured and interpreted in this way by Chance
et al.36 in the experiment illustrated in Fig. 2.5. The kinetics of the pentose
phosphate pathway in human red blood cells were also analyzed by 13 C NMR
by Kuchel’s group in Sydney, Australia, combined with computer modelling of
the enzyme reactions involved.17 Computer modelling is not always necessary,
as in Sheila Cohen’s measurements of the relative rates of the pyruvate kinase
reaction to gluconeogenic flux in perfused livers of normal and diabetic rats
using in vivo NMR studies of the redistribution of label from the substrate
[3– 13 C] alanine.46
NMR can also measure the rates of individual steps in metabolism in vivo
in favourable circumstances by magnetic labelling using a technique called
magnetization transfer. The sample is irradiated with radio waves of the
specific frequency that is absorbed by certain atoms of a particular compound,
for example the terminal phosphate of ATP. This causes the magnetic axis of
the nuclei to change their alignment in the magnetic field of the instrument,
and if the irradiation is strong enough, most of them will become aligned
in the same direction. If one of these atoms is transferred by a reaction
2.5 Measurement of metabolic fluxes 43

Fig. 2.5: 13 C NMR spectrum of a perchloric acid extract of rat heart perfused
with glucose and [2-13 C]acetate.
GluC2 , (GC2 ) etc are the groups of lines corresponding to carbon atoms 2, 3 and 4 of
glutamate (i.e. oxoglutarate), taken from Fig. 3, Chance et al. 36 The spectrum is the
average of 2000 scans collected over more than an hour. T 1 and T2 are carbon atoms
1 and 2 of naturally abundant taurine.

into another chemical grouping, it takes its alignment with it for a while,
until it spontaneously flips back. If enough of the atoms are transferring to
this other compound at a rate comparable to the rate at which the magnetic
alignment is lost, they affect the spectral line recorded from the new compound
in a manner that allows the rate of transfer to be recorded. There are a
number of ways the measurements can be taken, but one of the simplest
measures the rate at which the atoms transfer from the compound whose
spectrum is being measured to the compound being irradiated. (Perhaps
rather counterintuitively, since at first sight it looks as though it is the transfer
44 2 Methods for studying metabolism and its regulation

in the other direction that is being measured.) Unfortunately, this potentially

powerful technique is limited in its applicability by the sensitivity of NMR in
general, and also by the time window in which the technique operates, for it
can only detect processes that occur within a fairly narrow band of reaction
rates. For example, Kevin Brindle measured the rate at which inorganic
phosphate was being incorporated into ATP in yeast in vivo by irradiating
the γ-phosphate of ATP and measuring the spectrum of phosphate.22

2.6 Summary
1. Methodologies exist to study metabolism in systems at all levels of or-
ganization:

• whole multicellular organisms;

• isolated tissues and organs;
• tissue slices;
• isolated cells;
• permeabilized cells;
• cell–free systems;
• isolated organelles, and
• isolated enzymes.

2. Measurement of metabolite concentrations can be demanding because

of the low values and the complex mixtures that have to be analyzed.
Even then, there are further problems of experimental design and in-
terpretation in order to ensure the results are representative of in vivo
concentrations at the site of interest. These include:

• the need to prevent changes in metabolite levels during preparation

and analysis of samples;
• the need to relate the measured metabolites to known cellular and
subcellular compartments of known volume, and
• the difficulty of testing whether the metabolite exists mainly in the
free state or not.

3. The measurement of enzyme activities also raises difficult issues about

the relevance of in vitro measurements to in vivo conditions.

4. Measurement of overall metabolic fluxes is often a simple analytical

problem of measuring rates of change of input and output metabolites.
Measuring internal fluxes, however, can require ingenious use of isotopic
tracer methodologies, using detection by radioactivity, NMR or mass
spectrometry.
2.6 Summary 45

Further reading
1. Denton, R. M. & Pogson, C. I. Metabolic Regulation Chapman Hall
(1976).

2. Wilson, K. & Walker, J. (eds) Principles and Techniques of Practical

Biochemistry 4th ed. Cambridge University Press, Cambridge (1994).

3. Fraenkel, D. G. Genetics and intermediary metabolism, Ann. Rev.

Genet. 26, 159–177, (1992).

4. Bergmeyr, H. V. (ed) Methods of Enzymatic Analysis, 3rd ed. Vols. 1

& 2, Verlag Chemie, Weinheim (1983).

5. Zuurendonk. P. F., Tischler, M. E., Akerboom, T. P. M., van fer Meer,

R., Williamson, J. R. & Tager, J. M. Rapid Separation of Particulate
and Soluble Fractions from Isolated Cell Preparations. Methods in En-
zymology, 56, 207–223, (1979).

6. Badar–Goffer, R. & Bachelard, H. Metabolic Studies using 13 C NMR

Spectroscopy in Essays in Biochemistry, Tipton, K. F. (ed.), Vol. 26,
105–119 Portland Press, London (1991).

7. Cohen, S. M., In Vivo Studies of Enzymatic Activity. Methods in Enzy-

mology, 177, 417–434, (1989).
8. Cohen, J. S., Lyon, R. C. & Daly, P. F., Monitoring Intracellular Met-
abolism by NMR. Methods in Enzymology, 177, 435–452, (1989).
46 2 Methods for studying metabolism and its regulation
 3

Enzyme activity: the

molecular basis for its
regulation

3.1 Introduction
A description of metabolic regulation at the molecular level must include an
account of how the rates of the enzyme–catalyzed reactions vary with the
concentrations of the metabolites in the cell. The metabolites that most
obviously affect any enzyme are its substrates, and the best–known model for
enzyme action on a substrate gives rise to the Michaelis–Menten equation.
The dominance of this model in the context of metabolism is unfortunate,
for, as I will show, it is over–simplified for our purposes and can too easily
be misapplied. This chapter starts by reviewing Michaelis–Menten kinetics
and then presents some of the other aspects of kinetics relevant to metabolic
regulation, in particular:

• the effects of a product on enzyme rates;

• the kinetics of enzymes with two substrates;

• binding of metabolites by enzymes, and

• the kinetics of enzymes whose rates are affected by metabolites other

than their substrates and products.

3.2 Michaelis–Menten kinetics

The Michaelis and Menten model of enzyme action was developed to describe
how the initial rate of reaction of an enzyme–catalyzed conversion of a single

47
48 3 Enzyme activity: the molecular basis for its regulation

substrate to product(s) varied with substrate concentration. ‘Initial rate’ is

used here to specify that this is the rate at the beginning of the reaction before
substrate has been depleted and product started to accumulate; therefore it is
not necessary to consider the possibility that the product binds to the enzyme,
far less that catalysis of the reverse reaction occurs. There are two aspects of
the model that limit its applicability to metabolism in a cell:

• the majority of enzymes have more than one substrate (it is only amongst
the hydrolases and isomerases that single substrates are the norm), and

• an enzyme in a metabolic pathway is always operating in the presence

of its product(s), which are generally the substrates for other enzymes
in the cell.

Nevertheless, the Michaelis–Menten model, and the later derivation of the

same equation by Briggs and Haldane using a steady–state method, intro-
duce a number of useful concepts. Most importantly, the model does describe
the key outcome of initial rate studies on the variation of the rate of reac-
tion with substrate concentration: the rate depends non–linearly on substrate
concentration (in contrast with the usual linear dependence seen in chemical
kinetics) and exhibits saturation (Fig. 3.1) at high substrate concentrations.
The Michaelis–Menten equation that successfully describes such a curve is:

SV
v= (3.1)
S + Km

where V is the limiting rate (or maximum velocity, Vm ) and Km is known as

the Michaelis constant. The original derivation of this equation by Michaelis
and Menten in 1913 depended on the assumption that the enzyme (E) and
substrate (S) reacted together to form an enzyme-substrate complex (ES)
that in turn broke down to form free enzyme and product (P ), all according
to the laws of ordinary chemical kinetics, i.e. :
Scheme 3.1

1 k k2
E + S −→
←− ES −→ E + P
k−1

They assumed that the reactions between E, S and ES were so rapid that the
process was virtually at equilibrium (the rapid equilibrium assumption), with
the position of equilibrium essentially undisturbed by the much slower reaction
whereby ES broke down to E and P . If we draw an analogy with the problem
of understanding what determines the rate of a whole metabolic pathway,
then Michaelis and Menten were effectively declaring the breakdown of ES
to product to be the rate–limiting step of catalysis. Under these conditions,
Km in Eqn. [3.1] is equal to the equilibrium constant for the dissociation of
ES to E and S.
3.2 Michaelis–Menten kinetics 49

1.0

Rate, v
0.5

S = Km

0.0
0 5 10

Substrate concentration, S

Fig. 3.1: The dependence of enzyme rate on substrate concentration.

The Km and V have arbitarily been set to 1 for the purposes of illustration.

Briggs and Haldane later showed that the same equation could be derived
without any assumption about the relative rates of the reactions or their close-
ness to equilibrium. They used what they called the steady state assumption;
that is, they assumed that the rate of formation of the ES complex becomes
equal to its rate of breakdown shortly after mixing the enzyme and substrate
(c.f. Chapter 1.4.1). In Scheme 3.1 above, there is a single reaction form-
ing ES, which by the normal laws of chemical kinetics takes place at a rate
vf = k1 E S (i. e. rate equals a rate constant times the concentration of each
reactant). There are two routes of breakdown, one to E and S, the other to
E and P ; thus the breakdown rate, vb , is given by vb = k−1 ES + k2 ES. By
equating the rates of formation and breakdown, rearranging the equation and
making a few substitutions, the Michaelis–Menten equation can be derived, as
shown in most biochemistry texts. The effects of their change in assumptions
relative to those of Michaelis and Menten is buried in the interpretation of
Km : in the Michaelis-Menten derivation, Km equals the dissociation constant
for ES, or k−1 /k1 ; in the Briggs-Haldane derivation, Km = (k−1 + k2 )/k1 .
I have highlighted the distinction between these two derivations to empha-
size that neither of them is a factual description of enzyme activity; they are
conceptual and mathematical models. Any particular experimental observa-
tion that the rate of an enzyme-catalyzed reaction depends on the substrate
concentration in the way predicted by the Michaelis–Menten equation merely
demonstrates that either model is an adequate mathematical description of
the phenomenon for our purposes. We must look to other lines of evidence
(such as the existence of enzyme-substrate complexes) to support the molec-
50 3 Enzyme activity: the molecular basis for its regulation

ular details of the theory. Another characteristic of the models is that they
are mathematical approximations; both derivations assume that the substrate
concentration is very much higher than that of the enzyme, which is generally
true in laboratory experiments, but not always true in the intracellular envi-
ronment. Additionally, it is clear in the case of the Briggs–Haldane derivation
that the equation is an approximation because the steady state assumption
is not true on very short time scales (before the steady state concentration
of ES builds up). Indeed it is only approximately true the rest of the time
because there is a continual slow decline in the amount of ES, correspond-
ing to the gradual slowing down of the rate of reaction as the substrate is
exhausted, which in turn means that the rate of formation of ES must be
very slightly less than its rate of breakdown. Generally we can expect that
the errors introduced by these approximations will be much smaller than the
inherent observational inaccuracies in our experiments. This should not make
us forget that we have chosen to use a simple model that is adequate for our
purposes rather than a more elaborate one that would accurately reflect the
underlying molecular mechanisms. It is also worth noting that the derivation
involves the same assumptions of a time hierarchy of events that is necessary
in considering metabolic pathways, along with the same decision to ignore
events on very short and very long timescales (c.f. Chapter 1.3, p. 8).
For a more realistic molecular model of enzyme action, we could take into
account the reversibility of enzyme catalysis: the enzyme can catalyze the
conversion of product to substrate, so it must be able to form an EP complex
as well as an ES complex. The pathway from ES to EP passes through a
transient, intermediate form, the transition state ET , so if we also included
this, in initial rate conditions where product is absent, Scheme 3.1 would
become:
Scheme 3.2

1 k 2 3 kk4 k
E + S −→
←− ES −→
←− ET −→
←− EP −→ E + P
k−1 k−2 k−3

There is a lot more work involved in solving the initial rate equation, and the
result is to some extent disappointing because it turns out to be in the same
form as the Michaelis–Menten equation, except that Km is now an expression
involving all seven rate constants.

3.2.1 The measurement of Km and V

The advantage of describing an enzymic reaction by the Michaelis–Menten
equation is that it is possible to predict the rate at any substrate concentration
just by knowing the two parameters Km and V . The equation corresponds
to a rectangular hyperbolic relationship between the rate and the substrate
concentration (Fig. 3.1) with a limiting value, or asymptote, at infinite con-
centration corresponding to the limiting rate, V . The substrate concentration
giving a rate half this limiting value is the Km . However, plotting the results
3.2 Michaelis–Menten kinetics 51

in this form is not a very practicable method for determining the parameters
because an enzyme kinetics experiment does not give the curve directly; in-
stead it gives a small set of rate measurements made at particular starting
concentrations of substrate. Furthermore, these discrete points on the graph
include experimental errors which move them off the curve, and it is not pos-
sible to sketch an accurate rectangular hyperbola that passes as closely as
possible to the points.
In most branches of science until recently, the usual solution to the problem
of analysing a non–linear relationship like the Michaelis–Menten equation was
to rearrange the equation to give a linear relationship, so that the points could
be plotted as a graph and the best straight line drawn through them. It is
very unfortunate that the most popular linearization for enzyme kinetics has
been that proposed by Lineweaver and Burk:
µ ¶
1 Km 1 1
= + (3.2)
v V S V
This corresponds to the equation of a straight line, y = mx + c, with y = 1/v,
x = 1/S, the gradient m = Km /V and the y-intercept c = 1/V (Fig. 3.2).
The problem with this method is that using the reciprocals of v and S causes
extreme stretching of the axes at low substrate concentrations. This is il-
lustrated in Fig. 3.2 by showing what happens to a set of equally–spaced

5
1/v

0
0 1 2 3 4 5

1/S

Fig. 3.2: The Lineweaver–Burk plot for results of enzyme kinetics.

The Km and V have again been set to 1. The substrate concentrations are equally
spaced. The vertical bars show the range where an experimental point might be found
because of errors assumed proportional to the value of the rate.
52 3 Enzyme activity: the molecular basis for its regulation

S/v 4

0
0 1 2 3 4 5

Fig. 3.3: The Hanes plot for results of enzyme kinetics.

The Km and V have again been set to 1. The same equally–spaced substrate concentra-
tions are shown as in Fig. 3.2. The vertical bars show the range where an experimental
point might be found because of errors assumed proportional to the value of the rate.

concentration values. The experimental errors in the rate determinations are

differentially amplified in the same way, so that small errors in the slow rates
at low concentrations give a much bigger displacement of the point than an
equivalent error at high concentrations. The line that is drawn through a set
of actual experimental points is therefore unduly influenced by the chance
placing of the low concentration results, which makes this method perform
consistently badly in comparison tests along side other methods. The prob-
lem is not merely one of a poor subjective choice of line by the experimenter;
using linear regression to calculate the best–fit line does not help unless the
points are heavily weighted to compensate for the bias.
The best results with graphical linearization methods come from the plot
of S/v against S (the Hanes or Woolf plot) according to the following rear-
rangement of the Michaelis–Menten equation:

S 1 Km
= S+ (3.3)
v V V
This equation corresponds to a line of slope 1/V , with y–intercept Km /V
and x-intercept of −Km (Fig. 3.3). There is no distortion of the substrate
concentration spacing, since this is used unaltered as the x–axis, and the
errors in the velocity measurements are not unduly exaggerated or suppressed
in any part of the graph. The use of linear regression to determine the best-fit
straight line is therefore justifiable. This graph can equally as well be used
3.2 Michaelis–Menten kinetics 53

to diagnose inhibition effects as the Lineweaver–Burk plot. For example, a

competitive inhibitor (which increases the apparent Km of the enzyme for
its substrate but leaves the limiting rate unchanged) gives a parallel line on
the plot above that of the uninhibited enzyme. Because of the favourable
characteristics of this plot, it will be used later in the chapter for the analysis
of two–substrate enzyme kinetics.

1.0

0.8
2

0.6
Rate, v

0.4

0.2

0.0
0 1 2 3 4

Substrate concentration, S

Fig. 3.4: Computer fitting of enzyme kinetics results.

Curve 1 is the rectangular hyperbola predicted by the initial estimates of K m and V.
The arrows represent the distances from the observed rates to this curve. As explained
in the text, new estimates of the parameters are made giving the better curve 2.

There are two other approaches to determining Km and V that illustrate

other issues in the fitting of equations to experimental results. The first, made
possible only by the availability of computers to perform the large amount of
repetitive calculation, is direct fitting of the equation to the experimental
results, as suggested by Wilkinson275 in 1961. This avoids the distortion
involved in generating a linear graph from a non–linear function. Some ob-
jective criterion of the best fit to the results has to be used. Often, just as
in linear regression for fitting straight lines to results, the chosen criterion is
that sum of the squares of the distances of the points from the curve has been
minimized (see Fig. 3.4) but unlike linear regression, the calculation cannot
generally be done in one step. It is necessary to start with an approximate
estimate of Km and V . For each substrate concentration used, these estimates
are used to calculate an expected rate from the Michaelis–Menten equation,
and the difference between this expected rate and the experimental value is
taken and squared. One of a number of possible mathematical approaches is
then used to examine how these discrepancies would respond to alterations
54 3 Enzyme activity: the molecular basis for its regulation

in the estimates, and a prediction is made of a change in the values of Km

and V that would give a better overall fit to the results. However, this pre-
diction is based on an approximation, so the new values of Km and V are
used as new estimates, and the whole procedure is repeated. This iterative
improvement is continued until the suggested changes are too small to be
worth bothering with, at which point the current values are reported, usu-
ally along with estimates of their standard errors. Under most circumstances,
this method performs better than graphical methods, though there are some
pitfalls. Firstly, the procedure will attempt to continue even if the experi-
mental results do not fall on a rectangular hyperbola, so the user must take
responsibilty for checking that the fitted curve is an acceptable description
of the experimental results. Secondly, if the initial estimates are inappropri-
ate, or the results contain some points that are grossly in error (sometimes
because they have been mistyped into the computer!) the procedure may
fail to converge on a best–fit answer and get hopelessly lost. Details of the
mathematical implementation of this process are not given because of the
wide availabilty of ready-written computer packages, both freely-distributed
and commercial, that carry out non–linear least squares fitting, either for any
equation in general or for enzyme kinetics in particular.
The final approach is the direct linear plot method of Eisenthal and Cornish–
Bowden.65 Although this was devised as a simple graphical method, there are
computer implementations of it. This will not be described in detail because
its application to two–substrate kinetics is not widely used. However, in
essence it consists of taking all possible pairs of measurements. Suppose the
pair is v1 measured at substrate concentration S1 and v2 at S2 ; then these
values can be considered as a pair of simultaneous Michaelis–Menten equa-
tions that can be solved for Km and V . The graphical version is a simple
method of solving the equations. The estimates of Km and V from all possi-
ble pairs are then taken in order, and the median values taken as the result.
The significance of this is that the method does not make any assumptions
about the nature or type of error in the measurements, whereas least–squares
methods are implicitly based on the assumption that the errors are normally
distributed. In addition, the direct linear plot is less sensitive to occasional
very bad points, or outliers, which unduly influence the results obtained by
other methods. Outliers will arise occasionally in any experiment, just by a
rare chance conspiracy of all the contributing factors; everyone feels the temp-
tation to delete them, but it is difficult to justify this on statistical grounds.
The optimal design of an enzyme kinetics experiment for determining Km
and V has been investigated with a slightly surprising result.67 If it is already
known that the enzyme obeys the Michaelis–Menten rate law, then the best
design is to repeatedly measure just two points. One should be at as high a
substrate concentration as possible to give a rate close to V . The best value for
the other should be below Km , but depends on how the measurement errors
vary with the rate. Since the latter is often not known, Athel Cornish–Bowden
recommends using a fifth of the expected Km value.49 If the experiment must
also assess whether the Michaelis–Menten equation is applicable, then a range
3.2 Michaelis–Menten kinetics 55

of substrate concentrations should be used covering over 75% of the velocity

range. This requires that the highest velocity is at least 16 times the lowest
observed velocity, and will require a range of substrate concentrations from
less than 1/10th of the suspected Km to more than 4 times the Km .

3.2.2 Product inhibition

Most measurements of enzyme kinetics are based on initial rate measurements,
where only substrate is originally present, and the measurement is completed
before much product has accumulated. On the other hand, all enzymes in
cells operate in the presence of their products, and the lack of information
about the effects of product on the enzyme rate frequently limits our ability
to predict the behaviour of metabolic pathways. For reasons to be discussed
later (Chapter 6.2, p. 169 and Chapter 6.3.4, p. 192), even apparently weak
product inhibition can have a significant influence on pathway behaviour and
causes significant differences compared with the less probable case that there
is no inhibitory effect at all.
There are two contributions to the effect of product on an enzyme reaction.
Firstly, an enzyme is a catalyst, and cannot change the equilibrium constant
for the conversion of substrate to product, which is governed by the difference
in free energies between them. Therefore, since it catalyzes the conversion of
substrate to product, it must also be a catalyst for the conversion of product
to substrate; affecting one of the rates without affecting the other would
change the position of equilibrium. Secondly, since the enzyme can act as a
catalyst of the reverse reaction, product must presumably bind at the active
site as a first step. Therefore, even for a reaction with a large equilibrium
constant in favour of product formation, where the rate of the reverse reaction
from product to substrate will be relatively very small, there could still be a
significant inhibition of the enzyme by the product that is bound at the active
site, blocking the binding of the substrate. This inhibition is likely to be of
the competitive type, at least for single substrate enzymes.
The rate equation for a single substrate enzyme in the presence of its prod-
uct can be derived in a similar way to the usual Michaelis–Menten equation
to give:
¡ ¢¡ ¢
Vf /Km,S S − P/Keq
vnet = (3.4)
1 + S/Km,S + P/Km,P

where vnet is the net rate (positive for formation of P ), Vf is the limiting rate
in the forward direction, Km,S and Km,P are the Km values for substrate and
product and Keq is the equilibrium constant for the reaction. (This equation
can be found in other forms in some books because there is also a Vr for
the limiting rate in the reverse direction; however one of the parameters can
always be eliminated because of the constraint imposed by the equilibrium
constant known as the Haldane relationship: Keq = Vf Km,P /(Vr Km,S ).) An
illustration of the way the net rate of the reaction varies with both S and P is
shown in the three-dimensional plot of Fig. 3.5. The exact details of the form
56 3 Enzyme activity: the molecular basis for its regulation

would depend on the relative values of the constants and the equilibrium con-
stant. It is clear though, that once both substrate and product are present,
the response of an enzyme cannot necessarily be visualized by thinking about
where, on a rectangular hyperbola, the substrate concentration would fall rel-
ative to the Km,S . Later we will look at how we can characterize the position
on this 3-D graph that corresponds to particular intracellular concentrations
of S and P when we look at the concept of elasticity (Chapter 5.3). For the
moment, note how the diagram shows that a substrate concentration above
the Km value does not guarantee that the enzyme rate is approaching V ;
that will depend on the product concentration and the equilibrium constant.
An example where this has been overlooked is at the enzyme aldolase in the
glycolytic pathway. In certain cells, the concentration of substrate (fruct-
ose 1,6 bisphosphate) is greater than the Km , and this has been erroneously
interpreted to suggest that the enzyme is working near to its limiting rate.

7.5
Rate

5.0
2.5
0.0
-2.5

10
10 8
8 6
P 6 4 S
4 2
2 0
0

Fig. 3.5: Simultaneous dependence of enzyme rate on both substrate and product.
The parameters in Eqn. [3.4] have been set to: Km,S = 1; Vf = 10; Km,P = 2, and
Keq = 4.

3.3 Two–substrate enzymes

Approximately three quarters of all enzymes have two substrates, given that
coenzymes, such as NAD and ATP, count as substrates for this purpose.
The proportion of two substrate enzymes would be higher if water were to
be counted as a substrate, for it is a reactant in the reactions catalyzed by
the hydrolases such as trypsin and β–galactosidase, but it is not usual to
do so because of its very high concentration for enzymes that work in an
3.3 Two–substrate enzymes 57

aqueous environment. Besides, it is difficult (though not impossible) to vary

its concentration. Some enzymes have three substrates, and many have a
requirement for a cofactor, particularly a metal ion, that affects the rate in
a similar way to a substrate. Not all metal cofactor requirements involve a
direct interaction of the metal and the enzyme; kinases typically require Mg 2+
for activity, but this is because their true substrate is an ATP-Mg complex
(ATP4− .Mg2+ ) that forms spontaneously and reversibly between ATP and Mg
at physiological pH. Thus counting the number of substrates and deciding
what they really are may need more scientific detective work than at first
appears. Nevertheless, whatever they happen to be, there are two substrates
for the majority of enzymes, and in this section we shall look at the kinetics
of this important group. Of course, in the cell, these enzymes work in the
presence of their products (of which there will also be two in very many cases),
but I will not deal with the effects of products in this section because this
introduces unnecessary complications. Let it suffice to say that the enzyme
will catalyze a forward and a reverse reaction, and the numerator of the rate
equation will contain a positive and a negative term, similar to the single
substrate case (Eqn. [3.4]).
The obvious kinetics experiment to perform when a two–substrate en-
zyme is first investigated is to keep one substrate constant and to vary the
concentration of the other. The most common result from such an exper-
iment is that the enzyme behaves just like a single–substrate enzyme, and
the rate of reaction plotted against the concentration of the varied substrate
gives a hyperbolic curve. Indeed, it is usually possible to calculate a corre-
sponding Km and V , although these will generally turn out to be dependent
upon the concentration chosen for the other substrate. Nevertheless, the sim-
ilarity of the behaviour to that expected from the Michaelis–Menten model
suggests that the formation of enzyme–substrate complexes is involved. This
immediately poses a problem: it is quite easy to think of several ways that
enzyme–substrate complexes could form for two–substrate enzymes. Suppose
that the reaction is:
Scheme 3.3
A + B ←−
−→ P + Q
Even though we are only considering the reaction from left to right, in the
absence of products, the following might all be possibilities:
• a three body collision
• a compulsory order mechanism
• a random order mechanism
• a double–displacement mechanism
Of these, a three body collision is unable to explain catalysis because the
simultaneous collision of the enzyme molecule and the two substrates A and
B would be a relatively rare event.
58 3 Enzyme activity: the molecular basis for its regulation

The compulsory order mechanism would occur when the enzyme molecule
has a binding site for one of the substrates (say A), but not the other; when
this substrate binds, a conformational change in the enzyme (as proposed in
Koshland’s induced fit model of enzyme action) generates the binding site for
the second substrate, B. The ternary complex EAB therefore forms by an
obligatory sequence and converts to the ternary product complex, EP Q, by
the catalytic action of the enzyme. The product complex then breaks down
to release products, thus:
Scheme 3.4

E+A ←−
−→ EA
EA + B ←−
−→ EAB
EAB ←−
−→ EP Q
EP Q −→ E+P +Q

The random order mechanism also forms the ternary complex EAB in two
stages, but occurs when the enzyme has binding sites for both substrates so
that either one may bind before the other (though not necessarily with equal
probability).
Scheme 3.5
EA
%. -&
E EAB
-& %.
EB
EAB ←−
−→ EP Q
EP Q −→ E+P +Q

The double displacement mechanism differs from the previous two in that
a ternary complex is not formed. Instead, a chemical entity that is transferred
between the two substrates is temporarily parked on the enzyme molecule by
the first substrate, and then collected by the second substrate to complete the
catalytic cycle:
Scheme 3.6

E+A ←−
−→ EA
EA −→ EX + P
EX + B ←−
−→ EXB
EXB −→ E+Q

The scheme shows that B and P are related in that they both must have an
attachment point for X, and A and Q are related in that they both consist
of something linked to X (A = P + X and Q = B + X).
3.3 Two–substrate enzymes 59

Similar strategies and assumptions to those used in the derivation of the

Michaelis–Menten equation are used to obtain the equations that describe
how the initial rate of a two–substrate enzyme varies with the concentrations
of the substrates, though the algebra becomes more lengthy. As a result,
methods were devised to simplify the derivation, such as the graphical King–
Altman method133 (details of which can be found in some of the books listed
in the further reading at the end of the chapter). To a first approximation,
all three mechanisms lead to equations that are particular cases of a general
equation of the form:

V AB
v= (3.5)
KiA KB + KB A + KA B + AB
The meanings of the different parameters in this equation are as follows. V
is, just as for single substrate enzymes, the limiting rate approached as the
concentrations (but this time of both the substrates) increase to very high
values. Suppose the concentration B is fixed at a very high level and A is
varied; Eqn. [3.5] above can have both top and bottom lines of its right hand
side divided by AB to give:
V
v= (3.6)
KiA KB KB K
+ + A +1
AB B A
We can suppose that B can be raised high enough that the terms in the
denominator of the equation that are divided by B become very much smaller
than the others, so that we can ignore them and get the approximation:
V VA
v≈ = (3.7)
KA KA + A
+1
A
This is the same as the single substrate Michaelis–Menten equation (Eqn. [3.1]),
with A instead of S and KA in place of Km ; in other words, when B is fixed
at a high value and A is varied, the enzyme behaves like a single substrate en-
zyme, and KA is the Km for A under these conditions. By exchanging A and
B, we can conclude that at high concentrations of A, the rate will vary with
the concentration of B like a single substrate enzyme with KB being the Km
under these circumstances. Thus KA and KB can legitimately be regarded
as the Km values for these two substrates, except that we only know this to
be true when the concentration of the other substrate is very high, which is
unlikely to be the case in intracellular conditions for most enzymes. What
happens at lower concentrations of the other substrate? To illustrate this,
let us consider very low concentrations of B. As B decreases, the numerator
term of Eqn. [3.5] becomes smaller because the whole of it is multiplied by B;
the denominator, however, only shrinks to a certain extent. The denominator
terms mutiplied by B certainly become smaller and smaller, but the terms
without B are unaffected and, since the different terms are added together,
60 3 Enzyme activity: the molecular basis for its regulation

the denominator cannot be smaller than KiA KB + KB A, so that at very low

concentrations of B, Eqn. [3.5] approximates to:
µ ¶
B
V A
V AB KB
v≈ = (3.8)
KiA KB + KB A KiA + A

This means that the enzyme again behaves like a Michaelis–Menten enzyme
when A is varied. Comparison with Eqn. [3.1] shows that the apparent limiting
rate, Vapp is given by Vapp = V B/KB , and the Km is KiA . At intermediate
concentrations of B, it is possible to show that the apparent Km for A is
between KA and KiA . Thus although it is possible to study a two–substrate
enzyme by holding one substrate constant and varying the other, the results
have limited value. It is necessary to determine the values of all four param-
eters V , KA , KB and KiA to calculate how rapidly the enzyme will work at
any particular concentrations of A and B. How the parameters are found is
described next.

3.3.1 Experimental investigation

Since the way in which the velocity of a two substrate enzyme varies with
one substrate is likely to depend on the concentration of the other, the ex-
perimental design must involve simultaneous variation of the two substrates.
In addition, this must be done on a regular grid of combinations of the two
substrates (Table 3.1) if graphical analysis of the results is going to be used.
If a sufficiently wide range of both concentrations is used, the range of ve-
locities observed will be very large, with the lowest, v1,1 being as little as a
hundredth of the highest, v5,5 . It is not neccesary to fill the grid completely,
and if computer analysis of the results is to be used, it can be sparsely filled.
The graphical methods of analysing these results involve primary and sec-
ondary graphs: the results are plotted first against one substrate concentra-
tion, and measurements taken from that graph are plotted against the second
substrate concentration to make the secondary graphs. The method given in
many books is a variant of the Lineweaver–Burk plot; for the reasons given
earlier, this method is liable to error, and the two–substrate analogue of the
plot of s/v against s, originally described by Cornish–Bowden, is preferable 49 ).
In this case, one of the substrates is chosen as the independent variable (x-
axis) of the primary plot, say A from the table of results (Table 3.1). The
table is then rewritten with the entries consisting of A/v values, so that the
entry in the top left hand corner, for example, is A1 /v1,1 .
These values are then plotted on a graph of A/v against A, and a line
drawn through each set of points from the same row of Table 3.1, i.e. for sets
of points measured at the same value of B (Fig. 3.6). The y-axis intercept
and slope of the line is measured for each value of B. The intercepts give
values of KA,app /Vapp , where these are the apparent values of Km and V that
fit the variation of rate with A at each of the values of B. The slopes have
3.3 Two–substrate enzymes 61

Table 3.1: Two substrate kinetics:

schematic set of results.

Concn. Concentration of A
of B A1 A2 A3 A4 A5
B1 v1,1 v1,2 · · ·
A1 A2
v1,1 v1,2
B2 v2,1 ···
A1
v2,1
B3 ···

B4 ···

B5 ··· v5,5
A5
v5,5

The upper entry in each cell is the observed

rate and the lower is the derived value for
plotting Fig. 3.6.

values of 1/Vapp . Two secondary plots are then constructed. The first of
these involves plotting each slope value times the value of B at which it was
measured against B (Fig. 3.7a). The slope of this plot gives 1/V , and the y-
axis intercept is KB /V . (The x-axis intercept is −KB .) The other secondary
plot is each intercept value from the primary plot times the value of B at
which it was measured against B (Fig. 3.7b). The slope of this plot is K A /V ,
and since V has already been determined, this gives KA . The intercept on the
y-axis is KiA KB /V , and since KB /V has been determined, this gives KiA .
The values of the parameters obtained by this graphical analysis can them-
selves be reported as the properties of the enzyme. These days, they can also
be used as the initial estimates in an iterative computer fitting procedure, and
the results from this are then taken as the best estimates of the parameters.
In this case, the ultimate worth of the graphical analysis is that it gives visual
confirmation that the results do fit the two substrate equation. (If they do
not, there will be systematic deviation of the points from a straight line.)
What can the results tell us about the mechanism of the enzyme? We
need to look at this for each mechanism in turn.

3.3.2 Compulsory order mechanism

Enzymes following the compulsory order mechanism obey the general form of
the two substrate rate equation (Eqn. [3.5]). If A is the first substrate to bind,
then KiA is actually equal to the dissociation constant of the EA complex.
Without other evidence, it is not possible to tell which substrate binds first
62 3 Enzyme activity: the molecular basis for its regulation

6
....... B1
intercept
.. .
...
..........
.
........ ........ B2
...
.
..
............. .................
. .
........ .......... ......
...
.
............... ................. .................... 4 B3
. . .. .
........ .......... ................ .....................?. B4
...
.
............................. ..............4 ...
...... ............ B

A/v
. .. ...
....... ......... ............................?............................. ◦ 5
. ........ ......... ........4 .. ..
... ..................................................................?....................................◦....
. ..
.... ..
.... .....
4. . ......
. . .....
◦.... . .
....
.....................................................?............................ ...
.................................
..
......................................4
.....
..
...
.........
.. ◦
...
.
..
... . . ..
..
.. ... . . . . . . .
. .
. . . . . .
.?
◦.
.... slope=1/V
. .
. ..
.............. .............. ........... . . app
. . . .
.
..
.
.
.
.
.
..
.
.. .
. .
. ..
. -
. .
... ..........
.
.
...
..... A

Fig. 3.6: Primary plot of two substrate enzyme kinetics results.

Each line on the plot corresponds to points measured at a set value of B as in Ta-
ble 3.1. The slope (corresponding to 1/Vapp ) and intercept (KA,app /Vapp ) of each line
is measured for use in the secondary plots (Fig. 3.7).

from the initial rate kinetics experiments described above. Furthermore, the
results can equally be used to calculate a Ki term (i.e. KiB ) for the other
substrate even though, if B is the second substrate to bind, the EB complex
is not formed when enzyme and B are mixed together, so that a dissociation
constant for the EB complex cannot be measured. Therefore if a purified
sample of the enzyme is available, a comparison of the kinetic results with
binding studies (see p. 65) may show that the enzyme obeys a compulsory
order mechanism. Another experiment that can be used to diagnose this
mechanism is possible if there are a number of different substrates on which
the enzyme can act. For example, alcohol dehydrogenases will act on alcohols
with carbon chains of different lengths, though generally at different rates
with different substrates. Suppose that there are a number of substrates of
the B–type: B1, B2 etc. If the kinetics are determined using A and each of
these substrates in turn, KiA is the same whichever B substrate is used. On
the other hand, if the kinetics are studied with a number of different variants
of the A substrate, KiB will vary with the nature of the A substrate. This is
because the binding of the first substrate cannot be affected by the nature of a
substrate that has not yet bound, whereas the binding of the second substrate
is obviously influenced by the molecule already at the active site.
Aspartate transcarbamylase catalyzes the reaction:
carbamoyl phosphate + aspartate *
) carbamoyl aspartate + phos-
phate
The enzyme has some very interesting properties that will be discussed later
in the chapter in connection with allosteric enzymes (section 3.5), but its ki-
netic mechanism is ordered. Carbamoyl phosphate binds first, followed by
3.3 Two–substrate enzymes 63

aspartate. Products also leave in an ordered fashion: carbamoyl aspartate

first, then phosphate. Binding studies did detect aspartate binding to the
enzyme, though with a slightly high Kd . The binding of non–reactive aspar-
tate analogues, such as succinate and malate, has been found to be greatly
enhanced by the presence of carbamoyl phosphate.82

3.3.3 Random order mechanism

The random order mechanism gives rise to potential problems for our analy-
sis. The general two–substrate rate equation (Eqn. [3.5]) can be derived for
it by the ‘rapid–equilibrium’ method, that is, the equivalent of Michaelis and
Menten’s original method, assuming in this case that E, A and B are close
to equilibrium with EAB because the binding reactions are relatively faster
than the catalytic conversion of EAB to products. However, if no assumption
about the relative rates of reaction are made and the rate equation is derived
by the steady state assumption (as in the Briggs–Haldane method for sin-
gle substrate enzymes), a more complicated equation is obtained, containing
terms in A2 and B 2 . This equation does not necessarily predict hyperbolic
curves of rate against substrate concentration, nor straight lines in the graph-
ical analysis described above, though these are both observed as long as the
rate constants for certain steps of the mechanism have appropriate relative
magnitudes. Now there are certainly a significant minority of enzymes that
do not give hyperbolic rate curves but, in most cases, the reason for this is not

6 6
a) ¡ b)
B x intercepts

¡◦
B x slopes

?¡ ◦
¡ #
#
¡@I ?
#
@ # o
S
¡4 1/V 4
# S KA /V
¡ #
¡ #
¡ #
¡ #
¾ K /V #¾ KiA KB /V
¡ B
#
¡ - # -
6 B 6 B
−KB −KiB

Fig. 3.7: Secondary plots of two substrate enzyme kinetics results.

The point symbols are those of the corresponding line of Fig. 3.6 from which they were
derived. a) The secondary plot from the slopes gives K B and V. The units of the y–axis
are concentration/rate units. b) The secondary plot from the intercepts leads to K A
and KiA . The units of the y–axis are (concentration)2 /rate units.
64 3 Enzyme activity: the molecular basis for its regulation

that they obey a special form of the random order mechanism, but because
of cooperative binding that will be described later (p. 71). In practice, it
seems that most enzymes working by the random order mechanism do actu-
ally follow the two–substrate rate equation. One characteristic of the random
order mechanism is that binding of both substrates to the enzyme will be
demonstrable in binding studies, with the dissociation constants for the EA
and EB complexes being KiA and KiB respectively. Unlike the compulsory
order mechanism, these Ki terms would not be expected to be invariant if
different forms of the other substrate were used in the kinetic studies.
The effects of products in inhibiting the reaction can distinguish between
compulsory order and random mechanisms. Whether the inhibition by each
product with respect to each substrate is of the competitive or noncompetitive
type gives a set of results that differs between the two mechanisms.
One difficulty in distinguishing between the random order and compul-
sory order mechanisms is that in kinetic experiments, the dividing line may
be fuzzy. Although a random order enzyme may bind either substrate first,
it is unlikely that both routes will be equally probable. If one route is al-
ways dominant under the conditions investigated, then the enzyme is effec-
tively compulsory order, even though on the longer time scales of a binding
experiment, the binding of both substrates to the enzyme might be demon-
strable. This might explain why some enzymes have been difficult to classify,
with different laboratories reaching opposite conclusions. The hexokinase of
yeast is one example where contradictory conclusions were reached by dif-
ferent groups: some favoured an ordered mechanism in which glucose bound
first, whereas others thought that the mechanism was random. However,
the probable explanation is that the enzyme follows a steady state (rather
than rapid equilibrium) random mechanism, in which only the binary com-
plexes, enzyme–glucose and enzyme–ATP, are near to equilibrium with the
enzyme.202 (In these conditions, the random mechanism could show hybrid
characteristics with the compulsory order mechanism.)

3.3.4 Double displacement mechanism

The double displacement mechanism shows a clear difference from the other
two mechanisms in that the term KiA KB is not present in the denominator
of the rate equation (Eqn. [3.5]). This is because there is no ternary complex,
EAB. As a result, in the primary plot (Fig. 3.6), all the lines cross on the
y-axis at a value of KA /V .
The double displacement mechanism is common in group transfer reac-
tions, such as those catalyzed by the transaminases that transfer amino groups
between oxo–acids. The amino group is held on the pyridoxal phosphate that
is a prosthetic group in these enzymes.
Another feature of enzymes working by this mechanism is that they bring
about partial reactions in the presence of just one substrate. With small
amounts of enzyme, the amount of product formed may not be detectable
as the enzyme is unable to complete its catalytic cycle in the absence of the
3.4 Binding characteristics of enzyme sites 65

second substrate, and the net reaction stops when all the enzyme has been
converted into the modified form. If larger amounts of enzyme are available,
it may be possible to detect and characterize the modified form of the enzyme.
However, even when the amount of enzyme is small, so that the net product
formation is also small, the equilibrium between substrate and product is
being catalyzed, and it is possible to detect this reaction by isotope exchange.
Thus, suppose the partial reaction with one substrate AX is:
←− EX + P
E + AX −→

If isotopically unlabelled AX and labelled product P ∗ are added to the en-

zyme, then the reverse reaction of the above equilibrium will include:
EX + P ∗ −→ E + A∗ X
After a while, if AX is recovered from the reaction mixture, it will be found
to be isotopically labelled.
An example of both these features is given by the three–substrate, three–
product enzyme pyruvate carboxylase:
pyruvate + ATP + HCO− *
3 ) oxaloacetate + ADP + Pi

This has two partial reactions: in the first, ATP and HCO− 3 react with the
enzyme to form an enzyme–CO2 complex (in which the CO2 is attached to the
prosthetic group biotin), and in the second, the enzyme–CO2 complex reacts
with pyruvate to form oxaloacetate. The enzyme–CO2 complex has been
isolated after incubation with [14 C]HCO− 3 to demonstrate the first partial
reaction, and the exchange of [14 C] from pyruvate into oxaloacetate has been
detected to show the second partial reaction.252

3.4 Binding characteristics of enzyme sites

The aim of steady state approaches to enzyme kinetics is to avoid explicit
interpretation of the kinetics in terms of possibly invalid assumptions about
the role of the substrate binding equilibrium. Nevertheless, sometimes it is
easier to interpret an enzyme’s properties in terms of its binding equilibria
with its substrates. In some of these cases, direct measurements of the binding
equilibria are able to confirm that the kinetically measured properties are in
fact direct consequences of the binding behaviour. Before considering some
more complex enzyme kinetic behaviour, it is therefore necessary to consider
the types of binding that are observed between enzymes and substrates (or
inhibitors or products). The binding of substances (generically known as
ligands) by enzymes is, in essence, no different from the binding of molecules
by transport proteins, of hormones by receptors, or of antigens by antibodies,
except that there is the possibility of chemical reaction at the binding site.
However, for two–substrate enzymes (apart from the double displacement
type), there is no reaction if only one substrate is present, so it is perfectly
feasible to study the binding in isolation.
66 3 Enzyme activity: the molecular basis for its regulation

3.4.1 Identical independent binding sites

The simplest binding curve is given by a protein P that has just a single type
of binding site for a ligand L. If there is only one binding site per molecule
of P , the binding equilibrium is:

P + L ←−
−→ P L

Here P L is the protein–ligand complex, and its concentration is equal to

that of the bound ligand, Lb . The equilibrium constant for this reaction is
commonly expressed in terms of the dissociation constant, Kd , defined as:

[P ][L]
Kd =
[Lb ]

where Kd has the units of concentration. Furthermore, when the total pro-
tein Pt is divided equally between free protein, P , and protein–ligand complex,
P L = Lb , Kd equals the free ligand concentration [L]. Although the disso-
ciation constant has this advantage that its value corresponds to the concen-
tration of ligand needed to fill half the binding sites, it has the disadvantage
that strong binding (or high ligand affinity) corresponds to low values of K d ,
and weak binding corresponds to high values. (The equilibrium defined in
the opposite direction is the association constant, Ka ; Ka = 1/Kd , and large
values correspond to strong binding, but the units of Ka are the reciprocal of
concentration.) The equation for Kd can be rearranged with substitution of
[Pt ] = [P ] + [Lb ] to give:
[Pt ].[L]
[Lb ] =
Kd + [L]
This is analogous to the Michaelis–Menten equation (not surprisingly, as it was
the origin of their derivation) and therefore defines a rectangular hyperbolic
binding curve.
Many enzymes are composed of a number of identical subunits (most com-
monly four). Binding will involve the formation of complexes with 1, 2 . . . n
molecules of L at various stages of the binding curve:

P + L ←−
−→ P L ←−
−→ P L2 · · · ←−
−→ P Ln

If the binding sites do not interact in any way, then defining the concentration
of bound ligand, Lb as Lb = P L + 2P L2 · · · + nP Ln leads to essentially the
same equation:
n.Pt .L
Lb = (3.9)
Kd + L

Thus the binding curve can be characterized by the values of n and Kd , and
this is done by measuring Lb at various concentrations of L. Sometimes it
is easier to determine the fraction of binding sites occupied, the fractional
3.4 Binding characteristics of enzyme sites 67

2.0

1.5
b
Lb /(Pt L)
1.0

0.5 c

0.0
0.0 1.0 2.0 3.0 4.0

Lb /Pt

Fig. 3.8: Scatchard plot

(a) Four identical independent binding sites, K d = 2. (b) Two high affinity sites and
two low affinity sites. (c) Four interacting, cooperative sites.

saturation, Y , where
Lb L
Y = = (3.10)
n.Pt Kd + L
in which case it is still possible to determine Kd .
The analysis of the results of binding experiments has parallels with that
of enzyme kinetics. There are graphical methods based on the linearization of
the binding equation. For example, the Hughes–Klotz plot (an analogue of the
Lineweaver–Burk plot) consists of plotting 1/Lb against 1/Lf ; however this
has not been favoured and it suffers from the same defects as the Lineweaver–
Burk plot. The plot used most frequently is known as the Scatchard plot. It
consists of plotting Lb /(Pt .L) against Lb /Pt (Fig. 3.8) and is an analogue of
the enzyme kinetics plot of v/S against v, known as the Eadie–Hofstee plot.
The slope of the graph (line a) is −1/Kd and the intercept on the x-axis is
the number of binding sites, n.
However, the method of preference for analysing binding results is com-
puterized fitting of the binding equation. Since this is essentially the same
as the Michaelis–Menten equation, a computer program suitable for one task
may be satisfactory for the other.

3.4.2 Non–identical independent sites

Even if a protein has been purified, it is not unusual to find that it has more
than one type of binding site for a particular ligand. There are a number of
68 3 Enzyme activity: the molecular basis for its regulation

possible reasons for this. One is that there is a specific high–affinity site for a
particular ligand, but the ligand can also bind with lower affinity and speci-
ficity to another part of the protein through hydrophobic or ionic interactions.
Another possibility is that the protein sample contains a mixture of proteins
of different binding affinities, either because the protein exists in a number
of variant isoforms in vivo or because partial damage during preparation has
modified the binding properties of some of the molecules.
If the different sites do not interact in any way, they behave independently
and the total binding observed is just the sum of the binding at the separate
sites. Thus for a protein molecule with two types of site, of which there are
n1 and n2 per molecule, with dissociation constants Kd,1 and Kd,2 , the total
binding is:
µ ¶
n1 .L n2 .L
Lb = P t +
Kd,1 + L Kd,2 + L

In this case, the linear graphical plots are curved (e.g. Fig. 3.8, curve b).
There has been a general belief that the two components can be estimated by
drawing tangents to both ends of the curve on a Scatchard plot. In spite of the
many instances of this continuing practice in the biochemical and physiological
research literature, it has long been known that such a procedure gives grossly
incorrect estimates of the two sets of parameters.134, 176 For example, no part
of curve b in Fig. 3.8 can be extrapolated to 2 on the x–axis — the correct
value of the number of higher–affinity sites. In such cases computer fitting to
the equation above can often yield the parameters of the two components.

3.4.3 Identical interacting sites

Where proteins have a number of binding sites with an initially identical
binding affinity, there is the possibility that the binding of ligand at one site
affects the affinity of the other sites on the protein. This is referred to as
cooperativity in binding.
The most studied example of this phenomenon is the oxygen–binding pro-
tein haemoglobin from the blood of vertebrates. At the beginning of this cen-
tury, increasing accuracy in the determination of the oxygen binding curves
of blood had shown that the binding was not described by a simple binding
curve for a single type of binding site. Instead of a rectangular hyperbolic
binding curve, the curve was noticeably S–shaped, or sigmoid (Fig. 3.13).
At this time, the molecular mass of haemoglobin was not known, nor was it
known that haemoglobin was composed of four subunits. Hill proposed that
the binding curve could be explained if a haemoglobin molecule that was an
aggregate of n units bound n molecules of oxygen in a single step (at least in
the sense that a molecule either has zero or n oxygens bound, and no inter-
mediate number). The binding equation for the aggregates of size n in terms
of the fraction of oxygen binding sites occupied (the fractional saturation, Y )
3.4 Binding characteristics of enzyme sites 69

is:
O2n
Y = (3.11)
Kdn + O2n

This equation does give rise to a sigmoid curve if n is greater than 1; for

2
log [Y/(1 - Y)]

-1

-2 B

-3
A
-4

log S

Fig. 3.9: The Hill plot for cooperative binding.

A linear graphical form of the Hill equation, Eqn. [3.11]. Line A is the line predicted by
the equation. Curve B is the form usually obtained with experimental results. The Hill
coefficient, h, is the slope of the line, usually taken at the half-saturation point where
the y–axis value is zero.

n = 1, it gives the usual hyperbolic binding equation for a single type of

binding site (Eqn. [3.10]). The values of Kd and n can be determined with
a graph known as the Hill plot: if log {Y /(1 − Y )} is plotted against log O 2 ,
the equation is converted into a linear form with slope n. When O2 = Kd ,
Y = 0.5 and log {Y /(1 − Y )} = log 1 = 0 (Fig. 3.9). The oxygen binding curve
of blood generally gives a value for the slope n in the range 2.4–2.8. Hill knew
the result was not a whole number, and accounted for this by assuming that
the haemoglobin was a mixture of aggregates of different sizes, with 1, 2, 3
. . . sites. Since the slope of the Hill plot is not therefore guaranteed to give
the number of binding sites, it is usually called the Hill coefficient, nh . By the
1920s, the molecular mass of haemoglobin had been measured and it was clear
that it was a permanent aggregate of fixed size with four oxygen binding sites.
Adair proposed that the binding of oxygen to haemoglobin was a four step
process, i.e. that the intermediates Hb(O2 ), Hb(O2 )2 , Hb(O2 )3 and Hb(O2 )4
are formed successively. The curve is sigmoid because after the first site has
bound an oxygen, the affinity of the remaining sites becomes greater than
70 3 Enzyme activity: the molecular basis for its regulation

before, and so on at each stage. This requires that the binding of oxygen at
one site transmits an influence to the other sites to aid their binding; this
effect is known as cooperativity. However, for much of the century, little was
known about the molecular mechanism by which this could occur.
On the basis of energetic considerations, it was realized that the Hill equa-
tion was not a feasible description of cooperative binding, since it would re-
quire an infinite interaction energy to ensure that the affinity of all the sites
increased to such an extent that they immediately filled up after the first
oxygen had bound. With a finite interaction energy, the binding curve would
give a sigmoid curve in the Hill plot (see Fig. 3.9), having slopes of 1 at both
ends (i.e. at very low saturations and very high saturations). This is because
at very low concentrations of oxygen, only the reaction

←− HbO2
Hb + O2 −→

is being seen. Similarly, when the oxygen concentration is very high, the only
reaction being observed is

←− Hb(O2 )4
Hb(O2 )3 + O2 −→

Furthermore, the Hill coefficient could only have a maximum value equal to
the number of subunits, n, for an infinite interaction energy, and with a finite
interaction energy, nh would be less than n.
Thus the original basis of the Hill equation is no longer accepted. Nev-
ertheless, many proteins in addition to haemoglobin exhibit cooperativity in
their ligand binding, and the Hill plot is still widely used to characterize the
cooperativity. This is partly because the other forms of binding curve analy-
sis, such as the Scatchard plot, do not give linear plots for cooperative binding
and cannot therefore be used to analyse the results (e.g. see Fig. 3.8, curve
c). The Hill plot, on the other hand, tends to have a substantial linear por-
tion in the centre; the curvature towards a line with a slope of one is only
observable if the extreme ends of the curve can be measured. Furthermore,
the Hill coefficient, nh , can be used to diagnose cooperativity even when the
degree of sigmoidicity of the binding curve is difficult to detect by eye. If the
value of nh is greater than 1, the binding can be said to be cooperative (or
even positively cooperative). A value of 1 corresponds to ordinary hyperbolic
binding that should be analysable by methods such as the Scatchard plot.
However, it is possible to obtain values of nh of less than 1. For example, a
mixture of independent binding sites of different affinity gives this result. It is
also possible to conceive of the case where the first ligand to bind to a protein
causes a lowering of affinity of the other sites, so they become progressively
more difficult to fill. Since this is the opposite effect to the cooperativity
described above for haemoglobin, it is termed negative cooperativity. It gives
rise to non–hyperbolic binding curves and Hill coefficients of less than 1. The
difficulty is that the binding experiments themselves do not distinguish be-
tween these two causes for nh being less than 1. As a result there is sometimes
3.5 Allosteric enzymes 71

a dispute about which explanation is appropriate in a particular case, such as

with the binding of the hormone insulin by its cell surface receptor.

3.5 Allosteric enzymes

In 1956, the first two examples (threonine deaminase250 and aspartate tran-
scarbamylase281 ) were discovered of enzymes subject to feedback inhibition
(see Chapter 7.2.1, p. 205). These enzymes were at the beginnings of metabolic
pathways and were inhibited by metabolites produced further along. The
significance of this is that the inhibitory metabolites by that stage do not
chemically resemble the substrates or products of the enzyme, and there-
fore would not be expected to compete at the active site to cause inhibition.
Furthermore, for these examples the inhibition was highly specific, in that
analogous molecules from different metabolic pathways showed weaker or no
inhibition. Following the initial discoveries, many more instances were found,
and it became noticeable that these enzymes often did not give rectangular
hyperbolic plots of velocity against substrate concentration, so that it was not
possible to analyse their kinetics with the linear graphical methods described
earlier in this chapter. Instead, these enzymes often gave a sigmoidal plot
of rate against substrate concentration, like the binding curve of oxygen to
haemoglobin. It is possible to obtain a linear plot of their kinetics, at least
for the results from the middle of the curve, by using the Hill plot adapted
for enzyme kinetics. This consists of plotting log{v/(V − v)} against log S;
as with haemoglobin, the slope is the Hill coefficient, nh . The intercept with
the x-axis is the substrate concentration giving half the limiting rate, often
designated S0.5 .
By the time Monod, Changeux and Jacob160 reviewed the available evi-
dence in 1963, there were sufficient examples of these enzymes to justify the
following generalizations, which continue to be valid to this day:
• The enzymes are composed of a number of subunits, i.e. they are mul-
timeric. The actual number varies, but tetramers (4 subunits) are par-
ticularly common.
• The feedback inhibitor binds at a site distinct from the active site of the
enzyme. Jacob and Monod named this site an allosteric site because
the inhibitor is not a steric analogue of the substrates, and therefore the
site must have a different shape. From this term, these enzymes have
become known as allosteric enzymes. In some cases, the separation of
the active site and the inhibitor site extends to their being on different
subunits of the enzyme, giving catalytic and regulatory subunits.
• In some cases, there are activators as well as inhibitors of the enzymes.
(The general term for modifiers of enzyme activity covering both acti-
vators and inhibitors is effectors.) The occurrence of activators makes
it even more certain that the effectors bind at sites distinct from the
active site.
72 3 Enzyme activity: the molecular basis for its regulation

Rate, mmol·h-1 ·mg-1

4

0
0 5 10 15 20 25

[Aspartate], mM

Fig. 3.10: The kinetics of the allosteric enzyme aspartate transcarbamylase (AT-
Case).
The rate of reaction is shown with respect to one of its substrates, aspartate (¦). The
enzyme is inhibited in a feedback manner by the pathway end–product CTP (+). ATP
acts as an activator ( ). (Based on results of Pigiet, Yang & Schachmann. 82 )

• In the majority of cases, the enzyme rate gives a sigmoid curve when
plotted against the concentration of at least one of the substrates, cor-
responding to a Hill coefficient of greater than one. The effectors work
by changing the S0.5 of the enzyme for one of its substrates. Activators
decrease the S0.5 so that, at a given substrate concentration, the velocity
is higher in the presence of the activator (Fig. 3.10). On the other hand,
inhibitors increase S0.5 so that, at a given substrate concentration, the
enzyme rate is lower (Fig. 3.10). In both cases, the Hill coefficient is
altered by the effectors. Enzymes whose activities are regulated in this
way are classed as ‘K systems’. There is a minority class, the ‘V sys-
tems’, where the effectors alter the rate without changing the S0.5 , and
these do not necessarily exhibit sigmoid rate curves with respect to their
substrates.

The characteristics of allosteric enzymes are also shared by haemoglobin

in its oxygen binding behaviour. There is also evidence that the binding
of substrates and effectors to several allosteric enzymes is accompanied by
substantial conformational changes of the protein detectable by a variety of
physical methods. As a result, most attempts to explain how these enzymes
work have assumed that the sigmoidal rate curves reflect underlying cooper-
ativity in the substrate or effector binding. This is equivalent to making the
same assumption as Michaelis and Menten that the binding of substrate is
3.5 Allosteric enzymes 73

close to equilibrium because the rate of catalysis is relatively much slower. If

this simplification is not applied to allosteric enzymes, explanations of their
behaviour become very complicated.
There is just a handful of enzymes that show cooperative kinetics yet are
monomeric and so differ from the majority. The best known example is the
hexokinase IV from mammalian liver (also known as glucokinase51 ) which
exhibits positive cooperativity with respect to glucose. In this case, the origin
of the cooperativity is in an unusual kinetic mechanism.50

3.5.1 The concerted model

In 1965 Monod, Wyman and Changeux161 proposed a structural model of
allosteric enzymes to explain their cooperative binding properties and the
action of effectors. They assumed that an allosteric enzyme was necessarily a
protein with a number of subunits arranged in a symmetrical manner. They
developed their model on the assumption that there are two structural states,
or conformations of the subunits. The tense or T conformation is characterized
by a relatively low affinity for substrate, but is well stabilized by relatively
strong inter-subunit bonding. The relaxed or R conformation, on the other
hand, shows a higher affinity for the substrate, but is less tightly bonded.
Monod, Wyman and Changeux argued that the subunits in any one protein
molecule would be either all in the T state or all in the R state because
symmetrical structures would be intrinsically more stable than the asymmetric
ones that would result from a mixture of T and R conformations. Therefore,
when the subunits change conformation, they all change simultaneously in a
concerted manner.
Initially, in the absence of substrate, both conformations can be present,
but the T state might be expected to predominate because it is more highly
stabilized by the inter–subunit bonds. The R to T conversion has an equi-
librium constant designated L. If the state of the protein is considered at
successively higher substrate concentrations, then at first, most of the sub-
strate is binding with low affinity to the majority T state. However, with
one molecule of substrate S bound to the protein, the equilibrium between
T S and RS is somewhat more favourable for the R state because RS has
stronger, more stabilizing interactions with the substrate molecule than does
T S. If a second molecule of substrate binds, yet more substrate–binding en-
ergy is available to stabilize the R state relative to the T state. This can be
seen in Fig. 3.11 where the relative energy levels are shown for a tetrameric
enzyme. At some point in the substrate–binding curve, the R state becomes
more stable than the T state with the same number of substrate molecules
bound. Therefore, the R state comes to be the predominant form as the sub-
strate concentration increases. When the enzyme molecule changes from the
T state conformation to the R state conformation, all the remaining unfilled
substrate binding sites become high affinity sites. This increase in affinity
for the substrate after the binding of one or more substrate molecules to the
enzyme gives the characteristic feature of cooperative binding.
74 3 Enzyme activity: the molecular basis for its regulation

..........
..... ........
.... .. R
.... . .
............... .
∆G◦0 L ......... .
..... ........
.
......... ..
........ ....
..
.... .......... ... KR
... ..
... .
... T .........
... ..............
... ....... .... ......
.. . .... S
... ........KT ...... .... .. RS
.. . .................. ..................
...................
S ........
TS ....
....
........ ...
.
.
......... .........
...........
..... ......
TS2
S S ............................................ .... S S .... RS2
....
...................
.......... ........
.
......... ....
...
...
S S ..
TS3 ........
S ...................
................ ............
.
........ ...... ..... .....
. .... S S ....
........ ....S .... RS3
..............
S S .......
TS4 ....
S S ..
........... ...
...... ...
...... .
......
....... ........
.
..... ...........
..... ......
.... S S .... RS4
.... S S ..
.................

Fig. 3.11: The concerted allosteric enzyme model of Monod, Wyman & Changeux.
The T conformation is represented by squares and the R by circles. The different states
of the protein are placed vertically according to their relative standard free energy for
this specimen case. The exact relative relationships will vary with the values of the
equilibrium constants L, KR and KT .

Supplement: The Monod, Wyman & Changeux equation The equation

for the fractional saturation, Y , of an allosteric enzyme with n subunits that bind
the ligand S according to the Monod, Wyman and Changeux model is:

α(1 + α)n−1 + Lαc(1 + αc)n−1

Y = (3.12)
(1 + α)n + L(1 + αc)n
3.5 Allosteric enzymes 75

where L is the equilibrium constant corresponding to the T /R ratio in the absence

of substrate, α is the substrate concentration relative to its dissociation constant,
KR , for an R state subunit i.e. α = S/KR , and c is the ratio of the dissociation
constants for the binding of S to R and T state subunits, i.e. c = KR /KT . The
equation may look complicated, but in fact it has a simple interpretation. The
denominator is the sum of the concentrations of all the possible forms of the enzyme,
R+RS+. . . RSn +T +T S+. . . T Sn , relative to the concentration of R being arbitarily
assigned to 1. If the bracketed terms in the denominator are multiplied out, there is a
denominator term for each enzyme form. The numerator is the total concentration
of bound substrate molecules, RS + 2RS2 + . . . nRSn + T S + 2T S2 + . . . nT Sn ,
divided by n (again relative to R = 1) and again, the total number of terms in
the numerator is equal to the number of species with ligand. In other words, like
all binding equations for the fractional saturation, the equation is merely a specific
form of the general equation:
Sum of concentrations of occupied ligand sites
Y = (3.13)
n(Sum of concentrations of enzyme species)

where the denominator calculates the total concentration of ligand sites. Deriving
an equation like Eqn. [3.12] can be done by taking the diagram of the model (such
as Fig. 3.11) assigning one of the forms the concentration 1, calculating all the other
concentrations relative to it by using the equilibrium constants, and inserting these
concentrations into Eqn. [3.13].
The Monod, Wyman and Changeux equation (Eqn. [3.12]) therefore describes
cooperative binding using three parameters, L, c and KR ; it is assumed that n is
already known by other means. If c = 1 (corresponding to equal substrate affinity of
the R and T states) or L = 0 (corresponding to no observable T state) the equation
can be shown to reduce to an ordinary hyperbolic binding function like Eqn. [3.10]:

α S
Y = =
1+α KR + S
KR determines the left–right positioning of the top of the curve on the Hill plot
(Fig. 3.9). L and c determine the shape of the Hill plot curve. It can also be shown
that the maximum value of the Hill coefficient, nh , depends on the values of L and
c, but never exceeds the number of binding sites, n.

The model also accounts for the action of allosteric effectors. An inhibitor
is a molecule that binds preferentially to the T state and stabilizes it, just
as the substrate stabilizes the R state. Because inhibitors stabilize the T
state, they make it more difficult for the substrate to convert the molecule to
the R state. An extremely effective inhibitor would stabilize the T state so
much that the protein could not escape from it. The substrate would then
always bind to low affinity T –state sites, and there would be no cooperativity
in the substrate binding. An activator works the other way round by bind-
ing preferentially to the R state and therefore stabilizing it. This will favour
the transition to the R-state earlier in the substrate binding curve. Again,
a highly effective activator would stabilize the protein almost entirely in the
R state, so that the binding of substrate would be to the high affinity sites
and the cooperativity in substrate binding would be lost. This explanation
76 3 Enzyme activity: the molecular basis for its regulation

requires that the binding of the inhibitor and the activator both exhibit co-
operativity. Indeed, the interactions between the different ligands, known as
heterotropic effects, depend for their occurrence on the cooperativity of the
substrate and/or the effector binding, the homotropic effects.

Supplement: The equations for inhibition and activation In terms of the

binding equation (Eqn. [3.12]), the effects of activators and inhibitors are to change
the apparent value of L. For example, if an activator, B, binds exclusively to each of
the n subunits of the R state with a dissociation constant KR,B , whilst an inhibitor
G binds exclusively to the n subunits of the T state with a dissociation constant
KT,G , then L in Eqn. [3.12] is replaced by L0 given by
(1 + γ)n
L0 = L
(1 + β)n
where β is the relative concentration of the activator, B/KR,B and γ is the relative
concentration of the inhibitor, G/KT,G . If n is two or more, then this can make the
substrate binding curve of the enzyme very sensitive to the effectors in exactly the
K-system manner shown by the enzyme ATCase (Fig. 3.10).

3.5.2 The sequential model

Koshland and his colleagues137 developed a somewhat different theory for the
action of allosteric enzymes. As in the Monod, Wyman and Changeux model,
the cooperative kinetics are assumed to arise from the cooperative binding
properties of a multi–subunit enzyme. Again, the conformational changes
of the protein are an essential factor. However, in a link to Koshland’s
induced fit theory of enzyme action, the conformational change involved is
specifically induced by substrate binding. The original model developed by
Koshland, Nemethy and Filmer was termed the sequential model; later, Haber
and Koshland proposed a more flexible general model,93 which includes both
his sequential model and the Monod, Wyman and Changeux model as spe-
cial cases. However, the distinctive aspects of the Koshland models can be
illustrated with the original sequential model, which is described here.
In the sequential model, the subunits of the protein are assumed to adopt
two conformations, termed A and B. (T and R are not used because it is not
implied that one is more stabilized by bonds than the other, as is the case
with the tense and relaxed designations.) In this case, the A conformation is
characteristic of a subunit without substrate attached. The B conformation
is induced by the binding of substrate to the subunit; in the simple sequential
model, it is assumed that the B conformation does not arise unless substrate
binds, so there is no pre–existing equilibrium between subunits in the A and B
states in the absence of substrate. In addition, the conformational change oc-
curs one subunit at a time, in step with substrate binding, unlike the Monod,
Wyman and Changeux model where all the subunits must change their con-
formation together. The source of the effect of the substrate binding at one
3.5 Allosteric enzymes 77

site on the binding at another depends on the differences in the strengths of

the interactions between subunits according to their conformations. If the in-
teractions between a pair of adjacent subunits are considered, then the three
types of interaction are A–A, A–B and B–B. The behaviour of the model
is more complicated than that of the Monod, Wyman and Changeux model
for two reasons. Firstly, it is affected by whether the A–B interactions are
stronger or weaker relative to the A–A and B–B interactions. In particular, if
A–B interactions are relatively weak, then the initial phases of binding where
they are formed are not favoured, but the later stages of binding where the B–
B interactions form are promoted, and the model gives positive cooperativity.
However, if A–B interactions are strong, this assists the early stages of bind-
ing but not the later stages and the model generates negative cooperativity,
which cannot be explained by the concerted model. Secondly, it is affected by
the arrangement of the subunits in the protein, which determines how many
subunit–subunit interactions can occur. For example, four subunits could be
linked in a linear chain, in which case the two outer subunits have only one
interface with another, whereas the two inner ones each have two. (However,
this is thought to be an unlikely arrangement, since the protein should be able
to continue growing at the ends of the line; there does not seem to be any-
thing to stop the process at four subunits.) If each subunit interacts with two
neighbours, the arrangement could be schematically represented as a square
composed of four smaller squares joined along two edges (without implying
that the arrangement of subunits in space is geometrically square). If each
subunit interacts with three others, then the four subunits would have to be
arranged at the corners of a tetrahedron. The model is illustrated with the
square configuration in Fig. 3.12.
» ¾» ¾» ¾»
- S - S S - S S - S S
¾ ¾ ¾ ¾
½ ½¼
S S S

@ µ
¡
@
I ¡¡
@@ R
@ ¡¡
»
@ ª
¡
S

½
S

Fig. 3.12: The sequential allosteric enzyme model of Koshland, Nemethy & Filmer.
The A and B conformations are represented as squares and circle respectively. A tetramer
with square interaction geometry is illustrated. There are two possible distinct arrange-
ments of the tetramer with two substrates bound. 2 bars at the subunit interface show
an AA interaction, 1 an AB interaction and none a BB interaction.

Supplement: Equations for the Koshland models Binding equations can be

78 3 Enzyme activity: the molecular basis for its regulation

worked out using the principles described above in relation to Eqn. [3.13]. However,
in this case, different equations are obtained depending on the number of subunits
and the interaction geometry, whereas a single equation (containing the number of
subunits as a parameter) suffices for the concerted model. The original derivations
considered the equilibrium constant (or free energy change) at each step in a scheme
such as Fig 3.12 to be made up from a number of components: KS , the intrinsic
dissociation constant for a subunit in the B state; Kt , the equilibrium constant for
the tertiary conformational change of the A subunit into a B subunit; KAA , the
equilibrium constant for the interaction of two adjacent A subunits (i.e. part of the
quaternary conformational energy); KAB , the equilibrium constant for the quater-
nary interaction between an A subunit and a B subunit, and KBB the equilibrium
constant for the interaction of an adjacent pair of B subunits. At each step, the
change in the subunit–subunit interactions, brought about by the binding of S, has
to be taken into account, and this can be different if different spatial arrangments
of S on the protein subunits are possible (as in the case of two S molecules bound
in Fig. 3.12). Finally, it is found that in the resulting binding equation, these basic
equilibrium constants always occur in certain combinations, so that they cannot be
independently determined. For example, KS and Kt always occur multiplied to-
gether, so that any pair of values that give the same product would give the same
binding curve. In fact, for similar reasons, the final binding equation for a square
tetramer can be expressed in terms of just two constants, formed of collections of
the five constants. Unfortunately, the way the constants are collected proves to be
different for each case of number of subunits and interaction geometry. Thus the
assumed molecular interactions underlying the binding behaviour are not experi-
mentally accessible from the binding curves.

The action of allosteric effectors can be qualitatively explained by the se-

quential model. They can be thought of as inducing conformations that are
more B–like for activators or more A–like for inhibitors. There is no simple
way of writing activation and inhibition equations, though. Thus the flexi-
bility of the sequential model is offset by greater complexity, and since the
number of parameters in the equations is fewer than the number of physically
meaningful interactions that go into describing the model, effectively the val-
ues of some of the parameters are not accessible from binding experiments.
The scope for greater complexity is increased in Koshland’s general model,
which is like the sequential model except that it is assumed that the B con-
formation can exist before substrate binds. The Koshland models can be
thought of as emphasizing the tertiary structural changes of a protein (i.e.
the conformation of individual subunits) and also the linkage of the tertiary
structure to the quaternary structure (i.e. the subunit–subunit interactions).
The Monod, Wyman and Changeux model could be said to emphasize the
quaternary structure, whilst the tertiary structure has less importance.
Most other models of allosteric enzymes assume that the characteristic
kinetic behaviour depends on cooperativity in ligand binding and use the
concepts of the two models described above, though combined in different
ways. The only really different model for allosteric enzymes is the kinetic
model of Rabin. Here the cooperativity in kinetic behaviour does not depend
3.5 Allosteric enzymes 79

1.0

0.8

Fractional saturation 0.6

0.4

0.2

0.0
0 5 10 15 20

Oxygen partial pressure, kPa

Fig. 3.13: Oxygen binding by haemoglobin.
The curves were measured by my student Richard Penny on human red blood cells, with
an internal pH of 7.1. : No 2,3–bisphosphoglycerate. ¦: 5 mM 2,3 bisphosphoglycerate
in the cells.

on the equilibrium binding characteristics of the enzyme, but is a consequence

of the kinetic mechanism and the particular values of the rate constants for
its component steps. An occasional example of an enzyme apparently fitting
the Rabin model is encountered, but it is generally believed that it is more
common for allosteric behaviour to reflect binding properties, and in a number
of cases there is even direct or indirect evidence to support this view.

3.5.3 Specific examples

I will now give a few examples of how the properties of certain proteins can, or
cannot, be explained in terms of the allosteric models described above. They
show that both of the models have some degree of success in particular cases.

3.5.3.1 Haemoglobin
Haemoglobin is an oxygen binding protein rather than an enzyme, but other-
wise it has been regarded as an example of an allosteric protein. Myoglobin
and haemoglobin were the two first proteins to have their 3-dimensional struc-
tures determined by x–ray crystallography by Max Perutz and his colleagues
at Cambridge. Myoglobin is an oxygen storage protein that is found in partic-
ularly high levels in the muscles of diving mammals such as the sperm whale.
It is relevant here because it contains a haem group and its polypeptide chain
is folded in a virtually identical manner to the subunits of haemoglobin. Myo-
globin, however, occurs as single subunits showing a hyperbolic binding curve
80 3 Enzyme activity: the molecular basis for its regulation

for oxygen, with a much greater affinity for oxygen than has haemoglobin. The
oxygen affinity of haemoglobin at the top end of its oxygen saturation curve,
where the fourth oxygen is binding, is comparable to the oxygen affinity of
myoglobin, so it appears that the assembly of haemoglobin into a tetrameric
structure in some way inhibits the oxygen binding, as would be expected for a
T –state structure in the concerted model of Monod, Wyman and Changeux.
The concept that there were two distinct conformational states of haemoglobin
was supported by the crystallography studies. Crystals of deoxyhaemoglobin
were grown and the protein structure determined in this state. Significantly,
if oxygen is allowed into contact with the crystals, they crack because of the
conformational change on oxygen binding. Crystals of oxyhaemoglobin had to
be grown separately for determining its structure; in fact, during the growth
and x–ray analysis of the crystals, the iron atom in the haem group becomes
oxidized, and met–haemoglobin (which does not bind oxygen) is formed, but
its conformation is believed to be the same as oxyhaemoglobin. The con-
formations of the subunits themselves scarcely differ between deoxy and oxy
haemoglobin; the big difference was found to be the relative positioning of
the subunits and the bonding between them. In accordance with the con-
certed model, there was more hydrogen bonding and salt linkages between
the subunits in the deoxy state than in the oxy. Perutz’s model of the binding
of oxygen to haemoglobin assumes that there is a concerted conformational
transition between the deoxy (T ) state and oxy (R) state because it is as if the
zig–zag surface of the interface has moved across by a complete notch, and any
intermediate position would not be tenable. Nevertheless, x–ray crystallogra-
phy gives a relatively static view of protein structures, which are known from
other evidence to be flexible and mobile, and there is little direct evidence of
the intermediate states in oxygen binding because they cannot be prepared
as crystals, so the concerted transition is inferred rather than observed. How-
ever, there are well over a hundred known mutant haemoglobins that differ
from normal haemoglobin by different substitutions of a single amino acid,
and the effects of these on the affinity and the cooperativity of the oxygen
binding can generally be interpreted in terms of the impact of the structural
change on the concerted model (see question 2 at the end of this chapter).
In most mammalian red blood cells, there are high concentrations of 2,3–
bisphosphoglycerate. This is normally only present in low concentrations in
cells, as a cofactor for the glycolytic enzyme phosphoglycerate mutase, but in
humans and many other mammals it acts as an allosteric inhibitor of oxygen
binding. Its concentration is around 5 mM, comparable to the concentration
of haemoglobin protein, but it varies slowly with conditions that affect the
oxygenation of the blood such as the low oxygen pressure found at high al-
titudes. It acts as an inhibitor of oxygen binding, and is responsible for the
oxygen affinity of blood being approximately half that of isolated haemoglobin,
though it also slightly increases the cooperativity of binding. X-ray crystal-
lography has shown that the binding site for 2,3–bisphosphoglycerate is in
the central cavity of deoxyhaemoglobin between the four subunits; this cavity
closes down in oxyhaemoglobin and would be too small to allow binding. Di-
3.5 Allosteric enzymes 81

rect binding studies of 2,3–bisphosphoglycerate have confirmed that it binds

preferentially to deoxyhaemoglobin, but they do measure weaker binding to
oxyhaemoglobin, in spite of the apparent absence of the binding site in the
crystal structure. The inhibitor therefore appears to behave very much as
predicted in the concerted model. Proton binding, responsible for the pH–
dependence of oxygen binding and the release of protons on binding of oxygen
(the Bohr effect), does not appear to be a simple case of inhibition by pref-
erential binding to the T –state. Thus although many lines of evidence imply
that haemoglobin behaves in accordance with the predictions of the concerted
model, the equations of the model do not manage to describe the variation of
the oxygen binding curves at different pH and bisphosphoglycerate concen-
trations completely successfullly and modified versions of the theory have to
be used to describe such results.66

3.5.3.2 Aspartate transcarbamylase

Aspartate transcarbamylase was one of the first two enzymes shown to be
subject to feedback inhibition,281 and it has been studied intensively since
then. Its reaction

carbamoyl phosphate + aspartate *

) carbamoyl aspartate + phos-
phate

is inhibited by the end product of its pathway, CTP, and is activated by

ATP (Fig. 3.10). The enzyme is composed of six regulatory subunits (R)
that carry the binding sites for ATP and CTP and six catalytic subunits (C)
with the active sites. The regulatory subunits are present as three dimers
and the catalytic subunits as two trimers, so the structure is (R2 )3 (C3 )2 .
Various chemical and denaturing treatments specifically abolish the effects of
the regulatory subunits, leaving an enzyme that has hyperbolic kinetics with
respect to both its substrates and that is insensitive to ATP and CTP. Some
of these treatments, (e.g. with mercuric compounds) causes dissociation of
the catalytic trimers from the regulatory dimers. The binding curves of the
substrates and effectors to the enzyme fit well with the Monod, Wyman &
Changeux concerted model. The enzyme undergoes substantial changes in
conformation upon binding substrates or effectors that can be readily detected
by a variety of means (chemical reactivity of certain amino acid side chains,
spectroscopic properties and size and shape).82 These changes broadly fit
with the expected T *) R transitions predicted by the concerted model. Thus
aspartate transcarbamylase is probably the best established example of an
enzyme that obeys the concerted model.124

3.5.3.3 Glyceraldehyde–3–phosphate dehydrogenase

This glycolytic enzyme is an enigma, since its properties vary greatly depend-
ing on its source for reasons that are unknown. The enzyme from yeast binds
82 3 Enzyme activity: the molecular basis for its regulation

NAD+ with positive cooperativity, and there appears to be a concerted con-

formational change of the enzyme in accordance with the concerted model.
The enzyme from rabbit muscle, on the other hand, exhibits negative coopera-
tivity in the binding of NAD+ , and this cannot be explained by the concerted
model, though it can be by the sequential model. In fact, the negative coop-
erativity of this form of the enzyme is so marked that, after two molecules of
NAD+ have bound to the tetrameric enzyme, the affinities of the third and
fourth sites fall substantially. It is an example of a phenomenon since found
in some other multi–subunit enzymes called half of the sites reactivity, on the
basis that only a maximum of half the binding sites are occupied at any time.
What remains unclear to this day is whether there is any specific function in
metabolic regulation for negative cooperativity and half of the sites reactivity.

3.6 Summary
1. The well–known Michaelis–Menten equation is not a complete descrip-
tion of the behaviour of single–substrate enzymes in vivo because of the
the effects of product inhibition and reversibility of the reaction.

2. The majority of enzymes in metabolism have two substrates (and prod-

ucts) and catalysis can involve one of three basic mechanisms:

• compulsory order;
• random order, and
• double displacement.

3. Generally, for two–substrate enzymes, the apparent Km for one sub-

strate will vary with the concentration of the other, so prediction of the
rate at any particular set of substrate concentrations requires knowledge
of the parameters of the appropriate rate equation.

4. Many enzymes are composed of subunits. If the binding of a ligand, such

as a substrate, on one subunit affects the affinity of another subunit for
the same ligand, the binding is said to be cooperative and to involve
homotropic interaction.

5. Allosteric enzymes are a class of multisubunit (or multimeric) enzymes

that have sites for effectors as well as the active site. As well as ho-
motropic interactions, heterotropic interactions occur, whereby the ef-
fector can cause activation or inhibition of the enzyme by affecting sub-
strate binding.

6. Theories for the mechanisms of allosteric enzymes, such as the concerted

and sequential models, centre around linkages between ligand binding
and the conformational state of multimeric enzymes.
3.6 Summary 83

Further reading
Background to enzyme kinetics and allosteric enzymes: any biochemistry text
such as:

1. Mathews, C. K. & van Holde, K. E., Biochemistry The Benjamin/Cumm-

ings Publishing Co. Inc., Redwood City (1990).

2. Stryer, L. Biochemistry, 3rd ed. W. H. Freeman & Co., New York

(1988).

Enzyme kinetics, general:

3. Cornish–Bowden, A., Fundamentals of enzyme kinetics, 2nd ed. Port-

land Press, London (1995).

4. Cornish–Bowden, A. & Wharton, C. W., Enzyme kinetics IRL Press,

Oxford (1988).

5. Segel, I. H., Enzyme kinetics Wiley, New York (1975).

Computer programs
There are many programs available for fitting enzyme kinetics data. Hyper
by John Easterby is a Windows program for fitting Michaelis–Menten kinet-
ics, currently available as shareware on the Internet by anoymous FTP from
ftp.bio.indiana.edu in /molbio/ibmpc/hyper.zip, or from micros.hensa.ac.uk
in /mirrors/cica/win3/util/hyper.zip.
For fitting inhibition and two substrate kinetics, the program Leonora for
IBM PC-compatible microcomputers by Athel Cornish–Bowden is distributed
with his book explaining the theory behind the fitting procedures: Analysis
of enzyme kinetic data, Oxford University Press, Oxford, (1995).

Problems
1. A calcium–binding protein from bovine brain has a relative molecu-
lar mass of 17000. Its calcium–binding properties were investigated
by an equilibrium dialysis experiment, in which the protein solution (2
mg.cm−3 ) was placed in dialysis sacs that retain the protein and allowed
to come to equilibrium with outer solutions containing various amounts
of Ca2+ (in addition to buffer and salts). At equilibrium, the Ca2+
concentration is higher in the sacs than the outer solution since both
contain free Ca2+ at the same concentration, but the sacs also contain
protein–bound Ca2+ . The calcium concentrations at equilibrium were
measured by atomic absorption spectrophotometry, giving the results in
the table below. Determine the number of binding sites on the protein
for Ca2+ and their dissociation constant.
84 3 Enzyme activity: the molecular basis for its regulation

[Ca2+ ], µM
Outer solution Sac fluid
0.9 47
2.1 94
4.3 152
7.7 206
18.2 284
35.0 336

2. The following table contains data for the mid–regions of the oxygen bind-
ing curves of three haemoglobins measured in solution under identical
conditions. The oxygen concentration is reported as the partial pres-
sure of oxygen, PO2 , and the degree of oxygen binding by the fractional
saturation, S. The three haemoglobins are A (normal adult human),
Hiroshima (His 146 β → Asp) and Kansas (Asp 102 β → Thr). Plot a
graph to determine the Hill coefficient, nH , of each haemoglobin, and
speculate on how the effects on the cooperativity and oxygen affinity
could be interpreted in terms of the concerted model, given that the
mutations are at the subunit interfaces rather than the oxygen binding
sites.

A Hiroshima Kansas
P O2 S P O2 S P O2 S
0.67 0.044 0.133 0.045 1.33 0.193
1.33 0.243 0.267 0.179 2.67 0.371
2.00 0.500 0.667 0.620 4.00 0.500
2.67 0.691 1.33 0.822 5.33 0.592
4.00 0.874 2.00 0.948 6.67 0.660

3. Isocitrate dehydrogenase from a slime mould catalyzes the reaction:

isocitrate + NAD+ → 2-oxoglutarate + CO2 + NADH

The following rates of product formation (as µmol.s−1 .mg−1 ) were mea-
sured on the enzyme: Use the primary and secondary plot system to
estimate V and the other kinetic parameters of the two substrate rate
equation. Can you deduce anything about the possible mechanism of
this enzyme?
3.6 Summary 85

NAD+ , mM
isocitrate, mM 0.2 0.4 0.8 2
0.05 8.5 10.1 11.2 11.9
0.15 12.6 16.6 19.6 22.1
0.25 14.0 19.0 23.1 26.6
0.50 15.2 21.3 26.7 31.5
86 3 Enzyme activity: the molecular basis for its regulation
 4

Traditional approaches to
metabolic regulation

Until recently, biochemists tackled the problems of metabolic regulation and

control either by attempting to identify the molecular details of the under-
lying mechanisms or by trying to formulate qualitative descriptions of the
system’s behaviour (see Chapter 1.2). There was general agreement that the
explanations would centre around key regulatory, or rate–limiting, enzymes
(Chapter 1.4.3, p. 18). I also mentioned in Chapter 1 that this conceptual
framework was ultimately unsatisfactory and that the purpose of this book is
to show why and to explain an alternative theory. Before I do that, I want
to describe the experimental approaches that were used to identify regula-
tory enzymes, partly because the language and concepts are still in current
use, and partly to provide the context for the different approach of Metabolic
Control Analysis.
The most coherent descriptions of the lines of evidence used in the tra-
ditional approaches were written by Hales95 and by Eric Newsholme and his
colleagues in Oxford (including Gevers,169 Underwood172 and Rolleston201 ).
The plan of this chapter draws on their work and the influential textbook
arising from it, Regulation in Metabolism by Newsholme & Start.171 (Later,
Newsholme and Bernard Crabtree, were to develop their own form of quan-
titative systems theory.55–57 ) The following sections describe the lines of
evidence that were taken to point to regulatory enzymes.

4.1 The teleological approach

Teleological arguments assume that the system has the properties necessary
to fulfil its function; in other words, it assumes the design is appropriate to
the purpose. This can be made more biologically respectable by arguing that
evolution will have favoured the adoption of efficient solutions to the problems

87
88 4 Traditional approaches to metabolic regulation

of control and regulation. Of course, this assumes that we are able to identify
correctly the problems of constructing an efficient metabolism. The argument
applied in the search for regulatory enzymes is that it would be wasteful to
have ‘unnecessary’ metabolism, and if pathways were controlled near their
output ends, when they were slowed down, metabolites would still continue
to flow in at the input, causing an accumulation of metabolic intermediates.
Not only might this cause the expenditure of unnecessary metabolic energy,
it could also cause osmotic problems for the cell. Therefore, it would be much
better if a controlling enzyme were to be placed at the beginning of a pathway,
as suggested by Sir Hans Krebs.141
The argument seems seductively reasonable, and there are many appar-
ent examples in its favour. This is no doubt because the argument has been
influenced by the discovery of many enzymes with regulatory properties at
the points where a metabolic pathway branches off from the rest of metabol-
ism, as will be described later (Chapter 7.2). The problem is that there are
examples where the first enzyme in the pathway is not a regulatory enzyme
(e.g. mammalian serine metabolism, see Chapter 6.3.4.2, p. 193, in which the
regulatory enzyme is the last one in a three step pathway). In these cases, it is
explained that the regulatory enzyme is the first ‘committed step’. That is the
problem with teleological arguments: when they work, they seem reasonable;
when they don’t work, the story is changed slightly, and it’s now reasonable
that things should be different for this particular case.
A variant on this concept foreshadowed later developments: Rolleston 201
argued that the group of enzymes at the beginning of a pathway, including the
first non–equilibrium enzyme, could form a rate–limiting system, interacting
through substrate and coenzyme concentrations to form a functional unit.

4.2 Non–equilibrium enzymes

As discussed in Chapter 1.4.2, p.13, there was an expectation that the flux
through a pathway was not likely to be affected by changing the activity of
an enzyme catalyzing a reaction that was near equilibrium, but changing the
activity of an enzyme catalyzing a reaction displaced from equilibrium might
affect the flux. Of course, this principle does not explain which enzymes at
non–equilibrium steps can affect a metabolic flux, but it does suggest that
the field of candidates can be narrowed down by eliminating all the near–
equilibrium reactions.
A useful measure of the closeness of a reaction to equilibrium was defined
in Chapter 1.4.2 as the disequilibrium ratio ρ, which is the mass action ratio
relative to the equilibrium constant, Γ/Keq , as defined in Eqn. [1.6]. When ρ
is equal to 1, a reaction is at equilibrium, and reactions at steady state in a
metabolic pathway through which there is flow must have values less than 1.
The problem is to decide where the boundary lies between near–equilibrium
and non–equilibrium reactions. Rolleston suggested that the lowest value of ρ
for a reaction to be classed as near–equilibrium could be as high as 0.2 or as low
4.2 Non–equilibrium enzymes 89

10

1 .................................... ............ ............ ............ ............ ........................ ............ ............

............ ............
............ ............ ................................................
.................................... ............ ................................................
............ ........................
. . . . . . . . . . . .
............ ........................
.................................... ............ . . . . . . .
..................................... . . . . . . . . . . .
.................................... . . . . . ....................................
10-1 .................................... .................................... .................................... ....................................
.................................... .................................... .................................... ....................................
.................................... .................................... ....................................
10-2 .................................... .................................... ....................................
.................................... ....................................
.................................... ....................................
10-3
ρ

.................................... ....................................
.................................... ....................................
10-4 .................................... ....................................
.................................... ....................................
.................................... ....................................
10-5 .................................... ....................................
.................................... ....................................
.................................... ....................................
10-6 ....................................
............
10-7
Ph Pg Tr He Gl Pf Pm En Py

Glycolytic reaction

Fig. 4.1: Displacement of reactions from equilibrium in glycolysis in working rat

heart.
The measurements were of metabolites in working, perfused heart supplied with glucose
but no insulin. Both external glucose and endogenous glycogen were being used as
fuels. Equilibrium constants were corrected to measured intracellular conditions. Key:
Ph, phosphorylase; Pg, phosphoglucomutase; Tr, glucose transport; He, hexokinase; Gl,
glucose–6–P isomerase; Pf, phosphofructokinase; Pm, phosphoglycerate mutase; En,
enolase; Py, pyruvate kinase. Data from Kashiwaya et al. 125

as 0.05.201 Even if 0.05 is taken as the boundary between near–equilibrium and

non–equilibrium reactions, this often leaves a significant number of reactions
in the pathway that are non–equilibrium and therefore potential candidates
for being regulatory enzymes, since reactions with ρ between 0.01 and 0.0001
are relatively common.
Measurement of the disequilibrium ratio can pose a number of difficul-
ties. It depends upon measurements of metabolite concentrations, which are
subject to the problems referred to in Chapter 2.3. In addition, the equi-
librium constant of the reaction must be known from in vitro experiments.
It is possible for the constant to have been measured inaccurately or not to
have been adjusted appropriately for intracellular conditions. Equilibrium
constants of biochemical reactions may vary with pH, temperature and ionic
conditions, particularly where the reactants have weakly ionizable groups or
groups that bind metal ions such as K+ or Mg2+ . Phosphorylated metabolites
exhibit both properties. (See the article by Kwack & Veech cited at the end
of this chapter for examples of how equilibrium constants can be adjusted to
intracellular conditions.)
Some typical results are shown in Fig. 4.1 for the glycolytic pathway in
90 4 Traditional approaches to metabolic regulation

perfused working rat heart, taken from experiments performed recently in the
laboratory of Veech and Passonneau in the United States.125 This study was
undertaken to examine the proposal that, in the absence of the hormone in-
sulin, transport of glucose into the heart muscle cells is the rate–limiting step
for glycolysis.162 Since insulin stimulates the movement of the glucose trans-
porter molecules from the endoplasmic reticulum to the plasma membrane,
it had also been held in the past that transport ceases to be ‘rate–limiting’
in the presence of insulin.170 The design of these recent experiments included
a quantitative analysis of the distribution of control by Metabolic Control
Analysis. Later in this book (Chapter 5.5.1), I will return to the results of
that analysis and how they relate to the disequilibrium ratios. My reason
for choosing this set of results as an illustration, even though they were not
used to find a rate–limiting step, is that they were undertaken with unusual
thoroughness. Thus all the equilibrium constants used in the calcuations have
been corrected to intracellular conditions, such as pH and the free cytoplas-
mic phosphate (which had been measured by 31 P NMR) and the free Mg2+
(which had been calculated from its effect on the citrate:isocitrate ratio).
The results show that three reactions, phosphorylase, hexokinase and
phosphofructokinase are very far from equilibrium, as is also pyruvate ki-
nase, though to a lesser degree. Phosphoglucomutase, glucose–6–phosphate
isomerase and phosphoglycerate mutase are close to equilibrium; indeed, the
measured disequilibrium ratio for the latter is greater than 1, though not sig-
nificantly so given the errors in the metabolite ratios. Glucose transport and
enolase are on the boundary between the two groups. (The enzymes between
aldolase and phosphoglycerate kinase could not have their disequilibrium ra-
tios measured. This is partly because aldolase binds sufficient glyceraldehyde–
3–phosphate to distort the results and also because 1,3–bis phosphoglycerate
is both difficult to measure and subject to binding by enzymes.)
In the past, some researchers attributed significance to differences in the
disequilibrium ratio at a step in different metabolic states. However, it is not
easy to draw any conclusions from such results, and the separate changes in
substrate and product concentrations were often thought to be more informa-
tive, as will be discussed in a later section of this chapter (4.6).

4.3 Isotopic measurement of flux

Another method of distinguishing between a near–equilibrium reaction (for-
ward and reverse fluxes larger than net pathway flux) and a non–equilibrium
reaction (forward flux comparable to net flux) is to measure the compo-
nent forward and reverse reaction rates. This is possible because at near–
equilibrium steps, the isotopically labelled substrate is converted into the
product faster than would be expected from the net pathway flux. (At the
start of such experiments, although the pathway is at steady state with re-
spect to the concentrations of intermediates, it is not at isotopic equilibrium.)
This can be seen in the computer–simulated examples shown in Fig. 4.2. In
4.3 Isotopic measurement of flux 91

1.0 1.0
a) b)
0.8 0.8
Relative specific activity

Relative specific activity

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 0.0 1.0 2.0 3.0 4.0 5.0

Time Time

Fig. 4.2: Isotopic labelling kinetics of pathway.

The results have been produced by computer simulation of a pathway, and show the
spread of the isotope into the first three intermediates of the pathway after its addition in
the pathway source at time 0. Intermediates become isotopically labelled successively, so
the leftmost curve is the first intermediate and so on. a) The step between the first and
second intermediates is near–equilibrium, with ρ = 0.86, corresponding to the forward
rate’s being 7.2 times faster than the net pathway flux. The other steps are irreversible,
non–equilibrium steps. b) The step between the first and second intermediates is now
an irreversible, non–equilibrium step, but the kinetics of the pathway have been adjusted
to ensure that the pathway flux and intermediate concentrations are the same as in a).

the example where the first and second intermediates are linked by a near–
equilibrium enzyme, the isotope is slower to accumulate in the first inter-
mediate because of the high rate of transfer to the second intermediate. The
amount of isotope in the second intermediate closely follows the amount in the
first, and isotope accumulates in this intermediate faster than it does when
the reaction is irreversible and non–equilibrium, as in the other example,
Fig. 4.2b). Computer fitting of results such as those in Fig. 4.2a) allows the
forward and reverse reaction rates at the equilibrium step to be determined,
as well as the pathway flux.
Many investigators have used such techniques to measure the fluxes through
near–equilibrium reactions, particularly those in the tricarboxylic acid cycle
and the glycolytic and gluconeogenic pathways in a wide range of cells and
tissues.10, 19, 58, 130, 192 In a study of the oxidation of [1–14 C]acetate by perfused
rat heart, Sir Philip Randle and colleagues in Bristol found192 that the specific
activities (relative to the acetate) of successive tricarboxylic acid cycle inter-
mediates after 60 seconds labelling were: acetyl CoA, 63%; citrate (in the 2
carbons derived from acetate), 43%; 2–oxoglutarate, 23 %, malate, 13.5% and
oxaloacetate (from the 4 carbons contained in citrate), 4.7%. This illustrates
the progressive spread of label through the intermediates in a similar manner
92 4 Traditional approaches to metabolic regulation

to the example shown in Fig. 4.2. The activities in glutamate and aspartate
were 20% and 6%, showing that these two intermediates were near equilibrium
with 2–oxoglutarate and oxaloacetate respectively (as in Fig. 4.2a) through
the action of transaminases. These conclusions were confirmed by computer
modelling of the results and estimation of the rates of the individual steps.
Fig. 4.3 shows the relative rates of the non–equilibrium steps and the
forward and reverse reactions of the near–equilibrium reactions of the tricar-
boxylic acid cycle in the slime mould Dictyostelium discoideum as determined
by Barbara Wright’s group130 from fitting experimental results on the spread
of 14 C.

citrate
1
0.42 0.42

w
oxaloacetate 2–oxoglutarate
±

5.59 5.21 0.48

U °
malate succinate
o 7
0.73 1.04

1.25 z 1.55
fumarate )

Fig. 4.3: Fluxes in the tricarboxylic acid cycle of Dictyostelium discoideum.

Only part of the metabolic network that is used in the computer fit to the experimental
pattern of radioactive labelling is shown. The fluxes, in mM/min are not perfectly
balanced because of fitting errors and minor fluxes that are not shown.

4.4 Maximal enzyme activities

The maximal catalytic activities of enzymes (as limiting rate values) are often
in considerable excess over the rates of metabolic flux; values of 100 to 10,000
times higher are not unusual. Some enzymes do not appear to be present
in such ‘excessive’ amounts, and it is presumed that these are more likely to
4.4 Maximal enzyme activities 93

be regulatory. Making suitable measurements to carry out the comparison,

though, raises the problems referred to in the previous Chapter (2.4). For my
example, in Fig. 4.4, I have again drawn on the recent studies of glycolysis in
working rat heart carried out by Veech, Passonneau and their colleagues125
and used above to illustrate the displacement from equilibrium in a pathway.
Their measurements were not taken to identify rate–limiting enzymes by their
low activity, but they show the typical range of values observed in pathways,
and they have the advantage that the whole set was measured in the same
experiments and in constant assay conditions resembling the in vivo state
(pH 7.2, 38◦ C, 150 mM K+ and 5 mM Mg2+ ). The lowest activities are those
of hexokinase, phosphorylase and aldolase, followed by phosphofructokinase
and enolase. Apart from a lower activity of triose phosphate isomerase and
higher activity of phosphoglycerate kinase than usual, the activities of the
enzymes relative to one another are similar to those that have been reported
previously for mammalian tissues. (See, for example, Table 3.4 in Regulation
in Metabolism by Newsholme & Start.171 ) The consistently low measured
activities of aldolase and enolase seem at variance with the usual finding
that their reactions are not very far from equilibrium; this might point to a
systematic measurement problem, though if so, its nature is uncertain.

..............................
.......................
.......................
1000 .......................
.......................
.......................
....................... .......................
....................... ...............
................ ....................... .............................. .............. ...............................
100 ....................... .......
..................... ............... ............................... ............................... ............................ ...............................
V/Jgly

....................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
....................... ..................... ....................... ....................... ....................... ..................... ..............................
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.. .. .. .. .. .. .. . ..................... ....................... ....................... ....................... ........ ..................... .......................
....................... ....................... ....... ....... .......
....................... ....................... ....................... ........ ............................ ............................... ............................... ............................... ............................... ............................ ...............................
10 ................ ....................... ....................... ....................... ....................... ..................... ....................... ....................... ....................... ....................... ..................... .......................
....................... ....................... .............. ....................... ....................... ....................... ..................... ....................... ....................... ....................... ....................... ..................... .......................
....................... ....................... ..................... ....................... ....................... ....................... ..................... ....................... ....................... ....................... ....................... ..................... .......................
....................... ....................... ..................... ....................... ....................... ....................... ..................... ....................... ....................... ....................... ....................... ..................... .......................
....................... ....................... ..................... ....................... ....................... ....................... ..................... ....................... ....................... ....................... ....................... ..................... .......................
....................... ....................... ..................... ....................... ....................... ....................... ..................... ....................... ....................... ....................... ....................... ..................... .......................
1
Ph Pg He Gl Pf Al Tr Ga Pg Pu En Py La

Glycolytic reaction

Fig. 4.4: Relative enzyme activities in glycolysis in working rat heart.

The limiting rates (V) were all measured at pH 7.2 in the presence of 150 mM K + and
5 mM Mg2+ . They are given relative to the glycolytic flux, Jgly , measured for working
rat hearts using glucose as a fuel. Key: Ph, phosphorylase; Pg, phosphoglucomutase;
He, hexokinase; Gl, glucose–6–P isomerase; Pf, phosphofructokinase; Al, aldolase; Tr,
triose phosphate isomerase; Ga, glyceraldehyde–3–P dehydrogenase; Pg phosphoglycer-
ate kinase; Pu, phosphoglycerate mutase; En, enolase; Py, pyruvate kinase; La, lactate
dehydrogenase. Data from Kashiwaya et al.125
94 4 Traditional approaches to metabolic regulation

Further evidence for a problem in measuring aldolase activities has recently

come from Fraenkel’s laboratory at Harvard. He and his colleagues inserted
plasmids into E. coli to increase the expression of aldolase as much as 20 or 30
fold9 and measured glycolytic flux, metabolite levels and some isotopic fluxes.
Their results could only be explained quantitatively by assuming that the
enzyme in the cells was 7.5 times more active than indicated by the in vitro
assay. The same laboratory reached a similar conclusion about underestima-
tion of the activity of phosphoglucoseisomerase in yeast.14 Barbara Wright and
Kathy Albe in Montana have encountered similar problems in constructing
computer simulation models of metabolism of the slime mould Dictyostelium
discoideum; the models can only be made consistent with the extensive ex-
perimental evidence (collected over many years by Wright’s group) if many
of the values for limiting rates measured in vitro misrepresent the maximal in
vivo activities by factors of 100 or more.278, 279 (The problems of computer
simulation of metabolism will be revisited later in this book, in Chapter 6.2.)

Examination of the maximal enzyme activities was often extended to mea-

suring whether there were particular enzymes whose activities changed sig-
nificantly when changes in the metabolic fluxes were caused by longer term
changes in dietary or environmental conditions or drug or hormonal treat-
ments. In the catabolism of the amino acid tryptophan by rat liver, vari-
ous hormonal and dietary treatments that increase its rate (e.g. adminis-
tration of glucocorticoids73 ) specifically increase the amount of tryptophan
2,3–dioxygenase by activating its synthesis. This was interpreted as indicat-
ing a possible role for the enzyme in the control of tryptophan catabolism.
Later we shall see that there has been some confirmation of this view from
quantitative measurements of the influence of this enzyme on the tryptophan
breakdown flux (Chapter 6.1.2, p. 151).

In many other cases, the pattern of variation of enzyme activities is not

obviously informative about the potential of individual enzymes to control
metabolic flux. In the case of the enzymes of carbohydrate metabolism in rat
liver, three groups can be identified by their response to dietary and hormonal
treatments. These are the enzymes exclusive to the gluconeogenic pathway,
those exclusive to the glycolytic pathway, and those that are shared by the two
pathways. The gluconeogenic group consists of pyruvate carboxylase, phos-
phoenol pyruvate carboxykinase, fructose–1,6–bisphosphatase and glucose–6–
phosphatase (see Fig. 6.17) and their amounts vary in parallel.268 There is
good evidence, though, from more recent experiments (to be described later,
Chapter 6.3) that the enzymes in the group differ greatly in their degree of
control over the gluconeogenic flux. Nevertheless, this and other cases where
many of the enzymes of a pathway change in activity together to cause a
change in metabolic flux almost certainly carry a message about metabolic
control, though not one envisaged in this traditional view (see Chapter 8.1).
4.6 The cross–over theorem 95

4.5 Addition of intermediates

If a pathway is controlled by reactions near its beginning, it would seem rea-
sonable to expect that higher flux rates might be observed if intermediates
could be fed into the pathway below this point. For example, the synthesis
of cholesterol from 2–carbon precursors (acetyl CoA) leads first to 5–carbon
intermediates, which are condensed to 15–carbon intermediates, which are in
turn dimerized to give the 30–carbon steroid skeleton. Higher rates of choles-
terol synthesis are observed from the 5–carbon intermediate mevalonate than
from 2–carbon acetate, and this has been taken as evidence for regulation
of the rate of synthesis early in the pathway (at the hydroxymethylglutaryl-
CoA reductase step). The same result is found in whole tissues and cell–free
extracts, so the difference is not related to differential permeability.
Another example is the comparison of the rate of gluconeogenesis in rat
kidney slices using pyruvate as a substrate with the rates measured using
tricarboxylic acid cycle intermediates fumarate and 2–oxoglutarate that read-
ily give rise to the gluconeogenic intermediate oxaloacetate (see Fig. 6.17 for
the pathway). Sir Hans Krebs reported142 that the rates of glucose formation
were very similar from all three compounds, and that therefore it was unlikely
that the formation of oxaloacetate from pyruvate by pyruvate carboxylase is
the ‘pacemaker’ (as he termed a rate–limiting step). However, experiments on
gluconeogenesis in rat hepatocytes to be described later (Chapter 6.3) show
that this enzyme has a larger influence over the rate of the pathway than any
other.
The problem with such experiments on the relative rates with different
intermediates is that there can be other equally reasonable explanations. If
a pathway intermediate is used as a starting point, it must of necessity be
supplied at higher concentrations than are normally found. The pathway
would therefore be expected to operate faster merely because of this factor.
On the other hand, if the pathway failed to show an increased flux, this might
be because the intermediate is poorly permeable and does not reach the site
of its metabolism. Therefore this method has not been regarded as reliable.
Indeed, in Krebs’ experiments, poor rates of gluconeogenesis were obtained
from glutamate and aspartate (even though these might be expected to be
efficient sources of oxaloactetate) but he discounted these results and chose to
attach greater significance to the results with 2–oxoglutarate and fumarate.

4.6 The cross–over theorem

In the 1950s, Britton Chance and his colleagues in Philadelphia were studying
the electron transport chain and oxidative phosphorylation in mitochondria.
They measured the degrees of reduction and oxidation of various of the elec-
tron carriers in the mitochondria by spectrophotometric measurements on
working mitochondria. In the course of these experiments, they had noted
that when electron transport was partially inhibited, the carriers towards the
96 4 Traditional approaches to metabolic regulation

NADH end of the chain became more reduced relative to normal (i.e. there
was a build up of ‘substrates’ of the chain) whereas the carriers from the site
of inhibition towards oxygen became more oxidized (i.e. there was a reduc-
tion in the ‘substrate’).34 They then used this phenomenon of a cross-over
in the state of oxidation at the inhibition site to determine the sites of cou-
pling of electron transport to oxidative phosphorylation by comparing State
3 respiration (i.e. phosphorylating respiration in the presence of ADP, see
Chapter 6.1.4.2, p. 161) to State 4 respiration, which is inhibited by the lack
of ADP for phosphorylation. They formulated a number of cross-over the-
orems to guide the interpretation of these experiments35 and they validated
these theorems33 by computer simulations. The principal theorem was that a
cross–over from increased substrate (relatively more reduced carriers) to de-
creased substrate concentration (relatively more oxidized carriers) when the
pathway flux had decreased indicated the site of an inhibition. Subsidiary
theorems indicated that cross–overs in the opposite direction upon inhibition
were not significant, and dealt with identification of multiple sites of inhi-
bition (as obtained with the three coupling sites of electron transport upon
limitation of the phosphorylation rate).
Other investigators then applied the principle cross–over theorem to iden-
tifying the sites of action of regulatory signals, both inhibitors and activa-

200

150
Concentration, %

100

50

0
G6P F6P FBP DHAP GAP PGA PYR

Metabolite

Fig. 4.5: Cross–over plot of the glycolytic pathway in yeast.

The metabolite levels in anaerobic conditions, when glycolysis is stimulated, are shown
as percentages of their aerobic concentrations.144 Abbreviations: G6P, glucose–6–
phosphate; F6P, fructose–6–phosphate; FBP, fructose–1,6–bisphosphate; DHAP, di-
hydroxyacetone phosphate; GAP, glyceraldehyde–3–phosphate; PGA, 3–phosphoglycer-
ate, and PYR, pyruvate.
4.6 The cross–over theorem 97

tors, on metabolic pathways. A typical example is shown in Fig. 4.5 from a

study144 on the increase in rate of glycolysis that occurs when yeast that has
been catabolizing glucose aerobically is transferred to anaerobic conditions
(an effect known as the Pasteur effect, after the pioneering French microbiol-
ogist Louis Pasteur). The 80% increase in flux has been accompanied by a fall
in the levels of fructose–6–phosphate and all the preceding glycolytic inter-
mediates, but a 220% increase in fructose–1,6–bisphosphate; the cross–over is
therefore at phosphofructokinase. The subsequent crossings of the 100% line
are not statistically significant because of the errors in the metabolite mea-
surements. (In any case, one of the cross–over theorems33 states that such
reverse cross–overs should not be taken as evidence of an interaction site.)
The experiments were interpreted as showing the stimulation of phosphofruc-
tokinase by an effector; of the metabolites known to affect the yeast enzyme
(including fructose–2,6–bisphosphate), only changes in phosphate concentra-
tions correlated systematically with the occurrence of the Pasteur effect and
this cross–over.
The applicability of the cross–over theorem to its original domain — the
electron transport chain — has not been questioned. Indeed, it is possible to
prove that the occurrence of a cross–over must involve an inhibition at that
point on the chain. Rolleston201 pointed out that the proof did not necessarily
validate application of the theorem to other pathways, for the proof depended
on particular properties of the electron transport. In particular, the typical
reaction in the chain is:
Ared + Box *
) Aox + Bred
where A and B are successive carriers and red indicates the reduced state and
ox the oxidized state. The proof depends on the conservation of A and B,
that is, that their total concentrations remain fixed, introducing an obligatory
coupling between the substrate and product concentrations and ensuring that
they must change in opposite directions. This is not the case in a pathway
such as glycolysis, where the total concentration of intermediates is not fixed
when glucose and phosphate are freely available, so an increase in substrate
concentration does not necessarily force a decrease in product concentration
(or vice versa). Furthermore, the cross-over criterion is difficult to apply for
the very common case of two substrate, two product enzymes, since in this case
there is again nothing to force a correlation between the concentration changes
of the two substrates, which could feasibly change in opposite directions.
Now, Reinhart Heinrich & Tom Rapoport (who reappear in later chapters
as two of the founders of Metabolic Control Analysis) managed to show, us-
ing Metabolic Control Analysis arguments,100 that in a simple linear pathway,
the cross–over theorem could be proved to be applicable. However, they also
showed that it could not be guaranteed to work in a variety of other circum-
stances. These included pathways where the effector causing the changes acts
on more than one site in the pathway, and pathways containing more than
one feedback loop, in which the cross–over could feasibly occur at the wrong
enzyme.
98 4 Traditional approaches to metabolic regulation

Thus the cross–over plot cannot be regarded as reliable for metabolic path-
ways in general. In particular, the failure to observe a cross–over when there
has been a change in metabolic flux cannot be taken as proof that an effec-
tor did not act on an enzyme’s kinetics. Even when a cross–over has been
observed, as in the yeast glycolysis example above, the significance of the
finding needs careful thought. We shall see later in this chapter, section 4.8,
that it cannot be taken as proof that glycolysis has been accelerated by an
activation of phosphofructokinase alone, since experiments to increase that en-
zyme’s activity specifically and substantially have not resulted in any change
in glycolytic flux.
Sir Hans Krebs suggested in 1957141 that a site of regulatory action might
be identified by the observation of anomalous changes in the substrate con-
centration of a non–equilibrium enzyme when comparing two metabolic states
with different fluxes. That is, if the flux through the enzyme had decreased,
whereas its substrate concentration had increased (which would usually be
expected to cause an increase in the flux), then presumably the enzyme must
have been inhibited by some other factor. (Changes in product concentration
were rather summarily dismissed as a factor that could account for the change
in rate of a non–equilibrium enzyme on the grounds that its reverse reaction
rate is negligible; the inhibitory effect that a product can have even on an
irreversible enzyme reaction was, as usual, discounted.) This is effectively a
simplified version of the cross-over theorem and therefore shares some of its
limitations.

4.7 Enzyme properties

A non–equilibrium enzyme might be assumed to be regulatory if:

• kinetic studies reveal that it is inhibited or activated by pathway metabo-

lites other than the substrate (e.g. allosteric enzymes, Chapter 3.5,
involved in feedback or feed–forward loops (Chapter 7.2 ), or

• the enzyme is found to be subject to activity change by reversible mod-

ification reactions (Chapter 7.4).

The thinking behind this assumption is that such mechanisms are not acci-
dental and must serve some function. This may well be true, though covalent
modification of enzymes by phosphorylation seems very widespread in eu-
karyotic cells, and it is not known in every case what effect this has on the
target enzyme — phosphofructokinase is an example of a glycolytic enzyme
that undergoes phosphorylation in many mammalian tissues, but in some of
these it is unclear what difference the phosphorylation makes to its proper-
ties. Furthermore, even if the molecular mechanism is brought into play in
some metabolic circumstances, this does not prove that the effect has had a
significant role in the situations under study. If an enzyme has its activity
changed by an effector, then the proof that this interaction is of regulatory
4.9 Conclusion 99

significance would have to include demonstration that the effector changes in

concentration in a manner consistent with the changes in metabolic rate, that
the effector concentration in the cells is in the range that affects the enzyme
activity, and that changes in the activity of that enzyme can affect the overall
metabolic rate. All this would involve studies at the level of the metabolic
system and could not be inferred from the molecular properties of the enzyme
in isolation.
In other words, the characteristics of the enzyme may contribute to an
understanding of the mechanisms of regulation and control, but they cannot
stand alone as evidence; there must be other supporting evidence. Further-
more, I shall present arguments later in the book to show why enzymes that
are not ‘rate–limiting’ must even so be controlled by mechanisms such as
covalent modification when the pathway flux is controlled (Chapter 8.1).

4.8 Metabolic mutants

It is arguable whether the use of mutants should be listed as a traditional
technique for investigating metabolic regulation, since it is only the recent
rise in molecular genetics that has made this a much more feasible experi-
mental approach. Mutants can now more easily be isolated or made with
altered regulatory properties in a specific enzyme. The gene for an enzyme
in one organism can be replaced by one with different characteristics taken
from another organism. The amount of a selected enzyme can be specifically
varied by increasing the number of copies of its gene or changing its promoter.
Some of these experiments have been carried out recently as part of quanti-
tative investigations in Metabolic Control Analysis (Chapter 6.1.1.1–6.1.1.5),
but similar techniques have been used in qualitative investigations of the role
of particular enzymes in metabolic regulation. Traditional techniques have
been used to isolate mutants in the regulatory properties of key enzymes in
the amino acid biosynthesis pathways of bacteria, and the properties of the
mutants have often been consistent with expectations.251 On the other hand,
the common view that phosphofructokinase is the rate–limiting enzyme of gly-
colysis has been undermined because: a) mutants with different forms of the
enzyme in E. coli generally show little difference in their metabolism,76 and
b) changing the amount of the enzyme expressed in yeast makes no difference
to the glycolytic flux.96
Thus in principle, mutant studies can be used to examine traditional con-
cepts of metabolic control; in practice, the results obtained in this way are
actually contributing to the undermining of the theory.

4.9 Conclusion
None of the lines of evidence described above has a single, unambiguous mean-
ing. In the search for regulatory enzymes and ‘rate–limiting steps’, it was
100 4 Traditional approaches to metabolic regulation

Fig. 4.6: The quest for

the rate–limiting step
From Trends Biochem.
Sci. 11, p13 (1986) by A
B Tulp.

usual to seek corroboration from several different approaches. Unfortunately,

that did not always lead to a single, uncontested answer, and there were many
cases where groups in different laboratories produced incompatible theories
about the mode of regulation of a particular metabolic pathway. The cartoon
in Fig. 4.6 makes fun of this. How some of these arguments have been resolved
will be explained in the remainder of the book.

4.10 Summary
1. For most of the past half century, the main aim in the study of metabolic
control was to find the ‘rate–limiting’ enzyme of a pathway. Candidate
enzymes should show one or more of the following characteristics.

2. The teleological argument is that there should be a rate–limiting enzyme

at the first unique step of a pathway.

3. The rate–limiting step should be a non–equilibrium enzyme, since other-

wise it could not have a significant role in metabolic control. Candidates
4.10 Summary 101

should be identifiable by comparing the mass action ratio to the equi-

librium constant. Alternatively, isotopic measurements of forward and
reverse fluxes should distinguish between near–equilibrium and non–
equilibrium enzymes.

4. A ‘rate–limiting’ enzyme is expected to have less capacity to work faster

than other enzymes, so should show a relatively low limiting rate value
in assays.

5. The pathway flux would be expected to be higher from intermediates

added after the rate–limiting step.

6. A cross-over in relative metabolite levels between two metabolic states

could indicate where a regulatory signal had acted on a rate–limiting
enzyme to change the flux.

7. Regulatory enzymes should show changes in activity in response to

metabolites other than their own substrates.

8. Alteration of the activity or regulatory properties of a rate–limiting

enzyme should directly affect the metabolic flux.

9. The results of such investigations are not always consistent. Conflicts

can then arise that cannot be readily resolved within the framework of
the rate–limiting step.

Further reading
1. Denton, R. M. & Pogson, C. I. Metabolic regulation, Chapman Hall,
London (1976).

2. Hales, C. N. Some actions of hormones in the regulation of glucose met-

abolism in Essays in Biochemistry (Campbell, P. N. & Greville, G. D.
eds), 3, pp. 73–104, Academic Press, London (1967).

3. Newsholme, E. A. & Start, C. Regulation in metabolism, Wiley & Sons,

London (1973).

4. Newsholme, E. A. & Gevers, W. Control of glycolysis and gluconeogene-

sis in liver and kidney cortex, Vitamins & Hormones, 25, 1–87, (1967).

5. Rolleston, F. S. A theoretical background to the use of measured con-

centrations of intermediates in the study of the control of intermediary
metabolism, Curr. Top. Cell. Reg. 5, 47–75, (1972).

6. Kwack, H. & Veech R. L., Citrate: its relation to free magnesium con-
centration and cellular energy, Curr. Top. Cell. Reg. 33, 185–207
(1992).
102 4 Traditional approaches to metabolic regulation

7. Fraenkel, D. G. Genetics and intermediary metabolism, Ann. Rev.

Genet. 26, 159–177, (1992).
 5

Metabolic control analysis

5.1 The problems of the traditional approaches

Biochemists have used the investigative methods described in Chapter 4 ex-
tensively for the identification of regulatory enzymes in metabolic pathways.
In many cases, several such enzymes were found in one pathway and this led
to arguments between different groups of researchers who all agreed that there
could only be one rate–limiting step (see p. 18), but could not agree which
one it was. Is alcohol dehydrogenase the rate–limiting enzyme of human
ethanol metabolism? Is the adenine nucleotide transporter the rate–limiting
step in mitochondrial oxidative phosphorylation? These are just two exam-
ples of long–running disputes. It was of course possible that no group was
completely right, but that some of them were partially right, because there
was no single rate–limiting step. But if the control of rate was spread over
a number of steps, how could their relative contributions be compared? The
experiments described up to now cannot answer this question because they do
not generate any quantitative measure of the degree of control of an enzyme
over the rate, or flux, of a metabolic pathway. Apart from this practical diffi-
culty in applying the concept of rate–limitation usefully, there are theoretical
difficulties:

• The definition states that the rate–limiting step is the slowest step in a
pathway, but how is this to be interpreted? In a metabolic steady–state
(p. 10) all the steps along a linear pathway are going at the same rate. If
‘slowest step’ is interpreted as the one least able to go faster, the types
of observation described in Chapter 4 do not appear to measure this
inability.

• By the 1930s, it was known that the rate of a sequence of simple chem-
ical reactions could depend to varying degrees on the rate constants of
all the reactions. In 1964, Stephen Waley 260 showed that the rate of a
sequence of unsaturated enzymes (i.e. enzymes for which all the metabo-

103
104 5 Metabolic control analysis

lite concentrations are below their Km values) depended non–linearly

on the kinetic parameters of all the enzymes. Thus in these cases, even
though they are certainly simpler than a metabolic pathway in vivo,
there is no theoretical basis for expecting that a unique rate–limiting
step inevitably exists.
• If a rate–limiting step exists in a pathway, then varying the activity of
that step alone will change the flux in the pathway, and varying any
other activity will have no effect whatsoever. There are few definitive
experimental observations of such a phenomenon, but many (examples
of which will be discussed later in this chapter) of pathways affected
by the activities of several of the steps. Furthermore, there have been
attempts to increase the rate of a pathway by using gene cloning tech-
niques to increase the amount of the supposed rate–limiting enzyme.
For example, when this was applied to phosphofructokinase in yeast
(the enzyme many biochemistry books cite as the rate–limiting enzyme
of glycolysis), a 3.5–fold increase in the amount of enzyme had no sig-
nificant effect on the anaerobic glycolytic flux.96
If the concept of the rate–limiting step must be replaced because it is inad-
equate both experimentally and theoretically, what characteristics would an
alternative have to posess? It would have to allow that several enzymes might
affect the flux in a pathway, and ideally it would suggest a basis for quanti-
tatively comparing the effects that these enzymes have on the flux. Similar
problems in other branches of science, and economics, have been tackled with
sensitivity analysis. This involves the assessment of how strongly a variable
(such as a pathway flux) responds to a change in any one of the factors that
might conceivably affect it. Just such an approach has been introduced into
metabolic biochemistry by a number of different researchers, beginning in the
1960s with Joseph Higgins.104 In 1969, Michael Savageau started develop-
ing the sensitivity analysis that eventually became part of his Biochemical
Systems Theory.217 In the early 1970s, Henrik Kacser & Jim Burns in Ed-
inburgh119 and Reinhart Heinrich & Tom Rapoport in Berlin99, 100 indepen-
dently made developments of Higgins’ work that were eventually amalgamated
into Metabolic Control Analysis (or Metabolic Control Theory). Biochemical
Systems Theory and Metabolic Control Analysis are rather different in ap-
proach, though they are basically compatible where they overlap. In Biochem-
ical Systems Theory, sensitivity analysis is just one part of a mathematical
method for modelling and simulation of metabolic and physiological systems
at a high level of generality. Metabolic Control Analysis is not intended to
be a complete approach to the modelling of metabolism; its principal concern
is with sensitivity analysis, and it maintains a closer link to the individual
underlying enzymic reactions than does Biochemical Systems Theory. Since
the early 1980s, a larger body of experimental work has accumulated related
to Metabolic Control Analysis than to Biochemical Systems Theory. For this
reason, in the rest of this book I will use the approach of Metabolic Con-
trol Analysis, though occasionally I will mention concepts that were derived
5.2 Flux control coefficients 105

a) The specimen pathway:

xase % ydh
X0 −→ S1 −→ · · ·Y −→ S6 · · · → X1
&
source Jxase Jydh sink

∂Jydh .... c)
b)
.... .... .......... ∂lnJydh Jydh
...... ........................................
∂Exase =Cxase
.. . ∂lnExase
........ .... ..................
j
..
...
.......... .... ...
.............
...... j .... .......
........... .... . .................
................. ..... . ..
. .
Flux, Jydh

...............

ln(Flux)
.. ...................
...
.. .. .....
.
. . . ....
.. .
....
.
.... ....
. .
.
... .
.....
.
...
..
e e
Concentration of enzyme, Exase ln(Concentration of enzyme, Exase )

Fig. 5.1: The flux control coefficient

a) A hypothetical specimen pathway. b) Typical variation of the pathway flux measured
at a step ydh, Jydh , with the amount of an enzyme, xase. The flux control coefficient at
e,j is the slope of the tangent to the curve ∂Jydh /∂Exase times the scaling factor e/j. c)
On a double logarithmic plot of the same curve, the flux control coefficient is the slope
of the tangent to the curve.

within Biochemical Systems Theory.

5.2 Flux control coefficients

The fundamental difference between the rate–limiting step concept and Metabolic
Control Analysis is highlighted if we consider the question we are asking in the
two cases concerning the relationship between the flux through a metabolic
pathway and the activity of a particular enzyme. In the former we ask ‘Is this
enzyme rate–limiting?’, and our answer is supposed to be ‘Yes’ or ‘No’. In
the latter we ask ‘How does the metabolic flux vary as the enzyme activity is
changed?’. This obviously invites a more detailed answer and allows all pos-
sibilities between ‘Not at all’ to ‘Greatly’. Furthermore, it can be posed as a
quantitative question: ‘How much does the metabolic flux vary . . . ?’. At the
time of the rise of Metabolic Control Analysis, there were hardly any relevant
experimental studies, which seems surprising since answering the question
106 5 Metabolic control analysis

only requires measurements of flux and enzyme activity, which are standard
techniques in metabolic biochemistry. There were theoretical grounds, such
as Waley’s equation cited earlier, for expecting the relationship between flux
and enzyme activity to appear concave or hyperbolic as shown in Fig. 5.1b).
The few experimental studies available at the time were consistent with this
sort of relationship, and it has since been found in other investigations, some
of which are referred to later in this book. In addition, the increasing avail-
ability of computers had made possible the first attempts at simulating the
behaviour of sequences of enzymes, allowing theoretical investigation of their
behaviour.
The important feature of the dependence of flux on the concentration of
one particular enzyme (Fig. 5.1b) is that the response to a change in the
amount of enzyme will vary depending on the position of the starting point
of the change on the x-axis. Near the origin, there could be an almost pro-
portional increase in metabolic flux as the enzyme amount is increased; at
this point, there might be some justification for terming the enzyme ‘rate–
limiting’. If the starting point for increasing the enzyme is part way along
the x-axis, then the response of the flux to an increase in enzyme is notice-
able, but weaker than before. If the starting point is far to the right on the
x-axis, then an increase in the enzyme has a small or negligible effect on the
flux. There is continuous variation of the response of the flux to the amount
of enzyme between these extremes. Unless some suitable measurements have
already been made, we do not know where the content of a chosen enzyme in
a given cell type will map on this curve. Metabolic Control Analysis begins
by defining a coefficient that quantifies this variable response and then finds
means of measuring and using it.

5.2.1 Definition

In the pathway shown in Fig. 5.1a), suppose that a small change, δExase , is
made in the amount of enzyme Exase , and that this produces a small change,
δJydh , in the steady state pathway flux, J, measured at the step catalyzed
by ydh. If the change is made small enough, then the ratio δJydh /δExase
becomes equal to the slope of the tangent to the curve of Jydh against Exase
as shown in Fig. 5.1b). In mathematical notation this tangent is represented
as ∂Jydh /∂Exase . Obviously this represents the steepness of the response of
the flux to the amount of enzyme, but has the disadvantage that its numerical
value and its units will depend on the units used to measure the flux and the
enzyme. This problem can be avoided if we compare the fractional changes in
the enzyme and flux, i.e. δExase /Exase and δJydh /Jydh ; since the numerator
and denominator of each fraction are measured in the same units, the result is
dimensionless. (If multiplied by 100, each of these fractional changes can be
Jydh
regarded as percentage changes.) The flux control coefficient Cxase is given
5.2 Flux control coefficients 107

by the ratio of these fractional changes as δExase /Exase tends to zero:

Á
Jydh δJydh δExase
Cxase ≈ (5.1)
Jydh Exase

This can be rearranged so that the flux control coefficient is expressed as the
tangent to a curve such as that in Fig. 5.1b) times a scaling factor Exase /Jydh :

Jydh ∂Jydh Exase

Cxase = . (5.2)
∂Exase Jydh

In turn, this is mathematically identical to the following definition, which is

useful because it shows that the flux control coefficient is the slope of the
tangent on a plot of the logarithm of flux against the logarithm of the enzyme
amount, as in Fig. 5.1c):

J
ydh ∂ ln Jydh
Cxase = (5.3)
∂ ln Exase

(Whether logarithms to base 10 or natural logarithms are used does not matter
as long as the same type are used for both axes.)

Supplement: Other definitions. The definition of the flux control coefficient

given above has the advantages both of simplicity and of relevance to cellular events
that change the amount of active enzyme in a cell. There are, though, more precise
ways of defining it. In many circumstances, the different definitions would lead to the
same answer, but in a few instances, the results could differ. In these cases, the more
technical definitions should be preferred because they lead to greater consistency
of interpretation, whereas the simple definition in terms of enzyme concentration
would be less well-behaved. The reasons for this are given in the Appendix to this
Chapter, More about flux control coefficients, section 5.7, p. 130. However, the
simpler definition, as used initially by Kacser & Burns, works appropriately with
the experimental results described in this book and will be retained here in order to
avoid terminological differences with the research papers describing the results.

5.2.2 Interpretation
What values might we expect flux control coefficients to have, and what do
they signify? For the pathway shown in Fig. 5.1, typical values are shown
in Fig. 5.2. If the curve had flattened out so that flux could no longer be
increased by further addition of enzyme, as would have been seen if the graph
had been extended further to the right, the flux control coefficient would be
0. It therefore seems that the values can range between 0 and 1. A value of
1 would correspond to a proportional relationship between the pathway flux
and the amount of enzyme. (This can most easily be seen from the Eqn. [5.1],
108 5 Metabolic control analysis

..................
..
...
...
..................
..
............ ≈ 0.2
..
...
............
.
.
........
..
........
.
.
.....
.
......
.
.... ≈ 0.5
Flux, Jydh .....
..
...
.. .
.
..
....
.
...
...
.
..
... ≈ 1.0
Concentration of enzyme, Exase

Fig. 5.2: Values of the flux control coefficient.

The values shown are approximately correct for the curve shown in Fig.5.1b).

which shows that in this case, a given fractional change in enzyme amount
produces an equal fractional change in the flux.)
It is also possible for a flux control coefficient to have a negative value. For
example, in a branched pathway where one metabolite can go to two different
end products, an increase in the activity of an enzyme in one branch might
increase the flux of the metabolite down its own branch, and thereby decrease
the amount available to flow down the other branch. Therefore the enzyme
could have a negative flux control coefficient on the flux in the other branch.
There is also a theoretical possibility that a flux control coefficient could
have a value greater than 1 in certain control structures found in metabolism
(e.g. Chap. 7.3.3, pp. 223 & 248). No experimental measurements have yet
demonstrated that such values can be found in metabolism but there is an
important piece of evidence from the genetics of diploid organisms that shows
that many enzymes have small flux control coefficients. Diploids generally
have two functioning copies in each cell of the gene for an enzyme, but if
this copy number is changed, the amount of enzyme is usually proportional
to the gene dose. In addition, there are many known examples of mutations
in genes for enzymes, both naturally occurring and artificially induced, that
result in the production of enzymically inactive protein. However, the impor-
tant finding is that these mutations are almost invariably recessive, that is, in
heterozygotes that contain one normal and one mutant copy of the gene, the
organism’s properties (or phenotype) appear normal. The normal gene is said
to be dominant to the mutant form. For example, in 1928, the geneticist R A
Fisher reported that 94% of mutations in the fruit fly Drosophila melanogaster
were recessive, 6% were semi–dominant, and none were dominant. Now en-
zymes affect the phenotype through the flux in metabolic pathways, so the
5.2 Flux control coefficients 109

observation that the wild–type is dominant effectively means that there has
been no discernable change in flux in the heterozygote, even though it contains
only half the enzyme present in the wild–type homozygote. In a number of
cases, it has indeed been confirmed by direct measurement that the metabolic
flux is only marginally reduced in heterozygotes. This is only possible if the
content of the enzyme in the normal homozygote is high enough for the flux–
enzyme curve to have flattened out (c.f. Fig. 5.2) to such an extent that
halving the enzyme content has little effect on flux, which is equivalent to
the flux control coefficient being much less than 1 and close to zero. Since
it is so common for mutations in enzymes to be recessive, the implication is
that most enzymes have small flux control coefficients. Furthermore, Kacser
& Burns121 also pointed out that the phenomenon of dominance has a simple
explanation in terms of the systems behaviour of biochemical pathways. Pre-
viously, dominance had been difficult to explain, and mechanisms such as the
one Fisher proposed assumed that it had been selected during the evolution
of the control mechanisms of the pathways. Recent experiments on artificial
diploids created from a normally haploid organism, the alga Chlamydomonas
reinhardtii, showed that these new diploids also exhibited dominance of the
wild type genes over mutants in 90% of cases177 . Since the recessive nature of
mutants in the diploid could not have been subject to evolution in this haploid
organism, the experiments are consistent with the interpretation devised by
Kacser & Burns, but not with the earlier idea that dominance was an evolved
property.
In the following subsection on the summation theorem, we shall see that
there are some further restrictions on the values that the flux control coeffi-
cients can have.
If the value of a flux control coefficient is known, approximate predictions
can be made about how the metabolic flux will change if the amount of enzyme
is changed. Such changes might be brought about by changes in genetic
expression or by covalent modification of inactive enzyme (Chap. 7.4, p. 229).
(Other methods of changing enzyme activity involve changing the catalytic
activity whilst the amount of enzyme is unchanged; the flux control coefficient
is also involved in predicting the response to these perturbations, as will be
explained in the section on the response coefficient, p. 124.) If the enzyme
concentration changes from Exase,1 to Exase,2 , by an amount small enough
Jydh
for Cxase to be effectively constant, Eqn. [5.2] can be integrated to give:
Jydh
∆ ln Jydh = Cxase ∆ ln Exase (5.4)

where ∆ ln Exase = ln Exase,2 − ln Exase,1 . That is, the change in the loga-
rithm of the flux is equal to the flux control coefficient times the change in
the logarithm of the amount of enzyme. As pointed out by Higgins104 this
corresponds to the power law form:

J = aE C (5.5)

where for simplicity, subscripts and superscripts have been omitted, so that
110 5 Metabolic control analysis

J
ydh
J = Jydh , E = Exase , C = Cxase , and where a is a constant that takes the
value needed to make the equation exactly true at the enzyme and flux values
for which the control coefficient was measured. Eqns. [5.4] and [5.5] both
describe the tangent in Fig. 5.1c), so any prediction made with the flux control
coefficient of the metabolic flux at other enzyme concentrations becomes less
and less accurate the further away the new enzyme content is from the one
at which the tangent was measured. What is less obvious is that the tangent
in Fig. 5.1c) is generally a curve on the linear scales of Fig. 5.1b), that lies
between the tangent shown on that graph and the true curve. Thus Eqns. 5.4
and 5.5 are always a better approximation to the flux–enzyme relationship
than the following linear equation (based on Eqn. [5.1]), which has often been
advocated:
∆Jydh Jydh ∆Exase
= Cxase . (5.6)
Jydh Exase

Although the equations above have limited predictive power for large
changes in enzyme content, in 1993 Rankin Small & Henrik Kacser233 de-
vised a surprisingly simple but improved predictor. Strictly, their result only
applies in cases where the flux–enzyme relationship, such as Fig. 5.1b), is rect-
angular hyperbolic (as seen in Michaelis–Menten enzyme kinetics; Chap. 3.2,
p. 47). That is to say, the flux and enzyme amount have to be related by an
equation of the form:
AExase
Jydh =
B + Exase
where A and B are constants; if this is the case, a plot of 1/Jydh v. 1/Exase
will be linear. As it happens, a number of experimentally–determined results
appear to obey this relationship, although there can be no guarantee that
this will always be true. Nevertheless, when it is true in a linear metabolic
pathway, measurement of the two values of the flux Jydh , J1 and J2 , at two
widely separated levels of the enzyme Exase , E1 and E2 , allows calculation of
the flux control coefficient at enzyme level E1 as:

J (J2 − J1 )E2
CE = (5.7)
(E2 − E1 )J2

This differs from the small change approximation to the flux control coefficient
given earlier as Eqn. [5.6] in that the weighting factor is E2 /J2 , and not E1 /J1 .
Oddly, if E1 /J1 is used as the weighting factor, the resulting control coefficient
is that at enzyme level E2 ! This equation had also been derived by Mark Stitt
in experiments on the control of photosynthesis that are described later in this
book.
Rankin Small and Henrik Kacser also noticed that, with these equations,
the flux control coefficient measured at one point allows prediction of the flux
at a markedly different enzyme concentration with good accuracy. Suppose it
is possible to increase the enzyme amount r times; the pathway flux will be
5.2 Flux control coefficients 111

higher by a factor f given by the equation:

1
f= (5.8)
1 − r 1 Cxase
r − Jydh

This function is plotted in Fig. 5.3. What this shows is that by changing
the amount of a single enzyme, the effects on the pathway flux can be quite
limited, unless the flux control coefficient is greater than 0.5 to start with.
The maximum change in flux that can be obtained by a very considerable
increase in enzyme, when r À 1 (so that r ≈ r − 1), is:

1
f= ydh J
(5.9)
1 − Cxase
Jydh
Thus if Cxase = 0.5, the maximum flux that can be obtained is a factor of
Jydh
2.0 times greater. Only if Cxase is near to 1.0 are very large changes in flux
possible, but as mentioned previously, few measurements have been made of
enzymes with such large values for their flux control coefficients. This result
is very disappointing since it implies that biotechnologists who might have
hoped to increase significantly the flux through a chosen metabolic pathway
by engineering an increase in a single target enzyme, will rarely find it possible.
This limitation only applies, however, to increasing the metabolic flux; the
same restriction does not apply for a decrease in flux caused by a reduction
in the amount of active enzyme (i.e. r < 1), since this can always lead to a
substantial drop in flux.
Equation [5.8] and Fig. 5.3 also have important consequences for how cells
can control metabolic rate, and these will be discussed later (Chapter 8.1).

5.2.3 The summation theorem

When Kacser & Burns first developed Metabolic Control Analysis,119 they
found that, if all the enzymes that can affect a particular metabolic flux in a
cell or a metabolic system are taken and the values of their control coefficients
on that flux added up, the sum comes to 1. Thus if we number all the enzymes
involved from 1 to n, then:

C1J + C2J . . . + CnJ = 1 (5.10)

This relationship is known as the summation theorem for flux control coeffi-
cients, and in mathematical notation is expressed as:

n
X
CiJ = 1 (5.11)
i=1

Pn
where i=1 is the mathematical symbol for summing the n items indicated
by the following term. There are several ways of proving this theorem; an
112 5 Metabolic control analysis

10

Relative flux, f
6
50

20
4
10
5
2
2

0
0 0.2 0.4 0.6 0.8 1

Flux control coefficient

Fig. 5.3: The relative change of flux for large changes in enzyme amount.
The increases in flux predicted by Eqn. [5.8] against the value of the flux control coef-
ficient. The degree of amplification, r, of the enzyme amount is shown on each curve.
The graph can be used in reverse to give a quick estimate of the control coefficient if
the relative change in flux has been measured for a known change in enzyme amount.

example is given in the Appendix to this chapter, More about flux control
coefficients, p. 130.
The summation theorem shows that the enzymes of the pathway can share
the control of flux. In a linear pathway consisting of enzymes with normal
kinetics (i.e. where substrates stimulate and products inhibit the reaction
rate), all the flux control coefficients must be zero or positive, so the maximum
value any enzyme’s coefficient could have is 1, when all the other enzymes
would necessarily have flux control coefficients of zero. In this case, this one
enzyme could be said to be ‘rate–limiting’, since a flux control coefficient
of 1 corresponds to a proportional relationship between the activity of an
enzyme and the pathway flux. The summation theorem shows that this is
not a necessary feature of the pathway, as it is also possible for some or all
of the enzymes to have values greater than zero but less than 1. Note that
the definition states that the sum is for the flux control coefficients of all
the enzymes in the metabolic system; for example, this potentially implies
the whole cell. In practice, we would expect a pathway flux to be influenced
mainly by enzymes in that pathway, and perhaps by a few closely–connected
pathways, and that distantly–connected enzymes would have little influence at
all. In other words, the flux control coefficients of hundreds, even thousands,
of enzymes in a cell on our chosen flux will be zero, so even though the control
of flux is shared, it is not shared evenly.
5.3 Elasticities 113

However, another consequence of the highly branched and interconnected

nature of metabolism is that for any particular pathway, there will be other
pathways that could draw material or energy away from it causing its flux to
decrease, so that enzymes in these pathways could have negative flux control
coefficients on the flux of interest. Although this does not invalidate the sum-
mation theorem, nor the concept that control of flux is shared, it does mean
that it is not possible to state definitively that the values of individual flux
control coefficients can never exceed 1. This is because if there are negative
flux control coefficients, one or more flux control coefficients could have values
greater than 1. On the other hand, although negative flux control coefficients
have been measured experimentally, there are virtually no examples of an
enzyme with a flux control coefficient exceeding 1. One exception occurs in
mitochondrial oxidative phosphorylation at high respiration rates, and will be
described later (see Chap. 6.3.3.1, p. 186).
The summation theorem also shows that the flux control coefficient of an
enzyme is a system property. This can be seen when a flux control coefficient
varies with the amount of enzyme, as shown previously in Fig. 5.2. Since
the flux control coefficient decreases with an increase in the amount of the
enzyme, the summation theorem requires that the flux control coefficients of
some other enzymes must be increasing at the same time to maintain the sum
of all flux control coefficients constant at one. In addition, this must work the
other way round: the flux control coefficient of the enzyme we are considering
will be altered if the activity of some other enzyme is changed sufficiently to
affect its own flux control coefficient. This shows that an enzyme’s flux control
coefficient is not an intrinsic property of the enzyme itself; it is a property
of the whole system. Therefore, the value of a flux control coefficient cannot
be determined by considering the properties of the enzyme in isolation; the
characteristics and amounts of the other enzymes in the metabolic system will
affect the result. Furthermore, since the values of the flux control coefficients
can redistribute between enzymes according to circumstances, any particular
measured values apply only to the metabolic state in which they were deter-
mined. Examples showing the change in the pattern of control in different
metabolic states will be presented in the next chapter (e.g. Fig. 6.15, p. 165).

5.3 Elasticities
Although the last section has demonstrated the important conclusion that
the flux control coefficient of an enzyme is a system property that cannot be
related solely to the enzyme’s properties in isolation, there must of course
be links between the enzyme’s kinetic properties and its potential for flux
control. This can be illustrated by considering what happens in a pathway if
an enzyme activity is changed. For example, in the pathway:
114 5 Metabolic control analysis

Scheme 5.1
xase ydh zase
X0 −→ Y −→ Z −→ X1

suppose that an extra amount of ydh is added, to increase the rate of the
second step. What is the effect on the pathway? The increased amount of
ydh tends to lower the concentration of Y . The lower Y will:

• Increase the rate of xase because of reduced product inhibition.

• Decrease the rate of ydh because of lower substrate concentration.

The increased amount of ydh also tends to raise the concentration of Z. The
increased Z will:

• Decrease the rate of ydh because of increased product inhibition.

• Increase the rate of zase because of higher substrate concentration.

In conclusion:

• The effects of the increased amount of ydh involve the relative sizes of
the responses of the enzymes to the pathway metabolites.

• The effects on the metabolites could tend to counteract the change in

the amount of enzyme.

• The effects on the metabolites could tend to change the rates of neigh-
bouring enzymes to match the change in ydh.

This explanation has shown that the flux control coefficient of an enzyme
is likely to be linked to its kinetic responses to changed metabolite concen-
trations, as well as its ability to influence the concentrations of metabolites
in the pathway, a linkage that was shown first by Heinrich & Rapoport.99 To
be more specific about how the kinetics of enzymes relate to the flux control
coefficients, we need a means of describing their kinetic responses in a manner
that is compatible with the control coefficients. This measure is provided by
the elasticity coefficient.

5.3.1 Definition and examples

Unlike control coefficients, elasticities are properties of individual enzymes
and not of the metabolic system, even though we shall see that they are
defined in a very similar way. The elasticity of an enzyme to a metabo-
lite is related to the slope of the curve of the enzyme’s rate plotted against
metabolite concentration, taken at the metabolite concentrations found in the
pathway in the metabolic state of interest. Again, because the value of this
slope would depend upon the units of measurement, it is scaled to give a
5.3 Elasticities 115

a) ∂vxase .... b)
∂S .... ... ..........
..... ................................
∂lnvxase =εxase
............ ∂lnS S
. . ......
v .... ...... ....
.... ..
................
.... ...
..........
.
..... v .... .......
........... ..... ..............
. ..
Rate, vxase

. ............

ln (rate)
... .. ............
..... .
....
..
.
..
.....
.... .
....
... ......
..... ...
...
..
....
s s
Metabolite concentration, S ln (metabolite concentration, S )

Fig. 5.4: The elasticity coefficient

a) Typical variation of the rate of enzyme xase, vxase , with the concentration of metabo-
lite S. The elasticity coefficient, εxase
S , at s, v is the slope of the tangent to the curve
∂vxase /∂S times the scaling factor s/v. b) On a double logarithmic plot of the same
curve, the elasticity coefficient is the slope of the tangent to the curve.

dimensionless measure (Fig. 5.4a). It can be obtained directly as the slope

of the logarithm of the rate plotted against the logarithm of the metabolite
concentration (Fig. 5.4b).
The mathematical form of the definition is just like that of the flux control
coefficients given in Eqn. [5.2]. For example, the elasticity coefficient for the
effect of metabolite S on the velocity v of enzyme xase is the fractional change
in rate of the isolated enzyme for a fractional change in substrate S, with all
other effectors of the enzyme held constant at the values they have in the
metabolic pathway:

∂vxase S
εxase
S = .
∂S vxase
(5.12)
∂ ln | vxase |
=
∂ ln S

(The term ln | vxase | in this equation means that the logarithm of the abso-
lute value of vxase is used in this equation. That is, if vxase is negative, as
will be seen in Fig. 5.8, its positive value is used because it is not possible
to take the logarithm of a negative number.) Elasticities have positive values
for metabolites that stimulate the rate of a reaction (substrates, activators)
and negative values for those, like products and inhibitors, that slow the reac-
tion. Obviously, there must be some link to the results from enzyme kinetics
described in Chap. 3. In fact, the typical range of values that elasticities
116 5 Metabolic control analysis

have can be illustrated by examining some common equations from enzyme

kinetics.
1.620

1.610

1.600
log v

S = 0.5 mM
1.590

1.580
-0.335 -0.315 -0.295 -0.275

log S

Fig. 5.5: Graphical estimation of an elasticity from an enzyme rate law.

The rate of reaction has been calculated for the Michaelis–Menten equation with K m
= 0.75 mM and V = 100 µmol.min-1 for substrate concentrations from 95% to 105%
of the required value of 0.5 mM. The elasticity is 0.60, the slope of the tangent at S =
0.5 mM. The curve of rate against substrate concentration has not been drawn because
the curvature is so slight it barely separates from the tangent.

For example, suppose we want the elasticity with respect to substrate of

an enzyme that obeys the Michaelis–Menten equation (Eqn. [3.1]):
SV
v=
S + Km
If the substrate concentration at which the elasticity is required is 0.5 mM,
with the Km = 0.75 mM and V = 100 µmol.min−1 , we can calculate the rate
at 95%, 97.5%, 100%, 102.5% and 105% of the substrate concentration. When
the logarithms of these rate values are plotted against the logs of the substrate
concentrations, the elasticity is the slope of the graph at 100% of S = 0.5 mM
(Fig. 5.5). If you know the properties of logarithms, you may be able to spot
that the limiting rate, V , moves the curve up and down the y-axis but does
not change its slope. (This is because V is a constant multiplying factor in the
calculation for v, and thus becomes a constant additive term in log v.) Thus
it does not matter what value of V is used for the calculation; an arbitary
value of 1 suffices. Apart from the arithmetic convenience, this helps with the
biochemistry because it is not necessary to know the amount or activity of an
enzyme in a cell to know its elasticity. Elasticity values can also be obtained
5.3 Elasticities 117

1.0 ...
...
...
... . ..... .... ..
... .... .... .... .... .... ...
0.8 .... .... ....
... ... . ....
... ....
... ...
... ... .
0.6 ... ..
v/V or εvS ... .
....
........
.. ......
0.4
.. .....
......
.. .......
.........
.. ...........
.
. ..............
0.2 ..................
.. ...........................
.......................
..
0.0 ..
0 2 4 6 8 10
Relative concentration, S/Km,S

Fig. 5.6: Elasticity of an irreversible enzyme with respect to its substrate at varying
substrate concentrations.
The enzyme is an irreversible single–substrate enzyme operating in the presence of
its product, according to the equation v = V(S/Km,S )/(1 + S/Km,S +P/Km,P ), i.e.
Eqn. [3.4] with Keq = ∞ and P/Km,P = 0.2. : εvS . — — — : fractional velocity,
v/V.

from a rate equation by use of differential calculus; if you are familiar with
this, you will find more details, including the result for the Michaelis–Menten
equation, in the Appendix to this chapter More about elasticities, p. 134.
Examples of the elasticity values obtained for some simple enzyme mech-
anisms are shown in the next few figures. Fig. 5.6 shows how the substrate
elasticity of an irreversible Michaelis–Menten enzyme varies with substrate
concentration; its value goes from 1 at substrate concentrations much below
Km to 0 as the enzyme approaches its maximal velocity. Note that the val-
ues have been calculated for the presence of a fixed concentration of product,
which binds at the active site even if there is no appreciable reverse reaction.
This inhibitory effect of a product, even with an irreversible enzyme, means
that the enzyme has an elasticity with respect to the product, as illustrated
in Fig. 5.7. The value is negative, going as low as -1 when the enzyme is
strongly inhibited, because increasing the product concentration reduces the
rate.
For enzyme reactions that are near to equilibrium, the elasticities with
respect to substrate and product are mostly determined by the degree of dis-
placement of the reaction from equilibrium rather than by the kinetic details
of the reaction, and this can lead to much larger elasticity values than we have
seen so far (as was first pointed out by Kacser and Burns,119 and later de-
118 5 Metabolic control analysis

veloped in more detail by Bert Groen and his colleagues90, 273 ). For example,
consider the reversible Michaelis–Menten equation for the conversion of S to
P , Eqn. [3.4] where v is the net rate (positive for formation of P ). Algebraic
calculation of the elasticity with respect to substrate by differentiation and
scaling of this equation gives:

1 S/Km,S
εvS = −
1 − ρ 1 + S/Km,S + P/Km,P
1 vf
= − (5.13)
1−ρ Vf

where ρ is the disequilibrium ratio (Eqn. [1.6]) and vf the total forward rate.
The first term on the right hand side depends on the degree of displacement
from equilibrium, going from 1 in the absence of product to infinity at equilib-
rium, and the second term represents the fractional saturation of the enzyme
with S, going from 0 in the absence of S to a maximum value of 1, giving the
overall result illustrated in Fig. 5.8. (Note that S À Km,S is not, in this case,
a sufficient condition for the saturation term to reach 1, because the value
of P counts, particularly if P À Km,P ; ‘saturated’ can be a vague concept
with a near–equilibrium reaction.) In many practical cases, the elasticity of a
near–equilibrium reaction will be given with sufficient accuracy by the term

0.6
...
...
....
0.3 ..... .
... ..... .....
..... ..... ..... ..... .....
..... ..... ..... ..... ..... ..... .....
0.0 ...
...
v/V or εvP

...
...
...
-0.3 ....
....
.....
.....
......
.........
-0.6 ...........
.............
................
.......................
....................................
.......
-0.9
2 4 6 8 10
Relative concentration, P/Km,P

Fig. 5.7: Elasticity of an irreversible enzyme with respect to its product at varying
product concentrations.
The enzyme is an irreversible single-substrate enzyme operating in the presence of
its product, according to the equation v = V(S/Km,S )/(1 + S/Km,S +P/Km,P ), i.e.
Eqn. [3.4] with Keq = ∞ and S/Km,S = 1. : εvP . — — — : fractional velocity,
v/V.
5.3 Elasticities 119

10 ... .... ..
...
... ..... ..... ....
... .. .....
... ..... ..... ...
... . .....
5 ... ...
.. ....
... ..
... ... ..
.........
... ...................................
20v/V or εvS . .. .........................................................................
0 .......... ...
..... ...
.....
... .....
... ....
. ...
-5 . ...
...
...
...
...
..
-10
0 1 2 3
Relative concentration, S/Km,S

Fig. 5.8: Elasticity of a reversible Michaelis Menten enzyme with respect to its
substrate at varying substrate concentrations.
The enzyme is a reversible single–substrate enzyme operating in the presence of its
product, according to Eqn. [3.4] with Km,S =1; P = 5; Km,P = 5; Keq = 10. The
elasticity is undefined at equilibrium (S = 0.5). : ε vS . — — : 20 × fractional
velocity, v/V.

1/(1 − ρ), which generates very large values as ρ approaches 1. The product
elasticity is given similarly by:
−ρ P/Km,P
εvP = −
1 − ρ 1 + S/Km,S + P/Km,P
−ρ vr
= − (5.14)
1 − ρ Vr
Where the saturation terms are negligible, it follows that the two elasticities
are related as:
εvS + εvP = 1 (5.15)
Groen90 also showed that similar results can be obtained with multisubstrate
rate equations having numerators of similar form to that of Eqn. [3.4].
Large elasticity values can also be obtained with allosteric enzymes, even
when (as is often the case) their reactions are not close to equilibrium. With
an allosteric enzyme that has cooperative kinetics with respect to a substrate,
the elasticity can be greater than 1, although it must be less than the Hill
coefficient (see Eqn. [3.11]). In fact, Herbert Sauro showed when he was
a research student with me that, if the enzyme is an irreversible enzyme
whose rate of reaction is proportional to its fractional saturation, Y , with the
substrate S, then:
εvS = (1 − Y )nh (5.16)
120 5 Metabolic control analysis

3.0

2.5

2.0

εvS or Y or h 1.5

1.0

0.5

0.0
0 5 10 15 20

Fig. 5.9: The substrate elasticity of an allosteric enzyme.

The enzyme obeys the Monod, Wyman & Changeux model, Eqn. [3.12], with n = 4, L
= 1000, c = 0.01 and KT = 1 so that S = α. : εvS ; — — – : fractional saturation,
Y; · · · : the Hill coefficient, h.

In other words, the elasticity is similar to the Hill coefficient at low degrees of
fractional saturation, when Y is near zero, but becomes progressively smaller
as the degree of saturation increases. Since the Hill coefficient for an enzyme
showing positive cooperativity is 1 at low saturation, becomes perhaps as high
as 2 — 3 at mid–saturation, and then decreases to 1 again at high saturation,
the elasticity in the same range will start at 1, perhaps rise above 1 and
then fall away to zero (Fig. 5.9). Of course, if the Hill coefficient is greater
than 1, the enzyme will have a higher elasticity than it would if it were
non–cooperative at an equivalent degree of saturation. (For an irreversible
Michaelis–Menten enzyme, nh = 1, and the equation above gives the same
result for the elasticity as Fig. 5.6 and Eqn. [5.34] in the Appendix.)
These examples have shown how elasticity values can be obtained from
equations for enzyme kinetics. However, this can only be done for an enzyme
in a cell if:
• a complete kinetic equation is available that incorporates all the reac-
tants and effectors that occur in the cell and affect the enzyme;
• the intracellular concentrations of all these metabolites in the compart-
ment where the enzyme occurs are known, and
• all the kinetic parameters are known under intracellular conditions.
This is quite a demanding set of requirements, often difficult to satisfy because
5.3 Elasticities 121

of the patchy nature of the published results on enzymes. For instance, any
particular enzyme is likely only to have been studied in a few organisms, and
not necessarily in the one under study. Also, enzyme kinetics experiments
often do not report the right type of results. This is partly for reasons con-
sidered in more detail in the Appendix More about elasticities, p. 134. There
are also undiscovered reasons, examples of which have been given previously
(Chapter 4.4, p. 93). As an alternative to determination from kinetics, there
are methods for direct determination of elasticities in vivo (see Chapter 6.3,
p. 170).

5.3.2 Use and interpretation

The function of elasticities in Metabolic Control Analysis is as a quantita-
tive replacement for the vague concepts of responsiveness of an enzyme to a
metabolite that are used in qualitative explanations of metabolic regulation,
such as:
The substrate S controls the rate of reaction because its concen-
tration in the cell is well below the Km of the enzyme for S . . .
or
The rate of the enzyme will be relatively unresponsive to variations
in the concentration of S because it is well above the Km . . .
The values of elasticities are much more useful than such statements because
they are determined with the concentrations of other substrates, products
and effectors at their levels in the cell in the metabolic state being analysed.
In spite of biochemists’ attachment to simple Michaelis–Menten kinetics, the
rectangular hyperbola of the single–substrate enzyme in the absence of prod-
ucts is a poor guide to the behaviour of enzymes in vivo where the majority
have more than one substrate, are in the presence of appreciable concentra-
tions of products, and may be catalyzing a significant flux in both directions.
An elasticity, however, reflects the results of all these influences, and so it
would be a useful advance if biochemists agreed to use it as an indication
of the sensitivity of an enzyme to a metabolite, even in circumstances where
they had no intention of going any further into Metabolic Control Analysis.

5.3.3 The connectivity theorem

Now that we have defined how metabolites affect the activity of enzymes, we
can return to the question of how the flux control coefficients of enzymes can
be related to the kinetic properties of the enzymes. This link is provided by
the connectivity theorem 119 derived by Kacser & Burns in 1973. Suppose we
choose one pathway metabolite S and find all the enzymes in our metabolic
system whose rates respond to it; further suppose that we find there are three
and let us label these enzymes i, j and k. The connectivity theorem states
that, for each of these enzymes, if we form a term by taking its flux control
122 5 Metabolic control analysis

coefficient on a particular flux J and multiplying by its elasticity with respect

to S, then the sum of the terms is zero, i.e. :

CiJ εiS + CjJ εjS + CkJ εkS = 0 (5.17)

The mathematical statement of this theorem is:

n
X
CiJ εiS = 0 (5.18)
i=1

This is equivalent to the previous equation, even though the sum is now formed
of all the n enzymes in our system; this is because the enzymes that are not
affected by S will have elasticities of zero and will therefore not contribute
to the sum; perhaps only the enzyme that produces the metabolite and the
one that consumes it will have non–zero elasticities. To see how this works in
practice, consider a short section of glycolysis:
Scheme 5.2
enolase pyruvate kinase
· · · −→ PEP −→ ···

The specific connectivity theorem involving PEP is obtained by writing a

term for every enzyme that has a non–zero elasticity for PEP; in this case,
we assume this involves only enolase (denoted eno) and pyruvate kinase (pk),
giving:
J J pk
Ceno εeno
P EP + Cpk εP EP = 0 (5.19)
or
J
Ceno εpk
P EP
J
= − eno (5.20)
Cpk εP EP
Without knowing anything more about the rest of the pathway, we cannot tell
the magnitude of either flux control coefficient, but we can see that the rela-
tive values of two successive flux control coefficients depend on the elasticity
representing the product inhibition (i.e. PEP on enolase) and the substrate
activation (i.e. PEP on pyruvate kinase) by the intermediate metabolite. For
example, in working rat heart, supplied with glucose but no insulin, these two
J
elasticities have been estimated as about −0.69 and 0.86, implying that C eno
J
is ≈ 1.25 times Cpk . This ratio form of the connectivity theorem (Eqn. [5.20])
shows the tendency of large elasticities (e.g. for enzymes catalyzing reactions
near equilibrium) to be associated with small flux control coefficients, and
vice versa.

Supplement: More complicated connectivity relationships

There is a complication in the connectivity theorem when the metabolite is one
of a pair such as NAD+ and NADH, whose ratio can vary even though the total
of NAD units in the cell is effectively constant.69 A similar problem can arise with
larger ‘moiety–conserved’ groups, such as ATP, ADP and AMP. In essence, the right
5.3 Elasticities 123

hand side of Eqn. [5.18] for a single one of these metabolites is no longer equal to
zero, but the equations can be combined in pairs to give a zero result, with the side
effect that the number of connectivity equations is reduced by one.69 This technical
complication does not invalidate the general properties and uses of the connectivity
theorem.

Since we can choose any of the pathway metabolites to form a connectivity

equation, there are as many of them as there are variable metabolites in the
pathway. Do we gain any advantage from having all these different equations
involving control coefficients? The answer is ‘yes’ because the summation
theorem and the set of connectivity theorems for all the metabolites of a
linear pathway provide exactly the number of simultaneous equations needed
to solve for the flux control coefficients of all the enzymes in terms of the
elasticities. For example, we can consider a two–enzyme pathway with one
pathway intermediate, Y :
Scheme 5.3
xase ydh
X0 −→ Y −→ X1

This pathway has a summation theorem:

J J
Cxase + Cydh =1

and a connectivity theorem:

J
Cxase εxase
Y
J
+ Cydh εydh
Y =0

If the elasticities are known, these equations form a pair of simultaneous

equations in the two unknown flux control coefficients. In this case, the exact
solutions are:

J εydh
Y
Cxase =
εydh
Y − εxase
Y
J −εxase
Y
Cydh =
εydh
Y − εxase
Y

Thus we have shown that although the flux control coefficients are system
properties of the pathway, they are nevertheless explicable in terms of the
kinetic properties of the constituent enzymes.
This is not a special result for a simple pathway; for a three–enzyme linear
pathway there will be two intermediates, so the one summation theorem and
two connectivity theorems will give the three equations needed to express the
three flux control coefficients in terms of the elasticities. Whatever the length
of a linear pathway, the procedure will work, though this is not the case for
branched pathways or pathways containing certain types of cycle (such as
124 5 Metabolic control analysis

substrate cycles). For these pathways, my student Herbert Sauro and I69
proposed additional equations relating the flux control coefficients and the
relative fluxes through different parts of the system. These equations, known
as the branch–point and substrate cycle theorems, make the flux control coef-
ficients expressible in terms of the elasticities and the flux distribution in the
system. Since then, detailed mathematical analysis by other researchers (in
particular Christine Reder in Bordeaux) has shown that it is always possible
to write a soluble set of equations for any pathway, provided that the path-
way can reach a proper steady state. The branch–point equations will not be
described any further for the moment, but their use will be illustrated later in
an experimental study of oxidative phosphorylation in Chap. 6.3.3.1, p. 186.

5.4 Response coefficients

There are many control mechanisms that operate on metabolic pathways,
and some of these operate on the catalytically active amount of an enzyme,
others do not. In the former category are mechanisms for the induction and
repression of enzyme synthesis in response to hormonal and environmental
stimuli, the activation or inactivation of existing enzyme protein by covalent
modification reactions, and probably targeted protein degradation. In the
latter category are mechanisms that change the kinetic characteristics of the
enzymes, such as allosteric effectors changing the cooperativity or affinity of
an enzyme for its substrate, or physiological control mechanisms that alter the
concentration of the source metabolite for a pathway, or of an effector that
is external to the pathway. Some biochemists have therefore been concerned
that flux control coefficients seem only to relate to the first group of control
mechanisms and that they are not obviously relevant to describing the second,
very important, group. Later (Chapter 8.1) we will consider the possibility
that the functions of this second group are not primarily flux control, in
spite of what is usually claimed. For the moment, let us assume that the
conventional view of the second group is correct. In this case, the argument
for maintaining that flux control coefficients are still significant in spite of
the criticism of their relevance has two stages: firstly, it is accepted that it
is useful to define control coefficients for the response of metabolic fluxes or
concentrations to factors (parameters) other than enzyme concentration, but
secondly it is shown that the control coefficients we have already considered
always have an influence on these other types of coefficient.
Let us consider the first stage. Why not define a coefficient for the action
of some parameter P other than enzyme concentration on a metabolic flux?
There is no problem at all, provided that the parameter is just that: a factor
that influences the behaviour of the system, but that can be held constant
after it has been altered while the system reaches a new steady state. On this
basis, a metabolite that is synthesized and degraded by the system is not a
parameter, but a variable, because if some alteration is made to its concentra-
tion, this alteration will fade away as the system returns to its original steady
5.4 Response coefficients 125

state. The strength of the effect of the parameter P on, say, the flux Jydh is
defined in the same way as the control coefficient, except that it is generally
J
referred to as a response coefficient, RPydh , as originally named by Kacser and
119
Burns:
J ∂Jydh P
RPydh = . (5.21)
∂P Jydh
This equation for the response coefficient can also be reexpressed in the same
ways as those for the flux control coefficients, Eqns. [5.3–5.5] and has the same
graphical interpretation as shown in Fig. 5.1.

Supplement: Other definitions. An equivalent to the response coefficient

was defined by Crabtree and Newsholme55 (as a net sensitivity) and by Michael
Savageau214 (as a logarithmic sensitivity). The difference of emphasis between these
latter authors and the adherents of Metabolic Control Analysis comes in the second
step in the argument, originally made by Kacser & Burns119 and described below.

For the second stage of the argument, Kacser & Burns pointed out that,
if an external, constant metabolite P acts on the flux Jydh through being an
effector of the pathway enzyme xase, the response coefficient for the effect
of P is composed of the flux control coefficient with respect to xase and the
elasticity of xase with respect to P :

J J
ydh xase
RPydh = Cxase εP (5.22)

This equation has been proved in a number of ways, but in essence, it is a

mathematical necessity given the definitions of the terms. Its importance is
that it shows that the response of a pathway to an effector depends on two
factors:
• the sensitivity of the pathway to the activity of the enzyme that is the
target for the effector (given by the enzyme’s flux control coefficient),
and
• the strength of the effect of P on that enzyme (given by its elasticity).
It is necessary for both components to be non–zero for P to be able to affect
the pathway. Further, if there were a number of effectors that could act
on a pathway at a particular enzyme, their response coefficients would all
contain the same flux control coefficient as a component term. Thus the reason
that the flux control coefficient is valuable is that it can indicate whether
an enzyme has the potential to be a site at which a pathway is controlled
(i.e. its regulatory capacity,106 because the flux control coefficient is defined
without reference to the existence of any particular mechanism by which the
activity can be controlled. An effector cannot act on a pathway by changing
the activity of an enzyme with a near–zero flux control coefficient. (The
126 5 Metabolic control analysis

exception to this statement is when a very large increase occurs in an inhibitor

of a pathway; even if the inhibited enzyme had a flux control coefficient near
zero originally, it will increase as the inhibition builds up, so the response
coefficient to the inhibitor may start small, but will eventually rise. Inhibitor
titrations are considered in more detail in the next chapter, 6.1.4, p. 155.)
Another useful property of Eqn. [5.22] is that it is true regardless of the
mechanism by which the effector P acts on the enzyme xase. For example,
P might actually be a K–system allosteric inhibitor that exerts its effect by
raising the S0.5 of xase for its substrate (see Chap. 3.5, p. 71). The response
coefficient equation still applies in this case because the elasticity of xase with
respect to P takes into account that the change in activity of the enzyme
occurs at constant concentration of its substrate. On the other hand, if P did
directly affect the limiting rate of the enzyme, the equation would be equally
applicable; it would just be that the elasticity of the enzyme with respect to
P would be implemented by a different mechanism. If it seems surprising
that Metabolic Control Analysis can characterize the response of metabolism
to an effector without its being important how the effect is implemented at
the molecular level, remember that Control Analysis is intended to be a sys-
tems theory of metabolic control that aims to explain the general aspects of
metabolic behaviour without being dependent on the full details of molecular
mechanism.

Supplement: Response to a kinetic parameter. There is no particular

restriction on the factors for which a response coefficient can be defined. For ex-
ample, suppose we wanted to know how a metabolic pathway would respond to a
change in one of the kinetic parameters of enzyme xase, even though we had not
yet found a means of making this change. The response coefficient is once again the
flux control coefficient times the elasticity of the enzyme with respect to the kinetic
parameter. Elasticities can be defined with respect to a kinetic parameter in exactly
the same way as to a metabolite concentration274 and can be calculated from the
enzyme rate law. In many cases, if the kinetic parameter K is a Km or Ki value
for a substrate, product or effector molecule S, then εxase
K will equal -εxase
S . In fact,
in the Appendix to this Chapter (Section 5.7: More about flux control coefficients),
it is shown that the combination of the response coefficient to a kinetic parameter
and the elasticity to that same parameter leads to a more general definition of the
flux control coefficient, compared to the definition given earlier solely in terms of
enzyme concentration (Eqn. [5.2]).

What happens if an effector P acts on more than one enzyme in a metabolic

pathway? In this case, the total response will be the sum of the individual
responses from each enzyme affected.106 (This is only true for very small
changes in P , because the response coefficient, like the other coefficients, is
defined as a first order approximation, true for very small changes. For a large
change in P , the total effect will not be the sum of the effects on each en-
zyme because of the non–linear nature of the kinetics of metabolic systems.)
As with the connectivity relationship (Eqn. [5.18]), we can define the overall,
5.4 Response coefficients 127

multi–site response obtained from the n enzymes of the system as:

n
X
J
RP = CiJ εiP (5.23)
i=1

Contributions to the sum only come from the enzymes for which εiP is not
zero. This equation for the multi–site response coefficient does show the po-
tential value of control mechanisms in which an effector acts on more than
one site. In the case where the effector acts only on one enzyme, then it would
be rare for that enzyme to have a flux control coefficient as large as one; the
elasticity of the effector on the enzyme might be around 1, so overall, the re-
sponse coefficient is most likely to be less than 1, i.e. there will be less than a
1% change in flux for a 1% change in the effector concentration. If the effector
acts on several of the enzymes of the pathway that together account for most
of the flux control (i.e. their flux control coefficients add up nearly to 1), and
each of the elasticities is near 1, then the response coefficient can be close to
1. (If the effector acts cooperatively on allosteric enzymes, then the elasticity
terms could be greater than 1 in both cases considered.) Furthermore, there
is likely to be a further, hidden advantage in the multi–site response mode for
the activation of a metabolic pathway. Activating a metabolic pathway via an
effect on a single enzyme usually proves to be rapidly self–limiting, since even
if the enzyme has a relatively large flux control coefficient to start with, it
decreases, with control transferring to other enzymes, as the target enzyme is
activated. This was illustrated previously in connection with large changes in
enzyme amount in Fig. 5.3. At the other extreme, if an effector acted equally
on every enzyme in a pathway, the flux would continue to increase even for
large stimulations of activity without any changes in intermediate metabolite
concentrations or redistribution of control. (See the derivation of the sum-
mation theorem in the Appendix More about flux control coefficients at the
end of this chapter, p. 132, for a justification of this claim, and Chapter 8.1,
p. 263, for further consideration of its implications for metabolic control.) An
effector acting on several enzymes, each with a moderate flux control coeffi-
cient spaced along the pathway could be an adequate approximation to this
extreme case. Pure examples of this effect are difficult to give because there
are usually other control mechanisms that come into play as well. However, to
the extent that the nucleotide AMP can be considered an external effector of
glycogen catabolism in muscle, it activates phosphorylase b and phosphofruc-
tokinase. In the longer of the synthetic pathways for amino acids, there are
cases where the product amino acid (again, if it is appropriate to regard this
as an external metabolite) inhibits at more than one point along the sequence
(see Chap. 7.2.3, p. 210).
This equation for the multi–site response coefficient, Eqn. [5.23], shows
why we do not usually define response coefficients to metabolites that oc-
cur in the metabolic pathway under consideration. If P is in the pathway,
the equation contains the same terms as the connectivity relationship for P ,
128 5 Metabolic control analysis

J
Eqn. [5.18], which proves that the response coefficient in this case, RP , is zero.
The physical interpretation of this result is that if we attempt to change the
rate of the metabolic pathway by adding extra P , a metabolite in the path-
way, this extra material will flow away through the pathway, which will return
eventually to the steady state that existed before the addition, i.e. there are
no long–term consequences of adding a pathway metabolite. (The exception
to this is the case where the metabolite contains a conserved moiety that
cannot be metabolized away, but in this case, the connectivity relationship is
not equal to zero, but has a value that indicates the response to changing the
amount of the conserved material.)

5.5 Control Analysis and traditional approaches

The validity of some of the measures proposed for the identification of rate–
limiting steps (described in Chap. 4) can be examined using the concepts
developed in Metabolic Control Analysis. To do this, we will relax the inter-
pretation of a ‘rate–limiting step’ and seek to find out whether the proposed
criteria would discriminate between pathway enzymes with high and low flux
control coefficients. Two particular properties are worth special considera-
tion because they can be measured: displacement from equilibrium (Chapters
1.4.2 & 4.2) and relative limiting rates (Chapter 4.4). The only slight dif-
ficulty in carrying out such investigations is that we are no longer dealing
solely with the general properties of every metabolic system (like the summa-
tion theorem) that are true irrespective of the details of the enzyme kinetics;
now the equilibrium constants of the reactions, the kinetic equations of the
enzymes and the values of their kinetic parameters will all enter the result.
It is therefore too difficult to derive completely general algebraic solutions
to these problems. For this reason, Henrik Kacser and Jim Burns examined
a particularly simple case where it is possible to derive expressions for the
values of control coefficients in terms of factors such as displacement from
equilibrium and relative limiting rates:119 this was a linear pathway of sin-
gle substrate enzymes, all obeying the reversible Michaelis–Menten equation
(Eqn. [3.4]), and with steady state metabolite concentrations all lower than
the Km values. Under these circumstances, the expressions for the elasticities
of the enzymes become much simpler, and allow relative values of the flux
control coefficients to be calculated via the connectivity theorem.

5.5.1 Displacement from equilibrium

For their simple linear metabolic pathway, Kacser & Burns determined that
the relative values of the flux control coefficients of successive enzymes in the
chain would be:

C1J : C2J : C3J : . . . ≡ 1 − ρ1 : ρ1 (1 − ρ2 ) : ρ1 ρ2 (1 − ρ3 ) : . . . (5.24)

5.5 Control Analysis and traditional approaches 129

By considering particular sets of values of the disequilibrium ratios, it is pos-

sible to use this equation to show that:
• Relative values of the disequilibrium ratios, ρi , do not themselves show
the relative contributions of the enzymes to the control of flux, but the
terms in Eqn. [5.24] do.
• If any step i is at equilibrium (ρi = 1), then its flux control coefficient
becomes zero, because of the term (1 − ρi ). This justifies the view that
control cannot be exerted by an equilibrium reaction, though only so
far as for enzymes at equilibrium, which is not strictly possible for a
reaction carrying a net pathway flux.
• It is easy to create examples where the step nearest to equilibrium is
not the step with the smallest flux control coefficient.
• It is also possible to create examples where the step furthest from equi-
librium is not the one with the largest flux control coefficient.
• There is a tendency, in the case of this linear pathway, for the flux
control coefficients to be largest near the beginning of the pathway and
to decrease with the distance along. This is not invariant because it
is affected by the specific values of the disequilibrium ratios, but the
multiplication of an increasing number of terms, each of which is less
than 1, has this effect.
If the disequilibrium ratios cannot be relied upon to indicate the relative
influences of enzymes on the flux in such a simple pathway, there is no reason
to expect them to work better in more complex cases. Additional support
for this conclusion comes from Reinhart Heinrich & Tom Rapoport99 who
used a less restrictive derivation but also found that values of disequilibrium
ratios were not a reliable guide. Also, in the case of rat heart glycolysis, for
which the disequilibrium ratios are shown in Fig. 4.1, the relative flux control
coefficients of several of the enzymes have been determined,125 but in this real
example the disequilibrium ratios alone do not rank the enzymes correctly.
The equation given above, [5.24], does rank the enzymes in the right order,
though the numerical match is not very good.

5.5.2 Maximal enzyme activities

Kacser & Burns’ equation for the relative values of the flux control coefficients
of a linear pathway was:119
Km,1 Km,2 Km,3
C1J : C2J : C3J : . . . ≡ : : ... (5.25)
V1 V2 Keq,1 V3 Keq,1 Keq,2

Thus there is a tendency for the enzymes with the largest limiting rate to have
the smallest flux control coefficients and vice versa but, again, this cannot be
relied on to rank the enzymes according to their effects on the flux because
130 5 Metabolic control analysis

Km values and equilibrium constants enter the expressions. The equation has
not been applied in any real cases because these are always more complicated
(involving two substrate enzymes, for example), so it has no practical value
beyond demonstrating the unsuitability of this traditional criterion.

5.6 Summary
1. The qualitative categories of ‘rate–limiting’ and ‘not rate–limiting’are
replaced in Metabolic Control Analysis by a quantitative scale for the
influence of an enzyme on a metabolic flux: the flux control coefficient.

2. The flux control coefficient of an enzyme is a system property of a

metabolic pathway since its value can be affected by any or all of the
other enzymes. This is particularly apparent in the summation theo-
rem, a system constraint that requires the control coefficients of all the
enzymes on a particular flux to add to 1.

3. The influences of metabolites on enzymes are measured in terms of the

elasticity coefficients. These are related to the kinetic properties of the
enzymes, but are defined specifically for the conditions in the metabolic
pathway at steady state.

4. The connectivity theorem shows that the flux control coefficients of the
enzymes in a pathway have links to the elasticities, i.e. the system con-
trol properties can be related to the individual kinetic characteristics of
the enzymes.

5. The action of external effectors on a pathway flux can be measured as

a response coefficient, which can be shown to be the product of the
flux control coefficient of the affected enzyme and the elasticity of that
enzyme with respect to the effector. Both factors must be non–zero for
an effector to be able to influence a flux.

6. Metabolic Control Analysis shows that neither the degree of displace-

ment of a reaction from equilibrium nor the relative value of the limiting
rate of an enzyme are reliable guides to the degree of control an enzyme
can exert on a flux, even though these factors have been used for that
purpose in the past.

5.7 Appendix 1: More about flux control coeffi-

cients
The definitions given earlier for the flux control coefficient, Eqns. [5.2] &
[5.3], correspond to those originally used by Kacser & Burns (though at the
time they called them sensitivities), and have the advantage that they di-
rectly reflect the effects of changes in enzyme amounts caused by genetic or
5.7 Appendix 1: More about flux control coefficients 131

environmental factors. However, there are occasions where subtly modified

definitions are preferable, for example:

1. It is sometimes necessary to assign a metabolic flux a negative value, for

example if it is a reversible pathway flowing in the opposite direction to
that originally defined. In this case, Eqn. [5.3] cannot be used as it is
because it is not possible to have the logarithm of a negative number.
(Negative logarithms represent numbers between 0 and 1.) This turns
out not to matter because the following modified version works just as
well:
Jydh ∂ ln | Jydh |
Cxase =
∂ ln Exase
where | Jydh | signifies that the absolute value of Jydh is used, i.e. any
negative sign is ignored.

2. The logarithms in Fig. 5.1c) and Eqn. [5.3] are the natural logarithms,
but logarithms to base 10 can be used without any alteration to the
equations.

3. The flux control coefficient has been shown here as defined relative to
the concentration or amount of an enzyme, as mentioned above. How-
ever there is the potential for complications in cases where the activity
of the enzyme does not relate directly to its concentration. Such con-
ditions are not thought to be particularly common, but the difficulties
can be avoided by a variation of the definition that was proposed by
Reinhart Heinrich and his colleagues, in which the coefficient is defined
with respect to some parameter of the enzyme that acts on the enzyme’s
activity. Suppose this parameter is k; then we define the flux control
coefficient as the ratio of the response of the flux to k relative to the
effect of k on the enzyme activity, i.e. :
Á
Jydh k∂Jydh k∂vxase
Cxase = (5.26)
Jydh ∂k vxase ∂k

This equation is identical to the response coefficient relationship that

was described in Eqn. [5.22] and used in determining flux control coef-
ficients from inhibitor titrations (Eqns. [6.2,6.3]):
J
J Rkydh
Ck ydh = (5.27)
εxase
k

If the rate of enzyme xase depends proportionally on the parameter k,

then εxase
k will be 1 and the flux control coefficient will be equal to the
response coefficient to k. Generally, enzyme activities are expected to
be proportional to enzyme concentration, so a possible choice for k is
Exase and the flux control coefficient will be equal to the response of the
flux to the enzyme concentration. This has been the implicit assumption
132 5 Metabolic control analysis

in this book and the justification for the simple definition used in this
Chapter. There are possible exceptions, however, such as when the
enzyme concerned forms enzyme–enzyme complexes; then εxase k need
not necessarily equal 1 and the intrinsic flux control coefficient of the
step catalyzed by xase is not equal to the response coefficient of the flux
to the concentration of xase. In such cases, the definition in Eqn. [5.26]
above is preferable.
The summation theorem for flux control coefficients has been proved in
a number of different ways. The simplest to understand is probably the ar-
gument originally used by Kacser & Burns. Again, we have reverted to the
simple definition of the flux control coefficient relative to the amount of en-
zyme, Eqn. [5.2], on the assumption that for any enzyme xase, εxase Exase = 1.
Suppose that the amount of one of the enzymes in the pathway, say the first,
E1 , has its amount increased by a small fraction α = δE1 /E1 . Then according
to Eqn. [5.1], the change in flux is given by:
δJ J δE1 J
= CE = CE α
J 1
E1 1

However, if all the enzyme amounts could be simultaneously increased by

this same amount α, then all the rates would increase by this fraction α
because the rates are proportional to enzyme amount. If all the rates change
simultaneously by the same amount, the only change in the pathway will be
that the fluxes have increased by the fraction α. There will be no changes
in any metabolite concentrations because, for every metabolite, the rate of
synthesis has increased by the same amount as the rate of degradation. Now
the total fractional change in flux (α) is given by the sum of all the individual
changes as predicted by the equation above for each of the n enzymes i.e. :
J J J
CE 1
α + CE 2
α + . . . + CE n
α=α

Dividing through by α gives:

J J J
CE 1
+ CE 2
+ . . . + CE n
=1

which is the summation theorem, Eqn. [5.10].

5.8 Appendix 2: Concentration control coefficients

As well as flux control coefficients, there are other control coefficients de-
fined in the same way, such as the concentration control coefficients where
the variable affected by the chosen parameter, such as the enzyme xase, is a
metabolite concentration, say S:

S ∂S Exase ∂ ln S
Cxase = . = (5.28)
∂Exase S ∂ ln Exase
Being defined in the same way, they can be interpreted in the same way.
5.8 Appendix 2: Concentration control coefficients 133

5.8.1 Theorems for concentration control coefficients.

A set of theorems exist for the concentration control coefficients corresponding
to those for the flux control coefficients. Thus there is a summation theorem: 99
n
X S
Ci j = 0 (5.29)
i=1

where Sj represents any one of the variable metabolites of the pathway. This
reflects the result mentioned in the previous Appendix on the flux control
coefficients: simultaneously changing all the enzyme activities by the same
small fractional amount α has no effect on any metabolite concentration.
The connectivity relationship for concentration control coefficients is more
complex than for flux control coefficients272 in that it has one form when the
metabolite whose concentration is the subject of the control coefficients (say
A) is different from the one in the elasticities (say B):
n
X
CiA εiB = 0 (5.30)
i=1

but the following form when they are the same:

n
X
CiA εiA = −1 (5.31)
i=1

As with the flux control coefficients, the form of the equation changes when
metabolites in conserved cycles are involved.213
Generally, less attention has been given to concentration control coeffi-
cients. However, in Reinhart Heinrich & Tom Rapoport’s original work99 on
control analysis, they were called elements of the control matrix and used to
derive expressions for the flux control coefficients by means of the relation-
ships:
m
X S
CiJi = 1+ εiSj Ci j
j=1
m
X S
CiJk = εkSj Ci j (5.32)
j=1

These equations actually show how the systemic response of the flux to a
modulation of an enzyme can be broken into its components. Thus, the
first one shows that the response of the flux through step i to modulation of
enzyme i is composed of a proportional change from the change in the amount
of enzyme (the ‘1’), on which are superimposed the changes in activity of the
enzyme because of the changes in each of the metabolites Sj , with each of
these effects calculated from the concentration control coefficient of enzyme i
on substrate Sj (to show how much the steady state concentration changes)
134 5 Metabolic control analysis

and the elasticity coefficient for the effect a change in substrate Sj has on the
activity of enzyme i. In the second equation, because the effect of a change
in enzyme i on the flux at k is sought, there is no term for a direct effect of
the change in the amount of enzyme.

5.9 Appendix 3: More about elasticities

5.9.1 Elasticities and enzyme kinetics
In the main text, it was mentioned that it was possible to calculate elasticities
from enzyme kinetics equations. This raises the question of why it is thought
necessary to define elasticities at all if the information is available from the
results of traditional enzyme kinetics? In fact, a number of points of contrast
exist between the purposes for which enzyme kinetic information is usually
collected and the requirements for elasticities:

• The equation used in enzyme kinetics is related to the presumed mecha-

nism of action of the enzyme, and it is important that it fits the observed
rates across the range of metabolite concentrations used. For the pur-
poses of control analysis, it is not necessary that the function used for
the kinetics is soundly based mechanistically; it only has to describe the
enzyme’s responses to metabolites near their physiological concentra-
tions.

• Because the enzyme kinetic equations provide information about mech-

anism, the experiments simplify the problem by using the minimum
number of metabolites, for example, the substrates but not the prod-
ucts. The values of the elasticities must be obtained in the presence of
all cellular metabolites that affect the enzyme, but such measurements
are not usually taken in enzyme kinetics. The effects of products on en-
zymes, expressed as the product elasticities, are particularly important
in Metabolic Control Analysis, even when they correspond to apparently
weak product inhibition.

• If an enzyme responds to more than a very few metabolites, it becomes

virtually impossible to carry out an experiment in which they are all
varied and the results fitted to a single equation that, because it con-
tains parameters that account for how each metabolite affects the en-
zymes’s response to all of the others, can describe the rate under any
circumstances. Such detailed information is not required for elasticities
because they describe the enzyme’s response to each metabolite when
all the others are held constant.

• Enzyme kinetics experiments are performed under conditions that are

convenient for obtaining the required results. Analysis of the enzyme’s
mechanism does not necessarily require that the measurements are taken
at physiological conditions of pH, temperature and ionic composition.
5.9 Appendix 3: More about elasticities 135

On the other hand, the values of the elasticities must be valid for the
conditions in the metabolic system under study.
For these reasons, although enzyme kinetics could in principle supply the
information required as elasticities, the published information is frequently
not appropriate.

5.9.2 Algebraic evaluation of elasticities

Where a suitable rate function for an enzyme has been determined, then its
elasticities can be derived by analytical partial differentiation with respect to
each of the metabolites in accordance with Eqn. [5.12]. Although differen-
tiation of these functions is often tedious by hand, the wide availability of
symbolic algebra programs for computers has removed the need for a high
order of mathematical skill. For example, for the Michaelis–Menten equation
analysed graphically in the main text we have:
µ ¶
SV
∂
S S + Km
εvS = (5.33)
v ∂S
The result is surprisingly simple:

Km
εvS = (5.34)
Km + S

The final stage in the calculation of the required elasticity value is to insert
numerical values for the parameters in the equation, in this case only Km =
(0.75 mM), and the value for the concentration of the metabolite (0.5 mM),
to give 0.60 as obtained previously from the graph shown in Fig. 5.5.
The reason the results for most elasticities turn out to be simpler than
might be expected from the complexity of the functions to be differentiated is
because of the cancellations between numerator and denominator terms with
the scaling factor. It is possible to define an ‘elasticity calculus’ that simplifies
the problem of determining elasticities by avoiding generating terms that will
subsequently cancel. Most enzyme rate functions, F , have the form F = N/D,
where both N and D are functions of the metabolite concentration, S, for
which we want the elasticity.
S ∂F
εF
S =
F ∂S
SD ∂( N
D)
=
N µ∂S ¶
SD ∂N/∂S N ∂D/∂S
= −
N D D2
µ ¶
∂N/∂S ∂D/∂S
= S −
N D
136 5 Metabolic control analysis

Even when using computer algebra packages to perform the differentiation,

it is worth using the above result because many of them do not otherwise
manage to simplify the results fully.

Further reading
1. Kacser, H. and Burns, J. A. The control of flux. Symp. Soc. Exp. Biol.
27, 65–104 (1973); reprinted with modern notation and terminology as
Kacser, H., Burns, J. A. & Fell, D. A. The control of flux. Biochem.
Soc. Trans. 23, 341–366 (1995)

2. Kacser, H. and Porteous, J. W. Control of metabolism: what do we have

to measure. Trends Biochem. Sci.12, 5–14 (1987)

3. Kell, D. and Westerhoff, H. Metabolic Control Theory: its role in micro-

biology and biotechnology. FEMS Microbiol. Rev. 39, 305–320 (1986)

4. Kacser, H. Control of metabolism in The biochemistry of plants, Davies,

D. D. (ed.), Vol. 11, pp. 39–67, Academic Press, New York (1987)

5. Fell, D. A. Metabolic Control Analysis: a survey of theoretical and ex-

perimental developments. Biochem.J 286, 313–330 (1992)

Problems
1. Suppose an enzyme in a pathway follows Michaelis-Menten kinetics with
V = 100 units and Km = 0.05 mM:
SV
v=
S + Km

What is the elasticity of the enzyme with respect to its substrate a) at

a substrate concentration of 0.03 mM; b) at a substrate concentration
of 0.25 mM? (Hint: try graphical estimation, as in Fig. 5.5.)

2. The effect of a competitive inhibitor, I, on a Michaelis-Menten enzyme

can be described by the equation:

SV
v=
S + Km (1 + I/Ki )

Suppose V = 1, Km = 1 and Ki = 1 (this means the values of v, S and

I have been ’scaled’ so we can use dimensionless values for them). What
is the elasticity of the enzyme with respect to the inhibitor at I = 1 for
S = 0.5 and S = 5? (HINT: the solution is like that of the previous
question, except that for a graphical solution, it is now ln v against ln I
that is plotted.)
5.9 Appendix 3: More about elasticities 137

3. The enzyme fumarase catalyzes the reaction:

fumarate *
) malate

Its rate of reaction is described by the reversible Michaelis–Menten equa-

tion: µ ¶
mal
V f um −
Keq
v=
K mal
Kfum + f um + fum
Kmal
where V = 20 µmol.min−1 , Kfum = 0.9 mM, Kmal = 1.2 mM and Keq
= 11. What are the elasticities of the enzyme with respect to fumarate
and malate at fum = 0.4 mM and mal = 0.5 mM?

4. Consider the glycolytic pathway, particularly the successive enzymes

phosphofructokinase and aldolase:

PFK Ald
· · · Fru-6-P −→ Fru-1,6bisP −→ DHAP + GAP · · ·

The elasticity of phosphofructokinase (PFK) with respect to fru-1,6-bisP,

PFK ald
εF BP , is -0.01, whilst that of aldolase to the same metabolite ε F BP , is
2.5 in a particular cell. What is the ratio of the flux control coefficients
of these two enzymes on glycolysis? What is the flux control coefficient
PFK
of aldolase if εF BP is 0?
138 5 Metabolic control analysis
 6

Measuring control
coefficients

The previous chapter presented the main outlines of the theory behind Metabolic
Control Analysis. However, up to this point I have only shown the potential
of the theory: it has defined possible quantitative measures of the effects of
changes in enzyme amounts or of pathway effectors on the rates of metabolic
pathways, and there is the expectation that such measurements will show that
control is distributed through the pathway (although it is possible for it all
to be located on one enzyme). Now comes the point where it is necessary to
show that:

• Metabolic Control Analysis is not an abstract theory but can be applied

to measurements on metabolism, and further,

• when these measurements are made, they support the claims made ear-
lier.

There is no doubt that the measurements required for experiments in Metabolic

Control Analysis can be challenging, even though they involve many of the
same experimental procedures that were described in Chapter 2 and used in
earlier studies of metabolic regulation, as discussed in Chapter 4. The body
of experimental evidence only started to grow in the 1980s, ten years after
the theory was initially developed. This growth continues to accelerate as the
value of the approach becomes more widely appreciated. The experiments
that have been carried out involve a variety of pathways occurring in a range
of different organisms. This means that there is little coherence in an ac-
count organized chronologically, or by pathway, and I shall therefore group
the experiments according to the experimental approaches. The disadvantage
of this is that some of the scientific groups involved have quite rightly used
several different approaches to the same problem, either to obtain some com-
plementary information, or to validate their experiments by making the same

139
140 6 Measuring control coefficients

measurement in different ways. My method of presenting the experiments

will in these cases separate the parts of what was originally a coherent set of
experiments.
I will concentrate on experimental methods to measure flux control coef-
ficients, and these I will divide initially into direct methods, which involve
observation of the effects of making an alteration of an enzyme activity, and
indirect methods, which involve calculation of control coefficients from other
information, in particular elasticities. Obviously, looking at the indirect meth-
ods raises the question of how the elasticities can be measured.

6.1 Manipulation of enzyme activity

Since a control coefficient expresses the effect that a change in the amount
of an enzyme has on a system property such as metabolic flux or metabolite
concentration, the only direct method of determination of its value is to make
a change in the enzyme activity and observe the consequences whilst all other
conditions are kept constant. There are various ways to do this, but they
all share the problem that control coefficients are defined as the response to
an infinitesimally small perturbation, whereas the finite precision of any ex-
periment requires that the response is determined from changes large enough
to produce a measurable effect. One possible solution is to make a series of
graded changes and extrapolate the results to an infinitely small change; both
increases and decreases of enzyme activity should ideally be used to avoid
bias.
Some of the different methods of altering enzyme activity that have been
used are:
• Alteration of expressed enzyme activity by genetic means;
• Alteration of expressed activity by inducers or dietary and environmen-
tal means;
• Titration with purified enzyme, and
• Titration of enzymes by specific inhibitors.

6.1.1 Altering enzyme activity by genetic means.

In the previous chapter, when discussing the implications of dominance for
Metabolic Control Analysis (Chap. 5.2.2, p. 108), I mentioned that in diploid
organisms, heterozygotes carrying a mutant allele of a gene for an enzyme will
contain less enzymic activity, perhaps only 50% of the wild–type level if the
mutant allele codes an inactive enzyme. Phenotypically, these heterozygotes
usually appear normal, but small differences in flux may be discernable on
measurement. Nevertheless, a 50% reduction in the amount of the enzyme
hardly meets our need for small increases and decreases about the wild–type
level, so the results of such experiments are only likely to give an approximate
6.1 Manipulation of enzyme activity 141

value for the flux control coefficient. Fortunately there are cases where finer
genetic control of enzymic activity can be obtained.

6.1.1.1 Classical genetics: gene dosage

An example of this approach is given by the experiments on arginine synthesis
carried out by Henrik Kacser’s group in Edinburgh during the 1970s.74 In this
case, the design of the experiment was made more favourable by their choice,
instead of a diploid organism, of the fungus Neurospora crassa, which forms
mycelia lacking cross walls between the cells so that the cells effectively have
multiple nuclei surrounded by a common cytoplasm which is mixed by the
continuous cytoplasmic streaming. This fungus was the one used in the 1940s
by Beadle, G. W. & Tatum when they formulated their one gene, one enzyme
dictum on the basis of their genetic and biochemical analyses of mutants for
the synthesis of various amino acids; as a result of this and similar work, many
Neurospora mutants are available with known defects in particular enzymes.
By mixing spores of wild type and mutant strains, mycelia can be formed con-
taining wild–type and mutant nuclei in the same multi–nuclear cells (called
heterokaryons in this case because the nuclei are of different types). Further-
more, a range of ratios of wild–type and mutant nuclei can be obtained by
varying the ratios of spores in the mixture, whereas heterozygotes in diploid
organisms only allow a 1:1 ratio. The pathway from the amino acid glutamate
to arginine is shown in Fig. 6.1. Null mutants, that is mutants not forming any
active enzyme, were available for four of the enzymes. The results obtained
by Kacser’s group for the effect of two of the enzymes from this pathway on
the flux through the last enzyme to arginine are shown in Fig. 6.2. The hy-
perbolic curve I have fitted through the results for the heterokaryons of the
mutants for step 2, ornithine carbamoyltransferase, gives a flux control coeffi-
cient of 0.02 at the wild–type level of enzyme, increasing to 0.31 in the mutant

1 2 3 4
glutamate - ornithine - citrulline - arg–succ - arginine

Fig. 6.1: The arginine synthesis pathway of Neurospora crassa.

The four steps of the pathway studied in the experiments by Kacser’s group were: 1, the
four enzyme ornithine synthesis cycle, in which a mutant of the enzyme acetylornithine
aminotransferase was used; 2, ornithine carbamoyltransferase; 3, argininosuccinate syn-
thase, and 4, argininosuccinate lyase. Arg–succ has been used as an abbreviation for
argininosuccinate. The details of ornithine synthesis and the co-substrates of the path-
way have not been shown.
142 6 Measuring control coefficients

100
100
a
b
80
80
Flux, %

Flux, %
60 60

40 40

20 20

0 0
0 20 40 60 80 100 0 20 40 60 80 100

Ornithine carbamoyltransferase, % Argininosuccinate lyase, %

Fig. 6.2: The dependence of arginine synthesis flux in Neurospora on enzyme

levels.
The results are those of Flint et al.74 with my best–fit hyperbolic curves. Most of the
points were obtained by forming heterokaryons with different ratios of wild–type and
mutant nuclei. In each graph, however, the point at the lowest enzyme activity is not a
heterokaryon, but a partial revertant from the mutant. a) The dependence of the flux
to arginine through argininosuccinate lyase on the activity of ornithine carbamoyltrans-
ferase, both expressed as a % of wild–type levels. b) The dependence of the same flux
on the activity of argininosuccinate lyase itself.

with the least activity (4%) of this enzyme. Similarly, the hyperbolic curve
I have drawn through the results for the last step, argininosuccinate lyase,
corresponds to a flux control coefficient of 0.07 in the wild–type, increasing to
0.42 in the heterokaryon with 10% of the wild–type activity. Of the other two
enzymes examined, in step 1 acetyl–ornithine aminotransferase also appeared
to have a low flux control coefficient, around 0.06 in the wild–type organism,
whereas that for argininosuccinate synthase might have been as high as 0.2,
but was not very accurately determined. (Later experiments implied that
this enzyme’s flux control coefficient could not be so large.) Thus little of the
overall control of arginine synthesis can be accounted for by these enzymes
(given that the summation theorem implies that the remaining control to be
found is 1 minus the sum of the four flux control coefficients in the pathway).
However, the main importance of these experiments was the demonstration
that, just as Kacser & Burns had predicted, the influence of an enzyme on a
flux was not all or nothing, but variable depending on the activity level of the
enzyme.

6.1.1.2 Classical genetics: allozymes and heterozygotes

In diploid organisms, more subtle changes in the amount of active enzyme can
be generated where there are allelic forms of the enzyme, or allozymes with
6.1 Manipulation of enzyme activity 143

differing activity levels. The various homozygotes and heterozygotes that can
be formed can give a number of different levels of enzymic activity, though
interpretation of the results can become complicated if the allozymes differ
in kinetic properties such as Km as well as their limiting rates. Middleton
& Kacser determined the effect of varying activities of alcohol dehydrogenase
(EC 1.1.1.1) in this way on the catabolism of ethanol in the fruit fly, Drosophila
melanogaster .159 They used three naturally-occurring alleles for the enzyme:
S; F, and Fd . (The letters refer to the rate of migration in electrophoresis:
S for slow and F for fast. Fd is enzymically similar to F but is expressed in
larger amounts.) They bred four homozygous varieties: SS, FF, Fd Fd and
NN, where NN is the null mutant. In addition, they obtained the heterozy-
gotes SN, FN and FS. The measurement of relative enzymic activities was
complicated because the allozymes differ in both the expressed limiting rate,
V , and the Km for ethanol. At low, fixed substrate concentration relative to
the Km values, the activity of a two substrate enzyme such as alcohol dehy-
drogenase will be proportional to V /(Kethanol Ki,NAD ), as can be derived from
Eqn. 3.5, p. 59. This expression can therefore be used to compare the relative
activities of the allozymes whilst taking into account the differences in the
kinetic parameters. Accordingly, Middleton & Kacser made kinetic measure-
ments on extracts of the adult fruit flies and calculated V /(Kethanol KNAD )

100

80
Flux, dpm·h-1

60

40

20

0
0 0.2 0.4 0.6 0.8 1

V/Km , A·min-1 ·mg-1 ·mM-1

Fig. 6.3: The effect of alcohol dehydrogenase activity on the flux of ethanol
catabolism in Drosophila.
The genotypes giving the points are, from the left: NN; SN; FN; SS; FF, and F d Fd .
The points on the graph have been recalculated from the original results reported by
Middleton & Kacser,159 and a rectangular hyperbola fitted. The result for FS is not
shown as there is insufficient information to calculate its position on the ordinate.
144 6 Measuring control coefficients

100

Relative growth rate, %

80

60

40

20

0
0.001 0.01 0.1 1 10 100

β–galactosidase, %

Fig. 6.4: The effect of β–galactosidase on E. coli growth rate on lactose.

The results are those obtained by Dykhuizen et al. for constitutively–expressed wild–type
and mutant forms of the enzyme.61 The enzymic activity, relative to a value of 100 for
the wild–type in a constitutive mutant, has been shown on a logarithic scale because of
the wide range covered; the fitted hyperbolic curve appears sigmoidal because of this.

for the different genotypes. (Strictly, KNAD is not the correct parameter, but
there was no significant difference in its value between the strains.) However,
in order to show their results in Fig. 6.3, I have recalculated their kinetic
parameters and calculated the term V /Kethanol as the variable reflecting the
relative activity. This does not change the basic appearance of the graph,
nor their conclusion that the enzyme’s flux control coefficient is very close to
zero. This is consistent with their results on the flies’ relative abilities to tol-
erate ethanol (which they would naturally encounter in rotting fruit). All the
enzyme–containing strains had equal tolerance to the lethal effects of ethanol,
i.e. they exhibited an indistinguishable phenotype; only the homozygous re-
cessive, NN, was intolerant. These results indicate that it is unlikely that the
polymorphism at the alcohol dehydrogenase locus is maintained through se-
lective pressure related to alcohol metabolism, at least in the adult fly. They
are also a good demonstration of why mutations in alcohol dehydrogenase are
recessive: although the heterozygotes do contain less enzymic activity, they
have an essentially unchanged flux through the pathway.
The same technique is feasible in a haploid organism if there is a sufficient
number of alleles of differing enzymic activity. Dykhuizen et al.61 studied the
effect of varying activity of the β–galactosidase (EC 3.2.1.23) of Escherichia
coli on its rate of catabolism of lactose, which they assessed by the relative
growth rates of competing strains in a chemostat, where the steady–state
6.1 Manipulation of enzyme activity 145

lactose concentration will be low enough to be restricting growth. The β–

galactosidase is one of the products of the lac operon, which is induced by
the presence of lactose in the medium. One of the other products is the lac-
tose permease, which is the first step in the catabolic pathway because it
catalyzes entry of the lactose into the cell, where the galactosidase hydroly-
ses it into galactose and glucose that can then be catabolized. One way to
vary the activity of the β–galactosidase is to vary the degree of induction
of the lactose operon, but since the permease is transcribed from the same
operon, its activity then varies in parallel with that of the galactosidase. To
separate the effects of the two enzymes, Dykhuizen et al. used constitutive
mutants of E. coli, i.e. mutants that produce the permease and galactosi-
dase at their maximal rates regardless of the presence or absence of lactose.
To get a range of strains with different β–galactosidase activities, they used
revertants from a missense mutant for the galactosidase gene (that is, deriva-
tives of the mutant strain that had regained the ability to grow on lactose).
These revertants contained β–galactosidases with altered kinetic properties as
the result of amino acid substitutions relative to the wild–type. In addition,
for the very low activity end of the range, they used strains that contained
‘evolved’ galactosidase activity, that is, strains that had reverted by mutation
in another protein of unknown function that gave it some galactosidase ac-
tivity. Because the enzymes differed in both limiting rate and Km , relative
enzymic activity was expressed as V /Km . The results are shown in Fig. 6.4,
with the hyperbolic function fitted by the authors through the results. At the
maximally–induced or constitutive wild–type level of enzyme (100% on the
graph), the flux control coefficient of the enzyme on growth rate is effectively
zero, less than 0.005. The control coefficient of the permease will be referred
to later (Section 6.1.2).

Table 6.1: Additional examples of the use of heterozygotes in Metabolic Con-

trol Analysis.

Enzyme Organism % levels Flux to: Ref.

organelle

glucose–6–phosphate Clarkia xantiana 18, 36, 64 sucrose, starch 143

isomerase cytosol
glucose–6–phosphate Clarkia xantiana 50, 75 sucrose, starch 143
isomerase chloroplast
phosphoglucomutase Arabidopsis thaliana 0, 50 starch 167
chloroplast
glucose–1–phosphate Arabidopsis thaliana, 7, 50 starch 167
adenylyltransferase chloroplast

These measurements of flux control coefficients in plant photosynthesis were carried

out by Mark Stitt and his colleagues.
146 6 Measuring control coefficients

Mark Stitt and his colleagues, originally at the University of Bayreuth, but
now at Heidelberg, have been using a variety of approaches to estimate the
flux control coefficients for enzymes involved in the photosynthetic production
of sucrose and starch. Their studies using heterozygotes for various enzymes
are summarised in Table 6.1. Because the changes in enzyme activities in
these experiments were large, the flux control coefficients were calculated from
a difference equation that the authors had derived on the assumption of a
hyperbolic relationship between flux and enzyme content:
µ ¶
E2 J2
1−
J1 E1 J1
CE = µ ¶ (6.1)
1
J2 E2
1−
J1 E1

where the subscript 1 signifies the flux (J) and enzyme (E) levels at the first
point, and 2 those at the second. (This equation is equivalent to the ‘large
change’ equation derived later by Small & Kacser, Eqn. [5.8] and mentioned
in the previous chapter.)

6.1.1.3 Molecular genetics: gene dosage

The development of genetic engineering techniques in modern molecular bi-
ology has led to an increased number of methods for changing the amount
of a target enzyme that is expressed in a cell. For a while, the pioneering
experiments in this field were sufficiently difficult that achieving a change
in expressed enzyme activity was virtually an end in itself. Now that the
methodologies are established and becoming easier, molecular biologists are
discovering, by experiment, that an ability to manipulate enzyme activities
does not guarantee an ability to manipulate rates of metabolism: the need
to understand the factors controlling the rate of a metabolic pathway can-
not be side–stepped. However, the control of enzyme activities offered by
these genetic techniques brings new opportunities for the Metabolic Control
Analyst.
For example, an extra copy of a gene can be inserted in a plasmid that
is introduced into cells to increase the amount of enzyme expressed. This
still has some problems for Control Analysis purposes. In brewer’s yeast,
Saccharomyces cerevisiae, for instance, the plasmids that are commonly used
are present in multiple copies, but because of the way that yeast reproduces
by budding, the number of copies received by the daughter cells is variable; as
a result, the change in the amount of the enzyme carried on the plasmid can
be both large and variable from cell to cell. In fact, unless the plasmid confers
a specific advantage on the cells that contain it (such as conferring resistance
to an antibiotic or complementing a nutritional mutation in the parent cell),
it tends to be lost in culture because the cells that lose it grow faster. There
is also the danger of pleiotropic effects from the expression of large amounts
of one enzyme; that is, there can be phenotypic effects other than a simple
6.1 Manipulation of enzyme activity 147

change in the amount of the target enzyme. This can be simply because
diverting resources into synthesizing one enzyme in unusually large amounts
competes with the synthesis of other cellular components. Though this is not
likely to be aproblem for the small changes in enzyme amounts needed for
Control Analysis, the effect has been demonstrated in extreme cases. More
specifically though, the mechanisms that normally control the expression of
enzymes in a metabolic pathway may be activated by the increased amount
of one of the enzymes and cause a reduction in the synthesis of the others,
so that any observed change in flux cannot be ascribed solely to the enzyme
that has been added. Even if these problems are overcome, the experiments
can still have the same difficulty as classical genetic approaches: changes in
the gene dosage produce larger changes in enzyme amounts than we would
normally like for determining control coefficients.
In 1986, Heinisch96 overexpressed the allosteric glycolytic enzyme 6–phos-
phofructo–1–kinase in yeast cells about 3.5–fold, but observed no effect on
glycolytic flux to ethanol. This has come as rather a surprise to those who
believed biochemistry textbooks that state that phosphofructokinase is the
rate–limiting step of glycolysis on the basis of evidence such as that cited in
Chapter 4. Similar experiments with 8 of the 11 other glycolytic enzymes
have subsequently failed to show any changes in glycolytic rates either, even
though hexokinase activity, for example, was increased 13.9 fold.220 However,
Davies & Brindle59 showed in a similar experiment that although a 5–fold
excess of phosphofructokinase had no effect on anaerobic glycolysis, it did
stimulate anaerobic ethanol production in aerobic conditions. That is, the
glycolytic rate increased slightly, but the cells nullified the effects by reducing
the amount of pyruvate oxidized by the efficient aerobic route and getting rid
of the excess pyruvate as ethanol. Even so, if the flux control coefficient of
phosphofructokinase on glycolysis is calculated from their results using the
‘large enzyme change’ equation, Eqn. [5.8], it is found to be only about 0.3.

6.1.1.4 Molecular genetics: modulation of gene expression

In 1985, Walsh & Koshland262, 263 devised a method for getting finer control
over the amount of enzyme expressed by genes carried on a plasmid. Though
they had no intention of applying Metabolic Control Analysis to their results,
they were studying the effects of changing the activity of citrate synthase (EC
4.1.3.7), the first enzyme of both the tricarboxylic acid and glyoxalate cycles,
on the metabolism of Escherichia coli growing on acetate. They placed the
gene for citrate synthase on a plasmid, but they replaced the natural promoter
with the synthetic tac promoter (a hybrid of the promoters of the tryptophan,
trp, and lactose, lac, operons). The plasmid also carried the lac repressor so
that the expression of the enzyme could be altered by varying the amounts of
inducers of the lac operon to which the bacteria were exposed. As so often in
experiments with the lac operon, the actual inducer used was not a metabo-
lizable compound, such as lactose, which would interfere with the metabolism
under study, but the non–metabolizable gratuitous inducer isopropyl–thio–β–
148 6 Measuring control coefficients

D–galactoside (IPTG). (The reason for using the tac promoter rather than
just the lac promoter is that higher levels of enzyme synthesis can be driven
by the artificial version.)
Their methodology was then adopted for Metabolic Control Analysis by
Ruijter, Postma and van Dam203 in the Netherlands in 1991. They were
studying the first step in the metabolism of glucose by E. coli, which is its
simultaneous transport across the membrane and phosphorylation to glucose–
6–phosphate. As in a number of bacterial transport systems, the energy source
and phosphate donor driving this linked transport and phosphorylation is the
glycolytic intermediate phosphoenol pyruvate, which gets converted to pyru-
vate in the process. This phosphoenol pyruvate:carbohydrate phosphotrans-
ferase system (PTS) consists of several components: two cytoplasmic proteins
that are involved in accepting the phosphate from phosphoenol pyruvate and
that are shared between the different transfer systems, and the carbohydrate–
specific permeases that are located in the membrane and have one or two sub-
units. The glucose–specific PTS has two subunits, and the degree of control
of transport on glucose metabolism was investigated by measuring the flux
control coefficient of one of these, the membrane–bound permease with the
elegant name enzyme IIGlc (EC 2.7.1.69). Ruijter et al. constructed a plas-
mid containing the structural gene for enzyme IIGlc under the control of a tac
promoter and inserted it in an E. coli strain that lacked any chromosomally-
coded glucose transport systems; in this way, only cells containing the plasmid
could grow on glucose, and when the plasmid gene was expressed at low levels,
the content of enzyme IIGlc was below wild type levels. On the other hand,
by varying the concentration of the inducer IPTG, they could vary the level
of IIGlc between 20 and 600% of its wild–type level with little effect on the
expression of other proteins of the system. Slight variations in IIGlc activity
near wild–type levels in the presence of excess glucose had little effect on the
rates of glucose oxidation and growth, so the control coefficients were very
low. At the lowest enzyme levels studied, both rates showed dependence on
the enzyme content in the same sort of quasi–hyperbolic manner we have seen
earlier in this chapter. Of course, one possible explanation of the low flux con-
trol coefficients at wild–type levels could have been that although transport
does have some influence on glucose metabolism and growth, this control is
exerted by one of the other PTS components. This was excluded by studying
the flux in the transport step alone by measuring the rate of accumulation and
phosphorylation of the non–metabolizable sugar methyl–α–glucoside; enzyme
IIGlc had flux control coefficients of about 0.6 on these fluxes at wild–type lev-
els and therefore has a larger effect than the other three components, whose
flux control coefficients together cannot exceed 0.4 by the summation theo-
rem. Of course, it should be mentioned that the measurements of the effect of
the glucose permease on growth and glucose oxidation were made in just the
conditions, the presence of excess glucose, that would be likely to minimize
its influence on the fluxes. Lower glucose concentrations, as in chemostats at
steady state, would almost certainly increase the flux control coefficients of
the transport system.
6.1 Manipulation of enzyme activity 149

A related approach that avoids the disadvantages of expressing genes from

plasmids was reported in 1993 by Jensen, Westerhoff & Michelson.113–115
They replaced the natural promoter of a chromosomal gene in E. coli with
IPTG–inducible lac–type promoters (including tac). The enzyme they were
studying was the energy–transducing proton–translocating ATP synthase (EC
3.6.1.34), which is located in the plasma membrane of the bacterium and
clearly has a critical role in the transduction of energy obtained from aerobic
catabolism. (See the supplementary material on oxidative phosphorylation
later in this Chapter, p. 161.) They determined the flux control coefficients of
this enzyme by placing the atp operon, which codes for its subunits, under the
control of the lac–type promoters and varying its expression from 0.15 to 4.5
times the wild–type level by altering the level of IPTG. To do this, they had
to use a strain of E. coli that lacked the lactose permease, which catalyzes the
transport of IPTG into the cell, since otherwise the extent of expression of the
operon was too sensitively dependent on the inducer concentration. When the
bacteria were growing aerobically on succinate, the flux control coefficient of
the ATP synthase at wild–type levels on growth rate was virtually zero, and
on succinate consumption and respiration rate was actually slightly negative
at -0.25. The results with glucose as the substrate were very similar at the
same wild–type level of enzyme. It also appeared that the wild–type level of
ATP synthase in E. coli gives a minimal dependence of growth rate on the
enzyme content.

6.1.1.5 Molecular genetics: anti–sense RNA

Another method of varying the expression of a chromosomal gene is to insert
a gene that expresses anti–sense RNA relative to the normal gene transcript.
This results in hydrogen bonding between the two complementary strands and
formation of double–stranded RNA, which tends to be translated poorly and
degraded quickly. Thus the amount of active enzyme protein formed can be
reduced below wild–type levels. An example of the application of this tech-
nique in Metabolic Control Analysis was reported in 1991 by Mark Stitt and
his co–workers,189, 243 who used it to determine the flux control coefficients
of the enzyme responsible for CO2 fixation in plant photosynthesis: ribulose–
bisphosphate carboxylase (or rubisco, EC 4.1.1.39). Many plant biochemists
seem to assume that this enzyme is ‘rate–limiting’ for photosynthesis, partly
because it is present in an exceptionally high concentration, for an enzyme,
of about 4 mM in the chloroplasts of green plants; this concentration, cou-
pled with the large amounts of plant biomass, makes it the most abundant
enzyme in the biosphere. Biochemists are also dissatisfied with it, since some
2 billion years of evolution have not eliminated its side reaction with oxygen,
instead of CO2 , that leads to apparently wasteful photorespiration. On these
grounds, plant molecular biologists had started programmes to engineer im-
proved versions of the enzyme in the expectation that this would increase the
rate of carbon fixation by plants. In fact, at the time Stitt’s group started
their experiments, no–one knew whether increasing the activity of rubisco
150 6 Measuring control coefficients

1.0 1.0 4 4 4
4
0.8 0.8

0.6 0.6
Crubisco

Crubisco
J

J
0.4 0.4
a) b)
0.2 0.2

0.0 0.0
0 20 40 60 80 100 0 20 40 60 80 100
Rubisco, % wild-type level Rubisco, % wild-type level

1.0

0.8

0.6
Crubisco
J

0.4
c)
0.2

0.0
0 20 40 60 80 100
Rubisco, % wild-type level

Fig. 6.5: Flux control coefficients of rubisco on photosynthetic carbon assimilation

in transgenic tobacco plants.
The diagrams have been replotted from the results obtained by Stitt’s group. 243 The
amount of rubisco was reduced below wild–type levels by using transformed plants ex-
pressing antisense RNA. a) Variation of the flux control coefficient at (◦) high (1050
µmol. quanta.m-2 .s-1 ) and ( ) moderate (350 µmol. quanta.m−2 .s−1 ) light intensities.
Other conditions: 35 Pa. CO2 , 65% relative humidity, 22◦ C. b) Variation of the flux
control coefficient at ( ) above ambient (45 Pa.) and (4) below ambient (25 Pa.)
CO2 . Other conditions: 1050 µmol. quanta.m-2 .s-1 , 65% relative humidity, 22◦ C. c)
Variation of the flux control coefficient at (•) 85% and (4) 35% relative humidity.
Other conditions: 1050 µmol. quanta.m-2 .s-1 , 35 Pa. CO2 , 22◦ C.
6.1 Manipulation of enzyme activity 151

in plants would change the rate of photosynthesis, because the flux control
coefficient had not been measured. In the experiments, tobacco plants were
transformed with an ‘antisense’ gene to rbcS, the gene for the nuclear–encoded
small subunit of rubisco. The transformed plants exhibited varying degrees
of reduced expression of the enzyme, with only minor changes in the contents
of other photosynthetic enzymes. The flux control coefficients of the enzyme
on photosynthesis in the leaves were determined by comparing the rate of
carbon fixation in leaf disks cut from wild–type and transformed plants under
varying conditions of light, CO2 and humidity. In spite of rubisco’s reputa-
tion as the ‘rate–limiting’ enzyme of photosynthesis, the experiments showed
that the maximum flux control coefficient was about 0.8 for the wild–type
range of enzyme contents; the highest values were only observed with strong
illumination, high humidity and low CO2 , whereas they fell to about 0.1 at
high levels of CO2 or low light intensity (Fig. 6.5), and were intermediate at
average levels of these environmental parameters. Thus there must generally
be other steps that are also contributing to the control of the rate of photo-
synthesis, and the influence of environmental factors is very strong. Given
the values of the flux control coefficients, Fig. 5.3, which shows the expected
change in flux for a change in enzyme activity, reveals that very substantial
improvements in rubisco activity would have to be made to produce modest
changes in photosynthetic flux under most conditions, at least for tobacco
plants.

6.1.2 Natural alteration of expressed activity

The responses of an organism to changes in its diet or environment generally
include metabolic adaptations and, especially when the changes are sustained,
these often reflect alterations in the amounts of enzymes. In some pathways,
the amounts of several or even many of the enzymes change (see Chapter
8.1), but there are also cases where the number of enzymes affected may
be small, even just one. Indeed, in Chapter 4.7, I mentioned that this was
one of the pieces of evidence used in the qualitative identification of important
regulatory enzymes, and Metabolic Control Analysis shows that this has some
justification. For, as was seen in Fig. 5.3, a change in the amount of a single
enzyme is only likely to cause a significant change in flux if its flux control
coefficient is reasonably close to 1. Moreover, where the change in metabolic
flux in response to one of these natural signals can be correlated with the
change in a single enzyme, there is the possibility of going further than this
qualitative conclusion and of estimating the flux control coefficient of the
enzyme.
For example, Salter et al.207, 208 were interested in the factors determin-
ing the rate at which the mammalian liver broke down amino acids from the
blood. Not surprisingly, this biochemical system responds to the requirements
that the amount of protein in the diet and the physiological state of the mam-
mal place upon it. Accordingly, they used a range of dietary and hormonal
treatments that specifically change the degree of induction of an enzyme in
152 6 Measuring control coefficients

the pathway for the catabolism of the amino acid tryptophan — tryptophan
2,3–dioxygenase (EC 1.13.11.11) — in the livers of rats. Tryptophan that has
been brought into the cells by a transporter in the plasma membrane is con-
verted by this enzyme into kynurenine, which is in turn further catabolized by
kynureninase and other enzymes. Hepatocytes (i.e. liver cells) isolated from
these treated rats retained the induced changes in enzyme amount, and the
differences in their rate of breakdown of tryptophan could be measured under
standardized conditions. The hyperbolic relationship between the amount of
tryptophan 2,3–dioxygenase and the catabolic flux is shown in Fig. 6.6. The
enzyme’s flux control coefficient varied from 0.75 in the rats on a normal
diet (at the lower end of the range of enzyme content) to 0.25 in the max-
imally induced state. We shall see in a later section that they managed to
identify where most of the remaining control could be found by other means
(Section 6.1.4.3, p. 166 & Section 6.3, p. 171).
I have already the mentioned experiments that estimated the control coef-
ficient of β–galactosidase on the growth rate of E. coli. Dykhuizen et al.61 also
carried out further experiments in which they used variable induction of the
lac operon by the gratuitous inducer IPTG to produce parallel changes in the
activities of the lactose permease and the β–galactosidase, since the genes are
in the same operon under the control of a single promoter. As in their experi-
ments mentioned previously (p. 144), in which the activity of β–galactosidase
was varied at a constant level of the permease, the rate of lactose metabol-

10

8
Tryptophan flux

0
0 20 40 60 80 100 120

Tryptophan 2,3-dioxygenase, nmol·h-1 ·mg-1

Fig. 6.6: The effect of varying tryptophan 2,3–dioxygenase activity on the

catabolism of tryptophan in hepatocytes.
The enzyme activity was adjusted by various dietary and hormonal treatments of rats. 207
The smooth curve is a fitted rectangular hyperbola. 233
6.1 Manipulation of enzyme activity 153

100
75
Fitness, %

50
25
0

100
P, % 50 100
50
0 0 G, %

Fig. 6.7: Dependence of the growth of E. coli on variation of the permease and
β–galactosidase activities
The axes are expressed as percentages of metabolic flux and enzyme activities (P for
permease and G for galactosidase) obtained with maximal induction of the lac operon.
The experimental values at 100% permease activity are the same results as in Fig. 6.4.
The three–dimensional surface is that fitted by Dykhuizen et al.

ism was measured in terms of the relative growth rate in a chemostat under
conditions of lactose limitation. This time, however, they were attempting to
measure the flux control coefficient of the permease. Although in this case two
of the pathway enzymes changed simultaneously, the contribution from the
changes in β–galactosidase was known from the previous experiments. They
combined the results of the two experiments and fitted them to an equation
that assumed the metabolic flux depended on each enzyme in a rectangular
hyperbolic manner, as shown in Fig. 6.7. When the enzymes are only partly–
induced by the natural effect of the lactose in the chemostat, the two flux
control coefficients can be calculated to be Cperm = 0.53 for the permease
and Cβ−gal = 0.04 for the β–galactosidase. (These figures differ slightly from
those given by Dykhuizen et al. in their paper, but have been calculated from
their fitted function.) The coefficients at the maximally–induced level of ex-
pression are 0.11 and 0.004. One surprising conclusion from this numerical
analysis, which is qualitatively apparent in Fig. 6.7, is that the production
of permease would have to be increased 30–fold to reduce its flux control co-
efficient to the same level as that of the β–galactosidase, but the coordinate
expression of the proteins in the same operon prevents this happening. An-
other interesting feature is that, in these experiments, the lactose permease
had a relatively large flux control coefficient, but in the same organism, the
154 6 Measuring control coefficients

0.10

0.08

Glycolytic flux 0.06

0.04

0.02

0.00
0 0.5 1 1.5 2 2.5 3

Enzyme activity

Fig. 6.8: Dependence of glycolytic flux in a rat liver homogenate on added en-
zymes.
The enzymes added were hexokinase ( ) and phosphofructokinase (4). The results are
those of Torres et al.248 with computed best–fit rectangular hyperbolas. The leftmost
point on each curve represents the original activity in the homogenate. The phospho-
fructokinase activity has been multiplied by a factor of 10 for display purposes. Titration
with glucose–6–phosphate isomerase gave no change in flux.

experiments by Ruijter et al. described above showed the glucose PTS system
had a negligible flux control coefficient203 on growth rate. No doubt a relevant
factor is that in the lactose utilization experiments, the bacteria were growing
in a chemostat on limiting lactose, whereas in the others, the bacteria were
growing in batch mode on excess glucose.

6.1.3 Titration with purified enzyme

The experiments described so far have used a variety of methods to force
cells to make more or less of a chosen enzyme. But many enzymes can be
obtained in purified form, and are often available commercially. Why not just
take one of these purified enzymes and add it to a metabolizing system in
order to discover the effect of increasing the enzyme content? The obvious
difficulty is that the metabolites inside cells would not get out across the
plasma membrane to reach the enzyme, and the enzyme would be too large to
cross the plasma membrane into the cell. However, biochemists do not always
work on intact cells and tissues: cell homogenates often exhibit functioning
metabolic pathways even though the permeability barriers have been removed
by membrane disruption.
Enzyme addition has been tried on rat liver homogenates, in studies of
6.1 Manipulation of enzyme activity 155

their glycolytic activity carried out by Enrique Meléndez–Hevia’s group in

the Canary Islands, starting in the mid 1980s. The experiment was designed
to study the distribution of control in the upper part of the glycolytic pathway
in liver. To simplify the problem, the pathway was effectively limited to the
first three enzymes by adding an excess of (fructose–bis phosphate) aldolase
and glycerol–3–phosphate dehydrogenase.248 This traps all the fructose–1,6–
bis phosphate produced by hexokinase IV (glucokinase), glucose–6–phosphate
isomerase and 6–phosphofructokinase, and converts it into glycerol. Further-
more, the rate of utilization of NADH during glycerol production can be
determined spectrophotometrically and indicates the flux in the pathway.
Titrating this system with various additional amounts of the more readily
available yeast hexokinase (HK, for supplementing hexokinase IV) and 6–
phosphofructokinase (PFK) gave hyperbolic responses of the flux (Fig. 6.8),
but there was virtually no response to extra glucose–6–phosphate isomerase
(GPI). The flux control coefficients derived from the fitted hyperbolas for the
liver homogenate without added enzymes are233 CHK = 0.79, CGP I = 0.0 and
CP F K = 0.21. (Torres et al.248 reported marginally different values calculated
from parameters derived from double reciprocal plots of their results.) Note
that the results show no discrepancy from the summation theorem, as the co-
efficients add up to 1 (though this is somewhat fortuitous as the experimental
uncertainties are of the order of 5%).
These experiments were an interesting demonstration of the potential of
a simple method though their physiological relevance is doubtful. The rea-
sons include: liver is more typically gluconeogenic than strongly glycolytic;
the homogenate was more dilute than cytosol; the study was limited to just
three enzymes, and the method used to shorten the pathway is likely to lower
fructose–1,6–bis phosphate below in vivo levels. On the other hand, some very
rough enzyme supplementation experiments performed by other groups on red
blood cells, which are only capable of anaerobic conversion of glucose to lac-
tate, are consistent with significant control of the glycolytic flux to lactate
being shared predominantly by hexokinase and phosphofructokinase.

6.1.4 Titration of enzymes by specific inhibitors

Specific inhibitors of enzymes have long had a place in the study of metabol-
ism, because the involvement of an enzyme in the formation of a particular
metabolic product could be shown by demonstrating that large doses of the
inhibitor blocked the pathway. Blocking a pathway with inhibitors was also
a way of increasing the concentrations of metabolic intermediates to more
readily detectable levels during the initial discovery phase of metabolic bio-
chemistry.
In 1979 Rognstad suggested application of such inhibitors to the identi-
fication of rate–limiting steps in metabolic pathways.200 He expected that
if an inhibitor acted on the rate–limiting enzyme of a pathway, it would
cause an immediate hyperbolic inhibition of pathway flux, starting at the
lowest concentration that had an effect on the enzyme. For a strong non–
156 6 Measuring control coefficients

competitive inhibitor, Rognstad proposed that the graphical appearance of

the reciprocal of pathway flux against inhibitor concentration would be as
shown in Fig. 6.9a), on the basis that non–competitive inhibitors give this
type of linear plot with isolated enzymes. On the other hand, if the pathway
flux did not initially respond to an inhibitor of an enzyme at concentrations
known to affect the enzyme, but showed a stronger response at higher in-
hibitor concentrations, then the enzyme was originally ‘non–rate–limiting’
but became rate–limiting as its activity was reduced. This would give rise to
the ‘biphasic’ line shown in Fig. 6.9a), as there is supposedly no effect on flux
until the enzyme has been inhibited sufficiently to become the rate–limiting
enzyme, but thereafter, the plot follows the line expected for the enzyme
inhibition. Translated into a flux v. inhibitor graph (Fig. 6.9b), the rate–
limiting enzyme would give a hyperbolic plot, whereas the non–rate–limiting
enzyme would show a biphasic plot again (or sigmoidal if the sudden transi-
tion were rounded off). A somewhat more realistic plot, typical of the curves
observed in practice, is shown in Fig. 6.10. Unfortunately, if these are plotted
as reciprocal flux v. inhibitor, both of the cases give curved (or biphasic)
lines. Rognstad illustrated his proposal with results he obtained on the glu-
coneogenic pathway for the formation of glucose from lactate in isolated rat
hepatocytes. 3–Mercaptopicolinate, a specific inhibitor of the enzyme phos-
phoenol pyruvate carboxykinase (PEPCK, EC 4.1.1.32) gave the result he pre-

10.0
a) 1.0
8.0 b)
0.8

6.0
1/Flux

0.6
Flux

4.0
0.4

2.0 0.2

0.0 0.0
0 1 2 3 4 5 0 1 2 3 4 5

[Inhibitor] [Inhibitor]

Fig. 6.9: Expected effects of non–competitive inhibitors on pathway flux according

to Rognstad.
a) The reciprocal of pathway flux has been plotted against inhibitor concentration be-
cause this plot is linear for an isolated enzyme. The solid line shows the result expected
if the inhibitor acts on a rate–limiting enzyme, and the dashed line the biphasic response
expected if it acts on one that is not. The numeric scales are arbitary. b) The same
responses plotted as flux against inhibitor concentration.
6.1 Manipulation of enzyme activity 157

1.0

0.8

Flux 0.6

0.4

0.2

0.0
0 1 2 3 4 5

[Inhibitor]

Fig. 6.10: Action of an inhibitor on pathway flux.

The plot shows a more realistic graph of the response of the flux to a non–competitive
inhibitor than Fig. 6.9b). The solid line is for inhibition of an enzyme with a flux control
coefficient of 1, whereas the dashed line is for an enzyme with a coefficient of 0.05. The
curves were plotted with an approximate equation derived by Rankin Small 229 that gives
curves very similar to those experimentally observed.

dicted for a rate–limiting enzyme (Fig. 6.12). On the other hand, aspartate
transaminase (which is involved in the shuttle between the mitochondria and
the cytoplasm during the conversion of pyruvate to phosphoenol pyruvate)
gave biphasic responses with the inhibitors aminoxyacetate and cycloserine
and was therefore classed as not rate–limiting.
Readers should not by now be surprised to learn that Metabolic Control
Analysts were not content with Rognstad’s use of inhibitors to divide enzymes
into the two classes of rate–limiting and not rate–limiting. However, perhaps
just as important is the question of whether the shapes of the plots are reliable
qualitative indicators for making this classification anyway.

6.1.4.1 Theory
The application of Metabolic Control Analysis to the problem was first tackled
by the group of researchers at the University of Amsterdam who have given
experimental and theoretical studies of Metabolic Control Analysis such a
great boost from the early 1980s onwards. This group included, amongst
others, Tager, van Dam, Groen, Westerhoff, Wanders and Meijer, some of
whom have already had their work mentioned in this chapter. Rognstad’s
proposal was analysed by Bert Groen et al. in 1982,90 both in general terms,
and with reference to his specific conclusion about the role of PEPCK in
158 6 Measuring control coefficients

gluconeogenesis.

1.0
T(angent)
0.8

0.6
Flux

Imax
0.4

0.2

0.0
0 0.5 1 1.5

[Inhibitor]

Fig. 6.11: Titration with an irreversible noncompetitive inhibitor.

The curve is from inhibition of an enzyme with a flux control coefficient of 0.2 by an
inhibitor with very high affinity. The initial slope at zero inhibitor is shown by the dashed
line T. The linear region at the other end has been extrapolated to zero flux to give the
amount of inhibitor, Imax needed to bind all the enzyme.

The effect of an inhibitor, I, of some enzyme xase, on flux is described

by a response coefficient RIJ , as defined earlier in Eqn. [5.21]. As shown
in Eqn. [5.22], Chapter 5.4, this response is composed of the enzyme’s flux
control coefficient and the elasticity of the inhibitor I on xase, i.e. :

RIJ = Cxase
J
εxase
I (6.2)

Therefore, the flux control coefficient of the enzyme can be obtained from
the response of the metabolic flux to an inhibitor relative to the effect of the
inhibitor on the isolated enzyme:
J
Cxase = RIJ /εxase
I (6.3)

This is not, though, a suitable equation for experimental implementation; it

looks as though the way to evaluate it would be to measure the flux at a series
of inhibitor concentrations, and use these to estimate the response coefficient.
However, we want the value of the flux control coefficient in the absence of
inhibitor, and the response coefficient is not easy to obtain for I = 0. For
example, the response coefficient is the slope of a double logarithmic plot of
flux v. I, but log 0, for I = 0, cannot be plotted. The way round this devised
by Groen et al.90 is as follows. The equation above is rewritten with the
6.1 Manipulation of enzyme activity 159

response coefficient and elasticity written out in full, for the limit (lim I→0 ) as
I approaches zero:
Áµ ¶
J ∂J I ∂vxase I
lim Cxase = lim (6.4)
I→0 I→0 ∂I J ∂I vxase

The inhibitor concentrations in the numerator and denominator cancel to

give: Áµ ¶
J ∂J 1 ∂vxase 1
lim Cxase = lim (6.5)
I→0 I→0 ∂I J ∂I vxase
The two terms of this equation are easier to evaluate. The numerator, (∂J/∂I)(1/J)
is the slope of a graph of J v. I (in the limit that I = 0) times 1/J0 , where
J0 is the pathway flux in the absence of inhibitor. Let us call the value of
the slope T (for tangent) as in the example in Fig. 6.11; the numerator is
therefore T /J0 .
The denominator term, (∂vxase /∂I)(1/vxase ) in the limit where I = 0 (i.e.
limI→0 ), can be determined from the effect of the inhibitor on the enzyme’s
kinetics, but different types of inhibitor give different results. The simplest
case is the very strong inhibitor that can be considered effectively irreversible
because once bound to the enzyme, it dissociates very slowly. The degree of
inhibition of the enzyme is proportional to the amount of inhibitor added, up
to 100% inhibition when the amount of inhibitor Imax is equal to the amount
of enzyme sites. That is, the enzyme activity goes linearly from vxase at
I = 0 to 0 at I = Imax . The slope ∂vxase /∂I is therefore −vxase /Imax , so
our denominator term in Eqn. [6.5], (∂vxase /∂I)(1/vxase ) is −1/Imax . Imax
can often be estimated from the inhibitor titration graph itself as shown in
Fig. 6.11. If the values for the terms are now substituted into Eqn. [6.5],
the value of the flux control coefficient from a titration with an irreversible
inhibitor will be:90
J Imax
Cxase = −T (6.6)
J0
For other types of inhibitor, the denominator term can either be evaluated
experimentally by an inhibitor titration of the (isolated) enzyme activity, or
by calculation from the rate law and the inhibition constant. In the first case,
Douglas Kell and Hans Westerhoff pointed out that if the pathway flux and
the enzyme rate were both expressed as fractions of their uninhibited values
and plotted on the same graph, then Eqn. [6.5] is just the ratio of the initial
slopes of the two curves.129 Jean–Pierre Mazat and Thierry Letellier have
been using this method recently in Bordeaux for the study of the control of
oxidative phosphorylation in relation to diseases caused in humans by mito-
chondrial mutations; they used inhibitor titrations of both the flux and the
isolated enzyme activity. Their application of this method will be seen later
in Fig. 6.16.
As an example of the method using calculation from the enzyme rate law,
the denominator term for a noncompetitive inhibitor can be shown to be
160 6 Measuring control coefficients

1.0

0.8

Fractional flux or rate

0.6

0.4

0.2

0.0
0 10 20 30 40 50 60 70 80

[Mercaptopicolinate], µM

Fig. 6.12: Inhibition of gluconeogenesis from lactate by mercaptopicolinate.

The points are fractional rates of glucose formation by rate hepatocytes measured by
Rognstad200 and replotted from his graph of 1/Flux against the inhibitor concentration.
The solid curve through the points is a transformation of the line Rognstad fitted on
his graph. The dotted line is the fractional rate of PEPCK against mercaptopicolinate
calculated from its Ki value of 3µM.

−1/(I + Ki ), which when substituted with I = 0 into Eqn. [6.5] gives:

J Ki
Cxase = −T (6.7)
J0

The conclusion from this is that the flux control coefficient of an enzyme
is related both to the initial slope of the inhibitor titration curve and the
strength of the inhibitor effect on the isolated enzyme. The shape of the titra-
tion curve does not seem to give a reliable indication. This became apparent
when Groen et al.90 used Rognstad’s data200 on the inhibition of gluconeoge-
nesis by the action of mercaptopicolinate on PEPCK (phosphoenol pyruvate
carboxykinase) and applied their equation. Whereas Rognstad had classed
the enzyme as rate–limiting because of the shape of the inhibition curve, they
calculated that the flux control coefficient was only 0.08. We shall see later
that this value was corroborated by an entirely different type of experiment.
Rognstad’s results are shown in Fig. 6.12. Also plotted is the expected inhi-
bition curve for PEPCK based on the reported inhibition constant of 3µM; as
can be seen, the initial slope of the enzyme inhibition curve is about 10 times
steeper than the flux inhibition curve, corresponding to a flux control coeffi-
cient of about 0.1 on the basis of the Kell & Westerhoff slope ratio method.
6.1 Manipulation of enzyme activity 161

µ
¡
¡
1 2 3 ¡4
succinate - - - ∆µ̃H+
@ 5
@
R
@
ATPm
@ 6
@
R
@
ATPc
@7
@
R
Glc6P

Fig. 6.13: Schematic representation of mitochondrial oxidative phosphorylation

system.
The following steps were studied by Groen and his colleagues: 92 1, dicarboxylate carrier
for succinate transport; 2, bc1 complex in Complex III (the succinate dehydrogenase
(ubiquinone), Complex II, was not measured); 3, cytochrome–c oxidase; 4, proton leak;
5, ATP synthase (not measured in these experiments); 6, adenine nucleotide transloca-
tor, and 7, hexokinase. ∆µ̃H+ is the protonmotive force.

6.1.4.2 Experimental application to oxidative phosphorylation

The study of oxidative phosphorylation has been a very fruitful area for the
application of inhibitor titrations. This is partly because there are more strong
and highly specific inhibitors known for this system than for any other. How-
ever, there was also a powerful motivation to apply Metabolic Control Analysis
to oxidative phosphorylation because it was a notable example of how the con-
cept of the ‘rate–limiting step’ failed to achieve anything but the generation of
unresolved arguments. In particular, different research groups debated in the
literature about whether the ‘rate–limiting’ step was cytochrome–c oxidase or
the adenine nucleotide translocator when mitochondria were phosphorylating
ADP at maximal or near–maximal rates.

Supplement: Oxidative phosphorylation revision. Oxidative phospho-

rylation is most commonly studied on intact mitochondria isolated from cells and
consists of the steps from the oxidation of compounds such as NADH and succinate
by oxygen and the associated phosphorylation of ADP by phosphate to form ATP.
The rate of substrate oxidation is determined from the oxygen consumption, which
is measured with an oxygen electrode. Although mitochondria would normally gen-
erate their own oxidizable materials via pathways such as the tricarboxylic acid cycle
and β–oxidation, in the experiments, the oxidizable substrate is added externally.
This brings minor complications. For example, succinate is normally produced in-
ternally in the matrix of the mitochondrion in vivo, but has to be transported across
the mitochondrial membranes in the in vitro experiments. In addition, externally–
162 6 Measuring control coefficients

added NADH cannot be oxidized by animal mitochondria because it does not cross
the mitochondrial inner membrane and the NADH oxidase is only accessible from
the mitochondrial matrix. Therefore, to study the oxidation of NADH, it is neces-
sary to supply the mitochondria with metabolites that will be transported into the
mitochondrial matrix and be oxidized there by NAD+ –dependent dehydrogenases;
examples of suitable compounds include β–hydroxybutyrate, pyruvate plus malate
and glutamate plus malate.
The oxidation of the substrates by oxygen is carried out by the electron transport
chain, which is composed of a number of complexes of electron transport compo-
nents. Complex I is NADH:ubiquinone oxidoreductase (EC 1.6.5.3); Complex II is
succinate dehydrogenase (ubiquinone), EC 1.3.5.1. (The supposed succinate dehy-
drogenase, described by many textbooks, that reacts with FAD is merely a partial
reaction of the catalytic cycle of this complex.) The ubiquinol (or reduced Coen-
zyme Q) produced by these complexes is oxidized by Complex III, or ubiquinol–
cytochrome-c reductase (EC 1.10.2.2). Reduced cytochrome-c is oxidized with oxy-
gen by the cytochrome-c oxidase complex (EC 1.9.3.1) or Complex IV. In animal
mitochondria, Complexes I, III and IV are coupled to the phosphorylation of ADP,
but this does not imply, as previously assumed, that oxidation of an NADH gener-
ates 3 ATP molecules whereas oxidation of a succinate generates 2.
The linkage between oxidation and phosphorylation is described by Mitchell’s
chemiosmotic theory. The electron transport complexes coupled to phosphorylation
move protons out of the mitochondrial matrix across the inner mitochondrial mem-
brane, thereby creating both a concentration difference (or ∆pH) and a membrane
potential difference (or ∆ψ) that together constitute the protonmotive force (PMF
or ∆µ̃H + ). The PMF is the reservoir of energy that drives phosphorylation of ADP.
This reservoir is tapped by the H+ –transporting ATP synthase (EC 3.6.1.34), which
allows protons back into the mitochondrial matrix and captures their energy for the
formation of ATP from ADP and phosphate. The oxidation of NADH results in 4
protons being pumped out at Complex I, 2 at complex III and 4 at complex IV,
giving a total of 10 protons. It is thought that a total of 4 protons are needed to
take in ADP and phosphate from outside the mitochondria, turn them into ATP
and return the ATP outside the mitochondrion. Therefore, the P/O ratio, that is
the number of ATP molecules formed per oxygen atom used in oxidation of NADH
is 10/4 or 2.5. Since succinate is not oxidized via Complex I, its P/O ratio is 6/4
or 1.5.
Some of the PMF can dissipate without protons going through the ATP synthase.
The principal cause of this is the ‘leak’ reaction, which is most significant when
phosphorylation is going slowly and least significant when phosphorylation is rapid.
The ‘leak’ reaction can be stimulated by uncouplers, which increase the permeability
of the mitochondrial membrane to protons.
The phosphorylation process requires that ADP is transported into the mito-
chondrion and returned back to the outside as ATP. These two transport steps are
coupled as an exchange process catalyzed by the adenine nucleotide translocator.
There are two basic types of experiments performed on mitochondria in the study
of oxidative phosphorylation. In pulse experiments, mitochondria are supplied with
an oxidizable substrate and their rate of respiration measured. A known amount of
ADP is added; this causes a noticeable stimulation of respiration, (known as state 3
respiration) until all the ADP has been converted to ATP, when the respiration rate
returns close to its original rate (state 4). From the oxygen consumption profile,
it is possible to calculate the amount of oxygen used to phosphorylate the known
6.1 Manipulation of enzyme activity 163

amount of ADP. However, in this experiment, the steady states are only transient;
the mitochondria pass from zero phosphorylation to maximal phosphorylation rate
and back with relatively brief periods in the intermediate states. This is unlikely
to be similar to physiological conditions in most cell types. The alternative is to
set up a steady state phosphorylation rate. One way of doing this is to use the
enzyme hexokinase in the presence of glucose as a means of recycling the ATP back
to ADP (along with the formation of glucose–6–phosphate, which can be measured
to find out the phosphorylation rate) as shown in Fig. 6.13. By varying the amount
of hexokinase, the rate of phosphorylation, and hence oxygen consumption, can be
adjusted, and the steady state lasts until the oxygen runs out.
Most of the characteristics usually described in the textbooks relate to mam-
malian mitochondria, or more specifically, rat liver mitochondria. Experiments, in-
cluding Control Analysis studies, have been performed on mitochondria from other
sources, for example from yeast and from various plants, even though it is more
difficult to isolate the organelles succesfully from them. There are, however, some
significant differences between the details of the electron transport components and
coupling sites of these mitochondria from the animal ones.

Bert Groen and his colleagues in Amsterdam92 applied their inhibition

titration technique in the original Control Analysis study on rat liver mi-
tochondria. The system they analysed is shown in Fig. 6.13. Several of
the steps were titrated with specific inhibitors. For example, the adenine
nucleotide translocator was titrated with carboxyatractyloside, which is an
essentially irreversible inhibitor, at different respiration rates obtained by us-
ing different amounts of hexokinase to alter the rate of ADP regeneration.
Examples of the curves they obtained are shown in Fig. 6.14. From the ini-
tial slopes of the curves and the [Imax ] value for carboxyatractyloside, the
flux control coefficient of the translocator was measured at each respiration
rate. In addition, the dicarboxylate carrier (which transports the respiratory
substrate succinate into the mitochondria) was titrated with the competitive
inhibitor phenylsuccinate. The flux control coefficient of the hexokinase was
determined by the addition of extra amounts of the enzyme, i.e. the enzyme
titration method described previously. Finally, the flux control coefficient
of the proton leak was measured by an analogue of the inhibitor titration
method — an activator titration. In this, an uncoupler, carbonyl cyanide p-
trifluoromethoxyphenylhydrazone (FCCP) was added to stimulate respiration
by increasing the the leak of protons through the membrane. The flux control
coefficients of all these steps on respiration varied with the respiration rate as
shown in Fig. 6.15.
A further set of measurements was made at the maximal rate of phos-
phorylating respiration (State 3) by titrating cytochrome-c oxidase with the
non–competitive inhibitor azideand the bc1 complex with the non–competitive
inhibitor hydroxyquinoline–N–oxide. The results for the 6 steps studied are
given in Table 6.2.
These results are a powerful example of the utility of Metabolic Control
Analysis. For a start, they show the complete futility of the previous ar-
guments between different research groups as to whether the cytochrome-c
164 6 Measuring control coefficients

Fig. 6.14: The effect of carboxyatractyloside on oxygen uptake in rat liver mito-
chondria.
Mitochondria were incubated with succinate, ATP, glucose and varying amounts of hex-
okinase to give different initial respiration rates, as measured by oxygen consumption
with an oxygen electrode. At each respiration rate, the system was titrated with in-
creasing amounts of inhibitor. Carboxyatractyloside does not cause complete inhibition
of respiration, since the leak reaction is responsible for some oxygen consumption when
phosphorylation has been completely inhibited. The I max value for the inhibitor was
estimated from the curves. This graph is reproduced from Groen’s PhD thesis. 89

oxidase or the adenine nucleotide translocator was the rate–limiting step by

showing that neither of them was. Instead no step is rate–limiting because
the control of respiratory flux is shared between the different steps. In ad-
dition, the results show how that distribution changes according to the rate
of phosphorylation that the mitochondria are performing. In particular, the
ATP–consuming step, hexokinase, has a zero flux control coefficient at the
6.1 Manipulation of enzyme activity 165

lowest and highest respiration rates, but goes through a maximum of 0.52 at
intermediate rates. The step that comes closest to being rate limiting is the
leak reaction at the lowest respiration rates, where there is no phosphoryla-
tion and all the respiration is accounted for by the leak. However, the method
of calculation of this flux control coefficient from the activator titration has
been criticized, and so this flux control coefficient has almost certainly been
over–estimated in these experiments. Nevertheless, we shall see later that
a different method of performing this experiment has produced essentially
similar results. The results in Table 6.2 show that the measured flux control
coefficients account for 0.86 relative to the expected summation theorem total
of 1. Two steps have not been measured though: the succinate dehydroge-
nase (ubiquinone) (Complex II), and the proton–translocating ATP synthase.
These might account for the missing control, if indeed there is any missing as
opposed to some experimental error. Although Groen et al. did not measure
the flux control coefficient of the ATP synthase, they believed it to be low
because of the results of previous experiments on the inhibition of oxidative
phosphorylation by the ATP synthase inhibitor oligomycin.
Some very similar experiments carried out by Thierry Letellier and Jean–
Pierre Mazat in Bordeaux148 illustrate the application of the slope ratio

1.0

0.8
Flux control coefficient

0.6

0.4

0.2

0.0

0 20 40 60 80 100

Relative respiration rate, %

Fig. 6.15: The flux control coefficients of 4 steps in oxidative phosphorylation as

a function of respiration rate.
Mitochondria were incubated and the flux control coefficients determined by various
titrations,89, 92 as in Fig. 6.14. Results of inhibitor titrations are shown for: (4) the
translocator and ( ) the dicarboxylate carrier. (∗) is ATP utilization titrated with
hexokinase. The control coefficient for the leak ( ) was determined with an activator
titration method.
166 6 Measuring control coefficients

Table 6.2: Distribution of control in rat

liver mitochondria in State 3 respiration.

No. Step Ciresp

1 Dicarboxylate carrier 0.33

2 bc1 complex 0.03
3 Cytochrome-c oxidase 0.17
4 Proton leak 0.04
6 Adenine nucleotide
translocator 0.29
7 hexokinase 0
Total 0.86

The results are from Groen et al.92 The

numbers of the steps correspond to those
in Fig. 6.13. Ciresp is the control coefficient
of step i on the respiratory flux.

method. Fig. 6.16 shows the inhibition curve of mitochondrial respiration with
pyruvate and malate as substrate, using the Complex IV inhibitor cyanide;
on the same scale is shown the inhibition of the ‘isolated’ Complex IV step
under the same conditions. The ratio of the slopes leads to a value of the flux
control coefficient for Complex IV of 0.2.

6.1.4.3 Other experimental applications

The inhibitor titration method has been applied in a range of other pathways,
some examples of which are summarized in Table 6.3. The results shown
there for the control coefficient of tryptophan uptake by rat hepatocytes on
tryptophan catabolism complement the results cited earlier on the role of
tryptophan dioxygenase in this pathway (Section 6.1.2, p. 151). The flux
control coefficient of uptake is 0.22 when that of the dioxygenase is 0.75,
and is 0.70 when that of the dioxygenase is 0.25. Thus between them, the
control coefficients of these two steps sum to almost 1, and according to the
summation theorem, this leaves little scope for other steps to have a significant
influence.

6.1.4.4 Problems
The versatility of the inhibitor titration as a method of determining flux con-
trol coefficients is demonstrated by the examples I have cited. Of course,
the method is capitalizing on a long history in biochemistry of discovering
and using inhibitors of enzymes, not least because inhibitors are potential
6.1 Manipulation of enzyme activity 167

100

80

Flux or rate, % 60

40

20

0
0 5 10 15 20

[KCN], µM

Fig. 6.16: Inhibitor titration of mitochondrial respiration flux and Complex IV rate
with cyanide.
The substrate for oxidation was pyruvate + malate. : inhibition of mitochondrial respi-
ration. The solid line is a fitted curve. 4: inhibition of complex IV activity. The dashed
line is calculated from the parameters that fitted the other points for mitochondrial
respiration. Data supplied by Letellier.148

drugs. In spite of these successes, there is a difficulty with the inhibitor titra-
tion, connected with the need to determine the initial slope at zero inhibitor
concentration of the flux against inhibitor curve. Examination of the experi-
mental results in Figs. 6.14 and 6.16 shows why this is not easy. The response
of the flux is often significantly curved, so there is not necessarily an initial
linear region. In any case, the experimental variance in the flux measurements
makes it difficult to tell whether the initial region is a concave curve, linear or
a convex curve, all of which are possible shapes. In addition, one form of the
method requires an estimate of Imax from the graph, and another requires the
initial slope of the enzyme rate against inhibitor graph. My student Rankin
Small228 found that these problems could cause bias and uncertainty in the
estimates of the flux control coefficients. Both he229 and Gellerich and his
colleagues81 proposed that it was better to use a computer fit to the curve.
The difficulties of doing this are made greater because the exact equation
governing the curves is generally not known, and would in most cases be
very complex if it were. Instead, approximate equations have been devised,
matched to the type and properties of the inhibitor, but computer simulation
tests do show that this approach does work better. The titrations of oxidative
phosphorylation shown in Fig. 6.16 were fitted in this way; the curve drawn
through the flux measurements was computer–fitted to Gellerich’s equation.
168 6 Measuring control coefficients

Table 6.3: Additional examples of the use of inhibitor titrations in Control

Analysis.
J
Enzyme Inhibitor Pathway CE Ref.

PEPCK mercaptopicolinate gluconeogenesis 0.24 198

(hepatocytes)
diacylglycerol 2–bromooctanoate triacylglycerol synthesis 0.76 154
acyltransferase
ornithine carbamoyl– norvaline urea cycle < 0.02 265
transferase
carbonic anhydrase acetazolamide urea cycle < 0.02 265
alcohol dehydrogenase TMSOa & ethanol catabolism 0.5–0.7 179
isobutyramide (rat liver)
tryptophan uptake phenylalanine Trp catabolism 0.22–0.70 207
kynureninase amino–oxyacetate Trp catabolism < 0.01 239
tyr aminotransferase amino–oxyacetate Tyr catabolism 0.29–0.71 207
glutaminase glutamine γ–hyrazide Gln catabolism 0.96 151
glyceraldehyde–3–P iodoacetate yeast glycolysis ≈0 22
dehydrogenase
glucose phosphotrans– xylitol glycolysis, Clostridium < 0.2 264
ferase system pasteurianum
glucose transport 6–chloro–6– respiration, Agro– 0.1–0.3 47
deoxyglucose bacterium radiobacter
glucose transport 6–chloro–6– exopolysaccharide syn– 0.1–0.3 47
deoxyglucose thesis A. radiobacter

J
Where a range of values has been given for the flux control coefficient (C E ), this
is because the value varies with the experimental conditions.

a tetramethylene sulphoxide

A measure of the success of this approach is that the fitted curve predicts the
degree of inhibition of the isolated enzyme step, and this predicted inhibition
curve is the one drawn through the experimentally measured enzyme rate val-
ues on that same graph. In this example, there was relatively little difference
between the flux control coefficients obtained by the slope ratio method and
the computer fit, but the latter is regarded as more reliable.

6.2 Control coefficients from computer models

It is impossible to look at the enzyme kinetic equations for the set of enzymes
in a pathway and to guess how the system will behave, but for the past thirty
years it has been possible to use computers to numerically simulate what will
happen. Computer simulation of metabolism was pioneered by Garfinkel, 79
Park and Chance amongst others. More recently it has played an important
6.2 Control coefficients from computer models 169

role in the development of Metabolic Control Analysis by allowing the rapid

evaluation of the control properties of particular metabolic structures. The
experiment that can be so difficult to do in the laboratory — specifically
altering the activity of one enzyme whilst keeping all other conditions constant
— is so much simpler in a computer model, where it generally involves no more
than the alteration of a single number.
This would suggest that a simple way of finding the flux control coeffi-
cients of a metabolic pathway would be to use the experimental information
about the pathway to build a computer model, and then to use the model
to calculate the flux control coefficients. Unfortunately, the first stage has
proved to be very difficult, and there are still relatively few computer simu-
lations of metabolic pathways that are realistic and reliable. This is in spite
of the fact that there have been successive improvements in the computer
programs designed for simulating metabolism and calculating flux control co-
efficients, so that very little mathematical skill is needed to carry out the
simulations.52, 158, 209, 211 The problem seems to lie elsewhere: in the amount
and reliability of the experimental information and/or the assumption that
the simulation can use the kinetic equation for each enzyme. (Enzyme kinet-
ics equations are actually derived for conditions where enzyme concentrations
are very low relative to substrate concentrations, and this is often not the case
in cells.) It seems to be a common experience that when the experimentally–
determined information about the enzymes and their activities are incorpo-
rated into the metabolic model, it is rare for this to give results that match
well with the experimental observations of steady state concentrations and
fluxes. One particular finding is that it is only possible to make a computer
model behave in a realistic and stable manner by including some degree of
product inhibition on even the irreversible enzymes. (This is one of the lines
of evidence for the importance of product inhibition effects in spite of their
relative neglect in biochemistry. See also Chapter 6.3.4, p. 192.) Also, it
is frequently necessary to change the values of some of the enzyme activi-
ties by a hundred fold or more to make the simulation match the observed
results. This has been reported by Barbara Wright, for example,279 whose
research group has measured the amounts and properties of the enzymes of
carbohydrate metabolism in the slime mold Dictyostelium discoideum with
the express purpose of collecting the information needed to build a com-
puter model. Indeed, the model has been built, has succesfully represented
a number of the features of the slime mold metabolism, and has been used
to calculate control coefficients.2, 278 Nevertheless, the reasons for the discrep-
ancies in this and other examples (such as those mentioned in Chapter 4.4,
p. 93) are not completely clear, but probably include the existence of many
unknown weak interactions, between metabolic intermediates and enzymes,
that have not been characterized. There are also the possibilities of unknown
compartmentation of pathways and metabolites and of direct enzyme–enzyme
interactions.
Certainly, the small number of working simulations does show how far
the reductionist approach to biochemistry (building up explanations from the
170 6 Measuring control coefficients

molecular details; see Chap. 1.2.1) has to go before it can be regarded as truly
successful.
Many of the contributions from Reinhart Heinrich and his colleagues in
Berlin to the theory of Metabolic Control Analysis have been linked to to the
development of models of the metabolism of the human red blood cell from an
initial model with just the glycolytic pathway 193 through a series of versions to
one that includes the 2,3 bisphosphoglycerate bypass and, via the membrane
ATPase, ion fluxes and volume changes.97, 98, 194 Control coefficients for the
fluxes and metabolite concentrations have been calculated. Other models of
red cell metabolism, also developed in Berlin by Hermann–Georg Holzhütter’s
group, have included theoretical examination of the effects of inherited enzyme
deficiencies108, 109 and the regulation of the hexose monophosphate shunt.222
Of course, it is more reasonable to expect to be able to build a comprehensive
model of human erythrocyte metabolism because the cells contain no nulei or
mitochondria, so there is no nucleic acid and protein synthesis, nor oxidative
phosphorylation.
David Garfinkel, one of the pioneers of computer simulation of metabolism,
developed large models of catabolism in muscle cells. Kohn and colleagues
have applied Metabolic Control Analysis to some of these models.135, 136 How-
ever, it must be said that these models show one of the difficulties of large
computer models: once they reach a certain complexity, even if they are ex-
hibiting similar behaviour to the in vivo system, it can be almost as difficult
to understand why the model behaves as it does as it is for the real system.
A number of models of various aspects of photosynthesis have been pro-
posed to aid understanding of how its rate responds to environmental condi-
tions. Some of these have been subject to one or other types of sensitivity
analysis, including Metabolic Control Analysis.86, 87, 182, 276 Predictions made
by the models differ about the dependence of the control coefficients for the
photosynthetic flux on the conditions, but, like the various experimental stud-
ies reported earlier (e.g. Fig. 6.5), they show that whereas the flux control
coefficient for the CO2 –fixing enzyme, ribulose bisphosphate carboxylase, is
high in some conditions, in others, it declines and other steps have comparable
or higher coefficients.

6.3 Control coefficients from elasticities

There are three ways in which linkages with elasticities can be of use for the
experimental determination of control coefficients:
1. The connectivity theorem links the ratio of the control coefficients of
adjacent enzymes to the inverse ratio of their elasticities to their com-
mon metabolite. Thus if a control coefficient of one enzyme has been
determined by some means, knowledge of the elasticities of that enzyme
and its neighbour would allow the calculation of the unknown control
coefficient. This was used by Wanders et al. in Amsterdam265 to esti-
mate that the flux control coefficient of carbamoyl–phosphate synthase
6.3 Control coefficients from elasticities 171

on the flux to citrulline synthesis (as part of the urea cycle in rat hepa-
tocytes) was 0.96, relative to the control coefficient of the next enzyme
ornithine carbamoyltransferase, which had been determined by inhibitor
titration. (See Q. 1 at the end of this chapter.)

2. The response of a flux to an external metabolite (such as the pathway

source) is a relatively easy measurement to make; the response is equal
to the product of the flux control coefficient of the enzyme affected by
the external metabolite and the elasticity of that enzyme to the ex-
ternal metabolite. Thus if the elasticity value is combined with the
response coefficient, the flux control coefficient can be calculated from
Eqn. [5.21]. This was used by Groen et al.91 in their study of gluconeo-
genesis from lactate and pyruvate in hepatocytes. The response of the
flux to external pyruvate was divided by the elasticity to pyruvate to
obtain an estimate of the combined flux control coefficients of a block
of reactions at the start of the pathway. The result agreed well with
the results they obtained by a different method (see below). The same
technique was also used to determine the flux control coefficients of the
aromatic amino acid transporters on catabolism of the amino acids in
hepatocytes207 by measuring the responses of the rate of catabolism to
the external concentrations of tryptophan, tyrosine and phenylalanine
and dividing by the elasticity of the transporter. The resulting flux con-
trol coefficients varied between 0.2 and 0.9 depending on the amino acid
and the previous treatment of the hepatocytes. This formed part of the
study mentioned previously in Chapter 6.1.2, Fig. 6.6, Chapter 6.1.4.3
and Table 6.3.

3. As explained in the section on the connectivity theorem (5.3.3, p. 121),

all the control coefficients of a pathway are expressible in terms of elas-
ticities, and, for certain types of pathway, relative fluxes and concentra-
tions, by using the summation, connectivity and, if necessary, branch
point theorems. Therefore, it is in principle possible to measure the
elasticities and calculate all the control coefficients of a pathway. This
was first achieved by Bert Groen and his colleagues in Amsterdam in
their study of the control of gluconeogenesis from lactate in rat hepa-
tocytes,89, 91 which remains to this day the largest and most detailed
experiment of this type.

Of these three methods, the most informative is the third, but the control
coefficients obtained in this way are derived entirely indirectly, for the exper-
imental observations neither directly measure changes in flux as a function of
a known change in a single enzyme activity, nor relate the flux control coef-
ficients to another known flux control coefficient (as in the first method). In
addition, it involves certain assumptions, in particular:

• that all the relevant steps in the the metabolic system have been identi-
fied, and that all the significant influences of the metabolites on each of
172 6 Measuring control coefficients

lactate
mito
............................ NAD+ cytoplasm
..
0 ......................
....... ....
. NADH
pyruvate ............................1............... pyruvate
..........
.......
... ........
... ......
... .....
.....
2 .... ....
...
... ...
....
...
3
oxaloacetate ................................. oxaloacetate 5 ..
.
.
..
.... ...
...
. .
. .....
.
......
4 ..
...... . ..........
........ .
PEP
....
6 ....
...
..
6 .....
...
6 .....
..
..
6 .....
.
GAP ........................... ( DHAP )
... .........
..
7 .... .....................
..............
.....
.......
( FBP )
...
.
7 ....
.....
F6P
.....
8 ......
( G6P .. )
.
8 ....
....
glucose

Fig. 6.17: Gluconeogenesis from lactate in hepatocytes.

The source and sink are shown in boxes. The numbers on the reactions refer to the
groups used by Groen et al.91 for estimation of the flux control coefficients from the
elasticities. The 6 metabolites shown without brackets or boxes are the ones used for
elasticity estimates, though they were not all directly measured (see text). The reac-
tions are: 0, lactate dehydrogenase (assumed to be at equilibrium); 1, pyruvate trans-
port into the mitochondrion; 2, pyruvate carboxylase; 3, the malate and/or aspartate
shuttle systems for oxaloacetate transport; 4, PEPCK; 5, pyruvate kinase; 6, enolase,
phosphoglycerate mutase, phosphoglycerate kinase and glyceraldehyde–3–P dehydroge-
nase; 7, aldolase and fructose bisphosphatase; 8, glucose–6–P isomerase and glucose–6–
phosphatase. Hexokinase IV (glucokinase) and phosphofructokinase were inactive under
the conditions used.
6.3 Control coefficients from elasticities 173

these steps have been recognized. i.e. that all the non–zero elasticities
involved in the system are known.

• that the theorems of control analysis apply to the system. i.e. that
the metabolic system reaches a quasi steady–state (as defined in Chap.
1.4.1) and there are no features of the pathway that would require the
use of modified theorems or invalidate their application altogether. Such
problems are mostly beyond the scope of this book, though a brief ac-
count of some of them is given in the final Chapter 8.3, p. 280.

Provided that these conditions are satisfied, my research group and I showed
that it should be possible to carry out such analyses on any metabolic path-
way.69, 213 Still, in order to show that the assumptions are justified, it is best
if results obtained in this way are confirmed by direct measurement of one
of the control coefficients to check that similar results can be obtained by a
different method.
Before explaining some of the methods available to measure elasticities, I
will first give an example of how flux control coefficients have been determined
from them.

6.3.1 An experimental example

Groen et al.89, 91 determined the flux control coefficients for gluconeogenesis
with lactate as substrate in rat hepatocytes by measuring or calculating the
elasticity values. This was a new approach compared with previous studies of
gluconeogenesis, which had attempted to identify the rate–limiting step. As
we saw earlier, in section 6.1.4, Rognstad was one of the supporters of the
claims that the rate–limiting step was the enzyme phosphoenol pyruvate car-
boxykinase (PEPCK). There were other researchers who favoured pyruvate
carboxylase. The position of these enzymes in the pathway can be seen in
Fig. 6.17. The first problem that Groen and his colleagues had to solve was
ensuring that they could make observations in a quasi–steady state. They
solved this by perifusing the hepatocytes in a flowing, isotonic buffer. In this
way, the energy–providing substrate (the fatty acid oleate) and the substrate
for gluconeogenesis (lactate) were continually replenished. Along with the
lactate, it was necessary to provide sufficient pyruvate to prevent the cyto-
plasmic redox balance being perturbed. (This is because lactate, pyruvate,
NAD+ and NADH are kept close to equilibrium by the action of lactate dehy-
drogenase; see Fig. 6.17.) The build–up of glucose was prevented because it
was carried away in the buffer stream, and its steady state concentration was
less than the normal level in rat blood. The hepatocytes were contained in a
closed, stirred container and were prevented by a filter from escaping in the
outflow, though samples could be removed for analysis of the internal metabo-
lites. In the case of mitochondrial metabolites, this also involved rapid cell
fractionation by the digitonin method (Chapter 2.3.3, p. 35). The perifusate
was collected and analysed for lactate, pyruvate and glucose. In this way, it
174 6 Measuring control coefficients

was possible to tell when a steady state had been reached because the glucose
concentration in the perifusate would become constant (after about 20 – 30
min). Some change in the conditions, such as substrate concentration, could
then be made and the cells allowed to reach a new steady state, so that as
many as 6 experiments could be performed on a single sample of cells. The
concentration of the glucose in the perifusate also indicated the gluconeogenic
flux.
Experiments were carried out in the presence of the hormone glucagon,
which stimulates the liver to form glucose, and in its absence. One of the
major differences between these two sets of conditions is that, in the presence
of glucagon, the return loop from phosphoenol pyruvate, catalyzed by the
glycolytic enzyme pyruvate kinase, is suppressed by inactivation of the enzyme
by phosphorylation (Chapter 7.4.3.1). In the absence of the hormone, some
of the phosphoenol pyruvate is recycled to pyruvate.
The complete pathway from lactate to glucose involves about 15 steps
(excluding transport of lactate and glucose into and out of the cell). To mea-
sure the elasticities of all these steps would have required measurement of all
the metabolites that affected them, and this would have been an enormous
amount of work. In addition, not all the metabolic intermediates are easily
measurable. Groen and his colleagues therefore reduced the size of the prob-
lem by grouping some of the reactions together and treating them as a unit.
This means that it is not necessary to know the metabolite concentrations
inside the group. The control coefficient of a group must be the sum of the
control coefficients of the component enzymes (for the summation theorem to
remain valid). It is also possible to imagine that the group has elasticities
to its overall substrate and overall product.69 The relationship of these elas-
ticities to the component elasticities of the individual steps is more complex,
though it can be calculated. However, if the group elasticities are going to be
measured experimentally, this does not matter. The gluconeogenesis pathway
was therefore simplified to a set of 7 or 8 groups of steps between intracel-
lular (cytoplasmic) pyruvate and external glucose (Fig. 6.17). Seven steps
are involved when glucagon is present because it suppresses the flux through
pyruvate kinase. The interconversion of lactate to pyruvate was discounted
since the belief is that this enzymic reaction is always close to equilibrium.
The 8 groups interconvert 6 internal metabolites, which are mitochondrial
pyruvate and oxaloacetate, cytoplasmic oxaloacetate, phosphoenol pyruvate,
glyceraldehyde–3–phosphate and fructose–6–phosphate. These metabolites
were measured either directly or, as in the case of the oxaloacetate concen-
trations, indirectly from the malate concentrations on the assumption the
malate dehydrogenase reaction is near equilibrium. The other metabolites es-
timated indirectly were glyceraldehyde–3–P (GAP) from dihydroxyacetone–
P (DHAP) measurements, assuming the triose phosphate isomerase reaction
was close to equilibrium, and fructose–6–P (F6P) from glucose–6–P (G6P),
assuming the glucose–6–phosphate isomerase reaction was near equilibrium.
By a variety of means, some of which will be described in the following
sections, the elasticities of the groups were determined from these metabolite
6.3 Control coefficients from elasticities 175

measurements. This information was used to set up 7 (or 8) simultaneous

equations involving the 7 (or 8) flux control coefficients (depending on whether
pyruvate kinase was active or not). These equations were:

• The summation theorem;

• 6 connectivity theorem equations for each of the internal metabolites,

and

• in the absence of glucagon, a branch–point equation linking the control

coefficients of the groups between phosphoenol pyruvate and pyruvate
and between phosphoenol pyruvate and glucose with the relative fluxes
in these two branches.

Their results for 5 mM lactate are summarized in Table 6.4; the results at 0.5
mM and 1 mM lactate gave a similar pattern but are not shown.

tabsize

Table 6.4: Flux control coefficients in gluconeogenesis from lactate.

Flux control coefficient

Step + glucagon − glucagon

1 pyruvate transport 0.01 0.00

2 pyruvate carboxylase 0.83 0.51
3 oxaloacetate transport 0.04 0.02
4 PEPCK 0.08 0.05
5 pyruvate kinase 0.00 -0.17
6 enolase — PGK 0.00 0.29
7 TIM — FBPase 0.03 0.27
8 PGI + G6Pase 0.00 0.02

The results are those reported by Groen et al.91 for rat hepatocytes incubated with
5 mM lactate and 0.5 mM pyruvate, in the presence and absence of the hormone
glucagon. The enzyme groups correspond to those shown in Fig. 6.17, where the
full enzyme names are given. The flux control coefficients inevitably add up to
1, allowing for rounding error, because the summation theorem was one of the
equations used in the calculation.

Their conclusions were that, in the presence of glucagon, the largest flux
control coefficient was that of pyruvate carboxylase (0.83 – 0.89); no other
enzyme (including phosphoenol pyruvate carboxykinase) had a control coef-
ficient above 0.1. In the absence of glucagon, the largest control coefficient
was still that of pyruvate carboxylase (0.51 in the presence of 5 mM lactate),
but there were a number of other significant values: pyruvate kinase, -0.17;
176 6 Measuring control coefficients

the enolase to phosphoglycerate kinase group, 0.29, and the triose phosphate
isomerase to fructose bisphosphatase group, 0.27. PEPCK showed very small
control coefficients under all conditions. Furthermore, independent validation
of the calculations is provided by the re–analysis of the inhibitor titrations of
PEPCK, quoted earlier in section 6.1.4, that suggested a flux control coeffi-
cient of at most 0.1. Also, the response of the pathway to pyruvate was used
as described in the previous section to calculate that the flux control coeffi-
cient of the block of reactions (groups 1–4) converting pyruvate to PEP was
0.65 in the absence of glucagon. This is in close agreement with the figures
in Table 6.4.
Now that the potential value of determining control coefficients from elas-
ticities has been illustrated, I will describe some of the methods used both by
Groen and his colleagues and other research groups to obtain elasticity values
experimentally.

6.3.2 Experimental measurement in vivo by modulation.

A method for measuring elasticities in vivo, called the double modulation
method, was first suggested by Kacser & Burns120 a few years after they first
advocated measuring flux control coefficients. Let me illustrate their method
by showing how it might be applied to the enolase (eno) reaction in glycolysis:
Scheme 6.1
eno
· · · −→ 2PG −→ PEP −→ · · ·

The rate of enolase depends solely on the two glycolytic metabolites 2PG
(2-phosphoglycerate) and PEP (phosphoenol pyruvate). Mathematically we
state this as veno = f (2P G, P EP ), where f (. . .) means is a function of and
is used to indicate that, at this point, either we do not want to, or we cannot,
be more specific about what this relationship is. At steady state, veno = J,
where J is the glycolytic flux. In a control experiment, the flux Jc and the
concentrations 2P Gc and P EPc have to be measured, whilst in a parallel ex-
periment, the rate of glycolysis is perturbed in some way by a small amount,
for example by altering the input glucose level, and the new levels of flux, J e1
and concentrations, 2P Ge1 and P EPe1 measured at steady state. Mathemat-
ical theory for approximating the change in rate (∆J1 = Je1 − Jc ) in terms of
the changes in the metabolite concentrations gives the result:

∂veno ∂veno
∆J1 ≈ ∆2P G1 + ∆P EP1 (6.8)
∂2P G ∂P EP
Scaling this equation, by dividing throughout by Jc = veno and substituting
the definitions of the elasticities (Eqn. [5.12]) leads to:

∆J1 ∆2P G1 ∆P EP1

≈ εeno
2P G + εeno
P EP (6.9)
Jc 2P Gc P EPc
6.3 Control coefficients from elasticities 177

1 2 3 4
3PG ................................... 2PG ............................... PEP ..................................... pyruvate ................................. lactate
... . ..... ... ....
+... .... .... ... ...
... .. .. ... .
..
. ...
..
.... .. ... .
. . . NADH NAD+
2,3BPG ADP
.... ATP
.
..... .......
...... 5 ......
........................
.
......... ......
.....
... ....
.
... ..
Glc6P glucose

Fig. 6.18: A three enzyme in vitro segment of glycolysis.

This was the system used by Giersch85 to measure flux control coefficients from elas-
ticities and to compare them with the values obtained by enzyme titration. The three
enzymes are: 1. phosphoglycerate mutase; 2. enolase, and 3. pyruvate kinase. In
addition, 4. lactate dehydrogenase and 5. hexokinase are present in excess. Lactate de-
hydrogenase enables the flux to be continuously monitored by measuring the absorbance
at 340 nm by NADH. Hexokinase regenerates ADP. The variable metabolites of the sys-
tem are the ones not boxed, i.e. 2–phosphoglycerate (2PG) and phosphoenol pyruvate
(PEP).

If a second independent change can be made in the glycolytic flux, say by

inhibition of an enzyme downstream from PEP, leading to another equation:

∆J2 ∆2P G2 ∆P EP2

≈ εeno
2P G + εeno
P EP (6.10)
Jc 2P Gc P EPc

then the pair of equations can in principle be solved for the two unknown elas-
ticities. It is important for the accuracy of the method that the two imposed
changes result in mathematically independent equations (i.e. ∆2P G1 /∆P EP1 6=
∆2P G2 /∆P EP2 ) otherwise the result will be dominated by the experimental
error. Theoretical analyses by Rankin Small228 and Christoph Giersch84 have
suggested that the only viable experimental strategy is to make one change
upstream of the enzyme under investigation, and one change downstream.
This was the form of the double modulation method Groen et al.91 used in
the experiment described in the previous section on the control of gluconeo-
genesis in hepatocytes. They determined the elasticities of the transport of
oxaloacetate by the malate/aspartate shuttle between mitochondria and cyto-
plasm (see Fig. 6.17) by changing the amount of the pathway substrate lactate
and by inhibiting PEPCK with mercaptopicolinate as the two perturbations.
The substrate and product concentrations, mitochondrial and cytoplasmic
oxaloacetate, had to be estimated and the flux from lactate measured.
Although the double modulation equation (Eqn. [6.9]) has often been pre-
sented as shown above, the approximation formula used is not the best. Since
the elasticity is given directly by the slope of a graph of the logarithm of the
178 6 Measuring control coefficients

rate against the logarithm of the metabolite concentration, a better approxi-

mation is obtained starting from ln veno = f (ln 2P G, ln P EP ), to give:

∆ ln J1 ≈ εeno eno
2P G ∆ ln 2P G1 + εP EP ∆ ln P EP1 (6.11)

where ∆ ln J1 = ln Je1 −ln Jc . Furthermore, we can turn this into the basis of a
simple graphical analysis of the results of a modulation experiment. Dividing
through by ∆ ln J1 gives:

∆ ln 2P G1 ∆ ln P EP1
1 ≈ εeno
2P G + εeno
P EP
∆ ln J1 ∆ ln J1
In the limit as the change ∆ ln J1 tends to zero, the approximation becomes
the equality:
∂ ln 2P G ∂ ln P EP
1 = εeno
2P G + εeno
P EP (6.12)
∂ ln J ∂ ln J
The reason that this is particularly useful is that ∂ ln 2P G/∂ ln J is the slope
of a graph of ln 2P G against ln J, which can be obtained by modulating the
system by a series of varying perturbations and measuring 2P G and J each
time (e.g. Fig. 6.19). In the same experiments, if P EP is measured, then
the graph of ln P EP against ln J can also be plotted. The advantage of the
graphical analysis is that, by using a number of different sized modulations
and drawing a line or smooth curve through the results, the slopes can be more
precisely determined than by taking differences between a control and a single
modulated point. Thus, with an experiment of this type, the two elasticities
become the only unknown terms in Eqn. [6.12] above. Repeating the exper-
iment by applying a second type of modulation to varying degrees will give
two equations in the two unknowns. This analysis was originally proposed by
Rankin Small228, 230 and can be illustrated by applying it to some modulation
experiments performed by Christoph Giersch in Darmstadt.85 By perturbing
the in vitro model of a short section of glycolysis shown in Fig. 6.18, he ob-
tained the results plotted in Fig. 6.19. From the four slopes, the elasticities
of enolase to 2–phosphoglycerate and phosphoenol pyruvate can be calculated
as εeno eno
2P G = 2.58 and εP EP = -2.32.

Supplement: A complete analysis of Giersch’s modulation experiments.

Giersch’s experiment aimed to determine the flux control coefficients of an in vitro
segment of the glycolytic pathway between 3–phosphoglycerate and pyruvate, involv-
ing the three enzymes phosphoglyceromutase (PGM), enolase (eno) and pyruvate
kinase (PK). The system used is shown in Fig. 6.18. The experiment had two aims:
1. To show that flux control coefficients could be obtained from elasticities mea-
sured by modulation experiments, using a generalized version of the double
modulation method devised by Giersch,84 and
2. To validate the method by comparing the flux control coefficients obtained in
this way with the flux control coefficients obtained in the same experimental
system directly by the enzyme titration method.
6.3 Control coefficients from elasticities 179

2.1 2.0
a) b)
1.9
Log [PEP] or Log [2PG]

Log [PEP] or Log [2PG]

2.0

1.8
1.9
1.7

1.8
1.6

1.7 1.5
1.9 2.0 2.1 2.2 2.00 2.05 2.10 2.15

Log J Log J

Fig. 6.19: Modulation studies of an in vitro three enzyme segment of glycolysis.

The system under study is shown in Fig. 6.18. ¦, 2–phosphoglycerate (2PG); , phos-
phoenol pyruvate (PEP). a) Modulation of the third enzyme, pyruvate kinase, by vary-
ing the concentration of its second substrate ADP. The gradient of the regression line
through the 2PG points is -0.182, and that through the PEP points is -0.632. b) Modu-
lation of the first enzyme, phosphoglycerate mutase, by varying the concentration of the
obligatory cofactor 2,3 bisphosphoglycerate. The gradient of the regression line through
the 2PG points is 2.59 and that through the PEP points is 2.44. The results have been
replotted from experiments reported by Giersch.85

Giersch’s analysis of the modulation results is more flexible and extensive than the
simple double modulation method, and I am grateful to him for providing me with
his results and for giving me permission to analyse them in a different way. The
first enzyme, phosphoglycerate mutase, was modulated by varying the concentration
of the obligatory cofactor 2,3 bisphosphoglycerate (Fig. 6.19a). For the second
modulation, the third enzyme, pyruvate kinase, was subjected to variations in the
concentration of its second substrate ADP (Fig. 6.19b). The four lines on the figure
give the coefficients in the following two equations:

1 = −0.182εeno eno
2P G − 0.632εP EP

1 = 2.59εeno eno
2P G + 2.44εP EP

From these, we calculate that εeno eno

2P G = 2.58 and εP EP = -2.32.
To complete the analysis of Giersch’s experiment, we need to consider a simpli-
fication of the modulation equation, Eqn. [6.11]. If one term in eqn. 6.9 is known
or insignificant, only a single perturbation type is needed. Let us therefore term
this variant the single modulation method. It could apply either because one of the
elasticities is zero, or because one of the metabolite concentrations does not change.
For example, at the start of a pathway, the first enzyme has a pool metabolite as
its substrate, and the experiment will be arranged so either this concentration does
not vary significantly (so that the metabolite modulation term is zero), or the first
enzyme is saturated with this source metabolite (equivalent to an elasticity of zero).
A similar situation arises with the final step of the pathway, where the term in-
180 6 Measuring control coefficients

volving the sink metabolite should be zero. Both of these cases apply in the three
enzyme pathway studied by Giersch. Taking the experiment shown in Fig. 6.19a),
the enzyme modulated to alter the flux was pyruvate kinase. However, flux changed
at all three enzymes. In the case of enolase, as we have seen, the change in flux was
related to the changes in the concentrations of its substrate and product. In the
case of phosphoglycerate mutase, however, only the concentration of its product,
2PG, changed, so the double modulation equation reduces to:

∂ ln 2P G
1 = εP GM
2P G (6.13)
∂ ln J
where ∂ ln 2P G/∂ ln J is the slope of the 2PG line in Fig. 6.19a), which is -0.182. It
follows that the elasticity, εP GM
2P G = -5.49. Similarly, in Fig. 6.19b), where the first
enzyme PGM is modulated by varying the concentration of its cofactor, the flux
through pyruvate kinase varies solely because of the change in the concentration
of phosphenol pyruvate. The slope of the PEP line in this figure, 2.44, is 1/εP K
P EP ,
PK
giving εP EP = 0.411.
The analysis of these two modulation experiments has, in this case, given all
the information that is needed to determine the flux control coefficients of the three
enzyme system. There are three equations involving the flux control coefficients: a
summation theorem and two connectivity theorem equations with respect to 2PG
and PEP respectively:

CPJ GM + Ceno
J
+ CPJ K = 1
CPJ GM εP GM J eno
2P G + Ceno ε2P G = 0
J eno J PK
Ceno εP EP + CP K εP EP = 0

Subsituting the values obtained above for the four elasticities allows calculation of
the three flux control coefficients as CPJ GM = 0.07, Ceno
J
= 0.14 and CPJ K = 0.79.

Some additional examples of the use of modulation methods for measuring

elasticities are shown in Table 6.5. This is not comprehensive, because we shall
see in the next section that modulation methods have been incorporated into
an experimental protocol for measurement of flux control coefficients.

Supplement: Advanced modulation methods. If the activity of an en-

zyme is affected by three of the variable metabolites (for example, if the enzyme is
subject to feedback inhibition), then three independent perturbations are needed,
and this may be even more difficult to arrange. Giersch’s generalized modulation
method84 should be consulted for the method of calculation. His method could also
potentially find ‘unknown’ elasticities, that is, it could detect interactions between
metabolites and enzymes that had previously not been expected. Another multi-
ple modulation method that allows calculation of the elasticities in a system, again
detecting them for each metabolite on every enzyme, is the co–response coefficient
analysis of Cornish–Bowden & Hofmeyr,53 which extends and generalizes the use of
the slope measurements on log–log graphs of metabolite concentrations and fluxes
described above.
6.3 Control coefficients from elasticities 181

Table 6.5: Additional examples of the use of modulation methods for elasticity
measurements in Control Analysis.

Enzyme Pathway Modulation Ref.

Type Perturbation
pyruvate carboxylase gluconeogenesis single inhibition of PEPCK 91
pyruvate kinase gluconeogenesis single lactate/pyruvate ratio 91
ornithine carbamoyl– citrulline single ATP level (via TCA 265
transferase synthesis cycle inhibition)
carbamoyl–phosphate citrulline single inhibition of ornithine 265
synthase synthesis carbamoylt–ase
fructose bisphosphatase photosynthesis double light input
(spinach) sucrose concentration 242
sucrose–phosphate photosynthesis double light input
synthase (spinach) sucrose concentration 166, 242
fructose bisphosphatase photosynthesis double light input
(Clarkia) GPI enzyme level 165, 242
sucrose–phosphate photosynthesis double light input
synthase (Clarkia) GPI enzyme level 165, 242

These examples do not include the similar ‘top–down’ type of experiments described
in the next section. The examples above fall into two groups: those carried out on
hepatocyte metabolism by members of the Amsterdam group, and the photosyn-
thesis experiments by Mark Stitt and his coworkers.

6.3.3 Experimental measurement in vivo: the ‘top–down’ ap-

proach.
A problem about the use of elasticity determination for evaluating flux con-
trol coefficients is that the amount of experimental information needed for a
pathway appears daunting. In the gluconeogenesis example presented earlier,
it was necessary to reduce the number of elasticities and flux control coeffi-
cients under consideration by forming groups of enzymes and transport steps,
and even then it remained a large and complex experiment. If a pathway is
newly under investigation, how can it be decided how many steps can feasi-
bly be studied? Then if the conclusion implies the need to form groups of
enzymes, what groups should be formed? There are no general answers to
these questions; each case has to be decided on its merits, taking into ac-
count the relative difficulty of measuring different metabolite concentrations.
This has almost certainly deterred researchers who have been interested in
using Metabolic Control Analysis but have had difficulties devising a feasible
approach.
In Cambridge (England), a group of researchers consisting of Martin Brand,
Guy Brown, Patti Quant and their coworkers suggested that the solution to
182 6 Measuring control coefficients

L
µ
¡
¡4
1 2 3 ¡
succinate - - - ∆µ̃H+
C @ 5
@
R
@
ATPm
@ 6
@
R
@
ATPc
@7
P @
R
Glc6P

Fig. 6.20: Top–down analysis of mitochondrial oxidative phosphorylation

The steps were grouped by Hafner et al.94 into the three blocks outlined by dotted lines:
C, the respiratory chain; L, the proton leak, and P, the phosphorylation system. ∆µ̃ H+
is the protonmotive force. The numbered steps correspond to the components studied
by Groen et al. as described in Fig. 6.13.

these problems was to look at them from the other end. Instead of starting
with the components of a pathway and seeking how to group these into a
small number of components (a bottom–up approach), they proposed taking
the pathway and choosing a point at which it could be broken into two or three
components. They called this method the top–down approach to emphasize
that it worked in the opposite direction. For example, a linear pathway might
be divided into two about a single metabolite that is the product of one block
and the substrate of another, i.e. :
Scheme 6.2
Block 1 Block 2
X0 −→ S1 −→ X1

As far as the theory of Metabolic Control Analysis is concerned, this does not
raise any difficulties. We have already seen that it is legitimate to have flux
control coefficients of groups of reactions that must, by reason of the summa-
tion theorem, be formed of the sums of the flux control coefficients of their
component steps. There is also no difficulty about applying the definition
of the elasticity to a group of reactions rather than to a single one,69, 116, 213
though we would almost certainly have to measure its value experimentally,
since we cannot state general relationships between the kinetics of the com-
ponents and the elasticity in the same way that we can relate the kinetics of
an individal enzyme to its elasticity. In addition, it is necessary that the two
blocks are independent of one another, that is, no metabolite from one block
is a substrate or effector of any of the steps in the other block apart from the
single intermediate metabolite.
6.3 Control coefficients from elasticities 183

One of the advantages of the top–down approach is that it simplifies the

experimental procedure enormously. If we wanted the elasticities of the two
blocks in Scheme 6.2, then only two single modulation experiments are re-
quired. Thus, to determine the elasticity of Block 1 to S1 , it would be neces-
sary to measure the flux in the system and the level of S1 whilst altering the
activity of Block 2 (for example by using an inhibitor of one of its enzymes)
and keeping everything else constant, particularly X0 . The slope of a plot of
ln J against ln S1 is the elasticity εSBlock1
1
, by reason of the definition of the
elasticity and also by analogy with Eqn. [6.13]. The next step would be to
determine the elasticity of Block 2 with respect to S1 by modulating Block 1
(perhaps by varying the concentration of the source X0 ) and measuring the
flux and S1 again. With the two elasticities, we can find the flux control
coefficients of the two blocks by solving the pair of equations formed from the
summation theorem and the connectivity theorem with respect to S1 .
Would this help? Let us imagine the possible results of the experiment.
Either both blocks could have significant flux control coefficients, showing
that the control is distributed, or one could be near 1, when the other would
necessarily be near zero, showing that most of the control is somewhere inside
one of the blocks. In both cases, the next step is to choose a different central
metabolite, i.e. to form the blocks in a different way. In the first case, the
results of the flux control coefficients for these two different blocks could be
compared with the original set of results to allow deduction of the control
coefficients for three blocks: the beginning and end blocks, and the block
between the two chosen metabolites. In the second case, it might be found
that the second experiment narrows down even further the section of the
pathway where most of the control resides.

6.3.3.1 Experimental example: mitochondrial oxidative phosphorylation

Brand, Brown and coworkers applied this top–down methodology to the con-
trol of mitochondrial respiration and oxidative phosphorylation.21, 23, 94, 163
The system they studied was essentially similar to the one used by Groen et
al. for the inhibitor titration studies discussed earlier in Section 6.1.4 in which
isolated rat liver mitochondria oxidized succinate and drove an ATP-utilizing
system that converted the ATP formed in phosphorylation back to ADP, thus
allowing the system to attain a steady state. They chose the protonmotive
force as the central metabolite, which divides the system into three blocks, as
shown in Fig. 6.20. It might seem odd to regard the protonmotive force as
a metabolite, but it is the intermediate that is formed by respiration and is
used by the proton leak and phosphorylation. Furthermore, the protonmotive
force is logarithmically related to the proton concentration in the matrix, and
since logarithms of concentration usually appear in elasticities, the protonmo-
tive force itself can be used to define the elasticity. It is, of course, necessary
to measure the protonmotive force. The fluxes were measured in terms of
the oxygen consumption, measured for the respiration branch with an oxygen
electrode. The distribution of the flux at the branch point had also to be
184 6 Measuring control coefficients

1.0 1.0
a) b)
Flux control coefficient, CCJ

Flux control coefficient, CPJ

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0

0 20 40 60 80 100 0 20 40 60 80 100

Respiration rate, % Respiration rate, %

1.0
c)
Flux control coefficient, CLJ

0.8

0.6

0.4

0.2

0.0

0 20 40 60 80 100

Respiration rate, %

Fig. 6.21: Flux control coefficients for mitochondrial respiration rate as a function
of respiration rate.
The results of Hafner et al.94 are compared with the corresponding results of Groen
et al.92 a) The flux control coefficient of the respiratory chain (block C, Fig. 6.20).
Points from Hafner et al. are shown with the curve of Groen et al. for the dicarboxylate
carrier only (step 1 in block C, Fig. 6.20) taken from Fig. 6.15. b) The flux control
coefficient of the phosphorylation block (block P, Fig. 6.20). Again, the experimental
points are shown with the sum of the two curves through the the coefficients for the
adenine nucleotide translocator and hexokinase (steps 6 and 7, block P) from Fig. 6.15.
c) The flux control coefficient of the leak (block L). The experimental points are shown
with the corresponding curve from Fig. 6.15.
6.3 Control coefficients from elasticities 185

measured, i.e. the relative contributions of the leak and the phosphorylating
system to the consumption of the protonmotive force had to be determined. A
more detailed explanation of how this was done is given in the supplementary
material.
The elasticity of any one block with respect to the protonmotive force
was determined by inhibiting or stimulating one of the other blocks in order
to change the value of the protonmotive force. Thus if the elasticity of the
phosphorylating system was required, it was obtained from the slope of the
phosphorylation-coupled component of the respiration rate as the respiratory
chain block was titrated with the inhibitor malonate. Again this is covered in
more detail in the supplementary section.

Supplement: Experimental details The protonmotive force, ∆µ̃H + has 2

components: a membrane potential, ψ, across the inner mitochondrial membrane
and a pH difference, ∆pH. In these experiments, the membrane potential was
measured from the distribution of a lipophilic cation, methyltriphenylphosphonium
cation (Ph3 MeP+ ). A known amount was added to the mitochondrial suspension,
and, because it is lipophilic, it equilibrated rapidly across the membrane with the
equilibrium values of the inner and outer concentration dependent upon the mem-
brane potential as described by the Nernst equation. The external concentration
was continuously monitored by an electrode specific for the Ph3 MeP+ ion, which
allowed the matrix concentration, and hence the membrane potential, to be cal-
culated. The pH difference could be measured by its effect on the distribution of
the weak, permeant acid, [3 H] ethanoic acid between the mitochondrial matrix and
the exterior, but in a number of tests where the distribution of the radiaoctivity
from the 3 H was measured, it was found to be small and invariant relative to the
membrane potential component.
There are three fluxes in the system: the formation of the protonmotive force by
the respiratory chain, JC ; the dissipation of the protonmotive force by the leak of
protons across the inner mitochondrial membrane, JL , and the consumption of the
protonmotive force by the ATP–synthesizing branch, JP . These were all expressed
in terms of the rate of oxygen consumption for three reasons:
1. it was necessary for the analysis that all three fluxes were expressed on the
same scale, since then JC = JP + JL ;
2. the rate of oxygen consumption could be continuously monitored with an
oxygen electrode, which was non–invasive and made it easier to ensure that
flux measurements were made in a steady state, and
3. it avoids the uncertainties that would arise in converting the oxygen fluxes into
proton or phosphorylation fluxes because of uncertainties in the stoichiometric
ratios between these.
The experiments were performed on a series of steady states involving different
rates of respiration, between the resting rate in the absence of phosphorylation and
the maximal possible in a phosphorylating system (100% level), which were obtained
by using different amounts of hexokinase. At each steady state, the respiratory chain
flux JC and the protonmotive force ∆µ̃H + were measured as the reference point,
and a series of perturbations were made to measure the other variables.
186 6 Measuring control coefficients

Firstly, the phosphorylation and leak fluxes were measured as follows. In a

sample of the same mitochondria, phosphorylation was suppressed with the inhibitor
oligomycin, i.e. JP = 0, JC = JL . This resulted in a lower respiration rate but
higher protonmotive force than in the reference state. The protonmotive force and
the respiration rate were then successively reduced in a titration with the respiratory
chain inhibitor malonate (a competitive inhibitor of succinate dehydrogenase). A
graph of these titration results was plotted, and the respiration rate that gave the
same protonmotive force as in the reference state was read from the graph. This
was taken to be JL in the reference state, and the phosphorylation flux was found
by difference JP = JC − JL .
This same graph was also used to determine the elasticity of the leak block with
respect to the protonmotive force. The slope of the graph is dJL /d∆µ̃H + , and this
is scaled by the values of the rate and protonmotive force at the reference state to
give:
dJL ∆µ̃H +
εL∆µ̃H + =
d∆µ̃H + JL
A similar malonate titration experiment, except without oligomycin so that
phosphorylation was active, was carried out and the respiratory chain flux plot-
ted. The difference between the two graphs was taken as the phosphorylation flux
JP and this was in turn plotted against the protonmotive force. The gradient of this
graph gave the elasticity of the phosphorylation block to the protonmotive force,
i.e. :
dJP ∆µ̃H +
εP
∆µ̃H + =
d∆µ̃H + JP
Finally, the elasticity of the respiratory chain block was measured by modulating
one of the two branches and measuring JC and the protonmotive force. Two methods
were used: gradual increase of the leak flux by titration with an uncoupler, or
increase of the phosphorylation flux by titration with hexokinase. As in the previous
cases, the initial part of the titration gave a linear slope, and the gradient gave the
elasticity:
dJC ∆µ̃H +
εC
∆µ̃H + =
d∆µ̃H + JC

The results were calculated by the simultaneous solution of three equa-

tions. For example, the control coefficients on the respiratory chain flux came
from solving the summation equation:
JC
CC + CPJC + CLJC = 1; (6.14)

the connectivity equation:

JC C
CC ε∆µ̃H + + CPJC εP JC L
∆µ̃H + + CL ε∆µ̃H + = 0 (6.15)

and a branch point equation worked out according to a set of rules devised
by Herbert Sauro and me69

CPJC C JC
− L = 0. (6.16)
JP JL
6.3 Control coefficients from elasticities 187

1.0 1.0
a) b)
Flux control coefficient on JC

Flux control coefficient on JP

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0

0 20 40 60 80 100 0 20 40 60 80 100

Respiration rate, % Respiration rate, %

1.0
c)
Flux control coefficient on JL

0.5

0.0

-0.5

-1.0
0 20 40 60 80 100

Respiration rate, %

Fig. 6.22: Comparison of the distribution of control over the respiration, phos-
phorylation and leak fluxes in mitochondrial oxidative phosphorylation.
The results were derived by top–down control analysis of rat liver mitochondria. 94 a)
Control of the respiration flux, JC . Points from a single experiment from Fig. 6.21
summarized on one graph. b) Control of the phosphorylation flux, J P . One of the sets
of results from Fig. 6.21 recalculated. c) Control of the leak flux, J L , again recalculated
from results in Fig. 6.21. : flux control coefficients for the respiratory chain block. :
flux control coefficients for the phosphorylation block. 4: flux control coefficients for
the leak block. This figure is based on one by Brand. 20
188 6 Measuring control coefficients

The results obtained in this way by Hafner, Brown & Brand94 are shown
in Fig. 6.21. The range of the control coefficients on the respiration flux
between the minimum and maximum respiration rates was approximately:
CLJC = 0.9 – 0.0; CPJC = 0.0 – 0.5, and CC JC
= 0.1 – 0.5. The profiles between
these points resembles the results Groen et al.92 had obtained by inhibitor
titrations, although to see this, it is necessary to group these latter results
into the appropriate blocks. In order to do this, I used the smooth curves
drawn on the graph of their results (Fig. 6.15) and superimposed them on
Fig. 6.21. Under the circumstances, the degree of agreement between the
results of the Cambridge and Amsterdam groups is remarkably good, and it
can be taken as providing some independent validation of both experimental
methods: the inhibitor titration method, and the top–down control analysis
via block elasticities.
Once again, the results demonstrate that the distribution of the control
of flux varies according to the metabolic steady state. There is also another
feature of control that can be illustrated by further analysis of this system.
Because there are three fluxes in this branched system, we can ask whether
the distribution of control is the same for all three; we shall see that it is not.
Suppose that we wish to calculate the distribution of control over the flux in
the phosphorylation branch, JP . The flux control coefficients will be related
by a summation theorem, exactly the same as Eqn. [6.14] above, except with
JP instead of JC . The connectivity theorem equation will likewise be the
same as Eqn. [6.15] above. The third equation, the branch point equation, is,
however, different. It is now:
JP
CC C JP
+ L = 0. (6.17)
JC JL

Just as Eqn. [6.16], which applied to the respiratory chain flux JC , did not
JC
contain the control coefficient CC , the new equation does not contain the
control coefficient with respect to the phosphorylation block, CPJP . The sign
has also changed, because the two fluxes now involved have opposite orienta-
tions, in that JC produces the protonmotive force and JL uses it, whereas in
the previous case, both the fluxes related to utilization of protonmotive force.
The solutions of the set of three equations are consequently different, as can
be seen in Fig. 6.22b). Over most of the range of respiratory rates, the major-
ity of control of the phosphorylation flux lies in its own block. This top–down
experiment cannot tell the distribution of this control between the different
steps within the block. However, the share of the phosphorylation control ex-
erted by the adenine nucleotide translocator and the ATP utilization system
should be the same as their share of respiratory control as observed by Groen
et al. in Fig. 6.15. (This ratio of control is not affected by the change in the
reference flux.)
When the flux control coefficients are calculated for the control over the
leak flux, JL , a different pattern is seen again. Once again, the same summa-
tion and connectivity theorem equations apply, but the branch point equation
6.3 Control coefficients from elasticities 189

is different and does not contain JL :

JL
CC C JL
+ P = 0. (6.18)
JC JP
The results are shown in Fig. 6.22c). At high respiratory flux rates, the
solution is remarkable for having two flux control coefficients near 1, and the
third near -1. Under all circumstances, the leak itself has a high level of control
over its flux, but the degree of control of the other two blocks increases with
increasing rate of respiration.

6.3.3.2 Experimental example: hepatocyte energy metabolism

0.15 - 0.3 0.85 - 0.7

1 glucose −→ NADH −→ −→ −→

0.22
0.29
%
2 glucose −→ −→ −→ ∆µ̃H+
&

0.49

0.22
0.15 - 0.3 0 - 0.14
%
1&2 glucose −→ NADH −→ ∆µ̃H+
&

0.49

Fig. 6.23: Two overlapping top–down analyses of cellular respiration.

The two experiments used NADH and ∆µ̃H+ as the two central intermediates and are
shown in the first two lines. At the end of the second and third lines, the upward
arrow is the dissipation of protonmotive force by the leak and the downward arrow is
phosphorylation and ATP utilization. The third line combines the two results to infer a
flux control coefficient for the respiratory chain.

The studies on the control of oxidative phosphorylation in isolated mito-

chodria have led to a considerable advance over previous attempts to identify
the rate–limiting step. Nevertheless, to what extent can these findings be
applied to mitochondria working in cells? After all, in the experiments, large
amounts of succinate are made available externally, whereas in vivo, succinate
would be generated in smaller concentrations in the matrix of the mitochon-
dria. The ATP utilizing reactions in the cell are also more various than the
single reaction (hexokinase) used in the experiments. The Cambridge group
went on to consider this by applying the top–down experimental approach to
190 6 Measuring control coefficients

the control of respiration in isolated hepatocytes23 supplied with glucose as an

oxidizable substrate. In one set of experiments, they took the mitochondrial
NAD+ /NADH pair as the central metabolite, dividing the system into an
NADH supply block (glycolysis and the Krebs’ cycle) and an NADH demand
block (the respiratory chain, phosphorylation, etc). To do this, they had to
measure the mitochondrial NADH by fluorescence; oxygen consumption flux
was once again measured with an oxygen electrode. The flux control coeffi-
JC
cient of the NADH producers, Csupply , was estimated to be in the range 0.15
JC
– 0.3, with that of the NADH users, Cdemand , 0.85 – 0.7.
In another set of experiments, they used the mitochondrial protonmotive
force as the central ‘metabolite’, effectively repeating the mitochondrial ex-
periments described above, except that this time they were using whole cells.
As before, metabolism was divided into three blocks: a block generating the
protonmotive force by oxidizing glucose and the leak and phosphorylation
blocks consuming it. They showed that the flux control coefficient of the
phosphorylation block was about 0.48, of the generating block was 0.29, and
of the leak block was 0.21.
From the overlaps between these two sets of experiments, a more detailed
picture of the control can be built up as shown in Fig. 6.23. The two important
conclusions are firstly that the pattern of control over the rate of respiration
is very similar to that observed in the studies of isolated mitochondria for
intermediate respiration rates (e.g. Fig. 6.22a)). In particular, the block con-
taining the phosphorylation of ADP to ATP and the consumption of ATP by
cellular processes is the one exerting the largest amount of control. Much less
control is shared between catabolism and the respiratory chain. (This con-
centration of control in the lower end of the pathway is highly significant, for
reasons that will be analyzed later in Chpater 7.1.) Secondly, Control Analy-
sis is shown to be applicable to the totality of cellular energy metabolism, and
with two applications of the top–down approach, the section of metabolism
having the largest influence on respiratory rate has been identified.

6.3.3.3 Fatty acid metabolism

Top–down analysis is also being used by Patti Quant and her colleagues in
Cambridge to study the control of fatty acid utilization and ketone body for-
mation.153, 187, 188 In the case of ketogenesis, the traditional criteria for iden-
tifying rate–limiting enzymes had led some groups to favour carnitine palmi-
toyltransferase I and others to support hydroxymethylglutaryl CoA synthase.
The measurements of control coefficients so far show that their values vary
with metabolic state, but neither of these enzymes are fully rate–limiting in
any of the experiments. (Some of their results are presented in Problem 5 at
the end of this Chapter.)
6.3 Control coefficients from elasticities 191

6.3.4 Calculation of elasticities.

6.3.4.1 Theory

Perhaps readers may be surprised that the calculation of elasticity values

from metabolite concentrations and enzyme kinetic information was not one
of the first options to be discussed as a means of deriving control coefficients
from elasticities. After all, in Chapter 5.3 I introduced elasticities in terms of
their relationships to known enzyme kinetic functions, and in Appendix 3 to
that Chapter, More about elasticities, p. 134, I gave more detail about how
elasticities could be calculated from any enzyme rate law. However, in that
Appendix, I also mentioned some of the difficulties that arise in this approach
because the enzyme kinetics experiments recorded in the biochemical litera-
ture were not made with this application in mind and therefore often do not
include all the relevant measurements. Even so, a day in the library conduct-
ing a literature search sometimes yields some appropriate information, though
the chances that the measurements were made on the cell type you want to
study are not high unless it is Saccharomyces cerevisiae, Escherichia coli or
rat liver cells.
If the kinetic rate law and its parameters are known, the next set of in-
formation required is the concentrations of all the metabolites that influence
the enzyme rate, measured in the metabolic steady state of interest. This
raises all the usual problems of interpretation (see Chapter 2.3, p. 30). Are
the measurements of cellular metabolite concentrations representative of the
actual concentrations experienced by the enzyme in the cell? That is, do
they reflect the free concentrations of the metabolites or are they affected by
significant amounts of bound metabolites? Are the measurements applicable
to the compartment of the cell where the enzyme is located? Perhaps this
list of problems seems formidable, but even so, I do not believe we should
give up. None of them is unique to Metabolic Control Analysis; they apply
equally to other forms of biochemical explanation, and the reductionist ethos
of much biochemistry (see Chapter 1.2.1, p. 5) would be undermined if there
were no confidence in the applicability of such results. Would it be sensible
to believe that incorporating the available information into a logical, quanti-
tative framework such as Metabolic Control Analysis would give less reliable
results than applying the same information in a set of semi–quantitative prin-
ciples (cf Chapter 4) that we have seen can only be partially justified and can
often be misleading? For example, qualitative theories of control will often
assess the potential importance of a regulatory interaction by comparison of
a metabolite concentration to its Km , Ki or Ka value for a particular enzyme
on the assumption that the enzyme will respond if the two are of comparable
magnitude. We have seen, in the chapter on enzyme kinetics (for example,
sections 3.2 and 3.3) that a single kinetic parameter cannot serve to define the
effect of a metabolite concentration on an enzyme rate without considering
the concentrations of other relevant metabolites and their associated parame-
ters. The elasticity, on the other hand, does reflect the interaction of all these
192 6 Measuring control coefficients

factors. Furthermore, we know that the elasticity alone is not enough, its
influence on the flux control coefficient or response coefficient of the enzyme
has to be considered.
In spite of the admitted problems in getting reliable information for elas-
ticity calculations, there is another reason why it may still be worthwhile
attempting. This is that the results of Metabolic Control Analysis are not
necessarily particularly sensitive to inaccuracies in the information used in
the calculations either of the elasticities or, from them, of the flux control
coefficients. One reason can be seen by looking at the diagrams of elastic-
ity values against metabolite concentration (Figs. 5.6–5.9): for much of the
range, the elasticities are not changing very rapidly relative to substrate con-
centration, so errors in the substrate concentration (or equivalently, in the
corresponding Km value) do not cause errors of corresponding size in the
elasticity. Two cases where this is not so turn out to be rather different from
one another. One case is for enzymes near to equilibrium, where the elasticity
grows very rapidly as the metabolite concentrations tend to their equilibrium
values. Inaccuracy here is of little significance because, as summarized in
Chapter 5.5.1, Kacser & Burns showed119 that the flux control coefficients of
such enzymes tend to zero. The other concerns small inhibition elasticities;
studies with my coworkers Rankin Small and Simon Thomas have shown that
these are much more important.
In our work,228, 232, 246 they devised procedures for assessing the influence
that individual elasticity values have on the calculated values of the flux con-
trol coefficients. Some of our conclusions have been that:
• the values of the control coefficients are not equally sensitive to the
values of all the elasticities;
• the dependence of the value of any control coefficient on the value of
an elasticity follows a hyperbolic or inverse hyperbolic response, i.e.
the value of the control coefficient changes relatively rapidly with the
elasticity when the elasticity is small (much less than 1), but tends to a
constant value as the elasticity increases;
• this applies particularly to elasticities of near equilibrium reactions, be-
cause the elasticities are not only large in magnitude, but appear in
substrate–product pairs whose contributions tend to cancel, leading to
the tendency to small control coefficients mentioned above, and
• small feedback and product inhibition elasticities will tend to be par-
ticularly important, in that small changes in their values will have the
largest effects on the control coefficients. (See also Section 6.2, p. 169
for evidence from computer modelling that even weak product inhibition
effects are important in metabolism.)
The general conclusion is that using an approximate value for many of the
elasticities will not make a significant difference to the calculated flux control
coefficients, and the validity of this assumption can always be tested by the
6.3 Control coefficients from elasticities 193

procedures devised by Small and Thomas. Simon Thomas’ computer pro-

gram Metacon automates the calculations.245, 246 Therefore, as previously
mentioned, the elasticity values of near–equilibrium enzymes can frequently
be calculated with sufficient accuracy using only the displacement from equi-
librium terms in Eqns. [5.13] and [5.14]. The elasticity values that will cause
problems are small ones, particularly those associated with weak inhibition
effects; not only is it necessary to have an accurate value for them, it is es-
sential to know that they exist at all, and here there is a weakness in the
biochemical literature, because weak inhibition effects have not been of much
interest, since it is only recently that it has become apparent how important
they are.

6.3.4.2 Applications
Keith Snell drew my attention to the evidence that the control of the path-
way for synthesis of the amino acid serine in mammalian liver might be
rather unusual. The pathway branches off from the glycolytic intermedi-
ate 3–phosphoglycerate and involves three enzymes, phosphoglycerate dehy-
drogenase, phosphoserine aminotransferase, and phosphoserine phosphatase
catalyzing the sequence:
Scheme 6.3

3–phosphoglycerate −→ phosphohydroxypyruvate

−→ phosphoserine −→ serine
The first step is an NAD–dependent oxidation, the second a transamination
with glutamate as donor, and the third the hydrolysis of a phosphate group.
The pathway can be regarded as starting at 3–phosphoglycerate because the
flux to serine is so small relative to the glycolytic flux that the serine path-
way will have little effect on the 3–phosphoglycerate levels. What appeared
unusual about this pathway was firstly that published measurements of the
metabolite levels in rabbit liver suggested that the first two steps were nearer
to equilibrium than the final step, and secondly that the most obvious regu-
latory interaction known from previous enzyme kinetic studies was inhibition
of the third step, phosphoserine phosphatase, by the end–product serine.
At first sight, there seemed to be enough information to calculate the
elasticities of the enzymes and thus the flux control coefficients, but as so
often, not all the information was quite complete. For example, phosphohy-
droxypyruvate concentrations in liver are too low to be readily measurable;
however, since 3–phosphoglycerate and phosphoserine were nearly in equilib-
rium (taking into account the NAD+ , NADH, glutamate and 2–oxoglutarate
levels), the first two reactions could be combined as one near–equilibrium re-
action and its elasticities to 3–phosphoglycerate and phosphoserine calculated
from the displacement from equilibrium. The kinetic information on the other
step was also not ideal, in that the measurements had been made on chicken
liver and rat liver, not rabbit liver. However, we reanalysed the measurements
194 6 Measuring control coefficients

and showed that both sets were explained best by the serine inhibition’s being
of the uncompetitive type (Fig. 6.24). Although there were differences in the
Km values and inhibition parameters between the rat and chicken enzymes,
these were not so great as to give grounds for believing that the rabbit en-
zyme would be very different. With the equation for uncompetitive inhibition
and the parameters of the rat liver enzyme, we calculated the elasticities of
phosphoserine phosphatase to its substrate phosphoserine and the pathway
end–product, serine.
Thus the problem we analysed had become a two step pathway with a sin-
gle intermediate, phosphoserine, for which we had the elasticities. Examples
of the calculation method have already been given, so it will not be shown
here; Q. 2 in the Problems section asks you to calculate the solution. We
found71, 235 that, in normal rabbit liver where there is a relatively high biosyn-
thetic flux, phosphoserine phosphatase has the largest flux control coefficient
(0.97), whereas the first two enzymes are close to equilibrium and have negli-
gible flux control coefficients. Phosphoserine phosphatase is inhibited by the
pathway product serine, giving a calculated response coefficient of −0.63 for
the effect of serine on the pathway flux. In contrast, the response coefficient
of the flux to the pathway source, 3–phosphoglycerate, is about a tenth of this
value, so the flux is largely determined by the serine concentration through

100

80
Velocity, % maximum

60

40

20

0
0.0 0.1 0.2 0.3 0.4 0.5

[phosphoserine], mM

Fig. 6.24: Inhibition of rat liver phosphoserine phosphatase by serine.

The inhibition is of the uncompetitive type. Note that the curves are close together
at low substrate concentrations and diverge at higher ones. The solid curve is in the
absence of serine; the dashed one is at 1 mM, and the dotted one, 6 mM. Levels of
serine in rabbit liver are typically between these values, and intracellular phosphoserine
is about 0.4 mM.
6.3 Control coefficients from elasticities 195

inhibition of the final step. This is an exception to the common pattern where
the first enzyme after a pathway branches off is usually the enzyme subject to
end–product inhibition. The strong control exerted by serine on the pathway
flux is dependent on the strange properties of uncompetitive inhibition. In
uncompetitive inhibition, the inhibitor has contradictory actions: it inhibits
by reducing the apparent limiting rate (like a non–competitive inhibitor), but
at the same time it reduces the Km of the enzyme for its substrate, which
would normally cause an activation. This has the result that the stimulatory
and inhibitory effects are almost in balance at low substrate concentrations,
but inhibition predominates as the substrate concentration increases. As can
be seen from the graph of phosphoserine phosphatase kinetics (Fig. 6.24), the
degree of inhibition is higher at high substrate concentrations than at low
ones. This is the exact opposite of competitive inhibition, but is well adapted
to feedback inhibition, since the general effect of inhibiting a pathway enzyme
is to increase the concentration of its substrate, which could potentially nullify
the action of a competitive inhibitor. In fact, at intracellular levels of serine
in rabbit liver, phosphoserine phosphatase is effectively almost saturated with
its substrate.

6.3.4.3 Hybrid examples

Several research groups have adopted a hybrid strategy towards the determi-
nation of elasticity values, in that they have calculated elasticity values for
steps with well-characterized kinetics or that are not likely to have a signifi-
cant influence on the values of the flux control coefficients, but have measured
more critical elasticities, or those for which there is inadequate information.
In fact, some of the studies mentioned previously have relied on calculation
methods for some of the elasticities, for example, the study of the control
of gluconeogenesis in hepatocytes by Groen et al.89, 91 described earlier in
Section 6.3.1.
Galazzo & Bailey77 have used NMR (nuclear magnetic resonance) to mea-
sure the concentrations of intermediates in the glycolytic pathway of yeast
(Saccharomyces cerevisiae). They then used these concentrations in the ki-
netic equations of the glycolytic enzymes to calculate their elasticities and
then their flux control coefficients. The results varied for the different incu-
bation conditions they had used, though the largest control coefficients were
shown by glucose entry into the cell, phosphofructokinase, and ATP consump-
tion processes. However, the conclusions depend strongly on two elasticities
that are not well–characterized: the feedback inhibition of glucose transport
by glucose–6–phosphate and the group elasticity of ATP–utilizing processes
with respect to ATP.
Salter et al. also used some calculated elasticities in their study of the
control of aromatic amino acid metabolism by hepatocytes207 (see Sections
6.1.2 and 6.1.4.3, and Table 6.3). In particular, they calculated the elasticities
of the amino acid transporters to their amino acids from the displacement from
equilibrium, and combined them with the response coefficients of the catabolic
196 6 Measuring control coefficients

fluxes to the external levels of the amino acids to calculate the flux control
coefficients of the transport steps. These varied from about 0.25 for transport
of tyrosine and tryptophan at basal catabolic levels to 0.93 for phenylalanine
transport when catabolism was induced.

6.4 Summary
1. There is now a large and growing body of experimental measurements of
flux control coefficients in a range of different pathways and organisms.
Although the experiments require a good degree of accuracy in order
to obtain reasonably reliable values for flux control coefficients, they
are based on familiar experimental techniques from genetics, molecular
biology and biochemistry.

2. Direct methods of measuring flux control coefficients involve measuring

the change in flux when the amount or activity of an enzyme is ma-
nipulated by some means. Indirect methods, including the top–down
approach, involve calculation of the control coefficients from measured
elasticity values (or from calculated elasticities based on experimental
measurements of metabolites and enzyme kinetic parameters).

3. Values of flux control coefficients vary depending on the prevailing con-

ditions.

4. There are only a few cases where a flux control coefficient is very close to
1.0, and in some of these, in extreme rather than physiologically normal
conditions. Thus the control of a pathway by a single, rate–limiting
enzyme has been experimentally proved not to be the norm.

5. As predicted by the theory, the control of flux is distributed, with more

than one step in a pathway having some measure of control. However,
most flux control coefficients are found to be relatively small or zero.

Further reading
1. Fell, D. A. Metabolic Control Analysis: a survey of theoretical and ex-
perimental developments. Biochem.J 286, 313–330 (1992)

2. Quant, P. A. Experimental application of top–down Control Analysis to

metabolic systems. Trends Biochem. Sci. 18, 26–30 (1993)

3. Salter, M., Knowles, R. G. & Pogson, C. I. Metabolic control in Essays

in Biochemistry, Tipton, K F (ed), Vol 28, pp. 1–12, Portland Press,
London (1994)
6.4 Summary 197

Computer programs
Several programs are available to carry out simulation and metabolic con-
trol analysis of metabolic pathways. At the time of writing, the following
programs for PC compatible microcomputers are available over the Internet
by ‘anonymous FTP’ from my laboratory at bmsdarwin.brookes.ac.uk in the
directory pub/software/ibmpc:

• SCAMP, a package for carrying out simulation, steady state analysis

and Metabolic Control Analysis.209, 211

• Gepasi, a similar package to SCAMP with an interface for the Windows

operating system.158

• Metamodel, a package for steady state and control analysis of simple

pathways.106

• Metacon, a package for determining the algebraic expressions relating

control coefficients to the elasticities of a network and calculating the
uncertainties in the control coefficients that arise from the errors in the
elasticity measurements.245, 246

Problems
1. Citrulline is synthesised in mitochondria by the reactions:

(a) NH+ −
4 + HCO3 + 2ATP −→ carbamoyl phosphate + 2ADP + Pi

(b) carbamoyl phosphate + ornithine −→ citrulline + Pi

Reaction 1a is catalyzed by carbamoyl phosphate synthase (CPS) and

reaction 1b by ornithine transcarbamoylase (OTC). If the elasticities of
these two enzymes with respect to carbamoyl phosphate (cp) are εCP cp
S
OT C
= -0.011 and εcp = 0.92, what are the relative values of their flux
J J
control coefficients CCP S and COT C ?

2. In the serine biosynthesis pathway:

1 2
3-phosphoglycerate −→ phosphoserine → serine

the elasticity of the first step, ε1pser , is -1.43 in the liver of rabbits on
a normal low protein diet. (The first step is actually catalyzed by two
enzymes, but the elasticity is the ’combined’ elasticity for them both,
so they can be treated as a single step.) The elasticity of the second
step, ε2pser , is 0.041. What are the flux control coefficients, C1J and C2J ,
of the two steps?
198 6 Measuring control coefficients

The response coefficients for the responses of flux to changes in the

pathway source, phosphoglycerate (3pg), and the product, serine (ser),
are given by the Kacser & Burns combined response equations:
J
R3pg = C1J .ε13pg : Rser
J
= C2J .ε2ser
where ε13pg = 2.43 and ε2ser =-0.65. (‘J’ signifies the pathway flux,)
What does this suggest to you about the regulation of this pathway?
3. Stitt et al. determined how the rate of photosynthetic sucrose synthesis
in spinach cytosol depended on two groups of enzymes. Group 1 consists
of aldolase and fructose bisphosphatase and converts triose-P (delivered
by the chloroplasts) into hexose-P. Group 2 contains phosphoglucose iso-
merase, phosphoglucomutase, UDPG synthase, sucrose-P synthase and
sucrose-P phosphatase and converts hexose-P to sucrose. Elasticities
for hexose-P on the two groups were determined as ε1hexose−P = -0.8
and ε2hexose−P = 1.0. What are the flux control coefficients of the two
groups, C1J and C2J , on the flux J to sucrose?
Group 1 responds to the supply of triose-P from the chloroplasts with an
elasticity estimated as ε1triose−P = 1.6. What is the response coefficient,
J
Rtriose−P , for the effect of triose-P on the flux to sucrose?
4. Brand et al. considered the control of respiration in non–phosphorylating
mitochondria (i.e. in the absence of ADP). They argued that the system
can be considered as one in which the respiratory chain (rc) produces
an intermediate (the proton motive force, ∆µ) which is consumed by
leakage back across the membrane. It is possible to regard ∆µ as an
intermediate and estimate elasticities for the two reactions rc and leak,
i.e.:
rc leak
oxidizable substrate −→ ∆µ −→ dissipation
The fluxes were measured in terms of the rate of oxygen consumption,
JO , accompanying substrate oxidation. The measured elasticities were
εrc leak
∆µ = –15.2 and ε∆µ = 7.9. What are the two flux control coefficients
JO JO
Crc and Cleak ?
5. In experiments on the top–down analysis of ketogenesis, the acetyl
CoA/CoA couple can be regarded as the central metabolite, produced
by β–oxidation and utilized by ketogenesis. Both steps can be simulta-
neously modulated by using pyruvate to perturb the acetyl CoA level,
giving the following results:

[pyruvate] [acetyl CoA]:[CoA] Jβ –oxidation Jketogenesis

mM nmol.min−1 .mg−1 nmol.min−1 .mg−1
0.0 1.25 21.1 18.9
2.5 2.23 15.8 19.2
5.0 4.75 5.3 23.9
6.4 Summary 199

Make two log–log plots of the fluxes against [acetyl CoA]:[CoA] and
measure the slopes to determine the elasticities. Using these elasticity
measurements, determine the flux control coefficients of the acetyl CoA
producing block and the consuming block. (This calculation method
will not give exactly the same answer as that published by Quant and
colleagues using a different analysis method.188 )
200 6 Measuring control coefficients
 7

Control structures in
metabolism

So far, we have considered the potential that an enzyme might have for con-
trolling a pathway without having given particular consideration to its place
in the metabolic network. However, there is further organization in metabol-
ism beyond the serial sequences of reactions that we call metabolic pathways.
In this chapter, we shall look at some of the features of these higher levels
of organization and their potential roles in control, both as conventionally
interpreted and in the light of Metabolic Control Analysis. Once again we
will find that there are aspects of the conventional views that do not stand
up to close analysis.

7.1 Supply and demand

First let us consider the most basic problem of metabolic regulation and con-
trol: how to regulate the concentration of a metabolite when controls are
exerted to alter the rate of its production and consumption. The simplest
structure that we can analyze in metabolism that exhibits this feature is the
synthesis of some metabolite M from a source metabolite Xsource by a supply
pathway and its use by a demand pathway:
Scheme 7.1
supply demand
Xsource −→ M −→ Xsink

This is just the same simplification that is used in the ‘top–down’ method
of Metabolic Control Analysis considered in Chapter 6.3.3, p. 181, and as in
that method, the supply and demand steps can each be a composite of many
individual steps without changing the essential nature of the problem. For
example M might be an amino acid, in which case the supply step would
be its synthetic pathway and the demand step could be protein synthesis.

201
202 7 Control structures in metabolism

4
1
a)

3 0
or Csupp

or Cdem
-1
M

M
2
-2
Csupp

Cdem
J
J

-3
1
-4 b)

0 -5
0.1 0.7 1.3 1.9 2.5 -0.1 -0.7 -1.3 -1.9 -2.5
supp
εdem
M
εM

Fig. 7.1: Control of flux and regulation of concentration.

Relationships between the control coefficients and elasticities for the 2 step pathway
of Scheme 7.1 . a) Control by the supply pathway with ε supply
M
J
= -0.2: —, Csupply ;
M
· · ·, Csupply . b) Control by the demand pathway with εdemand
M
J
= 0.2: —, Cdemand ; · · ·,
M
Cdemand .

Alternatively, M might be a central metabolic intermediate such as acetyl

CoA, in which case supply would be a catabolic pathway such as glycolysis
and the demand might be further catabolism in the tricarboxylic acid cycle
or else fatty acid biosynthesis.
Jannie Hofmeyr and Athel Cornish–Bowden106 analyzed how regulation
of M interacts with the control of flux by looking at the factors that affect
the flux and concentration control coefficients and their relative values. Their
reason for doing this was that the flux control coefficients show how effective
the supply and demand pathways can be at controlling the flux (assuming
the existence of some mechanism for changing the overall activity of either of
these steps). At the same time, the concentration control coefficients show
how much M will change if control is exerted on either of the steps. The
method of deriving the flux control coefficients of the two steps in terms of
the elasticities was explained in Chapter 5.3.3, p. 123. For Scheme 7.1, the
solutions are:

J εdemand
M 1
Csupply = = (7.1)
εdemand
M − εsupply
M
1+Q

J −εsupply
M Q
Cdemand = = (7.2)
εdemand
M − εsupply
M
1+Q

where Q, which is equal to= -εsupply

M /εdemand
M , has been introduced to em-
phasize that it is the relative values of the two elasticities that determine the
7.1 Supply and demand 203

flux control distribution rather than their absolute values. (The minus sign
gives Q a positive value because εMsupply , the product inhibition elasticity, is
typically negative.) The control coefficients on the concentration M can be
derived in a similar fashion (see Chapter 5.8.1, p. 133) to give:

M 1 1 1
Csupply = = (7.3)
εdemand
M − εsupply
M
εdemand
M 1+Q
M −1 −1 1
Cdemand = = (7.4)
εdemand
M − εsupply
M
εdemand
M 1 + Q
M M
Obviously Csupply = -Cdemand , but it is not possible to express the concen-
tration control coefficients solely in terms of the elasticity ratio. In fact, the
magnitude of the concentration control coefficients are inversely proortional to
(εdemand
M − εsupply
M ), the sum of the magnitudes of the two elasticities (because
supply
εM is itself negative).
Why is this important? It is because there are many instances where
pathways exhibit large changes in flux accompanied by relatively very much
smaller changes in metabolite concentrations. The extreme example is the
changes in glycolytic flux in animal muscle on going from the resting to the
working state, which can be between 100 and 1000–fold, yet the changes in
the concentrations of glycolytic intermediates are trivial.24, 101, 107, 204 There
are significant advantages to metabolite homoeostasis during flux changes:
1. if pathway intermediates are also participants in other metabolic path-
ways that are not required to change in rate at the same time as the
pathways being controlled, it is evidently better to minimize distur-
bances in the common intermediate concentrations;
2. having to make large changes in intermediate levels slows down the rate
at which a pathway can respond to a control signal with a change in
flux,62 so faster responses can be made if metabolite concentrations are
kept as near–constant as possible, and
3. avoiding large concentration changes during large flux changes min-
imizes the possibility of sudden adverse changes in cellular osmotic
strength.6
If control is exercised on a metabolic pathway, effective control of the flux com-
bined with good homoeostasis of the intermediates requires that the pathway
element acted on by the control mechanisms must have a high flux control
coefficient (so it can exhibit a strong response to the effector signal) but low
concentration control coefficients, so that the response of the concentrations
to the same signal is small. The conditions that have to be met to achieve
this are different depending on whether control is exerted in the supply block
or the demand block.
J
In control by the supply block, a high flux control coefficient, Csupply , is
demand supply
obtained if εM is significantly greater than | εM |, corresponding to
204 7 Control structures in metabolism

a low value of Q (see Fig. 7.1a and Eqn. [7.1]). The concentration control
coefficient only becomes smaller than the flux control coefficient if εdemand
M is
greater than 1 (Fig. 7.1a and Eqn. [7.3]). The elasticities we are considering
are composite elasticities for a block of enzymes rather than those of single
enzymes. For single enzymes, elasticities lie mainly in the range -1 to 1 for
those showing simple kinetics, or from around -3 to 3 for allosteric enzymes,
except where the reaction is near to equilibrium (Chapter 5.3). The way
in which the elasticities of the component enzymes combine to give block
elasticities generally makes a block elasticity smaller in magnitude than any of
its components, and since the blocks we are considering are pathway segments,
they would not be near equilibrium. A small (negative) value for εsupply M
would not be difficult to obtain since εsupply
M is a product inhibition elasticity,
and these are usually smaller in magnitude than substrate elasticities such
as εdemand
M in the case of single enzymes. When we are dealing with a block
of enzymes, this difficulty tends to be even worse, since the overall product
inhibition elasticity of the block is expected to be much weaker than those
of the component enzymes (reflecting a rather poor backward transmission of
the signal from the metabolite M ). However, too small a value of | εsupply M |
(e.g. giving Q < 0.1) would not be an advantage, since the larger it is, the
smaller the concentration control coefficient (Eqn. [7.3]).
If the elasticity of the demand block could be dominated by the elastic-
ity of an enzyme with cooperative substrate kinetics, it might be as large
as 3. My investigations of this problem with my colleague Simon Thomas
have shown that this may indeed be possible, for example if the cooperative
enzyme is the first enzyme in the demand block. The cost of this, though,
is poorer metabolite homoeostasis within the demand block. Improving on
this probably requires unusual pathway structures not often seen in practice
(e.g. sequences of several cooperative enzymes). Thus in spite of the tra-
ditional view (Chapter 4.1) that control should be exerted at the beginning
of a pathway (i.e. in the supply block), it is not easy to see how this can be
implemented with the available components to give a well–regulated pathway,
i.e. one that shows good metabolite honoeostasis during flux changes.
What about control by demand? An experimental example where control
has been shown to be shifted to the demand end of the pathway is hepatocyte
energy metabolism (Chapter 6.3.3.2). The requirement for control by the
J
demand block is that Cdemand J
exceeds Csupply , which requires that | εsupply
M |
demand
should be significantly greater than εM , giving a high value for Q (see
M
Fig. 7.1b and Eqn. [7.2]). The concentration control coefficent, Cdemand , only
supply
becomes smaller than the flux control coefficient if εM is more negative
than -1 (Fig. 7.1b). A low value of εdemand
M will be obtained if the demand
block is approaching saturation with M . This will also make the flux in the
demand block relatively insensitive to changes in the level of M , which may
well be an advantage if there are other processes that can affect it. The large
supply
negative value required for εM might seem to be a problem, especially as
I have just explained that the product elasticity of a block of enzymes tends
7.2 Feedback inhibition 205

to be weaker than those of any of its component enzymes. This would create
an obstacle to efficient homoeostasis in the supply–demand sequence, were it
not that metabolism has a solution to this difficulty, as we shall see in the
next section.

7.2 Feedback inhibition

7.2.1 Discovery and relationship to allosteric enzymes

X0 1 ................. S1 2 ................. S2 3 ................. S3 4 ................. X4

..
...... ..
..
ª.
....................................................................................................................

Fig. 7.2: A feedback inhibition loop.

This is a generalized feedback loop, with the enzymes represented by numbered boxes.
Metabolite S3 inhibits enzyme 1.

If bacteria such as E. coli are growing on [14 C]glucose together with a nitro-
gen source, the amino acids incorporated into their proteins are radioactively
labelled showing that they are being synthesized from the glucose carbon. If
one of these amino acids is added to the medium, it was established in the mid
1950s that the synthesis of this amino acid is selectively discontinued. Part of
the mechanism of this effect has been found to involve specific repression of
the expression of the synthetic enzymes in the presence of end product, but
this cannot provide a complete explanation, since growing bacteria degrade
very little of their proteins and therefore repression cannot account for the
existing synthetic enzymes not making the amino acid after it has been added
to the medium. The other part of the mechanism was discovered in 1956
when Umbarger250 found that the first enzyme in the pathway to isoleucine,
threonine deaminase, is specifically and strongly inhibited by isoleucine. This
feedback inhibition by the end product of a pathway is highly specific since the
structurally similar amino acids leucine and valine are not effective inhibitors
of the enzyme.
In the same year, Yates and Pardee281 also reported feedback inhibition of
aspartate transcarbamylase in E. coli: this first step in the synthesis of pyrim-
idine nucleotides from aspartate, is inhibited by the end product CTP. Once
these two examples had been discovered, further ones followed and it became
clear that feedback loops onto the first irreversible enzyme of a pathway are
a common structural feature in metabolism. In Chapter 3.5, p. 71, we saw
that the unusual inhibition kinetics of these enzymes led to the recognition
that they form a special class of allosteric enzymes. Although it was clearly
established for some of the early examples, particularly in amino acid synthe-
sis, that the feedback does operate to slow the rate of the synthetic pathway,
in many other cases it is a presumption that the purpose of the feedback is
206 7 Control structures in metabolism

to regulate the rate of the pathway. In the case of histidine biosynthesis in E.

coli, the in vivo concentration of histidine is too low to operate the feedback
inhibition on the first enzyme of the pathway. Again in E. coli, mutants do not
differ greatly in growth rates and yield when the normal allosteric, feedback–
regulated phosphofructokinase is replaced by a non–inhibited form.76 The
function of feedback inhibition is therefore not always as simple and obvious
as has been assumed.

7.2.2 Feedback inhibition and control analysis

1.0

0.8
Flux control coefficient

b
0.6

0.4

0.2 a

0.0
-5 -4 -3 -2 -1 0

log (inhibition constant)

Fig. 7.3: Feedback alters the control distribution.

The flux control coefficients of enzymes 1 (curve a) and 4 (b) of the pathway in Fig. 7.2
are plotted against the inhibition constant for S 3 on enzyme 1 in this simulated example.
The inhibition weakens from left to right.

It is in the interpretation of feedback inhibition that conventional bio-

chemical thought and Metabolic Control Analysis confront one another with
particular force. After all, the concept that there is no single rate–limiting
step but instead several steps with varying degrees of influence does not neces-
sarily undermine conventional thinking; superficially it seems that biochem-
istry can continue as before provided that the label ‘rate–limiting’ can be
replaced by a flux control coefficient that numerically measures the degree of
rate-limitingness. However, Metabolic Control Analysis of feedback inhibi-
tion shows that there is a much deeper and more fundamental contradiction
between the old and the new.
Right back at the origin of Metabolic Control Analysis, Henrik Kacser &
Jim Burns119 had considered the effects on the flux control coefficients of the
7.2 Feedback inhibition 207

pathway caused by feedback inhibition by a metabolite near the end of a linear

pathway onto an enzyme near its beginning (e.g. Fig. 7.2). They used rather
simplified kinetic equations so that the elasticity values could be easily calcu-
lated, but were able to show that feedback inhibition makes the flux control
coefficient of the regulated enzyme smaller, and the flux control coefficients of
the steps following the metabolite exerting feedback larger. This is illustrated
by a simulated example in Fig. 7.3 where the effect of changing the strength of
the feedback inhibition on the control coefficients of the enzymes is shown. (In
the example, the first enzyme follows a Monod, Wyman & Changeux allosteric
rate law, and the others normal Michaelis–Menten kinetics.) Strong feedback
inhibition has the effect of transferring flux control away from the beginning of
the pathway, where it would otherwise tend to be, to the enzyme (enzyme 4 in
the example) consuming the feedback metabolite (S3 ). As the strength of the
inhibition weakens, the flux control coefficient of enzyme 1 increases. On the
basis of the traditional criteria for the identification of rate–limiting enzymes
(Chapter 4), a highly regulated enzyme near the start of a metabolic pathway
would be a prime candidate for the rate–limiting step, so the suggestion that
it wpould have a low flux control coefficient was particularly unwelcome and
led to considerable resistance to the concepts of Metabolic Control Analysis
and the flux control coefficient in particular. However, the Control Analysis
interpretation has been vindicated by experiments, such as those quoted in
Chapter 6 that showed that substantial increases in the activity of just such
an enzyme (phosphofructokinase) has negligible effects on its pathway flux
(glycolysis).
The only element of Kacser & Burns’ analysis119 that could be regarded as
supporting traditional ideas was their demonstration that the transfer of con-
trol by feedback inhibition would be most effective if the enzyme subject to the
inhibition would otherwise have been the rate–controlling step of the sequence
up to the feedback metabolite. That is, if the pathway could be terminated
at the feedback metabolite and there was some way to keep its concentration
constant, then the kinetic and pathway characteristics that would cause the
inhibited enzyme to have the largest possible flux control coefficient in this
case would also lead to the most effective transfer of flux control in the full
pathway with feedback operating. It is therefore a reasonable hypothesis that
an enzyme subject to feedback inhibition is the dominant enzyme in rate con-
trol in its pathway up to the feedback metabolite, and of course this enzyme
would probably have been identified as the ‘rate–limiting step’ by traditional
criteria. However, it is important to note that an experimental test of this
hypothesis would be difficult and has not been performed. Furthermore, the
hypothesis is little more than a theoretical curiosity that focusses on the point
where traditional thought and Metabolic Control Analysis diverge. In prac-
tice, the inhibited enzyme will be found to have a small flux control coefficient
under the usual circumstances where the feedback metabolite is free to vary
in concentration.
An extensive theoretical analysis of the properties of pathways with feed-
back inhibition was carried out by Michael Savageau in Michigan using his
208 7 Control structures in metabolism

Biochemical Systems Theory.215–217 This has some differences from Metabolic

Control Analysis, but does involve concepts equivalent to elasticities and con-
trol coefficients. He showed that feedback inhibition can reduce the concen-
trations of the feedback metabolite and of all the intermediates between it
and the inhibited enzyme. Feedback inhibition also tends to reduce the vari-
ation of the feedback intermediate (and that of the other metabolites in the
loop) in response to fluctuations in the concentration of the source metabolite.
These effects are proportional to the strength of the feedback (as represented
by its elasticity on the inhibited enzyme). However, they do depend on how
the comparison between the inhibited system and the uninhibited system is
made, since there is not an obvious single choice for the ‘reference state’ that
is the basis of the comparison. A more generally valid conclusion concerns
the effects of feedback on the concentration control coefficients (though he
expressed this in terms of his equivalent measures, the parameter sensitivi-
ties of the concentrations). He showed that feedback inevitably reduces the
dependence of the concentrations of the feedback metabolite and the inter-
mediates in the loop upon the enzyme activities, to an extent that depends
on the strength of the inhibition. Although Savageau established methods
that could examine the effects of feedback on the pathway flux, he did not
pursue this aspect of the analysis. He did show though that feedback inhibi-
tion causes a pathway to reach a new steady state faster after a perturbation
than an uninhibited pathway does;216 a similar conclusion was also reached
by a different theoretical route (the study of the transient times of metabolic
pathways) by John Easterby.63
Although these results emphasize the stabilizing effects of feedback inhi-
bition, Savageau215 and many other researchers have identified circumstances
where it becomes destabilizing. This happens if the feedback loop is long but
the inhibition is strongly cooperative. The pathway then never settles to a
steady state but oscillates about it. In intermediate cases, the pathway does
reach the steady state relatively rapidly after a perturbation to the rate, but
goes through a series of diminishing (or damped) oscillations first. Oscillations
can be explained by the delay in the transmission of the feedback signal; if the
feedback enzyme is running at the wrong rate but the pathway is long, there is
a time lag before the concentration of the feedback metabolite changes enough
to readjust the rate of the inhibited enzyme. Then, if the adjustment is too
severe (because the cooperativity means the rate change is disproportionately
much greater than the metabolite change), there is again a delay before this
causes a change in the feedback metabolite again, so the rate keeps swinging
above and below that required. (The same phenomenon occurs in national
economics: there is a delay caused by information gathering and government
inertia between some economic imbalance arising and any policy changes be-
ing made to counter it; severe measures are then taken to redress the problem,
ensuring the inevitable continuation of ‘boom’ and ‘bust’ cycles.) Metabolic
oscillations have been observed in glycolysis in yeast and muscle as well as
in photosynthesis and mitochondrial respiration.103 Feedback inhibition on
phosphofructokinase by ATP (coupled to activation by AMP) can account for
7.2 Feedback inhibition 209

the oscillations in glycolysis, which have been studied extensively by Britton

Chance and Benno Hess and their colleagues.
Jannie Hofmeyr and Athel Cornish–Bowden106 used Metabolic Control
Analysis to study the effects of feedback inhibition on flux and concentration
control coefficients to complement their study of supply and demand struc-
tures described earlier (section 7.1). I pointed out there that achieving a
high value of the elasticity ratio Q, -εsupply
M /εdemand
M , presents a difficulty that
has to be overcome in order to operate the supply and demand pathways in
the domain where there is effective homoeostasis of the intermediate metabo-
lite M . This is because the product inhibition elasticity εsupply
M tends to be
weaker than any of the product inhibition elasticities of its component en-
zymes. However, Hofmeyr & Cornish–Bowden showed that when the supply
pathway begins with an enzyme that is feedback–inhibited by M , the alge-
braic expression for εsupply
M contains a new term that comes from the feedback
elasticity on the first enzyme, ε1M . Furthermore, the supply block elasticity is
approximately equal to this elasticity (εsupply
M ≈ ε1M ) under conditions where
the other contributions are small. (As in Kacser & Burns’ original analysis,
this ensures that the inhibited enzyme is the most controlling enzyme in the
supply block.) So, if the feedback–inhibited enzyme is an allosteric enzyme
that can have large–valued elasticities because of cooperativity (e.g. Fig. 5.9),
supply
it is possible to achieve the large Q value and large negative value of εM
that is required for effective homoeostasis of M . This implies that feedback
inhibition loops function primarily to ensure homoeostasis of metabolite con-
centrations, whereas conventional biochemical explanations have emphasized
their function in rate control.
Hofmeyr and Cornish–Bowden continued their Metabolic Control Analysis
of feedback inhibition in a supply–demand pathway to examine just this point.
They varied the degree of cooperativity of the feedback of M in a computer–
simulated pathway and studied the effects this had on the behaviour of the
system as the demand was varied. The changes had relatively little impact
on the steady state fluxes and the flux control coefficients, but there was a
big difference in the variation of M and its concentration control coefficients,
M
with the smallest range of M and smallest value of Cdemand being achieved
with the greatest degree of cooperativity of feedback by M . A high degree of
cooperativity in the feedback inhibition generates the risk of oscillations, so
its existence implies that it has been selected for in evolution in spite of this
disadvantage, and the only demonstrable unique advantage that cooperative
feedback inhibition has is the improvement of metabolic homoeostasis.
Putting the theoretical analyses together, our picture of feedback inhibi-
tion now looks like this:

1. Feedback inhibition is an antidote to the tendency of reactions at the

start of a pathway to have the greater control of flux. It transfers con-
trol to the reactions using the feedback metabolite from the enzyme it
inhibits, which is probably the most rate–controlling enzyme in the sup-
ply of the metabolite, though this control is of lesser significance in the
210 7 Control structures in metabolism

full pathway.

2. Feedback inhibition improves homoeostasis of the concentration of the

feedback metabolite (and all metabolites in the feedback loop), and
increased cooperativity of the inhibition specifically enhances this effect.

3. Feedback inhibition improves the stability of pathways in that it speeds

up the return to a steady state after some random perturbation, or the
rate of reaching a new steady state when external conditions change.

For the reasons given in the earlier section on supply and demand (section 7.1),
the first two effects are in fact inextricably linked. Thus enzymes subject
to inhibition are regulatory enzymes in the sense in which regulation was
defined in Chapter 1.1, p. 1; they are not, contrary to common belief, effective
control sites for changing the pathway flux because they have small flux control
coefficients.
Of course, part of the reluctance to accept this conclusion arises because
a regulatory enzyme subject to feedback inhibition is demonstrably inhibited,
and the degree of inhibition does change with metabolic circumstances (as
shown by cross-over plots that identify phosphofructokinase as a site of inhi-
bition in glycolysis, Fig. 4.5). In fact, this confers no special status on the
feedback–inhibited enzyme; every enzyme in the pathway has also changed in
rate under the same circumstances. The only difference is that the feedback–
inhibited enzyme had the potential to have a more dominant role in control,
but feedback inhibition has suppressed that and made it subordinate to the
steps that utilize the feedback metabolite.

7.2.3 Patterns of feedback inhibition in branched pathways

Because anabolism radiates outwards from a relatively small number of metabolic
precursors in the core of metabolism, biosynthetic pathways are usually branch-
ed so that the early part of the pathway leads ultimately to two or more
end–products. The regulation of the pathway then has to be able to cope
with differential variations in the net demand for the different end products.
For example, if the end products were two different amino acids in a bac-
terial cell, it might occur that in some circumstances neither is available in
the environment and they both have to be synthesized in the amounts re-
quired for protein synthesis, but in other circumstances, enough of one of
them is present in the medium to require very little synthesis of it whereas
the other must be synthesized fully. In both cases, the common part of the
pathway must ensure homoeostasis of the common branch–point intermediate.
The biosynthetic pathways of bacteria exhibit a variety of different molecu-
lar mechanisms employed in the regulation of such branched pathways. 237
However, when Michael Savageau continued his theoretical study of feedback
inhibition using Biochemical Systems Analysis,217 he concluded that there are
only two functionally distinct types, though one of them can be implemented
7.2 Feedback inhibition 211

by a number of different molecular mechanisms. The two basic forms are

nested and sequential feedback.

7.2.3.1 Theory of nested and sequential feedback

........................................... ........
.............. ........X2
a)
. ........................ ª .................. .........X2 b)
ª.........
...... ... . . ..
....ª...... .. . .. 6 ... . .
. 6
. .
.. .... .. ....
.. . S .... . S
... ..... ......... 4 ..
.. ... ...
.. ª ..............
.
.... ......... 4
.
..... ...
4 ..... ......
...... ...... 4
... .
.. .... .
... .
. .
X0 .
1 . S1..
... ......
2 . S2... X0 .
1 . S1.
.... 2 ..S2...
.
.....
...... ..... .....
.. 3 3
.. ..... .......... .... .......
..
... .
.. S3... ... ...S3.
...ª .. ..... .. .....
...... .. .. ..
....... .... 5 .. ... 5 ..
.......... ..
................ ª ................................. ª ........
............................
........................................X1 X1

Fig. 7.4: Feedback inhibition patterns in branched pathways.

a) Nested inhibition; b) sequential inhibition. Any step shown above can be regarded
as being composed of an arbitary number of enzyme reactions.

Savageau’s studies of the two basic types of feedback mechanism in branched

pathways, illustrated in Fig. 7.4, showed that the sequential system seemed
to be the more reliable in that it generally behaves as required when the
source and end–product metabolites vary, though its degree of effectiveness
in ensuring homoeostasis varies with the particular values of the kinetic and
inhibition constants of the enzyms involved. The nested pattern can be both
better and worse, for the type of behaviour it exhibits varies with the values
of the kinetic and inhibition constants. It can give more effective homoeosta-
sis of end–product concentrations in response to variations in the supply of
source metabolite and the kinetic characteristics of the enzymes. However,
this behaviour is less stable because in some circumstances it changes to one of
two undesirable forms. Translating Savageau’s results into Control Analysis
terminology, the provision of one end product (say, S2 in Fig. 7.5) can cause
the concentration and rate of synthesis of the other (S3 ) to drop unless:

J1 ε12
<1 (7.5)
J2 ε22

Since the flux before the branch point (J1 ) is greater than that in the branch
(J2 ), this requires the product inhibition elasticity of S2 on its own branch
to be significantly stronger (as indicated by the elasticity) than its inhibition
of the common pathway. In fact, the equation can be interpreted as a re-
quirement that a given change in the end product, ∆S2 , should potentially
212 7 Control structures in metabolism

cause a bigger change in the rate of the branch than of the common pathway
(∆v2 > ∆v1 ) if an increase in S2 is not to cause a fall in S3 . (I say potentially
because these are the predictions for the isolated steps, as they are derived
from the elasticities, not for the whole system.) Of course, there is an equiv-
alent need for the other end product to inhibit its own branch more strongly
than the common pathway. Because of the velocity ratio in Eqn. [7.5] and the
relatively limited range of inhibition elasticity values, it seems that satisfying
both requirements simultaneously will be unlikely unless the two branch fluxes
have approximately similar shares of the common feed flux. However, several
of the observed molecular mechanisms to be described below may exist to en-
sure the correct relative relationships of the inhibition strengths. On the other
hand, incorrect operation of these systems is been observed experimentally in
cases where provision of an excess of one end–product (typically an amino
acid) results in growth inhibition by causing a deficiency in the synthesis of
another.
This requirement on the inhibition strengths has the incipient danger that
one of the end products might not inhibit the common pathway sufficiently
so that when its concentration is increased, first S1 and then the other end–
product start to accumulate excessively. This in fact is Savageau’s other
undesirable form of behaviour of the nested feedback inhibition pattern: there
is the possibility of the breakdown of the steady state if

J1 ε12
¿1 (7.6)
J2 ε22

Thus the more effective homoeostasis possible with nested inhibition can only
be reliably exploited where the relative inhibition strengths, represented by
the inhibition elasticities scaled by the relative fluxes, fall within a restricted
range. However, in spite of this apparent unreliability, nested feedback inhi-
bition structures are found in metabolism.

2 - S2 4 - X4

X0 1 - S1

3 - S3 5 - X5

Fig. 7.5: A generalized branched pathway.

S2 andS3 are the ‘end products’ of the biosynthetic pathways, but are consumed by
incorporation into biomass. S2 and S3 can exert feedback inhibition of the nested or
sequential type.
7.2 Feedback inhibition 213

pep + ery-4-P
.......
... ....ª .....
.....
DAHP ....
...... ...
..... ...
.... ..
..
shikimate ..
... ... ..
... ... ............. ..
... ... ª ...... ...
... ... .. .
.. ..
shikimate-3-P ... .
..... ... ...
. .
...... ......
.
...................... chorismate
. .
ª ...................... .... .... ........ ......
. ....
...
.. .
... ......... . .. ....
..... ...... ...... ........
.... ..anthranilate
. .... ......................
.. ........... ..................ª................................. .............................ª
prephenate
......
tryptophan........ ...
...
. .. .......... ....
. .
.. .
.. ..
..... phenylpyruvate .
. hydroxyphenylpyruvate ..
..
.
. ....
... ..
............
.. . ................ ......... ....
.
phenylalanine tyrosine

Fig. 7.6: Feedback in aromatic amino acid synthesis.

The pathways in B. subtilis exhibit sequential feedback inhibition.

7.2.3.2 Sequential feedback inhibition

Sequential feedback inhibition was discovered in the pathways synthesizing
the aromatic amino acids tryptophan, tyrosine and phenylalanine in Bacil-
lus subtilis (Fig. 7.6). The final common intermediate of this pathway is an
equilibrium mixture of chorismate and prephenate. These intermediates can
inhibit the first step in their synthesis, the joining of the glycolytic interme-
diate phosphoenol pyruvate and the 4–carbon sugar erythrose–4–phosphate
catalyzed by an enzyme sometimes termed DHAP synthase (because its true
name, and that of its product, are too long to be memorable). On the other
hand, none of the aromatic amino acids inhibits this enzyme. Each of them
specifically inhibits the first step of its own branch.

7.2.3.3 Nested feedback inhibition

As mentioned previously, Michael Savageau regarded a set of apparently dif-
ferent feedback inhibition systems as different molecular mechanisms for im-
plementing a single functional type: nested feedback inhibition (Fig. 7.4a).
214 7 Control structures in metabolism

The following examples illustrate the different categories of mechanism.

aspartate
................. 1 .........ª
ª ....................
...............
. ... .......
........ ... ... ......
.........
. aspartyl–P .....
. .
. ...
.. ..
.. . ..
..
.. . ... ..
..
.. aspartate semialdehyde
. ..
.. ..... ª.....
.... ....... .
...
.... ...
...
.. .........ª
....... .. ..
.... ....... ...
.. .
... .
..
........ . .... ........... ..
... ....... ..
.
..........
.. ...
...
. . . ..
.. .. . .
.....
. 2 ..
. .
... ... ...
. . dihydropicolinate .. ..
.. homoserine ...... . .
... ... .............. ........ .... ...
.. .. .......... ....... .. ..
. . ... . .
lysine threonine

Fig. 7.7: Nested inhibition with multiple enzymes in threonine and lysine synthesis.
Three aspartate kinase isoenzymes (step 1) occur in this pathway in E. coli., though the
one repressed by methionine (which is formed from homoserine) has not been shown
here. Step 2 is homoserine dehygrogenase.

7.2.3.3.1 Enzyme multiplicity A common method of ensuring that the

inhibited step in the common pathway responds to each of the end–products,
without any one of them having too powerful an effect (which must be avoided
as shown by Savageau’s conditions on the effectiveness of nested inhibition),
is to have separate isozymes, each of which is inhibited by one of the products.
(This is also generally linked to separate controls on synthesis of each of the
isoenzymes, such as specific repression by the end–product that also inhibits
it.) This is seen in the synthesis of threonine and lysine from aspartate in
E. coli (Fig. 7.7) where there are separate aspartate kinases inhibited by
each product. Furthermore, work by Georges Cohen and his colleagues41
showed that the isoenzymes aspartate kinase I and homoserine dehydrogenase
I occurred on a single bifunctional molecule and that both activities were
inhibited in parallel by a single threonine inhibitory site. This seems to fit
in with Savageau’s theory since it should ensure that the required degree of
relative inhibition of the common and branch point steps is ensured by the
protein structure. After all, there is no other obvious metabolic advantage
in having these two different activities in a single molecule as they are not
consecutive steps that could transfer their common intermediate between the
sites.
In reality, the regulation of this system is even more complex, since methio-
nine and isoleucine also share parts of this pathway from aspartate, and there
7.2 Feedback inhibition 215

is a third aspartate kinase and a second homoserine dehydrogenase whose

syntheses are repressed by methionine.
Another illustration of the complexity of the patterns of control in branched
biosynthetic pathways is that the aromatic amino acid synthesis pathway cited
above as an instance of sequential control exhibits control by nested feedback
with enzyme multiplicity in E. coli. There are three separate DHAP syn-
thases, each inhibited by one of the products.

.................................................................
.......................... ......................X2 ...
........
. .
. ...... ......
. ..
.
...... ª .
... .. ..
..
... .. 6 ..
. .... ª .. ...... ..
.. .. .. .
.. .
.. ....S4
... ..... ...
. .
.. ..
..... .... 4
.
. ... ...
X0 .
1 ..........S1 2 ..........S2... .............
.....
3
..... ..........
... S3.....
.. ...
... 5.
.
..
ª ......... ..
............................
X1

Fig. 7.8: Schematic diagram of the concerted form of nested feedback inhibition.

7.2.3.3.2 Concerted feedback inhibition Concerted feedback inhibition

(Fig. 7.8) is a form of nested feedback inhibition where the enzyme in the
common pathway has the property that it is not inhibited by either of the
end–products separately. Inhibition by one requires that the other be present.
Like the other versions of nested inhibition listed below, this is a distinction
based on the in vitro kinetics, since both products would normally be present
in vivo and therefore both would be inhibitory. This is probably a molecular
mechanism for minimizing the chances of the nested inhibition displaying the
bad aspects of its behaviour: if the inhibition by the product in excess were
stronger than necessary, the consequent tendency of the concentration of the
other to drop would be limited by its fall causing a weakening of the inhibition
by the one in excess. Conversely, if the inhibition by the product in excess
were too weak, the tendency of the other to rise in concentration would cause
a strengthening of the inhibition.
This type of inhibition occurs in threonine and lysine synthesis in Bacillus
polymyxa, which does not have multiple aspartate kinase isoenzymes as E.
coli does.

7.2.3.3.3 Cumulative feedback inhibition Cumulative feedback inhibition

(Fig. 7.9) is effectively just like the standard form of nested feedback inhibi-
216 7 Control structures in metabolism

..................................................................
....................................................................................X2 ....
............................. .. .. .
....... ........ ..
..
................ ª
.... . . ..
..
.
........ . 6 ..
.. .. .. .
... ..
... S ...
ª
.. .... . ª .
..
...... ..........
4
...... ...
. .
.......... ....
4
....
.
.. . ....
X0 1 .........S1 2 .........S2... ..........
...... .....
.. 3
..
.. ..... ..........
.. .
... ... S3.....
...ª
...... .. ...
....... ... 5.
.......... ... .
................ ª ....................... ........
...............................................X1

Fig. 7.9: Cumulative feedback inhibition.

A schematic representation of this form of nested feedback inhibition.

tion, with the additonal property that the inhibition of the common enzyme
in the presence of both end–products is greater than that caused by either
separately. However, this greater inhibition is exactly that expected from the
combination of the effects of each when they act independently. For example,
if one caused one–half inhibition, and the other caused one–third, the expected
outcome for independent effects is the product of the fractional activities, i.e.
1/2 × 2/3 = 1/3, or two–thirds inhibition.
This was first suggested to be the case for the glutamine synthase (glut-
amate–ammonia ligase) of E. coli:

ATP + glutamate + NH3 −→ ADP + Pi + glutamine

Glutamine is the nitrogen donor in the synthesis of a range of end–products,

and so can be regarded as the common metabolite of a number of diverging
pathways (considered in terms of nitrogen flow, rather than carbon flow as nor-
mal), including synthesis of AMP, CTP, histidine, tryptophan, glucosamine–
6–phosphate and carbamoyl phosphate. Stadtman and his colleagues 197 estab-
lished that all of the above (except tryptophan) and alanine and glycine had
separate inhibition sites on the enzyme, each of which separately only caused
partial inhibition even at saturating concentrations of the inhibitor. Together,
the effects were cumulative and all eight together caused almost complete in-
hibition. However, the original observations were made on enzyme that was
a mixture of adenylated and non–adenylated forms, and the interconversion
between the two (see section 7.4.4 later in this Chapter) causes changes in
the inhibition characteristics; from later studies it appears that there may be
some interactions between the sites under some circumstances, though prob-
ably not enough to change the essential conclusion. (Nor is some uncertainty
in the kinetics of an enzyme with six substrates and products, eight inhibitors
7.3 Substrate or ‘futile’ cycles 217

and a requirement for divalent cations surprising; Savageau,217 as quoted in

the Introduction, Chapter 1.2.1, pointed out the practical impossibility of a
complete kinetic characterization.)

7.2.3.3.4 Synergestic feedback inhibition Synergestic feedback inhibition

is intermediate between concerted and cumulative feedback inhibition. Un-
like concerted inhibition, each end–product exhibits some partial inhibition
on its own account, but the inhibition of both end–products together is much
greater, and greater than would be expected for cumulative inhibition, i.e.
there is interaction between the separate inhibition sites. Synergestic inhi-
bition was first observed in purine biosynthesis on the first enzyme of the
common pathway to AMP and GMP, amidophosphoribosyltransferase:

glutamine + 5–phosphoribose–1–diphosphate + H2 O −→ gluta-

mate + 5–phosphoribosylamine + pyrophosphate

This allosteric enzyme has separate inhibitory sites for AMP and GMP, but
both of them together interact to cause greater inhibition than would be
expected. Similar synergestic inhibition is seen with the glutamine synthase
of Bacillus lichiniformis, which is only slightly inhibited by low concentrations
of glutamine, histidine or AMP, but almost completely inhibited by AMP plus
histidine or glutamine plus histidine.

7.3 Substrate or ‘futile’ cycles

7.3.1 Definition
Apart from branches, other common network elements in metabolism are
the various types of cyclic pathway. Some, like the tricarboxylic acid cycle,
are inherently cyclic. In fact, the reactions of the tricarboxylic acid cycle
cannot themselves account for the the initial generation of the di– and tricar-
boxylic acid intermediates, so in the absence of other anaplerotic reactions to
synthesize them, the intermediates form a moiety–conserved cycle in which
the total quantity of intermediates is fixed. In other cases, the existence of
the cyclic pathway is not inevitable; a circular route can be traced on the
metabolic chart, but will only occur if all the reactions are simultaneously
active. Amongst these potential cycles are those that appear to be wasteful
or unnecessary, and which were therefore termed futile cycles. Later, when
it was conjectured that these cycles might fulfil certain specific functions, the
more neutral term substrate cycles came to be preferred.
A substrate cycle could potentially exist in metabolism wherever a reaction
(or set of reactions) that converts metabolite S1 into S2 is opposed by a
second set that reconverts S2 to S1 . If, under certain circumstances, both
sets of reactions are active simultaneously, then there can be a cyclic flux
whereby an appreciable proportion, perhaps all, of the S1 that is converted
to S2 reverts to S1 (Fig. 7.10). Both sets of reactions must be intrinsically
218 7 Control structures in metabolism

A1 A2
..... .........
...... .. .
.....
........
........... ..
....... .......
...................................................
.............. .................
................ v2a
.........
.......
.
.... ...... .
v1 .... ......... v3
X0 - S1 S2 - X1
.......... ...
. ...... .
.
....... ....
.......... v .......
.................... 2b .............................
.... ....
....... ..............
............... ..........
.......... ........
...
........ ......
........ ....
...
A4 A3

Fig. 7.10: A general substrate cycle.

The substrate cycle reactions are made favourable by coupling to changes involving
metabolites A1 to A4 . For example, A1 could be ATP, A2 , ADP, A3 , H2 O and A4 ,
phosphate, so the overall effect of the cycle is to hydrolyze ATP.

favourable (i.e. exergonic) even though they link the same two metabolites in
opposite directions, and for this to be the case, at least one of the sets must
be coupled to some other exergonic process (for example, phosphorylation by
ATP). Hence the cyclic flow leads to no net change other than the dissipation
of energy by a net flux through the coupled process that drives the cycle. This
characteristic - the occurrence of two oppositely directed sets of reactions that
would operate to achieve no change other than dissipation of energy - is often
given as the definition of a substrate cycle126, 196, 237 and is the particular
justification for the term ‘futile cycle’.
The early evidence that cycling was not just a possibility but did actually
occur in cells under certain circumstances was provided by:

• Cahill and colleagues26 in 1959 in respect of cycling between glucose

and glucose–6–phosphate in liver;

• Steinberg241 in 1963 in connection with cycling between fatty acids and

triacylglycerol in adipose tissue, and

• Newsholme and Underwood172 in 1966 for cycling between fructose–6–

phosphate and fructose–1,6–bisphosphate in kidney cortex.

These specific examples of substrate cycles do, however, illustrate a problem

with the definition given above: the definition is too broad and does not
represent all the features that are characteristic of actual substrate cycles.
Thus the previous definition is applicable to all the types of cycle shown in
Fig. 7.11, yet:
7.3 Substrate or ‘futile’ cycles 219

a) b)

............. .
.................v..............................
............... v ..................... .......
. .. ....... 2a ......
.. .... 1a ......
..
v1 v3 v2
X0 - S1 S2 - X1 X0 S1 - X1
........ ... ........ ...
........ v .
...
.. ........ v .....
.
.. ..
....................
....................2b .....................
...................1b

c) d)
X0
@ v1
@
.................v..............................
.
.................v..............................
.
....
....... 1a ...... @R
@ ....
....... 3a ......
.. ..
S1 S2 S1 S2
........ ........
........ v ..........
. ¡ ................ v3b ...................
.
.
....................
....................1b
¡ ........................
¡
¡
ª v2
X1

f)
e)
X0 X0 X2
@ v1 @ v1 v4 ¡
@ @ ¡
..............v.............................. ..............v..............................
@
R
@ ......
........ 3a ....... @R
@ ......
........ 3a ....... ¡
¡
ª
.. ..
S1 X2 S1 S2
........ . ........ .
¡ ................ v3b ................... ¡ ................ v3b ................... @
........................ ........................
¡ ¡ @
¡
¡
ª v2 ¡
¡
ª v2 v5 @
R
@
X1 X1 X3

Fig. 7.11: Topologies of some cyclic structures in metabolism.

Versions of these structures have been referred to as substrate cycles in the literature.
Second substrates used for driving intrinsically endergonic reactions (e.g. ATP) have
not been shown. X is used to denote a pool metabolite, and S to represent a variable
metabolite.

• only a and f are indisputably substrate cycles;

• b and e are possible substrate cycles (though because one of the substrate
cycle metabolites is a fixed pool, their properties are essentially those
of a simple branched pathway);
• c is a moiety–conserved cycle because the sum of S1 + S2 remains con-
stant at all times, and
• d is a dead–end cycle that cannot affect the steady–state properties of
the linear pathway.
220 7 Control structures in metabolism

Definition is important because cycles occur embedded in the highly inter-

connected network of metabolism where they can be less easy to identify and
classify than the idealized cases in Fig. 7.11. Even counting the number of
cycles can be difficult.
Newsholme and Crabtree stated that a distinguishing feature of a system
with a substrate cycle was that there were two distinct fluxes: one a net
conversion of S1 to S2 , matching the rates at which S1 is produced by input
reactions to the cycle and S2 is consumed by output reactions, and the other
the cyclic flux itself. This amounts to stating that the pathway must be
representable as Fig. 7.11a, if necessary by regarding each step as standing for
a large block of reactions. Although their definition arguably covers the ‘open
futile cycle’ defined by Stein & Blum240 or Reich & Sel’kov196 (Fig. 7.11f, this
could conceivably have a cyclic flux in the absence of a net flux through the
cycle (when v1 = v2 and v4 = v5 ). I therefore believe their definition should
be generalized as follows. A substrate cycle exists where:

1. The flux pattern in the network cannot be fully described as the com-
bination of the minimum number of linear paths needed to account for
the mass flows connecting the inputs and outputs. (This can apply to
larger, more complex networks like Fig. 7.11f.)

2. One of the additional fluxes needed to complete the description is a

feasible, internal cyclic route. (All the reactions of the cycle must be
exergonic; the requirement that the cycle is internal excludes cycles
Fig. 7.11b and e.)

3. There is one step of the cycle that can be deleted in principle and still
leave a network capable of connecting the observed input fluxes to the
observed output fluxes. (This is the criterion that shows the cycle is
intrinsically unnecessary; it eliminates conserved cycles like Fig. 7.11c
and hypothetical oddities like Fig. 7.11d.)

There is obviously plenty of scope to disagree with my proposals about which

of the cycles in Fig. 7.11 is classed as a substrate cycle. The important issue
is that there is a number of ways these cycles can be distinguished, and that
therefore it is unwise to assume that they all have similar properties.

7.3.2 Evidence for substrate cycling

The obvious difficulty about proving the occurrence of substrate cycles by
measuring their rates is that they produce no net change in the amounts of
their constituents. This is not true, of course, for the coupling reaction, and
cycling can be measured if the formation of products by the coupling reac-
tions can be detected. This is the case for the triacylglycerol:fatty acid cycle
in mammalian white adipose tissue during net triacylglycerol storage: glyc-
erol phosphate (derived from glycolysis) is needed for the conversion of fatty
acids to triacylglycerol, but the hydrolysis of triacylglycerols to fatty acids
7.3 Substrate or ‘futile’ cycles 221

a) ATP ADP b) ATP ADP

1 > 3 >
¡ ¡
¡
~ ª
¡ ¡
~ ª
¡
¾- glc glc–6–P ¾- fru–6–P fru–1,6–bisP
I @ I @
@
R @
R
= 2 = 4
Pi H2 O Pi H2 O

c)
6
?
pep
ADP
Á @
@
GDP Y 5@ - ATP
CO2 @ ¡
k @
R
@ ¡
ª
7 pyruvate

¡ CO@
2
GTP 6 ¡ ATP @
R
¡
¡
¡ W
¡
ª ADP
oxaloacetate
1
³ PP
³³ q
P
Fig. 7.12: The three substrate cycles of glycolysis and gluconeogenesis.
a) The glucose:glucose–6–phosphate cycle catalyzed by hexokinase IV (1) and glucose–
6–phosphatase (2). b) the fructose–6–phosphate:fructose–1,6–bisphosphate cycle cat-
alyzed by phosphofructokinase (3) and fructose–1,6–bisphosphatase (4). c) the phos-
phoenolpyruvate:pyruvate cycle catalyzed by pyruvate kinase (5), pyruvate carboxylase
(6) and PEPCK (7).

releases glycerol. White adipose tissue metabolizes glycerol poorly owing to

the virtual absence of glycerol kinase, so the formation of glycerol is a good
indicator of the rate of hydrolysis, which can be combined with measurement
of the net rate of formation of fatty acids (during net lipolysis) to calculate the
degree of cycling.241 These and similar experiments for studying the glycerol
balance can show cycling rates as high as 10–20 times the net rate of lipolysis
or lipogenesis in isolated fat tissue. This is so even though there is a poten-
tial mechanism for suppressing the cycle via control of the hormone–sensitive
triacylglycerol lipase, which is activated by protein phosphorylation by the
cyclic AMP–dependent protein kinase (section 7.4.3.1).
Several methods for the estimation of the degree of cycling in the three sub-
strate cycles of glycolysis and gluconeogenesis (Fig. 7.12) depend on monitor-
222 7 Control structures in metabolism

ing the fate of the radioactive hydrogen isotope, tritium (3 H). This is because
certain of the reactions of the pathway specifically cause the loss of 3 H at-
tached to a particular carbon atom of the metabolites. For example, glucose–
6–phosphate isomerase causes the loss of 3 H from [2–3 H]glucose–6–phosphate
when it is converted to fructose–6–phosphate; since the 3 H enters water and
is greatly diluted, it is not re–incorporated when the reverse reaction occurs.
Therefore, if [2-3 H]glucose is supplied to mammalian liver or kidney cortex,
the loss of 3 H from the 2 position of the glucose can be used as a mea-
sure of cycling because it shows that the glucose has been phosphorylated to
glucose–6–phosphate (which rapidly equilibrates with fructose–6–phosphate,
losing the 3 H, because of the high rate of the glucose–6–phosphate isomerase
reaction), and then dephosphorylated again back to glucose. Because of the
other metabolic reactions occurring simultaneously, the relative loss of 3 H
compared with the 14 C content of a doubly–labelled glucose molecule gives a
better measure of the cycling. Interpretation of such experiments is difficult
and often controversial; the topic is discussed at length in some of the arti-
cles listed under Further reading at the end of the Chapter. Randomization
of 14 C label between different positions of the glucose molecule can also be
used. For example, the fructose–6–phosphate:fructose–1,6–bisphosphate cycle
can be monitored in this way since, when aldolase splits the latter into two
trioses, the fragments can be reassembled at opposite ends of the hexose from
their original position.
Measuring metabolic fluxes by monitoring the spread of isotope through
a pathway (Chapter 2.5) can give the rates of the component reactions of
the cycle. This is because the cycle can give rise to faster migration of the
label between intermediates than would be expected from the net pathway
flux, just as near–equilibrium reactions do (Fig. 4.2, Chapter 4.3). This ap-
proach has been used extensively by Jacob Blum’s group at Duke University,
particularly in connection with the measurement of substrate cycling in rat
hepatocytes.10, 58, 190 The experiments are much more informative than mon-
itoring isotope loss or rearrangement in a single compound, but involve much
more work, both experimentally and computationally.
The results from these measurements differ between the three cycles of gly-
colysis and gluconeogenesis in rat liver and hepatocytes and between different
metabolic states. In general, though, the highest relative rates of cycling are
observed with the glucose:glucose–6–phosphate cycle, for which the cycling
rate can be many times the net flux of glucose uptake or release.261 This is
consistent with the lack of any obvious mechanism that can suppress the cycle
by ensuring that the enzymes are not active simultaneously. In the liver of
some mammalian species (including rats, mice and humans) the hexokinase
is present as an isoenzyme (hexokinase IV, also commonly but somewhat
erroneously known as glucokinase51 ) that is competitively inhibited by the
reversible binding of a complex between a specific regulatory protein256, 257
and fructose–6–phosphate, the concentration of which varies in parallel with
glucose–6–phosphate. There is, however, no known regulatory mechanism
acting on the glucose–6–phosphatase (although it is spatially separated by its
7.3 Substrate or ‘futile’ cycles 223

location in the endoplasmic reticulum from the cytosolic hexokinase).

The flux round the fructose–6–phosphate:fructose–1,6–bisphosphate cycle
can be comparable to net gluconeogenic flux in hepatocytes, though in many
cases it is much less.146 The cycle also occurs in isolated liver, though the
results showing that the cycling rate can be several times gluconeogenic flux
use a disputed method,111, 127 and other results suggest lower cycling rates,
certainly much less than for the glucose:glucose–6–phosphate cycle. There
are also a number of regulatory and control mechanisms that act on the cy-
cle enzymes in opposite senses and which could therefore limit cycling. For
example, AMP, phosphate and fructose–2,6–bisphosphate activate phospho-
fructokinase but inhibit fructose bisphosphatase. (The older view that phos-
phofructokinase is activated by its product fructose–1,6–bisphosphate now
seems wrong; the activation observed in assays was because the product binds
weakly at the fructose–2,6–bisphosphate activator site, but when the activator
is present at physiological levels, product inhibition dominates.11, 186, 247, 255 )
Fructose–2,6–bisphosphate is a third messenger of hormone action, since its
synthesis and degradation in liver are controlled by glucagon and α–adrenergic
agents that change the activity of a bifunctional enzyme, phosphofructo–
2–kinase/fructose–2,6–bisphosphatase, by reversible phosphorylation brought
about by the cyclic AMP–dependent protein kinase185 (see Section 7.4.3.1).
In addition, the cycle enzymes themselves may be phosphorylated by cyclic
AMP–dependent protein kinase, but the significance of this is uncertain be-
cause the effects of phosphorylation on the kinetics of the enzymes that have
been demonstrated to date seem quite small.
Measurements on the phosphoenol pyruvate:pyruvate cycle often show pyr-
uvate kinase fluxes of the same order as gluconeogenic flux, though this has
often been in the presence of unphysiologically high levels of gluconeogenic
substrates. The gluconeogenic hormone glucagon seems to suppress the cycle
by causing inactivation of pyruvate kinase via reversible protein phosphory-
lation.

7.3.3 Suggested functions

The reason for preferring the term substrate cycle to futile cycle is the possi-
bility that cycling is intrinsically advantageous, and several different functions
have been attributed to cycling:
1. heat production (thermogenesis) in brown adipose tissue of mammals
and flight muscle of bumble bees;168, 234
2. more sensitive regulation of the net flux through the pathway by regula-
tion of the enzymes carrying the cycle flux,168, 169, 171 perhaps involving
properties typical of switching and trigger devices;196
3. control of the direction of flow at branch points and in bidirectional
pathways,112, 126, 169, 237 and
4. buffering of metabolite concentrations.26, 112, 171
224 7 Control structures in metabolism

Of these proposals, (1) is straightforward and capable of experimental confir-

mation without any theoretical problems, (2) and (3) have been the subject
of continuing discussion which will be considered below, and (4) has not been
developed in a quantitative manner nor subject to any extensive theoretical
analysis,.

7.3.4 Thermogenesis
It seems surprising that the generation of body heat by metabolism should
need any special explanation, since catabolism is characteristically exother-
mic. However, the coupling mechanisms of metabolism create a difficulty,
because part of the energy released during catabolic reactions is conserved
by the phosphorylation of ADP to ATP and the regulatory mechanisms of
catabolism ensure it slows down unless this ATP is hydrolyzed. The anal-
yses of the control of supply and demand pathways and feedback inhibition
presented earlier (Sections 7.1–7.2) imply that better homoeostasis of ATP
may be obtained if control by demand predominates over control by supply
(catabolism), and this expectation is supported by the results of the top–
down analysis of hepatocyte energy metabolism (Chapter 6.3.3.2). Therefore,
if there is a requirement for extra heat generation, either the demand for ATP
must be increased or the coupling of catabolism to ATP production must be
weakened. There is evidence that both approaches are adopted, and one of
the ways that the demand is increased is by substrate cycling.
Eric Newsholme and Bernard Crabtree suggested that the fructose-6–phos-
phate:fructose–1,6–bisphosphate cycle is used by bumble bees to raise the tem-
perature of their flight muscles in cold weather.168 They had made measure-
ments that showed that flight muscles of several species of bumble bees contain
unusually high levels of fructose–1,6–bisphosphatase, which is required for the
gluconeogenic pathway and is not usually present in significant amounts in a
highly glycolytic tissue like muscle. Other species of bee that have relatively
little of this enzyme are unable to fly at such low ambient temperatures (10 ◦ C)
as bumble bees. Even at these temperatures, bumble bees can maintain their
thoracic temperature at the 30◦ C necessary for them (and many other insects)
to fly. The hypothesis was supported by measurements made by a different
laboratory39, 40 that showed that cycling could be measured in bumble bee
muscles at low temperatures, but the cycling decreased with temperature and
had disappeared at 27◦ C. Furthermore, the cycling was prevented in flight,
probably by inhibition of the fructose bisphosphatase by the rise in Ca 2+ that
occurs on stimulation of the muscle. However, Newsholme and Crabtree also
calculated that the energy yield of this substrate cycle was not enough by itself
to account for the total heat generation, so there must be other mechanisms
as well.
This also seems to be true of another tissue for which substrate cycling
could make a contribution to thermogenesis: brown adipose tissue. Brown
adipose tissue is a form of adipose tissue in which the cells have more cyto-
plasm and mitochondria, and contain several small lipid droplets rather than
7.3 Substrate or ‘futile’ cycles 225

the single large one found in the more abundant white adipose tissue. Brown
adipose tissue is present in neonatal mammals, which generally are less ef-
fective than adults at temperature regulation, and in hibernating mammals.
In both cases, brown adipose tissue has a significant role in heat generation.
The triacylglycerol:fatty acid cycle is active in brown adipose tissue, though
calculations imply that it could only account for a small proportion of the
heat production by this tissue.168 The mitochondria of brown adipose tissue
contain specific mechanisms that uncouple respiration from ATP synthesis
and make a much larger contribution.

7.3.5 Sensitivity of control

.........................
.............v =100....................
.
.........
. 2a ......
v1 =10 .. ..... v3 =10
X0 - S1 S2 - X1
........
....... ....
.......... v2b =90 ...............
......................................

Fig. 7.13: Increased sensitivity of control by a substrate cycle.

The fluxes shown on each step are illustrative. If stimulation of the forward enzyme of
the cycle increases v2a 10% to 110, and the reverse reaction remains unchanged at 90,
then v1 and v3 must increase 100% to 20.

Around 1970, Eric Newsholme and his coworkers in Oxford168, 169, 171 de-
veloped the proposal that substrate cycles could be devices for increasing the
sensitivity of the regulation of pathway flux by an effector acting on one or
both of the pathway enzymes. The initial form of their argument169, 171 is
shown in Fig. 7.13; generalized algebraically, it led to the proposal that an
effector that makes a given percentage change in the activity of enzyme 2a
will cause a percentage change in the net flux (= v1 = v3 ) that will be larger
by a factor (1 + v2b /v1 ) than would occur if the pathway were linear with
v2b = 0.168 Similarly, the response of the net flux to a change in the activ-
ity of enzyme 2b would be amplified by the factor −v2b /v1 (with the minus
sign showing that an increase in the enzyme activity produces a decrease in
the net flux). If the same effector acted to stimulate one cycle enzyme and
to inhibit the other, these effects would be additive. This could be particu-
larly relevant to the fructose–6–phosphate:fructose–1,6–bisphosphate cycle in
which AMP and fructose–2,6–bisphosphate activate phosphofructokinase but
inhibit fructose–1,6–bisphosphatase.
Stein and Blum240 pointed out that the basis of the theory was paradox-
ical since it was derived on the assumption that the concentrations of the
226 7 Control structures in metabolism

intermediates do not change when the flux is altered, yet

• The enzyme 2a dominates the flux control of the pathway, i.e. the
pathway flux always responds fully to an increase in the activity of this
enzyme by matching increases in the rate of enzymes 1 and 3. This
requires that the concentration of S1 falls and that of S2 rises.
• When the flux is increased through 2a, there is no compensating alter-
ation in the flux through the other limb of the cycle, 2b, which requires
the assumption of no change in the intermediate concentrations to be
true.
They performed three separate computer simulations of open substrate cycles
(Fig. 7.11 f) based on each of those found in carbohydrate metabolism and
concluded that the gain in sensitivity of control was small in physiologically
feasible conditions and that other functions for the cycles (to be discussed
later) seemed more likely.
Crabtree and Newsholme reanalyzed their proposal more rigorously with
their own form of metabolic sensitivity analysis54, 55 and found that their
original equation over–states the gain in sensitivity but that there is still some
gain. Herbert Sauro and I69 used Metabolic Control Analysis to study the
same question (see Appendix 1 to this chapter, Control Analysis of substrate
cycles for details). We calculated the gain in sensitivity by asking how much
the flux control coefficient of the cycle enzyme 2a is increased by cycling,
relative to an equivalent pathway with the same net flux and intermediate
concentrations. This factor, r, depends on the elasticities of the enzymes and
the cycle fluxes in the following way:
1
r= (7.7)
v2b (ε11 ε32 − ε11 ε2b 3 2b
2 − ε 2 ε1 )
1−
v2a (ε11 ε32 − ε11 ε2a 3 2a
2 − ε 2 ε1 )

With normal substrate kinetics, where elasticities with respect to substrates

are ≥ 0, and with respect to products are ≤ 0, the cycle increases the control
coefficient if:
ε2b
2 ε2b
1
+ <1 (7.8)
ε32 ε11
Since each of the two terms on the left of this equation must be less than
1 for the condition to be true, this show that the output enzyme 3 must be
less saturated with S2 than is enzyme 2b, and that the product inhibition of
enzyme 1 by S1 must be stronger than that of 2b. For r to reach Newsholme
and Crabtree’s maximum value of (1 + v2b /v1 ), it is also necessary that:
¯ 2a ¯ ¯ 2a ¯
¯ ε2 ¯ ¯ ¯
¯ ¯ + ¯ ε1 ¯ ¿ 1 (7.9)
¯ ε3 ¯ ¯ ε1 ¯
2 1

(The bars around the terms indicate that only their magnitude is considered;
any negative signs, such as in the product elasticity ε11 are ignored.) Here
7.3 Substrate or ‘futile’ cycles 227

the first term on the left shows that the product inhibition of 2a must be
much weaker than the elasticity of 3 with respect to its substrate, and the
second that the substrate elasticity of 2a must be much smaller in magnitude
than the product inhibition of 1 (a condition most likely to be met when 2a
is saturated, as noted by Crabtree54 ). These same conditions also maximize
the magnitude of the negative flux control coefficient of the reverse limb of
the cycle, 2b.
In conclusion, the conditions under which substrate cycles can generate
greatly increased flux control coefficients (which are necessary to give in-
creased responsiveness to effectors of the cycle enzymes) are quite restric-
tive. Also, an effector that acts to increase the net flux through the cycle
also reduces the flux control coefficients of the cycle enzymes, and therefore
its own response coefficient. At present, there is no indisputable example
of the phenomenon; the best candidate is the fructose–6–phosphate:fructose–
1,6–bisphosphate cycle in mammalian liver glycolysis/gluconeogenesis, but it
is not certain that the cycling rates are high enough and the relevant elasticity
values are unknown.

7.3.6 Switching the direction of flux

Fig. 7.12 shows that the substrate cycles of carbohydrate metabolism are
of the type termed an ‘open futile cycle’ (Fig. 7.11f) and occur at complex
metabolic ‘crossroads’. The computer simulations carried out by Stein and
Blum240 and mentioned in the previous section certainly suggested that the
cycles aided the switching of flux in bidirectional pathways as the relative flows
changed in the inputs and outputs or modifiers acted on the cycle enzymes.
Sel’kov and his colleagues196 also concluded that a substrate cycle can function
as a switch, with a minimum of cycling near the crossover in the direction of
the net flux through the cycle. Since the direction of flux does switch direction
in these pathways in vivo, the proposal seems relatively uncontroversial. What
is difficult to quantify is the cost:benefit analysis. The costs are difficult to
measure accurately, but in liver cells each of the cycles probably only uses a
few percent of the cell’s total energy production.10, 146, 190 There is no measure
for the improvement in switching obtained with a given degree of cycling.

7.3.7 Buffering metabolite concentrations

Cahill’s original demonstration26 of the glucose:glucose–6–phosphate cycle led
him to propose that it helped to regulate the blood glucose concentration.
Newsholme and Crabtree suggested that the triacylglycerol:fatty acid cycle in
adipose tissue could help to buffer the levels of fatty acids in the blood during
lipolysis.168 At one time it was suggested that the cycling of Ca2+ between
cytosol and mitochondria could buffer the cytosolic Ca2+ concentration and
allow sensitive control12, 112, 174 though nowadays the role of the mitochondria
is not thought to be so significant.155, 183 The cycle involves uptake of Ca2+
by the mitochondrial matrix driven by the membrane potential component of
228 7 Control structures in metabolism

the protonmotive force and export from the matrix by a Ca2+ /2H+ antiport
(i.e. exchanger) driven by the pH gradient component of the protonmotive
force, so both components are exergonic and effectively unidirectional.
The problem with the concept that buffering is a specific property of sub-
strate cycles is revealed by considering the Ca2+ cycle, even though this is
strictly a conserved cycle (type Fig. 7.11c) if Ca2+ exchange with the en-
doplasmic reticulum12 is ignored. If we take the cytoplasmic Ca2+ , then
its steady–state value is that which makes the rate of the supply process
(mitochondrial export) equal the rate of the demand process (mitochondrial
import). This is true even though the supply and demand processes meet
in the mitochondrial matrix to form a cycle, and is in fact a general state-
ment about any metabolite at steady state. Any buffering therefore reflects
the dynamic equilibrium, or non–equilibrium steady state, determined by the
specific kinetics of the process. Several conclusions follow:
• Any buffering effect that does exist cannot be a specific function of a
substrate cycle.
• There is no specific effect of cycling rate, since the rates of the demand
and supply processes can be multiplied by any arbitary factor and the
same steady state concentration is obtained.
• The question of homoeostasis of the concentration can be addressed
by the supply–demand analysis of Hofmeyr and Cornish–Bowden that I
described at the beginning of this Chapter in section 7.1. The sensitivity
of the control is represented by the concentration control coefficient; the
kinetic features of the pathway, but not its flux, will determine whether
the sensitivity is high or low.
• If buffering is meant to imply the tendency of the system to return
to the same steady state after an arbitary perturbation of the concen-
trations, then this is a common property of metabolic systems with
a dynamically–stable steady–state, and again the cycling is irrelevant.
However, the rate of return to the steady state after a perturbation is
related to the turnover rate of the intermediates, and will be shorter if
the flux is high.
Another demonstration that substrate cycling does not affect the ho-
moeostasis of metabolites follows from the Metabolic Control Analysis ap-
plied to the substrate cycle of Fig. 7.13 in Section 7.3.5 and Appendix 1 of
this Chapter. I have used the same technique that shows that the flux control
coefficients depend on the cycling rate to show that the concentration control
coefficients do not.
Is there no truth in the concept that the glucose:glucose–6–phosphate cy-
cle helps to regulate blood glucose concentration? I believe there is, and the
computer simulations carried out by Stein and Blum240 suggested it worked.
The explanation cannot be given just in terms of the effect arising from the
cycle. Firstly, if blood glucose was determined solely by the balance of supply
7.4 Regulation by covalent modification of enzymes 229

by absorption from the intestine and demand by the tissues, it would exhibit
poor homoeostasis because the inhibition of absorption by blood glucose is
too weak (Section 7.1). Through the glucose:glucose–6–phosphate cycle, blood
glucose is connected to other supply and demand processes, and the home-
ostatic characteristics are now those of the total system. The homoeostasis
is therefore related to the switching properties of the cycle, whereby it can
bidirectionally connect the input and output processes for blood glucose to
liver carbohydrate metabolism. As noted by many authors,111, 171 the steady–
state level of blood glucose is close to the calculated value at which the rates
of liver hexokinase IV and glucose–6–phosphatase would be equal and there
would be no net flux through the cycle. This is because the half–saturation
value of the sigmoidal hexokinase IV rate curve is in the 1–10 mM range
characteristic of mammalian blood glucose levels. When blood glucose level
changes, the substrate cycle automatically operates as a switch with no need
for any other mechanism. Thus a rise in blood glucose level, which rapidly
equilibrates with liver cytosol glucose causes an increases in the rate of phos-
phorylation but has no significant effect on the glucose–6–phosphatase rate
(since the demand and supply system for glucose–6–phosphate in the liver
is fairly effective at ensuring homoeostasis of its concentration), so the cycle
switches to cause net phosphorylation and glucose uptake by the liver. The
reverse happens when blood glucose level falls and the rate of phosphorylation
drops below that of the phosphatase. Thus the homoeostasis results from the
kinetic balance of supply and demand at a complex metabolic crossroads, in
which the switching property of the cycle is a relevant factor. There is no
need for (and no point in) invoking a specific buffering action of the cycle in
itself.

7.4 Regulation by covalent modification of enzymes

Much of this book so far has dealt with the extent to which particular en-
zymes can control metabolic fluxes. Only indirectly does this address the
question of what controls metabolism, since it identifies sites at which control
might be exerted most effectively, and predicts to some extent the effect that
will be produced if the enzyme’s activity is changed. The response coefficient
(Chapter 5.4) indicates the control that could be exerted via an allosteric
enzyme by an allosteric effector external to the pathway; the implication is
that this gives a viable method of control, but I can give no example where
this has been demonstrated to be quantitatively important in the same way
that Keith Snell and I showed that serine could control its rate of synthesis
(Chapter 6.3.4.2). Induction and repression of enzyme synthesis, for exam-
ple in response to nutritional or hormonal signals, are mechanisms of control
that can be interpreted in terms of the flux control coefficients of the enzymes
involved, and examples of this were given in Chapter 6.1.2. Yet the meta-
bolism of organisms shows substantial and specific responses to signals such
as environmental factors, hormones, growth factors and nerve stimulation
230 7 Control structures in metabolism

that precede any effects these signals may have on synthesis and degradation
of enzymes, and without direct interactions between the signal molecules and
specific pathway enzymes. This is clearly one of the most important aspects of
metabolic control. In this section I will present examples that show that many
of these controls are exerted via mechanisms that involve post–translational,
covalent modifications of proteins that affect their enzymic activity. Since
the processes involve changes in enzymic activities, they can be interpreted
in Metabolic Control Analysis through flux control coefficients.

7.4.1 Irreversible and cyclic cascades

There are many types of post–translational modifications of enzymic proteins
that have effects on their activity. Some of these are irreversible and not
primarily mechanisms of control. For example, proteolytic cleavage of inactive
precursors (or zymogens) can be a mechanism for safely synthesizing and
storing a potentially hazardous product such as a digestive enzyme until it is
released at its site of action. Nevertheless, there is the potential for control
in such mechanisms, as illustrated by blood clotting. Here there is a cascade
of zymogens in which each factor, upon activation, proteolytically cleaves the
next zymogen in the sequence, until prothrombin is cleaved to thrombin, which
cleaves soluble fibrinogen to yield the clot–forming fibrin fibres. The sequence
exhibits great amplification through using catalysts that create catalysts, and
converts the minute intitating signal given by tissue injury into a macroscopic
response within a few tens of seconds. However, for the enzymes involved,
it is a once–only mechanism as the clotting can only be stopped by their
proteolytic inactivation.
Another irreversible modification of enzymes, this time of intracellular en-
zymes in metabolic pathways, is the attachment of ubiquitin.102 This marks
the enzyme for energy–dependent degradation by ubiquitin–dependent pro-
teases and may well play a role in metabolic control by adjusting enzyme
amounts (Chapter 1.4.1), since different enzymes in the same metabolic path-
way often have very different turnover rates in eukaryotic cells. Much remains
to be learnt about the role of enzyme degradation in the control of metabol-
ism.
The covalent modifications that play the greatest role in metabolic con-
trol are reversible or cyclic, so that the enzyme is interconverted between
two forms that differ in activity, either because of effects on the kinetics with
respect to substrates or of altered sensitivity to effectors. The first exam-
ple of such a process was assembled over about 20 years from the combined
efforts of several groups of researchers139, 244 studying the enzyme phospho-
rylase, which breaks down glycogen to form glucose–1–phosphate. The work
of the Coris in the 1940s had shown that there were two forms of phosphory-
lase that differed in activity: phosphorylase a was the more active, whereas
phosphorylase b required AMP activation and was inhibited by glucose–6–
phosphate. Furthermore, these two forms were interconverted in some way.
In 1950, Sutherland discovered that the activation of phosphorylase in liver
7.4 Regulation by covalent modification of enzymes 231

brought about by the hormone adrenaline was mediated by a heat–stable

factor, though it was not until 1959 that this was identified as cyclic AMP
(Fig. 7.14). Meanwhile, Edwin Krebs’ group was studying phosphorylase in
muscle, and in 1955 they discovered that the interconversion was phospho-
rylation of phosphorylase b by ATP, catalyzed by an enzyme, phosphorylase
kinase. In the same year, Sutherland and Wosilait showed that inactivation
of phosphorylase in the liver involved formation of phosphorylase b by the
action of a phosphatase. Krebs’ laboratory went on to establish by 1959 that
phosphorylase kinase itself was controlled by phosphorylation and dephospho-
rylation, showing that the control of a metabolic enzyme could be organised
in a cacscade fashion. Sutherland’s group went on to discover the formation
of cyclic AMP by adenylate cyclase in 1962, and in 1965 and 1966 put forward
the concept of cyclic AMP as a second messenger or intracellular mediator of
(extracellular) hormone action.
Following these initial discoveries, it has become clear that there are several
types of covalent modification in addition to phosphorylation, and that there
are other signals and second messengers apart from cyclic AMP.

7.4.2 Types of reversible modification

The description of any reversible covalent modification system involves a tar-
get protein, an enzyme for modifying it and one for reversing the modification,
and mechanisms (such as allosteric effectors or further covalent modification
systems) for controlling the balance of the activities of the two interconvert-
ing enzymes. The complexities of classifying the large number of systems now
known will become increasingly apparent as this section progresses, but stem
from the multiplicity of effects and the many interactions between systems.
Thus:

• some modifying enzymes are themselves targets for modification (as

implied above), thus forming cascade systems;

• some targets undergo multiple modifications, either of the same type

but at different sites by different enzymes, or of different types;

• different signals can have overlapping mechanisms and share a modifying

cycle, and

• one signal can interfere with or potentiate the actions of another by

cross–reactions between the modifying cycles or by interactions between
the modified sites on the target.

The principal types of covalent modification known to reversibly affect

enzymes are summarized in Table 7.1, based on data in reference [226]. Phos-
phorylation is the most common modification in eukaryotic cells and is carried
out by several different systems, as will be described below. It is probable
that there are many more targets than have been identified currently, since
232 7 Control structures in metabolism

Table 7.1: Types of covalent modification of enzymes

Type of Donor Amino acids No. of Comments

modification modified targets
(approx)
Phosphorylation ATP, GTP Ser, Thr, Tyr 90 More prevalent in eukaryotes
OH–Lys than prokaryotes.

ADP–ribosylation NAD+ Arg, Glu, Lys 20

Nucleotidylation ATP, UTP Tyr, Ser 4 In prokaryotes.

Methylation S–adenosyl– Asp, Glu, Lys, 14 Prokaryotes and eukaryotes,

methionine His, Gln but less prevalent in the
latter than phosphorylation.

Based on data in reference [226], surveyed in 1986.

as many as 1 in 4 polypeptide chains have been estimated to be phosphory-

lated in mouse lymphoma cells37 (though this average figure takes no account
of the multiple phosphorylation of some chains, nor the relative abundance
of different proteins). In contrast, levels of phosphorylation are much lower
in prokaryotic cells,75 but even so, a significant role for phosphorylation in
metabolic control is being discovered.205 Although there are relatively few
identified methylation targets in eukaryotic cells, and it is about 50–fold less
extensive than phosphorylation, electrophoresis of radiolabelled mouse lym-
phoma cell extracts showed at least 24 different methylated protein bands. 37
To illustrate the reversible modification systems further, I will only deal with
phosphorylation and, to a lesser extent, nucleotidylation.

7.4.3 Phosphorylation
It is impossible to know where to start to produce a simple, coherent descrip-
tion of such a complex, interacting and interlocking network as the eukaryotic
phosphorylation systems. The difficulty is illustrated by the general history
of discovery of a system. This usually stems from the observation that a par-
ticular target enzyme undergoes phosphorylation. In turn, this leads to the
identification of a kinase that is responsible, and which is known initially by
the name of its target (even when this leads to comic names like phosphory-
lase kinase kinase for the kinase that phosphorylates phosphorylase kinase).
Generally, it later emerges that this kinase is similar, or even identical, to
one known by a different name because it has been shown to phosphorylate
7.4 Regulation by covalent modification of enzymes 233

a different target, and the kinase is named after the signal that stimulates it
to act rather than by its target.
The position is even more complicated with the protein phosphatases, since
these do not show unique associations with target enzymes, specific protein
kinases or signal systems, though they are affected by all of these.
Furthermore, although phosphorylation of an enzyme can lead to activity
changes, many enzymes are subject to multiple phosphorylations at differ-
ent sites, and some of the sites appear to affect ease of phosphorylation or
dephosphorylation at other sites rather than to directly affect activity.

7.4.3.1 Kinases and their control signals

Table 7.2: Protein serine and threonine kinases in metabolism.

Signal Enzyme

Cyclic nucleotides cyclic AMP–dependent protein kinases (Types I & II)

cyclic GMP–dependent protein kinase

Ca2+ and calmodulin Ca2+ /calmodulin multiprotein kinase

phosphorylase kinase/glycogen synthase kinase 2

diacylglycerol protein kinase C

AMP AMP–activated kinase

Metabolic intermediates and Many target–specific kinases, including:

other ‘local’ effectors pyruvate dehydrogenase kinase;
branched–chain ketoacid dehydrogenase kinase, and
glycogen synthase kinases 3 & 4

Based on the classification by Krebs139 with the addition of the AMP–

activated protein kinase28, 29 characterized by Grahame Hardie’s group in
Dundee.

The major groups of protein kinases that affect intermediary metabolism

are summarized in Table 7.2, arranged by control signal. Thus for simplicity,
phosphorylation of structural proteins like histones, enzymes of protein syn-
thesis and muscle contractile apparatus have been ignored. Tyrosine protein
kinases have not been considered here since these activities seem to be as-
sociated with the receptors for certain hormones such as insulin and growth
factors and, important as they are, the possible links between phosphorylation
and any metabolic effects only now beginning to be unravelled.145
234 7 Control structures in metabolism

NH2

C
N
N C

CH
HC C
N
N
5’ CH2 O

C C
O 1’
H H

O P C C
3’

O- O O
OH

Fig. 7.14: Cyclic AMP

Like all adenine nucleotides, cyclic AMP has an adenine ring attached to the 1 0 carbon
of the ribose sugar. AMP itself has a phosphate on the 5 0 carbon atom of the ribose
ring, but in 30 ,50 –cyclic AMP, this phosphate is bonded to both the 3 0 and 50 positions.

7.4.3.1.1 Cyclic AMP In section 7.4.1 I mentioned the discovery by Suther-

land’s group that cyclic AMP (Fig. 7.14) is an intermediary in the hormonal
activation of phosphorylase. They later showed that it was produced by the
enzyme adenylate cyclase, located on the cytoplasmic surface of the plasma
membrane of eukaryotic cells:
ATP −→ cyclic AMP + pyrophosphate
The concentrations of cyclic AMP are generally low, in the region of 0.1–
1 µM. It is hydrolyzed to AMP by a variety of phosphodiesterase enzymes
that differ in their intracellular location, their Km for cyclic AMP (broadly
classified as ‘high’ and ‘low’), and the physiological effectors and drugs that
modulate their activities. Cyclic AMP is not a metabolic intermediate for
the synthesis of any other compound; its function appears to be solely as a
signal that controls enzyme activities. Since adenylate cyclase is stimulated
by hormones that bind to receptors on the outer surface of cells, causing the
intracellular level of cyclic AMP to vary in a hormone–dependent fashion,
Sutherland proposed the term second messenger to indicate its role in the
intracellular relay of a hormonal message.
The hormonal activation of adenylate cyclase is one specific instance of a
trans–membrane signal transduction pathway, many of the details of which
were determined by Martin Rodbell and his colleagues (see ref. [199]). This
involves the interactions of three classes of components in the plasma mem-
7.4 Regulation by covalent modification of enzymes 235

brane. There are specific receptors at the extracellular surface that can bind
stimulatory or inhibitory signal molecules (Rs and Ri respectively). The in-
teraction of these receptors with the catalytic subunit is mediated by guanyl–
binding regulatory proteins, again of stimulatory and inhibitory types (G s
and Gi ). The G–proteins bind GTP when they interact with the receptor–
hormone complex, and the G–protein complex with GTP in turn interacts
with adenylate cyclase (or other target protein in the general case). The
action of the G–protein terminates when it hydrolyzes its bound GTP. Hor-
mones that act on adenylate cyclase through a stimulatory G–protein in-
clude adrenaline (epinephrin), adrenocorticotropin, thyroid–stimulating hor-
mone (TSH), luteinizing hormone (LH), secretin and glucagon, not necessarily
all in the same tissue though those mentioned do all stimulate adenylate cy-
clase in adipose tissue, eventually causing lipolysis. The action of adrenaline
is complex because it acts on adenylate cyclase via a class of receptors called
β–adrenergic receptors, found particularly in muscle, but via α–adrenergic
receptors (for example, in liver) it acts to raise intracellular Ca2+ concentra-
tions. Adenosine inhibits adenylate cyclase in adipose tissue via an inhibitory
G–protein.
Stimulation of adenylate cyclase leads to a rise in the intracellular concen-
tration of cyclic AMP, the principal, perhaps only, effect of which is stimula-
tion of cyclic AMP–dependent protein kinase (protein kinase A). The amount
by which the cyclic AMP concentration increases must depend on the activity
and kinetics of the phosphodiesterase enzymes present in the cells, but since
there are phosphodiesterases with Km values above the intracellular levels of
cyclic AMP, it is likely that the concentration changes are at best propor-
tional to the stimulation of the cyclase. Protein kinase A occurs as a set
of isoenzymes, but there are two basic classes: Types I and II.13 Both are
tetrameric proteins formed of two regulatory subunits (R) and two catalytic
subunits (C) in the inactive state. Activation involves binding of four cyclic
AMP molecules by the regulatory subunits and release of free, active catalytic
subunits:
R2 C2 + 4 cyclic AMP *
) R2 (cyclic AMP)4 + 2C
The two sites for cyclic AMP per regulatory subunit interact cooperatively
so that the Hill coefficient for kinase activation with respect to cyclic AMP is
about 1.6.13 The difference between the Type I and Type II kinases is that
the latter undergo autophosphorylation of the regulatory subunit, though the
physiological significance of this, and the respective physiological roles of the
isoenzymes are not very clear.
There is a large number of targets for phosphorylation by protein kinase
A. Not all protein serine or threonine residues can be phosphorylated, nor all
proteins, so there must be specific sites recognised by the kinase. A common
sequence at phosphorylation sites involves two basic amino acids (lysine or
arginine) to the N–terminal side, for example arg-arg-X-ser and arg-lys-X-ser.
The effects on the target enzyme can be inhibitory or activatory; in general
where antagonistic enzymes in potential substrate cycles are phosphorylated,
236 7 Control structures in metabolism

one is activated and one inhibited.42 Glycogen phosphorylase and glycogen

synthase are one such pair, and the evidence is that substrate cycling is not
significant between glucose–1–phosphate and glycogen because these two en-
zymes are not both in their most active forms simultaneously. Philip Cohen, 42
whose group in Dundee has carried out much of the work on the phosphory-
lation of glycogen synthase and on the phosphoprotein phosphatases has also
suggested that enzymes catalyzing biodegradation are generally activated by
phosphorylation and biosynthetic enzymes are inhibited. Phosphorylase, tri-
acylglycerol lipase (in adipose tissue) and phenylalanine–4–monooxygenase
(in liver) are clearly catabolic enzymes that are activated by phosphorylation.
Anabolic enzymes that are inhibited by phosphorylation include glycogen syn-
thase and acetyl CoA carboxylase (the first enzyme of lipid synthesis). Other
instance are less clear cut because of tissue differences both in metabolism
and the effects of hormones. Thus whereas in muscle tissue adrenaline pro-
motes glycolysis, in liver, it and glucagon promote gluconeogenesis and here
the glycolytic enzyme pyruvate kinase is inhibited by phosphorylation.
Another example of tissue–dependent effects involves phosphofructokinase,
which in liver forms a substrate cycle with fructose bisphosphatase (Sec-
tion 7.3.2, p. 223). In mammalian heart164 and adipose tissue,206 phospho-
fructokinase is phosphorylated after adrenaline stimulation (perhaps not by
protein kinase A) to give an enzyme that is less sensitive to the action of
its inhibitors such as ATP and more sensitive to its activators, in particular
fructose–2,6–bisphosphate. As described earlier, fructose–2,6–bisphosphate is
a third messenger of hormone action, since it is purely a signal metabolite
and its production and consumption by the bifunctional phosphofructo–2–
kinase and fructose–2,6–bisphosphatase is affected by cyclic AMP–dependent
phosphorylation. However, in heart muscle, it seems that adrenaline causes
an increase in the concentration of fructose–2,6–bisphosphate by activation of
the 2–kinase,164 which is the exact opposite of its effect in the liver. Thus in
glucose–consuming tissues, it appears that, directly or indirectly, hormone–
sensitive phosphorylation activates phosphofructokinase. In liver, glucagon
initiates cyclic AMP–dependent phosphorylation of the bifunctional enzyme,
causing inhibition of the 2–kinase, activation of the phosphatase and a fall in
fructose–2,6–bisphosphate that has the effect of inhibiting phosphofructoki-
nase and activating fructose–1,6–bisphosphatase.
Even greater complexity in the cyclic AMP–dependent systems will emerge
in the following sections, partly when we encounter the role of the protein
phosphatases, but also as the interactions with other covalent modification
systems become apparent.

7.4.3.1.2 Ca2+ and calmodulin Even though calcium is prevalent in the

environment and is present in our blood at 2 mM, the levels of free Ca2+
in cells are maintained at an extraordinarily low average level of about 10−8
M. Undoubtedly this is necessary in part to avoid the formation of insoluble
complexes with the many compounds in the cell that contain phosphate. How-
7.4 Regulation by covalent modification of enzymes 237

ever, eukaryotic cells have adopted Ca2+ as a second messenger in response

to a wide variety of external and internal regulatory signals acting on a whole
range of cellular activities as well as metabolism. In all eukaryotes, many of
these functions are mediated through the regulatory protein calmodulin, first
discovered in 1970 as a Ca2+ –dependent activator of the enzyme cyclic AMP
phosphodiesterase by Cheung, Kakiuchi and Yamazaki.45, 266
Unlike cyclic AMP, which acts exclusively through its protein kinase, the
Ca2+ –calmodulin system acts through multiple mechanisms, both general and
specific, by interaction with particular enzymes of metabolism. The common
element is the nature of the interaction with calmodulin. Calmodulin is a sin-
gle polypeptide chain with four calcium–binding domains that have differing
calcium affinities; its dissociation constants range from 1 to 100 µM. Calmod-
ulin binds to the enzymes it affects, but with substantially higher affinity when
it has Ca2+ bound than when it is Ca2+ –free; by energy conservation prin-
ciples, this means that formation of the calmodulin–enzyme complex causes
an increase in the affinity of the calmodulin in the complex for Ca2+ so that
the dissociation constants are µM or less. At the µM levels of calmodulin
in the cell, many of the target enzymes will not complex with calmodulin in
the absence of Ca2+ , but do in its presence. However, in two Ca2+ –activated
enzymes, phosphorylase kinase and phosphoprotein phosphatase 2B, calmod-
ulin is present as a permanent subunit in the structure. In virtually all known
cases, enzymes that interact with Ca2+ –calmodulin are activated.
Ca2+ –calmodulin can stimulate phosphorylation via a calmodulin–dependent
multiprotein kinase,45 which was previously thought to be a number of sep-
arate activities. The target enzymes that this phosphorylates include glyco-
gen synthase in muscle and liver (causing deactivation) and tryptophan–5–
monooxygenase and tyrosine–3–monooxygenase in brain (where they are parts
of the pathways to the neurotransmitters dopamine and serotonin). Other
enzymes can be phosphorylated by this kinase in vitro, but often it is not
clear whether this is physiologically significant. Increased Ca2+ levels cause
glycogen phosphorylase in both muscle and liver to be activated by phospho-
rylation by phosphorylase kinase, which contains calmodulin as a subunit. In
muscle, this links the activation of glycogen breakdown to muscle contraction,
since Ca2+ is released into muscle cytoplasm from the sarcoplasmic reticulum
to activate the contractile apparatus when a nerve impulse arrives. (Ca2+ –
calmodulin and troponin C, which is related to calmodulin, also have a range
of effects on the contractile apparatus and on Ca2+ transport that are not of
immediate relevance here.)
Apart from the effects that Ca2+ –calmodulin has on its own second mes-
senger system through its activation of Ca2+ transport, it also interacts with
phosphorylation induced by the cyclic AMP system. Thus in brain it stim-
ulates adenylate cyclase, but in most other tissues it stimulates one of the
cyclic nucleotide phosphodiesterase isoenzymes. In addition, as will be dis-
cussed below, Ca2+ –calmodulin specifically activates one of the phosphopro-
tein phosphatases (2B) that acts on targets phosphorylated by protein kinase
A. All this has led to the conclusion that Ca2+ can attenuate the actions of
238 7 Control structures in metabolism

cyclic AMP.44

7.4.3.1.3 Diacylglycerol The G–protein signal transduction system also

couples some hormones to membrane phospholipases that break down inosi-
tol phospholipids in the membrane to yield inositol triphosphate and diacyl-
glycerol.199 The inositol triphosphate is involved in promoting the hormone–
induced rise in intracellular Ca2+ , probably by causing its release from internal
stores. The diacylglycerol activates protein kinase C, which also requires Ca 2+
and phospholipid for activity. Protein kinase C catalyzes phosphorylation of a
wide range of proteins and influences a range of cellular phenomena including
signal transduction itself, hormone release, gene expression, cell proliferation
and muscle contraction, but seems less involved in the direct control of in-
termediary metabolism. It may have roles in steroidogenesis in the adrenal
cortex and in glycogen metabolism.

7.4.3.1.4 AMP The kinases I have described so far respond to external

signals; it is therefore particularly significant that David Carling, Grahame
Hardie and their colleagues in Dundee recently discovered that there was a
protein kinase that responded to an internal signal of metabolic state: AMP. 28
There are two fundamental issues about this that need to be answered:

10

1
b

0.1
ANP, mM

0.01

0.001

0.0001
0 1 2 3 4

ATP, mM

Fig. 7.15: The adenylate kinase equilibrium.

The curves show the concentrations on a logarithimic scale of a) ATP, b) ADP and
c) AMP at equilibrium at a given ATP level when the total concentration of adenine
nucleotides is 5 mM and Keq = 1.12.258
7.4 Regulation by covalent modification of enzymes 239

Relative sensitivity
-5
a

-10

-15

-20

-25
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Fractional ATP content

Fig. 7.16: Relative sensitivities of the adenine nucleotides to a change in ATP

level.
The fractional changes in a) AMP and b) ADP in response to a fractional change in ATP
as a result of the adenylate kinase equilibrium, calculated from equations in ref. [69].
The response of ATP is not shown as it is always 1. Note that the relative sensitivity
of ADP changes sign at an ATP fraction of about 0.3, which is the point at which the
sensitivity to AMP becomes nearer zero than that to ATP. Again, AN P = 5 mM and
Keq = 1.12.

• What is the advantage in having a covalent modification system that

responds to an internal metabolite when this is what allosteric enzymes
do?

• How does AMP indicate internal metabolic state?

Covalent modification systems can exhibit different responses to effec-

tors compared with allosteric enzymes, as will be discussed in a later section
(7.4.5). Even so, many of the advantages that covalent modification has as a
transducer of the effect of an external signal present at very low levels would
not seem to be relevant for responding to an intracellular metabolite that
could bind directly to allosteric sites on enzymes. However, I will argue later
(Chapter 8.1) that large changes in a pathway flux require simultaneous and
proportional action on a number of the pathway enzymes to avoid disruption
of cellular homoeostasis. Activation of a kinase that can operate on a number
of enzymes might be one way to achieve this. A common response of several
enzymes to the same allosteric effector might be another possibility, but would
only work well if the allosteric effects were nearly balanced, and might there-
fore be difficult to implement reliably. However, ATP and AMP are possible
240 7 Control structures in metabolism

candidates as allosteric signals that operate on a number of enzymes, as they

do affect a number of enzymes in carbohydrate metabolism for example.
This leads to the second issue: why AMP is involved in allosteric effects
and activation of a protein kinase. It is because AMP amplifies the effects
of small changes in ATP concentration, as was explained by Newsholme &
Start.171 The theory invovles three general observations about cellular meta-
bolism:

• The total quantity of the adenine nucleotides, ATP, ADP and AMP, is
fixed, at least in the short term, by the availability of their common
adenosine moiety.

• Cells regulate their energy metabolism so that the concentration of ATP

exceeds the other two in nearly all circumstances.

• Cells contain adenylate kinase, which catalyzes the reaction:

2 ADP *
) ATP + AMP

and maintains the reaction close to equilibrium at all times.

The effect of this is illustrated in Fig. 7.15, which shows the equilibrium
relationship between the three nucleotides as a function of ATP for an assumed
total of 5 mM adenine nucleotides. Measurement of total adenine nucleotide
levels in cells commonly shows that ATP can be maintained at over 80%
of the total, even in muscle after a 100–fold increase in energy expenditure.
Nevertheless, an increase in ATP utilization in the cell does cause a drop in
its concentration, however slight measurements show this to be. What the
Figure shows is that, in these regions of high relative ATP content, a very
slight fall in ATP is accompanied by a rapid rise in AMP from its very low
level. It is true that ADP concentrations also show an almost as rapid rise,
but ADP has a disadvantage as a signal: any particular ADP concentration
could correspond to two different levels of ATP. The qualitative impression
of the greater sensitivity of AMP levels to an ATP change is confirmed by
my quantitative calculations of the sensitivities69 shown in Fig. 7.16. These
show that sensitivity to AMP change is always the greatest provided ATP
concentration exceeds that of AMP (i.e. for fractional ATP contents greater
than about one third), and that the factor becomes very large indeed when
most of the nucleotide exists as ATP. When corrections are made for bound
ADP and AMP contents in cells, the ATP increases to as much as 98% of
the total free adenine nucleotides.164, 258 This would suggest the possibility
of extremely high sensitivities to ADP and AMP, though the exact values are
hard to predict as the binding of ADP and AMP affects the graphs (without
changing the qualitative picture).
The AMP–dependent protein kinase was discovered as a kinase responsible
for the phosphorylation, and inactivation, of acetyl CoA carboxylase, the first
step in the pathways of fatty acid and cholesterol synthesis.28 It then became
7.4 Regulation by covalent modification of enzymes 241

apparent that this kinase was the same as one that had been identified as re-
sponsible for the phosphorylation (and inactivation) of hydroxymethylglutaryl
CoA reductase, the first step in the branch to cholesterol and other steroids.
It has also been found that the AMP–dependent kinase phosphorylates the
hormone–sensitive triacylglycerol lipase of adipose tissue; in this case, the
phosphorylation does not affect the enzyme’s activity directly, but it prevents
its phosphorylation (and activation) by protein kinase A, thus preventing re-
lease of fatty acids and cholesterol from the fat stores. The AMP–dependent
kinase is activated in two ways by AMP: firstly, AMP is an allosteric activa-
tor of the enzyme, and secondly, AMP makes the enzyme more susceptible
to phosphorylation by a kinase (called kinase kinase!) that remains attached
to the the AMP–dependent kinase throughout nearly all purification steps. 270
Together, the effects cause up to a 50–fold increase in activity of the kinase in
response to AMP, providing further amplification of the effect of a change in
ATP. The metabolic consequence is that fat metabolism is inhibited in condi-
tions where demand for ATP is causing its level to drop, as signalled by the
rise in AMP.
The activity of the AMP–dependent protein kinase is also stimulated by
very low concentrations of fatty acyl CoA esters, so it probably also limits the
synthesis of fatty acids and cholesterol and the hydrolysis of triacylglycerols
under conditions where fatty acyl CoA is already accumulating. On this basis,
it could also be regarded as an example of the next class of protein kinases.

7.4.3.1.5 Metabolic intermediates The pyruvate dehydrogenase enzyme

complex is present in the mitochondrial matrix of eukaryotes and oxidizes
pyruvate to acetyl CoA. There is also a similar branched–chain α–keto acid
dehydrogenase complex that oxidizes the carbon skeletons from the essential
branched chain amino acids (valine, leucine and isoleucine). Both complexes
are subject to inactivation by phosphorylation by specific kinases that are
tightly bound to the enzyme complexes. The specific phosphatases that ac-
tivate the enzymes also seem to be tightly bound to the complexes. The
kinase activities are allosterically activated by metabolites; for example, in
the case of pyruvate dehydrogenase kinase, the phosphorylating activity is
stimulated by metabolites that will increase in concentration when mitochon-
drial oxidative capacity is being stretched, such as acetyl CoA and NADH,
and inhibited by metabolites that indicate a need for more oxidative meta-
bolism, such as ADP. (ADP is present at higher levels in the mitochondria
than in the cytoplasm.) The phosphatase is stimulated by Ca2+ in the mito-
chondrial matrix. This may enable the phosphorylation state to respond to
this second messenger, since the mitochondria accumulate Ca2+ relative to
the cytosol, so the dehydrogenase complexes may not be solely regulated by
metabolic intermediates.
242 7 Control structures in metabolism

Table 7.3: Phosphoprotein phosphatases.

Type Principal targets Specific properties

1 glycogen metabolism, muscle con- 1. Inhibited by phosphorylated In-

tractility hibitor–1.
2. Particulate, e.g. attached to
glycogen.
3. Inhibited by okadaic acid

2A glycolysis, gluconeogenesis, aro- 1. Cytosolic

matic amino acid catabolism, fatty 2. Inhibited by okadaic acid.
acid synthesis.

2B Protein kinases, protein phos- 1. Activity dependent

phatases and Inhibitor–1 that have on Ca2+ and stimulated 10–fold by
been phosphorylated by protein ki- Ca2+ /calmodulin.
nase A. 2. Poor action on metabolic en-
zymes.
3. May cause Ca2+ attenuation of
the effects of cyclic AMP.

2C AMP–dependent protein kinase, 1. Cytosolic.

and possibly its targets in fatty acid 2. Requires Mg2+ for activity.
and cholesterol metabolism.

The table is largely based on information from references [42–44].

7.4.3.2 Phosphatases

The action of protein kinases is reversed by the action of phosphoprotein

phosphatases. These generally have very broad specificities when tested in
vitro, which has made classification and determination of their physiological
functions difficult. However, Philip Cohen’s group in Dundee have assigned
the phosphatases into four main groups42–44 that are generally present in
vertebrate and invertebrate animals, and probably in most eukaryotes. The
main properties of the four groups are summarized in Table 7.3. (The in-
hibitor okadaic acid that is referred to there is useful in distinguishing the
physiological roles of the phosphatases and is a toxin produced by marine
dinoflagellates. It is also known for accumulating in the shellfish that feed on
them and causing diarrhetic shellfish poisoning.) The Table shows that there
is no simple correspondence between the kinases and the phosphatases. There
are some important links though:
7.4 Regulation by covalent modification of enzymes 243

• Inhibitor–1, which specifically inhibits protein phosphatase 1, only does

so when it has been phosphorylated by protein kinase A. Thus activation
of this kinase can thereby lead to a reduction in the dephosphorylating
activity, so augmenting the increase in phosphorylation that it promotes
directly by its kinase action.
• Phosphoprotein phosphatase 2B, which is activated by Ca2+ and Ca2+ /cal-
modulin, acts specifically on the kinases, phosphatases and Inhibitor–1
that have been phosphorylated by protein kinase A. It therefore forms
another link between the cyclic AMP and Ca2+ systems. Like the effects
of Ca2+ on cyclic AMP phosphodiesterase, the result is to cause Ca2+
attenuation of the action of cyclic AMP.42
In addition, the action of phosphatase 2B in dephosphorylating (and thereby
inactivating) Inhibitor–1, will activate phosphatase 1 (by relieving its inhibi-
tion) and may be an example of a phosphatase ‘cascade’.42
Protein phosphatase 1 is found in a variety of forms, mainly associated
with cellular structures. For example, in muscle cells it is attached to glyco-
gen, sarcoplasmic reticulum and myofibrils. The glycogen–bound form con-
sists of a complex of the phosphatase with a glycogen–binding subunit. The
formation of this complex causes a specific 5–10 fold stimulation in the activ-
ity of the phosphatase towards phosphorylase and glycogen synthase, but not
towards other substrates. Since the phosphatase 1 bound to myofibrils has
enhanced activity against phosphorylated myosin light chains, it seems that
the phosphatase is targetted towards its substrates by the binding proteins.

7.4.3.3 Phosphorylation cascades

The activation of glycogen phosphorylase, shown in Fig. 7.17, illustrates a
covalent modification cascade system where there is more than one stage to
the process. Phosphorylase b, the unphosphorylated form, only shows sig-
nificant enzymic activity towards glycogen when allosterically activated by
AMP. Phosphorylase a is active without the need for activation by AMP. The
phosphorylation of phosphorylase is dependent upon the action of a specific
phosphorylase kinase. This itself exists in a less active, dephosphorylated
form and an active, multiply–phosphorylated form. Both forms are also acti-
vated directly by Ca2+ through the calmodulin–like subunit. In muscle this
enables phosphorylase activation to occur within about a second of sustained
muscular contractions because of the raised cytosolic Ca2+ levels. Phosphory-
lase kinase is phosphorylated by protein kinase A, which in turn is activated
by hormones that stimulate the synthesis of cyclic AMP, such as adrenaline.
The activation of phosphorylase by adrenaline in muscle is slower than by
contraction.
A further element to the cascade that has been shown in Fig. 7.17 has
details that are tissue–specific. The phosphorylated forms of phosphorylase
kinase and phosphorylase are primarily dephosphorylated by protein phos-
phatase 1. In muscle, this phosphatase is inhibited by the phosphorylated
244 7 Control structures in metabolism

hormone + receptor *
) hormone–receptor complex
...
.. ⊕
..... PDE.
ATP ...................... cAMP ........................AMP
...
.
.....⊕
protein kinase A
...
...
... ... ...
.. .. .
... ... ATP..
........... .....⊕ ..........ADP ...
... ... ..... .
..
..
.
..
...
...
.. ..
................. .
.. .. ...... ..........
.. ... ........ ........
... ... ..... ....
.. .. phos kinase phos kinase–P
... ... ⊕
... ....... ... ...
ATP .
... ADP .......... .........
..
... ... ...................................... .....
.. ....................................... .
... ...
.
. ..
......
. . ..
....
........... .. .
.... .. . . ... ... .
...
.......
........ ........ ....... ... ...... ...
..
... ..... ...... Pi .. ⊕ H2 O ....
...... .. .
...
... Inhib 1
ª
Inhib 1–P ............................................ ... ...⊕
....
....... . . .
..
.. ..... prot Pase 1 ...
... .........
............. .
....
....
... .......... .
... ... ........... ...
.. ............ .... ............ .
.. ADP
.
... Pi . ⊕ H O ATP
.......... ... .......... .
... ... 2
....................................
. ............
..
... prot Pase 2B
...... ........... ........
... ⊕ .. . ..
. .......
.. phos b phos a
..
ATP
......... ...... ........ADP ... ........
........ ...
. .. . . .
.
.....
..................................
.... .
...
...
...
............
. . . .. .. .
.... ....
.. . . . . ................ .
...
... .........
........ ....... Pi ....
. ........... ... ............... .
..
..... .... ª
............................................. . . .. . ⊕ H O ..
ª
GBS–P ...................................................................................................................
...... ... 2 .
GBS . ..⊕
......
........ ... ...... prot Pase 1
. ...
............ .....
................................. ..
............... ..... ............... ....
Pi ... ⊕ H2 O .....
. glycogen + Pi ........................... G1P
prot Pase 2B

Fig. 7.17: The phosphorylase activation cacsade in muscle.

Abbreviations: PDE, cyclic AMP phosphodiesterase; prot Pase, protein phosphatase;
Inhib 1, inhibitor 1; phos, phosphorylase, Pi , inorganic phosphate; GBS, glycogen binding
subunit.

form of Inhibitor 1; since Inhibitor 1 is phosphorylated by protein kinase A,

the cyclic AMP signal acts to enhance the phosphorylation and inhibit the
dephosphorylation of both phosphorylase and its kinase. In addition, the
inhibition of dephosphorylation is further strengthened because protein ki-
nase A phosphorylates the glycogen binding subunit that both holds protein
phosphatase 1 to the glycogen particle and prevents Inhibitor 1 acting on it.
The phosphorylation of this subunit releases protein phosphatase 1 from the
glycogen, reduces its activity towards phosphorylase and allows Inhibitor 1 to
bind. Mechanisms with similar effects operate in liver and brain.
7.4 Regulation by covalent modification of enzymes 245

When cyclic AMP levels fall, phosphorylase will tend to return towards
the dephosphorylated b state. This will in part be caused by the reduced
rate of phosphorylation of phosphorylase kinase, Inhibitor 1 and glycogen
binding subunit. Inhibitor 1 and glycogen binding subunit will return more
towards their dephosphorylated state because they are acted on by protein
phosphatase 2B, which has not been inhibited. This relieves the inhibition on
protein phosphatase 1, so that the rates of dephosphorylation of phosphorylase
kinase and phosphorylase are enhanced.
The same mechanism that activates phosphorylase inactivates glycogen
synthase, which in its phosphorylated form is only active when allosterically
stimulated by raised concentrations of glucose–6–phosphate. The dephospho-
rylated form is active without glucose–6–phosphate. Protein kinase A phos-
phorylates glycogen synthase, but the details of the system are much more
involved than for phosphorylase because glycogen synthase undergoes multiple
phosphorylations and a range of protein kinases are responsible. So although
it is certainly the case that cyclic AMP–linked hormones inactivate glycogen
synthase, the relative importance of the different components in physiological
conditions has been difficult to establish. In liver there is an additional link
to the reciprocal relationship between phosphorylase and glycogen synthase:
protein phosphatase 1 is inhibited by phosphorylase a, so that glycogen syn-
thase cannot be dephosphorylated to its active form until virtually all the
phosphorylase has been returned to the relatively inactive b form.
There are a number of points in the phosphorylase cascade where Ca2+ can
act. It directly activates phosphorylase kinase, as mentioned above, regard-
less of phosphorylation state. However, it also activates protein phosphatase
2B, which will lower the amounts of the inhibitory forms of Inhibitor 1 and
glycogen binding subunit so that protein phosphatase 1 becomes more active.
In addition, by activating one of the cyclic AMP phosphodiesterases, Ca 2+
will tend to lower cyclic AMP levels and reduce the degree of activation of
protein kinase A.

7.4.4 Nucleotidylation
Whereas phosphorylase was the mammalian enzyme that led to the discov-
ery of cyclic covalent modification systems, in bacteria it was the glutamine
synthase (properly glutamate–ammonia ligase) of E. coli that furnished the
first example. Glutamine synthase has a central role in nitrogen metabolism,
since approximately 12% of cellular nitrogen comes from the amide nitrogen of
glutamine, and most of the remainder largely from the α–amino group of glu-
tamate rather than directly from ammonia. In bacteria and higher plants, the
net synthesis of glutamate from 2–oxoglutarate and ammonia in turn depends
on a cycle involving both glutamine synthase and glutamate synthase:

glutamate + NH3 + ATP −→ glutamine + ADP + Pi

2–oxoglutarate + glutamine + NADPH −→ 2 glutamate + NADP+
Sum: 2–oxoglutarate + NH3 + NADPH + ATP −→ glutamate +
246 7 Control structures in metabolism

...
..
UTP ......... ......⊕ .........PPi
.................................
. . ..
..
.. .......... ..... ................
.
..... ................. ......
PII .... UT .... PII (UMP)4 ...
.......
...... ............
.....UR
....
...
.......... ...... ............... ...
..
. .
................................ ...
......... ...... ....... ...
UMP ⊕..
.. ....⊕
H2 O
.... .. ...
.. ..
glutamine .. .. 2–oxoglutarate
.. ...
.. .. ..
.. .. ..
Pi. ª.. .. ...⊕ .ADP
......... ..... ..... .... .........
.. .
................................
.
............. ..... ...............
.... ..... ...... ......
...ATd.....
GS(AMP)12 ... .
. GS
....... ....ATa .............
....... . .
......
........... ..... ..............
................................
......... ...... .......... .......
⊕ PPi . .... ..ª ATP
⊕ .. ..

Fig. 7.18: The bicyclic glutamine synthase cascade.

This system is a central part of the control of nitrogen metabolism in E. coli. The fig-
ure is based on that in ref. [38] but modified so that activation reactions are above
deactivations and enzymically active forms are to the right. Abbreviations: GS, glu-
tamine synthase; UT, uridylyl transferase; UT, uridylyl removing enzyme; ATa, adenylyl
transferase; ATd, the deadenylylating activity of the transferase; PP i , inorganic py-
rophosphate; Pi , inorganic phosphate.

NADP+ + ADP + Pi
The problem that organisms have to resolve is the need to maintain appropri-
ate levels of glutamate and glutamine to meet synthetic requirements without
withdrawing all the 2–oxoglutarate from the tricarboxylic acid cycle, and to
manage to do this across the range from conditions where usable nitrogen is
in short supply to those where it is in excess. The enzyme has already been
mentioned in Chapter 1.2.1 and earlier in this chapter (Section 7.2.3.3.3) be-
cause it is an allosteric enzyme affected by 8 different feedback inhibitors.
(The ninth effector is Mg2+ .)
It was Earl Stadtman’s group at Bethesda that was largely responsible
for establishing that glutamine synthase in E. coli was subject to inactivation
by adenylylation (addition of AMP groups) of tyrosine residues38 catalyzed
by an adenylyl transferase. Glutamine synthase is composed of 12 identical
subunits, and each of these has one modifiable tyrosine. The enzymic activity
of a glutamine synthase molecule is inversely proportional to the number of
subunits that have been modified. Reactivation of the enzyme is by phospho-
rolysis of the adenylyl groups to give free enzyme and ADP:
7.4 Regulation by covalent modification of enzymes 247

12 ATP + GS12 −→ GS12 (AMP)12 + 12 PPi

GS12 (AMP)12 + 12 Pi −→ GS12 + 12 ADP

(where PPi is pyrophosphate and Pi is phosphate). The adenylylation is

catalyzed by glutamine synthase adenylyltransferase (properly glutamate–
ammonia ligase adenylyltransferase, EC 2.7.7.42). The de–adenylylation re-
action is a phosphorolysis, and is an activity of the same enzyme as the
adenylylation, but at a separate active site.

The relative levels of the two activities of the adenylyl transferase are recip-
rocally modulated by two types of factor (Fig. 7.18). Firstly, there are metabo-
lites that affect the activity. Of these, glutamine itself and 2–oxoglutarate
(product and substrate of the synthase) have oppositely directed effects with
glutamine stimulating the inactivation and inhibiting the activation, whereas
2–oxoglutarate does the opposite. The second factor is the tetrameric regu-
latory protein PII discovered by Shapiro. This also exists in two forms. The
unmodified form is an activator of the adenylylation (inactivation) of glu-
tamine synthase. It is modified by uridylylation (Fig. 7.18), and the uridy-
lated form is an activator of the de–adenylylation (activation) of glutamine
synthase. The interconversion reactions affecting PII mirror the modification
of glutamine synthase itself. The protein–PII uridylyl transferase and the
uridylyl–removing enzyme (which this time is a hydrolysis reaction to form
UMP) are separate activities on the same polypeptide chain (EC 2.7.7.59).
Glutamine stimulates the uridylyl–removing activity, thereby leading to ad-
ditional activation of the adenylylation of glutamine synthase (in addition
to its direct affect on this latter modifying reaction). Again, 2–oxoglutarate
has the opposite activity, since it stimulates the uridylyl transferase, forming
the modified PII that stimulates the de–adenylylation of glutamine synthase.
Thus the control of glutamine synthase involves a cascade system of two cyclic
modification reactions, but these cycles are doubly coupled by the reciprocal
activities of the two forms of PII . The four modifying activities between them
also respond to over 20 other assorted metabolites, including nucleotides and
intermediates of the glycolytic pathway and the tricarboxylic acid cycle. (A
further layer of complexity is added by the role of PII in modulating the tran-
scription of glutamine synthase. This will not be discussed further here, but
is reviewed in reference [197].)

The detailed properties of a system like this with such a network of inter-
acting effects are not immediately obvious from the structure of the cascade.
Stadtman and Chock, and other members of their group, therefore undertook
an extensive theoretical analysis of the possible behaviours of cyclic modifica-
tion cascades. Their conclusions are discussed in the next section, as they are
applicable to cyclic modification schemes in general (including phosphorylase)
and not just glutamine synthase.
248 7 Control structures in metabolism

Ki + N *
) Ka
. ..
..
ATP ... ADP
. ....... ...
.......................................
........ .. .
.. .
........
.... ..........
........ ......
........ ..... .
... .....
E E–P
....... ...
.....
....... ..
........
........... .
...
.... ............
................................................
..... ... ..
Pi .. H2 O
..
Pi + F *
) Pa

Fig. 7.19: Model of a covalent modification scheme.

In Stadtman & Chock’s model,238 E is the unmodified enzyme, E–P is its phosphorylated
form, K is the kinase and P the phosphatase. i and a indicate the inactive and active
forms of the latter. N is the effector of the oN reaction and F is the effector of the oFf
reaction.

7.4.5 Properties of cyclic modification systems

Stadtman & Chock226, 238 used the basic model shown in Fig. 7.19 to in-
vestigate how the amount of enzyme in one of the modified forms (say the
phosphorylated form) depended at steady state on the effectors of the two
modifying reactions. Although a comparison of the model with the examples
discussed in this Chapter shows that it posesses the fundamental character-
istics of covalent modification systems, it is much simplified, in that the con-
verting enzymes are only active when associated with their effector molecules
and that the effector binding is simple, non–cooperative binding. The model
is illustrated as a phosphorylation, but the type of modification is not really
important because ATP, ADP and phosphate are not considered as specific
variables in the model. Stadtman and Chock derived equations that relate
the fraction of the target enzyme in the modified form, when the cycle has
reached steady state, to the concentration of one of the effectors of the con-
verting enzymes. From these, they determined the following different aspects
of the behaviour of these cycles.

7.4.5.1 Catalytic amplification

Since the modification cycle is catalyzed, it is an obvious property that a small
amount of converter enzyme can act on a much larger amount of target en-
zyme. Thus the catalytic power can be amplified in the sense that the limiting
rate (V ) of the maximally–activated target enzyme can be many–fold greater
than that of the modifying enzyme. Of course, if the specific activities of the
two enzymes are relatively similar, the catalytic amplification largely reflects
7.4 Regulation by covalent modification of enzymes 249

the difference in the amounts of the two enzymes. This is undoubtedly an im-
portant aspect of many modification cycles, such as cases where small amounts
of hormones (at perhaps 10−9 M) combine with relatively few receptors to act
eventually on relatively abundant enzymes (perhaps at concentrations around
10−5 M). Another example is in the irreversible modification cascade of blood
clotting, where the initial signal is tiny but a macroscopic blood clot is formed.
Important as it may be, there is little mystery about catalytic amplifica-
tion. It should be noted that a high degree of catalytic amplification is only
obtained at a price that is not evident in steady state studies: the larger the
amount of target enzyme relative to modifying enzyme, the longer the system
will take to respond to a change in the amount of an effector.

7.4.5.2 Signal amplification

1.0

0.8 c
b
Fractional modification

0.6 a

0.4

0.2

0.0
-3 -2 -1 0 1 2

log N

Fig. 7.20: Degree of activation of the target enzyme.

Curve c shows the fraction of the kinase activated by binding its effector, N (see
Fig. 7.19). Curves a and b show the fraction of the target enzyme phosphorylated
for different values of the parameters in Stadtman & Chock’s model.

A very significant feature of cyclic modification schemes is that the degree

of activation of converting enzyme by an effector does not correspond directly
to the amount of modified target enzyme that it forms. This is illustrated in
Fig. 7.20 with some theoretical curves calculated with Stadtman & Chock’s
model. Curve a has the parameters of the model set so that half conversion of
the enzyme to the phosphorylated form is achieved at the same concentration
of effector N that gives half–saturation of the kinase (curve c). Curve b has a
16–fold greater kinase:phosphatase activity ratio and a 2.5–fold decrease in the
250 7 Control structures in metabolism

activation of the phosphatase by its effector F . This causes the concentration

of N at half conversion to have fallen to nearly 10−2 , or nearly 100-fold lower
than the concentration of N required to half–activate the kinase, which is
unchanged at 1. Since N is the signal that initiates phosphorylation, the
modification cycle shows signal amplification. In this case it can be quantified
as:
N required for half activation of kinase
(7.10)
N required for half phosphorylation of target
The signal amplification arises because the cycle is catalyzed and therefore a
small amount of kinase can convert a larger amount of target, but it is not
the same as the catalytic amplification referred to above. The significance of
signal amplification is that it allows the system to produce an effect on the
target enzyme at a signal concentration that is well below the dissociation
constant for the binding of the signal to the enzyme it activates.

7.4.5.3 Amplitude
Fig 7.20 also shows that the modification cycle cannot necessarily produce
complete conversion of the target to its phosphorylated form. The symmetry
of the cycle implies that there will also be circumstances where there cannot
be complete conversion to the dephosphorylated form. The amplitude is the
range of fractional modification that is covered in going from zero level of an
effector to saturating levels. For curve b of Fig. 7.20 it is nearly 1.0, whereas
for curve a it is only 0.67. Variation of the parameters of the model can
produce amplitudes anywhere from 0 to 1.

7.4.5.4 Sensitivity
Stadtman & Chock were concerned to know whether their model of a covalent
modification cycle produced a curve that was steeper or less steep than the
familiar rectangular hyperbola for enzyme kinetics and ligand binding. To
measure this they introduced a measure of sensitivity. They took as their ref-
erence the change in ligand concentration required to go from 10% to 90% sat-
uration for a normal rectangular hyperbolic binding curve. From Eqn. [3.10],
Chapter 3.4, it is possible to calculate that this requires an increase in the
ligand concentration equal to 8.89 times the dissociation constant (or ligand
concentration that gives 50% binding). They therefore defined sensitivity, S,
as:
8.89 N0.5
S= (7.11)
N0.9 − N0.1
where N0.5 is the concentration of the effector N that gives 50% of the maximal
amplitude of modification. This sensitivity function differs from those defined
in Metabolic Control Analysis (i.e. control coefficients and elasticities) in that
it is measured over a finite interval rather than from the gradient of a curve,
but it does share the properties of being dimensionless and of measuring the
steepness of response. It is actually more like the Hill coefficient (Chapter 3.4,
7.4 Regulation by covalent modification of enzymes 251

p. 69) in that values of S greater than 1 indicate that the curve relating the
fractional modification to effector concentration is steeper than a rectangular
hyperbola, whereas values less than 1 show it is less steep. As it happens,
Stadtman & Chock’s relatively simple model for the modification cycle gives
sensitivity values equal to 1 unless an effector acts at more than one site.
For example, if N activates the kinase and inhibits the phosphatase, then the
sensitivity is greater than 1. (The result is similar to that seen in curves a
and c of Fig. 7.21, though these were calculated for a different model.) This
is equivalent to positive cooperativity in its outcome, though it is obtained
without any cooperative effects in the binding of the effector. An example of
an effector that acts in opposite directions on the two converting enzymes of
the cycle is glutamine in the modification of glutamine synthase (Fig. 7.18).
There are other ways that the sensitivity can be altered that will be discussed
further below.

7.4.5.5 Biological integration

Cyclic modification systems involve more enzymes and therefore naturally cre-
ate more sites at which effectors can act. This is well illustrated by glutamine
synthase, which responds directly as an allosteric enzyme to about 15 ligands,
but indirectly to about another 20 that act through one or more of the con-
verting enzymes. In addition, Fig. 7.19 represents just one pattern of effector
action in a covalent modification system; as mentioned above, the sensitivity
of the response is changed if one effector acts on both converting reactions.
Thus these systems have the potential to show a range of behaviours and to
integrate a wide range of signals. The phosphorylase system is a particularly
good example of signal integration, since hormones, via cyclic AMP, and other
hormones or nerve impulses via Ca2+ , can separately activate phosphorylase,
but when they occur together, the responses do not simply multiply together
because there is interaction between the signals.

7.4.5.6 Multiple cyclic cascades

One of the ways in which covalent modification is often more complex than
the simple cyclic scheme shown in Fig. 7.19 is that multiple cycles can form a
cascade. These can either be open like the phosphorylase cascade (Fig. 7.17),
where a cascade at one level has only a single point of action on the next
level, or closed, like the glutamine synthase cascade (Fig. 7.18) where the two
interconverted forms of the enzyme at one level both act on the next level,
but on opposite sides.
In essence, the properties of a single modification cycle become even more
pronounced in the multiple cycles. Thus the model of Stadtman & Chock
shows that the signal amplification can increase dramatically with the number
of cycles in the cascade, basically because the amplifications of each stage
multiply together. Similarly, the sensitivity increases in proportion to the
number of stages that the effector N modulates (cf Fig. 7.21c).
252 7 Control structures in metabolism

Another property that is affected by the number of cycles is the rate of

modification of the final target enzyme. If the initiating effector N is increased
in concentration, there is a lag before the target enzyme is modified. This
lag increases with the number of cycles in the cascade, not unreasonably
as this increases the number of stages through which the signal has to be
transmitted. However, the rate of modification is increased at each stage,
corresponding to rate amplification. The rate amplification is cumulative in
multiple cycle cascades, so after a lag, there is an almost explosive increase
in the amount of the modified target enzyme. With realistic estimates of the
kinetic constants, Stadtman & Chock’s models show that it is feasible for
covalent modification cascades to operate in the millisecond to second time
span. Indeed, measurements of the rate of formation of phosphorylase a in
frog sartorius muscle upon electrical stimulation, carried out by C F Cori
and his colleagues in the 1960s, showed that the half time of activation was
700 ms. Similar measurements on mouse muscle gave a half time of 1 s. 101
However, since phosphorylase is activated on muscle contraction by a single
cycle cascade (Ca2+ activation of phosphorylase kinase), and the effects of
adrenaline on phosphorylase in muscle are slower, even though they involve a
multi–stage cascade, these experiments are not an indubitable illustration of
rate amplification.

7.4.5.7 Ultrasensitivity
The next development in the theoretical understanding of covalent modifica-
tion systems came from Albert Goldbeter of the Free University of Brussels
and Daniel Koshland of the University of California. Whereas the analy-
sis described above had used simple chemical kinetics for the modification
reactions, Goldbeter and Koshland represented them as true enzymic reac-
tions with the dependence of rate on the concentration of their substrates
(the unmodified and modified forms of the target enzyme) described by the
Michaelis–Menten equation.88 For their sensitivity measure, they used the
measure earlier developed by Koshland and his colleagues for describing the
steepness of cooperative binding curves in allosteric enzymes.137 For a factor
N that affects the degree of modification, their sensitivity RN is defined as:
N0.9
RN = (7.12)
N0.1
N could be either an effector, or the ratio of limiting rates of the two convert-
ing enzymes. Unlike Stadtman and Chock’s sensitivity measure (Eqn. [7.11]),
or the Hill coefficient, this value decreases with increasing steepness of the
curve, from a value of 81 for a rectangular hyperbolic curve, to a minimum of
1 for an infinitely steep transition. The use of the different definitions means
that neither the equations derived by the two groups nor the numerical mea-
sures of sensitivity are exactly comparable, though in qualitative terms they
are referring to the same concept. (Such differences in the operational defini-
tion of sensitivity, of which these are merely two instances, was the main rea-
7.4 Regulation by covalent modification of enzymes 253

1.0

c a
Fractional modification
0.8 b

0.6

0.4

0.2

0.0
0.01 0.1 1 10 100

[Effector], F

Fig. 7.21: Zero–order ultrasensitivity.

a) Fractional phosphorylation of the target enzyme according to Goldbeter &
Koshland’s88 model with unsaturated modifying enzymes and F a non–competitive in-
hibitor of the phosphatase (Fig. 7.19). b) as a), but showing ultrasensitivity with the
target enzyme 10–fold higher than the Km values of both modifying enzymes. c) as a)
except F activates the kinase and inhibits the phosphatase.

son that Kacser and Burns abandonned their original term sensitivity for the
flux control coefficient.) The analysis carried out by Goldbeter and Koshland
showed that the sensitivity was much more variable than in Stadtman and
Chock’s studies, where it only changed in response to the number of steps at
which an effector acted.
Their novel finding was that, if one or both of the converting enzymes
is operating close to saturation, the fractional modification of the target en-
zyme can become an extremely sensitive function of the ratio of the limiting
rates of the converting enzymes. They termed this zero order ultrasensitivity,
since a saturated enzyme shows near zero order kinetics in the terminology
of chemical kinetics, and it was possible to show theoretical curves with a
steepness equivalent to a Hill coefficient, nH , of 13 or more. The effect is
illustrated in Fig. 7.21, where curve a has been calculated with their model
for concentrations of the target enzyme well below the Km values of the mod-
ifying kinase and phosphatase. Curve b is far steeper, but the only change is
that the total amount of target enzyme has been increased to 10 times the
level of both Km values. For comparison, the smaller increase in steepness
obtained with an effector that acts on both kinase and phosphatase is shown
in curve c. Goldbeter and Koshland also showed that the ultrasensitivity can
be amplified in a multicycle cascade if each component cascade exhibits ultra-
254 7 Control structures in metabolism

sensitivity. Furthermore, a non–competitive inhibitor of a modifying enzyme

shows a greater degree of ultrasensitivity than a competitive inhibitor, since
the latter increases the apparent Km of the modifying enzyme and moves it
from the zero–order region towards the first order region.
Unlike the comparable conjecture for increased sensitivity in substrate cy-
cles (Section 7.3.3), there is some experimental support for the occurrence
of ultrasensitivity in covalent modification. A reconstituted, in vitro system
of phosphorylase plus protein phosphatase and phosphorylase kinase showed
a response that could be fitted by the equations derived by Goldbeter and
Koshland at phosphorylase concentrations comparable to those found in mus-
cle.156 The degree of ultrasensitivity is roughly equivalent to a Hill coefficient
of 2.3. The only problem with this interpretation was that the activation of
phosphorylase kinase (whether by Ca2+ or by phosphorylation) was believed
at the time to be caused by a decrease in the Km for phosphorylase, which
would not be an effective stimulus in the zero–order sensitivity region of oper-
ation. However, a subsequent re–investigation of the kinetics of phosphorylase
kinase has shown that the predominant action of these effectors is to increase
the limiting rate.173

7.4.5.8 Control Analysis

So far, the analysis of the properties of cyclic modification cascades has not
made use of the concepts of Metabolic Control Analysis. However, there
are three reasons why it is important to apply Control Analysis to covalent
modification:
• covalent modification represents an important class of control mecha-
nisms, and therefore Control Analysis would not be complete if it did
not take account of it;
• the analyses presented above have been very dependent on the particular
types of molecular mechanism used in the model, which need not be the
case with Control Analysis, and
• the analyses have only considered whether the degree of activity of a
single pathway enzyme can be changed substantially,and not whether
this will have a significant effect on pathway flux.
Marı́a Luz Cárdenas and Athel Cornish–Bowden27 in Marseilles used the re-
sponse coefficient of Metabolic Control Analysis (Eqn. [5.21], Chapter 5.4) as
a measure of the dependence of the fraction of a modifiable target enzyme
in the active form upon the concentration of an effector of the modifying en-
zymes. The advantage of this is that a value of the response coefficient greater
than 1 can be regarded as useful enhanced sensitivity, on the basis that a value
up to 1 could be achieved by the direct action of an effector on an enzyme
without the need for additional enzymes and the energy expenditure involved
in a covalent modification cycle. On this basis, they were able to show that
an effector that acts only on the Km values of the modifying enzymes can
7.4 Regulation by covalent modification of enzymes 255

only generate enhanced sensitivity if it acts on both enzymes and if it inhibits

the deactivating enzyme at lower concentrations than are required to stimu-
late the activating enzyme. Otherwise, the response to the effector is actually
poorer than would be obtained by direct action of the effector on the target
enzyme.
They also carried out a numerical search on a general model of a cova-
lent modification cycle to determine the circumstances under which the model
showed the maximum attainable ultrasensitivity, with the effector acting on
both modifying enzymes. This confirmed Goldbeter and Koshland’s observa-
tion that the modifying enzymes should both be as near to saturation with
target enzyme as possible. In addition, the actions of the effector on the
limiting rates of the enzymes must predominate over its effects on the K m
values. Again, they also found that it was important that the inhibition of
one modifying enzyme occurred at much lower concentrations of effector than
were required to activate the other. Under the most favourable combination
of circumstances, the sensitivity of the the target enzyme to the effector could
be equivalent to a Hill coefficient as high as 800. This represents the potential
of the system, since there is no experimental evidence for a system showing
an effect as large as this.
At about the same time, Rankin Small and I were also applying the con-
cepts of Metabolic Control Analysis to covalent modification of an enzyme
embedded in a metabolic pathway, so that we could consider the effects on
the pathway flux rather than just the amount of activated enzyme.231 One
conclusion we reached was that the flux control coefficient of the modifiable
enzyme (defined in terms of the total amount of that enzyme) could be in-
creased, decreased or unchanged by the occurrence of the modification cycle.
For it to be increased, the modifying enzymes would have to have small elas-
ticities with respect to the target enzyme; this is equivalent to the requirement
for zero–order ultrasensitivity, except that it has been reached for the pathway
flux, and it is general in that it does not depend on any particular assumptions
about the kinetic equations of the modifying enzymes. It is also a reminder of
an important factor that has been almost lost sight of: whatever the effect of
the modification cycle on the activity of the target enzyme, there will be no
effect on flux unless the enzyme has a non–zero flux control coefficient. We
also extended the concept of the response coefficient to define the response of
pathway flux (J) to an effector (N ) that acts on a modifying enzyme (a) of a
target enzyme (t) as:
∗
J
RN = CtJ Cat εaN (7.13)
∗
where Cat is the concentration control coefficient of the modifying enzyme a
on the concentration of the active form t∗ of the target enzyme. This confirms
that an enhanced response of the pathway flux to the effector cannot be guar-
∗
anteed by a large value for one of the terms, such as Cat (which does have a
large value under the conditions favourable for ultrasensitivity). Small values
of the other coefficients can nullify the advantage. Again, algebraic analysis
∗
of the term Cat shows that it will be less than 1 if the modifying enzymes are
256 7 Control structures in metabolism

unsaturated with the target enzyme, and hence under these circumstances,
the covalent modification cycle actually acts to reduce the response of the
flux to the effector. We were also able to confirm and extend the conclusions
reached by Cárdenas and Cornish–Bowden about the conditions that would
lead to ultrasensitivity, with the advantage that many of our arguments de-
pend on relative values of elasticities, and not on the molecular mechanisms
underlying them.
Our overall conclusions were that only a quite restricted set of conditions
generate ultrasensitivity and ensure it is translated into a more sensitive re-
sponse of pathway flux to an effector. This is even more so in multicyclic
cascades, where additional terms appear in the response coefficient, but each
of these with a value less than 1 contributes to an unavoidable attenuation of
the sensitivity of the flux to the signal. Nevertheless, the actual examples of
covalent modification cycles presented above show that the theoretical mod-
els studied so far are relatively simple and that there is plenty of potential
for complex properties to emerge from covalent modification cascades. Also,
most of the analysis has been concerned with steady state behaviour, but the
transient behaviour in the change between metabolic states is almost certainly
very important as well. However, the weakest link in all the studies, whether
practical or theoretical, has been that little attention has been paid to the
link between the change in modification state and the effect on the metabolic
pathway of which the target enzyme is part. (Modular Control Analysis is a
recent development of Control Analysis that deals more systematically with
this problem.123 ) Metabolic Control Analysis has shown that enzymes that
are truly rate–limiting, with flux control coefficients of 1, are not generally
found. It has also shown that for an enzyme with a flux control coefficient
less than 1, the scope for increasing flux by activating the enzyme is relatively
limited (see Chapter 5.2.2). How then, can we explain covalent modifica-
tion mechanisms that only appear to exist to cause a significant change in
flux by making large changes in the amount of active enzyme in response
to small changes in signal molecules, when the evidence suggests this cannot
be effective? The traditional paradigm of covalent modification systems was
that they control metabolism by acting on ¡rate–limiting enzymes. This is no
longer tenable; a new view of how control of metabolism is actually exerted
in the light of the analyses presented in this Chapter will be developed in the
next and final chapter.

7.5 Summary
1. Traditional theories of metabolic control assumed there were obvious
advantages of controlling the rate of a pathway near its start. Analysis
of the control of supply of and demand for a metabolite shows that
better homoeostasis of metabolite concentrations is achieved when the
greater flux control is exerted by the demand reactions, i.e. towards the
end.
7.6 Appendix 1: Control Analysis of substrate cycles 257

2. Feedback inhibition loops are common regulatory structures in metabolic

pathways. Analysis of their properties suggests their primary effect is
homoeostasis of metabolite concentrations rather than flux control, as
previously believed. This is because feedback loops transfer control to
the demand reactions after the metabolite exerting feedback.
3. In branched pathways, both sequential and nested patterns of feedback
inhibition are found. Sequential feedback inhibition is the more reliable,
though nested feedback inhibition can give better homoeostasis over a
limited range of conditions.
4. Cyclic pathways that dissipate energy potentially exist in metabolism,
and measurements confirm that some of these routes are indeed active.
Such substrate or futile cycles could fulfil a number of functions, of which
thermogenesis and switching fluxes at complex metabolic crossroads are
the best established.
5. Covalent modification cycles are another common control structure in
metabolism, linking it to a range of internal and external signals. They
amplify the effects of small, initial signals in a variety of ways and al-
low metabolism to integrate a range of different, possibly conflicting,
signals. Theoretical analysis reveals the possibility that they show ul-
trasensitivity, where the response to a stimulus becomes sharpened, and
there is evidence that the activation system for glycogen phosphorylase
exhibits this.

Further reading
1. Stadtman, E. R. Mechanisms of enzyme regulation in metabolism in The
Enzymes, Boyer, P. D. (ed.) 3rd ed. Vol. 1, pp. 397–459, Academic
Press, New York (1970)
2. Katz, J. & Rognstad, R. Futile cycles in the metabolism of glucose,
Current Topics in Cell Regulation, 10, 237–289, (1976)
3. Hue, L. The role of futile cycles in the regulation of carbohydrate meta-
bolism in the liver, Adv. Enzymol. 52, 247–331, (1981)
4. Hue, L. Futile cycles and regulation of metabolism in Metabolic Com-
partmentation, Sies, H. (ed.) pp. 71–97, Academic Press, London
(1982)
5. Boyer, P. D. & Krebs, E. G. (eds.) Control by Phosphorylation, Parts A
& B in The Enzymes, Vols. XVII & XVIII, Academic Press, New York
(1986)
6. Shaltiel, S. & Chock, P. B. (eds) Modulation by Covalent Modification.
Current Topics in Cell Regulation, 27, (1985)
258 7 Control structures in metabolism

7.6 Appendix 1: Control Analysis of substrate cy-

cles
For the substrate cycle pathway shown in Fig. 7.13, the following theorems of
Metabolic Control Analysis apply. There is a summation theorem:
J
C1J + C2a J
+ C2b + C3J = 1

There are two connectivity theorems with respect to S1 and S2 :

J 2a
C1J ε1S1 + C2a J 2b
εS1 + C2b ε S1 = 0
J 2a J 2b
C2a ε S2 + C2b ε S2 + C3J ε3S3 = 0

Finally, as explained in Chapter 5.3.3, p. 121, there is a need for an additional

theorem, the substrate cycle summation relationship, originally developed by
myself and Herbert Sauro.69 Its form varies depending on the choice of which
flux is used in the flux control coefficients, but for J = J1 = J3 it is:
J
C2a CJ
+ 2b = 0
v2a v2b

These relationships form a set of four simultaneous equations that can be

expressed in matrix form69 as:
     
1 1 1 1 C1J 1
     
     
 ε1 ε2a ε2b 0   J   0 
 S1 S1 S1   C2a   
 . = 
   J   
 0 ε2a ε2b ε3S2   C2b   0 
 S2 S2     
     
0 1/v2a 1/v2b 0 C3J 0

Algebraic solution of these equations gives each of the flux control coef-
ficients in terms of the elasticities and the rates of the cycle enzymes. 69 The
flux control coefficient of enzyme 2a is:

J ε11 ε32 (1 + v2b /v1 )

C2a =
D
where

D = ε11 ε32 − (1 + v2b /v1 )(ε11 ε2a 3 2a

2 + ε 2 ε1 )
1 2b 3 2b
+ (v2b /v1 )(ε1 ε2 + ε2 ε1 )

The factor, r, by which cycling increases the flux control coefficient of

enzyme 2a is obtained by comparing two cases with the same net flux and
7.6 Appendix 1: Control Analysis of substrate cycles 259

intermediate concentrations, except that in the reference state, the cycling

rate, v2b , is zero:
1
r= (7.14)
v2b (ε1 ε2 − ε11 ε2b
1 3 3 2b
2 − ε 2 ε1 )
1−
v2a (ε11 ε32 − ε11 ε2a 3 2a
2 − ε 2 ε1 )
260 7 Control structures in metabolism
 8

Conclusion

At first glance, the impact of Metabolic Control Analysis on conventional

approaches to biochemical control can appear limited and benign: the con-
cept of a single rate–limiting step is replaced by the concept that a number
of steps could share control without any of them being fully rate–limiting.
On this interpretation, the search for rate–limiting enzymes is replaced by
the measurement of flux control coefficients to determine which enzymes have
the coefficients with the largest values, as illustrated in Chapter 6. However,
once a rigorous approach to the systematic analysis of metabolic control is
embarked upon, it becomes increasingly clear that the whole edifice of conven-
tional biochemical thinking about metabolic control is very shaky, in spite of
the very impressive level of knowledge about the molecular properties of the
underlying components. One after another, cherished ideas have been found
to be untenable.
First were the conventional criteria used to identify rate–limiting enzymes
(Chapter 4). These have been shown to be a poor guide to identifying enzymes
with large flux control coefficients (e.g. Chapter 5.5.1). In fact, they are an
inconsistent mix of criteria, since some of them identify enzymes with regula-
tory properties that can be shown (in the study of feedback inhibition effects,
Chapter 7.2.2) to result in those enzymes being poor sites of flux and con-
centration control. In this case the properties relate to improved regulation,
showing that regulation and control can be distinguished. To the surprise of
critics of the theory, the experimental results on phosphofructokinase (Chap-
ter 6.1.1.3) have confirmed that the theoretical view is correct: the enzyme is
regulatory but cannot be used to control flux.
Another doubtful conventional concept is that the allosteric control of reg-
ulatory enzymes is responsible for rapid responses in flux control on shorter
time–scales than either covalent modification or enzyme synthesis and degra-
dation.5, 139 However, the transition time for pathways to reach a steady state
flux is typically in the region of seconds to minutes,63, 196, 249 which is the time
scale on which covalent modification works (as mentioned in Chapter 7.4.5).

261
262 8 Conclusion

Thus there is no obvious ‘functional gap’ in flux control that allosteric effects
have to fill so long as covalent modification mechanisms are operating.
Even the teleological principle, that it is advantageous to control a pathway
near its beginning, is open to doubt. The study of supply and demand path-
ways (Chapter 7.1), combined with the effects of feedback inhibition, shows
that metabolic control structures exist to give good regulation (i.e. strong flux
control combined with relatively weaker effects on metabolite concentrations)
if flux control is concentrated at the demand end of the pathway. It is not not
known at present whether there are equivalent control structures to ensure
such good regulation if control is exerted near the beginning of the path-
way, but the evidence is against it. In this context, it is most remarkable that
control of flux in cells is often achieved with so little change in metabolite con-
centrations. For example, when a 1000–fold increase in the glycolytic rate of
rat muscle was obtained by stimulating the muscle, the measured metabolites
increased less than 10–fold in concentration.107 These and similar experi-
ments led Bücher and Rüssmann24 to speculate that metabolic pathways are
controlled at more than a single enzyme in order to ensure homeostasis of
metabolite concentrations, as mentioned in Chapter 1.4.3. Similarly, Helmre-
ich and Cori101 noted the constancy of metabolite levels with large glycolytic
flux changes in vertebrate muscle and also concluded that glycolysis must be
controlled as a unit, particularly as they could not account for the changes in
rates of several of the enzymes by the small changes in concentrations of their
substrates and effectors.
Metabolic Control Analysis has confirmed these authors’ suspicions that
there was a problem still to be solved about the control of glycolysis. Even
if the fallacy of the rate–limiting step is made apparently more respectable
by supposing that control is exerted by action at the enzyme with the largest
flux control coefficient, control at a single site has a clear problem: Small and
Kacser’s theory of large changes (Chapter 5.2.2) shows that it is impossible
to achieve anything other than small total changes in the flux in response to
massive changes in an enzyme’s activity if its flux control coefficient is only
slightly less than 1. Therefore, if large changes in flux occur in cells (such as
in glycolytic flux in muscle upon changes in mechanical activity, or in pho-
tosynthesis in plants upon changes in illumination) they cannot be achieved
by controls acting at a single site; there must be widespread sites of action
of any control signal. Furthermore, even though the covalent modification of
phosphorylase apparently provides an obvious mechanism for flux control of
muscle glycolysis, the analysis of homoeostasis and regulation makes it un-
likely that control of flux solely from the beginning of the pathway can explain
such good homeostasis of metabolite concentrations as is observed.
Thus although this book has covered a systems theory of the control of
metabolism and has examined the range of control and regulatory mecha-
nisms available in metabolism, we have arrived at its final chapter without a
satisfactory answer to the question: how is the flux of a metabolic pathway
controlled in vivo? In the next section, I will suggest an answer, developed
with my colleague Simon Thomas,72 that is consistent with the theory and
8.1 A new view of the physiological control of flux 263

the experimental evidence.

8.1 A new view of the physiological control of flux

' $
3 Fig. 8.1: Schematic cell
? metabolism for the Universal
2
S - G Method.
The metabolic inputs are grouped
as N, the desired product is P,

& %
and all the other products and
1 cell biomass are G. S is the
metabolite at which synthesis of
P branches from the rest of met-
? abolism. Each arrow represents a
P complex metabolic network.

The unresolved problem of metabolic control is how to achieve a large

change in metabolic flux in response to some environmental or hormonal sig-
nal without disruption of other areas of metabolism and whilst maintaining
metabolite levels relatively stable. In fact, Henrik Kacser found an answer to
this problem, not by studying control systems in vivo but by asking how a
biotechnologist could stimulate a pathway that makes a useful product (such
as an amino acid or the antibiotic penicillin). Practical attempts to do this
by using genetic engineering techniques to increase the amount of a partic-
ular enzyme in a pathway have tended to be unsuccessful. In part, this is
because the genetic engineers have selected enzymes that are likely to have
small flux control coefficients, such as feedback–inhibited enzymes. (The ex-
ample of phosphofructokinase in yeast has been mentioned previously.) In
any case, the predictions that Rankin Small and Henrik Kacser have made
of the effects of large changes in a single enzyme show that this procedure is
likely to give disappointing results.
Therefore in 1993 Kacser and Luis Acerenza studied how biotechnologists
could design a change in metabolism to cause an arbitrarily large flux increase
in a specific pathway, whilst leaving all concentrations and most other fluxes
unperturbed.117, 118 They termed their result the ‘Universal Method’, since
it does not depend on the details of the kinetics and should be possible, at
least in principle, under most circumstances. (Whether it is universally prac-
ticable remains to be seen.) To illustrate their method, assume the target
pathway is one producing an end–product of metabolism, P (Fig. 8.1). The
method proposes making proportional increases in all the enzymes leading
264 8 Conclusion

directly to this output from the input(s), with the largest changes in activity
being made in all the enzymes in the final linear sequence to the output, and
with progressively smaller changes being made in each preceding branch so
that the fluxes in colateral branches remain unchanged. In Fig. 8.1, suppose
that the rate of production of P (J1 ) is to be increased 5–fold. All the en-
zymes along that branch would have to be increased by a factor of 5. This
creates an additional consumption of S equal to 4J1 . To ensure that the con-
centration of S, and hence the rate of synthesis of G, remain constant, the
rate of production of S has to be increased by the same amount. That is
J3 has to be increased to J3 + 4J1 . The fractional increase required in the
enzymes synthesising S is therefore 4J1 /J3 . Since J1 is much less than J3 ,
the change required in J3 is quantitatively less significant, and probably the
need to change enzyme levels dies away as the pathway is traced back to the
inputs. In effect, the method exploits the different behaviour of fluxes and
concentrations when the activities of a sequence of enzymes are all increased
in the same proportion. According to the summation theorem for flux control
coefficients (Chapter 5.2.3), the flux will increase in proportion to the enzyme
activities, whereas the metabolite concentrations will remain constant in ac-
cordance with the summation theorem for concentration control coefficients
(Chapter 5.8.1). It might not be necessary to increase every enzyme along
each branch affected provided that any enzymes omitted have flux control
coefficients very near zero.
The potential for success of this strategy of multiple enzyme manipula-
tions, compared with the failure of manipulating single enzymes in a pathway,
has been illustrated by Peter Niederberger and his colleagues for the trypto-
phan biosynthesis pathway of yeast.117, 175 There are 5 enzymes from cho-
rismate, the common precursor of the amino acids tryptophan, tyrosine and
phenylalanine, to tryptophan itself, coded by the genes TRP1 to TRP5. The
approximate control coefficients of these 5 enzymes were estimated from mea-
surements of tryptophan synthesis flux in mutant/wild–type heterozygotes in
tetraploid yeast. None of the enzymes had a large control coefficient, and
the sum of the five was 0.26. The five enzymes were overexpressed in yeast
by incorporating their genes, either singly or in combination, into a yeast
plasmid. As might have been expected from the low flux control coefficients,
overexpression of any of the enzymes singly, even up to 50–fold, had barely
any effect on the flux. The more interesting result was the demonstration
that the flux increase achieved when all the enzymes were changed together
was far greater than the product of the flux changes obtained by increasing
each enzyme separately. The results are summarized in Table 8.1, where it
is clear that the best increase in flux is obtained when all five enzymes are
overexpressed. Even so, an average 23–fold excess of the enzymes still only
produces a 9–fold increase in flux because no changes have been made to in-
crease the supply of chorismate. Incidentally, the experiments also show that
abolishing the allosteric feedback on anthranilate synthase (TRP2) does not
significantly stimulate tryptophan synthesis.
The Universal Method is a theoretical proposal for engineering a desired
8.1 A new view of the physiological control of flux 265

change in metabolic flux. Although the experiments on tryptophan synthesis

are not a complete implementation of it, the results indicate that relatively
large changes of flux can be obtained by activating all the enzymes in a path-
way. Whilst writing this book, it occurred to me to examine the possibility
that cells had already implemented this method in the control of their own
metabolism.

8.1.1 Evidence for multisite modulation in cells

If cells do control their metabolism by simultaneously changing the activities
of many enzymes (multisite modulation), then it should be possible to see the
principles of the Universal Method being implemented in vivo. For example:
1. When enzyme amounts change in response to physiological or environ-
mental signals, the relative proportions of pathway enzymes would re-
main constant.
2. The common factor by which the amounts of the pathway enzymes
change would be equal to the factor by which the flux changes.
3. The levels of change would be greatest in the main branch of the path-
way being controlled, although coordinated, but smaller, changes would

Table 8.1: Metabolic engineering of tryptophan synthesis in yeast.

Genes overexpressed Mean fold Relative

TRP2 TRP4 TRP1 TRP3 TRP5 increase flux to trp

- - - - - 1 1.0
- - + + - 58 2.0
+ + - + - 35 2.4
+∗ - + + - 34 1.2
+∗ + + + - 30 2.1
+ + - + + 19 8.2
+∗ + + + + 23 8.8

Results from Niederberger et al.175 The genes are arranged in order of

their enzymes between chorismate and tryptophan (see Fig. 7.6). Each
row represents a separate experiment where ‘+’ indicates the enzyme
was overexpressed from plasmid–borne genes; ‘-’ indicates wild–type
level. For TRP2, ‘+∗ ’ indicates that a mutant allele, resistant to
feedback inhibition by tryptophan, was overexpressed. The mean fold
increase column gives the average overexpression of the engineered
genes.
266 8 Conclusion

occur in more distal branches.

4. In addition to enzyme induction, other control mechanisms that act on
the pathway would also operate on a similar set of multiple target sites.
The first point is well illustrated by the organization of the genes for path-
way enzymes in operons (i.e. as adjacent genes with a common control of
expression) and also in regulons 152 where the genes are not all adjacent and
do not share control elements yet still respond to the same signals. Srere 236
has recently noted that the concept of control by rate–limiting steps is contra-
dicted by the many examples (glycolysis, the tricarboxylic acid cycle, photo-
synthesis and the syntheses of fatty acids, urea, nucleotides and amino acids)
where environmental or physiological signals cause coordinated induction of
all the enzymes in a pathway. I shall therefore concentrate on some examples
that illustrate the other three characteristics listed above.

8.1.1.1 Lipogenesis
Lipogenesis, or triacylglycerol synthesis, (Fig. 8.2) exemplifies all three. One
type of genetic obesity in mice maps to a single locus, the obese gene. This
codes for a secreted protein, specifically made by adipose tissue, that ap-
pears to have signalling functions and the absence of which causes profound
obesity.282 The obese phenotype is linked to the alteration of enzyme levels
in adipose tissue and liver relative to wild–type levels25 as summarized in
Table 8.2. The mutant mouse appears to be capable of a rate of lipid syn-
thesis per unit lean tissue some three fold higher than normal mice.25 This
is consistent with the 2–4 fold activation of eight of the major enzymes of
lipid and carbohydrate metabolism as shown in Table 8.2 (point 2). (Effec-
tively it is more enzymes than this, because fatty acid synthase is itself a
multi–enzyme complex of the seven enzymes needed to add malonyl–CoA to
a growing acyl–CoA chain.) In connected parts of carbohydrate metabolism,
Table 8.2 shows that four enzymes are elevated by about 50%, and two adi-
pose tissue lipases are significantly depressed25 (point 3). Of the enzymes
in this table, acetyl CoA carboxylase, fructose–1,6–bisphosphatase, hormone–
sensitive lipase, phosphofructokinase and pyruvate kinase are known to be
subject to control, directly or indirectly, by protein phosphorylation and de-
phosphorylation reactions that respond to dietary status105 (point 4).
Two lines of evidence suggest that this is not an odd result from a par-
ticular type of mutant mouse. Firstly, Bulfield and his colleagues obtained
essentially similar results in a later study of the differences in enzyme contents
between lean and fat lines of mice that had been selected over 26 generations
from the same base population.4 For most of the enzymes in Table 8.2, there
is good evidence that the controls on their levels of expression respond to
starvation, refeeding or high carbohydrate diets105 in rats. On the other
hand, there is a lack of experimental evidence supporting any system–wide
role for allosteric effectors in physiologically significant control of lipogenesis
by dietary factors.105
8.1 A new view of the physiological control of flux 267

8.1.1.2 Urea cycle

The rate of urea synthesis in rats responds proportionately to the amount

of protein in the diet, and over a range of protein intakes, the amounts of
all four of the urea cycle enzymes and carbamyl phosphate synthase also vary
proportionately.221 A comparison of the enzyme contents of rats on diets that
can cause a 4–fold difference in urea outputs showed that eight of the enzymes

cytosol

glucose

NADP+ NADPH
8
? ¸
glucose-6-P - pentose-P pathway
7

? 6
- glycerol-P

?
P-enolpyruvate

5 NADPH NADP+
? ]
pyruvate ¾ malate ¾

$
² 4
CO2
oxac
? 6
pyruvate
- triacylglycerol
J 3
citrate
J
^
J
acetyl–CoA 3́
´
´ ?
´
´ acetyl–CoA
? q ´
oxac citrate 2
¸
TCA ?
cycle malonyl–CoA

® 1
}
?
fatty acid
mitochondrion

Fig. 8.2: Pathways of lipogenesis.

The numbered enzymes are: 1, fatty acid synthase; 2, acetyl–CoA carboxylase; 3, ATP
citrate lyase; 4, malic enzyme; 5, pyruvate kinase; 6, glycerol–3–phosphate dehydroge-
nase; 7, glucose–6–phosphate dehydrogenase; 8, hexokinase. oxac: oxaloacetate.
268 8 Conclusion

Table 8.2: Relative enzyme levels in the obese mouse.

Tissue Enzyme % change

Group 1: elevated
adipose glycerol kinase 371
liver hexokinase IV 357
liver acetyl–CoA carboxylase 335
liver fatty acid synthase 267
liver malic enzyme 254
liver ATP citrate lyase 231
liver fructose–1,6–bisphosphatase 213
liver glycerol–3–phosphate dehydrogenase 210
Group 2: slightly elevated
liver glucose–6–phosphatase 163
liver phosphofructokinase 160
adipose glucose–6–phosphate dehydrogenase 151
liver pyruvate kinase 150
liver glucose–6–phosphate dehydrogenase 139
Group 3: depressed
adipose hormone–sensitive lipase 73
adipose monoacylglycerol lipase 20

The % change in activity represents the amount in obese

mice relative to lean controls; data from ref. [25]. Hex-
okinase IV is also known as glucokinase.51 Malic enzyme
is malate dehydrogenase (oxaloacetate–decarboxylating)
(NADP+ ), E.C. 1.1.1.40.

measured increased significantly. All the enzymes of the urea cycle increased
2–3 fold in activity, which is sufficient to account for the flux change if there is
also a slight stimulation by increased substrate supply (point 2 above). There
were also large increases in alanine transaminase and glucose–6–phosphate
dehydrogenase.

8.1.1.3 Carbohydrate metabolism

George Weber and his colleagues carried out extensive investigations in the
1960s into the effects of dietary changes, starvation and hormone treatments
on carbohydrate metabolism in rat liver. The hormones whose effects were
covered included glucocorticoid steroids, which stimulate gluconeogenesis, and
insulin, which suppresses it.268, 269 They concluded that the liver enzymes fall
into three groups: the exclusively gluconeogenic enzymes (in the left hand
8.1 A new view of the physiological control of flux 269

Table 8.3: Control of expression of enzymes of gluconeogenesis

and glycolysis in mammalian liver.

Expression stimulated by insulin

Expression inhibited by insulin, or glucose, inhibited by cyclic
stimulated by cyclic AMP AMP

glucose–6–phosphatase hexokinase IV (glucokinase)

fructose–1,6–bisphosphatase phosphofructo–1–kinase
phosphoenol pyruvate carboxykinase phosphofructo–2–kinase /
fructose–2,6–bisphosphatase
aldolase
pyruvate kinase

See Fig. 6.17 for the positions of most of these enzymes in gluconeo-
genesis. The effects shown are ones for which the molecular mech-
anisms have recently been established.147 But the existence of these
two groups of oppositely–regulated enzymes has been known for 30
years.268

column of Table 8.3), the exclusively glycolytic enzymes (those in the right
hand column of Table 8.3 except aldolase), and the dual–function enzymes
(the rest). The amounts of the members of a group change coordinately,
but the directions and magnitudes of the changes differ between the groups,
exemplifying points 1 and 3 above. For several of these enzymes, the molecular
mechanisms of control of enzyme amount have been shown to include the
actions of insulin, cyclic AMP and glucocorticoids on transcription.147, 184
Some support for a similar pattern of control by other mechanisms (point 4)
comes from the effects of protein phosphorylation. The actions of glucagon
and α–adrenergic agents in stimulating gluconeogenesis in rat liver involve
the inhibition of pyruvate kinase by phosphorylation, and the activation of
fructose–1,6–bisphosphatase and the inhibition of phosphofructo–1–kinase by
changes in the level of fructose–2,6–bisphosphate brought about by protein
phosphorylation. As it is possible that as many as 1 in 4 mammalian cell
proteins can be reversibly phosphorylated,37 it is even conceivable that there
are more undiscovered sites of phosphorylation involved in the stimulation of
gluconeogenesis.

In other tissues, protein phosphorylation stimulates glycolysis. For exam-

ple, in the different types of animal muscle, stimulation of contraction may
be associated with activation of some or all of phosphorylase, phosphofructo–
1–kinase, pyruvate dehydrogenase and myosin ATP–ase.
270 8 Conclusion

8.1.1.4 Steroid synthesis

acetyl–CoA cytoplasm

†
† uptake synthesis µ
?ª secretion
cholesterol ¾
1
cholesterol esters
∗
> µ

'
‡
† transport
$

?
cholesterol cortisol CS
6 6
NADPH
1P45011β ∗ DC
?
adrenodoxin
¸

ª R DC) 3
∗ P450SCC P45011β ∗ 11–DOC
6
11-DOC9 2 NADPH
© H
¼
© j
H
?
pregnenolone - pregnenolone
mitochondrion endoplasmic reticulum

Fig. 8.3: Pathways of steroidogenesis in adrenal cortical cells.

Key: 11–DOC, 11–deoxycortisol; DC, deoxycorticosterone; CS, corticosterone; 1, choles-
terol ester hydrolase; 2, 3–β–hydroxysteroid dehydrogenase + P–45- C21 + P–45017α ; 3,
3–β–hydroxysteroid dehydrogenase + P–45-C21 ; ∗, ACTH activates via cyclic AMP; †,
ACTH stimulated, mechanism unclear; ‡, possibly inhibited by ACTH. Enzymes involved
in NADPH production are also activated.

The synthesis of the several corticosteroid hormones in the adrenal cortex

(Fig. 8.3) can be stimulated by adrenocorticotropic hormone (ACTH) both
in the short term (seconds to minutes) and over periods of several days 181, 267
increasing production up to 10–fold relative to basal rates.138, 191 Cyclic AMP
is involved on both time scales. The pathway provides another example of
parallel multisite changes both in the rapid–acting modulation of activity (by
mechanisms including protein phosphorylation) and the control of enzyme
amounts. The rapid acting effects increase the availability of cholesterol at the
mitochondrial site of cytochrome P–450SCC (where SCC signifies side chain
cleavage). Cholesterol is stored in the cells largely as cholesterol esters, and
the cholesterol esterase is activated by protein phosphorylation induced by
cyclic AMP. However, uptake of cholesterol into the cell and its transfer from
the cytoplasm into the mitochondrial matrix are also both rapidly stimulated
8.1 A new view of the physiological control of flux 271

by mechanisms not entirely clear. Esterification of cholesterol is possibly

also inhibited by ACTH, but this is controversial. Both mitochondrial P–450
enzymes involved in the pathway (P–450SCC and P–45011β ) appear to be
rapidly activated by ACTH; the latter by a mechanism involving cyclic AMP,
the former by a mechanism that is still unknown but might involve cyclic
AMP.
The longer term effects of ACTH involve action on all aspects of meta-
bolism that affect the amount of the common precursor steroid, cholesterol,
its conversion to the common intermediate pregnenolone, and many sites in
the final formation of the various products by oxidation.181, 267 The synthesis
rates are increased of the four cytochrome P–450 enzymes (P–450SCC , P–
45011β , P–45017α and P–450C21 ) and of their associated electron transport
elements: the Fe–S protein adrenodoxin; the enzyme adrenodoxin reductase
that catalyzes its reduction by NADPH in the mitochondria, and NADPH–
cytochrome P–450 reductase in the endoplasmic reticulum. ACTH also exerts
long–term activation of the uptake of cholesterol into the cell and increases
the amounts of cholesterol esterase. . In other related branches of metabol-
ism, esterification of cholesterol for storage in lipid droplets is inhibited whilst
increased amounts of isocitrate dehydrogenase (NADP+ ), malic enzyme and
glucose–6–phosphate dehydrogenase provide NADPH for the hydroxylation
reactions.138 The synthesis of cholesterol from acetate is also stimulated over
about 36 hours; increased activity of one of the early enzymes of the pathway,
hydroxymethyl glutaryl–CoA reductase, has been demonstrated but does not
match the time course of the stimulation of the pathway flux, which could be
evidence that synthesis of other enzymes is required as well.191 Significantly,
hydroxymethyl glutaryl–CoA reductase is an enzyme known to be controlled
by reversible phosphorylation (Chapter 7.4.3.1.4).

8.1.1.5 Photosynthesis
Finally, light–dependent activation of plant photosynthesis in the chloroplasts
and associated metabolism in the plant cytoplasm illustrates rapid control
mechanisms acting at multiple sites throughout a metabolic network (point
4). As well as the diurnal cycle, plants are subject to continual fluctuations in
illumination from light flecks (shifting patches of sunlight filtering through the
leaf canopy) and changing orientations of their leaves as they bend and shake
in the wind. These transient increases in illumination can be responsible for a
significant fraction of daily carbon fixation and can cause sudden increases in
assimilation of CO2 of the order of 10–fold on a time scale of a few seconds.180
There are several components to plant photosynthesis. For the C3 plants
typical of temperate regions these are:
• the light reactions in the chloroplast, where photosystems I and II har-
ness the light energy on the thylakoid membranes and generate ATP
and NADPH;
• the so–called dark reactions in the chloroplast that use the ATP and
272 8 Conclusion

NADPH to fix CO2 with ribulose bisphosphate carboxylase (rubisco),

regenerate the ribulose bisphosphate by the Calvin cycle and store pho-
tosynthate as starch, and

• the processing of photosynthate in the cytosol, by the formation and

export primarily of sucrose but also of amino acids.

Short–term variations in the rate of the light reactions largely follow the
availability of light; however, the balance between the production of ATP
and NADPH is kept by regulating the ratio between non–cyclic photophos-
phorylation (which uses both photosystems and produces both products) and
cyclic photophosphorylation, which uses only photosystem I and produces
only ATP. This regulation involves reversible phosphorylation of the light–
harvesting component of photosystem II by a protein kinase that responds to
the redox state of the intermediate electron transport chain between the two
photosystems,15 but is not the aspect I wish to focus on here. Rather, I want
to look at how the assimilation of carbon in the chloroplast and cytoplasm
(Fig. 8.4) is controlled in accordance with the fluctuations in the supply of

$
cytosol CO2

' $
pep 4
⊕
- oxaloacetate
chloroplast
NH+ -NH+ - amino -
4 4 acids
6
CO2
1
RuBP
NO− ¾ NO− ¾ 3
NO− ¾
2 2 3
7
⊕ ⊕ 6
6
⊕
N
Ru5P PGA protein dephosphorylation
O
⊕ ⊕ ⊕
1 5
) - triose P -hexose P ?
- sucrose P
triose P
2
? - Pi
?
hexose P
sucrose -
6
ª
?

& %
starch

Fig. 8.4: Effects of light on photosynthesis in C3 plants.

Key: ⊕, light activation by a variety of mechanisms; ª, light inactivation. Ab-
breviations: pep, phosphoenol pyruvate; PGA, phosphoglycerate; Ru5P, ribulose–5–
phosphate; RuBP, ribulose–1,5–bisphosphate. Enzymes: 1, pentose pathway — ac-
tivable enzymes are fructose bisphosphatase and sedoheptulose bisphosphatase; 2, su-
crose phosphate synthase; 3, nitrate reductase; 4, phosphoenol pyruvate carboxykinase;
5, glyceraldehyde–3–phosphate dehydrogenase; 6, rubisco; 7, ribulose–5–phosphate ki-
nase.
8.1 A new view of the physiological control of flux 273

ATP and NADPH from the light reactions. The levels of ATP and NADPH
change relatively little during the day, so there is obviously control on the de-
mand reactions (i.e. the dark reactions) in accordance with the requirement
for effective homeostasis in supply–demand pathways.
The activation state of four of the enzymes of the Calvin cycle was discov-
ered by Buchanan to depend on light–mediated reduction carried out by small
proteins called thioredoxins.3 This is a covalent modification reaction involv-
ing the oxidation state of cysteine sulphydryl groups in the enzymes. When
they are oxidized to form disulphide linkages, the enzymes are inactive; when
they are reduced to the sulphydryl form, the enzymes are active. Thioredox-
ins are abundant in chloroplasts and themselves contain adjacent sulphydryl
groups that can be oxidized to disulphides. They are oxidized by atmospheric
oxygen, but are reduced by ferredoxin, the electron transport component
that is itself reduced by photosystem I and that in turn reduces NADP+ to
NADPH. Thus the redox state of thioredoxin senses the activation of the light
reactions and transmits this to activate phosphoribulokinase, glyceraldehyde–
3–phosphate dehydrogenase, fructose–1,6–bisphosphatase and sedoheptulose
bisphosphatase. The system can also respond to the state of the Calvin cycle
because the susceptibility of the enzymes to reduction is also modulated by
levels of the Calvin cycle intermediates.80 Ribulose–bisphosphate carboxylase
itself, the enzyme that actually fixes the CO2 , is controlled by multiple effects
including a covalent modification: an activation by carbamylation of a lysine
group promoted by rubisco activase, which is itself activated by light. 80 In
some plants, rubisco is inhibited by carboxyarabinitol–1–phosphate, which is
synthesized in the dark and degraded in the light;180 again, rubisco activase is
involved in releasing rubisco from this inhibition. Calvin cycle enzymes also
respond to the light–induced changes in stromal pH and Mg2+ concentra-
tion277 that occur as the light reactions pump protons out of the stroma into
the thylakoid space. In C4 plants, such as tropical grasses, the activity of one
of the key enzymes in their alternative mode of CO2 fixation, pyruvate phosph-
ate dikinase is controlled by phosphorylation in a light–dependent manner. 110
In addition, light–dependent changes in protein phosphorylation activate en-
zymes of the assimilatory pathways in the cytoplasm (Fig. 8.4), including
sucrose phosphate synthase, nitrate reductase and phosphoenol pyruvate car-
boxykinase110 presumably so that the relative amounts of starch, sucrose and
amino acids being synthesized are kept in balance both with one another and
with the amount of CO2 being fixed.
Much remains to be learnt about the molecular mechanisms underlying the
several different light–activated control systems operating in photosynthesis.
There are also differences between different types of plant, such as C3 and C4
plants. However, even though the account given above is not exhaustive in
terms of the enzymes whose activities are controlled by light, it is sufficient to
demonstrate that the rate of carbon assimilation is very far from being con-
trolled by a single enzyme such as rubisco. Indeed, the measurements of the
flux control coefficient of rubisco, made by Mark Stitt’s group and described
in Chapter 6.1.1.5, p. 149, reveal how unlikely this would be: under most
274 8 Conclusion

circumstances the control coefficient is not large enough to allow significant

stimulation of carbon assimilation by activation of rubisco alone.

8.1.2 Implications of multisite modulation

As seen from the examples, there are undoubtedly cases where environmental
or physiological signals that induce large changes in metabolic flux do so by
acting on many, or even all, of the enzymes in a pathway. It is difficult to
find examples of the negative case: large changes in flux indubitably initiated
solely by action on a single enzyme. One possible contradictory example is the
stimulation of tryptophan catabolism by nutritional or hormonal induction of
tryptophan dioxygenase, as discussed in Chapter 6.1.2. Here the catabolic
flux does seem to be stimulated by a selective increase in the amount of this
pathway enzyme alone; on the other hand, the degree of stimulation of the
flux is relatively small and the pathway is short and soon enters the main areas
of catabolism which are of far higher capacity. So if, as the theory suggests,
multisite modulation is the only effective way to achieve large changes in
metabolic flux in the majority of cases, some uncomfortable conclusions follow:
• The rate–limiting step concept is more misguided than even Metabolic
Control Analysis initially suggested. Under its influence, biochemists
have searched for a single or a small number of key control points when
the important control mechanisms are those that have multiple sites of
action in metabolism.
• If control is exerted by multisite modulation, the effects are synergestic.
There will be deleterious effects from interfering with the response at
any of the control sites, and the response from a single control site
operated in isolation will always be relatively poor. Characterizing the
importance of individual steps to an overall flux response will be very
difficult since the overall response is a property of the system itself, not
of its components.
• If the physiological control mechanisms are interlinked throughout a
pathway, experiments in Metabolic Control Analysis will not be easy to
design because it will be difficult to find ways of perturbing a specific
point in the pathway without the possibility that the control mechanisms
will transmit the perturbation to other parts of the pathway by routes
that have not been taken into account.
Addressing these problems is not a simple issue, even within Metabolic
Control Analysis. Its suitability to take on this challenge is discussed in the
next section.

8.2 Limitations of Control Analysis

I have neither presented all aspects of Metabolic Control Analysis nor its full
complexity in this book. This would obviously have been inappropriate in
8.2 Limitations of Control Analysis 275

an introductory text. Nevertheless, I believe I have presented enough of the

central aspects of the theory for the reader to consider, and form a judgement
on, some of the criticisms of Metabolic Control Analysis as I have described
it here. I shall consider these criticisms in two groups:

• Objections to the formulation and general approach of Control Analysis

as a whole, and

• Criticisms of over–simplifications in Control Analysis, which have led to

changes and extensions to the theory.

8.2.1 Objections to Control Analysis

There have been arguments against both the name and the concept of the
control coefficients. Some of the principal grounds cited have been that:

1. Enzyme concentration is not a parameter that is particularly relevant

to metabolic control e.g.,8, 56 compared with the action of effectors that
bind to allosteric enzymes.

2. The sensitivity of a flux to an enzyme concentration is not a measure

of whether that enzyme is a control or regulatory enzyme,8, 56, 218 so the
term is misleading.

3. The flux control coefficient has limited predictive value because it is

only valid under the conditions of measurement, and as the state of the
system changes, the values of the flux control coefficients change.56, 259

The first of these criticisms has been made particularly forcefully by both
Eric Newsholme and Dan Atkinson. Newsholme, with Bernard Crabtree, has
developed an alternative theoretical approach to metabolic control which in-
volves characterizing the net sensitivity of a pathway to an effector of an
enzyme.55 If necessary, a hypothetical effector is assumed to exist in order
to allow definition of the net sensitivity. In Metabolic Control Analysis, the
equivalent measure is the response coefficient (Chapter 5.4), so the key dif-
ference between the approaches is that in Control Analysis the flux control
coefficient is separated out as an individual entity that can be combined with
different elasticities to indicate the responses of the pathway flux to a range
of effectors whereas in Newsholme’s theory the net sensitivity is the primary
entity. Atkinson8 objected that the key feature in control of metabolism is the
action of effectors of allosteric enzymes which frequently change the affinity of
an enzyme for its substrates, not the catalytic activity. Again, the response
coefficient of the flux to the effector, formed from the flux control coefficient
of the enzyme and the elasticity of the enzyme to the effector, provides a mea-
sure of the strength of the response. Thus it is the belief of the advocates of
Metabolic Control Analysis, admittedly not shared by the objectors, that the
measures they seek are provided in Control Analysis. A more fundamental
rebuttal of their objection would be that control by allosteric effectors is not
276 8 Conclusion

the only control mechanism; the equally important effects of genes, DNA ma-
nipulation, environment, diet and hormones on metabolism can be mediated
through the change in the concentration of the active form of an enzyme, and
the flux control coefficient indicates what the effect of such a change will be.
The second objection returns to a theme that has run through this book.
There is no doubting that regulatory enzymes are often predicted by Control
Analysis to have small flux control coefficients, and the argument is there-
fore that there is little use in a measure that does not identify the important
characteristics of an enzyme. The reasons why regulatory enzymes subject to
feedback regulation have small flux control coefficients has been considered at
length in the Chapter on feedback inhibition (Chapter 7.2). I have sympathy
for the concept that there ought to be a measure of the strength of the regula-
tory role of an enzyme; in fact, some measures have been suggested but have
not been widely adopted. However, as long as molecular biologists are over-
expressing regulatory enzymes in the expectation of altering metabolic flux
and are being disappointed by the results, I believe the priority is to empha-
size the distinction between regulation and control and to persuade molecular
biologists to pay more attention to the flux control coefficient, which could
have been used to predict their disappointment.
There is truth in the third objection, made by Newsholme & Crabtree and
Savageau & Voit. The flux control coefficient is only valid under the condi-
tions of its measurement, and therefore it can make only limited predictions
about what will happen if circumstances change greatly. Everybody would
be happier if it was possible to make accurate predictions about the effects
of large perturbations induced by factors such as hormonal stimulation, but
this is extremely difficult. If we had complete models of metabolism, it would
be possible, but the information needed to build completely accurate models
is not, and may never be, available. If it were, we would not need Metabolic
Control Analysis. Henrik Kacser was sensitive to the force of this criticism
and his response to it was expressed in his characteristically entertaining way
in the last article he completed117 before his death in March 1995. In re-
sponse, he cited his work on the relationship between flux control coefficients
and large changes in enzyme activity (summarized in Chapter 5.2.2), and also
the Universal Method, referred to earlier in this chapter (section 8.1) for ob-
taining large changes in flux. As he pointed out, these go beyond the realm
where Metabolic Control Analysis is strictly valid, but have depended on the
insights from Control Analysis. Savageau’s criticism reflects the different aims
he had in developing his Biochemical Systems Theory, which is intended as
a complete tool for the modelling and analysis of stability and sensitivity. 217
Furthermore, he claims that Metabolic Control Analysis is only a limited sub-
set of his theory;218, 219 I accept that it is possible to view the two theories in
this light, but I and other advocates of Control Analysis feel that our theory
avoids some of the aspects of Biochemical Systems Theory that we regard as
undesirable.48
There have also been criticisms of the roles of the theorems, such as the
summation and connectivity theorems, in Metabolic Control Analysis. Kacser
8.2 Limitations of Control Analysis 277

& Burns119 gave them a central role in the development of the theory, and I
have used them to provide the explanations in this book. Savageau and col-
leagues219 have claimed that they are unnecessary because they are implicit
in the mathematical description of a metabolic system at steady state. It is
also true that Christine Reder in Bordeaux, Marta Cascante in Barcelona and
Christoph Giersch in Darmstadt have shown that pathways can be analyzed
in terms of Metabolic Control Analysis without assuming the existence of the
theorems,31, 32, 83, 195 and that the theorems then emerge from the analysis.
This is extremely useful work because it provides deeper mathematical foun-
dations for Control Analysis. It is not the same as saying the theorems have
no use; they summarize important properties that assist in explanations of
how control is distributed and make a link to a biochemical interpretation of
the mathematics. But at this stage, I am content to abide by the reader’s
judgement: take a look at some of the papers cited in this section, and see if
you prefer their approach.

8.2.2 Extensions of Control Analysis

Only the simplest and most straightforward aspects of Metabolic Control
Analysis have been presented in this book. However, over the last 10 years
there has been considerable development of the theory. Much of this has been
driven by the urge to examine claims that Metabolic Control Analysis would
not be able to deal with metabolic pathways containing particular features
that have not been taken into account during the development of the theory.
In most cases it has proved possible to use the definitions and formalisms of
Metabolic Control Analysis to incorporate these features, but at the expense
of greater complexity in the mathematics and a loss of validity of the the
theorems. There are three main areas where potential problems could arise:
• the degree of connectedness of the metabolic network;
• the relationship between enzyme amount and catalytic activity, and
• the validity of metabolite concentration measures.
I shall briefly summarize the complications to which each of these can lead.

8.2.2.1 Connectedness of metabolic networks

In the treatment of Metabolic Control Analysis given in this book I did not
choose to show the proof that the summation theorem applies over a network
that is fully connected by mass flows; in other words, it is a statement about
a reaction network where the material flowing in can be traced all the way to
the outputs, and every reaction in the network is carrying some of that flow.
On this basis, two pathways occurring simultaneously that involve completely
separate metabolites and have uncorrelated fluxes should be completely in-
dependent and each should have its own summation theorem with respect to
the enzymes that catalyze its reactions. But what if the metabolites from
278 8 Conclusion

one influenced the rates of the other, even though they did not appear as
substrates or products of its reactions? It is not difficult to find examples:
covalent modification cascades consist of a hierarchy of pathways where the
substrates and products of one pathway are the catalysts of the next path-
way down, but do not participate in the flux of that pathway. Transcription
and translation of genes to form enzymes, which then catalyze pathways is
another example where the mass flows at the protein synthesis level might
be weakly coupled to the mass flows at pathway level, but again there is a
catalytic relationship between them, and even the possibility of influence in
the reverse direction through effects of metabolites from the pathway on the
transcription and translation of the enzymes.
One possible solution in simple cases is to define the response of the path-
way to events occurring at a higher level in an ad hoc way. An example is
the treatment that Rankin Small and I gave to covalent modification reac-
tions, described in Chapter 7.4.5. However, Daniel Kahn in Toulouse and
Hans Westerhoff in Amsterdam have developed a Modular (or hierarchical)
Control Analysis that uses the same mathematical formulation as Metabolic
Control Analysis, but which can handle these problems systematically. 123, 224
In essence, the solutions contain terms that represent the standard Control
Analysis of the separate pathway components, but these are modified by in-
teraction terms between the pathways. The approach has much potential,
but is not for the mathematically faint–hearted; experimental applications
are under way, but are not yet concluded.

8.2.2.2 Enzyme properties

As was briefly mentioned in Chapter 5.7, the Appendix More about flux control
coefficients, the simple form of Metabolic Control Analysis presented here has
defined flux control coefficients in terms of the relative change in flux for a
relative change in enzyme concentration. A hidden assumption is that the
rate of an enzymic process is proportional to the enzyme concentration, and
that one enzyme acts uniquely on a single metabolic step. There are a number
of conceivable cases where this assumption is not obeyed:
• Where an enzyme catalyzes more than one step in the same metabolic
pathway,30 as occurs in purine metabolism and the pentose phosphate
pathway.
• Where the enzyme undergoes association equilibria, either with other
molecules of the same enzyme, or molecules of a different enzyme, and
the aggregates have different catalytic properties from the free enzymes. 122, 212
In the first case, the relationship between the enzyme concentration and
the rate of the step is not linear; in the second, the two enzymes involved
could each affect the rate of the other’s step.
• Where the enzyme is subject to covalent modification,231 unless the
enzyme concentration is defined in terms of the active form only, rather
than the total.
8.2 Limitations of Control Analysis 279

The effect of most of these is that:

• The value of the flux control coefficient will depend on whether it is
defined relative to enzyme concentration or to some other parameter
that does affect enzymic rate linearly.
• Different experimental approaches could give different results for the
flux control coefficient.223 For example, a genetic manipulation that
changes the total amount of an enzyme might not give the the same
result as an inhibitor titration that changes the activity of the enzyme.
• The summation theorem may be invalidated.
However, each of these difficulties can be formulated in terms of Metabolic
Control Analysis; it is just that the resulting analysis becomes more complex
than the cases illustrated in this book. Are these potential problem cases
frequent enough and severe enough to invalidate the application of Metabolic
Control Analysis? They have not shown up in any of the experimental ap-
plications studied so far, but this might be because it is difficult to do the
experiments with sufficient accuracy to detect the small deviations that might
occur. My subjective impression is that until the accuracy and precision of
the methods for measuring flux control coefficients has been improved sub-
stantially, these problems will remain a minor issue.

8.2.2.3 Metabolite concentrations

The third and most difficult of the problems is the appropriateness of metabo-
lite concentration measures. Here I am not referring to the experimental dif-
ficulties in making meaningful measurements discussed in Chapter 2.3, since
these are neither better nor worse for Control Analysis than for any other
mode of interpreting metabolic control and regulation. This is a deeper prob-
lem relating to the underlying assumption in Metabolic Control Analysis that
there exists a single value for the concentration of a metabolite that enters the
equations and determines the values of coefficients such as the elasticity coef-
ficients. In other words, the concentration must be adequately represented by
a single value that is equally applicable to all the enzyme molecules. Again,
the problem here does not come from the possibility of compartmentation in
the cell, provided that a unique value can be assigned to the concentration
in each compartment. It is more the possibility that the concentration of a
metabolite varies with its position within a single compartment. This could
happen in a system where the reactions do not occur uniformly throughout
the space, and the rate of diffusion is not sufficiently fast relative to the re-
action rates to cause completely uniform mixing. I have calculated that this
could occur with cyclic AMP in some cells, for it is synthesized only at the
plasma membrane but is hydrolyzed throughout the cell, and this could give
rise to cyclic AMP concentration gradients in the cell.68 Such systematic
spatial variation in metabolite concentration has not been taken into account
in Control Analysis. On the other hand, diffusion of metabolites is a fast
280 8 Conclusion

process on the scale of cellular dimensions, and is likely to be able to keep

concentration gradients in the aqueous compartments small in most cases.
A related problem, but one that also involves aspects of enzyme–enzyme
interaction is the effect of channelling. This refers to cases where the common
metabolite between two enzymes is not released into the medium but is di-
rectly transferred between them, either in a static channel (one that involves
a relatively permanent association of the enzymes) or a dynamic channel (one
where the metabolite is transferred in a transient association between the
enzymes). This is a controversial area (see, for example, ref. [178] and the
papers in the same issue of that journal), with respect to the frequency of
occurrence, the mechanisms involved, and the effects (if any) that channelling
has on metabolite concentrations in the bulk medium. Again, it is possible to
modify the theory of Metabolic Control Analysis to cover this case131, 132, 210
but the question remains how often any effects will be large enough to be
detectable with current experimental techniques.
One final worry about metabolite concentrations has turned out not to be
a great cause for concern at all. In some pathways where the enzymes are
relatively abundant (glycolysis, and rubisco in photosynthesis for example),
the enzyme concentrations can be higher than the metabolite concentrations. 1
Some people were concerned that if the enzyme levels were not negligible
compared with the metabolites, the theory would be invalidated. In fact, there
should be no problem in terms of the steady state behaviour of the pathway
provided that it is the free and not the total metabolite concentrations that
appear in any of the equations. With Herbert Sauro, I identified one case
where the effect might matter,70 and this was where an enzyme present at
relatively high concentration was binding a significant fraction of a moiety–
conserved grouping (e.g. the adenosine of the adenine nucleotides, or the
NAD+ /NADH pair). Even so, deviations in the summation theorems and
discrepancies between different definitions of the flux control coefficients only
become significant if the amount of complexed metabolite on the enzyme is a
significant fraction of the total of the metabolites in the conserved pool (not a
single member of the group). This circumstance is likely to be much more rare
than simple comparability between enzyme and metabolite concentrations,
and again there are no experiments suggesting a problem at the moment.

8.3 Applications of Metabolic Control Analysis

Finally, what is the use of Metabolic Control Analysis? I hope that by now
I have persuaded you that Metabolic Control Analysis provides a framework
within which it is possible to carry out a rational and quantitative exam-
ination of the regulation and control of metabolism. This can be by the
design and execution of appropriate experiments, as described in Chapter 6,
or by theoretical analysis of common metabolic structures as illustrated in
Chapter 7 and this chapter. Either way, the results have been a challenge to
much traditional thinking in biochemistry and the debate will undoubtedly
8.3 Applications of Metabolic Control Analysis 281

continue.
Because this is a biochemistry book, it would be easy to overlook the
important implications of Metabolic Control Analysis for genetics, which
were mentioned in Chapter 5.2.2 in connection with its relevance to the phe-
nomenon of dominance. Putting biochemistry and genetics together links
Metabolic Control Analysis with the evolution of metabolic pathways and even
with population ecology. These aspects are again more theoretical than prac-
tical, but Jean–Pierre Mazat’s group in Bordeaux has been using Metabolic
Control Analysis to diagnose and understand the genetic disorders known as
mitochondrial cytopathies.149, 150 These are a diverse group of inborn errors
of metabolism with variable symptoms. They are caused by mutations in the
mitochondrial genome, but the variability arises because there are multiple
copies of the mitochondrial DNA in every human cell, so that the proportion
of mutant to wild–type genes can be almost continuously variable, in con-
trast to nuclear genes where there are generally just two copies per cell. The
variability in the symptoms reflects the non–linear relationship between the
amount of an enzyme and the pathway flux. For many of the enzymes, the
content can be reduced significantly with limited effects on flux, and therefore
with little in the way of symptoms, but beyond a certain point (an apparent
threshold), the flux decreases more rapidly with enzyme content, and the res-
piration rate of the mitochondria becomes progressively more limited by the
step with the mutation.
Another potential area of application of Control Analysis in medicine is the
identification of particularly suitable sites for the manipulation of metabolism
with drugs. This is not a new idea; Hans Krebs141 advocated it in 1957,
but he considered that the answer would be to find the pacemaker enzymes.
The concept has been reformulated in Metabolic Control Analysis terms by
Chris Pogson and his group at the Wellcome pharmaceutical company.208
They reasoned that even though it was unlikely that true pacemaker, or rate–
limiting, enzymes existed, it would nevertheless still be worthwhile in the
drug–discovery process to target the enzymes with the highest flux control
coefficients. Very often, the aim with a drug is to inhibit a metabolic process,
such as an essential pathway in a parasite or cancer cell, and in this case, any
enzyme in the process is a potential target because every enzyme in a sequence
is essential for the sequence to work. Even so, the effects on metabolism are
likely to be obtained with lower concentrations of drug if an enzyme with a
high flux control coefficient is being inhibited, rather than one with a low
coefficient.
The area where Metabolic Control Analysis could make the biggest impact
in the near future is the area of biotechnology known as metabolic engineering.
Modern genetic engineering techniques have the potential to sidestep the tra-
ditional genetic approaches of selective breeding for organisms with a desired
characteristic by permitting directed, rational changes that could modify the
metabolism of organisms in a desired direction. Generally, increasing the flow
in a specific pathway is the more usual goal, and as discussed earlier in this
Chapter, increasing the flux is a more difficult problem than decreasing it. In
282 8 Conclusion

this area, Metabolic Control Analysis is already winning the argument over
traditional concepts of metabolic control, because so–called rate–limiting en-
zymes have already been over–expressed in cells with little effect. Identifying
enzymes with large flux control coefficients would allow a more rationally tar-
getted approach, but the two limitations of this approach have already been
discussed:

• It is conceivable that no enzyme in a pathway has a large flux control

coefficient.

• Even where an enzyme has a moderately large flux control coefficient,

the theory of finite changes in enzyme amounts predicts that the effects
on metabolic flux of overexpressing it may be very modest.

There is no doubting the incentives to solve these problems. For example,

plant metabolism might be manipulated to allow sustainable production from
renewable resources of alternative raw materials to those currently derived
from oil (such as plastics from starch). Advocates of Metabolic Control Anal-
ysis should not expect to be popular for suggesting that our knowledge of
metabolic control shows such goals may be much more difficult to attain than
is commonly thought. This was one of the problems that Henrik Kacser’s
‘Universal Method’118 (discussed earlier in section 8.1) was designed to solve,
but the procedures it predicts to be necessary in order to change flux will not
be simple to carry out. There is much scope here for further work.

8.4 Summary
1. Control of flux cannot be exerted by one or a few enzymes at the be-
ginning of a pathway when large changes in pathway flux are seen to be
associated with much smaller relative changes in metabolite concentra-
tions.

2. In biotechnological applications, to ensure a significant increase in a

metabolic flux without a marked change in metabolite concentrations,
Kacser’s Universal Method predicts that it is necessary to simultane-
ously activate many, even most, of the pathway enzymes.

3. Examples of significant flux changes in a number of pathways provide

evidence that these too are achieved by multisite modulation, i.e. the
parallel activation of several, many, or even most of the enzymes of the
pathway.

4. Control Analysis alone may not be able to tackle all conceivable prob-
lems in regulation and control. Nevertheless, it is a better starting point
for aquiring a deeper understanding than the qualitative principles of
conventional biochemistry.
8.4 Summary 283

Further reading
1. Fell, D. A. (ed.) The control of flux: 21 years on. Biochem. Soc. Trans.
23, 341–391 (1995).
284
References

1. Albe, K. R., Butler, M. H. and Wright, B. E. (1990) J. Theor. Biol. 143, 163–195
2. Albe, K. R. and Wright, B. E. (1994) J. Theor. Biol. 169, 243–251
3. Anderson, L. E. (1986) Adv. Bot. Res. 12, 1–46
4. Asante, E. A., Hill, W. G. and Bulfield, G. (1991) Genet. Res. Camb. 58, 123–127
5. Ashworth, J. M. and Kornberg, H. L. (1963) Biochim. Biophys. Acta 73, 519–522
6. Atkinson, D. E. (1969) Curr. Top. Cell. Regul. 1, 29–43
7. Atkinson, D. E. (1977) Cellular Energy Metabolism and its Regulation., Academic
Press, New York
8. Atkinson, D. E. (1990) in Cornish-Bowden, A. and Cárdenas, M. L., eds., Control
of Metabolic Processes, pp. 3–11, New York, Plenum Press
9. Babul, J., Clifton, D., Kretschmer, M. and Fraenkel, D. G. (1993) Biochemistry 32,
4685–4692
10. Baranyi, J. M. and Blum, J. J. (1989) Biochem. J. 258, 121–140
11. Bartrons, R., Van Schaftingen, E., Vissers, S. and Hers, H.-G. (1982) FEBS Lett.
143, 137–140
12. Becker, G. L., Fiskum, G. and Lehninger, A. L. (1980) J. Biol. Chem. 255, 9009–9012
13. Beebe, S. J. and Corbin, J. D. (1986) in Boyer, P. D. and Krebs, E. G., eds., The
Enzymes, vol. XVII, pp. 43–111, Academic Press, New York, 3rd edn.
14. Benevolensky, S. V., Clifton, D. and Fraenkel, D. G. (1994) J. Biol. Chem. 269,
4876–4882
15. Bennett, J. (1991) Annu. Rev. Plant. Physiol. Plant Mol. Biol. 42, 281–311
16. Berry, M. N. and Friend, D. S. (1969) J. Cell Biol. 43, 506–520
17. Berthon, H. A., Bubb, W. A. and Kuchel, P. W. (1993) Biochem. J. 296, 379–389
18. Blackman, F. F. (1905) Ann. Bot. 19, 281
19. Blum, J. B. and Stein, R. B. (1982) in Goldberger, R., ed., Biological Regulation
and Development, vol. 3A, chap. 3, pp. 99–125, Plenum Press, New York
20. Brand, M. D. (1993) in Hochachka, P. W., Lutz, P. L., Sick, T., Rosenthal, M.
and van den Thillart, G., eds., Surviving Hypoxia: Mechanisms of Control and
Adaptation, chap. 20, pp. 295–309, CRC Press, Boca Raton
21. Brand, M. D., Hafner, R. P. and Brown, G. C. (1988) Biochem. J. 255, 535–539
22. Brindle, K. M. (1988) Biochemistry 27, 6187–6196
23. Brown, G. C., Lakin-Thomas, P. L. and Brand, M. D. (1990) Eur. J. Biochem. 192,
355–362
24. Bücher, T. and Rüssmann, W. (1964) Angew. Chem., Internat. Ed. 3, 426–439,
Originally published as Angew. Chem., 73, 881 in 1963.
25. Bulfield, G. (1972) Genet. Res. Camb. 20, 51–64
26. Cahill, G. F., Ashmore, J., Renold, A. E. and Hastings, A. B. (1959) Am. J. Med.
26, 264–282
27. Cárdenas, M. L. and Cornish-Bowden, A. (1989) Biochem. J. 257, 339–345
28. Carling, D. J., Zammit, V. A. and Hardie, D. G. (1987) FEBS Lett. 223, 217–222
29. Carling, D. J., Clarke, P. R., Zammit, V. A. and Hardie, D. G. (1989) Eur. J.
Biochem. 186, 129–136
30. Cascante, M., Canela, E. I. and Franco, R. (1990) Eur. J. Biochem. 192, 369–371
31. Cascante, M., Franco, R. and Canela, E. I. (1989) Math. Biosci. 94, 271–288

285
286 References

32. Cascante, M., Franco, R. and Canela, E. I. (1989) Math. Biosci. 94, 289–309
33. Chance, B., Holmes, W., Higgins, J. J. and Connelly, C. M. (1958) Nature (London)
182, 1190–1193
34. Chance, B. and Williams, G. R. (1955) J. Biol. Chem. 217, 409–427
35. Chance, B., Williams, G. R., Holmes, W. F. and Higgins, J. (1955) J. Biol. Chem.
217, 439–451
36. Chance, E. M., Seeholzer, S. H., Kobayashi, K. and Williamson, J. R. (1983) J. Biol.
Chem. 258, 13785–13794
37. Chelsky, D., Ruskin, B. and Koshland Jr., D. E. (1985) Biochemistry 24, 6651–6658
38. Chock, P. B., Shacter, E., Jurgensen, S. R. and Rhee, S. G. (1985) Curr. Top. Cell.
Regul. 27, 3–12
39. Clark, M. G., Bloxham, D. P., Holland, P. C. and Lardy, H. A. (1973) Biochem. J.
134, 589–597
40. Clark, M. G., Kneer, N. M., Bosch, A. L. and Lardy, H. A. (1974) J. Biol. Chem
249, 5695–5703
41. Cohen, G. N. (1969) Curr. Top. Cell. Regul. 1, 183–231
42. Cohen, P. (1985) Curr. Top. Cell. Regul. 27, 23–37
43. Cohen, P. (1989) Annu. Rev. Biochem. 58, 453–508
44. Cohen, P. and Chen, P. T. W. (1989) J. Biol. Chem. 264, 21435–21438
45. Cohen, P. and Klee, C. B., eds. (1988) Calmodulin. Molecular Aspects of Cellular
Regulation, Vol. 5., Amsterdam, Elsevier
46. Cohen, S. M. (1987) Biochemistry 26, 573–580
47. Cornish, A., Greenwood, J. A. and Jones, C. W. (1988) J. Gen. Microb. 88, 3111–
3122
48. Cornish-Bowden, A. (1989) J. Theor. Biol. 136, 365–377
49. Cornish-Bowden, A. (1995) Fundamentals of Enzyme Kinetics, Portland Press, Lon-
don, 2nd edn.
50. Cornish-Bowden, A. and Cárdenas, M. L. (1987) J. Theor. Biol. 124, 1–23
51. Cornish-Bowden, A. and Cárdenas, M. L. (1991) Trend Biochem. Sci. 16, 218–282
52. Cornish-Bowden, A. and Hofmeyr, J.-H. S. (1991) Comp. Appl. Biosci. 7, 89–93
53. Cornish-Bowden, A. and Hofmeyr, J.-H. S. (1994) Biochem. J. 298, 367–375
54. Crabtree, B. (1976) Biochem. Soc. Trans. 4, 999–1002
55. Crabtree, B. and Newsholme, E. A. (1985) Curr. Top. Cell. Regul. 25, 21–76
56. Crabtree, B. and Newsholme, E. A. (1987) Biochem. J. 247, 113–120
57. Crabtree, B. and Newsholme, E. A. (1987) Trends Biochem. Sci. 12, 4–12
58. Crawford, J. M. and Blum, J. J. (1983) Biochem. J. 212, 585–598
59. Davies, S. E. C. and Brindle, K. M. (1992) Biochemistry 31, 4729–4735
60. Denton, R. M. and Pogson, C. I. (1976) Metabolic Regulation., Chapman Hall,
London
61. Dykhuizen, D. E., Dean, A. M. and Hartl, D. L. (1987) Genetics 115, 25–31
62. Easterby, J. S. (1981) Biochem. J. 199, 155–161
63. Easterby, J. S. (1986) Biochem. J. 233, 871–875
64. Easterby, J. S. (1990) Biochem. J. 269, 255–259
65. Eisenthal, R. and Cornish-Bowden, A. (1974) Biochem. J. 139, 715–720
66. El-Yassin, D. I. and Fell, D. A. (1982) J. Mol. Biol. 156, 863–889
67. Endrenyi, L. and Chan, F.-Y. (1981) J. Theor. Biol. 90, 241–263
68. Fell, D. A. (1980) J. Theor. Biol. 84, 361–385
69. Fell, D. A. and Sauro, H. M. (1985) Eur. J. Biochem. 148, 555–561
70. Fell, D. A. and Sauro, H. M. (1990) Eur. J. Biochem. 192, 183–187
71. Fell, D. A. and Snell, K. (1988) Biochem. J. 256, 97–101
72. Fell, D. A. and Thomas, S. (1995) Biochem. J. 311, 35–39
73. Fiegelson, M. and Fiegelson, P. (1965) Adv. Enzyme Regul. 3, 11–27
74. Flint, H. J., Tateson, R. W., Bartelmess, I. B., Porteous, D. J., Donachie, W. D.
and Kacser, H. (1981) Biochem. J. 200, 231–246
75. Forsberg, H., Zetterqvist, O. and Engström, L. (1969) Biochim. Biophys. Acta 181,
171–175
76. Fraenkel, D. G. (1992) Annu. Rev. Genet. 26, 159–177
References 287

77. Galazzo, J. L. and Bailey, J. E. (1990) Enzyme Microb. Technol. 12, 162–172
78. Garfinkel, D. (1971) Comput. Biomed. Res. 4, 18–42
79. Garfinkel, D. (1981) Trends Biochem. Sci. 6, 69–71
80. Geiger, D. R. and Servaites, J. C. (1994) Annu. Rev. Plant Physiol. Plant Mol. Biol.
45, 235–256
81. Gellerich, F. N., Kunz, W. S. and Bohnensack, R. (1990) FEBS Lett. 274, 167–170
82. Gerhart, J. C. (1970) Curr. Top. Cell. Regul. 2, 275–325
83. Giersch, C. (1988) J. Theor. Biol. 134, 451–462
84. Giersch, C. (1994) J. Theor. Biol. 169, 89–99
85. Giersch, C. (1995) Eur. J. Biochem. 227, 194–201
86. Giersch, C., Lämmel, D. and Farquhar, G. (1990) Photosynth. Res. 24, 151–165
87. Giersch, C., Lämmel, D. and Steffen, K. (1990) in Cornish-Bowden, A. and Cárdenas,
M. L., eds., Control of Metabolic Processes, pp. 351–361, New York, Plenum Press
88. Goldbeter, A. and Koshland, D. E. (1981) Proc. Natl. Acad. Sci. U.S.A. 78, 6840–
6844
89. Groen, A. K. (1984) Quantification of control in studies on intermediary metabolism.,
Ph.D. thesis, University of Amsterdam
90. Groen, A. K., van der Meer, R., Westerhoff, H. V., Wanders, R. J. A., Akerboom,
T. P. M. and Tager, J. M. (1982) in Sies, H., ed., Metabolic Compartmentation, pp.
9–37, Academic Press, London
91. Groen, A. K., van Roermund, C. W. T. Vervoorn, R. C. and Tager, J. M. (1986)
Biochem. J. 237, 379–389
92. Groen, A. K., Wanders, R. J. A., Westerhoff, H. V., van der Meer, R. and Tager,
J. M. (1982) J. Biol. Chem. 257, 2754–2757
93. Haber, J. E. and Koshland, D. E. (1967) Proc. Natl. Acad. Sci. U.S.A. 58, 2087–2093
94. Hafner, R. P., Brown, G. C. and Brand, M. D. (1990) Eur. J. Biochem. 188, 313–319
95. Hales, C. N. (1967) in Campbell, P. N. and Greville, G. D., eds., Essays in Biochem-
istry, vol. 3, pp. 73–104, The Biochemical Society / Academic Press, London
96. Heinisch, J. (1986) Mol. Gen. Genet. 202, 75–82
97. Heinrich, R. (1985) Biomed. Biochim. Acta 44, 913–927
98. Heinrich, R. (1990) in Cornish-Bowden, A. and Cárdenas, M. L., eds., Control of
Metabolic Processes, pp. 329–342, New York, Plenum Press
99. Heinrich, R. and Rapoport, T. A. (1974) Eur. J. Biochem. 42, 89–95
100. Heinrich, R. and Rapoport, T. A. (1974) Eur. J. Biochem. 42, 97–105
101. Helmreich, E. and Cori, C. F. (1965) Adv. Enzyme Regul. 3, 91–107
102. Hershko, A. and Ciechanover, A. (1992) Annu. Rev. Biochem. 61, 761–807
103. Hess, B. (1977) Trends Biochem. Sci. 2, 193–195
104. Higgins, J. (1963) Ann. N. Y. Acad. Sci. 108, 305–321
105. Hillgartner, F. B., Salati, L. M. and Goodridge, A. G. (1995) Physiol. Rev. 75, 47–76
106. Hofmeyr, J.-H. S. and Cornish-Bowden, A. (1991) Eur. J. Biochem. 200, 223–236
107. Hohorst, H. J., Reim, M. and Bartels, H. (1962) Biochem. Biophys. Res. Commun.
7, 137–141
108. Holzhütter, H.-G., Jacobasch, G. and Bisdorff, A. (1985) Eur. J. Biochem. 149,
101–111
109. Holzhütter, H.-G., Schuster, R., Buckwitz, D. and Jacobasch, G. (1990) Biomed.
Biochim. Acta 49, 791–800
110. Huber, S. C., Huber, J. L. and McMichael, R. W. (1994) Int. Rev. Cytol. 149, 47–98
111. Hue, L. (1981) Adv. Enzymol. 52, 247–331
112. Hue, L. (1982) in Sies, H., ed., Metabolic Compartmentation, pp. 71–97, Academic
Press, London
113. Jensen, P. R., Michelsen, O. and Westerhoff, H. V. (1993) Proc. Nat. Acad. Sci.
U.S.A. 90, 8068–8072
114. Jensen, P. R., Westerhoff, H. V. and Michelsen, O. (1993) EMBO J. 12, 1277–1282
115. Jensen, P. R., Westerhoff, H. V. and Michelsen, O. (1993) Eur. J. Biochem. 211,
181–191
116. Kacser, H. (1983) Biochem. Soc. Trans. 11, 35–40
117. Kacser, H. (1995) Biochem. Soc. Trans. 23, 387–391
288 References

118. Kacser, H. and Acerenza, L. (1993) Eur. J. Biochem. 216, 361–367

119. Kacser, H. and Burns, J. A. (1973) Symp. Soc. Exp. Biol. 27, 65–104, Reprinted in
Biochem. Soc. Trans. 23, 341–366, 1995.
120. Kacser, H. and Burns, J. A. (1979) Biochem. Soc. Trans. 7, 1149–1160.
121. Kacser, H. and Burns, J. A. (1981) Genetics 97, 639–666
122. Kacser, H., Sauro, H. M. and Acerenza, L. (1990) Eur. J. Biochem. 187, 481–491
123. Kahn, D. and Westerhoff, H. V. (1991) J. Theor. Biol. 153, 255–285
124. Kantrowitz, E. R. and Lipscomb, W. N. (1990) Trends Biochem. Sci. 15, 53–59
125. Kashiwaya, Y., Sato, K., Tsuchiya, N., Thomas, S., Fell, D. A., Veech, R. L. and
Passonneau, J. V. (1994) J. Biol. Chem. 269, 25502–25514
126. Katz, J. and Rognstad, R. (1976) Curr. Top. Cell. Regul. 10, 237–289
127. Katz, J. and Rognstad, R. (1976) Curr. Top. Cell. Regul. 10, 237–289
128. Katz, J., Wals, P. A. and Rognstad, R. (1978) J. Biol. Chem. 253, 4530–4536
129. Kell, D. B. and Westerhoff, H. V. (1986) FEMS Microbiol. Rev. 39, 305–320
130. Kelly, P. J., Kelleher, J. K. and Wright, B. E. (1979) Biochem. J. 184, 589–597
131. Kholodenko, B. N., Cascante, M. and Westerhoff, H. V. (1993) FEBS Lett. 336,
381–384
132. Kholodenko, B. N., Cascante, M. and Westerhoff, H. V. (1994) Mol. Cell Biochem.
133, 313–331
133. King, E. L. and Altman, C. (1956) J. Phys. Chem. 60, 1375–1378
134. Klotz, I. M. (1985) Quart. Rev. Biophys. 18, 227–259
135. Kohn, M. (1983) Ann. of Biomed. Eng. 11, 533–549
136. Kohn, M. C. and Chiang, E. (1983) J. Theor. Biol. 100, 551–565
137. Koshland, D. E., Némethy, G. and Filmer, D. (1966) Biochemistry 5, 365–385
138. Kowal, J. (1970) Recent Prog. Horm. Res. 26, 623–687
139. Krebs, E. G. (1986) in Boyer, P. D. and Krebs, E. G., eds., The Enzymes., vol. XVII,
pp. 1–20, Academic Press, London, 3rd edn.
140. Krebs, H. A. (1946) Enzymologia 12, 88–100
141. Krebs, H. A. (1957) Endeavour 16, 125–132
142. Krebs, H. A. (1963) Adv. Enzyme Regul. 1, 385–400
143. Kruckeberg, A. L., Neuhaus, H. E., Feil, R., Gottlieb, L. D. and Stitt, M. (1989)
Biochem. J. 261, 457–467
144. Lagunas, R. and Gancedo, C. (1983) Eur. J. Biochem. 137, 479–483
145. Lawrence, J. C. (1992) Annu. Rev. Physiol. 54, 177–193
146. Leiser, J. and Blum, J. J. (1985) Cell Biophysics 11, 123–138
147. Lemaigre, F. P. and Rousseau, G. G. (1994) Biochem. J. 303, 1–14
148. Letellier, T. (1992) Contrôle du métabolisme cellulaire: Application aux oxydations
phosphorylantes normales et pathologiques., Ph.D. thesis, Université de Bordeaux
II
149. Letellier, T., Heinrich, R., Malgat, M. and Mazat, J.-P. (1994) Biochem. J. 302,
171–174
150. Letellier, T., Malgat, M. and Mazat, J.-P. (1993) Biochim. Biophys. Acta 1141,
58–64
151. Low, S. Y., Salter, M., Knowles, R. G., Pogson, C. I. and Rennie, M. J. (1993)
Biochem. J. 295, 617–624
152. Maas, W. K. and Clark, A. J. (1964) J. Mol. Biol. 8, 365–370
153. Makins, R. A., Drynan, L. F., Zammit, V. A. and Quant, P. A. (1995) Biochem.
Soc. Trans. 23, 288S
154. Mayorek, N., Grinstein, I. and Bar-Tana, J. (1989) Eur. J. Biochem. 182, 395–400
155. McCormack, J. G. and Denton, R. M. (1986) Trends Biochem. Sci. 11, 258–262
156. Meinke, M. H., Bishop, J. S. and Edstrom, R. D. (1986) Proc. Natl. Acad. Sci.
U.S.A. 83, 2865–2868
157. Meléndez-Hevia, E., Torres, N. V., Sicilia, J. and Kacser, H. (1990) Biochem. J. 265,
195–202
158. Mendes, P. (1993) Comp. Appl. Biosci. 9, 563–571
159. Middleton, R. J. and Kacser, H. (1983) Genetics 105, 633–650
160. Monod, J., Changeux, J.-P. and Jacob, F. (1963) J. Mol. Biol. 6, 306–329
References 289

161. Monod, J., Wyman, J. and Changeux, J.-P. (1965) J. Mol. Biol. 12, 88–118
162. Morgan, H. E., Randle, P. J. and Regen, D. M. (1959) Biochem. J. 73, 573–579
163. Murphy, M. P. and Brand, M. D. (1987) Biochem. J. 243, 499–505
164. Narabayashi, H., Lawson, J. W. R. and Uyeda, K. (1985) J. Biol. Chem. 260, 9750–
9758
165. Neuhaus, H. E., Kruckeberg, A. L., Feil, R. and Stitt, M. (1989) Planta 178, 110–12
166. Neuhaus, H. E., Quick, W. P., Siegl, G. and Stitt, M. (1990) Planta 181, 583–592
167. Neuhaus, H. E. and Stitt, M. (1990) Planta 182, 445–454
168. Newsholme, E. A. and Crabtree, B. (1976) Biochem. Soc. Symp. 41, 61–109
169. Newsholme, E. A. and Gevers, W. (1967) Vitamins and Hormones 25, 1–87
170. Newsholme, E. A. and Randle, P. J. (1964) Biochem. J. 93, 641–651
171. Newsholme, E. A. and Start, C. (1973) Regulation in Metabolism., Wiley and Sons,
London
172. Newsholme, E. A. and Underwood, A. H. (1966) Biochem. J. 99, 24C–26C
173. Newsholme, P. and Walsh, D. A. (1992) Biochem. J. 283, 845–848
174. Nicholls, D. G. and Crompton, M. (1980) FEBS Lett. 111, 261–268
175. Niederberger, P., Prasad, R., Miozzari, G. and Kacser, H. (1992) Biochem. J. 287,
473–479
176. Norby, J. G., Ottolenghi, P. and Jensen, J. (1980) Analyt. Biochem. 102, 318–320
177. Orr, H. A. (1991) Proc. Natl. Acad. Sci. U.S.A. 88, 11413–11415
178. Ovadi, J. (1991) J. Theor. Biol. 152, 1–22, See also the following discussion by
various authors on pp23–141.
179. Page, R. A., Kitson, K. E. and Hardman, M. J. (1991) Biochem. J. 278, 659–665
180. Pearcy, R. W. (1990) Annu. Rev. Plant Physiol. Plant Mol. Biol. 41, 421–453
181. Pedersen, R. C. and Brownie, A. C. (1986) Biochem. Action Horm. 13, 129–166
182. Pettersson, G. and Ryde-Pettersson, U. (1988) Eur. J. Biochem. 175, 661–672
183. Pietrobon, D., Di Virgilio, F. and Pozzan, T. (1990) Eur. J. Biochem. 193, 599–622
184. Pilkis, S. J. and Claus, T. H. (1991) Annu. Rev. Nutr. 11, 465–515
185. Pilkis, S. J., Claus, T. H. and El-Maghrabi, M. R. (1988) Adv. Second Mess. Phos-
phoprotein Res. 22, 175–191
186. Przybylski, F., Otto, A., Nissler, K., Schellenberger, W. and Hofmann, E. (1985)
Biochim. Biophys. Acta 831, 350–352
187. Quant, P. A. and Makins, R. A. (1994) Biochem. Soc. Trans. 22, 441–446
188. Quant, P. A., Robin, D., Robin, P., Girard, J. and Brand, M. D. (1993) Biochim.
Biophys. Acta 1156, 135–143
189. Quick, W. P., Schurr, U., Scheibe, R., Schulze, E.-D., Rodermel, S. R., Bogorad, L.
and Stitt, M. (1991) Planta 183, 542–554
190. Rabkin, M. and Blum, J. J. (1985) Biochem. J. 255, 761–786
191. Rainey, W. E., Shay, J. W. and Mason, J. I. (1986) J. Biol. Chem. 261, 7322–7326
192. Randle, P. J., England, P. J. and Denton, R. M. (1970) Biochem. J. 117, 677–695
193. Rapoport, T. A., Heinrich, R., Jacobasch, G. and Rapoport, S. (1974) Eur. J.
Biochem. 42, 107–120
194. Rapoport, T. A., Heinrich, R. and Rapoport, S. M. (1976) Biochem. J. 154, 449–469
195. Reder, C. (1988) J. Theor. Biol. 135, 175–201
196. Reich, J. G. and Sel’kov, E. E. (1981) Energy Metabolism of the Cell, Academic
Press, London
197. Rhee, S. G., Chock, P. B. and Stadtman, E. R. (1989) Adv. Enzymol. 62, 37–92
198. Rigoulet, M., Leverve, X. M., Plomp, P. J. A. M. and Meijer, A. J. (1987) Biochem.
J. 245, 661–668
199. Rodbell, M. (1992) Curr. Top. Cell. Regul. 32, 1–47
200. Rognstad, R. (1979) J. Biol. Chem. 254, 1875–1878
201. Rolleston, F. S. (1972) Curr. Topics Cell Regul. 5, 47–75
202. Rudolph, F. B. and Fromm, H. J. (1971) J. Biol. Chem. 246, 6611–6619
203. Ruijter, G. J. G., Postma, P. W. and van Dam, K. (1991) J. Bacteriol. 173, 6184–
6191
204. Sacktor, B. and Wormser-Shavit, E. (1966) J. Biol. Chem. 241, 624–631
205. Saier, M. H., Wu, L.-F. and Reizer, J. (1990) Trends Biochem. Sci. 15, 391–395
290 References

206. Sale, E. M. and Denton, R. M. (1985) Biochem. J. 232, 897–904

207. Salter, M., Knowles, R. G. and Pogson, C. I. (1986) Biochem. J. 234, 635–647
208. Salter, M., Knowles, R. G. and Pogson, C. I. (1994) in Tipton, K. F., ed., Essays in
Biochemistry, vol. 28, pp. 1–12, Portland Press, London
209. Sauro, H. M. (1993) Comput. Applic. Biosci. 9, 441–450
210. Sauro, H. M. (1994) Biosystems 33, 55–67
211. Sauro, H. M. and Fell, D. A. (1991) Mathl. Comput. Modelling 15, 15–28
212. Sauro, H. M. and Kacser, H. (1990) Eur. J. Biochem. 187, 493–500
213. Sauro, H. M., Small, J. R. and Fell, D. A. (1987) Eur. J. Biochem. 165, 215–221
214. Savageau, M. A. (1971) Arch. Biochem. Biophys. 145, 612–621
215. Savageau, M. A. (1974) J. Mol. Evol. 4, 139–156
216. Savageau, M. A. (1975) J. Mol. Evol. 5, 199–222
217. Savageau, M. A. (1976) Biochemical Systems Analysis: a Study of Function and
Design in Molecular Biology, Addison–Wesley, Reading, Mass.
218. Savageau, M. A., Voit, E. O. and Irvine, D. H. (1987) Math. Biosci. 86, 127–145
219. Savageau, M. A., Voit, E. O. and Irvine, D. H. (1987) Math. Biosci. 86, 147–169
220. Schaaff, I., Heinisch, J. and Zimmerman, F. K. (1989) Yeast 5, 285–290
221. Schimke, R. T. (1962) J. Biol. Chem. 237, 459–468
222. Schuster, R., Holzhütter, H.-G. and Jacobasch, G. (1988) Biosystems 22, 19–36
223. Schuster, S. and Heinrich, R. (1992) BioSystems 27, 1–15
224. Schuster, S., Kahn, D. and Westerhoff, H. V. (1993) Biophys. Chem. 48, 1–17
225. Serrano, R., Gancedo, J. M. and Gancedo, C. (1973) Eur, J, Biochem. 34, 479–482
226. Shacter, E., Chock, P. B., Rhee, S. G. and Stadtman, E. R. (1986) in Boyer, P. D.
and Krebs, E. G., eds., The Enzymes., vol. XVII, pp. 21–42, Academic Press, New
York, 3rd edn.
227. Shalwitz, R. A., Beth, T. J., MacLeod, A. M. K., Tucker, S. J. and Rolison, G. G.
(1994) Amer. J. Physiol. 266, E433–E437
228. Small, J. R. (1988) Theoretical aspects of metabolic control, Ph.D. thesis, Oxford
Polytechnic
229. Small, J. R. (1993) Biochem. J. 296, 423–433
230. Small, J. R. (1994) Microbiology 140, 2439–2449
231. Small, J. R. and Fell, D. A. (1990) Eur. J. Biochem. 191, 405–411
232. Small, J. R. and Fell, D. A. (1990) Eur. J. Biochem. 191, 413–420
233. Small, J. R. and Kacser, H. (1993) Eur. J. Biochem. 213, 613–624
234. Smith, R. E. and Horwitz, B. A. (1969) Physiol. Rev. 49, 330–425
235. Snell, K. and Fell, D. A. (1990) Adv. Enzyme Regul. 30, 13–32
236. Srere, P. A. (1993) Biol. Chem. Hoppe–Seyler 374, 833–842
237. Stadtman, E. R. (1970) in Boyer, P. D., ed., The Enzymes, vol. 1, pp. 397–459,
Academic Press, New York, 3rd edn.
238. Stadtman, E. R. and Chock, P. B. (1978) Curr. Top. Cell. Regul. 13, 53–95
239. Stanley, J. C., Salter, M., Fisher, M. J. and Pogson, C. I. (1985) Arch. Biochem.
Biophys. 240, 792–800
240. Stein, R. B. and Blum, J. J. (1978) J. Theor. Biol. 72, 487–522
241. Steinberg, D. (1963) Biochem. Soc. Symp. 24, 111–143
242. Stitt, M. (1989) Phil. Trans. R. Soc. Lond. B 323, 327–338
243. Stitt, M., Quick, W. P., Schurr, U., Schulze, E.-D., Rodermel, S. R. and Bogorad,
L. (1991) Planta 183, 555–566
244. Sutherland, E. W. (1972) Science 177, 401–408
245. Thomas, S. and Fell, D. A. (1993) Biochem. J. 292, 351–360
246. Thomas, S. and Fell, D. A. (1994) J. Theor. Biol. 167, 175–200
247. Tornheim, K. (1985) J. Biol. Chem. 260, 7985–7989
248. Torres, N. V., Mateo, F., Melendez-Hevia, E. and Kacser, H. (1986) Biochem. J.
234, 169–174
249. Torres, N. V. and Meléndez-Hevia, E. (1992) Mol. Cell. Biochem. 112, 109–115
250. Umbarger, H. E. (1956) Science 123, 848
251. Umbarger, H. E. (1978) Annu. Rev. Biochem. 47, 533–606
252. Utter, M. F. and Scrutton, M. C. (1969) Curr. Top. Cell. Regul. 1, 253–296
 291

253. Van Schaftingen, E., Hue, L. and Hers, H. G. (1980) Biochem. J. 192, 887–895
254. Van Schaftingen, E., Hue, L. and Hers, H. G. (1980) Biochem. J. 192, 897–901
255. Van Schaftingen, E., Jett, M.-F., Hue, L. and Hers, H.-G. (1981) Proc. Natl. Acad.
Sci. U.S.A. 78, 3483–3486
256. VanderCammen, A. and Van Schaftingen, E. (1990) Eur. J. Biochem. 191, 483–489
257. Vandercammen, A. and Van Schaftingen, E. (1993) Biochem. J. 294, 551–556
258. Veech, R. L., Lawson, J. W. R., Cornell, N. W. and Krebs, H. A. (1979) J. Biol.
Chem. 254, 6538–6547
259. Voit, E. O. and Savageau, M. A. (1987) Biochemistry 26, 6869–6880
260. Waley, S. G. (1964) Biochem. J. 91, 514–517
261. Wals, P. A. and Katz, J. (1994) J. Biol. Chem. 269, 18343–18352
262. Walsh, K. and Koshland Jr., D. E. (1985) Proc. Natl. Acad. Sci. U.S.A. 82, 3577–
3581
263. Walsh, K., Schena, M., Flint, A. J. and Koshland Jr., D. E. (1987) Biochem. Soc.
Symp. 54, 183–195
264. Walter, R. P., Morris, J. G. and Kell, D. B. (1987) J. Gen. Microbiol. 133, 259–266
265. Wanders, R. J. A., van Roermund, C. W. T. and Meijer, A. J. (1984) Eur. J.
Biochem. 142, 247–254
266. Wang, J. H., Pallen, C. J., Sharma, R. K., Adachi, A. M. and Adachi, K. (1985)
Curr. Top. Cell. Regul. 27, 419–436
267. Waterman, M. R. and Simpson, E. R. (1989) Recent Prog. Horm. Res. 45, 533–566
268. Weber, G., Singhal, R. L. and Srivastava, S. K. (1965) Adv. Enzyme Regul. 3, 43–75
269. Weber, G., Singhal, R. L., Stamm, N. B., Lea, M. A. and Fisher, E. A. (1966) Adv.
Enzyme Regul. 4, 59–81
270. Weekes, J., Hawley, S. A., Corton, J., Shugar, D. and Hardie, D. G. (19994) Eur. J.
Biochem. 219, 751–757
271. Welch, G. R. (1985) J. Theor. Biol. 114, 433–446
272. Westerhoff, H. V. and Chen, Y.-D. (1984) Eur. J. Biochem. 142, 425–430
273. Westerhoff, H. V., Groen, A. K. and Wanders, R. J. A. (1984) Biosci. Rep. 4, 1–22
274. Westerhoff, H. V. and Kell, D. B. (1987) Biotechnol. Bioeng. 30, 101–107
275. Wilkinson, G. N. (1961) Biochem. J. 80, 324–332
276. Woodrow, I. E. (1986) Biochim. Biophys. Acta 851, 181–192
277. Woodrow, I. E., Murphy, D. J. and Latzko, E. (1984) J. Biol. Chem. 259, 3791–3795
278. Wright, B. E. and Albe, K. R. (1994) J. Theor. Biol. 169, 231–241
279. Wright, B. E., Butler, M. H. and Albe, K. R. (1992) J. Biol. Chem. 267, 3101–3105
280. Wright, B. E. and Reimers, J. M. (1988) J. Biol. Chem. 263, 14906–14912
281. Yates, R. A. and Pardee, A. B. (1956) J. Biol. Chem. 221, 757–770
282. Zhang, Y., Proenca, R., Maffei, M., Barone, M., Leopold, L. and Friedman, J. M.
(1994) Nature (London) 372, 425–432
283. Zuurendonk, P. F. and Tager, J. M. (1974) Biochim. Biophys. Acta 333, 393–399
284. Zuurendonk, P. F., Tischler, M. E., Akerboom, T. P. M., van der Meer, R.,
Williamson, J. R. and Tager, J. M. (1979) Methods Enzymol. 56, 207–233
292
Index

acetazolamide, 168 over–expression, 94

acetyl–CoA, 91, 241 allosteric effector, 9, 72, 72, 78
acetyl–CoA carboxylase, 267, 268 activator, 76
phosphorylation, 236, 240, 266 control of flux, 124, 125, 229, 267,
acetylornithine aminotransferase, 141, 142 275
activator, see allosteric effector elasticity, 125, 126
Adair, G. S., 70 heterotropic, 76
addition of intermediates, 95 homotropic, 76
adenine nucleotide translocator, 161, 162 inhibitor, 75, 80
flux control coefficient, 163–165, 184, proof of physiological role, 98
188 response coefficient, 125
rate–limiting?, 103, 161 unknown, 38
adenylate cyclase, 234–235 allosteric enzyme, 4, 9, 71–82
calmodulin and, 237 characteristics, 71–72
discovery, 231 concerted model, 73–76, 79, 81, 82
adenylate kinase equilibrium, 238–240 equation, 75, 76
adenylation, see covalent modification, nu- conformational change, 73, 78, 81
cleotidylation cooperativity, see cooperativity
adenylyltransferase, see glutamine syn- elasticity, 119–120
thase examples, 79–82
adipocytes, 28 feedback inhibition, 205, 209
adipose tissue, 218, 220, 227, 235, 236, flux control, 261
241, 266, 268 general model, 78
brown, 223, 224 K system, 72, 126
ADP kinetic model, 73, 79
assay, 32 kinetics, 71–72, 81
bound, 36, 37, 240 regulatory, 98
from creatine kinase equilibrium, 37 sequential model, 76–78, 82
mitochondrial uptake, 162 equation, 77
ADP–ribosylation, see covalent modifi- V system, 72
cation allozyme, 142
adrenaline, 231, 235, 236, 243, 252 amidophosphoribosyl transferase
adrenergic response, 223, 235, 269 feedback inhibition, 217
adrenocorticotropic hormone, 235, 270– amino acid catabolism, see phenylalanine,
271 tryptophan, tyrosine and urea
adrenodoxin, 270, 271 amino acid synthesis, 99, 141, 205, 212,
alanine transaminase, 268 266, 272, see also individual
Albe, K. R., 94 compounds
alcohol dehydrogenase, 62, 143 amino–oxyacetate, 157, 168
allozymes, 143 AMP, 127, 209
flux control coefficient, 143–144, 168 adenylate kinase equilibrium, 240
rate–limiting?, 103 as signal metabolite, 233, 238–241
aldolase, 56, 90, 269 synthesis, 216, 217
activity level, 93 AMP–dependent protein kinase, 233, 240–
flux control coefficient, 175 242
low activity, 93–94 amplification, see covalent modification

293
294 Index

animal experiments binding site, 65–71

inbred strains, 24 interacting, 68–71, see cooperativ-
need for, 24, 27, 28 ity
non–invasive, 24 multiple, identical, 66–68
tissues and organs, 25–27 non–identical, 68
whole organism, 24–25 Biochemical Systems Theory, 104–105, 276
anthranilate synthase, 213, 264 feedback inhibition, 208–209, 211
antibodies, see immunoassay biotechnology, 2, 111, 263, 281
antisense RNA, 150, 151 2,3–bisphosphoglycerate, 80–81, 170, 179
aquorein, 36 Blackman, F. F., 18
Arabidopsis thaliana, 145 blood, 24
arginine synthesis, 141–142 cell, red, 27, 42, 80, 155, 170
argininosuccinate lyase, 141, 142 clotting, 230, 248
flux control coefficient, 142 glucose, 3, 227–229
argininosuccinate synthase oxygen binding, 68–70, 81
flux control coefficient, 142 Blum, J. J., 41, 220, 222, 225, 227, 228
aromatic amino acids Bohr effect, 81
catabolism, 195, 242 branch point theorem, 124, 171, 175, 186,
flux control coefficient, 171 188, 189
synthesis, 213, 215 branched–chain α–keto acid dehydroge-
aspartate kinase, 215 nase complex, 241
bifunctional enzyme, 214 branched–chain ketoacid dehydrogenase
enzyme multiplicity, 214, 215 kinase, 233, 241
aspartate shuttle, 172, 177 Brand, M. D., 181–189, 198
aspartate transaminase, 157 Briggs & Haldane, 48–50, 63
aspartate transcarbamylase, 62, 71, 72, Brindle, K. M., 34, 44, 147
81–82, 205 bromo–octanoate, 168
association constant, 66 Brown, G. C., 181–189
Atkinson, D. E., 13, 16, 22, 275 Buchanan, B. B., 273
ATP, 16 Bücher, T., 19, 262
assay, 33–35 Buchner, H. & E., 29
homoeostasis, 224, 240 buffering concentrations, see homoeosta-
Mg2+ complex, 57 sis, metabolites
turnover measurement, 44 Bulfield, G., 266
ATP citrate lyase, 267, 268 Burns, J. A., 104, 107, 109, 111, 117, 121,
ATP consumption 125, 128–130, 132, 142, 176,
flux control coefficient, 164, 195 192, 198, 206, 207, 209, 253,
atp operon, 149 277
ATP synthase, 162
flux control coefficient, 148–149, 165, 13 C,24, 34, 41–43
190 14 C,24, 32, 35, 40, 41, 65, 91, 92, 205,
azide, 163 222
Ca2+ , 36, see calmodulin
Bailey, J. E., 195 mitochondrial uptake, 227–228, 241
Beadle, G. W. & Tatum, E. L., 141 muscle contraction, 224, 243
bee flight, 224 second messenger, 235, 237, 238, 241,
Berry, M. N., 28 243, 245, 251
bidirectional switching, 223, 227, 229 Cahill, G. F., 218, 227
bifunctional enzyme calmodulin, 233, 236–238, 242, 243
aspartate kinase / homoserine de- Calvin, M., 32
hydrogenase, 214 cAMP, see cyclic AMP
glutamine synthase adenylyltrans- cancer, 27, 281
ferase, 247 carbamoyl–phosphate synthase
phosphofructo–2–kinase / fructose– flux control coefficient, 170, 181
2,6–bisphosphatase, 223, 236, carbohydrate metabolism, 6, see named
269 pathways
protein–PII uridylyl transferase, 247 carbonic anhydrase
Index 295

flux control coefficient, 168 226–228

carboxyarabinitol 1–phosphate, 273 concentration control coefficient, 132–134,
carboxyatractyloside, 163, 164 202–204, 208, 209, 228, 255
Cárdenas, M. L., 254, 255 algebraic solution, 203
Carling, D. J., 238 concerted model, see allosteric enzyme
carnitine palmitoyltransferase I conformational change, see allosteric en-
flux control coefficient, 190 zyme and enzyme kinetics, in-
cascade, see covalent modification duced fit
Cascante, M., 277 connectivity theorem, 121, 121–124, 127,
cell 128, 170, 171, 277
culture, 27 application, 175, 180, 183, 186, 188,
fractionation, 30 258
rapid, 36, 173 concentration control coefficients, 133
homogenate, 29, 37, 154 control, 3, 1–3
isolated, 27–28 distributed, 112, 139, 188, 262–274
organelles, 29–30, see compartmen- mechanisms, 8–9, 124–128, 201–256
tation theories of, 18–20
permeabilized, 29 traditional approach, 87–100, 128–
cell–free system, 29 130, 206–207, 261–262, 274, 282
Chance, B., 95, 168, 209 control matrix, 133
Chance, E. M., 42, 43 cooperativity, 68, 70–71, 73–82, 119–120,
Changeux, J. P., 71, see allosteric en- 127, 208–209, 251
zyme, concerted model negative, 71, 77, 82
channelling, 280 co–response coefficient, 180
chemiosmotic hypothesis, 162 Cori, C. F., 252, 262
Cheung, W. Y., 237 Cori, C. F. & G. T., 230
Chlamydomonas reinhardtii, 109 Cornish–Bowden, A., 2, 3, 54, 60, 180,
Chlorella, 32 202, 209, 228, 254, 255
chloroplast, 30, 271–273 corticosteroid synthesis, 270–271
Chock, P. B., 247–253 covalent modification, 9, 229–256
cholesterol, see corticosteroid synthesis ADP–ribosylation, 232
synthesis, 95, 240 amplification
cholesterol esterase, 271 catalytic, 248–249
chorismate, 213, 264 rate, 251–252
chromatography, 33, 42 signal, 249–251
citrate, 90, 91 amplitude, 250
citrate cleavage enzyme, see ATP citrate carbamylation, 273
lyase cascade, 230, 231, 243–245, 278
citrate synthase, 147 multicyclic, 251–252
citrulline synthesis, 171, 181 control of flux, 223, 266, 269, 270
Clarkia xantiana, 145, 181 control potential, 98–99, 109
Cohen, G. N., 214 cyclic, see reversible
Cohen, P., 236, 242 discovery, 230–231
Cohen, S. M., 42 enzyme assay and, 39
committed step, 88 irreversible, 230
compartmentation, 29, 35–36, 44, 120, kinases, 233–241
169, 191, 279 Metabolic Control Analysis and, 254–
Complexes I–IV, see electron transport 256, 278–279
chain methylation, 232
compulsory–order, see enzyme kinetics, nucleotidylation, 232, 245–247
mechanism discovery, 246
computer phosphatases, see phosphoprotein phos-
fitting, 41, 53–54, 60, 61, 67, 68, 83, phatase
91, 92, 167 phosphorylation, 231–245
program, 83, 193, 197 photosynthesis, see photosynthesis,
simulation, 6, 41, 42, 90, 91, 94, 96, light activation
106, 168–170, 192, 197, 209, reversible, 230, 230–256
296 Index

sensitivity, 250–251, 253, 254 digitonin, 36, 173

signal integration, 251 dinoflagellates, see okadaic acid
thiol oxidation, 273 direct linear plot, 54
time scale, 9, 252, 261 disequilibrium ratio, 14, 14–17, 37, 128,
ultrasensitivity, 252–256 193, 195
zero–order, 253, 255 elasticity and, 118
Crabtree, B., 87, 125, 220, 224, 226–227, flux control coefficients and, 128–
275, 276 129
creatine kinase, 37 measurement, 88–90
cross–over theorem, 95–98, 210 dissociation constant, 48, 49, 62, 64, 66,
crossroads 68, 75–77, 250
metabolic, 227, 229 dominance, 108–109, 140, 281
CTP dopamine, 237
synthesis, 205, 216 double modulation method, see elastic-
cyanide, 166 ity, measurement
30 ,50 –cyclic AMP, 234–236 double–displacement, see enzyme kinet-
Ca2+ and, 237, 243, 251 ics, mechanism
discovery, 231 Drosophila melanogaster
free, 36, 37 alcohol dehydrogenase, 143
gradient, 280 mutations, 108
immunoassay, 33 drugs, 2, 24, 94, 167, 281
second messenger, 231, 243–245, 269– Dykhuizen, D. E., 144–145, 152, 154
271 dynamic channel, 280
cyclic AMP–dependent protein kinase, see dynamic equilibrium, see equilibrium, dy-
protein kinase A namic
30 ,50 –cyclic–nucleotide phosphodiesterase, dynamic range, 8, 9, 9
234, 235, 244
Ca2+ –activation, 237 Eadie–Hofstee plot, 67
cycloserine, 157 Easterby, J. S., 15, 19, 83, 208
cytochrome–bc1 complex, 161, 163, 166 effector, see allosteric effector
cytochrome P450 , 270–271 Eisenthal, R., 54
cytochrome–c oxidase, 161–163 elasticity, 56, 114, 114–121, 134–136
flux control coefficient, 166 accuracy, 192–193
rate–limiting?, 164 algebraic derivation, 135–136
cytoplasmic concentration, 35–36 allosteric effector, 126, 127
cytoplasmic redox state, 173 block or group, 174, 182, 204, 209
cytoskeleton, 29 calculation, 116, 191–195
concentration control coefficients and,
dark reactions, see photosynthesis 203–205
Davies, S. E. C., 147 connectivity and, see connectivity
Denton, R. M., 19 theorem
deoxyhaemoglobin, see haemoglobin enzyme kinetics and, 116–121, 134–
dephosphorylation, protein, see phospho- 135
protein phosphatase flux control coefficients and, 134, 170–
deuterium, see isotope, stable 196, 226–227, 258–259
DHAP synthase, 213, 215 inhibitor, 158, 208, 209, 211–212
diacylglycerol, 233 kinetic parameter, 126
second messenger, 238 measurement
diacylglycerol acyltransferase double modulation, 176–179
flux control coefficient, 168 generalized modulation, 180
diaphragm muscle, 26 single modulation, 179–180
diarrhetic shellfish poisoning, see okadaic top–down method, see flux con-
acid trol coefficient, measurement
dicarboxylate carrier, 161 product, 204, 209
flux control coefficient, 163, 165, 166 importance, 134, 192
Dictyostelium discoideum, 41, 92, 94 irreversible reaction, 117
diffusion, rate of, 279 reversible reaction, 119
Index 297

response coefficient and, 125, 171 substrate binding, 15, 58, 62–64, see
substrate also binding sites
irreversible reaction, 116–117, 135 substrate complex, 48
reversible reaction, 117 subunit structure, 66, 71, 73, 76,
electron transport chain, 31, 95, 162 77, 81
intermediate, 272 titration, see flux control coefficient
electrophoresis enzyme kinetics, 47–85
cell proteins, 232 induced fit, 58, 76
isoenzymes, 143 inhibition, see inhibitor
endoplasmic reticulum, 30, 90, 223, 228, initial rate, 8, 48, 55
270, 271 limitations of, 5–6, 169, 191
energy maximal activity, 129
dissipation, 88, 218, 227 mechanism, 57–59
metabolism, 189–190, 204, 240 compulsory–order, 58, 62–63
transduction, 149 double–displacement, 58–59, 64–
enolase, 93, 177 65
disequilibrium, 89, 90 random–order, 58, 63–64
elasticity, 122, 176, 178, 179 Michaelis Menten, 47–56
computer fitting, 53–54, 83
flux control coefficient, 122, 175, 176,
178–180 equation, 48
graphical analysis, 50–54
low activity, 93
reversible, 55
enzyme
multi–enzyme sequence, 103, 106
activity, 5, 24, 92–94
optimal design, 54, 60
changing, 2, 8, 9, 18, 94, 99, 124,
partial reactions, 65
269
permeabilized cells, 29
large changes, see flux, large changes
product inhibition, 17, 39, 55–56,
measurement, 37–39
64
metabolic flux and, see flux con-
importance, 169, 192
trol coefficient
rate equation, 134, 135
allosteric, see allosteric enzymes
two–substrate, 56–65, 97
assay, 32–33
computer fitting, 61, 83
coupled, 32, 33 experiment, 60
association, 132, 169, 278, 280 graphical analysis, 60–61
catalysing several steps, 278 rate equation, 59–60
covalent modification, see covalent enzyme multiplicity, 215
modification feedback inhibition, 214–215
degradation, 8, 230 enzyme IIGlc
elasticity, see elasticity flux control coefficient, 148
expression epididymal fat pad, 26
controllable, 147–149, 151, 152 epinephrin, see adrenaline
high in vivo concentration, 280 equilibrium, 13
isolated, 30 chemical, 10, 13
mutations, 140, 141, 170, 281, see constant, 14, 55, 66, 73, 77, 128
dominance in vivo correction, 89–90
near–equilibrium, see equilibrium, displacement from, see disequilib-
near–equilibrium reaction rium ratio
non–equilibrium, see equilibrium, non– dynamic, 10, 13, 228
equilibrium reaction near–equilibrium reaction, 15–17, 37,
phosphorylation, see covalent mod- 48, 64, 88, 90–92, 117, 122,
ification 129, 173, 174, 192, 193, 204,
production, 2 222
regulatory, 19–20, 87–100, 210, 275, non–equilibrium reaction, 15–17, 19,
276 88–92, 98, 129
mutant, 99 erythrocyte, see blood, cell, red
properties, 98, 99, 261 Escherichia coli, 8, 94, 99, 144, 147–149,
reversibility, 55 152, 205, 206, 214–216, 245,
298 Index

246 generalized, 126, 131–132

ethanol effect of covalent modification, 255
catabolism, 103, 143–144, 168 elasticities and, see connectivity the-
synthesis in yeast, 29, 39, 147 orem
eukaryote, 29, 35, 98, 230, 231, 234, 237, finite change equation, 146, 147, 262,
241, 242 263, 282
external metabolite, 125, 127, 171, 229 interpretation, 107–114, 125, 229
large values, 108, 113, 281, 282
fat cells, see adipocytes limiting rates and, 129–130
fatty acid measurement, 139–196
catabolism, 190 allozymes, 142–146
synthase, 266, 268 computer models, 168–170
synthesis, 240, 242, 266 enzyme titration, 154–155
fatty acyl–CoA, 241 from elasticities, 170–180, 191–
FCCP, 163 195
feedback inhibition, 9, 71, 97, 98, 195, gene dosage, 141–147
205–217, 263 gene expression, 147–154
abolishing, 264 inhibitor titration, 155–168
branched pathways, 210–217 mixed methods, 195–196
concerted, 215 top–down analysis, 181–190
cumulative, 215–217 predictive value, 109–111, 124, 126,
discovery, 71, 81, 205–206 202, 275, 276
Metabolic Control Analysis, 180, 192, response coefficient and, 125–126,
206–210 171
metabolite concentrations and, 208– small values, 122, 210
210 expected, 108–109, 112
nested, 211–217 summation theorem, see summation
oscillation, 208, 209 theorem
sequential, 211–213 Fraenkel, D. G., 94
stability, 208, 210 free energy, 13–15, 77
synergistic, 217 notation, 21
transient time and, 208, 210 standard, 14–16
Fell, D. A., 124, 173, 186, 192, 194, 197, freeze clamping, 31, 35, 39
204, 226, 229, 255, 258, 263, fructose 1,6–bisphosphate, 15, 38, 40, 56,
265, 278, 280 96, 97, 155, 223
fibrin, 230 fructose 2,6–bisphosphate, 97, 223, 225,
Filmer, D., see allosteric enzyme, sequen- 236, 269
tial model discovery, 38
Fischer, E., 9 third messenger, 223, 236
Fisher, R. A., 108 fructose 6–phosphate, 40, 96, 97, 174,
fitness, 3 222
flux, 4 fructose 6–phosphate:fructose 1,6–bisphosphate
control, 19, 20, 88, 94, 113, 124, cycle, 218, 221, 223–225, 227,
202, 229, 262, 265 236
large changes, see multisite modu- measurement, 222
lation fructose–1,6–bisphosphatase, 94, 172, 176,
achieving, 111, 239, 262, 263, 274, 181, 221, 223–225, 236, 266,
276, 282 268, 269
engineering, see Universal Method light activation, 272, 273
examples, 203, 262, 266–274 fructose–2,6–bisphosphatase, see bifunc-
measurement, 39–44, 90–92, 190 tional enzyme
substrate cycles, 222 fumarate, 95
pathway, 11, 13, 17 functional readiness, 19
flux control coefficient, 105–113, 130–132 futile cycle, see substrate cycle
algebraic solution, 123–124, 174, 180,
186, 202, 258 G–protein, 235, 238
definition, 106–107 galactose
Index 299

bacterial growth on, 9 glucose–6–phosphate isomerase, 89, 90,

β–galactosidase, 144–145 145, 154, 172, 174, 222
evolved, 145 glutamate, 95
flux control coefficient, 145, 152–154 labelling, 92
Galazzo, J. L., 195 NMR spectrum, 43
Γ, see mass action ratio synthesis, 245
Gancedo, J. M. & C., 29 glutamate–ammonia ligase, see glutamine
Garfinkel, D., 6, 168, 170 synthase
Gellerich, F. N., 167 glutaminase, 168
gene expression, 8, 94, 109, 146, 205, 238, glutamine, 247
264, 266, 269, see also flux con- glutamine synthase, 6
trol coefficient, measurement adenylyltransferase, 246–247
and enzyme, expression covalent modification, 245–247, 251
genetic disease, 281 feedback inhibition, 216, 217, 251
genetic engineering, 2, 99, 146, 147, 149, glyceraldehyde–3–phosphate, 174
263, 281 enzyme–bound, 90
genetic obesity, 266 glyceraldehyde–3–phosphate dehydrogenase,
genetics 93, 172
classical, 99, 108–109, 141, 142, 281 flux control coefficient, 168
molecular, see genetic engineering kinetics, 82
Gevers, W., 20, 87 light activation, 272, 273
Giersch, C., 177–180, 277 glycerol
glucagon, 3, 174, 174, 175, 176, 223, 235, fatty acid metabolism, 220
236, 269 glycerol–3–phosphate dehydrogenase, 267,
268
glucocorticoid hormones, 94, 269
glycogen
glucokinase, see hexokinase IV
bound enzymes, 39, 243, 244
gluconeogenesis, 155, 236, 242, 269
metabolism, 237, 238, 242
control analysis, 160, 168, 171, 173–
glycogen phosphorylase, 93, 230, 245
177, 181, 195
a, 243
enzymes of, 94, 172, 269
b, 127, 243
fluxes in, 41, 42, 91
covalent modification, 39, 236, 237,
rate–limiting step of?, 95, 156, 160
243–245, 251, 262
substrate cycles, 42, 174, 221–223, discovery, 9, 230
227 rate, 252
glucose ultrasensitivity, 253
blood, see blood glucose kinase, see phosphorylase kinase
catabolism, 39–41 glycogen synthase
control analysis, 190 covalent modification, 39, 236, 237,
glucose 6–phosphate cycle, 218, 221, 245
222, 227, 228 glycogen synthase kinase, 233
transport glycogen–binding subunit, 243–245
A. radiobacter, 168 glycolysis, 242
C. pasteurianum, 168 computer model, 170
E. coli, 148, 154 control analysis, 122, 154–155, 168,
heart, 90 195
yeast, 195 cross–over, 96–98, 210
glucose 1–phosphate, 236 diet and hormones, 236, 266, 269
glucose 6–phosphate, 96, 230, 245 disequilibrium ratios, 15, 89, 129
glucose–1–phosphate adenylyltransferase, enzyme levels, 93, 280
145 enzymes of, 94, 269
glucose–6–phosphatase, 94, 172, 222, 229, failure to change flux, 99, 104, 147,
268, 269 207
glucose–6–phosphate flux measurement, 91
flux control coefficient, 154–155 in vitro model, 178
glucose–6–phosphate dehydrogenase, 267, large flux changes, 19, 203, 262
268, 271 oscillations, 208
300 Index

substrate cycles, 221, 222, 224 homogenate, see cell homogenate

glyoxalate cycle, 147 homotrotropic effector, see allosteric ef-
Groen, A. K., 118, 119, 157, 158, 160, fector
161, 163–166, 171, 173–177, 183, hormone, 2, 3, 9, 20, 25, 38, 94, 124, 151,
184, 188, 195 229, 231, 238, 248, 251, 263,
guanine–nucleotide–binding protein, see 269, 276
G–protein receptor, 8, 66, 71, 233, 234
HPLC, 33, 40–42
3 H, 24, 35, 41, 185 Hue, L., 38
measuring substrate cycle flux, 222 Hughes–Klotz plot, 67
Haber, J. E., 76 hydroxymethylglutaryl–CoA reductase, 95,
haemoglobin, 68, 69, 72, 79–81 271
Hafner, R. P., 188 phosphorylation, 241, 271
Haldane, see Briggs & Haldane hydroxymethylglutaryl–CoA synthase, 190
relationship, 55 hydroxyquinoline–N–oxide, 163
Hales, C. N., 87 5–hydroxytryptamine, 237
half of the sites reactivity, 82
Hanes–Woolf plot, 52 immunoassay, 33
Hayflick limit, 27 induced fit, see enzyme kinetics
heart muscle, 31 inhibitor
glycolysis, 89, 93, 129, 236 allosteric, see allosteric effector, in-
model, 6 hibitor
perfusion, 26, 42, 43 competitive, 53, 163, 186, 195, 222,
tricarboxylic acid cycle, 91 253
Heinisch, J., 147 irreversible, 159, 163
Heinrich, R., 15, 97, 104, 114, 129, 131, non–competitive, 155, 159, 163, 253
133, 170 product, see enzyme kinetics, prod-
Helmreich, E., 262 uct inhibition
hepatocyte, 28, 36, 40, 41 uncompetitive, 194, 195
amino acid catabolism, 152, 166, 171, inhibitor titration, see flux control coef-
195 ficient, measurement
energy metabolism, 190 Inhibitor–1, 242–245
gluconeogenesis, 41, 95, 156, 171, initial rate, see enzyme kinetics
173–176, 222, 223 inositol triphosphate, 238
Hers, H. G., 38 insulin, 3, 90, 269
heterokaryon, 141 receptor, 71, 233
heterotropic effector, see allosteric effec- intracellular volume, see water, intracel-
tor lular
heterozygote, 108, 140–146, 264 inulin, 35
hexokinase, 29, 41, 64, 89, 90, 93, 267 iodoacetate, 168
IV, 73, 221, 222, 229, 268, 269 IPTG, 147, 149, 152
flux control coefficient, 155 isobutyramide, 168
over–expression, 147 isocitrate, 90
Hierarchical Control Analysis, see Mod- isocitrate dehydrogenase, 271
ular Control Analysis isoenzyme, 222, 235, 237, see also allozyme,
Higgins, J., 104, 109 electrophoresis and enzyme mul-
Hill tiplicity
coefficient, 69, 70–72, 75, 119, 120, isoleucine
235, 250, 253–255 synthesis, 205, 214
equation, 69, 69 isopropylthiogalactoside, see IPTG
plot, 69–71, 75 isotope, 24
histidine synthesis, 206, 216 exchange, 65
Hofmeyr, J.–H. S., 2, 3, 180, 202, 209, in flux measurement, 41, 90, 91, 222
228 radioactive, 32, 41
homoeostasis, 2, 3, 19, 239 stable, 34, 42
metabolites, 19, 203–205, 209–212,
228–229, 262, 273 Jacob, F., 8, 71
Index 301

Jensen, P. R., 149 forward and reverse, 55

isoenzymes, 143, 145
Km , 48 measurement, 50–55, 60–61
Km , 38, 55–57, 59–60, 104 Lineweaver–Burk plot, 51, 53, 60, 67
and dissociation constant, 48–50, 62 lipogenesis, see fatty acid synthesis
elasticity and, 118, 121, 135, 191, liver, 222
192 cells, see hepatocytes
flux control coefficients and, 129, 143– homogenate, 154
145 mitochondria, 30, 163
measurement, 50–55, 60–61 control analysis, 163
K system, see allosteric enzyme perfusion, 26
Kacser, H., 104, 107, 109–111, 117, 121, slices, 27
125, 128–130, 132, 141–143, 146, luciferase, 33
176, 192, 198, 206, 207, 209, luteinising hormone, 235
253, 262, 263, 276, 277, 282 lysine synthesis, 214, 215
Kahn, D., 278
magnetization transfer, see NMR, flux
Kakiuchi, S., 237
measurement
Katz, J., 40
malate, 63, 91
Kell, D. B., 159, 160
shuttle, 172, 177
ketogenesis
malate dehydrogenase, 174
control analysis, 190
‘malic’ enzyme, 267, 268, 271
kidney cortex, 27, 218, 222
mass action ratio, 14, 88
King–Altman method, 59
mass spectrometer, 42
Kohn, M. C., 170
maximal activity, see enzyme kinetics, max-
Koshland, D. E., 58, 76, 147, 253–255
imal activity
Koshland, Nemethy & Filmer model, see
maximal velocity, see limiting rate
allosteric enzyme, sequential model
Mazat, J.–P., 159, 165, 281
Krebs, E. G., 9, 231
Meijer, A. J., 157
Krebs, H. A., 17, 29, 88, 95, 98, 281
3–mercaptopicolinate, 156, 160, 168, 177
Kuchel, P., 42
Metabolic Control Analysis, 4, 15, 18, 20
kynureninase, 152 applications, 280–282
flux control coefficient, 168 comparison with
kynurenine, 152 Biochemical Systems Theory, 104–
105, 276, 277
lac rate–limiting step, 105–106, 128–
operon, 8, 145, 147, 152 130, 161, 163, 204, 206–207,
promoter, 149 210, 261–262, 274
repressor, 147 criticisms of, 39, 191, 275–277
lactate, 40 experiments, 90, 139–196
lactate dehydrogenase, 32, 93 extensions, 277–280, see Modular
near–equilibrium, 173, 174 Control Analysis
lactose permease, 145, 149 channelling, 280
flux control coefficient, 152–154 enzyme association, 278
large change theory, see flux control ceof- genetic implications, see dominance
ficient, finite change equation and genetics, classical
least–squares fitting, 53–54 origins of, 104, 133
Letellier, T., 159, 165, 167 theory, 97, 103–136, 201–212, 226–
ligand, 66 227, 254–256, 258–259, 263–
binding, 66–71, 75, 79, 250 264
light reactions, see photosynthesis metabolic engineering, 111, 263–265, 281
limiting rate, 38, 48, 50, 53, 59, 71, 248, metabolism, 1
255 metabolite
in vivo, 92–94 bound or free, 36–37, 280
apparent, 60 concentration gradient, 280
elasticity and, 116 concentration in vivo, see compart-
flux control coefficients and, see flux mentation and metabolite, bound
control coefficient or free
302 Index

low concentration, 32, see osmotic NAD+

limit ADP–ribosyl donor, 232
measurement, 30–37 NAD+ /NADH, see moiety–conserved cy-
methionine synthesis, 214 cle
methyltriphenylphosphonium ion, see pro- NADH
tonmotive force, measurement top–down analysis of metabolism,
mevalonate, 95 190
Mg2+ , free near–equilibrium, see equilibrium and dis-
citrate/isocitrate ratio, 90 equilibrium ratio
Michaelis constant, see Km Némethy, G., see allosteric enzyme, se-
Michaelis Menten kinetics, see enzyme quential model
kinetics Neurospora crassa, 141
Middleton, R., 143 Newsholme, E. A., 19, 20, 87, 93, 125,
mitochondria, 30 218, 220, 224–227, 240, 275,
Ca2+ cycling, see Ca2+ 276
cholesterol uptake, 271 Niederberger, P., 264
cytopathies, 159, 281 nitrate reductase, 272, 273
isolated, 30 NMR, 24, 33–36, 42, 43, 90, 195
metabolite levels, 35–36, 173, 174, magnetization transfer, 42–44
190 non–equilibrium, see equilibrium and dis-
metabolite transport, 161 equilibrium ratio
models norvaline, 168
computer, see computer simulation nucleotidylation, see covalent modifica-
mathematical, 5–7, 104, 276 tion
Michaelis Menten equation as, 47,
49 obese mutation, see mouse, obese
verbal, 6 okadaic acid, 242
Modular Control Analysis, 278 oligomycin, 165, 186
modulation methods, see elasticity, mea- one gene, one enzyme, see Beadle & Tatum
surement operon, 266, see atp, lac
moiety–conserved cycle, 122, 128, 133, organ
217, 219, 220, 228, 280 isolated, 25–27
molecular biology, see genetic engineer- perfusion, 26
ing ornithine carbamoyltransferase, 141, 142,
monoacylglycerol lipase, 268 181
Monod, J., 8, 71 flux control coefficient, 141, 168, 171
Monod, Wyman & Changeux model, see oscillation, see feedback, glycolysis
allosteric enzyme, concerted model osmotic limit on metabolites, 11, 13, 15,
mouse 88, 203
lean and fat lines, 266 oxaloacetate, 91, 174
obese, 266, 268 transport, see aspartate, malate shut-
mulitprotein kinase tles
calmodulin–dependent, 233, 237 turnover, 31
multicyclic cascades, see covalent modi- oxidative phosphorylation, 95, 161–163
fication control analysis, 149
multisite modulation, 265–274 inhibitor titration, 161–166
multisite response, see response coefficient, top–down analysis, 183–189
multisite hepatocytes
muscle, see diaphragm,heart and sarto- control analysis, 189–190
rius rate–limiting step?, 103, 161
glycolysis, 19, 203, 208, 236, 237, 2–oxoglutarate, 91, 95, 245, 247
243, 262, 270
mutants, 80, 99, 108, 264, see enzyme P/O ratio, 162
mutations; mitochondria, cy- PII , 247
topathies; mouse, obese 31 P, 34, 35, 90

constitutive, 145 32 P, 41

myoglobin, 79 pacemaker, 95, 281, see rate–limiting step

Index 303

parameter, 50, 59, 75, 124, 124, 125, 126, phosphohydroxypyruvate, 193
131 phosphoprotein phosphatase, 233, 241–
Pardee, A. B., 205 244, 254
partial reactions, see enzyme kinetics 1, 242, 243, 245
Passonneau, J. V., 90, 93 2A, 242
Pasteur, L., 97 2B, 242, 243, 245
PEPCK, 94, 160, 172, 269 calmodulin, 237, 243
flux control coefficient, 160, 168, 175– 2C, 242
177 phosphoribulokinase, 273
light activation, 272, 273 phosphorylase, see glycogen phosphory-
rate–limiting?, 156, 173 lase
perchloric acid, 31, 36 phosphorylase kinase, 231–233, 237, 243–
perfusion, see heart, organ 245, 254
permeabilized cells, see cell, permeabi- calmodulin, 237
lized phosphorylation, protein, see covalent mod-
Perutz, M., 80 ification
phenotype, 108, 144, 266 phosphoserine
phenylalanine, 168 aminotransferase, 193
catabolism phosphatase, 193, 195
control analysis, 171 flux control coefficient, 194
synthesis, 213 phosphotransferase system, see phospho-
transport enolpyruvate
flux control coefficient, 196 photorespiration, 149
phenylalanine–4–monooxygenase photosynthesis, 266, 271–274
phosphorylation, 236 control analysis, 145, 149–151, 181
phenylsuccinate, 163 computer models, 170
phosphodiesterase, see cyclic–nucleotide dark reactions, 272
phosphodiesterase large flux changes, 271
phosphoenolpyruvate light activation, 273–274
assay, 32 light reactions, 272
phosphotransferase system, 148 oscillations, 208
phosphoenolpyruvate carboxykinase, see photosystem II, 272
PEPCK phosphorylation, 272
phosphoenolpyruvate:pyruvate cycle, 42, ping–pong, see enzyme kinetics, mecha-
174, 221, 223 nism, double–displacement
phosphofructo–2–kinase, see bifunctional plasmid, 94, 146–149, 264, 265
enzyme pleiotropic effects, 146
phosphofructokinase, 38, 93, 221, 268, 269 Pogson, C. I., 19, 281
activation, see fructose 2,6–bisphosphate polymorphism, 144
AMP activation, 127, 208, 223, 225 Postma, P. W., 148
ATP inhibition, 208, 236 power law, 109
cross–over, 97, 210 pregnenolone, 270, 271
disequilibrium, 89, 90 Prigogine, I., 13
flux control coefficient, 154, 155, 195 product inhibition, see enzyme kinetics
non–inhibited, 206 promoter, see lac, tac
over–expression, 104, 147 protein kinase A, 36, 221, 223, 233–236,
phosphorylation, 98, 236, 266, 270 241–243, 245
product activation, apparent, 38, 223 cooperativity, 235
rate–limiting?, 99, 147 specificity, 235
phosphoglucomutase, 89, 90, 93, 145 Types I & II, 235
phosphoglucoseisomerase, see glucose–6– protein kinase C, 238
phosphate isomerase protein kinase, redox–dependent, 272
phosphoglycerate dehydrogenase, 193 protein phosphatase, see phosphoprotein
phosphoglycerate kinase, 93, 172, 176 phosphatase
phosphoglycerate mutase, 80, 89, 90, 172, proteolysis, 230, see enzyme degradation
177, 179 prothrombin, 230
flux control coefficient, 178–180 proton leak, 161, 162, 182, 183, 185
304 Index

elasticity, 186 regulatory enzyme, see enzyme, regula-

flux control coefficient, 163, 165, 166, tory
184, 187–190 regulon, 266
flux measurement, 186 Reich, J. G., 220
protonmotive force, 162, 183, 190, 228 respiratory chain, see electron transport
measurement, 185 chain
purine synthesis, 217 response coefficient, 124–128, 158, 171,
pyrimidine synthesis, see CTP 194, 195, 229, 254–256, 275
pyruvate kinetic parameter, 126
assay, 32 multisite, 126–127
pyruvate carboxylase, 65, 94, 172 response time, see transition time
elasticity, 181 ρ, see disequilibrium ratio
flux control coefficient, 175 ribulose–bisphosphate carboxylase, see ru-
rate–limiting?, 95, 173 bisco
pyruvate dehydrogenase complex, 241 Rodbell, M., 234
phosphorylation, 270 Rognstad, R., 155–157, 160, 173
pyruvate dehydrogenase kinase, 233, 241 Rolleston, F. S., 87, 88, 97
pyruvate kinase, 42, 93, 172, 177, 179, rubisco, 280
267–269 activase, 273
disequilibrium, 89, 90 antisense RNA, 149
elasticity, 122, 180, 181 flux control coefficient, 149–151, 170
flux control coefficient, 122, 175, 176, light activation, 272
178–180 Ruijter, G. J. G., 148
phosphorylation, 223, 236, 266, 269 Rüssmann, W., 19, 262
pyruvate phosphate dikinase, 273
pyruvate transport, 172 S0.5 , 71, 72, 126, 250
flux control coefficient, 175 S–adenosylmethionine, 232
Saccharomyces cerevisiae, see yeast
Quant, P. A., 181–183, 190 sartorius muscle, 252
quaternary structure, 78 Sauro, H. M., 119, 124, 186, 226, 258,
quenching metabolism, 31–32 280
Savageau, M. A., 6, 104, 125, 207–208,
R state, 73 210–214, 217, 276, 277
Rabin, B. R., 79 Scatchard plot, 67, 68, 70
radioactivity, see isotope, radioactive second messenger, 231, 234, see also Ca2+ ,
Randle, P. J., 91 cyclic AMP
random–order, see enzyme kinetics, mech- sedoheptulose bisphosphatase
anism light activation, 272, 273
rapid equilibrium assumption, 48, 63, 64, Sel’kov, E. E., 220, 227
73 semi–dominant, 108
Rapoport, T. A., 15, 97, 104, 114, 129, sensitivity
133 in covalent modification, see cova-
rate–limiting step, 4, 18, 17–20, 48, 87, lent modification
88, 90, 95, 99, 105, 149, 256 logarithmic, 125
flux control coefficient and, 106, 112 net, 125, 275
inhibitor titration, 155–157 term for flux control coefficient, 253
objections, 103–104, 161, 261, 266, sensitivity analysis, 104, 104, 170, 226,
see Metabolic Control Analy- 276
sis sequential model, see allosteric enzyme
rectangular hyperbola, 50, 56, 57, 64, 66, serine synthesis
68–70, 110, 121, 145, 146, 152, control analysis, 193–195, 229
153, 155, 250, 253 serotonin, see 5–hydroxytryptamine
red blood cell, see blood, cell, red Serrano, R., 29
Reder, C., 124, 277 Shapiro, B. M., 247
reductionism, 1, 5–6 sigmoid curve, 68, 69–73, 156, 229
regulation, 1–3, 13, 201–210, 261, 262, signal, 3, 9, 19, 20, 96, 101, 203, 229, 231,
276 263
Index 305

amplification, see covalent modifi- proof, 132

cation, amplification supply and demand, 201–205, 209, 262
integration, see covalent modifica- flux control, 202, 203
tion, signal integration homoeostasis, 203, 228
transduction, 234–235, 238, 274 Sutherland, E. W., 230–231, 234
simulation, see computer simulation switch, metabolic, see bidirectional switch-
sink, 10 ing
slime mould, see Dictyostelium discoideum systems, 4
Small, J. R., 110, 146, 167, 177, 178, 192, closed, 13
193, 255, 263, 278 open, 13
Snell, K., 193, 194, 229 theory, 262
source, 10, 11 qualitative, 6–7
specific radioactivity, 41, 91 quantitative, 7–8, 87, 104, 126
spectrophotometry
dual beam, 31 T state, 73
fluorescence, 31 tac promoter, 147–149
NMR, see NMR Tager, J. M., 36, 157
Srere, P. A., 266 teleological argument, 87–88, 262
Stadtman, E. R., 216, 246–253 ternary complex, 58
starch synthesis, 146, 272, 273 tertiary structure, 77, 78
State 3, 96, 163, 166 Tetrahymena pyriformis, 41
State 4, 96, 162 thale cress, see Arabidopsis thaliana
steady state, 10–11, 13, 41, 88, 90, 103, thermodynamics, 21
106, 124, 128, 163, 174, 185, classical, 13
277 irreversible, see thermodynamics, non–
assumption, 48, 49, 50, 63, 65 equilibrium
breakdown, 212 metabolism, 13–17
quasi, 11, 173 non–equilibrium, 13, 21
stable, 10, 228 thioredoxin, 273
unstable, 208 third messenger, see fructose 2,6 bisphos-
Stein, R. B., 220, 225, 227, 228 phate
Steinberg, R. D., 218 Thomas, S., 192, 193, 204, 263
Stitt, M., 145, 149, 181, 198, 274 threonine
subcellular fractionation, see cell fraction- deaminase, 71, 205
ation synthesis, 214, 215
substrate cycle, 217–229 thyroid–stimulating hormone, 235
as metabolic switch, see bidirectional time scale, 8–9, 11–12, 50, 261
switching tissue culture, see cell, culture
definition, 217–220 tissue slices, 27, 95
evidence for, 220–221 tobacco
flux measurement, see flux, mea- transgenic, 150, 151
surement top–down approach, see flux control co-
metabolite homoeostasis, 227–229 efficient, measurement
open, 220, 226, 227 Torres, N. V., 155
sensitivity of control, 225–227 transcription, 278
summation theorem, 258 transition state, 50
thermogenesis, 224–225 transition time, 15, 261
succinate–ubiquinone oxidoreductase, see triacylglycerol
electron transport chain fatty acid cycle, 218, 220, 225, 227
sucrose phosphate synthase, 181, 272, 273 lipase
sucrose synthesis, 146, 273 hormone–sensitive, 221, 236, 241,
summation theorem, 111–113, 123, 264, 266, 268
277 synthesis, 266–267
application, 175, 180, 186, 258 tricarboxylic acid cycle, 217, 246, 266
concentration control coefficients, 133 discovery, 29
experimental illustration, 155 labelling studies, 91, 92
limitations, 278–280 triglyceride, see triacylglycerol
 306 Index

triose phosphate isomerase, 93, 174, 176 Wollenberger clamp, 31

tritium, see 3 H Woolf plot, see Hanes–Woolf plot
TRP genes, 264, 265 Wosilait, W. D., 231
tryptophan Wright, B. E., 6, 41, 92, 94, 169
2,3–dioxygenase, 94 Wyman, J., see allosteric enzyme, con-
flux control coefficient, 151, 166 certed model
5–mono–oxygenase, 237
catabolism, 94, 151, 166, 168, 274 Yamazaki, R., 237
synthesis, 213, 216 Yates, R. A., 205
control analysis, 264 yeast
engineering, 265 glycolysis, see ethanol, glucose trans-
uptake port
flux control coefficient, 166, 168, glycolytic oscillations, 208
171, 196 multicopy plasmid, 146
turnover time, 31, 228 Pasteur effect, 97
two–substrate kinetics, see enzyme kinet- permeabilized, 29
ics tryptophan synthesis, see tryptophan
tyrosine
3–mono–oxygenase, 237 zero–order ultrasensitivity, see covalent
aminotransferase modification
flux control coefficient, 168 Zuurendonk, P. F., 36
catabolism, 168 zymogen, 230
protein kinase, 233
synthesis, 213
uptake
flux control coefficient, 171, 196

ubiquitin, see enzyme degradation

ultrasensitivity, see covalent modification
Umbarger, H. E., 205
uncoupler, 162, 163, 186
Underwood, A. H., 87, 218
Universal Method, 263–265, 276, 282
urea cycle, 266
control analysis, 168, 171
flux change, 268
uridylylation, see covalent modification,
nucleotidylation

Vmax , see limiting rate

V system, see allosteric enzyme
van Dam, K., 148, 157
Van Schaftingen, E., 38
variable, 2, 104, 124, 132
Veech, R. L., 89, 90, 93
Voit, E. O., 276

Waley, S. G., 103, 106

Walsh, K., 147
Wanders, R. A., 157, 170
water
extracellular, 35
intracellular, 35
Weber, G., 268
Welch, G. R., 21
Westerhoff, H. V., 149, 157, 159, 160, 278
Wilkinson, G. N., 53

View publication stats