Phil. Trans. R. Soc.

B (2006) 361, 1761–1776

doi:10.1098/rstb.2006.1912
Published online 7 September 2006

The origin of replicators and reproducers

Eörs Szathmáry1,2,*
1
Collegium Budapest (Institute for Advanced Study), 2 Szentháromság utca, 1014 Budapest, Hungary
2
Department of Plant Taxonomy and Ecology, Institute of Biology, Eötvös University,
1/c Pázmány Péter sétany 1117 Budapest, Hungary
Replicators are fundamental to the origin of life and evolvability. Their survival depends on the
accuracy of replication and the efficiency of growth relative to spontaneous decay. Infrabiological
systems are built of two coupled autocatalytic systems, in contrast to minimal living systems that must
comprise at least a metabolic subsystem, a hereditary subsystem and a boundary, serving respective
functions. Some scenarios prefer to unite all these functions into one primordial system, as illustrated
in the lipid world scenario, which is considered as a didactic example in detail. Experimentally
produced chemical replicators grow parabolically owing to product inhibition. A selection
consequence is survival of everybody. The chromatographized replicator model predicts that such
replicators spreading on surfaces can be selected for higher replication rate because double strands
are washed away slower than single strands from the surface. Analysis of real ribozymes suggests that
the error threshold of replication is less severe by about one order of magnitude than thought
previously. Surface-bound dynamics is predicted to play a crucial role also for exponential replicators:
unlinked genes belonging to the same genome do not displace each other by competition, and
efficient and accurate replicases can spread. The most efficient form of such useful population
structure is encapsulation by reproducing vesicles. The stochastic corrector model shows how such a
bag of genes can survive, and what the role of chromosome formation and intragenic recombination
could be. Prebiotic and early evolution cannot be understood without the models of dynamics.
Keywords: replicator; origin of life; ribozyme; autocatalysis; compartmentation; error threshold

1. INTRODUCTION molecular and non-molecular replicators with limited

The replicator, as introduced by Dawkins (1976), has or unlimited hereditary potential.
become one of the central concepts in evolutionary Oparin (1961) defined any system capable of
theory. He identified two types of replicator with replication and mutation as alive. Most evolutionary
unbounded evolutionary potential, namely genes and biologists would agree with this view. Systems with
memes (memes were meant to be hereditary units of these properties can evolve complex adaptations
cultural rather than genetic evolution). These ideas (purposeful functions) in the natural world, highly
have turned out to be extremely fruitful: they have characteristic of living beings. Yet some authors
elicited renewed interest in the philosophy of evolution (including Gánti 1971, 1978) have raised doubts
(e.g. Hull 1980) and led to the recognition of other concerning such an approach. The acid test is whether
types of replicators with the most important role in viruses are alive or not. Gánti (1971) argued that to
evolution (Maynard Smith & Szathmáry 1993, 1995). regard viruses as living amounts to a conceptual
A classification of replicators was presented by mistake equating programs with computers. In the
Maynard Smith & Szathmáry (1995) and it has been full analogy, the virus corresponds to a program,
refined a number of times (Szathmáry 1995, 2000). written in a decodable language, which says to the
Most widely known replicators, including genes, are computer: ‘Print me again and again, even if you
strongly tied to the world of chemistry: this is obviously disintegrate as a result of doing so!’ The active part is
not true for memes. Some replicators have only limited obviously the computer and not the program. The
heredity (Maynard Smith & Szathmáry 1995), imply- computer can do many things without such a malign
ing that the number of possible types is smaller than or program. In sharp contrast, the program cannot do
roughly equal to the number of individuals (copies, anything on its own. The living cell is thus analogous to
sequences, etc.) in a plausible (realistic) system. the computer. Since everyone regards the cell in its
Conversely, in the case of unlimited hereditary active state alive, life as such in the example rests with
replicators, the number of types by far exceeds that of the cell rather than the virus.
individuals in the population (Szathmáry & Maynard Yet viruses evolve. In fact, they have become one of
Smith 1997). This shows that a classification of
the most accessible test systems for evolutionary
replicators is not naturally hierarchical: there exist
hypotheses (e.g. Poon & Chao 2004). Computer
programs can also evolve (e.g. Bedau et al. 2000).
*[email protected] What is the relationship between units of evolution and
One contribution of 19 to a Discussion Meeting Issue ‘Conditions for units of life? To give a tentative answer, both the
the emergence of life on the early Earth’. concepts must be defined first with sufficient clarity,

1761 This journal is q 2006 The Royal Society

1762 E. Szathmáry Origin of replicators and reproducers

and only after this the two notions can be compared.

Units of evolution must: (i) multiply, (ii) have heredity
and (iii) heredity must not be totally accurate
(variability). Furthermore, some of the inherited traits
must affect the chance of reproduction or of survival of
the units. If all these criteria are met, then in a
population of such entities, evolution by natural
selection can take place (Maynard Smith 1986). Note
that this definition does not refer to living systems. Any
system satisfying these criteria can evolve in a
Darwinian manner.
Units of life as such are less well studied, although
cells and organisms are widely known and analysed.
Gánti (1971, 1979, 1987, 2003) has refined his ‘life
criteria’ that living systems must meet. He observed,
correctly, that for the individual living state, reproduc-
tion is neither necessary nor sufficient. Many cells
and organisms are commonly regarded alive even if formaldehyde glycolaldehyde
they cannot reproduce (any longer). The so-called
potential life criteria must be met only if the population Figure 1. The autocatalytic core or seed of the formose
reaction (Fernando et al. 2005). Each circle represents a
of units is to be maintained and evolved. Then, the
chemical group including one carbon atom.
correct relationship between units of evolution and
units of life is that of two partially overlapping sets flow reactor where total concentration is kept constant,
(Szathmáry 2002). then the rate equations for growth and competition
Some regard the concept of a replicator more become
informational, detached from real processes of replica-
tion, reproduction and development. The elegant dx=dt Z xKQKxF; ð2:1aÞ
concept of a reproducer (Griesemer 2000, 2002) is dy=dt Z yk C xK ð1KQÞKyF; ð2:1bÞ
meant to fill this gap. A reproducer is a unit of where x and y are concentrations of wild-type and
multiplication, hereditary variation and development. mutant, respectively, FZxKCyk and total concen-
A reproducer must have at least a minimum develop- tration is (without loss of generality) unity. It is easy to
mental capacity required for further multiplication. see that in equilibrium, when both templates are
There is not only an informational link but also present in non-zero concentration, it holds that
material overlap between generations of reproducers.
Thus, genes in an organism are replicators but not ðKQKkÞ
xZ ; ð2:2Þ
reproducers. Conversely, an organism is not a repli- ðK KkÞ
cator but reproducer. In the course of prebiotic and where it must be true that QOk/K. If there are n
early biological evolution, replicators ganged up to digits in the sequence, QZqn can be approximated by
yield reproducers. We shall consider in detail how this eKn(1Kq), where q is the copying fidelity per base per
could have happened. replication. From this we obtain
lnðK =kÞ
n! ; ð2:3Þ
2. SURVIVAL CRITERIA FOR INFORMATIONAL ð1KqÞ
REPLICATORS
Informational replicators, such as genes, have unlim- which is Eigen’s error threshold of replication. Non-
ited heredity. The earliest informational replicators enzymatic replication implies low q, so n!100 is
must have faced at least two severe constraints. Serious probable for prebiotic chemistry, which is about the
considerations suggest that primordial nucleic acids size of a tRNA molecule. Therefore, early genomes
(or their analogues) must have been rather short must have consisted of independently replicating
molecules owing to excessive noise in their copying. entities. But they would compete with each other and
Another consideration emphasizes the fact that repli- the one with the highest fitness would win (Eigen
cators must have a growth rate high enough to 1971). Hence, the ‘Catch-22’ of molecular evolution:
compensate for spontaneous decay. I consider these no enzymes without a large genome and no genome
two aspects in turn. without enzymes (Maynard Smith 1983).

(a) The error threshold (b) The decay threshold

Eigen (1971) called attention to the fact that the length Consider, for a change, a non-informational replicator,
of molecules (number of nucleotides) maintained in such as any intermediate in the formose reaction
mutation–selection balance is limited by the copying (figure 1). Note that such an autocatalytic cycle differs
fidelity. We recapitulate the simplified treatment by markedly from Kauffman’s (1993) reflexively auto-
Maynard Smith (1983). Imagine two sequences with catalytic protein nets: in the former, each elementary
replication rate constants K and k(!K ), respectively. reaction is stoichiometric rather than catalytic. There is
The first sequence mutates into the second with a a severe problem with the formose reaction: deadly side
mutation rate (1KQ). If we assume that they are in a reactions drain it to such an extent that the

Phil. Trans. R. Soc. B (2006)

 Origin of replicators and reproducers E. Szathmáry 1763

intermediates of the cycle disappear ultimately (e.g. hurdle (Szathmáry 2000). The same considerations
Shapiro 1986). This may have been different for cycles necessarily apply to the fittest cycles. If they coexist,
on surfaces, but we do not know (yet). As King (1982, ecology tells us that they must occupy different niches
1986) pointed out, the smaller the cycle, the better the in abstract space, such as requiring different com-
chances for its propagation. Suppose that there is a bination of raw materials.
simple autocatalytic cycle of p steps (similar to the An alternative way of maintaining a variety of cycles
system in figure 2, where pZ4). At each step, the is a high mutation rate (low copying fidelity). This is
legitimate reaction leads to the next cycle intermediate, true, but low copying fidelity does not allow the
and a number of side reactions drain the system. The selection for the fittest, because the system gets below
latter give rise to all sorts of unwanted by-products. Let the error threshold of replication (see §2a). In such a
the specificity of a reaction at step i be si, which is the case, the cycles would cease to be selectable individuals:
rate of legitimate reaction divided by the total rate of all they would rather form a single, un-evolvable network.
(legitimateCside) reactions. Successful growth of the Orgel (1992) called attention to the fact that the
cycle is guaranteed if intermediates of formose reaction are not informational
Yp replicators. In the prebiotic context, Wächtershäuser
2 si O 1; ð2:4Þ (1992) called attention to the possibility that there
iZ1 could be, in principle, a limited set of metabolic
or if we calculate with the geometric mean s of the replicators. These replicators could have limited
specificities heredity, allowing some evolution by natural selection.
sp O 1=2; i:e: p!Klogð2Þ=logðsÞ: ð2:5Þ This possibility is intriguing, but it is without any direct
experimental support at present: nobody has seen a
This shows that the viable system size p increases metabolic replicator, other than the formose reaction,
hyperbolically with specificity. Let us apply Eigen’s that would run without enzymes. In contemporary
(1971) full dynamical formalism to this problem systems, such cycles (the Calvin cycle, the reductive
(Szathmáry 2002) by assuming that there can be a citric acid cycle) are well above the damage threshold
number of alternative cycles such as the formose outlined here, owing to the rate-enhancing effect of
reaction that occasionally can produce each other’s evolved enzymes. Thus, the requisite degree of metabolic
intermediates: channelling is one of the biggest (if not the biggest) hurdles of
X
n the origin of life.
x_i Z ðRi Qi KDi Þxi C wij xj Kxi F; ð2:6Þ
jsi

where xi is the concentration of species i; Ri, the rate of

3. INFRABIOLOGICAL SYSTEMS AND THE LIPID
replication irrespective of the correctness of the off-
WORLD SCENARIO
spring; Qi, the fidelity of replication; Di, the rate of
We do not know where RNA came from. Some people
spontaneous decomposition; wij, the mutation rate
think that the first replicators were not even template-
from species j to species i; and F, an outflow ensuring
based; as we shall see reproducing compartments
that the total concentration remains unity. Here, the
(vesicles, micelles) are favoured by some. Others see
different ‘species’ mean the catalytic seeds of different
the crucial steps in the linking of different autocatalytic
alternative cycles (if their existence is feasible, see
systems that ultimately could evolve into primitive
below), and ‘mutation’ refers to the ‘macromutation’,
living systems.
producing an intermediate of another autocatalytic
cycle. Spontaneous decay corresponds to irreversible
side reactions; in the case of DNA, it means damage (a) Infrabiological systems
(rather than mutation; damaged DNA is chemically no Gánti (e.g. 2003) emphasized that contemporary living
longer DNA). systems always have: (i) some metabolic subsystem,
When is species i viable? It means that it can increase (ii) some systems for heritable control and (iii) some
in concentration when rare. If we forget about selection boundary system to keep the component together. So
of, and mutations to, this species for a moment, from I consider it unlikely that a chemical system satisfying
equation (2.6) we obtain all the constraints from this abstraction could have
Ri Qi KDi O 0; or Ri Qi O Di ; ð2:7Þ appeared just out of chemical chaos. This observation
led to the formulation of the concept of infrabiological
which after rearrangement yields systems (Szathmáry 2005; Fernando et al. 2005).
1O Qi O Di =Ri O 0; ð2:8Þ Infrabiological systems always lack one of the key
components just listed. For example, in the original
where it also holds that formulation of Ganti (1971), a model of minimal life
did not include a boundary system. The combination
Ri O Di : ð2:9Þ
of a metabolic cycle and a membrane was conceived
Lack of enzymatic catalysis implies that the decay also by Gánti (1978), and called a self-reproducing
rate is rather high. Inequalities (2.8) and (2.9) suggest microsphere. In contrast, Szostak et al. (2001)
that copying fidelity must be high. Fortunately, this fits, conceived a protocell-like entity with a boundary and
since mutations are expected to be very rare in the template replication but no metabolic subsystem. Such
systems composed of cycles of small molecules (most systems show a crucial subset of necessary biological
fluctuations cannot propagate their own kind). Thus phenomena. The three subsystems can be combined to
for autocatalytic cycles, damage is the most severe yield three different doublet systems (figure 2).

Phil. Trans. R. Soc. B (2006)

1764 E. Szathmáry Origin of replicators and reproducers

The process imagined is shown in figure 3. It

metabolism MB displays a reflexively autocatalytic micelle with many
components. The incorporation of amphiphile Li may
boundary MT MBT be catalysed by amphiphile Lj at rate enhancement bij
(the ratio of catalysed and uncatalysed reaction rates).
The crucial question is this: where can one obtain the
template BT values of bij, considering the fact that no such system
has been realized so far (the experimental cases are all
Figure 2. Elementary combinatorics of infrabiological directly autocatalytic and show no heredity; see
systems (Fernando et al. 2005). The chemoton is a biological Fernando et al. 2005 for review)? The authors suggest
minimal system comprising three qualitatively different translating the model developed for molecular recog-
subsystems (metabolism, membrane and template). nition between receptors and ligands (Segré et al.
1998). If catalysis depends on recognition of substrate
by catalyst, the reasoning is sound implying that
(b) Composomes and the graded autocatalytic catalysis is a graded phenomenon. From this empiri-
replication domain model cally constrained theoretical distribution, the authors
An interesting line of research has been initiated by obtain the distribution of bij values in their model.
Doron Lancet with his group, conveniently referred to It is imagined that every micelle (or vesicle) is a
as the ‘lipid world’ scenario (Segré et al. 2001a). The sample with replacement of a set of possible lipid
basic idea is as follows. We know that lipids (more molecules. Some samples will contain mutually auto-
generally, amphiphilic compounds with a hydrophobic catalytic gangs, but not others. The latter ones will not
tail and a hydrophilic head) tend to form supramole- be able to grow. The former will grow and then
cular structures, such as bilayers, micelles and vesicles. fragment/divide by some spontaneous process.
They can grow autocatalytically. Now imagine that we Micelles containing more efficient gangs (characterized
have a mixture of molecules in any one vesicle. Some of by higher bij values) will take over. Such sets have some
them may act as catalysts of certain reactions. It is heredity; the gangs maintain and propagate their
theoretically possible that some will catalyse their own identity by virtue of their mutual catalytic activity.
incorporation (direct autocatalysis), or there will be a What are the major concerns apart from the lack of
gang of molecules each exerting some catalytic an experimental basis (at this moment) of this model?
function; thus as a net result, the incorporation of all In the light of the foregoing, I see the following
members of the gang is ensured by the gang (reflexive difficulties:
autocatalysis). If this idea holds water, membrane
heredity in the lipid world, and natural selection of (i) This model works only if the bij values are drawn
vesicles without a genetic subsystem, would be feasible. from a lognormal, rather than a normal
The different, reflexively autocatalytic gangs would distribution. In the latter case, there is no
constitute compositional genomes or ‘composomes’ interesting composome population.
(Segré et al. 2001b). Note that the model does not deal (ii) The absolute magnitude of the bij values will
also matter. Side reactions, as in many other
with the formation of the lipid constituents: they are
prebiotic models, are neglected in the lipid
assumed to be there in the surrounding soup.
world scenario. If the catalytic values are too
Now, there is nothing mysterious about compo-
low, then composomes may shrink below the
sitional genomes in the first place. Although relying
decay threshold, even if without decay very
on direct autocatalysis at the molecular level, the
interesting dynamics may unfold.
genome of the stochastic corrector (see §7) is also (iii) Even if the decay threshold is not reached,
a compositional genome in which the genes are composomal replication may be so inaccurate
unlinked and the genome is characterized by gene that fitter composomes cannot be maintained by
composition. Formally, each protocell can be charac- selection; thus the system may be above the
terized by a genome vector with entries denoting the corresponding error threshold.
number of copies of the ith gene in that vesicle. The
change in this number is a stochastic process, which can I hope the fascinating scenario of the lipid world
be characterized by mean and variance. A crucial scenario will be complemented by theoretical investi-
difference is that, in the stochastic corrector model, gations along these lines. Experimental validation is
we are dealing with a bag of template replicators: there another formidable problem.
are no genes in Lancet’s model.
A similar approach is possible while considering (c) Limited heredity in composomes
questions in the lipid world; however, the issue is Contemporary DNA-based organisms have an unlim-
complicated by the fact that we need to tackle the ited hereditary potential, since the number of types that
problem of reflexive autocatalysis. This has also one can construct from the purely informational point
precedence in the literature: the reflexively autocataly- of view greatly exceeds the number of individuals that
tic protein networks (e.g. Kauffman 1993) are perhaps the Earth can maintain. What is the hereditary
the best-known example. I hasten to point out that potential of composomes? They can have limited
nobody has seen real reflexively autocatalytic protein heredity only (Szathmáry 2000). First of all, it is only
sets. Let us see whether one can be more hopeful the composition rather than the steric configuration of
regarding autocatalytic lipid sets. the system that is maintained. In order to appreciate

Phil. Trans. R. Soc. B (2006)

 Origin of replicators and reproducers E. Szathmáry 1765

bij
Li

ki
Lj k –i

Figure 3. The graded autocatalytic replication domain or composome model: catalysed micelle growth and fission (Segré et al.
2001a,b). Li and Lj molecules are different amphiphilic compounds, ki and kKi are rate constants for spontaneous insertion and
emigration of amphiphile Li, and bij is the rate enhancement of getting in and out of this molecule from the micelle, catalysed by
Lj. Note that the model does not deal with the primary origin of Li molecules per se.

this point, consider n types of molecules that we use to simple Malthusian growth: hyperbolic and parabolic
build our replicator of size k. In the case of template growth are faster and slower than Malthusian growth,
(digital, see later) replication, all possible sequences are respectively. Hyperbolic growth was thought to be
potential replicators; Hence, their number is given by relevant for hypercycles (mutualistic molecular repli-
cators), whereas parabolic growth was experimentally
Ns Z nk ; ð3:1Þ demonstrated to happen with small synthetic replica-
as it follows from elementary combinatorics. In the case tors. The consequences for selection in a competitive
of ensemble replicators, the positions do not matter setting are remarkable: survival of the common for
and hence the upper bound for the number of possible hyperbolic growth and survival of everybody for
types is parabolic growth. In this section, I focus mainly on
! parabolic growth and its consequences.
n C kK1 ðn C kK1Þ!
Nc Z Z : ð3:2Þ
k ðnK1Þ!k!
(a) Growth laws and selection consequences
The simplest reproduction process is the binary
This is clearly an upper bound since every possible
fission of the parent object, of which the formal
subset cannot be realized by the alternative attractors
stoichiometry is
associated with the system. For the same n and k, Ns is
always larger than Nc, usually by orders of magnitude. A C S/ 2A C W ;
Indeed, by the application of the Stirling formula for
factorials, one can deduce an approximate equation for where A is a replicator, and S and W are source and
the proportion of the number of types waste materials, respectively (here I follow the treat-
Ns pffiffiffiffiffiffi ment of Szathmáry & Maynard Smith, 1997). The
zkkC1=2 ðnK1ÞnK1=2 nk ðn C kK1Þ1=2KkKn 2p; ð3:3Þ associated kinetic equation describes a Malthusian
Nc
growth process
which, for sufficiently large n and k, further approxi- dx
mates to Z x_ Z kx; ð4:1Þ
dt
Ns pffiffiffiffiffiffi
zkk nkCn ðn C kÞKkKn 2p: ð3:4Þ which means that growth of x (the concentration of A)
Nc
is exponential with a per capita rate constant k,
Note that the number of attractors for such collective provided the concentration of S is kept stationary.
replicators has not been analytically calculated yet. In When two replicators with different rate constant grow
any case, the ratio (3.4) showing the advantage of together, the one with larger k will outgrow the other.
modular template replicators is definitely underesti- This is, of course, elementary. For didactic purposes,
mated. A satisfactory answer must take two consider- let us express this outcome through the ratios of the
ations into account: (i) the number of attractors in sets of growing concentrations
unlimited size (Kauffman 1993) and (ii) finite size k for
realistic systems (Segré et al. 1998). x1 ðtÞ x1 ð0Þek1 t
Z Z Cegt ; g Z k1 Kk2 O 0; ð4:2Þ
x2 ðtÞ x2 ð0Þek2 t

4. PARABOLIC GROWTH, SURVIVAL OF showing that even in a freely growing system, the worse
EVERYBODY AND THE APPEARANCE growing population is diluted out in the limit. This is a
OF DARWINIAN SELECTION very simple demonstration of differential survival.
In the field of prebiotic evolution, non-conventional Departures from this simple scheme are easily
growth laws, such as hyperbolic and parabolic, have imaginable. A minimum complication is that two
been widely discussed. Both represent departures from individuals are necessary to produce a third one

Phil. Trans. R. Soc. B (2006)

1766 E. Szathmáry Origin of replicators and reproducers

(akin to sexual reproduction), such as: and in the limit

2A C X / 3A C Y ; x1 ðtÞ k21
lim Z 2: ð4:8Þ
t/N x2 ðtÞ k2
and the associated growth equation reads
x_ Z kx2 ; ð4:3Þ Thus, ‘survival of everybody’ (Szathmáry 1991a) is
guaranteed, as shown by selection equations in
which is called hyperbolic growth, the selection Szathmáry & Gladkih (1989).
consequences of which are very interesting (Eigen But what kind of molecular mechanism could
1971). In order to see this, let us replace the exponent 2 underlie such an odd type of growth? von Kiedrowski
by p and solve the equation by separation to obtain (1986) and Zielinski & Orgel (1987) were the first to
show that oligonucleotide analogues follow a square-
xðtÞ Z ½ktKkpt C xð0Þ1Kp
1=ð1KpÞ : ð4:4Þ root growth law in the appropriate medium. The
When pO1, defining hyperbolic growth, the system has reason, in a simplified form, is as follows. A template
a finite escape time, i.e. it reaches infinite concentration molecule A reacts with the source materials whereby a
in finite time. As it is easy to check, for pZ2 the new copy of A is made, which remains associated with
asymptote lies at tZ1/[x(0)k]. The smaller the time of the template.
unbounded explosion, the larger x(0)k. Among the a
competitors, the one with the highest initial concentration A C A% AA;
b
times the growth rate constant wins. Thus, initial c
A C X $$% AA:
conditions also determine the outcome of selection
and this phenomenon has been called the ‘survival of Crucial is the ordering of the rate constants a[bO
the common’, where intrinsic fitness is masked by the c, i.e. association of two template molecules is faster
growth law (Michod 1983, 1984). than their dissociation, and replication per se is rate
The relevance of hyperbolic growth and survival of limiting. Note that the immediate product of copying is
the common may be as follows. Eigen (1971) proposed the replicationally inert AA complex. Thus, replication
that the hypercycle might have been a link between in this way is self-limiting. The higher the concen-
solitary genes and bacterial genomes. It is a cycle of tration of A, the stronger this self-limitation is. Note
replicators in which any member catalyses the replica- also that this type of replication is conservative: there is
tion of the next. Each member undergoes a replication no material overlap between copy and template, and
cycle as an autocatalyst, and there is the superimposed template and copy are exactly identical as well as
cyclic network of heterocatalytic aid, hence the term complementary (this can be achieved by palindromes).
hypercycle. Under simplifying kinetic assumptions, the As it is apparent from the above reaction scheme, the
members of the hypercycle grow coherently and rate of replication is determined by the concentration of
hyperbolically (e.g. Eigen 1971; Eigen & Schuster free A, and at high enough total concentration of A
1977). Thus, among a set of rival hypercycles, the (denoted by x) and AA (denoted by y), the former is
already common is likely to win. This dynamics was negligible since association is stronger than dis-
claimed to have been important in the fixation of sociation. The formation and dissociation of AA are
chirality and the genetic code (e.g. Küppers 1983). Yet in quasi-equilibrium, thus
this assumption is unwarranted (Szathmáry 1989a), pffiffiffiffiffiffiffiffiffi pffiffiffi
briefly because: (i) parallel simple autocatalytic replica- ax2 zby; x z by=a zr z; z Z x C y; ð4:9Þ
tion modifies invadability, (ii) stochastic effects allow
uncommon, but intrinsically fitter hypercycles to and therefore,
invade and (iii) spatially distinct habitats would have
dz
allowed for diversity anyway. Thus, although hyper- Z z_ zkz1=2 ; ð4:10Þ
cyclic systems may have played some role in prebiotic dt
evolution, it is unlikely that their hyperbolic growth was which is formally identical with equation (4.5).
very important (cf. Szathmáry et al. 1988). Owing to self-limitation based on molecular com-
Parabolic growth ensues when in the equation plementarity, AA and BB complexes (where A and B
are two different replicators) are stronger than AB
x_ Z kx p ; 0! p! 1; ð4:5Þ
complexes. Hence, each species limits its own growth
the solution of which is also given by equation (4.4). more strongly: this condition for joint survival is also
When pZ1/2, it is reduced to found in traditional Lotka–Volterra competitive
systems. This is the ultimate cause for survival of the
xðtÞ Z ½kt=2 C x1=2 ð0Þ
2 ; ð4:6Þ common in parabolic systems (Szathmáry 1991a).
which is why this type of growth is called parabolic. In the meantime, several more replicators obeying
Parabolic growth entails survival of everybody in a the same type of growth dynamics have been con-
competitive situation. To see this, consider the relative structed among others by Rebek (1994) and Sievers &
concentration of two parabolically growing replicators von Kiedrowski (1995). (In the latter case, the single-
in the same environment stranded templates are not self-complementary.)
A detailed kinetic theory for parabolic growth of
pffiffiffiffiffiffiffiffiffiffiffi 2 minimal replicators was worked out by von Kiedrowski
x1 ðtÞ x1 ð0Þ C k1 t=2
Z pffiffiffiffiffiffiffiffiffiffiffi 2 ; ð4:7Þ (1993). It seems that parabolic growth is a rather robust
x2 ðtÞ x2 ð0Þ C k2 t=2 phenomenon among these replicators, although with

Phil. Trans. R. Soc. B (2006)

 Origin of replicators and reproducers E. Szathmáry 1767

the appropriate ‘molecular gymnastics’ nearly exponen-

r
tial growth can be achieved (Kindermann et al. 2005).
One of the important steps of prebiotic evolution
must have been the emergence of replicators with R
exponential growth. Incidentally, this is very likely to
have opened up the possibility of a transition from
ki
limited to unlimited heredity as well.
Ai

(b) A nontrivial consequence of exponential decay ai

Szathmáry & Gladkih (1989) realized that parabolic Bi
growth as expressed in equation (4.5) results in
bi di
coexistence whenever replicators are in a competitive
situation. The system they used was:
Ai
X
x_i Z ki xip Kxi kj xjp ; ð4:11Þ
di
j

which implies a constraint of constant total population

size (cf. Eigen 1971). The strange result of the analysis
of this system was ‘survival of everybody’ (Szathmáry
1991) in contrast to the classical (Darwinian) case of Figure 4. Stoichiometric scheme of the simplified system with
exponential growth ( pZ1), where survival of the fittest differential decay rates for the double and single strands (von
prevails. This result was mathematically confirmed by Kiedrowski & Szathmáry 2000). The resource R is fed into
Varga & Szathmáry (1997) who, by finding an the system at a constant rate r. The assumption d[d
appropriate Liapunov function, demonstrated that corresponds to that of the more complicated case when the
there was a single internal, globally stable rest point double strand is retained much more strongly than the single
of the system (4.11). strand by the chromatography column.
Lifson & Lifson (1999) recently extended these
findings by demonstrating that if single strands
decompose by spontaneous (exponential) decay, coex- (c) Theory before experiment: the chromatogra-
istence is not possible any more and ‘selection of the phized replicator model
unfittest’ sets in. Independently, von Kiedrowski A common problem of non-enzymatic artificial repli-
(1998) announced that in a simulated chromato- cator systems is product inhibition leading to parabolic
graphic system of competing self-replicators natural instead of exponential amplification. Exponential
selection could happen, despite the fact that this would chemical replication of oligonucleotides was achieved
not be possible in the spatially homogeneous case, by an iterative stepwise procedure, which employs the
modelled by equation (4.11). surface of a solid support and was called Surface
Let us first point out that it is not the system (4.11) Promoted Replication and Exponential Amplification
that the Lifsons modified. If you introduce decay rates of DNA analogues (SPREAD; Luther et al. 1998).
into the model, you get I review theoretical insights (von Kiedrowski &
X Szathmáry 2000) into the design of an autonomous
x_i Z ki xip Kdi xi Kxi ðkj xjp Kdj xj Þ; ð4:12Þ variant of the SPREAD procedure. The corresponding
program simulates a given set of chemical reactions
for which survival of everybody is still guaranteed, coupled to a chromatographic process, where the
despite the specific decay rates di. Using essentially the chromatographic column is treated as a series of
original rationale of Szathmáry & Gladkih (1989) one connected cells. The crucial step is a template-directed
finds that
reaction occurring at the surface: thus it is assumed that
" !#
X   two parabolic replicators compete for their building
x_ i Zxip ki Kx1Kp
i di C ðkj xjp Kdj xi Oxip ki Kx1Kp
i kmax ; blocks in the chromatographic column. A simplified
j
semi-analytic treatment confirms that competing
ð4:13Þ parabolic replicators, which spread on mineral surfaces
wh4ich means that the time derivative is positive if the are amenable for Darwinian selection under a wide
concentration xi is sufficiently low (Scheuring & range of parameter values.
Szathmáry 2001). Now my aim is to demonstrate by a semi-analytically
In their model, the Lifsons assume that ‘double soluble simplified model that differential retention can
strands do not replicate and are resistant to decom- lead to competitive exclusion (von Kiedrowski &
position’ (cf. their equations (3.2) and (4.15)). Their Szathmáry 2000). Consider a single compartment
assumption that double strands do not decompose at with a constant nutrient (raw material) inflow and
all is unrealistic. In the following, I review results by assume that single strands have a higher decay rate than
von Kiedrowski & Szathmáry (2000) that competitive double strands. This is meant to substitute for the
coexistence is still possible under a range of parameter higher retention of double strands on the chromatog-
values for self-replicators with a parabolic growth raphy column. The scheme of reactions is displayed in
tendency, even if decay of strands is taken into account. figure 4. For two species, we have the following

Phil. Trans. R. Soc. B (2006)

1768 E. Szathmáry Origin of replicators and reproducers

ordinary differential equation system: disappears—for replication limits the length of the
9 genome that can be maintained by selection; see
dR=dt Z rKRðk1 A1 C k2 A2 Þ >
> equation (2.3). Primordial replication must have been
  =
dAi =dt Z 2 bi Bi Kai A2i KAi ðki R C di Þ ; ð4:14Þ error-prone, so early replicators are thought to have
>
> been necessarily short (Eigen 1971). The error
;
dBi =dt Z ai A2i Kbi Bi C ki RAi Kdi Bi ; threshold also depends on the fitness landscape. In an
where R is the common resource and Ai, Bi are the RNA world (Gilbert 1986), there will be many neutral
single and double strands of species i, respectively and compensatory mutations that can raise the
threshold, below which the functional phenotype,
(iZ1, 2). We are interested in the conditions under
rather than a particular sequence, is still present.
which invasion by the inferior species when rare is not
A comparative analysis of two extensively mutagenized
possible, i.e. we have competitive exclusion. A crucial
ribozymes has shown that with a copying fidelity of
relation is the following:
0.999 per digit per replication, the phenotypic error
d
RO 2 : ð4:15Þ threshold rises well above 7000 nucleotides, which
k2 permits the selective maintenance of a functionally rich
Thus, when R1 maintained by species 1 alone satisfies ribo-organism with a genome over 100 different genes
condition (4.15), invasion by species 2 is possible, the size of a tRNA (Kun et al. 2005a,b). This ‘only’
otherwise it is impossible. Obviously, if A2 is to invade, requires an order of magnitude improvement in the
then the rate of its template ligation must be large and accuracy of in vitro generated polymerase ribozymes
that of its decay must be small. A symmetric treatment ( Johnston et al. 2001; Müller & Bartel 2003).
applies to invasion by species 1 if species 2 is the Incidentally, this genome size coincides with that
resident one. The significant fact is that the threshold estimated for a minimal cell achieved by top-down
R1 depends on the decay rates of the single strand (d1) analysis (comparative analysis of the genomes of
and the double strand (d1) of the resident species 1 as reduced organisms: Gil et al. 2004) minus the genes
well. dealing with translation.
Competitive exclusion (survival of the fittest) is Eigen’s insight of an error threshold quantifies the
compatible with problem. Following (2.3), we have
d [ d; ð4:16Þ
ln s
n! ; ð5:1Þ
but not the other way round. In the chromatographic ð1KqÞ
case, this corresponds to a high retention factor for the
double strand and low for the single strand. Note that where sZK/k is the so-called selective superiority of the
an increase in d easily throws the system into the region fittest (master) sequence. In this simplified treatment,
of coexistence. all mutants share the same replication rate, neutral
I believe that the chromatographized replicator mutations of and back mutations to the master are
model is relevant to the origin of life on Earth. The ignored.
chromatographic column is equivalent to a tunnel or a The error threshold was first defined in relation to a
riverbed of minerals in which water containing the particular genotype. However, it is obvious that in an
resources is continuously running through. Although RNA world there will be many neutral and compensa-
our model, so far, refers to an isothermal reaction tory mutations, which allow the preservation or the
system, it can be easily extended to account for a restoration of the fittest phenotype rather than of a
gradient of increasing temperature along the direction single genotype. Other things being equal, this will
of the column. As long as parabolic replicators need modify the error threshold by increasing it (thus longer
high temperatures whereas short replicators work at genomes will become maintainable). Since in an RNA
low temperatures (von Kiedrowski 1993), long repli- world the functional ribozymes will have the strongest
cators may grow from the consumption of shorter ones effect on fitness, one should gather the pertinent data
synthesized at the entry of the column where the from known ribozymes. As we shall see, there is just
temperature is low. The chromatographized replicator enough empirical evidence to formulate an encoura-
model can be simplified by means of attributing ging statement.
individual desorption rates to individual decay rates. To construct a fitness/functionality landscape of a
Moreover, the findings from the simplified reaction ribozyme: (i) its secondary structure has to be
model, viz. that both selection and coexistence can experimentally determined, (ii) this secondary structure
occur, has been independently confirmed by cannot contain a pseudo-knot, a special structural
simulations based on the original model. element that conventional RNA folding algorithms
The case presented is an unusual one in that theory cannot satisfactorily cope with, (iii) mutagenesis experi-
makes a clear prediction for experiment. Moreover, ments have to reveal all important sites and nucleotides
experimental realization of the model should be and (iv) the size of the ribozyme should not be very long,
relatively straightforward. otherwise any calculation would be practically unfea-
sible. The first requirement excludes most of the known
ribozymes, since apart from the function only the
5. REAL RIBOZYMES AND A RELAXED sequence has been determined. The naturally occurring
ERROR THRESHOLD ribozymes generally fulfil the third requirement, but
The error threshold—the critical copying fidelity Hepatitis Delta Virus fails to meet the second require-
below which the fittest genotype deterministically ment and Group I and II introns, as well as RNAase P,

Phil. Trans. R. Soc. B (2006)

 Origin of replicators and reproducers E. Szathmáry 1769

Figure 5. Secondary structures of (a) Neurospora VS ribozyme and (b) hairpin ribozyme indicating different regions (Kun et al.
2005a,b). Position numbering follows standard convention. Capitalized nucleotides specify those sites that have been subjected
to mutagenesis experiments, and enzymatic activities of mutants are available. A total of 183 mutants for the VS ribozyme
affecting 83 out of 144 positions, excluding insertions and deletions, were considered. For the hairpin ribozyme, the survey was
based on 142 mutants affecting 39 out of 50 positions of the ribozyme and some part of the substrate region. Nucleotides marked
in bold are the critical sites.

fail to meet the fourth. This leaves the hammerhead, the This is the first time that the fitness landscape in
hairpin and the Neurospora VS ribozymes as possible terms of functionality has been inferred from real
candidates. Kun et al. (2005a) chose the hairpin and the ribozymes (see also Kun et al. 2005b). The phenotypic
Neurospora VS ribozymes for our study (figure 5). Both error threshold thus inferred alleviates Eigen’s paradox.
are relatively short, naturally occurring self-cleaving This relates to the finding that the fitness landscapes
ribozymes, which can be divided into a trans-acting are sufficiently similar. Inequality (5.1) cannot be used
enzyme/substrate system where the trans-acting enzyme to assess the effect of the landscape on the error
part does not contain a pseudo-knot. threshold owing to its restrictive preconditions.
The construction of the fitness/functionality land- A recently derived expression (Takeuchi et al. 2005)
scape is based on four general observations: (i) the offers a much more pertinent approximation:
maintenance of the secondary structure is a major factor
Kln s
in retaining enzymatic activity, but the nature of most n! ; ð5:2Þ
individual base pairs is not important and many can be lnðq C lKqlÞ
reversed or replaced by a different pair without major
where l is the fraction of neutral single substitutions.
loss of activity so long as a base pair is retained at a given
For the VS ribozyme nZ144, qZ0.947, lZ0.26; and
position, (ii) the structure can have slight variations
for the hairpin ribozyme nZ50, qZ0.856, lZ0.22.
which in most cases manifest in some mismatch base Thus, for ln s we obtain 5.761 and 5.957, respectively.
pairs and/or some deletions or elongation in a helical The fitness values obtained allow us to reconsider
region, (iii) there are critical regions in the molecule, Eigen’s paradox. Although it was shown that within-
where the nature of the base located there is also gene recombination could raise the error threshold to
important and (iv) the effect of multiple mutations is some extent, it has been unknown until recently what
multiplicative, i.e. the product of the activities of single would be the required accuracy of a sufficient replicase
mutants provides the activity of the multiple mutants. ribozyme in a ribo-organism. Substituting an accuracy
From the fitness/functionality landscapes, the esti- of qZ0.999 in the lower bound of viral RNA replicases
mated phenotypic error thresholds are mZ _ 0:0533 and into inequality (5.2), and using the two obtained values
_ 0:144 for the VS and hairpin ribozymes, respect-
mZ for l, we find that nZ7000–8000; namely, such a
ively, where m_ is the effective mutation rate per ribozyme could replicate a genome consisting of more
nucleotide per replication. As expected, these figures than 100 different genes each of length 70 nucleotides
are substantially higher than those inferred from fitness or more than 70 different genes each of length 100.
landscapes that do not take into account the secondary This would be sufficient to run a functionally rich ribo-
structure of the ribozymes but include information on organism, estimated to harbour about this number of
single mutational effects. genes (Jeffares et al. 1998). Incidentally, a recent

Phil. Trans. R. Soc. B (2006)

1770 E. Szathmáry Origin of replicators and reproducers

analysis of a core minimal bacterial gene set gives about linked to the fact that such a hypercycle without
200 genes (Gil et al. 2004). This shows that if we take population structure shows sustained oscillation in
away the genes coding for the whole contemporary time. Each wing of a rotating spiral looks a bit like the
translation system, we are again in the same ballpark. arm of a galaxy, and is dominated by templates of the
The artificial template-dependent RNA polymerase same membership in the hypercycle. Parasites are
ribozyme selected by Johnston et al. (2001) has an unable to kill the hypercycle in that system. This
average fidelity qZ0.97. Using formula (5.2) and the finding was attributed to the dynamics of spirals. Two
fitness/functionality landscape obtained for the VS and questions emerge: Are spirals necessary? What happens
hairpin ribozymes (an admitted leap), it was concluded if one models other systems in the same way (i.e. by
that the accuracy of this ribozyme would allow the cellular automata)?
maintenance of replicators with length around 250, The dynamics of the non-spatial version of the
which means that this ribozyme could replicate itself if metabolic system looks as follows.
other conditions (such as processivity) were favourable.
In order to eliminate the burden of Eigen’s paradox, a dxi
replicase with an error rate of 10K3 per nucleotide per Z xi ½ki MðxÞKFðxÞ
; ð6:1Þ
dt
replication might have been sufficient to provide the
minimal life requirements in the RNA world. where xi stands for the concentrations of template Ii,
and x is the vector of these concentrations. M(x) is a
multiplicative function of the concentrations of all the
6. REPLICATOR EVOLUTION ON THE SURFACE templates, and F(x) is an outflow term representing a
It is a common experience in theoretical ecology selection constraint (constant total concentration).
and evolutionary biology that population structure This formulation is formally identical to that given by
promotes coexistence and favours the spread of Eigen & Schuster (1978) for a ‘minimum model of
altruism. Importantly, theoretical investigations in primitive translation’. As they noted correctly, the fact
the field of early evolution have paved the way for that replication of any template is impossible without
such investigations to a considerable extent. Without the presence of all the others does not prohibit the
the aim of completeness, I survey some interesting system from undergoing competitive exclusion: M(x) is
relevant examples. same in all the equations, hence the system essentially
behaves as a collection of Malthusian competitors,
(a) Metabolic ribozymes coexist on surfaces whose dynamics are influenced by a common time-
Imagine a non-hypercyclic, so-called ‘metabolic’ dependent factor.
system (cf. figure 45 in Eigen & Schuster 1978). It is assumed that the replicators Ii have dual
Undoubtedly, we are here comfortably in the RNA functionality: as templates they are necessary for their
world: we assume that informational replication and own replication (autocatalysis), and as ‘ribozymes’
selection for enzymatic function has already been (RNAs able to act as enzymes) they contribute to
achieved. The templates are assumed to contribute to metabolism producing the monomers.
metabolism via enzymatic aid; metabolic products are Now we assume that replication takes place on
in turn used up by the templates for replication at the surface of a mineral (possibly pyrite) substrate.
different rates. Although all templates contribute to The replicator molecules themselves are of a finite
metabolism (‘the common good’), they are able to use size; therefore the number of replicators bound to a
it with different efficiency. Thus in a spatially unit area of the substrate is constrained. We consider
homogenous environment, competitive exclusion a two-dimensional square lattice of binding sites as
follows despite the metabolic coupling (Eigen & the scene of the replication–diffusion process; each of
Schuster 1978). the sites can harbour a single macromolecule at
Interesting selection dynamics occurs when most. The lattice is toroidal (the opposite edges of
molecules are bound to the surface without being the grid are merged in both dimensions) to avoid
washed away regularly. This problem was modelled by edge effects.
the use of ‘cellular automata’ (Czárán & Szathmáry At tZ0, half of the sites are occupied by n different
2000). Without becoming too technical, it suffices to types of macromolecules (we call n the system size).
say that each square of a grid is assumed to be occupied The replicator types are equally abundant in the initial
by a single molecule (template), or be empty. pattern and individual molecules are randomly
Templates can do two things: to replicate (put an assigned to sites. The other half of the sites are empty
offspring into a neighbouring empty cell if available) initially. Time is discrete; replication, decay and
and hop away into empty sites nearby. Replication may diffusion take place in each generation of the
depend on the composition of the few neighbouring simulation.
cells. In the case of a hypercycle, for example, the The effect of monomer-producing metabolism is
template and a specimen of the preceding cycle implicit in the model, itself directly acting on the
member must be present in the same small area if replication process through a local metabolic function. It
replication of the former is to occur. This of course is local in the sense that its arguments are the copy
makes perfect chemical sense. numbers f(i) of replicator types i (iZ1, ., n) within
Boerlijst & Hogeweg (1991) simulated hypercycles certain localities (neighbourhoods) of the lattice.
on a surface exactly in this way. They found that In accordance with the assumption that the presence
rotating spirals on the surface appear, provided the of a complete set of replicators is necessary for
hypercycle consists of more than four members. This is metabolism to produce monomers for replication, the

Phil. Trans. R. Soc. B (2006)

 Origin of replicators and reproducers E. Szathmáry 1771

metabolic function must be a multiplicative form of more likely to be complemented by a metabolically

within-neighbourhood copy numbers f(i). A simple sufficient set of replicators in its neighbourhood than a
option for the concrete form of the metabolic function common one.
M( fs) at a site occupied by a replicator s is the
geometric mean of the copy numbers fs(i ) within the (b) Reciprocal altruism on the rocks and
metabolic neighbourhood of s, i.e. the evolution of replicases
" #1=n Although the question where the first RNA molecules
Y
n
came from is still unsolved, it is nevertheless assumed
Mð f s Þ Z fs ðiÞ : ð6:2Þ
iZ1
that catalytic RNA enzymes (ribozymes) with replicase
function emerged at some stage of early evolution.
Note that M( fs) is zero if any of the replicator types Eigen’s finding of the error threshold demonstrates that
is missing from the metabolic neighbourhood of s, and the length of templates maintained by selection is
that the larger and more uniform the copy numbers of limited by the copying fidelity; therefore, other things
the different replicator types within the metabolic being equal, an increase in template length is
neighbourhood, the more efficient the metabolism at disadvantageous. On the contrary, longer molecules
the given locality. By choosing (6.2) as the metabolic are expected to be better replicases—a feature not
function, we assume that the conspecific replicators incorporated in the original model. An iterative
within the same neighbourhood help replication and scenario for longer and longer molecules with better
that the focal replicator supports its own replication. and better replicase function has been suggested
The first assumption can be interpreted as metabolism ( James & Ellington 1999; Poole et al. 1999) and
being somewhat faster locally in the presence of more analysed mathematically (Scheuring 2000). A crucial
catalysts. The actual effect should be rather weak and it open question is whether parasites (efficient templates
should vanish with the copy number increasing; this that are inefficient replicases) can ruin the system.
feature is properly reflected in the metabolic function Absorption to mineral surfaces was hypothesized to
(6.2): if a replicator type is already present in a help replicases find their useful colleagues in the
replication neighbourhood, then its successive copies immediate neighbourhood ( Joyce & Orgel 1999).
do not add too much to the replication chance of the A cellular automaton simulation revealed that copying
focal template. Implicit in the second assumption is fidelity, replicase speed and template efficiency could
that the time-scale of metabolite diffusion out of the increase by evolution, despite the presence of molecular
neighbourhood in which it was produced is longer than parasites, essentially owing to reciprocal altruism on
that of the catalysed reactions of metabolism. The the surface, thus making the scenario for a gradual
‘habitat’ of the reaction-diffusion system being an improvement of replicase function more plausible
absorptive mineral surface is again straightforward to (Szabó et al. 2002).
assume. The size of the metabolically effective Consider a population of macromolecules, adsorbed
neighbourhood is an implicit measure of metabolite to a surface and built of four different monomers: A, B,
and monomer diffusivity: larger neighbourhoods rep- C and D. Owing to their catalytic activity, macro-
resent faster diffusion of the intermediate metabolites molecules located on neighbouring sites of the surface
and the monomers. can template-replicate each other, which means build-
Czárán & Szathmáry (2000) managed to show that ing a new macromolecule from free monomers by
given such a spatial setting, non-hypercyclic systems copying an existing one. In each replication process, two
are once again viable alternatives. The fundamental replicator molecules are involved: one is the template
difference between their model and that of Boerlijst & and the other acts as a replicase enzyme. We attribute
Hogeweg (1991) is the following: the dynamical link two main properties to replication events, speed and
among the replicators is realized through a common fidelity, which in turn depend on three parameters of the
metabolism, instead of the direct, intransitive hyper- two replicators involved in the process:
cyclic coupling. Using the cellular automaton model of
the metabolic system, the aim was to show that (i) replicase activity expresses how fast the molecule
can add a monomer to a primer while acting as a
(i) metabolic coupling can lead to coexistence of replicase,
replicators in spite of an inherent competitive (ii) replicase fidelity measures the accuracy of replica-
tendency, tion per monomer when the molecule acts as a
(ii) parasites cannot easily kill the whole system and replicase and
(iii) complexity can increase by natural selection. (iii) template efficiency defines an average ‘affinity’ of
the molecule behaving as a template against
The result that there is coexistence without any others.
conspicuous pattern (i.e. something like spirals) is robust
and counter-intuitive. It is owing to the inherent The authors assumed that these traits are in a three-
discreteness (i.e. the corpuscular nature of the way tradeoff: there were no free lunches. Replication
replicator molecule populations) and spatial explicit- speed depends on the activity of the replicase and the
ness of the model, which grasp essential features of the quality of the template: higher replicase activity and
living world in general, and macromolecular replicator template efficiency result in faster replication. Given
systems in particular. An inferior (i.e. more slowly two neighbouring replicator molecules, L and M, on
replicating) molecule type does not die out since there the surface, one of the two different replication events
is an advantage of rarity in the system: a rare template is can occur between them: either L as replicase copies

Phil. Trans. R. Soc. B (2006)

1772 E. Szathmáry Origin of replicators and reproducers

and M as a template, or the other way round.

Mutations allowed not only point mutations but also I1
additions and deletions of one nucleotide
The outcome was a bimodal population: efficient
replicases evolved and short parasites could not ruin
the system. This result, together with the chromato-
graphized replicator model, emphasizes the import- R2 R1
ance of surface dynamics in prebiotic evolution. It also
raises the idea that compartmentation offered by
vesicles could have been an even more efficient means
to evolve more efficient and accurate systems, a
possibility to which I now turn.
I2
replication
7. BAGS OF GENES: THE STOCHASTIC
translation
CORRECTOR MODEL
It is true that the hypercyclic link ensures indefinite Figure 6. The hypercycle with translation. Ri is a replicase
ecological survival of all member replicators. However, protein enzyme coded for by gene Ii.
problems arise when mutations are taken into account.
In order to consider them, it is worthwhile to look at a (iii) Replication of templates is described by
diagram where auto- and heterocatalytic aids are stochastic means. Since the number of genes
functionally clearly separate, such as in a hypercycle in any compartment is small (up to a few
with protein replicases (figure 6). Mutants providing hundred), their growth is affected by the luck
stronger heterocatalytic aid to the next member are not of the draw. Ecologists would express this as
selected. In contrast, increased autocatalysis is always demographic stochasticity.
selected, irrespective of its concomitant effect on (iv) There is no individual regulation of template
heterocatalytic efficiency. This is the well-known copy number per protocell.
problem of parasites in the hypercycle (Maynard (v) Templates are assorted randomly into off-
Smith 1979). As Eigen et al. (1981) observed, putting spring cells upon protocell division.
hypercycles into reproducing compartments helps,
because ‘good’ hypercycles (with efficient heterocata- I must emphasize that in the stochastic corrector
lysis) can be favoured over ‘bad’ ones. The following model, the templates are not coupled to one another
two questions arise out of this: through a reflexive (intransitive) cycle of replicational
aid, since it would be a hypercycle. Instead, we assume
(i) Are there other means whereby parasites can be that they contribute to the ‘common good’ of the
selected against? protocell by catalysing steps of its metabolism. Within
(ii) Are there non-hypercyclic systems that function each compartment, the templates are free to compete
well in a compartment context? because they can reap the benefits of a common
metabolism differently. (A similar situation can arise
The answers turned out to be ‘yes’ to both of these among chromosomes and plasmids in contemporary
questions; I discuss them below. bacteria.) Despite the fact that templates compete, the two
sources of stochasticity generate between-cell variation in
(a) Group selection of early replicators template copy number on which natural selection (between
The phase of evolution just outlined refers to the protocells) can act. This is an efficient means of group
pre-cellular level. Later in evolution, protocells selection of templates, since it is the protocells that are
must have appeared. It turns out that cellularization the groups obeying the stringent criteria: (i) there are
offers the most natural, and at the same time most many more groups than templates, (ii) each group has
efficient, resolution to Eigen’s paradox. It also leads only one ancestor and (iii) there is no migration between
to the appearance of linkage, i.e. the origin of the groups (cf. Leigh 1983). Grey et al. (1995) gave a
chromosomes. The dynamics of genes encapsulated fully rigorous re-examination of the stochastic corrector
in a reproductive protocell is described by the model. The two mentioned sources of stochasticity
stochastic corrector model (Szathmáry & Demeter effectively lead to the correction of a malign within-
1987; Szathmáry 1989a,b; Grey et al. 1995; Zintzaras protocell trend of harmful competition of the templates.
et al. 2002; Fontanari et al. 2006). It rests on the It cannot be too strongly emphasized that the stochastic
following assumptions (figure 7). corrector is not, contrary to common misunderstand-
ing, a hypercyclic system. Hypercycles need compartments
(i) Templates contribute to the fitness of the but compartments can live without hypercycles. It is
protocell as a whole and there is an optimal interesting to see that genuine group selection is likely
proportion of the genes. Concretely, we to have aided a major transition from naked genes to
assume that the genes encode enzymatic aid protocells. Group structure is provided by the physical
given to the intracellular metabolism. boundaries of cells.
(ii) Templates compete with each other within the Within the same context, the origin and establish-
same protocell. As before, replication rates ment of chromosomes (linked genes) in the popu-
may differ from gene to gene. lation have also been analysed (Maynard Smith &

Phil. Trans. R. Soc. B (2006)

 Origin of replicators and reproducers E. Szathmáry 1773

stochastic replication
stochastic fission
Figure 7. The stochastic corrector model. Different templates (labelled by open and closed circles) contribute to the well being
of the compartments (protocells) in that they catalyse steps of metabolism, for example. During protocell growth, templates
replicate at differential expected rates stochastically. Upon division, there is chance assortment of templates into offspring
compartments. Stochastic replication and reassortment generate variation among protocells on which natural selection at the
compartment level can act and oppose to (correct) internal deterioration owing to within-cell competition.

Szathmáry 1993). A chromosome consisting of two the stochastic corrector model. The results on real
genes takes about twice as long to be replicated as ribozymes (§5) alleviate, but do not solve, the
the single genes. It turns out that chromosomes are problem. Lehman (2003) raised the issue that
strongly selected for at the cellular level even if they recombination, a frequently ignored player in models
have this twofold within-cell disadvantage. Linkage of early evolution, could have been crucial to build up
reduces intracellular competition (genes are necess- primeval genomes of sizeable length. In the article that
arily replicated simultaneously) as well as the risk of coined the phrase ‘the RNA world’, Gilbert (1986)
losing one gene by chance upon cell division (a gene already speculated that ‘the RNA molecules evolve in
is certain to find its complementing partner in the self-replicating patterns, using recombination and
same offspring cell if it is linked to it). The molecular mutation to explore new functions and to adapt to
biology of the transition from genes to chromosomes new niches’. In this context, Riley & Lehman (2003)
has also been worked out (Szathmáry & Maynard have shown that Tetrahymena and Azoarcus ribozymes
Smith 1993). can promote RNA recombination.
This capability of RNA recombination to potentially
(b) Sex and protocells reduce the burden imposed by the error threshold has
The results on coexistence leave one (one could say been recently analysed by Santos et al. (2004). They
the original) question in the dark: does the error assumed that the recombination in protocells took
threshold increase or decrease in various systems? place via copy-choice means, i.e. the replicase switched
Although it was shown that the stochastic corrector between RNA-like templates, as occurs frequently in
model performs better than the compartmentalized RNA viruses and is crucial for retroviral replication
hypercycle under a high error rate (Zintzaras et al. during reverse transcription. The numerical results
2002), we still do not know the selectively maintain- showed that there is a quite intricate interplay between
able genome size (or the number of different genes) in mutation, recombination and gene redundancy, but

Phil. Trans. R. Soc. B (2006)

1774 E. Szathmáry Origin of replicators and reproducers

the conclusion from the fitness function they used was Fernando, C., Santos, M. & Szathmáry, E. 2005 Evolution-
that the informational content could have increased by ary potential and requirements for minimal protocells.
25% by keeping the same mutational load as that for a Top. Curr. Chem. 259, 167–211.
population without recombination. Fontanari, J. F., Santos, M. & Szathmáry, E. 2006
Coexistence and error propagation in pre-biotic vesicle
The consequences of imperfect replication in vesicle
models: a group selection approach. J. Theor. Biol. 239,
models are puzzling. For small mutation rates, 247–256. (doi:10.1016/j.jtbi.2005.08.039)
increased level of polyploidy favours the persistence Gánti, T. 1971 The principle of life (in Hungarian). Budapest,
of protocell lineages since the random loss of essential Hungary: Gondolat.
genes after fission is attenuated. However, for large Gánti, T. 1978 The principle of life (in Hungarian). Budapest,
mutation rates, the situation is reversed: those lineages Hungary: Gondolat.
with low levels of polyploidy are better able to cope with Gánti, T. 1979 A theory of biochemical supersystems. Baltimore,
higher mutation rates, particularly when recombina- MD: University park press.
tion is allowed. This means that gene redundancy was Gánti, T. 1987 The principle of life. Budapest, Hungary:
indeed costly. Therefore, selective forces favouring the OMIKK.
linkage of genes to make the first chromosomes would Gánti, T. 2003 The principles of life. Oxford, UK: Oxford
University Press.
eventually outweigh the advantage of faster replicating
Gil, R., Silva, F. J., Peretó, J. & Moya, A. 2004 Determination
single genes, because linked genes are less likely to be of the core of a minimal bacterial gene set. Microbiol. Mol.
lost by random assortment when protocells divide Biol. Rev. 68, 518–537. (doi:10.1128/MMBR.68.3.518-
(Maynard Smith & Szathmáry 1993). 537.2004)
The role of the number of gene copies in a primitive Gilbert, W. 1986 The RNA world. Nature 319, 818. (doi:10.
cell was investigated by Koch (1984), who pointed out 1038/319618a0)
the existence of two conflicting forces: (i) higher copy Grey, D., Hutson, V. & Szathmáry, E. 1995 A re-examination
numbers act as a safeguard against random loss of all of the stochastic corrector model. Proc. R. Soc. B 262,
copies of a gene but (ii) such copy numbers slow down 29–35.
adaptive evolution because a newly arisen favourable Griesemer, J. 2000 The units of evolutionary transition.
mutant is diluted out and cannot be ‘seen’ efficiently by Selection 1, 67–80. (doi:10.1556/Select.1.2000.1-3.7)
Griesemer, J. 2002 What is “epi” about epigenetics? Ann. NY
natural selection acting on cells. He further observed that
Acad. Sci. 981, 97–110.
a moderately high (less than 100) copy number per gene Hull, D. L. 1980 Individuality and selection. Annu. Rev. Ecol.
is not only optimal, but also confers some additional Syst. 11, 311–332. (doi:10.1146/annurev.es.11.110180.
evolvability by the ‘duplication and divergence’ scenario, 001523)
as first emphasized by Ohno (1970). James, K. D. & Ellington, A. D. 1999 The fidelity of template-
directed oligonucleotide ligation and the inevitability of
This work was supported by the Hungarian Scientific Research
polymerase function. Orig. Life Evol. Biosph. 29, 375–390.
Fund (OTKA T 047245) and the National Office for Research
and Technology (NAP 2005/KCKHA005). Helpful comments (doi:10.1023/A:1006544611320)
by two anonymous referees are gratefully acknowledged. Jeffares, D. C., Poole, A. M. & Penny, D. 1998 Relics from the
RNA world. J. Mol. Evol. 46, 18–36. (doi:10.1007/
PL00006280)
Johnston, W. K., Unrau, P. J., Lawrence, M. S., Glasen, M. E.
REFERENCES & Bartel, D. P. 2001 RNA-catalyzed RNA polymerization:
Bedau, M. A., McCaskill, J. S., Packard, N. H., Rasmussen, accurate and general RNA-templated primer extension.
S., Adami, C., Green, D. G., Ikegami, T., Kaneko, K. & Science 292, 1319–1325. (doi:10.1126/science.1060786)
Ray, T. S. 2000 Open problems in artificial life. Artif. Life Joyce, G. F. & Orgel, L. E. 1999 Prospects for understanding
6, 363–376. (doi:10.1162/106454600300103683) the origin of the RNA world. In The RNA world (ed. R. F.
Boerlijst, M. C. & Hogeweg, P. 1991 Spiral wave structure in Gesteland, T. R. Cech & J. F. Atkins) 2nd edn., pp. 49–77.
pre-biotic evolution—hypercycles stable against parasites. Plainview, NY: Cold Spring Harbor Lab. Press.
Physica D48, 17–28. Kauffman, S. A. 1993 The origins of order. Oxford, UK:
Czárán, T. & Szathmáry, E. 2000 Coexistence of replicators Oxford University Press.
in prebiotic evolution. In The geometry of ecological von Kiedrowski, G. 1986 A self-replicating hexadeoxy
interactions: simplifying spatial complexity (ed. U. Dieck- nucleotide. Angew. Chem. Int. Ed. Engl. 25, 932–935.
mann, R. Law & J. A. J. Metz), pp. 116–134. Wien, (doi:10.1002/anie.198609322)
Austria: IIASA and Cambridge University Press. von Kiedrowski, G. 1993 Mimimal replicator theory I:
Dawkins, R. 1976 The selfish gene. Oxford, UK: Oxford parabolic versus exponential growth. Bioorg. Chem.
University Press. Frontiers 3, 113–146.
Eigen, M. 1971 Self-organization of matter and the evolution von Kiedrowski, G. 1998 120. Versammlung Deutscher Ärtze
of biological macromolecules. Naturwissenschaften 58, und Nafurforscher. Berlin, 21st September 1998.
465–523. (doi:10.1007/BF00623322) von Kiedrowski, G. & Szathmáry, E. 2000 Selection versus
Eigen, M. & Schuster, P. 1977 The hypercycle: a principle of coexistence of parabolic replicators spreading on surfaces.
natural self-organization. Part A: emergence of the Selection 1, 173–179. (doi:10.1556/Select.1.2000.1-3.17)
hypercycle. Naturwissenschaften 64, 541–565. (doi:10. Kindermann, M., Stahl, I., Reimold, M., Pankau, W. M. &
1007/BF00450633) von Kiedrowski, G. 2005 Systems chemistry: kinetic and
Eigen, M. & Schuster, P. 1978 The hypercycle: a principle of computational analysis of a nearly exponential organic
natural self-organization. Part C: the realistic hypercycle. replicator. Angew. Chem. 117, 6908–6913. (doi:10.1002/
Naturwissenschaften 65, 341–369. (doi:10.1007/BF0043 ange.200501527) Angew Chem. Int. Ed. Engl. 44:
9699) 6750–6755.
Eigen, M., Schuster, P., Gardiner, W. & Winkler-Oswatitsch, King, G. A. M. 1982 Recycling, reproduction, and life’s
R. 1981 The origin of genetic information. Sci. Am. 244, origin. BioSystems 15, 89–97. (doi:10.1016/0303-2647
78–94. (82)90022-3)

Phil. Trans. R. Soc. B (2006)

 Origin of replicators and reproducers E. Szathmáry 1775

King, G. A. M. 1986 Was there a prebiotic soup? J. Theor. Biol. Santos, M., Zintzaras, E. & Szathmáry, E. 2004
123, 493–498. (doi:10.1016/S0022-5193(86)80216-8) Recombination in primeval genomes: a step forward
Koch, A. L. 1984 Evolution vs the number of gene copies per but still a long leap from maintaining a sizeable
primitive cell. J. Mol. Evol. 20, 71–76. (doi:10.1007/ genome. J. Mol. Evol. 59, 507–519. (doi:10.1007/
BF02101988) s00239-004-2642-7)
Kun, Á., Santos, M. & Szathmáry, E. 2005a Real ribozymes Scheuring, I. 2000 Avoiding Catch-22 of early evolution by
suggest a relaxed error threshold. Nat. Genet. 37, stepwise increase in copying fidelity. Selection 1, 135–145.
1008–1011. (doi:10.1038/ng1621) (doi:10.1556/Select.1.2000.1-3.13)
Kun, A., Maurel, M.-C., Santos, M. & Szathmáry, E. Scheuring, I. & Szathmáry, E. 2001 Survival of replicators
2005b Fitness landscapes, error thresholds, and cofac- with parabolic growth tendency and exponential decay.
tors in aptamer evolution. In The Aptamer handbook J. Theor. Biol. 212, 99–105. (doi:10.1006/jtbi.2001.
(ed. S. Klussmann), pp. 54–92. Weinheim, Germany: 2360)
Wiley-Vch. Segré, D., Lancet, D., Kedem, O. & Pilpel, Y. 1998 Graded
Küppers, B.-O. 1983 Molecular theory of evolution. Berlin, autocatalysis replication domain (GARD): kinetic analysis
Germany: Springer. of self-replication in mutually catalytic sets. Orig. Life Evol.
Lehman, N. 2003 A case for the extreme antiquity of Biosph. 28, 501–514.
recombination. J. Mol. Evol. 56, 770–777. (doi:10.1007/ Segré, D., Ben-Eli, D., Deamer, D. W. & Lancet, D.
s00239-003-2454-1) 2001a The lipid world. Orig. Life Evol. Biosph. 31,
Leigh, E. G. 1983 When does the good of the group override 119–145.
the advantage of the individual? Proc. Natl Acad. Sci. USA Segré, D., Shenhav, B., Kafri, R. & Lancet, D. 2001b The
80, 2985–2989. (doi:10.1073/pnas.80.10.2985) molecular roots of compositional inheritance. J. Theor.
Lifson, S. & Lifson, H. 1999 Models of prebiotic replication: Biol. 213, 481–491.
survival of the fittest versus extinction of the unfittest. Shapiro, R. 1986 A skeptic’s guide to the creation of life on Earth.
J. Theor. Biol. 199, 425–433. (doi:10.1006/jtbi.1999. New York, NY: Summit Books.
0969) Sievers, D. & von Kiedrowski, G. 1995 Self-replication of
Luther, A., Brandsch, R. & von Kiedrowski, G. 1998 Surface- complementary nucleotide-based oligomers. Nature 369,
promoted replication and exponential amplification of 221–224. (doi:10.1038/369221a0)
DNA analogues. Nature 396, 245–248. (doi:10.1038/ Szabó, P., Scheuring, I., Czárán, T. & Szathmáry, E. 2002 In
24343) silico simulations reveal that replicators with limited
Maynard Smith, J. 1979 Hypercycles and the origin of life. dispersal evolve towards higher efficiency and fidelity.
Nature 280, 445–446. (doi:10.1038/280445a0) Nature 420, 360–363.
Maynard Smith, J. 1983 Models of evolution. Proc. R. Soc. B Szathmáry, E. 1989a The emergence, maintenance, and
219, 315–325. transitions of the earliest evolutionary units. Oxf. Surv.
Maynard Smith, J. 1986 The problems of biology. Oxford, UK: Evol. Biol. 6, 169–205.
Oxford University Press. Szathmáry, E. 1989b The integration of the earliest genetic
Maynard Smith, J. & Szathmáry, E. 1993 The origin of information. Trends Ecol. Evol. 4, 200–204. (doi:10.1016/
chromosomes I. Selection for linkage. J. Theor. Biol. 164, 0169-5347(89)90073-6)
437–446. (doi:10.1006/jtbi.1993.1165) Szathmáry, E. 1991 Simple growth laws and selection
Maynard Smith, J. & Szathmáry, E. 1995 The major transitions consequences. Trends Ecol. Evol. 6, 366–370. (doi:10.
in evolution. Oxford, UK: Freeman & Co. 1016/0169-5347(91)90228-P)
Michod, R. 1983 Population biology of the first replicators: Szathmáry, E. 1995 A classification of replicators and
on the origin of genotype, phenotype, and organism. Am. lambda-calculus models of biological organization. Proc.
Zool. 23, 5–14. R. Soc. B 260, 279–286.
Michod, R. 1984 Constraints on adaptation, with special Szathmáry, E. 2000 The evolution of replicators. Phil.
reference to social behaviour. In A new ecology (ed. P. W. Trans. R. Soc. B. 355, 1669–1676. (doi:10.1098/rstb.
Price, C. N. Slobodchikoff & W. S. Gaud), pp. 253–278. 2000.0730)
New York, NY: Wiley. Szathmáry, E. 2002 Units of evolution and units of life. In
Müller, U. F. & Bartel, D. P. 2003 Substrate 2 0 -hydroxyl Fundamentals of life (ed. G. Pályi, L. Zucchi & L. Caglioti),
groups required for ribozyme-catalyzed polymerization. pp. 181–195. Paris, France: Elsevier.
Chem. Biol. 10, 799–806. (doi:10.1016/S1074-5521(03) Szathmáry, E. 2005 Life: in search of the simplest cell. Nature
00171-6) 433, 469–470. (doi:10.1038/433469a)
Ohno, S. 1970 Evolution by gene duplication. Berlin, Germany: Szathmáry, E. & Demeter, L. 1987 Group selection of early
Springer. replicators and the origin of life. J. Theor. Biol. 128,
Oparin, A. I. 1961 Life, its nature, origin and development. New 463–486.
York, NY: Academic Press. Szathmáry, E. & Gladkih, I. 1989 Sub-exponential growth
Orgel, L. E. 1992 Molecular replication. Nature 358, and coexistence of non-enzymatically replicating tem-
203–209. (doi:10.1038/358203a0) plates. J. Theor. Biol. 138, 55–58.
Poole, A., Jeffares, D. & Penny, D. 1999 Early evolution: the Szathmáry, E. & Maynard Smith, J. 1993 The origin of
new kids on the block. BioEssays 21, 880. (doi:10.1002/ chromosomes II. Molecular mechanisms. J. Theor. Biol.
(SICI)1521-1878(199910)21:10!880::AID-BIES11O3. 164, 447–454. (doi:10.1006/jtbi.1993.1166)
0.CO;2-P) Szathmáry, E. & Maynard Smith, J. 1997 From replica-
Poon, A. & Chao, L. 2004 Drift increases the advantage of sex tors to reproducers: the first major transitions leading
in RNA bacteriophage Phi6. Genetics 166, 19–24. (doi:10. to life. J. Theor. Biol. 187, 555–571. (doi:10.1006/jtbi.
1534/genetics.166.1.19) 1996.0389)
Rebek, J. 1994 Synthetic self-replicating molecules. Sci. Am. Szathmáry, E., Kotsis, M. & Scheuring, I. 1988 Limits of
271, 34–40. hyperbolic growth and selection in molecular and
Riley, C. A. & Lehman, N. 2003 Generalized RNA-directed biological populations. In Mathematical ecology (ed. T. G.
recombination of RNA. Chem. Biol. 10, 1233–1243. Hallam, L. Gross & S. Levin), pp. 46–68. Singapore,
(doi:10.1016/j.chembiol.2003.11.015) Singapore: World Scientific.

Phil. Trans. R. Soc. B (2006)

1776 E. Szathmáry Origin of replicators and reproducers

Szostak, J. W., Bartel, D. P. & Luisi, P. L. 2001 Wächtershäuser, G. 1992 Groundworks for an evolutionary
Synthesizing life. Nature 409, 387–390. (doi:10.1038/ biochemistry: the iron–sulphur world. Prog. Biophys. Mol.
35053176) Biol. 58, 85–201. (doi:10.1016/0079-6107(92)90022-X)
Takeuchi, N., Poorthuis, P. H. & Hogeweg, P. 2005 Zielinski, W. S. & Orgel, L. E. 1987 Autocatalytic synthesis of
Phenotypic error threshold; additivity and epistasis in a tetranucleotide analogue. Nature 327, 346–347. (doi:10.
RNA evolution. BMC Evol. Biol. 5, 9. (doi:10.1186/1471- 1038/327346a0)
2148-5-9) Zintzaras, E., Mauro, S. & Szathmáry, E. 2002 Living under
Varga, Z. & Szathmáry, E. 1997 An extremum principle for the challenge of information decay: the stochastic
parabolic competition. Bull. Math. Biol. 59, 1145–1154. corrector model versus hypercycles. J. Theor. Biol. 217,
(doi:10.1016/S0092-8240(97)00048-7) 167–181. (doi:10.1006/jtbi.2002.3026)

Phil. Trans. R. Soc. B (2006)