J. Theoret. Biol.

(1967) 14,74-101

Theory of the Chemostat-Study of the Conditions under

which a Population of Micro-organisms and its Mutant
will Coexist in the Chemostat
C. RENNEBOOG-SQUILBIN~

Laboratoire de Physiologie Animale, Universite’ Libre de Bruxelles, Belgium

(Received 30 July 1966)

When a population of micro-organisms giving rise to non-reverting mutants
is allowed to grow in a chemostat and when the mutation rate is assumed
to be proportional either to the growth rate or to absolute time, three
different final situations will occur. The first one is that both populations
are washed out because the flow rate was badly chosen (too fast). The second
is that the mutant having a selective advantage replaces the normal popula-
tion in the culture vessel. The third possibility is that both populations
grow simultaneously and that between their respective concentrations, a
constant ratio gets established. We have studied this last case in detail
and have shown the importance of the flow rate in the establishment of a
stable ratio.

1. Introduction
When a population of micro-organisms (which will be called “normal”) gives
rise to mutant organisms, two eventualities are to be considered. One is that
the proportion of mutants produced per unit time is proportional to the
specific growth rate of the cells generating them as it is usually admitted
(Braun, 1953), the other is that the proportion of mutant cells appearing per
unit time is constant as it has been established by Novick & Szilard (195Oa.
195 1) in the case of mutation to phage T5 resistance.
The specific growth rate ~1is defined as
1 drr
I-l = n’ht’
where n is the concentration of the cells and t the time (Herbert, Elsworth &
Telling, 1956). As shown by Monod (1942), y is an increasing function of
the concentration of an essential growth rate limiting substrate S, which is
expressed by
M.S
‘l = k+S’
t Aspirant du Fonds National de la Recherche Scientifique.
74
 THEORY OF THE CHEMOSTAT 75
M being the growth rate constant (i.e. the maximum value of p at saturation
levels of substrate) and k a saturation constant which depends on the affinity
of the cells for the substrate. When k is small, the affinity is great. Moser
(1957) has given another form for the dependence of ,u upon the concentration
of a limiting growth factor:
M.SX

where x is a numerical constant (exponent of S). The general conclusions we

shall find in this paper do not depend on the expression we choose as
the unknown S (or S’) will not appear in the equation expressing the ratio
--
m/n as function of the dilution rate D (section 2(B)), provided that the
values of x are identical for both organisms (normal and mutant). Moser
(1957) showed that this assumption was essentially true for Escherichia
coli strain S/l t and its prevalent mutant B/ 1tfwhen grown under tryptophan
control.
We shall study the behavior of a population of microorganisms growing
in a chemostat and giving rise to nonreverting mutants. The chemostat is a
culture vessel whose volume remains constant by means of an overflow and
whose nutrient medium is constantly renewed. This nutrient medium contains
an excess of all growth factors except one, which is limiting. The basic
principles of the chemostat have been given for the first time by Monod (1950)
and Novick & Szilard (1950aJ).
We can ask under what conditions a normal population and a derived
mutant form may coexist in equilibrium in the culture vessel keeping in mind
that we have only the possibility of imposing the flow rate of the nutrient
medium and the initial concentration of the limiting substrate.
The choice of the appropriate flow rate is of the highest importance because
it determines the specific growth rates of both populations and, consequently,
controls the proportion of mutants present at the steady state. We call the
steady state, the state where the concentration of normal cells and the
concentration of mutated cells no longer vary. The fact that a steady state is
possible does not, however, imply that the system will ever reach this state
(Volterra, 1931), but it is possible to demonstrate in this particular case (cf.
Appendix) that when the proportion of mutants approaches the steady state
value, this value will finally be reached.
We were prompted to this study by experiments with Saccharomyces
cerevisiae and its “petite cytoplasmique” mutant. As this mutation is very
frequent (about 0402 mutant per organism and per generation time, Ephrussi,
1951), the accumulation of mutants in the chemostat may occur after a short
time even when the latter have no great selective advantage. We shall first
76 C. RENNEBOOG-SQUILBIN

consider the case which is admitted to apply to this situation, i.e. the mutation
rate is proportional to the growth rate and then more briefly, the case where
the mutation rate is constant per unit time.

2. The Mutation Rate is Proportional to the Growth Rate

(A) EQUATIONS OF THE PHENOMENON

We may write
2 = YP,
where I is the proportion of mutant cells which appears per unit time (time- ‘)
and y a positive constant always inferior to the unity. Let n be the concentra-
tion of normal cells in the culture vessel (cells x ml.-I); m, the concentration
of mutant cells in the culture vessel (cells x ml.-‘); ,uI, the specific growth
rate of the normal population, which is the mean growth rate of the cells of
the population (time-‘); pZ, the specific growth rate of the mutant popula-
tion (time-‘); D, the flow rate divided by the culture volume, which is called
the dilution rate (time-r) (Herbert et al., 1956); S,, the initial concentration
of the limiting substrate in the nutrient medium (moles xml.-‘); S, the
concentration of the limiting substrate in the nutrient medium (moles x ml. - ‘);
l/Y,, the yield constant of the normal population for the limiting substrate
(cells x mole-‘), (Monod (1942) showed that there is a simple relationship
between growth and utilization of substrate dn/dt = - I/ Y ds/dt, where l/ Y
is known as the yield constant); l/ Y,, the yield constant of the mutant popu-
lation for the limiting substrate (cells x mole-‘).
Mathematically, the problem may be stated as follows. During a time
interval At which is supposed very small, the normal population gives rise to
(1 -y)pl n.At
new normal cells per unit volume. During that same time, an amount normal
cells proportional to II, D and At is removed from the culture vessel. The
variation of n is then
AFI = (l-y)pl~~.At--nD.At, (11
The mutant population is increased by division of mutant cells and by
mutation of normal cells. We get
(&n~+y~l~~).At
new mutant cells per unit volume. During that same time, a fraction of the
mutant cells is removed. Finally, the variation of TFZ is
AFIT = (~c,m+~~L1~~).At-~)~D.At. (2)
For the variation of the concentration of the limiting substrate, we know
 THEORY OF THE CHEMOSTAT 77
that during the time interval At and per unit volume, the normal cells are
responsible for
A 1 S=-I’n Ar
Yl .

and the mutant cells are responsible for

A 2 S= -h, At
Y, . *
A concentration S, of limiting substrate enters the culture vessel, while a
concentration 5’ leaves it, hence
A,S = D(S,-S).At.
The total variation of the limiting substrate concentration is thus

AS= D(S,-S)++n
1 2 1.At.

From equations (l), (2) and (3), we deduce the following differential equations
(3)

f dn
dt= Rl-~h-Dl~~ (4)
dm

= +D(S,-S) _ !!;.! _ !!?. (6)

1 y2

The general solution of this system is

We see that the general solution of this system of differential equations is

extremely complicated. Consequently, we shall not study the evolution of
the phenomenon toward the steady state but the conditions which must be
fulfilled by the different factors and especially by the dilution rate D if the
78 C. RENNEBOOG-SQUILBIN

steady state with the two populations present is to be possible. The steady
state will be reached when at the same time
r -=
drz
0
dt

Equations (4), (5) and (6) become

[jY1(l-y)-D]fi = 0
(j2 - D)iG+ yj& ii: = 0
I -- __
D(S,- S) - F;! _ 5; = 0.
I 1 2

The letter with a bar represents the value of the variable at the steady state. The
interesting caseswhere we satisfy conditions (13), (14) and (15)are the following :
--
p2fn-.-.
ii=& j2=D, s=S,-
Y2D
The normal population has been washed out of the culture vessel. Only the
mutant population remains in the chemostat.
--
(ii) (l-y),&=D, D-p,=j’&;, &s,-S-E.
1 2
We shall develop this last case in which the two populations remain in the
chemostat. In the Appendix we shall show that, when n, m and S approach
the steady state values, continuing oscillations about the steady state are
excluded and the steady state will be attained.

(B) STUDY OF THE INFLUENCE OF THE DILUTION RATE D ON THE STEADY STATE
WHENCE # 0~~~16 # 0
Case (ii) gives us the equations

ri
D-J2 =yjilz. (17)

From equation (16) we may define a superior limit for D. We have a maximum
value S, for S. Consequently, the maximum value that pr may reach is
MI.%
plr = 6%’
 THEORY OF THE CHEMOSTAT 79
with M1 as growth rate constant of the normal cells and kr as saturation
constant of the normal cells for the limiting substrate. It follows that the
dilution rate we choose must be inferior to the value
D -M,(~-Y)S,
r- (18)
k,+S, *
--
Equation (17) leads to another limit for D. The term r&(n/m) being
positive, D must be greater than &. (It is recalled that & is positive by
definition.) Let us develop the condition D > &.
Relation (16) becomes
M,.S 1
-_ zz -DD.
k,i-S l-y
We know that
_ M,.S
(20)
” =k,+S
with M2 as growth rate constant of the mutant cells and k, as saturation
constant of the mutant cells for the substrate. Putting the value of 3 derived
from equation (19) in equation (20), we find

It is easy to see that the curve of jiz as function of D is a hyperbola. The

asymptotes

and the slope of the tangent at the origin

define, as shown by Figs 1,2,3 and 4, the conditions which the dilution rate D
must fulfil to be superior to j&.
 80 C. RENNEBOOG-SQUILBIN

r-

FIG. 1. The specific growth rates at the steady state of the normal and the mutant popula-
k,
tions, fil (-) and pa (-), as functions of the dilution rate D when - < 1 and
ka
M1 F < M1(l - y). The intercept of p, and of the dashed line (----) pz = D determines
2
the limiting value D, of the dilution rate,
M+MI(I - :,)
D, _ ..m+-m~- -,
g-1

FIG. 2. The specific growth rates at the steady state of the normal and the mutant popula-
tions, p1 and pz, as functions of the dilution rate D when 2 > 1 and Ma : < Ml(l --I+.
a a
For the signifkation of the curves, see Fig. 1.
 THEORY OF THE CHEMOSTAT 81

0 D

FIG. 3. The specific growth rates at the steady state of the normal and the mutant popula-
tions, ,c, and pz, as functions of the dilution rate D when z < I and Ma : > M, (1 - r).
a a
For the signification of the curves, see Fig. I.

v / //
I
//
//
/ /

b D<P,
0
1 47
D=-l;i2 ,

FIG. 4. The specitic growth rates at the steady state of the normal and the mutant popula-
kl
tions, p1 and pz, as functions of the dilution rate D when -k2>1andM.$-M+~).
For the signification of the curves, see Fig. 1.

T. B. 6
82 C. RENNEBOOG-SQUILBIN

We notice that equation (16) has as a consequence that ,Er is always greater
than ji2 when D > ,ii2.
In the particular case where k,/k, = 1, equation (21) becomes
-~ M2 D
” = M,(l-y)
The inequality D > ,i& is satisfied only when M, < M,(l - y).
We may summarize the data of Figs 1, 2, 3 and 4 as follows : a steady state
with E/E different from infinity is possible only in the three following cases :
(i) when
k,.-~ <1
kz
M,$ < M,(l -y).
2

provided that D < D, and D < D, with

M,; - M,(l -r)

D,= -..2 ~.-
k,
k,-’
(ii) when k
-1 > 1
kz
M2$ < M,(l-y).
2
provided that D < D,;
(iii) when
k, >1
li;
M+ >M,(l -:‘,J,
2
provided that D, < D -c D,.
Putting the value of p1 from equation (16) and of --p2 from equation (21) in
equation (17) we obtain the relation between the ratio m/n and the dilution rate :

Figures 5,6 and 7 represent the variations of Ejfi as function of D in the only
three possible cases.
 THEORY OF THE CHEMOSTAT 83
1
I
I
I
I
I
I
I

I
I
I
I
I
0 4 L

FIG. 5. The proportion of mutants at the steady state, fijfi as function of the dilution rate
D when $1 < 1 and M, L: < M1(l - y). (--) ?%/A; (----) asymptotes which
limit the ra:ge of dilution rates for which the steady state is possible. It is recalled that D
must also be inferior to D, = &Clp--, - Y)S The existence of this second limiting value
kz+S,
depends on the intrinsic constants of the.normal population.

I------

FIG. 6. The proportionof mutants at the steady state, S’z/rias function of the dilution rate
MI (1 - YW, For the signification of the
Dwhen~>IandM2~iM1(l-;~).D~=
2 k,+Sr ’
curves, see Fig. 5.
84 C. RENNEBOOG-SQUILBIN
i-~----.-. ._-
I
/ !
1 ‘1\
ET, ‘,

FIG. 7. The proportion of mutants at the steady state, ti/ti as function of the dilution
rate D when 2 > 1 and M, F > M 1(1 - ;A). D, :- M1$-&!@. For the signification of
2 I T
the curves, see’Fig. 5.

--
We notice that, when k,/k, = 1, the fraction m/n is independent of
D. Any value of dilution rate is thus allowed provided that it is smaller
than

A suitable dilution rate being chosen, let us briefly see the influence of the
other constants (which we cannot change because they represent inherent
properties of the considered micro-organism) on the value reached by the
ratio E/E at the steady state. It is easy to demonstrate from equation (22)
- -.
that in the three cases cited above, m/n is an increasing function of y between
the y values zero and 1. Taking the derivatives of iii/@ as function of k,/kz, we
find that it is always positive and thus, the greater is k,/k, (i.e. the lower is
the affinity of the normal population for the limiting substrate in respect to
the one of the mutant population), the greater is the ratio FE/R at the steady
state, all the other constants remaining the same. In the same manner,
the derivatives of G/E as function of M, and as function of Mz show us
that the proportion of mutants increases with the growth rate constant of
the normal population and decreases when the growth rate constant of
the mutant population increases. These results were of course logically
expected.
 THEORY OF THE CHEMOSTAT 85

(C) NUMERICAL EXAMPLES

We shall illustrate those theoretical data by some numerical examples. In

all those examples, we have deliberately chosen an abnormally high mutation
rate. Notwithstanding this, conditions of culture can still be found compatible
--
with a ratio m/n different from infinity.
In Fig. 8, we have chosen very close values for the growth rate constants of
both populations (M, = 0+3/hr, M, = 0+294/hr) and we have taken different

FIG. 8. The proportion of mutants at the steady state, rTi/iias function of the dilution rate
D for different kl/k2 ratios. S, = 40.10-Bmoles/l., y = lOma, kl = 10-Bmoles/l.,
MI = 0.3/hr, Ma = 0.294/hr, D, = 0.290/hr. For curves (l), (2), (3), (4), (5), (6), (7), (8)
and (9), the ratios kI/kz are respectively 0.1, 0.5, 0.91, 0.99, I, lW1, 1.0102, 1.1 and 10.

k,/k,. When the normal population has a very low affinity for the limiting
substrate compared with those of the mutant population (kl/k2 S l), the
dilution rate must be very large if both populations are to coexist, but this
condition is often incompatible with its superior limit (0, = 0*29/hr). When
on the contrary, the normal population has a very high affinity (kl/k2 G l),
--
it is logical that the fraction m/n remains very small. For values of kl/kz close
to 1, appreciable quantities of both populations will be present at the steady
state. When k,/kd is slightly greater than 1, very slow dilution rates give high
X6 C. RENNEBOOG-SQUILBIK

levels of mutants while when k,jk, is slightly smaller than I. they induce a
ratio Z/E smaller than 1, but in both casesdilution rates near the superior
limit D, lead to a proportion of mutants of about 1.
If the growth rate constant Mz of the mutant population is much smaller
--
than that of the normal population, the ratio of mutant cells m/n at the steady
state will be very small except for large values of k,/k, and within a limited
range of dilution rates (Fig. 9).

FIG. 9. The proportion of mutants at the steady state, N2jfi as function of the dilution
rate D for different kl/ka ratios. Same S,, y, k,, MI and D, as in Fig. 8. M, = 0,03/hr. For
curves (1) and (2), the ratios kI/kz are respectively 10 and 100.

If the growth rate constant M, of the mutant population is much higher

than that of the normal population, washing out of the normal population
will result except for very low values of k,/k, (Fig. 10).
The influence of M, on the physionomy of the curves expressing iir/ii as
function of D appears clearly by comparing Figs 8, 9 and 10.
We shall now seein more detail in Fig. 11, the effect of the constants in the
only interesting case,i.e. when Mz approachesM, and when k,/k, hasa value
close to 1. The comparisons betweencurves (1) and (2) between curves (3) and
(4) or between curves (6) and (7) shows the influence of slight differences
 !I
I

FIG. 10. The proportion of mutants at the steady state, fi/ti as function of the dilution
rate D for different kl/kz ratios. Same S,, y, k,, MI and D, as in Fig. 8. Mz = 3/hr. For
curves (l), (2), (3) and (4) the ratios kI/kz are respectively 10e6, 10-a, 0.09 and 0.098.

FIG. 11. The proportion of mutants at the steady state, rii/ri as function of the dilution
rate D. Same S,, y and kI as in Fig. 8. The other characteristics of the curves are given in
Table 1.
 C. RENNEBOOG-SQUILBIN

TABLE 1
Characterisks of the curves of‘ Fig. I1

ML! LA
Curve k,/k,
(hr-I) (hr- ‘)

I 0.3 0.2967 0.9; 0.290

2 0.3 0.294 0.99 0.290
3 0.3 0.2967 I.0102 0.290
4 0.3 0.294 I.0102 0.290
5 1 .5 I.47 1.0102 I.45
6 0.3 0.3003 0.91 0.290
7 0.3 0.2967 0.91 0.290
8 0.3 0.2967 I 0.290

between the M,. The way of the shift of the curve is that we expect logically.
For curve (3) we see that dilution rates smaller than 0.2667 result in the wash-
ing out of all normal cells. The effect of k,/k, is made evident by curves (1).
(3) and (7), or (2) and (4). In curve (5), M, and M, are respectively fivefold
greater than in curve (4). We see that as long as the difference between
M,(l-y) and M2(kl/k2) is small compared with D(k,/k, - l), the ratio
E/E is fivefold greater in curve (5) than in curve (4).

(D) CONCLUSION

The conclusion of this study is that the choice of the dilution rate following
the particularities (growth rate constant, affinity for the limiting substrate,
mutation rate) of the two populations present is of the greatest importance
when their affinity for the limiting substrate is not the same. On the contrary,
the dilution rate (provided it is inferior to the maximum rate) has no effect
when the affinities of both populations for the substrate are identical.

3. The Mutation Rate I is Constant per Unit Time

(A) EQUATIONS OF THE PHENOMENON

The equations of the phenomenon are established in the same manner as in

section 2. They become
r dil
/ 5 = (p1-a-D,J7

dm
- = (/A2 - D)m + in
dt
 THEORY OF THE CHEMOSTAT 89
At the steady state,
(Jil -A-D)ii = 0 (26)
I (,E,-D)Ei+Aii = 0 (27)

As in section 2, the interesting cases where we satisfy conditions (26), (27)

and (28) are the following:
--
P2 J'l
Jl=o, ji2= D, s=s,----
Y,D'
Only the mutant population remains in the chemostat.

(ii)

Here, the number of normal and mutant cells differs from zero. We shall
develop this case.

(B) STUDY OF THE INFLUENCE OF THE DILUTION RATE ON THE STEADY

when fi # 0 AND
STATE 15i # 0
The equations we shall study are thus
D=&-A (29)

Equation (29) gives an immediate condition for the dilution rate. The
substrate concentration being inferior to S,., ,iI1 will never exceed
MI .S,
k,
so that D has a maximum value

D, = MI .Sr
k,+s,-l-
For reasons of convenience, we shall not express the value of Ci/ti as func-
tion of D, as in section 2, but the value of ii/Z as function of D/L
In the particular case of kl/kz = 1, Z/Kz as function of D/A is a straight line

rij??i is positive provided that M2 < M, and D//z > M,/(M, -MJ.
T.B. 7
90 C. RENNEBOOG-SQUILBIN

If we say

we find from equation (29) and (30)

ii
--=X-y---. L2
Ei 1+ 3. (31)
x+1
- -.
We immediately see that n/m IS the difference between a straight line whose
equation is a = x and a hyperbola which corresponds to fi,/A and whose
equation is
y = -LL.
1 +;$!j

The graphic representation of such functions is very easy.

The asymptotes
y = L,
and x = - L, - 1 define the conditions which the dilution rate must satisfy in
order that the ratio la/k is positive at the steady state.
There exist four cases, illustrated in Figs 12, 13, 14 and 1.5.

0 :j A
FIG. 12. The specific growth rates at the steady state of the normal and the mutant
populations expressed as &/A and izz/J. as functions of the dilution rate expressed as
x = D/Awhen k,/kz P 1, The proportion of normal cells at the steady state ii/t% (curve (3))
is directly obtained by subtracting curve (1): y = &/A from the dashed line )I = D/j..
The solid straight line (2) represents PI/l.
 THEORY OF THE CHEMOSTAT 91

FIG. 13. The speciiic growth rates at the steady state of the normal and the mutant popula-
tions expressed as p,/A and Q,I as functions of the dilution rate expressed as x = D/l when
kJkz < 1 and IL,I < 1. For the signification of the curves, see Fig. 12.

FIG. 14. The specific growth rates at the steady state of the normal and the mutant popula-
tions expressed as pJ,I and ii,/1 as functions of the dilution rate expressed as x = D/Awhen
kl/kl < 1, jLll > 1 and (1 +L1 --L,)“-I-~LL~ > 0. For the signification of the curves,
see Fig. 12.
92 C. RENNEBOOG-SQUILBIK

0 ‘:

FIG. 15. The specific growth rates at the steady state of the normal and the mutant popula-
tions expressed as &/,I and bz/A as functions of the dilution rate expressed as .Y = D/A when
kl/kz < 1, ]Lll > 1 and (l+L,-LL,)2+4Lz < 0. For the signification of the curves,
see Fig. 12.

--
Since n/m must be positive, the only dilution rates allowed are those at
which y is smaller than X. For that reason, the case of Fig. 15 is excluded
while the case of Fig. 13 must be rejected because y (i.e. &/A) is negative.
In Fig. 12, x must be superior to x1 and in Fig. 14, x must be comprised
between x1 and x2. x1 and x2 are defined as the solutions to the equation

x2 = ---(1+L1-L,)-J(1+L,-LJ2+4L2 ____-
2
It must be noticed that at the steady state the specific growth rate ,Ei of the
normal population necessarily exceeds the specific growth rate fiZ of the
mutant population when D > iI2 (cf. equation (29)).
The situation is summarized in the following manner. When the affinity of
the normal population is smaller than that of the mutant population (Fig. 12),
--
the fraction n/m will increase with the dilution rate provided that this latter
does not exceed the maximum value allowed by the conditions of the experi-
ment. This is comprehensible because the difference between x and )? (i.e.
 THEORY OF THE CHEMOSTAT 93
also between jil and j&) increases with the dilution rate. The difference
between Figs 14 and 15 is less immediate. It lies only in the sign of the func-
tion F = (1 +Li -L2)2 +4L,, which is responsible for the existence or non-
existence of intercepts between the hyperbola y and the straight line 01.
(i) Function F may be expressed as function of k1/k2

The denominator being always positive except when k,/k, = 1, we are only
interested in the numerator N which is the equation of a parabola. Its
derivative N’ is first negative and then positive. That means that the concavity
of the parabola is directed towards the top and that function F(kl/k2) is
either always positive or negative for k1/k2 values comprised between
(k,/k,), and (kl/k2)2 and positive for k,/k, values outside those limits,
(k,/k,), and (k,/k2)2 being the roots of N.

The influence of M,, M2 or y on F, all the other terms remaining respec-

tively constant, may be studied in the same manner as the k,/k, influence.
We find that M1 values must be outside M1, and M12, with

M,,=M2$($1)2+2/~-l)i~M2,

M,,=M,$($+,/-~:)~
94 C. RENNEBOOG-SQUILBIN

M, values must be outside M,, and MZ2, with

kz
and y values must be outside yi and y2, with
.---
42M,M2
__.----2
k1
-- 1
kz

4;M,M,
1

in order that the steady state with n/m different from zero may exist.
In other terms, certain values only of the intrinsic constants of the popu-
lations are compatible with the proposed state.

(C) NUMERICAL EXAMPLES

Example 1
M, = l+/hr, M, = 1*47/hr, A = O*Ol/hr. The roots (k,/k,), and (kl/k2)2
are imaginary numbers, hence F(k,/k,) is positive for any k,/k, (cf. Fig. 14).
In Table 2 we find the abscissae of the intercepts of y and CI for different
k,/k,. It is recalled that the dilution rate must be inferior to 1*45341/hr (or
x < 145.341) if we choose S, = 40.10m6 moles/l. and k, = low6 moles/l.
and that the permitted values of x are comprised between x1 and x2 when
k,/k2 is smaller than 1, or are greater than xi when k,/k, is greater than 1.
We see that when k,/k, values reach the neighborhood of 10, x1 becomes
greater than 145.341. Hence we can say if the affinity of the mutant popula-
tion is tenfold higher than that of the normal population, no dilution rates
are compatible with an equilibrium with both populations in presence.
--
Figure 16 shows the ratio n/m as function of D for some k,/k, values of
Table 2.
 TABLE! 2

Abscissae x, and x2 of the intercepts of the straight line u and the hyperbola y for different values of k,/k,. Note: when
k,/k, tends to zero, x, tends to ((MI/A) - I) and x2 to zero. When k,lk2 tends to (+ oo), x1 tends to M2/1 and x2 to
(-1). When k,/k, increases to I, x, tends to (+ CO) and x2 to M,/(M, - M,), while when k,/k, decreases to I, x1 z
tends to M2/(MI - M,) and x2 to (- CO)
:
‘x
k&z 0 0~001 0.1 0.9 0.99 0.999 1 1 aoo1 1.01 1.1 10 1000 to *
Xl 149 149W2 149.2 168 411 3099 + co 49 49 67 128.6 146.8 146.99 147 $
X2 0 O*OOl 0.109 7.9 35.4 47.4 49 - c~3 -29903 -221 - 12.6 -1.11 -1+4M -1 1
2

:
TABLE 3 E
Abscissae x, and x2 of the intercepts of the straight line OLand the hyperbola y for d@erent values of k,/k, 8
9”
k&z 0 0.1 0.5 0.9 0.914 (kJk,), 0.99 (k,/k& 0.999 1 1 +OOl 1.01 1.1 10 1000 03 +I
Xl 29 28.95 28.6 22.79 18.33 17.7 - 80.28 577.54 + co 49 49.2 40.94 32.39 2944 29.4004 29.4
x2 0 0.11 1.025 11.61 17 17.7 - 80.28 50.85 49 - co -59756 -72.54 -9.98 -1.11 -IQ01 -1
96 C. RENNEBOOG-SQUILBIN

FIG. 16. The proportion of normal cells at the steady state, ti/tii as function of the dilution
rate expressed as x = D/1.. S, = 4O.1O-6 moles/l., 1. =: 10e2/hr. k, =: lo-” moles/l..
M1 = 1.5/hr, M, = 1.47/hr. For curves (l), (2), (3) and (4), the ratios k,/kz are respec-
tively 0.1, 0.9, 1.01 and 1.1.

To illustrate which conditions of M,, M, and y must be satisfied for the

given value k,/k, = O-1, we shall search the limits outside of which one
must stay in order to have the studied steady state. These limits are
M1, = 90*8/hr and M,, = 89*4/hr, M,, = 91*6/hr and M,, = 91*4/hr,
Yi = 0.78 and yZ = 2.88. The given values M, = 1..5/hr, M, = 1*47/hr and
y = 0.01 are outside those limits as expected.

Example 2
M1 = O-3/hr, M, = 0*294/hr, I = O-Ol/hr. Those values lead to a
(k,/k,), = 0.99546 and a (kl/kJ2 = 0.91416. Inside this interval, F(kl/kZ)
being negative (cf. Fig. 15), the steady state with both populations in the
culture vesseldoes not exist. Outside this interval, as the maximum dilution
is 0*283/hr (or x,,, = 28.3), the steady state will only be attained for k,/k,
values smaller than (k,/k,) = O-91416. In Table 3 llustrated by Fig. 17 for
some k,/k,, we find the abscissaeof the intercepts of y and CIfor different
k,/kz when these intercepts exist.

4. General Conclusion
To summarize the influence of the dilution rate D on the proportion of
mutants at the steady state, it must first be recalled that the specific growth
rates of the normal and of the mutant populations, ,Er and ,C2are increasing
 THEORY OF THE CHEMOSTAT 97

20/---

x
FIG. 17. The proportion of normal cells at the steady state, A/E as function of the dilution
rate expressed as x = D/l. SameS,, 1, k, as in Fig. 16. M, = 0.3, M, = 0.294. For curves
(I), (2) and (3), the ratios kl/kz are respectively 0.1, 0.5 and 0.9.

functions of D and that obviously jiI must always be superior to /Yi2if the
normal population is to coexist with the mutant population.
When the mutation rate per unit time L is constant (section 3), the propor-
--.
tion of normal organisms at the steady state (n/m) mcreases or decreases with
the difference A,u = ,i& -j& (cf. Figs 12 and 14).
When 1 is proportional to the growth rate (section 2), two factors interact
to influence the proportion of mutants at the steady state when the dilution
rate D increases. These are Ap, as above, but also /I, which is proportional to
p1 and hence to D. This latter fact explains why, in Fig. 5, for low dilution
rates E/E decreases (the inverse Z/E increases) though Ap increases (cf.
Fig. 1). The most striking effect of the interaction between 2 and AP is that
Z/C is independent of the dilution rate when the affinities of both populations
for the limiting substrate are the same (kl/k2 = I), ~1and /1 being in this case
both proportional to D.
A last thing to observe is that only the ratio k,/k, of the saturation con-
stants occurs in the equations. It follows that the absolute value of k, does
not matter while the absolute value of k, sets a limit to the maximum dilution
rate allowed.
We may say in all generality that some pairs of microorganisms formed by
a given population and a mutant derived from it will never coexist in the
chemostat because of the values of the intrinsic constants which characterize
them, while other pairs will coexist within a limited range of dilution rates.

We thank Dr R. Hamers for his helpful advice and we are very grateful to
Dr J Renneboog for his mathematical help and his encouragement.
98 C. RENNEBOOG-SQUILBIN

REFERENCES
BRAUN, W. (1953). “Bacterial Genetics”. Philadelphia, London: W. B. Saunders Company.
EPHRUSSI. B. (1951). In “Genetics in the Twentieth Centurv”. Dunn. New York: The
Macmillan Company.
HERBERT, D., ELSWORTH, R. & TELLING, R. C. (1956). J. grn. hficrobiol. 14, 601.
MONOD, J. (1942). “La croissance des cultures bactkriennes”. Paris: Hermann et Cie.
MONOD, J. (1950). Ann. Inst. Pasteur, 79, 390.
MOSER, H. (1957). Cold Spring Harb. Sytnp. quant. Biol. 22, 121.
NO~KK, A. & SZILARD, L. (195Oa). hoc. natn. Acad. Sri. U.S.A. 36, 708.
NOVICK. A. & SZILARD. L. (1950b). Science. N.Y. 112. 715.
NOWZK; A. & SZKARD; L. <1951): Cold Spthg Horb. ‘Syn7p. quant. Biol. 16, 337.
VOLTERRA, V. (1931). “Lqons sur la theorie mathematique de la lutte pour la vie”. Paris:
Gauthier-Villars.

Appendix-Demonstration of the Stability of the Steady State

(A) CONDITIONS OF STABILITY OF THE STEADY STATE
Near the steady state, we have
An= n-ii, with An G ii;
Am = no- iG, with Anz < ni;
AS = S-S, with AS < S.
The equations (4), (5) and (6) may be written

if =fl(S,n)
; !f =f,(S,II,tn)

i -$ =fAS,t1,m>,
since [(l - y),ul - D] n is a function of S, and n, (p2 - D)m+ ypl n and

D(S,- S) - $ n - F Ii1
1 2
are functions of S, n and tn.
We know that
r dn dAi1
dt dt
!4 dtn dAt7t
~..~~_
= __
~ dt dt
 THEORY OF THE CHEMOSTAT 99
Considering that An, Am and AS are very small respectively compared with
fi, Z and S, the Taylor’s series corresponding to fi(S, n), fi(S, n, m) and
,&(S, n, m) may be limited to the terms of the first degree. Hence
dAn
dt ~j-~(S,ii) + @An + @AS (32)

dm
dt =jl(S, E, E) +

dS
dt Nfl(S,ti,%) + e)An + ($)A, + g)A& (34)
i
where

represent respectively the values at the steady state of the partial differentials
of the functionsfr, f2 and& with respect to n, m and S.
If we say

c=ji2-D

d=.,(g)+YR(!$

g-“;
2

h =- D

where

($$) and ($$)

represent respectively the values at the steady state of the derivatives of the
 100 C. RENNEBOOG-SQUILBIN

functions pI and ,uZwith respect to S, the equations (32), (33) and (34) become
f dAn
(35)
I dAm
IryaAS
1 dr 2: bbrr + cAm + dAS (36)

(37)

The resolution of this system of three homogeneous linear differential

equations with constant coefficients can classically be reduced to the resolution
of one differential equation of the third order with constant coefficients. By
eliminating Am and AS, we obtain the following equation for An:
d3An
-- _ (c + h) ::T + (hc - dg - aj) f$i + a(fc - bg)Arl N 0. (38)
dt3
In order that An = eXt is a solution of the equation (38), x must be a root of
the equation
lx3 - (c + h)%2 $ (hc -dg-CrJ‘)cc+a(fc-hg)=O.
The Routh-Hurwitz criterion defines the following conditions of stability:
- (c + h) > 0 (39)
a(fc - by) > 0 (40)
a(bg +Jk) - (c + h)(ch - dg) > 0. (41)
(B) ANALYSIS OF THE THREE CONDITIONS OF STABILITY
- -

(9 - (42)

Seeing that -
($!) and (i%)
are respectively-equal
. to

which are always positive (because x > 0, S > 0, jI, or ji2 > 0, PI or
j& < M, or M,), the expression (42) is positive if D is superior to ,ii2, a
condition that we have established in section 2(B).
G-l
(ii) a(fc - bg) = (1 - y)E as (D - &)!$ -t ypl F . (43
( >[ 1 -12
 THEORY OF THE CHEMOSTAT 101
For the same reasons, the expression (43) is always positive.
(iii) That the inequality (41) is satisfied when l/Y, > r/Y, results from
the fact that then (bg+fh) is positive and the second term (c+h)(ch-dg) is
negative.
When l/Y, < y/Y,, the verification of the inequality (41) implies rather
laborious calculations. After some transformations, the inequality (41) is
equivalent to

-ti /* ;uJ - P2) -t- (& - :-> P2P - P2)] (44)

( as’>[ 1 2 1
which is always verified for (D > ji2).

(C) CONCLUSION
The three conditions of stability being satisfied, we may say that the
considered system (i.e. two populations of microorganisms, one of which
derives from the other), when near the steady state, will tend to approach it
more and more. Thus oscillations about the steady state without reaching it
are excluded.