BIOTECHNOLOGY AND BIOENGINEERING,

VOL. XVII, PAGES 1301-1312 (1975)

A Model for Diauxic Growth

G. VAN DEDEhI and M . MOO-YOUNG,* Department of Chemical

Engineering, Biochemical and Food Engineering Group, University of
Waterloo, Waterloo, Ontario, Canada

Summary
A model is developed to predict batch and continuous culture behavior of
fermentations on two different carbon and energy sources. The basic assumption
of the model is that the permease for the favored substrate is constitutive, whereas
the permease for the second substrate is subject to induction and catabolite
repression. Simulations of the model show features of diauxic growth described
in the literature.

INTRODUCTION
The phenomenon of diauxic growth, the sequential utilization of
carbon sources in a batch culture by microorganisms has been de-
scribed by Monod.' A phenomenological model was described by
Kono and Asai;2 recently, a detailed study of growth on two different
carbon sources in batch and continuous culture, specifically the ap-
pearance of transients after feed rate shift-ups and feed composition
changes, was reported by Standing e t aL3 This phenomenon is im-
portant in biological waste treatment where the biological oxygen
demand (BOD) may consist of complex mixtures of carbon sources.
A better understanding of diauxic growth is also required for the
analysis of microbial degradation of insoluble or polymeric substrates
where the rate of the extracellular enzyme synthesis often depends
on the concentration of the oligomers that are the products of the
enzymatic breakdown of the polymeric material. The extracellular
cellulase complex of the mold Trichoderma viride is induced b y the
disaccharide sophorose and strongly repressed b y g l u ~ o s e . ~Sophor-
ose may be a transglucosidation product in the enzymatic breakdown
of c e l l u l o ~ ewhereas
,~ the major breakdown products of cellulose by
T . viride cellulase are glucose and cellobiose. Cellulase activity is
only released after cell growth has virtually stopped. Lactose also
* To whom correspondence should be sent.
1301
@ 1975 by John Wiley & Sons, Inc.
1302 VAN DEDEM AND MOO-YOUNG

induces cellulase activity in T . ~ i r i d e . ~When T . viride is grown on

a mixed carbon source of glucose and lactose, glucose is consumed
first and lactose subsequently. Only during the latter period is
cellulase activity induced and expressed.
The extracellular lipase of Candida lipolytica appears to be induced
by fatty acids.6 When the yeast is grown on glucose as the only
carbon source, less lipase activity is induced. When grown on a
mixed carbon source of oleic acid and glucose, a typical diauxic pattern
is observed, favoring the fatty acid. Lipase activity is most rapidly
expressed during the transition period.
I n the following, we will attempt to quantitatively describe the
diauxic growth phenomenon, using the concepts of induction and
repression of enzyme synthesis applied to the permease of the second
substrate. The results of computer simulations will be presented and
discussed.

THE MODEL
The model is based on the following assumptions. 1) The permease
for the favored carbon source is constitutive. 2) The permease of the
second carbon source is subject to induction by this substrate and to
catabolite repression. 3) The product of the two permeases is a
common intermediate, the flux into the cells of which is controlled by
the formation rate of the permeases, and thus by the concentration of
either substrate and by the specific activity of the inducible permease.
4) The common intermediate is the substrate of a branched pathway,
leading t o either cell mass or a high energy intermediate (e.g., ATP).
5) The two enzymes controlling the branched pathway are allosteric
and under heterotropic control by ATP. 6) Induction of the in-
ducible permease is dependent on the rate of uptake of the second
substrate according to the expression derived earlier. 7) Catabolite
repression of the expression of the inducible permease is dependent
on the intracellular ATP concentration according to a n inhibition
function.
A schematic outline of the model is presented in Figure 1. The
intracellular components are designated as F, the common inter-
mediate, A , the high energy compound, El,the constitutive permease
for substrate 1, Ez, the inducible permease for substrate 2 , R , the
messenger RNA for EZ. The bulk constituents are the following:
S1,the favored substrate; Sz, the second substrate; X , the cell mass.
The rate of change of ( F ) is given by
dFldt = k17r1(S~,EI) + hwZ(S2, - ~ I S T A~ )F-, k16r4(F, A )
( 1)
 A MODEL FOR DIAUXIC GROWTH 1303

I
I
I I
ENZYME I I
SY NTHESIS(E2) I

Fig. 1. Schematic representation of the model.

The rate terms r1 and r2 are functions of S1,El and S2,E2, respec-
tively, as indicated. It is assumed here that the permeases follow
simple Michaelis-Menten kinetics, so that rl and r 2 are given by

The two key enzymes E4 and E3 regulate the magnitudes of the energy
and cell mass generating fluxes, respectively. ,To avoid accumulation
of either intermediate, F, or high energy compound, A , the two
enzymes have to be finely attuned to each other. I n this case, this
is thought to be brought about by allosteric control of enzymes E3
and E 4 through compound A . Accumulation of A will thus cause
inhibition of enzyme Eq through a negative heterotropic interaction,
and stimulation of E 3 through a positive heterotropic reaction.
Monod, Wyman, and Changeux12 derived the following expressions
for positive [eq. (4)] and negative [eq. ( 5 ) ] heterotropic interactions :
1304 VAN DEDEM AND MOO-YOUNG

If we designate b y R , the concentration of messenger RNA (mRNA)

specific for the inducible permease Ez, we may describe its rate of
change as9:
dR/dt = klk:5r3Q - k z R (6)
where Q is the fraction of free operators on the operon specific for
E z and ki5r3is the specific growth rate term on which the rate of RNA
synthesis depends.
I n a previous article, the authors assumed proportionality between
the intracellular and the medium inducer concentration. Recent
studies on the nature of the natural inducer of the lac-operon in
E. coli by Jobe and B o ~ r g e o i shave
' ~ shown that it is a by-product of
lactose permeation, i.e., allolactose (1-6-0-@-~- galactopyranosyl-D-
glucose). It seemed therefore a better assumption t o make the intra-
cellular inducer concentration proportional to the rate of uptake of
the substrate of the inducible enzyme, i.e., to r 2 . By analogy to the
derivation of the relation between the inducer concentration and the
fraction of free operators (Q) given in our previous deduction19we
obtain:

The rate of Ez-activity depends on the extent of catabolite re-

pression, as given by Qz:

dE2/dt = k24QzR (8)

where Qz is given as:

The inhibition function [eq. (9)] is arbitrary and must be seen as a

short-cut approximation as long as the mechanism of catabolite re-
pression is insufficiently understood.
The rate of change of A depends on the rates of both E3 and Eq:
E4 is assumed to be the rate limiting enzyme in the formation of A ;
the rate of E3 determines the biosynthetic rate and thus the rate of
consumption of A . For a large number of microorganisms, there is
a n almost fixed stoichiometric ratio between the cell mass produced
per number of moles of ATP consumed of about 10 g dry wt/rnole.'O
Therefore,
dA/dt = k6r4(F,A ) - k5r3(F,A ) (10)
 A MODEL FOR DIAUXIC GROWTH 1305

The rate of bulk cell growth depends on the flux of material into cell
mass and on the bulk cell concentration X :
dX/dt = k:5r3(F, A ) X (11)
So far, all concentrations have been expressed in mol/liter of cell
volume, except S1 and Sz, which are expressed in g/liter or any other
convenient dimension, provided ksl and ks2 are expressed in the same
units. To obtain the total quantity El or E 2 per liter of culture
medium, one has to know the cell volume per liter of broth, which
equals X (expressed in g/liter) times the specific cell volume V ,
(liter/g). The expressions for the rate of change of the bulk medium
concentrations of S1 and S z therefore become:

dSl/dt = k;7r1(S1, E l ) X V , (12)

dS2/dt = ki8rZ(S2,
E2)XV, (13)
I n chemostat culture, eqs. (1 I)-( 13) take the form :
dX/dt = - DX + k L r 3 ( F 1A)X (14)
dSl/dt = D(S1,o - Si) - k:’r~i(Si,
Ei)XV, (15)
dSz/dt = D(S2,o - Sz) - k:a~z(Sz, E z ) X V , (16)
The coefficients ki7 and k & are combined rate constants and con-
version factors and have the dimension g/mol hr.
T o convert the above equations into dimensionless forms, the fol-
lowing quantities were assumed as reference quantities: Sl,o Szvo+
for all quantities relating to bulk medium substrate concentrations;
V , for the bulk cell concentration; k&, for all rate constants; El for all
intracellular concentrations. We now obtain the following dimen-
sionless expressions :
 1306 VAN DEDEM AND MOO-YOUNG

x = xv,

e = kist

EZ
tl=-
Ei

Thus, the following equations are obtained:

Throughout the calculations, the following parameter values were

assumed :
K17/KF = KM/KF = 2.5 X 104; K15 = 1.4 x 104; K16 = 4 x 104;
K5 = 7 x 10’; K6 = 2 x lo4; K2 = 1.0; Kz4 = 1.0;
Ki7 = Ki8 = 1.0; K1 = 1.0
 A MODEL FOR DIAUXIC GROWTH 1307

An unequivocal justification for these parameter values cannot be

given, but the parameters were chosen based on reasonable estimates
of a number of quantities. K17 and K 1 8 , for example, were chosen on
the basis of the magnitude of substrate influx into 1 g/liter of cells,
growing a t a maximum specific growth rate of 0.25 hr-' on a carbo-
hydrate carbon source. The influx of substrate is then approxi-
mately 0.5 g/liter hr, which equals about 3 mol/liter cell vol hr. An
estimate for the concentration of El was taken to be 5 X lop4mol/
liter cell vol. Therefore, K 1 7 becomes 0.6 X lo4 hr-l, K ; S equals 0.25
hr-' (because it is equivalent to the maximum specific growth rate),
so that K17 = 2.5 X lo4. k F may be taken as equal to El, so that we
obtain the value of K 1 7 / K F mentioned above.
It has t o be pointed out that the enzymes E3 and E4do not represent
real enzymes (although of course similar branch points abound in
biological systems), but the overall processes of cell synthesis and
energy generation. Intracellular levels of intermediates such as
glucose-6-phosphate and ATP are usually of the order of magnitude
of lop3 mol/liter cell v01i4J5and do not change dramatically in the
course of a growth cycle. Since the flux of substrate is of the order of
magnitude of a t least 1 mol/liter cell vol hr, there is an extremely
rapid material flux through a very small pool. This causes the differ-
ential eqs. (17) and (18) to become very stiff and therefore difficult
to solve by straightforward explicit methods. As a first approach,
however, a steady state condition was assumed, setting the two
right-hand side expressions equal t o zero and in effect solving a
system of two nonlinear algebraic equations. After solving for the
remainder of the system [eqs. (19)-(23)], the new values of ul,u z ,and
7 are substituted and the system is solved for the next time value.
I n Figure 2 a batch growth curve is represented for growth on two
carbon and energy sources. The typical diauxic pattern is obtained,
which is very clearly shown in the substrate depletion curves.
Growth on the second substrate alone is shown in Figure 3 for a n
initial concentration of r) of 0.1. A pronounced lag phase is observed;
for comparison a batch growth curve on u1 alone is drawn in the same
figure. Transients in continuous culture, as described by Standing
e t aLJ3were simulated by iterating a t different values of D',starting
with the steady state value of D' = 0.04 as the initial condition.
The results are shown in Figure 4 for the change in the concentration
of the second substrate. I n agreement with the experimental results
of Standing et al.,3 the peak value of u2 becomes higher when the step
in dilution rate is increased.
1308 VAN D E D E M AND MOO-YOUNG

1
9
Fig. 2 . Batch growth curve for two carbon and energy sources. Initial values
are: z = 0.001; U , = 0.5; u2 = 0.5; p = 0.001; 7 = 0.1.

A step change from u1 to u2 was also simulated (Fig. 5). The

transient increase in u2 and decrease in x,that were found by Standing
e t al.,3 is reproduced a t least qualitatively.

DISCUSSION
As pointed out before, there is some difficulty in obtaining inde-
pendent estimates of the parameters in the model presented here.
The main uncertainty of this model is caused by the rather unsatis-
factory way in which the pnenomenon of catabolite repression has
previously been modeled. Although it has been shown for the lac-
operon of E. coli that cyclic AMP (c-AMP) cancels the inhibitory
effect of glucose on its transcription, the mechanism that regulates the
intracellular level of c-AMP is as yet unknown.16 The role of catabo-
lite repression in the diauxy phenomenon seems a necessary postulate,
for otherwise the excess of the second substrate would induce enough
of its permease in a short time in batch culture. This is not observed
in practice. Also, there is usually a fixed sequence in which carbon
sources are consumed, the pattern being that the carbon source that
gives rise t o the highest specific growth rate is consumed first. This
is also implicit in our model since the second substrate will allow
lower ATP levels than the preferred one, leading to a lower specific
growth rate.
 A MODEL FOR DIAUXIC GROWTH 1309

" O c

5 10 15 20
e
Fig. 3. Batch growth curve for S, only (left-hand curve) and for Sz only.
Initial values are: ul = 1.0; uz = 1.0; other initial values are the same as specified
for Fig. 2.
I

Fig. 4. Transient responses of a chemostat culture growing on S1and Sz after

step changes in D'. Tbe step changes in D' are from bottom to top in uz param-
eter: from 0.04 to 0.2, 0.24, and 0.36. The u1 concentrations (not shown) are one
order of magnitude less than uZ and change smoothly to the new steady state
values without reaching extrema.
1310 VAN DEDEM AND MOO-YOUNG

a2

0.05

0
0 5 10 15
9
Fig. 5 . Transient response of a chemostat culture after a shift in substrate from
only S1 to only S2. The dimensionless dilution rate D' = 0.2.

No attempt has been made t o exactly match the data of Standing

e t aL3 Nevertheless, our simulation shows virtually all the features
found in that experimental work. The batch curve shows the typical
diauxic pattern and sequential substrate utilization. The transient
response of a chemostat in which cells are grown on X1 after a step
change to only Sz is fairly well reproduced. Also, transients after
step changes in the dilution rate of a chemostat culture growing on
mixed substrates show most of the features found experimentally,
with the exception of the dip in the cell mass (which in our simulation
is only significant when large shifts-up in the dilution rate are applied).
An additional feature of our model is the prediction of a lag phase
when X2 is the only carbon source and the cells have not been pre-
viously adapted. Indeed this is often found, e.g., when bacteria,
previously grown on glucose, are adapted to phenol as the only
carbon source.ll
I n our present model, we have succeeded in simulating the features
of experimental observations on diauxic growth. This gives us con-
fidence in the usefulness of the model for the simulation of the more
complex case of microorganisms growing on polymeric and/or water-
insoluble substrates. The practicality of the model would be im-
proved if a better understanding was available about the relationship
between the extent of catabolite repression and measurable external
 A MODEL FOR DIAUXIC GROWTH 1311

conditions, such as substrate concentration and composition, specific

growth rate, etc. The usefulness of this model in designing cellulose
fermentation processes is being pursued.
The authors appreciate the financial support given to this study by the Na-
tional Research Council of Canada.

Nomenclature
intracellular ATP-concentration (mol/liter3)
dilution rate (T-l)
dimensionless dilution rate
intracellular concentration of the constitutive permease (mol/liter3)
intracellular concentration of the inducible permease (mol/liter3)
intracellular concentration of intermediate (mol/liter3)
.. constants
binding constant for intermediate (mol/liter3)
binding constant for ATP (mol/liter3)
allosteric constants
exponents in rate equation of allosteric enzymes
fraction of free operators on the operon specific for permease 2
catabolite repression function
intracellular concentration of mltNA for permease 2 (mol/liter3)
.. rate expressions (mol/liter3) (T-1)
bulk medium concentration of substrate 1 (mol/liter3)
bulk medium concentration of substrate 2 (mol/liter3)
time (T)
specific cell volume (liter3/mol)
bulk medium cell mass concentration (mol/liter$)
dimensionless cell mass concentration
dimensionless ATP concentration
Ez/Ei
dimensionless time
.. dimensionless constants
d X / d t ' 1/X
dimensionless intermediate concentration
dimensionless mlENA-concentration
dimensionless substrate concentrations

References
1. J. Monod, Annu. Rev. Microbiol., 3, 371 (1949).
2. T. Kono and T. Asai, J . Ferment. Technol., 49, 128 (1971).
3. C. N. Standing, A. G. Frederickson, and H. M. Tsuchiya, A p p l . Microbiol.,
23, 354 (1972).
4. T. Nisizawa, H. Suzuki, M. Nakayama, and K. Nisizawa, J. Biochem.,
70, 375 (1971).
5. M. Mandels and E. T. lteese, J. Bacteriol., 73, 269 (1957).
6. M. Moo-Young and G. van Dedem, Proceedings of Global Impacts of
Applied Microbiology, Brazilian Society for Microbiology, Sao Paulo, Brazil, 1975.
1312 VAN DEDEM AND MOO-YOUNG

7. A. White, D. Handler, and E. L. Smith, Principles of Biochemistry, 4th ed.,

McGraw-Hill, New York, p. 783.
8 . S. Roseman, in Metabolic Pathways, vol. 4, Academic Press, New York,
1972, p. 41.
9. G. van Dedem and M. Moo-Young, Biotechnol. Bioeng., 15, 419 (1973).
10. A. H. Stouthamer, in Methods i n Microbiology, vol. 1, J. R. Norris and
D. W. Ribbons, Eds., Academic Press, New York, 1969.
11. G. Hill, M.A.Sc. Thesis, University of Waterloo, 1974.
12. J. Monod, J. Wyman, and J. P. Changeux, J . Mol. Biol., 12, 88 (1965).
13. A. Jobe and S. Bourgeois, J . Mol. Biol., 69, 397 (1972).
14. E. S. Polakis and W. Bartley, Biochem. J., 99, 521 (1966).
15. H. A. Cole, J. W. T. Wimpenny, and D. E . Hughes, Biochim. Biophys. Acta,
143, 445 (1967).
16. I. Pastan and R. Perlman, Science, 169, 339 (1970).

Accepted for Publication Lpril 15, 1975