You are on page 1of 15

Critical Reviews in Biotechnology, 20(1):115 (2000)

Penicillin Fermentation: Mechanisms and Models


for Industrial-Scale Bioreactors*
Pratap R. Patnaik**

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

Institute of Microbial Technology, Sector 39-A, Chandigarh-160 036, India


ABSTRACT: Even after many years of research and industrial practice, the production of penicillin
G in fed-batch fermentation by Penicillium crysogenum continues to attract research interest. There are
many reasons: the commercial and therapeutic importance of penicillin and its derivatives, the complexity
of cell growth, and the impact of engineering variables, the last of which are significant in large bioreactors
but are not yet fully understood. Extensive research has generated new information on the mechanisms
of cellular reactions and morphological features of the mycelia and their role in the synthesis of the
product. Given a choice of mechanisms, models of different degrees of complexity, for both cellular
differentiation and bioreactor performance, have been proposed. The more complex models require and
provide more information. They are also more difficult to evaluate and apply in automatic control systems
for production-scale bioreactors. The present review considers the evolution of recent knowledge and
models from this perspective.
KEY WORDS: penicillin fermentation, mechanisms, models, morphology, industrial bioreactors.
* IMTECH communication no. 010/99.
** Fax : ++91-(0)172-690585, 690632. E-mail: pratap@imtech.ernet.in

I. INTRODUCTION
With increasing experience of the fermentation, there has been an improved understanding
of the mechanisms and models for the production of penicillin G. The evolution of new models has been both useful and problematic. It is
useful because improved experimental methods
and models have provided more detailed understanding of not just a specific important fermentation but of the morphology and biochemistry
of filamentous fungi in general. Ironically, this
is also the source of some difficulties because
research has not progressed uniformly on all aspects of the fermentation. Information on some
aspects is more detailed than may be necessary
in industrial application, while there is inadequate

knowledge in some key areas. This poses the


problem of how best to combine different degrees of detail to obtain a comprehensive model
with balanced and sufficient detail without being
too complex.
The term industrial application embodies
a central aspect of this review, and it is an important but difficult feature of penicillin fermentation for two reasons. First, the more information
a model contains, the more complex it becomes,
thus limiting its usefulness for automation. Secondly, while biochemical and morphological
aspects of the growth and productivity of Penicillium crysogenum are now fairly well known,
there is less understanding of the engineering aspects. The effects of gases (mainly oxygen and
carbon dioxide), agitation, heterogeneity of the

0738-8551/00/$.50
2000 by CRC Press LLC

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

cells, nonisotropicity of the broth, and changes


in rheological properties have begun to be investigated only in recent years.
Because the engineering variables influence
the metabolic processes as well as fungal morphology (see Nielsen, 1992; and Prosser and Tough,
1991 for reviews), models that incorporate all
features become prohibitively complex. They are
also difficult and expensive to verify in real (nonideal) situations. The presence of many parameters poses the risk that a model that fits the data
may not necessarily be either the simplest possible or mechanistically correct (Patnaik, 1999).
Thus, while structural detail of intracellular processes and hydrodynamics of the mycelial broth
are important determinants of the performance
of the fermentation, a judicious extent of lumped
approximations are useful in practical applications for on-line optimization and control (Menezes
et al., 1994; Nestaas and Wang, 1983). This review surveys recent developments in the mechanisms and models for the growth of P. crysogenum and the production of penicillin G in order
to understand how they may be improved, integrated, and possibly simplified for cost-effective applications on an industrial scale. The early
work in this direction has already been reviewed
(Hersbach et al., 1984; Schutt, 1986). Since then,
there has been significant progress in understanding of the morphology, transport processes, and
reactor dynamics, which form the main subjects
of this review.

II. UNSTRUCTURED MODELS


Despite criticism and weaknesses, the seminal work by Bajpai and Reuss (1980) continues
to be used in different forms to this day. Their
model was based on the following assumptions:
(1) cell growth is limited by the substrate (glucose) and oxygen and follows Contois kinetics,
(2) the entire biomass is able to grow and synthesize penicillin, (3) product formation is inhibited by the substrate, and (4) maintenance
requirements are constant. Then, mass balances

for fed-batch fermentation lead to the simple,


and often workable, model shown below.

dX
FX
= X
dt
V
dS
FS FS
X X
=

mX + f
dt
Yxs Yps
V
V
dP
FP
= X
dt
V
dV
= F( t )
dt

(1)
(2)
(3)
(4)

where X = biomass concentration, S = substrate


concentration, P = product concentration, V =
volume of broth in the bioreactor, F = feed rate,
Sf = substrate concentration in the feed, = specific growth rate of biomass, = specific rate of
product formation, m = maintenance coefficient,
Yxs = yield of biomass per unit mass of substarte,
and Yps = yield of product per unit mass of
substrate.
The specific growth rate was modelled by
the Contois equation
=

max S
KsX + S

(5)

which implies that, as experimentally observed,


the rate slows down with increasing biomass concentration. While the substrate does not inhibit
cell growth according to Eq. 5, it does for penicillin synthesis, which is expressed by
=

max S
K p + S + S2 K i

(6)

In Eqs. 5 and 6, Kp is the Monod constant, Ki


the inhibition constant, and max and max are
the maximum values of and .
Some of these assumptions have understandably been questioned later. The result has been
the growth of two classes of models: (1) lumped
(or unstructured) models, which have retained
the basic Bajpai-Reuss framework but modified
some aspects, and (2) structured models, which

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

have radically changed or discarded some of


the assumptions and included more cellular
detail.
A primary difference between models that
include cellular structure and those that do not
is the recognition of two or more kinds of cells,
of which some are capable of penicillin synthesis and others are not. Even though this idea was
introduced by Nestaas and Wang in 1983, some
studies conducted later prefered to modify the
Bajpai-Reuss model, retaining the concept of a
uniformly productive biomass. Nicolai et al. (1991)
proposed a flexible model that combines the
Bajpai-Reuss model and its forerunner (Heijnen
et al., 1979). These two approaches have two
main differences. First, Heijnen et al. (1979) used
Blackman type of kinetics for the specific formation rate, , of penicillin, and a Monod equation for the specific growth rate, , of biomass;
Bajpai and Reuss (1980), on the other hand,
employed Haldane substrate inhibition kinetics
for and Contois equation for . Secondly, the
former authors adopted an endogenous metabolism viewpoint while the latter used a maintenance metabolism concept.
Nicolai et al. (1991) pointed out that neither
approach is valid over a sufficiently wide range
of operating conditions. Endogenous metabolism can be justified at low substrate concentrations, while maintenance metabolism is dominant
at high concentrations. This is more than a conceptual difference. In fed-batch operation for
penicillin G production, initially a high substrate
concentration is maintained to promote cell growth,
and later the substrate feed rate is controlled so
that its concentration in the bioreactor is maintained sufficiently low (but above a critical level) such that inhibition of penicillin synthesis is
minimized. This indicates that cell growth is
governed by maintenance metabolism and penicillin synthesis by endogenous metabolism. To
avoid thus having to switch from one kind of
model to another during the fermentation, the
combined model of Nicolai and co-workers allows a smooth transition, with a predominantly
maintenance metabolism slowly being transformed to the endogenous form.

More recently, Menezes et al. (1994) also


utilized the Bajpai-Reuss formalism for an industrial strain of Penicillium crysogenum, but
modified it to suit their observations. The main
modification was to have a Monod type of
equation for the specific rate of formation, , of
penicillin. This may be contrasted with the
Blackman kinetics of Heijnen et al. (1979) and
the substrate inhibition model of Bajpai and
Reuss (1980). While the absence of catabolite
repression by glucose is unusual, it does seem
to depend on the particular microbial culture
because Tiller et al. (1994) also reported a
similar observation. However, as both strains
were procured from industry and were not identified, it is difficult to speculate on the reasons.
Like Nestaas and Wang (1983), Menezes et al.
recognized two kinds of cells; however, because their concentrations during the fermentation were not measured but estimated from
material balances for the reactor, the correctness of the values depends on the correctness
of the model itself.

III. STRUCTURED MODELS


Like that of Bajpai and Reuss (1980), the
structured model proposed by Nestaas and Wang
(1983) became a trendsetter for models of its
kind. All subsequent structured models considered two or three kinds of cells and mechanisms
for hyphal growth. The basic tenet of Nestaas
and Wang was that the culture contains productive cells and nonproductive cells; the former
have cytoplasm and the latter do not. This is an
important distinction because the authors could
correlate penicillin synthesis activity of the mycelial culture to the total cytoplasmic content of
the cells. Such a relationship is also supported
by later studies (Paul and Thomas, 1996; Prosser
and Tough, 1991). Another key concept was the
growth of hyphae, which was expressed by postulating a special kind of cells at the hyphal tips.
Productive cells may degenerate to the nonproductive type, but, of course, the reverse pro-

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

cess does not occur. How this happens biochemically was clarified later (Aynsley et al., 1990;
Nielsen, 1993). Nestaas and Wang merely assumed that degeneration is negligible during the
biomass growth phase (Figure 1) and expressed
the rates of change of hyphal concentration and
productive cell concentration as linear functions
of the former. The fact that the latter rate depends
linearly on the hyphal concentration and not on
the cell concentration is mechanistically significant because it implies that growth occurs only
by extension, generation, and fragmentation of
hyphae, and that a cell may have a variable number of hyphae. This idea, prescient at that time,
has been experimentally confirmed later (Paul et
al., 1994a; Yang et al., 1992).
During the subsequent production phase (Figure 2) the inflow rate of substrate is reduced,
thus lowering the growth rate. The transition
from growth to production was postulated to
pass through a set of metabolic intermediates;
because of comparable time constants, they were
lumped for modeling into one chemical component. The concept of a metabolic intermediate
prior to penicillin formation was also employed
by Aynsley and associates (1990). Similar to

Nestaas and Wang (1983), they considered active biomass and dead (nonproductive) biomass
but neglected the latter because it is difficult to
measure. This is a more serious weakness than
that of Menezes et al. (1994), who also recognize experimental difficulty but nevertheless estimated the inactive biomass computationally.
While the metabolic intermediate concept
leads to models that agree with the overall performance of the fermentation, the inability to
specify its physical identity or that of an intermediate enzyme that helps transform the precursor
into the cell wall (Aynsley et al., 1990) leaves
open the possibility of these species being modeling artifacts. Such a possibility for even the
division of biomass into an active (productive)
form and an inactive form has been alluded to
by Kluge et al. (1992), who observed a shift
from nitrogen to carbon limitation after transition from the growth phase to the production
phase. As described earlier (Nicolai et al., 1991),
this shift also corresponds to a change from maintenance metabolism to endogenous metabolism;
the correspondence is physically plausible because maintenance becomes less significant as
the build-up of cellular material slows down.

FIGURE 1. Postulated reaction mechanism during the biomass growth phase. (Reproduced from Nestaas and
Wang [1983] with permission from John Wiley & Sons 1983.)

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

FIGURE 2. Postulated reaction mechanism during the penicillin production phase. (Reproduced from Nestaas and
Wang [1983] with permission from John Wiley & Sons 1983.)

IV. HYPHAL GROWTH


AND FRAGMENTATION
The hyphae play a key role in both cell growth
and penicillin synthesis, and hence most structured models have addressed this phenomenon
in considerable detail. While all researchers recognize the special nature of the tips of the hyphae, they differ in their interpretations of its
involvement in the transport of nutrients and
the biochemical reactions that result in cellular
growth. Aynsley et al. (1990) assumed that an
unspecified enzyme at the tips contributes to
the synthesis of cell wall material and that the
process is not limited by mass transfer. Both
assumptions have difficulties in acceptability.
If the enzyme is confined to the tips, one may
ask how Aynsley et al. also assumed that the
full length of the fungal filament contributes to
growth. Secondly, mass transfer can indeed be

a limiting factor because the broth may be viscous (Pedesen et al., 1993) and oxygen transfer
is a key factor controlling both growth and product formation (Nielsen, 1992; Prosser and Tough,
1991).
While the exact relationships between the
metabolic reactions and cellular growth continue to be investigated, the special role of the
tips of the hyphae has been well established.
The biochemical mechanisms are discussed in
the references cited by Prosser and Tough (1991)
and Nielsen (1993) and are not the subject of
this review. The broad consensus that emerges
from the mechanisms is that cell growth models should distinguish the cells at the tips from
those elsewhere. Accordingly, Aynsley et al.
(1990) distinguished apical cells (at the tips or
apexes) from the rest of the fungal biomass,
Nielsen (1993) identified apical cells, subapical cells (just beyond the apex), and hyphal cells

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

(toward the roots of the hyphae) (Figure 3), and


Paul and Thomas (1996) demarcated each hypha into five regions.
As the hyphae are critical to the penicillin
production process, it is important to have some
measure of their growth. Caldwell and Trinci (1973)
defined a hyphal growth unit (HGU) as the total
mycelium length divided by the number of tips.
Calling this a length unit, lhgu, Nielsen (1993) refers to a volume growth unit, vhgu (Robinson and
Smith, 1979), and a mass growth unit, mhgu (Nielsen,
1992), which can be defined in a similar manner. The three units are related as (Nielsen, 1992):
m hgu = (1 w)v hgu
= (1 w)( 4)d 2 l hgu

(7)

where d is the hyphal diameter, is the mycelium density, w is the water content of the mycelium and = 3.14159. The mass growth unit is
approximately constant at different environmen-

tal conditions for a number of filamentous microorganisms, but the hyphal diameter varies;
hence, by Eq. 7, lhgu also varies. The growth of
biomass is favored when lhgu is small, implying
a large number of apical tips; therefore, during
fermentation the value of lhgu increases initially
when the cells are growing and then decreases
during the penicillin synthesis phase (Aynsley
et al., 1990; Nielsen, 1993).
Because the hyphae are long, narrow, and intermeshed in the mycelial mass, as they elongate
they send to break and to develop stagnant regions or vacuoles. These processes are lucidly
captured by a model proposed by Paul and Thomas
(1996). They divided a hyphal filament into (Figure 4): (1) an actively growing apical region, Ao;
(2) nongrowing cytophasm, A1; (3) vacuolar spaces,
A2; (4) a degenerate region A3; and (5) an autolysed region, A4. The growth of hyphae takes place
in the apical regions, whereas new tips may be
generated either by branching from the A1 regions

FIGURE 3. Structure of a densely packed hyphal element. The tip section has been enlarged to show the apical,
subapical, and hyphal cells. (Reproduced from Nielsen [1993] with permission from John Wiley & Sons 1993.)

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

FIGURE 4. Schematic representation of five regions of a hyphal filament (above) and the internal reaction
mechanisms (below). (I) Hyphal extension; (II) branching; (III) penicillin production; (IV) vacuolation (V; V) degeneration; (VI) autolysis. (Reproduced from Paul and Thomas [1996] with permission from John Wiley & Sons 1996.)

or by fragmentation of hyphae. It is interesting to


note that the two modes of tip generation occur
under very different substrate conditions; branching occurs at high concentrations and fragmentation at low concentrations. Although new tips
arise through fragmentation, their usefulness is
nullified by autolysis. This happens because
fragmentation takes place as a consequence of
the vacuoles becoming large, with the result that
the cell wall is weak and easily ruptured by shear
(Paul et al., 1994b).
While the concentrations of substrate, penicillin, and the total biomass may be expressed
in the mass balances of either lumped (Bajpai
and Reuss, 1980; Menezes et al., 1994) or structured (Kluge et al., 1992; Paul et al., 1998) mod-

els through their measured macroscopic values,


the properties of the hyphae have to be statistically averaged. The averaging may be done either
experimentally or by population balance methods. Image analysis (Paul et al., 1994a; Tucker
et al., 1992) offers a convenient and accurate
technique to obtain estimates of average values
of hyphal length, number of tips per hyphal element, length of the hyphal growth unit, etc. The
models of Aynsley et al. (1990), Kluge et al.
(1992), Tiller et al. (1994), and Viniegra-Gonzalez
et al. (1993) are formulated in terms of such
average properties.
Average property models, by definition, ignore variations among the cells in a mycelial
population, which population balance models
7

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

describe. The latter may be more accurate and


informative, but they are also more difficult to
verify and apply. Most models of this kind have
some deterministic and some stochastic components. Krabben and Nielsen (1998) have summarized the salient features of some population
balance models. Because their complexity has
restricted the industrial application of these models, one solution is to derive average properties
from them for use in simpler deterministic models. For instance, Nielsen (1993) derived the equations given below for the average values of the
number of hyphal elements, e, the mass of hyphae, m, and the number of actively growing
tips, n, per unit volume of broth.

with the initial conditiony

q( r,0) = f ( r )

(13)

and the boundary condition


q( r0 , t ) = g( t )

(14)

In Eqs. 13 and 14, f(r) is the distribution of valuole


sizes at the time of inoculation (t = 0), r0 is the
(average) radius of a vacuole at birth, g(t) is the
variation with time of the population of new
vacuoles, and is a dispersion coefficient.

V. MIXING AND AERATION


de
= ( m D)e
dt
dm
= ( m )m
dt
dn
= ( n )m
dt

(8)
(9)
(10)

D is the conventional dilution rate, F/V. is the


specific branching frequency, that is, the number of tips formed per hyphal element unit mass
per unit time:
= k1z

(11)

where z is the concentration of subapical cells,


while and k1 are physiological constants.
(t) is the specific rate of fragmentation; it is
determined by the circulation pattern and the
shear force distribution in the bioreactor but
may be approximated to be a function of only
the energy input (van Suidjam and Metz, 1981).
The growth of vacuoles may also be described
by a similar approach. The number, q(r,t), of
vacuoles of radius r at time t evolves with time
according to (Paul and Thomas, 1996):
q
q
2 n
= ks
+ 2
t
r
r

(12)

It may be recalled that during the penicillin


synthesis phase the substrate concentration should
be low to avoid catabolite repression (Hersbach
et al., 1984; Ryu and Hospodka, 1980). However,
because low concentrations also promote fragmentation and lysis, a minimum concentration
has to be maintained. Because the duration of
a fermentation batch is typically 150 to 180 h
and the rheological properties (Pedesen et al.,
1993), mass transfer coefficient for oxygen (Ju et
al., 1991), and the mixing intensify in the broth
(Pedersen et al., 1994) vary during this period,
it is not easy to decide and maintain an optimal
substrate concentration at all times on the basis
of a bioreactor model. One way to overcome this
difficulty is by the application of a control strategy that does not require a process model, such
as through neural networks (DiMassimo et al.,
1992; Ignova et al., 1996; Montague et al., 1992).
These features become increasingly pronounced as the size of the fermentation vessel
and the concentration of biomass increase. In a
preliminary study, Pedesen et al. (1994) observed
that the time to attain 63.2% of complete mixing in a 41 L fermentation vessel equipped with
Rushton tubines increased threefold and the
average viscosity increased from that of water
to more than 300 mPa during the course of fermentation. Smith et al. (1990) reported qualitatively similar results from a 150-L vessel filled

with 80 L of the medium at the beginning of the


fermentation. However, they calculated the circulation time asy
t c = (0.76 N )(H D t )

0.6

(D t

Di )

2.7

(15)

whereas Pedersen et al.s mixing time was

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

t m = (1 k ) ln E

(16)

In Eq. 15, N is the impeller tip speed (m/s), Di


is the impeller diameter (m), Dt is the vessel
diameter (m) and H is the vessel height (m).
Equation 16 was obtained from radioactivity monitoring of 113mIn (half life = 99.5 min) tracer, for
which E is the exit age distribution. Although
Pedesen et al. (1994) and Smith et al. (1990)
obtained comparable values (1 to 3 s), it is difficult to see a quantitative relation between tc
and tm.
Nevertheless, agitation affects the morphology, the rheological properties of the broth, and
the rate of oxygen transfer. Studies of the effects of agitation on the mycelia have tried to
ensure that it is good enough to avoid limitations
due to mass transfer and mixing. However, the
minimum stirring speed needed for this is not
clear because the minimum requirement of dissolved oxygen concentration to overcome mass
transfer control varies from 8% (Konig et al.,
1981) to 30% (Vardar and Lilly, 1982) of the
saturation value. To be safe, most investigators
have chosen the upper value. However, apart
from increasing the energy requirement, excessive aeration results in more of the carbon source
being converted to CO2 rather than penicillin
(Konig et al., 1981). While this is undesirable,
recent studies have shown that sparging with
CO2 (along with air) increases the branching
frequency of hyphae (Smith and Ho, 1985) and
reduces the non-Newtonian (pseudoplastic)
character and the viscosity of the broth (Ju et al.,
1991). Thus, there appear to be some optimal
concentrations for both oxygen and CO2. However, their exact values seem to depend on the
characteristics of the strain, the medium and the

reactor, thus making it difficult to provide general recommendations.


The requirement of oxygen is obviously linked
to the cellular metabolism, but details about this
are just beginning to emerge. Henriksen and coworkers (1997) observed that above a dissolved
oxygen concentration of 0.08 mM (the same as
in Konig et al., 1981) there was no further improvement in specific penicillin productivity and
byproducts such as -(L-a-aminoadipyl)-cysteinyl-D-valine, isopenicillin N, and 6-aminopenicillanic acid decreased with increasing oxygen
concentration. On the contrary, the formation of
glutathione and cysteine increased. Metabolic
control analysis revealed that at low dissolved
oxygen concentration the flux control is mainly
by isopenicillin N synthetase, whereas a more
even distribution of flux control by this enzyme
and -(L-a-aminoadipyl)-L-cysteinyl-D-valine
synthetase is possible at higher concentrations.
While increased agitation provides better gas
dispersion and a reduction in viscosity by shifting the pseudoplastic nature of the broth toward
Newtonian character (Allen and Robinson, 1990;
Justen et al., 1996; Pedesen et al., 1993), the greater shear generated results in shorter hyphae and
a large number of branches. While this may favor penicillin synthesis, increased shear also
causes greater fragmentation and autolysis of
hyphae, with a consequent loss of intracellular
enzymes. Thus, from this perspective, the specific penicillin production rate is expected to
increase initially with increasing agitation but
should soon decrease as the speed or intensity
of agitation becomes large enough to make its
detrimental effects more significant, as demonstrated in many experiments (Konig et al., 1981;
Makagiansar et al., 1993; Smith et al., 1990; Vardar
and Lilly, 1982). It is necessary to distinguish
between speed and intensity of agitation because different types of impellers produce different shear stresses at the same speed, and, consequently, the cell growth rate, oxygen transfer
rate, and penicillin production rate differ (Justen
et al., 1996, 1998).
The expected decrease in productivity at
high intensities of mixing is also supported by

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

observations (Paul and Thomas, 1996) that it


generates vacuoles in the distal regions, where
penicillin synthesis is believed to be localized.
However, this inference may not be definitive,
as shown by their later results. With Rushton
turbine, paddle type, and pitched blade impellers,
Justen et al. (1998) observed that the percentage
of hyphal vacuolation decreased with time during
the rapid growth phase (about 24 h) and was sensibly independent of the agitation conditions. This
seemingly anomalous behavior was attributed
to the high rate of glucose inflow, which earlier
had been shown (Paul et al., 1994a) to reduce
hyphal vacuolation and degeneration. In other
words, the physiological effect of glucose more
than counterbalances the impact of agitation.
However, in the penicillin production phase
there was no common trend. For Rushton turbines, an initial rise in vacuolation was followed
by a continuing rise (600 rpm) or a plateau
(1000 rpm) or a decline (1400 rpm). For other
types of agitators, the extent of vacuolation remained practically constant beyond the growth
phase. The productivity of penicillin was highest
with a Rushton turbine at 600 rpm and a paddle
impeller at 280 rpm, which produced very different amounts of vacuolation and empty cells.
These results are a cautionary indication that a
generally valid relation between performance and
operating conditions in realistic situations is
still not known; this strengthens the argument
in favor of methods that directly utilize performance data and artificial intelligence for on-line
optimization and control (DiMassino et al., 1992;
Ignova et al., 1996; Montague et al., 1992).
Given that the morphology varies with time
and the operating conditions, and that penicillin
synthesis is linked with the mycelial structure,
it becomes useful to correlate morphology with
some index of the operating environment. Based
on the work of Smith et al. (1990), Makagiansar
et al. (1993), and their own, Justen et al. (1996)
suggested that an energy dissipation/circulation function of the form (P/kD3) (1/tc) was
the only good correlator. Here P is the power
input, D is the impeller diameter, tc is the circulation time and K is a parameter that depends
on the type of impeller. Extending Eq. 15, from
10

Smith et al. (1990), they proposed a generalized form for the circulation time:
t c = V ( Fl G ND3 )

(17)

where N is the impeller speed, V is the working


volume and FlG is the gassed flow number
which may be obtained from the ungassed flow
number, Fl, as:

Fl G = Fl( Po G Po)

(18)

Po and PoG are the ungassed and gassed power


numbers, and for Fl Justen et al. (1996) recommended different correlations for axial flow and
radial flow. Data from bioreactors with working
volumes of 1.4 L, 20 L, and 180 L showed that
the mean total hyphal length and its mean projected area decreased with increasing energy dissipation and time, and the correlation coefficients
were better than 0.80.
While these trends corroborate earlier observations (Paul and Thomas, 1996; Prosser and
Tough, 1991; Smith et al., 1990) of a shift in
morphology from clumped (or pelletized) form
to freely dispersed mycelia as fermentation progresses, they also have interesting implications
for bioreactor scale-up. From Eq. 17 and with
geometric similarity and scale-up at constant
tip speed of the impeller, it can be shown (Justen
et al., 1996) that:

(P k D )(1 t ) D
3

2 3

(19)

Therefore, the larger the impeller diameter,


the smaller is the energy dissipation function
and hence the less is the fragmentation. In other
words, mycelial breakage will be less in large
reactors than in small ones, a very useful deduction for industrial production.

VI. DIFFUSION AND REACTION


IN AGGREGATED MYCELIA
Although both freely dispensed mycelia and
aggregated mycelia (clumps or pellets) are pres-

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

ent, most of the studies have considered only


the former. While this simplifies modeling because diffusion through the aggregates does not
have to be considered, it is not quite realistic because free mycelia usually constitutes no more
than 20% of the biomass on the basis of projected area measurements by image analysis
(Tucker et al., 1992).
Emersons (1950) cube-root law is still considered a good basis to describe the growth of
mycelial pellets (Prosser and Tough, 1991). It
takes the form

M1

= M10 3 + kt

(20)

where M0 is the initial mass of culture and M


is its value at a time t. To determine the value
of k, Pirt (1966) considered a culture containing
spherical pellets of equal radius r and density
, and assumed that growth progressed outward
through a peripheral shell of thickness w. With
this model,
k = ( 4 n 3) w
13

(21)

where is the specific growth rate.


According to the foregoing model, the rate
of increase of pellet radius r will be:
dr dt = w.

When is constant, as in the latter phase of fedbatch fermentation (Konig et al., 1981; Nielsen,
1992; Tiller et al., 1994) or in continuous cultivation (Prosser and Tough, 1991),
r = r0 + wt

(22)

with r0 being the starting value of r.


When inward diffusion of nutrients through
the pellet is limited by hyphal packing, growth
ceases at a critical radius rc at which the substrate concentration at the center of the pellet
becomes zero.
rc = (6DYxs )

12

(23)

Here D is the diffusion coefficient of the substrate and Yxs is the yield coefficient (g/g) for
biomass from substrate.
Although cube-root kinetics have been observed experimentally, the same data can sometimes be described by more than one model. For
example, Trincis (1970) data for Aspergillus nidulans can be modeled by a cube-root law as well
as by a modified logistic equation (Koch, 1975).
Herein lies a common problem in model discrimination. Sometimes two or more differential
equation models may fit a set of data within
acceptable errors. Then it becomes difficult to
choose the most plausible model, and one may be
inclined to opt for the most simple model. However, a simple model may not have adequate mechanistic detail, thus limiting its range of validity and
the ability to manipulate the model for different
situations. Considering the present example, although the logistic equation is less simple than
the cube-root equation, it includes the cytoplasm
content of the cells, which determines penicillin synthesis activity (Nestaas and Wang, 1983;
Paul and Thomas, 1996), describes different kinds
of growth, and relates growth on solid and in
liquid medium. Thus, when adequate data are available, the most plausible model may not be the
most simple. When equation-based methods fail
to identify one such model, alternate methods of
discrimination based on the topological structures of competing mechanisms may be successful (Patnaik, 1993, 1999).
Neither the cube-root formulation nor the
logistic equation rigorously incorporates mass
transfer and diffusion. Oxygen is a key growthlimiting nutrient. Its steady-state diffusion through
a spherical pellet may be described by a Fickian
form (Nielsen, 1992).

d 2 c 2 dc
D 2 +
= x( r )
dr
r dr

(24)

In Eq. 24, c is the oxygen concentration and x


the biomass concentration at a radial distance r
from the center. To obtain the intrapellet oxygen
profile, Eq. 24 should be solved together with
a pellet growth model (cube-root or logistic or any
other) and a model for the specific growth rate .
11

Agitation can damage mycelial pellets just


as it does to hyphae. This aspect has not yet been
studied in sufficient detail. Nevertheless, the early
work of Taguchi (1971) provides a good basis
to construct more detailed, and possibly more
accurate, models. He proposed that the mean pellet diameter DP and the number of nondisrupted
pellets n change with time according to the following semiempirical equations.

Equation 27 considers roughly that there is


enhanced transport by convection through the
sparse outer regions of the pellet (upper condition) and hindrance of transport in inner dense
regions. This perception is supported by image
analysis photographs of P. crysogenum growth
(Paul et al., 1994a; Tucker et al., 1992).

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

VII. CONCLUDING REMARKS


dD p

= k c ( ND i ) D 5p.7
5.5

dt
dn
= D3p.2 N 6.65D8i .75
dt

(25)
(26)

N and Di denote the speed and the diameter,


while kc and are constants. An obvious way
to generalize this model would be to allow the
exponents to vary and determine their values
along with those of kc and . However, it would
be more useful to have a model derived from
fundamental principles, similar to the modeling
of noncatalytic fluid-solid reactions but accounting for pellet-size frequency distribution, which
itself changes with time (Justen et al., 1996,
1998), and changes in density and diffusivity
as a result of growth, movements of filaments,
and hyphal fragmentation.
Although Tiller et al. (1994) applied their
model to a high-yielding industrial strain of P.
crysogenum with significant pellet formation,
there was no explicit consideration of the properties of the pellets. This is important because
the size distribution of pellets is strongly influenced by shear forces (Hoptop et al., 1993), and
the density of a pellet varies along its radius
and with the operating conditions (van Suidjam
and Metz, 1981). Recently, Meyerhoff et al.
(1995) expressed some of these features in an
effective diffusivity coefficient De that depended
on the cell density in the manner:
5Dmax if < crit and r > rcrit

D e () = 2 Dmax (1 )(1 + )
(27)

if > crit or r < rcrit

12

Over the last 2 decades, experimental studies have revealed considerable detail of the metabolic reactions inside fungal cells, mainly of the
Penicillium, Aspergillus, and Streptomyces genus,
and of the growth and morphological features
of the mycelia. Most of these studies have been
in laboratory and pilot-scale bioreactors, which
may not truly represent industrial conditions.
The importance of simulating industrial conditions in laboratory experiments and models
for penicillin biosysnthesis is highlighted by
recent studies, which show that during the long
course (150 to 180 h) of fermentation, the metabolism, morphology, and rheology of the broth
change considerably, with consequent effects
on productivity. Much of the research has, however, concentrated on deriving mechanisms for
cellular and morphological changes under somewhat ideal conditions, and phenomena that become significant on a large scale of operation
have not received sufficient attention. The latter group includes incomplete and changing mixing patterns (Pedersen et al., 1994), the effect of
dissolved oxygen and carbon dioxide on rheology, mass transfer and metabolism (Henriksen
et al., 1997; Ju et al., 1991), diffusion and reaction in mycelial aggregates of varying characteristics (Prosser and Tough, 1991), and the
role of reactor and agitator design (Justen et al.,
1996, 1998).
Because of the disparity between detailed
models of microbial physiology and morphology on the one hand and more conservative models of bioreactor behavior on the other, control
of the fermentation, which relies mainly on
macroscopic observations, has to employ some

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

lumping of the metabolic and morphologic equations (Montague et al., 1989; Modak, 1993). This
poses a fundamental dilemma: how much detail
should be included for a particular application?
Unstructured models are simple but contain little
information on intracellular processes and the
effects of operating conditions on morphology
and productivity. Moreover, they are valid only
if the microrganism has adapted to its environment (Roels, 1983), which is unlikely in fedbatch penicillin fermentations (Paul and Thomas,
1996). Structured models, on the other hand,
may become too complex for easy automation.
The large numbers of parameters they contain
make these models difficult to evaluate and to
update during the course of fermentation. Even
simple structured models proposed earlier (Megee
et al., 1970; Nestaas and Wang, 1983) contained
more than 40 parameters.
The development of workable models, which
contain just the required degree of complexity,
include all relevant features and can readily be
applied in industrial control systems will probably remain a prime challenge for mycelial fermentations. At present, there are two approaches
to this problem: either reduce the number of parameters by judicious lumping (Paul and Thomas,
1996) or adopt model-free methods utilizing performance data and artificial intelligence (DiMassimo
et al., 1992; Ignova et al., 1996; Montague et
al., 1992). Both approaches have limitations; so
eventually the best approach may combine phenomenological models of some aspects of the
process with intelligent software for the more
difficult features (Schubert et al., 1994; Preusting
et al., 1997).

REFERENCES

Bajpai, R. K. and Reuss, M. 1980. A mechanistic


model for penicillin production. J. Chem. Technol.
Biotechnol. 30: 332344.
Caldwell, I. Y. and Trinci, A. P. J. 1973. The growth
unit of the mould Geotrichum candidum. Arch.
Microbiol. 88: 110.
DiMassimo, C., Montague, G. A., Willis, M. J., Tham,
M. T., and Morris, A.J. 1992. Towards improved
penicillin fermentation via artificial neural networks.
Comput. Chem. Eng. 16: 283291.
Edelstein, L. and Hadar, Y. 1983. A model for pellet
size distributions in submerged mycelial cultures.
J. Theor. Biol. 105: 427452.
Emerson, S. 1950. The growth phase in Neurospora
corresponding to the logarithmic phase in unicellular organisms. J. Bacteriol. 60: 221.
Heijnen, J. J., Roels, J. A., and Stouthamer, A. H. 1979.
Application of balancing methods in modelling the
penicillin fermentation. Biotechnol. Bioeng. 21: 2175
2201.
Henriksen, C. M., Nielsen, J., and Villadsen, J. 1997.
Influence of dissolved oxygen concentration on the
penicillin biosynthetic pathway in steady-state cultures of Penicillium crysogenum. Biotechnol. Prog.
13: 776782.
Hersbach, G. J. M., Van der Beek, C. P., and Van Dijk,
P. W. M. 1984. The penicillins: properties, biosynthesis and fermentation. In: Vandamme, E., Ed.,
Biotechnology of Industrial Antibiotics. Marcel
Dekker, New York, pp. 45140.
Hoptop, S., Moller, J., Niehoff, J., and Schugerl, K.
1993. Inflluence of the preculture conditions on
the pellet size distribution of Penicillium crysogenum cultivations. Process Biochem. 28: 99104.
Ignova, M., Paul, G. C., Glassey, J., Ward, A. C., Montague,
G. A., Thomas, C. R., and Karim, M. N. 1996.
Towards intelligent process supervision: industrial
penicillin fermentation case-study. Comput. Chem.
Eng. 20: S545S550.

Allen, D. G. and Robinson, C. W. 1990. Measurement


of the rheological properties of filamentous fermentation broths. Chem. Eng. Sci. 45: 3748.

Ju, L-K., Ho, C. S., and Shanahan, J. F. 1991. Effects


of carbon dioxide on the rheological behavior and
oxygen transfer in submerged penicillin fermentations. Biotechnol. Bioeng. 38: 12231232.

Aynsley, M., Ward, A. C., and Wright, A. R. 1990. A


mathematical model for the growth of mycelial
fungi in submerged culture. Biotechnol. Bioeng.
35: 820830.

Justen, P., Paul, G. C., Nienow, A. W., and Thomas, C.


R. 1996. Dependence of mycelial morphology on
impeller type and agitation intensity. Biotechnol.
Bioeng. 52: 672684.

13

Justen, P., Paul, G. C., Nienow, A. W., and Thomas, C. R.


1998. Dependence of Penicillium crysogenum growth,
morphology, vacuolation and productivity in fedbatch fermentations on impeller type and agitation
intensity. Biotechnol. Bioeng. 59: 762775.

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

Kluge, M., Siegmund, D., Diekmann, H., and Thoma,


M. 1992. A model for penicillin production with
and without temperature shift after the growth phase.
Appl. Microbiol. Biotechnol. 36: 446451.
Koch, A. L. 1975. The kinetics of mycelial growth. J.
Gen. Microbiol. 89: 209.
Konig, B., Seewald, Ch., and Schugerl, K. 1981. Process engineering investigations of penicillin production. Europ. J. Appl. Microbiol. Biotechnol. 12:
205211.
Krabben, P. and Nielsen, J. 1998. Modeling the mycelium
morphology of Penicillium species in submerged cultures. Adv. Biochem. Eng./Biotechnol. 60: 125152.
Makagiansar, H. Y., Shamlou, P. A., Thomas, C. R.,
and Lilly, M. D. 1993. The influence of mechanical forces on the morphology and penicillin production of Penicillium crysogenum. Bioprocess Eng.
9: 8390.
Megee, R. D., Kinoshita, S., Fredrikson, A. G., and
Tsuchiya, H. M. 1970. Differentiation and product
formation in molds. Biotechnol. Bioeng. 12: 771
801.
Menezes, J. C., Alves, S. S., Lemos, J. M., and deAzevedo,
S. F. 1994. Mathematical modeling of industrial
pilot-plant penicillin-G fed-batch fermentations. J.
Chem. Technol. Biotechnol. 61: 123138.
Meyerhoff, J., Tiller, V., and Bellgardt, K-H. 1995.
Two mathematical models for the development of
a single microbial pellet, Parts I and II. Bioprocess
Eng. 12: 305, 315.

Nicolai, B. M., Van Impe, J. F., Vanrollegem, P. A.,


and Vandewalle, J. 1991. A modified unstructured
mathematical model for the penicillin G fed-batch
fermentation. Biotechnol. Lett. 13: 489494.
Nielsen, J. 1992. Modeling the growth of filamentous
fungi. Adv. Biochem. Eng./Biotechnol. 46: 187223.
Nielsen, J. 1993. A simple morphologically structured
model describing the growth of filamentous microorganisms. Biotechnol. Bioeng. 41: 715727.
Patnaik, P. R. 1993. Steady-state multiplicity in enzyme
deactivation: application of a simple graph theoretic
method. Enzyme Microb. Technol. 15: 162166.
Patnaik, P. R. 1999. Penicillin fermentation revisited:
a topological analysis of kinetic multiplicity. Process Biochem. 34: 737743.
Paul, G. C. and Thomas, C. R. 1996. A structured model
for hyphal differentiation and penicillin production
using Penicillium crysogenum. Biotechnol. Bioeng.
51: 558572.
Paul, G. C., Kent, C. A., and Thomas, C. R. 1994a.
Image analysis for characterizing differentiation
of Penicillium crysogenum. Trans. Inst. Chem. Engrs.,
Part C. 72: 95105.
Paul, G. C., Kent, C. A., and Thomas, C. R. 1994b.
Hyphal vacuolation and fragmentation in Penicillium crysogenum. Biotechnol. Bioeng. 44: 655660.
Paul, G. C., Syddall, M. T., Kent, C. A., and Thomas,
C. R. 1998. A structured model for penicillin production on mixed substrates. Biochem. Eng. J. 2:
1121.
Pedesen, A. G., Bundgaard-Nielsen, M., Nielsen, J.,
and Villadsen, J. 1994. Characterization of mixing
in stirred tanks equipped with Rushton turbines.
Biotechnol. Bioeng. 44: 10131017.

Modak, J. M. 1993. Choice of control variable for optimization of fed-batch fermentation. Biochem. Eng.
J. 52: B59B69.

Pedesen, A. G., Bundgaard-Nielsen, M., Nielsen, J.,


Villadsen, J., and Hassager, O. 1993. Rheological
characterisation of media containing Penicillium
crysogenum. Biotechnol. Bioeng. 41: 162164.

Montague, G. A., Morris, A. J., and Tham, M. T. 1992.


Enhancing bioprocess operability with generic software sensors. J. Biotechnol. 25: 183201.

Pirt, S. J. 1966. A theory of the mode of growth of fungi


in the form of pellets in submerged culture. Proc.
R. Soc. Ser. B 166: 369.

Montague, G. A., Morris, A. J., and Ward, A. C. 1989.


Fermentation monitoring and control: a perspective. Biotechnol. Genet. Eng. Rev. 7: 147188.

Preusting, H. J., Noordover, J., Simutis, R., and Lubbert,


A. 1997. The use of hybrid modeling for the optimization of the penicillin fermentation process. Chimia
50: 416417.

Nestaas, E. and Wang, D. I. C. 1983. Computer control


of the penicillin fermentation using the filtration
probe in conjunction with a structured process
model. Biotechnol. Bioeng. 25: 781796.

14

Prosser, J. I. and Tough, A. J. 1991. Growth mechanisms and growth kinetics of filamentous microorganisms. Crit. Revs. Biotechnol. 10: 253274.

Robinson, P. M. and Smith, J. M. 1979. Development


of cells and hyphae of Geotrichum candidum in
chemostat and batch culture. Proc. Brit. Mycol.
Soc. 72: 3947.
Roels, J. A. 1983. Energetics and Kinetics in Biotechnology. Elsevier, Amsterdam.

Critical Reviews in Biotechnology Downloaded from informahealthcare.com by UCA Univeristy of Cadiz on 10/24/11
For personal use only.

Ryu, D. D. Y. and Hospodka, J. 1980. Quantitative


physiology of Penicillium crysogenum in penicillin fermentation. Biotechnol. Bioeng. 22: 289
298.
Schubert, J., Simutis, R., Dors, M., Havlik, I., and
Lubbert, A. 1994. Bioprocess optimization and
control: application of hybrid modeling. J. Biotechnol. 35: 5168.
Schutt, H. 1986. Penicillin. Chemica, Medicinal Chemical Monographs No.1, Michael Barber, Coulsdon,
U.K.
Smith, J. J., Lilly, M. D., and Fox, R. I. 1990. The
effect of agitation on the morphology and penicillin production of Penicillium crysogenum. Biotechnol. Bioeng. 35: 10111023.
Smith, M. D. and Ho, C. S. 1985. The effect of dissolved carbon dioxide on penicillin production:
mycelial morphology. J. Biotechnol. 2: 347363.
Taguchi, H. 1971. The nature of fermentation fluids.
Adv. Biochem. Eng./Biotechnol. 1: 130.

Tiller, V., Meyerhoff, J., Sziele, D., Schugerl, K., and


Bellgardt, K-H. 1994. Segregated mathematical
model for the fed-batch cultivation of a highproducing strain of Penicillium crysogenum. J.
Biotechnol. 34: 119131.
Trinci, A. P. J. 1970. Kinetics of the growth of mycelial pellets of Aspergillus nidulans. Arch.
Microbiol. 73: 353.
Tucker, K. G., Kelly, T., Delgrazia, P., and Thomas, C.
R. 1992. Fully-automatic measurement of mycelial morphology by image analysis. Biotechnol.
Prog. 8: 353359.
van Suidjam, J. C. and Metz, B. 1981. Influence of
engineering variables upon the morphology of filamentous molds. Biotechnol. Bioeng. 23: 111148.
Vardar, F. and Lilly, D. 1982. Effect of cycling dissolved oxygen concentrations on product formation in penicillin fermentations. Eur. J. Appl.
Microbiol. Biotechnol. 14: 203211.
Viniegra-Gonzalez, G., Sanchez-Castaneda, G., LopezIsunza, F., and Favela-Torres, E. 1993. Symmetric branching model for the kinetics of mycelial
growth. Biotechnol. Bioeng. 42: 110.
Yang, H., Reichl, U., King, R., and Gilles, E. D. 1992.
Measurement and simulation of the morphological development of filamentous microorganisms.
Biotechnol. Bioeng. 39: 44-48.

15

You might also like