Bioresource Technology 137 (2013) 98105

Contents lists available at SciVerse ScienceDirect

Bioresource Technology
journal homepage: www.elsevier.com/locate/biortech

Development of a mathematical model for the growth associated

Polyhydroxybutyrate fermentation by Azohydromonas australica
and its use for the design of fed-batch cultivation strategies
Geeta Gahlawat, Ashok K. Srivastava
Department of Biochemical Engineering and Biotechnology, Indian Institute of Technology Delhi, Hauz Khas, New Delhi 110016, India

h i g h l i g h t s
 Batch mathematical model for the Polyhydroxybutyrate (PHB) production was developed.
 Kinetic model was extrapolated to fed-batch for enhanced PHB production.
 Proposed model helped in designing various nutrients feeding regimes.
 Constant feed rate cultivation showed 3.6 fold improvement in PHB production.

a r t i c l e

i n f o

Article history:
Received 23 December 2012
Received in revised form 4 March 2013
Accepted 6 March 2013
Available online 21 March 2013
Keywords:
Polyhydroxybutyrate
Azohydromonas australica
Modeling
Process parameters
Fed-batch culture

a b s t r a c t
In the present investigation, batch cultivation of Azohydromonas australica DSM 1124 was carried out in a
bioreactor for growth associated PHB production. The observed batch PHB production kinetics data was
then used for the development of a mathematical model which adequately described the substrate limitation and inhibition during the cultivation. The statistical validity test demonstrated that the proposed
mathematical model predictions were signicant at 99% condence level. The model was thereafter
extrapolated to fed-batch to identify various nutrients feeding regimes during the bioreactor cultivation
to improve the PHB accumulation. The distinct capability of the mathematical model to predict highly
dynamic fed-batch cultivation strategies was demonstrated by experimental implementation of two
fed-batch cultivation strategies. A signicantly high PHB concentration of 22.65 g/L & an overall PHB content of 76% was achieved during constant feed rate fed-batch cultivation which is the highest PHB content
reported so far using A. australica.
2013 Elsevier Ltd. All rights reserved.

1. Introduction
Polyhydroxybutyrate (PHB) is a natural biodegradable polymer
which is usually synthesized by various microorganisms as energy
reserve material under the limitation of essential nutrients such as
nitrogen, phosphorus and magnesium in the presence of excess
carbon source (Zou and Chen, 2007). In addition to its complete
biodegradability this microbially synthesized polymer also has
various interesting properties such as biocompatibility, thermo
processability, piezoelectricity, and nonlinear optical activity with
many promising applications particularly in food packaging,
wound management, tissue engineering and drug delivery (Zinn
et al., 2001). Another important feature of microbial production
of biopolymer is that it can be produced from renewable feedstocks such as plant oils, corn syrup, glycerol and carbon dioxide

Corresponding author. Tel.: +91 11 26591010; fax: +91 11 26582282.

E-mail addresses: ashokks@dbeb.iitd.ac.in, ashokiitd@gmail.com (A.K. Srivastava).
0960-8524/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.biortech.2013.03.023

and this microbial route can reduce the dependency of society on

disappearing fossils fuels (Loo and Sudesh, 2007). However, the
ongoing impediment to the widespread use of PHB is its much
higher cost of production as compared to conventional plastics.
To date various approaches have been used to tackle this problem and achieve cost-effective PHB production. Several researchers have focused on reducing the production cost by utilization
of inexpensive substrates (Yezza et al., 2007; Omar et al., 2001),
high yielding strains (Grothe and Chisti, 2000; Yamane et al.,
1995), and various process optimization strategies (Ramadas
et al., 2010; Berenjian et al., 2011; Ahn et al., 2000). A bioprocess
engineering approach to address this problem could be to develop
a mathematical model which not only helps in the understanding
of the system but also predicts various cultivation strategies to
facilitate the optimization of a fermentation process, saving much
of the time and cost for performing experiments (Xu et al., 2011;
Liu et al., 2003; Voll et al., 2011). In the present investigation Azohydromonas australica was utilized for PHB production using a
mathematical kinetic modeling approach because it produces

G. Gahlawat, A.K. Srivastava / Bioresource Technology 137 (2013) 98105

99

Nomenclature
a
KS1
KS2
K1
KI
mS1
mS2
S1
S2
Sm

exponent indicating type of relationship between S

(nitrogen) and l
saturation constant for sucrose consumption (g/L)
saturation constant for nitrogen consumption (g/L)
growth associated product formation constant (g/g)
substrate inhibition constant for sucrose consumption
(g/L)
maintenance coefcient (gsucrose/gcells h)
maintenance coefcient (gnitrogen/gcells h)
sucrose concentration (g/L)
nitrogen concentration (g/L)
nitrogen concentration at which complete inhibition occurs (g/L)

PHB in a growth-associated manner, grows fast and can accumulate high amounts of PHB (up to 80% of cell dry weight) during
growth phase of cultivation which could improve the PHB yield
on the carbon source and efciency of PHB recovery process (Hrabak 1992; Gahlawat et al., 2012; Choi and Lee, 1999). Moreover A.
australica is able to utilize an inexpensive carbon source, sucrose
thereby indicating the possible utilization of various renewable
resources rich in sucrose such as beet molasses, cane molasses
and maple sap for PHB production (Yezza et al., 2007; Zafar
et al., 2012).
The main aim of the present study was to propose a simple kinetic model that not only adequately describes the observed batch
kinetics of cultivation but also elucidates substrate limitation and
inhibition under different scenarios of nutrient feed conditions of
fed-batch cultivations. Fed-batch cultivation strategies have been
used by different researchers to facilitate enhanced biomass and/
or product formation for maintenance of non limiting and non
inhibitory concentrations of substrate in the bioreactor but these
could not yield best results primarily due to the trial and error approach adopted for the fresh nutrient feeding in the reactor (Grothe
and Chisti, 2000; Sayed et al., 2009). The trial and error approach of
nutrient feeding can easily result in under and/or overfeeding of
the substrate which can cause cell starvation or inhibition problems leading to the formation of unwanted products. In such a case,
a mathematical model seems to be a valuable, fast and reliable tool
to predict the various nutrient feeding regimes for fed-batch cultivation in order to maximize the PHB productivity (Xu et al., 2011;
Kaur et al., 2012). The model has the capability to predict the outcome of feeding of different concentrations of fresh substrates on
the culture growth and their rates off-line which not only removes
the limitation of substrate but also overcomes the inhibition problem caused by the substrate (Kaur et al., 2012).
No such attempt has been made in the literature so far on mathematical modeling of growth-associated PHB production using A.
australica. In the present investigation, a batch kinetic model was
proposed and its model (process) parameters were identied by
minimizing the difference between the model simulation and original experimental batch kinetics data acquired from the bioreactor cultivation. The accuracy of the model was demonstrated by
the statistical validity test. The model was then used to simulate
various fed-batch cultivation strategies off-line (on computer)
which would feature maintenance of optimal concentrations of
major nutrients (sucrose and nitrogen) in the reactor for the entire
cultivation process to achieve much higher product concentration
and productivity than the batch cultivation. Therefore the mathematical model was used to identify various nutrients feeding regimes to increase the PHB production.

Y(X+P)S1
YX/S2
qS1
qS2
qp
X
n1

yield of biomass and product based on sucrose (g/g)

yield of biomass based on nitrogen (g/g)
specic sucrose consumption rate (h1)
specic nitrogen consumption rate (h1)
specic product formation rate (h1)
biomass concentration (g/L)
constant in equation

Greek symbol
lmax
maximum specic growth rate (h1)
l
specic growth rate (h1)

2. Methods
2.1. Microorganism and maintenance
In the present investigation, A. australica DSM 1124 (previously
known as Alcaligenes latus DSM 1124) was procured from German
Collection of Microorganisms and Cell Cultures (DSMZ), Germany.
The strain was maintained on nutrient agar (HiMedia, India) slants
at 4 C and sub cultured monthly.
2.2. Medium and culture conditions
In the present study previously statistically optimized media
recipe was used for cell growth and PHB production (Gahlawat
and Srivastava, 2012). The statistical optimization of media was
done by Plackett-Burman Plackett and Burman (1946) and Central
Composite Design protocol using Design Expert (version 5.0.9) software (Stat-Ease Corporation, USA). Trace elements solution (TES)
consisted of: 6 g/L Ammonium Fe (III) citrate, 10 g/L CaCl22H2O,
0.3 g/L H3BO3, 0.2 g/L CoCl26H2O, 0.1 g/L ZnSO47H2O, 0.03 g/L
MnCl24H2O, 0.03 g/L Na2MoO42H2O, 0.02 g/L NiSO47H2O and
0.01 g/L CuSO45H2O. Phosphate solution was autoclaved separately
to prevent precipitation and TES was sterilized by ltration. All the
medium components were mixed aseptically in a laminar ow hood
before inoculation. Medium pH was adjusted aseptically to 7.0 using
2 N NaOH/HCl. Fifty milliliters sterile medium was taken in 250 mL
conical ask and inoculated aseptically with 2 loop full of culture
grown on nutrient agar slant. The conical asks were kept in an orbital shaker at 200 rpm and 33 C (for 48 h) for culture inoculum
development. Before the operation of the bioreactor, another inoculum adaptation step was performed by transferring the growing culture (5% v/v) into 1 L asks containing 200 mL media. The ask
cultivation was continued for 24 h until log phase was reached
and actively growing culture was used to inoculate the bioreactor.
2.3. Preliminary substrate inhibition studies
To study the effects of different concentrations of nutrients (sucrose and nitrogen) on the growth of A. australica preliminary inhibition studies were carried out in 500 mL shake asks containing
100 mL medium. The effect of increasing sucrose concentration
on the growth of A. australica was studied by monitoring the maximum specic growth rate (lmax) in the medium with the varying
concentration of sucrose (ranging from 10 to 120 g/L) while keeping the concentration of rest of the medium components constant
at their optimized values. The effects of varying the concentration
of nitrogen (0.1313 g/L) on the culture growth were also studied

100

G. Gahlawat, A.K. Srivastava / Bioresource Technology 137 (2013) 98105

in a similar manner. Shake asks containing 100 mL medium were

inoculated with 5% inoculum and kept in a rotary shaker at 33 C
and 200 rpm for 24 h. Samples were withdrawn at regular interval
of 2 h and analyzed for biomass.
2.4. Batch-cultivation in the bioreactor
Batch kinetics was studied in a 7 L stirred tank reactor (STR)
(Applikon Dependable Instruments, The Netherlands) containing
4 L statistically optimized medium. The reactor was sterilized at
121 C for 20 min, cooled and then inoculated with 5% inoculum
(v/v). Agitation in the reactor was carried out by using a conventional at blade turbine-type impeller. The temperature was controlled at 33 C using chilled water circulator unit (Julabo FP50,
Germany). Medium pH was maintained at 7.0 by automatic addition of 2 N NaOH/HCl. Air was sparged from the bottom of the reactor using perforated stainless steel L-shaped sparger. The dissolved
oxygen concentration in the bioreactor was determined by an
in situ dissolved oxygen probe (Applisens, The Netherlands). The
dissolved oxygen level was maintained above 30% of the saturation
value by manually adjusting the agitation speed and/or aeration
rate. Fermentation samples were collected at an interval of 3 h
and analyzed for biomass, PHB sucrose and nitrogen. Batch experiments were done in triplicates for the reproducibility of the
results.

nutrients concentration. Residual sucrose concentration in an

appropriately diluted sample was measured by the dinitrosalicylic
acid (DNS) method (Miller, 1959). 1 g/L sucrose solution was used
as a stock solution to prepare the standard curve. Sucrose is a nonreducing sugar; therefore it was rst hydrolyzed into reducing sugars using 2 N HCl solution. After adding 2 N HCl the samples were
boiled in water bath for 10 min and then treated with 2 N NaOH to
neutralize the reaction mixture. For the analysis of residual ammonia nitrogen, Kjeldahl method was used (Horwitz, 1980). The cell
pellet (obtained after centrifugation) was dried at 90 C in a hot
air oven and DCW was calculated. PHB content was determined
by gas chromatography (GC 2010 Shimadzu Co., Japan) using benzoic acid as an internal standard (Riis and Mai, 1988). 1,2-Dichloroethane (DCE) was used as the solvent for the extraction of PHB.
40 mg of dried cells were dissolved in 2 mL of DCE, 2 mL of acidied propanol (1:4::hydrochloric acid:propanol) and 200 mL of
internal standard. The reaction mixture was then kept in an incubator at 100 C for 2 h. PHB was thus depolymerized and transesteried into propyl ester of hydroxybutyric acid using acidied
propanol. After cooling to room temperature, 4 mL of distilled
water was added and the heavier phase (DCEpropanol) was injected into the gas chromatograph column. Each PHB sample was
injected three times into the column and amount of PHB was
determined from averaged values in the present study.

2.5. Modeling and simulation procedure (optimization program)

3. Results and discussion

The batch kinetics data and independently obtained substrates

inhibition data were used for the development of the mathematical
model for the PHB fermentation process. Model parameters were
estimated by minimizing the difference between experimental
observations and model simulations. The optimization program
for kinetic parameters estimation was based on the original algorithm of Rosenbrock (1960) and the computer programs and methodology described by Volesky and Votruba (1992) which was later
on adopted by Kaur et al., 2012.

3.1. Batch cultivation in the bioreactor

2.6. Model-guided fed-batch cultivation strategy

Model-based fed-batch cultivation was carried out in a 7 L stirred tank bioreactor (Applikon Dependable Instruments, The Netherlands) with initial working volume of 2 L. The bioreactor
containing 2 L optimized medium with initial sucrose concentration of 25 g/L was rst cultivated under batch mode. At 20 h when
culture was in actively growing stage, sucrose and nitrogen feeding
(along with other medium components) was started. In case of
decreasing feed rate strategy, both sucrose (S01 = 125 g/L) and
nitrogen (S02 = 2.8 g/L) were fed at a decreasing rate (ranging from
40 to 20 mL/h) using a peristaltic pump. The fresh nutrient feeding
was stopped at 36 h when no residual nutrients were left in the
reactor. In constant feed rate fed-batch cultivation, the sucrose
(125 g/L) and nitrogen (2.8 g/L) were fed at a rate of 100 mL/h
using a peristaltic pump. Nutrient feeding was carried out for
15 h (during 2035 h). After which the fermentation was operated
in a batch mode (3538 h) again for the consumption of residual
nutrients in the reactor.

Batch kinetics of A. australica was studied in detail in STR under

controlled environmental conditions with respect to biomass and
PHB accumulation. Fig 1 illustrates typical time course of the batch
cultivation of A. australica in the bioreactor. From Fig. 1 it becomes
clear that after a lag phase of 3 h, the culture starts growing and
PHB production also started after lag phase because A. australica
produces PHB in a growth-associated manner. At the end of log
phase maximum biomass of 8.71 g/L and PHB concentration of
6.24 g/L was obtained and PHB content (P/X) as high as 72% of
DCW was accumulated in 36 h of cultivation period. Upon GC analysis, PHB was found to be the main product of fermentation without any by- products from A. australica. To the best of our
knowledge from literature survey it was observed that PHB is the
main metabolite or product of sucrose fermentation produced by
this bacterium (Yezza et al., 2007; Grothe and Chisti, 2000; Hrabak,

2.7. Analytical methods

The optical density of the samples was monitored spectrophotometrically at 600 nm (OPTIZEN model 3220UV, Mecasys, Korea)
after suitable dilution of fermentation broth. Biomass was calculated by a standard plot between OD600nm vs. dry cell weight
(DCW). Culture samples were centrifuged at 10,000 rpm for
15 min at 4 C and the supernatant was used to determine residual

Fig. 1. Batch kinetics of PHB fermentation using A. australica. Smooth lines illustrate
the model simulations and data points represent the experimental values. Process
variables show the average values of experimental data at different time intervals
(run in triplicates).

101

G. Gahlawat, A.K. Srivastava / Bioresource Technology 137 (2013) 98105

1992). During the fermentation period 23.3 g/L sucrose concentration was metabolized from its initial concentration of 25 g/L and
only 1.7 g/L sucrose was left unconverted in the culture broth. Culture fermentation resulted in a maximum PHB production rate and
PHB yield of 0.75 g/h and 0.29 g PHB/g sucrose respectively.
3.2. Batch kinetic model development
The batch kinetic data and substrate inhibition data was utilized for the development of a mathematical model for PHB fermentation. The model was based on following assumptions:
(1) Sucrose and nitrogen were the only limiting substrates
affecting growth and PHB production.
(2) There was no process limitation by phosphorus and other
medium components (e.g. trace elements solution) and
these were in excessive amount in the fermentation
medium.
(3) The temperature (33 C) and pH (7.0) of culture broth
remain constant throughout the fermentation.
The mathematical model consisted of following differential
mass balance equations which adequately describes the batch fermentation dynamics of PHB production:

l flmax S1 =S1 K S1 S2 n1 =S2 n1 K S2 n1 

dX=dT lmax S1 =S1 K S1 S2 n1 =S2 n1 K S2 n1 gX

1
2

Fig. 3. Effect of increasing initial sucrose concentration on maximum specic

growth rate of A. australica. Value of KI was determined by tting the sucrose
inhibition data to Eq. (3) using a non-linear regression technique.

Eq. (2) represents that the biomass formation rate was found to
be limited by two major nutrients, sucrose (S1) and nitrogen (S2)
which were expressed by Monod and Sigmoidal kinetics respectively (Fig 2A and B). The approximate values of parameters lmax,
KS1 and KS2 were calculated from the plots of l vs S and thereafter
non linear regression was done to minimize the carefully dened
objective function SSWR (explained in detail in Eq. (13)) to arrive
at the experimental trends of depletion of substrate as described
in Fig 2. From independent inhibition studies, it was established
that the lmax started to decrease after a particular concentration
of sucrose, however complete inhibition of culture growth was
not achieved even at sucrose concentration of 120 g/L (Fig. 3).
Therefore an empirical correlation given by Ierusalimsky (1967)
and later on followed by Aiba and Shoda (1969) was more appropriate to account for observed experimental inhibition by sucrose
as given below:

l lmax K 1 =K 1 S1 

On the other hand, growth inhibition studies in case of nitrogen

demonstrated a slow initial drop in maximum specic growth rate
followed by a rapid decrease to zero (at 13 g/L) with increasing initial nitrogen concentration (Fig. 4). Therefore inhibition of culture
growth by the nitrogen was represented by an empirical correlation suggested by Luong (1985) as below:

l lmax 1  S2 =Sm a 

The exponent a in Eq. (4) describes the type of relationship between lmax and S2. The Sm shows the concentration of nitrogen that
would cause complete inhibition of growth. The values of model
parameters KI, Sm and a were determined by tting the independently obtained substrate inhibition data (shown in Figs. 3 and
4) to Eqs. (3) and (4) respectively. Thereafter, the substrate inhibition terms were also incorporated in the expression for biomass
formation rate and overall differential mass balance equation for
growth was described as follows:

dX=dT flmax S1 =S1 K S1 S2 n1 =S2 n1 K S2 n1 K I =K I S1 

1  S2 =Sm a gX

5
n1

n1

n1

l lmax S1 =S1 K S1 S2 =S2 K S2 K I =K I S1 

Fig. 2. Variation in specic growth rate as a function of concentration of limiting
substrates (A) sucrose and (B) nitrogen. In case of sucrose, Monod model equation
was able to t the experimental data very well, whereas in case of nitrogen
sigmoidal kinetics provides a close t to the experimental data. Smooth lines
represent the tted values.

1  S2 =Sm a 

Sucrose was assumed to be utilized for the growth of biomass,

product formation and maintenance of the cell. Therefore specic
rate of sucrose consumption was as follows:

102

G. Gahlawat, A.K. Srivastava / Bioresource Technology 137 (2013) 98105

Table 1
Values of optimized model parameters of PHB fermentation process of A. australica.

Fig. 4. Effect of increasing initial nitrogen concentration on maximum specic

growth rate of A. australica. Values of parameters Sm and a were determined by
tting the substrate inhibition data to Eq. (4).

qS1 1=Y XP=S1 l mS1 

dS1 =dt 1=Y XP=S1 l mS1 X

Here Y XP=S1 is a lumped parameter representing the yield of

both biomass and product on sucrose. Nitrogen was assumed to
be utilized for the growth of biomass and maintenance only. Following equation represents the specic nitrogen consumption
rate:

qS2 1=Y X=S2 l mS2 

dS2 =dt 1=Y X=S2 l mS2 X

9
10

Here, Y X=S2 represents the yield of biomass with respect to nitrogen. The product formation was observed during growth phase
only, therefore specic rate of product formation was adequately
described by the growth associated component only as follows:

qp K 1 l

11

dP=dt K 1 lX

12

where K1 is the growth associated product formation constant.

Hence, Eqs. 6, 7, 9, and 11 represent the mathematical model equations for PHB fermentation.

Model parameters (units)

Optimized
parameters values

Maximum specic growth rate, lm (h1)

Sucrose afnity constant, Ks1 (g/L)
Nitrogen afnity constant, Ks2 (g/L)
Biomass and product yield on sucrose, 1/Y(X+P)/S1 (g/g)
Maintenance coefcient, ms1 (gsucrose/gcells h)
Biomass yield on nitrogen, 1/YX/S2 (g/g)
Maintenance coefcient, ms2 (gnitrogen/gcells h)
Growth associated product constant, K1 (g/g)
Sucrose inhibition constant, KI (g/L)
Critical nitrogen concentration, Sm (g/L)
Exponent indicating type of relationship between S2
and l, a (dimensionless)
Constant in equation, n1 (dimensionless)

0.41
9.21
0.35
4.11
0.058
0.45
0.003
2.9
90
13
1.102
1.26

difference between the model and experimental value (ymodel  yexpt). Thus the optimized values of kinetic parameters (Table 1)
were obtained by minimizing the objective function (SSWR). The
optimized parameters values were then used to simulate the model
Eqs. 6, 7, 9, and 11. Fig. 1 shows the comparison of the model simulation (smooth lines) and experimental data points and a good
agreement between the two was clearly reected. As seen from
Fig 1 the developed model could successfully simulate the original
experimental kinetics.
3.4. Model validation using statistical test
The proposed model was evaluated for the degree of reliability
using a statistical method suggested by Bard (1974). The mean
residual of each variable Dj was calculated as follows:

Dj

n
1X
Dij
n i1

14

where n is the total number of experimental data points and Dij is

the difference between the experimental value and its corresponding model simulation value. The variance of the error of a residual
(Sj) was then calculated as follows:

Sj

n
1 X
Dij  Dj2 ;
n  1 i1

j 1; m

15

3.3. Estimation of model parameters

The optimal values of parameters were obtained by minimizing
the deviation between the experimental data and model simulations using a non-linear regression technique (Bard, 1974; Volesky
and Votruba, 1992). For the estimation of model parameters a system of differential equations was solved using numerical integration program based on the RungeKutta Method of 4th order.
The optimization program for the direct search of the minimum
of the multivariable function (or objective function) was based
on the algorithm of Rosenbrock (1960) which was later on followed by Patwardhan and Srivastava, 2008 and Kaur et al., 2012.
The minimization criteria used in the program was:

SSWR

n X
m
X
D2ij
i1 j1

W 2j

13

where SSWR represents Sum of the Square of the Weighed Residues; i and j represents the number of experimental data points
and number of variables, respectively; Wj is the weight of each variable (selected as the maximum value of each variable) and Dij is

where m is the number of process variables. For data characterization, the F-test was used to calculate the k which has the F(m,nm)
distribution as below:
m
n  mn X
Dij
n  1m j1 Sj

16

The k-value was calculated to be 6.28 which was less than the
F(4,12) value (observed from F table) for 99% condence level of
the model. Therefore statistical F-test further conrmed the validity of proposed model.
3.5. Model based - fed-batch fermentation
For further improvement in PHB concentration yield and productivity, the batch mathematical model was extrapolated to
fed-batch cultivation by taking the mass balance around the bioreactor and incorporating the dilution terms. The fed-batch model
equations were as follows:

103

G. Gahlawat, A.K. Srivastava / Bioresource Technology 137 (2013) 98105

Fig. 5. Model-based fed-batch cultivation with varying (decreasing) feed rate

strategy. Comparison between model predictions (smooth lines) and experimental
values (data points). (020 h batch: 2036 h fed-batch at decreasing feed rate).
Experimental data points show the average values of the samples.

dV=dt F1 F2 F;

17

D F=V

18

dX=dt flm S1 =S1 K S1 S2 n1 =S2 n1 K S2 n1 

K I =K I S1 1  S2 =Sm a gX  DX

19

dS1 =dt 1=Y XP=S1 l mS1  DS01  S1

20

dS2 =dt 1=Y X=S2 l mS2  DS02  S2

21

dP=dt K 1 lX  DP;

22

D is the dilution rate which is F/V. Feed rate F encompasses the

ow rate for both sucrose, F1 and nitrogen, F2 and V is the working
volume of the bioreactor; S01 and S02 are the inlet concentration of
sucrose and nitrogen respectively in the feed bottle. Off-line computer simulations of the mathematical model equations for the
fed-batch cultivation were done on the computer to identify suitable nutrient feeding regimes for the addition of the disappearing
limiting nutrients: sucrose and nitrogen. Two simple fed-batch cultivation strategies (decreasing and constant feed rate strategy)
indicating signicant growth enhancement and improved PHB
accumulation were selected and experimentally implemented.

3.5.1. Decreasing feed rate strategy

Different decreasing feed rate strategies were simulated off-line
to gure out a particular nutrient feeding regime for enhancement
in the PHB production over batch cultivation. For achieving this
programmed cultivation in the bioreactor, the feed rate of nutrients (sucrose and nitrogen) addition to the reactor was gradually
reduced with time. This ensured maintenance of non-limiting
nutrients concentration during the entire cultivation and also featured low residual substrate availability at the end of fermentation.
Fed-batch cultivation was simulated at a decreasing feed rate of

4020 mL/h with an initial sucrose and nitrogen concentration of

125 and 2.8 g/L respectively. During this strategy, cultivation was
rst carried out batch wise until the sucrose concentration had
reached to a value of 12.5 g/L. At 20 h, when culture was in exponential phase and biomass in the reactor had reached to a concentration of 45 g/L, feeding of sucrose and nitrogen was started at a
feed rate of 40 mL/h using a peristaltic pump (Fig 5). The ow rate
was then reduced by 10 mL/h after every 6 h until the feed rate
reached 20 mL/h. The ow rate of 20 mL/h was maintained for
4 h (3236 h) and the feeding was stopped at 36 h when no nutrients were left in the reactor. This cultivation strategy was then
experimentally validated.
In this type of nutrient feeding prole there was no need of secondary batch cultivation as the end of substrate feeding coincided
with the end of fermentation. This strategy had demonstrated a
considerable improvement in PHB concentration 14.67 g/L and
overall PHB production rate 1.43 g/h respectively as compared to
batch cultivation. Fig. 5 shows the comparison of model simulation
(smooth lines) and experimental data (points) for decreasing feed
rate strategy. It was invariably observed that the experimental
trends of all process variables (biomass, sucrose, nitrogen and
PHB) were in satisfactory agreement with the model predictions.
This showed that the model was satisfactorily robust to elucidate
the observed experimental data other than which was used for
model development and identication.
3.5.2. Constant feed rate strategy
Several feeding strategies were simulated off-line by varying the
sucrose concentration from 25 to 150 g/L and nitrogen concentration from 1.4 to 5.6 g/L in the feed at different time intervals. These
fed-batch feeding strategies were simulated to achieve an
improvement in PHB concentration and productivity over that in
batch cultivation. Table 2 shows the some of the best possible
strategies demonstrating high biomass and PHB accumulation.
Main objective of all the simulations was to ensure high biomass
particularly full of intracellular PHB with minimum increase in
the reactor volume. Among all the possible simulated strategies
for constant feed rate best results were obtained when feeding of
125 g/L sucrose and 2.8 g/L nitrogen was initiated at a feed rate
of 100 mL/h (shown in Table 2, strategy 4). This constant feed rate
strategy was then experimentally implemented to test the accuracy of developed fed-batch model.
The culture was initially cultivated in batch mode and the nutrient (sucrose and nitrogen) feeding was started at 20 h when the
sucrose and nitrogen concentration levels dropped to 12.5 and
0.31 g/L respectively (as mentioned above). The batch kinetics indicated that the culture was in actively growing stage at 20 h, therefore it was expected that supplementation of nutrients at this stage
would result in much faster biomass growth and rapid sucrose
consumption. The model simulation facilitated the development
of feeding strategy such that substrate accumulation is below
inhibitory level and overall culture metabolism in fed-batch cultivation features enhanced biomass and PHB accumulation. The
feeding of sucrose and nitrogen was continued for 15 h (from 20

Table 2
Off-line computer simulations of some best possible constant feed rate strategies.

S. No.

Fed-batch cultivation strategy (Constant feed rate strategies)

1
2
3
4
5

Sucrose
Sucrose
Sucrose
Sucrose
Sucrose

125 g/L and nitrogen* 2.8 g/L feeding at 50 mL/h (during 2035 h)
150 g/L and nitrogen 3.5 g/L feeding at 50 mL/h (during 2442 h)
150 g/L and nitrogen 3.5 g/L feeding at 50 mL/h (during 2035 h)
125 g/L and nitrogen 2.8 g/L feeding at 100 mL/h (during 2035 h)
125 g/L and nitrogen 2.8 g/L feeding at 100 mL/h (during 2442 h)

Fermentation
time (h)

Maximum
Biomass (g/L)

PHB
(g/L)

PHB production
rate (g/h)

38
44
40
38
44

19.14
21.83
23.71
29.71
26.04

14.74
16.98
18.09
22.65
20.18

1.06
1.12
1.24
2.13
1.74

Nitrogen feeding was done in the form of ammonium sulfate. 2.8 g/L nitrogen concentration basically translates to 13.2 g/L ammonium sulfate.

104

G. Gahlawat, A.K. Srivastava / Bioresource Technology 137 (2013) 98105

only one report on mathematical modeling of PHB production by

A. australica and particularly none on the use of mathematical
model for designing nutrient feed strategies for growth associated
PHB production using A. australica. A combined metabolic/polymerization kinetic model was proposed by Penloglou et al.
(2010) to describe the intracellular accumulation prole and
molecular weight distribution of PHB inside the cells. The proposed
metabolic model was found to be satisfactory and could be used as
a valuable tool for the design of various process operating proles
for the production of desired polyhydroxyalkanoates (PHA) grades.
But the use of proposed model for fed-batch operation strategy did
not result in much improvement either in PHB concentration or in
average molecular weight of PHB as compared to batch cultivation.
Several companies have focused on producing PHAs with high productivity and high yield by using different approaches to reduce
the overall costs. In 1970, Imperial Chemical Industries Ltd. (ICI,
UK) initiated the production of PHB on an industrial scale
(200,000 L). The process utilized a mutant of Cupriavidus necator,
NCIB 11599 (Chanprateep, 2010). The fermentation was carried
out in a two-step fed-batch culture with glucose as the sole carbon
source, and phosphate as the limiting element to enhance PHB production. The nal biomass was 100 g/L with a productivity of 2.5 g/
L h. In 1996, the technology was sold to Monsanto Company in the
U.S.A. which stopped the production of PHAs at the end of 1998. In
2001 Metabolix, Inc. had acquired Biopol (copolymer of hydroxybutyrate and hydroxyvalerate) assets from Monsanto (Avrous and
Pollet, 2012). Metabolix biopolymers are made by fermentation
using renewable carbon based feedstocks. In 2008, Metabolix,
Inc. announced the combined production of PHA Bio-based Polymers and Biomass Energy with a target to obtain PHB from switchgrass at a level of 20% of dry-cell weight, 75% of which could be
recovered (Somleva et al., 2008). In 2010, Telles, a joint venture
company formed by the Archer Daniels Midland Company (ADM)
and Metabolix, Inc. opened the rst commercial-scale plant to produce a corn syrup-based PHA resin, Mirel, in Clinton, Iowa, U.S.A.
It was expected that this plant would produce 50,000 tons of resin
per year but this joint venture stopped in January 2012 (Avrous
and Pollet, 2012).
Another industrial process for the production of PHB was developed by Chemie Linz, Austria (later Biotechnologische Forschungsgesellschaft or btF, Austria). The process was based on A.
latus btF-96, which can accumulate PHB up to 80% of dry-cell
weight (Hrabak, 1992; Chanprateep, 2010). The fermentation was
carried out in a one-step fed-batch culture using a mineral-salt
medium with sucrose as the sole carbon source. Although a biomass density of 60 g/L was achieved, the company stopped the production of PHB in 1993 because of the high production cost. The
PHB production and processing technology are now owned by Biomer, Germany (Avrous and Pollet, 2012). Several researchers have
also attempted to increase the PHB concentration and productivity
using A. latus by implementing different feeding strategies during
fed-batch cultivations (Grothe and Chisti, 2000; Sayed et al.,
2009). Grothe and Chisti (2000) have attempted various strategies
of nutrients feeding for fed-batch cultivation using A. latus DSM

Fig. 6. Model-based fed-batch cultivation at constant feed rate strategy. Comparison between model predictions (smooth lines) and experimental values (data
points). (020 h-batch; 2035 h-fed-batch at constant feed rate of 100 mL/h; 35
38 h-batch fermentation). Data points represent the average values of the samples.

to 35 h as shown in Fig 6) at a constant rate during the fed-batch

cultivation in order to maintain the exponential growth of culture
and rates and specic rates of biomass and PHB formation at significantly higher values. On the other hand the typical batch cultivation was associated with the major disadvantage of the limiting
(disappearing) nutrient availability at the time when the biomass
concentration was extremely high (at hour 20) which eventually
featured decreased rates of biomass and product accumulation.
The model predicted that in fed-batch cultivation this peculiar
nutrient limitation and inhibition problem can be easily addressed
and concentrations of biomass and PHB can be signicantly improved. The reactor was again operated in batch mode (secondary
batch) for the duration 3538 h for the complete consumption of
residual substrates. Fig 6 illustrates the comparison of experimental data points with the model simulations (smooth lines) for
fed-batch cultivation using constant feeding strategy. The experimental observations of all the process variables were in close
agreement with the model simulations as shown in Fig 6. Maximum biomass and PHB concentration of 29.71 and 22.65 g/L was
obtained in 38 h as compared to model predicted values of 30.82
and 23.08 g/L respectively. The constant feed rate strategy had
demonstrated 3.6-fold improvement in PHB production. This strategy featured a signicant improvement in PHB production rate
(2.13 g/h) as compared to batch (0.75 g/h) cultivation. At the end
of fed-batch cultivation an overall PHB content of 76% of DCW
was obtained. Table 3 shows the comparison of batch and fedbatch cultivations in terms of nal biomass and PHB concentration.
The mathematical modeling approach not only helps in understanding the process behavior but also facilitates process optimization by designing various nutrient feeding regimes. In fact these
mathematical models are empirical type in nature nevertheless
they appropriately describe the fermentation process limitations
and inhibitions caused by the substrates. To date there has been

Table 3
Model predictions and experimental data of batch and fed-batch cultivation of A. australica.
Cultivation strategy

Batch
Fed-batch
Fed-batch
a
b

Feeding mode

No feeding
Decreasing feed rate (40mlL/h to 20 mL/h for 16 h)
Constant feed rate (100 mL/h for 15 h)

Mod model predicted values.

Exp results of experimental observations.

Total fermentation time (h)

36
36
38

Maximum biomass (g/L)

Maximum PHB (g/L)

Moda

Expb

Mod

Exp

8.9
20.83
30.82

8.71
19.86
29.71

6.51
15.23
23.08

6.24
14.65
22.65

G. Gahlawat, A.K. Srivastava / Bioresource Technology 137 (2013) 98105

1123 but best results were achieved with the continuous constant
rate fed-batch cultivation. This strategy yielded a maximum biomass and PHB concentration of 14.9 and 9.4 g/L respectively with
an overall PHB productivity of 1.15 g/L h. Although PHB productivity in this strategy was high but PHB concentration (9.4 g/L) and
content (63%) were very less which ultimately would lead to increase in the product recovery cost. Thereafter Sayed et al.
(2009) reported two feeding strategies for PHB production by
fed-batch cultivation using different strain, A. latus ATCC 29712.
In case of continuous feeding strategy, it featured a PHB concentration of 8.84 g/L with an overall PHB content of 58.12% using sucrose as carbon source. In yet another fed-batch, Sayed et al.
applied pulsed-feeding strategy and obtained a maximum PHB
concentration and production rate of 12.41 g/L and 0.069 g/g h
respectively. Moreover PHB content as high as 66.15% of DCW
was obtained which was lower than the present study. Hence,
the present work is mainly signicant considering the high PHB
content (76%), concentration (22.65 g/L) and production rate
(2.13 g/h) obtained by A. australica using constant feed rate fedbatch cultivation strategy. The developed mathematical model
helped in designing the various nutrient feeding regimes during
the fed-batch by avoiding the trial and error approach of substrate
feeding and demonstrated signicant improvement in both biomass and PHB production. The developed mathematical model
could also serve as a guiding tool in designing more reactor operating strategies to further improve the PHB concentration and/or
productivity.
4. Conclusions
A batch mathematical model was proposed and its parameters
were identied by minimizing the difference between model simulation and experimental kinetic data. The statistical F test conrmed the validity of proposed model. Applicability of the
developed mathematical model was further demonstrated by predicting and verifying the dynamic fed-batch cultivation strategies
for the enhanced PHB production and accumulation inside the
cells. The constant feed rate strategy demonstrated an enhanced
PHB concentration of 22.65 g/L with a maximum PHB content of
76% of DCW. Mathematical model was extremely useful in predicting the various nutrient feeding regimes during fed-batch cultivation of A. australica using sucrose as carbon source.
Acknowledgements
The Senior Research Fellowship (SRF) award by the Department
of Biotechnology (DBT), Govt of India, New Delhi for the execution
of the project is gratefully acknowledged by one of the authors (Ms
Geeta Gahlawat).
References
Ahn, W.S., Park, S.J., Lee, S.Y., 2000. Production of poly (3-hydroxybutyrate) by fedbatch culture of recombinant Escherichia coli with a highly concentrated whey
solution. Appl. Environ. Microbiol. 66, 36243627.
Aiba, S., Shoda, M., 1969. Reassessment of the product inhibition in alcohol
fermentation. J. Ferment. Technol. 47, 790794.
Avrous, L., Pollet, E., 2012. Biodegrdable Polymers, in: Environmental Silicate
Nano-biocomposites. Springer-Verlag, London, pp. 1339.
Bard, Y., 1974. Nonlinear Parameter Estimation. Academic Press, New York.
Berenjian, A., Mahanama, R., Talbot, A., Bifn, R., Regtop, H., Valtchev, P., Kavanagh,
J., Dehghani, F., 2011. Efcient media for high menaquinone-7 production:
response surface methodology approach. New Biotechnol. 28, 665672.
Chanprateep, S., 2010. Current trends in biodegradable polyhydroxyalkanoates.
Biosci. Bioeng. J. 110, 621632.

105

Choi, J., Lee, S.Y., 1999. Factors affecting economics of polyhydroxyalkanoate

production by bacterial fermentation. Appl. Microbiol. Biotechnol. 51, 1321.
Gahlawat, G., Sengupta, B., Srivastava, A.K., 2012. Enhanced production of poly (3hydroxybutyrate) in a novel airlift reactor with in situ cell retention using
Azohydromonas australica. J. Ind. Microbiol. Biotechnol. 39, 13771384.
Gahlawat, G., Srivastava, A.K., 2012. Estimation of fundamental kinetic parameters
of polyhydroxybutyrate fermentation process of Azohydromonas australica
using statistical approach of media optimization. Appl. Biochem. Biotechnol.
168, 10511064.
Grothe, E., Chisti, Y., 2000. Poly (b-hydroxybutyric acid) thermoplastic production
by Alcaligenes latus: behavior of fed-batch cultures. Bioprocess. Eng. 22, 441
449.
Horwitz, W., 1980. Ofcial Methods of Analysis of the Association of Ofcial
Analytical Chemist, thirteenth ed. AOAC Methods, Washington, DC.
Hrabak, O., 1992. Industrial production of poly-b-hydroxybutyrate. FEMS Microbiol.
Rev. 103, 251256.
Ierusalimsky, N.D., 1967. Bottlenecks in metabolism as growth rate controlling
factor. In: Powel, E.O., Eans, C.G.P., Strange, R.E., Tempest, D.W. (Eds.), Microbial
Physiology and Continuous Culture. HMSO, London, pp. 2333.
Kaur, G., Srivastava, A.K., Chand, S., 2012. Mathematical modeling approach for
concentration and productivity enhancement of 1,3-propanediol using
Clostridium diolis. Biochem. Eng. 68, 3441.
Liu, J.Z., Weng, L.P., Zhang, Q.L., Xu, H., Ji, L.N., 2003. A mathematical model for
gluconic acid fermentation by Aspergillus niger. Biochem. Eng. 14, 137141.
Loo, C.Y., Sudesh, K., 2007. Polyhydroxyalkanoates: bio-based microbial plastics and
their properties. Malays. Polym. J. 2, 3157.
Luong, J.H.T., 1985. Kinetics of ethanol inhibition in alcohol fermentation.
Biotechnol. Bioeng. 27, 280285.
Miller, G.L., 1959. Use of dinitrosalicylic acid reagent for determination of reducing
sugar. Anal. Chem. 31, 426428.
Omar, S., Rayes, A., Eqaab, A., Vob, I., Steinbuchel, A., 2001. Optimization of cell
growth and poly (3-hydroxybutyrate) accumulation on date syrup by a Bacillus
megaterium strain. Biotechnol. Lett. 23, 11191123.
Patwardhan, P., Srivastava, A.K., 2008. Fed-batch cultivation of Wautersia eutropha.
Bioresour. Technol. 99, 17871792.
Penloglou, G., Roussos, A., Chatzidoukas, C., Kiparissides, C., 2010. A combined
metabolic/polymerization kinetic model on the microbial production of poly(3hydroxybutyrate). New Biotechnol. 27, 358367.
Plackett, R.L., Burman, J.P., 1946. The design of optimum multifactorial experiments.
Biometrica 33, 305325.
Ramadas, N.V., Soccol, C.R., Pandey, A., 2010. A statistical approach for optimization
of polyhydroxybutyrate production by Bacillus sphaericus NCIM 5149 under
submerged fermentation using central composite design. Appl. Biochem.
Biotechnol. 162, 9961007.
Riis, V., Mai, W., 1988. Gas chromatographic determination of polyb-hydroxybutyric
acid in microbial biomass after hydrochloric acid propanolysis. J. Chromatogr. A
445, 285289.
Rosenbrock, H.H., 1960. An automatic method of nding the greatest or the least
value of a function. Comput. J. 3, 175184.
Sayed, E.I., Azhar, A., Abdelhady, H.M., Abdel Hafez, A.M., Khodair, T.A., 2009. Batch
production of polyhydroxybutyrate (PHB) by Ralstonia eutropha and Alcaligenes
latus using bioreactor different culture strategies. J. Appl. Sci. Res. 5, 556564.
Somleva, M.N., Snell, K.D., Beaulieu, J.J., Peoples, O.P., Bradley, R., Garrison, B.R.,
Patterson, N.A., 2008. Production of polyhydroxybutyrate in switchgrass, a
value-added co-product in an important lignocellulosic biomass crop. Plant
Biotechnol. 6, 663678.
Volesky, B., Votruba, J., 1992. Mathematical Model Identication, in: Modeling and
Optimization of Fermentation Process. Elsevier, Amsterdam, The Netherlands,
pp. 3854.
Voll, F., Krger, R.L., Castilhos, F.D., Filho, L.C., Cabral, V., Ninow, J., Corazza, M.L.,
2011. Kinetic modeling of lipase-catalyzed glycerolysis of olive oil. Biochem.
Eng. 56, 107115.
Xu, L.J., Liu, Y.S., Zhou, L.G., Wu, J.Y., 2011. Modeling of Fusarium redolens Dzf2
mycelial growth kinetics and optimal fed-batch fermentation for beauvericin
production. J. Ind. Microbiol. Biotechnol. 38, 11871192.
Yamane, T., Fukunaga, M., Lee, Y.W., 1995. Increased PHB productivity by high-celldensity fed-batch culture of Alcaligenes latus, a growth-associated PHB
producer. Biotechnol. Bioeng. 50, 197202.
Yezza, A., Halasz, A., Levadoux, W., Hawari, J., 2007. Production of poly-bhydroxybutyrate (PHB) by Alcaligenes latus from maple sap. Appl. Microbiol.
Biotechnol. 77, 269274.
Zafar, M., Kumar, S., Kumar, S., Dhiman, A.K., 2012. Articial intelligence based
modeling and optimization of poly(3-hydroxybutyrate-co-3-hydroxyvalerate)
production process by using Azohydromonas lata MTCC 2311 from cane
molasses supplemented with volatile fatty acids: A genetic algorithm paradigm.
Bioresour. Technol. 104, 631641.
Zinn, M., Witholt, B., Egli, T., 2001. Occurrence, synthesis and medical applications
of bacterial polyhydroxyalkanoates. Adv. Drug Deliv. Rev. 53, 521.
Zou, X.H., Chen, G.Q., 2007. Metabolic engineering for microbial production and
applications of copolyesters consisting of 3-hydroxybutyrate and medium chain
lengths 3-hydroxyalkanoates. Macromol. Biosci. 7, 174182.