A publication of

CHEMICAL ENGINEERING TRANSACTIONS

The Italian Association
VOL. 27, 2012 of Chemical Engineering
Online at: [Link]/cet
Guest Editors: Enrico Bardone, Alberto Brucato, Tajalli Keshavarz
Copyright © 2012, AIDIC Servizi S.r.l.,
ISBN 978-88-95608-18-1; ISSN 1974-9791

Mathematical Modeling as a Tool to Describe and

Optimize Heterologous Protein Production by Yeast Cells
in Aerated Fed-Batch Reactor
Lucia Paciello*a, Carmine Landia, Elisabetta de Alteriisb Palma Parascandolaa
a
Dept of Industrial Engineering, Università degli Studi di Salerno, via Ponte Don Melillo, 84084 Fisciano, Salerno;
b
Dept of Structural and Functional Biology, Università degli Studi di Napoli “Federico II”, via Cinthia, 80100 Napoli.
lpaciello@[Link]

In this work, two recombinant yeast strains, the prototrophic non-conventional Zygosaccharomyces
bailii [pZ3KlIL-1β] and the auxotrophic Saccharomyces cerevisiae BY4741[PIR4-IL1β], both producing
human interleukin-1β, have been cultured in aerated fed-batch using glucose as limiting substrate. A
mathematical model of the fed-batch reactor has been developed, based on mass balance equations
of the main process variables -biomass, glucose and product- and implemented with kinetic
expressions to explain the yeast behaviour within the aerated fed-batch reactor. In the case of Z. bailii,
the mathematical model evidenced the suitability of the fermentative inoculum with respect to the
respiratory one at the start of the exponential feeding. In the case of the auxotrophic S. cerevisiae
BY4741, the modellistic approach has permitted to highlight a strong deviation from the expected
behaviour and quantify the glucose amount that is spent for maintenance rather than for growth, thus
impairing the outcome of the bioprocess.

1. Introduction
The aerated fed-batch is the cultural system mainly employed in the production of recombinant proteins
with glucose-sensitive yeasts (Mendoza-Vega et al., 1994; Porro et al., 2005). Indeed, fed-batch
provides, through limited supply of one nutrient (generally the carbon and energy source), a suitable
strategy to avoid over-flow metabolism, promote fully respiratory pathway and high yield of biomass
and product of interest. Furthermore, fed-batch mode allows the proliferating biomass to be
accumulated. This is a prerequisite to maximize volumetric productivity i.e. the amount of biomass
and/or product in a given volume within a certain time, which is the most plausible target for
optimization.
Mathematical modeling of bioprocess is an useful tool to describe microbial cell growth, product
formation and to optimize culture conditions. General attempts to model fed-batch processes have
been described (Sinclair et al., 1987). Notwithstanding this, modeling of yeast high-cell density cultures
and optimization of recombinant protein production need to be further developed, considering the
peculiar environment represented by the aerated fed-batch reactor. In this concern, unstructured and
non-segregated models, which describe the rate of growth based on the availability of a single
substrate, may be easier and faster to develop and optimize with respect to the more sophisticated
structured and segregated models.
In this work, an unstructured non-segregated model has been developed to describe the fed-batch
cultures of two glucose-sensitive yeast strains, the non-conventional prototrophic Zygosaccharomyces

79
bailii, and the auxotrophic Saccharomyces cerevisiae BY4741, both engineered for interleukin-1β (IL-
1β) production. In all the experiments, a medium properly formulated has been used, and the feeding
strategy consisted in an exponentially increasing feed covering the entire run, which allowed the yeast
strain to grow at a constant value of specific growth rate. The proposed unstructured and non-
segregated model proved to be able of accurately describing and predicting key aspects of the
fermentations, experimentally observed during yeast proliferation in the fed-batch reactor.

2. Materials and Methods

2.1 Strains
The strain Z. bailii [pZ3KlIL-1β], kindly provided by prof. D. Porro, (UNIMIB-Italy), carried the plasmid
pZ3KlIL-1β containing the human IL-1β gene expressed under the constitutive S. cerevisiae TPI
promoter and, as selective marker, the resistance to geneticin (G418) (Vigentini et al., 2005). The S.
cerevisiae BY4741[PIR4-IL1β] strain was obtained according to Paciello et al. (2010) by transformation
of S. cerevisiae BY4741 (MATa,ura3Δ0, leu2Δ0, met15Δ0, his3Δ1) with the expression vector pIA1,
containing URA3 as selectable marker and the human IL-1β gene functionally fused with a portion of
PIR4 ORF.
2.2 Inocula preparation
Strain samples, from the frozen cultures (-80 °C in 12.5% (v/v) glycerol), was grown at 30 °C in 500 ml
flasks containing 100 ml of a defined mineral medium (Verduyn et al., 1992), pH 5.0. and 1% (w/v)
casamino acids (BD Bacto™ Casamino Acids, BectonDickinson & Co., Sparks, MD 21152 USA) and
-1
made selective with 200 mg L G418 in the case of Z. bailii [pZ3KlIL-1β]. Initial α-D glucose
concentration was 5 and 2% w/v for Z. bailli and S. cerevisiae respectively.
2.3 Fed-batch cultures
Fed-batch cultures have been performed at 30 °C in a 2.0 L working volume of a stirred fermenter,
Bioflo 110 (New Brunswick Scientific). The fermenter initially contained 1 L of the defined mineral
medium above mentioned. The fermenter was inoculated to give an initial O.D.590 of 0.04. As regards
S. cerevisiae, fed-batch culture started after 15 h when glucose in the batch was exhausted, whereas
for Z. bailli, fermentative and respiratory inocula were obtained with 18 and 30 h of batch phase,
respectively. Then, an exponentially increasing feed was applied to allow the biomass to proliferate
-1 -1
with a constant value of specific growth rate (0.13 h and 0.16 h for Z. bailii [pZ3KlIL-1β] and S.
cerevisiae BY4741[PIR4-IL1β], respectively), lower than the 60% of the maximum specific growth rate
of the strain (Enfors, 2001). The feeding solution contained glucose (50% w/v), salts, trace elements,
glutamic acid, vitamins, and casamino acids, the concentration of which was calculated according to
Paciello et al. (2010), taking into account the value of biomass yield for the given amino acid under
aerobic conditions (Pronk, 2002). Oxygen was supplied by air sparging (DOT 30% air saturation).The
culture pH was maintained at 5.0 by automatic addition of 2 N KOH during batch phase and 10% v/v
NH4OH during exponential phase. The foam level in the fermenter was controlled by the automatic
addition of the antifoam B (Sigma Aldrich) (dil. 1:10).
2.4 Determination of biomass, cell viability and specific death rate
Total biomass was determined by optical density (O.D.590) and dry weight. The calibration curve
relating O.D.590 values to biomass density provides a correlation factor of 2.0, and 2.45 O.D.590 per mg
mL-1 for Z. bailii [pZ3KlIL-1β] and S. cerevisiae BY4741 [PIR4-IL1β], respectively.
Viable cell density during fed-batch runs was determined by viable count (in triplicate) on YPD
(1%Yeast Extract, 2% Peptone, 2% w/v Destrose) agar plates incubated at 30 °C for 48 h, and
calculated according to:

() =  ∙ 

(1)


It was assumed that, at the start of feeding (t = 0), all the yeast cells were viable. The specific death
-1
rate (kd,) was evaluated as a first order kinetic constant by plotting the ratio CFU mL / O.D.590 vs. time,
where CFU corresponds to the colony forming units originated by viable cell count.

80
2.5 Analyses
Samples withdrawn from fed-batch cultures were filtered on 0.45 μm GF/A Millipore filters and
analyzed to determine residual glucose, ethanol and IL-1β concentrations in the culture medium
(Paciello et al., 2010). All samples were analyzed in triplicate and the values of standard deviation
obtained varied between 1 and 2%.

3. Mathematical model

The unstructured non-segregated mathematical model was developed on the basis of component
mass balances, starting from the differential equation written below which describes the change with
-1
time of the variable of interest (y, concentration, g L ).
 ()

= () [ − ()] ±  · () (2)

qy is the specific rate of production or consumption referred to the generic variable y. The differential
equation was numerically solved, by Eulero method, starting from given initial values. The first mass
-1
balance developed was that on the biomass x(g L ):
 ()

=− ()
∙ () +  · () (3)

-1 -1 -1
The specific rate qx (h ) includes the specific growth rate (μ, h ) and the specific death rate (kd, h )
(Tab.1). This equation was combined with the mass balance on glucose, the limiting substrate:
 ()

= () [ − ()]−  · () (4)
-1
Glucose specific consumption rate (qs, h ) is the overall specific rate of glucose consumption, including
-1 -1
consumption for both growth (qg, h ) and maintenance (qm, h ). IL-1β production was modeled,
considering that it is a growth-linked product :
 ()

=− ()
∙ () +  · () (5)
-1
where qp (h ) is the IL-1β specific production rate. This latter is represented by the product between
specific growth rate ( ) and product yield coefficient on biomass ( ).

Table.1: Kinetic expressions for fed-batch reactor with recombinant yeast strains
Strain !" !# !$ Validity range

) ∀ t
β]
Z. bailii [pZ3KlIL-1β  ( =  ∙  
*+⁄
,

 − % ( + 
S. cerevisiae BY4741  ∙   0≤t<t1
[PIR4-IL1β]
 ∙ exp[−&( − ' )]-% ( ∙ -[−.( − ' )] + [/( − ' )0 + 1( − ' ) + 2] { ∙ exp[−&( − ' )]} ∙   ≥ '


Model parameter 6 , ' , % , ( , ⁄ ,  7 (Tab.1) values were obtained setting up ad hoc

experiments. Particularly, the μ value was evaluated as the derivative of biomass profile over time; '
was the time at which the specific growth rate and the glucose consumption began to vary; ( value
was evaluated from the ratio between  value and ⁄; % was evaluated by viable count on agar
plate assuming that the death kinetics was a first order kinetics; the yield coefficients ⁄ and 

were determined experimentally from biomass or product obtained per glucose unit consumed. To
assign other parameter (&, ., /, 1, 2,  ) values, simulations were compared to the experimental data,
so as to find a parameter set which gave the best fit of the model to the experimental data. Fitting

81
evaluation was done by minimizing the sum of squared errors between the model and the experimental
data. The exponential profile of flow rate 8() was obtained from the mass balance on limiting
substrate and calculated according to Enfors and Haggstrom (1998) throughout the assumption of a
quasi-steady state on the glucose balance.

3.2 Nomenclature
- / constant for parabolic increase of specific glucose consumption for maintenance for  ≥ ' [h-3]
-2
- 1 constant for parabolic increase of specific glucose consumption for maintenance for  ≥ ' [h ]
- γ constant for parabolic increase of specific glucose consumption for maintenance for  ≥ ' [h ]
-1

-1
- 8 glucose feed rate at time t [L h ]
-1
- θ constant for exponential increase of specific glucose consumption rate for growth [h ]
-1
- % specific death rate [h ]
-1
-  residual glucose concentration at time t [g L ]
-1
-  glucose concentration in the inlet [g L ]
-1
-  specific growth rate [h ]
-1
-  interleukin-1β concentration at time t [g L ]
-1 -1
- ( specific glucose consumption rate for growth [g g h ]
-1 -1
-  specific glucose consumption rate for maintenance [g g h ]
-  time [h]
-1
- t1 time at which the variation of specific growth rate and glucose consumption starts [h ]
-1
- & constant for exponential decrease of specific growth rate [h ]
- 9 initial culture volume [L]
- 9 culture volume at time t [L]
-1
-  viable biomass concentration at time t [g L ]
-1
-  initial viable biomass concentration [g L ]
-1
-  concentration of variable of interest at time t [g L ]
-1
-  concentration of variable of interest in the inlet [g L ]
- ⁄ biomass yield on glucose
-  IL-1β yield on biomass


4. Results
4.1 Modeling of aerated fed-batch culture with Z. bailii [pZ3KlIL-1ββ]
The model developed to describe growth and IL-1β production with Z. bailli [pZ3KlIL-1β] (Table 1)
considered that yeast cells remained viable over the entire fermentation run. This assumption was
supported by the experimental determination of % which resulted zero. The unfitting between the
Residual glucose [g L-1]

20 40
Residual glucose [g L-1]

20 40 120 5
Biomass, residual

A B
glucose [g L-1]

C
Biomass [g L-1]
Biomass [g L-1]

16 4
IL-1β [mg L-1]

30 16 30 90
12 12 3
20 20 60
8 8 2
10 10 30 1
4 4
0 0 0 0 0 0
0 4 8 12 0 4 8 12 0 10 20 30
Time [h] Time [h]
Time [h]

Figure 1: Z. bailii [pZ3KlIL-1β] growing in the aerated fed-batch reactor: respiratory inoculum (A),
fermentative inoculum (B), and run of 24 h with fermentative inoculum (C). Simulation curves
(continuous lines) and experimental data refer to biomass (full rhombus), residual glucose (full
triangle), and product (empty rhombus) concentrations  = 4.9 : ; <' ,  = 0 : ; <' ,  = 0.35 =: ;<' ,
⁄ = 0.37,  = 4 × 10<> .


82
simulation curves and the experimental data regarding biomass and residual glucose (Figure 1A, B)
highlighted that μ of Z. bailli [pZ3KlIL-1β] kept constant at the given value (0.13 h-1) only when the
inoculum came from a fully fermentative batch culture (Figure 1A). During the fed-batch run carried out
with the fermentative inoculum (Figure 1C), a good fitting between simulation curves and experimental
data was observed. Glucose did not accumulate in the medium and ethanol was not produced (data
not shown), indicating that Z. bailii [pZ3KlIL−1β] displayed a fully respiratory metabolism. A cell density
of more than 100 g L-1 and a IL-1β concentration of 4 mg L-1 was achieved after 24 h, with a IL-1β
productivity of 0.15 mg L-1 h-1.

4.2 Modeling of aerated fed-batch culture with S. cerevisiae BY4741[PIR4-IL1β]

Differently from the bioprocess carried out with Z. bailii, the model for S. cerevisiae BY4741 (Table 1)
considered that yeast cells did not remain viable over the entire fermentation run, since kd was
-1
significantly high (kd = 0.028 h ). Figure 2 shows a good agreement existing between experimental
data and simulation curves. It is evident that the specific growth rate (μ) chosen to build up the
exponential feeding profile, was maintained in the time interval 0 ≤ t< t1 and exponentially decreased
when t ≥ t.1 (see Table 1), where t1 corresponds to 7 h of feeding. Maximum of IL-1β productivity was
-1 -1
achieved after 17 h (0.08 mg L h ) of feeding, then it diminished because cell density diminished, due

20 3
Residual glucose [g L-1]

15
2
β [mg L-1]
Biomass,

10

1
IL-1β

0 0
0 10 20
Time [h]

Figure 2: S. cerevisiae BY4741[PIR4-IL1β] growing in the aerated fed-batch reactor: simulation curves
(continuous lines) and experimental data refer to biomass (full rhombus),residual glucose (full triangle),
and product (empty rhombus) concentrations. = 3.1 : ; <' ,  = 0 : ; <' ,  = 1.1 =: ;<' , ⁄ =
0.50,  = 1 × 10<?


to both cell death and culture dilution which prevailed on proliferation. The final biomass was
-1
significantly lower (14 g L ) than that achieved by the prototrophic Z. bailii. Glucose was completely
consumed (Figure 2), without ethanol production (data not shown). In an attempt to describe the
complete glucose consumption coupled with the lowering of specific growth rate, modeling considered
that the specific glucose consumption rate ( ) was split into two components, ( and  (Table 1),
which account for energy and precursors for growth and energy demand for maintenance, respectively.
More precisely, mathematical model considered that ( was constant during the first seven hours of
run, then it decreased exponentially with time, similarly to what happened to the specific growth rate
(Table 1), whereas the specific glucose consumption for maintenance ( ), negligible in the first seven
hours, increased according to a parabolic law (Table 1).
As expected, IL-1β profile follows that of biomass and IL-1β concentration achieved a maximum of 2.4
-1
mg L (Figure 2).

5. Discussion
The unstructured non-segregated mathematical model developed in parallel with the experimental runs
carried out with the two producer strains examined, Z. bailli [pZ3KlIL-1β] and S. cerevisiae BY4741
[PIR4-IlL1β], revealed to be of basic importance for both bioprocess optimization and understanding
the physiological characteristics of the strains. In fact, with Z. bailii [pZ3KlIL-1β], the mathematical

83
model gave the possibility to optimize the bioprocess achieving a significantly high productivity,
provided that a fermentative inoculum is employed. Contrarily to Z. bailii [pZ3KlIL-1β], the viability of
auxotrophic S. cerevisiae BY4741[PIR4-IlL1β] drastically reduced during the run and growth
unavoidably arrested. The mathematical model, based on the viable biomass, lead to the specification
and quantification of the maintenance coefficient ( ), which increased over time. Apparently, unlike
the robust prototrophic Z. bailii, the auxotrophic S. cerevisiae strain used most of the available carbon
source to satisfy an increasing energy demand for maintenance during the fermentation run, thus
impairing the overall biomass/product productivity. This behaviour was ascribed to the peculiar
environment which arose in the aerated fed-batch reactor, able to affect the overall performance of the
auxotrophic strain. In conclusion, the modellistic approach presented in this work revealed of
fundamental importance to understand the strain physiology and consequently to predict its
performance under conditions which resemble those of production.

6. Acknowledgements
This research was supported by the University of Salerno (FARB 2010-2011 to Palma Parascandola).

References
Enfors, S.O., 2001. Baker’s yeast, p. 377-389. In: Ratledge, C. and Kristiansen, B. (ed.) Basic
Biotechnology Cambridge University Press,
Enfors S.O., Haggstrom L. 1998. Bioprocess technology: fundamentals and applications,
Hogskoletryckeriet, Stockholm.
Mendoza-Vega O., Sabatiè J., Brown S.W. 1994. Industrial production of heterologous proteins by fed-
batch cultures of the yeast Saccharomyces cerevisiae. FEMS Microbiol. Rev. 15: 369-410.
Paciello, L., Andrès, I., Zueco, J., Bianchi, M.M., de Alteriis, E., Parascandola, P. 2010. Expression of
human interleukin-1β in Saccharomyces cerevisiae using PIR4 as fusion partner and production in
aerated fed-batch reactor. Ann. Microbiol. 60, 719-728.
Porro D., Sauer M., Branduardi P., Mattanovich D. 2005. Recombinant protein production in yeasts. Mol.
Biotechnol. 31:245-259.
Pronk J.T. 2002. Auxotrophic yeast strains in fundamental and applied research. Appl. Environ.
Microbiol. 68: 2095-2100.
Sinclair C.G., Kristainsen B., Bu’Lock JD. 1987. Fermentation kinetics and modelling Milton Keynes:
Open University Press.
Verduyn C., Postma E., Scheffers W.A., Van Dijken J.P. 1992. Effect of benzoic acid on metabolic fluxes
in yeasts: a continuous-culture study on the regulation of respiration and alcoholic fermentation.
Yeast 8:501-517.
Vigentini I., Brambilla L., Branduardi P., Merico A., Porro D., Compagno C. 2005. Heterologous protein
production in Zygosaccharomyces bailii: physiological effects and fermentative strategies. FEMS
Yeast Res. 5:647–652.

84