Computer simulation for large scale bioprocess design

S.A. Rouf, P.L. Douglas , M. Moo-Young, J.M. Scharer

Department of Chemical Engineering, University of Waterloo, Waterloo, Ont., Canada N2L 3G1
Received 15 February 2000; accepted 19 March 2001

Abstract
Two bioprocess simulators, Aspen BPSTM and SuperPro Designer , were used to simulate and compare two flowsheets for production
of tissue plasminogen activator (t-PA) from Chinese hamster ovary (CHO) cells. The flowsheets were designed to compare the economics
of using serum and serum-free medium in the bioreactor. The base case flowsheet used a medium with 10% serum while the alternative
flowsheet used a serum-free medium.
Affinity chromatography was the major step in the downstream processing of the base case flowsheet. The alternate flowsheet was
designed for two-step cell culture growth and production. Ion-exchange chromatography was considered sufficient for purification of the
low protein medium for this case. At a price of US$ 25/l for serum-free medium, the alternate flowsheet was found to have an return on
investment (ROI), of 115% compared to 65% for the base case flowsheet. 2001 Elsevier Science B.V. All rights reserved.
Keywords: Process simulation; Bioprocess design; Aspen BPSTM ; SuperPro designer

1. Introduction
The advantages of process simulation for development,
evaluation and scale up of bioprocesses have long been
realised [13]. Process simulators offer the opportunity to
shorten the time required for process development. They
allow comparison of process alternatives on a consistent
basis so that a large number of process ideas can be synthesised and analysed interactively in a short time. Simulation
of integrated processes also enables the study of interactions
that exist between upstream and downstream processes. In
this study, two commercial bioprocess simulators, BioProcess SimulatorTM (BPS, [4]) and SuperPro Designer (SPD,
[5]), were assessed for evaluation of large-scale production
processes of a biopharmaceutical from a recombinant mammalian cell culture.
Serum, which is a major component of most cell culture
medium, has the disadvantage of introducing variability to
the different batches of medium. It also increases the possibility of medium contamination with viruses, mycoplasma,
bacteria, fungi, endotoxin, etc. [26]. The presence of other
proteins in the serum complicates purification of the desired
protein and the high cost of quality serum increases the
price of the medium. Therefore, efforts have been directed
Corresponding author. Tel.: +1-519-888-4567/2913;
fax: +1-519-746-4979.
E-mail address: [email protected] (P.L. Douglas).

towards the formulation of a serum-free medium supplemented with growth factors and other essentials. The most
important advantage of a process using serum-free medium
is the simplification of the downstream processing, which
can account for 7080% of the overall cost for high value
products [26]. The economics of using serum-free medium
was investigated by simulating alternate flowsheets.
2. Process description
Production of recombinant tissue plasminogen activator
(t-PA) from CHO cells represents one of the largest scale
productions of a bio product from cell culture [6]. A proteolytic enzyme, t-PA finds application in the treatment of
cardiovascular and cerebrovascular obstructions (i.e. heart
attack and stroke). Commercial production of t-PA is currently carried out by Genentech [7], who developed and obtained FDA approval for its large-scale production in 1987.
Information on the patented large scale production process for t-PA is not available so the proposed flowsheets are
based on bits of information scattered in the technical literature combined with engineering judgement and experience
with other recombinant products [8]. The generation of the
flowsheets for simulation was also largely based on the laboratory scale production methods.
The process information and operating conditions required for simulation were gathered from the literature. The
simulated bioreactors were stirred tank type with the CHO

1369-703X/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 1 3 6 9 - 7 0 3 X ( 0 1 ) 0 0 1 1 2 - 7

230

S.A. Rouf et al. / Biochemical Engineering Journal 8 (2001) 229234

Fig. 1. (a) Process flowsheet for t-Pa production from CHO Cells. (b)
Alternate flowsheet for t-Pa production from CHO Cells.

cells grown in suspension culture, [9]. The batch charge

consisted of 10% Fetal Bovine Serum (FBS), 90% Hams
F-12 medium and 4 mM Glutamine for the base case. The
bioreactor was inoculated with 1000 l of inoculum having a
concentration of 2 105 cells/ml. The contents of the bioreactor were aerated with 0.02 VVM sterile air. The base case
flowsheet is shown in Fig. 1a and the alternate flowsheet,
Fig. 1b, uses separate bioreactors for growth and production. Hams F-12 medium was replaced with HB-CHO basal
mixture and a proprietary supplement (Immucor Canada
Inc.) in the serum-free medium for the production bioreactor of this flowsheet. Since, the actual composition of
the serum-free medium is not known, the concentration of
BSA-like protein in this medium was assumed to be 10%
of that in the serum-containing medium, [9]. Because of
this low background protein concentration, purification with
ion-exchange chromatography was assumed to be sufficient
to reach the desired purity. In the laboratory, purification of
t-PA from serum-free medium using a single ion-exchange
column followed by gel filtration has resulted in a purification factor of 58 and an overall yield of 61% [10].

The permeate and retentate pressure for the membrane filtration units were chosen to yield a transmembrane pressure
of <1 psi, as recommended by Rudolph [11] and Maiorella
et al. [12] for animal cell harvesting. A resistance-in-series
model was chosen as the flux model for the system. Membrane resistance and gel resistance parameters were obtained
from [13,14,27,28].
The affinity ligand recommended for purification of
t-PA is Lysine Sepharose, which has a capacity of >0.6 mg
plasminogen/ml gel (Pharmacia LKB Biotechnology). It
is widely used in laboratory scale purification of t-PA because of it biospecificity [15]. A Langmuir type equilibrium
relationship was assumed for the system since, this relationship has been found satisfactory for many ligand-protein
combinations [16]. The Langmuir adsorption isotherm
constant was assumed to be 120 103 cm3 /g. This value
seems reasonable when compared to the reported value of
162 103 cm3 /g for antibenzenearsonate monoclonal antibody on Sepharose 4B [16]. Simulation was carried out for
a 200 l column having a height to diameter ratio of 0.5 and
a superficial loading velocity of about 40 cm/h.
The ion-exchange column in the alternate flowsheet was
simulated under similar conditions. The ion-exchange resins
used in the laboratory scale purification of t-PA are strong
cations such as SP Sephadex C-SO and Mono S [10,17].
The total ionic capacity of SP Sephadex C-SO is found to
be between 2.0 and 2.6 mmol/g (Pharmacia LKB Biotechnology). An estimated surface area of 1 cm2 /cm3 was used
in the simulation. The operation was simulated at a pH of
4.5, close to the isoelectric point of BSA (5.0) as described
in the laboratory scale purification methods.
Sephadex G-100, G-150 and G-200 are the gel filtration
media used in the laboratory scale purification techniques
for t-PA, with Sephadex G-150 being most popular. Gel
filtration was simulated in a 147 l column with a length to
diameter ratio of 1.5. The sample volume was taken as 5% of
column volume as recommended [18] and the flow velocity
was 16 cm/h.

3. Simulation and results

The process was designed to produce 11,000 g of purified
t-PA per year [19]. To meet this production level, five 6000 l
bioreactors are required (Table 1). One of these five bioreactors with its associated downstream processing units was
simulated using Aspen BPSTM .
The purity requirement of therapeutic pharmaceuticals
like t-PA is very stringent. Downstream processing, therefore, is an important part and represents a dominant cost
element (between 70 and 80%). In general, the final product should be free of infectious agents and should meet the
limits specified for other contaminants such as DNA, cell
and media proteins, endotoxin, etc. [20]. In this simulation,
clearance of only foreign proteins was investigated. Among
these contaminating proteins, the clearance of bovine serum

S.A. Rouf et al. / Biochemical Engineering Journal 8 (2001) 229234

Table 1
Design basis
World-wide cases (per year)
Treatable cases (40%) (per year)
Market penetration (10%) (g/year)
Amount oft-PA per dose (mg)
Production of purified t-PA (g/year)
Recovery yield (%)
Product titer before purification (mg/l)
Plant capacity (l/year)
Over design (%)
Plant capacity (l/year)
Time/one batch (days)
No. of batches/year
Batch volume (l)

231

Table 2
Simulation results for base case flowsheet and alternate flowsheet
2.8 106
1.1 106
110000
100
11000
40
33.5
820000
20
1025000
10
35
5 6000

albumin (BSA) was considered to be critical, since, t-PA

and BSA have very similar molecular weights (about 68,000
and 65,000, respectively). The desired level of retrovirus and
DNA removal were assumed to be achieved by the processes,
because validated processes have shown that a combination
of two chromatographic steps normally results in the desired
reduced level of these contaminants [21].
For simulation, required process information about the
flowsheet and streams, composition of each stream, properties and values of operating variables and parameters of
the different unit operation blocks, were supplied through
input files written according to the specifications outlined
in the BPSTM manual. The BPSTM has a built-in data bank
for components and properties. It allows three types of
components; (conventional, non-conventional and biocomponents). Components such as CHO cells are handled as
non-conventional, and macromolecules such as t-PA and
other proteins as biocomponents. The built-in property estimator BIOENTE calculates enthalpy from heat of formation
and a temperature dependent heat capacity term. BIODEN
calculates density as a second-order polynomial of temperature. The BPSTM also allows addition of user supplied
properties. The simulator carries out mass and energy balances for the different blocks when necessary information
is provided by the user. There are built-in correlations
for estimation of different coefficients. It accepts userdefined kinetics and correlations in the form of FORTRAN
subroutines.
Input information required to carry out simulation with
Aspen BPSTM and SuperPro Designer are different because they use different unit operation models (rigorous
and shortcut models). Most of the information required by
BPSTM was available in the literature or could be estimated
from the built-in correlations. However, some information
required by some of the unit operation blocks in SuperPro
needed to be obtained from experiment. For example,
simulation of the chromatographic column requires the
knowledge of binding percentage and recovery of each component. For these situations, results from BPSTM simulation
were used as input to SuperPro .

Block

Size

Recovery (%)

Purity (%)

Base case flowsheet

Bioreactor
Micro filter
Ultra filter
Affinity Chrom.
Ultra filter
Affinity Chrom.
Ultra filter
Gel filtration

6000 l
80 m2
60 m2
200 l 3
10 m2
85 l
2.5 m2
147 l 6

93.3
83
73.6
94
90.687
94.13
99.997

0.55
0.55
7.65
99.98

99.9977

99.9998

Alternate flowsheet
Growth Bioreactor
Micro filter
Production Bioreactor
Micro filter
Ultra filter
Ion-exchange
Ultra filter
Ion-exchange
Ultra filter
Gel filtration

6000 l
80 m2
5000 l
7 Gm2
50 m2
200 l 2
15 m2
113 l
5 m2
147 l 5

84.5
84.76
83.45
92.37
87.16
96.26
99.997

6.86
6.86

99.93

99.999

99.9998

It was assumed with the two-step fermentation flowsheet,

that the cell concentration in a growth-inducing environment
[22,23] reaches almost twice the concentration reached in the
single step growth-production bioreactor of the base case.
Therefore, the number of production bioreactors was half the
number of growth bioreactors [19] and the number of associated downstream units was half of that required by the base
process. Comparison of the simulated flowsheets will therefore demonstrate the economic effect of using serum-free
medium under these conditions.
From the results it is found that both the flowsheets are
capable of producing t-PA of required purity (99.9998%).
The overall recoveries for the two flowsheets are 45.7 and
46.3%, respectively. These are well above the recovery yield
(40%) assumed in the basis. The batch time for the base
case is 218 h or about 9.08 days. For the alternate flowsheet
the total time required is 221 h or about 9.2 days. These are
also within the batch time of 10 days, assumed in the basis.
Table 2 shows simulation results for the two flowsheets.
The base case requires five sets of the equipment
(Table 2) to meet the total production of 30,000 l. The alternate flowsheet, on the other hand, requires five sets for the
first two steps and then only three sets afterwards because
of the assumption stated in the previous section.

4. Economic evaluation
Aspen BPSTM and SuperPro Designer are both
equipped with economic evaluators. However, the economic
evaluator of Aspen BPSTM , which has been derived from
the chemical process simulator AspenPlusTM , is geared
more towards chemical processes. Its calculation mode is

232

S.A. Rouf et al. / Biochemical Engineering Journal 8 (2001) 229234

rigorous and requires a lot of data from a real project to

make appropriate use of the available options. SuperPro
Designer on the other hand, has an economic evaluator
that is specifically developed for bioprocesses. It is simple
and easy to use. For preliminary screening of these conceptual flowsheets and in the absence of detailed information,
economic evaluation with SuperPro was thought to be
sufficient in this study.
SuperPro generates an equipment list along with the
size and number of each item as a part of economic
evaluation. Then it uses the built-in power law model to
calculate the purchase cost of the equipment (CPE ). The
total (CTC ) and fixed capital (CFC ) costs are then calculated
by multiplying the purchase equipment cost by appropriate
cost factors. For the calculation of the annual operating
cost, SuperPro calculates the utility requirement and the
number of labour-hours. It also calculates the amount of
different input streams required from material balances. If
the treatment cost of the output streams is supplied, it will
also calculate the waste treatment cost for the process.
The flowsheets were compared on the basis of return on
investment (ROI) and Gross Margin (GM). The annual operating cost (AOC) was calculated following the procedure
outlined by Datar and Rosen [24]. Table 3 shows the calculation of the economic parameters.
The alternate flowsheet uses fewer (three) sets of equipment following cell separation compared to the base case
(five). Cell culture equipment accounts for 29% of the total
equipment purchase cost for this flowsheet compared to
14% for the base case (Fig. 2a and b). This results in significant saving (US$ 5.5 million) in the equipment purchase
cost and consequently lowers the capital investment. The
expensive affinity columns of the base case are also replaced
with cheaper ion-exchange columns. However, the purchase

Table 3
Comparative economic analysis (in millions of dollars)

CPE
CFC = 4.6 CPE
CTC = 5.5 CPE
Revenue, R
AOC
Gross profit, GP = R AOC
Net profit, NP (40% tax)
Net cash flow = NP + Dep.
ROI (%)
GM = (GP/R) (%)

Base case

Alternate

14.2
65.32
78.1
176
101
75
45
51
65
43

8.8
40.48
48.4
176
92.9
83
49.86
55.86
115
47

Fig. 2. (a) Breakdown of equipment purchase cost for the base case
flowsheet; (b) breakdown of equipment purchase cost for the alternate
flowsheet.

Fig. 3. (a) Sensitivity to serum-free medium Price; (b) sensitivity to replacement frequency.

S.A. Rouf et al. / Biochemical Engineering Journal 8 (2001) 229234

233

Fig. 4. Sensitivity to serum price.

cost obtained from SuperPro for these columns is almost

the same and no savings in purchase cost were realised from
this source.
The operating cost for the alternate flowsheet is found to
be significantly higher which is mainly due to the cost of the
media required for its eight bioreactors (five growth bioreactors and three production bioreactors). Media cost accounts
for over 94% of the total process chemicals and media cost
and cost of serum-free media accounts for 62% of the total media cost. The cost of serum-free medium is estimated
to be US$ 25/l. It is clear that the profitability of this flowsheet is very sensitive to this price. The replacement cost
of the chromatographic resin for the ion-exchange column,
however, is much cheaper (Fig. 3a and b). The higher media
cost and lower resin replacement cost tend to balance out
the raw material and media cost for this flowsheet.
A sensitivity analysis of the economics of the alternate
flowsheet with respect to serum-free medium cost showed
that, a 50% reduction in this price will improve the ROI
of the process by about 9% (Fig. 4). Therefore, with the
on-going efforts to formulate low cost serum-free medium,
this process can become an even more attractive alternative
to the base case. However, the economics of the process
is also based on the assumption that resulted in fewer sets
of equipment for this flowsheet. The process economics is
found to be a much stronger function of equipment purchase
cost than raw material cost [25] and this limitation should be
emphasised while reaching conclusions about the economics
of these processes.

5. Conclusion
Bioprocess simulators were used to design and evaluate two alternate flowsheets for t-PA production. The flowsheet employing two-stage cell culture and using serum-free
medium in the production bioreactor is found to be an attractive alternative to the base case. This process has both,
lower equipment purchase cost and annual operating cost
compared to the base case. The economics of this flowsheet is sensitive to the assumption of higher cell concentration reached with the two-stage operation. The use of

ion-exchange columns in the absence of background protein

also economises the downstream purification of the product. Simulation helped to quantify the anticipated gain of
the process.
References
[1] L.B. Evans, R.P. Field, Requirements of a biotechnology process
simulator, in: Proceedings of the World Congress II of Chemical
Engineers, 1986, pp. 890893.
[2] C.L. Cooney, D. Petrides, M. Barrera, L. Evans, Computer-aided
design of a biochemical process, ACS Symp. Ser. 39 (1988) 3961.
[3] D. Petrides, Biopro designer: an advanced computing environment
for modeling and design of integrated biochemical processes,
Comp. Chem. Eng. 18 (1994) S621S625.
[4] Aspen Technology Inc., ASPEN BPS Reference Manual, Unit
Operation Models, Vol. 1, Release 9.2, Cambridge, Massachusetts,
1995.
[5] Intelligen, Inc., 1995. SuperPro Designer-Users Guide, Scotch
Plains, New Jersey.
[6] A.A. Lin, R. Kimura, W.M. Miller, Production of t-Pa in recombinant
CHO cells under oxygen-limited conditions, Biotechnol. Bioeng. 42
(1993) 339350.
[7] A.D. Levinson, D. Pennica, W.J. Kohr, G.A. Vehar, D.V. Goeddel,
C.C. Simonsen, Human TPA Production Using Vectors for DHFR,
US Patent 5,268,291, Genentech Inc., San Francisco, CA, 1993.
[8] D. Petrides, E. Sapidou, J. Calandranis, Computer-aided process
analysis and economic evaluation for biosynthetic human insulin
production a case study, Biotechnol. Bioeng. 48 (1995) 529541.
[9] R.L. Dutton, J.M. Scharer, M. Moo-Young, Descriptive parameter
estimation in mammalian cell culture, Cytotechnol 26 (1998) 139
152.
[10] E.K.O. Kruithof, W.-D. Schleuning, F. Bachmann, Human tissue-type
plasminogen activator production in continuous serum-free cell
culture and rapid purification, Biochem. J. 226 (1985) 631.
[11] E.A. Rudolph, et al., Tangential flow filtration systems for
clarification and concentration, in: B.K. Lydersen, N.A. Delia, K.L.
Nelson (Eds.), Bioprocess Engineering Systems, Equipment and
Facilities, Wiley, New York, 1994.
[12] B.L. Maiorella, G. Dorm, A. Canon, D. Harano, Cross-flow
microfiltration of animal cells, Biotechnol. Bioeng. 37 (1990) 121
126.
[13] D.W. Lee, Hollow-fibre cross-flow microfiltration of E. coli suspensions, MSc Thesis, University of Waterloo, Waterloo, Canada, 1989.
[14] D.N. Wood, Harvesting of E. coli cells using cross-flow membrane
filtration, MSc Thesis, University of Waterloo, Waterloo, Canada,
1988.
[15] Y.D. Clonis, Process affinity chromatography, in: J.A. Asenjo (Ed.),
Separation Process in Biotechnology, Marcel Decker, New York,
1990.

234

S.A. Rouf et al. / Biochemical Engineering Journal 8 (2001) 229234

[16] F.H. Arnold, H.W. Blanch, C.R. Wilke, Analysis of affinity

separations I. Predicting the performance of affinity adsorbers, Chem.
Eng. J. 30 (1985) B9B23.
[17] W.F. Prouty, Process chromatography in production of recombinant
products, in: C. Horvath, L.S. Ettre (Eds.), Chromatography in
Biotechnology, ACS Symposium Series, 1993, p. 529.
[18] S.M. Wheelwright, Protein Purification: Design and Scale up of
Downstream Processing, Hanser, New York, 1991.
[19] R.V. Datar, T. Cartwright, C. Rosen, Process economics of animal
cell and bacterial fermentations: a case study analysis of tissue
plasminogen activator, Biotechnology 11 (1993) 349357.
[20] V.R. Anicetti, B.A. Keyt, W.S. Hancock, Purity analysis of protein
pharmaceuticals produced by recombinant DNA technology, Trends
Biotechnol. 7 (12) (1989) 342349.
[21] D. Brady, J. Bonnerjea, C.R. Hill, Purification of monoclonal
antibodies for human clinical use: validation of DNA and retroviral
clearance, in: D.L. Pyle (Ed.), Separations for Biotechnology 2,
Elsevier, London, 1990, pp. 472479.

[22] R. Kimura, W.M. Miller, Effects of elevated pCO2 and/or

osmolability on growth and recombinant t-PA production of CHO
cells, Biotechnol. Bioeng. 52 (1996) 152160.
[23] G.B. Nyberg, R.R. Balcarcel, B.D. Follstad, G. Stephanopoulos, D.I.C. Wang, Metabolism of peptide hamster ovary cells
grown in complex medium, Biotechnol. Bioeng. 62 (1999) 324
335.
[24] R.V. Datar, C.G. Rosen, Downstream process economics, in: J.A.
Asenjo, (Ed.), Separation Process in Biotechnology, Marcel Dekker,
New York, 1990.
[25] S.A. Rouf, Evaluation of Process Simulators for an Integrated
Bioprocess Design, PhD Thesis, University of Waterloo, Waterloo,
Ont., Canada, 1999.
[26] T. Cartwright, Animal Cells as Bioreactors, Cambridge University
Press, Cambridge, UK, 1994.
[27] M. Cheryan, Ultrafiltration Handbook, Technomic Publishing Co.
Inc., Lancaster, Pennsylvania, USA, 1986.
[28] M.I. Tamer, Personal communication, 1995.