Optimization of virus yield as a strategy to improve rabies vaccine production by Vero cells in a bioreactor.

Abstract

To improve rabies vaccine production by Vero cells, we have developed a strategy based on high cell density culture and optimization of virus yield. We have first optimized cell growth in spinner flask using a Taguchi's L8 experimental design. We analyzed the effects of the following factors: initial glucose and glutamine concentrations, Cytodex 1 concentration and the regulation of glucose level at 1 g l(-1). We have also investigated the effect of the following factor interactions: Cytodex 1 concentration/glutamine concentration, Cytodex 1 concentration/glucose concentration and glucose concentration/glutamine concentration. Statistical analysis of the collected data pointed to the initial glucose concentration, the regulation of glucose level at 1 g l(-1) and the interactions between Cytodex 1 concentration/initial glucose concentration and Cytodex 1 concentration/initial glutamine concentration as the parameters that affected cell growth. Using the optimal conditions determined earlier, we have studied Vero cell growth in a 7-l bioreactor and in batch culture, and obtained a cell density level equal to 3.6 +/- 0.2 x 10(6) cells ml-1. Cell infection with rabies virus (LP 2061/Vero strain) at a multiplicity of infection (MOI) of 0.3 using M199 medium supplemented with 0.2% bovine serum albumin (BSA), yielded a maximal virus titer equal to 8 +/- 1.6 x 10(7) Fluorescent Focus Units (FFU) ml-1. We have also studied Vero cell growth in a 7-l bioreactor using recirculation as a perfusion culture mode during cell proliferation step and perfusion for virus multiplication phase. In comparison to batch culture, we reached a higher cell density level that was equal to 10.1 +/- 0.5 x 10(6) cells ml-1. Cell infection under the conditions previously indicated, yielded 14l of virus harvest that had a virus titer equal to 2.6 +/- 0.5 x 10(7) FFU ml-1. The activity of the inactivated virus harvest showed a protective activity that meets WHO requirements.

Full text

Page 1

Journal of Biotechnology 121 (2006) 261–271
Optimization of virus yield as a strategy to improve rabies
vaccine production by Vero cells in a bioreactor
Khaled Trabelsi, Samia Rourou, Houssem Loukil, Samy Majoul, H´ ela Kallel∗
Viral Vaccines Research and Development Unit, Institut Pasteur de Tunis 13, Place Pasteur, BP 74, 1002 Belv´ ed` ere, Tunis, Tunisia
Received 14 January 2005; received in revised form 9 May 2005; accepted 11 July 2005
Abstract
To improve rabies vaccine production by Vero cells, we have developed a strategy based on high cell density culture and
optimization of virus yield. We have first optimized cell growth in spinner flask using a Taguchi’s L8 experimental design.
We analyzed the effects of the following factors: initial glucose and glutamine concentrations, Cytodex 1 concentration and
the regulation of glucose level at 1gl−1. We have also investigated the effect of the following factor interactions: Cytodex
1 concentration/glutamine concentration, Cytodex 1 concentration/glucose concentration and glucose concentration/glutamine
concentration.
Statisticalanalysisofthecollecteddatapointedtotheinitialglucoseconcentration,theregulationofglucoselevelat1gl−1and
the interactions between Cytodex 1 concentration/initial glucose concentration and Cytodex 1 concentration/initial glutamine
concentration as the parameters that affected cell growth.
Using the optimal conditions determined earlier, we have studied Vero cell growth in a 7-l bioreactor and in batch culture,
and obtained a cell density level equal to 3.6±0.2×106cellsml−1. Cell infection with rabies virus (LP 2061/Vero strain) at a
multiplicity of infection (MOI) of 0.3 using M199 medium supplemented with 0.2% bovine serum albumin (BSA), yielded a
maximal virus titer equal to 8±1.6×107Fluorescent Focus Units (FFU) ml−1.
WehavealsostudiedVerocellgrowthina7-lbioreactorusingrecirculationasaperfusionculturemodeduringcellproliferation
step and perfusion for virus multiplication phase. In comparison to batch culture, we reached a higher cell density level that was
equal to 10.1±0.5×106cellsml−1. Cell infection under the conditions previously indicated, yielded 14l of virus harvest that
had a virus titer equal to 2.6±0.5×107FFUml−1. The activity of the inactivated virus harvest showed a protective activity that
meets WHO requirements.
© 2005 Elsevier B.V. All rights reserved.
Keywords: Vero cell; Cytodex 1; Rabies vaccine; Taguchi experimental design; Perfusion culture
∗Corresponding author. Tel.: +216 71 783 022;
fax: +216 71 791 833.
E-mail address: hela.kallel@pasteur.rns.tn (H. Kallel).
1. Introduction
Rabies is a preventable and curable disease.
However, in many parts of the world, particularly
0168-1656/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.jbiotec.2005.07.018

Page 2

262
K. Trabelsi et al. / Journal of Biotechnology 121 (2006) 261–271
in developing countries, rabies continues to cause
damage to human and animal health. According to
the World Health Organization (WHO), there are
approximately 50,000 cases of fatal rabies worldwide
each year (Martinez, 2000).
Safe and potent rabies vaccines produced on differ-
ent cell lines such as MRC5 cells (a human fibroblast
cell line), heteroploid cells like Vero cells (a contin-
uous cell line of vervet monkey kidney) and chicken
embryo cells were developed two decades ago and are
currentlyavailableindevelopedcountries.Today,Vero
cells are considered as a more suitable substrate for
the production of viral vaccines. This cell line presents
several advantages over primary and diploid cell sub-
strates (Prem Kumar et al., 2002). Vero cells can be
used in microcarrier and suspension cultures for large
scale production in bioreactors. Moreover, virus titer
achievedishigherthanthatreachedusingothertypesof
cell substrates (Montagon et al., 1981; Duchene et al.,
1990). These properties greatly facilitate the transfer
of vaccine-producing capability to developing coun-
tries, which is an important goal of the World Health
Organization.Additionally,Verocellshavebeenexten-
sively characterized to evaluate their oncogenic poten-
tial. Several studies have shown that this cell line is
free from oncogenic property and is not presenting
any threat to human health when used as a substrate
forbiologicalproduction(Vincent-Falquetetal.,1989;
Horaud, 1992; Montagnon et al., 1999).
Therefore, Vero cells have been accepted for viral
vaccinationproduction(WHO,1998)andarecurrently
used in the manufacture of rabies (Wiktor et al., 1987),
polio(Montagnonetal.,1984)andJapaneseencephali-
tis (Sugawara et al., 2002) vaccines.
However, most of the developing countries where
rabies is endemic, still use vaccines of the first gener-
ation, which are locally produced and prepared from
brain tissue infected with a fixed strain of rabies virus
(Yusibov et al., 2002; Fu et al., 1997). This type of
vaccinemayinducesevereneurologicalcomplications,
and vaccine failures are more frequent than with cell
culture vaccines (Bagchi, 2004; Meslin and Kaplan,
1996).
Developing countries face different problems to use
efficient vaccine for human protection. Most of these
countries could not import rabies cell culture vaccines
because of their high cost. Furthermore, developing
countries lack local resources to develop their own
cell-culture vaccine. Indeed process development of
such vaccines requires qualified personnel as well as
high cost of goods (microcarriers, fetal calf serum,
etc.).
To lower the economic burden of modern rabies
vaccines produced using cell-culture techniques, sev-
eral studies have shown that modifying the traditional
five-doseEssenregimenforpostexposuretreatmentby
reducing the volume of the injected vaccine and using
more economic regimens such as the Thai Red Cross
intra-dermal regimen, is efficient and reduces the cost
of post exposure prophylaxis treatment (Quiambo et
al., 2005; Goswami et al., 2005).
Optimizing virus yield is an alternative strategy for
the development of a low cost rabies cell-culture vac-
cine. Such strategy improves the productivity of the
upstream process and therefore increases the number
of vaccine doses produced per run.
Several approaches can be used to optimize a
process. For conventional optimization strategies,
each process variable is tested in an independent run.
Thus, to perform an optimization study, it is necessary
to conduct a large number of costly and labor intensive
experiments. However, the Taguchi approach uses a
number of progressive trials to examine simultane-
ously several factors and allow the rapid identification
of those that have major effect (Taguchi, 1986). These
control factors are then used to predict a combination
that will lead to the optimal performance (Aujame et
al., 2000; Han et al., 1998). If these results are satisfac-
tory, then further experiments are unnecessary. This
approach has been successfully applied at laboratory
scale in the optimization of PCR (Cobb and Clarkson,
1994), the enzyme linked immunosorbent assay
(ELISA) (Jeney et al., 1999), protein expression in the
baculovirus system (Burch et al., 1995) and mono-
clonalantibodyproductionbyhybridoma(Kalleletal.,
2002).
The aim of this work is to optimize the productivity
of the upstream process of a rabies vaccine produced
onVerocellsintendedforhumanuse.Forthispurpose,
we optimized rabies virus production by Vero cells
grown on Cytodex 1 in MEM supplemented with fetal
calf serum. To achieve a high virus yield, we have
optimized cell growth using the Taguchi method. We
have also investigated Vero cells growth and rabies
virus production using perfusion and conventional
batch cultures under the optimal conditions deduced

Page 3

K. Trabelsi et al. / Journal of Biotechnology 121 (2006) 261–271
263
from the experimental design that we have carried out
following the Taguchi’s approach.
2. Material and methods
2.1. Cell line
Vero cells obtained from National Vaccine Institute
(Bilthoven, The Netherlands) were used in this study.
Cells were cultivated in MEM+10% FCS at 37◦C and
5% CO2. Once a week, cells were trypsinized and sub-
cultivated at 8×104cellscm−2.
2.2. Virus strain
Pasteur 2061 strain adapted to grow on Vero cells
(LP 2061/Vero) and provided by Institute Pasteur
(Paris, France) was used throughout this study.
2.3. Culture media
MEMandM199mediaweresuppliedbyInvitrogen
(Glasgow, UK). Fetal calf serum (FCS) was provided
by Hyclone (Logan, USA). Bovine serum albumin,
methionine and serine were obtained from Sigma (St.
Louis, USA).
2.4. Microcarrier preparation
Cytodex 1 microcarriers from Amersham Pharma-
cia Biotech (Uppsala, Sweden) were used throughout
thisstudy.Theywerepreparedandsterilizedaccording
to the manufacturers’ instructions.
2.5. Cell growth in spinner flasks
Cultures were carried out in 250ml spinner flasks
containing 200ml of cultured cells, at 37◦C in a 5%
CO2 incubator. Cells were cultivated in MEM sup-
plemented with 10% FCS and enriched with 0.2mM
serine and 0.2mM methionine. The stirring speed was
maintained at 30rpm. The spinners were inoculated
with 2×105cellsml−1. The experiments were car-
ried out in duplicate. Samples were taken daily for
cell count, estimation of microcarriers loading and for
the measurement of glucose, lactate, ammonia and pH
levels.
2.6. Bioreactor cultures
The cultures were performed in a 7-l bioreactor
(BioBraun, Melsungen, Germany) containing 4l as a
working volume, equipped with a marine impeller and
aspinfilter(poresize:75?m)fixedontheaxis.During
cell culture proliferation step, the following conditions
were applied: pH set at 7.2 (regulated by CO2sparging
or addition of NaHCO3at 88gl−1), pO2maintained
at 50% air-saturation by injecting air or pure oxygen
when required, temperature at 37◦C and agitation rate
at 100rpm.
Recirculation culture mode was used during cell
growth phase. This culture mode is a perfused cul-
ture mode where the outlet medium is re-injected into
the feeding bottle that contains three reactor volumes
of culture medium (MEM+5% FCS). Recirculation
was started after 2 days of culture, recirculation rate
was daily increased. Residual glucose level was daily
adjusted to 1gl−1.
For rabies virus production step, pH was main-
tained at 7.4, pO2at 30% air-saturation, agitation
rate at 100rpm and temperature at 34◦C. Once the
temperature reached 34◦C, recirculation and agita-
tion were stopped and microcarriers were left to settle
down.CellswerethenwashedtwicewithM199+0.2%
BSA and infected at an MOI of 0.3 in 50% of the
working volume to increase virus infection efficiency
(Yokomizo et al., 2004; Perrin et al., 1995). Two hours
aftercellinfection,thebioreactorvolumewasadjusted
to 4l and perfusion was restarted at a rate equal to 0.5
reactor volume day−1.
Samples were taken daily to determine the follow-
ingparameters:celldensity,cellviability,microcarriers
load, virus titer, cell infection, glucose, lactate and
ammonia levels.
All bioreactor cultures were repeated twice.
2.7. Cell counting
FivemillilitesofVerocellculturewerewashedthree
timeswithphosphatebufferedsaline(PBS)thentreated
in 5ml of 0.1M citric acid (Sigma) containing 0.1%
crystal violet (Sigma) and 0.1% Triton X-100 (USB,
Cleveland, OH, USA) and incubated at least for 1h at
37◦C.Thereleasednucleiwerecountedusingahema-
cytometer (HBG, Germany).

Page 4

264
K. Trabelsi et al. / Journal of Biotechnology 121 (2006) 261–271
The specific growth rate, µ (h−1) was estimated by
the following equation:
µ =LnXn− LnXn−1
tn− tn−1
where X represents the viable cell density per ml, t the
time points of sampling expressed in hour; the sub-
scripts n and n−1 stand for two succeeding sampling
points.
2.8. Microcarriers load evaluation
SamplesofVerocellsculturewereexaminedmicro-
scopically using an inverted light microscope (Olym-
pus, Japan). The percentage of microcarriers bearing
cells was evaluated by counting the number of micro-
carrierstotallyorpartiallycoveredwithcellsoruncov-
ered. For a statistical significance, 100 microcarrier
beads (three different microscopic fields) were exam-
ined per sample.
2.9. Metabolite analysis
Glucose and lactate concentrations were monitored
by enzymatic assays, using specific assay kits from
Chronolab(Switzerland,Cat.no.:101-0014,101-0040
for glucose and lactate, respectively). Ammonia con-
centration was quantified enzymatically using a kit
from r-Biopharm (Darmstadt, Germany, Cat. no. 111-
12732035).
2.10. Cell infection
Aliquots of 5ml of infected microcarriers cul-
ture were washed three times with PBS then resus-
pended in a final volume of 2ml. Two times 50?l
were put into a microscope slide and let to dry at
room temperature in a laminar flow cabinet. The
slide was then fixed with cold 80% acetone and
stained with fluorescein-labelled antirabies nucleo-
capsid immunoglobulins (Sanofi Diagnostic Pasteur,
Marnes la Coquette, France, Cat. no. 72114).
2.11. Rabies virus titration
Virus titer was determined according to a modi-
fiedRapidFluorescenceFocusInhibitionTest(RFFIT)
(Smithetal.,1973)andexpressedinFluorescentFocus
Units per ml (FFUml−1).
2.12. Virus harvest inactivation
Rabiesvirusharvestswerefirstclarifiedbyfiltration
through8?mfiltertheninactivatedby?-propiolactone
[final dilution 1/4000 (v/v)] (Morgeaux et al., 1993).
2.13. Potency test
The protective activity of inactivated rabies virus
harvests was determined according to the NIH test
(Wilbur and Aubert, 1996). Swiss mice were intraperi-
tonealy immunized at days 0 and 7 and then intracere-
brally challenged using the Challenge Virus Standard
strain (CVS strain). An internal reference calibrated
to the international reference vaccine was also used.
The potency is expressed in International Unitml−1
(IUml−1).
2.14. Taguchi’s orthogonal design
The Taguchi experiments followed an experimen-
tal layout called the L8 array. We have chosen the L8
array because it allows investigation of the effect of
four factors and three interactions at once. The levels
ofthefactorsstudiedandthelayoutoftheL8Taguchi’s
orthogonal array are shown in Tables 1 and 2.
To determine the significance of the effect of each
factor and interaction studied, an ANOVA table was
constructed. The variance of each parameter was cal-
culated. The F-value was used to test the significance
of the factor. A factor is considered as having a major
effectifthecalculatedvalue(F)ishigherthenthetabu-
lated value (F?), determined from the Fisher–Snedecor
table at the p=0.05 level.
Table 1
Levels assigned to the factors studied according to the L8 Taguchi
array for the optimization of Vero cell growth
Level
+
−
6
No
3
4
Cytodex 1 concentration (A) (gl−1)
Regulation of glucose level at 1gl−1(B)
Initial glucose concentration (gl−1) (C)
Initial glutamine concentration (mM) (D)
3
Yes
1
2

Page 5

K. Trabelsi et al. / Journal of Biotechnology 121 (2006) 261–271
265
Table 2
TheL8arrayTaguchilayoutandmaximalcelldensitylevelobtained
for each run
ABC ACD ADCDMaximal cell
density level
(106ml−1)
2.70±0.21
0.99±0.10
1.10±0.15
1.80±0.17
2.03±0.18
2.80±0.23
3.68±0.27
4.10±0.3
Run 1
Run 2
Run 3
Run 4
Run 5
Run 6
Run 7
Run 8
+
+
−
−
+
+
−
−
+
−
−
+
−
+
+
−
+
+
+
+
−
−
−
−
+
+
−
−
−
−
+
+
+
−
+
−
+
−
+
−
+
−
−
+
+
−
−
+
+
−
+
−
−
+
−
+
The factor levels are shown in each experiment, the letters at the top
of the columns indicate the factors/interactions (variables).
3. Results
3.1. Optimization of Vero cell growth in spinner
flasks
To optimize Vero cells growth in spinner flask, we
investigated the effect of the following four factors:
Cytodex 1 concentration (factor A), regulation of
glucose level at 1gl−1during the culture (factor B),
initial glucose concentration (factor C) and initial
glutamine concentration (factor D). Furthermore, we
studied the interactions between Cytodex 1 concen-
tration/initial glucose concentration (AC), Cytodex
1 concentration/initial glutamine concentration (AD)
and initial glucose concentration/initial glutamine
concentration (CD).
For each factor and interaction studied, we calcu-
latedtheeffect,determinedthevarianceandcalculated
the F-values which were compared to the tabulated
values retrieved from the Fisher–Snedecor table (F?).
It should be noted that studying four factors and three
interactions using an L8 orthogonal array, does not
leave any degree of freedom to estimate the residual
error (R).
Table 2 shows the highest cell density level
obtained for each of the experiments conducted
according to the L8 Taguchi array. The lowest cell
yields were reached during experiments 2 and 3.
In these conditions, maximal cell density level was
equal, respectively, to 0.99±0.10×106cellsm−land
1.1±0.15×106cellsml−1. On the other hand, the
highest cell density yield was achieved during experi-
ment 8 and was equal to 4.1±0.3×106cellsml−1.
The calculation of the variance for each factor and
interaction studied shows that the lowest value was
obtained for factor D (initial glutamine concentration).
This indicates that this factor did not have a significant
effect on cell density. Thus, it was neglected and its
effect was considered as a residual error (R). The
calculations based on these observations are presented
in Table 3. The analysis of the ANOVA table (Table 3)
showed that only factor B (glucose level regulation
at 1gl−1), and factor C (initial glucose regulation) as
well as the interactions between Cytodex 1 concentra-
tion/initial glutamine concentration (AC) and Cytodex
1 concentration/initial glucose concentration (AD)
have a significant effect on the cell growth.
The cell growth was related to the significant vari-
ables (factors and interactions) using the following
linear equation:
celldensity(cellsml−1)
= X + hBB + hCC + hACAC + hADAD
Table 3
Analysis of the different factors and interactions effects on Vero cell growth
Factor/interaction effectVariance
S2
S2
S2
S2
S2
S2
S2
Degree of freedom
FF?
161
161
161
161
161
161
hA=−269500
hB=344500
hAD=394500
hC=−752000
hAC=468000
hD=−23000
hCD=274500
A=5.81×1011
B=9.49×1011
AD=1.24×1012
C=4.52×1012
AC=1.75×1012
CD=6.02×1011
R=4.23×109
1
1
1
1
1
1
1
137.29
224.35
294.19
1069
414.03
142.43
hI=?(Yj×level of I)/8, Variance I=8×(hI)2; I indicates the factor or the interaction studied. Yjindicates the viable cell density obtained for
the experiment j. The level used for each factor at the corresponding experiment is indicated Table 2.

Page 6

266
K. Trabelsi et al. / Journal of Biotechnology 121 (2006) 261–271
whereXistheaveragecelldensityobtainedintheeight
experiments. hB, hC, hACand hADstand, respectively,
for the effects of factors B, C and the interactions
between factors A and C, A and D.
Therefore to maximize cell density level, fac-
tor A (Cytodex 1 concentration) should be set at
level—(6gl−1), factor B (glucose regulation at
1gl−1)+(i.e. glucose level maintained at 1gl−1)
factor C (initial glucose concentration) at level—(i.e.
3gl−1) and factor D (initial glutamine concentration)
at level—(4mM). The selection of the optimal level
for each factor depends upon its effect. For instance,
the effect of factor B (maintaining glucose level at
1gl−1) (hB) is equal to 344,500 (Table 3), therefore
to maximize cell density level the term “hBB” should
have a positive value, then factor B should be set at
level +.
3.2. Bioreactor cultures
3.2.1. Batch mode
WestudiedVerocellsgrowthinbatchcultureandin
a 7-l bioreactor, under optimal conditions determined
according to the Taguchi model.
Tostudyrabiesvirusproduction,cellswereinfected
with the virus strain LP 2061/Vero, at day 4 at an MOI
of 0.3.
Data shown in Fig. 1A indicate that the cells
exhibit an exponential cell growth with an aver-
age specific growth rate equal to 0.0207h−1. Micro-
scopic examination of Cytodex 1 microcarrier beads
shows that 24h post inoculation, at least 6–10
cells are attached to each microcarrier bead. Four
days after the start of the culture, 92% of the
microcarriers beads were totally covered with cells
(Fig. 2A).
With regard to by-products production, lac-
tate and ammonia levels were high, they reached
23.8±2.5mM and 3.76±0.3mM, respectively. At
day4,wereachedacelldensitylevelof3.6±0.2×106
cellsml−1.
After cell infection, cells continued to grow for
one more day, then we observed a sharp decrease
of cell density with a concomitant secretion of the
virus into the culture medium. The highest virus
titer was reached four days post infection and was
equal to 8±1.6×107FFUml−1(Fig. 1B). Daily
monitoring of cell infection indicates that at day 4
Fig. 1. Kinetics of rabies virus production by Vero cells grown in
batch culture mode in a 7-l bioreactor under the optimal conditions
determined according to the L8 Taguchi array (Cytodex 1: 6gl−1,
initialglucoselevel:3gl−1,initialglutaminelevel:4mM,regulation
ofglucoselevelat1gl−1).Evolutionof(A)celldensity,lactate,glu-
cose, ammonia concentration and (B) virus titer, throughout culture
time. (
) Indicates cell infection.
post infection, almost 100% of the cells were infected
(Fig. 3A).
3.3. Recirculation/perfusion mode
Cells were grown on Cytodex 1 at 6gl−1in MEM
enriched with glucose (3gl−1) and glutamine (4mM)
andsupplementedwith0.2mMserine,0.2mMmethio-
nine and 10% FCS during batch growth and 5% during
recirculation.
Fig. 4 shows that starting recirculation at day
2 enhanced cell growth. Furthermore, continuous
increaseoftherecirculationrateasindicatedinFig.4C,
resulted in an exponential cell growth. At day 6, cell

Page 7

K. Trabelsi et al. / Journal of Biotechnology 121 (2006) 261–271
267
Fig. 2. Monitoring of Cytodex 1 microcarriers load throughout culture time during Vero cells growth in a 7-l bioreactor using batch (A) and
recirculation (B) as culture mode. Cells were grown under the optimal conditions determined according to the L8 Taguchi array (Cytodex 1:
6gl−1, initial glucose level: 3gl−1, initial glutamine level: 4mM, regulation of glucose level at 1gl−1). For recirculation culture, cells were
grown in batch mode then in recirculation using MEM+5% FCS. Cells were observed with an inverted light microscope without staining
at 140× magnification (batch culture) and at 80× magnification, after staining with hematoxylin at 0.2% (recirculation culture). The arrows
indicate bridges of cells between Cytodex 1 beads.
density level was equal to 10.1±0.5×106cellsml−1.
As compared to batch culture, metabolites production
was lower, lactate and ammonia concentrations
were equal to 14.35±1.5mM and 2.64±0.2mM,
respectively.
Daily monitoring of microcarriers load indicates a
homogenous cell distribution 1-day post inoculation,
i.e.almost100%ofmicrocarriersarebearing6–10cells
per bead. At day 4, the percentage of microcarriers
totally covered with cells is equal to 95. From days
4 to 6, cells grow on microcarriers by forming multi-
layers. We also notice bridges formed of cells between
microcarriers beads (Fig. 2B).
At day 6, cells were infected with the virus strain
LP 2061/Vero at an MOI of 0.3, in M199+0.2% BSA.
Virus production step was performed using perfusion
culture mode.
During the virus production phase, cell behavior
was comparable to that observed during the previ-
ous culture. Cells grew for a further day then cell
density level decreased sharply while virus titer
increased. Seven days post infection, almost 100%
of the cells were infected resulting in a maximum
virus titer of 7.7±1.5×107FFUml−1(Figs. 3B
and 4B). The culture was stopped immediately
after we observed a decrease of the percentage of
infected cells, at the end of the culture virus titer
was equal to 3±0.6×107FFUml−1. Virus harvests
obtained from day 3 post infection until the end of
the culture were pooled, we obtained 14l of virus

Page 8

268
K. Trabelsi et al. / Journal of Biotechnology 121 (2006) 261–271
Fig. 3. Cell infection as evaluated by immunofluorescence, during rabies virus production by Vero cells. (A) Cytodex 1 beads 4 days post
infection, cells were grown in batch mode in a 7-l bioreactor under the conditions indicated in Fig. 2; (B) Cytodex 1 beads 7 days post infection,
cells were grown in a 7-l bioreactor using recirculation/perfusion culture modes under the conditions indicated in Fig. 4. The arrows indicate
infected cells. Cells were observed with an inverted fluorescence microscope at a 140× magnification.
suspension at a virus titer equal to 2.6±0.5×107
FFUml−1.
3.4. Potency test
The pooled virus harvests were clarified and inac-
tivated by beta propiolactone. The inactivated harvests
showed an activity of 4UIml−1, this result indicates
that this product fulfils WHO requirements for vaccine
potency.
4. Discussion
Cell culture products such as viral vaccines, ther-
apeutic proteins and monoclonal antibodies are cur-
rently produced either using heterogeneous systems
such as roux flasks and roller bottles or homogenous
systems like bioreactors.
For biologicals produced using adherent cells, the
use of microcarrier technology is widely spread at an
industrialscale.Mostoftheprocessesarecarriedoutin
aseveralhundred-literbioreactorwithalowconcentra-
tion of microcarrier beads (3–5gl−1) and using batch
culture mode (Berry et al., 1999; Moran, 1999; Werner
etal.,1992;Montagnonetal.,1984).Increasingmicro-
carrier concentration results normally in a higher cell
density. Nevertheless, processes based on the use of a
high microcarrier concentration face different techni-
calhurdlessuchasacellgrowthlimitationbydissolved
oxygen. Furthermore, special design of culture media
is also required for high microcarrier concentration
processes. Indeed the medium should supply all the
nutrientsneededwithoutimpairingcellgrowth.Partic-
ular consideration should be given to osmolarity since
this parameter is considered as the most important
physical property of cell culture media and could have
a detrimental effect on cell growth (Freshney, 1994).
Osmolarity level should be in the range of 280–320
mOsmkg−1.
In this work, we demonstrated that by using the
Taguchiapproachbasedontheuseofalimitednumber
of experiments, we showed that initial glucose con-
centration as well as regulation of glucose level at
1gl−1during cell growth, are the most critical factors
that affect cell growth whereas Cytodex 1 and initial
glutamine concentrations did not have any significant
effect on cell growth. These results are in agreement
with those reported by Mendonc ¸a et al. (2002) who
showed that regular feeding of Vero cell culture with
glutamine and galactose had prevented apoptosis and
resulted in a high cell density level. We also showed
that interactions between microcarrier concentration
and initial glucose or glutamine concentrations, had

Page 9

K. Trabelsi et al. / Journal of Biotechnology 121 (2006) 261–271
269
Fig. 4. (A–C) Vero cells growth and rabies virus production in a
7-l bioreactor under the optimal conditions determined according
to the L8 Taguchi array (Cytodex 1: 6gl−1, initial glucose level:
3gl−1, initial glutamine level: 4mM, regulation of glucose level at
1gl−1). During cell growth phase, cells were first grown in batch
mode in MEM+10%FCS then in recirculation mode, in MEM+5%
FCS. Virus production was carried out in M199+0.2% BSA using
perfusion culture mode. Cells were infected at day 6 at an MOI of
0.3. (
infection.
) Indicates the start of recirculation, ( ) indicates cell
a significant effect on cell growth. Hence, it appears
that medium enrichment with substrates like glucose
and glutamine enhances cell growth unless there is
a limitation by the available surface area and there-
fore by microcarrier concentration. In the case of cell
growth limitation by the available surface area, excess
of nutrients such as glutamine and glucose, would be
convertedtoby-products:mainlylactateandammonia,
that could accumulate to toxic levels and hamper cell
growth (Butler, 1987; Doverskog et al., 1997). There-
fore,microcarrierconcentrationandsubstratelevelsare
crucial parameters that are closely related and which
are of utmost importance for a successful microcarrier
culture.
In this study, we also showed that increasing ini-
tial glucose and glutamine concentration to 3gl−1and
4mM, respectively, while using 6gl−1of Cytodex 1,
we obtained 4±0.3×106cellsml−1during a spin-
ner culture. This performance is higher than the data
reportedbyMendonc ¸aandPereira(1995)whoreached
2.7×106cellsml−1during Vero cells on 10gl−1
Cytodex 1, in a benchtop bioreactor and using batch
culture mode.
Our data showed that batch growth of Vero cells in
a 7-l bioreactor under optimal conditions determined
according the Taguchi model, resulted in a cell density
level equal to 3.6±0.2×106cellsml−1while recir-
culation culture under the same conditions, yielded
10.1±0.5×106cellsml−1.
Compared to batch culture, by-products concentra-
tions were lower. Indeed recirculation culture allows
continuous renewal of the medium that results in a
lower level of the metabolites accumulated in the
medium. These data correlate with those that we
obtained in a previous study (Trabelsi et al., 2005)
where we investigated the effect of different culture
modes on Vero cell growth.
Furthermore, cell density level reached during
recirculation culture was higher than the level
reported in a previous work (Trabelsi et al., 2005).
We obtained 4.3×106cellsml−1when Vero cells
were grown in a 2-l bioreactor, using recirculation
as a culture mode and 3gl−1Cytodex 1. Hence, by
doubling microcarrier concentration and modifying
medium composition, cell density level was 2.3-fold
higher.
Ourdataalsoindicatethatcelldensitylevelobtained
during the recirculation culture was significantly

Page 10

270
K. Trabelsi et al. / Journal of Biotechnology 121 (2006) 261–271
higher then that reported by Mendonc ¸a and Pereira
(1998) and Aycardi (2002) who reached 6×106and
12×106cellsml−1, respectively, although they used
a higher concentration of Cytodex 1: 10 and 25gl−1,
respectively.
Aycardi (2002) also reported a specially designed
media formulation to achieve such high cell density
concentration; this step was necessary for an efficient
control of pH and to avoid osmolarity increase at a
level that impairs cell growth. In our case, no partic-
ular medium design was needed. During recirculation
culture, we had only increased glucose and glutamine
concentrations of the feeding medium to 3gl−1and
4mM,respectively.Suchmodificationwouldnotresult
in a dramatic increase of osmolarity. Under these con-
ditions and compared to the data obtained by Aycardi
(2002), we reached a slightly lower cell density level
although we used a microcarrier concentration 4.16-
fold lower.
Regarding virus production phase, the use of
perfusion as a culture mode, increases the total
amount of virus harvested compared to batch culture.
Under perfusion culture conditions, the total amount
of virus recovered was 1.7-fold higher than batch
culture.
Achieving a high cell density level at a good
metabolic activity also enhances virus titer as well as
total virus yield recovered. In this study, we showed
that pooling the harvests obtained from day 3 post
infection until the end of the culture, resulted in
14l of virus at a titer of 2.6±0.5×107FFUml−1.
Furthermore, the inactivated pooled crude harvests
showed an activity equal to 4IUml−1. Therefore, we
increased the productivity of the upstream process
that we have developed as compared to our previ-
ous data (Trabelsi et al., 2005) where we obtained
3.25l of virus harvests that had an activity equal to
2.9IUml−1. We also achieved a higher productiv-
ity than Mendonc ¸a et al. (1993) who showed that
cell supernatants harvested during Vero cell infec-
tion by the PV rabies virus strain, had a potency
equal to 0.65–0.9IUml−1. Hence, we also expect to
increase the number of vaccine doses that would be
produced and therefore decrease the cost of vaccine.
Nevertheless, we need to develop a process for virus
purification (downstream process) to obtain a vac-
cine intended for human use that fulfils all of WHO
requirements.
Acknowledgment
The authors are grateful to Dr. Dahmani Fathallah
for critical reading and revision of the manuscript.
References
Aujame, L., Seguin, D., Droy, C., Hessler, C., 2000. Experimental
design optimisation of filamentous phage transfect mammalian
cells by cationic lipids. Biotechniques 28, 1202–1208.
Aycardi,E.,2002.Producinghumanrabiesvaccinesatlowcost.Gen.
Eng. News 22, 52.
Bagchi, S., 2004. Preventable rabies deaths because of wrong vac-
cine. Lancet Neurol. 3, 639.
Berry, J.M., Barnab´ e, N., Coombs, K.M., Butler, M., 1999. Produc-
tionofreovirustype-1andtype-3fromVerocellsgrownonsolid
and macroporous microcarriers. Biotechnol. Bioeng. 62, 12–19.
Burch, G.J., Ferguson, C.H., Catwright, G., Kwong, F.Y., 1995.
Application of Taguchi experimental design to the optimiza-
tion of baculovirus expression system. Biochem. Soc. Trans. 23,
107S.
Butler, M., 1987. Growth limitations in microcarrier cultures. In:
Fiechter, A. (Ed.), Advances Biochem. Eng. Springer-Verlag,
Berlin, Heidelberg, pp. 58–83.
Cobb, B.D., Clarkson, J.M., 1994. A simple procedure for optimisa-
tionthepolymerasechainreaction(PCR)usingmodifiedTaguchi
methods. Nucleic Acids Res. 22, 3801–3805.
Duchene, M., Peetermans, J., D’Hondt, E., Harford, N., Fabry, L.,
Stephenne, J., 1990. Production of polio vaccines: past, present
and future. Viral Immunol. 3, 243–272.
Doverskog, M., Ljunggren, J.,¨Ohman, L., H¨ aggstr¨ om, L., 1997.
Physiology of cultured animal cells. J. Biotechnol. 59, 10–115.
Freshney, R.I., 1994. Culture of Animal cells, third ed. Wiley-Liss,
New York.
Fu, Z.F., Dietzschold, B., Plotkin, S.A., Rupprecht, C.E., 1997.
Improved vaccines against rabies. In: Levine, M.M., Woodrow,
G.C.,Kaper,J.B.,Cobon,G.S.(Eds.),NewGenerationVaccines.
Marcel Dekker, New York, pp. 607–617.
Goswami, A., Plun-Favreau, J., Nicoloyannis, N., Sampath, G., Sid-
diqui, M.N., Zinsou, J.A., 2005. The real cost of rabies post-
exposure treatments. Vaccine 23, 2970–2976.
Han,J.J.,Yang,T.H.,Rhee,J.S.,1998.Optimizationofreactionvari-
ables for sucrose monoester production using lipase in solvent
freesystembyTaguchi’smethod.Biotechnol.Tech.12,295–299.
Horaud, F., 1992. Absence of viral sequences in the WHO-Vero cell
bank: a collaborative study. Dev. Biol. Stand. 76, 43–46.
Jeney, C., Dobay, O., Lengyel, A., Adam, E., Nasz, I., 1999. Taguchi
optimization of ELISA procedures. J. Immunol. Methods 223,
137–146.
Kallel, H., Zairi, H., Rourou, S., Essafi, M., Barbouche, R., Dellagi,
K., Fathallah, D.M., 2002. Use of Taguchi’s methods as a basis
to optimize hybridoma cell line growth and antibody production
in a spinner flask. Cytotechnol. 39, 9–14.
Martinez, L., 2000. Global infectious disease surveillance. Int. J.
Infect. Dis. 4, 22210–22228.

Page 11

K. Trabelsi et al. / Journal of Biotechnology 121 (2006) 261–271
271
Mendonc ¸a, R.Z., Arrozio, S.J., Antoniazzi, M.M., Ferreira Jr.,
J.M.C., Pereira, C.A., 2002. Metabolic active-high density Vero
cell culture on microcarriers following apoptosis prevention by
galactose/glutamine feeding. J. Biotechnol. 97, 13–22.
Mendonc ¸a, R.Z., Ioshimoto, L.M., Mendonc ¸a, R.M.Z., De-Franco,
M., Valentini, E.J.G., Bec ¸ak, W., Raw, I., Pereira, C.A., 1993.
Preparation of human rabies vaccine in Vero cell culture using a
microcarrier system. Braz. J. Med. Biol. Res. 26, 1305–1317.
Mendonc ¸a, R.Z., Pereira, C.A., 1995. High density Vero cell cul-
ture on microcarriers in a cell bioreactor. Bioprocess Eng. 12,
279–282.
Mendonc ¸a, R.Z., Pereira, C.A., 1998. Cell metabolism and medium
perfusion in Vero cell cultures on microcarriers in a bioreactor.
Bioprocess Eng. 18, 213–218.
Meslin, F.X., Kaplan, M.M., 1996. General considerations in the
production of brain-tissue and purified chicken embryo vaccines
for human use. In: Meslin, F.X., Kaplan, M.M., Koprowsky, H.
(Eds.), Laboratory Techniques in Rabies. WHO, Geneva, pp.
360–368.
Montagnon, B., Vincent-Falquet, J.C., Fanget, B., 1984. Thousand
liter scale microcarrier culture of Vero cells for killed polio virus
vaccine. Dev. Biol. Stand. 55, 37–42.
Montagnon, B.J., Vincent-Falquet, J.C., Saluzzo, J.F., 1999. Experi-
ence with Vero cells at Pasteur M´ erieux Connaught. Dev. Biol.
Stand. 98, 137–140.
Montagon, B.J., Fanjet, B., Nicolas, A.J., 1981. The large-scale cul-
tivation of Vero cells in microcarrier culture for virus vaccine
production: preliminary studies for killed poliovirus vaccine.
Dev. Biol. Stand. 47, 55–64.
Moran, E., 1999. A microcarrier-based cell culture process for
the production of a bovine respiratory syncytial virus vaccine.
Cytotechnol. 29, 135–148.
Morgeaux, S., Tordo, N., Gontier, C., Perrin, P., 1993. ?-
propiolactonetreatmentimpairsthebiologicalactivityofresidual
DNA from BHK-21 cells infected with rabies virus. Vaccine 11,
82–90.
Perrin, P., Madhusudana, S., Gontier-Jallet, C., Petres, S., Tordo,
N., Merten, O.W., 1995. An experimental rabies vaccine pro-
duced with a new BHK-21 suspension cell culture process: use
ofserum-freemediumandperfusion-reactorsystem.Vaccine13,
1244–1250.
PremKumar,A.A.,BhaskaraRao,Y.U.,Joseph,A.L.W.,Mani,K.R.,
Swaminathan, K., 2002. J. Biosci. Bioeng. 94, 375–383.
Quiambo, B.P., Dimaano, E.M., Ambas, C., Davis, R., Banzhoff, A.,
Malerczyk, C., 2005. Reducing the cost of post-exposure rabies
prophylaxis: efficacy of 0.1 ml PCEC rabies vaccine admin-
istrated intradermally using the Thai Red cross post-exposure
regimen in patients severely exposed to laboratory-confirmed
rabid animals. Vaccine 23, 1709–1714.
Sugawara, K., Nishiyama, K., Ishikawa, Y., Abe, M., Sonoda, K.,
Komatsu, K., Horikawa, Y., Takeda, K., Honda, T., Kuzuhara,
S., Kino, Y., Mizokami, H., Mizuno, K., Oka, T., Honda, K.,
2002. Development of Vero cell-derived inactivated Japanese
encephalitis vaccine. Biologicals 30, 303–314.
Smith, J.S., Yager, P.A., Baer, G.M., 1973. A rapid tissue culture test
for determining rabies neutralizing antibody. In: Kaplan, M.M.,
Koprowsky, H. (Eds.), Laboratory Techniques in Rabies. WHO,
Geneva, pp. 354–357.
Taguchi, G., Asian Productivity Organization, 1986. Introduction to
quality engineering. UNIPUB, New York.
Trabelsi, K., Rourou, S., Loukil, H., Majoul, S., Kallel, H., 2005.
Comparisonofvariousculturemodesfortheproductionofrabies
virus by Vero cells grown on microcarriers in a 2-l bioreactor.
Enzyme Microb. Technol. 36, 514–519.
Vincent-Falquet,J.C.,Peyron,L.,Souvras,M.,Moulin,J.C.,Tektoff,
J.,Patet,J.,1989.QualificationofworkingcellbanksfortheVero
cell line to produce licensed human vaccines. Dev. Biol. Stand.
70, 153–156.
Werner, R.G., Walz, F., Noe, W., Konrad, A., 1992. Safety and eco-
nomic aspects of continuous mammalian cell culture. J. Biotech-
nol. 22, 51–68.
Wilbur, L.A., Aubert, M.F.A., 1996. The NIH test for potency. In:
Meslin, F.X., Kaplan, M.M., Koprowsky, H. (Eds.), Laboratory
Techniques in Rabies. WHO, Geneva, pp. 360–368.
Wiktor, T.J., Fanget, B.J., Fournier, P., Montagnon, B.J., 1987. Pro-
cess for the large scale production of rabies vaccine. US Patent
No. 4,664,912.
World Health Organization, 1998. Requirements for the use of ani-
mal cells as in vitro substrates for the production of biologicals.
In: WHO Technical Report Series, vol. 878. WHO, Geneva, pp.
20–53 (Annex. 1).
Yokomizo, A.Y., Antoniazzi, M.M., Galdino, P.L., Azambuja Jr., N.,
Jorge, S.A.C., Pereira, C.A., 2004. Rabies virus production in
high Vero cell density cultures on macroporous microcarriers.
Biotechnol. Bioeng. 85, 506–515.
Yusibov, V., Hooper, D.C., Rupprecht, C.E., Spitsin, S.V., Fleysh,
N., Kean, R.B., Mikheeva, T., Deka, D., Karasev, A., Cox,
S., Randall, J., Koprowski, H., 2002. Expression in plants and
immunogenicity of plant virus-based experimental rabies vac-
cine. Vaccine 20, 3155–3164.