Chemical Engineering Science 66 (2011) 2431–2439

Contents lists available at ScienceDirect

Chemical Engineering Science

0009-25

doi:10.1

n Corr

E-m

journal homepage: www.elsevier.com/locate/ces

Continuous CHO cell cultures with improved recombinant proteinproductivity by using mannose as carbon source: Metabolic analysis andscale-up simulation

Julio Berrios a,n, Claudia Altamirano a, Nelson Osses b, Ramon Gonzalez c,d

a School of Biochemical Engineering, Pontificia Universidad Catolica de Valparaıso, Av. Brasil 2147, Valparaıso, Chileb Instituto de Quımica, Pontificia Universidad Catolica de Valparaıso, Chilec Department of Chemical and Biomolecular Engineering, Rice University, Houston, TX, USAd Department of Bioengineering, Rice University, Houston, TX, USA

a r t i c l e i n f o

Article history:

Received 1 December 2010

Received in revised form

9 March 2011

Accepted 9 March 2011Available online 16 March 2011

Keywords:

Bioprocessing

Biochemical engineering

Biopharmaceuticals

Metabolism

Recombinant protein

Scale-up

09/$ - see front matter & 2011 Elsevier Ltd. A

016/j.ces.2011.03.011

esponding author. Tel.: þ56 32 227 3823; fax

ail address: julio.berrios@ucv.cl (J. Berrios).

a b s t r a c t

The replacement of glucose by mannose as a means to improve recombinant protein productivity was

studied for the first time in continuous cultures of Chinese hamster ovary (CHO) cells producing human

recombinant tissue plasminogen activator (t-PA). Steady-state operation at two hexose levels in the

inlet (2.5 and 10 mM) allowed comparing the effect of sugar type and concentration on cell metabolism

and t-PA production independently of changes in specific growth rates produced by different culture

conditions. An increase in biomass concentration (15–20%) was observed when using mannose instead

of glucose. Moreover, specific hexose consumption rates were 20–25% lower in mannose cultures

whereas specific production rates of lactate, an undesirable by-product, were 25–35% lower than in

glucose control cultures. The volumetric productivity of t-PA increased up to 30% in 10 mM mannose

cultures, without affecting the sialylation levels of the protein. This increase is manly explained by the

higher cell concentration, and represents a substantial improvement in the t-PA production process

using glucose. Under this condition, the oxygen uptake rate and the specific oxygen consumption rate,

both estimated by a stoichiometric analysis, were about 10% and 25% lower in mannose cultures,

respectively. These differences lead to significant differences at larger scales, as it was estimated by

simulating cell cultures at different bioreactor sizes (5–5000 L). By assuming a set of regular operating

conditions in this kind of process, it was determined that mannose-based cultures could allow culturing

CHO cells up to 3000 L compared to only 80 L in glucose cultures at the same conditions. These facts

indicate that mannose cultures may have a significant advantage over glucose cultures not only in

terms of volumetric productivity of the recombinant protein but also for their potential application in

large-scale productive processes.

& 2011 Elsevier Ltd. All rights reserved.

1. Introduction

Processes based on mammalian cell cultures for recombinantprotein production with therapeutic applications have beenincreasing in the last decades (Ferrer-Miralles et al., 2009) sincethese cells, in contrast to many microorganisms, are able tosynthesise more complex proteins that often requires extensivepost-translational modifications such as glycosylations (Grillbergeret al., 2009). Glucose, the commonly used carbon source in culturesof mammalian cell lines, is mainly metabolised through theglycolytic pathway producing lactate as by-product (Tsao et al.,2005). This causes glucose to be inefficient in supporting the

ll rights reserved.

: þ56 32 227 3803.

energy requirements of the cell, thus an additional energy source,usually glutamine or glutamate must be added, generating ammo-nium as by-product (Altamirano et al., 2000, 2004). Both lactateand ammonium have a detrimental effect on cell growth andproductivity (Wagner, 1997; Chen and Harcum, 2006).

Alternative carbon sources in mammalian cell cultures have beenstudied in the last years (Altamirano et al., 2006; Wlaschin and Hu,2007; Berrios et al., 2009). Although metabolic pathways for glucoseand mannose utilisation are very similar, mannose and glucoseimpact differently the levels of UDP-glucosamine and UDP-galacto-samine (Ryll et al., 1994). The addition of mannose or its derivativesat low concentrations (o5 mM) together with glucose has also beenproposed for improving sialic acid amount in protein glycosylation(Gu and Wang, 1998; Follstad, 2004), but its effect on cell metabolismas the main carbon source replacing glucose has not been extensivelyevaluated yet. Recently, the effect of mannose as carbon source has

J. Berrios et al. / Chemical Engineering Science 66 (2011) 2431–24392432

been studied in batch cultures at different temperatures and mannoseconcentrations and a linear relationship between specific growth rateand the specific recombinant protein production rate has been found(Berrios et al., 2009). However, the effect of mannose concentrationcannot be assessed independently from changes in the specificgrowth rate in batch cultures, because the specific growth rate cannotbe controlled throughout the course of the culture.

Oxygen consumption rate (OUR) and CO2 evolution rate (CER)are widely used in microbial cultures for monitoring and controlthe fermentation (Zhang et al., 2005; Garcia-Ochoa et al., 2010;Liang et al., 2010). The respiratory quotient (RQ) quantitativelydescribes the relation between pathways involved in CO2 produc-tion and those mediating O2 consumption, and has also beenproposed for monitoring and controlling cell cultures (Xiao et al.,2006; Xiong et al., 2010). In contrast, this information is scarce formammalian cell cultures, especially because of difficulties withprecise measurement of CO2 in both gas and liquid phases and theeffect of CO2-enriched atmospheres normally used in thesecultures (Bonarius et al., 1995). A stoichiometric analysis basedon material and energy balances is a simple but effective way toestimate the oxygen uptake and CO2 production in mammaliancell cultures, which has been shown to agree with several reportsin the literature (Xiu et al., 1999; Xie and Zhou, 2006; Selvarasuet al., 2010). Therefore, OUR, CER and RQ are useful parameters tostudy culture behaviour and assess aeration requirements inbioreactors at different scales.

Oxygen transfer is a major issue in the scale-up of mammaliancell cultures. Although the oxygen uptake rate of these cultures is atleast two orders of magnitude lower than microbial cultures,mammalian cells are very sensitive to shear stress and air bubbles(Ma et al., 2006). For these reasons, the operation conditions interms of agitation speed and aeration rate are restricted to lowlevels, obtaining typical kLa values between 0.5 and 20 h�1, whereasin microbial systems kLa is normally 4100 h�1 (Langheinrich et al.,2002; Matsunaga et al., 2009; Acevedo, 2002). These facts must beconsidered when mammalian cell culture processes are scaled-up.In this regard, several different criteria for scaling-up mammaliancell processes have been proposed, including constant specificpower input, kLa and impeller tip speed (Varley and Birch, 1999;Xing et al., 2009). Still, the information available about scaling-upmammalian cell cultures is scarce, making simulation a useful toolfor estimating the operation conditions at different bioreactor scales.In this regard, most of industrial processes of therapeutic proteinshave been implemented in either fed-batch or perfusion culture(Xie and Zhou, 2006; Kompala and Ozturk, 2006). However, thoughstudies in continuous cultures have not been implemented forproductive processes, they provide useful data such as operationconditions and parameters that have been successfully applied todesign fed-batch or perfusion cultures (Fenge and Lullau, 2006).

The objective of this work was to evaluate the effect of replacingglucose by mannose as a carbon source on protein production andcell metabolism at a defined specific growth rate using steady-statecontinuous cultures. This strategy also allowed an assessment of theeffect of varying concentrations of mannose on cell growth and t-PAproduction independently of changes in specific growth rates. Usinga stoichiometric analysis, the effect of glucose replacement bymannose on the oxygen requirements of the cultures was estimated.A scaling-up simulation at different bioreactor scales as well aspotential advantages of mannose cultures is also presented.

2. Materials and methods

2.1. Cell line and culture medium

The strain CHO TF 70R of Chinese Hamster Ovary (CHO) cellsproducing human recombinant tissue plasminogen activator

(t-PA) was used (Pharmacia & Upjohn, Stockholm, Sweden). Cellviability was determined by trypan blue exclusion method. Cellbiomass was determined by dry weight method, washing thesamples twice with PBS buffer and removing most of the bufferwith a micropipette to avoid interferences from the dissolvedsalts in the buffer. The culture medium used was glutamine-freeBIOPRO1 (Lonza, Belgium), supplemented with 0.75 mM serine,0.65 mM asparagine, 0.45 mM proline and 6 mM of glutamate(Altamirano et al. 2006). Glucose or mannose was supplementedas indicated.

2.2. Continuous cultures

Continuous cultures were carried out in 250 mL Spinner flasks(Techne, USA) with 150 mL of culture medium. The flasks werespecially conditioned by inserting through one of the lateral capthree sealed access, for fresh medium inlet, cell culture mediumoutlet and a connection for a sterile filter that allowed the gasexchange. Cultures at each experimental condition were carriedout in duplicate and started from a freshly thawed cryovial, scaled-up in T-flasks and later transferred to one of the conditionedspinner flasks containing 10 mM of either mannose or glucose. Thecultures were carried out under controlled atmosphere at 37 1C, 5%CO2 and 95% relative humidity. Experiments using mannose wereperformed by feeding fresh medium containing 2.5 mM (Man-L) or10 mM (Man-H). Control cultures were run with 2.5 mM glucose(Glc-L) or 10 mM glucose (Glc-H) glucose. The dilution rate, kept at0.015 h�1 for all the cultures, was controlled by a low-flowperistaltic pump (Ismatec). 2 mL samples were taken every24–72 h for viable cell quantification, centrifuged and the super-natant immediately frozen at �20 1C for analytic measurements. Itwas considered that a culture reached a steady-state when, after atleast five residence times, both the number of viable cells andlactate concentration were constant in two consecutive samples(Altamirano et al., 2001b; Takuma et al., 2007).

2.3. Analytical measurements

Glucose, lactate and glutamate concentrations were determinedusing an YSI 2700 automatic analyser (Yellow Spring Instruments,USA). Mannose concentration was measured by HPLC with aPerkin-Elmer 200 Series fluorescence detector (excitation 360 nm,emission 425 nm) and a C-18 reversed-phase column (Waters,Ireland), derivatising the samples with anthranilic acid (Du andAnumula, 1998). Amino acids were measured by fluorescence in aHPLC (Perkin-Elmer 200 Series) using AccuTag kit (Waters, USA)according to manufacturer instructions.

2.4. t-PA measurement, purification and sialic acid level

t-PA concentration was measured by a commercial ELISA kit(Imulyse t-PA, Biopool, USA). For the determination of sialic acidquantity in t-PA, this protein was first purified by affinitychromatography in a fast protein liquid chromatography (FPLC,Pharmacia, Sweden) using erythrin-trypsin-inhibitor immobilisedin CNBr-activated Sepharose 4 Fast Flow (Amersham Biosciences)according to Heussen et al. (1984). The purification process waschecked by SDS-PAGE with 10–15% gradient gels (PhastSystem,Pharmacia, Sweden). The sialic acid was removed from glycosyla-tions of purified t-PA using the enzyme neuraminidase andtransformed into the corresponding mannosamine by the enzymeneuraminic acid aldolase (Fu and O’Neill, 1995). Each sample wasincubated at 37 1C for one hour with 0.5 U mL�1 of each enzymesimultaneously. The mannosamine produced was subsequentlyquantified by the above-mentioned method for determination ofmannose.

J. Berrios et al. / Chemical Engineering Science 66 (2011) 2431–2439 2433

2.5. Calculation of specific rates

Specific rates of production or consumption of i (qi) werecalculated by a mass balance made for the reactor

qi ¼DðCi

i�Coi Þ

X� 109

ð1Þ

where Cii is the concentration of i in the inlet (mmol/L), Ci

o is theconcentration of i in the outlet (mmol/L), X is the biomassconcentration (g/L) and D is the dilution rate of the culture (1/h).

2.6. Elemental composition of cells

The elemental composition of the cell was measured from7-8 mg of dried samples in an EA 1110 CHNS elemental analyser(CE Instruments, Italy). The empirical cell formula is usuallyexpressed per C atom (C-mol) containing only the major elements(Cinar et al., 2003; Najafpour, 2007) and can be calculated as

sE ¼%E

%C

12

mEð2Þ

where s is the subscript of the element E in the formula, %E is themass percentage of E in the dry cell, %C is the mass percentage ofcarbon in the dry cell, and mE is the atomic mass of E.

2.7. Stoichiometric analysis

The global equation for cell growth is presented below, usingGreek characters to represent stoichiometric coefficients:

a C6H12O6

Hexose

þbO2þX

j

gjCajHbj

OcjNdj

Amino acids

-CHeOf Ng

Cell

þdC3H5O�3Lactate

þeNHþ4 þl CO2þy H2O ð3Þ

For simplicity, the j amino acids are all presented in thesubstrate-side of the equation. In the case of Ala, Asp and Gly,which were produced by the cells in our experiments, thestoichiometric coefficients were considered to be negative in theequation. Since recombinant protein levels are low compared tothe other components of Eq. (2), this term was neglected. Thestoichiometric coefficients can be determined from the specificconsumption or production rates of the different molecules,according to this expression:

SCi ¼qi

qcellð4Þ

where SCi is the stoichiometric coefficient of i represented byGreek letters in Eq. (3), and qcell the specific production rate ofcells. Seventeen of the 20 main amino acids were successfullydetermined by chromatography the method described above.Glutamine could be measured but it was not present in theculture formulation nor detected in any culture. The amino acidsnot measured were Asn, Cys and Trp. Both Cys and Trp can beneglected since their contribution to biomass formation andenergy metabolism is considerably low compared to the otheramino acids (Gambhir et al., 2003). On the other hand, Asn ismore important in terms of its contribution to biomass so itcannot be neglected. For this reason, its coefficient was alsocalculated. Thus, a, gj, d and e were experimentally determined(except for gAsn) whereas the calculated stoichiometric coeffi-cients were b, gAsn, l and y. This was done by using elementalbalances for carbon (Eq (5)), hydrogen (Eq. (6)), oxygen (Eq. (7))and nitrogen (Eq. (8)) in the considered molecules:

C:6aþX

j

ajgj�1�3d�l¼ 0 ð5Þ

H:12aþX

j

bjUgj�e�5d�4e�2y¼ 0 ð6Þ

O:6aþ2bþX

j

cjgj�f�3d�2l�y¼ 0 ð7Þ

N:X

j

djgj�g�3e¼ 0 ð8Þ

These four equations, along with Eq. (3), constitute the systemof equations solved to determine the unknown coefficients. Thisover-determined system was solved by minimising the squaredifferences for the mass of each main element (C, N, O, and N) inreactants and products. In order to check the calculationsobtained by the former equations, a reductance balance was alsocarried out (Blanch and Clark, 1997; Xiu et al., 1999). Thereductance of the molecules or the cell was obtained from thefollowing equation:

Oi ¼ 4sCi þs

Hi �2sO

i �3sNi ð9Þ

where Oi is the reductance of the molecule i, sEi is the number of

atoms of the element E in the molecule i. The latter wererepresented in the global balance equation (Eq. (3)) by lower caseLatin letters. In this way, the previously calculated stoichiometriccoefficients were tested in the reductance balance shown below:

aOHexþbOO2þX

j

gjOAa ¼OCellþdOLacþeONHþ4ð10Þ

Reductances of CO2 and water are equal to zero.

2.8. Scale-up simulation

The effect of the observed differences between glucose andmannose cultures on aeration requirements of bioreactors atdifferent scales was evaluated through a scale-up simulation thataccounted for the metabolic differences between the two cul-tures. Several criteria have been used for process scale-up fromlaboratory (0.1–5 L) to industrial (10–10,000 L) scale in cellcultures, especially those related to aeration and agitation condi-tions. Geometric similarity is often assumed in most theoreticalcases, but this rarely occurs in real situations, as the H/T ratio ofthe bioreactor increases at larger scales (Varley and Birch, 1999;Xing et al., 2009). In this work, we have estimated the thresholdkLa (kLa*) in order to satisfy the demand of a culture running atthe same growth rate of our experiments (0.015 L/h). This may beachieved not only in continuous cultures but also in fed-batch orperfusion cultures, where controlled m and high cell concentra-tions (1.0–2.0�107 cells/mL) can be obtained (Xie and Zhou,2006; Kompala and Ozturk, 2006). The variation of kLa at differentscales was estimated and compared to kLa* to determine whetherthe operation conditions satisfy the oxygen requirements. Impel-ler tip speed has been used as a scale-up criterion in larger scalesystems (Xing et al., 2009). In this simulation, we have use a tipspeed of 1 m/s that is similar to reported values (0.5–1.0 m/s,Matsunaga et al., 2009). Thus, it can be assumed that thesimulation considers a bioreactor operating at the highest possi-ble agitation speed without causing cell damage.

Assuming that the oxygen transfer rate (OTR) in the bioreactormust be at least equal to the oxygen uptake rate (OUR) in order tosatisfy the oxygen demand of the culture, and that the contribu-tions of the dissolved oxygen in the medium inlet and outlet arenegligible, then

OUR¼OTR¼ kLaðC�L�CLÞ ð11Þ

A gas flow consisting of pure oxygen was considered inthe simulation, with. C*L¼1.07 mmol/L. In mammalian cell cul-tures, the dissolved oxygen tension is normally kept around 50%.

J. Berrios et al. / Chemical Engineering Science 66 (2011) 2431–24392434

Thus, kLa* can be calculated from OUR as follows:

kLa*¼OUR

0:5C�Lð12Þ

Eq. (12) allows calculating the minimum kLa value (kLa*) tosatisfy the oxygen requirements of the culture in terms of OUR. Inorder to estimate the kLa at different bioreactor scales, Van’t Rietcorrelation was used:

kLa¼ AeBT vC

s ð13Þ

where A¼7.2, B¼0.7, C¼0.2, (Langheinrich et al., 2002) andeT¼eIþeg, where

eI ¼NPrnN3D5

i

2:16� 105Vð14Þ

eG ¼ vsgr ð15Þ

and

vs ¼4Qg

pT2ð16Þ

In microbial cultures, the term eg can be neglected since eIbeg.However, eg can be comparable to eI in mammalian cell culturesbecause of their lower agitation speed and agitation power input(Langheinrich et al., 2002). The configuration and proportions ofbioreactor dimensions were assumed to be constant, with similargeometry to an actual industrial large-scale bioreactor (Xing et al.,2009): H/T¼2.5, Di/T¼0.5, three marine propellers (NP¼0.35 forRe4103). The aeration rate was assumed 0.005 vvm (Gorenfloet al., 2002; Langheinrich et al., 2002; Matsunaga et al., 2009).

For estimating the cell damage by gas bubbles in cell cultures,Tramper et al. (1988) proposed the following semi-empiricalcorrelation that can be used to estimate the first-order death rateconstant (kd) as a function of gas flow (Qg):

kd ¼4QgX0

p2D2i H

ð17Þ

where X0 is the specific hypothetical killing volume, which ischaracteristic for each system and is usually between 2�10�3

and 2�10�2 (Ma et al., 2006).

2.9. Statistical analysis

Continuous cultures at each condition were performed induplicate and two independent samples were taken at each timepoint for every culture with analytical measurements carried outseparately. Values are expressed as mean7standard error. One-way analysis of variance was used to compare the results,followed by Dunnett’s post-hoc testing using an evaluation copyof the software GraphPad InStat version 3.05 (Graphpad Software

Table 1Comparison of continuous cultures of CHO cells performed at high and low concentrat

Units Glc-L

Nvb 105 cell/L 6.9570.09

X g/L 0.15170.02

Residual hexose mM 0.2470.01

qHexc mmol/(h g) �231.772.6

YCell/Hex g/mol 66.570.6

Residual lactate mM 2.4270.11

qLac mmol/(h g) 248.171.2

YLac/Hex mol/mol 1.0770.03

a Hexose in the inlet medium stream; Man-L: mannose 2.5 mM; Man-H: mannoseb Viability was higher than 95% in all culturesc Negative values indicate consumption.

Inc., USA). Differences were considered statistically significantwhen po0.05.

3. Results and discussion

3.1. Cell growth, sugar utilisation and lactate synthesis in

continuous cultures with mannose or glucose as carbon source

Continuous cultures using mannose as carbon source wereperformed together with glucose cultures used as controls andcontaining the same inlet sugar concentration (2.5 or 10 mM,representing low and high concentrations, respectively). Atsteady-state, the viable cell concentration (Nv) obtained in man-nose cultures was 25% higher than their control glucose cultures,whereas a small but significant increase was observed in cultureswith higher hexose concentration (Table 1). However, the varia-tion in biomass concentration (X) was only 14% and 19% for Man-L and Man-H compared to their glucose controls. This indicatesthat cell mass varies depending on the experimental conditions(in pg per cell: Glc-L 216; Glc-H 281; Man-L 196; Man-H 272).Cell number concentration instead of biomass concentration isoften used in mammalian cell cultures. However, we have shownsome results in biomass units (g/L) in order to be consistent withthe mass balance, though they can be converted to cell concen-tration units by using the cell mass reported above.

Mannose metabolism seems to be more efficient than glucose inthe synthesis of biomass, since the hexose-into-cells yield (YCell/Hex)was 23–36% higher in mannose cultures than glucose controlcultures (Table 1). The increase in residual hexose concentration inMan-H and Glc-H compared to Man-L and Glc-L, respectively, is anindication that cultures are not hexose-limited at 10 mM. Furthersupporting this reasoning, the YCell/Hex decreases at 10 mM com-pared to 2.5 mM. This finding is in agreement with previous reports,where glucose has been shown to be the limiting nutrient incontinuous culture of CHO cells when feed concentrations are lowerthan 2.6 mM (Altamirano et al., 2001b; Takuma et al., 2007).

We have recently reported that the specific rate of hexoseconsumption (qHex) depends on m in batch cultures (Berrios et al.,2009). However, different mannose concentrations also led todifferent values of m. Since the steady-state culture operation fixesm at a defined value, the differences observed are only caused by thechange in sugar type or concentration. qHex in mannose cultureswere 20–25% lower than the corresponding glucose control cultures(Table 1), indicating a significant difference between mannose andglucose metabolism under these conditions. The residual lactateconcentration was 26% lower in Man-L and 13% lower in Man-H,compared to the corresponding glucose controls, whereas qLac valueswere 63% and 74% of their controls for Man-L and Man-H, respec-tively. In addition, hexose-to-lactate yield (YLac/Hex) in Man-L wasalmost 20% lower than control, whereas in Man-H it was not

ions of glucose and mannose. Values are from the steady-statea.

Glc-H Man-L Man-H

7.4870.14 8.6870.13 9.2570.11

0.21070.04 0.17370.03 0.24970.08

4.3470.07 0.4270.01 5.0570.11

�396.178.6 �183.372.6 �295.5714.1

37.170.9 81.870.6 50.870.2

7.2770.47 1.8070.08 6.3570.20

508.875.2 158.671.0 379.179.1

1.2870.06 0.8770.04 1.2870.04

10 mM. Control cultures: Glc-L: glucose 2.5 mM; Glc-H: glucose 10 mM.

Fig. 1. Specific consumption/production rates of amino acids and ammonium (qi).

Hexose in the inlet medium stream; Man-L: mannose 2.5 mM; Man-H: mannose

10 mM. Control cultures: Glc-L: glucose 2.5 mM; Glc-H: glucose 10 mM.

J. Berrios et al. / Chemical Engineering Science 66 (2011) 2431–2439 2435

significantly different from the control, demonstrating that withmannose it is possible to achieve a more efficient hexose utilisation/metabolism. These results suggest that differences in both qHex andqLac values between mannose and glucose also imply a differentmetabolic response of cells to these sugars, even when growing atthe same specific growth rate.

The available information about the use of mannose instead ofglucose is limited only to a few publications. It has been observedthat mannose has an effect on the activity of some enzymes suchas glucosamine-6-phosphate isomerase (Ryll et al., 1994; C- ayliet al., 1999). However, this does not explain the differencesobserved in central metabolism, particularly in glycolysis andlactate production, nor in the cell biomass. The experimentalevidence indicates that differences in biomass levels observedbetween mannose and glucose cultures are a result of differentmetabolic responses. This is clearly observed in yield values ofTable 1. A possible explanation, similar to what is observed insugar transporters where there are hexose transporters andothers more specific for glucose or fructose (Nelson and Cox,2004), is that the metabolic response to mannose or glucose atpathway control level might be different. The existence ofdifferences at alosteric regulation level seems to be more likelythan genetic modulation, since it has been observed in batchcultures that the mannose growth does not require any previousadaptation of the cells to this sugar (Altamirano, 2000).

3.2. Cell elemental composition

The elemental composition of the biomass obtained is shownin Table 2. Small but significant variations were observed in C, Hand O percentage of dry biomass in response to changes in hexoseconcentration, whereas N did not change except in Glc-H cultures.On the other hand, changes in response to hexose type are notsignificant suggesting that the increase in biomass concentrationobserved with mannose might not significantly change the cellcomposition. Elemental composition is also expressed by using anempirical formula and molecular weight of the cell on C-mol basis(Table 2). The empirical formula and elemental composition ofthe cell are valuable tools in process engineering when designingculture media or solving mass balances in cell cultures (Europaet al., 2000; Gambhir et al., 2003; Najafpour, 2007). Data regard-ing the elemental composition of biomass in mammalian cells isscarce. For example, Europa et al. (2000) reported an empiricalformula of CH1.96O0.49N0.26, with a molecular weight of 25.47 g/molfor hybridoma, without assessing the effect of experimentalconditions on it. In this work, we have shown that the experi-mental conditions indeed have a small but significant effect onelemental composition. These measurements are more accuratethan macromolecular composition measurements, where higherexperimental variations are observed (Nielsen et al., 2003).

3.3. Amino acids metabolism

The specific consumption or production rates of the measuredamino acids and ammonium were calculated in steady-state

Table 2Elemental composition and empirical formula of CHO cells.

% of element in the cell Glc-L

C 43.370.5

H 6.370.1

O 31.170.5

N 11.270.1

Empirical formula CH1.76O0.54N0.22

Molecular weight of the cell (g/mol) 25.49

(Fig. 1). Most amino acids are consumed, except for Ala, Gly,Asp, in agreement with previous reports (Altamirano et al. 2006).The specific consumption rate of glutamate (qGlu) in steady-statedecreased 40–50% at higher hexose concentration, whereas smal-ler differences (10–17%) were observed between mannose andcontrol glucose cultures. Lower values of qGlu and ammoniumlevels at higher hexose concentration can be explained by higherhexose-dependent energy availability in Glc-H and Man-H cul-tures (Altamirano et al., 2000). The decrease in qGlu related tohigher glucose availability in the medium has been describedpreviously (Martinelle et al., 1998; Altamirano et al., 2001a) asenergy requirements of the cell are supported by both glucoseand glutamate metabolism (Altamirano et al., 2001b).

Ammonium concentrations in all the evaluated conditionswere lower than 0.9 mM (data not shown), and hence belowharmful levels for protein production and cell growth (Wagner,1997). High hexose concentrations led to a reduction of 40–50% inammonium levels with smaller differences between mannose andglucose cultures (0.44 mM in Glc-H and Man-H). In contrast to thebehaviour observed for sugar consumption, changes in qAmm wereobserved between mannose cultures and their controls, indicatinga more important effect of hexose concentration rather than thehexose type (Fig. 1). Glutamate-into-ammonium yield (YAmm/Glu)was 30% lower in Man-H compared to Man-L (0.3970.03 and0.5770.01, respectively), while it was 50% lower in Glc-Hcompared to Glc-L (0.4570.03 and 0.6870.02, respectively).

3.4. Recombinant protein production

The production of the recombinant protein t-PA was evaluatedat steady-state through both specific and volumetric productionrates (Table 3). The specific production rate of t-PA (qt-PA) inMan-L is about 30% lower than the glucose control, and exhibits a2.3-fold and 1.6-fold increase in Man-H and Glc-H, respectively,reaching the same levels for both sugars. In contrast, no signifi-cant effect of mannose was observed on volumetric productionrate of t-PA in low-hexose cultures (Table 3), whereas Man-H

Glc-H Man-L Man-H

46.270.2 44.270.4 45.970.2

6.970.6 5.970.3 6.470.5

27.070.3 30.770.4 28.570.1

12.970.2 11.270.3 11.470.6

CH1.78O0.44N0.24 CH1.61O0.52N0.22 CH1.67O0.47N0.21

24.12 25.00 24.12

Table 3Specific and volumetric productivity of t-PA, yield of hexose-into-t-PA and sialic

acid levels of t-PA for steady-state continuous cultures of CHO cells.

Units Glc-L Glc-H Man-L Man-H

qt-PA mg/(h g) 32.9970.05 40.8270.68 26.6170.08 44.0670.12

Qt-PA mg/(L h) 4.9670.17 8.5370.35 4.5370.54 11.0270.65

Yt-PA/Hex mg/g 0.7970.04 0.5770.02 0.8170.05 0.8370.11

SA:t-PAa mol/mol 2.570.1 3.370.2 3.170.1 3.570.2

a SA: Sialic acid.

Table 4Experimentally measured and calculated stoichiometric coefficients for 1 mol of

biomass.

Compound SC Glc-2.5 Glc-10 Man-2.5 Man-10

MeasuredHexose a 0.3834 0.6499 0.3054 0.4750

Glu gGlu 0.1781 0.1102 0.2129 0.1075

Arg gArg 0.0034 0.0092 0.0029 0.0048

His gHis 0.0017 0.0011 0.0015 0.0058

Ile gIle 0.0102 0.0103 0.0132 0.0106

Leu gLeu 0.0068 0.0092 0.0073 0.0106

Lys gLys 0.0119 0.0103 0.0059 0.0125

Met gMet 0.0051 0.0092 0.0044 0.0106

Phe gPhe 0.0051 0.0069 0.0103 0.0058

Pro gPro 0.0068 0.0046 0.0073 0.0067

Ser gSer 0.0577 0.0677 0.0397 0.0470

Thr gThr 0.0119 0.0092 0.0088 0.0096

Tyr gTyr 0.0051 0.0080 0.0073 0.0086

Val gVal 0.0102 0.0115 0.0118 0.0086

Lactate d 0.4106 0.8347 0.2643 0.6094

Ammonium e 0.1205 0.0505 0.1219 0.0422

Alaa gAla �0.0102 �0.0276 �0.0147 �0.0250

Aspa gAsp �0.0322 �0.0310 �0.0176 �0.0125

Glya gGly �0.0017 0.0126 �0.0044 �0.0134

CalculatedAsn gAsn 0.0227 0.0270 0.0138 0.0081

O2 b 1.500 1.689 1.610 1.242

H2O y 1.477 1.922 1.518 1.424

CO2 l 1.513 1.564 1.651 1.143

a Negative coefficient for these amino acids indicate that they are produced.

Table 5Respiratory quotient and specific and volumetric consumption/production rates of

O2 and CO2.

Glc-L Glc-H Man-L Man-H

RQ 1.01 0.93 1.03 0.92

qO2(mmol/(g h)) 0.906 1.029 0.966 0.773

qCO2(mmol/(g h)) 0.914 0.953 0.991 0.711

OUR (mmol/(L h)) 0.136 0.216 0.164 0.194

CER (mmol/(L h)) 0.137 0.200 0.169 0.179

YCO2=Hex (mol/mol) 3.95 2.41 5.41 2.41

J. Berrios et al. / Chemical Engineering Science 66 (2011) 2431–24392436

increased 2.4-fold compared to Man-L, and it was 1.3-fold higherthan its glucose control (Glc-H). This increase is mainly explainedby a higher number of cells in Man-H (Table 1).

The hexose-into-t-PA yield (Yt-PA/Hex) remained unchanged athigher mannose concentration while it decreased 25% in glucosecultures (Table 3). The sialic acid (SA) content in glycosylatedt-PA, expressed as molar ratio of SA to t-PA (SA:t-PA), is alsoshown in Table 3 and indicates that using mannose instead ofglucose does not affect sialylation at high hexose concentrationswhereas it improves the sialylation levels at low hexose concen-tration. SA quantity in protein glycosylations is an importantvariable to be considered in therapeutic protein production, sinceit has an important effect on biological activity of t-PA byreducing the blood clearance of the protein (Joziasse et al.,2000). Additional studies that are out of the scope of this worksuch as glycosylation patterns analysis could contribute to abetter understanding of this behaviour. In contrast to glucosecultures, both Yt-PA/Hex and glycosylation ratio do not changesignificantly in mannose cultures in response to variations in theconcentration of this hexose in the feeding. These findingstogether with the improved productivity make mannose culturessuitable as a more stable and productive process compared toglucose cultures.

The factor affecting glycosylations and particularly the sialicacid content are not completely clear. In our experiments,differences observed between mannose and glucose culturesmight be explained by different metabolic responses to eachsugar. As mentioned before in Section 3.1, mannose producedchanges in metabolic behaviour of the cell culture. It is possiblethat these changes could affect the glycosylations synthesispathways, as it has been established for the glucosamine-6-phosphate isomerase (C- ayli et al., 1999). This enzyme is directlyinvolved in the synthesis of glycosylations intermediaries suchGlcNAc and sialic acid. Mannose enters to mammalian cellsthrough a hexose transporter (Rodriguez et al., 2005), where itis converted to mannose-6-phosphate by a hexokinase. Themannose flux into the cell could increase the concentration ofthese metabolites and affect the sialic acid levels in the glycosyla-tions of t-PA. Further research in this direction could elucidate thecauses of the observed behaviour.

3.5. Stoichiometric analysis

Values of steady-state concentration of metabolites and cellswere used to calculate the stoichiometric coefficients of Eq. (3)directly from the experimental measurement using Eqs. (2) and (4),and are presented in Table 4. The non-measured coefficients gAsn, land r of Asn, O2, CO2 and H2O, respectively, were calculated usingthe elemental balances for C, H, O and N and are also presented inTable 4.

The respiratory quotient was calculated from the stoichio-metric coefficient of CO2 and O2 in each experimental conditionand is shown in Table 5. Changes in hexose concentrationproduced a shift in RQ from about 0.92 at low hexose levels to

above 1.01 at high hexose levels, but without significant differ-ences between mannose and glucose control cultures. RQ valuesare within the ranges obtained by other authors (Bonarius et al.,1995, 1996; Xiu et al., 1999). Specific consumption rate of O2 (qO2

)and production of CO2 (qCO2

) were calculated from stoichiometriccoefficients. Unlike the trend of RQ as a function or hexoseconcentration, qO2

and qCO2in mannose cultures decrease when

increasing the hexose concentration in the inlet, whereas it wasthe opposite in the control cultures. Moreover, at high hexoseconcentration, qO2

and qCO2in mannose cultures are about 25%

lower than glucose controls, whereas at low hexose their valuesin mannose are about 7% higher than the control cultures.

The calculation of the oxygen uptake rate (OUR) allowsestimating the oxygen requirements of the bioreactor, and isshown in Table 5. OURs varied with a similar trend than thatobserved for qO2

in response to hexose type, although to adifferent extent. The shift from glucose to mannose at lowconcentration leads to a decrease of 7% in qO2

, but a 20% decreasein OUR. In mannose cultures, OUR increases 18% at high concen-trations, whereas it increases 60% in control cultures, revealing animportant metabolic difference between these two sugars inresponse to changes in feeding concentration. For both mannoseand control cultures, the values of qO2

are in the range reportedfor mammalian cell cultures (Xiu et al., 1999). The hexose-into-CO2 yield (YCO2=Hex) is 55% lower in Man-H than in Man-L,

J. Berrios et al. / Chemical Engineering Science 66 (2011) 2431–2439 2437

indicating that a higher proportion of carbon from the hexose isrouted to biomass or, in lower amount, to lactate, consequently alower proportion of the sugar is oxidised to CO2. This finding canalso be supported by considering the values of X and YLac/Hex

values (Table 1).

3.6. Scale-up simulation of mannose and glucose cultures

The potential for using mannose in large-scale CHO cellcultures was evaluated by simulating the kLa required to supportthe oxygen demand of bioreactors ranging in size between 5 and5 000 L (Fig. 2A). The bioreactor characteristics and scale-upcriteria used in the simulation are explained in Materials andmethods. The simulation led to N values between 280 and 28 rpmfor bioreactor sizes between 5 and 5000-L, respectively indicatinga good approximation to real systems, since N is normallybetween 100 and 300 rpm at scale o10 L (Tsao et al., 2005;Frahm et al., 2002; Tisu and Pavko, 2010; Weber et al., 2005) anddecreases with the bioreactor size to values around 30 rpm at5000 L (Xing et al., 2009). The kLa values obtained by thesimulation for different bioreactor sizes are similar to thoseusually observed in mammalian cell cultures (Langheinrichet al., 2002; Xing et al., 2009; Nienow, 2006). The minimum kLa

(kLa*) required to satisfy the OUR of Glc-H and Man-H cultures arealso included in Fig. 2A, considering a cell concentration of12�106 cells/mL as an example that is within the range of valuesreported (Gorenflo et al., 2002; Kompala and Ozturk, 2006; Xinget al., 2009). Under the given conditions, the kLa of the system isable to satisfy the oxygen requirements of the glucose cultureonly up to a scale of about 80 L, whereas mannose culture can be

Fig. 2. Scale-up simulation obtained at different bioreactor sizes for conditions

described in the main text: (A) values of kLa and N at different bioreactor size(s).

The values of kLa* for both glucose and mannose cultures at cell concentration of

12�106 cells/mL are also shown. (B) Theoretical maximum cell concentration (Nv)

that the bioreactor system may support in cultures using glucose or mannose.

supported up to 3000 L with the same operation conditions. If theglucose culture is to be implemented at 3000-L scale, higher OTRshould be achieved to satisfy the OUR under this condition. Sincein this simulation the gas flow consist of pure oxygen, and theagitation speed is already the highest possible before shear stressdamage is produced, other strategies should be implemented.

A possible strategy to increase OTR is an increase in gas flow (Qg).Eq (13) predicts that an increase from 0.005 to 0.024 vvm isneeded to achieve the kLa for glucose culture at 3000-L scale. Thismeans that Qg must be increased from 2.5�10�4 to1.19�10�3 m3/s (4.8-fold increase). However, it has been wellestablished that gas bubbles may cause severe cell damage, increas-ing the cell death rate and decreasing productivity (Ma et al., 2006).For a given bioreactor, it can be assumed from Eq. (17) that kd¼cQg,where c is a constant. Thus, a 4.8-fold increase in kd is produced at3000-L scale, that could lead to a significant reduction in theviability of the culture.

Since the former analysis was made for one cell concentration,it is interesting to determine how the systems would behave inother conditions. Fig. 2B shows an estimation of the maximumcell concentration that could theoretically be achieved withglucose and mannose culture. For both hexoses the profile iscoincident with the calculated kLa (Fig. 2A). Mannose-basedcultures allows to reach higher cell concentrations than glucoseones at any scale. The simulation indicates that the glucose-basedsystem might support the oxygen demand of 18.8�106 cells/mLin 5-L bioreactor and 9.2�106 cells/mL in a 5000-L bioreactor. Incontrast, mannose-based system can support 26�106 cells/mL in5-L bioreactor and 12.7�106 cells/mL in a 5000 L bioreactor.These differences, along with the other evidence presented in thiswork, make mannose-based cultures a new and interesting alter-native for recombinant protein production in mammalian cellcultures. The improvement achieved using mannose as the onlycarbon source allows obtaining higher productivities than glu-cose-base cultures. For example, for volume scale of 100 L thevolumetric productivity of mannose-base culture is 44% higherthan the obtained with glucose. Besides, the capacity of operatingat much higher scales with mannose-based cultures, as men-tioned before, is an advantage for industrial production, where theproduction scale is a key aspect to be considered. In summary,mannose-based cultures rise as a new and promising strategy forrecombinant protein produced by CHO cells in industrialprocesses.

4. Conclusions

The effect of replacing glucose with mannose as the carbonsource in CHO cell cultures was evaluated in steady-state contin-uous cultures. To our knowledge, this is the first report ofcontinuous cultures of mammalian cells using mannose as carbonsource. The use of the same dilution rate allowed studying theeffect of the type of hexose independently of changes in specificgrowth rate. At high mannose concentration in the feeding(10 mM), the volumetric rate of production of the recombinantprotein increased 30% compared to a glucose control, withoutaffecting its sialic acid level. This improvement is mainly explainedby a higher cell concentration rather than for an increase in thespecific production rate. In addition, the metabolic behaviour usingmannose differed in several aspects from that obtained in glucose,leading to lower specific hexose consumption and lactate produc-tion rates than the glucose control culture. Using a stoichiometricanalysis it was determined that, although no differences in RQwere observed at high hexose concentration, both the qCO2

and qO2

were considerably lower in mannose cultures. The impact of thesedifferences evaluated by modelling the system behaviour at

J. Berrios et al. / Chemical Engineering Science 66 (2011) 2431–24392438

different scales indicates that mannose cultures can be performedat higher cells concentrations and volumetric productivities thanglucose cultures when scaled-up. One potential drawback of usingmannose is the higher cost of this sugar, although the increasedproductivity, high product price, and potential application athigher scales could compensate that situation.

Nomenclature

Cii concentration of i in the inlet, mmol/L

Cio concentration of i in the outlet, mmol/L

CL oxygen concentration in the liquid, mmol/LCL* maximum oxygen solubility in the liquid, mmol/LCER carbon evolution rate, mmol/(L h)D dilution rate of the culture, 1/hDi impeller diameter, mE element used in the mass balance (C, H, O or N)g gravitational constant, 9.81 m/s2

H medium height in the bioreactor, mmE atomic mass of E, g/molkd first-order death rate constant, 1/skLa volumetric oxygen transfer coefficient, 1/hkLa* volumetric oxygen , 1/hn number of impeller in the bioreactorN stirrer speed, rpmNp power numberOTR oxygen transfer rate, mmol/(L h)OUR oxygen uptake rate, mmol/(L h)Qg gas flow (pure oxygen in this work), m3/sqi specific consumption/production rate of i, mmol/(h g)Re Reynolds numberRQ respiratory quotientSA sialic acidSCi stoichiometric coefficient of the molecule i

T bioreactor diameter, mV bioreactor volume, m3 or Lvvm aeration rate, L/(L min)X biomass concentration (g/L)X0 specific hypothetical killing volumeYx/y Yield of x-into-y

Greek letters

a stoichiometric coefficient of hexose, molb stoichiometric coefficient of oxygen, molgj stoichiometric coefficient of the amino acid j, mold stoichiometric coefficient of lactate, mole stoichiometric coefficient of ammonium, moleG energy dissipation rate from aeration, W/m3

eI energy dissipation rate from agitation, W/m3

eT total energy dissipation rate, W/m3

y stoichiometric coefficient of water, moll stoichiometric coefficient of CO2, molr liquid density, kg/m3

sEi number of atoms of the element E in the molecule i.

sE subscript of the element E in a compound formulaOi reductance of i

Acknowledgments

This work has been supported by FONDECYT 1070337, VRIEA-PUCV BINAF project 037.109/2008, and doctoral thesis grant fromCONICYT. Technical support of Veronica Figueroa and CatalinaMorales is also acknowledged.

References

Acevedo, F., 2002. Transferencia de masa, momento y calor en fermentaciones.In: Acevedo, F., Gentina, J.C., Illanes, A. (Eds.), Fundamentos de IngenierıaBioquımica, Ediciones Universitarias de Valparaıso, Valparaıso, pp. 217–251.

Altamirano, C., 2000. Analisis del metabolismo de celulas cho recombinantes ydesarrollo de estrategias de cultivo en medios definidos. Ph.D. Thesis, Uni-versitat Aut �onoma de Barcelona, Spain.

Altamirano, C., Cairo, J., G �odia, F., 2001a. Decoupling cell growth and productformation in Chinese hamster ovary cells through metabolic control. Biotech-nology and Bioengineering 76, 351–360.

Altamirano, C., Illanes, A., Becerra, S., Cairo, J., G �odia, F., 2006. Considerations onthe lactate consumption by CHO cells in the presence of galactose. Journal ofBiotechnology 125, 547–556.

Altamirano, C., Illanes, A., Casablancas, A., Gamez, X., Cairo, J., G �odia, F, 2001b.Analysis of CHO cells metabolic redistribution in a glutamate-based definedmedium in continuous culture. Biotechnology Progress 17, 1032–1041.

Altamirano, C., Paredes, C., Cairo, J., G �odia, F., 2000. Improvement of CHO cellculture medium formulation: simultaneous substitution of glucose andglutamine. Biotechnology Progress 16, 69–75.

Altamirano, C., Paredes, C., Illanes, A., Cairo, J.J., G �odia, F., 2004. Strategies for fed-batch cultivation of t-PA producing CHO cells: substitution of glucose andglutamine and rational design of culture medium. Journal of Biotechnology110, 171–179.

Berrios, J., Dıaz-Barrera, A., Bazan, C., Altamirano, C., 2009. Relationship betweentissue plasminogen activator production and specific growth rate in ChineseHamster Ovary cells cultured in mannose at low temperature. BiotechnologyLetters 31, 1493–1497.

Blanch, H., Clark, D., 1997. Biochemical Engineering. Marcel Dekker Inc.,New York.

Bonarius, H.P., de Gooijer, C.D., Tramper, J., Schmid, G., 1995. Determination of therespiration quotient in mammalian cell culture in bicarbonate buffered media.Biotechnology and Bioengineering 45, 524–535.

Bonarius, H.P., Hatzimanikatis, V., Meesters, K.P., de Gooijer, C.D., Schmid, G.,Tramper, J., 1996. Metabolic flux analysis of hybridoma cells in differentculture media using mass balances. Biotechnology and Bioengineering 50,299–318.

C- ayli, A., Hirschmann, F., Wirth, M., Hauser, H., Wagner, R., 1999. Cell lines withreduced UDP-N-acetylhexosamine pool in the presence of Ammonium. Bio-technology and Bioengineering 65, 192–200.

Chen, P., Harcum, S., 2006. Effects of elevated ammonium on glycosylation geneexpression in CHO cells. Metabolic Engineering 8, 123–132.

Cinar, A., Partulekar, S., Undey, C., 2003. Batch Fermentation, Modelling, Monitor-ing and Control. Marcel Dekker, New York.

Du, P., Anumula, K., 1998. Quantitative analysis of monosaccharides fromglycoproteins by fast HPLC with highly sensitive fluorescence detection.Journal of Biomolecular Techniques 9, 31–33.

Europa, A., Gambhir, A., Fu, P, Hu, W.-S., 2000. Multiple steady states with distinctcellular metabolism in continuous culture of mammalian cells. Biotechnologyand Bioengineering 67, 25–34.

Fenge, C., Lullau, E., 2006. Cell culture bioreactors. In: Ozkurt, S.S., Hu, W.-S. (Eds.),Cell Culture Technology for Pharmaceutical and Cell-based. Taylor Francis,London, pp. 155–224.

Ferrer-Miralles, N., Domingo-Espın, J., Corchero, J.L., Vazquez, E., Villaverde, A.,2009. Microbial factories for recombinant pharmaceuticals. Microbial CellFactories 8, 17.

Follstad, B., 2004. Methods and media for controlling sialylation of proteinsproduced by mammalian cells. U.S. Patent No. 2004/00810.

Frahm, B., Blank, H.-C., Cornand, P., Oelßner, W., Guth, U., Lane, P., Munack, A.,Johannsen, K., Portner, R., 2002. Determination of dissolved CO2 concentrationand CO2 production rate of mammalian cell suspension culture based on off-gas measurement. Journal of Biotechnology 99, 133–148.

Fu, D., O’Neill, R., 1995. Monosaccharide composition analysis of oligosaccharidesand glycoproteins by high-perfomance liquid chromatography. AnalyticalBiochemistry 227, 377–384.

Gambhir, A., Korke, R., Lee, J., Fu, P., Europa, A., Hu, W.-S., 2003. Analysis of cellularmetabolism of hybridoma cells at distinct physiological states. Journal ofBioscience and Bioengineering 95, 317–327.

Garcia-Ochoa, F., Gomez, E., Santos, V.E., Merchuk, J.C., 2010. Oxygen uptake ratein microbial processes: an overview. Biochemical Engineering Journal 49,289–307.

Gorenflo, V.M., Smith, L., Dedinsky, B., Persson, B., Piret, J.M., 2002. Scale-up andoptimization of an acoustic filter for 200 L/day perfusion of a CHO cell culture.Biotechnology and Bioengineering 80, 438–444.

Grillberger, L., Kreil, T.R., Nasr., S., Reiter, M., 2009. Emerging trends in plasma-freemanufacturing of recombinant protein therapeutics expressed in mammaliancells. Biotechnology Journal 4 186–201.

Gu, X., Wang, D.I.C., 1998. Improvement of interferon-g sialylation in Chinesehamster ovary cell culture by feeding of N-acetylmannosamine. Biotechnologyand Bioengineering 58, 643–648.

Heussen, C., Joubert, F., Dowdle, E.B., 1984. Purification of human tissue plasmino-gen activator with erythrina trypsin inhibitor. The Journal of BiologicalChemistry 259, 11635–11638.

Joziasse, D., Lee, R., Lee, Y., Biessen, E., Schiphorst, W., Koeleman, C., van denEijnden, D., 2000. a3-Galactosylated glycoproteins can bind to the hepatic

J. Berrios et al. / Chemical Engineering Science 66 (2011) 2431–2439 2439

asialoglycoprotein receptor. European Journal of Biochemistry 267,6501–6508.

Kompala, D.S., Ozturk, S.S, 2006. Optimization of high cell density perfusionbioreactors. In: Ozkurt, S.S., Hu, W.-S. (Eds.), Cell Culture Technology forPharmaceutical and Cell-based. Taylor Francis, London, pp. 387–416.

Langheinrich, C., Nienow, A.W., Eddleston, T., Stevenson, N.C., Emery, A.N., Clayton,T.M., Slater, N.K.H., 2002. Oxygen transfer in stirred bioreactors under animalcell culture conditions. Trans IChemE 80, 39–44.

Liang, J.-G., Chu, X.-H., Chu, J., Wang, Y.-H., Zhuang, Y.-P., Zhang, S.-L., 2010.Oxygen uptake rate (OUR) control strategy for improving avermectin B1a

production during fed-batch fermentation on industrial scale (150 m3). AfricanJournal of Biotechnology 9, 7186–7191.

Ma, N., Mollet, M., Chalmers, J.J., 2006. Aeration, mixing and hydrodynamics inbioreactors. In: Ozkurt, S.S., Hu, W.-S. (Eds.), Cell Culture Technology forPharmaceutical and Cell-based. Taylor Francis, London, pp. 225–248.

Martinelle, K., Doverskog., M., Jacobbson, U., Chapman, B., Kuchel, P., Haggstrom,L., 1998. Elevated glutamate dehydrogenase flux in glucose-deprived hybri-doma and myeloma cells, evidence from 1H/15N NMR. Biotechnology andBioengineering 60, 508–517.

Matsunaga, N., Kano, K., Dobashi, T., 2009. Culture scale-up studies as seen fromthe viewpoint of oxygen supply and dissolved carbon dioxide stripping.Journal of Bioscience and Bioengineering 107, 412–418.

Najafpour, G., 2007. Biochemical Engineering and Biotechnology. Elsevier, Oxford.Nelson, D., Cox, M., 2004. Lehninger Principles of Biochemistry, fourth ed. W.H.

Freeman, New York.Nielsen, J., Villadsen, J., Liden, G., 2003. Bioreaction Engineering Principles, second

ed, Kluwer Academics, New York.Nienow, A., 2006. Reactor engineering in large scale animal cell culture. Cyto-

technology 50, 9–33.Rodriguez, P., Rivas, C., Godoy, A., Villanueva, M., Fischbarg, J., Vera, J., Reyes, M.,

2005. Redefining the facilitated transport of mannose in human cells: absenceof a glucose-insensitive, high-affinity facilitated mannose transport system.Biochemistry 44, 313–320.

Ryll, T., Valley, U., Wagner, R., 1994. Biochemistry of growth inhibition byammonium ions in mammalian cells. Biotechnology and Bioengineering 44,184–193.

Selvarasu, S., Kim, D.Y., Karimi, I.A., Lee, D.-Y., 2010. Combined data preprocessingand multivariate statistical analysis characterizes fed-batch culture of mousehybridoma cells for rational medium design. Journal of Biotechnology 150,94–100.

Takuma, S., Hirashima, C., Piret, J., 2007. Dependence on glucose limitation of thepCO2 influences on CHO cell growth, metabolism and IgG production.Biotechnology and Bioengineering 97, 1479–1488.

Tisu, M., Pavko, A., 2010. Oxygen transfer in laboratory stirred tank bioreactorduring mammalian cell culture cultivation. Acta Chimica Slovenica 57,123–128.

Tramper, J., Smit, D., Straatman, J., Vlak, J.M., 1988. Bubble column design forgrowth of fragile insect cells. Bioprocess Engineering 3, 37–41.

Tsao, Y., Cardoso, A., Condon, R., Voloch, M., Lio, P., Lagos, J., Kearns, B., Liu, Z., 2005.Monitoring Chinese hamster ovary cell culture by the analysis of glucose andlactate metabolism. Journal of Biotechnology 118, 290–303.

Varley, J., Birch, J., 1999. Reactor design for large scale suspension animal cellculture. Cytotechnology 29, 177–205.

Wagner, R., 1997. Metabolic control of animal cell culture processes. In: Hauser,H., Wagner, R. (Eds.), Mammalian Cell Biotechnology in Protein Production.Walter de Gruyter, Berlin, pp. 233–276.

Weber, W., Rimann, M., de Glutza, F.-N., Weber, E., Memmert, K., Fussenegger, M.,2005. Gas-inducible product gene expression in bioreactors. Metabolic Engi-neering 7, 174–181.

Wlaschin, K., Hu, W.-S., 2007. A scaffold for the Chinese hamster genome.Biotechnology and Bioengineering 98, 429–439.

Xiao, J., Shi, Z., Gao, P., Feng, H., Duan, Z., Mao, Z., 2006. On-line optimization ofglutamate production based on balanced metabolic control by RQ. Bioprocessand Biosystems Engineering 29, 109–117.

Xie, L., Zhou, W., 2006. Fed-Batch cultivation of mammalian cells for theproduction of recombinant proteins. In: Ozkurt, S.S., Hu, W.-S. (Eds.), Cellculture Technology for Pharmaceutical and Cell-based. Taylor Francis, London,pp. 349–386.

Xing, Z., Kenty, B.M., Li, Z.J., Lee, S.S., 2009. Scale-up analysis for a CHO cell cultureprocess in large-scale bioreactors. Biotechnology and Bioengineering 103,733–746.

Xiong, Z.-Q., Guo, M.-J., Guo, Y.-X., Zhuang, Y.-P., Wang, N.S., Zhang, S.-Y., 2010.RQ feedback control for simultaneous improvement of GSH yield and GSHcontent in Saccharomyces cerevisiae T65. Enzyme and Microbial Technology46, 598–602.

Xiu, Z.-L., Deckwer, W.-D., Zeng, A.-P., 1999. Estimation of rates of oxygen uptakeand carbon dioxide evolution of animal cell culture using material and energybalances. Cytotechnology 29, 159–166.

Zhang, C., Shi, Z., Gao, P., Duan, Z., Mao, Z., 2005. On-line prediction of productsconcentrations in glutamate fermentation using metabolic network model andlinear programming. Biochemical Engineering Journal 25, 99–108.