Effect of Bench-Scale Culture Conditionson Murine IgG Heterogeneity

Maria Marino, Angelo Corti,* Agostino Ippolito, Giovanni Cassani,Giorgio Fassina

Protein Engineering, TECNOGEN S.C.p.A., Parco Scientifico, 81015 Piana diMonte Verna (CE), Italy; telephone: 39-823-612-214; fax: 39-823-612-230e-mail: fassina@technogen.it

Received 12 April 1996; accepted 17 July 1996

Abstract: A stable murine hybridoma cell line, secretingIgG1 antibodies (7H3) against the soluble type I receptorfor Tumor Necrosis Factor (sTNF-R1), was cultivated intwo different bioreactor systems, a hollow fiber and astirred tank fermentor, in order to evaluate the effect ofculture conditions on antibody structural and functionalheterogeneity. Conventional serum-supplemented andserum-free media were chosen for fermentation instirred tank bioreactor, whereas only serum-supple-mented media were used for hollow fiber cultivation. Ex-tent of glycosylation, determined by lectin binding as-says, and charge heterogeneity of murine monoclonalantibodies displayed relevant variations according to thefermentation system used. After complete sugars re-moval by N-glycosidase F treatment, charge heterogene-ity were still observed suggesting the occurrence of ad-ditional modifications at the protein level. In vitro culturein serum-supplemented media carried out with the hol-low fibre system led to higher productivity but greaterantibody charge heterogeneity and differences in lectin-binding profile than cultivation in the stirred tank biore-actor.

Results cumulatively indicated that hybridoma cultiva-tion methods, but also cultivation time, influence anti-body heterogeneity, both in the protein and sugar moi-eties. © 1997 John Wiley & Sons, Inc. Biotechnol Bioeng 54:17–25, 1997.Keywords: hybridoma cell culture; fermentation; MAbheterogeneity

INTRODUCTION

Monoclonal antibodies represent one of the most promisingnew approaches to disease detection and therapy. In thepast, monoclonal antibodies have been produced exclu-sively by growing hybridoma cells in the abdominal cavityof suitable animals, mainly mice (Fike et al., 1991). Themain advantage of this procedure is that small amounts ofhighly concentrated antibody can be made easily and with-out special equipment and technology. However, the prod-uct obtained may contain pathogens and foreign mouse im-munoglobulins which make the purification more problem-atic. Recently, the introduction of hybridoma technologyallowed monoclonal antibody (MAb) production in biore-

actors with a productivity adequate for the industrial devel-opment of diagnostic assays, in vivo imaging immuno-therapy, and immunopurification (Co et al., 1991). One ofthe problems associated with the use of hybridomas forMAb production resides in product homogeneity, since pro-teins expressed by eukaryotic cells are subjected to a varietyof posttranslational modifications (i.e., glycosylation, phos-phorylation) which are strictly dependent on culture condi-tions and which may modify MAbs structure and function-ality. This problem is very important for MAbs with thera-peutic applications, since one of the strictest requirements ofregulatory agencies (such as the Food and Drug Adminis-tration (FDA) in the United States and the Committee forProprietary Medical Productions (CPMP) in the EuropeanCommunity) is that proteins of therapeutic interest haveproduct consistency.

Only a limited number of papers have been published onthe effect of culture conditions on antibody structural andfunctional homogeneity (Maiorella et al., 1993; Patel et al.,1992). Investigations on murine IgM production indicatedthat antibody structure, activity, and pharmacokinetics weredependent on culture conditions, resulting in wide differ-ences in binding activity and clearance rate according to theproduction method used (Maiorella et al., 1993). In a dif-ferent study, a transfectant clonal derivative of the murinemyeloma cell line, NSO, expressing the heavy and lightchain genes of a recombinant human anti-HIV gp120 IgG1,was cultivated in a fed-batch mode, and MAb characteriza-tion from different time points in the process suggested thepresence of differential antibody glycosylation (Robinson etal., 1994). Other works focused mainly on the evaluation ofbioreactors productivity, with little attention to the qualityof the resulting products (Unterluggauer et al., 1994;Kurkela et al., 1993). In this study, we extended the analysisof the effect of culture condition on MAb consistency com-paring the characteristics of a murine IgG1 antibody (7H3)against the soluble form of human Tumor Necrosis Factortype I Receptor (sTNF-RI) (Corti et al., 1994), produced bya hybridoma cell line using two different cultivation sys-tems. First, a hollow fiber bioreactor was used as a hetero-geneous system with serum-supplemented medium, andsecond, a perfusion-stirred tank bioreactor was chosen as a

* Present address:DIBIT, via Olgettina 58, 20132 Milan, ItalyCorrespondence to:G. Fassina

© 1997 John Wiley & Sons, Inc. CCC 0006-3592/97/010017-09

homogeneous system to culture the same cell line with se-rum-supplemented and serum-free medium. Both proce-dures were compared in terms of antibody productivity andafter purification, 7H3 was characterized in terms of chargeheterogeneity, immunoreactivity, glycosylation by bindingexperiments with a set of different lectins, and charge het-erogeneity after complete enzymatic deglycosylation.

MATERIALS AND METHODS

Cell Line and Medium

The cell line secreting 7H3 (IgG1/k) used for this study hasbeen generated by fusing spleen cells of one BALB/c mouseimmunized with sTNF-R1 with NSO myeloma cells as pre-viously described (Corti et al., 1994). Cultivation in hollowfiber and stirred tank bioreactors were performed with aserum-supplemented Dulbecco’s Modified Eagle’s Medium(JRH Biosciences) containing 4.5 g/L glucose, 4.0 mM L-glutamine, 110 mg/L sodium pyruvate and supplementedwith 3.7 g/L sodium bicarbonate (Sigma), 1% non-essentialamino acids (JRH Biosciences), 7% fetal calf serum (JRHBiosciences). A serum-free Ex-Cell 300 (JRH Biosciences),supplemented with 1.6 g/L sodium bicarbonate was used foranother cultivation in stirred tank bioreactor. Preculturepropagation was performed in spinner flasks with a workingvolume of 500 mL (Techne, Cambridge, UK). Spinnerflasks were agitated at 40 rpm in a 37°C, 5% CO2 incubator.

Culture Conditions in Hollow Fiber Bioreactor

Cells were inoculated into the extracapillary space of a hol-low fiber cartridge with a molecular cut-off 10 kDa in anautomated tissue culture system (AcuSyst Maximizer 1000,Endotronics, Inc., Coon Rapids, MN). Optimal culture con-ditions were maintained by computer-controlled regulationof pH, media feed rates, temperature, and dissolved oxygen.The temperature was kept at 37°C ± 0.2°C, the pH wascontrolled with carbon dioxide and maintained at 7.1 ± 0.1.Serum was added only in the extracapillary space. The tis-sue culture system was installed according to the manufac-turer’s instructions. The cartridge was inoculated with 7 ×108 viable cells. Cell growth was monitored by measuringthe glucose consumption and lactate production in intracap-illary space samples every day using an automated glucoseand lactate analyzer (YSI Model 2700 SELECT, YellowSprings Instrument Co., Inc., Yellow Springs, Ohio USA).A lactate concentration below 1 g/L and glucose concentra-tion above 1 g/L were maintained by increasing perfusionrate in the intracapillary space. Samples from extracapillaryspace were taken every day and analyzed for antibody pro-duction.

Culture Conditions in Perfusion Stirred TankBioreactor

Cells were inoculated at a density of 2.0 × 105 viable cells/mL into the 2-liter stirred bioreactor equipped with a perfu-

sion system containing a double stirrer membrane whichcarried the hydrophobic aeration membrane as well as thehydrophilic microfiltration membrane for medium perfusion(B-Braun Biotech International, Melsungen, AG). The stir-rer speed was adjusted to 30-40 rpm and the oxygen contentwas maintained at 40% of air saturation using an automaticgas mixing station with pO2 controller. The temperature waskept at 37°C ± 0.2°C. the pH was controlled with carbondioxide and maintained at 7.1 ± 0.1. The gas passed into theculture medium through a hydrophobic membrane and dif-fused in a bubble-free manner. Cell growth was monitoredby counting viable cells by trypan blue exclusion by meansof a hemocytometer, and measuring the glucose consump-tion and lactate production. A lactate concentration below 1g/L and glucose concentration above 1 g/L were maintainedby increasing perfusion rate. Samples were taken every dayand analyzed for antibody production.

Purification Procedure

Antibodies were first fractionated by ammonium sulphatetreatment. The proteins precipitated at 50% saturation werecollected by centrifugation, dissolved in the smallest pos-sible volume of 20 mM Tris/HCl buffer pH 8.2 (buffer A)and then dialyzed extensively against the same buffer. Thedialyzed samples, clarified by filtration with a 0.2mm mem-brane were loaded into a column (2.6 × 40 cm) of DEAE-Sepharose Fast Flow (Pharmacia, Uppsala, Sweden) previ-ously equilibrated with buffer A at flow rate of 4 mL/min.After sample application and washing, the column waseluted with a linear gradient of 0-0.5 M NaCl in buffer A.MAb was eluted between 0.20-0.35 M NaCl.

The partially purified MAb from hollow fiber culturewere further purified by gel permeation chromatography byusing a Sephacryl S200 HR (Pharmacia, Uppsala, Sweden)column (2.6 × 70 cm). Monoclonal antibodies from perfu-sion stirred tank bioreactor in serum-supplemented mediumcondition were further purified by affinity chromatographyon a Protein A-Sepharose Fast Flow (Pharmacia, Uppsala,Sweden) column (2.6 × 4.0 cm). Samples from ammoniumsulphate precipitation were first dialyzed against 1.5 M gly-cine buffer pH 8.9, containing 3M NaCl, clarified by filtra-tion with a 0.2mm membrane and loaded into the Protein Acolumn previously equilibrated with the same buffer. Aftersample application and washing, the column was elutedwith 0.1 M Na Citrate pH 4 and fractions were immediatelyneutralized with NaOH.

Determination of Protein Concentration

Protein concentration was determined by the Lowry assay,using bovine gamma globulin in the range from 0.2 mg/mLto 1.5 mg/mL for the determination of calibration curves(Lowry et al., 1951; Peterson et al., 1979).

18 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 54, NO. 1, APRIL 5, 1997

Analytical Methods

Isoelectric Focusing

Charge heterogeneity was assessed by isoelectric focusingon a pH 5–8 gradient polyacrylamide gel under denaturingand native conditions using thin-layer polyacrylamide gelsand a Fast System apparatus (Pharmacia, Uppsala, Sweden).In denaturing conditions, dry gel was pre-equilibrated for 30min at room temperature with the following solution: 10 mLH2O, 0.5 mL Nonidet P-40, 9 g urea, and 2 mL Pharmalyte5–8 (Pharmacia, Uppsala, Sweden); gel was then run for900 volthour (Vh) according to the manufacturer’s instruc-tions. In native conditions, samples were loaded on the gelafter a prefocusing step in which the pH gradient wasformed. Gel was run for 600 Vh according to the manufac-turer’s instructions. The following proteins were used as pIstandards: phycocyanin (pI 4.45–4.75),b-lactoglobulin B(pI 5.1), bovine carbonic anhydrase (pI 6.0), human car-bonic anhydrase (pI 6.5), equine myoglobin pI (6.8–7.0),human hemoglobin A (pI 7.1), human hemoglobin C (pI7.5), lentil lectin (pI 7.8–8.2), cytochrome c (pI 9.6).

Gels were stained by the silver nitrate method, washedwith 10% ethanol and 5% acetic acid at 50°C, and fixedwith 20% trichloroacetic acid in water for 5 min at 20°C.Proteins were sensitized with a solution of 8.3% glutaral-dehyde for 6 min at 50°C. After washing with deionizedwater at the same temperature, gels were treated with asolution of 0.4% of silver nitrate for 10 min at 40°C andthen washed again with deionized water at 30°C for 1 min.Gels were then developed at 30°C for 4 min with a solutionof 0.015% formaldehyde in 2.5% sodium carbonate andthen washed with deionized water. All step were carriedout in according to protocols from the manufacturer (Phar-macia, Uppsala, Sweden).

SDS-PAGE and Lectin-Blot

SDS-PAGE and Lectin-Blot analysis, with and without ad-dition of 2-mercaptoethanol, were performed in 4% acryl-amide stacking gel and 12% or 15% acrylamide resolvinggel. The gels were stained with a solution of Coomassie blueR-250 0.05%, acetic acid 10%, isopropyl alcohol 25% atroom temperature for 1 hour and then destained overnightwith a solution of 10% acetic acid, 30% methanol. Proteinsused as molecular weight standards were: myosin (200 kD),E. coli b-galactosidase (116.25 kD), rabbit muscle phos-phorylase b (97.4 kD), bovine serum albumin (66.2 kD),hen egg white ovalbumin (45 kD), bovine carbonic anhy-drase (31 kD), soybean trypsin inhibitor (21.5 kD), hen eggwhite lysozyme (14.4 kD). For Lectin blot, gels were elec-troblotted on a nitrocellulose membrane in Tris/glycinebuffer. Membranes were then probed with digoxigenin-labeled lectins which selectively recognize terminal sugars.The recognition was detected by incubation with anti-digoxigenin-AP and incubation in a staining solution of5-bromo-4-chloro-3-indolyl-phosphate and 4-nitroblue tet-

razolium chloride according to protocols from the manufac-turer (Boehringer Mannheim).

N-Glycosidase F Treatment

Purified IgG (10mg) in 30ml phosphate buffer was treatedwith 2 ml N-Glycosidase F (200U/mL) (Boehringer Mann-heim), in the presence of 0.1% SDS. Samples were boiledfor 2 min and diluted with 70ml of 10 mM EDTA contain-ing 1% Nonidet P-40. After incubation for 24 h at 37°C,samples were subjected to SDS-PAGE and electroblotted ona sheet of nitrocellulose membrane. Sugars were revealedby Dig-Glycan Detection Kit based on oxidation of carbo-hydrate with sodium metaperiodate, and labelling with di-goxigenin-succinyl-e-amido-caproic acid hydrazide. Afterwashing, the membrane was incubated with a solution ofanti-digoxigenin-AP for 1 h and incubated in a stainingsolution of 5-bromo-4-chloro-3-indolyl-phosphate and 4-ni-troblue tetrazolium chloride according to protocols from themanufacturer (Boehringer Mannheim GmbH, Mannheim,Germany).

ELISA of MAb 7H3

Antibody determination was carried out as follows: PVCmicrotiter plates were incubated with 40mg/mL goat anti-mouse (IgG) immunoglobulin solution (GAM) (50ml/well)in 0.05 M sodium phosphate, 0.15 M NaCl pH7.2 solution(PBS), overnight at 4°C. After washing three times withPBS the plates were blocked with 200ml of PBS containing3% BSA and incubated for 2 h at 37°C. The plates werethen washed with PBS containing 0.05% Tween 20 (PBS-T)and filled with 50ml of 7H3 samples, previously dilutedwith PBS-T containing 1% BSA, 1% normal goat serum(PBS-BTN). After incubation for 1 h at 37°C plates werewashed again with PBS-T and incubated for 1 h at 37°Cwith 50 ml/well of HRP-labeled GAM solution, diluted 1:1000 with PBS-BTN. The plates were then left for 1 h at37°C and washed with PBS-T. The chromogenic reactionwas carried out by filling each well with 100ml of 2.2-Azino-di-[3-ethylbenzthiazoline sulfonate] (ABTS) chro-mogenic substrate solution freshly prepared according toprotocols from the manufacturer (Boehringer Mannheim).The color was allowed to develop for 30 min and the ab-sorbance at 405 nm of each well was determined with aELISA plate reader (Labsystems Multiskan BICHRO-MATIC, Labsystems Oy, Helsinki, Finland).

Receptor binding activity was determined essentially asdescribed before using PVC microtiter plates coated withsTNF-R1 (2mg/mL, 50 ml/well) in 100 mM Na2CO3 pH9.6, overnight at 4°C.

RESULTS

Cell Culture

Cells for perfusion stirred tank bioreactor (2 L workingvolume) were propagated in spinner flasks and then inocu-

MARINO ET AL.: CULTURE CONDITIONS AND MONOCLONAL ANTIBODY HETEROGENEITY 19

Figure 1. Stirred tank culture of 7H3 cell line. Panel A: viable cell concentration (m) andmonoclonal antibody concentration (h) in perfusion stirred tank culture carried out with serum-supplemented medium. Panel B: viable cell concentration (m) and monoclonal antibody concen-tration (h) in perfusion stirred tank culture carried out with serum-free medium. Panel C: % cellviability in perfusion stirred tank-culture carried out with (m) and without (n) serum.

20 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 54, NO. 1, APRIL 5, 1997

lated with a density of 2.0 x 105 viable cells/mL. Cultureconditions were monitored by counting viable cells and bymeasuring glucose consumption and lactate production. Instirred tank culture (Fig.1A) cells grew in a serum-supplemented medium to a maximum viable cell density of1 × 106 cells/mL in batch mode, after that perfusion modestarted up. In perfusion mode, cells were allowed to reach amaximum viable cell density of 11 × 106 cells/mL. Cellswere then maintained in a steady state making a continuousdilution of culture and keeping viability between 88% and95% (Fig. 1C). MAb production was low during the first lagphase and the following log phase but increased when thestationary phases was reached. During this time the produc-tivity was in the range of 0.5–1.4 g/day. Culture with serum-supplemented medium was carried out for 26 days with atotal production of 11.8 g of monoclonal antibody.

When serum-free medium was used cell growth de-creased (Fig. 1B). Viable cells grew to a maximum of 8.3 ×105 cells/mL in batch mode, followed by perfusion mode toreach a density of 7.4 × 106 cells/mL, and then cells weremaintained in the stationary phase. Cell viability of culturein serum-free medium was lower than under serum-supplemented conditions, ranging from 58% to 80% (Fig.1C). Monoclonal antibody production decreased during pro-liferation phase but increased again when the stationaryphase was reached. During this time, culture productivitywas in the range of 0.3-1.2 g/day with a total production of2.3 g of product.

Cells for hollow fiber bioreactor were precultured in spin-ner flask until a total number of 7 × 108 viable cells wasreached and then inoculated into the extracapillary space ofan hollow fiber cartridge. Inoculated cells had a long lagphase as shown by the glucose and lactate concentration

profile (Fig. 2). Maximum cell density was reached afterabout 30 days and extracapillary space was completelyfilled with cells. At the cycle end, 7H3 total production wasof 10.9 g. In both systems, the perfusion rate was adjustedduring cultivation in order to maintain lactate productionbelow 1 g/L and glucose consumption above 1 g/L. Figure3 shows the comparison between the productivity of the twoculture systems.

Analytical Studies

Monoclonal antibody, produced from different culture sys-tems and sampled at different time intervals, was purified asdescribed in Material & Methods and characterized by SDS-PAGE analysis, isoelectric focusing, and functional activity.

As shown in Figure 4, analysis by non-reducing SDS-PAGE indicated that 7H3 preparations isolated from eachtime-point were of roughly equivalent purity (greater than95%) and showed equivalent electrophoretic mobilities. Noheterogeneity was observed by SDS-PAGE analysis underreducing conditions in the apparent molecular weight oflight and heavy chains (Fig. 4).

Isoelectric focusing gel electrophoresis (Fig. 5) showeddifferences in the charge heterogeneity of samples derivingfrom different culture systems and different culture time-intervals. Samples from stirred tank culture carried out withserum-supplemented media showed different species ofMAb with apparent pIs ranging from 6.90 to 7.75 (Fig. 5C),while pIs varying from 6.95 to 7.75 were shown by MAbscultured in serum-free conditions (Fig. 5B). The pI rangedisplayed by 7H3 produced with the hollow fiber systemwas wider, ranging from 6.70 to 7.75 (Fig. 5A). Similar

Figure 2. Hollow fiber culture of 7H3 cell line. Glucose consumption (d) and lactate production (s).

MARINO ET AL.: CULTURE CONDITIONS AND MONOCLONAL ANTIBODY HETEROGENEITY 21

results were obtained for MAbs isolated from two indepen-dent cultures of each type (data not shown). Different pro-tocols were used for MAb purification, but recovery yieldsfrom supernatants were high (> 60%) and similar for all the

three different production systems investigated (Table I).This suggested that enrichment of one charge or glycoformpopulation over the other did not occur.

Glycosylation studies were carried out by lectin-blotanalysis using a panel of different lectins. Lectins used inthis study wereSambucus nigra Agglutinin(SNA) thatbinds to terminal sialic acid which isa 2→6 linked togalactose residues,Galanthus nivalis Agglutinin(GNA) thatbinds to terminal mannose residues,Aleuria aurantia(AAA) that binds to fucose (a1→6) GlcNAc, and Conca-navalin A (Con A) that binds to mannose cores containingbiantennary complex-type structures of sugar chains. Lec-tin-blots of heavy chains of 7H3 deriving from differentculture conditions and at different time intervals are shownin Figure 6. The binding of SNA to 7H3 heavy chain in-creased according to the hollow fiber culture age, whereasno terminal sialic acid was detected in stirred tank products.The degree of GNA binding to 7H3 heavy chain was dif-ferent for samples deriving from different culture systems,showing higher levels of truncated structures with terminalmannose residues for IgG obtained from hybridoma cul-tured in stirred tank bioreactor under serum-supplementedconditions than IgG from other cultivation systems. Thebinding of lectins as AAA and ConA to 7H3 heavy chainswas unrelated to culture conditions and time, since allsamples reacted similarly with the two lectins, showing that7H3 antibody was fucosylated (data not shown).

7H3 reactivity towards the antigen was kept unaffectedindependently from the production system used, thus sug-gesting that glycosylation had a limited role in antibodyreceptor recognition. A data comparison on the character-istics of 7H3 from different culture conditions is shown in

Figure 3. MAb production of 7H3 hybridoma cells in different culture systems. Daily MAb yield from perfusion stirred tank culturecarried out with serum-supplemented medium (j), serum-free conditions (h), and hollow fibre culture (l).

Figure 4. Apparent molecular weight of MAb 7H3 isolated from varioustime-points from different culture systems. Upper panel: SDS-PAGEanalysis under non-reducing conditions. Lower panel: SDS-PAGE underreducing conditions. Lane markers indicate: from 1 to 6 samples fromhybridoma cells cultured in hollow fiber bioreactor taken at different times(days 21, 31, 38, 40, 46, 49); from 7 to 9 samples from hybridoma cellscultured in serum-supplemented suspension taken at different times (days15, 19, 25); samples 10 and 11 from hybridoma cultured in serum-freestirring bioreactor taken at different times (days 6, 12). Each lane wasloaded with 5mg of protein.

22 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 54, NO. 1, APRIL 5, 1997

Table II. Extent of antibody heterogeneity was further ex-amined, preparing fully deglycosylated 7H3 by treatmentwith N-glycosidase F. As shown in Figure 7A, isoelectro-focusing of enzyme-treated samples indicated the presenceof charge heterogeneity even at the protein level. Deglyco-sylated antibodies showed a measurable acidic pI shift, sug-

gesting that sugars removal led to the occurrence of struc-tural modifications affecting the protein isoelectric point.This effect has been observed before for other deglyco-sylated proteins (Righetti, 1990). Samples from hollow fibercultivation displayed a wider pI range than samples derivingfrom stirred tank fermentors, as seen before for non-deglycosylated antibodies. Enzyme treatment was effectivein fully removing all the sugar moiety since deglycosylatedsamples were unreactive to staining with the Dig-Glycandetection kit (Figure 7B).

DISCUSSION

Cell growth conditions play a crucial role in protein post-translational processes such as glycosylation. Glucose con-centration and the presence of proteins or growth factors(added to increase cell productivity in serum-free medium)can influence protein glycosylation (Gooche¨e and Monica,1990). A variety of other environmental components caninfluence oligosaccharide processing of glycoproteins, in-cluding EDTA (West and Brownstein, 1988), HEPES(Megaw and Johnson, 1979) and hydrogen ion concentra-tion (Matlin et al., 1988; Borys et al., 1993). Changes onglycosylation have many consequences on the structure andbiological activity of glycoproteins. The presence of gly-cans on glycoproteins confers good solubility avoiding ag-gregation (Arakawa et al., 1991) while deglycosylation canbe followed by complete biological inactivation (Oh-Eda etal., 1990). In addition, sugars influence many biologicalproperties of glycoproteins such as immunogenicity, eitherbecause carbohydrates can be a part of the epitope itself(Feizi and Childs, 1987), or because they can mask existingantigenic sites on the peptide backbone. In vivo, sugar

Figure 6. Lectins binding to MAb 7H3 isolated from various time-pointsin hollow fiber and stirred tank bioreactors. Lanes from 1 to 6: samplesfrom hybridoma cells cultured in hollow fiber bioreactor taken at differenttimes (days 21, 31, 38, 40, 46, 49); lanes from 7 to 9: samples from cultureperformed in perfusion stirred tank bioreactor in serum-supplemented con-ditions taken at different times (days 15, 19, 25); lanes 10 and 11: samplesfrom hybridoma cultured in serum-free stirring bioreactor taken at differenttimes (days 6, 12). Lectin-blot was performed as described in Materials andMethods using carboxypeptidase Y as positive controls for GNA, trans-ferrin as positive control for SNA, while asialofetuin was used as negativecontrol. Lanes were loaded with 5mg of protein as determined by theLowry assay.

Table I. 7H3 purification yields.

Culture mode Day of culturePurification yieldsa

%

Serum-supplementedstirred tank 5 62

15 63.319 60.625 62

Serum-freestirred tank 6 61.3

10 64.312 62.915 62.2

Serum-supplementedhollow fiber 21 60

31 61.538 60.240 61.646 6249 60.3

aDetermined by ELISA assay on antigen coated plates.

Figure 5. Isoelectric focusing of MAb 7H3 isolated from various time-point from different culture systems. Lanes from 1 to 6: samples fromhybridoma cells cultured in hollow fiber bioreactor taken at different times(days 21, 31, 38, 40, 46, 49); lanes from 7 to 10: samples from hybridomacells cultured in serum-free suspension taken at a different times (days 6, 10,12, 15); lanes from 11 to 14: samples from hybridoma cultured in serum-supplemented stirring bioreactor taken at different times (days 5, 15, 19,25). Each lane was loaded with 200 ng of protein.

MARINO ET AL.: CULTURE CONDITIONS AND MONOCLONAL ANTIBODY HETEROGENEITY 23

chains of serum IgGs from various species are characterizedby sialylated or non-sialylated biantennary complex-typestructures in which the presence or absence of fucose, N-acetylglucosamine and galactose residues produce relevantmicro-heterogeneity (Takasaki et al., 1982). Although thereis one consensus sequence on each IgG Fc H chain for anasparagine-linked sugar chain, the variable region of L andH chain may also be glycosylated (Abel et al., 1968; Spiel-berg et al., 1970). The carbohydrate on the Fc portion ofimmunoglobulin has been shown to be involved in differentevents like complement activation through the binding toClq (Duncan and Winter, 1988), Fc receptor binding, anti-

body-dependent cellular cytotoxicity, and feedback immu-nosuppression, in many cases without affecting antigenbinding (Maiorella et al., 1993; Robinson et al., 1994).

Over the last few years, many researchers have comparedthe glycosylation profiles of immunoglobulins resultingfrom different methods of cell culture of hybridomas. Sev-eral studies have shown that monoclonal antibodies pro-duced in ascites tumors lack sialic acid, whereas hybrido-mas cultured in serum-supplemented media lead to under-galactosylated glycan structures (Patel et al., 1992). In otherstudies, hybridomas producing IgM have been grown inascites, serum-free suspension, or serum-containing hollow-fiber perfusion culture, detecting wide differences in theextent of antibody glycosylation and differences in pharma-cokinetic parameters (Maiorella et al., 1993). In our study,7H3 MAb prepared from hollow fiber culture showed avariable pl distribution, depending on culture time, withmore acidic bands in samples taken at the early stage than atthe end of culture, while lectin-blot analysis indicated thepresence of more sialic acid in MAbs produced in the latestage of culture than in the early stage. MAbs producedfrom stirred tank technology and especially in serum-freeconditions were more homogeneous in charge profile, sincepl range variation was considerably limited. Since it isknown that glycosylation affects clearance via the hepaticasialo-glycoprotein receptor (Ashwell and Harford, 1982),reticuloendothelial cell mannose/N-acetylglucosamine re-ceptor or fucose receptor (Pay et al., 1980; Lehrman et al.,1986), the different pharmacokinetic profile displayed bymonoclonals obtained from different cultivation systems(Maiorella et al., 1993) it is not surprising.

No significant differences in antigen binding activitywere detected in the different preparation. Complete sugarremoval from these samples did not completely eliminatecharge heterogeneity, suggesting the presence of additionalsources of heterogeneity, such as differential modificationof the carboxyl-terminus of the protein moiety, or differ-ences in phosphorylation or deamidation.

Our results support existing data in the literature on an-

Figure 7. Isoelectric Focusing of deglycosylated MAb7H3 isolated fromdifferent culture systems. Panel A: Heterogeneity of deglycosylatedsamples. Lanes 1, 3, and 5 correspond to deglycosylated samples, whilelanes 2, 4, and 6 to non-deglycosylated samples from hollow fiber (day 21,Lanes 1 and 2), serum-supplemented culture (day 25, lanes 3 and 4), andserum-free culture (day 12, lanes 5 and 6). Panel B: Evaluation of N-glycosidase F sugars removal. Enzyme treated (Lanes 1, 3, and 5) anduntreated (+) samples of MAb 7H3 were subjected to SDS-PAGE and afterblotting, stained with the Dig-glycan detection kit. Lanes were loaded with200 ng of protein as determined by the Lowry assay.

Table II. Characterization of MAb 7H3 from different culture modes.

Culture modeMolecularWeighta

pIrangeb

Receptorbindingactivityc

Mannosecontentd

Sialicacid

contentd

Serum-supplementedstirred tank 50 & 25 KDa 6.90–7.75 unaffected high low

Serum-freestirred tank 50 & 25 KDa 6.95–7.75 unaffected medium low

Serum-supplementedhollow fiber 50 & 25 KDa 6.70–7.75 unaffected medium medium

aDetermined by SDS-PAGE analysis onb-mercaptoethanol treated samples.bDetermined by isolelectric focusing.cDetermined by ELISA.dDetermined by lectin-blot analysis comparing band intensity.

24 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 54, NO. 1, APRIL 5, 1997

tibody heterogeneity (Maiorella et al., 1993; Patel et al.,1992; Robinson et al., 1994; Takasaki et al., 1982) andprovide additional indications that extent of antibody gly-cosylation depends not only on the cultivation process used,but also on the cultivation time, which needs to be carefullycontrolled to maintain product consistency. Furthermore,data obtained in this study suggest that culture conditionsaffect antibody heterogeneity even at the protein level. Thisadditional source of antibody heterogeneity has not beenreported before.

Special care should, therefore, be taken in planning scale-up and production development for therapeutic antibodies toevaluate, from the beginning, bioreactors and growth con-ditions that provide maintenance of product characteristics.

This research was supported in part by a contract of the Ministerodell’Universita e della Ricerca Scientifica e Tecnologica as-signed to TECNOGEN S.C.p.A. within the Programma Nazio-nale Biotecnologie Avanzate.

References

Abel, C. A., Spielberg, H. L., Grey, H. 1968. The carbohydrate content offragments and polypeptide chains of humangG-myeloma patients ofdifferent heavy chain. Biochem.7: 1271–1278.

Arakawa, T., Yphantis, D. A., Lary, J. W., Narhi, L. O., Lu, H. S., Pre-strelski, S. J., Clogston, C. L., Zsebo, K. M., Mendiaz, E. A., Wypych,J. 1991. Glycosylated and unglycosylated recombinant-derived humanstem cell factors are dimeric and have extensive regular secondarystructure. J. Biol. Chem.266: 18942–18948.

Ashwell, G., Harford, J. 1982. Carbohydrate-specific receptors of theLiver. Ann. Rev. Biochem.51: 531–554.

Borys, M. C., Linzer, D. I. H., Papoutsakis E. T. 1993. Culture pH affectsexpression rates and glycosylation of recombinant mouse placentallactogen proteins by chinese hamster ovary (CHO) cells. Bio/Technol.11: 720–724.

Co, M.-S., Deschamps, M., Whitley, R. J., Queen, C. 1991. Humanizedantibodies for antiviral therapy. Proc. Natl. Acad. Sci.88: 2869–2873.

Corti, A., Bagnasco, L., Cassani, G. 1994. Identification of an epitope ofTumor Necrosis Factor (TNF)-Receptor Type 1 (p55) recognized by aTNF-a-antagonist monoclonal antibody. Lymphokine and CytokineResearch13, 3: 183–190.

Day, J. F., Thornburg, R. W., Thorpe S. R., Baynes, J. W. 1980. Carbohy-drate-mediated clearance of antibody antigen complexes from the cir-culation. J. Biol. Chem.255: 2360–2365.

Duncan, A. R., Winter, G. 1988. The binding site for C1q on IgG. Nature332: 738–740.

Feizi, T., Childs, R. A. 1987. Carbohydrates as antigenic determinants ofglycoproteins. Biochem.245: 1–11.

Fike, R., Jayme, D., Weiss, S. A. 1991. Monoclonal antibody enhancementin protein-free and serum supplemented hybridoma culture. Amer. Imt.Biotech. Lab. June 1991, p. 40–42.

Goochee, C. F., Monica, T. 1990. Environmental effects on protein glyco-sylation. Bio/Technol.8: 421–427.

Kurkela, R., Fraune, E., Vihko, P. 1993. Pilot-Scale production of murinemonoclonal antibodies in agitated, ceramic-matrix or hollow-fiber cellculture systems. BioTechniques15: 674–683.

Lehrman, M. A., Pizzo, S. V., Imber, M. J., Hill, R. L. 1986. The bindingof fucose-containing glycoproteins by hepatic lectins. J. Biol. Chem.261: 7412–7418.

Lowry, O. H., Rosebrough, N. J., Farr, A. L., Randall, R. J. 1951. Proteinmeasurement with the Folin-phenol reagent. J. Biol. Chem.193:265–275.

Maiorella, B. L., Winkelhake, J., Young, J., Moyer, B., Bauer, R., Hora,M., Andya, J., Thomson, J., Patel, T., Parekh, R. 1993. Effect ofculture conditions on IgM antibody structure, pharmacokinetics andactivity. Bio/Technol.11: 387–392.

Matlin, K. S., Skibbens, J., McNeil, P. L. 1988. Reduced extracellular pHreversibly inhibits oligomerization, intracellular transport, and pro-cessing of the influenza hemaglutinin in infected Madine-Darby caninekidney cells. J. Biol. Chem.263: 11478–11485.

Megaw., J. M., Johnson, L. D. 1979. Glycoprotein synthesized culturedcells: effects of serum concentrations and buffers on sugar content.Proc. Soc. Exp. Biol. Med.161: 60–65.

Oh-Eda, M., Hasegawa, M., Hattori, K., Kuboniwa, H, Kojima, T., Orita,T., Tomonou, K., Yamazaki, T., Ochi, N. 1990. O-Linked sugar chainof human granulocyte colony-stimulating factor protects it against po-lymerization and denaturation allowing it to retain its biological ac-tivity. J. Biol. Chem.265: 11432–11435.

Patel, T. P., Parekh, R. B., Moellering, B. J., Prior, C. P. 1992. Differentculture methods lead to differences in glycosylation of a murine IgGmonoclonal antibody. Biochem. J.285: 839–845.

Peterson, Gary L., 1979. Review of the polin phenol protein quantitationmethod of Lowry, Rosebrough, Farr, and Randall. Analy. Biochem.100: 201–220.

Righetti, P. G., 1990. Immobilized pH gradients: theory and methodology.Elsevier, Amsterdam.

Robinson, D. K., Chan, C. P., Yu Ip, C., Tsai, P. K., Tung, J., Seamans,T. C., Lenny, A. B., Lee, D. K., Irwin, J., Silberklang, M. 1994. Char-acterization of a recombinant antibody produced in the course of a highyield fed-batch process. Biotech. & Bioeng.44: 727–735.

Spielberg, H. L., Abel, C. A., Fishkin, B. C., Grey, H. 1970. Localizationof the carbohydrate within the variable region of light and heavychains of humangG-myeloma patients. Biochem.9: 4217–4223.

Takasaki, S., Mizuochi, T., and Kobata, A. 1982. Hydrazinolysis of as-paragine-linked sugar chains to produce free oligosaccharides. Meth.Enzym.83: 263–268.

Unterluggauer, F., Doblhoff-Dier, O., Tauer, C., Jungbauer, A., Gaida, T.,Reiter, M., Schmatz, C., Zach, N., Katinger, H. 1994. Stable, continu-ous large-scale production of human monoclonal HIV-1 antibody us-ing a computer-controlled pilot plant. BioTechniques16: 140–147.

West, C. M., Brownstein, S. A. 1988. EDTA treatment alters protein gly-cosylation in the cellular slime moldDictyostelium discoideum.Exp.Cell Res.175: 26–36.

MARINO ET AL.: CULTURE CONDITIONS AND MONOCLONAL ANTIBODY HETEROGENEITY 25