culture temperature modulates aggregation of recombinant antibody in cho cells

ARTICLE

Culture Temperature Modulates Aggregation ofRecombinant Antibody in CHO Cells

Natalia Gomez,1 Jayashree Subramanian,1 Jun Ouyang,2 Mary D.H. Nguyen,2

Matthew Hutchinson,3 Vikas K. Sharma,4 Andy A. Lin,1 Inn H. Yuk1

1Department of Early Stage Cell Culture, Genentech, Inc., 1 DNA Way, MS 32,

South San Francisco, California 94080-4990; telephone: 650-467-2539; fax: 650-225-2006;

e-mail: [email protected] of Protein Analytical Chemistry, Genentech, Inc., South San Francisco,

California3Department of Purification Development, Genentech, Inc., South San Francisco, California4Department of Early Stage Pharmaceutical Development, Genentech, Inc.,

South San Francisco, California

Received 11 April 2011; revision received 16 July 2011; accepted 26 July 2011

Published online 30 September 2011 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/bit.23288

ABSTRACT: During production of therapeutic monoclonalantibodies (mAb), it is highly desirable to remove andcontrol antibody aggregates in the manufacturing processto minimize the potential risk of immunogenicity topatients. During process development for the productionof a recombinant IgG in a CHO cell line, we observedatypical high variability from 1 to 20% mAb aggregatesformed during cell culture that negatively impacted anti-body purification. Analytical characterization revealed theIgG aggregates were mediated by hydrophobic interactionslikely caused by misfolded antibody during intracellularprocessing. Strikingly, data analysis showed an inverse cor-relation of lower cell culture temperature producing higheraggregate levels. All cultures at 378C exhibited �5% aggre-gates at harvest. Aggregate levels increased 4–12-fold in 338Ccultures when compared to 378C, with a corresponding 2–4-fold increase in heavy chain (HC) and light chain (LC)mRNA. Additionally, 378C cases showed a greater excess ofLC to HC mRNA levels. Endoplasmic reticulum (ER) chap-erone expression and ER size also increased 25–75% at 338Cversus 378C but to a lesser extent than LC and HC mRNA,consistent with a potential limiting ER folding capacity at338C for this cell line. Finally, we identified a 2–5-foldincrease in mAb aggregate formation at 338C comparedto 378C cultures for three additional CHO cell lines. Takentogether, our observations indicate that low culture temper-ature can increase antibody aggregate formation in CHOcells by increasing LC and HC transcripts coupled withlimited ER machinery.

Biotechnol. Bioeng. 2012;109: 125–136.

� 2011 Wiley Periodicals, Inc.

KEYWORDS: aggregates; CHO; ER; chaperone; light chain;heavy chain

Introduction

Therapeutic recombinant monoclonal antibodies are com-monly manufactured in Chinese Hamster Ovary (CHO)cells and chromatographically purified to produce drugsubstance. Since protein aggregates can induce immuno-genic responses in animals and humans (Braun et al., 1997;Gamble, 1966; Moore and Leppert, 1980; Rosenberg, 2006),antibody aggregates in the drug product need to becontrolled.

During process development for monoclonal antibody Ausing a stably transfected CHO cell line, we observedatypical high variability in antibody aggregates formed inproduction culture. The aggregates varied from 1 to 20% atharvest throughout multiple bioreactor runs. Althoughdownstream purification steps could effectively reduceaggregates to an acceptable level, understanding the natureof the aggregates and cell culture conditions that decreaseaggregation could further improve process yield, control,and robustness.

Production of monoclonal antibodies in CHO cells canyield antibody aggregates that vary with amino acidsequence, cell line characteristics, clonal variation, andcell culture medium and conditions. In general, antibodyaggregates can be of different types depending on solubility,covalent/non-covalent interactions, reversibility, andmonomer denaturation (Cromwell et al., 2006; Philo andArakawa, 2009).

Previous studies reported aggregate formation for arecombinant protein or antibody in CHO cell culture.Cromwell et al. showed extracellular non-covalent antibodyaggregate formation in cell culture, and disulfide-mediatedantibody aggregates formed by unpaired thiols (CromwellCorrespondence to: N. Gomez

� 2011 Wiley Periodicals, Inc. Biotechnology and Bioengineering, Vol. 109, No. 1, January, 2012 125

et al., 2006). Lee et al. found low mRNA LC/HC ratioscorrelated with increased antibody aggregates (Lee et al.,2009). Schroder et al. reported increased aggregation ofantithrombin III with elevated expression, and saturation offolding capacity with the highest amplified cell line(Schroder and Friedl, 1997; Schroder et al., 2002). Pendseet al. also showed that secretion efficiency decreased andintracellular degradation increased with increased copynumber (Pendse et al., 1992).

Antibody formation depends on correct folding andassembly of two light chains (LC) and two heavy chains(HC) in the endoplasmic reticulum (ER). In particular, HCfolding needs chaperone BiP, oxidoreductase PDI and thepresence of LC (Hendershot, 1990; Kaloff and Haas, 1995;Lee et al., 1999; Mayer et al., 2000; Vanhove et al., 2001).Studies on plasma cell differentiation showed increased ERmachinery supports folding and assembly of immunoglo-bulins (Kirk et al., 2010; Shaffer et al., 2004). Finally, ERquality control eliminates misfolded chains that dislocate tocytoplasm and degrade in proteosomes (Fagioli et al., 2001).

This manuscript describes the studies on aggregateformation during CHO cell culture. Analytical characteri-zation revealed the antibody aggregates were dissociableunder denaturing conditions but not under reducing or highsalt environments, and formed intracellularly. Based onthese results, we proposed aggregation was caused bymisfolded domains during intracellular processing. Weinvestigated LC and HC expression, ER size, chaperonecontent, and correlations of aggregates with cell cultureconditions and performance. Overall, this investigationprovided insights on aggregate formation caused byenhanced transcription and limited folding capacity.Furthermore, cell culture temperature modulated aggregateformation and could minimize the impact on downstreampurification of the therapeutic antibody.

Materials and Methods

CHO Cell lines

Stable cell line X secreting IgG1-based antibody A wasderived from a suspension-adapted CHO dihydrofolatereductase-deficient (DHFR-) host. The host cells weretransfected with a bicistronic DNA plasmid encoding genesfor DHFR, LC, and HC and cultured with a proprietarymedium containing methotrexate (MTX). Stable cell linesW, Y and Z secreting a non-glycosylated version of antibodyA, antibody A (sister clone of cell line X), and a differentantibody B, respectively, were developed similarly.

Production Cell Culture

Cells were cultured for production experiments in duplicateshake flasks (SF) (125 or 250mL, Corning, Corning, NY) orduplicate 2 L sparged vessels (Applikon, Foster City, CA)

with digital control units (B. Braun Biotech, Allentown, PA)for 12–14 days. Cells were seeded at 1� 106 cells/mL in aproprietary basal medium (derived from Ham’s F12/Dulbecco’s Modified Eagle’s Medium). Additional nutrients(20% v/v) were fed as a single bolus on day 3. SF werecultured at constant 378C in a humidified incubator with5% CO2, or subsequently moved to a 31–358C incubator48 h post-inoculation. Bioreactor cultures were maintainedat constant 378C, pH 6.9, and 30% of air saturation, orsubsequently shifted to lower temperature (33–358C) 48 hpost-inoculation. SF and 2 L bioreactors were sampled (days3, 6, 9, and 13 for SF and daily for bioreactors) and analyzedfor viable cell concentration (VCC) and viability using aVi-Cell XR (Beckman Coulter, Fullerton, CA), and for pH,glucose and lactate using a Bioprofile 400 (Nova Biomedical,Walthan, MA). Supernatant samples were taken on the samedays for titer analysis using protein A HPLC and aggregatequantification using size exclusion chromatography (SEC).Cell pellets were collected on days 3, 6, and 9 for mRNAanalysis and western blotting. Integrated viable cellconcentration (IVCC) and cell-specific productivity(qp) were calculated as previously described (Gomezet al., 2010).

For large-scale production, 400 L (ABEC, Bethlehem, PA)and 2,000 L (B. Braun Biotech, Allentown, PA) bioreactorswere used. Temperature was shifted from 37 to 338C after48 h, pH was maintained at 6.9 and DO at 30% for theduration of the run.

For cell-free supernatant incubation studies, 25mL of cellculture from days 3, 6, and 9 was removed aseptically fromthe SF, centrifuged to remove cells, and transferred toTubeSpins (Trasadingen, Switzerland). These cell-freesupernatants were further incubated in the TubeSpins at33–378C and 250 rpm in the humidified incubator for therest of the experiment.

For mRNA stability studies, actinomysin D (Amresco,Solon, OH) was added on day 6 post-inoculation to 338C or378C SF cultures for a final concentration of 5mg/mL. Cellpellets were collected 4, 8, 24, and 32 h post-addition formRNA extraction and analysis.

qRT-PCR

Culture samples were centrifuged (4� 106 cells, 5,000 rpm,2min) and cell pellets were stored (�708C) for PCR analysis.mRNA extraction and purification were performed using aRNeasy kit (Qiagen, Valencia, CA). For qRT-PCR, mRNA(12.5mg) was added to a reaction mixture prepared withRT-PCR kit (Applied Biosystems, Foster city, CA) contain-ing forward and reverse primers (0.3mM each) and probe(0.2mM). Primers were prepared in-house for LC, HC, PDI,BiP, GRP94, ATF6, XBP, and beta-2-microglobulin (B2m,used as the housekeeping gene). An ABI Prism 7700Sequence Detection System (Applied Biosystems, Fostercity, CA) was used. mRNA concentrations were quantifiedfrom the Ct with the calibration curve method.

126 Biotechnology and Bioengineering, Vol. 109, No. 1, January, 2012

Western Blotting of Cell Lysates

Day 9 cell pellets (1� 106 cells, 5,000 rpm, 2min) stored at�708C were lysed with a lysis buffer (Cell Signaling, Boston,MA) containing protease inhibitor cocktail (Roche,Manheim Germany). Total protein content was determinedwith a Bradford assay (Bradford Assay Kit, Pierce, Rockford,IL). For reduced samples, 0.1M DTT was added. Tris-glycine SDS sample buffer was added to cell lysates andheated. Cell lysate (10mg) was loaded onto a 1mm, Nu-PAGE 4–12% Bis-Tris gel (Invitrogen, Carsbad, CA) andrun in a Nu-PAGE electrophoresis apparatus (Invitrogen,Carsbad, CA) using 1� MOPS buffer (Invitrogen, Carsbad,CA). Proteins were transferred to a nitrocellulose membrane(BioRad, Hercules, CA) with the XCell IITM Blot Module(Invitrogen, Carsbad, CA). Membranes were blocked andincubated overnight with either a goat anti-human IgG Fcantibody HRP conjugated (Bethyl Laboratories,Montgomery, TX), goat anti-human Kappa LC HRPconjugated (Bethyl Laboratories, Montgomery, TX) oranti-b-actin antibody (Sigma-Aldrich, St. Louis, MO). Forb-actin staining, additional secondary antibody incubationusing an anti-mouse IgG antibody conjugated to HRP (GEHealthcare Biosciences, Piscataway, NJ) was performed(1 h). Membranes were incubated with an ECL kit(Amersham-GE Healthcare Biosciences, Piscataway, NJ),followed by film exposure.

Fluorescence Microscopy

Day 9 culture samples were centrifuged (3� 106 cells,1,500 rpm, 5min, 48C) and cells fixed with�208Cmethanolfor 3.5min. Cells were centrifuged and washed twice withPBS. Cells were incubated in 1% BSA (Surmodics, EdenPrairie, MN) for 30min at 378C, followed by centrifugationand re-suspension in 1% BSA containing mouse anti-PDI(Thermo-Pierce, Rockford, IL) and incubated overnightwith agitation. Cells were washed with PBS three times. Goatanti-mouse, Alexa 488 conjugated antibody (Invitrogen,Carlsbad, CA) was added in 1% BSA and incubated for20min at room temperature. Cells were washed with PBSthree times, re-suspended in PBS and added to poly-L-lysine(0.01% Sigma, St. Louis, MO) coated chamber slides (NalgeNunc, Rochester, NY). After 2 h incubation, PBS wasremoved and slides mounted (Vectashield with DAPI,Vector, Burlingame, CA) and sealed with coverslips. Cellswere imaged with an inverted phase-contrast and fluores-cence microscope (Olympus IX81). Images were capturedwith a color CCD video camera and analyzed with ImageJ(NIH).

Size-Exclusion Chromatography (SEC)

SEC was conducted using an Agilent 1100 high performanceliquid chromatography (HPLC) system equipped witha binary pump (Agilent Inc. Santa Clara, CA). Sample

was injected onto a TSK-GEL G3000SWxl column(7.8� 300mm) (Tosoh Bioscience, Montgomeryville,PA). The molecular size variants were separated at ambienttemperature and eluted isocratically with a mobile phaseconsisting of 0.2M potassium phosphate, 0.25M potassiumchloride, pH 6.2 with a flow rate of 0.5mL/min. Proteinelution was monitored by 280 nm absorbance. The nature ofthe aggregates was studied using modified SEC methods.In one experiment, neat injections (�30mg) of samplescontaining low, medium, and high levels of aggregates wereloaded on the SEC column and eluted with the mobile phasecontaining 6M guanidine. The second experiment involveda 30-min incubation of samples with 10mM DTT at 378C,followed by neat injections to the SEC column and elutedwith the mobile phase containing 10mM DTT. The thirdexperiment used the mobile phase containing either 0.5Mor 1.0M KCl. Samples were injected neat.

Multi-Angle Light Scattering (MALS)

An aggregate-enriched (87% aggregates, 13% monomer)and monomer-enriched (99.8% monomer, 0.2% aggregate)fractions were characterized with MALS to determine molarmass distribution. An Agilent 1100 HPLC equipped with adiode array detector was used and configured with dual SECcolumns connected inline to a Wyatt Dawn Heleos MALSdetector (using 658 nm laser, 17 detectors) and a WyattOptilab Rex refractive index detector (Wyatt Technology,Santa Barbara, CA).

Ion Exchange Chromatography (IEC–HPLC)

The aggregate and monomer-enriched fractions wereanalyzed for differences in charge. IEC was conductedusing an Agilent 1100 HPLC system equipped with a binarypump. Fifty microgram of sample was injected onto aDionex ProPac WCX-10 (weak cation exchange-10,4� 250mm) column. Samples were separated accordingto their apparent positive charge on the surfaces and elutedwith a gradient of increasing sodium chloride (from 10 to200mM) in the buffer system of 20mMHEPES, pH 7.9. Theflow rate was 0.5mL/min and the column temperature was

Table I. Summary of aggregate dissociation studies of antibody A using

SEC.

Samplea

% Aggregates with SEC Method

Control

DTTb

(10mM)

Guanidineb

(6M)

KClb

(0.5M)

KClb

(1M)

High aggregates 16.7 16.2 0.5 16.2 16.4

Middle aggregates 9.8 9.6 0.2 9.8 9.7

Low aggregates 1.0 0.8 0.7 1.0 0.8

aRepresentative Antibody A samples (Protein A purified) with a rangeof % aggregates at harvest.

bAdditive was added to the SEC mobile phase. For DTT, the sampleswere also incubated with 10mM DTT at 378C prior to SEC.

Gomez et al.: Aggregates in Cell Culture 127

Biotechnology and Bioengineering

Figure 1. Analytical characterization of Antibody A aggregate-enriched fraction (Fraction 1: 87% aggregates) compared to monomer-enriched fraction (Fraction 2: 99.8%

monomer). A: SEC chromatogram; (B) IEC chromatogram; (C) HIC chromatogram; (D) CE–SDS electropherogram (LC, light chain; HL, one heavy and light chains; HH, two heavy

chains; HHL, two heavy chains and one light chain; HHLL, antibody monomer; HMWS, high molecular weight species); (E) near UV CD spectra.

Figure 2. Aggregate levels at harvest for cell line X as a function of (A) IVCC, (B) titer, (C) qp, and (D) cell culture temperature. All data points in (A)–(C) correspond to the

average of biological duplicates� standard deviation.�, 318C;*, 338C;~, 358C;&, 378C. For (D), n¼ 4 for 318C, n¼ 33 for 338C, n¼ 8 for 358C, and n¼ 12 for 378C. �P< 0.001 for a

two-tailed t test for averages.


kept at 408C. Protein elution was monitored by 280 nmabsorbance.

Hydrophobic Interaction Chromatography (HIC–HPLC)

The aggregate and monomer-enriched fractions wereanalyzed for differences in hydrophobicity. HIC wasconducted using an Agilent 1100 HPLC system equippedwith a binary pump. Fifty microgram of sample was injectedonto a TOSOH Butyl-NPR (4.6� 35mm) column (TosohBioscience, Montgomeryville, PA). Samples were separated

at ambient temperature by retention on the hydrophobicstationary phase and eluted with a gradient of decreasing saltconcentration from 1.5M to 0M ammonium sulfate.Protein elution was monitored by 280 nm absorbance.

Capillary Electrophoresis (CE–SDS)

The purified aggregate and monomer fractions wereanalyzed with CE–SDS with laser-induced fluorescence(LIF) to quantify the molecular size distribution underdenaturing conditions. Prior to fluorescent labeling, thesamples were diluted and exchanged into 100mM sodiumphosphate (pH 6.7) buffer. The desalted samples weretreated with SDS and N-ethylmaleimide (NEM) at 708C for5min. To label the samples, amine-reactive fluorogenic dye,3-(2-furoyl)quinoline-2-carboxaldehyde was added in thepresence of potassium cyanide and incubated at 508C for10min. The dye-labeled samples were directly diluted with a1% SDS sample buffer and injected into a Beckman CoulterPA800 Plus (Beckman Coulter, Brea, CA).

Near UV Circular Dichroism (CD)

CD spectra were collected using a Jasco J-815 CDSpectrometer (Easton, MD) equipped with a tempera-ture-controlled holder. All samples contained 0.7mg/mLantibody in buffer. CD measurements were collected atintervals of 0.5 nm from 340 to 250 nm with a 4 s responsetime and a 1 nm bandwidth. All spectra were collected in a1 cm quartz cuvette and the temperature was controlled at258C using a thermocouple inserted in a reference cell. Threescans were collected and averaged and the buffer data weresubtracted.

Statistical Analysis

Statistical tests were performed using JMP 7.0 software (SASInstitute Inc, Cary, NC). Statistical correlations betweenvariables were obtained from standard least-square linearregressions. P-values correspond to the t-test for thecoefficient of the independent variable. For comparingmeans, a two-tailed t-test with respect to controls wasperformed. Statistical significance was defined by P-values<0.05.

Results and Discussion

Analytical Characterization

We extensively characterized the antibody aggregates with avariety of analytical techniques. SEC analysis of threerepresentative samples with high, medium, and lowaggregates showed aggregates were dissociable underdenaturing conditions (guanidine) but not reducing

Figure 3. Aggregate levels at harvest for cell line X as a function of (A) mRNA

LC/B2m, (B) mRNA HC/B2m, (C) mRNA LC/HC. qRT-PCR ratios are an average of day 3,

day 6, and day 9 values in each one of the experiments. All data points correspond to

the average of biological duplicates� standard deviation. �, 318C;*, 338C;~, 358C;&, 378C.



(DTT) or high salt conditions (Table I). These resultsindicated antibody aggregates were mediated by hydro-phobic interactions.

For further characterization, both the monomer andaggregates were purified into separate fractions using cationexchange chromatography (Fraction 1, aggregate-enriched:87% aggregates, 13% monomer; Fraction 2, monomer-enriched: 0.2% aggregate, 99.8% monomer). Analysis ofthese two fractions with MALS-SEC showed aggregates were63% dimer, 27% trimer, and 10% mixture of tetramerand pentamer (Fig. 1A). Aggregates appeared to be morepositively charged on the surface (Fig. 1B) and morehydrophobic than the monomer (Fig. 1C). CE–SDS analysisconfirmed aggregates were readily dissociable underdenaturing conditions showing 93% monomer peak andonly 3% non-dissociable aggregates in fraction 1 (Fig. 1D).Glycosylation of monomer and aggregates was comparable(results not shown). CD measurements showed differencesin the near UV spectrum between monomer and aggregates(Fig. 1E). Because near UV CD probes the microenviron-ment around the aromatic amino acids tryptophan,tyrosine, and phenylalanine, the difference in spectraindicated the tertiary conformation of aggregates wasaltered.

Retrospective Analysis of Effect of Cell CultureConditions on Antibody Aggregates

We conducted a retrospective statistical analysis of aggregatelevels as a screening exercise to identify the most relevant cellculture parameters. The univariate analysis of 7 productionexperiments (22 cultures in 2 L bioreactors, 14 cultures inshake flask satellites, 2 cultures in 400 L bioreactors, and 2cultures in 2,000 L bioreactors), showed no statistically

significant correlations of aggregates with titer (P¼ 0.34),IVCC (P¼ 0.15), qp (P¼ 0.39) or bioreactor vessel or scale(P¼ 0.60 for shake flask vs. 2 L bioreactors). Although wedid not directly compare the effect of pH, aggregates in 2 Lbioreactors correlated well with shake flask satellites withno pH control (R2¼ 0.91), indicating a minor impact of cellculture pH in the 6.4–6.9 range. Cell culture temperatureshowed a statistically significant correlation with aggregates(P¼ 0.004), showing that aggregate formation decreasedwith higher cell culture temperature.

Effect of Temperature on Aggregate Formation and Lightand Heavy Chain Expression

Based on the retrospective analysis results, the effect of cellculture temperature on aggregates was further investigated.We conducted shake flask studies at 31–378C and measuredLC and HC mRNA since their ratios were previouslyreported to increase antibody aggregates in CHO cells (Leeet al., 2009). Analysis of 57 production cultures confirmedour previous results showing no statistically significantcorrelation of aggregates with IVCC (P¼ 0.8), titer(P¼ 0.1), or qp (P¼ 0.08) (Fig. 2A–C). Temperatureshowed the strongest correlation with low aggregates incultures at 378C (Fig. 2D) (P< 0.001 for all temperaturecomparisons excepting 318C vs. 338C).

Increased mRNA encoding HC and LC (mRNA LC/B2m,mRNA HC/B2m) significantly increased aggregates(P¼ 0.001 for mRNA LC/B2m, P< 0.001 mRNA HC/B2m) (Fig. 3). Additionally, mRNA HC/B2m ratios <1.0and mRNA LC/HC ratios �1.5 produced <5% aggregates,indicating that excess transcript levels in combinationwith a less favorable LC/HC ratio increases aggregation.High recombinant protein expression increases protein

Figure 4. Time-course analysis during cell culture for two different experiments. A: % aggregates, (B) mRNA LC/B2m, and (C) mRNA HC/B2m; &, 338C for experiment I,&,

378C for experiment I, D, 338C for experiment II;~, 378C for experiment II. D: Time-course of % aggregates in experiment I for supernatant samples without cells sourced from cell

culture on day 3 (^^) day 6 (**) and day 9 (�). All data points correspond to the average of biological duplicates� standard deviation.


aggregation in CHO cells (Schroder and Friedl, 1997;Schroder et al., 2002). Additionally, a mRNA LC/HC ratio>1.5 was correlated to low aggregates in previous studies(Lee et al., 2009). Other reports also highlighted theimportance of excess LC over HC to increase antibodyproduction (Schlatter et al., 2005). Gonzalez et al. showedthe shortest antibody assembly time is for LC/HC ratios>1.67 (Gonzalez et al., 2002).

The kinetics of aggregate formation were investigated intwo sets of experiments using a time-course study ofaggregates as a function of production culture duration(days 3–13). Aggregates increased with culture duration at

338C but not at 378C (Fig. 4A). The increased aggregates at338C correlated with a two-fold average increase in mRNALC/B2m and a 3.3-fold average increase in mRNA HC/B2mon day 6 when compared to 378C cultures (Fig. 4B–C).Finally, the amount of aggregates after secretion in themedium did not change over time in the absence of cells(Fig. 4D). Therefore, aggregates formed intracellularly.

To further confirm the difference in aggregate formationat different temperatures, cultures at either 338C or 378Cuntil day 6 were then split into three cases for days 6–13 inproduction: (A) Case 1 continued cell culture at the initialtemperature; (B) case 2 was shifted to the opposite

Figure 5. Time-course in cell culture during split experiment of (A)–(B) % aggregates; (C)–(D) mRNA LC/B2m; (E)–(F) mRNA HC/B2m. &, 338C constant;^, 338C until day 6,

378C after day 6;D, 338C until day 6, 378C after day 6 and medium exchanged;&, 378C constant;^, 378C until day 6, 338C after day 6;~, 378C until day 6, 338C after day 6 and mediumexchanged. All data points correspond to the average of biological duplicates� standard deviation.



temperature; (C) case 3 was also shifted to the oppositetemperature and completely medium exchanged aftercentrifugation and cell re-suspension.

For the temperature shift from 338C to 378C on day 6,aggregates at harvest decreased for case 2 (12%, P¼ 0.007)and 3 (5%, P< 0.001) with respect to case 1 (19%) (Fig. 5A).In case 3, the change in aggregates correlated with a four-fold decrease in mRNA HC/B2m on day 9 (P¼ 0.01)(Fig. 5E). For the temperature shift from 378C to 338C onday 6, cases 2 (7%, P¼ 0.04) and 3 (12%, P¼ 0.003)significantly increased aggregates with respect to case 1 (5%)(Fig. 5B). This increase correlated with three-fold moremRNA HC/B2m on day 9 (P� 0.05 for both cases 2 and 3)(Fig. 5F). These results showed that culture temperaturemodulated aggregate formation.

Messenger RNA level is a function of transcription anddegradation rates, usually modeled as zero order function oftranscription and first order function of degradation (Foxet al., 2005):

dCðtÞdt

¼ kT�kDCðtÞ (1)

where t is the time, C(t) is mRNA concentration, kT istranscription rate, and kD is the degradation rate.

Wemeasured kD and half-life (t1/2) for LC andHC at 338Cand 378C by adding the transcription inhibitor actinomysinD (Fox et al., 2005; Oguchi et al., 2006; Ross, 1995).LC mRNA showed increased half-life at 338C (t1/2, LC-338C¼26 h, t1/2, LC-378C¼ 14 h) but not HC mRNA (t1/2, HC-338C¼20 h, t1/2, HC-378C¼ 18 h). Assuming steady state conditionsin Equation (1) and using the experimental kD, the ratios oftranscription rates for LC and HC were estimated. For two-fold increase in mRNA LC/B2m at 338C versus 378C (basedon day 6 average in Fig. 4B), kT-LC,338C/kT-LC,378C� 1. For3.3-fold increase in mRNA HC/B2m at 338C versus 378C(based on day 6 average in Fig. 4C), kT-HC,338C/kT-HC,378C� 3. Comparing the transcription rates for LC and HC,kT-LC,338C/kT-HC,338C� 0.7 and kT-LC,378C/kT-HC,378C� 1.8.These results estimated HC mRNA was transcribed threetimes more at 338C than at 378C and 1.4 times more than LCmRNA at 338C.

Previous investigations on recombinant protein produc-tion in CHO cells reported higher mRNA levels at sub-physiological temperatures. In some studies, the increasedmRNA levels were a result of increased mRNA stability atlow temperature (Fox et al., 2005; Oguchi et al., 2006; Yoonet al., 2003). In our results, kD was lower at 338C for LC(0.03 h�1 vs. 0.05 h�1 at 378C) and a two-fold increase inmRNA levels at 338C was caused by increased half-life.However, 3.3-fold increase in HC mRNA at 338C wascorrelated to increased kT. Culture of CHO cells at lowtemperatures triggers increased expression of a wide varietyof genes (Baik et al., 2006; Kumar et al., 2008). Additionally,inherent random integration of transgenes in the cell hostgenome could significantly affect recombinant proteinexpression depending on the integration site. O’Callaghanet al. (2010) proposed position effects and transcriptionalinterference as plausible causes for differences in LC and HCtranscription rates, despite the use of a single bicistronicvectors with independent promoters for each chain. Recentstudies on fragmentation of bicistronic vectors during cellline development (Ng et al., 2010) could also indicateindependent integration of transfected genes.

Analysis of western blots for HC in cell lysates showed asecond band above the HC monomer in the high aggregatesamples (Fig. 6A). This shifted band was less evident in the�7% aggregates samples. We speculate this band corre-sponds to either incompletely folded HC or HC withadditional fragments. Studies with a simplified HC reportedincompletely folded HC domains lose mobility in the SDS–PAGE gel and migrate at an apparent higher molecularweight (Lee et al., 1999). High aggregate samples alsoshowed more Fc fragments, indicating higher degradation ofHC. This increased degradation could also explain thedouble band as a HC with a small fragment. Based on theseresults, we speculate that increased HC transcription in highaggregate cultures leads to folding limitations in the ER. Theexcess HC is either misfolded and targeted for degradationin proteosomes via ER-Associated Protein degradation,

Figure 6. Western blots of day 9 cell lysates for three samples at 338C and two

samples at 378C. A–B: Anti-FC staining (unreduced and reduced with 0.1M DTT); (C)–

(D) anti-LC staining (unreduced and reduced with 0.1M DTT).


ERAD (Fagioli et al., 2001), or misfolded but still assembledwith LC to produce defective antibody monomer. Theantibodies with misfolded HC could expose hydrophobicresidues, mediating the formation of aggregates intracellu-larly that are subsequently secreted into the medium.Western blots for LC (Fig. 6B) did not show majordifferences among the samples studied.

ER Analysis

We studied ER proteins that participate in antibody foldingincluding BiP, PDI, and GRP94 (Gonzalez et al., 2002; Leeet al., 1999; Lenny and Green, 1991; Mayer et al., 2000) or inthe Unfolded Protein Response (UPR) such as XBP or ATF6(Adachi et al., 2008; Yamamoto et al., 2007). For this study,mRNA levels at 338C and 378C were compared in twoindependent experiments (Fig. 7A–B). We observed anoverall increased expression of these ER proteins at 338Cversus 378C (25–75% increase, P< 0.05 for PDI, GRP94,and ATF6 in experiment II). As additional supportive data

to the transcript analysis, we also estimated the ER size byfluorescently staining for PDI (Fig. 7C–E). ER size increasecorrelates with higher capacity for immunoglobulinproduction in plasma cells (Kirk et al., 2010; Shafferet al., 2004). The average ER size for cells at 338C expanded32% (P< 0.001), indicating a potential increase in foldingmachinery and consistent with the slight increase inchaperone mRNA content. However, this difference inboth the chaperone mRNA levels and ER volume was smallcompared to the 2–4-fold increase in LC and HC mRNAlevels at 338C, supporting the potential limiting foldingcapacity at lower temperature.

Based on these results, some additional cell engineeringstrategies could potentially reduce antibody aggregation.Increasing LC content by either super-transfecting LC orusing different promoter strengths could enhance HCfolding. However, even if increased LC facilitates CH1domain folding, chaperones are still needed. Studies ontransfection of PDI or BiP into CHO cells showedconflicting results on increased antibody production, butthe effects on antibody aggregation were not reported (Borth

Figure 7. ER analysis for two different experiments. mRNA content of ER proteins in (A) Experiment I; (B) Experiment II. Values correspond to an average of mRNA protein/

B2m on day 3, day 6, and day 9 for each sample.&, 338C;&, 378C. All data points correspond to the average of biological duplicates� standard deviation. Fluorescence microscopy

(anti-PDI staining in green, blue for nuclei) photomicrographs of day 9 samples at (C) 338C and (D) 378C; (E) quantification of ER volume on day 9 at 338C (n¼ 239) versus 378C(n¼ 273). �P< 0.05 for a two-tailed t test for averages. [Color figure can be seen in the online version of this article, available at http://wileyonlinelibrary.com/bit]



et al., 2005; Davis et al., 2000; Hayes et al., 2010). Otherstudies showed over-expression of XBP-1 increased ER sizeand recombinant protein production in CHO cells (Beckeret al., 2008; Tigges and Fussenegger, 2006), but the impacton protein aggregation was not described.

Effect of Temperature on Aggregates in Other Cell Lines

We investigated aggregate formation at 338C and 378C inadditional three cell lines that also exhibited >10% IgGaggregates (Fig. 8). All cell lines showed comparable orincreased titer and qp at 37 versus 338C (cell line Y, P< 0.01for both titer and qp) (Fig. 8A and B). Although limited dataon these cell lines was available, we observed a statisticallysignificant 2–5-fold decrease in aggregates at 378C in allcases (P< 0.005) (Fig. 8C). HC/B2m mRNA significantlydecreased for two of the cell lines at 378C (P¼ 0.04 for cell

lines W and Z) (Fig. 8D). Taken together, these resultsindicate an overall benefit in decreasing antibody aggrega-tion in 378C cultures. However, these results may varydepending on type of aggregates and cell line behavior.

Conclusion

We identified a strong inverse correlation of antibodyaggregate levels with cell culture temperature. Aggregates at378C were �5% in all experiments, whereas aggregatesincreased up to 20% at 338C. These results indicate aphysiological mechanism where reduced cell culturetemperature significantly increases HC and LC transcripts,which combined with a less favorable mRNA LC/HC ratioand limited increase in ER machinery leads to proteinmisfolding and subsequent aggregate formation.

Figure 8. Evaluation of cell lines W, Y, and Z for (A) titer, (B) qp, (C) % aggregates, (D) mRNA LC/B2m (solid bars) and HC/B2m (stippled bars).&, 338C;&, 378C. All data pointscorrespond to the average of biological duplicates� standard deviation. �P< 0.05 for a two-tailed t test for averages.


Nomenclature

ATF6 activating transcription factor 6

B2m beta-2 microglobulin

BiP immunoglobulin heavy chain binding protein

C mRNA concentration

CHO Chinese hamster ovary

ER endoplasmic reticulum

GRP94 glucose regulated protein 94

HC heavy chain

HPLC high-performance liquid chromatography

IVCC integrated viable cell concentration

kD mRNA degradation rate

kT mRNA transcription rate

LC light chain

mRNA messenger ribonucleic acid

PDI protein disulfide isomerase

qp cell-specific productivity

qRT-PCR quantitative reverse transcriptase-polymerase chain reaction

t time

VCC viable cell concentration

XBP-1 X-box binding protein 1

The authors thank Laura Simmons for insightful scientific discussions

and Adrian Nava for support with ER chaperone oligos; AthenaWong

for support with microcopy protocols; James Giulianotti for support

with western blots; John Joly and Steve Meier for data discussion and

support; the cell line development groups for the stable cell lines

evaluated; analytical operations for performing assays for titer and

aggregates; media preparation group for all the media and solutions

used during cell culture; cell banking group for supplying ampules for

the cell lines evaluated and Genentech oligo synthesis for primer

preparation.

References

Adachi Y, Yamamoto K, Okada T, Yoshida H, Harada A, Mori K. 2008.

ATF6 is a transcription factor specializing in the regulation of quality

control proteins in the endoplasmic reticulum. Cell Struct Funct

33(1):75–89.

Baik JY, LeeMS, An SR, Yoon SK, Joo EJ, Kim YH, Park HW, Lee GM. 2006.

Initial transcriptome and proteome analyses of low culture tempera-

ture-induced expression in CHO cells producing erythropoietin. Bio-

technol Bioeng 93(2):361–371.

Becker E, Florin L, Pfizenmaier K, KaufmannH. 2008. An XBP-1 dependent

bottle-neck in production of IgG subtype antibodies in chemically

defined serum-free Chinese hamster ovary (CHO) fed-batch processes.

J Biotechnol 135(2):217–223.

Borth N, Mattanovich D, Kunert R, Katinger H. 2005. Effect of increased

expression of protein disulfide isomerase and heavy chain binding

protein on antibody secretion in a recombinant CHO cell line.

Biotechnol Prog 21(1):106–111.

Braun A, Kwee L, Labow MA, Alsenz J. 1997. Protein aggregates seem to

play a key role among the parameters influencing the antigenicity of

interferon alpha (IFN-alpha) in normal and transgenic mice. Pharm

Res 14(10):1472–1478.

Cromwell ME, Hilario E, Jacobson F. 2006. Protein aggregation and

bioprocessing. AAPS J 8(3):E572–E579.

Davis R, Schooley K, Rasmussen B, Thomas J, Reddy P. 2000. Effect of PDI

overexpression on recombinant protein secretion in CHO cells.

Biotechnol Prog 16(5):736–743.

Fagioli C, Mezghrani A, Sitia R. 2001. Reduction of interchain disulfide

bonds precedes the dislocation of Ig-mu chains from the endoplasmic

reticulum to the cytosol for proteasomal degradation. J Biol Chem

276(44):40962–40967.

Fox SR, Tan HK, Tan MC, Wong SC, Yap MG, Wang DI. 2005. A detailed

understanding of the enhanced hypothermic productivity of interfer-

on-gamma by Chinese-hamster ovary cells. Biotechnol Appl Biochem

41(Pt 3):255–264.

Gamble CN. 1966. The role of soluble aggregates in the primary immune

response of mice to human gamma globulin. Int Arch Allergy Appl

Immunol 30(5):446–455.

Gomez N, Vinson A, Ouyang J, NguyenM, Chen X, Sharma V, Yuk I. 2010.

Triple light chain antibodies: Factors that influence its formation in cell

culture. Biotechnol Bioeng 105(4):748–760.

Gonzalez R, Andrews BA, Asenjo JA. 2002. Kinetic model of BiP- and PDI-

mediated protein folding and assembly. J Theor Biol 214:529–537.

Hayes NV, Smales CM, Klappa P. 2010. Protein disulfide isomerase does not

control recombinant IgG4 productivity in mammalian cell lines.

Biotechnol Bioeng 105(4):770–779.

Hendershot LM. 1990. Immunoglobulin heavy chain and binding protein

complexes are dissociated in vivo by light chain addition. J Cell Biol

111(3):829–837.

Kaloff CR, Haas IG. 1995. Coordination of immunoglobulin chain folding

and immunoglobulin chain assembly is essential for the formation of

functional IgG. Immunity 2(6):629–637.

Kirk SJ, Cliff JM, Thomas JA, Ward TH. 2010. Biogenesis of secretory

organelles during B cell differentiation. J Leukoc Biol 87(2):245–255.

Kumar N, Gammell P, Meleady P, Henry M, Clynes M. 2008. Differential

protein expression following low temperature culture of suspension

CHO-K1cells. BMC Biotechnol 8:42.

Lee CJ, Seth G, Tsukuda J, Hamilton RW. 2009. A clone screening method

using mRNA levels to determine specific productivity and product

quality for monoclonal antibodies. Biotechnol Bioeng 102(4):1107–

1118.

Lee YK, Brewer JW, Hellman R, Hendershot LM. 1999. BiP and immuno-

globulin light chain cooperate to control the folding of heavy chain and

ensure the fidelity of immunoglobulin assembly. Mol Biol Cell

10(7):2209–2219.

Lenny N, Green M. 1991. Regulation of endoplasmic reticulum stress

proteins in COS cells transfected with immunoglobulin mu heavy

chain cDNA. J Biol Chem 266(30):20532–20537.

MayerM, Kies U, Kammermeier R, Buchner J. 2000. BiP and PDI cooperate

in the oxidative folding of antibodies in vitro. J Biol Chem 275(38):

29421–29425.

Moore WV, Leppert P. 1980. Role of aggregated human growth hormone

(hGH) in development of antibodies to hGH. J Clin Endocrinol Metab

51(4):691–697.

Ng SK, LinW, Sachdeva R, Wang DI, YapMG. 2010. Vector fragmentation:

Characterizing vector integrity in transfected clones by Southern

blotting. Biotechnol Prog 26(1):11–20.

O’Callaghan PM, McLeod J, Pybus LP, Lovelady CS, Wilkinson SJ, Racher

AJ, Porter A, James DC. 2010. Cell line-specific control of recombinant

monoclonal antibody production by CHO cells. Biotechnol Bioeng

106(6):938–951.

Oguchi S, Saito H, Tsukahara M, Tsumura H. 2006. pH Condition in

temperature shift cultivation enhances cell longevity and specific hMab

productivity in CHO culture. Cytotechnology 52(3):199–207.

Pendse GJ, Karkare S, Bailey JE. 1992. Effect of cloned gene dosage on cell

growth and hepatitis B surface antigen synthesis and secretion in

recombinant CHO cells. Biotechnol Bioeng 40(1):119–129.

Philo JS, Arakawa T. 2009. Mechanisms of protein aggregation. Curr Pharm

Biotechnol 10(4):348–351.

Rosenberg AS. 2006. Effects of protein aggregates: An immunologic

perspective. AAPS J 8(3):E501–E507.

Ross J. 1995. mRNA stability in mammalian cells. Microbiol Rev 59(3):423–

450.

Schlatter S, Stansfield SH, Dinnis DM, Racher AJ, Birch JR, James DC. 2005.

On the optimal ratio of heavy to light chain genes for efficient



recombinant antibody production by CHO cells. Biotechnol Prog

21(1):122–133.

Schroder M, Friedl P. 1997. Overexpression of recombinant human anti-

thrombin III in Chinese hamster ovary cells results in malformation

and decreased secretion of recombinant protein. Biotechnol Bioeng

53(6):547–559.

Schroder M, Schafer R, Friedl P. 2002. Induction of protein aggregation in

an early secretory compartment by elevation of expression level.

Biotechnol Bioeng 78(2):131–140.

Shaffer AL, Shapiro-Shelef M, Iwakoshi NN, Lee AH, Qian SB, Zhao H, Yu

X, Yang L, Tan BK, Rosenwald A., et al. 2004. XBP1, downstream of

Blimp-1, expands the secretory apparatus and other organelles, and

increases protein synthesis in plasma cell differentiation. Immunity

21(1):81–93.

Tigges M, Fussenegger M. 2006. Xbp1-based engineering of secretory

capacity enhances the productivity of Chinese hamster ovary cells.

Metab Eng 8:264–272.

Vanhove M, Usherwood YK, Hendershot LM. 2001. Unassembled Ig heavy

chains do not cycle from BiP in vivo but require light chains to trigger

their release. Immunity 15(1):105–114.

Yamamoto K, Sato T, Matsui T, Sato M, Okada T, Yoshida H, Harada A,

Mori K. 2007. Transcriptional induction of mammalian ER quality

control proteins is mediated by single or combined action of

ATF6alpha and XBP1. Dev Cell 13(3):365–376.

Yoon SK, Song JY, Lee GM. 2003. Effect of low culture temperature on

specific productivity, transcription level, and heterogeneity of erythro-

poietin in Chinese hamster ovary cells. Biotechnol Bioeng 82(3):289–

298.


culture temperature modulates aggregation of recombinant antibody in cho cells

Documents