www.elsevier.com/locate/jbiosc

Journal of Bioscience and BioengineeringVOL. xx No. xx, 1e8, 2014

Efficient folding/assembly in Chinese hamster ovary cells is critical for high quality(low aggregate content) of secreted trastuzumab as well as for high production:

Stepwise multivariate regression analyses

Yoichi Ishii,1,3,* Junko Murakami,1 Kazue Sasaki,1 Masayoshi Tsukahara,2 and Kaori Wakamatsu3

Bio-process Research and Development Laboratories, Kyowa Hakko Kirin Co., Ltd., 100-1 Hagiwara-machi, Takasaki-shi, Gunma 370-0013, Japan,1 Kyowa Hakko Kirin Co., Ltd., 1-6-1Ohte-machi, Chiyoda-ku, Tokyo 100-8185, Japan,2 and Graduate School of Engineering, Gunma University, 1-5-1 Tenjin-cho, Kiryu-shi, Gunma 376-8515, Japan3

Received 3 October 2013; accepted 27 January 2014Available online xxx

* CorrespondKyowa Hakko KJapan. Tel.: þ8

E-mail add

1389-1723/$http://dx.doi

Please citeaggregate c

When developing cell culture processes for therapeutic antibodies, the low content of aggregated proteins is the mostcritical because administering aggregated antibody molecules might result in adverse effects such as immunogenicity.To characterize cells with high productivity and quality, we determined factors that are closely related to antibody titer,which is a productivity indicator, and the area percentage of high molecular weight species in cultivated media, which isequivalent to aggregate content and is used as a quality indicator. We examined the factors influencing antibody titerand aggregate content using various data from 28 cell lines throughout their culture periods from growth to deathphases. Our study using correlation analysis revealed that statistically significant correlations between factors and in-dicators changes with sampling points, hence we thought that various factors would influence each indicator simul-taneously. To understand the relationship between these factors and titer/aggregates contents, we performed stepwisemultiple linear regression analyses and deduced a multiple linear model for each indicator. The titer was found topositively associate with specific growth rate and specific production rate and negatively with intracellular heavy chaincontent. The aggregate content was found to positively associate with protein disulfide isomerase mRNA level andnegatively with light chain secreted into culture media, specific production rate, intracellular light chain content, andspecific growth rate. Our observations suggest that correct and efficient assembling and/or folding of an antibodymolecule in an endoplasmic reticulum are important for high titer and low aggregates contents.

� 2014, The Society for Biotechnology, Japan. All rights reserved.

[Key words: Therapeutic antibody; Antibody productivity; Antibody quality; Cell line; Antibody aggregate; Chinese hamster ovary cell; Endoplasmicreticulum stress; Protein disulfide isomerase; Heavy chain binding protein; Stepwise multiple linear regression]

Therapeutic monoclonal antibodies (mAbs) have been widelyused owing to their high antigenic specificity, long serum half-life,and low incidence of undesirable side effects. Moreover, the marketfor mAbs is growing annually worldwide (1).

Because the administration of high doses of therapeutic mAbsis required, production processes with high productivity (titer) ofmAbs are needed to reduce the expense to patients. Titers of re-combinant antibodies produced by cell lines have been improvedby selecting hosts, improving expression vectors, optimizinggene-coding sequences, selecting cell lines suitable for large-scaleproduction, and optimizing culture media (2e5). Although severalmammalian cell lines are used for manufacturing glycoproteins,therapeutic mAbs are commonly manufactured using Chinesehamster ovary (CHO) cell lines for which clinical safety has beenestablished and high productivity has been reported. Finaltiters of 1e5 g/L mAbs are currently achieved with CHO cell lines(5e7).

ing author at: Bio-process Research and Development Laboratories,irin Co., Ltd., 100-1 Hagiwara-machi, Takasaki-shi, Gunma 370-0013,1 27 353 7436; fax: þ81 27 352 4977.ress: yoichi.ishii@kyowa-kirin.co.jp (Y. Ishii).

e see front matter � 2014, The Society for Biotechnology, Japan..org/10.1016/j.jbiosc.2014.01.013

this article in press as: Ishii, Y., et al., Efficient folding/assemontent) of secreted trastuzumab as well as for..., J. Biosci. Bioe

In addition to their productivity, the quality of therapeutic mAbsis also critical because mAbmolecules produced from CHO cells areheterogeneous owing to aggregation (8,9) and various post-trans-lational modifications (10,11) such as oxidation (12), fragmentation(13,14), deamidation (15,16), epimerization (17,18), glycation(19,20), and glycosylation (21e23). These aggregations and modi-fications of antibodies might lead to reduction of biological activityand/or increase of undesirable effects under certain circumstances(8,24e27).

Among various qualities of mAbs that produce heterogeneity,the content of aggregated species is the most important becauseprotein aggregates might induce immunogenic responses andcause adverse events on administration (28,29). The relation be-tween aggregate contents in intravenous immune globulin andvarious side effects has been reported (30). U.S. Food and DrugAdministration (FDA) recommends that it is critical for manufac-turers of therapeutic protein products to minimize protein aggre-gation to the extent possible (31). Therefore, antibody aggregatelevels must to be controlled precisely (8,24e27).

Because the productivity and quality of mAbs depend on cellline properties, selecting cell lines suitable for large-scale produc-tion is a very important stage in process development for mAbproduction. In fact, this selection process is time consuming as well

All rights reserved.

bly in Chinese hamster ovary cells is critical for high quality (lowng., (2014), http://dx.doi.org/10.1016/j.jbiosc.2014.01.013

2 ISHII ET AL. J. BIOSCI. BIOENG.,

as labor intensive (32), and various factors must be evaluatedduring this screening. Factors affecting antibody production in thecultivation processes of cell lines are found to be cell growth rate,viable cell density (VCD), and specific production rate (Qp); inparticular, cell lines suitable for large-scale production have beenreported to exhibit high VCD and Qp (5).

The final goal in developing a cell culture process for therapeuticmAbs is high quality and high productivity. To achieve this goal, theeffects of various factors (e.g., Qp, VCD, levels of heavy/light chains,levels of proteins involved in protein folding/assembly) on thequality and productivity should be analyzed simultaneously andsystematically. Although the effects of these factors may changewith culture period, the correlations between an indicator andvarious factors were analyzed based on the data obtained at a singleculture point in previous reports (32,33). In fact, we found that eventhe sign of many correlations changes with culture period asdescribed below. In addition, simultaneous analysis on the corre-lation of various factors with aggregate contents has not beenmadeas far as we know, and it was not known which factor dominantlyaffects the productivity and quality. So we carried out stepwisemultivariate regression analyses on the titer and quality at theculture end point using many factors at the time points that in-fluences most strikingly on these indices.

In this study, we determined factors closely related to titer,which is a productivity indicator, and the area percentage of highmolecular weight species [HMWS(%)], which is equivalent toaggregate content and is used as a quality indicator, to characterizecells that have high productivity and low aggregates contents.Twenty-eight stable CHO cell lines that produce trastuzumab[trade names Herclon, Herceptin (34)] were generated and theirproperties were analyzed, such as titer, HMWS(%) in culture me-dia, Qp, specific growth rate (m), low molecular weight species(LMWS) in culture media, heavy chain (HC)/light chain (LC) mRNAlevels, protein disulfide isomerase (PDI)/heavy chain bindingprotein (BiP) mRNA levels, and intracellular HC/LC protein contentlevels. To understand the relationship between various factors andtiter/HMWS(%), we performed stepwise multiple linear regressionanalyses. The cell lines that exhibited a high titer were charac-terized by high m, high Qp, and low HC protein content. The celllines that exhibited low HMWS(%) were characterized by a low PDImRNA level, high LMWS(%), high Qp, high m, and high LC proteincontent.

The contributions of high Qp and m to high productivity havebeen demonstrated by many reports (2,5,32). However, althoughthe effect of low intracellular HC content on titer has been sug-gested, a clear associate between them was demonstrated in thisstudy for the first time. On the other hand, the researches that re-ported the factors affecting the aggregate content of mAbs arelimited: Lee et al. (32) suggested the effect of the LC/HCmRNA ratioon aggregation, Gomez, et al. (33) indicated the effect of tempera-ture on aggregation, and Bhoskar et al. (35) reported that free lightchain in culture media reflects antibody productivity and quality. Inthis study, various factors that related to the efficient assemblingand/or folding of an antibody heterotetramer in an endoplasmicreticulum (ER) were demonstrated to be crucial for the high mAbquality, by stepwise multivariate regression analyses. In addition,the contribution of high titer to low HMWS(%) was demonstrated.

MATERIALS AND METHODS

Cell culture Twenty-eight single clonal cell lines were created from the CHOcells that produce trastuzumab. Clonal isolation was carried out using limiting dilu-tion. These cellswere inoculatedat 0.3�106 cells/mLwith aworking volumeof 30mLin 125 mL Erlenmeyer flasks. Basal medium and feed-culture medium (serum-free)were prepared in-house respectively. The cultures were shaken at 100 rpm at 37�Cand under a 5% CO2 atmosphere. On day 3, daily feeding (the volume equivalent to 3%

Please cite this article in press as: Ishii, Y., et al., Efficient folding/assemaggregate content) of secreted trastuzumab as well as for..., J. Biosci. Bioe

of the media that remained in the flask at the time) of the feed mediumwas started.Sampling for analysis was performed on days 5, 7, 10, 12, and 14.

Antibody concentration determination The titer of the antibody in amultiple-culture media was determined using the Protein A column (4.6 � 50 mm,Applied Biosystems, Foster City, CA, USA) at ambient temperature. Mobile phase Aconsisted of 20 mM sodium phosphate and 300 mM sodium chloride (pH 7.0), andmobile phase B consisted of 20mM sodium phosphate and 300mM sodium chloride(pH 2.8). The Protein A column was first equilibrated with 100% mobile phase A for0.3 min, then eluted with 100% mobile phase B. Detection of antibodies was per-formed at 214 nmwith an ultraviolet (UV) detector, and their titers were determinedfrom the calibration curve that was calculated using standard samples.

Calculation of growth and production rates m was calculated using thefollowing formula:

m ¼ln�Xt2Xt1

�

t2 � t1(1)

where Xti is the viable cell density on day ti.Qp was calculated using the following formula:

Qp ¼ r ¼ ðPt2 � Pt1ÞZt2

t1

Xdt

(2)

where Pti is the concentration of the production on Day ti and Xti is the viable celldensity on Day ti. The trapezoidal rule was used for estimating an approximate valueof the integral, and the area under the growth curve from Time t2 to Time t1 wasdetermined using the following equation.

Zt2

t1

XdtzS ¼X ðt2 � t1ÞðXt2 � Xt1Þ

2(3)

Product concentrationwas plotted against the integral viable cell density (IVCD)from day 0 to 7, and Qp was determined as the slope calculated by a least-squaremethod. To consider the effects of cell death in the late stage of the culture, Qp ineach culture period was also calculated.

mRNA analysis of heavy/light chains and chaperones A culture fluidcontaining 1 � 106 cells was centrifuged at 1500 rpm for 1 min, the supernatantwas removed, the cell pellets were re-suspended in phosphate buffered saline(PBS) and centrifuged again, the supernatant was discarded, and the precipitatedcells were stored at �20�C. RNA extraction was performed with the EZY RNA Cellmini-kit (Qiagen, Venlo, Netherlands) according to the manufacturer’s protocol,and its amount was determined using NanoDrop (Thermo Fisher Scientific, SaltLake City, UT, USA). Primers matching the designed TaqMan probes weredeveloped using a Primer-Express software (Applied Biosystems). TrastuzumabLC primers: forward, 50-TCACTTGTCGGGCGAGTCA-30 and reverse, 50-TGCCTGGTTTCTGCTGATACC-30 . Trastuzumab HC primers: forward, 50-GGACAAGAAAGTTGAGCCCAAA-30 and reverse, 50-GGTCCCCCCAGGAGGAGTTCA-30 . BiPprimers: forward, 50-ACTACAGCCTGTTGCTGGACTTC-30 and reverse, 50-GCCACCATAGGGAACTTCATCT-30 . PDI primers: forward, 50-TGATGGCAACCTGAAGAGATACC-30 and reverse, 50-TTTCTGCTACCACAACCTTGACA-30 . TM probe sequences were asfollows. LC: 50-ACGTGAACACCGCCGTGGCC-30; HC: 50-TGACAAAACTCACATGCCCACCG-30; BiP: 50-AGACTGCAGACGGACCGACCGC-30; PDI: 50-CAAGTCTGAACCTATCCCAGAGACCAACGA-30 . The TM probes were labeled with 6-carboxyfluorescein at the 50 end and with 6-carboxytetramethyl rhodamine at the 30

end. For real-time-PCR (RT-PCR) analysis, TaqMan one-step RT-PCR Master MixReagent (Applied Biosystems) and TM Ribosomal RNA Control Reagent (AppliedBiosystems) were used according to the manufacturer’s protocol. Each mRNAexpression levels were normalized to the housekeeping gene 18S rRNA. RT-PCRsamples were processed using the HT7900 system (Applied Biosystems), and RT-PCR conditions were a reverse transcription at 48�C for 30 min and an activationat 95�C for 10 min followed by 40 cycles at 95�C for 0.25 min and at 60�C for1 min. The relative fold change of expression of the transcripts of target geneswas quantified as 2�(DCt), where DCt was Ct (target genes) e Ct (housekeepinggene, 18S rRNA).

Size-exclusion chromatography To analyze the content of aggregates andlow molecular weight protein components and antibody molecules in culture media,size-exclusion chromatography (SEC) analyses were performed. The culture mediawere filteredwith a 0.22 mm filter andwere stored at�80�C until analysis. Preliminaryexperiments have confirmed that the SEC patterns of the culturemedia do not changesignificantlyby this storageprocedure (datanot shown). Inculturemedia, SECwasusedto estimate thepercentagesofHMWS(substanceswithmolecularweightsgreater thanthat of the antibody), LMWS(substanceswithmolecularweights lower than thatof theantibody), and themain species exhibiting the samemolecularweight as the antibody.

SEC was performed with a TSK-gel SuperSW3000 column (4.0 mm i.d. � 30 cm;Tosoh, Tokyo, Japan)precededbyaguardcolumn.Themobilephaseconsistedof50mMsodiumphosphate, 500mMsodium chloride, and 5% v/v ethanol (pH7.0), and theflowrate was 0.175 mL/min. Detection was performed with the UV detector set at 215 nm.

bly in Chinese hamster ovary cells is critical for high quality (lowng., (2014), http://dx.doi.org/10.1016/j.jbiosc.2014.01.013

VOL. xx, 2014 EFFICIENT FOLDING IS CRITICAL FOR HIGH QUALITY ANTIBODY 3

Western blotting To determine the intracellular content of HC and LC,Western blotting analyses were performed. A culture fluid containing 1 � 106 cellswas centrifuged at 1500 rpm for 1min, the supernatant was removed, the cell pelletswere re-suspended in PBS and centrifuged again, the supernatant was discarded,and the precipitated cells were stored at �80�C. The protein in the pellet preparedfrom the cells was solubilized using a Qproteome Mammalian Protein Prep Kit(Qiagen), according to the manufacturer’s protocol. The solubilizing solution(1011 mL) consisted of 1000 mL of the Lysis buffer, 1 mL of the Lysis buffer containing1 U/mL Benzonase Nuclease, and 10 mL of the protease inhibitor. The resultant lysateswere used as samples for semiquantitative analysis. NuPAGE LDS Sample Buffer (�4,20 mL) (Invitrogen, Foster City, CA, USA), 8 mL Reducing Agent (�10) (Invitrogen), and39 mL of purified water were added to 13 mL of the sample, and they were heated at65�C for 10 min before loading to NuPAGE 4%e12% BiseTris Gel (1.0 mm � 26 well,Invitrogen), followed by separation with MOPS SDS Running Buffer (Invitrogen)containing NuPAGE Antioxidant (Invitrogen) at 200 V for 45 min. Separated proteinswere transferred to nitrocellulose membranes by electroblotting using iBlot TransferStacks, Regular (Invitrogen), according to the manufacturer’s protocol. LC and HCpolypeptides of trastuzumab were detected by anti-human kappa light chain rabbitantibody (product no. K1255, Sigma-Aldrich, St. Louis, MO, USA) and anti-human g-chain specific rabbit antibody (product no. I9764, Sigma-Aldrich), respectively, usingNovex ECL Chemiluminescent Substrate Reagent Kit (Invitrogen). Immunoreactivesignals were measured with LAS-1000 plus Luminoimage Analyzer (Fujifilm, Tokyo,Japan) and semiquantitative analysis was performed using MultiGauge, ver. 2.2(Fujifilm).

Statistical data analysis Spearman’s rank-correlation analysis and stepwisemultiple linear regression analyses were performed by using SPSS (IBM Corporation,Armonk, NY, USA) to evaluate associations. P < 0.05 was considered significant.

RESULTS AND DISCUSSION

Kinetic parameters of cell lines The time courses of the ki-netic parameters for 28 cell lines are illustrated in SupplementalFig. S1. All analyzed properties (VCD, m, titer, and Qp) exhibitedwide variations among cell lines, although these were derivedfrom the same cell line. The growth profiles of the cells weredivided into three phases: (i) growth, (ii) stationary, and (iii)death (Supplemental Fig. S1A). The titers in the death phase (onday 14), in which mAbs were harvested, exhibited wide variationsof 1.67e5.30 g/L (Supplemental Fig. S1C). The titer stoppedincreasing in the death phase, except for the cell lines with lowtiter. The variability of Qp values increased with the cultureperiod, but the values of all cell lines fell drastically on Day 14(Supplemental Fig. S1D). We analyzed various properties of all 28cell lines, which included those with poor growth or productionin contrast to other studies reported. The data obtained generateda large sample space from which significant trends could bereliably deduced.

Correlation of titer with HMWS(%) and LMWS(%) To char-acterize cells that exhibited high titer and low aggregates contents,HMWS(%) and LMWS(%) were analyzed using SEC. Typical SECchromatograms of supernatants from cell cultures on days 5, 7, and12 are shown in Supplemental Fig. S2. Western blotting analysisshowed that HMWS mainly contained aggregated antibodies, andLMWS mainly contained a monomer and homodimer of LC,

FIG. 1. (A) Correlation between antibody titer and HMWS(%) at day 5 (open circles), day 7circles), (B) correlation between antibody titer and LMWS(%) at day 5 (open circles), day 7circles), and (C) correlation of LMWS(%) on day 14 versus specific production rate.

Please cite this article in press as: Ishii, Y., et al., Efficient folding/assemaggregate content) of secreted trastuzumab as well as for..., J. Biosci. Bioe

whereas the main peaks corresponded to a heterotetramerconsisting of two HC and two LC polypeptides (data not shown).The peaks appearing later than LMWS (not shown) were due tomedium components and were ignored in the peak-areapercentage calculation. Consistent with these observations, thearea percentage of HMWS before and after purification usingprotein A of the culture media was correlated, whereas the LMWSpeak almost disappeared after the purification (data not shown).LC is important for antibody folding/assembling, and cell linesthat secrete a large amount of LC were reported to show highproductivity (3,36e38). The presence of aggregates maycontribute to unwanted immunogenic responses againsttherapeutic proteins (28e30). Therefore, to determine LMWS(%)(corresponding to LC) and HMWS(%) (corresponding toaggregates), SEC analyses were performed for the culture medianot pretreated with protein A. The correlation between the titersof the antibody and HMWS(%) or LMWS(%) in media changedwith the culture period length, as shown in Fig. 1A and B,respectively. Although LMWS(%) values increased with the cultureperiod length in many clones, HMWS(%) essentially stayed thesame. There were negative relationships between HMWS(%) andthe titers on Days 5 and 7 (Spearman’s rank correlationcoefficient (rs) ¼ �0.590, P ¼ 0.001 on Day 5; rs ¼ �0.498,P ¼ 0.007 on Day 7). However, there was no significantcorrelation between HMWS(%) and the titers on days 10, 12, and14. The rs values of various correlations are summarized inSupplemental Table S1.

There were positive relationships between LMWS(%) and the ti-ters on days 5 and 7 (rs ¼ 0.520, P ¼ 0.005 on day 5; rs ¼ 0.561,P¼ 0.002 on day 7), but no significant correlation onDays 10 and 12.Negative relationships emerged on day 14 (rs ¼ �0.455, P ¼ 0.015).

Correlation of Qp with LMWS(%) Previously, cell lines thatsecrete a large amount of LC were believed to show high produc-tivity (3,34e37). However, too much secretion of LC was found todecrease productivity (titer), as described below. Beforedescribing the results, we will briefly comment on the folding ofHC and LC into mature antibody molecules. Although matureantibody molecules are secreted as heterotetramer, consisting oftwo HC and two LC polypeptides, monomer and homodimerproteins of LC are also secreted. However, monomers or dimers ofHC polypeptides are not secreted. When HC polypeptides areproduced in an ER in the absence of sufficient LC polypeptides,HC polypeptides remain in the ER as a tight complex (in apartially unfolded state) with BiP and are not secreted from theER (3,36,38). Although LC polypeptide also interacts with BiP, LCpolypeptide that was not associated with HC is known to besecreted from the ER (3,35e37).

Fig. 1C shows the relationship between Qp and LMWS(%) on day14; a strong negative relationship was observed (rs ¼ �0.567,P < 0.002). Quite similar relationship was observed for other days

(open squares), day 10 (open triangles), day 12 (open rhombuses), and day 14 (filled(open squares), day 10 (open triangles), day 12 (open rhombuses), and day 14 (filled

bly in Chinese hamster ovary cells is critical for high quality (lowng., (2014), http://dx.doi.org/10.1016/j.jbiosc.2014.01.013

4 ISHII ET AL. J. BIOSCI. BIOENG.,

(Supplementary Table S1). Although sufficient LC polypeptides arerequired for high Qp, too much production might also be disad-vantageous for Qp.

mRNA of HC and LC The changes in mRNA levels of HC andLC in 28 cell lines during the time of their culture are shown inFig. 2A and B. Although HC mRNA levels did not vary significantly

FIG. 2. The mRNA levels of (A) HC and (B) LC on day 5 (open bars), day 7 (filled bars), day 10 (cell lines, (C) HMWS(%) and (D) LMWS(%) versus the LC/HC mRNA ratio on day 14, (E) correla10 (open triangles), day 12 (open rhombuses), and day 14 (closed circles), and (F) correlation(open squares), day 10 (open triangles), day 12 (open rhombuses), and day 14 (closed circl

Please cite this article in press as: Ishii, Y., et al., Efficient folding/assemaggregate content) of secreted trastuzumab as well as for..., J. Biosci. Bioe

among cell lines, LC mRNA levels exhibited large differencesamong cell lines.

For each cell line, we calculated, for each culture period, the ratiobetween LC and HC mRNA levels and their LC/HC mRNA ratio toevaluate their correlation with antibody impurities (HMWS andLMWS). We found a strong negative relationship between the LC/HCmRNA ratio and HMWS(%) on Day 14 (rs¼�0.622, P< 0.001), as

hatched bars), day 12 (lightly shaded bars), and day 14 (horizontally striped bars) for 28tions of titer and the LC/HC mRNA ratio at day 5 (open circles), day 7 (open squares), days of specific production rate, Qp, and the LC/HC mRNA ratio at day 5 (open circles), day 7es).

bly in Chinese hamster ovary cells is critical for high quality (lowng., (2014), http://dx.doi.org/10.1016/j.jbiosc.2014.01.013

VOL. xx, 2014 EFFICIENT FOLDING IS CRITICAL FOR HIGH QUALITY ANTIBODY 5

shown in Fig. 2C; quite similar relationship was observed for otherdays (Supplementary Table S1). Therefore, the balance of LC vs. HCmRNA levels seems to affect aggregate formation in mAb expres-sion and secretion. Because cell lines that do not secrete a largeamount of HMWS are advantageous in the following purificationsteps, the cell lines with such properties and high productivity aresuitable for large-scale production. On the other hand, LMWS(%) onday 14 exhibited a strong, positive correlation (rs¼ 0.718, P< 0.001)with the LC/HC mRNA ratio (Fig. 2D) in contrast to the negativecorrelation between HMWS(%) and the LC/HC mRNA ratio; quitesimilar relationship was observed for other days (SupplementaryTable S1).

Positive correlation was observed between LC/HC mRNA ratioand antibody titers on days 5 and 7 (rs ¼ 0.536, P ¼ 0.003; andrs ¼ 0.601, P < 0.001), but there was no significant correlation ondays 10, 12 and 14 (Fig. 2E). We observed negative correlationsbetween LC/HC mRNA ratio and Qp on days 12 and 14 (rs ¼ �0.508,P¼ 0.006; and rs¼�0.567, P¼ 0.002), while such analysis found nosignificant correlations on days 7 and 10 (Fig. 2F). These resultsshow that correlations changedwith sampling points. Although thecell lines with high LC/HC mRNA in the growth phase showed hightiter, such correlation was not observed after the stationary phase(Supplemental Table S1).

mRNA of PDI and BiP To evaluate the contribution of pro-teins involved in protein folding and assembly to titer and aggre-gate content, we analyzed cellular mRNA levels of various proteinsrelated to endoplasmic reticulum quality control such as activatingtranscription factor 4, C/EBP-homologous protein, X-box bindingprotein, and endoplasmic reticulum-degradation enhancing a-mannosidase-like protein in addition to PDI and BiP mRNA, inpreliminary experiments. BecausemRNAs that exhibited significantdifferences among cell lines were those of PDI and BiP (data notshown), we measured the mRNA level of the two proteins in detail.We found that HMWS(%) is positively correlated with the mRNAlevels of PDI and BiP (Fig. 3A and B), which are involved in theassembling/folding of antibodies in the ER. Strongly positivecorrelations were observed between PDI mRNA levels andHMWS(%) on days 5, 7, and 10 (rs ¼ 0.742, P < 0.001; rs ¼ 0.867,P < 0.001; and rs ¼ 0 0.463, P ¼ 0.013, respectively), as shown inFig. 3A. Significant correlations were observed also between BiPmRNA levels and HMWS(%) on days 5, 7, 10, and 12 (rs ¼ 0.823,P < 0.001; rs ¼ 0.692, P < 0.001; rs ¼ 0.399, P ¼ 0.036; andrs ¼ 0.397, P ¼ 0.037, respectively), as shown in Fig. 3B.Consistent with these observations, there were strongcorrelations between PDI mRNA levels and BiP mRNA levels onDays 5, 7, and 10 (rs ¼ 0.914, P < 0.001; rs ¼ 0.818, P < 0.001; andrs ¼ 0.805, P < 0.01, respectively), as shown in Fig. 3C. Theseobservations suggest that the environment of the ER but not ofthe culture media is related to aggregate formation. If antibody

FIG. 3. Correlations of (A) the PDI mRNA level and HMWS(%) at day 5 (open circles), day 7 (op(open circles), day 7 (open squares), day 10 (open triangles), day 12 (open rhombuses), andsquares), and day 10 (open triangles).

Please cite this article in press as: Ishii, Y., et al., Efficient folding/assemaggregate content) of secreted trastuzumab as well as for..., J. Biosci. Bioe

aggregates are always formed in culture media from themonomeric antibody molecules secreted from cells, culture mediawith a higher antibody concentration should always exhibithigher HMWS(%) (8). However, we found that cell lines producinglarge amount of antibodies tended to exhibit low HMWS(%) ondays 5 and 7 (Fig. 1A). Because antibody molecules areheterotetrameric proteins, it might not be easy to assemble andfold the four polypeptides correctly and efficiently.

It has been reported that ER stress, i.e., accumulation of unfoldedprotein in the ER, activates unfolded protein response (UPR) andinduces expression of UPR genes such as BiP and PDI via the acti-vation of transcription factors such as X-box binding protein-1 andactivating transcription factor 6 (39e41). One might expect theinduction of UPR in cells in which antibody is not correctly foldedefficiently, yet our report is the first one that clearly demonstratedthe correlation between the amounts of aggregated speciessecreted into medium and the mRNA levels of BiP/PDI in cells. Inaddition, although mRNA levels of BiP and PDI were high in celllines with high HMWS(%), the high mRNA levels are not consideredto be the cause of HMWS generation, but its effect. The UPR has alsobeen reported to optimize the efficiency of mAbs secretion in B cells(42). Therefore, the concentrations of HC and LC proteins within thecells were determined for each cell line, as described below.

Intracellular protein content ratio of LC to HC Weobserved negative correlations between intracellular LC/HC proteinratio and the percent of HC degradation products on days 5 and 7(rs ¼ �0.609, P < 0.001; and rs ¼ �0.548, P ¼ 0.003) (Fig. 4A), whilesuch analysis found no significant correlations on Days 10, 12 and14 (only the data of day 10 are shown in Fig. 4A for clarity). Theseresults indicate again that correlations changed with samplingpoints (Supplemental Table S1).

Correlation of intracellular HC protein content withHMWS(%) Intracellular HC protein content exhibited a positivecorrelation with HMWS(%) in media on days 5, 7, 10, and 12(rs ¼ 0.612, P < 0.001; rs ¼ 0.594, P < 0.001; rs ¼ 0.470, P ¼ 0.012;and rs ¼ 0.512, P < 0.001, respectively) (Fig. 4B). This observation isin agreement with our expectation that some of the partiallymisfolded antibody molecules, in those cells where LC supply wasinsufficient, are secreted into media. The partially misfoldedantibody molecules may cause aggregates in the culture mediumor the cells. In fact, it has been reported that the antibodyaggregates observed by lowering culture temperature are formedintracellularly (33).

Correlation of intracellular HC protein content withLMWS(%) Intracellular HC protein content was negativelycorrelated with LMWS(%) in media on days 5, 7, 10, 12, and 14(rs ¼ �0.650, P < 0.001; rs ¼ �0.742, P < 0.001; rs ¼ �0.730,P < 0.001; rs ¼ �0.749, P < 0.001; and rs ¼ �0.567, P < 0.001,

en squares), and day 10 (open triangles), (B) the BiP mRNA level and HMWS(%) at day 5(C) the BiP mRNA level and the PDI mRNA level at day 5 (open circles), day 7 (open

bly in Chinese hamster ovary cells is critical for high quality (lowng., (2014), http://dx.doi.org/10.1016/j.jbiosc.2014.01.013

FIG. 4. (A) Intracellular protein content of LC/HC ratio on day 5 (open bars), day 7 (filled bars), and day 10 (hatched bars) for 28 cell lines, and (B) correlations of intracellular HCcontent and HMWS(%) at day 5 (open circles), day 7 (open squares), day 10 (open triangles), day 12 (open rhombuses), and day 14 (closed circles), and (C) correlations of intra-cellular LC content and LMWS(%) at day 5 (open circles), day 7 (open squares), day 10 (open triangles), day 12 (open rhombuses), and day 14 (closed circles).

6 ISHII ET AL. J. BIOSCI. BIOENG.,

respectively) (Supplemental Fig. S3A) in agreement with theobservation that LMWS primarily consists of LC polypeptides.

Intracellular LC protein content exhibited a negative correlationwith LMWS(%) in media on Days 10, and 12 (rs ¼�0.494, P¼ 0.008;and rs ¼ �0.466, P ¼ 0.012, respectively), but there were no cor-relations on days 5, 7 and 14 (Fig. 4C). These results suggest thatexcess LC in the cells is secreted as LMWS, but the correlation be-tween intracellular LC protein content and LMWS(%) differ amongsampling points, and other factors relevant to the status of the cellmay be related to these differences.

Correlation of titer with LC/HC protein ratio andintracellular HC content The LC/HC protein ratio exhibitedpositive correlations with the titers of antibodies during the growthphase (rs ¼ 0.402, P ¼ 0.034 on Day 5; rs ¼ 0.564, P ¼ 0.002 on Day7) (Supplemental Fig. S3B). Intracellular HC content exhibitednegative correlations with the titers of antibodies on days 5 and 7(rs ¼ �0.498, P ¼ 0.007, and rs ¼ �0.748, P < 0.001, respectively),but there were no correlations on days 10, 12, and 14(Supplemental Fig. S3C). The rapidly dividing cells in the growthphase may be too susceptible to the stress from excess HC tosecrete antibodies efficiently.

Indicators and factors for stepwise multiple linearregression analysis We selected the titer as the productivityindicator and HMWS(%) as the quality indicator, and evaluated therelationships of these indicators with factors that might affect theseindicators: Qp, LMWS(%), and LC/HC mRNA levels. Because the titerand HMWS(%) during the death phase are critical indicators forproduction, we have to elucidate the correlation of these indicators

TABLE 1. Stepwise multipl

R2 Adjusted R2

Dependent variable: titer 0.603 0.549Independent variablesSpecific growth rate (m)Specific production rate (Qp)Intracellular HC content on day 14Constant

Titer of antibody (g/L) ¼ 13.126x1 þ 0.0840x2 � 0.004x3 � 5.337 (model A)

Dependent variable: HMWS(%) 0.863 0.829Independent variablesPDI mRNA level on day 14LMWS(%) on day 14Specific production rate (Qp)Intracellular LC content on day 14Specific growth rate (m)Constant

HMWS(%) ¼ 109.597x4 � 0.217x5 � 0.145x2 � 0.010x6 � 11.878x1 þ 26.577 (model B

R: correlation coefficient; SE: standard error; P: significance probability.

Please cite this article in press as: Ishii, Y., et al., Efficient folding/assemaggregate content) of secreted trastuzumab as well as for..., J. Biosci. Bioe

with factors during the death phase. For example, although thecorrelation between titer and LMWS(%) is apparent at the growthphase, the correlation is not obvious at later phases. These obser-vations strongly suggest that each indicator is affected by multiplefactors. Consequently, we analyzed the correlations of the two in-dicators with the factors above during the death phase using astepwise multiple linear regression. The values of m and Qp that weadopted were calculated using the data from day 0 to 7.

Stepwise multiple linear regression analysis of titer Celllines with high m, high Qp, and low intracellular HC protein contentwere found to exhibit high titer. Table 1 shows the multiple linearregression model of the antibody titer during the death phase,which was obtained using 13 independent variables [m, Qp, L/HmRNA ratio, LMWS(%), PDI mRNA level, BiP mRNA level, HCmRNA level, LC mRNA level, degradation product of HCpolypeptide, intracellular HC protein content, intracellular LCprotein content, LH/HC protein ratio, and HMWS(%)].

The following multiple linear model was obtained:

Titer of antibody¼ 13.126x1 þ 0.0840x2 � 0.004x3 � 5.337 (model A)

where x1, x2, and x3 are values of m, Qp, and intracellular HC proteincontent, respectively.

The sufficiently large R2 value (0.603) warrants the significanceof the correlation, and its small difference with the adjusted R2

value suggests that the number of samples in model A is satisfac-tory. The problem of multicollinearity in model A can be ignoredfrom the small values of the variance inflation factor (VIF, <4). The

e regression analysis.

Coefficient SE Standardizedcoefficient

P

13.126 2.408 1.139 <0.0010.084 0.018 1.170 <0.001

�0.004 0.001 �0.449 0.020�5.337 1.761 0.006

109.597 20.877 0.474 <0.001�0.217 0.031 �0.793 <0.001�0.145 0.037 �0.590 0.001�0.010 0.004 �0.223 0.019

�11.878 5.204 �0.300 0.03426.557 4.243 <0.001

)

bly in Chinese hamster ovary cells is critical for high quality (lowng., (2014), http://dx.doi.org/10.1016/j.jbiosc.2014.01.013

FIG. 5. Possible involvement of the amount of LC and UPR in the secretion of correctly folded and misfolded (and then aggregated) antibodies. Abbreviations used: BiP, heavy chainbinding protein; PDI Ox, protein disulfide isomerase in an oxidized state; ER, endoplasmic reticulum; UPR, unfolded protein response.

VOL. xx, 2014 EFFICIENT FOLDING IS CRITICAL FOR HIGH QUALITY ANTIBODY 7

predicted values calculated from the deduced model A were inagreement with the measured values (Supplemental Table S2).

The contribution of each factor to the titer value was shown as astandardized coefficient, and its magnitude was in the order ofQp > m > intracellular HC protein content. This observation is inagreement with previous reports: the factors that affect antibodyproduction in the cultivation processes of cell lines were reportedto be cell growth rate, VCD and Qp (5). However, the negative effectof the accumulation of intracellular HC protein along with thepositive effect of Qp and m on the titer was revealed for the first timeby using the present stepwise multivariate regression method. Thecell lines that exhibit high intracellular HC content due to theirdifficulties in the assembling/folding process in the ER are consid-ered to exhibit decreased titer.

Stepwise multiple linear regression analysis of HMWS(%)Cell lines with low HMWS(%) were characterized by a low PDImRNA level, high LMWS(%), high Qp, high m, and high intracellularLC protein content. Table 1 indicates the multiple linear regressionanalysis results of HWMS(%) obtained using 13 independentvariables. The following multiple linear model was obtained:

HMWS ¼ 109.597x4 � 0.217x5 � 0.145x2 � 0.010x6 � 11.878x1þ 26.577 (model B)

where x4, x5, x2, x6, and x1 are values of PDI mRNA level, LMWS(%),Qp, intracellular LC protein content, and m, respectively.

The sufficiently largeR2 value (0.863)warrants the significanceofthe correlation, and its small difference with the adjusted R2 valuesuggests that the number of samples in model B is satisfactory. Theproblem of multicollinearity in model B can be ignored from thesmall values of VIF (<4). The predicted values calculated from thededuced model B were in agreement with the measured values(Supplemental Table S3). This model indicates that many parame-ters contribute to the formation of HMWS, in contrast to the case ofthe titer, and the cell lines that show a low PDI mRNA level, highLMWS(%), high Qp, high m, and high intracellular LC protein contentexhibit low HMWS(%). Such novel findings will be quite useful forsuppressing aggregate formation in the mAbs manufacturing pro-cess (Fig. 5). Quite recently, Bhoskar et al. (35) reported that free LCsecreted into culture media is related to mAb productivity andquality (aggregate), in agreement with our results. Our study is,

Please cite this article in press as: Ishii, Y., et al., Efficient folding/assemaggregate content) of secreted trastuzumab as well as for..., J. Biosci. Bioe

however, more exhaustive in that antibody proteins retainedwithincells are also analyzed and in that the mRNA levels of proteinsinvolved in unfolded protein response expected for protein aggre-gates within cells are analyzed. In addition, while they reported theeffect of culture media on the productivity and the aggregate con-tent, our study revealed the factors that are involved in the differ-ence in the productivity and aggregates contents among cell linesunder optimal culture media conditions.

In this study, we created 28 stable cell lines that producedtrastuzumab using CHO cells and analyzed their growth charac-teristics, antibody productivity, H/LMWS(%) in culture media,mRNA levels of HC/LC/PDI/BiP, and HC/LC intracellular proteincontent. These analyses revealed that statistically significant cor-relations between factors change with sampling points(Supplemental Table S1). To characterize the cells with high pro-ductivity and high quality, we analyzed how two indicators, titer forproductivity and HMWS(%) for quality, were associated to the fac-tors listed above. We found that high titer was associated to high m,high Qp, and low intracellular HC protein content using stepwisemultivariate regression analyses. LowHMWS(%) was associated to alow PDI mRNA level, high LMWS(%), high Qp, high intracellular LCprotein content, and high m. These observations will provide usefulinformation for enhancing mAb titer and suppressing mAb aggre-gate formation. Although further investigation is required to clarifythe detailedmechanisms of mAb aggregate formation, these resultswill be used for the selection of high-productivity cell lines withlow aggregate contents, which are suitable for large-scaleproduction.

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.jbiosc.2014.01.013.

ACKNOWLEDGMENT

We would like to thank Risa Ogawa, Keina Yamaguchi, KousukeKuroda and Noriko Komatuzaki for their valuable help.

References

1. Elvin, J. G., Couston, R. G., and van der Walle, C. F.: Therapeutic antibodies:market considerations, disease targets and bioprocessing, Int. J. Pharm., 440,83e98 (2013).

2. Chusainow, J., Yang, Y. S., Yeo, Y. H. M., Toh, P. C., Asvadi, P., Wong, N. S. C.,and Yap, M. G. S.: A study of monoclonal antibody-producing CHO cell lines:

bly in Chinese hamster ovary cells is critical for high quality (lowng., (2014), http://dx.doi.org/10.1016/j.jbiosc.2014.01.013

8 ISHII ET AL. J. BIOSCI. BIOENG.,

what makes a stable high producer? Biotechnol. Bioeng., 102, 1182e1196(2009).

3. Davies, S. and James, D.: Engineering mammalian cells for recombinantmonoclonal antibody production, pp. 153e173, in: Al-Rubeai, M. (Ed.), Cell linedevelopment. Cell engineering, vol. 6. Springer, Netherlands (2009).

4. Kalwy, S., Rance, J., and Young, R.: Toward more efficient protein expression,Mol. Biotechnol., 34, 151e156 (2006).

5. Yu, M., Hu, Z. L., Pacis, E., Vijayasankaran, N., Shen, A., and Li, F.: Under-standing the intracellular effect of enhanced nutrient feeding toward hightiter antibody production process, Biotechnol. Bioeng., 108, 1078e1088(2011).

6. Arunakumari, A., Dai, X. P., Goldstein, J., Kloth, C., Ghebremariam, H.,Macisaac, G., andWagner,M.: The impact of cell culturemediumon cell line andprocess development timelines and strategies, Biopharm. Int. Suppl., 2 June2009(2009).

7. Jayapal, K. R., Wlaschin, K. F., Hu, W. S., and Yap, M. G. S.: Recombinantprotein therapeutics from CHO cells e 20 years and counting, Chem. Eng. Prog.,103, 40e47 (2007).

8. Cromwell, M. E. M., Hilario, E., and Jacobson, F.: Protein aggregation andbioprocessing, AAPS J., 8, E572eE579 (2006).

9. Pan, H., Chen, K., Pulisic, M., Apostol, I., and Huang, G.: Quantitation of sol-uble aggregates in recombinant monoclonal antibody cell culture by pH-gradient protein A chromatography, Anal. Biochem., 388, 273e278 (2009).

10. Harris, R. J.: Heterogeneity of recombinant antibodies: linking structure tofunction, Dev. Biol. (Basel), 122, 117e127 (2005).

11. Wang, W., Singh, S., Zeng, D. L., King, K., and Nema, S.: Antibody structure,instability, and formulation, J. Pharm. Sci., 96, 1e26 (2007).

12. Gaza-Bulseco, G., Faidu, S., Hurkmans, K., Chumsae, C., and Liu, H. C.: Effectof methionine oxidation of a recombinant monoclonal antibody on the bindingaffinity to protein A and protein G, J. Chromatogr. B, 870, 55e62 (2008).

13. Cordoba, A. J., Shyong, B. J., Breen, D., and Harris, R. J.: Non-enzymatic hingeregion fragmentation of antibodies in solution, J. Chromatogr. B, 818, 115e121(2005).

14. Cohen, S. L., Price, C., and Vlasak, J.: b-Elimination and peptide bond hydro-lysis: two distinct mechanisms of human IgG1 hinge fragmentation uponstorage, J. Am. Chem. Soc., 129, 6976e6977 (2007).

15. Chelius, D., Rehder, D. S., and Bondarenko, P. V.: Identification and charac-terization of deamidation sites in the conserved regions of human immuno-globulin gamma antibodies, Anal. Chem., 77, 6004e6011 (2005).

16. Li, X. J., Cournoyer, J. J., Lin, C., and O’Cormora, P. B.: Use of 18O labels tomonitor deamidation during protein and peptide sample processing, J. Am. Soc.Mass Spectrom., 19, 855e864 (2008).

17. Wakankar, A. A., Borchardt, R. T., Eigenbrot, C., Shia, S., Wang, Y. J.,Shire, S. J., and Liu, J. L.: Aspartate isomerization in the complementarity-determining regions of two closely related monoclonal antibodies, Biochem-istry, 46, 1534e1544 (2007).

18. Hambly, D. M., Banks, D. D., Scavezze, J. L., Siska, C. C., and Gadgil, H. S.:Detection and quantitation of IgG 1 hinge aspartate isomerization a rapiddegradation in stressed stability studies, Anal. Chem., 81, 7454e7459 (2009).

19. Quan, C., Alcala, E., Petkovska, I., Matthews, D., Canova-Davis, E., Taticek, R.,and Ma, S.: A study in glycation of a therapeutic recombinant humanizedmonoclonal antibody: where it is, how it got there, and how it affects charge-based behavior, Anal. Biochem., 373, 179e191 (2008).

20. Brady, L. J., Martinez, T., and Balland, A.: Characterization of nonenzymaticglycation on a monoclonal antibody, Anal. Chem., 79, 9403e9413 (2007).

21. Huhn, C., Selman, M. H. J., Ruhaak, L. R., Deelder, A. M., and Wuhrer, M.: IgGglycosylation analysis, Proteomics, 9, 882e913 (2009).

22. Read, E. K., Park, J. T., and Brorson, K. A.: Industry and regulatory experienceof the glycosylation of monoclonal antibodies, Biotechnol. Appl. Biochem., 58,213e219 (2011).

23. Li, H. J. and d’Anjou, M.: Pharmacological significance of glycosylation intherapeutic proteins, Curr. Opin. Biotechnol., 20, 678e684 (2009).

Please cite this article in press as: Ishii, Y., et al., Efficient folding/assemaggregate content) of secreted trastuzumab as well as for..., J. Biosci. Bioe

24. Carpenter, J., Cherney, B., Lubinecki, A., Ma, S., Marszal, E., Mire-Sluis, A.,Nikolai, T., Novak, J., Ragheb, J., and Simak, J.: Meeting report on proteinparticles and immunogenicity of therapeutic proteins: filling in the gaps in riskevaluation and mitigation, Biologicals, 38, 602e611 (2010).

25. den Engelsman, J., Garidel, P., Smulders, R., Koll, H., Smith, B., Bassarab, S.,Seidl, A., Hainzl, O., and Jiskoot, W.: Strategies for the assessment of proteinaggregates in pharmaceutical biotech product development, Pharm. Res., 28,920e933 (2011).

26. Cordoba-Rodriguez, R. V.: Aggregates in mAbs and recombinant therapeuticproteins: a regulatory perspective, Biopharm. Int., 21, 44 (2008).

27. Hermeling, S., Crommelin, D. J. A., Schellekens, H., and Jiskoot, W.: Struc-ture-immunogenicity relationships of therapeutic proteins, Pharm. Res., 21,897e903 (2004).

28. Berkowitz, S. A., Engen, J. R., Mazzeo, J. R., and Jones, G. B.: Analytical toolsfor characterizing biopharmaceuticals and the implications for biosimilars, Nat.Rev. Drug. Discov., 11, 527e540 (2012).

29. Tovey, M. G., Legrand, J., and Lallemand, C.: Overcoming immunogenicityassociated with the use of biopharmaceuticals, Expert Rev. Clin. Immunol., 4,623e631 (2011).

30. Knezevic-Maramica, I. and Kruskall, M. S.: Intravenous immune globulins: anupdate for clinicians, Transfusion, 43, 1460e1480 (2003).

31. U.S. Department of Health and Human Services, Food and Drug Adminis-tration, Center for Drug Evaluation and Research (CDER), and Center forBiologics Evaluation and Research (CBER): Guidance for industry; immuno-genicity assessment for therapeutic protein products. Draft guidance. U.S.Department of Health and Human Services, Washington, D.C. (2013).

32. Lee, C. J., Seth, G., Tsukuda, J., and Hamilton, R. W.: A clone screening methodusing mRNA levels to determine specific productivity and product quality formonoclonal antibodies, Biotechnol. Bioeng., 102, 1107e1118 (2009).

33. Gomez, N., Subramanian, J., Ouyang, J., Nguyen, M. D. H., Hutchinson, M.,Sharma, V. K., Lin, A. A., and Yuk, I. H.: Culture temperature modulates ag-gregation of recombinant antibody in cho cells, Biotechnol. Bioeng., 109,125e136 (2012).

34. Harris, R. J., Kabakoff, B., Macchi, F. D., Shen, F. J., Kwong, M., Andya, J. D.,Shire, S. J., Bjork, N., Totpal, K., and Chen, A. B.: Identification of multiplesources of charge heterogeneity in a recombinant antibody, J. Chromatogr. B,752, 233e245 (2001).

35. Bhoskar, P., Belongia, B., Smith, R., Yoon, S., Carter, T., and Xu, J.: Free lightchain content in culture media reflects recombinant monoclonal antibodyproductivity and quality, Biotechnol. Prog., 29, 1131e1139 (2013).

36. Dinnis, D. M. and James, D. C.: Engineering mammalian cell factories forimproved recombinant monoclonal antibody production: lessons from nature?Biotechnol. Bioeng., 91, 180e189 (2005).

37. Dul, J. L., Aviel, S., Melnick, J., and Argon, Y.: Ig light chains are secretedpredominantly as monomers, J. Immunol., 157, 2969e2975 (1996).

38. Strutzenberger, K., Borth, N., Kunert, R., Steinfellner, W., and Katinger, H.:Changes during subclone development and ageing of human antibody-pro-ducing recombinant CHO cells, J. Biotechnol., 69, 215e226 (1999).

39. Lee, J. H., Won, S. M., Suh, J., Son, S. J., Moon, G. J., Park, U. J., and Gwag, B. J.:Induction of the unfolded protein response and cell death pathway in Alz-heimer’s disease, but not in aged Tg2576 mice, Exp. Mol. Med., 42, 386e394(2010).

40. Yoshida, H., Okada, T., Haze, K., Yanagi, H., Yura, T., Negishi, M., and Mori, K.:ATF6 activated by proteolysis binds in the presence of NF-Y (CBF) directly tothe cis-acting element responsible for the mammalian unfolded proteinresponse, Mol. Cell. Biol., 20, 6755e6767 (2000).

41. Yoshida, H., Matsui, T., Yamamoto, A., Okada, T., and Mori, K.: XBP1 mRNA isinduced by ATF6 and spliced by IRE1 in response to ER stress to produce ahighly active transcription factor, Cell, 107, 881e891 (2001).

42. Gunn, K. E., Gifford, N. M., Mori, K., and Brewer, J. W.: A role for the unfoldedprotein response in optimizing antibody secretion, Mol. Immunol., 41,919e927 (2004).

bly in Chinese hamster ovary cells is critical for high quality (lowng., (2014), http://dx.doi.org/10.1016/j.jbiosc.2014.01.013