Case 1:18-cv-01363-CFC Document 79-4 Filed 03/22/19 Page 1 of 11 PageID #:
`9459
`
`ARTICLE
`
`Mechanism of Antibody Reduction in Cell Culture
`Production Processes
`
`Yung-Hsiang Kao,1 Daniel P. Hewitt,1 Melody Trexler-Schmidt,2 Michael W. Laird3
`1Protein Analytical Chemistry, Genentech, 1 DNA Way, South San Francisco,
`California 94080; telephone: 650-225-8943; fax: 650-225-3554; e-mail: yhk@gene.com
`2Late Stage Purification, Genentech, South San Francisco, California
`3Late Stage Cell Culture, Genentech, South San Francisco, California
`
`Received 5 February 2010; revision received 7 May 2010; accepted 14 June 2010
`
`Published online 29 June 2010 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/bit.22848
`
`ABSTRACT: We recently observed a significant disulfide
`reduction problem during the scale-up of a manufacturing
`process for a therapeutic antibody using a CHO expression
`system. Under certain conditions, extensive reduction of
`inter-chain disulfide bonds of an antibody produced by
`CHO cell culture may occur during the harvest operations
`and/or the protein A chromatography step, resulting in the
`observation of antibody fragments (light chain, heavy chain,
`and various combination of both) in the protein A pools.
`Although all conditions leading to disulfide reduction have
`not been completely identified, an excessive amount of
`mechanical cell lysis generated at the harvest step appears
`to be an important requirement for antibody reduction
`(Trexler-Schmidt et al., 2010). We have been able to deter-
`mine the mechanism by which the antibody is reduced
`despite the fact that not all requirements for antibody
`reduction were identified. Here we present data strongly
`suggesting that the antibody reduction was caused by a
`thioredoxin system or other reducing enzymes with thior-
`edoxin-like activity. The intracellular reducing enzymes and
`their substrates/cofactors apparently were released into the
`harvest cell culture fluid (HCCF) when cells were exposed to
`mechanical cell shear during harvest operations. Surpris-
`ingly, the reducing activity in the HCCF can last for a long
`period of time, causing the reduction of inter-chain disulfide
`bonds in an antibody. Our findings provide a basis for
`designing methods to prevent the antibody reduction during
`the manufacturing process.
`Biotechnol. Bioeng. 2010;107: 622–632.
`ß 2010 Wiley Periodicals, Inc.
`KEYWORDS: antibody; disulfide; thioredoxin; reduction;
`mechanism
`
`Introduction
`
`A large number of recombinant monoclonal antibodies
`(mAb) have been approved by regulatory agencies for
`
`Correspondence to: Y.-H. Kao
`
`treating human diseases during the last 15 years (Reichert,
`2001, 2002; Reichert and Pavolu, 2004; Reichert et al., 2005).
`Many more are currently in various stages of development
`and expected to become available in the near future. A very
`important structural feature of an antibody is the disulfide
`bonds that link its light and heavy chains (inter-chain
`disulfides) together to form a quaternary complex (Davies
`et al., 1975). In a mammalian cell culture system for large-
`scale therapeutic antibody production (e.g., CHO), all of the
`disulfide bonds (both inter- and intra-chain) of the antibody
`are correctly paired before the product is secreted into the
`cell culture fluid (CCF). Recently, we encountered a
`significant antibody disulfide reduction problem during
`the manufacturing process of a therapeutic antibody using a
`CHO expression system (Trexler-Schmidt et al., 2010).
`Under specific conditions, extensive reduction of
`the
`antibody’s inter-chain disulfides bonds was observed after
`the harvest operations (centrifugation and filtration) and/or
`the first purification step (i.e., protein A chromatography).
`On one such instance, it was estimated that as little as 10% of
`the antibody remained intact after the protein A step. This
`antibody reduction is not a unique issue to a specific
`antibody or a particular cell line as this phenomenon has
`been observed for multiple antibodies both at manufactur-
`ing and laboratory scales.
`The exact conditions resulting in antibody reduction
`remain elusive as the reduction phenomenon is not always
`reproducible at the manufacturing scale. However, our
`investigations have shown that an excessive amount of
`mechanical cell lysis generated at the harvest step was an
`important
`factor for the antibody reduction and the
`reducing activity was observed in the protein A load
`flow-through (Trexler-Schmidt et al., 2010). In this report,
`we present evidence indicating that the reduction was caused
`by an active thioredoxin (Trx) system or other reducing
`enzymes with thioredoxin-like activity in the harvested cell
`culture fluid (HCCF; post-centrifugation and filtration).
`The Trx
`system,
`consisting of
`thioredoxin (Trx),
`
`622 Biotechnology and Bioengineering, Vol. 107, No. 4, November 1, 2010
`
`ß 2010 Wiley Periodicals, Inc.
`
`

`

`Case 1:18-cv-01363-CFC Document 79-4 Filed 03/22/19 Page 2 of 11 PageID #:
`9460
`thioredoxin reductase (TrxR), and NADPH, is a ubiquitous
`Ward Hill, MA); reduced glutathione (GSH; Mallinckrodt
`antioxidative enzymatic system that plays an important role
`Baker, Phillipsburg, NJ); monobromobimane
`(mBB;
`in maintaining the cellular redox balance and keeping
`Sigma–Aldrich); histidine (Mallinckrodt Baker); sodium
`intracellular protein disulfides generally reduced (Gromer
`sulfate (Mallinckrodt Baker); thioredoxin (Trx; Sigma–
`et al., 2004). This system is also known to be involved in
`Aldrich); thioredoxin reductase (TrxR; Sigma–Aldrich). All
`various cellular processes including gene expression, signal
`chemicals and reagents were used as received with no further
`transduction, proliferation, and apoptosis (Arne´r and
`purification.
`Holmgren, 2000; Matsuo et al., 2002). Considering the
`importance of the Trx system and the cell lysis caused by
`the harvest operations, the presence of the Trx system in the
`HCCF is not surprising. However, it was surprising that the
`Trx system, fueled by NADPH, remained active for an
`extended period of time after the cell culture harvest. Our
`data indicated that the NADPH required for Trx and TrxR
`activities was provided by glucose-6-phosphate dehydro-
`genase (G6PD) activity, which generated NADPH from
`þ
`. Furthermore,
`glucose-6-phosphate (G6P), and NADP
`G6P is likely produced from glucose and ATP by hexokinase
`(phosphorylation of glucose; the first step of glycolysis).
`Together, these cellular enzymes (Trx system, G6PD, and
`hexokinase) along with their substrates are released into the
`HCCF upon cell lysis allowing the reduction event to occur.
`The observed antibody reduction is an outcome of a
`highly coupled reaction network in HCCF. Its kinetics is a
`very complex problem and dependent on many factors, such
`as viability of the production cell culture, the additional cell
`lysis from harvest operation, and the dissolved oxygen (DO)
`level in HCCF. While the antibody reduction kinetics is an
`important subject, the focus of this study is to determine
`the reduction mechanism. The kinetics of antibody
`reduction will be addressed in future studies.
`Moreover, we also describe strategies that can be applied
`to prevent undesired disulfide reduction in the manufactur-
`ing process for recombinant antibodies. Based on the
`identified reduction mechanism, any inhibitors, methods, or
`processes that can eliminate the activities of Trx system,
`G6PD, or hexokinase can be included as a mode to prevent
`disulfide reduction of recombinant proteins.
`
`Generation of Cell Culture Fluid (CCF)
`
`Mammalian CCF containing an antibody produced by
`Chinese hamster ovary (CHO) cells were generated using
`a representative small-scale fermentation process according
`to the methods described by Chaderjian et al. (2005). Cell
`culture process indicators (e.g., pH,
`temperature, DO,
`agitation rate) were monitored on-line while other culture
`indicators such as glucose, lactate, ammonium, glutamine,
`glutamate, and sodium were measured daily. Samples were
`also taken to monitor cell growth, viability, and rMAb
`concentration every 24 h.
`
`Harvested Cell Culture Fluid (HCCF) Preparation
`
`At the end of the production culture, complete mechanical
`lysis of CCF was achieved by high-pressure homogenization
`using a Microfluidics (Newton, MA) HC-8000 homogeni-
`zer. The pressure regulator of the instrument was set to
`4,000–8,000 psi, and complete cell lysis (membrane break-
`age) was achieved after a single pass, as determined by a
`lactate dehydrogenase (LDH) assay. The homogenate was
`centrifuged in a Sorval (Thermo Scientific, Asheville, NC)
`RC-3B rotor centrifuge at 4,500 rpm for 30 min at 208C. The
`centrate was decanted, depth filtered, and then sterile filtered
`(0.22 mm) to generate HCCF. For a non-homogenized
`control, a CCF sample was centrifuged at lab-scale and the
`centrate was sterile filtered (0.22 mm).
`
`Materials and Methods
`
`Materials
`
`Materials and devices used in the experiments described in
`this manuscript include: 50 and 55 cm3 stainless steel vials
`(mini-tanks; Flow Components, Dublin, CA); dialysis
`tubing (6,000–8,000 MWCO; Spectrum Laboratories,
`0.22 mm filter
`Rancho Dominguez, CA);
`(Millipak
`Gamma Gold; Millipore, Billerica, MA); phosphate-buffered
`saline (PBS; EDM Chemicals, Gibbstown, NJ); ethylene-
`diaminetetraacetic acid (EDTA; Sigma–Aldrich, St. Louis,
`MO); a-nicotinamide adenine dinucleotide phosphate
`(NADPH; EDM Chemicals); dehydroepiandrosterone
`(DHEA; TCI, Portland, OR); cupric sulfate (Sigma–
`Aldrich), G6P (EDM Chemicals); aurothioglucose (ATG;
`USP, Rockville, MD); aurothiomalate (ATM; Alfa Aesar,
`
`Dialysis Experiment
`
`A dialysis experiment was carried out in order to determine
`whether the components causing reduction of the antibody
`were small molecules or macromolecules (i.e., enzymes).
`A sample of 3 mL of purified and formulated antibody
`(30.2 mg/mL) was dialyzed against 1 L of PBS (10 mM,
`pH 7.2) at 258C for 24 h and the PBS was changed after 8 h.
`Dialysis tubing was hydrated overnight in a 0.05% azide
`solution and rinsed with sterile water prior to use. After
`dialysis, the concentration of the antibody sample was
`determined by the absorbance at 280 nm and adjusted to
`1 mg/mL. Aliquots were stored at 708C prior to use. The
`HCCF, obtained from homogenization of CCF from a 3-L
`fermentor, was thawed and filtered through a 0.22 mm
`Millipak filter using a peristaltic pump. Six 50 cm3 mini-
`tanks were filled with 30 mL of HCCF each. In each mini-
`tank, a purified antibody sample (500 mL) placed in a
`
`Kao et al.: Mechanism of Antibody Reduction
`
`623
`
`Biotechnology and Bioengineering
`
`

`

`Case 1:18-cv-01363-CFC Document 79-4 Filed 03/22/19 Page 3 of 11 PageID #:
`9461
`TrxR solution (4 mM). The reaction was incubated at room
`dialysis bag was submerged in the HCCF. The mini-tanks
`were sealed and loaded into a bench top mixer (Lab-Line
`temperature for 24 h in the absence of light. Aliquots of
`20 mL were taken at each sampling time point and stored at
`MAX Q 4000; Thermo Scientific) operating at 35 rpm and
`708C until analyzed by the microchip-based capillary
`ambient temperature. For each time point, one mini-tank
`was removed from the mixer, and aliquots of the HCCF (in
`electrophoresis assay. Control reactions were performed to
`the mini-tank) and the antibody sample (in the dialysis bag)
`determine if the enzymatic pathway was active when an
`were taken and stored at 708C until analyzed with the free
`enzyme was omitted by substituting an equal volume of
`thiol assay and the microchip based capillary electrophoresis
`PBS for either Trx and/or TrxR in the reaction mixture.
`assay (described below).
`Inhibition of the Trx/TrxR activity was demonstrated using
`the same reaction conditions described above with the
`addition of 5 mL ATG (10 mM) or ATM (10 mM). To
`demonstrate the inhibition of Trx system by Cu2þ
`, 2.5 mL of
`CuSO4 (10 mM) was added to reaction mixture using the
`same enzymes but a different buffer (10 mM histidine,
`10 mM Na2SO4, 137 mM NaCl, 2.5 mM KCl, pH 7.0) to
`prevent formation of insoluble Cu3(PO4)2.
`
`Liquid Chromatography–Mass Spectrometry (LC–MS)
`Analysis of Reduced Antibody
`
`Measuring Free Thiol Contents in HCCF Samples
`
`A series of reduced GSH standard samples were prepared in
`PBS (10 mM, pH 6.0) in order to generate a standard curve
`for quantifying free thiols. From a 110 mM GSH solution,
`standards were prepared at concentrations of 0, 5.5, 11, 22,
`44, 55, 110, and 550 mM through serial dilution. From an
`acetonitrile stock solution of mBB (10 mM stored at
`208C), a 100 mM solution of mBB was prepared in PBS
`(10 mM, pH 10.0) and stored protected from light.
`In a flat-bottomed 96-well plate, 100 mL of mBB was
`dispensed into each well. A 10 mL aliquot of GSH standard
`solution or test sample was then added to the wells. All wells
`were prepared in triplicate. The plate was incubated at room
`temperature for 1 h and then analyzed using a fluorescence
`plate reader (SpectraMax Gemini XS; Molecular Devices,
`Sunnyvale, CA) with an excitation wavelength of 390 nm
`and an emission wavelength of 490 nm. A linear standard
`curve was generated using the average result of the three
`standard wells plotted versus GSH concentration. Free thiol
`levels in samples were calculated from the linear equation of
`the standard curve using the average value of the three
`sample wells.
`
`Microchip-Based Capillary Electrophoresis Assay
`
`Reduction of antibody was monitored by capillary electro-
`phoresis using the Agilent 2100 Bioanalyzer (Agilent
`Technologies). Sample preparation was carried out as
`described in the Agilent manual (Protein 230 Assay
`Protocol) with minor changes. For HCCF samples a 1:4
`dilution was performed on the samples prior to preparation.
`At the denaturing step, in addition to mixing the test sample
`with 2 mL denaturing solution as described in the protocol, a
`24-mL solution containing 50 mM iodoacetamide (IAM)
`and 0.5% SDS solution was added. Digital gel-like images
`were generated using Agilent 2100 Expert software.
`
`The protein A elution pool from a manufacturing run
`that showed reduced antibody product was analyzed by
`reversed-phase high-performance liquid chromatography
`(RP-HPLC) using a Poroshell column (300SB-C8 1.0 mm 
`75 mm, 5 mm, Agilent Technologies, Santa Clara, CA). The
`sample was eluted with a formic acid/trifluoroacetic acid/
`acetonitrile gradient for direct on-line electrospray ioniza-
`tion-mass spectrometry (ESI-MS) using a QStar Pulsar i
`mass spectrometer (Applied Biosystems/MDS Sciex, Foster
`City, CA). Spectra were derived from multiple charged ions
`and deconvoluted using the Analyst QS 1.0/BioAnalyst 2.0
`software package (Applied Biosystems/MDS Sciex). The
`sample was also analyzed after
`treated with tris(2-
`carboxyethyl)phosphine (TCEP) to fully reduce all disulfide
`bonds in the antibody.
`
`HCCF Incubation Experiments to Probe the Role of
`NADPH and Glucose-6-Phosphate and to Test Inhibitors
`for Reduction
`
`A 55 cm3 mini-tank was filled with 27 mL of HCCF.
`Depending on the experiment design, various reagents such
`as NADPH, G6P, inhibitors of G6PD, or inhibitors of TrxR
`were added to the desired concentration, and the final
`volume in the mini-tank was brought to 30 mL with PBS
`(10 mM, pH 7.2). The mini-tanks were sealed and loaded
`into a bench top mixer running at 35 rpm and ambient
`temperature. At each time point for sampling, the exteriors
`of the mini-tanks were sterilized with 70% isopropyl alcohol
`and opened in a laminar flow hood for the removal of an
`aliquot. The mini-tanks were then re-sealed and loaded back
`into the bench top mixer. All aliquots were stored at 708C
`until analyzed with the free thiol assay and microchip-based
`capillary electrophoresis assay.
`
`In Vitro Trx/TrxR Studies
`
`In a polypropylene 1.5 mL microcentrifuge tube, 437 mL
`PBS, 25 mL NADPH (20 mM), 16 mL formulated the
`antibody solution (30.2 mg/mL), and 5 mL human Trx
`solution (500 mM in 10 mM PBS, pH 7.2) were gently mixed.
`The reaction was initiated by the addition of 17.5 mL rat liver
`
`624
`
`Biotechnology and Bioengineering, Vol. 107, No. 4, November 1, 2010
`
`

`

`Results and Discussion
`
`Dialysis Experiment
`
`Case 1:18-cv-01363-CFC Document 79-4 Filed 03/22/19 Page 4 of 11 PageID #:
`9462
`dialysis bag did not reach the exact same level as observed
`outside the bag, most likely because some free thiols were
`present on macromolecules and unable to enter the dialysis
`bag. Nevertheless, the levels of free thiols inside and outside
`the dialysis bag became comparable in <5 h after the
`incubation was initiated, suggesting that the free thiol
`containing small molecules in the HCCF entered the dialysis
`bag without restraints and small molecule components from
`the HCCF were able to reach equilibrium inside and outside
`the dialysis bag. Thus, the antibody reduction occurred
`outside the dialysis bag was not caused by the free thiol
`containing small molecules. Since the reduction was
`observed only outside the dialysis bag with a MWCO of
`7,000 Da, the molecular weight of the reducing molecule(s)
`must be >7,000 Da. The result from this dialysis experiment
`suggested that an enzymatic reaction was responsible for the
`observed antibody reduction.
`
`The reducing activity in HCCF was found to remain in the
`protein A load flow-through as evidenced by the reduction
`of various purified intact antibodies spiked into the protein
`A load flow-through, indicating that reducing components
`could be readily removed from the antibody manufacturing
`process (Trexler-Schmidt et al., 2010). However, it was not
`immediately clear whether the reducing components were
`small molecules or macromolecules (e.g., enzymes). To
`elucidate this question, a dialysis experiment was designed
`to determine the nature of reducing molecules involved in
`antibody reduction. In this dialysis experiment, an intact
`antibody was placed in a dialysis bag with a MWCO of
`7,000 Da and incubated in HCCF contained in a stainless
`steel mini-tank. As shown in Figure 1, the antibody inside
`the bag was not reduced after the incubation period
`(Fig. 1a), whereas the antibody outside the bag in the HCCF
`was significantly reduced soon after the incubation started as
`indicated by the loss of intact antibody (150 kDa) and the
`formation of the antibody fragments (various combinations
`of heavy and light chains) (Fig. 1b).
`The free thiol measurement (by using mBB to derivatize
`and quantify the sulfhydryl groups) showed that no free
`thiols were present inside the dialysis bag at the beginning of
`the incubation (Fig. 2). The levels of free thiols inside the
`
`Liquid Chromatography–Mass Spectrometry (LC–MS)
`Analysis of Reduced Antibody
`
`An LC–MS analysis was performed on the protein A elution
`pool
`from a manufacturing run that showed reduced
`antibody product. Under non-reducing assay conditions,
`the major species detected were antibody light chains and
`heavy chains with very little intact antibody observed,
`indicating the presence of reduced antibody in the protein A
`
`Figure 1. Dialysis experiment: digital gel-like images obtained from the microchip-based capillary electrophoresis analysis (each lane representing a time point: 0, 1, 3, 21, 25,
`and 29 h from lanes 2–7). a: The antibody inside the dialysis bag remained intact during the incubation period. b: The antibody outside the dialysis bag was reduced during the
`incubation period as evidenced by the loss of intact antibody (150 kDa) and the formation of antibody fragments. The band appearing just above the 28 kDa marker arises from the
`light chain of antibody. There was a significant amount of free light chain already present in the HCCF before the incubation began. The presence of excess free light chain and
`dimers of light chain in the HCCF is typical for the cell line producing antibody. [Color figure can be seen in the online version of this article, available at wileyonlinelibrary.com.]
`
`Kao et al.: Mechanism of Antibody Reduction
`
`625
`
`Biotechnology and Bioengineering
`
`

`

`Case 1:18-cv-01363-CFC Document 79-4 Filed 03/22/19 Page 5 of 11 PageID #:
`9463
`the intra-chain disulfides are generally located in the interior
`of an antibody and inaccessible to the reducing enzymes.
`
`Reduction of Antibody by Trx/TrxR In Vitro
`
`The Trx system (i.e., Trx and TrxR) is one of the two
`intracellular enzymatic systems (the other is the glutar-
`edoxin [Grx] system) that regulates the cellular redox status
`and maintains a reducing environment
`in the cytosol
`(Gromer et al., 2004). The Trx system can reduce disulfide
`bonds within a protein at the expense of NADPH (Fig. 3).
`Since the Grx system activity is only limited to the reduction
`of S-glutathionylated substrates (i.e., GSH-mixed disulfides)
`(Johansson et al., 2004), the Trx system is most likely the
`primary enzymatic system that reduces the inter-chain
`disulfides in the antibody. An in vitro experiment was
`conducted to test if the Trx system can reduce the antibody
`by incubating the intact antibody with Trx, TrxR, and
`NADPH. As expected, the microchip-based capillary electro-
`phoresis results indicate that the antibody can be reduced in
`vitro by the Trx system (Fig. 4). Similar to the reduction
`observed in the HCCF, the LC–MS analysis also showed that
`only inter-chain disulfide bonds were reduced in vitro by the
`Trx system (data not shown).
`There are many known Trx and TrxR inhibitors (Gromer
`et al., 2004). For example, gold complexes are among the
`
`Figure 3.
`Thioredoxin system and other reactions involved in antibody reduc-
`tion: the thioredoxin system, composed of thioredoxin (Trx), thioredoxin reductase
`(TrxR), and NADPH, is a hydrogen donor system for reduction of disulfide bonds in
`proteins. Trx is a small monomeric protein with a CXXC active site motif that catalyzes
`many redox reactions through thiol-disulfide exchange. The oxidized Trx can be
`reduced by NADPH via TrxR. The reduced Trx is then able to catalyze the reduction of
`disulfides in proteins. The NADPH required for thioredoxin system is provided via
`reactions in pentose phosphate pathway and glycolysis. [Color figure can be seen in
`the online version of this article, available at wileyonlinelibrary.com.]
`
`Figure 2.
`Free thiol levels from dialysis experiment: free thiols inside (dashed
`line) and outside (solid line) the dialysis bag reach comparable levels within a few
`hours,
`indicating a good exchange of small molecule components in the HCCF
`between inside and outside the dialysis bag. The error bars are based on the standard
`deviation of three independent free thiol measurements of the each sample.
`
`pool. The observed mass for the light chains in the protein A
`pool sample was 23,234 Da. Since the heavy chain in an IgG
`molecule contains an N-linked glycsoylation site, several
`masses arising from the heterogeneity of glycan structures
`were observed for the heavy chains in the protein A pool
`sample. The predominant heavy chain species had a mass
`of 50,996 Da. For light chain, the observed mass was lower
`than the expected mass of the fully reduced light chain
`(23,238 Da) by 4 Da. For heavy chain, the main observed
`mass was lower than the expected mass of the fully reduced
`heavy chain containing an asialo, agalacto biantennary
`oligosaccharide structure with a core fucose (G0 glycoform;
`51,004 Da) by 8 Da. In an IgG molecule, there are two and
`four intra-chain disulfide bonds in the light chain and heavy
`chain, respectively. The mass differences observed in the
`LC–MS analysis suggested that the light chain and heavy
`chain in the protein A pool sample must contain two and
`four disulfide bonds, respectively, because the formation of
`each disulfide would reduce the mass number by 2 Da (loss
`of two protons). These mass spectrometry data clearly
`indicated that only the inter-chain disulfide bonds were
`reduced, resulting in the presence of light chain and heavy
`chain in the protein A pool. The intra-chain disulfide bonds
`in the light chain and heavy chain, however, were still intact.
`Subsequent peptide map analysis with LC–MS detection
`also confirmed that reduction occurred only at the inter-
`chain disulfide bonds. The fact that only the inter-chain
`disulfide bonds were reduced is consistent with the results
`from dialysis experiment showing that antibody reduction
`in the HCCF was caused by reducing macromolecules since
`
`626
`
`Biotechnology and Bioengineering, Vol. 107, No. 4, November 1, 2010
`
`

`

`Case 1:18-cv-01363-CFC Document 79-4 Filed 03/22/19 Page 6 of 11 PageID #:
`9464
`
`Figure 4.
`In vitro activity of thioredoxin system: digital gel-like image from the
`microchip-based capillary electrophoresis analysis (each lane representing a time
`point: 0, 0.5, 1, 2, 3, 21, and 23 h from lanes 2–8). The incubation of intact antibody (1 mg/
`mL) with 0.1 mM TrxR (rat liver), 5 mM Trx (human), and 1 mM NADPH in PBS resulted in
`antibody reduction (completely reduced in <21 h). [Color figure can be seen in the
`online version of this article, available at wileyonlinelibrary.com.]
`
`most effective and selective inhibitors of mammalian TrxRs
`known to date. Two commercially available specific
`inhibitors of TrxR, ATG, and ATM, were tested for their
`ability to inhibit the Trx system in vitro and the antibody
`reduction. As expected, both ATG and ATM can effectively
`inhibit the antibody reduction in the assay described above
`(Fig. 5).
`Cupric sulfate is known for its ability to provide oxidizing
`redox potential and has been used in the cell culture
`processes to minimize free thiol (i.e., minimize unpaired
`cysteine)
`levels
`in recombinant
`antibody molecules
`(Chaderjian et al., 2005). We have also tested whether
`cupric sulfate can inhibit the Trx system in vitro and the
`subsequent reduction of the antibody. In this in vitro
`reduction experiment, the buffer system was changed from
`PBS to histidine sulfate in order to avoid the formation
`of insoluble Cu3(PO4)2. Figure 6 showed that the antibody
`was readily reduced by the Trx system in the histidine sulfate
`buffer (even faster than in PBS buffer), but the addition
`of CuSO4 to this reaction clearly inhibited the antibody
`reduction.
`
`Inhibition of Antibody in HCCF by ATG and ATM
`
`The two gold compounds shown to be capable of inhibiting
`reduction of the antibody by the Trx system in vitro are
`specific inhibitors for TrxR (Fig. 5). If the Trx system was
`active in the HCCF and caused the antibody reduction in the
`failed antibody manufacturing runs and in the lab scale
`reduction experiments, both gold compounds (ATG and
`ATM) should be able to inhibit the reduction of antibody in
`
`Figure 5.
`In vitro activity of thioredoxin system inhibited by ATG: the addition of
`ATG at a concentration of 1 mM to the reaction mixture as described in the caption for
`Figure 4 effectively inhibited the antibody reduction as shown in the digital gel-like
`image from the microchip-based capillary electrophoresis analysis (each lane repre-
`senting a time point: 0, 0.5, 1, 2, 3, 21, and 23 h from lanes 2–8). Similarly, the addition
`of aurothiomalate (ATM) to the reaction mixture has the same inhibitory effect on
`antibody reduction (data not shown). [Color figure can be seen in the online version of
`this article, available at wileyonlinelibrary.com.]
`
`HCCF as well. Figure 7 showed that the antibody was readily
`reduced in a homogenized HCCF generated from a 3-L
`fermentor after a period of
`incubation. However,
`the
`antibody reduction event was completely inhibited when
`either 1 mM ATG or ATM was added to the HCCF. These
`results demonstrated that the Trx system was active in the
`HCCF and directly responsible for the antibody reduction.
`
`The Source of NADPH for Trx System Activity and the
`Roles of G6P and Glucose in the Reduction Mechanism
`
`The reduction of disulfides by the Trx system requires the
`reducing equivalents from NADPH (Fig. 3). The main
`cellular metabolic pathway that provides NADPH for all
`reductive biosynthesis reactions is the pentose phosphate
`pathway. For the antibody reduction event to occur, the
`enzymes in this pathway must be still active in the HCCF in
`order to keep the Trx system active. At a minimum, the first
`step in the pentose phosphate pathway (catalyzed by G6PD)
`þ
`to NADPH while
`must be active to reduce NADP
`converting G6P to 6-phosphogluconolactone. In addition,
`G6P is most likely produced from glucose and adenosine
`50-triphosphate (ATP) by the hexokinase activity in HCCF.
`The overall mechanism of the antibody reduction in HCCF
`is summarized in Figure 4.
`The reducing activity in the HCCF appeared to be
`transitory in some cases and may be lost over time under
`certain storage conditions or after multiple freeze/thaw
`cycles. The HCCF that had lost reducing activity actually
`
`Kao et al.: Mechanism of Antibody Reduction
`
`627
`
`Biotechnology and Bioengineering
`
`

`

`Case 1:18-cv-01363-CFC Document 79-4 Filed 03/22/19 Page 7 of 11 PageID #:
`9465
`
`Figure 6.
`In vitro activity of thioredoxin system inhibited by CuSO4: digital gel-like images from the microchip-based capillary electrophoresis analysis (each lane
`representing a time point: 0, 0.5, 1, 2, 3, 21, and 23 h from lanes 2–8). a: The incubation of intact antibody (1 mg/mL) with 0.1 mM TrxR (rat liver), 5 mM Trx (human), and 1 mM NADPH in
`10 mM histidine sulfate buffer resulted in antibody reduction in <1 h. b: The addition of CuSO4 at a concentration of 50 mM to the reaction mixture effectively inhibited the antibody
`reduction. [Color figure can be seen in the online version of this article, available at wileyonlinelibrary.com.]
`
`provided an opportunity to explore the role of NADPH and
`G6P in the antibody reduction by Trx system. An HCCF
`which already experienced several freeze/thaw cycles was
`found to have lost its reducing activity (Fig. 8a) despite
`that the antibody reduction was seen previously in the
`
`freshly thawed HCCF from the same fermentation (data not
`shown). To determine if the Trx system was still active in this
`non-reducing HCCF, NADPH was added to the HCCF at a
`concentration of 5 mM. The antibody reduction event
`was observed again after the addition of NADPH (Fig. 8b).
`
`Figure 7.
`Inhibition of the antibody reduction in HCCF by aurothioglucose: digital gel-like images from the microchip-based capillary electrophoresis analysis (each lane
`representing a time point: 0, 0.5, 1, 2, 19, 21, and 23 h from lanes 2–8). a: The antibody was reduced in an incubation experiment using an HCCF generated from homogenized CCF from
`a 3-L fermentor. b: Addition of 1 mM aurothioglucose to the HCCF effectively inhibited antibody reduction. The addition of aurothiomalate (ATM) to the HCCF has the same inhibitory
`effect on antibody reduction (data not shown). [Color figure can be seen in the online version of this article, available at wileyonlinelibrary.com.]
`
`628
`
`Biotechnology and Bioengineering, Vol. 107, No. 4, November 1, 2010
`
`

`

`Case 1:18-cv-01363-CFC Document 79-4 Filed 03/22/19 Page 8 of 11 PageID #:
`9466
`
`Figure 8.
`Losing and restoring reduction activity in HCCF: digital gel-like images obtained from the microchip-based capillary electrophoresis analysis (each lane
`representing a time point: 0, 1, 2, 4, 19, and 21 h from lanes 2–7). a: The HCCF from one of the manufacturing runs that already experienced several freeze/thaw cycles was used in an
`incubation experiment. Surprisingly, no antibody reduction was observed in the microchip-based capillary electrophoresis analysis despite the antibody reduction seen previously
`in the freshly thawed HCCF from this same fermentation (data not shown). b: The reduction of antibody was observed again after the addition of NADPH at a concentration of 5 mM
`into the HCCF whose reduction activity had been previously lost. c: The reduction of antibody was also observed again in the microchip-based capillary electrophoresis assay after
`the addition of G6P at a concentration of 10 mM into the HCCF. [Color figure can be seen in the online version of this article, available at wileyonlinelibrary.com.]
`
`These results suggested that the Trx system was still intact in
`the HCCF that had lost it reducing activity. In addition, the
`reducing activity was lost in this HCCF over time because
`the NADPH source was depleted, presumably as a result of
`the oxidation of NADPH by all of the reductive reactions
`that competed for NADPH. Another experiment to support
`this hypothesis involved the addition of G6P (10 mM) to the
`same non-reducing HCCF. The results showed that the G6P
`addition was also able to reactivate the Trx system and
`subsequently reduce the antibody in the HCCF incubation
`experiment (Fig. 8c). This is a very important result as it
`indicated that the antibody reduction in HCCF was caused
`by the activities of both the Trx system and G6PD.
`Furthermore, G6PD was still active in the non-reducing
`HCCF and the loss of reduction activity in this HCCF
`appeared to be due to the depletion of G6P, which disabled
`þ
`to NADPH.
`the conversion of