Case 1:18-cv-01363-CFC Document 79-5 Filed 03/22/19 Page 1 of 10 PageID #:
`9470
`
`ARTICLE
`
`Identification and Prevention of Antibody Disulfide
`Bond Reduction During Cell Culture Manufacturing
`
`Melody Trexler-Schmidt,1 Sandy Sargis,1 Jason Chiu,2 Stefanie Sze-Khoo,1 Melissa Mun,3
`Yung-Hsiang Kao,4 Michael W. Laird3
`1Late Stage Purification, Genentech, Inc., DNA Way, South San Francisco, California 94080;
`telephone: 650-225-5137; fax: 650-225-3880; e-mail: schmidt.melody@gene.com
`2Early Stage Purification, Genentech, Inc., South San Francisco, California
`3Late Stage Cell Culture, Genentech, Inc., South San Francisco, California
`4Protein Analytical Chemistry, Genentech, Inc., South San Francisco, California
`
`Received 31 July 2009; revision received 23 December 2009; accepted 8 February 2010
`
`Published online 22 February 2010 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/bit.22699
`
`ABSTRACT: In the biopharmaceutical industry, therapeutic
`monoclonal antibodies are primarily produced in mamma-
`lian cell culture systems. During the scale-up of a mono-
`clonal antibody production process, we observed excessive
`mechanical cell shear as well as significant reduction of the
`antibody’s interchain disulfide bonds during harvest opera-
`tions. This antibody reduction event was catastrophic as the
`product failed to meet the drug substance specifications and
`the bulk product was lost. Subsequent laboratory studies
`have demonstrated that cells subjected to mechanical shear
`release cellular enzymes that contribute to this antibody
`reduction phenomenon (manuscript submitted; Kao et al.,
`2009). Several methods to prevent this antibody reduction
`event were developed using a lab-scale model to reproduce
`the lysis and reduction events. These methods included
`modifications to the cell culture media with chemicals
`(e.g., cupric sulfate (CuSO4)), pre- and post-harvest
`chemical additions to the cell culture fluid (CCF) (e.g.,
`CuSO4, EDTA, L-cystine), as well as lowering the pH and
`air sparging of the harvested CCF (HCCF). These methods
`were evaluated for their effectiveness in preventing disulfide
`bond reduction and their impact to product quality. Effec-
`tive prevention methods, which yielded acceptable product
`quality were evaluated for their potential to be implemented
`at manufacturing-scale. The work described here identifies
`numerous effective reduction prevention measures from
`lab-scale studies; several of these methods were then success-
`fully translated into manufacturing processes.
`Biotechnol. Bioeng. 2010;106: 452–461.
`ß 2010 Wiley Periodicals, Inc.
`KEYWORDS: antibody; disulfide; reduction; centrifugation;
`shear; lysis
`
`Correspondence to: M. Trexler-Schmidt
`
`Introduction
`
`Recombinant monoclonal antibodies (rMAb) have become
`prevalent as a clinical therapy over the last 20 years for
`oncology and immunological diseases (Reichert, 2001, 2002;
`Reichert and Pavolu, 2004; Reichert et al., 2005). Currently
`for commercialized products in the biopharmaceutical
`industry, rMAbs are typically produced in mammalian cell
`culture in large stainless steel fermentors up to 25-kL in scale
`(Benton et al., 2002; Kelley, 2007). After initiating a
`production bioreactor, various process additions and
`parameter manipulations are performed to maximize
`growth and antibody production and yield suitable product
`quality (Andersen and Krummen, 2002; Andersen and
`Reilly, 2004; Birch et al., 2005; Birch and Racher, 2006;
`Wurm, 2004). The rMAbs are expressed and secreted
`extracellularly into the cell culture fluid (CCF). At the end of
`the production phase, the feedstock is usually harvested by
`disc stacked centrifugation followed by depth filtration or by
`tangential flow microfiltration (Kempken et al., 1995; Roush
`and Lu, 2008). The resulting harvested cell culture fluid
`(HCCF) is then purified using protein A affinity chromato-
`graphy, a series of alternative chromatography steps, and
`finally formulated by ultrafiltration/diafiltration or size
`exclusion chromatography (Fahrner et al., 2001; Kelley,
`2007).
`During process development of the centrifugation step,
`clarification efficiency during harvest operations is typically
`determined by measurement of centrate turbidity, particle
`size distribution, and/or filterability (Kempken et al., 1995;
`Roush and Lu, 2008). In addition, mechanical cell lysis is
`also an important consideration during large-scale harvest
`operations. Excessive mechanical shear during harvest may
`impact product quality (e.g., aggregates) (Hutchinson et al.,
`2006) and/or
`the release of undesirable intracellular
`components into the HCCF that could degrade the product
`
`452 Biotechnology and Bioengineering, Vol. 106, No. 3, June 15, 2010
`
`ß 2010 Wiley Periodicals, Inc.
`
`

`

`Generation of CCF and Production of rMAb
`
`Mammalian cell culture fluids derived from Chinese hamster
`ovary (CHO) cells were generated using a representative
`small-scale fermentation process similar to the methods
`described previously (Chaderjian et al., 2005). Cell culture
`process indicators (e.g., pH, temperature, dissolved oxygen
`(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.
`
`Lab-Scale HCCF Preparation
`
`At the end of the production culture, complete mechanical
`lysis of CCF was achieved by high pressure homogenization
`using a Microfluidics HC-8000 homogenizer. The pressure
`regulator of the instrument was set to 4,000–8,000 psi, and
`complete cell lysis (membrane breakage) was achieved after
`a single pass, as determined by a lactate dehydrogenase
`(LDH) assay. The homogenate was then blended with the
`original CCF (non-mechanically lysed) in order to obtain
`final pools with desired target amounts of total lysis levels.
`The blends of homogenate and CCF were centrifuged in a
`Sorval 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).
`
`Case 1:18-cv-01363-CFC Document 79-5 Filed 03/22/19 Page 2 of 10 PageID #:
`9471
`Materials and Methods
`of interest or be difficult to clear downstream. Sources of cell
`shear during harvest could be from the centrifuge
`equipment alone including feed zones or from any flow
`restrictions between the fermentor and centrifuge such as an
`inlet feed pump and inlet flow control valve.
`During scale-up of a rMAb process, we observed
`significant reduction of the antibody’s interchain disulfide
`bonds during harvest operations as well as significant
`mechanical cell lysis. Subsequent laboratory studies demon-
`strated that cells subjected to mechanical shear release
`cellular components
`that contribute to this antibody
`reduction event. When a sufficient amount of these active
`enzymes were present, the extent of product disulfide bond
`reduction was then dependent on HCCF incubation time
`(duration), temperature, and dissolved oxygen level. The
`intrachain disulfide bonds were not reduced as determined
`by both mass spectrometry and reversed-phase HPLC
`analysis. Previous work by Zhang and Czupryn (2002)
`identified the presence of
`free sulfhydryls and non-
`covalently associated antibody fragments after purification
`from CHO-derived HCCF. However, the authors proposed
`that the origin of these species was related to incomplete
`disulfide bond formation in the endoplasmic reticulum as
`a consequence of inefficient assembly of heavy and light
`chains and not
`the result of degradation via cellular
`components.
`The disulfide reduction event described here in this work
`was somewhat difficult to monitor in the HCCF, but was
`readily analyzed in the purified protein A pool when protein
`A affinity chromatography was used as the capture step.
`The enzymatic reduction activity was established through
`observations of an increase in the level of free thiols during
`HCCF incubation as well as the required involvement of
`macromolecules in a dialysis study (data not shown). The
`currently understood cellular mechanism for rMAb reduc-
`tion is discussed in a subsequent manuscript (manuscript
`submitted; Kao et al., 2009).
`Here in this work, we describe that mechanical cell shear
`of highly viable cells is required to induce the antibody
`reduction phenomenon observed during large-scale man-
`ufacturing. Multiple approaches were tested in a lab-scale
`reduction susceptibility model to determine effective con-
`ditions for preventing disulfide reduction during large-scale
`manufacturing. The methods capable of preventing disulfide
`reduction were evaluated for their process feasibility as well
`as their impact to other product quality attributes. Our
`strategy is to inactivate the reducing enzymes in the HCCF
`to prevent antibody reduction. It is not desirable to oxidize
`the reduced interchain disulfides after allowing antibody
`reduction to occur because the impact of uncontrolled re-
`oxidation of the antibody poses a potential risk to product
`quality. For example, reduced antibody might react with
`other thiol containing molecules in the HCCF or form
`aggregates through cross linking between reduced cysteines
`in two different antibodies. This report discusses the
`antibody reduction events, prevention, and implementation
`in large-scale manufacturing.
`
`Chemical Inhibitor Additions
`
`The following separate stock solutions were used in the lab-
`scale HCCF hold time studies: (1) 250 mM EDTA, pH 7.4
`prepared using EDTA, disodium dihydrate ( Sigma-Aldrich,
`St. Louis, MO) and EDTA, tetrasodium dihydrate (Sigma);
`(2) 50 or 80 mM cupric sulfate pentahydrate (CuSO4)
`(Sigma); (3) 1.0 M acetic acid solution (Mallindkrodt,
`Hazelwood, MO); and (4) 200 mM L-cystine (Fluka, Sigma-
`Aldrich, St Louis, MO). The EDTA, CuSO4, acetic acid, or L-
`cystine stock solutions were added to either the CCF prior to
`homogenization or directly to the HCCF (post-harvest) to
`evaluate a range of final concentrations to prevent antibody
`disulfide reduction. Additions made prior to homogeniza-
`tion were to mimic a pre-harvest addition in a manufactur-
`ing setting.
`
`HCCF Pool Incubation in Mini-Tanks
`
`The HCCF pools were generated from either manufactur-
`ing-scale runs via large-scale centrifugation and depth
`filtration or from lab-scale studies as discussed in the
`previous section. The HCCF was held in 50 mL 316L
`stainless steel mini-tank containers (Flow Components,
`
`Trexler-Schmidt et al.: Antibody Disulfide Bond Reduction
`
`453
`
`Biotechnology and Bioengineering
`
`

`

`Percent cell lysis was determined by measuring the level of
`LDH, an enzyme that catalyzes the oxidation of lactate to
`pyruvate (Babson and Babson, 1973; Legrand et al., 1992).
`All sample aliquots were stored in 0.1 g/L saponin at
`<708C until analysis. The lysis percentage in the CCF
`supernatant, centrate, and HCCF samples was calculated by
`dividing the LDH level in the selected sample by the LDH
`level in the whole cell sample.
`
`Monoclonal Antibody Protein Concentration Assays
`
`For HCCF samples, rMAb concentration was assayed by an
`HPLC-based protein A method to measure rMAb titer
`values. The antibody concentration in the purified protein A
`pool was measured using UV spectrometry at 280 nm and
`the known extinction coefficient for that particular rMAb.
`
`Case 1:18-cv-01363-CFC Document 79-5 Filed 03/22/19 Page 3 of 10 PageID #:
`9472
`Cell Lysis Measurements
`Dublin, CA). The stainless steel mini-tanks were filled with
`approximately 30–50 mL HCCF, sealed with a clamp to
`avoid air exposure, and stored at the desired incubation
`temperature (usually ambient, 18–228C). The solutions
`held in the mini-tanks were not aerated or agitated. For
`experiments with only HCCF sample analysis, samples
`were taken at pre-determined time-points and immediately
`frozen at <708C until analysis. For experiments with
`protein A purification, at pre-determined time points
`(usually after 1- and 2-day holds), the HCCF solution was
`removed from the mini-tank and immediately purified
`over a lab-scale protein A affinity column. The generated
`protein A pools were immediately frozen at <708C until
`analysis.
`
`HCCF Air Sparging
`
`steel vessels
`sparging studies, 15-L stainless
`For air
`(Sartorius Stedim Biotech, Aubagne, France) were used.
`Approximately 4 L of HCCF was 0.22 mm sterile filtered into
`each sterilized vessel. HCCF temperature was controlled to
`208C, and the HCCF was agitated during incubation at a rate
`of 50 rpm (15-L fermentor). Dissolved oxygen, oxidative-
`reduction potential (ORP), and pH were monitored on-line
`throughout
`each study using three
`separate probes
`manufactured by Broadley-James, Irvine, CA. The pH of
`the HCCF was not controlled in these studies. The HCCF
`was constantly sparged with either air to increase the
`dissolved oxygen level or with nitrogen (control) to remove
`any dissolved oxygen in solution. Gas flow to each vessel
`ranged between 0.01 and 0.02 vvm (50 mL/min) in order
`to be representative of a potential future flow rate for a
`manufacturing process. At pre-determined time points,
`50 mL samples were removed from both vessels and
`purified over a lab-scale protein A affinity column prior
`to analysis.
`
`Protein A Processing
`
`Antibody purification from the HCCF samples was achieved
`by Protein A affinity chromatography (Millipore, Billerica,
`MA, Prosep-vA High Capacity or GE Healthcare, Uppsala,
`Sweden, MabSelect SuRe). The resin was packed in a 0.66 cm
`inner diameter glass column (Omnifit, Diba Industries,
`Danbury, CT) with a 14–20 cm bed height resulting in a 4.8–
`6.8 mL final column volume. Chromatography was per-
`formed using an AKTA Explorer 100 chromatography
`system (GE Healthcare) at ambient
`temperature. The
`protein A purification process varied, depending on the
`rMAb of interest and the resin. An acidic buffer (pH 3) was
`used for elution of product, and each pool was pH adjusted
`to pH 5–7 using a stock solution of Sodium HEPES or Tris
`base. Protein A elution pool samples aliquots were stored at
`<708C until analysis by the Bioanalyzer assay to quantitate
`the percentage of non-reduced antibody at 150 kDa.
`
`454
`
`Biotechnology and Bioengineering, Vol. 106, No. 3, June 15, 2010
`
`Disulfide Bond Reduction Assay
`
`For non-reduced SDS-PAGE analyses (Laemmli, 1970),
`samples were mixed with 4 NuPAGE LDS sample buffer
`(Life Technologies, Invitrogen, Carlsbad, CA), heated 5–
`10 min at 708C, and then loaded onto a precast NuPAGE 4–
`12% Bis-Tris gel (Invitrogen) with MOPS running buffer.
`The molecular weight marker was Mark12 Unstained
`Standard (Invitrogen, LC5677). Gels were stained with
`1
`either Coomassie Blue (12 mg protein per well) or SYPRO
`Ruby stain (2–3 mg product per well) with a 5–7 s exposure.
`Microchip capillary electrophoresis (CE) (Agilent 2100
`Bioanalyzer) was used to quantitate the level of non-reduced
`antibody. Sample preparation was carried out as described
`in the Agilent Protein 230 Assay Protocol (G2938-90052)
`with minor changes defined here. Samples were diluted to
`1 g/L with water prior to preparation, and then 0.5% SDS
`without iodoacetamide (IAM) and 2 mL of denaturing
`solution were used in the denaturing step. Samples for CE
`analysis were heated for 5 min at 708C.
`
`Results
`
`Detection of Antibody Disulfide Reduction
`
`During a manufacturing campaign for a clinical IgG1
`monoclonal antibody product,
`two out of five runs
`exhibited interchain disulfide reduction in the protein A
`pool (Fig. 1). In the runs where the antibody was reduced
`(Runs C and E), a decrease in the amount of the intact IgG
`band (150 kDa) and an increase in the amount of heavy
`chain (50 kDa) and light chain (25 kDa) fragments was
`observed. There was also a slight increase in the heavy-
`heavy-light band (125 kDa) compared to Runs A, B, and D.
`Given that only the Fc portion of a monoclonal antibody
`should bind to the protein A ligand, it was unexpected to
`observe the presence of light chain in the protein A pool.
`This occurrence is most
`likely due to a non-covalent
`
`

`

`Case 1:18-cv-01363-CFC Document 79-5 Filed 03/22/19 Page 4 of 10 PageID #:
`9473
`the reducing enzymes released by the lysed cells
`that
`are removed during the load phase of
`the protein A
`operation and therefore removed from the manufacturing
`process. In addition, spiking of various purified antibodies
`into the active protein A load flow-through solution also
`resulted in disulfide reduction confirming that the event was
`not molecule structure specific to this particular IgG1 product
`(data not shown). When the protein A load flow-through,
`collected using manufacturing-scale HCCF, was incubated at
`208C in 50 mL Falcon tubes for 4 days, the activity of reducing
`enzymes was lost. However, after incubation of protein A load
`flow-through at <708C and 58C for 4 days, the reducing
`activity was preserved as evident in disulfide reduction of
`antibody after the spike (data not shown).
`
`Figure 1. Non-reduced SDS-PAGE (Coomassie Blue stain) analysis of five
`protein A pools from large-scale manufacturing. Lane 1, Molecular weight marker
`(Invitrogen, LC5925): 191, 97, 64, 51, 39, 28, 19, 14 kDa; Lane 2, Run A; Lane 3, Run B;
`Lane 4, Run C; Lane 5, Run D; Lane 6, Run E.
`
`interaction of the heavy and light chain of the antibody even
`with reduced interchain disulfide bonds. Future work will
`investigate the role of the protein A step in disulfide
`reduction including addition of an alykating agent such as
`iodoacetamide (IAM) to the HCCF post-incubation and
`prior to protein A purification to eliminate the purification
`process from influencing the amount of reduced antibody.
`
`Isolation of Reducing Component(s) Activity
`
`HCCF generated from a manufacturing run was purified
`over a lab-scale protein A column and in addition to the
`collection of the elution pool, the load flow-through was
`also collected in order to isolate the reducing enzymes
`during the protein A process. The lab-scale protein A pool
`was adjusted from pH 5.3 to pH 7.0 to maintain a similar pH
`to the HCCF pH where antibody reduction was typically
`observed. Pure antibody was spiked into the protein A load
`flow-through at a concentration of 1 g/L to mimic the
`concentration in the original HCCF. All three solutions
`(original HCCF; proA pool; antibody spiked protein A load
`flow-through) were held separately in stainless steel mini-
`tanks at 258C for 22 h. Two mini-tanks were used per
`solution to start the ‘‘t ¼ 0’’ time point at different times of
`the day in order to obtain hold time data at more frequent
`intervals. Figure 2 shows non-reduced SDS-PAGE for all
`three holds where the loss of intact IgG (150 kDa) and
`additional fragments were observed in both the HCCF and
`the antibody spiked protein A load flow-through samples
`after 14 h of incubation. Excess light chain (25 kDa) was
`also observed and is the result of excessive expression during
`the cell culture process. Antibody reduction was not
`observed in the protein A pool incubation, thus confirming
`
`Lab-Scale Antibody Reduction Model
`
`In order to develop reduction prevention strategies that
`could be implemented into the manufacturing process,
`it was imperative to first develop a lab-scale model to
`accurately reproduce the event. Although the HCCF
`generated from the manufacturing process can exhibit
`disulfide reduction during lab-scale incubation, the redu-
`cing components may lose activity during long-term HCCF
`storage at 58C, after HCCF freeze/thaw, and/or after
`exposure to oxygen from air surface transfer. In addition,
`reductant-active HCCF from the manufacturing process
`may have been generated just on the threshold of lysis
`required for disulfide reduction, thereby limiting the release
`of
`intracellular reducing components. This resulted in
`experimental inconsistencies when reproducing the reduc-
`tion event in a lab setting. These complications make it
`difficult to troubleshoot and inhibit the disulfide reduction
`event that occurred during large-scale harvest operations.
`Since conducting studies at large-scale is not practical, a
`lab-scale reduction susceptibility model was developed to
`reproduce the large-scale reduction event and eliminate
`most, if not all, of the variables previously observed.
`Homogenization was used to fully lyse the mammalian
`cells (100% lysis) once the cell culture process was complete.
`The homogenate was then diluted with CCF supernatant
`(non-homogenized) to achieve the desired levels of total cell
`lysis. Figure 3a shows the reduction susceptibility curves
`using CCF from different cell culture runs for two different
`products at various amounts of total
`lysis. The HCCF
`generated from the blending procedures was incubated at
`208C for 1 day and then purified using protein A chromato-
`graphy. For Product A, the final cell culture viability
`measured by Trypan Blue cell counts ranged between 70%,
`80%, and 90% and the initial cell lysis measured by the LDH
`assay was 25%, 20%, and 10%, respectively. For Product B,
`both cell culture materials tested were 90% viable by cell
`count and 10% initial cell lysis by LDH. Therefore, cell
`death measurements using both assays matched as expected.
`From this data, Product A exhibited disulfide bond
`reduction at 60–70% total
`lysis for the three different
`cultures, while Product B reduced at 40% total lysis for the
`
`Trexler-Schmidt et al.: Antibody Disulfide Bond Reduction
`
`455
`
`Biotechnology and Bioengineering
`
`

`

`Case 1:18-cv-01363-CFC Document 79-5 Filed 03/22/19 Page 5 of 10 PageID #:
`9474
`
`Figure 2. Non-reduced SDS-PAGE analysis (SYPRO1 Ruby stain) during mini-tank incubation at 258C. Molecular weight marker (Invitrogen, LC5677; 200, 116.3, 97.4, 66.3, 55.4,
`36.5, 31.0, 21.5, 14.4, 6.0 kDa) is shown and then time of pool incubation (h) is listed above each gel. Duplicate times represent samples taken from a duplicate mini-tank.
`a: Manufacturing-scale generated HCCF. b: Lab-scale generated protein A pool. c: Lab-scale generated protein A load flow-through spiked with pure antibody.
`
`two different cultures shown. Note that different rMAb
`processes as well as different cell culture conditions and/or
`cell culture performance within each rMAb process can lead
`to differences in susceptibility to disulfide bond reduction. It
`has also been observed that healthier, actively growing cells
`contain more reducing components than cultures contain-
`ing mostly non-viable cells and cultures exhibiting sharp
`declines in cell viability. With so many factors affecting
`reduction susceptibility, comparing products in terms of
`percent
`lysis is currently the most practical approach
`although future analysis could include normalization to
`other factors such as cell density or total cellular protein.
`For Product A, disulfide reduction was observed during
`manufacturing runs at 50% total lysis which is slightly
`lower than that observed in the lab-scale model. This could
`be due to differences in the HCCF dissolved oxygen level in
`the manufacturing-scale HCCF hold tank and the lab-scale
`mini-tank.
`Another way to evaluate these results is to plot the
`additional lysis instead of total lysis since this represents the
`mechanical lysis contribution from the harvest operations in
`a manufacturing run (Fig. 3b). This data plot is then heavily
`dependent on the initial level of cell lysis for each study.
`Product A reduced at 50% additional lysis for the three
`different cultures shown, while Product B reduced at 30%
`additional
`lysis for the two different cultures shown.
`
`456
`
`Biotechnology and Bioengineering, Vol. 106, No. 3, June 15, 2010
`
`For Product A, disulfide reduction was observed during
`manufacturing runs at 30% additional
`lysis, again
`suggesting differences in the HCCF dissolved oxygen level
`between manufacturing-scale and lab-scale. One disadvan-
`tage of the lab-scale blending model to determine reduction
`susceptibility is that homogenization requires taking the
`cells to extreme lysis and then diluting back to a particular
`percentage compared to the manufacturing process where
`mechanical cell lysis from the harvest operation is added to
`the starting level of lysis at the end of the cell culture process.
`
`Chemical Inhibition of Disulfide Reduction
`
`Chemical additions were evaluated as a method to prevent
`disulfide reduction during HCCF incubation. Since rMAb
`reduction is caused by reducing enzymes, inhibitors for
`any of the enzymes involved in the pathway may prevent
`reduction. Chemical inhibition methods that were evaluated
`included (1) an increase in the CuSO4 concentration in the
`cell culture basal media as previously evaluated in a separate
`study (Chaderjian et al., 2005), (2) the addition of EDTA
`to pre-harvest CCF or HCCF, (3) the addition of CuSO4
`to pre-harvest CCF or HCCF, (4) lowering the pH of the
`pre-harvest CCF or HCCF to pH 5.0–5.5, and (5) the
`addition of L-cystine to the pre-harvest CCF. As a known
`
`

`

`Case 1:18-cv-01363-CFC Document 79-5 Filed 03/22/19 Page 6 of 10 PageID #:
`9475
`
`Figure 3. Reduction susceptibility curves for two different rMAb products.
`Protein A pools were analyzed after 1 day of HCCF incubation at 208C. a: Graphed
`lysis on the x-axis. b: Graphed with additional mechanical
`with total
`lysis post-
`fermentation on the x-axis. Total
`lysis is calculated by combining the end of
`fermentation natural
`lysis value (e.g., an 80% viable culture exhibits 20% natural
`lysis) plus the added lysis value obtained from the mechanical shear. The percent
`intact antibody is calculated via the Bioanalyzer analysis.
`
`metal chelator and protease inhibitor, EDTA was added
`to sequester the metal
`ions that that may be required
`for activities of enzymes involved in rMAb reduction.
`The addition of copper can prevent rMAb reduction by
`maintaining the reducing components in their oxidized
`form and/or acting as a direct enzyme inhibitor. L-Cystine
`may function as a competitive inhibitor against the reducing
`components. Finally,
`lowering the CCF/HCCF pH to
`below pH 6.0 stabilizes the protonated form of free thiols
`and thus decreases the reduction activity.
`HCCF generated from a manufacturing run was
`incubated in mini-tanks for 32 h in the absence of EDTA
`or for 72 h in the presence of 12 mM EDTA. In the absence of
`EDTA, disulfide reduction was observed after only 7 h of
`incubation at 208C via a noticeable decrease in the 150 kDa
`band (intact IgG) and an increase in the 50 kDa band (heavy
`chain) as shown in Figure 4. By 16 h, there was complete
`reduction of antibody. However, in the presence of EDTA,
`reduction was inhibited for the entire 72 h incubation. In
`
`Figure 4. Non-reduced SDS-PAGE analysis (SYPRO1 Ruby stain) during HCCF
`incubation in mini-tanks at 208C. Time of incubation (h) is listed above each gel. a: No
`EDTA addition. b: 12 mM EDTA addition to HCCF prior to incubation.
`
`a separate study, CuSO4 was added to manufacturing-
`generated HCCF and incubated in mini-tanks at 208C for
`48 h. At selected time points, the HCCF was purified over
`protein A to analyze the product in the absence of reducing
`components. Figure 5 shows the inhibition from the
`addition of 50 mM CuSO4 compared to the control case
`where no CuSO4 was added. While analysis of crude HCCF
`samples was faster and more time points were able to be
`obtained for analyzing reaction kinetics, it was determined
`that protein A purification of the incubated HCCF samples
`was optimal for robust reduction analysis by avoiding the
`presence of reducing components that could interfere with
`the analytical measurement.
`Once a lab-scale reduction susceptibility model was
`implemented and proven to be robust and reproducible,
`chemical inhibitors were then tested in this model. Lab-scale
`generated HCCF via cell lysis not only can enable testing
`inhibitor robustness at higher levels of cell lysis, but the
`large-scale manufacturing generated HCCF tested usually
`has lower reducing activity due to freeze/thaw. In order to
`test chemical inhibitors in the lab-scale model, additions can
`occur to the cell culture basal media, pre-homogenization,
`or
`to the final blended HCCF itself. As
`shown in
`Figure 6a, four different inhibition methods: (1) addition
`
`Trexler-Schmidt et al.: Antibody Disulfide Bond Reduction
`
`457
`
`Biotechnology and Bioengineering
`
`

`

`Case 1:18-cv-01363-CFC Document 79-5 Filed 03/22/19 Page 7 of 10 PageID #:
`9476
`
`Figure 5. CuSO4 was added to manufacturing-generated HCCF and incubated in
`mini-tanks at 208C for 48 h. HCCF samples were taken and frozen as well as purified
`over protein A chromatography at the shown time points. No CuSO4 was added to the
`control conditions. Percent 150 kDa peak was quantitated for both the HCCF and
`protein A pools, and all results were normalized to the t ¼ 0 h time point
`for
`consistency.
`
`of EDTA pre-harvest, (2) addition of CuSO4 to the cell
`culture batch medium, (3) addition of CuSO4 pre-harvest,
`and (4) adjustment of the HCCF from pH 7.0 to 5.5, used
`independently for one feedstock, were effective in prevent-
`ing disulfide reduction at 75% total lysis during 2 days of
`HCCF hold at 208C. Without an inhibitor addition, the
`product was completely reduced after just 1 day of HCCF
`incubation at 208C. A separate study demonstrated that
`concentrations of 20 mM EDTA and 50 mM CuSO4 added
`pre-homogenization were not robust at completely inhibit-
`ing reduction when tested at the extreme condition of 100%
`total lysis (data not shown).
`For another rMAb product, L-cystine was also evaluated
`as a competitive inhibitor for the reducing enzymes as
`shown in Figure 6b. L-Cystine concentrations 2.2 mM or
`CuSO4 concentrations 75 mM were effective at preventing
`lysis, whereas 1.1 mM L-cystine
`reduction at 60% total
`was not effective. Maximum expected levels of
`total
`lysis expected in the manufacturing process with the
`products shown is 50–60% (equivalent to 30% additional
`lysis for 70–80% viable cell cultures) under standard harvest
`lysis (equivalent to 10%
`conditions, or 30–40% total
`additional
`lysis for 70–80% viable cell cultures) under
`improved harvest conditions that minimize cell shear. Since
`each product’s final cell viability at the end of the cell culture
`production phase will vary, it is also important to track and
`evaluate additional cell lysis during harvest operations.
`To evaluate chemical additive robustness to inhibit
`reduction across multiple feed streams,
`three different
`feedstocks consisting of three separate starting lab-scale cell
`culture runs with the same product and cell culture process
`parameters were processed through the lab-scale lysis model.
`Levels of total lysis targeted in these three studies were
`
`458
`
`Biotechnology and Bioengineering, Vol. 106, No. 3, June 15, 2010
`
`Figure 6. Protein A pool results are shown where different reduction inhibition
`methods were tested for two different products. Stock solutions of CuSO4, EDTA, or L-
`cystine were added to the initial cell culture medium or pre-homogenization. To
`decrease the HCCF pH, a stock solution of glacial acetic acid was added to the HCCF to
`decrease the pH from pH 7.0 to pH 5.5. The final HCCF pools were incubated at 208C
`and purified using protein A chromatography after 1- and 2-day holds. a: Results from
`one product tested at 75% total lysis. The 2-day time point from the HCCF without
`additions was not analyzed. b: Results from a second product tested at 60% total
`lysis. The 0-day time point was not analyzed for any of the conditions tested.
`
`approximately 65% and/or 75% total lysis. Stock solutions
`of either EDTA or CuSO4 were added either pre-
`homogenization or to the HCCF directly to achieve final
`concentrations of 20 mM EDTA or 30–35 mM CuSO4 in the
`HCCF. As shown in Figure 7, both 20 mM EDTA and 30–
`35 mM CuSO4 were robust at inhibiting reduction during
`lab-scale incubations at 208C. Feedstock variability in terms
`of reduction susceptibility was observed in which Feedstock
`B did not completely reduce to <5% 150 kDa peak at either
`65% or 75% total cell lysis compared to Feedstocks A and C.
`Lowering the pre-harvest CCF or the HCCF pH from pH
`7.0 to 5–6 causes a large amount of precipitation of host
`cell proteins and DNA (Lydersen et al., 1994; Roush and Lu,
`2008). Additional studies were conducted to evaluate if the
`loss of reducing activity was due to removal of reducing
`components via this precipitation event instead of proto-
`nated thiols at the lower pH. After pH adjustment and
`
`

`

`Case 1:18-cv-01363-CFC Document 79-5 Filed 03/22/19 Page 8 of 10 PageID #:
`9477
`dissolved oxygen concentration above levels permissive of
`disulfide reduction. HCCF generated from 100% lysed CCF
`was held in two separate stainless steel 15-L fermentors and
`either air or nitrogen sparged for 36 h at 208C. The results
`showed that approximately 85% intact antibody was present
`in the i