Case 1:18-cv-01363-CFC Document 79-8 Filed 03/22/19 Page 1 of 9 PageID #: 9523
`
`ARTICLE
`
`Air Sparging for Prevention of Antibody
`Disulfide Bond Reduction in Harvested
`CHO Cell Culture Fluid
`
`Melissa Mun,1 Stefanie Khoo,2 Aline Do Minh,1 James Dvornicky,2
`Melody Trexler-Schmidt,2 Yung-Hsiang Kao,3 Michael W. Laird1
`1Late Stage Cell Culture, Genentech, Inc., 1 DNA Way, South San Francisco, California
`94080; telephone: 650 467 4596; e-mail: mlaird@gene.com
`2Purification Development, Genentech, Inc., San Francisco, California
`3Protein Analytical Chemistry, Genentech, Inc., San Francisco, California
`
`ABSTRACT: During the scale-up of several Chinese Hamster Ovary
`(CHO) cell monoclonal antibody production processes, significant
`reduction of the antibody interchain disulfide bonds was observed.
`The reduction was correlated with excessive mechanical cell shear
`during the harvest operations. These antibody reduction events
`resulted in failed product specifications and the subsequent loss of
`the drug substance batches. Several methods were recently developed
`to prevent antibody reduction, including modifying the cell culture
`media, using pre- and post-harvest chemical additions to the cell
`culture fluid (CCF),
`lowering the pH, and air sparging of the
`harvested CCF (HCCF). The work described in this paper further
`explores the option of HCCF air sparging for preventing antibody
`reduction. Here, a small-scale model was developed using a 3-L
`bioreactor to mimic the conditions of a manufacturing-scale harvest
`vessel and was subsequently employed to evaluate several air
`sparging strategies. In addition, these studies enabled further
`understanding of the relationships between cell lysis levels, oxygen
`consumption, and antibody reduction. Finally, the effectiveness of air
`sparging for several CHO cell lines and the potential impact on
`product quality were assessed to demonstrate that air sparging is an
`effective method in preventing antibody reduction.
`Biotechnol. Bioeng. 2015;112: 734–742.
`ß 2014 Wiley Periodicals, Inc.
`KEYWORDS: antibody; disulfide;
`dissolved oxygen; air
`
`lysis;
`
`sparging;
`
`reduction;
`
`stainless steel bioreactors (Andersen and Krummen, 2002; Wurm,
`2004) and the product is separated from the cells via centrifugation
`(Kempken et al., 1995; Roush and Lu, 2008). The product then
`undergoes further downstream purification (Fahrner et al., 2001;
`Kelley, 2007) prior to the bulk formulation steps. During the
`centrifugation process, cells are subjected to mechanical shear
`(Hutchinson et al., 2006) which can lead to the disruption of cell
`membrane integrity. Reduction of antibody interchain disulfide
`bonds has been observed in the scale-up of CHO rMAb production
`processes at Genentech as a result of excessive cell shear during the
`centrifugation process (Trexler-Schmidt et al., 2010). The reduction
`events were caused by intracellular components,
`identified as
`thioredoxin (Trx/TrxR) or thioredoxin-like enzymes, their asso-
`ciated enzyme pathway intermediates (e.g., glucose-6-phosphate
`dehydrogenase, hexokinase), and an energy source (e.g., NADPH)
`(Kao et al., 2010; Koterba et al., 2011), which were released upon cell
`lysis. Studies have shown that disulfide bond reduction is correlated
`to levels of mechanical cell lysis of viable, actively growing cells
`(Hutterer et al., 2013; Trexler-Schmidt et al., 2010) which results in
`the release of active thioredoxin system components. In contrast,
`reduction enzymes released upon cell death may not be active at the
`time of harvest and the energy source may be depleted. Multiple
`strategies have been developed to prevent this antibody reduction
`event including the addition of chemical inhibitors to the CCF and
`HCCF, as well as maintaining a minimum dissolved oxygen (dO2)
`level
`in the HCCF via air sparging to promote an oxidizing
`environment for the antibody (Trexler-Schmidt et al., 2010).
`Mammalian cells rely on several systems to maintain the
`oxidation reduction potential of each intracellular compartment at
`the appropriate state. Some of the sulfhydryl-containing oxido-
`reductase systems commonly studied include thioredoxin/thiore-
`doxin reductase (Trx/TrxR) and glutathione/glutathione disulfide
`(GSH/GSSG). These systems regulate cellular events such as cell
`signaling (Filomeni et al., 2002), formation of disulfide bonds
`(Cumming et al., 2004; Jessop and Bulleid, 2004) and gene
`transcription (Sen and Packer, 1996), and also serve to protect the
`cells from reactive oxygen species (ROS) (Linke and Jakob, 2003;
`Nordberg and Arner, 2001; Shen et al., 2005). Oxygen plays a role as
`
`Introduction
`
`(rMAb) are commonly
`Recombinant monoclonal antibodies
`produced in CHO cells in the biotherapeutics industry. In large-
`scale manufacturing of rMAb, cells are generally cultured in
`
`The present address of Aline Do Minh is Genipro (STNH), Inc.
`Correspondence to: M. W. Laird
`Received 24 June 2014; Revision received 24 September 2014; Accepted 3 November
`2014
`Accepted manuscript online 11 November 2014;
`Article first published online 23 December 2014 in Wiley Online Library
`(http://onlinelibrary.wiley.com/doi/10.1002/bit.25495/abstract).
`DOI 10.1002/bit.25495
`
`734 Biotechnology and Bioengineering, Vol. 112, No. 4, April, 2015
`
`ß 2014 Wiley Periodicals, Inc.
`
`

`

`Case 1:18-cv-01363-CFC Document 79-8 Filed 03/22/19 Page 2 of 9 PageID #: 9524
`
`the terminal electron acceptor in these oxidoreductase systems
`(Shimizu and Hendershot, 2009) and oxygen supplementation has
`previously been applied in protein refolding to promote the correct
`formation of disulfide bonds (Fischer et al., 1993; Menzella et al.,
`2002). With this understanding in mind and coupled with the
`knowledge gained from the studies described by Trexler-Schmidt
`et al. (2010), the dO2 trends of HCCF derived from manufacturing-
`scale harvests were evaluated. Surprisingly, the dO2 level dropped to
`0% within a few hours. This appears to be linked to mechanical cell
`lysis and plays an important role in the subsequent disulfide bond
`reduction event. Therefore, implementation of air sparging in the
`HCCF vessel to promote an oxidizing environment should aid in
`preventing reduction of disulfide bonds.
`Dissolved oxygen control strategies are widely applied in cell
`culture bioreactors. A common sparging strategy employs a dO2
`probe which provides feedback to a PID control loop to scale the air
`and/or oxygen sparge output to a rate appropriate to maintain the
`desired dO2 set-point. Many factors are considered in bioreactor
`design in order to provide adequate gas transfer to maintain high
`cell densities while minimizing potential harmful shear effects from
`agitation or bubble rupture that could damage the cells or protein
`(Al-Rubeai et al., 1995; Chisti, 2001; Merchuk, 1991; Trinh et al.,
`1994).
`This paper explores the method of air sparging as an antibody
`reduction mitigation strategy through the development and
`utilization of a small-scale model to evaluate process requirements
`and characteristics. First, a small-scale model was established using
`a 3-L bioreactor to mimic the conditions of an HCCF tank. This
`small-scale model was then used to evaluate several air sparging
`strategies for effectiveness at preventing disulfide bond reduction
`and potential impact on product quality. Several CHO cell lines were
`tested with this method to ensure robustness for manufacturing and
`to determine if this could be universally applied to prevent antibody
`reduction. These studies also permitted further understanding of
`the relationships between cell lysis levels, oxygen consumption, and
`antibody reduction.
`
`Materials and Methods
`
`Generation of CCF and Production of rMAb in Bioreactors
`
`CHO cells were cultured in 3-L glass stirred tank bioreactors
`(Applikon) in conditions similar to those previously described by
`Chaderjian et al. (2005). Culture conditions (e.g., temperature,
`pH, dO2, and agitation) were controlled and monitored on-line. Off-
`line measurements of pH, dissolved gases (pO2, pCO2), sodium, and
`metabolite concentrations
`(glucose,
`lactate, ammonia) were
`obtained with a NOVA Bioprofile Analyzer. Daily samples were
`taken to monitor cell growth, viability, and titer.
`
`Small-Scale Lysis Model to Generate HCCF With
`Reduction Activity
`
`CCF at the end of the production culture was homogenized using a
`Microfluidics HC-8000 homogenizer, centrifuged, and filtered as
`described previously by Trexler-Schmidt et al. (2010). The non-
`mechanically lysed material (also referred to as “non-lysed”) has
`
`cell lysis levels present from the end of the production culture.
`“Non-lysed” cell lysis levels were consistently between 10–30% for
`the feedstocks used for these studies. The lysis levels reported in
`the data figures throughout this work reflect total
`lysis levels
`(“non-lysed”þ mechanical lysis).
`
`HCCF Pool Incubation in Small-Scale Vessels for
`Sparging Studies
`
`HCCF pools were incubated at a 2-L working volume in sterile 3-L
`glass stirred tank bioreactors (Applikon) with downflow pitched
`blade impellers for up to two days. Conditions were monitored with
`dO2 (Mettler Toledo), pH (Broadley James), and temperature
`probes. Digital control units (B. Braun) were used to control
`agitation rate and maintain dO2 at or above the specified set-points
`by delivering air through an open pipe sparger as required. Air
`sparge rates ranged from 0–50 standard cubic centimeters per
`minute (sccm), which corresponds to 0–0.025 vvm. No N2 sparge
`was used to decrease the dO2 level if it exceeded set-point. The pH
`levels were maintained at specified set-points if called for by the
`experiment design by addition of CO2 or 1 M Na2CO3. HCCF was
`held at ambient temperature without temperature control (18–
`22C). The agitation rate was controlled at 50–100 rpm. Unless
`otherwise noted, a 50 sccm N2 overlay was supplied for all cases
`with dO2 control, while no N2 overlay was employed for non-
`sparged holds. After 1 and 2 day holds, HCCF was purified over a
`lab-scale Protein A affinity column and the pools were frozen at <
`70C until analysis.
`
`HCCF Pool Incubation in Mini-Tanks and Protein A
`Processing
`
`HCCF was held in 50 mL 316-L stainless steel mini-tank containers
`(Flow Components, Dublin, CA) as described previously by Trexler-
`Schmidt et al. (2010). After 1 and 2 day holds, the HCCF was
`purified over a lab-scale Protein A affinity column as described by
`Trexler-Schmidt et al. (2010).
`
`Disulfide Bond Reduction Assay
`
`Microchip capillary electrophoresis (CE) (Agilent 2100 Bioanalyzer)
`was used to quantitate the level of non-reduced antibody as
`described previously by Trexler-Schmidt et al. (2010). The non-
`reduced, intact antibody migrates at 150 kDa.
`
`Mass Transfer Measurements
`
`Volumetric gas transfer coefficients (kLa) were determined for the
`sparge vessel conditions using a dynamic method. Sparge vessels
`were filled with 2-L of cell-free culture media and set to the desired
`agitation and sparge rates at ambient temperature (18–22C). N2 or
`air was added to the bioreactor through the sparger and/or
`headspace to drive the dO2 to ~50 or ~100%, respectively. dO2
`readings were measured with a probe and recorded using
`FermWorks software (Jova Solutions). Based on the equation
`ln (C*CL)¼ kLa tþ constant, data were plotted in a semi-log
`fashion with (C*CL) versus time. Linear regressions were
`
`Mun et al.: Air Sparging for Prevention of Antibody Reduction
`
`735
`
`Biotechnology and Bioengineering
`
`

`

`Case 1:18-cv-01363-CFC Document 79-8 Filed 03/22/19 Page 3 of 9 PageID #: 9525
`
`performed with the slope representing the kLa value for each
`condition.
`
`Cell Lysis Measurements (LDH Assay)
`
`Percent cell lysis was determined by measuring the level of lactate
`dehydrogenase (LDH) as described previously by Trexler-Schmidt
`et al. (2010).
`
`Monoclonal Antibody Protein Concentration Assays
`
`For HCCF samples, an HPLC-based Protein A method was used to
`determine rMAb concentration in order to calculate purification
`load density and yield. The rMAb concentration in the purified
`Protein A pool was measured using UV spectrometry at 280 nm and
`the extinction coefficient for that particular rMAb.
`
`IEC-HPLC Assay
`
`Ion-exchange chromatography (IEC) was used to assess charge
`heterogeneity. Samples were treated with carboxypeptidase B to
`remove C-terminal
`lysine before analysis by cation-exchange
`chromatography using
`a Dionex ProPac WCX-10 column
`(4  250 mm). The mobile phase used in the separation consisted
`of a potassium phosphate/potassium chloride pH 6.9 buffer.
`
`Perturbation Oxygen Uptake Rate (OUR) Measurements
`
`HCCF pools were incubated in 3-L glass stirred tank bioreactors
`with dO2 controlled at 30%. At specified intervals throughout the
`hold period (e.g., every 2–4 h), dO2 control and air sparging were
`turned off and the rate of change in dO2 was calculated over a 5%
`drop. Agitation remained at set-point throughout the hold period
`and during the perturbation.
`
`Results
`
`Establishment of a Small-Scale Sparge Model
`
`An initial experiment was performed to investigate hold conditions
`for the small-scale sparge model. Lysed HCCF (85% lysis) was held
`at a 2-L working volume in 3-L bioreactors at two conditions:
`50 rpm agitation with 50 sccm N2 overlay and 100 rpm agitation
`with no N2 overlay. The rMAb reduction levels in these two hold
`conditions were measured using the Agilent Bioanalyzer assay and
`compared to the control mini-tank hold. As shown in Figure 1,
`rMAb reduction was observed in the mini-tank control after a two-
`day hold time. In contrast, no rMAb reduction occurred with the
`100 rpm agitation and no N2 overlay conditions and reduction was
`accelerated with the 50 rpm agitation and N2 overlay conditions,
`with reduction occurring after a one-day hold time. In addition, a
`significant difference in dO2 levels was observed between the two
`sparge vessel hold conditions due to surface transfer from the
`headspace. The presence of a N2 overlay led to removal of O2 from
`the liquid and helped drive the dO2 levels to 0%. While 0% dO2 was
`maintained for the entire hold period, re-formation of the antibody
`disulfide bonds was observed once the reduction activity (i.e.,
`
`736
`
`Biotechnology and Bioengineering, Vol. 112, No. 4, April, 2015
`
`Figure 1.
`Lysed HCCF (85% total lysis; 65% mechanical lysis) was held in small-
`scale sparge vessels at room temperature (18–22C) at two different hold conditions.
`Levels of rMAb reduction and dO2 were compared between conditions as well as to the
`control mini-tank hold.
`
`intracellular components) was depleted after the two-day hold
`period. Re-formation of the disulfide bonds could be the result of
`the antibody attaining a more stable conformation of the lowest
`energy for its given environment. However, other as yet to be
`defined mechanisms may also be at work here. In the case without a
`N2 overlay, dO2 levels increased over time due to surface transfer
`from air in the headspace, which was also facilitated by the higher
`agitation rate. Based on these results, 50 rpm was selected as the
`agitation rate for future studies and a N2 overlay was employed for
`most sparge experiments, unless otherwise specified, since it
`represented a worst case environment for antibody reduction. As an
`exception, no N2 overlay was used for non-sparged hold cases where
`dO2 trends were being observed. Surface transfer is expected to be
`less significant in a manufacturing-scale tank due to a smaller
`surface area-to-volume ratio.
`Gas transfer in bioreactors occurs through two interfaces, one
`between the liquid and sparged bubbles and the other between the
`liquid surface and tank headspace. To characterize the small-scale
`sparge model, kLa values were measured at varying agitation and
`sparge rates through both the sparger and headspace. As shown in
`Table I, the sparger kLa was 2.5-fold higher at a sparge rate of
`50 sccm versus 10 sccm. In addition, the lower agitation rate
`decreased the headspace kLa slightly but did not significantly
`impact gas transfer through the sparger. These results further
`confirm the selection of a 50 rpm agitation rate in order to minimize
`the surface transfer yet provide adequate mixing. No difference was
`observed in the headspace kLa with a 50 sccm or 250 sccm overlay,
`so 50 sccm was chosen for future studies. Overall, the headspace kLa
`values were relatively low compared to the sparger kLa values within
`the range of sparge rates discussed in this work. This suggests that
`sparge rates were the main driver in the observed results and that
`headspace transfer played a relatively minor role.
`In order to optimize the sparge vessel conditions to adequately
`maintain a dO2 set-point of 30%, a study was performed to compare
`a 10 or 50 sccm maximum air sparge rate. Lysed HCCF (85% lysis)
`was held in sparge vessels with an agitation rate of 50 rpm and a
`
`

`

`Case 1:18-cv-01363-CFC Document 79-8 Filed 03/22/19 Page 4 of 9 PageID #: 9526
`
`Table I. Sparger and headspace kLa values in the small-scale sparge
`model at varying agitation and sparge rates.
`
`Agitation
`(rpm)
`
`Sparge rate
`(sccm)
`
`Sparger kLa
`(hr1)
`
`Headspace kLa
`(hr1)
`
`50
`
`100
`
`10
`50
`250
`10
`50
`250
`
`0.28
`0.75
`—
`—
`0.68
`—
`
`—
`0.12
`0.13
`—
`0.15
`0.16
`
`50 sccm N2 overlay and dO2 trends were monitored for each
`condition. As shown in Figure 2, a maximum sparge rate of 10 sccm
`provided insufficient gas transfer to maintain the dO2 set-point
`early in the hold, while a maximum sparge rate of 50 sccm provided
`immediate and reliable control at the dO2 set-point. Therefore, a
`maximum sparge rate of 50 sccm, which corresponds to a kLa of
`0.75 hr1, was chosen for future studies.
`
`Profile of Lysed HCCF in the Sparge Model
`
`To characterize the behavior of lysed HCCF, material at several lysis
`levels was held in small-scale sparge vessels with 50 rpm agitation,
`no air sparge, and no N2 overlay. Correlations were observed
`between cell lysis levels, the amount of oxygen consumed, and
`rMAb reduction for several CHO production cell lines. Figure 3A
`shows that an increase in cell lysis levels resulted in an increase in
`oxygen consumption. For the non-lysed (10% lysis) and 20% lysis
`cases, oxygen consumption was minimal and the levels of dO2
`remained relatively high. Conversely, at higher total cell lysis levels
`of 45 and 65%, the oxygen demand was high and resulted in the
`depletion of dO2. These trends suggest that HCCF dO2 levels could
`be used as an indicator for lysis levels and the potential for rMAb
`reduction at manufacturing-scale. However, absolute values from
`the small-scale model cannot be applied to manufacturing-scale
`due to differences in behavior and operations between scales (e.g.,
`
`Figure 2. dO2 trends for HCCF (85% total lysis; 65% mechanical lysis) controlled at
`30% dO2 set-point with two maximum sparge rates in the small-scale sparge vessel.
`
`Figure 3. Results showing the correlation between cell
`lysis levels and (A)
`Dissolved oxygen levels; (B) rMAb reduction. Lysis values listed reflect total lysis levels
`(‘‘non-lysed’’ (10%)þ mechanical lysis). In this study, the CCF was harvested at 90%
`viability, which was confirmed by the LDH assay and represented 10% lysis. Data were
`collected over a two-day hold in the sparge vessels.
`
`surface transfer, overlay pressure, hydrostatic pressure). The
`maximum oxygen consumption rate observed in the small-scale
`model was approximately 60%/h for the same cell line tested in
`Figure 3A, using a perturbation oxygen uptake rate (OUR)
`algorithm at 100% total
`lysis (data not shown). Concurrent
`implementation of other reduction mitigation strategies, such as
`lower temperature or chemical inhibitors, has also been shown to
`impact dO2 levels (data not shown).
`In addition to dO2 levels, cell lysis levels were also found to
`impact the level of rMAb reduction, which is consistent with
`previous findings (Trexler-Schmidt et al., 2010). Different threshold
`lysis levels are required for rMAb reduction to occur in different cell
`lines (data not shown). In Figure 3B, the results after a one-day hold
`show that intact rMAb levels were lower at the higher lysis levels.
`Higher levels of intact rMAb were observed after the two-day hold
`presumably due to a decrease in sample reduction activity following
`the depletion of enzyme pathway reducing agents after an extended
`hold time, which allowed the rMAb to re-oxidize to its lowest energy
`state. Results from previous experiments also support that the
`maximum reduction activity is observed within the first 24 h of
`incubation with active reducing components (data not shown).
`
`Mun et al.: Air Sparging for Prevention of Antibody Reduction
`
`737
`
`Biotechnology and Bioengineering
`
`

`

`Case 1:18-cv-01363-CFC Document 79-8 Filed 03/22/19 Page 5 of 9 PageID #: 9527
`
`While high lysis conditions result in both oxygen depletion (Fig. 3A)
`and rMAb reduction (Fig. 3B), oxygen depletion alone is not
`sufficient to trigger reduction. Material with low reduction activity
`was sparged with N2 to drive down the oxygen levels to 0% and no
`rMAb reduction was observed (data not shown). Therefore, a
`combination of sufficient reduction activity and low dO2 levels are
`required for the reduction event to occur.
`
`Impact of Air Sparging Strategies on rMAb Reduction
`Inhibition
`
`Two potential air sparging strategies were explored: constant sparge
`rates with variable dO2 levels and constant dO2 levels with variable
`sparge rates. In order to determine the minimum level of air
`sparging required to inhibit rMAb reduction, a range of air flow
`rates of 5–50 sccm and a range of dO2 levels of 10–50% were tested
`with lysed HCCF (65–85% lysis) that was shown to be susceptible to
`disulfide bond reduction in the non-sparged control vessel. The dO2
`control was one-sided with air supplied up to a maximum sparge
`rate of 50 sccm to maintain the set-point. All cases had a N2 overlay
`except for the 50 sccm constant sparge case. Figure 4A shows the
`levels of intact rMAb over a two-day hold period for the various
`sparge strategies tested. Slight reduction was observed for the
`5 sccm constant sparge case after a two-day hold time, however, the
`degree of reduction was decreased compared to the non-sparged
`control. All other sparge rates and dO2 set-points were found to be
`sufficient to prevent disulfide bond reduction.
`As shown in Figure 4B, the dO2 levels for the non-sparged control
`case and the 5 sccm sparge case were depleted within one day. This
`indicates that 5 sccm air sparge was unable to provide sufficient
`oxygen supply to maintain the dO2 level above 0% in competition
`with the rate of oxygen consumption by the lysed HCCF. All other
`sparge cases maintained dO2 levels >0% since the sparge rates met
`or exceeded the oxygen consumption rates. Climbing dO2 levels
`were observed for the 50 sccm sparge case as the oxygen delivery
`rate exceeded the consumption rate.
`As shown in Figure 4C, the HCCF pH increased by up to 1.3 units
`over two days as a result of CO2 stripping. The rate of pH increase
`was accelerated at the higher air sparge rates or higher dO2 set-
`points due to increased CO2 mass transfer. Lower rates of CO2
`stripping are expected in a manufacturing-scale HCCF tank because
`the lower liquid surface-to-volume ratio reduces CO2 mass transfer
`(Matsunaga et al., 2009).
`When choosing an appropriate HCCF air sparging strategy, it is
`desirable to minimize the amount of air sparge delivered to decrease
`CO2 stripping and the subsequent increase in pH and potential
`impact on product quality. The use of a minimum dO2 set-point
`provides feedback control and scales the air output appropriately to
`meet demand. This ensures that adequate air is supplied to
`maintain dO2 levels >0% and prevent antibody reduction, while
`also preventing the addition of excess air sparge. Therefore, a sparge
`strategy of dO2 control was selected as the preferred antibody
`reduction mitigation strategy for further evaluation. A 30%
`minimum dO2 set-point was chosen because it is commonly used in
`mammalian cell culture production bioreactors and has been shown
`to be robust and achievable, leaving a sufficient safety factor on the
`lower end, while not over-sparging.
`
`738
`
`Biotechnology and Bioengineering, Vol. 112, No. 4, April, 2015
`
`Figure 4. HCCF with 65–85% total lysis (40–55% mechanical lysis) was tested with
`multiple air sparge strategies. A N2 overlay was used in all cases except for the
`50 sccm constant sparge case. Results show trends of (A) rMAb reduction, (B)
`Dissolved oxygen, (C) pH change.
`
`Air Sparge Requirement Per Lysed Cell
`
`Figure 5A,B shows an assessment of the amount of air that is
`required to maintain a 30% dO2 set-point for a single cell line at
`three lysis levels. As seen in Figure 3A, the amount of oxygen
`consumed is dependent upon the level of cell
`lysis. In this
`experiment, lysed HCCF was held in sparge vessels with or without
`a N2 overlay to assess extreme conditions with respect to additional
`
`

`

`Case 1:18-cv-01363-CFC Document 79-8 Filed 03/22/19 Page 6 of 9 PageID #: 9528
`
`oxygen transfer from the liquid surface. These conditions bracket
`the expected surface oxygen transfer rate of a manufacturing-scale
`HCCF tank. As predicted, larger volumes of air were required to
`maintain the 30% dO2 set-point for cases with higher levels of cell
`lysis (Fig. 5A). In the presence of a N2 overlay, additional air sparge
`was required to maintain the dO2 set-point. Most of the air sparge
`was required during the first 24 h of the hold time, corresponding to
`the period of maximum reduction activity.
`The air sparge requirement through a 24-hour hold is plotted in
`Figure 5B versus the number of mechanically lysed viable cells. For
`this cell line, the sparge requirement per lysed cell in the small-scale
`model is 3.3 109 L air/lysed cell or 6.9 1010 L O2/lysed cell.
`This value could be used to determine oxygen demand for a
`particular lysis level or gauge relative reduction activity between cell
`lines or culture conditions. Additionally, since a linear correlation is
`observed between required air sparge and the number of lysed cells,
`these data provide confirmation that
`the oxygen-consuming
`
`compounds in the HCCF are being released from the cells upon
`lysis. An alternative approach to defining a dissolved oxygen control
`strategy at manufacturing-scale would be to measure the oxygen
`uptake rate of the cell lysate in order to estimate the required mass
`transfer coefficient in a HCCF vessel.
`
`Air Sparging Effectiveness for Multiple Cell Lines
`
`Since air sparging was found to be a robust and effective method in
`preventing rMAb reduction for one product,
`three additional
`products produced in CHO cell lines were tested in the small-scale
`model and the results are summarized in Figure 6. For all four
`products, air sparging to maintain a minimum 30% dO2 level was
`able to fully prevent reduction and maintain >90% intact rMAb.
`These results show that air sparging can be used as a mitigation
`strategy to prevent rMAb reduction for multiple cell lines with
`varying levels of reduction activity.
`
`Air Sparging Impact on Product Quality
`
`In the manufacture of rMAbs, appropriate process controls are
`important to ensure consistent product quality. Protein product
`quality can be sensitive to variations in cell culture process
`parameters (e.g., temperature, pH, or dO2) and changes could result
`in a shift in product attributes such as glycan distribution, aggregate
`formation, or charge variant profile (Andersen and Goochee, 1994;
`Cromwell et al., 2006; Yoon et al., 2004). Additionally, product
`quality can be impacted by cell-free hold conditions during recovery
`operations or in the final formulation buffer (Cromwell et al., 2006;
`Peters and Trout, 2006; Wang, 1999; Wakankar and Borchardt,
`2006). Implementation of air sparging during the HCCF hold step
`has the potential to impact product quality. Therefore, as with any
`
`Figure 5.
`Totalized volume of air sparge required to maintain a 30% dO2 set-point
`for varying lysis levels in the sparge model with and without a N2 overlay. Lysis values
`listed reflect total lysis levels (‘‘non-lysed’’ (10%)þ mechanical lysis). (A) Plotted with
`time on the x-axis; (B) Totalized air sparge over 24 h on the y-axis and number of
`mechanically lysed cells on the x-axis. The slope of each line represents the volume of
`air sparge required over 24 h per mechanically lysed cell to maintain a 30% dO2 set-
`point.
`
`Figure 6. Results showing the effectiveness of air sparging on preventing
`reduction with four different rMAb products. The dashed bars indicate the no sparge
`condition and the solid bars indicate the sparge condition (30% dO2 set-point with up to
`a maximum 50 sccm sparge rate). Products were tested at a total lysis level conducive
`to reduction (65–85% total lysis; 55–65% mechanical lysis). The HCCF was sampled
`from the sparge vessels and purified over Protein A after 0, 1, and 2 days.
`
`Mun et al.: Air Sparging for Prevention of Antibody Reduction
`
`739
`
`Biotechnology and Bioengineering
`
`

`

`Case 1:18-cv-01363-CFC Document 79-8 Filed 03/22/19 Page 7 of 9 PageID #: 9529
`
`other change to a process, the appropriate product quality analyses
`should be performed to ensure minimal impact.
`Sparged material purified by Protein A was evaluated using SEC-
`HPLC, IEC-HPLC, CE-glycan, and peptide map to evaluate any
`changes to aggregate levels, charge variant distribution, glycan
`distribution, and oxidation, respectively. Based on these analyses,
`air sparging was found to have no significant impact with respect to
`aggregate, glycan, or oxidation profiles (data not shown). However,
`a decrease in the levels of main peak by IEC-HPLC was observed
`with a concurrent
`increase in the levels of acidic variants.
`Experiments were performed to identify the cause of these acidic
`variants by specifically evaluating the effects of sparging and cell
`lysis. Figure 7 displays the effect of 0–50 sccm air sparging on main
`peak levels for HCCF material with a low lysis percentage of 20%
`total
`lysis (10% mechanical
`lysis), which is expected to be
`representative of the lysis levels of manufacturing-generated HCCF.
`Lower main peak was observed with increasing sparge rate. Overall,
`the non-sparged control material had a drop of 1% main peak per
`day while the sparged cases showed an additional 0.5–1.5%
`decrease per day. Previous studies observed a rate of ~2% decrease
`per day which was not considered significant (Trexler-Schmidt
`et al., 2010).
`Next, the effect of cell lysis on the charged variant profile was
`examined. In this experiment, HCCF at different
`lysis levels,
`including a non-lysed control, was sparged at a constant rate to
`ensure that the product in each case had equivalent air exposure. A
`high sparge rate of 50 sccm was selected to assess worst case. The
`results shown in Figure 8 demonstrate that cell lysis contributes to
`the decrease in main peak with as much as a 2% additional decrease
`per day over the non-lysed control for the 100% lysis case. Taken
`together, the results from Figures 7 and 8 indicate that the highest
`levels of acidic variant increase (main peak decrease) would be
`expected for material with both high lysis and high sparge rates.
`
`Figure 7. Non-lysed þ 10% mechanical lysis (20% total lysis) material was tested
`at different levels of constant air sparge (no N2 overlay) to assess the impact of air
`sparging on IEC-HPLC main peak. The HCCF was sampled from the sparge vessels and
`purified using Protein A after 1 and 2 days and submitted for IEC-HPLC analysis. The
`data were compared to non-sparged hold conditions.
`
`740
`
`Biotechnology and Bioengineering, Vol. 112, No. 4, April, 2015
`
`Figure 8. Material at different lysis percentages was sparged at a constant
`sparge rate of 50 sccm (no N2 overlay) to assess the impact of cell lysis on IEC-HPLC
`main peak. Lysis values listed reflect total lysis levels (‘‘non-lysed’’ (10%) þ mechanical
`lysis). The HCCF was sampled from the sparge vessels and purified over Protein A after
`1 and 2 days and submitted for IEC-HPLC analysis. The data were compared to a non-
`lysed control.
`
`Additional studies to help understand the source of the acidic
`variant formation are ongoing. One factor that has been explored is
`the possibility that the pH increase during air sparging, due to CO2
`stripping from the HCCF in the small-scale sparge vessel, leads to
`accelerated deamidation and therefore an increase in acidic variants
`(Patel and Borchardt, 1990; Peters and Trout, 2006).
`
`Discussion
`
`Air sparging has been shown to be a robust and universal mitigation
`strategy to prevent rMAb reduction. While the complete mechanism
`of rMAb reduction prevention has not yet been fully elucidated, the
`consumption of oxygen by cellular metabolic pathways results in the
`exhaustion of nutrient and energy supplies. For example, oxygen
`may convert NADPH, the proposed primary fuel for the reducti