Impact of Aeration Strategy on CHO Cell Performance During Antibody Production

M. Lourdes Velez-Suberbie, Richard D. R. Tarrant, and Andrew S. Tait

The Advanced Centre of Biochemical Engineering, Dept. of Biochemical Engineering, University College London, Torrington Place, London, WC1E 7JE, U.K.

Daniel I. R. Spencer
Ludger Ltd, Culham Science Centre, Abingdon, Oxfordshire, OX14 3EB, U.K.

Daniel G. Bracewell
The Advanced Centre of Biochemical Engineering, Dept. of Biochemical Engineering, University College London, Torrington Place, London, WC1E 7JE, U.K. DOI 10.1002/btpr.1647 Published online November 17, 2012 in Wiley Online Library (wileyonlinelibrary.com).

Stirred tank bioreactors using suspension adapted mammalian cells are typically used for the production of complex therapeutic proteins. The hydrodynamic conditions experienced by cells within this environment have been shown to directly impact growth, productivity, and product quality and therefore an improved understanding of the cellular response is critical. Here we investigate the sub-lethal effects of different aeration strategies on Chinese hamster ovary cells during monoclonal antibody production. Two gas delivery systems were employed to study the presence and absence of the airliquid interface: bubbled direct gas sparging and a non-bubbled diffusive silicone membrane system. Additionally, the effect of higher gas ow rate in the sparged bioreactor was examined. Both aeration systems were run using chemically dened media with and without the shear protectant Pluronic F-68 (PF-68). Cells were unable to grow with direct gas sparging without PF-68; however, when a silicone membrane aeration system was implemented growth was comparable to the sparged bioreactor with PF-68, indicating the necessity of shear protectants in the presence of bubbles. The cultures exposed to increased hydrodynamic stress were shown by ow cytometry to have decreased F-actin intensity within the cytoskeleton and enter apoptosis earlier. This indicates that these conditions elicit a sub-lethal physiological change in cells that would not be detected by the at-line assays which are normally implemented during cell culture. These physiological changes only result in a difference in continuous centrifugation performance under high ow rate conditions. Product quality was more strongly affected C 2012 American Institute of by culture age than the hydrodynamic conditions tested. V Chemical Engineers Biotechnol. Prog., 29: 116126, 2013. Keywords: Chinese hamster ovary cells, monoclonal antibody, aeration, F-actin, glycosylation

Introduction
Over the past 30 years there has been a signicant increase in the large-scale manufacture of therapeutic proteins produced in mammalian cells largely due to demand for monoclonal antibody (mAb) therapeutics. This has been accomplished by the application of cell lines such as Chinese hamster ovary cells (CHO) which provide the appropriate post-translational modications necessary for therapeutic efcacy.1 As well as this they are also associated with good expression levels2 and have been adapted for robust growth in suspension up to 25,000 L in stirred tank reactors (STR).3 In this environment they are exposed to various hydrodynamic forces as a result of agitation and aeration.46 The
Correspondence concerning this article should be addressed to D. G. Bracewell at d.bracewell@ucl.ac.uk.
116

magnitude and exposure time to these forces is therefore critical to bioreactor design.7 Mechanical agitation can cause cell damage by cellcell collision, cell collision with the vessel walls or stationary surfaces in the vessel, cell collision with the impeller, or by the turbulence experienced within the media.4,8,9 Across the vessel the hydrodynamic forces vary. For example, 70% of the total energy of dissipation is located in the area of the impeller (approximately 10% of the volume vessel).7 In order to reduce the damage caused by the mechanical agitation in mammalian cell culture low power inputs and pitched-blade impellers (usually 45 ) are commonly used.5 As it has been shown previously that bubbles have a detrimental effect on mammalian cells,9,10 the effects of direct sparging in CHO cell culture must be considered. The events that can be responsible for cell damage in sparged
C 2012 American Institute of Chemical Engineers V

Biotechnol. Prog., 2013, Vol. 29, No. 1

117

bioreactors are the formation of bubbles at the sparger, bubble detachment, bubble rise through the media, bubble coalescence, and break-up inside the culture broth or the events at the airliquid media interface (e.g., foaming).8 However, it has been shown that bubble burst at the airliquid interface is the main cause of damage in cells grown in suspension.911 The damage associated with bubbles in bioreactors can be reduced by the addition of antifoam and shear protectant agents to the culture medium.9,1214 In the presence of antifoam and shear protectant agents, such as Pluronic F-68, cells are less likely to be attached to bubbles and therefore be trapped in the liquid surrounding the bubbles.15,16 Besides the damage associated with bubble rupture and foaming CO2 removal is another concern in aerated bioreactors.6 The rate at which CO2 can be removed is dependent on the aeration system. In the case of direct sparged systems CO2 accumulation is not a big concern as the CO2 and O2 mass transfer coefcients are very similar; however when a silicone membrane aeration system is used the CO2 mass transfer coefcient is lower than the O2, causing a reduction in CO2 removal. The accumulation of CO2 in the culture medium can affect cell growth, productivity, and product quality.17,18 The shear sensitivity of cells is determined by many factors such as cell type and size, and growth environment.5,8 The effect of shear can be lethal (apoptosis and necrosis) or sub-lethal (changes in cell growth, cell viability, oxygen uptake rate, and alterations in cell morphology).5,19 Apoptosis is an aspect of cell physiology that has received signicant attention, because it can be a factor that limits productivity in mammalian cell culture. Apoptosis is a regulated physiological process in which the cells respond to stress stimuli, activating a cascade of events resulting in cell death.8,13 The cytoskeleton, which determines cell structure, is dynamic and has the capability of reorganizing itself as the cell shape changes or undergoes mitosis. It also responds to environmental changes and is responsible for maintaining the cell integrity during repeated deformations.20,21 The cytoskeleton is formed by different regulatory proteins and polymers including actin laments, microtubules, and intermediate laments.22 Actin can be present in two forms: as a monomer (G-actin) or a lament (F-actin). Actin laments are continuously assembled and disassembled in response to environmental signaling and stress. To respond quickly cells maintain a pool of G-actin which can be polymerized into F-actin when required.22,23 To understand how the bioreactor environment inuences the cytoskeleton we used a F-actin staining to follow these changes. As well as the cell line selected, it is the time course, media composition, and choice of operating mode (whether batch, fed-batch or continuous) which inuences cell density and condition during production.5 These strategies may put cells under considerable stress and result in low viabilities at time of harvest and have been shown to directly impact on the downstream process, e.g. solids removal during primary recovery.24 They also impact on the glycosylation prole which is critical for many therapeutic proteins including mAbs, playing vital role in biological activity, efcacy, and stability.2527 As maintaining a consistent glycosylation prole is essential in mAb manufacture the effect of bioreactor environment on the glycoform is included as a key part of this study. Silicone membrane systems have been used to provide aeration into the culture medium;16,28 however no direct comparison to direct gas sparged systems has been done before. The objective of this work is to characterize the sub-lethal effects

that the bioreactor environment has upon CHO cell physiology and product quality during batch and fed-batch culture. To achieve this, the effect of aeration strategy on cell and product quality was studied. Direct gas sparged and silicone membrane aeration systems were used to compare a bubbled and bubble-free environment with and without shear protectant (Pluronic F-68). In addition, the effect that aeration conditions might have on downstream processing (DSP) was studied using ultra scale-down (USD) techniques.24

Materials and Methods

Cell culture and media formulation Experiments were carried using a Chinese hamster ovary cell line (GS-CY01, generously provided by Lonza Biologics, Slough, UK), expressing an IgG4 mAb. Cells were cultured in 250 mL disposable shake asks with vent caps (Corning, NY) with a working volume of 100 mL. Cells were incubated (Sanyo, Loughborough, UK) on a shaker at 150 rpm (IKA 260, Wolf Laboratories, York, UK) at 37 C and 5% CO2. Cells were subcultured every 3 or 4 days (up to 25 passages) until required for inoculation in a chemical dened medium (CD CHO medium, Life Technologies, Paisley, UK), and using 25 lM methionine sulfoximine (MSX) (Sigma-Aldrich, Gillingham, UK) to maintain the selection of the recombinant gene. Stirred tank reactor cultures Experiments were carried out in two 5 L bioreactors (Biostat B-DCU control unit, Sartorius, Epsom, UK), running in parallel with a working volume of 3.5 L. Agitation was provided by a three-blade segment (in a 45 angle) single impeller, rotating at 260 rpm. Temperature was controlled at 37 C. pH was controlled at 7.1 0.1 by sparging CO2 and a sodium carbonate buffer (100 mM Na2CO3, 100 mM NaHCO3). Dissolved oxygen tension (DOT) was maintained by gas blending at 30% 1 by sparging air, oxygen, or nitrogen using a horse shoe gas sparger (bubble) or a silicone membrane aeration system (bubble free). The vessel was sterilized at 121 C for 20 min, with 2.5 L of water, which was removed aseptically after sterilization. Cells were seeded at a density 2 105 cells mL1. A chemical dened medium (CD CHO, Life Technologies) with and without a shear protectant agent (Pluronic F-68, Life Technologies) was used. The medium was supplemented with a 1% (wt/ vol) solution of antifoam C emulsion (Sigma Aldrich). Glucose concentration was monitored daily using a NOVA bioanalyzer 400 (Nova Biomedical, Deeside, UK). For the fedbatch cultures the reactor was fed once a day to maintain a concentration of 2 g L1 of glucose, using 10 times concentrated CD CHO medium with glucose added to a concentration of 150 g L1. Aeration systems Direct Gas Sparging. Aeration was provided using a tube (8 mm outer diameter) with a horse shoe sparger which has 14 drilled holes (1 mm diameter). The sparger was located below the impeller and reactors were operated in down-pumping mode at a different gas ow rates; 500 mL min1 (0.14 vvm) or 100 mL min1(0.03 vvm). For high gas ow rate, air or nitrogen was sparged whilst for the low ow rate oxygen or nitrogen was used.

118

Biotechnol. Prog., 2013, Vol. 29, No. 1

Figure 1. Silicone membrane aeration system and stainless steel rack for the 5 L STR bioreactor.

Silicone Membrane Aeration System. A 15 m silicone tube (SILASTIC Rx-50, medical grade, Dow Corning, Midland, USA), with an outside diameter 3.18 mm and a wall thickness of 0.60 mm was wrapped around a stainless steel cage as shown in Figure 1.28,29 Gas ow rate through the tube was maintained at 100 mL min1. Analytical techniques Viable Cell Density and Cell Size. Viable cell density and cell size were monitored daily. Cell density and viability were determined using a Vi-Cell XR (Beckman Coulter, High Wycombe, UK) based on the exclusion of trypan blue. Cell diameter was determined using a CASY analyzer (Innovatis, Bielefeld, Germany). 2050 lL of culture supernatant was diluted in 10 mL of Casyton buffer (Innovatis, Bielefeld, Germany) and then analyzed. The CASY was used with a 150 lm orice, and set to measure up to 40 lm with a ve times repeat measurement, from which an average was reported. Product Concentration by high performance liquid chromatography (HPLC) Analysis. The mAb concentration was determined by HPLC (Agilent Technologies, South Queensferry, UK). 100 lL of each sample was loaded on to a 1 mL HiTrap Protein G HP column (GE Healthcare, Buckingham-

shire, UK). The column was washed with sodium phosphate buffer (20 mM, adjusted to pH 7.0) and eluted with glycine buffer (20 mM, adjusted to pH 2.8). The mAb concentration was determined by integrating the elution peak at 280 nm and using a standard curve of puried mAb. Product Purication, N-Glycan Release, Labelling, Clean-Up and Analysis. Supernatant samples were taken on the harvest day and puried using 40 lL PhyTyps (PhyNexus, CA), packed with MabSelect SuRe resin (GE Healthcare, Pittsburgh, PA). The column was equilibrated with 10 column volumes (CVs) 20 mM sodium phosphate/150 mM sodium chloride, pH 7.2 and was then loaded to a capacity of 20 lg lL1. The column was then washed with 10 CVs of 20 mM sodium phosphate/150 mM sodium chloride, pH 7.2, and eluted into 1 mL using 0.1 M sodium citrate, pH 3.50. The elute samples were immediately adjusted to pH 7.0 with 1.0 M Tris-HCl, pH 9.0, and were immediately frozen. Triplicates of each antibody sample were taken through an in-solution PNGase-F release protocol according to the manufacturers and instructions (QA-Bio, Parm Dessert CA). Briey, 100 lL of each protein A puried sample, containing approximately 100 lg of protein, was dried down in a Thermo Savant centrifugal evaporator (Thermo Ltd, Hampshire, UK) and re-suspended in 35 lL Milli-Q water (Millipore, Oxfordshire, UK). 10 lL of 250 mM sodium phosphate pH 7.5 and 2.5 lL of a denaturation solution containing 2% sodium dodecyl sulfate and 1 M b-mercaptoethanol were added to the antibody mix, and the sample heated in a closed Eppendorf tube at 100 C for 5 min. The samples were cooled and then 2.5 lL of 15% Triton-X100 was added, mixed using a vortexer followed by 2 lL of PNGaseF solution in 20 mM Tris-HCl, pH 7.5 (0.01 U of enzyme, according to manufacturer). Samples were incubated at 37 C for 3 hours in order to release the glycans. Glycans were enriched and puried using Ludger clean EB10 devices (Ludger Ltd, Oxfordshire, UK) according to manufacturers instructions. The devices were equilibrated with 0.1% triuoroacetic acid (TFA). The PNGaseF incubated antibody samples were bound onto the EB10 columns, washed with 0.1% TFA and glycans eluted with 50% acetonitrile in 0.1% TFA. Samples were dried in the Thermo Savant centrifugal evaporator. Fluorophore labeling of the released glycans was performed with a LudgerTag 2-aminobenzamide (2-AB) glycan labeling kit (Ludger) according to manufacturers instructions and as detailed in Bigge et al. (1995)30 Excess 2-AB reagent was removed from the samples using LudgerClean T1 cartridges on a vacuum manifold system (Ludger). Namely, each 2-AB labeled sample was diluted with 0.2 mL acetonitrile, added to the cartridge, and allowed to settle into the cartridge for 5 min. The cartridge was then washed with 3 1 mL of 96% acetonitrile, using the vacuum manifold to pull the wash through the cartridge. 2-AB labeled glycans were eluted from the cartridge using 0.5 mL Milli Q water under a slow vacuum (approximately 2 min elution time). Samples were dried using the Thermo Savant centrifugal evaporator and then re-suspended in 100 lL of 78% acetonitrile in water. 25 lL of each sample was analyzed using a Thermo U3000 Ultimate ultra high performance liquid chromatography (UHPLC) instrument (Thermo Ltd, Hampshire, UK) equipped with a U3000 uorescence detector and a Waters ethylene bridge hybrid (BEH) Glycan UHPLC column, 2.1 mm 150 mm (Waters, Hertfordshire, UK).

Biotechnol. Prog., 2013, Vol. 29, No. 1

Table 1. Experimental Set-Up and Key to Growth Proles in Figure 2 Direct gas sparging (100 mL min1) Pluronic F-68 Without With
  

119

Silicone membrane aeration system (100 mL min1)

  

(500 mL min1)
 

(Not shown) (Figures 2A,D) Batch (Figures 2C,F) Fed-batch

(Figures 2A,D) Batch (Figures 2C,F) Fed-batch

(Figures 2B,E) Batch (Figures 2B,E) Batch (Figures 2C,F) Fed-batch

Conditions tested. Conditions not tested because cells were not able to grow when direct gas sparging and Pluronic F-68 free medium were used.

Gradient conditions were similar to those stated in the manufacturers instructions. The UHPLC column was kept at 60 C. Solvents were A: 50 mM ammonium formate (Ludger) and B: acetonitrile (Romil, Cambridge, UK). A gradient was run at 05 min78% acetonitrile, 22% 50 mM ammonium formate; 538.5 min78% acetonitrile to 55.9% acetonitrile. Flow rate used was 0.5 mL min1. Fluorescence detector settings were as follows: excitation wavelength 330 nm, emission wavelength 420 nm, and sensitivity setting 8. Chromatogram data were analyzed using Chromeleon 6.8 software (Thermo, Hampshire, UK). Apoptosis Assay. Progression into apoptosis was determined from day 5 onwards using a commercially available Annexin V-FITC/7ADD kit PN IM3546 (Beckman Coulter, High Wycombe, UK) and Coulter Epics XL-MCL Flow Cytometer (Beckman Coulter). Data were collected and analyzed using EXPO 32 ADC XL Color software (Beckman Coulter). Cell culture samples were prepared following the manufacturers protocol. Samples were analyzed using 488 nm excitation and detected using 525 nm and 675 nm band-pass lters for Annexin V and 7-ADD (7-amino-actynomycin D) respectively, and collected for 300 s or 10,000 events. To verify instrument optical alignment and uidics, a ow check (Flow-Check Fluorospheres, Beckman Coulter) was carried out before every experiment. All samples were analyzed within 30 min of staining. Positive control samples were analyzed to ensure that appropriate gating was applied. F-Actin Staining for Flow Cytometry and Confocal Microscopy. The cytoskeleton staining of F-actin was carried out from day 5 onwards. Cells were uorescently stained with Alexa Fluor 488 phalloidin (Invitrogen, Paisley, UK) and analyzed using a Coulter Epics XL-MCL Flow Cytometer and Perkin Elmer Spinning Disk Confocal microscope driven by Volocity Acquisition (Perkin Elmer, Cambridge, UK). Cells were simultaneously xed, permeabilized, and stained. Cells were re-suspended in a solution of 3.7% formaldehyde with 0.2% of Tween 20 (Sigma-Aldrich) to a concentration of 1 1062 106 cells mL1. About 510 units of uorescent phallotoxins were added (2550 lL of methanolic stock solution) to label for F-actin. The stained solution of cells was incubated for 20 min at 4 C. The cell suspension was washed twice with phosphate buffered saline (PBS) and centrifuged for 5 min at 300g at 4 C, the supernatant was discarded and the cell pellet was re-suspended in PBS to a concentration of 1 106 to 2 106 cells mL1. For ow cytometry stained samples were analyzed using 488 nm excitation and detected using 525 nm band-pass lter. Samples were collected for 10,000 events or 300 s. For confocal microscopy, 300 lL of 300 nM DAPI suspension (40 ,6-diamidino-2-phenylindole, dihydrochloride) (Invitrogen) was added to the washed and stained cell suspension and incubated for 5 min at room temperature. About 10 lL of the stained samples were mounted on a slide. For long-term storage

of samples a drop of ProLong Gold antifade agent (Invitrogen) was added. The corners of the cover slip were sealed with nail polish. The samples were cured for 24 h at room temperature in the dark, and then the edges of the cover slip were completely sealed with nail polish and stored at 4 C. Slides were analyzed by spinning disk confocal microscope using a 450 nm laser for blue (DAPI) and a 510 nm for green (F-actin). The images were obtained by Volocity Acquisition software (PerkinElmer). Determination of kLa Oxygen mass transfer resistance (kLa) was determined using the gassing out method as described in Wise (1951).31 The optrode was calibrated to 100% air saturation and to 0% air saturation by sparging air and nitrogen, respectively. kLa was determined in a 5 L bioreactor lled with 3.5 L chemically dened culture medium supplemented with a 1% solution of antifoam C emulsion. Temperature was controlled at 37 C and nitrogen was sparged until the level of oxygen had fallen to zero. The gas supply was then changed to air and the increase in oxygen concentration was recorded over time. For direct gas sparged system, the kLa was determined at two gas ow rates (100 and 500 mL min1) and for the silicone membrane aeration system different gas ow rates were tested to obtain a matching kLa to the direct sparged system. USD shear and centrifugation studies A rotating shear device developed at University College London (UCL) was used to investigate the effect that shear has upon the break up of cells over the duration of the culture and the subsequent impact this has upon centrifugation performance. The shear typically experienced in the feed zone of industrial centrifuges was mimicked by application of different rotational speeds within the device (previously characterized using computational uid dynamics, Boychyn et al. (2004)32). The energy dissipation experienced using a hermetic feed zone (0.055 106 W kg1) can be equated to the disc-stack centrifuge used in this study. Cells taken at different stages during cell culture were exposed to shear at different levels (0.055, 0.18, 0.53, and 1.2 106 W kg1) for 20 s, to provide sufcient time for full break-down of the material as may occur in a centrifuge feed zone.33 Prediction of large-scale centrifugation was performed using an Eppendorf 5180R benchtop centrifuge (Eppendorf, Cambridge, UK) with an A-4-62 swing-out rotor and 96-well plate format using the correlation between laboratory-scale centrifugation (V/ c.t.R) and large-scale centrifugation (Q/c.R) which has been described previously.24 Briey, centrifugation was undertaken using ll volumes of 0.5, 0.9, or 1.35 mL and the plates were centrifuged at 3,000 rpm for 5 or 10 min to give a range of V/c.t.R values between 1.67 and 5.69 108 ms1. The supernatant from each well was removed and the OD600 prior to and after centrifugation used to determine the % clarication achieved.

120

Biotechnol. Prog., 2013, Vol. 29, No. 1

Figure 2. Growth and productivity proles from batch and fed-batch fermentations of mAb producing CHO cells (GS-CY01) grown using two aeration systems.
Cells were grown in 5 L stirred bioreactor (3.5 L working volume), using media with Pluronic F-68 (unless otherwise stated). Viable cell concentration () and % cell viability (--) were determined daily. A: Batch mode using direct sparging (DS). B: Batch mode using silicone membrane aeration system (SM). C: Fed-batch mode using DS (~,*) and SM (n) (average of three cultures (*)) n 3 (replicate cell counts) s.d. The product titer (--) was determined by protein G HPLC analysis. D: Batch mode using DS. E: Batch mode using SM. F: Fed-batch mode using DS (~,*) SM (n) (average of three cultures (*)) n 2 (replicate cell counts) s.d

Results and Discussion

It has been previously shown that direct gas sparging can have detrimental effects on mammalian cells.8,34,35 Here we investigate the sub-lethal physiological effects that aeration has on a commercially relevant cell line, as well as the potential impact of changes in the bioreactor operation on DSP and to product quality. In order to look into the effects that bubbles have during fermentation two aeration systems were used: direct gas sparging and a silicone membrane aeration system (Figure 1).13,28,29 The conditions tested are summarized in Table 1.

Impact of aeration system on cell growth and productivity The effect of aeration conditions on mammalian cells was studied using a base case scenario with a gas ow rate of 100 mL min1 (0.03 vvm) and compared to a culture run with a ve times greater ow rate (500 mL min1, 0.14

vvm). The high gas ow rate results in a greater number of bubbles within the reactor system at any one time and will increase the frequency that cells are exposed to the airliquid interface. The difference in ow rate will also result in a change to the kLa, which were determined to be 2.3 h1 and 5.6 h1 (using the gassing out method, with CD CHO medium containing antifoam) for the low and high gas ow rates, respectively. To ensure the oxygen levels seen by the cells in the bioreactors remained constant at both ow rates, the dissolved oxygen was maintained at 30% using gas blending, therefore eliminating the effect of the difference in kLa. The peak viable cell density was observed to be 15% higher in the culture operated with low gas ow rate (100 mL min1), indicating that cells cultured in the presence of fewer bubbles were able to grow more efciently (Figure 2A).9,34,35 However, there was a minor increase in the antibody productivity of the culture with an elevated gas ow rate (500 mL min1) (Figure 2D, Table 2) compared to the one at low gas ow rate

Table 2. Specic Productivity, Volume Specic Productivity, and Volumetric Productivity from CHO cells (GS-CY01) Grown Using Two Aeration Systems in Batch Mode Time (days) (100 mL min )
1

Direct Gas Sparging (100 mL min )

1

Silicone Membrane Aeration System (500 mL min ) 16.3 N/A 18.5 103 N/A 0.42 N/A
1

(100 mL min1) without Pluronic F-68 13.3 9.9 19.1 103 10.3 103 0.48 0.38

(100 mL min1) 7.6 6.3 12.4 103 8.01 103 0.27 0.31

Overall cell specic productivity (pg cell1 day1)* 08 12.1 7.3 010 6.2 N/A Overall volume cell specic productivity (pg lm3 d1) 10.5 103 08 10.8 103 010 7.48 103 N/A Volumetric productivity (g L1) 08 0.25 0.34 010 0.30 N/A

N/A: Bioreactors were harvested on day 8 * Determined from the linear slope of the plot of product concentration versus cIVC

Biotechnol. Prog., 2013, Vol. 29, No. 1

121

Figure 3. Apoptosis proles of mAb producing CHO cells (GS-CY01) grown in 5 L stirred bioreactors (3.5 L working volume).
Bioreactors were operated in batch mode using Pluronic F-68 containing media (unless otherwise stated). The viable cell population (), and late apoptotic population (---) were determined by ow cytometry. A: Gas ow rate 100 mL min1 (l and * two independent batches) and gas ow rate of 500 mL min1 (~). B: Gas ow rate 100 mL min1 in medium with Pluronic F-68 (n) or Pluronic F-68 free medium (n). Trends are shown by a third order linear regression. In all cases there was a constant early apoptotic population of 5%.

(100 mL min1), which is consistent with an increase in productivity in cells under environmental stress.3638 To avoid the airliquid interfaces and hydrodynamic forces created by bubbles aeration was delivered via a silicone membrane aeration system (Figure 1) at a gas ow rate of 100 mL min1 (0.03 vvm). The kLa of the system was 2.3 h1 obtained by the gassing out method in growth medium (CD CHO media with Pluronic F-68 containing antifoam). The silicone membrane aeration system proved to be an efcient aeration system to culture the cells, with the cell density, percentage viability (Figure 2B), product titer (Figure 2E), and cell specic productivity (PCD) (Table 2) being comparable to those obtained by direct gas sparging (100 mL min1). The effect of shear protectant on the cell physiology was determined utilizing both aeration systems with a chemical dened medium with and without Pluronic F-68.12 Both systems were operated at a gas ow rate of 100 mL min1, where the kLa had been determined to be 2.3 h1 by the static gassing out method in culture medium supplemented with antifoam. When direct gas sparging was used to provide aeration into the culture medium the cell viability decreased to 30% within 24 hours of inoculation (data not shown), showing that in absence of Pluronic F-68 and with direct gas sparging cells are not capable of growing in this environment.8,11,39,40 On the contrary, when aeration was provided using the silicone membrane aeration system in the absence of Pluronic F-68, the cell density and percentage viability obtained (Figure 2B) were similar to those obtained with direct gas sparging (100 mL min1) with Pluronic F-68 (Figure 2A). This indicates that protecting cells against the airliquid interface created by bubbles is critical and it is this effect which prevents growth under conditions of direct sparging without Pluronic F-68. In terms of productivity, the product titer (Figure 2E) and cell specic productivity (Table 2) were similar when the cells were grown with a shear protectant free medium. A decrease in product titer was observed on the last day of the culture (Figure 2E and Table 2), and may be caused by degradation or aggregation of the product in the bioreactor,26 which in turn could be inuenced by surfactants such as those used as shear protectants. In the experiment carried out without a shear protectant, this effect was accentuated, having a signicant impact on the overall PCD which

decreased approximately 25% between days 8 and 10. In order to maximize the productivity and avoid undesired changes in the product, such as to the glycosylation prole the length of the culture could be reduced.27 Fed-batch mode is widely used in industry for the production of therapeutic proteins as it increases the cell density and product titer.5 In order to increase the product titer and identify the effect that aeration has on the cell physiology and product quality fed-batch cultures were carried out and the following conditions were investigated: direct sparging at low (100 mL min1) or high (500 mL min1) gas ow rate and a silicone membrane aeration system (100 mL min1) (Table 1). The fed-batch cultures were extended from 10 to 14 days, showing consistent growth with the batch cultures prior to implementing the feeding strategy (Figure 2C). A greater cell density was achieved, the cell productivity increased (approximately 50% for the three conditions studied) and there was 2-fold increase in product titer (Figure 2F). In terms of cell growth (Figure 2C) and productivity (Figure 2F), the same trends as observed for the batch cultures (Figures 2A,B,D, and E) were observed. Impact of aeration system on CO2 and surface aeration The effect of the aeration system on CO2 stripping from the culture medium was considered as accumulation of CO2 can have detrimental effects on cells and product quality.18 In systems where pH is controlled accumulation of CO2 can lead to an increase in the osmolality of the medium.41 The medium osmolality was monitored throughout the culture, increasing with culture age. However at harvest point the medium osmolality was similar in all the conditions tested, demonstrating that the aeration system did not have an impact on the osmolality of the culture medium (data not shown). The maximum level of surface aeration was determined via head space aeration at 400 mL min1 and gave a kLa of 0.7 h1, which is not adequate to support the cells oxygenation requirements.18 Even in this optimized set-up with no foaming, this represents less than one-third of the poorest kLa of aeration system studied, but in reality would be considerably lower due to foaming. To further minimize any differences due to surface aeration, no head space aeration was provided and identical bioreactor geometries and volumes were used.

122

Biotechnol. Prog., 2013, Vol. 29, No. 1

Figure 4. F-actin proles obtained from ow cytometry of mAb producing CHO cells (GS-CY01) grown in 5 L STR, operated in batch mode using two aeration systems. F-actin was stained with Alexa Fluor 488 (phalloidin).

Impact of aeration system on cell physiology To characterize the effects that aeration has on the CHO cells physiology, the progression into apoptosis was determined from day 5 of the culture until the day of harvest. Both aeration systems presented a transition from a viable cell population to a late apoptotic population as the culture progressed (Figure 3) and a constant population of early apoptotic cells was present (data not shown).

In the case of direct gas sparging, the high gas ow rate (500 mL min1) resulted in a reduction in the viable cell population and an increase in late apoptotic population (Figure 3A) compared to low gas ow rate (100 mL min1), indicating that when cells are exposed to a greater number of bubbles they enter apoptosis earlier in the culture.8,9,34,35 When aeration was provided via the silicone membrane system the apoptosis proles (Figure 3B) obtained were

Biotechnol. Prog., 2013, Vol. 29, No. 1

123

Figure 5. Example of confocal images of mAb producing CHO cells (GS-CY01) grown in 5 L STR, operated in batch mode.
In green, F-actin stained with Alexa Fluor 488 (phalloidin) and in blue, the nucleus stained with DAPI (40 ,6-diamidino-2-phenylindole, dihydrochloride). Gas ow rate of 100 mL min1 (AB; EM) or 500 mL min1 (CD). Gas ow rate of 100 mL min1 in a Pluronic F-68 free medium (HJ) or medium with Pluronic F-68 (KM). The scale bar represents 10 lm.

similar to the ones when direct gas sparging was utilized. However in shear protectant free medium there was a decrease in the viable cell population at an earlier stage of the culture process. It has been hypothesized that Pluronic F-68 has different protection mechanism, one is that it lowers the plasma membrane uidity of the cells, which causes a decrease in shear sensitivity,11,42 and the other that Pluronic F-68 is internalized by the cells and its shear protection may be derived from the alteration and stiffening of the mechanical properties of the cells cytoplasm.43 When cells were subjected to increased hydrodynamic stress (greater presence of bubbles using shear protectant and

direct sparging) they had a faster progression into apoptosis as well as lower viable cell density, conrming what has been previously reported.8,34,35 To identify the effect aeration has on the cell structure, F-actin, which is one of the main cytoskeletal proteins, was uorescently stained from day 5 of the culture onwards and analyzed by ow cytometry and confocal microscopy. F-actin was stained with Alexa Fluor 488 phalloidin, and proles obtained from ow cytometry are shown in Figure 4. In all aeration conditions tested, there was a decrease in the intensity of F-actin as the culture progressed (the cell population shifted down in the density plot). When the bioreactor

124

Biotechnol. Prog., 2013, Vol. 29, No. 1

was aerated by direct sparging at a high ow rate (500 mL min1) the decrease in F-actin intensity was more accentuated compared to the one aerated at lower gas ow rate (100 mL min1) (Figure 4A). A decrease in F-actin intensity was also observed with the use of the silicone membrane system (Figure 4B). However, this effect occurred at a later stage in the culture when the cell viability decreased and a greater population of cells was late apoptotic. During apoptosis the cytoskeleton reorganizes itself and actin forms clusters of short laments,44,45 which could account for the loss of the F-actin intensity observed by ow cytometry. It has been suggested that when the cells are xed the blebs, which are protrusions of the plasma membrane formed when the cells undergo apoptosis,19 might separate into apoptotic bodies,46 possibly accounting for the observation in the density plots. A population with low F-actin intensity and small size was present on left bottom side of the plot, as the culture progressed and more cells became apoptotic as the size of this population increased suggesting that more apoptotic bodies were detached from the cells when xed for the staining. The uorescently stained cells were mounted onto slides and images were obtained using confocal laser scanning microscopy. The images are shown in Figure 5 where the Factin, shown in green, and the nucleus which was stained with DAPI is shown in blue. In the case of direct sparging, the cells were more symmetrically spherical (Figures 5AG) and this shape was maintained over the course of the culture. The actin laments were clearly visible and evenly distributed throughout the cells, although there were a few regions with higher actin concentration. When aeration was delivered with the silicone membrane system the actin laments were less dened as the culture progressed and the cells tended to have more irregularities such as peripheral indentations and protrusions (Figures 5HM). In addition, approximately 30% of the population experienced late apoptotic effects, changes in cell shape and F-actin aggregate formation.44,47 These may have contributed to the shape change of the lamentous structure as shown in Figures 5G, I, J, and M where F-actin was localized around the membrane and account for the prominent onset of blebbing in Figure 5M. For the fed-batch cultures progression into apoptosis and in F-actin intensity reduction was observed to be the same as the batch cultures. We can conclude that the shear protectant agent, higher gas ow rate in direct sparged bioreactors, apoptotic state of the cells, and culture length have an effect on the cytoskeleton, as seen by decreases in F-actin intensity and changes in cell shape.

Figure 6. Clarication performance of fed-batch CHO cultures operated using two aeration systems.
Samples taken from day 14 of fed-batch cultures operated with silicone membrane aeration system (*) and direct gas sparging with 500 mL min1 (~) were exposed to shear environments that represent those typically experienced in industrial scale centrifuges. The predicted clarication performance was determined by centrifugation of sheared and non-sheared samples in 96-well plates to give V/c.t.R values of 2.05 108 m s1 (--) and 4.15 108 m s1 (). Data represent three replicate centrifugation samples s.d.

Performance on primary recovery An ultra scale-down centrifugation mimic was used to determine whether the differences in the cell culture performance and cell physiology observed under different gassing regimes inuence how cells respond to the shear typically experienced in primary recovery. Under the aeration systems tested for the batch cultures no signicant difference in the clarication performance measured using USD centrifugation was observed for conditions that mimic hermetically and non-hermetically sealed feed zones (energy dissipation rates 0.055 106 to 1.22 106 W kg1 were examined) and centrifuge clarication characteristics (V/c.t.R, equivalent to Q/c.R) ranging from 1.67 to 5.69 108 m s1 (data not shown). Cells that were not subjected

to shear were also examined for reference. Cells taken from fed-batch culture and centrifuged at lower V/c.t.R (2.05 108 m s1) as expected showed improved clarication compared to higher V/c.t.R, but also showed no signicant difference (Figure 6, dashed lines) between the different aeration systems. However, if the V/c.t.R is increased (4.15 108 m s1), to increase the ow rate for greater throughput for example, a difference is observed (Figure 6, solid lines). Under these conditions the cell culture material taken from the silicone membrane aerated culture has poorer clarication performance compared with that from culture sparged at 500 mL min1 with up to 4% greater solids remaining observed. This difference is observed under all the shear conditions tested, but not where no shear is applied [(1.58 0.12)% and (1.75 0.12)% solids remaining respectively] indicating this phenomenon is shear related. This difference in clarication performance was also observed in a pilotscale disc stack centrifuge set up with a low shear feed zone (Q/c.R 4.15 m s1) where the clarication performance was observed to improve from 3.4% solids remaining postcentrifugation using silicone membrane aeration, to 2.85% solids remaining with direct gassing at 500 mL min1. This data indicates that cells taken from cultures operated under conditions where the shear forces are high, but sub-lethal, result in better clarication performance. It is hypothesized that this is due to a change in the physiological state of the cells, with those exposed to long-term high shear in the culture being more resistant to damage in DSP operations. This could be related to the different F-actin network distributions observed here, as recent reports suggest that a change in the F-actin network intensity can lead to different responses to external shear stimuli.22,45 However, cytoskeleton mechanics are still not fully understood and it is therefore difcult to draw any rm conclusions from the work presented here, but it does indicate that a deeper understanding of the physiological response of cells to culture conditions could help to improve the whole process performance.

Biotechnol. Prog., 2013, Vol. 29, No. 1

125

Figure 7. Glycosylation prole at harvest point from 5 L stirred bioreactor (3.5 L working volume) of mAb producing CHO cells (GS-CY01).
Ratio of fucosylated moieties (nG0F, nG1F, and nG2F). n 3 replicates of antibody sample s.d.

ity to a standard sparged system with Pluronic F-68 supplemented medium, demonstrating that when shear protectant agents can not be used a silicone membrane aeration system is an effective way to deliver gas into the culture medium. Direct sparging in the absence of Pluronic F-68 prevented cell proliferation, highlighting the vital role of shear protectant and the signicant impact of bubble bursting. Also, towards the end of the culture with the silicone tube aeration system, actin laments were less dened giving a more shrunken morphology to the cells. In addition, the performance of DSP was not signicantly affected by the different aeration conditions investigated, except from the direct sparging at high ow rate, for which a better clarication performance was observed. Finally, the gas delivered system had a slight effect on the glycosylation prole, whilst harvest point had a greater impact on the relative abundance of glycosylated species.

Acknowledgments
Authors are grateful to Lonza Biologics plc for providing the cell line utilized in this study. Authors thank Graeme Smith from UCL Biochemical Engineering workshop for his help in the construction of the silicone aeration framework. Support for this work was from the Mexican Council on Science and Technology (CONACyT) for MLVS scholarship, Biotechnology and Biological Sciences Research Council, Engineering and Physical Sciences Research Council and Bioprocessing Research Industry Club (Initiative Ref. BB/G010307/1).

Effect of aeration system on product quality The effect of the aeration strategy on product quality at harvest time was determined by N-glycan analysis on days 8 and 10 for batch cultures and days 10 and 14 for fed-batch cultures. It was found that the aeration delivery method, gas ow rate, and harvest time did not have an impact on the ratio of non-fucosylated glycans, with all the conditions tested having similar proportions of G0, G1, and G2 moieties. The proportions of fucosylated species (G0F, G1F, and G2F) were not signicantly changed by the gas delivered system into the culture or in the absence of Pluronic F-68 in the culture medium (Figure 7), except when the bioreactor was directly sparged at a high gas ow rate in batch mode and there was an increase in G0F abundance. G0F percentage was 55% compared to 48% at the low gas ow rate in batch mode. Under this operating condition there was higher hydrodynamic stress caused by the greater presence of bubbles and this has been previously shown to have an effect on the resulting glycosylation prole.25 A minor difference in the ratio of G0F, G1F, and G2F was also observed when the bioreactor was operated in fed-batch, with an increase in the percentage of G1F and G2F compared to that run in batch. This agrees with previous reports that the resulting glycoform can be affected by the production method.25 It can also be seen that harvest time has an impact on the ratio of fucosylated species (Figure 7). As the culture progressed there was an increase in the simple glycan form (G0F) and a decrease in the more complex species (G1F and G2F), agreeing with what has been reported previously.27

LITERATURE CITED
1. Whitford W. Fed-batch mammalian cell culture in bioproduction. BioProc Int. 2006;4:3040. 2. Butler M. Animal cell cultures: recent achievements and perspectives in the production of biopharmaceuticals. Appl Microbiol Biotechnol. 2005;68:283291. 3. Kelley B. Very large scale monoclonal antibody purication: the case for conventional unit operations. Biotechnol Prog. 2007;23:9951008. 4. Sandadi S, Pedersen H, Bowers JS, Rendeiro DA. comprehensive comparison of mixing, mass transfer, Chinese hamster ovary cell growth, and antibody production using Rushton turbine and marine impellers. Bioprocess Biosyst Eng. 2011;34:819832. 5. Nienow AW. Reactor engineering in large scale animal cell culture. Cytotechnology. 2006;50:933. 6. Hu W, Berdugo C, Chalmers JJ. The potential of hydrodynamic damage to animal cells of industrial relevance: current understanding. Cytotechnology. 2011;63:445460. 7. Godoy-Silva R, Chalmers JJ, Casnocha SA, Bass LA, Ma NN. Physiological responses of CHO cells to repetitive hydrodynamic stress. Biotechnol Bioeng. 2009;103:11031117. 8. Chisti Y. Animal-cell damage in sparged bioreactors. Trends Biotechnol. 2000;18:420432. 9. Lakhotia S, Bauer KD, Papoutsakis ET. Damaging agitation intensities increase DNA synthesis rate and alter cell-cycle phase distributions of CHO cells. Biotechnol Bioeng. 1992;40:978990. 10. Chalmers JJ, Bavarian F. Microscopic visualization of insect cell bubble interactions 2. the bubble lm and bubble rupture. Biotechnol Prog. 1991;7:151158. 11. Clincke MF, Guedon E, Yen FT, Ogier V, Roitel O, Goergen JL. Effect of surfactant pluronic F-68 on CHO cell growth, metabolism, production, and glycosylation of human recombinant IFN-gamma in mild operating conditions. Biotechnol Prog. 2011;27:181190. 12. Kilburn DG, Webb FC. The cultivation of animal cells at controlled dissolved oxygen partial pressure. Biotechnol Bioeng. 1968;10:801814. 13. Michaels JD, Petersen JF, McIntire LV, Papoutsakis ET. Protection mechanisms of freely suspended animal cells (CRL 8018)

Conclusion
The use of a silicone membrane aeration system and direct gas sparging with gas blending allowed creation of bubblefree and varied bubbled conditions with constant dissolved gas environments. The apoptosis proles during culture showed that when cells are subjected to stress they are more likely to enter apoptosis in an earlier stage of cell culture and have decreased F-actin intensity. On the other hand, these two operating conditions led to an increase in product titer. The use of a silicone membrane aeration system with Pluronic F-68 free medium allowed comparable growth and productiv-

126
from uid-mechanical injury. Viscometric and bioreactor studies using serum, pluronic F-68 and polyethylene glycol. Biotechnol Bioeng. 1991;38:169180. Goldblum S, Bae YK, Hink WF, Chalmer J. Protective effect of methylcellulose and other polymers on insect cells subjected to laminar shear stress. Biotechnol Prog. 1990;6:383390. Garcia-Briones MA, Chalmers JJ. Flow parameters associated with hydrodynamic cell Injury. Biotechnol Bioeng. 1994;44:10891098. Michaels JD, Nowak JE, Mallik AK, Koczo K, Wasan DT, Papoutsakis ET. Interfacial properties of cell culture media with cell-protecting additives. Biotechnol Bioeng. 1995;47:420430. Henzler HJ, Kauling DJ. Oxygenation of cell cultures. Bioprocess Eng. 1993;9:6175. Ozturk SS. Engineering challenges in high density cell culture systems. Cytotechnology. 1996;22:316. Arden N, Betenbaugh MJ. Regulating apoptosis in mammalian cell cultures. Cytotechnology. 2006;50:7792. Ko KS, McCulloch CAG. Partners in protection: Interdependence of cytoskeleton and plasma membrane in adaptations to applied forces. J Gen Microbiol. 2000;174:8595. Pender N, McCulloch CAG. Quantitation of actin polymerization in 2 human broblast subtypes responding to mechanical stretching. J Cell Sci. 1991;100:187193. Fletcher DA, Mullins D. Cell mechanics and the cytoskeleton. Nature. 2010;463:485492. Ballestrem C, Wehrle-Haller B, Imhof BA. Actin dynamics in living mammalian cells. J Cell Sci. 1998;111:16491658. Tait AS, Aucamp JP, Bugeon A, Hoare M. Ultra scale-down prediction using microwell technology of the industrial scale clarication characteristics by centrifugation of mammalian cell broths. Biotechnol Bioeng. 2009;104:321331. Hossler P, Khattak SF, Li ZJ. Optimal and consistent protein glycosylation in mammalian cell culture. Glycobiology. 2009;19:936949. Kessler M, Goldsmith D, Schellekens H. Immunogenicity of biopharmaceuticals. Nephrol Dial Transpl. 2006;21:912. Reid CQ, Tait AS, Baldascini H, Mohindra A, Racher, A Bilsborough S, Smales CM, Hoare M. Rapid whole monoclonal antibody analysis by mass spectrometry: an ultra scale-down study of the effect of harvesting by centrifugation on the posttranslational modication prole. Biotechnol Bioeng. 2010;107:8595. Qi HN, Goudar CT, Michaels JD, Henzler HJ, Jovanovic GN, Konstantinov K B. Experimental and theoretical analysis of tubular membrane aeration for mammalian cell bioreactors. Biotechnol Prog. 2003;19:11831189. Blackie J, Wu P, Naveh D. Membrane oxygenation of mammalian cell culture fermenters using Dupont Teon AF-2400 tubing. W, Wurm F, editors. Animal Cell In: Bernard A, Grifths B, Noe Technology: Products from Cells, Cells as Products. Kluwer Academic Publishers: The Netherlands, 1999:299301. Bigge JC, Patel TP, Bruce JA, Goulding PN, Charles SM, Parekh RB. Nonselective and efcient uorescent labeling of glycans using 2-amino benzamide and anthranilic acid. Anal Biochem. 1995;230:229238.

Biotechnol. Prog., 2013, Vol. 29, No. 1

31. Wise WS. The measurement of the aeration of culture media. J Gen Microbiol. 1951;5:167177. 32. Boychyn M, Yim SS, Bulmer M, More J, Bracewell DG, Hoare M. Performance prediction of industrial centrifuges using scaledown models. Bioprocess Biosyst Eng. 2004;26:385391. 33. Hutchinson N, Bingham N, Murrell N, Farid S, Hoare M. Shear stress analysis of mammalian cell suspensions for prediction of industrial centrifugation and its verication. Biotechnol Bioeng. 2006;95:483491. 34. Al-Rubeai M, Singh RP, Goldman MH, Emery AN. Death mechanisms of animal cells in conditions of intensive agitation. Nephrol Dial Transpl. 1995;45:463472. 35. Meier SJ, Hatton TA, Wang DIC. Cell death from bursting bubbles: role of cell attachment to rising bubbles in sparged reactors. Biotech Bioeng. 1999;62:468478. 36. Lee GM, Park SY. Enhanced specic antibody productivity of hybridomas resulting from hyperosmotic stress is cell line specic. Biotechnol Lett. 1995;17:145150. 37. Sunley K, Butler M. Strategies for the enhancement of recombinant protein production from mammalian cells by growth arrest. Biotechnol Adv. 2010;28:385394. 38. Tharmalingam T, Sunley K, Butler M. High yields of monomeric recombinant beta-interferon from macroporous microcarrier cultures under hypothermic conditions. Biotechnol Prog. 2008;24:832838. 39. Papoutsakis ET. Media additives for protecting freely suspended animal cells against agitation and aeration damage. Trends Biotechnol. 1991;9:316324. 40. Zhang Z, Al-Rubeai M, Thomas CR. Effect of Pluronic F-68 on the mechanical properties of mammalian cells. Enzyme Microb Technol. 1992;14:980983. 41. Zhu MM, Goyal A, Rank DL, Gupta SK, Vanden Boom T, Lee SS. Effects of elevated pCO2 and osmolality on growth of CHO cells and production of antibody-fusion protein B1: a case study. Biotechnol Prog. 2005;21:7077. 42. Ramirez OT, Mutharasan R. The role of the plasma membrane uidity on the shear sensitivity of hybridomas grown under hydrodynamic stress. Biotechnol Bioeng. 1990;36:911920. 43. Gigout A, Buschmann MD, Jolicoeur M. The fate of Pluronic F-68 in chondrocytes and CHO cells. Biotechnol Bioeng. 2008;100:975987. 44. Ndozangue-Touriguine O, Hamelin J, Breard J. Cytoskeleton and apoptosis. Biochem Pharmacol. 2008;76:1118. 45. Stricker J, Falzone T, Gardel ML. Mechanics of the F-actin cytoskeleton. J Biomech. 2010;43:914. 46. Endresen PC, Prytz PS, Aarbakke J. A new ow cytometric method for discrimination of apoptotic cells and detection of their cell-cycle specicity through staining of F-actin and DNA. Cytometry. 1995;20:162171. 47. Gourlay CW. Ayscough KR. The actin cytoskeleton: a key regulator of apoptosis and ageing?. Nat Rev Mol Cell Biol. 2005;6:583589. Manuscript received Jun. 29, 2012, and revision received Oct. 12, 2012.

14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24.

25. 26. 27.

28.

29.

30.