SARS-CoV-2 Production in a Scalable High Cell Density Bioreactor

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic has demonstrated the value of pursuing different vaccine strategies. Vaccines based on whole viruses, a widely used vaccine technology, depend on efficient virus production. This study aimed to establish SARS-CoV-2 production in the scalable packed-bed CelCradleTM 500-AP bioreactor. CelCradleTM 500-AP bottles with 0.5 L working volume and 5.5 g BioNOC™ II carriers were seeded with 1.5 × 108 Vero (WHO) cells, approved for vaccine production, in animal component-free medium and infected at a multiplicity of infection of 0.006 at a total cell number of 2.2–2.5 × 109 cells/bottle seven days post cell seeding. Among several tested conditions, two harvests per day and a virus production temperature of 33 °C resulted in the highest virus yield with a peak SARS-CoV-2 infectivity titer of 7.3 log10 50% tissue culture infectious dose (TCID50)/mL at 72 h post-infection. Six harvests had titers of ≥6.5 log10 TCID50/mL, and a total of 10.5 log10 TCID50 were produced in ~5 L. While trypsin was reported to enhance virus spread in cell culture, addition of 0.5% recombinant trypsin after infection did not improve virus yields. Overall, we demonstrated successful animal component-free production of SARS-CoV-2 in well-characterized Vero (WHO) cells in a scalable packed-bed bioreactor.


Introduction
The coronavirus disease 2019  pandemic caused by the novel beta-coronavirus severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [1] has had a major impact worldwide with >173 million confirmed infections and >3.7 million deaths (as of 8 June 2021) [2]. Infection may be asymptomatic or cause a wide range of symptoms ranging from mild symptoms of the upper respiratory tract to severe pneumonia and multisystemic disease. Furthermore, following acute disease manifestations, individuals may suffer from post-acute sequelae of SARS-CoV-2 infection [3][4][5][6]. Vaccine development has been moving

Virus Isolate
The virus seed stock used in this study was a passage 4 virus stock derived from the virus isolate SARS-CoV-2/human/Denmark/DK-AHH1/2020 [15]. Passage 1 and Vaccines 2021, 9,706 3 of 17 2 were carried out in Vero E6 cells maintained in DMEM supplemented with 10% fetal bovine serum (Sigma), 100 U/mL penicillin and 100 µg/mL streptomycin (Sigma) [15]. Passage 3 and 4 were carried out in Vero (WHO) cells maintained in serum-free medium as described above. The passage 4 virus stock was characterized by infectivity and genome titers as well as deep sequencing as described below.

Cell Culture and Virus Production in the CelCradle TM 500-AP
The CelCradle TM 500-AP (Esco Aster Pte. Ltd., Singapore) is a single-use bioreactor with BioNOC™ II macrocarriers comprising the packed bed. This bioreactor applies the tide motion principle exposing the cells growing on BioNOC™ II macrocarriers to alternating cycles of nutrition and aeration. Sterile CelCradle TM 500-AP bottles with working volumes of 500 mL used in this study were provided pre-packed with approximately 850 BioNOC™ II carriers. Initially, Vero (WHO) cells were expanded in T175 and T500 flasks (Nunc). Cell seeding was carried out as recommended [16]. Briefly, cells were harvested, centrifuged for 5 min at 200× g and resuspended in 50 mL cell culture medium. The CelCradle TM 500-AP bottle was pre-incubated with 450 mL cell culture medium for 30-60 min at 37 • C and 5% CO 2 prior to seeding of 1.4 × 10 8 -2.3 × 10 8 cells (as specified for individual experiments). CelCradle TM 500-AP stage settings during cell attachment were: Rising rate 2 mm/s, top holding time 20 s, down rate 2 mm/s and bottom holding time 0 s. Attachment was evaluated after 2.5-3.5 h using Equation (1): attachment efficiency = 1 − total cells in suspension total cells seeded × 100.
CelCradle TM 500-AP stage settings during cell culture and infection were rising rate 1 mm/s, top holding time 10 s, down rate 1 mm/s and bottom holding time 10 s. The aim was to maintain the glucose concentration >8.3 mM, and the pH value in the range of 7.0-7.5. A combination of fed-batch and medium exchange was used to maintain glucose concentrations and pH levels in the cultures. Prior to infection, cell culture supernatant glucose concentration, lactate concentration, pH value, and in selected experiments glutamine (supplied as GlutaMAX, the measurements represent total glutamine, both free and available as GlutaMAX) and ammonia concentrations, were evaluated in the morning and evening with the Cedex Bio Analyzer (Roche, Mannheim, Germany) and the FiveGo F2 pH meter (Mettler Toledo, Columbus, OH, USA). Medium was exchanged up to twice per day (morning and evening) to maintain glucose concentration. When required, the pH value and glucose concentration were further adjusted by the addition of 45% glucose solution (Sigma) and 8% NaHCO 3 solution (Sigma). Sampling and measurements were carried out either before or both before and after medium exchange and/or adjustment of glucose and pH, as specified. After infection, glucose concentrations were measured with the GlucCell System glucose meter (Esco Aster Pte Ltd.), and pH values were evaluated with pH-indicator strips (Merck, Darmstadt, Germany). Cell culture glucose consumption between two time points was calculated as the difference between the total amount of glucose available in the cell culture supernatant upon each measurement. Glucose consumption recorded for one day was summed to calculate glucose consumption rate (GCR). The cell-specific GCR (csGCR) was calculated by dividing the total glucose consumption per day by the total cell number per day.
Cells were counted every day until the time of infection. To this end, five carriers were sampled from the 500-AP culture bottle and incubated 20-35 min with Trypsin-EDTA solution (Sigma) for cell detachment. Detached cells were collected in DMEM containing 20% fetal bovine serum. Cells were counted with a Scepter™ 2.0 Cell Counter (Merck) or with a hemocytometer with Trypan Blue (Gibco, Grand Island, NY, USA). The number of cells detached per carrier was multiplied with the estimated number of carriers in the 500-AP culture bottle (assuming 850 carriers upon start of culture, according to the product manual, and subtracting 5-6 carriers per day removed to obtain cell counts) to obtain total cell number in the CelCradle TM 500-AP culture. Infection was carried out when the total cell number exceeded 2 × 10 9 cells with a prior medium exchange. The virus inoculum (passage 4 virus seed stock diluted in a total volume of 50 mL cell culture medium) was added to the culture at a multiplicity of infection (MOI) of 0.006. Then, cultures were maintained at 37 • C or 33 • C and harvests (approximately 90% medium exchange) were carried out 1 or 2 times per day as specified for individual experiments. In one specified experiment, the culture medium contained 0.5% (v/v) TrypLE Select 10× (Gibco, Grand Island, NY, USA) from the time of infection. Virus infectivity titers in the CelCradle TM 500-AP harvests were determined using 50% tissue culture infectious dose (TCID 50 ) assays as described below.
The number of TCID 50 units obtained from each harvest was calculated from the geometric mean of three replicate harvest infectivity titrations and the harvest volume. The accumulated TCID 50 value obtained from each experiment was determined as the sum of TCID 50 units in each harvest. The values given are rounded to one decimal. The cell-specific virus yield (CSVY) was calculated dividing the accumulated TCID 50 units by the total cell number in the CelCradle TM 500-AP recorded on the day of infection. Non-rounded accumulated TCID 50 values were used for these calculations. Similarly, the fold-differences of accumulated TCID 50 values obtained from different experiments were calculated based on non-rounded values.

RNA Titration
Pre-diluted virus-containing cell culture supernatant was mixed 1:3 with Trizol LS (Life Technologies, Carlsbad, CA, USA) and RNA was extracted with chloroform (Sigma) in 5PRIME Phase Lock Gel Heavy tubes (Quantabio, Beverly, MA, USA). RNA was purified with the RNA Clean and Concentrator-5 kit (Zymo Research, Irvine, CA, USA) according to the manufacturer's instructions and eluted in nuclease-free water (Ambion, Austin, TX, USA). Primers and probes have been described elsewhere [21] and the reaction composition was adapted to TaqMan Fast Virus 1-Step Master Mix (Thermo Fisher). Briefly, 400 nM E_Sarbeco_F (5 -ACAGGTACGTTAATAGTTAATAGCGT-3 ), 400 nM E_Sarbeco_R (5 -ATATTGCAGCAGTACGCACACA-3 ) and 200 nM E_Sarbeco_P (FAM-5 -ACACTAGCCATCCTTACTGCGCTTCG-3 -BHQ1) and 2.5 µL purified RNA was used. In each assay, a negative control and RNA standards ranging from 10 1 to 10 5 RNA copies/µL (Twist Bioscience, South San Fransisco, CA, USA) were included. Two technical replicates of all experimental samples and controls were analyzed. Thermal cycling was performed at 55 • C for 10 min for reverse transcription, followed by 95 • C for 3 min and RNA titers (copies/mL) were calculated by interpolation of cycle threshold (Ct) values in the standard curve generated from the standard panel, using the LightCycler 96 software version 1.1.0.1320 (Roche).

S1 ELISA
The amount of S1 obtained from harvests of experiment 7 was determined. Virus harvests collected at 36-122 hpi were pooled and inactivated chemically using β-propiolactone (details on inactivation will be reported elsewhere). The sample pools were evaluated by ELISA. 96-well Maxisorp plates (Nunc) were coated with 2.5 µg/mL capture antibody (Sino Biologicals #40150-D003) diluted in PBS. Plates were shaken at 500 rpm for 1 min and incubated overnight at 4 • C and washed three times with PBS 0.1% tween-20. Blocking was done by incubation with PBS containing 2% (w/v) bovine serum albumin (Roche) for 2 h at room temperature. The plates were washed three times with PBS 0.1% tween-20 followed by the addition of serially diluted recombinant S1 standards (Sino Biologicals #40591-V08H) or samples in duplicates. Plates were incubated at room temperature for 1.5 h with shaking at 500 rpm and washed five times with PBS 0.1% tween-20. Detection antibody (Sino Biologicals #40150-D001-H) diluted 1:5000 in PBS containing 2% (w/v) bovine serum albumin and 0.1% tween-20 was added, and plates were incubated at room temperature for 1.5 h with shaking at 500 rpm followed by washing five times with PBS 0.1% tween-20. Plates were incubated 5-10 min with TMB substrate and stop solution was added. Absorbance was measured at 450 nm (OD450) with a plate reader (BIO-TEK Instruments Inc., Winooski, VT, USA). Absorbance of the negative controls was subtracted from all measurements prior to analysis. The standard curve was generated using a four-parameter logistic regression in GraphPad Prism 9, and S1 sample concentrations were calculated based on the standard curve.

Deep Sequencing
Viral RNA was extracted as described above for RNA titration. Reverse transcriptase polymerase chain reaction (RT-PCR) and PCR conditions and primers are described elsewhere [15]. Amplicons were purified with the DNA clean and concentrator-25 kit (Zymo Research) according to the manufacturer's instructions. The NEBNext Ultra II DNA library Prep kit (New England Biolabs, Ipswich, MA, USA) was used for library preparation and sequence analysis was carried out as described elsewhere [15,22,23]. Mutations prevalent in <5% of the virus population are not reported, and mutations with <50% prevalence are considered minor.

Similar Cell Expansion Obtained Comparing Two Different Cell Numbers for Seeding of the CelCradle TM 500-AP (Experiment 1, 2 and 3)
For all CelCradle TM 500-AP experiments described in this study, following the seeding of the specified number of cells, the cell attachment efficiency was ≥98% after 2.5-3.5 h (Equation (1)). For all CelCradle TM 500-AP cultures, either not subjected to infection (experiments 1-3) or prior to infection with SARS-CoV-2 (experiments 4-8), the medium was exchanged every day from the first or second day post cell seeding (dpcs) and twice per day from 4 dpcs. Additionally, NaHCO 3 solution and glucose were supplied to the culture medium as required. In non-infected cultures, the pH value and concentrations of glucose, lactate, and in selected experiments glutamine and ammonia were measured and reported twice per day. Each time, the cell culture supernatant was sampled before and after medium exchange or adjustment of pH or glucose for calculation of consumption of nutrients and production of waste products between two time points.
Vero cells were first cultivated in the CelCradle TM 500-AP bioreactor system without SARS-CoV-2 infection to evaluate cell growth and peak total cell numbers. Two seed cell numbers (1.4 × 10 8 , Seed Low , experiment 1 and 2.3 × 10 8 , Seed High , experiment 2) were Vaccines 2021, 9, 706 6 of 17 initially compared ( Figure 1). In the Seed Low culture, the glucose concentration was in the range of 7.7-22 mM throughout the time of culture and was below the lower target of 8.3 mM before medium exchange at 5 and 7 dpcs. pH values were in the range of 6.9-7.5 throughout the time of culture, thus pH values were below the lower target of 7 before medium exchange at 5-8 dpcs ( Figure 1A, Table 1). Daily peak concentrations of lactate increased to 24 mM at 9 dpcs ( Figure 1B, Table 1). Glutamine concentrations (supplied as GlutaMAX) were in the range of 3.5-4.2 mM and daily consumption was relatively low peaking at 0.14 mmol at 9 dpcs. Ammonia concentrations up to 0.4 mM were observed at 9 and 10 dpcs ( Figure 1C, Table 1). In the Seed High culture, similar cell culture supernatant parameters were recorded ( Figure 1, Table 1).
glucose, lactate, and in selected experiments glutamine and ammonia were measured and reported twice per day. Each time, the cell culture supernatant was sampled before and after medium exchange or adjustment of pH or glucose for calculation of consumption of nutrients and production of waste products between two time points.
Vero cells were first cultivated in the CelCradle TM 500-AP bioreactor system without SARS-CoV-2 infection to evaluate cell growth and peak total cell numbers. Two seed cell numbers (1.4 × 10 8 , SeedLow, experiment 1 and 2.3 × 10 8 , SeedHigh, experiment 2) were initially compared ( Figure 1). In the SeedLow culture, the glucose concentration was in the range of 7.7-22 mM throughout the time of culture and was below the lower target of 8.3 mM before medium exchange at 5 and 7 dpcs. pH values were in the range of 6.9-7.5 throughout the time of culture, thus pH values were below the lower target of 7 before medium exchange at 5-8 dpcs ( Figure 1A, Table 1). Daily peak concentrations of lactate increased to 24 mM at 9 dpcs ( Figure 1B, Table 1). Glutamine concentrations (supplied as GlutaMAX) were in the range of 3.5-4.2 mM and daily consumption was relatively low peaking at 0.14 mmol at 9 dpcs. Ammonia concentrations up to 0.4 mM were observed at 9 and 10 dpcs ( Figure 1C, Table 1). In the SeedHigh culture, similar cell culture supernatant parameters were recorded ( Figure 1, Table 1).

Cell culture supernatant parameters prior to infection
Peak GCR, mmol/day (dpcs) 11 (9) 12 (8) 13 (7) 13 (7) 13 (7) 12 (7) 11 (7) 13 (7) Glucose range, mM 7 Comparing experiment 1 and experiment 2 with different seed cell numbers, there was only a minor difference in total cell numbers throughout the time of culture ( Figure 1D, Table 1). Total cell numbers peaked at 9 dpcs with 2.8 × 10 9 cells (3.5 × 10 6 cells/carrier) and 2.7 × 10 9 cells (3.4 × 10 6 cells/carrier) in the Seed Low and Seed High culture, respectively. Similarly, only a minor difference in GCR was seen, with the Seed High culture having a slightly higher GCR throughout most of the culture period. The GCR peaked at 9 dpcs with 11 mmol/day in the Seed Low culture and at 8 dpcs with 12 mmol/day in the Seed High culture ( Figure 1D, Table 1). The csGCR was highest on the early days post cell seeding and then seemed to plateau at 4-5 pmol/cell/day from 6 dpcs onwards ( Figure 1E). A third culture (experiment 3) (Table 1) initiated with a low seed cell number (1.5 × 10 8 cells) demonstrated similar cell growth and culture parameters compared to the first experiment with peak total cell numbers of 2.7 × 10 9 (3.4 × 10 6 cells/carrier) at 9 dpcs. Thus, we concluded that there were no differences for the two seed cell numbers tested regarding peak total cell numbers and the number of days required for cell expansion. Since glutamine did not seem to be a limiting factor, glutamine and ammonia concentrations were not monitored in subsequent experiments.

SARS-CoV-2 Production with One Harvest Per Day (Experiment 4)
In all the following experiments, CelCradle TM 500-AP cultures for SARS-CoV-2 infection and production were seeded with 1.5 × 10 8 cells. The monitored pH values, concentrations of glucose and lactate, as well as total cell numbers and GCR, showed similar trends as to what was observed in initial experiments until the time of infection (Figure 2A-C, Table 1). Infection was carried out at the MOI of 0.006 at 7 dpcs when the total cell numbers exceeded 2 × 10 9 (2.5 × 10 6 cells/carrier) ( Figure 2C, Table 1). This MOI was selected based on observations in another study [24], as well as on a pilot experiment in T25 flasks. In cultures infected with this MOI, virus-induced cell death was generally not observed one day post-infection (dpi); therefore, infections of the CelCradle TM 500-AP cultures were carried out before peak total cell numbers were expected based on the initial CelCradle TM 500-AP experiments.

SARS-CoV-2 Production with One Harvest Per Day (Experiment 4)
In all the following experiments, CelCradle TM 500-AP cultures for SARS-CoV-2 infection and production were seeded with 1.5 × 10 8 cells. The monitored pH values, concentrations of glucose and lactate, as well as total cell numbers and GCR, showed similar trends as to what was observed in initial experiments until the time of infection (Figure 2A-C, Table 1). Infection was carried out at the MOI of 0.006 at 7 dpcs when the total cell numbers exceeded 2 × 10 9 (2.5 × 10 6 cells/carrier) ( Figure 2C, Table 1). This MOI was selected based on observations in another study [24], as well as on a pilot experiment in T25 flasks. In cultures infected with this MOI, virus-induced cell death was generally not observed one day post-infection (dpi); therefore, infections of the CelCradle TM 500-AP cultures were carried out before peak total cell numbers were expected based on the initial CelCradle TM 500-AP experiments. The virus seed stock used to infect the CelCradle TM 500-AP cultures in this study was a passage 4 polyclonal stock derived from a SARS-CoV-2 isolate from a Danish patient [15]. Compared to the original patient isolate, viruses in the passage 4 seed stock had acquired four minor (prevalence between 5 and 50%) substitutions all localized in the S protein. These substitutions were H66R with an estimated prevalence of 47% in the viral population, D215G with a prevalence of 12%, H245R with a prevalence of 17%, and S247R with a prevalence of 5% (Table 2). The virus seed stock was produced in a T175 flask and had an infectivity titer of 6.4 log 10 TCID 50 /mL and an RNA titer of 9.7 log 10 copies/mL. The first CelCradle TM 500-AP culture infected with SARS-CoV-2 for virus production was maintained at 37 • C during the virus production phase. Virus harvest and medium exchange were carried out once per day on 1-4 dpi. After infection, pH values and glucose concentrations were only measured before harvest and medium exchange, and lactate concentrations were not evaluated ( Figure 2B). Additional NaHCO 3 and glucose were added with each harvest and medium exchange. This was insufficient to maintain pH values or glucose concentrations above the lower target levels ( Figure 2A, Table 1). It was expected that the addition of even larger volumes of NaHCO 3 would leave pH values inappropriately high in the first hours after harvest and medium exchange.

Improved Virus Yields by Two Harvests per Day (Experiments 5 and 6)
To evaluate whether increasing the harvest frequency to two harvests per day would increase virus yield from the CelCradle TM 500-AP culture, virus harvest and medium exchange were carried out twice per day from 1-4 dpi. After infection, pH values and glucose concentrations were only measured before harvest and medium exchange. Under these conditions, smaller volumes of NaHCO 3 and glucose were required to maintain pH values and glucose concentrations at similar levels as in experiment 4 ( Figure 2A, Table 1); however, pH values and glucose concentrations remained difficult to control at 1 and 2 dpi.
A peak infectivity titer of 6.4 log 10 TCID 50 /mL was observed for the harvest at 52 hpi for experiment 5 ( Figure 2D). Three harvests had infectivity titers ≥6 log 10 TCID 50 /mL (harvests at 38, 52 and 62 hpi). The accumulated TCID 50 from all harvests (3.9 L) was 9.5 log 10 TCID 50 , resulting in a CSVY of 1.3 TCID 50 /cell ( Table 3). The virus yield from this culture with two harvests per day was thus approximately 1.6-fold increased, compared to the yield from the first culture with one harvest per day (experiment 4).
In a similar experiment with two harvests and medium exchanges per day at 1-3 dpi (experiment 6), a peak infectivity titer of 7.0 log 10 TCID 50 /mL was observed for the harvest at 60 hpi ( Figure 2D). Three harvests had infectivity titers ≥6.5 log 10 TCID 50 /mL (harvests at 36, 49, and 60 hpi), and two additional harvests had infectivity titers in the range 6-6.5 log 10 TCID 50 /mL (harvests at 24 and 73 hpi). The accumulated TCID 50 from all harvests was 10.1 log 10 TCID 50 (3 L), resulting in a CSVY of 5.9 TCID 50 /cell ( Table 3). The virus yield from this second culture with two harvests per day (experiment 6) was approximately 4.3-fold higher than that of the first culture with two harvests per day (experiment 5) and 7.1-fold higher when compared to the yield of the culture with one harvest per day (experiment 4). Deep sequencing of virus from experiment 5 harvested at 38 hpi showed viral genetic stability. Compared to the seed virus there were only two substitutions in the Spike protein, R685S and S686R, both recorded with frequencies of 5-10% (Table 2). Overall, two harvests per day allowed better control of glucose concentration and pH values and increased virus yields.

Virus Production at 33 • C Further Enhanced Virus Yield (Experiment 7)
It has been described for both protein and virus production that a reduction in temperature may have beneficial effects on product yields [25][26][27]. Therefore, after cell seeding, expansion, and infection as described above ( Figure 3A-C), a culture was maintained at 33 • C with virus harvest as well as medium exchange being carried out twice per day at 1-6 dpi. In this and the following experiment pH values and glucose concentrations were measured both before and after harvest and medium exchange. Compared to the previous experiments, following infection, the pH values and glucose concentrations were better maintained above the lower target levels with a cultivation temperature of 33 • C ( Figure 3A, Table 1). temperature may have beneficial effects on product yields [25][26][27]. Therefore, after cell seeding, expansion, and infection as described above ( Figure 3A-C), a culture was maintained at 33 °C with virus harvest as well as medium exchange being carried out twice per day at 1-6 dpi. In this and the following experiment pH values and glucose concentrations were measured both before and after harvest and medium exchange. Compared to the previous experiments, following infection, the pH values and glucose concentrations were better maintained above the lower target levels with a cultivation temperature of 33 °C ( Figure 3A, Table 1). A peak infectivity titer of 7.3 log 10 TCID 50 /mL was observed for the harvest at 72 hpi ( Figure 3D). Thus, six harvests had infectivity titers ≥6.5 log 10 TCID 50 /mL (harvests at 36, 48, 59, 72, 84 and 98 hpi), and three additional harvests had infectivity titers in the range 6-6.5 log 10 TCID 50 /mL (harvests at 111, 122 and 136 hpi). The accumulated TCID 50 from all harvests (5.2 L) was 10.5 log 10 TCID 50 , resulting in a CSVY of 12.7 TCID 50 /cell ( Table 3). The total virus yield was thus increased approximately 9.4-and 2.2-fold, when compared to the yields obtained with two harvests per day at 37 • C in experiments 5 and 6, respectively. The peak RNA titer of 11 log 10 copies/mL was observed for the harvest at 98 hpi ( Figure 3D). Harvests at 36 to 111 hpi had RNA titers in the range of 10.7-11 log 10 copies/mL, and a total of 2.4 × 10 14 copies were obtained from this experiment. Approximately 3.7 mg Spike protein subunit S1 in total was obtained from a pool of eight harvests with the highest infectivity titers from 36 to 122 hpi according to S1 quantification in chemically inactivated samples by ELISA as described in Materials and Methods. Deep sequencing of an inactivated virus pool (experiment 7 virus pool) consisting of these eight harvests from experiment 7 collected at 2-5 dpi and spiked with virus from experiment 5 harvested at 62 hpi indicated one substitution in the Spike protein, S686R, compared to the seed virus with a frequency of 5% (Table 2).

Addition of TrypLE to the Cell Culture Medium Did Not Enhance Virus Production (Experiment 8)
It has been suggested that trypsin activity is important for SARS-CoV-2 entry and that exogenously added trypsin may enhance SARS-CoV-2 spreading in cell culture [28][29][30]. To confirm these results using our virus and cell line, we tested in T25 flasks whether the addition of TrypLE, a recombinant trypsin replacement product, to the cell culture medium had an effect on virus yields at 37 • C. TrypLE concentrations in the range of 0.25-1% (v/v) were evaluated. A concentration of 0.5% (v/v) TrypLE improved virus spreading and resulted in slightly higher infectivity titers, whereas a concentration of 1% (v/v) TrypLE led to significant cell death and reduced infectivity titers compared to the culture without TrypLE. A concentration of 0.25% (v/v) TrypLE resulted in similar cell death and infectivity titers as observed in the culture without TrypLE (data not shown). Also, in non-infected cells maintained in T25 flasks at 33 • C, 1% (v/v) TrypLE resulted in moderate cell death whereas 0.5% (v/v) TrypLE did not result in apparent cell death. Trying to further improve the culture conditions so far resulting in the highest virus yields (experiment 7), we evaluated if 0.5% (v/v) TrypLE also enhanced infectivity titers and overall virus yield in a CelCradle TM 500-AP culture maintained at 33 • C after the time of infection. For this, cells were expanded in the CelCradle TM 500-AP as described above and infected at 7 dpcs. From the time of infection, the culture was maintained at 33 • C and 0.5% (v/v) TrypLE was added to the culture medium upon each harvest and medium exchange. Harvest and medium exchange were carried out twice per day at 1-5 dpi. Additional NaHCO 3 would have been required to keep pH above the lower target level at 1 dpi, however, pH values and glucose concentrations were generally well maintained as was also the case during the previous 33 • C virus production (experiment 7) ( Figure 3A, Table 1).
A peak infectivity titer of 7.2 log 10 TCID 50 /mL was observed for the harvest at 59 hpi ( Figure 3D). Five harvests had infectivity titers ≥6.5 log 10 TCID 50 /mL, (harvests at 35, 48, 59, 72 and 83 hpi), and one additional harvest had infectivity titers in the range 6-6.5 log 10 TCID 50 /mL (harvest at 95 hpi). The accumulated TCID 50 from all harvests (4.3 L) was 10.4 log 10 TCID 50 , resulting in a CSVY of 11.1 TCID 50 /cell (Table 3). Overall, the virus yield from this culture maintained at 33 • C with 0.5% TrypLE in the cell culture medium during the virus production phase was not improved compared to the culture without TrypLE (experiment 7), which had approximately 1.2-fold higher yield.

Discussion
Efficient virus production is critical for vaccines based on whole viruses but also has applications in basic virus research, such as for ultrastructural studies of viral particles, animal immunogenicity studies, vaccine challenge trials, and in development of diagnostic assays. In this study, we established efficient SARS-CoV-2 production in animal componentfree conditions using a well-characterized cell line cultivated in the CelCradle TM 500-AP bioreactor, a scalable cell culture system useful for process development and optimization studies for production of vaccines. Further, the established procedures could be of value for generation of virus seed stocks for subsequent pilot and production scale vaccine manufacturing.
Traditionally, adherent cells have been cultivated in monolayers applying roller bottles or multilayer cell culture flasks for production purposes. For adherent cells that are difficult to cultivate on microcarriers in stirred tank bioreactors, options for large-scale high-density cell culture have been limited. Packed-bed bioreactors facilitating large-scale cultivation of adherent cells have recently been developed. The CelCradle TM 500-AP (Esco Aster Pte. Ltd.) is a packed-bed bioreactor with a 0.5 L working volume operated by tide-motion for mixing with a bed consisting of BioNOC™ II macrocarriers. It thus supports cultivation of adherent cells and operates with a gentle mixing mode. The CelCradle TM 500-AP has been demonstrated to successfully support production of Japanese encephalitis virus in Vero cells [31], hepatitis D virus-like particles in baby hamster kidney cells [32], and influenza A virus in Madin-Darby canine kidney cells and Vero cells [33]. Importantly, the CelCradle TM 500-AP used in this study is a laboratory-scale version of the TideXcell system available with packed-bed capacities of up to 100 L, thus applying the cultivation principles of the CelCradle TM 500-AP at production scale [34]. A linear scalability from the CelCradle TM to the 10 L TideXcell system was recently reported for influenza A virus in Madin-Darby canine kidney cells [33]. Apart from the TideXcell systems, other large-scale packed-bed bioreactors available include the iCELLis 500+ system (Pall Life Sciences, New York, NY, USA) and scale-X nitro (Univercells Technologies, Nivelles, Belgium), both with culture volumes of 60-70 L [14].
According to recent reports, Vero (WHO) cell culture-derived SARS-CoV-2 for the inactivated whole virus vaccines from Sinovac [24] and Sinopharm [35] were produced in a 25 L wave bag bioreactor (Cytiva) [36] and a 10 L basket bioreactor [35,37], respectively. The inactivated whole virus vaccine from Bharat Biotech was produced in unspecified bioreactors [38,39], while the vaccine developed by Valneva currently going through clinical trials is produced in roller bottles [40]. Basket bioreactor cultures were reported to yield harvest titers of around 7 log 10 TCID 50 /mL, but no process details have been published describing virus production in these bioreactors. The wave bag bioreactor technology might be scaled to working volumes of up to 100 L, and while the basket bioreactor system is now used in large-scale vaccine manufacturing, specific information on reactor size and working volume is not publicly available. More detailed reports of cell-culture based SARS-CoV-2 production could guide improvements of virus production processes.
SARS-CoV-2 production in this study was based on a highly characterized Vero (WHO) cell bank also available for GMP production [41]. It is derived from the WHO Vero cell bank which is widely accepted for vaccine production and used in several licensed viral vaccines based on cell-derived antigen [42], including the inactivated SARS-CoV-2 vaccines mentioned above. These cells show high permissiveness to SARS-CoV-2 infection and can readily be grown in serum-free medium. The total cell numbers of Vero cells in the CelCradle TM 500-AP cultures obtained with the culture conditions of this study were similar to those in previous reports on Vero cell expansion in the CelCradle TM 500-AP bioreactor [31,43]. Different harvest strategies and production conditions were tested in this study, and the highest yield was obtained with two harvests per day and a change of temperature to 33 • C at the time of infection (experiment 7). While production at 33 • C was reported to increase yield in other bioprocesses, it is possible that the improved virus yield observed at 33 • C in this study was at least partly attributed to improvements in pH control due to changes in cell metabolism at the reduced temperature [25,27,44]. In T25 flask experiments in our laboratory, typically peak infectivity titers of up to 6.5 log 10 TCID 50 /mL and CSVY of up to 7 have been observed. Thus, virus production at improved conditions in the CelCradle TM 500-AP appeared more efficient than production in small monolayer cell cultures.
Limited consumption of glutamine was observed, and thus glucose was the main carbon source of the culture, which was also reflected in the low levels of ammonia and relatively high levels of lactate measured in the culture supernatant. Lactate levels in the range of 20-30 mM, measured in this study, have been reported not to considerably interfere with cell growth in other studies [45,46]. However, cell culture parameters with undetectable effects on cell growth could potentially still affect cell culture productivity [44]. Additionally, during cell expansion, it became challenging to maintain pH within the desired range by manual addition of NaHCO 3 . Thus, further improved productivity might be achieved in a bioreactor supporting closed-loop pH control to compensate for the acidification by the lactate. Recently developed improved versions of the CelCradle TM 500-AP and the larger scale TideXcell systems are equipped with closed-loop pH control and thus enable a stable pH during the period of cell culture and virus propagation.
Whereas trypsin-like protease activity plays a role for SARS-CoV-2 cell entry [28][29][30], addition of 0.5% (v/v) TrypLE to the culture medium (experiment 8) from the time of infection did not increase virus production in the CelCradle TM 500-AP culture maintained at 33 • C. It should be mentioned that TrypLE activity is sensitive to pH and temperature [47]. The low pH values measured at 1 dpi, possibly in combination with the cultivation temperature of 33 • C, might thus reduce TrypLE activity and potentially reduce the effect on virus yield when compared to the pilot experiment in T25 flasks at 37 • C. In future experiments, it would be relevant to test the effect of TrypLE addition on virus production in a CelCradle TM 500-AP culture maintained at 37 • C.
In addition to application of closed-loop pH control, other strategies to further improve the virus yield from Vero cells in this packed-bed bioreactor could target cell culture medium formulation or additives [25,48]. Further, optimized medium refreshment strategies during cultivation to reduce waste accumulation and maintain nutrient levels, such as optimized schemes for introduction of partial medium exchange, cell culture medium recirculation or perfusion, might contribute to increased productivity [33,46,[49][50][51][52]. These feeding modes could similarly benefit pH control. A further cell culture adapted SARS-CoV-2 virus inoculum might also facilitate higher production yields. It is possible that other SARS-CoV-2 strains show different growth characteristics in Vero cells, thus, producing other strains of SARS-CoV-2 in the CelCradle TM 500-AP could require adjustments of the protocols described in this study to optimize yield.
The SARS-CoV-2 virus strain used in this study was isolated from a Danish patient [15] and had six amino acid changes compared to the Wuhan-Hu-1 SARS-CoV-2 virus strain [53] ( Table 2). The Danish isolate carries the D614G substitution, presenting the po-lymorphism that has been dominating at this position globally since the spring of 2020 [54]. Of interest for vaccine antigen production, the D614G substitution has been reported to increase virion Spike protein content and to mediate increased infectivity [55]. The four substitutions in the S protein in the CelCradle TM 500-AP virus seed stock with at least 5% prevalence acquired during four passages in cell culture were represented in the GISAID database of patientderived virus sequences [56], however, at frequencies <0.2% (total number of registered sequences were 415,516, as of 12 April 2021). These substitutions were not located in the critical receptor binding domain of the Spike protein. It will be relevant to investigate immunogenicity of the CelCradle TM 500-AP produced SARS-CoV-2 and to evaluate how substitutions in the vaccine antigen Spike protein might influence immunogenicity and protection against COVID-19 conferred by relevant virus variants [57].
Taken together, our data show that with a small footprint laboratory-scale bioreactor SARS-CoV-2 can be produced at high titers of up to 7.3 log 10 TCID 50 /mL, with an accumulated TCID 50 of 10.5 log 10 TCID 50 , and obtaining 3.7 mg S1 from one experiment. Assuming a vaccine dose of 7 log 10 TCID 50 or 1 ug S1 and 50% loss in downstream processing, approximately 1500 doses could be obtained from one CelCradle TM 500-AP culture (1 dose / 3 mL bioreactor harvest) and several hundred thousand vaccine doses could potentially be obtained from a 100 L TideXcell bioreactor. Assumptions on the vaccine dose are based on data for other antiviral vaccines [36,37,[58][59][60][61][62] and observations from preliminary immunogenicity studies in mice (unpublished). Virus manufacturing based on the described and similar processes might be of interest for inactivated but also attenuated second-generation COVID-19 vaccines [63,64].