The recombinant chimeric spike soluble (CSs) protein (Genbank QJE37812.1, Amanat et al., 2020) comprises the Sars-CoV-2 S1/S2 spike protein with two stabilizing proline mutations (K986P, V987P) and a mutation of the polybasic cleavage site (RRAR) together with a thrombin cleavage site. The chimeric foldon domain of the T4 phage was added at the C-terminus of the protein to promote trimerization of the molecule (Meier et al., 2004) followed by a 6x His tag. The mature CSs monomeric chain, thus, consists of 1,242 amino acids, 30 cysteine residues, and 22 potential N-glycosylation sites. The calculated molecular weight of the monomeric CSs chain is 180 kDa, assuming 2.5 kDa per glycosylation site. Similarly, a theoretical molecular weight of 540 kDa is calculated for the full trimeric CSs spike protein. Amino acid sequence information is presented.

The original plasmid pCAGGS-CSs (Amanat et al., 2020) was received from Prof. Krammer at the Mount Sinai Medical School (New York City, NY, USA) and was co-transfected with pCIneo (Promega) that harbors the G418 selection marker. For stable cell line generation, the open reading frame of CSs was cloned into the AscI-linearized pL mammalian expression vector (Mader et al., 2012; Zboray et al., 2015), resulting in pL-CSs. The backbone of pL includes a chimeric CAGGS promoter (CMV enhancer, chicken beta-actin promoter, chimeric intron of chicken beta-actin and rabbit beta-globin), bovine growth hormone (bGH) poly(A) signal, phosphoglycerate kinase promoter driving expression of the neomycin phosphotransferase gene (NeoR/KanR) for G418 resistance, a bacterial origin of replication (ori), bacterial ampicillin resistance gene (AmpR) and a 5′ and 3′ homology region (5′HR and 3′HR) for potential recombination into bacterial artificial chromosomes (BACs).

A suspension CHO-K1 cell line adapted to ActiPro medium (Cytiva) + 6 mM L-gln was chosen as the preferred host cell line for recombinant protein expression. It was previously demonstrated that this host cell line is particularly suitable for recombinant protein production (Reinhart et al., 2018). Alternatively, these cells can be cultivated in CDCHO medium (Gibco) without negatively impacting cellular growth. For transfection, 10 µg of plasmid DNA was resuspended in 200 µl of CD-DG44 medium (Gibco). Five micrograms of pCAGGS-CSs and 5 µg of pCIneo were mixed and transfected to generate initial transfection pools and early protein production. Similarly, 10 µg of pL-CSs was transfected for the establishment of monoclonal stable high-producing cell lines. Lipofectin, 50 µl, was mixed with 150 µl of CD-DG44, incubated for 35 min, then pooled with the plasmid solution and incubated for 10 min. The Lipoplex solution was then added to 2 × 10⁶ viable CHO-K1 cells in 3.6 ml CD-DG44 + 8 mM stable L-gln + 15 mg/L phenol red and incubated for 24 h at 220 rpm. At 24 h post-transfection, the cells were centrifuged and resuspended in 10 ml of CDCHO + 8 mM stable L-gln + 15 mg/L phenol red. A 2-day post-transfection (dpt) selection was started by centrifugation of the cells and resuspension in 4 ml of CDCHO + 8 mM stable L-gln + 15 mg/L phenol red + 0.5 mg/ml G418 and seeded at 2,500 cells/well in 50 µl in a 384 deep-well plate. The plate was fed with 50 µl of selection medium 10 dpt, and growing wells were expanded to 96-well plates and finally to 50 ml of shake tubes starting at 16 dpt. High-producing subclones were then generated by the second round of single-cell limiting dilution subcloning in 384-well plates to select homogenous monoclonal subclones.

The initial cell pool of pCAGGS-CSs/pCIneo co-transfected CHO-K1 cells was cultivated in 200-ml shake tubes (ST200) at 50-ml working volume in ActiPro supplemented with Cell Boost 1 and 3 (Cytiva) and 8 mM L-glutamine with a daily medium exchange to increase the maximum cell density, called semi-continuous perfusion process. Tangential flow filtration with a MWCO of 100 kDa was used to concentrate the supernatant and exchange the buffer to 20 mM sodium phosphate + 0.5 M NaCl + 40 mM imidazole. A HisTrap Fast Flow (FF) column (Cytiva) was used for IMAC purification and CSs was eluted by 20 mM sodium phosphate + 0.5 M NaCl + 500 mM imidazole. Finally, a PD10 column was used for buffer exchange to PBS, and a 15-kDa Amicon column to concentrate the sample. The protein solution was quantified by total A280 absorbance and frozen to −80°C.

The parental stable CSs clone 131A4 and the final subclone CSs C11 (short “C11”) were cultivated in different bioprocesses and different media. C11 was cultivated in lab-scale fed-batch bioreactors or subjected to an optimized semi-continuous perfusion protocol.

For semi-continuous perfusion cultivation, 50 ml of cell suspension was seeded at 10 × 10⁶ viable cells/ml in a ST200 in CDCHO + 8 mM L-gln and incubated at 220 rpm, 50 mm shaking amplitude, 37°C, humid 7% CO₂ atmosphere in a Kuhner shaker incubator. Exactly every 24 h, cells were centrifuged at 1,300 rpm for 7 min, the supernatant was collected and stored at 4°C, and the cell pellet was disrupted and resuspended in fresh pre-warmed medium. An exact protocol for semi-continuous perfusion cultivation can be found in Mayrhofer & Kunert (2020).

For the integrated perfusion experiment of clone 131A4, a ReadyToProcess WAVE25 bioreactor (Cytiva) was used with ActiPro medium supplemented with Cell Boost 1 and 3 (Cytiva) and 8 mM L-glutamine. An Äkta Pure system equipped with HisTrap Excel was directly integrated with the perfusion harvest line. Fed-batch cultivation was performed in a 500-ml DASGIP system and CDM4HEK293 basal medium fed with Cell Boost 7a and 7b (Cytiva).

The supernatant was concentrated and buffer exchanged to running buffer containing 20 mM phosphate buffer, 150 mM NaCl, 20 mM imidazole, pH 7.4 by TFF using a 100-kDa membrane. Stepwise elution was initiated by 20%, 40%, 70%, and 100% of elution buffer containing 20 mM phosphate buffer, 150 mM NaCl, 500 mM imidazole, pH 7.4. Pooled elution fractions were loaded onto Amicon Ultra-15 centrifugal filter devices with a 100-kDa cutoff, and buffer was exchanged to 20 mM Tris-HCl, 150 mM NaCl, pH 8.0. Protein content was quantified by total A280 absorbance or quantitative CSs-specific ACE2/anti-His sandwich ELISA. Samples were 0.2 µm filtered and snap frozen to −80°C.

A quantitative ACE2/anti-His sandwich ELISA was developed for the quantification of CSs samples. The 96-well MaxiSorp plates (Thermo Fisher) were coated with 5 μg/ml of ACE2 (R&D systems) in PBS overnight at 4°C or 2 h at room temperature. Unknown samples or 1 μg/ml of purified CSs standard were serially diluted in PBS with 0.1% Tween (TPBS) and 1% BSA. Coated plates were washed with TPBS, and the applied samples were incubated for 1 h at room temperature. Biotinylated anti-His (Thermo Scientific) detection antibody was diluted at 1:1,000 in TPBS + 1% BSA and incubated on the plate for 1 h at room temperature. Streptavidin–HRP conjugate (Roche) was diluted at 1:5,000 in TPBS + 1% BSA and incubated on the plate for 30 min at room temperature. The color reaction was developed using TMB substrate and stopped with concentrated H₂SO₄ followed by OD 450-nm measurement at 620-nm reference. The EC50 value of CSs binding to ACE2 was determined by fitting a non-linear four-parametric sigmoidal log model on the absorbance values at different CSs molar concentrations. To evaluate the binding affinity of the anti-SARS monoclonal antibody CR3022, purified CSs protein (5 μg/ml) was coated, and serially diluted biotinylated CR3022 was prepared starting at 2 μg/ml (13 nM) in TPBS + 1% BSA. After incubation for 1 h at room temperature, the detection was performed with streptavidin–HRP conjugate (1:5,000) and TMP substrate.

A similar ELISA setup was used to confirm the CSs protein quality with vaccinated donor samples. CS-coated plates were blocked for 1 h at room temperature with TPBS containing 3% of non-fat skim milk powder (MP). Samples were diluted in TPBS with 3% MP and incubated for 2 h at room temperature. Anti-human gamma-HRP detection antibody (Invitrogen) was diluted at 1:5,000 and incubated for 1 h at room temperature. A color reaction was developed using the TMB substrate.

EC50 values were determined by fitting a non-linear four-parametric sigmoidal log model on the absorbance values at different CSs molar concentrations.

SARS-CoV-2 convalescent plasma (Takeda) was thawed and cryoprecipitated at 4°C for 2 h followed by centrifugation at 5,000 × g for 15 min. Supernatant, 1 ml, was diluted with 2 ml of PBS, and 2.45 ml of saturated ammonium sulfate was added under continuous stirring, and the mixture was incubated at 4°C for 5 h. After centrifugation at 1,000 × g for 15 min, the precipitate was washed with 5 ml of 45% saturated ammonium sulfate solution and centrifuged at 1,000 × g for 15 min. The pellet was resuspended in 1 ml PBS and centrifuged again at 1,000 × g for 2 min. The supernatant was dialyzed against PBS overnight at 4°C and centrifuged again at 10,000 × g for 5 min. Following 0.22-µm filtration and storage at 4°C, 1 ml of cryo- and ammonium sulfate precipitated hyperimmune serum was diluted with 9 ml of phosphate buffer and loaded onto a protein A affinity column. Polyclonal hyperimmune IgGs were eluted with 50 mM sodium acetate pH 3.5 and neutralized with Tris-HCl pH 8.0. Hyperimmune IgG, 1.54 mg, was biotinylated using the EZ-Link NHS-PEG4-Biotin kit (ThermoFisher).

Bis-Tris NuPAGE polyacrylamide gels, 4%–12%, were run in MOPS SDS running buffer. Samples were incubated with NuPAGE LDS sample buffer for 10 min at 70°C for non-reducing conditions or additionally mixed with NuPAGE reducing agent and incubated for 5 min at 98°C for reducing conditions. The gels were run at 200 V for approximately 1 h at room temperature. A PVDF membrane was activated for 1 min with methanol (MeOH) and rinsed with NuPAGE transfer buffer in 20% MeOH for Western blotting. Proteins were blotted from the SDS-Gel onto the PVDF membrane using a PerfectBlue “Semi-Dry” Electro Blotter (peqlab) at 10 V, 3 mA/cm² for 1 h. The blotted membrane was blocked with 5% BSA in TPBS for 1 h or overnight and washed with TPBS. Detection was performed for 1 h with primary detection antibody in TPBS with 0.5% BSA. Secondary detection was performed with streptavidin–HRP conjugate in TPBS with 0.5% BSA for 30 min. Chemiluminescence was induced with the SuperSignal West Pico Plus substrate (Thermo Fisher).

Preliminary material of a new protein of interest is needed to rapidly develop analytical methods supporting the selection of high-producing subclones during clone development. The originally received plasmid (pCAGGS-CSs) (Amanat et al., 2020) was co-transfected with a vector (pCIneo) containing a suitable G418 selection marker in a quick and straightforward effort. Transfection pools were generated by lipofection, G418 pool selection and after recovery of the viability above 95%, expansion to higher cell concentration using shake tube bioreactors. A semi-continuous perfusion culture starting already 25 days post-transfection was used to generate enough protein to develop analytical methods. Using a daily medium exchange protocol, high cell densities of 50 × 10⁶ viable cells/ml were achieved and passaged once to maintain high viability (Figure 1A). The culture supernatant for optimization of the IMAC purification step was collected from harvests of the semi-continuous perfusion culture of the initial transfection pool.

The purified CSs preparation (Figures 1B, C) was suitable to support essential analytical methods for further clone development (Table 1). The analytical workflow aims to screen, identify, and select high-producing clones. The rapid production of a new recombinant protein such as the SARS CoV-2 spike protein faces several challenges, particularly the availability of suitable reagents (Esposito et al., 2020). One strategy is to use generic analytic methods, e.g., total protein visualization by SDS gel electrophoresis, or ELISA-based methods that detect commonly used protein tags (e.g., 6xHis). However, these methods only provide limited information about the quality of the expressed protein. Thus, several strategies to detect the functional spike protein were pursued in parallel and are summarized in Table 1. Already quite early in the pandemic, the monoclonal antibody CR3022 binding to the receptor binding domain (RBD) of SARS-CoV-2 spike protein was available. This mAb was initially discovered to be specific for SARS CoV (van den Brink et al., 2005) and is commercially available. To be independent of the global market, we also harvested culture supernatants from the early transfection/selection pool and of subclones of recombinant CR3022 CHO-K1 cells and cultivated them in fed-batch bioreactors to yield low but sufficient 10 mg/L of mAb titer at harvest. The ACE2 receptor was commercially available and used for analytical purposes. Using these and other commonly available reagents, various ELISA setups were tested and optimized and the ACE2/anti-His sandwich ELISA gave the best reproducible outcomes.

The pL-CSs construct was used for the generation of recombinant CHO-K1 cell lines stably expressing SARS-CoV-2 spike protein. Purified CSs protein from the initial pool experiment (Figure 1) was used as reference material to select the best producing CSs subclones, as shown in the following ELISA experiments (Figure 2). Parental clone 131A4 reproducibly gave the highest titers and good growth behavior and was selected as the leading parental cell clone.

Finally, the parental clone 131A4 was selected for limited dilution subcloning. Three subclones were selected by ELISA quantification of cell culture supernatants and further expanded into shake cultures to characterize growth behavior and productivity (Figure 3). Subclone C11 showed high growth rates of 0.8 days⁻¹ comparable with the parental clone and highest titers of 6 mg/L after 3 to 4 days of routine cultivation (Figures 3A, B). All clones showed the correct mass of the monomeric spike protein in Western blotting under reducing and denaturing conditions detected with anti-His or hyperimmune IgGs (Figure 3C).

The parental CSs clone 131A4 was used in an integrated 500 ml perfusion bioreactor directly connected to the Äkta Pure system equipped with a HisTrap Excel column (Figure 4) to generate large harvest volumes compensating for the relatively low product titer. The perfusion rate of 0.2 vessel volumes per day (vvd) was kept constant from days 3 to 8 followed by a stepwise increase to 2 vvd on day 11 to prevent nutrient limitation. In the steady-state of the perfusion, the viable cell density (VCD) reached 80 × 10⁶ viable cells/ml, maintained by a continuous cell bleed starting on day 8 (Figure 4A). This adjustment translates into optimized cell-specific perfusion rates between 10 and 40 pl medium per cell per day (pL/c/d) for minimized medium consumption and increased product titer. Integrated direct purification was initiated on day 5. The continuous perfusion harvest was mixed continuously with 50% (v/v) of phosphate buffer and loaded onto a HisTrap Excel column. This column minimizes metal stripping off the column caused by extended exposure to the cell culture medium. The first IMAC elution fraction was collected the day after loading and dialyzed against 20 mM Tris-HCl, 150 mM NaCl, pH 8.0. High-molecular weight species were visible in silver-stained SDS-PAGE (Figure 4B) that were enriched in the dialyzed product (lane 4) compared with the IMAC elution fraction (lane 3). These aggregates were partly resolved under reducing conditions (lane 9). Inefficient binding of the CSs protein led to significant amounts of CSs protein in the flow-through fraction (FT, lane 1 and 6) and wash fraction (WS2, lane 2 and 7). Thus, although the perfusion process in the bioreactor was successful and high performing concerning the cell culture parameters, further optimization would be required for the integrated downstream purification process for optimized yield and purity in the eluate fraction.

In another experiment, the final CSs clone C11 was cultivated in lab-scale fed-batch bioreactors. Various fed-batch strategies were tested to boost final spike protein titers in the culture supernatant. Valproic acid (VPA) is a small molecule compound that improves protein expression in mammalian cells, acting as an inhibitor of histone deacetylases (iHDAC). Furthermore, hypothermic cultivation temperatures limit cellular growth but increase cell-specific recombinant protein expression. In transient spike protein expression, a beneficial effect of lowered culture temperatures was demonstrated that led to titer enhancements of more than 10-fold in previous studies (Esposito et al., 2020). In another study, hypothermic conditions and VPA addition increased product titers by a factor of five in stable CHO pools (Johari et al., 2021). Using the selected production clone C11 in bioreactor fed-batch cultivation, we observed that viability dropped to 80% on day 11 at standard 37°C cultivation temperature (Figure 5A). Culture longevity was induced by a hypothermic temperature shift to 32°C after day 6 or a single shot of valproic acid on day 6. A maximum VCD of 20 × 10⁶ viable cells/ml was reached for both the standard 37°C and hypothermic 32°C cultures, but only 15 × 10⁶ viable cells/ml for the culture supplemented with VPA. Hypothermic conditions prolonged process duration up to 18 days at viabilities above 80%. Despite the induction of higher cell-specific productivities, VPA addition was not suitable to obtain higher CSs titers because of reduced VCD.

Interestingly, hypothermic treatment has induced exceptionally high product titers of up to 500 mg/L. Despite similar peak VCD, the temperature shift has induced a harvest titer 10-fold higher than the default fed-batch process of 37°C. In order to eliminate the possibility of artifacts in the ELISA quantification due to misfolded CSs from the hypothermic condition, equal supernatant volumes from the different fed-batch conditions were applied on an SDS-PAGE for selected culture days (days 3 and 9). An enriched product band confirmed higher product titers under the hypothermic fed-batch condition (Figure 5B, lane 7 and 14).

Alternatively, we developed a small-scale semi-continuous perfusion culture that keeps cells in a healthy state for prolonged periods at exceptionally high cell densities of 40 × 10⁶ viable cells/ml in uncontrolled 200-ml shake tubes. This procedure follows a daily medium exchange of one reactor volume per day and provides daily supernatant harvests of up to 15 mg/L spike protein (Figure 6). Compared with the fed-batch process, cells at higher VCD are kept at higher viabilities, and consequently, the best product quality can be expected.

Immobilized metal affinity chromatography (IMAC) was chosen as the preferred purification method of the CSs protein by using a chelating HisTrap column loaded with Ni-ions and a stepwise increased imidazole gradient. In the first experiment, elution buffer was increased in 5% increments and fractions analyzed by ELISA. In fractions with low imidazole concentration, the bound spike protein coelutes with host-cell proteins, whereas, at higher imidazole concentrations, the spike protein can be isolated in a pure form. Consequently, a stepwise elution profile at 20%, 40%, 70%, and 100% of 500 mM imidazole concentration was defined in 20 mM phosphate buffer with 150 mM NaCl, pH 7.4 and was used to purify the culture supernatant from the semi-perfusion process (Figure 7A). Different HisTrap elution fractions were analyzed by non-reducing SDS-PAGE/silver stain (Figure 7B), Western blotting with hyperimmune IgG (Figure 7C), and SEC-HPLC (Figure 7D ). This confirmed that CSs protein can be isolated with high purity from the 40% gradient step (fraction B9–C6) (Figure 7A).

IMAC purification shows several disadvantages concerning low binding affinity (Riguero et al., 2020), coelution of host-cell proteins, ion stripping, and incompatibility of cell culture media components (Farnós et al., 2020; Johari et al., 2021). Thus, many laboratories later switched to alternative affinity chromatography resin (e.g., Strep-Tactin). Although we developed a suitable purification protocol for highly pure spike protein, a significant fraction of the expressed protein was lost in the first wash step. Furthermore, significant amounts of CSs protein were detected in the flow-through. This was also observed by Stuible et al. (2021) for a Ni-Sepharose excel resin and attributed the reduced binding capacity to the size and/or shape of the spike construct, which may prevent diffusion into the pores. Further optimization of the IMAC procedure would be required to improve the final yield by optimizing the step gradient, choosing a gradient elution profile, or alternative purification methods (Cibelli et al., 2021).

The IMAC purified CSs protein was characterized in more detail by SDS-PAGE, Western blotting, ELISA, and N-glycan analysis (Figure 8 and Figure 9). A single defined band with the correct size of the CSs monomeric chain was observed in non-reducing SDS-PAGE and Western blotting using biotinylated hyperimmune IgG, CR3022, or anti-His mAbs (Figure 8A). The SEC profile showed the characteristic double peak with the correct molar mass of a fully glycosylated and trimeric CSs protein (Figure 8B). EC50 against ACE2 or CR3022 was determined at 0.1 and 0.4 nM, respectively (Figures 8C, D) and agrees with previously published literature (Tian et al., 2020; Walls et al., 2020). The purified CSs spike protein was then successfully applied in serological ELISAs against COVID negative and positive sera or vaccinated individuals. The produced spike protein generated a strong response for binding of anti-SARS mAb CR3022 and anti-SARS-CoV-2 hyperimmune IgG present in convalescent sera of COVID patients or vaccinees after one or two doses of the AstraZeneca vaccine (Figure 8E).

N-glycosylation was qualitatively analyzed by LC-ESI-MS/MS analysis (Figure 9). The identified glycosylation patterns are in agreement with previously published glycan patterns of the spike protein. Specifically, for sites 2, 4, 7, 11, 16, and 17, a significant fraction of oligomannose type was reported in the literature (Watanabe et al., 2020; Allen et al., 2021; Hoffmann et al., 2021).