ABSTRACT
We describe scalable and cost-efficient production of full length,
His-tagged SARS-CoV-2 spike glycoprotein trimer by CHO cells that can be
used to detect SARS-CoV-2 antibodies in patient sera at high specificity
and sensitivity. Transient production of spike in both HEK and CHO cells
mediated by PEI was increased significantly (up to 10.9-fold) by a
reduction in culture temperature to 32ºC to permit extended duration
cultures. Based on these data GS-CHO pools stably producing spike trimer
under the control of a strong synthetic promoter were cultured in
hypothermic conditions with combinations of bioactive small molecules to
increase yield of purified spike product 4.9-fold to 53 mg/L.
Purification of recombinant spike by Ni-chelate affinity chromatography
initially yielded a variety of co-eluting protein impurities identified
as host cell derived by mass spectrometry, which were separated from
spike trimer using a modified imidazole gradient elution. Purified CHO
spike trimer antigen was used in ELISA format to detect IgG antibodies
against SARS-CoV-2 in sera from patient cohorts previously tested for
viral infection by PCR, including those who had displayed COVID-19
symptoms. The antibody assay, validated to ISO 15189 Medical
Laboratories standards, exhibited a specificity of 100% and sensitivity
of 92.3%. Our data show that CHO cells are a suitable host for the
production of larger quantities of recombinant SARS-CoV-2 trimer which
can be used as antigen for mass serological testing.
Keywords: bioproduction; SARS-CoV-2; COVID-19; spike trimer;
serological assay
Immune response represents the first line of defense against severe
acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection that has
caused the coronavirus disease 2019 (COVID-19) pandemic. The spike
glycoprotein that protrudes from the surface of the virus is highly
immunogenic with the receptor-binding domain being the target of many
neutralizing antibodies (Yuan et al., 2020). Utilizing a stabilized
version of the full-length SARS-CoV-2 spike protein, a very robust and
accurate serological enzyme-linked immunosorbent assay (ELISA) for
antibodies in patient sera has recently been developed (Amanat et al.,
2020) and approved for use by the US FDA. However, very low production
titers (1–2 mg/L) of the spike trimer were reported using the human
embryonic kidney (HEK) Expi293 Expression system (Esposito et al.,
2020), therefore effectively limiting its widespread utilization as a
preferred antigen in serological assays for COVID-19. The low production
titer is not surprising considering that the SARS-CoV-2 spike is a large
homotrimer (~600 kDa) with 22 N -linked
glycosylation sites per monomer (Watanabe et al., 2020). In this work,
we describe the development of spike manufacturing platforms utilizing
Chinese hamster ovary (CHO) cells as a preferred production host with
established gene amplification methods and improved engineering
strategies. Transient expression was initially employed to fast-track
the production of CHO spike to perform biophysical analyses and early
clinical evaluation of the material, as well as to evaluate product
manufacturability and refine process conditions. Even though CHO cells
possessed a higher amount of heparan sulfate proteoglycans (HSPGs)
compared to HEK cells (Lee et al., 2016) we developed an effective
affinity purification procedure, and further validated the CHO-derived
spike serological assay for COVID-19 antibody.
We have previously shown that for difficult-to-express (DTE) proteins,
transient production processes need to be tailored to negate the
protein-specific negative effects of recombinant gene overexpression in
host cells (e.g., unfolded protein response (UPR) induction, limited
cell-specific productivity (qP); Johari et al., 2015). Using the plasmid
construct from the Krammer Laboratory (Amanat et al., 2020), HEK
Expi293F and CHO-S cells were transiently transfected with the
CAG-driven expression plasmid using PEI at an optimal gene dosage for
spike production in both cases (data not shown). Further, we utilized a
mild hypothermic condition, an effective process engineering
intervention for production of DTE proteins (e.g., Estes et al., 2015;
Johari et al., 2015) and to extend culture longevity. As shown in Figure
1A, the qP of HEK cells increased 2.4-fold from 0.20 pg/cell/day to 0.48
pg/cell/day when the culture temperature was lowered from 37°C to 32°C.
Additionally, the prolonged batch culture duration at 32°C (Figure 1B)
enabled a 4.1-fold increase in titer, yielding 10.2 mg/L of purified
spike. Greater enhancement was observed with CHO cells where mild
hypothermia resulted in an 8.5-fold higher qP than that at 37°C, and a
further increase in titer (10.9-fold, 5.4 mg/L) was obtained via
increased cell accumulation (Figure 1A,B). We anticipate that improved
CHO systems (e.g., ExpiCHO-S cell line and ExpiCHO medium) as well as
co-expression of genetic effectors and chemical chaperone addition
(Cartwright et al., 2020; Johari et al., 2015) would significantly
increase spike transient production in CHO cells. The CHO-derived spike
exhibited a monomeric molecular mass of ~190 kDa by
SDS-PAGE (Figure 1C) and a trimeric mass of 619 kDa was measured using
analytical size exclusion chromatography (Supplementary Figure S1). The
material was further validated using peptide mapping in conjunction with
mass spectrometry analysis (Supplementary Figure S2, Supplementary Table
S1). Critically, the preliminary COVID-19 antibody serological test
demonstrated its suitability for the ELISA (data not shown) thus
permitting development of CHO stable production platform.
For DTE proteins, very low yielding transient expression systems can be
an early indication of reduced stable production (Mason et al., 2012),
where particular engineering strategies may be required to obtain stable
cells with desirable production characteristics. To generate CHO cells
stably expressing recombinant spike trimer, we tested two in-house CHO
synthetic promoters, namely 40RPU (~90% CMV activity)
and 100RPU (~220% CMV activity) promoters (see Brown et
al., 2017; Johari et al., 2019). Although the use of extremely strong
promoters may be counterintuitive for DTE proteins, we reasoned that
only those transfectants harboring a sub-UPR threshold productivity, and
thus capable of proliferation would survive. Thus, if cell proliferation
attenuation and apoptosis occur as ER functional capacity is exceeded,
this is a condition that would directly deselect poorly performing
stable transfectants. The promoters and spike gene were inserted into a
vector construct encoding glutamine synthetase (driven by an SV40
promoter) and the electroporated CHO-S host cells were subjected to a
single round of selection at 25 or 50 μM methionine sulfoximine (MSX),
using suspension culture. After 19 days, recovered CHO cell populations
were screened for the ability to produce spike in 3-day batch culture
(Figure 3A). These data showed that transfectant pools derived from
genetic constructs harboring the strong 100RPU promoter expressed
recombinant spike whereas those using the weaker 40RPU promoter did not.
More stringent selection conditions (50 µM MSX) yielded transfectant
pools exhibiting lower productivity. Together, these data imply that
higher glutamine production may protect cells constitutively expressing
DTE spike trimer (e.g. via glutathione production to alleviate oxidative
stress). Accordingly, CHO cell pools harboring the 100RPU promoter under
25 μM MSX were taken forward for the manufacturing process.
In order to rapidly produce recombinant spike, stable transfectant pools
(rather than clonally derived populations) were employed. Based on CHO
transient process data (Figure 1), we tested the hypothesis that an
optimal 10-day fed-batch stable production process could be executed at
32°C and further enhanced by chemical chaperone additives (e.g., Johari
et al., 2015). We compared this strategy to an alternative approach
utilizing reduced culture temperature implemented at maximal cell
density (biphasic), as well as constant 37°C as control. These data are
shown in Figure 2B and C, whilst the screening data for 8 small molecule
chemical additives is shown in Supplementary Figure S3. Compared to the
37°C control culture, hypothermia resulted in a clear (33%) initial
reduction in cell-specific proliferation rate over the first 5 days of
culture (Figure 2B). However, the qP of the latter was 3.4-fold higher
over control and addition of valproic acid (VPA) at Day 6 further
enhanced qP 5.1-fold, yielding 51 mg/L of spike after purification by
immobilized metal affinity chromatography (IMAC; Figure 2C). Similar
enhancement was observed with betaine although there was no synergistic
effect when the two molecules were utilized together. Reduction in
culture temperature after 3 days culture improved the integral of viable
cell density (IVCD) 1.4-fold and when combined with VPA addition at Day
4, 53 mg/L of spike was attained after IMAC purification. These data
demonstrate that the optimal process engineering intervention for
recombinant spike production identified for rapid transient gene
expression was translatable to the stable production process. Further,
as low-level, sub-UPR threshold expression is likely required to permit
adequate cell growth, we reasonably expect that application of mammalian
inducible expression technology (e.g. cumate; Poulain et al., 2017) to
switch on spike production using an intensified biphasic culture system
would be particularly useful to maximize stable production.
To purify spike protein from culture supernatant, IMAC was initially
performed using a step-elution of 250 mM imidazole according to
Stadlbauer et al. (2020). Figure 3A shows SDS-PAGE of eluted proteins,
and reveals the presence of protein impurities not derived from
recombinant spike (Figure 3B), which were identified using tandem mass
spectrometry as CHO host cell derived proteins (HCPs; see Supplementary
Table S1). Whilst all of the identified extracellular HCPs have
previously been shown to be present in CHO cell culture supernatant
(Park et al., 2017), HSPG has been reported to occur in CHO cells at
relatively higher level than HEK cells (Lee et al., 2016) and shown to
interact with spike protein via heparin binding (Mycroft-West et al.,
2020). In order to increase recombinant spike purity, a revised gradient
elution profile up to 250 mM imidazole was implemented. As shown in
Figure 3C, HCPs were eluted at a lower imidazole concentration than
recombinant spike, permitting recovery of high purity
(>95%) product for use in serological assays (Figure
3D).
COVID-19 antibody tests would help reveal the true scale of the pandemic
in a population and the persistence of immunity, whether vaccines (many
of which are based on the production of neutralizing antibodies against
spike protein) designed to protect from infection are effective, as well
as identify highly reactive human donors for convalescent plasma
therapy. The CHO-spike anti-SARS-CoV-2 ELISA was developed based on the
Krammer Laboratory’s assay, and validated to ISO 15189 Medical
Laboratories standards. Initially, we tested a panel of 234 negative
samples taken pre-COVID-19 outbreak (June–August 2019) and 26 positive
samples taken during the COVID-19 outbreak (≥15 days post-positive PCR
test). ELISAs were performed by 1/20 dilution of the individual serum
samples and the cut-off index of 1.4 was determined using the cut-off OD
value (ROC curve with 100% specificity) and the negative control. In
this particular evaluation, the assay had an overall specificity of
100% and sensitivity of 92.3% as illustrated in Figure 4A. To
establish the reproducibility of the ELISA, a positive sample was tested
on 5 separate assays over 2 days at 3 different dilutions to determine
the inter-assay variations. The data (Figure 4B) shows that the assay
performed within the standard range for precision with inter-assay %CV
of ≤5%. To be able to interpret serosurveys correctly, the ELISA was
evaluated for potential cross-reactivity from individuals with other
medical conditions where zero positives were observed in all cases
(Supplementary Table S2).
Overall, our work serves as an exemplar for a development process of
characteristically difficult-to-express spike manufacturing platform
utilizing CHO cells. This itself is a significant and useful finding, as
many DTE recombinant proteins cannot be produced using this industry
standard production vehicle — e.g., a recent study reported that for
over 2,200 human genes encoding secreted proteins expressed in CHO
cells, almost 50% did not yield target protein (Uhlen et al., 2018). On
the other hand, the low spike production in HEK cells was highly
dependent on the very expensive Expi293 medium (we note that spike
production using FreeStyle 293 medium resulted in an even lower titer
(<40%); data not shown). With one mg of spike providing
serological assays for approximately 3,500 patient samples, the rapid,
scalable transient platform was adequate for local population antibody
tests and research studies. To enable large, constant clinical supply of
spike, we showed that it was possible to generate CHO stable
transfectants expressing the very complex glycoprotein, whilst high
titers could be achieved via a combination of process engineering
approaches designed for both high qP and cell biomass accumulation. The
refinement of the IMAC affinity purification process permitted greatly
enhanced purity of the CHO spike product following an extended 10-day
culture, ensuring suitability for use in serological immunity testing.
The assay has been implemented at local hospitals with
~7,200 staff tested (as of 31 July 2020) which resulted
in ~16% positive COVID-19 antibody detection, thus
supporting the global effort to limit and mitigate the impact of
SARS-CoV-2. Furthermore, it is highly likely that the cell and process
engineering interventions designed for SARS-CoV-2 spike production is
generically applicable to spike from different coronavirus strains.