An inactivated SARS-CoV-2 vaccine based on a Vero cell culture-adapted high-titer virus confers cross-protection in small animals

Abstract

Rapidly waning immunity against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) requires continued global access to affordable vaccines. Globally, inactivated SARS-CoV-2 vaccines have been widely used during the SARS-CoV-2 pandemic. In this proof-of-concept study we adapted an original-D614G SARS-CoV-2 virus to Vero cell culture as a strategy to enhance inactivated vaccine manufacturing productivity. A passage 60 (P60) virus showed enhanced fitness and 50-fold increased virus yield in a bioreactor compared to the original-D614G virus. It further remained susceptible to neutralization by plasma from SARS-CoV-2 vaccinated and convalescent individuals, suggesting exposure of relevant epitopes. Monovalent inactivated P60 and bivalent inactivated P60/omicron BA.1 vaccines induced neutralizing responses against original-D614G and BA.1 viruses in mice and hamsters, demonstrating that the P60 virus is a suitable vaccine antigen. Antibodies further cross-neutralized delta and BA.5 viruses. Importantly, the inactivated P60 vaccine protected hamsters against disease upon challenge with original-D614G or BA.1 virus, with minimal lung pathology and lower virus loads in the upper and lower airways. Antigenicity of the P60 virus was thus retained compared to the original virus despite the acquisition of cell culture adaptive mutations. Consequently, cell culture adaptation may be a useful approach to increase yields in inactivated vaccine antigen production.

Introduction

Worldwide, the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic had major socioeconomic impacts. SARS-CoV-2 (original: Wuhan-Hu-1 isolate) was identified in late 2019, and SARS-CoV-2 with the D614G mutation in the spike protein (referred to as original-D614G in this study) became dominant during the spring of 2020. Since then, variants have continuously emerged due to adaptation to the human host and immune escape^1,2. A number of effective vaccines based on different technologies were developed against original SARS-CoV-2 and received regulatory approval, including inactivated, mRNA, viral vector, and protein subunit vaccines³. Inactivated vaccines developed by Sinovac Biotech (CoronaVac) and Sinopharm (BBIBP-CorV) were among the first to be widely distributed^4,5. Importantly, they were made available also in low-income countries otherwise experiencing poor vaccine accessibility. A large proportion of the Coronavirus disease 2019 (COVID-19) vaccine doses administered by late 2021 were thus inactivated vaccines, which are generally cheap, easy to produce, and have less strict cold chain requirements than other types of vaccines^6,7. Historically, inactivated vaccines have proven effective against various virus pathogens, including hepatitis A virus, tick-borne encephalitis virus, and rabies virus, with good safety profiles³. The inactivated COVID-19 vaccines were reported to be safe with vaccine effectiveness of 65–83% against symptomatic infection across different vaccine trials^4,8,9. Neutralizing antibodies are a correlate of protection against COVID-19, and the spike protein is the main target of neutralizing antibodies. Mutations acquired in the spike protein have thus affected vaccine effectiveness against emerging SARS-CoV-2 variants¹. As observed for other COVID-19 vaccines based on original SARS-CoV-2, inactivated vaccine effectiveness was significantly reduced against the initial omicron variants BA.1 and BA.2, circulating when this study was initiated, as well as against emerging omicron sub-lineages. Three doses of inactivated vaccines based on original SARS-CoV-2 were reported to have an effectiveness of at least 68% against severe COVID-19 and death during waves of omicron BA.1 and BA.2^10,11. Studies conducted during waves of omicron sub-lineages BF.7 and BA.5 also reported up to 65% of protection^12,13.

Immunity induced by vaccination against COVID-19 and by infection wanes after several months. Consequently, booster vaccination programs against COVID-19 were implemented by the end of 2021. Appropriate vaccine accessibility thus remains important, in particular for populations at high risk of severe complications, such as the elderly¹⁴.

In this proof-of-concept study, as a strategy to enhance inactivated vaccine production, we adapted an original-D614G SARS-CoV-2 to grow to high titers in the World Health Organization (WHO) vaccine cell line VeroCCL81. Virus fitness, virus yields obtained from a small scalable bioreactor culture, and susceptibility to neutralization by plasma from SARS-CoV-2 vaccinated and convalescent individuals were evaluated for a high-titer polyclonal virus and a recombinant virus containing its specific cell culture adaptive mutations. To investigate if the high-titer virus was suitable as a vaccine antigen despite mutations in the structural proteins, we studied immunogenicity and protection conferred by an inactivated vaccine based on this virus, including a bivalent vaccine additionally containing an inactivated omicron BA.1 virus, in mice and hamsters.

Results

Accumulation of mutations during serial passaging in Vero cells

The original SARS-CoV-2 with the D614G mutation in the spike protein (original-D614G)¹⁵ was serially passaged in VeroCCL81 cells (African green monkey kidney cell line). From passage (P) 4 onwards, this was done in serum-free conditions, and supernatants from P4, P23, P37, P59, P70, and P88 were analyzed by deep sequencing. Analysis of missense mutations revealed that in P59, 11 nucleotide changes encoding 10 amino acid changes became dominant in the viral population at a frequency of ≥ 50%. Of these amino acid changes, there were 6 in the spike protein, 1 in the envelope protein, 2 in non-structural protein (nsp) 3, and 1 in nsp15 (Table 1 and Supplementary Text). In the P70 viral population, there was 1 additional change at a frequency of > 50% in nsp3, 1 in nsp9, and 3 in spike. Only two additional changes reached a frequency of ≥ 50% in the P88 population, 1 in nsp5 and 1 in nsp14. Some of the low-frequency changes observed in P59 remained at < 50% through almost 30 passages until P88. The only change in the receptor binding domain (RBD) of the spike protein had a frequency of 21–27% in virus populations in P59, P70, and P88 and encoded the amino acid change E484D.

Table 1 Missense mutations acquired by SARS-CoV-2 during serial passaging.

Characterization of the Vero cell-adapted SARS-CoV-2 virus

The peak infectious titer of the P59 virus was several-fold higher than that of the initial P4 virus. We therefore prepared a P60 virus stock for characterization and bioreactor-based virus production. For vaccine manufacturing, a well-defined recombinant virus is considered more attractive than a polyclonal virus stock. Therefore, we also engineered a recombinant virus containing the missense mutations present at a ≥ 50% frequency in P70, the P70 recombinant (P70rec) virus. All mutations included in the P70rec virus were reported in the GISAID database of patient isolate sequences, but some at very low prevalence (Supplementary Table S1)¹⁶. The P70rec virus was genetically stable, maintaining all mutations inserted in the plasmid during 4 passages in cell culture while not acquiring any new mutations at a frequency of ≥ 5% (Supplementary Table S1).

To characterize the P60 and P70rec viruses, we first investigated their fitness compared to the original-D614G virus in a kinetic experiment, also including a polyclonal P71 virus for comparison. The polyclonal P71 virus stock had the same mutations and similar mutation frequencies as the P70rec virus. From day 2 post-infection (dpi), infectious titers were highly similar for P60, P70rec, and P71 viruses, and their peak infectious titers of 8.7–8.8 log₁₀ 50% tissue culture infectious doses (TCID₅₀)/mL were 117–162-fold higher than that of the original-D614G virus, which had a peak infectious titer of 6.6 log₁₀ TCID₅₀/mL (Fig. 1a). Further, virus RNA titers at 2–3 dpi were highly similar for P60, P70rec, and P71 viruses, and their peak RNA titers of 11.1–11.5 log₁₀ copies/mL were around 16–38-fold higher than that of the original-D614G virus, which had a peak RNA titer of 9.9 log₁₀ copies/mL (Fig. 1b). Moreover, P60, P70rec, and P71 viruses showed similar cytopathic effect (CPE) at 1–3 dpi (Fig. 1c). However, the CPE was more pronounced for polyclonal P60 and P71 than for the recombinant P70rec virus after dpi 3. In an independent experiment, the CPE data were validated with images and luminescence-based cell viability assays (Supplementary Fig. S1).

Figure 1

Comparison of the SARS-CoV-2 polyclonal original-D614G, passage 60 and 71, and the recombinant P70rec viruses. (a–c) VeroCCL81 cells were infected at a multiplicity of infection of 0.001 with polyclonal passage (P) 4 (original-D614G virus), P60, or P71 virus stocks or the recombinant virus P70rec (n = 1 culture per virus), supernatant was collected at day 1–6 post-infection, and cultures were evaluated for (a) supernatant infectious titers (50% tissue culture infectious dose (TCID₅₀)/mL) with dots representing geometric means of four technical replicates with standard deviations, (b) supernatant virus genome titers (RNA copies/mL), and (c) cytopathic effect evaluated by microscope inspection. (d) Original-D614G, P60, and P70rec viruses as specified above the graphs were produced in a fixed-bed bioreactor. Infectious titers were determined for the virus-containing medium collected upon each medium-replacement, bars represent geometric means of two to three technical replicates with standard deviations (left y-axis). The accumulated yield of TCID₅₀ units obtained from each culture was calculated based on the infectious titers and the volume of collected virus-containing medium (right y-axis). (e) 50% inhibitory dilution (ID₅₀) values against P60, P70rec, and P71 viruses for plasma from infection-naïve SARS-CoV-2 vaccinated (n = 6) or vaccination-naïve convalescent (n = 6) individuals. (f) 50% inhibitory concentration (IC₅₀) values for neutralization by monoclonal antibody bebtelovimab. ID₅₀ and IC₅₀ values were calculated based on curves shown in Supplementary Fig. S2. For (e) the three groups were compared in a Friedman test (line) with a post-hoc Dunn’s multiple comparison test (line with brackets). **p < 0.005; ns, p > 0.05. LLOQ, lower limit of quantification. #, cytopathic effect not evaluated due to low numbers of cells because of SARS-CoV-2 induced cell death.

Further, the P60 and P70rec viruses were compared in bioreactor-based virus productions¹⁷. Compared to a production of the original-D614G virus reaching a peak infectious titer of 7.6 log₁₀ TCID₅₀/mL and a total yield of 5.8·10¹⁰ TCID₅₀ units, the P60 and P70rec viruses had higher peak infectious titers of 9.2–9.3 log₁₀ TCID₅₀/mL and similar total yields of 270–300·10¹⁰ TCID₅₀ units, around 50-fold higher than that of the original-D614G virus (Fig. 1d).

As spike protein mutations may affect neutralization susceptibility¹, we also tested if P60, P70rec, and P71 viruses were similarly sensitive to neutralization by plasma from SARS-CoV-2 vaccinated or convalescent individuals in cell-based in vitro neutralization assays. Using plasma from vaccinated individuals (prime boost with original mRNA vaccine, Comirnaty, Pfizer/BioNTech¹⁸), the mean 50% inhibitory dilution (ID₅₀) values were 1436, 588, and 1966 for P60, P70rec, and P71 viruses, respectively. For plasma from convalescent individuals, hospitalized with COVID-19 (exposure to original SARS-CoV-2 with the D614G mutation in the spike protein¹⁹), the mean ID₅₀ values were 5285, 3778, and 4269, for P60, P70rec, and P71 viruses, respectively, showing no major differences between the three viruses (Fig. 1e). Additionally, they were similarly susceptible to neutralization by the therapeutic monoclonal antibody (mAb) bebtelovimab (Fig. 1f)²⁰.

Vero cell-adapted P60 and P71 viruses and the corresponding engineered P70rec virus thus showed similarly enhanced fitness compared to the original-D614G virus, also resulting in higher yields in bioreactor-based virus production for P60 and P70rec viruses. There were no major differences between P60, P70rec, and P71 viruses in susceptibility to neutralization by bebtelovimab or plasma from vaccinated or convalescent individuals, although P70rec showed a slightly reduced sensitivity to neutralization compared to P71 by plasma from individuals vaccinated against COVID-19.

Mono- and bivalent inactivated vaccines based on the P60 virus were immunogenic in mice

Next, the P60 virus was tested as vaccine antigen. Mice were immunized with a SARS-CoV-2 vaccine based on inactivated P60 virus (I-P60) administered with the MF59-like adjuvant AddaVax. At the time of the study design, there was a focus on bivalent vaccine formulations, and the I-P60 was additionally tested in a bivalent vaccine regimen. For the bivalent vaccine, inactivated omicron BA.1 virus (I-BA.1) was selected as the second antigen, as this was the dominating variant at the time²¹. The P60 and BA.1 viruses were produced in a small fixed-bed bioreactor, inactivated, and purified as described previously²² to generate the I-P60 and I-BA.1 antigen preparations. Serum was collected throughout the immunization experiment as indicated in Fig. 2a, and serum neutralization of the original-D614G virus was first evaluated. The I-P60 and I-P60/BA.1 immunized groups showed similar neutralization capacities. For these groups, mean ID₅₀ values increased after each immunization, reaching 2759 and 2275, respectively, at day 56 of the experiment (Fig. 2b). The group immunized with I-BA.1 alone showed no neutralization of the original-D614G virus at day 56. Neutralization of the BA.1 virus was subsequently analyzed. At day 56, the I-BA.1 and I-P60/BA.1 immunized groups reached mean ID₅₀ values of 453 and 185, respectively. The I-P60 immunized group reached a mean ID₅₀ value of 815, however, if excluding the one animal with an extraordinarily high ID₅₀ value against BA.1 in repeated independent assays, the mean ID₅₀ value was 89. Cross-neutralization was investigated for samples collected at day 56. Serum from I-P60 and I-P60/BA.1 immunized groups both neutralized the delta virus, and at least one animal per group neutralized the omicron sub-lineage BA.5, while serum from the I-BA.1 immunized group did not cross-neutralize these viruses. Serum samples from ovalbumin protein (OVA) immunized animals were tested against the original-D614G virus and showed no neutralization.

Figure 2

Neutralizing responses induced by SARS-CoV-2 mono- and bivalent inactivated vaccines in mice. (a) Mice were immunized three times with monovalent I-P60 (n = 4) or I-BA.1 (n = 4), bivalent I-P60/BA.1 (n = 4), or Ovalbumin (OVA) (n = 4) on day 0, 21, and 42, and serum was sampled on day 0, 14, 37, and 56. (b) Neutralization of homologous and heterologous viruses by serum was analyzed (left and right graphs for each group, respectively). In vaccines containing I-P60, original-D614G virus was used as the homologous virus in this assay. Dots represent values from individual animals, bars represent group means. < indicates that at least one value used to calculate the mean is below LLOQ; the LLOQ was determined as the first dilution tested in the assay, 1:12.5. Statistical analysis was only carried out between groups where all values were above the LLOQ. A Kruskal–Wallis test was done to compare more than two groups with the post-hoc Dunn’s multiple comparisons test (I-P60, I-P60/BA.1, and I-BA.1 serum from day 56 against BA.1). A Mann–Whitney test was used to compare two groups (at each timepoint, ID₅₀ values against each virus were compared for the inactivated vaccine groups). None of the tests showed significant differences. ID₅₀ values were calculated based on curves shown in Supplementary Fig. S3. ID₅₀, 50% inhibitory dilution. LLOQ, lower limit of quantification.

In line with previous findings²², antigen re-stimulation of splenocytes followed by cytokine profile analysis showed vaccine antigen-specific responses for interferon (IFN)-γ, interleukin (IL)-10, IL-5, and tumor necrosis factor (TNF)-α, suggesting a mixed Th1/Th2-type T cell response (Supplementary Fig. S4), despite some background cytokine secretion in the inactivated vaccine groups (I-P60, I-P60/BA.1, and I-BA.1)²². The OVA immunized group showed low levels of TNF-α secretions when stimulated with inactivated vaccine antigen.

Mice immunized with I-P60 inactivated vaccine antigen thus showed serum neutralization against the original-D614G virus and BA.1, both alone and in a bivalent preparation with I-BA.1, as well as cross-neutralization of the delta and BA.5 viruses. Mice immunized with the I-BA.1 vaccine antigen showed serum neutralization of the homologous virus but without cross-neutralization. However, similar patterns of cytokine secretion upon stimulation with inactivated original-D614G virus, I-P60, or I-BA.1 were observed for the three inactivated vaccine groups.

Mono- and bivalent I-P60 vaccines protect hamsters from weight loss and lung tissue damage

The protective potential of the I-P60 monovalent and I-P60/BA.1 bivalent inactivated vaccines were evaluated in a hamster infection model. First, infectious challenge with the original-D614G virus and omicron BA.1 was investigated in a pilot study (Fig. 3a). Using equivalent infectious doses of 2000 TCID₅₀ units for the two viruses resulted in detectable infectious titers in swab and lung samples, with lower titers in animals challenged with BA.1 (Fig. 3b–c). Accordingly, weight loss was only detected in the group challenged with the original-D614G virus (Fig. 3d).

Figure 3

Investigation of SARS-CoV-2 challenge dose in hamsters according to weight and infection. (a) Hamsters were challenged on day 0 with 2000 TCID₅₀ units of the original-D614G virus (n = 4) or the BA.1 virus (n = 4) or remained non-challenged (n = 4). Oropharynx swabs were obtained every day after challenge, lungs were collected on day 5 upon euthanasia. (b) Virus genome titers of oropharynx swabs. (c) Virus genome titers in lungs. For (b) and (c) symbols represent individual animals and lines represent group medians. < indicates that at least one value used to calculate the median is below LLOQ. (d) Body weight relative to the day of challenge, symbols represent group means with standard deviation, and symbols for the non-challenged group were nudged by − 0.1 on the x-axis. Statistical analysis was only carried out between groups where all values were above the LLOQ. For (b) and (c), a Mann–Whitney test was used to compare two groups. Only statistically significant differences are shown. For (d) a Kruskal–Wallis test (significance indicated above the symbols) was used at each timepoint; a post-hoc Dunn’s multiple comparison test showed that the significant difference was between the original-D614G and the non-challenged groups. *p < 0.05. LLOQ, lower limit of quantification.

A dose of 2000 TCID₅₀ of original-D614G and 6000 TCID₅₀ of BA.1 was selected for the vaccine experiment. Hamsters were immunized twice with either I-P60 or I-P60/BA.1 prior to challenge with the original-D614G virus or BA.1 on day 44 after the first immunization (Fig. 4a). Control animals were immunized with OVA and similarly either challenged or not challenged. In contrast to the non-challenged OVA immunized group, the OVA immunized group challenged with the original-D614G virus showed progressive weight loss from the day after challenge and until day 49 (day 5 post-challenge) when animals were euthanized, whereas the OVA immunized group challenged with BA.1 followed a similar pattern until day 47 and then started to regain weight (Fig. 4b). Compared to the respective OVA immunized groups, immunization with either I-P60 or I-P60/BA.1 resulted in reduced weight loss following challenge with either the original-D614G or the BA.1 virus. I-P60 or I-P60/BA.1 immunized groups challenged with the original-D614G virus showed a minor dip in body weight on day 46–47 but showed similar weight as the non-challenged OVA immunized group on day 48–49, and the I-P60 or I-P60/BA.1 immunized groups challenged with BA.1 showed no weight loss.

Figure 4

Weight and tissue pathology in inactivated vaccine-immunized hamsters following SARS-CoV-2 challenge. (a) Hamsters were immunized twice with monovalent I-P60 (n = 8), bivalent I-P60/BA.1 (n = 8), or OVA (n = 12) on day 0 and 21, and plasma was sampled on day 0, 21, 42, and 49. On day 44, hamsters were challenged with 2000 TCID₅₀ units of the original-D614G virus or 6000 TCID₅₀ units of BA.1 (n = 4 per immunization group), or remained non-challenged (n = 4 for the OVA immunized group). (b) Body weight relative to the day of challenge, symbols represent group means with standard deviation. The non-challenged group is depicted in both graphs. (c) Lung inflammation scores determined as the percentage of lung tissue area affected by inflammation, dots represent animal means (determined from three tissue sections), bars represent group means also given as numbers above the bars. Mean scores for individual animals are given in Supplementary Table S2. (d) The number of animals per group with the indicated pathological changes in the lungs. Observations for individual animals are given in Supplementary Table S4. For (b) and (c) a Kruskal–Wallis test was used and significance is indicated above the symbols in (b) or as lines in (c) with a post-hoc Dunn’s multiple comparison test (line with brackets). Only statistically significant differences are shown. ****p < 0.0001; ***p < 0.0005; **p < 0.005; *p < 0.05. OVA, Ovalbumin.

Inflammation and pathology in the lungs were evaluated on the last day of the experiment, on day 49, 5 days after challenge. In the OVA immunized group challenged with the original-D614G virus, there were clear lung inflammatory responses, with a mean inflammation score of 25.5% (% area of analyzed lung tissue affected by inflammation) (Fig. 4c, Supplementary Table S2). By comparison, the inactivated vaccine groups showed mean inflammation scores of 0.01–0.1% after challenge with the original-D614G virus. The OVA immunized group challenged with BA.1 virus had a markedly lower score (0.4%) than the OVA immunized group challenged with the original-D614G virus (25.5%), with two animals having a score of 0%. Nevertheless, both inactivated vaccine groups had a two– to fourfold lower inflammation score than the OVA immunized group after BA.1 virus challenge. All animals of the OVA immunized group challenged with the original-D614G virus presented with sites of tissue necrosis, type II pneumocyte hyperplasia, and syncytial cell formation (Fig. 4d). Tissue necrosis and type II pneumocyte hyperplasia were observed in one and two animals, respectively, of the OVA immunized group challenged with BA.1. None of these pathological changes were observed in the inactivated vaccine groups.

Immunization with I-P60 or I-P60/BA.1 thus conferred protection against weight loss after challenge with either original-D614G or BA.1 virus in hamsters, and furthermore conferred protection from lung inflammation and tissue damage after challenge with the original-D614G virus. In line with other reports of reduced pathogenicity of the BA.1 virus^23,24, challenge with the BA.1 virus in this experiment resulted in limited lung pathology, and thus, protection against BA.1-induced lung pathology could not be exhaustively evaluated.

Hamsters immunized with mono- or bivalent I-P60 vaccines have reduced virus titers in upper and lower airways after challenge

For animals challenged with the original-D614G virus, upper airway virus RNA titers of the I-P60 immunized group were slightly reduced at all timepoints until euthanasia on day 49 (Fig. 5a). For the I-P60/BA.1 immunized group, virus RNA titers resembled those of the OVA immunized group on day 45–47 and dropped on day 48–49 compared to the OVA immunized group. For animals challenged with BA.1, the I-P60 and I-P60/BA.1 immunized groups had around 1 log₁₀ copies/mL lower median virus RNA titers on day 46–48 which dropped further on day 49 compared to the OVA immunized group (Fig. 5c).

Figure 5

SARS-CoV-2 titers in upper and lower airways of inactivated vaccine-immunized hamsters following SARS-CoV-2 challenge. (a–d) Oropharynx swabs were obtained daily following challenge on day 44, according to the experiment outlined in Fig. 4, until euthanasia on day 49. (a, c) Virus genome titers following challenge with the original-D614G (a) or BA.1 (c) virus. The OVA immunized non-challenged group is included in both graphs for comparison. Curves represent group medians and dots represent individual values, n = 4 for each group except for the BA.1 challenged OVA immunized group on day 46 with n = 3 due to low sample recovery. (b, d) Infectious titers following challenge with the original-D614G (b) or BA.1 (d) virus; I-P60 and I-P60/BA.1 symbols in (b) and (d) were nudged by +/− 0.04, respectively, on the y-axis. The OVA immunized non-challenged group is included in both graphs for comparison. (e) Virus genome titers in lung tissue obtained on day 49, n = 4 for each group, LLOQ is based on measurements from the OVA immunized non-challenged group (mean plus three times the standard deviation). < indicates that at least one value used to calculate the median is below LLOQ, colored according to group. Statistical analysis was only carried out between groups where all values were above the LLOQ. A Kruskal–Wallis test was used (significance indicated above the symbols) with a post-hoc Dunn’s multiple comparison test (line with brackets). A Mann–Whitney test was used when only two groups were above LLOQ (not significant). Only statistically significant differences are shown. *p < 0.05. OVA, ovalbumin. TCID₅₀, 50% tissue culture infectious dose. LLOQ, lower limit of quantification.

Evaluating upper airway infectious titers, for animals challenged with the original-D614G virus, the median titers in the I-P60 and I-P60/BA.1 immunized groups were more than 1 log₁₀ TCID₅₀/mL reduced compared to the OVA immunized group on day 45–46 and dropped to the lower limit of quantification (LLOQ) on day 47 in contrast to the OVA immunized group (Fig. 5b). For animals challenged with BA.1, the median titers of the I-P60 and I-P60/BA.1 immunized groups were 1 log₁₀ TCID₅₀/mL reduced compared to the OVA immunized group on day 46 and dropped below the LLOQ on day 47 (Fig. 5d).

Lung virus RNA titers were detected on day 49 in the OVA immunized group challenged with the original-D614G virus, and for three out of four animals in the OVA immunized group challenged with the BA.1 virus, whereas titers in all I-P60 and I-P60/BA.1 immunized animals were below the LLOQ (Fig. 5e).

Immunization of hamsters with I-P60 or I-P60/BA.1 thus led to faster clearance of infection after challenge with either the original-D614G virus or BA.1 when compared to the challenged OVA immunized groups. Furthermore, I-P60 or I-P60/BA.1 immunized hamsters showed reduced virus loads in the upper airways, and no detectable virus in the lower airways at day 49, day 5 post-challenge.

Mono- and bivalent I-P60 vaccines induce neutralizing responses in hamsters

Prior to challenge, plasma from hamsters receiving monovalent or bivalent inactivated vaccines neutralized the original-D614G virus with similar efficacy. The I-P60 immunized group had mean ID₅₀ values of 354 and 741 after the first and second immunization, respectively, and the I-P60/BA.1 immunized group had mean ID₅₀ values of 308 and 739 after the first and second immunization, respectively (Fig. 6a). Neutralization of the BA.1 virus was also similar in the two groups with mean ID₅₀ values below or close to the LLOQ after the first immunization, and at 169 and 100 after the second immunization for the I-P60 and I-P60/BA.1 immunized groups, respectively.

Figure 6

Neutralizing responses induced by SARS-CoV-2 mono- and bivalent inactivated vaccines in hamsters pre- and post-challenge. Plasma collected during the experiment outlined in Fig. 4 at the indicated time points were analyzed for neutralization of the original-D614G and BA.1 viruses. (a) Plasma collected prior to challenge (I-P60 group n = 8, I-P60/BA.1 group n = 8, OVA group n = 12). (b) Plasma collected after challenge, n = 4 for each group. Dots represent values from individual animals, bars represent group means. < indicates that at least one value used to calculate the mean is below LLOQ; the LLOQ was determined as the first dilution tested in the assay, 1:25. Statistical analysis was only carried out between groups where all values were above the LLOQ. For (a), a Mann–Whitney test was done to compare groups (I-P60 against original-D614G vs. I-P60/BA.1 against original-D614G on day 21 and day 42). None of the tests showed significant differences. For (b), a Kruskal–Wallis test was used (line) with a post-hoc Dunn’s multiple comparison test (line with brackets). ID₅₀ values were calculated based on curves shown in Supplementary Fig. S5. *p < 0.05; ns, p > 0.05. ID₅₀, 50% inhibitory dilution. LLOQ, lower limit of quantification. OVA, ovalbumin.

Challenge of the inactivated vaccine groups with either original-D614G or BA.1 virus led to increased neutralization titers (Fig. 6b). After challenge with the original-D614G virus, mean ID₅₀ values against this virus reached 2739 and 2290 for the I-P60 and I-P60/BA.1 immunized groups, respectively. After challenge with the BA.1 virus, mean ID₅₀ values against the original-D614G virus reached 1688 and 1442 for the I-P60 and I-P60/BA.1 immunized groups, respectively. Similarly for neutralizations of BA.1, there were no major differences between the inactivated vaccine groups. Mean ID₅₀ values were 1819 and 823, respectively, for I-P60 and I-P60/BA.1 immunized groups after challenge with the original-D614G virus, and 891 and 990, respectively after challenge with the BA.1 virus.

The OVA immunized group challenged with the original-D614G virus reached a mean ID₅₀ value of 427 against the original-D614G virus, with no detectable cross-neutralization of the BA.1 virus (Fig. 6b). In contrast, the OVA immunized group challenged with BA.1 did not mount any detectable neutralization against BA.1 or the original-D614G virus.

I-P60 and I-P60/BA.1 vaccines thus induced similar levels of neutralizing responses in hamsters, with higher neutralization titers observed against the original-D614G virus than against the BA.1 virus. Following challenge with either original-D614G or BA.1 virus, neutralization titers against both viruses were boosted in the inactivated vaccine groups. In both inactivated vaccine groups, neutralization titers were better following two immunizations than after challenge alone. Challenge alone with original-D614G or BA.1 virus, when evaluated 5 days post-challenge, induced lower or no neutralizing responses, respectively, without observable cross-neutralization.

Discussion

This study provides proof-of-concept that an inactivated vaccine based on a high-titer, cell culture-adapted SARS-CoV-2 virus induced protective immune responses in small animals. The high-titer virus was based on an original-D614G isolate and the inactivated vaccine successfully induced neutralizing responses against the original-D614G virus, as well as against delta, BA.1, and to a lesser extent, BA.5 viruses. Importantly, hamsters immunized with a vaccine based on the high-titer virus and challenged with either original-D614G or BA.1 virus were protected from disease. This was evidenced by reduced weight loss, accelerated virus clearance, reduced virus load in the upper airways, and absence of virus in lower airways with minimal lung pathology at day 5 post-challenge in inactivated vaccine immunized hamsters when compared to OVA immunized hamsters. The monovalent high-titer virus vaccine and a bivalent vaccine additionally containing inactivated BA.1 virus performed similarly regarding induction of immune responses in mice and hamsters and conferred similar protection in hamsters.

SARS-CoV-2 commonly acquires mutations when passaged in cell culture, which has been shown to increase virus fitness²⁵. There is some overlap between detected mutations appearing during cell culture adaptation and in variants circulating in the human population. However, 10 out of the 15 mutations acquired during serial cell culture passaging in our study and included in the P70rec virus are observed below a 0.1% frequency in human samples (GISAID database, assessed Dec 5, 2023, sequences: 13,604,325)¹⁶. This likely reflects that immune evasion is an important driver of mutational escape in circulating variants, in contrast to infection in cell culture, where lack of immune surveillance drives viral evolution in a different direction²⁶. We found a unique set of mutations following serial passaging of the SARS-CoV-2/human/DNK/DK-AHH1/2020 isolate¹⁵ in VeroCCL81 cells which is further discussed in the Supplementary Text. These mutations mostly localized to the spike protein, but also to the envelope protein and non-structural proteins, and resulted in increased virus fitness and virus-induced CPE. Some of these mutations were previously observed by others^{15,25,27,28,29}, also reporting enhanced fitness and CPE during cell culture infection with serially passaged viruses^15,25,29.

Despite putative cell culture adaptive mutations in the spike protein, the P60, P70rec, and P71 viruses remained susceptible to neutralization by plasma from SARS-CoV-2 vaccinated and convalescent individuals. This could be expected given the absence of mutations in the spike protein RBD and is important when considering the utility of these viruses as vaccine antigens. They were additionally susceptible to the mAb bebtelovimab, one of the few therapeutic mAb also approved for treatment of the early omicron variants, with contact residues in the receptor binding-motif of the spike protein RBD²⁰.

The inactivated vaccine regimens tested in this study induced both neutralizing responses and cytokine responses indicative of cellular immune responses in small animals. Given that immunization with monovalent I-P60 resulted in neutralizing responses against the original-D614G, delta, BA.1, and to some extent, BA.5 viruses, the I-P60 preserved an important feature reported for other inactivated vaccines based on original SARS-CoV-2, which also induce neutralizing responses against early omicron variants BA.1, BA.2, and BA.5^30,31. Comparing neutralizing titers to those induced by commercially available vaccines across studies is challenged by differences in immunization regimens and neutralization assays. Nevertheless, reported data suggests that I-P60 induces similar levels of homologous neutralizing titers in BALB/c mice as the subunit vaccine NVX-CoV2373 (Novavax) (titer of ~ 2000³²), lower titers than the mRNA vaccine Comirnaty (Pfizer/BioNTech) (titer of ~ 43,000³³), and higher titers than commercially available inactivated vaccines CoronaVac (Sinovac Biotech) and Covaxin (Bharat Biotech) (titers of ~ 200–1500 and ~ 400, respectively^34,35). In hamsters, neutralizing titers induced by I-P60 were comparable to those induced by Comirnaty (titers of ~ 500–2700 in different studies^33,36), but lower than those reported for NVX-CoV2373 and Covaxin (titers of ~ 29,000 and ~ 10,000, respectively^37,38). Taken together, this suggests that I-P60 can induce clinically relevant immunological responses.

On its own, I-BA.1 induced lower neutralization titers than I-P60, and narrower neutralizing responses, and when combined with I-P60, it did not appear to contribute to the immunogenicity of the bivalent I-P60/BA.1 vaccine. BA.5 is antigenically different from the earlier omicron variants BA.1 and BA.2. This is illustrated by ~ tenfold lower neutralizing titers against BA.5 than against BA.1 following immunization with a monovalent BA.1 mRNA vaccine, as measured in a pseudovirus-infection assay³⁹. Together with the overall lower neutralizing titers induced by I-BA.1, as measured in a virus-infection assay, this could explain why no cross-neutralization of the omicron sub-lineage BA.5 was detected for serum from the I-BA.1 immunized group. When comparing the monovalent I-P60 and bivalent I-P60/BA.1 vaccines in hamsters, they conferred similar levels of protection upon BA.1 challenge, possibly due to the cross-protection conferred by I-P60 alone. Inactivated mono- and bivalent vaccines based on antigenically related early variants, e.g., original-delta or original-theta were also protective upon BA.1 challenge in animal models, both as mono- and bivalent vaccines⁴⁰.

In line with findings by others, we report that, in small animals, infection or vaccination with an omicron variant (in our case BA.1) result in lower homologous neutralization titers than infection or vaccination with previously circulating variants^23,31,41. In the context of infection, lower immunogenicity could be due to the reduced pathogenicity of BA.1²³. Indeed, we used a BA.1 challenge dose threefold higher than that of the original-D614G virus to obtain development of disease. Nevertheless, disease following challenge with the original-D614G virus was still more pronounced and resulted in induction of neutralizing responses in plasma, which was not observed after BA.1 challenge. Protection conferred by I-P60 and I-P60/BA.1 against lung pathology after BA.1 challenge could thus not be thoroughly evaluated. Considering the body weight loss following BA.1 challenge in hamsters, assessing lung pathology at day 3 after challenge might have been advantageous. It should be mentioned that the BA.1 virus stocks used in this study had a mutation at position 685 in the spike protein’s furin cleavage site, which has been reported to reduce pathogenicity of original SARS-CoV-2⁴².

While this study was ongoing, concerns were raised that omicron break-through infections and variant-updated bivalent booster vaccines induced lower neutralization titers against the circulating omicron variants than against original SARS-CoV-2^41,43. Importantly, repeated omicron exposure enhances omicron-specific antibodies^44,45. Further, as reported here, there is some evidence from another study in mice that primary immunization with an inactivated omicron vaccine resulted in slightly lower neutralizing responses than an inactivated original SARS-CoV-2 vaccine³¹. Compared to other types of vaccines, the antigen in inactivated vaccines resembles the infectious virus encountered by the immune system during infection, and these observations could point to lower immunogenicity of newer variants. Nevertheless, in preclinical studies with mRNA and protein vaccines, omicron-specific vaccines did not appear less immunogenic than original vaccines in the context of primary vaccination, and induced cross-neutralizing responses against non-homologous omicron variants, but with poor or undetectable neutralization of original SARS-CoV-2^32,39. Effective variant-specific immunogenicity of omicron vaccines likely depends on the mode of antigen presentation, dose, and previous SARS-CoV-2 exposures.

Aiming to achieve stronger immune responses against currently circulating variants, the WHO now recommends variant-updated monovalent boosters¹⁴. Following updated bivalent mRNA vaccines targeting original/omicron BA.1 (2022) and original/omicron BA.4/5 (2022) variants, a monovalent inactivated BA.1 vaccine was developed by Sinovac⁴⁶, and for 2023/2024, monovalent XBB.1.5 vaccines were developed by Moderna and Pfizer/BioNTech (mRNA vaccines), as well as by Novavax (subunit vaccine)^14,32.

In individuals previously exposed to original strains of SARS-CoV-2 through infection or vaccination against COVID-19, omicron-based monovalent vaccines, to some extent, boosted neutralization titers against both the original variant and homologous as well as non-homologous omicron sub-lineages^47,48, supporting a monovalent booster vaccination strategy. Nevertheless, as shown in this and other studies, neutralizing responses induced by primary omicron infection or vaccination show poor cross-neutralization of previously circulating variants^31,32,39,41. Bivalent vaccines could thus be relevant for SARS-CoV-2 naïve individuals in times when two highly antigenically divergent variants are co-circulating at high prevalence⁴⁹, such as during the transitions from delta to omicron, and from early omicron variants to XBB.1 and its sub-lineages.

In conclusion, SARS-CoV-2 could be further adapted to Vero cells while retaining good immunogenicity as an inactivated vaccine antigen. These observations could be of particular relevance for some of the newer SARS-CoV-2 variants, including the BA.1 virus, showing lower fitness than original SARS-CoV-2 in Vero cells, one of few cell lines used in the production of vaccines approved for human use^45,50. Additionally, cell culture-adaptive mutations may reduce SARS-CoV-2 pathogenicity, another advantage considering a vaccine production setting⁴². The need for serial passaging would increase response time, however, Vero cell adaptive mutations are acquired already within the first 10 passages^28,29,51. This approach to vaccine development could thus enable efficient manufacturing of variant-specific inactivated vaccines and represent a strategy to overcome vaccine shortage.

Methods

Cell lines

VeroE6 cells (RRID: CVCL_0574) (a gift from Jean Dubuisson, University of Lille) were maintained in Dulbecco’s Modified Eagle Medium (DMEM) (Thermo Fisher Scientific #31966021) supplemented with 10% fetal bovine serum (FBS) (v/v) (Sigma-Aldrich #F7524), 100 U/mL penicillin, and 100 µg/mL streptomycin (PS) (Sigma #P4333). Cells were sub-cultured every 2–4 days, using Trypsin (Sigma-Aldrich #T3924) for detachment, and maintained at 37 °C and 5% CO₂ as described elsewhere^15,52. VeroE6 cells were used for all cell culture-based experiments with the exceptions mentioned below.

VeroCCL81 cells (ECACC Cat #88020401, RRID:CVCL_JF53) (Nuvonis) were maintained in OptiPRO SFM medium (Thermo Fisher Scientific #12309019) supplemented with 4 mM GlutaMAX Supplement (Thermo Fisher Scientific #35050061) and PS, and kept at 37 °C and 5% CO₂ as described elsewhere¹⁷. Cells were sub-cultured every 2–4 days, using TrypLE Express (1x) (Thermo Fisher Scientific #12604013) for detachment, and maintained at 37°C and 5% CO₂ unless otherwise specified. VeroCCL81 cells were used for bioreactor-based virus production, serial passaging of original-D614G SARS-CoV-2, and kinetic experiments comparing original-D614G, P60, P70rec, and P71 viruses.

SARS-CoV-2 viruses

Cultured SARS-CoV-2 viruses isolated from patients as described¹⁵ were used in the study: Original SARS-CoV-2 with the D614G mutation in the spike protein (original-D614G) (SARS-CoV-2/human/DNK/DK-AHH1/2020; GenBank: MZ049597)¹⁵, delta (SARS-CoV-2/human/DNK/DK-AHH3/2021; GenBank: OP271297)²¹, omicron BA.1 (SARS-CoV-2/human/DNK/DK-AHH4/2021; GenBank: OP271296)²¹, and omicron BA.5 (SARS-CoV-2/human/DNK/DK-AHH6/2022; Genbank: OP722492)²¹. Second passage virus stocks were subjected to deep sequencing prior to use in experiments as described in the section “Deep sequencing”¹⁵. Virus stocks derived from SARS-CoV-2/human/DNK/DK-AHH1/2020 used in this study are detailed in Supplementary Table S3. The omicron BA.1 virus stock used to inoculate the bioreactor was a third passage serum-free stock derived from SARS-CoV-2/human/DNK/DK-AHH4/2021, with no changes in consensus sequence compared to the sequence deposited in the Genbank database (GenBank: OP271296).

Serial passaging of SARS-CoV-2

VeroCCL81 cells were seeded in a T25 flask (2·10⁶ cells) or a T80 (6.4·10⁶ cells) flask (Nunc Cell Culture Treated Flasks with Filter Caps, Thermo Fisher Scientific) the day prior to infection. Cultures were inoculated with 1–5 µL virus-containing supernatant. Cultures were inspected with a Primovert inverted cell culture microscope (Zeiss) (4 × objective) to monitor development of CPE. CPE was scored according to the amount of rounded and detached cells, and the integrity of the cell monolayer (1: detectable rounding or detachment of cells, 2: moderate levels of rounding or detachment of cells, 3: significant detachment of cells, or 4: massive detachment of cells and a clear disruption of the cell monolayer). After receiving a score of at least 2 at 2–4 days after infection, cell culture supernatant was collected for continued virus passaging and stored at – 80 °C, and cultures were closed. Selected passages were analyzed by deep sequencing as described in the section “Deep sequencing”. Virus stocks were prepared from selected passages. Briefly, VeroCCL81 cells were inoculated, and when the culture got a CPE score of 3, the supernatant was collected and 0.45 µm-filtered prior to storage at – 80 °C in smaller aliquots. Thus, the P60 virus stock was inoculated with supernatant from serial passage 59 and is independent of the serial passage 60 collected during virus passaging. The infectious titer of virus stocks was determined in replicate assays prior to use in experiments as described in the section “Virus infectious titers” and sequenced as described in the section “Deep sequencing”.

Reverse genetics

The recombinant virus was generated using a bacterial artificial chromosome (BAC) based reverse genetics system described previously^53,54. The plasmid representing the SARS-CoV-2/human/DNK/DK-AHH1/2020⁵³ isolate and custom synthesized DNA fragments (Genscript) (fragment 1: nucleotide positions 4325–12,804, fragment 2: nucleotide positions 19,667–26,272) containing the mutations selected from the P70 virus (Supplementary Table S1) were combined using an InFusion cloning kit (Takara #638948). The resulting plasmid was amplified and purified with the QIAprep Spin miniprep (Qiagen #27106) and HiSpeed Plasmid Maxi Kits (Qiagen #12663). Generation of in vitro transcribed RNA and transfections were carried out as described^21,53. Briefly, the mMESSAGE mMACHINE T7 Transcription Kit (Invitrogen #AM1344) was used for in vitro transcription, and VeroCCL81 cells were transfected using Lipofectamine 2000 (Invitrogen #11668030). A first virus passage was inoculated with supernatant collected at day 2 post-transfection. A second passage virus stock was inoculated with supernatant from passage one collected at day 2 post-infection. The stock was sequenced as described in the section “Deep sequencing”. Virus genomes were similarly sequenced to confirm genetic stability after three and four passages.

Kinetic comparison of viruses

VeroCCL81 cells were seeded in T25 flasks (2·10⁶ cells) (Nunc Cell Culture Treated Flasks with Filter Caps, Thermo Fisher Scientific) the day prior to infection. Cultures were inoculated with original-D614G, P60, P70rec, or P71 viruses at a multiplicity of infection (MOI) of 0.001. Each day post-infection, supernatant was collected to determine infectious titers as described in the section “Virus infectious titers” and CPE was scored as described in the section “Serial passaging of SARS-CoV-2”. In one experiment, RNA titers were determined as described in the section “Virus RNA titers”, and in another experiment, images of the T25 flask cultures were taken with an Evos Floid microscope (Invitrogen) (20 × objective). Immunostaining was done with primary anti-spike protein antibody (Sino Biological #40150-D004; RRID: AB_2827983) at 1:3000 in BSK (phosphate buffered saline (PBS) containing 1% (w/v) bovine serum albumin (BSA) (Roche #10735086001) and 0.2% (w/v) skimmed milk (Easis #801300)) incubated overnight at 4 °C or 2 h at room temperature and secondary antibody Alexa Fluor 488 conjugated antibody (Thermo Fisher Scientific #A-11013; RRID:AB_2534080) at 1:500 in BSK incubated for 20 min at room temperature and inspected using an Axio Vert.A1 inverted microscope (Zeiss) (20 × objective) as described^21,52. To quantify cell viability¹⁵, VeroCCL81 cells were seeded in 96-well plates (Thermo Fisher Scientific #167008) at 19,000 cells/well, one plate per timepoint, and infected at an MOI of 0.001 with original-D614G, P60, P70rec, and P71 viruses, 4 wells per virus per plate, and with 8 non-infected wells per plate. Cell viability was measured with the Viral ToxGlo Assay (Promega #G8943) according to the manufacturer’s instructions, and luminescence was measured with a Synergy XL multimode reader (Biotek). The luminescence in SARS-CoV-2-infected wells relative to the mean luminescence in non-infected wells was calculated.

Virus production in scalable bioreactor

Virus was produced in the CelCradle bioreactor mounted with 500AP culture flasks (Esco Aster Pte. Ltd. #1400004) containing BioNOCII macrocarriers, with a 100 mL bed volume and a 500 mL working volume^17,55. Briefly, VeroCCL81 cells were expanded in T175 flasks, and 1.5·10⁸ cells were seeded in the 500AP flask with CelCradle stage settings: Rising rate 2 mm/s, top holding time 20 s, downwards rate 2 mm/s, and bottom holding time 0 s for 2.5–3 h at 37 °C and 5% CO₂, and subsequently cultivated with CelCradle stage settings: Rising rate 1 mm/s, top holding time 10 s, downwards rate 1 mm/s, and bottom holding time 10 s. Medium was exchanged once or twice daily to maintain glucose concentrations above 8.3 mM (Cedex Bio Analyzer, Roche) and pH above 7 (FiveGo F2 pH meter, Mettler Toledo). The culture was infected at day 7 after cell seeding at an MOI of 0.007 (D614G, P60, and P70rec viruses) or 0.008 (BA.1). From the time of infection, cultures were maintained at 33 °C and 5% CO₂, medium exchange (virus harvest) was carried out twice daily for 5–6 days, and virus harvests were stored at – 80 °C until further processing. The infectious titer of virus harvests was determined as described in the section “Virus infectious titers”. Sequences of the virus stocks used for inoculation are outlined in Supplementary Table S3.

Inactivation and purification of virus for inactivated vaccines

Virus inactivation was carried out as described²². Pooled virus harvests containing 0.5 M HEPES at a pH of 7.5 were incubated with beta-propiolactone at 1:2000 (v/v) for 16 h at 4 °C and 3 h at 37 °C, sequentially clarified by filtration (Sartopure PP3 filters, cutoff 5 µm and 0.65 µm, Sartorius #5051342P5—OO—B and #5051305P5—OO—B), and spiked with sucrose (Sigma-Aldrich #S1888-500G) for a final concentration of 5% (w/v) prior to storage at – 80 °C. To confirm inactivation, VeroE6 cells seeded in T25 (0.7·10⁶ cells) flasks were inoculated with inactivated, clarified virus and monitored according to CPE and SARS-CoV-2 spike protein immunostaining every 2–4 days for two weeks. Immunostaining was done as described in the section “Kinetic comparison of viruses”. Purification of inactivated, clarified virus was carried out by steric exclusion chromatography^22,56. Briefly, for DNA digestion, the virus was incubated with 50 U/mL Denerase (c-Lecta #20804-100k), 2 mM MgCl₂ (Sigma-Aldrich #M8266-1 KG), and 0.05% (w/v) NaN₃ (Sigma-Aldrich #08591) for 6 h at room temperature on a magnetic stirrer at 250 RPM and filtered through 0.2 µm membranes (Cytiva #10410314). Virus capture was carried out with membrane-based steric exclusion chromatography with capsules assembled in-house^22,56 using the Äkta Pure 25M chromatography system (Cytiva) controlled by the Unicorn v6.3 software, and the presence of virus particles was monitored with a NICOMP 380 particle analyzer (Particle Sizing Systems). The virus was mixed in-line with polyethylene glycol (PEG) 6000 (Sigma-Aldrich #81260) in PBS for a final concentration of 8% during loading (w/v), followed by washing with 8% PEG6000, elution in PBS, and dialysis overnight at 4 °C into PBS in a 300 kDa cutoff cellulose ester dialysis tubing (SpectraPor #GZ-02890-77). Dialyzed samples were spiked with sucrose for a final concentration of 5% (w/v) prior to storage at – 80 °C.

Virus infectious titers

Virus infectious titers were determined as described previously^15,17,52. Serial dilutions of samples were prepared in supplemented DMEM, and 100 µL of each dilution was added to quadruplicate wells in a 96-well plate (Thermo Fisher Scientific #167008) seeded with 10,000 cells/well the previous day and incubated at 37 °C and 5% CO₂. After 48 h, cells were fixed in cold methanol (vwr #20847.320) for 20 min, followed by immunostaining with primary anti-spike protein antibody (Sino Biological #40150-D004; RRID:AB_2827983) diluted 1:5000 in BSK incubated overnight at 4 °C or 2 h at room temperature, secondary antibody F(ab’)2-Goat anti-Human IgG Fc Cross-Absorbed Secondary Antibody, horseradish peroxidase (HRP) (Thermo Fisher Scientific #A24476; RRID:AB_2535945) diluted 1:2000 in BSK incubated 1 h at room temperature, and visualization with the Bright-DAB solution kit (Immunologic #BS04-500). Plates were imaged using an Immunospot series 5 UV analyzer (CTL Europe GmbH)⁵⁷. The TCID₅₀ was determined according to the Reed-Muench method¹⁷. The LLOQ was determined by the first dilution included in the assay.

Virus RNA titers

Virus RNA titers were determined as described previously^15,17,52. Oropharynx swab samples were clarified by centrifugation for sample preparation. Clarified oropharynx swab samples or cell culture supernatant samples were mixed with 3 volumes of Trizol LS Reagent (Thermo Fisher Scientific #15596018) and RNA was extracted by addition of 0.2 volumes of chloroform (Sigma-Aldrich #C2432) and phase separation in 5PRIME Phase Lock Gel Heavy tubes (Quantabio #2302830). RNA was purified from the aqueous phase using the Zymo Research RNA clean and concentrator-5 kit (Zymo Research #R1014) according to the manufacturer’s instructions and eluted in nuclease-free water (Ambion #AM9930). Small pieces of lung tissue (42–88 mg per sample) were added to 1 mL of Trizol Reagent (Thermo Fisher Scientific #15596026) in MagNA Lyser Green bead tubes (Roche #03358941001) and homogenized in a MagNA Lyser (Roche). RNA was extracted by addition of 0.2 volumes of chloroform and phase separation in 5PRIME Phase Lock Gel Heavy tubes, purified using the Zymo Research RNA clean and concentrator-25 kit (Zymo Research #R1017) according to the manufacturer’s instructions, eluted in nuclease-free water, and quantified using a Nanodrop 1000 (Thermo Scientific) and a Qubit 3.0 (Invitrogen) with the AccuBlue Broad Range RNA Quantification Kit (Biotium #31073-T). The quantitative polymerase chain reaction (qPCR) probes and primers were described elsewhere and were adapted to use with TaqMan Fast Virus 1-Step Master Mix (Thermo Fisher #4444434) (LightCycler 96 System, Roche)¹⁷. Each analysis included duplicates of RNA standard (Twist Bioscience #102024) in the range of 10¹–10⁵ copies/µL, sample RNA, and negative control. A standard curve was generated in the LightCycler 96 software version 1.1.0.1320 (Roche) for determination of sample RNA titers. The LLOQ was determined as the mean value obtained from medium only samples plus three times the standard deviation. Where indicated in the figure legend, the LLOQ for lung tissue samples was determined as the mean value obtained from non-infected lung samples plus three times the standard deviation.

Protein quantification

The total protein concentration of inactivated, purified virus preparations was quantified using a Pierce BCA Protein Assay Kit (Thermo Fisher Scientific #23227) according to the manufacturer’s instructions. A standard curve was generated in GraphPad Prism 9 (4-parameter logistic non-linear regression analysis) and used to determine sample protein concentrations.

Mouse immunizations

Mouse experiments were carried out in certified animal facilities at the University of Copenhagen (Copenhagen, Denmark) in accordance with the ARRIVE-10 questionnaire and national Danish guidelines under animal study proposal 2020-15-0201-00586 and approved by the Danish Animal Experiments Inspectorate. Female BALB/c mice (Taconic #BALB-F MPF; RRID: IMSR_TAC:balb) 6–8 weeks of age arrived at the animal facility at least one week prior to initiation of experiments. Mice were immunized with inactivated, purified I-P60 (6 µg), I-BA.1 (6 µg), or I-P60/BA.1 (6 µg/6 µg), or OVA (100 µg) (Invivogen #vac-pova) mixed 1:1 with AddaVax adjuvant (Invivogen #vac-adx-10) (n = 4 per group) and administered subcutaneously on day 0, 21, and 42; for inactivated vaccine preparations, the dose (µg) relates to total protein. Mouse immunization doses of inactivated SARS-CoV-2²² and OVA^22,58 were selected based on previous studies. Blood was sampled on day 0, 14, 37, and 56, and serum was stored at – 80 °C. For euthanasia, animals were subjected to cervical dislocation. Upon euthanasia, total blood and spleens were collected. Mice were randomly assigned to immunization groups and animal care takers responsible for sampling and day-to-day handling were blinded regarding assigned groups.

Hamster immunizations

Hamster experiments were carried out in certified animal facilities at Statens Serum Institut (Copenhagen, Denmark) in accordance with the ARRIVE-10 questionnaire and national Danish guidelines under animal study proposal 2020-15-0201-00718 and approved by the Danish Animal Experiments Inspectorate. Male Syrian hamsters Mesocricetus auratus (RjHan:AURA, Janvier) arrived at the animal facility at least one week prior to initiation of experiments. In the pilot experiment, hamsters (~ 12 weeks of age) were challenged intranasally with 100 µL virus in DMEM, corresponding to 2000 TCID₅₀ units of either original-D614G virus (n = 4) or BA.1 (n = 4) in a BSL3 facility, or mock challenged with PBS (n = 4) in a BSL1/2 area. Oropharynx swabs were collected every day after challenge. Swab sticks were stored in 400 µL of transport medium (DMEM supplemented with 10% FBS and Antibiotic–Antimycotic 100× (Thermo Fisher Scientific #15240062)). Body weight was recorded every day from the day of challenge. For euthanasia, animals were dosed with either 5% isoflurane with a calibrated vaporizer or administered a combination of ketamine (200 mg/kg), xylazine (15 mg/kg), and acepromazine (5 mg/kg) intraperitoneally. Following full anesthesia (ascertained by loss of pedal retraction reflex upon stimulation), the animals were subjected to cervical dislocation. Upon euthanasia, the left lung was stored in RNAlater (Sigma #R0901) at – 80 °C.

In the vaccine experiment, hamsters (5–7 weeks of age) were immunized with inactivated, purified I-P60 (20 µg, n = 8), I-P60/BA.1 (20 µg / 30 µg, n = 8), or OVA (100 µg, n = 12) mixed 1:1 with AddaVax adjuvant and administered subcutaneously on day 0 and 21; for inactivated vaccine preparations, dose (µg) relates to total protein. The hamster immunization doses of I-P60 and OVA were selected based on a previous study²²; compared to I-P60, a higher dose of I-BA.1 was chosen due to the limited immunogenicity of I-BA.1 observed in the mouse immunizations. On day 44, hamsters were challenged intranasally with 100 µL virus in DMEM, corresponding to either 2000 TCID₅₀ units of the original-D614G virus (n = 4 per immunization group) or 6000 units of BA.1 (n = 4 per immunization group) in a BSL3 facility. One OVA-immunized group (n = 4) was mock challenged with PBS and was kept in a BSL1/2 area. Blood was sampled on day 0, 21, and 42, and plasma was stored at – 80 °C. Oropharynx swabs were collected on day 0, 21, 44, and every day after challenge. Body weight was recorded every day from the day of challenge. Upon euthanasia, total blood was collected for preparation of plasma, the left lung was stored in RNAlater (Sigma #R0901) at – 80 °C, and the right lung was formalin fixed as described in the section “Tissue pathology”. Hamsters were randomly assigned to immunization and challenge groups and animal care takers responsible for sampling and day-to-day handling were blinded regarding assigned groups, but not to challenge status as the mock challenged group was kept in a BSL1/2 facility²².

Mouse serum neutralization assay

Neutralization assays with mouse serum were carried out as described previously²². Serum was heat inactivated for 30 min at 56 °C and serially diluted in DMEM. In five replicate wells in a 96-well pre-plate, 5 µL of diluted serum was mixed with 12 µL of virus master mix for an MOI producing a fully stained cell layer as determined in pilot experiments. MOI of 0.03–0.05 (original-D614G virus), 0.03–0.06 (BA.1), 0.05–0.06 (delta), or 0.03–0.04 (BA.5) were used. A SARS-CoV-2 neutralizing antibody (Sino Biological #40591-MM43; RRID: AB_2857934) served as a positive control for neutralization of original-D614G and delta viruses, no positive control was included for neutralization of BA.1 or BA.5 viruses. Each plate included eight virus-only wells without serum, and six blank wells with neither serum nor virus. Following 1 h incubation at 37 °C and 5% CO₂, DMEM was added to each well for a total volume of 100 µL, which were then transferred to a 96-well plate seeded with 10,000 VeroE6 cells per well on the day prior to the experiment. Cells were incubated at 37 °C and 5% CO₂ for 48 h followed by fixation and immunostaining as decribed in the section “Virus infectious titers”. Plates were imaged and spike protein-positive cells were counted using an Immunospot series 5 UV analyzer (CTL Europe GmbH)⁵⁷. The percentage of SARS-CoV-2 positive cells in experimental wells relative to the mean number of SARS-CoV-2 positive cells in virus-only wells was calculated. The ID₅₀ was determined in GraphPad Prism 9 with a Sigmoidal dose–response (variable slope) regression with bottom and top constrains of 0 and 100, respectively. LLOQ was determined as the first dilution tested in the assay, 1:12.5.

Hamster plasma, human plasma, and bebtelovimab neutralization assays

Neutralization assays with hamster plasma were carried out as described previously^19,22. Hamster plasma was heat inactivated for 30 min at 56 °C, with or without a subsequent centrifugation step at 2500 RCF at 4 °C, and serially diluted in DMEM. The centrifugation step did not influence assay results as evaluated in head-to-head experiments. In a pre-plate, diluted plasma was mixed 1:1 with virus master mix for an MOI producing a fully stained cell layer as determined in pilot experiments. MOI of 0.012–0.035 (original-D614G virus) or 0.02–0.13 (BA.1) were used. A SARS-CoV-2 neutralizing antibody (Sino Biological #40592-MM57; RRID: AB_2857935) served as a positive control for neutralization of the original-D614G virus, no positive control was included for neutralization of BA.1. Each plate included eight virus-only wells without plasma, and four blank wells with neither plasma nor virus. Following 1 h incubation at room temperature, 100 µL of plasma-virus mix was transferred to quadruplicate wells in a 96-well plate seeded with 10,000 VeroE6 cells per well on the day prior to the experiment. Cells were incubated at 37 °C and 5% CO₂ for 48 h, followed by fixation, immunostaining, and SARS-CoV-2 positive cell counting as described in the section “Mouse serum neutralization assay”. The percentage of SARS-CoV-2 positive cells in experimental wells relative to the mean number of SARS-CoV-2 positive cells in virus-only wells was calculated. The ID₅₀ and 50% inhibitory concentration (IC₅₀) (Bebtelovimab) were determined in GraphPad Prism 9 with a Sigmoidal dose–response (variable slope) regression with bottom and top constrains of 0 and 100, respectively. LLOQ was determined as the first dilution tested in the assay, 1:25. Neutralization assays with human plasma and bebtelovimab (Cell Sciences #CPC539B) were done as described above¹⁹ with minor modifications. Human plasma samples were not centrifuged after heat inactivation, and human plasma and bebtelovimab dilutions were prepared in PBS. MOI values of 0.1 were used for P60, P70rec, and P71 viruses, and plates were fixed after 24 h of incubation. Immunostaining was done with primary SARS-CoV-2 neutralizing antibody (Sino Biological #40592-MM57; RRID: AB_2857935) at 1:5000 and secondary antibody Anti-mouse IgG, HRP linked whole antibody (from sheep) (Cytiva #NA931; RRID: AB_772210) at 1:1000 with diluent and incubation time as described in the section “Virus infectious titers”.

Human plasma samples

Human plasma samples from the Clinical, Virological, and Immunological (CVIC) study^18,19 were collected from infection-naïve individuals vaccinated against COVID-19 (1 month after second vaccination) with the original mRNA vaccine, Comirnaty (Pfizer/BioNTech) or from vaccination-naïve individuals hospitalized with COVID-19 (1 month after symptom onset; infected with original SARS-CoV-2 with the D614G mutation in the spike protein). The CVIC study was conducted according to the Helsinki guidelines and was approved by the Regional Ethics Committee and the Data Protection Agency (reference numbers H-20025872 and P-2020-357, respectively). Written informed consent was obtained from all participants of the study^18,19.

Meso scale discovery assay

The assay was done as described previously²². Spleens were collected from mice upon euthanasia and passed through a Falcon 100 µm sterile nylon cell strainer (Fisher Scientific #10282631). Red blood cells were lysed with RBC Lysis Buffer (Thermo Fisher Scientific #00-4333-57) and washed with PBS. Isolated mouse splenocytes were stimulated with inactivated SARS-CoV-2 at 1 µg/mL (original-D614G, P60, or BA.1) or recombinant proteins: Original-D614G spike protein (Sino Biological #40589-V08B4), omicron BA.1 spike protein (Sino Biological #40589-V08B33), original nucleocapsid protein (Sino Biological #40588-V08B), original envelope protein (Sino Biological #40609-V10E3), or OVA protein (Invivogen #vac-pova) all at 1 µg/mL, or medium. After incubation for 96 h at 37 °C and 5% CO₂, cells were pelleted by centrifugation and supernatant was transferred to a clean 96-well plate and stored at – 80 °C until analysis in the multiplex Meso Scale Discovery assay (mouse U-plex assay for cytokines INF-γ, IL-10, IL-5, IL-13, IL-17, TNF-α, Meso Scale Discovery) according to the manufacturer’s instructions. One replicate per sample was analyzed. Samples were analyzed with a Sector Imager 2400 system (Meso Scale Discovery) and cytokine concentrations were calculated according to the standard curve generated in the Discovery Workbench 4.0.12 software with a 4-parameter logistic non-linear regression analysis.

Tissue pathology

Lungs were collected from hamsters on the day of euthanasia and prepared for histological analysis as described previously^22,59. The right lung was inflation fixed with 10% formalin for 24–48 h and stored in 70% ethanol until trimming, tissue processing, and paraffin embedding for sectioning. Continuous 5 µm lung sections were collected and 3 individual sections, 100 µm apart in the z-plane displaying both the cranial and the caudal lung lobes, were deparaffinized, stained with Hematoxylin and Eosin (HE), and mounted for histological evaluation. All HE stained sections of the lungs were scanned to digital slides (Hamamatsu NanoZoom S360 with a 40 × objective with a digital resolution of 0.23 μm/pixel). The scanned sections were evaluated using the Qupath software version 0.4.2⁶⁰. The area of each lung lobe was outlined and measured (µm²) using the wand brush annotation tool. Similarly, lung tissue affected by inflammation (i.e., broncho-interstitial pneumonia) was outlined and measured (µm²), and the percentage of the total lung area affected by inflammation was determined (inflammation score) (Supplementary Table S2). For each animal, the mean scores of both lung lobes from all three sections were calculated and used to determine the group mean (Fig. 4c). In addition, sections were evaluated microscopically, and lesions identified in the lung sections were scored (Supplementary Table S4).

Deep sequencing

Deep sequencing sample preparation and analysis were carried out as described previously^15,53. RNA was extracted from cell culture supernatant as described in the section “Virus RNA titers”, and reverse transcribed using Maxima H Minus Reverse Transcriptase (Thermo Scientific #EP0752), dNTP mix (Thermo Fisher Scientific #18427089), RNAsin Plus RNase inhibitor (Promega #N2611), and a cDNA primer with subsequent RNase H treatment (Thermo Fisher Scientific #EN0201). The genome was amplified in five PCR reactions using Q5 Hot Start 2X Master Mix (NEB #M0494S), and PCR product was purified with Zymo Research DNA Clean and Concentrator-5 (Zymo Research #D4034) according to the manufacturer’s instructions. Indexed libraries were prepared with NEBNext Ultra II FS DNA library prep kit (NEB #E7805L) and sequence analysis was done as described previously^15,53.

Statistics

Graphs, data analysis, and statistical analysis were carried out using GraphPad Prism version 9. Group size information (n = 4 in most cases) is indicated in figure panels and legends. Generally, individual groups were compared in Mann–Whitney tests, Kruskal–Wallis tests with the post-hoc Dunn’s multiple comparisons test were used to compare more than two groups, and Friedmann tests with the post-hoc Dunn’s multiple comparisons test were used for paired analysis of more than two groups. LLOQ are indicated in figures and legends. For calculation of graph means and medians, values below LLOQ were given the value of the LLOQ. If one or more values used for calculations of means or medians were below LLOQ, this is indicated in the figure by “ < ” above the relevant symbols. Statistical analysis was only carried out between group means or medians where all values were above LLOQ. Where indicated in the figure legends, only statistically significant differences are shown in the figures.