Identification of an Optimized Receptor-Binding Domain Subunit Vaccine against SARS-CoV-2

This article is featured in Top Reads, p. 905

Severe acute respiratory syndrome coronavirus 2, the agent of the COVID-19 pandemic, has caused 10s of millions of deaths worldwide. Current levels of disease activity predict that COVID-19 will remain a major respiratory disease following its initial emergence in December 2019. SARS-CoV-2 has joined influenza A and respiratory syncytial virus to form a new trinity of endemic respiratory viral pathogens (1, 2). COVID-19 is projected to be the major cause of mortality among this trinity.

Vaccines are essential for control of COVID-19. Enormous scientific, public health, and private sector resources led to the successful development of multiple COVID-19 vaccines in record time. According to the World Health Organization, >300 COVID-19 vaccines were initially evaluated under preclinical or clinical conditions, with 11 vaccines ultimately approved for global public health use (3). Approved vaccines used four distinct vaccine platforms: inactivated virus vaccines, mRNA vaccines, adenovirus vector–based vaccines, and protein adjuvant vaccines (4). In the United States, four vaccines have been authorized with either full approval or for emergency use. These include two mRNA vaccines (Moderna mRNA-1273 and Pfizer BNT162b2), the adenovirus vector–based vaccine Ad26 COV2.S manufactured by Johnson and Johnson, and the protein adjuvant vaccine NVX-CoV2373 manufactured by Novavax. Despite the availability of these vaccines, considerable challenge to full vaccine control of COVID-19 remains. These include waning vaccine immunity (5), inequitable global vaccine supply, and vaccine hesitancy (6). This has led to high rates of global infection from which highly transmissible viral variants emerge that partially escape neutralizing Abs induced by vaccine or prior infection.

A vaccine approach using just the receptor-binding domain (RBD) region of the spike protein as Ag has the advantage of focusing immunity to the key protective determinant. Depletion experiments demonstrate that 90% or more of the neutralizing activity present in the plasma of convalescent subjects (7) and up to 99% of neutralizing activity in mRNA-vaccinated subjects target the RBD (8). The RBD has multiple Ab epitopes including both conserved and variable sequence regions. Conserved regions include the receptor-binding sequences and Ab responses focused on the receptor-binding sequences have the potential to provide cross-variant protection. Thus, RBD protein with a suitable adjuvant is an attractive vaccine strategy for multiple variants of SARS-CoV-2. In fact, a study by Cohen et al. (9) demonstrated that mice immunized with multimeric RBD proteins on nanoparticles elicited cross-reactive binding and neutralization responses supporting selection of RBD as a pan-sarbecovirus vaccine.

Host cell attachment of SARS-CoV-2 depends on the surface-exposed trimeric spike glycoprotein. The total length of SARS-CoV-2 spike protein is 1273 aa, which consists of a signal peptide (1–13 aa), the S1 subunit (14–685 aa), and the S2 subunit (686–1273 aa) (10). The S1 subunit is responsible for receptor binding and the S2 subunit mediates viral cell membrane fusion. The S1 subunit contains four domains: an N-terminal domain (NTD, 14–305 aa), an RBD (331–527 aa), and two subdomains (SD1, 528–590 aa and SD2, 591–685 aa) (Fig. 1A) (11). SARS-CoV-2 RBD is the key part of the spike protein that allows the protein to dock to the host cell receptor, angiotensin-converting enzyme 2 (ACE2), to gain entry into cells and lead to infection.

The atomic structure of the SARS-CoV-2 spike trimer has been determined by cryogenic electron microscopy, revealing two different conformations for the RBD with either an open or closed state (12) (Fig. 1B). The spike protein binds to ACE2 via the open conformation of the S1 subunit (Fig. 1C). Arase and colleagues (13) demonstrated that Abs that bind the NTD of the spike protein can induce an open conformation of the RBD, resulting in enhanced binding to the virus receptor ACE2 and enhanced infectivity. In a screen of Abs isolated from patients with COVID-19, a subset of infection-enhancing Abs against the N-terminal domain were detected among patients with severe COVID-19 (13). Thus, it is possible that full-length spike Ag-based vaccines induce both neutralizing Abs as well as anti-NTD Abs that may enhance infectivity. In principle, an RBD-based vaccine avoids the potential for adverse effects resulting from anti-NTD Abs.

In addition to important biological properties including ACE2 binding, key neutralizing Ab epitopes, and independent folding, RBD also possesses important manufacturing properties such as high thermal stability, cell-based production, and ease in scalability (14). Thus, there are substantial practical reasons to develop an RBD-based subunit vaccine.

In this study, we constructed various versions of RBD protein and conducted several mouse trials to evaluate the immunogenicity of these vaccines that are formulated in aluminum hydroxide (Alum) or the CD4 Th1 adjuvant DDA (dimethyldioctadecylammonium bromide)/TDB (d-(+)-trehalose 6,6’-dibehenate) (15). Moreover, we generated human ACE2 (hACE2) knock-in mice to assess in vivo vaccine efficacy in a virus challenge model. We compared RBD-based vaccines to a spike protein-based vaccine to identify an optimized RBD-based subunit vaccine suitable for human clinical trials.

Plasmids containing a prefusion-stabilized soluble spike by introducing six proline substitutions (HexaPro) (Spike-HP) (16) and RBDSD1 (residues 319–591, Wuhan strain) with a monomeric human IgG1 Fc tag (HEK_RBDSD1mFc) in a pαH mammalian expression vector were a gift from Dr. Jason McLellan (University of Texas, Austin, TX). SARS-CoV-2 Beta, B.1.351 (K417N, E484K, and N501Y) mutations in RBD were introduced to RBDSD1mFc by site-directed mutagenesis (HEK_RBDSD1Fc_B1.351). RBDSD1 with a dimeric Fc fusion (HEK_RBDSD1dFc) was designed in-house and purchased from Twist Biosciences (South San Francisco, CA). Extraction and purification of mammalian expression plasmids were performed with a PureLink HiPure plasmid maxiprep kit (Thermo Fisher Scientific, Waltham, MA). For Sf9 secreted expression of soluble RBD, residues 319–541 were cloned into pFHMSP-LIC with a honeybee melittin signal sequence and a C-terminal 6× His tag (RBD_SF9; Twist Biosciences). Yeast expression of RBD (residues 332–527; Pichia_RBD206) was done in Pichia pastoris using a pαH vector; the vector was shortened by site-directed mutagenesis.

Expression of HEK-derived RBDSD1 was carried out using the FreeStyle 293 and the Expi293F cell expression system (Thermo Fisher Scientific). Cells were grown at 37°C, 8% CO₂ in a >80% humidified chamber, and under shaking conditions. Protein was generally harvested at 96 h posttransfection as per the supplier’s protocols. RBDSD1mFc and RBDSD1dFc were purified using HiTrap protein A high performance (Cytiva) followed by size-exclusion chromatography on a Superdex 200 Increase 10/300 equilibrated with 20 mM HEPES (pH 7.5), 150 mM NaCl. RBDSD1 tagless was purified from supernatant by ammonium sulfate precipitation followed by size-exclusion chromatography step on a Superdex 75 Increase equilibrated with 20 mM HEPES (pH 7.5), 150 mM NaCl. Fractions containing RBDSD1 were pooled and diluted in half with 20 mM MES (pH 6.5) prior to loading onto a HiTrap SP XL column equilibrated with 20 mM MES (pH) 6.5. The flowthrough was collected and injected onto a Superdex 75 Increase equilibrated with 20 mM HEPES (pH 7.5), 150 mM NaCl. Aliquots of purified RBDSD1 were flash-frozen and stored at −80°C.

Spike-HP was expressed in Expi293F cells and purification was carried out at 25°C. The supernatant was collected and filtered; 20 mM imidazole, complete protease inhibitor (Roche), and 5 mM 2-ME were added prior to loading onto Ni-immobilized metal affinity chromatography resin. Eluted Spike-HP was pooled and concentrated and then subjected to size-exclusion chromatography on Superose 6 Increase column equilibrated with 20 mM HEPES (pH 8.0), 200 mM NaCl, 0.2 mM TCEP at 25°C. Aliquots of purified Spike-HP were flash-frozen and stored at −80°C.

Sf9 suspension cells were grown in Sf900 SFM III (Thermo Fisher Scientific) at 28°C and 0% CO₂. P3 baculovirus stocks expressing RBD were made as per the baculovirus expression vector systems manual (Invitrogen). Sf9 suspension cells were infected with P3 stock at a multiplicity of infection of 2. At 3–4 d postinfection, the supernatant was collected and clarified prior to loading onto a Ni Sepharose excel column (Cytiva) equilibrated with 10 mM HEPES (pH 7.2), 150 mM NaCl. Protein was eluted with 250 mM imidazole, then concentrated and subjected to size-exclusion chromatography with Superdex 200 Increase equilibrated in 10 mM HEPES (pH 7.2), 150 mM NaCl. Peaks containing monomeric RBD were pooled, concentrated, flash-frozen, and stored at −80°C.

Pichia pastoris was transformed as per the pPICZα A, B, and C manual (MAN0000035) from Invitrogen. Suspension cultures were grown in yeast extract peptone dextrose with Zeocin (500 µg/ml) at 30°C. Expression was performed in buffered glycerol-complex medium and cultures were induced in buffered methanol- complex medium (3% methanol). The supernatant was collected and filtered prior to the addition of ammonium sulfate to a final concentration of 2 M. The supernatant was then loaded onto a HiTrap phenyl Sepharose 6 fast flow (low sub) equilibrated with 20 mM HEPES (pH 7.5), 2 M ammonium sulfate, 500 mM NaCl. Protein was eluted with a gradient to 20 mM HEPES (pH 7.5), 500 mM NaCl (1 ml/min, 1 h). The eluate was concentrated and subjected to size-exclusion chromatography with a Superdex 75 Increase column equilibrated with 20 mM HEPES (pH 7.5), 150 mM NaCl. Peaks containing monomeric RBD were pooled, concentrated, flash-frozen, and stored at −80°C.

C57BL/6 wide-type mice (6 wk old, female) were purchased from Charles River Canada (Saint Constant, QC, Canada). Human ACE2 knock-in C57BL/6 mice that are susceptible to SARS-CoV-2 infection were generated in Princess Margaret Cancer Centre University Health Network (Toronto, ON, Canada). All of the animal experiments were reviewed and approved by the Committee on the Ethics of Animal Experiments of the University of British Columbia, the University Health Network (Toronto), and the University of Toronto.

C57BL/6 wild-type mice were used in the three mouse trials (Table I, experiments 1–3) to evaluate immunogenicity of vaccine candidates. Human ACE2 knock-in C57BL/6 mice were used in the fourth mouse trial (Table I, experiment 4) to evaluate vaccine protection elicited by the RBDSD1Fc vaccine, the best RBD vaccine optimized from experiments 1–3.

The groups of vaccine candidates in each experiment are listed in Table I. Ten micrograms of protein was formulated in a Th1-biased adjuvant DDA/TDB except those groups with different doses and formulated in the adjuvant Alum (Alhydrogel adjuvant 2%, InvivoGen) as noted in experiment 1 in Table I. DDA (product no. 890810P) and TDB (product no. 890808P) were produced by Avanti Polar Lipids and purchased from Sigma-Aldrich. The formulation of DDA/TDB was prepared as described previously (15). Each inoculation dose of 100 μl for immunization contained 250 μg of DDA and 50 μg of TDB.

For experiments 1–3, six C57BL/6 wild-type (WT) mice of each group were immunized three times s.c. at the base of the tail at 2-wk intervals. Two weeks after the final immunization, blood was collected to assess Ab responses and spleens were harvested to assess cellular immune responses. For experiment 4, eight hACE2 knock-in mice and eight C57BL/6 WT mice (as a control) were immunized with RBDSD1Fc vaccine three times s.c. at the base of the tail at 2-wk intervals. Eight nonvaccinated hACE2 knock-in (KN) mice were set up as a negative control. Two weeks after the final immunization, all three groups of mice in experiment 4 were challenged with SARS-CoV-2 virus to evaluate the vaccine efficacy.

Mouse sera from different vaccine groups were harvested to test anti-RBD or spike IgG responses by ELISA. Then, 96-well polystyrene microtiter plates (Corning, 3369) were coated with 0.1 μg per well of RBD or spike recombinant proteins and incubated with serially diluted murine sera from different vaccinated groups. This was followed by incubation with 1:2000 peroxidase-conjugated donkey anti-mouse IgG (Jackson ImmunoResearch, 715-035-150) and substrate (0.5 μg/ml ABTS in 0.1 M citrate buffer [pH 4.2] containing 0.03% H₂O₂). The reactions were read at 405 nm with Microplate Manager (Bio-Rad Laboratories). IgG Ab titers are expressed as the reciprocal of the highest dilution resulting in a reading of 2 SD above the value of negative control (PBS) sera.

The IFN-γ ELISPOT assay was performed as described previously (17). Briefly, 96-well MultiScreen-HA filtration plates (Millipore) were coated overnight at 4°C with 2 μg/ml murine IFN-γ–specific mAb (clone R4-6A2; BD Pharmingen). Splenocytes isolated from mice in complete RPMI 1640 medium (Sigma-Aldrich) were added to the coated plates at 10⁶ cells per well in the presence of tested RBD or spike recombinant protein (1 μg/ml). After 20 h of incubation at 37°C and 5% CO₂, the plates were thoroughly washed and then incubated with biotin anti-mouse IFN-γ (clone XMG1.2; BD Pharmingen) at 2 μg/ml. This was followed by incubation with streptavidin–alkaline phosphatase (BD Pharmingen) at a 1:1,000 dilution. The spots were visualized with a substrate consisting of 5-bromo-4-chloro-3-indolyl phosphate and NBT (Sigma-Aldrich). The number of IFN-γ–secreting cells was expressed as the difference between the number of spots per 10⁶ cells in Ag (protein)-stimulated wells and the number of spots per 10⁶ cells in nonstimulated wells.

Splenocytes isolated from different vaccine groups were stimulated with 1 μg/ml specific vaccine protein and 2 μg/ml Ab to CD28 in complete RPMI 1640 for 4 h at 37°C. Brefeldin A, an inhibitor of protein secretion, was added at a final concentration of 1 μg/ml, and cells were incubated for an additional 12 h before intracellular cytokine staining. Cells were surface stained for CD3 (PerCP-Cy5.5, clone 145-2C11), CD4 (Pacific Blue, clone RM4-5), and CD8 (allophycocyanin-H7, clone 53-6.7) as well as with the viability dye, green fluorescent reactive dye (Molecular Probes, L23101), followed by staining for IFN-γ (allophycocyanin, catalog no. 554413), TNF-α (PE-Cy7, catalog no. 557644), and IL-17 (PE, catalog no. 559502) using the BD Cytoperm Plus fixation/permeabilization kit according to the manufacturer’s instructions. All Abs and all reagents for intracellular cytokine staining were purchased from BD Pharmingen. We acquired 100,000 live lymphocytes per sample by using a BD FACSVerse flow cytometer and analyzed the data by using FlowJo software (Tree Star). The gating strategy used was to first gate on live cells followed by selection of cells expressing the T lymphocyte markers CD3, CD4, and CD8. The cytokine IFN-γ–, TNF-α–, or IL-17–producing cells were further gated from the CD3⁺CD4⁺ T cell population. Lastly, the fraction of IFN-γ/TNF-α double-positive cells from total IFN-γ–producing CD4 T cells was assessed (Supplemental Fig. 1).

Anti-spike IgG and its neutralizing capacity against SARS-CoV-2 WT and variants B.1.1.7, B.1.351, B.1.617, B.1.617.1, B.1.617.2, B.1.617.3, P.1, P.2, B.1.526.1, BA.1, BA.2, and BA.2.12.1 and the percent inhibition against BA.2.75 and BA.5 were quantified using the V-PLEX SARS-CoV-2 panel 13 and V-PLEX SARS-CoV-2 key variant spike panel 1 (ACE2) kits and a MESO QuickPlex SQ 120MM instrument (Meso Scale Diagnostics, Rockville, MD). The assays were performed according to the manufacturer’s instructions. Results for the Meso Scale Diagnostics assays were reported in picograms per milliliter derived from back fitting the measured signals for samples to a calibration curve, except for BA.2.75 and BA.5 Omicron variants not supported by a calibration curve. Alternatively, results for BA.2.75 and BA.5 Omicron variants were reported as percent inhibition, that is, % inhibition = (1 − sample ECL signal/blank ECL signal) × 100%.

The virus-neutralizing Ab assay was described by our group in 2005 (18). Briefly, the assay consists of mixing 100 infectious units of the virus with serial 2-fold dilutions of the test serum and, after a 2-h incubation, adding 100 μl of each mixture to a respective cell culture. The cultures were examined after 3 d for the presence of a viral cytopathic effect denoting the growth of the virus. The serum dilution immediately before the one in which virus growth was detected was referred to the titer of the serum. This assay was compared with the plaque reduction neutralization assay with the collaboration of the National Microbiology Laboratory and found to be substantially concordant.

To evaluate the vaccine efficacy elicited by the RBDSD1Fc vaccine (experiment 4), 2 wk after the final immunization, eight hACE2 knock-in mice and eight C57BL/6 WT mice immunized with RBDSD1Fc vaccine as well as eight KN mice were intranasally (i.n.) challenged with 4 × 10⁵ PFU of SARS-CoV-2. Oropharyngeal swabs were taken on days 0, 2, 3, 5, and 7 postinfection (p.i.). Half of each group of mice were sacrificed on days 3 and 7 p.i., respectively. Blood was collected to measure neutralizing Ab in sera. Lungs were collected and homogenized. Viral titers in lung homogenates were determined by median tissue culture–infective dose (TCID₅₀). Viral copy numbers in lung homogenates and oropharyngeal swabs were measured by real-time quantitative PCR (RT-qPCR). All manipulations with live SARS-CoV-2 in experiment 4 were performed in the Combined Containment Level 3 Unit at the Toronto High Containment Facility at the University of Toronto.

All data were analyzed with the aid of GraphPad Prism software. The Kruskal–Wallis test was performed to analyze data for IgG Ab titer and virus-neutralizing Ab titers from multiple groups, and the Mann–Whitney U test was used to compare medians between two vaccine groups. Comparisons of cytokine productions as determined by ELISPOT assay and flow cytometry, and virus titers, viral copy numbers determined by TCID₅₀, and RT-qPCR were analyzed using a Student two-tailed t test to compare means between two vaccine groups. A p value <0.05 was considered significant. Data are presented as means ± SEM.

Three eukaryotic systems (Pichia pastoris, Sf9, and HEK293) were used to produce candidate spike Ags designed around the RBD and SD1 domain (Fig. 1D, 1E). Each expression system displayed differences in protein production, glycosylation, and behavior in solution impacting the proteins selected for further study (data not shown). Different sequences covering the RBD were successfully expressed and purified from Pichia, Sf9, and HEK293 cells (Fig. 1D, 1E). Proteins including the SD1 domain immediately downstream to the RBD were also produced in HEK293 cells alone (tagless; RBDSD1) or with a C-terminal fused monomeric (16) (HEK_RBDSD1mFc) or dimeric (HEK_RBDSD1dFc; designed based on previously approved Ab-based therapeutics aflibercept and etanercept [19]) human IgG1 Fc region. The addition of Fc to a protein vaccine candidate is known to enhance immunogenicity as well as prolong the plasma half-life (20). The mutations occurring in the RBD region of the Beta version of SARS-CoV-2 strain B1.351 were introduced by site-directed mutagenesis (HEK_ RBDSD1Fc_B1.351). As a benchmark comparator, a prefusion-stabilized recombinant spike protein (Spike-HP) (16) was produced and assessed alongside the RBDSD1 Ags.

In these studies, we performed a series of animal trials to test versions of RBD recombinant protein with different Ag doses and adjuvants. The experimental protocols to evaluate vaccine immunogenicity are illustrated at the top of Fig. 2. Mice were immunized with vaccine candidates three times at 2-wk intervals. Two weeks after the final immunization, blood was collected to assess Ab responses including anti-RBD and whole spike protein Ab responses and anti–virus-neutralizing Ab responses. Spleens were harvested to assess cellular immune responses by IFN-γ ELISPOT assays and intracellular cytokine staining followed by flow cytometry.

We initially tested two SARS-CoV-2 RBD proteins: SF9 (insect)-derived RBDSD1 (RBD_SF9) and HEK293F (mammalian) cell–derived RBDSD1 (RBD_HEK). Three different doses of each protein (10, 5, and 2.5 μg) were formulated with Alum (2% Alhydrogel) or DDA/TDB (Table I, experiment 1). Alum is known to be effective at enhancing Ab responses and is the adjuvant in many commercial vaccine formulations. The DDA/TDB adjuvant has been found to be a particularly promising adjuvant capable of inducing both strong CD4 Th1 and Th17 responses and high Ab titers (15, 21).

We measured anti-RBDSD1 IgG responses by ELISA in sera from different vaccine groups. HEK-produced RBDSD1 was superior to SF9-produced vaccine, as significantly higher Ab responses (Fig. 2B) and CD4 Th1 responses including IFN-γ (Fig. 2C, 2E), TNF-α (Fig. 2F), and IL-17 (Fig. 2G) responses were observed. We also found that DDA/TDB was superior to Alum, as the DDA/TDB vaccine groups not only induced CD4 Th1 responses but also showed higher Ab responses compared with the corresponding Alum vaccine groups. No CD4 Th1 responses were observed in the Alum vaccine groups (Fig. 2D). We evaluated 10, 5, and 2.5 μg of RBD_HEK and RBD_SF9 protein in the vaccine formulations and found that the three doses induced similar levels of IgG Ab responses in all Alum and DDA/TDB formulations (Fig. 2B). However, the 10 μg of RBD_HEK + DDA/TDB vaccine group induced significantly higher CD4 Th1 responses compared with the 5 and 2.5 μg vaccine groups (Fig. 2E, 2F). Overall, mice immunized with 10 μg of RBD_HEK formulated in DDA/TDB adjuvant produced high anti-RBDSD1 Ab titers and strong IFN-γ responses in which 87.7 ± 2.3% of IFN-γ CD4 T cells secreted TNF-α (Fig. 2H, 2I).

We next performed the SARS-CoV-2 microneutralization assay to evaluate neutralizing Ab against SARS-CoV-2 virus infection on Vero E6 cells. We used the prevaccination and postvaccination serum from an mRNA-vaccinated human volunteer as a positive control in our assays. The positive serum was standardized using the World Health Organization reference standard serum. The neutralizing Ab titer of the positive serum was between 1:80 and 1:160 in multiple assays, similar to titers found in most fully mRNA-vaccinated people (our unpublished data). The results showed that the HEK-derived RBDSD1 protein induced higher neutralizing Ab responses compared with the SF9-derived protein, and the adjuvant DDA/TDB significantly enhanced the production of neutralizing Abs compared with Alum (Fig. 2J). Of note, the three groups of SF9-derived RBDSD1 protein vaccines formulated with Alum failed to induce any virus-neutralizing Ab. Ten micrograms of HEK RBD protein formulated with DDA/TDB produced 2-fold higher neutralizing Ab titers compared with the spike protein control.

To optimize the protein dose in the vaccine formulation, we repeated the experiment using 20, 10, and 5 μg of RBD_HEK protein in the DDA/TDB formulations. We observed that the 20-μg protein dose did not enhance cellular immune responses or Ab responses compared with 10 and 5 μg formulations (data not shown). Ten micrograms of protein in DDA/TDB was selected as the optimal vaccine formulation against which to test other protein candidates.

In the second experiment, we evaluated more versions of RBD recombinant proteins with the adjuvant DDA/TDB. We tested three HEK293F cell–derived RBD recombinant proteins: RBD (205 aa) (HEK_RBD205), RBD with proximal SD1 domain (281 aa) (HEK_RBDSD1), and RBDSD1 with a human IgG1 monomeric Fc fragment (581 aa) (HEK_RBDSD1mFc). An RBD recombinant protein (206 aa) produced in Pichia yeast cells (Pichia_RBD206) was also tested. Spike-HP (16) was evaluated as the comparator vaccine Ag.

Two weeks after the final immunization, blood and spleens were collected to evaluate Ab and cellular immune responses. We measured Ag-specific IgG responses by ELISA (Fig. 3A). The results demonstrated that RBD produced in HEK293 cells generated significantly higher Ab titers than did RBD produced in Pichia cells. Importantly, RBD with SD1 dramatically enhanced Ab responses compared with RBD alone. The Ab titers in the HEK_RBDSD1 group were 12-fold higher than those in the HEK_RBD205 group. Moreover, the results showed that the addition of monomeric Fc fragment onto RBDSD1 (RBDSD1mFc) significantly enhanced Ab responses, that is, 2.5-folds higher than that in the HEK_RBDSD1 group. The Spike-HP group showed the highest IgG responses. Because the spike protein was the Ag used in these serological assays, the results may represent the presence of additional non-RBD epitopes, resulting in higher measured immunogenicity.

Next, we measured Ag-specific IFN-γ responses in the different vaccine groups by ELISPOT assays (Fig. 3B). The splenocytes from different vaccine groups were stimulated with Spike-HP, HEK_RBDSD1, HEK_RBDSD1mFc, HEK_RBD205, or Pichia_RBD206 protein and Ag-specific IFN-γ–producing splenocytes were assessed. The results showed that mice immunized with HEK_RBDSD1 or HEK_RBDSD1mFc induced strong IFN-γ responses to the stimulation by both RBD and Spike-HP protein. Notably, mice immunized with Spike-HP showed strong IFN-γ responses only to stimulation with Spike-HP protein, but not with RBD protein, suggesting that the IFN-γ responses induced by spike protein were specific to regions of spike protein outside the RBD. Compared to the HEK_RBDSD1 vaccine, mice immunized with RBD without the SD1 domain (HEK_RBD205) almost totally lacked the capacity to induce IFN-γ responses, confirming that SD1 is also key to maintaining CD4 T immunogenicity of RBD. It was not surprising that RBD derived from Pichia (Pichia_RBD206) failed to induce IFN-γ responses, as it also induced weak Ab responses. We further analyzed the cytokine-producing CD4 T cells in the different vaccine groups (Fig. 3C–H) and found that Spike-HP, HEK_RBDSD1, or HEK_RBDSD1mFc induced high IFN-γ– or TNF-α–producing CD4 T cells. Again, most IFN-γ–producing CD4 cells coexpressed TNF-α. HEK_RBD205 and Pichia_RBD206 did not induce IFN-γ– or TNF-α–producing CD4 T cells. Overall, we concluded that the addition of SD1 or SD1mFc fragment into HEK_RBD205 significantly increased cytokine production.

We next evaluated virus-neutralizing Ab responses in the different vaccine groups (Fig. 3I). Spike-HP vaccine generated high anti–virus-neutralizing Ab responses with the median titers of 1:4800. HEK-derived RBDSD1 with an mFc fragment also developed high neutralizing Ab responses with the median titer of 1:1120 that was significantly higher when compared with HEK-derived RBDSD1 with the median titer of 1:200. RBD alone derived from HEK developed low neutralizing Ab with the median titer of 1:22; no neutralizing Ab was detected in RBD derived from Pichia. Based on these data, we conclude that 1) HEK-produced RBD is superior to Pichia (yeast)- and SF9 (insect)-produced protein, 2) the SD1 domain is essential for RBD to induce optimal B and T cell immune responses and 3) RBDSD1 with mFc induced the highest neutralizing Ab responses.

In the third experiment, we further optimized RBDSD1. RBDSD1mFc, RBDSD1 plus dimeric Fc (RBDSD1dFc), and RBDSD1 plus monomeric Fc with the Beta version B1.351 (RBDSD1Fc_B1.351) were produced from HEK293F cells and formulated in DDA/TDB for the evaluation of vaccine immunogenicity.

Immune sera and spleens were collected 2 wk after the final immunization to assess Ab and cellular immune responses, respectively. The IgG ELISA results showed that RBDSD1 with all tested immune fragments (RBDSD1mFc, RBDSD1dFc, RBDSD1 3C_B1.351, RBDSD1Fc_B1.351) induced significantly higher Ab responses than did RBDSD1 without any tag (tagless) (Fig. 4A). The IFN-γ ELISPOT results (Fig. 4B) demonstrated that the addition of monomeric Fc onto RBDSD1 with either the WT or Beta variant B1.351 version significantly increased IFN-γ production. RBDSD1 that dimerized with the addition of dimeric Fc failed to enhance IFN-γ production although the dimeric Fc fragment enhanced Ab production. Similar to results observed in the second trial, Spike-HP induced strong IFN-γ responses only to stimulations with Spike-HP protein, but not to RBDSD1 protein, again suggesting that the CD4 T cell responses induced by spike protein were mainly contributed by regions outside of the RBDSD1 domain.

We next used intracellular staining followed by flow cytometry to measure cytokine-producing CD4 T cell responses. We found that RBDSD1 with all tested immune fragments except RBDSD1dFc generated significantly higher numbers of IFN-γ– or TNF-α–producing cells than did RBDSD1 tagless (Fig. 4C, 4D). More than 80% of IFN-γ–producing CD4 T cells coexpressed TNF-α (Fig. 4F). The IL-17 responses were detected in the group of RBDSD1 tagless, RBDSD1mFc, RBDSD1Fc_B1.351, and Spike-HP. The IL-17–producing CD4 T cells in the Spike-HP but not the RBDSD1mFc, RBDSD1Fc_B1.351 group were significantly higher than that in the RBDSD1 tagless group (Fig. 4E).

Next, we tested virus-neutralizing activity efficacy against SARS-CoV-2 virus using sera from different vaccine groups. Any version of RBDSD1 significantly enhanced the production of neutralizing Ab (Fig. 4G). Among all of the RBD Ags we tested, RBDSD1Fc_B1351+DDA/TDB generated the highest virus-neutralizing Ab responses, which were comparable to those observed with Spike-HP+DDA/TDB. The median neutralizing Ab titer of RBDSD1Fc_B1351+DDA/TDB versus Spike-HP+DDA/TDB were 1:4300 versus 1:4480.

To determine whether the vaccine was able to provide a broad neutralizing activity against various virus variants, we performed an ACE2 neutralization assay to test Abs that block binding of ACE2 to the spike Ags of the original SARS-CoV-2 virus and 14 SARS-CoV-2 variants using mouse sera collected after immunization with RBDSD1Fc_B.1.351+DDA/TDB and Spike-HP+DDA/TDB. Immunization with either RBDSD1Fc_B.1.351+DDA/TDB (RBDSD1Fc Imm) or Spike-HP+DDA/TDB (Spike-HP Imm) produced strong Abs responses that blocked binding of ACE2 to the original and 14 SARS-CoV-2 variants (Fig. 4H). The results showed that RBDSD1Fc_B.1.351 formulated in DDA/TDB (RBDSD1Fc vaccine) compared with other RBD vaccines generated the highest neutralizing Abs not only against WT SARS-CoV-2 but also against other SARS-CoV-2 variants including five Omicron variants. These Ab concentrations induced by RBDSD1Fc were equivalent to those induced with Spike-HP. It is not surprising that the ACE2 neutralizing Abs to Omicron variants were lower compared with other variants because of the large sequence divergence between Omicron and other variants of concern. Importantly, the immune sera showed substantial activities that inhibited the binding of ACE2 to all Omicron variants.

To evaluate vaccine protective efficacy, we generated hACE2 knock-in mice that were susceptible to infection with SARS-CoV-2. Based on data obtained from three previous mouse experiments to evaluate vaccine immunogenicity, the RBDSD1Fc vaccine (RBDSD1Fc_B1.351+DDA/TDB) was selected as the best RBD vaccine candidate. Next, we performed the fourth experiment to evaluate vaccine protection elicited by RBDSD1Fc vaccine in hACE2 knock-in mice. WT C57BL/6 mice and vaccinated hACE2 knock-in (KV) mice were vaccinated s.c. three times using the same vaccination protocol. A second set of KN mice were set up as a negative control. All three groups (WT, KV, and KN) were i.n. challenged with 4 × 10⁵ PFU of SARS-CoV-2. Oropharyngeal swabs were taken on days 0, 2, 3, 5, and 7 p.i. Half of each group of mice were sacrificed on days 3 and 7 p.i. Blood and lungs were collected to measure neutralizing Ab in blood and viral abundance in lung. The animal protocol to evaluate vaccine protection is illustrated in Fig. 5A.

We measured neutralizing Ab titers in the sera of vaccinated WT, KV, and KN mice at days 3 and 7 postinfection (Fig. 5B). Vaccinated WT and hACE2 knock-in mice exhibited similar neutralizing Ab responses. The median titers at days 3 and 7 p.i. were 1:2560 and 1:1506 in the KV mice, and 1:1920 and 1:2560 in the vaccinated WT mice. As expected, no and very low virus-neutralizing Ab was detected at days 3 and 7 p.i. in the KN mice.

We quantified virus titers in lung homogenates by TCID₅₀ (Fig. 5C). TCID₅₀ was measured by observing cytopathic effects on Vero E6 cells. The results showed that the virus titers determined by TCID₅₀ per milligram of lung tissue at days 3 and 7 p.i. were 1266 ± 181 and 317 ± 70, respectively, in the KN mice. In contrast, the virus titers in the KV mice at days 3 and 7 p.i. were only 0.056 ± 0.0027 and 0.061 ± 0.0058, respectively, comparable to the assay background 0.067 ± 0.0005 (day 3) and 0.058 ± 0.0081 (day 7) tested in the WT mice. The results demonstrated that the optimized RBDSD1Fc vaccine was able to completely protect the vaccinated mice from SARS-CoV-2 infection in the lung.

We next performed RT-qPCR to measure viral copy number in lungs and nasal swabs of WT, KV, and KN mice at indicated time points. Viral copy number was determined by a standard curve using SARS-CoV-2 open reading frame (ORF)1ab PCR primers. We found that the virus copy number in lungs of KN at days 3 and 7 p.i. reached to 902,568 ± 321,763 and 292,274 ± 103,307, respectively, whereas the viral copy number in lungs of KV mice showed 17 ± 14 and 1.4 ± 0.2 at days 3 and 7 p.i., similar to background levels in WT mice (Fig. 5D). The results of nasal swabs further revealed that high viral copy number was detected in KN mice at days 2, 3, 5, and 7 p.i. whereas no viral copy was detected at days 3, 5, and 7 p.i. in KV or WT mice. Due to the virus challenge, low viral copy number (<10) was transiently found in a few mice from the KV and WT groups only at day 2 p.i. (Fig. 5E). Therefore RT-qPCR analysis of the SARS-CoV-2 ORF1 gene in lungs and nasal swabs confirmed the excellent vaccine efficacy of RBDSD1Fc in providing complete protection against the virus challenge.

Protein adjuvant vaccines are one of medicine’s most reliable vaccine platforms from both a clinical and manufacturing perspective. Our goal was to produce a protein adjuvant vaccine to prevent COVID-19 disease and SARS-CoV-2 infection.

We focused on the RBD of the spike protein and compared the immunogenicity of a variety of RBD versions to Spike-HP as the gold standard. We found that RBDSD1 recombinant protein produced in mammalian HEK cells elicited better immunogenicity than did the identical protein produced in insect (SF9) or yeast (Pichia) cells, suggesting the importance of protein glycosylation to immunogenicity. RBDSD1 combined with the DDA/TDB adjuvant produced higher neutralizing Ab responses and much stronger CD4 T cell responses than did the identical protein combined with Alum. In fact Alum adjuvant elicited virtually no CD4 T cell responses. We also observed that the SD1 domain was essential for the RBD to elicit maximal B and T cell immunogenicity. The production of an RBDSD1 fusion protein with a monomeric human Fc receptor significantly enhanced immunogenicity and the production of neutralizing Ab. We compared RBDSD1mFc protein (Beta version) versus spike HP protein with DDA/TDB adjuvant and noted that both vaccines elicited nearly identical high titers of neutralizing Ab (>1:4000), ∼25-fold greater than convalescent sera seen in a post–mRNA-vaccinated individual and 100-fold greater than that observed in convalescent sera from naturally infected persons. The Beta variant of RBCSD1Fc and spike HP generated Abs that were able to block ACE2 binding with the Wuhan strain as well as 14 variants of concern, including Omicron. We generated hACE2 knock-in mice to evaluate protective efficacy in a virus challenge model. We found that RBDSD1Fc with DDA/TDB provided complete protection against virus challenge in the knock-in mouse model. Thus, we have identified an optimized RBD-based subunit vaccine suitable for clinical trials.

Currently, multiple vaccines using new technologies such as modified mRNA encapsulated in lipid nanoparticles and replication-deficient adenoviral vectors have been developed. Initial trials showed vaccine effectiveness >90% against symptomatic disease (22) and decreased rates of transmission among vaccinated individuals (23). However, these vaccines have become less effective over time due to waning immunity (24) and continued viral mutational evolution with increased risk of transmission (25).

Protein vaccines alone or in combination with mRNA or adenovirus-vectored vaccines may be able to induce higher titer and longer-lived Ab responses and thus limit selection of antigenic variants. This heightened immunity is seen with hybrid immunity where an individual receives a mRNA vaccine course followed by a subsequent SARS-CoV-2 infection. Thus, mRNA prime and protein boost could be an alternative vaccine strategy that builds on the success of mRNA vaccines to prevent further morbidity and mortality and reduce the evolution of SARS-CoV-2. Recently Mao et al. (26) developed a vaccination strategy in mice that uses systemic priming with mRNA followed by intranasal boosting with spike protein. The results showed that the strategy induced resident mucosal memory B and T cell responses, IgA at the respiratory mucosa, boosted systemic immunity, and completely protected mice from lethal SARS-CoV-2 infection. The strategy was durable, leading to protection from lethal SARS-CoV-2 challenge for as long as 118 d from vaccination. The strategy was also protective in hamsters and was superior to a mRNA–lipid nanoparticle prime-boost at blocking transmission. Thus, protein-based SARS-CoV-2 vaccines have the potential to significantly augment the current mRNA vaccine platform.

For acute SARS-CoV-2 infection, neutralizing Abs are critical for blocking infection. However, a combination of Ab and cellular immune responses most likely controls viral replication p.i. and prevents progression to severe disease (27). For a highly transmissible SARS-CoV-2 variant that escapes neutralizing Abs, cellular immunity may be particularly important for protection against severe disease (28). Multiple studies have shown that while neutralizing Abs induced by primary mRNA vaccine regimens show reduced cross-reactivity with the Omicron variant of SARS-CoV-2, T cell responses exhibit very good (>80%) cross-reactivity to Omicron (29, 30) and to other variants (31). Thus, mRNA vaccines remain highly effective at preventing hospitalization and death even though they are failing to completely block transmission. We hypothesize that a RBD vaccine by generating Abs to multiple RBD epitopes including the conserved ACE2 receptor binding sequences could offer advantages in preventing disease and blocking transmission not seen with mRNA vaccines.

In this study, we used a Th1-biased adjuvant DDA/TDB, also called CAF01, and observed that DDA/TDB was superior to Alum in terms of T and B cell immunogenicity. DDA/TDB promoted a broad and complex immune response characterized by multifunctional T cells with a CD4 Th1 profile. We also detected CD4 Th17 responses that may augment mucosal immune defenses (32). IL-17 responses were about one third of the magnitude of those of IFN-γ or TNF-α. DDA/TDB adjuvant has been tested in several phase I clinical trials to evaluate safety, tolerability, and immunogenicity for the development of vaccines against tuberculosis, malaria, pandemic influenza, and Chlamydia (33, 34). We expect this adjuvant to be increasingly used in human vaccines including COVID-19 vaccine because of its superior adjuvanticity and defined synthetic chemistry compared with Alum.

We determined that the most immunogenic form of RBD protein is expressed in mammalian cells and concluded that this likely reflects the immunological importance of glycosylation patterns. Glycosylation is important for proper folding of protein and for Ab recognition. Multiple studies have shown that glycosylation can increase the conformational stability of proteins to prevent rapid clearance from the circulation. Other SARS-CoV-2 protein vaccines are being produced in plant (35) or insect cells (36), but based on our immunogenicity data we expect our Ag produced in mammalian cells will have superior immunological properties.

Our data also showed that the SD1 domain was essential for RBD to elicit maximal immunogenicity. SD1 as an RBD-adjacent domain might be necessary to stabilize the native conformation of RBD for virus–host interaction, and/or to enhance immunogenicity by the addition of 63 aa from the SD1 fragment. Recently, a class of broadly neutralizing Abs that bind an epitope on SD1 were identified (37). This SD1 epitope is conserved between current SARS-CoV-2 variants and SARS-CoV-1. Thus, the addition of SD1 offers several potential advantages to an RBD vaccine.

To further optimize the RBD vaccines, we generated RBDSD1-Fc–based candidates. We found that addition of human IgG1 Fc fragment in either the monomeric or dimeric format onto RBDSD1 dramatically enhanced the Ag-specific IgG responses and virus-neutralizing Ab responses. There are at least two advantages for Fc-fusion proteins. First, the Fc fragment of human IgG can extend the half-life of the immunogen via increased size and FcR recycling (38). Fc fusion is one of the most clinically successful strategies for extending protein half-life and has been used in the development of hormones, growth factors, blood proteins, and other therapeutic proteins or peptides. Second, the Fc fragment of IgG when fused with an antigenic protein provides more efficient delivery to APCs. It has been reported that the Fc portion of the fusion protein can increase immunogenicity by ∼1000-fold (20). Although both RBDSD1 with monomeric Fc and RBDSD1 with dimeric Fc vaccines significantly increased Ab responses compared with the RBDSD1 tagless vaccine, only RBDSD1 with monomeric Fc but not RBDSD1 with dimeric Fc vaccine enhanced cellular immune responses. The reason for the difference in eliciting cellular immune responses between monomeric and dimeric Fc remains unknown. Based on these results we elected to move forward with evaluation of the monomeric Fc fusion protein Ag for protective efficacy in an animal model system.

As hACE2 but not murine ACE2 (mACE2) can bind SARS-CoV-2 spike protein, several mouse models genetically engineered to express hACE2 have been established for testing of vaccines and therapeutics and defining aspects of SARS-CoV-2 pathogenesis. The K18-hACE2 transgenic mouse that expresses hACE2 under an ectopic cytokeratin K18 promoter is commonly used and provides a model of severe infection in vivo (39). However, expression of the hACE2 transgene in these mice is nonphysiological with both a high number of hACE2 transgene insertions as well as ectopic expression in cells that do not normally express ACE2. This has raised issues regarding their relevance to human infection, as the hACE2 receptor shows tissue-specific expression patterns. In this study, we generated hACE2 knock-in mice in which hACE2 was expressed under an endogenous promoter in place of mACE2. We preserved the endogenous mACE2 signal sequence (aa 1∼17) and 3′ untranslated region to not affect the gene expression of corresponding pathways in the mouse. The region from aa 18 to aa 805 of mACE2 was replaced with the region from aa 18 to aa 805 of hACE2. To improve expression in the mouse, the entire amino acid hACE2 coding region was codon-optimized using the GeneArt tool, which improved expression ∼8-fold compared with nonoptimized gene sequences. As hACE2 expression is driven by the endogenous ACE2 regulatory elements, our hACE2 knock-in mouse model should model the ACE system in a more physiologically relevant manner and thus be more suitable for evaluation of viral pathogenesis and vaccine efficacy compared with hACE2 transgenic mouse models.

Our results showed that after intranasal inoculation, high levels of viral titers and viral copy numbers were detected in lungs of the KN mice whereas no or extremely low levels of virus were detected in the WT mice, suggesting that the hACE2 knock-in mice we generated are sensitive and specific for SARS-CoV-2 infection. Importantly, the hACE2 knock-in mice vaccinated with our RBDSD1Fc vaccine produced high neutralizing Ab responses and engendered complete protection from the SARS-CoV-2 infection, further validating our hACE2 knock-in mice as mouse model for testing of COVID-19 vaccines.

In conclusion, we have identified the Beta version of RBDSD1Fc with the adjuvant DDA/TDB as an RBD-based subunit vaccine that is able to produce high neutralizing Abs against multiple SARS-CoV-2 variants as well as strong CD4 Th1 immune responses and that provides complete protection against virus challenge. The RBDSD1Fc vaccine induced potent neutralizing Abs comparable to the full-length spike vaccine but with the potential of avoiding adverse effects such as enhanced infectivity induced by the non-RBDSD1 region in spike. We chose the RBDSD1Fc Beta version as our best vaccine because it induced Ab to a broad range of antigenic variants. We believe this vaccine candidate is suitable for clinical trials. This RBD vaccine can be updated or combined with other RBD variants of concern based on epidemiological trends. Considering high-yield production and manufacturing potential, RBD-based vaccines offer an advantage of temperature-stable doses at an affordable cost, which is especially beneficial for populations in low- and middle-income countries. With waning immunity and the need for booster doses, an RBD-based vaccine such as this one could be a promising vaccination strategy for both previously immunized and naturally infected individuals in a renewed global effort to control SARS-CoV-2 spread.

The authors have no financial conflicts of interest.