Development of recombinant COVID-19 vaccine based on CHO-produced, prefusion spike trimer and alum/CpG adjuvants

COVID-19 pandemic has severely impacted the public health and social economy worldwide. A safe, effective, and affordable vaccine against SARS-CoV-2 infections/diseases is urgently needed. We have been developing a recombinant vaccine based on a prefusion-stabilized spike trimer of SARS-CoV-2 and formulated with aluminium hydroxide and CpG 7909. The spike protein was expressed in Chinese hamster ovary (CHO) cells, purified, and prepared as a stable formulation with the dual adjuvant. Immunogenicity studies showed that candidate vaccines elicited robust neutralizing antibody responses and substantial CD4+ T cell responses in both mice and non-human primates. And vaccine-induced neutralizing antibodies persisted at high level for at least 6 months. Challenge studies demonstrated that candidate vaccine reduced the viral loads and inflammation in the lungs of SARS-CoV-2 infected golden Syrian hamsters significantly. In addition, the vaccine-induced antibodies showed cross-neutralization activity against B.1.1.7 and B.1.351 variants. These data suggest candidate vaccine is efficacious in preventing SARS-CoV-2 infections and associated pneumonia, thereby justifying ongoing phase I/II clinical studies in China (NCT04982068 and NCT04990544).


Introduction
The pandemic of coronavirus disease 2019 (COVID- 19), which is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [1], has severely impacted the public health and global economy. Since the first cases of COVID-19 were reported in December 2019 [2], numerous researchers have taken great efforts to control this disease. Development of prophylactic vaccines against SARS-CoV-2 is a proven strategy to prevent and to terminate the unprecedented pandemic [3]. Currently, different types of vaccines have been developed or under development against SARS-CoV-2 [4]. Several of them have completed Phase III clinical trials and demonstrated to be efficacious in preventing SARS-CoV-2 infections and/or reducing the occurrence of severe symptoms, hospitalization rate, and death caused by SARS-CoV-2 infections. As the demand for SARS-CoV-2 vaccines is huge, current approved SARS-CoV-2 vaccines cannot meet the requirement of removing COVID-19 pandemic with rapidity. In addition, vaccinees may choose different type of vaccines according to the age, health status and affordability. Thus, it is necessary to develop SARS-CoV-2 vaccines with different platforms. Here we report the development of a modified prototype spike protein-based vaccine combined with Alum/CpG dual adjuvant system. SARS-CoV-2 invades into host cells by engaging the receptor binding domain (RBD) of spike glycoprotein with angiotensinconverting enzyme 2 (ACE2) on host cell surface [5]. Based on the cell entry mechanism, spike glycoprotein is a reasonable vaccine target. In line with this hypothesis, it was confirmed that plenty of neutralizing antibody (nAb) epitopes reside in spike glycoprotein [6][7]. Spike glycoproteins are displayed on the exterior of SARS-CoV-2 virion as a trimer. We hypothesized that spike trimer in the prefusion conformation is highly antigenic, a lesson learned from the vaccine development targeting respiratory syncytial virus (RSV), MERS, and SARS [8][9]. Therefore, we designed a prefusion-stabilized spike trimer as the vaccine target, named SDTM. Chinese hamster ovary (CHO) cell expression system was used to express the target antigen, as it has sophisticated glycosylation system, which may be essential to SDTM's immunogenicity.
To elicit maximum immune responses, we incorporated a dual adjuvant system into our candidate vaccine, which contains aluminium hydroxide (Alum) and CpG 7909 (CpG). Aluminium salts have been used in vaccines for approximately 100 years with an excellent record of safety and effectiveness. Though the mechanisms of action of alumunium adjuvants are controversial, it is demonstrated that they can help induce T helper type 2 (Th2) cell-associated antibody responses [10]. In addition, aluminium salts are able to absorb and stabilize antigens formulated in vaccines [11], which contributes to the stability of vaccine immunogens and benefits the process of vaccine production. CpG 7909 is a synthetic oligonucleotide, a ligand of Toll-like receptor 9 (TLR9) [12]. By binding to endogenous TLR9 in B cells, dendritic cells (DCs), or macrophages, CpG 7909 activates MyD88 signal pathway and induce proinflammatory immune responses [12][13]. In addition, CpG 7909 activates DC to upregulate costimulatory molecules and activation markers to promote their homing to draining lymph nodes [14]. As a result, CpG 7909 help organisms to induce Th1biased cellular and humoral immune responses, which confer the protection against infection. As the dual adjuvant system possesses the advantages of both aluminium salts and CpG oligonucleotide, vaccine targets adjuvanted with this adjuvant system are likely to induce high level of antibody responses associated with Th1biased immunity. In terms of development of SARS-CoV-2 vaccine, Th1-biased immune responses may reduce the potential of vaccine-enhanced diseases (VED) [15][16][17], though no VED was reported in completed clinical trials and post clinical trial studies so far.
In this study, we report the excellent antigenicity of immunogens, immunogenicity of the vaccine candidate in rodents and nonhuman primate (NHP) models as well as the efficacy of the vaccine candidate in a hamster challenge study. We also analyzed the cross-neutralizing activity of the immune sera of NHP. The results suggest our vaccine candidates are promising and support further clinical development.

Immunogen design and characterization
As it is the initial binding site of SARS-CoV-2 during its invasion into host cells by interacting with ACE2, spike protein is an ideal choice of vaccine immunogen. Here we designed both SDTM and RBD as vaccine components shown in Fig. 1A. SDTM derives from ectodomain of spike protein with two mutation sites and conjugates with a T4 fibritin trimerization motif, whereas monomer RBD is a part of S1 domain, including RBD region and SD1 region. Both recombinant proteins were expressed by CHO cells, and purified by multi-step chromatography. The purified proteins were characterized by SDS-PAGE, size-exclusion chromatography (SEC), and biolayer interferometry (BLI). SDS-PAGE analysis showed that the purity of both SDTM and RBD are over 90% (Fig. 1B). The SEC results showed that retention time of purified SDTM was similar with a 670 kDa standard protein, thyroglobulin, indicating the SDTM was in its trimeric form (Fig. 1C). The binding affinity of RBD and SDTM to human ACE2 was determined by BLI, with dissociation constant (K D ) of 1.53 Â 10 -8 M and 3.46 Â 10 -9 M, respectively ( Fig. 1D and 1E). In addition, both RBD and SDTM were recognized by highly diluted human convalescent sera (HCS) (Fig. 1F), indicating their good antigenicity.

Immunogenicity of candidate vaccine in mice
Purified subunit proteins alone are poorly immunogenic as they are lack of immunostimulatory capacity. Adjuvants are capable of enhancing vaccine-induced adaptive immune responses by triggering innate immune responses. In addition, as it was reported nAb epitopes mainly locate in the RBD region of spike protein, we expected to compare the immunogenicity of SDTM and RBD. We performed a mouse study to select the best combination of antigen and adjuvant. In this study, BALB/c mice were immunized intramuscularly twice at a 3-week interval with 5 lg SDTM or RBD, formulated with Alum, CpG or Alum/CpG ( Fig. 2A). Binding antibody titers and nAb titers were detected by ELISA and pseudovirus-based neutralization assay 2 weeks post second immunization, respectively ( Fig. 2B and 2C). Results showed that all SDTM groups induced high titers of anti-RBD antibody with geometric mean titers (GMT) ranging from 3.9 Â 10 4 to 1.8 Â 10 6 . In contrast, RBD adjuvanted with Alum alone or Alum/ CpG induced high titers of anti-RBD antibody, 9.6 Â 10 4 (GMT) and 2.0 Â 10 6 (GMT), respectively. In consistent with ELISA results, all adjuvanted SDTM groups induced high titers of nAbs with GMT ranging from 624 to 51939. SDTM alone without adjuvant showed relatively poor immunogenicity. Unexpectedly, all RBD groups elicited weak nAb responses with the highest nAb titers of 3429 (GMT) in Alum/CpG adjuvanted group. These results indicate SDTM as vaccine component has advantage over RBD in terms of immunogenicity.
We then detected vaccine-induced nAb responses in SDTM groups via live virus-based neutralization assay (Fig. 2D). SDTM alone without adjuvant elicited weak neutralizing activity. By contrast, SDTM adjuvanted with either Alum, CpG or Alum/CpG elicited high titers (GMT) of nAbs, ranging from 1660 to 10240. In detail, the GMTs of nAb induced by SDTM/Alum/CpG group were 3-fold and 6-fold of those by SDTM/CpG group and SDTM/Alum group, respectively. Next, we determined the IgG subclasses of sera from SDTM groups (Fig. 2E). Results showed the ratio of IgG2a/ IgG1 in SDTM/CpG group or SDTM/Alum/CpG was reversed compared with SDTM alone group or SDTM/Alum group, which suggest CpG is able to divert SDTM induced humoral immune responses into Th1-associated.
Vaccine-induced cellular immune responses are essential to facilitate the production of high-quality antibodies and kill pathogen-invaded host cells. We sacrificed 5 mice in each SDTM groups to evaluate T cell responses 2 weeks post second immunization ( Fig. 2F and 2G). Compared with Alum/CpG group, neither SDTM alone group nor SDTM/CpG group elicited antigen-specific CD4 + T cell responses; whereas SDTM/Alum group and SDTM/ Alum/CpG group did, though the responses in SDTM/Alum group were weak. In this study, we rarely detected CD8 + T cell responses in Alum adjuvanted groups.
We then monitored the persistence of vaccine-induced immune responses, as this characteristic determines the quality of candidate vaccine in long term protection against SARS-CoV-2. We measured the nAb titers of vaccinated mouse sera from SDTM groups at different timepoints by pseudovirus-based neutralization assay ( Fig. 3A). Results showed that nAb responses peaked at 3 weeks post second immunization and declined thereafter. However, the nAb titers in SDTM/Alum/CpG group maintained at 12,380 (GMT) even 6 months post second immunization, which indicates that vaccine-induced immune responses are with strong magnitude and good longevity in mice. Subsequently, we sacrificed 5 mice from each group and detected the SDTM-specific memory B cells 6 months post second immunization ( Fig. 3B and 3C). In accordance with data from nAb responses, mice in SDTM/Alum/CpG group maintained the highest number of SDTM-specific memory B cells in the spleen, which produce nAbs against SARS-CoV-2.
In summary, SDTM combined with adjuvants elicited higher titers of nAbs than that elicited by adjuvanted RBD. Furthermore, SDTM formulated in Alum/CpG adjuvants induced significantly higher levels of both antibody responses and T cell responses than those with either Alum or CpG alone. This result is consistent with previous report [18] and supports the use of dual adjuvants. In addition, candidate vaccines, especially SDTM formulated in Alum/CpG adjuvants, showed strong and long-lasting immune responses.

Immunogenicity of candidate vaccine in nonhuman primates
As the immune system of nonhuman primate is closer to that of human being, we next evaluated candidate vaccines with three different formulations in cynomolgus monkeys (Fig. 4A). Formulation 1 contains 50 lg SDTM/500 lg Alum/500 lg CpG, Formulation 2 contains 25 lg SDTM/500 lg Alum/500 lg CpG, and Formulation 3 contains 50 lg SDTM/500 lg Alum/250 lg CpG. Each formulation has a volume of 500 lL. In this study, each group of 5 cynomolgus monkeys were immunized twice at a 3-week interval with these 3 formulations. Blood was collected from individual monkeys 1 week ahead of immunization, 3 weeks, 5 weeks, 15 weeks and 27 weeks post immunization. Binding antibody and nAb titers were detected by ELISA and neutralization assays, respectively, at different timepoints ( Fig. 4B-D). Results showed that all three formulations induced high titers of both binding and neutralizing antibodies 3 weeks post first immunization. Two weeks after the second shot, the antibody responses became much stronger, ranging from 6566 (GMT) to 9060 (GMT) of binding antibody titers, from 7760 (GMT) to 10,240 (GMT) of live virusbased nAb titers. However, there was no significant difference among these three formulations in antibody responses. As a reference point, the live virus-based nAb titers of human convalescent sera (HCS) were 1881 (GMT).
We then evaluated the cellular immune responses in vaccinated NHPs. PBMCs were separated from whole blood of individual monkeys, which was collected one week before the first immunization and two weeks post the second immunization. SDTM specific T cell responses were detected by intracellular cytokine staining assay (Fig. 4E). Results showed that all three formulations induced significant CD4 + T cell responses after immunization. However, there was no SDTM-specific CD8 + T cells elicited, which is in accordance with the results in mouse study. At the beginning of development of SARS-CoV-2 vaccine, vaccine-enhanced disease was concerned as a potential risk, which was suggested to be related with vaccine-induced Th2-associated responses. Here, we tested the secretion of Th1-associated cytokines (IFN-c, IL-2, TNF-a) and Th2-associated cytokines (IL-4, IL-5, IL-6) from stimulated PBMC supernatants (Fig. 4F). Results showed that stimulated PBMC mainly secret IFN-c, IL-2 and TNF-a. No Th2 associated IL-4 or IL-5 was detected. Though the secretion of IL-6 in supernatants   were detected, there was no significant difference between samples processed pre-and post-immunization. We suggested the origin of IL-6 was from innate immune cells activated by unknown factors. These results demonstrated the vaccine candidates induce only Th1-associated T cell responses.
We also monitored the persistence of vaccine-induced immune responses by pseudovirus-based neutralization assay in NHPs (Fig. 4G). Compared with the nAb responses 2 weeks post second immunization, the nAb titers (GMT) in each group declined rapidly 12 weeks post second immunization, ranging from 6859 to 7744. However, the nAb titers (GMT) only declined slightly at 24 weeks post second immunization, ranging from 5379 to 6382, compared with that at 12 weeks post second immunization. The results indicate that our candidate vaccines are able to induce persistent nAb responses in NHPs.

Immunogenicity and protective efficacy of candidate vaccines in SARS-CoV-2/hamster model
We demonstrated the candidate vaccines are able to elicit robust immune responses in mice and NHPs. However, the ultimate goal of developing SARS-CoV-2 vaccine is to prevent viral infections and associated disease. Thus, we conducted a challenge study in golden Syrian hamsters (Fig. 5A), which is a good model to evaluate the protective efficacy of candidate vaccines [19][20][21][22]. In this study, group of 10 hamsters were immunized intramuscularly with Alum/CpG adjuvanted SDTM, Alum/CpG adjuvanted RBD or Alum/CpG alone on Day 0 and Day 21, and challenged with SARS-CoV-2 (10 5 TCID 50 , i.n.). Blood was collected on Day 35 to prepare serum samples, and subjected to both ELISA and live virus-based neutralization assay ( Fig. 5B and 5C). Results showed that RBD combined with Alum/CpG hardly induced binding antibody responses, whereas SDTM combined with Alum/CpG induced high titers of binding antibodies (GMT = 62500) in golden Syrian hamsters. In addition, only Alum/CpG adjuvanted SDTM group induced high nAb titers (GMT = 2195). For reference, the virus nAb titers of human convalescent sera were 5100 using the same live virus-based neutralization assay performed by the same laboratory. Subsequently, hamsters were sacrificed seven days post infection (Day 49), virus loads of lungs and nasal turbinates were measured by quantitative real-time polymerase chain reaction (qRT-PCR) (copies/mL) (Fig. 5D and 5E). We observed a significant reduction in virus titers in lungs (3.00 log 10 ) and nasal turbinates (0.99 log 10 ) 7 days post intranasal virus infection in Alum/CpG adjuvanted SDTM group. However, there were no significant differences in virus titers in lungs and nasal turbinates between Alum/CpG adjuvanted RBD group and Alum/CpG alone group.
We then evaluated the lung histopathology 7 days post intranasal virus infection by H&E staining. Hamsters in all groups showed varying degrees of lung inflammation with thickened alveolar septa. The lung pathology was scaled as mild, moderate or severe in this study (Fig. 5F). In Alum/CpG alone group, as a control, 4 of 6 hamsters showed severe lung inflammation; 2 of 6 hamsters showed moderate lung inflammation (Fig. 5G). In Alum/CpG adjuvanted SDTM group, by comparison, 3 of 6 hamsters showed moderate lung inflammation; 3 of 6 hamsters showed mild lung inflammation (Fig. 5G). However, there were no significant reduction in lung pathology in Alum/CpG adjuvanted RBD group compared with Alum/CpG alone group, 4 hamsters developing severe lung inflammation and 2 hamsters developing moderate lung inflammation (Fig. 5G).
In summary, Alum/CpG adjuvanted SDTM induced robust nAb responses. And the candidate vaccine can significantly reduce the viral loads in lungs and nasal turbinates after SARS-CoV-2 infection in hamsters. In addition, hamsters showed reduced lung pathology after immunized with Alum/CpG adjuvanted SDTM. Taken together, these results demonstrated significant efficacy of Alum/ CpG adjuvanted SDTM in golden Syrian hamster challenge model.

Cross-reactivity with variants
As the COVID-19 pandemic goes viral, new variants emerged. Some of the new variants showed enhanced infectivity and/or immune escape [23]. This phenomenon is probably a challenge for the development of efficacious SARS-CoV-2 vaccines. Here we thawed sera collected 2 weeks after second immunization from NHPs and evaluated the cross-reactivity of vaccine-induced neutralizing antibodies against pseudoviruses that bear spike proteins from variants, including B. 1.1.7, B.1.351, B.1.617, B.1.429 and P.1 strains (Fig. 6A). In the pseudovirus-based neutralization assay, results showed candidate vaccine-induced neutralizing antibodies cross-reacted with all these variants. However, the nAb titers against variants reduced in different degree compared with that of wild type (Fig. 6B)  data are in line with other studies that variants bearing E484K mutation in spike protein showed immune evasion in larger extent [24][25][26]. B.1.617 variant, first arising in India and currently circulating in many countries extensively, also showed significant decrease in neutralization relative to wild type. These results indicate SARS-CoV-2 viruses evolved and are still evolving to escape immune system.

Discussion
Currently, several SARS-CoV-2 vaccines have been approved and demonstrated to take into effect in controlling the COVID-19 pandemic. However, there is still a huge gap between the limited supply and the large population, especially in the developing countries. Subunit vaccines have been widely developed with a good record of safety and efficacy. Here we comprehensively evaluated the immunogenicity and protective efficacy of a recombinant spike protein vaccine adjuvanted with aluminium hydroxide and CpG 7909 in mouse, hamster and cynomolgus monkey models. Preclinical data demonstrated the vaccine candidate is promising, as it is able to induce potent nAb responses in all these animals and showed protective efficacy in hamsters after SARS-CoV-2 challenge. These data are in accordance with published data of similar vaccines developed by Clover Biopharmaceuticals and Medigen Vaccine Biologics Corporation [27][28][29][30], in which CpG 1018 was used.
Both CpG 7909 and CpG 1018 are type B CpG ODN, which enhance vaccine-induced adaptive immune responses by stimulating human B cells and plasmacytoid dendritic cells. Previous studies on HBV vaccine development showed that when HBsAg adjuvanted with high dose CpG 1018 (3 mg/dose) elicited higher titers of antibody responses than adjuvanted with CpG 7909 [31][32]. However, the immune persistence of CpG 7909 adjuvanted vaccine seems to be better than CpG 1018 adjuvanted [31][32]. Thus, we chose CpG 7909 as the vaccine component. In preclinical study, our candidate vaccine induced comparable nAb responses with those adjuvanted with CpG 1018 at the peak. As there is no available data on longitudinal analysis of CpG 1018 adjuvanted SARS-CoV-2 vaccine-induced immune responses so far, we cannot confirm whether CpG 7909 adjuvanted SARS-CoV-2 vaccine could induce more durable immune responses than CpG 1018 adjuvanted. However, our candidate vaccines present excellent immune persistent in both mice and non-human primate models. Outstandingly, the decline of nAb titers from 12 weeks post second immunization to 24 weeks post second immunization is approximately 20% in our NHP study. In comparison, a RBD nanoparticle-based study adjuvanted with AS03 showed more than 2.5-fold decrease in nAb titers from 11 weeks post second immunization to 23 weeks post second immunization [33].
In the course of SARS-CoV-2 vaccine development, both Spike protein and RBD were utilized as vaccine immunogens and demonstrated to be immunogenic [27,[34][35][36][37][38]. In this study, we compared the immunogenicity between trimeric spike protein and monomer RBD. Trimeric spike protein induced more potent nAb responses than monomer RBD in both mouse and hamster models. Besides, only trimeric spike protein showed protective efficacy in hamster challenge model. These results are not unexpected as spike protein contains more neutralizing epitopes than RBD. In addition, spike protein as a trimer resembles its native structure, making it highly immunogenic. The unexpected is that RBD-induced nAbs were detected in mice but not golden Syrian hamsters. This phenomenon maybe explained by the difference of MHC class II molecule diversity and/or TCR repertoire between hamsters and mice, as the production of antibodies needs help from CD4 + T cells [39]. Our results are in line with a previous report [40], which showed RBD has limited immunogenicity in mice. In addition, the report found RBD is more immunogenic in non-human primates than in mice, analogous to the finding that RBD is more immunogenic in mice than golden Syrian hamsters in our study.
The emergence of new variants of SARS-CoV-2 is of great concern. Our results showed prototype SARS-CoV-2 vaccine can cross-neutralize variants. However, the nAb titers against P.1, B.1.351 and B.1.617 variants are significantly reduced compared with that against prototype. Other studies also reported the similar phenomenon [26]. These results suggest that vaccineinduced neutralizing antibodies may not be fully broadly neutralizing. Thus, it is necessary to develop a COVID-19 vaccine that is effective to prevent infections of major variants of SARS-CoV-2 continuously.
In summary, our vaccine candidate showed high immunogenicity and good protective efficacy in preclinical studies. These positive results support further clinical trials and paved the way for developing next generation vaccine against variants of SARS-CoV-2.

Protein production and characterization
To express the prefusion spike protein, a gene encoding the ectodomain of SARS-CoV-2 S (GenBank: MN908947) with proline substitutions at residues 986 and 987, a ''GGSG" substitution at the furin cleavage site (residues 682-685), a C-terminal T4 fibritin trimerization motif was synthesized and cloned into the mammalian expression vector. To express the RBD (RBD-SD1), residues 332-591 of SARS-CoV-2 spike protein were cloned into a different vector and a 6 Â His tag was added to the N terminus. Both expression vectors were transfected into Chinese hamster ovary (CHO) cells. SDTM was purified from clarified supernatant through low pH for preventative viral inactivation (VI), followed by three different chromatography steps to remove host cell DNA, host cell proteins, and any other impurities, and finally nanofiltration as a preventative viral removal (VR) step. While his-tagged RBD was purified using immobilized metal affinity chromatography (IMAC). SDS-PAGE and Size-Exclusion HPLC were run to check the purity of both proteins and the trimeric conformation of SDTM.
Biolayer interferometry (BLI) assays were performed on a GatorPrime (GatorBio) instrument at 30°C with shaking at 1,000 rpm. The Fc-tagged human ACE2 (Sino Biological, Cat. 10108-H02H) was immobilized to human Fc (HFC) Probes (Gator Bio, Cat. 160003) at 10 lg/mL in 10 Â Kinetics Buffer. SDTM and RBD were two-fold and three-fold serially diluted in 10 Â kinetics buffer respectively prior to the measurement of association and dissociation rate. The data were baseline subtracted prior to fitting performed using a 1:1 binding model. Mean k on , k off values were determined with a global fit applied to all data.

Mouse and nonhuman primate studies
Female specific pathogen free (SPF) BALB/c mice, 6-8 weeks of age, were obtained and housed in Shanghai SIPPR-BK Lab Animal Co., Ltd. All mouse experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Shanghai SIPPR-BK Lab Animal Co., Ltd. All mice were intramuscularly immunized on Day 0 and Day 21 respectively. Blood was collected at different timepoints to evaluate antibody responses via ELISA and neutralization assays. For some experiments, a subset of mice from groups was sacrificed to assess T cell responses by intracellular cytokine staining (ICS) and B cell memory by B cell ELISPOT, respectively.

Enzyme-linked immunosorbent assay (ELISA)
For the measurement of binding antibody titers, RBD or SDTM proteins were diluted in phosphate-buffered saline (PBS) at 1 lg/ mL and used to coat 96-well microplates (Corning, Cat. 9018) at 100 lL/well overnight at 2-8°C. Plates were washed 5 times with PBST (PBS containing 0.05% Tween 80) and blocked with 5% milk in PBST at room temperature for 2 h. Sera were 2-or 5-fold serially diluted with 2% milk in PBST. Diluted sera were added into antigen coated plates and incubated at 37°C for 1 h, then incubated with horseradish peroxidase (HRP) conjugated Goat anti-mouse IgG (Bio-Rad, Cat. 1706516) at 1:10000 dilution, Goat anti-hamster IgG (Invitrogen, Cat. PA1-28823) at 1:15000 dilution or Goat anti-monkey IgG (BETHYL, Cat. A140-102P) at 1:5000 dilution at 37°C for 1 h. The plates were washed 5 times with PBST after each incubation. Then 100 lL TMB substrate system (SeraCare, Cat. 5120-0038) was added into each well and incubated for approximately 15 min. The color development was stopped by adding 50 lL 2 M sulfuric acid. Plates were read at 450 nm and 620 nm for absorbance by using a microplate reader (Molecular Devices, SpectraMax iD3). The endpoint titer is defined as the reciprocal of the highest serum dilution providing an OD 450nm-620nm value above 2.1-fold of the negative control. In situations where the OD 450nm-620nm value of the negative control is<0.05, regard it as 0.05.
For detecting IgG2a and IgG1 of sera from immunized mice, all procedures are the same as the above, except the following. After diluted sera were incubated with coated antigen, plates were washed 5 times with PBST and added anti-mouse IgG2a or IgG1 (Sigma, Cat. IS02) at 1:1000 dilution. Then, incubated another 1 h at room temperature before the reaction with HRP conjugated rabbit anti-goat IgG (Sigma, Cat. A5420) at 1:5000 dilution.

Pseudovirus-based neutralization assay
Pseudoviruses containing spike proteins from SARS-CoV-2 were prepared using a replication-deficient VSV-based rhabdoviral pseudotyping system expressing firefly 485 luciferase (VSV-dGfluc), which was obtained from State Key Laboratory of Virology, Wuhan University. SARS-CoV-2 pseudovirus were generated according to a previously described protocol [41]. Briefly, Vero-E6 cells were transfected with the plasmids overexpressing SARS2-CoV-2 spike proteins (pCAGGS-SARS2-S-dc) using Lipofectamine 2000 reagent (Invitrogen, Cat. 11668-027). After 48 h, the transfected cells were transduced with VSV-dG-fluc reporter viruses for 1 h at 37°C. Transduced cells were washed with PBS for 5 times and then replenished with fresh culture medium with anti-VSV monoclonal antibody (1:500 dilution) to neutralize the infectivity of the residual VSV-dG-fluc. SARS-CoV-2 pseudoviruses containing culture supernatants were collected 24 h later. The supernatants then subjected to centrifugation for 5 min at 2,000 rpm, and stored at À 80°C. The titration of SARS-CoV-2 pseudovirus was performed by the 50% tissue culture infective dose method (TCID 50 ) according to Reed-Muench.
For neutralization assay, 60 lL of heat inactivated sera were 2fold serially diluted and added into 96-well plates (Corning, Cat. 3599), and 60 lL SARS-CoV-2 pseudoviruses diluted to contain 800 TCID 50 were added into the plates. The mixture was incubated at 37°C for 1 h, and then added to BHK-21-hACE2 cells in a 96-well white plate with clear bottom (Corning, Cat. 3610). Luciferase activity was measured 24 h later using luciferin-containing substrate (PerkinElmer, Cat. 6066769). The neutralizing titer was calculated by the dilution number of 50% inhibition condition. The neutralizing titer was calculated according to Reed-Muench method.

Live virus-based neutralization assay
The Live virus-based neutralization assay was performed by Biosafety Level 3 Laboratory, Shanghai Medical College, Fudan University and the Institute of Laboratory Animal Science, Chinese Academy of Medical Sciences (for hamster study only). All SARS-CoV-2 live virus-related experiments were conducted in the BSL-3 laboratory. Briefly, medium containing serum at varying dilutions in 96-well plates was pre-incubated with an equal volume of live SARS-CoV-2 solution diluted to contain 100 TCID 50 . After neutralization in a 37°C for 1 h, the mixture was added to Vero E6 cells in a 96-well plate. After 3-5 days' incubation at 37°C, cytopathic effect (CPE) of each well was recorded under microscopes, and the neutralizing titer was calculated by the dilution number of 50% protective condition.

Intracellular cytokine staining (ICS) assay
Splenocytes and PBMCs were prepared from mice and NHPs respectively. Splenocytes or PBMCs (1 Â Briefly, the splenocytes were pre-incubated for 72 h in the presence of 1 lg/mL R848 and 10 ng/mL rmIL-2 to enrich memory B cells. Wells of PVDF-based membrane plates were coated with 150 lg/mL SDTM protein overnight at 2-8°C. Then plates were washed and blocked with RPMI 1640 containing 10% FBS. 3 Â 10 5 stimulated splenocytes were added into each well after removing secreted antibodies by two-round wash to incubated at 37°C 5% CO 2 incubator. Following a 24-hour incubation, the cells were removed, and the plates were washed 5 times with PBS and incubated with biotinylated anti-mouse IgG antibodies for 2 h. Next, plates were washed 5 times and incubated with alkaline phosphatase (AKP) conjugated streptavidin (BD, Cat.554065) at 1:2000 dilution for another 1 h. Then plates were washed 5 times with PBS and spots were visualized by adding 100 lL/well BCIP/ NBT substrate solution (Thermo Fisher Scientific, Cat. 34042). Plates were left to dry in the dark. Spots were counted using ELI-SPOT reader (CTL, S6).

Cytometric bead array (CBA)
Cytokine secretion from stimulated PBMCs was determined by Non-Human Primate Th1/Th2 Cytokine kit (BD, Cat. 557800) according to manufacturer's instruction manual, with minor modifications. Briefly, similar to ICS assay above, 1 Â 10 6 NHP PBMCs were seeded into 96-well plates and stimulated with 1% DMSO, peptide pool of spike protein or PMA & Ionomycin. After 18-hour incubation, cell culture was centrifugated and supernatant was harvested. Then 50 lL standards or supernatants from each sample were incubated with 20 lL mixture of Capture Beads and 25 lL Detection Reagent at room temperature for 3 h, protected from light. The samples were washed with 1 mL Wash Buffer. Finally, samples were resuspended in 150 lL Wash Buffer and analyzed samples with BD FACSCantoII flow cytometry.

Hamster challenge study
SPF golden Syrian hamsters, 6-10 weeks of age, were obtained from Beijing Vital River Laboratory Animal Technology Co., Ltd. and housed in Institute of Laboratory Animal Science, Chinese Academy of Medical Sciences (CAMS). All hamster experiments were approved by the IACUC of the Institute of CAMS. Hamsters were intramuscularly immunized twice on Day 0 and Day 21 respectively. On Day 35, blood from all hamsters was collected to detect antibody responses. On Day 42, all hamsters were challenged intranasally with 10 5 TCID 50 SARS-CoV-2, isolate of SARS-CoV-2/WH-09/human/2020/CHN. On Day 49, a subset of hamsters in each group was euthanized for detecting viral loads of lungs and nasal turbinates by qRT-PCR and evaluating lung histopathology by hematoxylin and eosin (H&E) staining. Both qRT-PCR and H&E staining were performed using published protocols [19], described briefly in the following paragraphs.
To detect tissue viral loads by qRT-PCR, total RNA was extracted from lung and turbinate tissue homogenates by the RNeasy Mini Kit (Qiagen) according to manufacturer instructions. Reverse transcription was completed by the PrimerScript RT Reagent Kit (TaKaRa) and qRT-PCR was conducted utilizing the PowerUp SYBR Green Master Mix Kit (Applied biosystems). The forward and reverse primers targeting SARS-CoV-2 envelope protein (E) gene for qRT-PCR were 5 0 -TCGTTTCGGAAGAGACAGGT À3 0 and 5 0 -GCGCAGTAAGGATGGCTAGT À3 0 , respectively. qRT-PCR was conducted by an ABI 3730 DNA sequencer (Applied Biosystems) under the following reaction conditions: 1) 50°C for 2 min; 2) 95°C for 2 min; 3) 40 cycles of 95°C for 10 s, and 60°C for 30 s; 4) 60°C for 1 min; and 5) 95°C for 45 s.
To evaluate lung histopathology by H&E staining, lung tissues from hamsters were fixed by 10% buffered formalin and processed for paraffin embedding. Paraffin blocks were cut into 5-lm sections and stained with hematoxylin and eosin. Lung pathology, including overall lesion extent, pneumocyte hyperplasia, and inflammatory infiltrates, was assessed and classified into 3 types: mild, moderate and severe lung inflammation.

Statistical analysis
All statistical analyses were performed using GraphPad Prism. Antibody titers and viral loads were log 10 transformed prior to statistical analysis. Differences among two groups were analyzed using Mann-Whitney test. Values of p < 0.05 are considered statistically significant.