All experiments and methods were performed in accordance with relevant guidelines and regulations. Experiments using mice were approved by the Research Ethics Review Committee of the Shanghai Public Health Clinical Center Affiliated to Fudan University.

The full-length s gene sequence of the reference SARS-CoV-2 strain was optimized according to the preference of human codon usage and synthesized by Genewiz (Genewiz Biotech Co., Ltd., Suchow, China). The codon optimized spike gene was subcloned into a eukaryotic expression vector (pJW4303, kindly gifted by Dr. Shan Lu at the University of Massachusetts). And the sequence of inserted gene was verified by Sanger sequencing (Sangon Biotech Co., Ltd., Shanghai, China). An EndoFree plasmid extraction kit (Cat# 12391, Qiagen, Hilden, USA) was used to prepare the recombinant plasmid for mouse vaccination.

Purified SARS-CoV-2 S protein (Cat# 40591-V08H, Sino Biological, China) was reconstituted with PBS at a concentration of 80μg/ml. Then, it was mixed with a commercialized alum adjuvant (Cat# 77161, ThermoFisher Scientific Co., Ltd., USA) at a ratio of 1:1 (v/v). The final concentration of the purified S protein was 40μg/ml.

Female C57BL/6 mice, 6 to 8 week-old, were housed under a specific-pathogen free (SPF) environment. As shown in Figure 1 , the mice were immunized intramuscularly with the pJW4303-CMV-S DNA vaccine (50μg/mouse) or the alum adjuvanted S protein subunit vaccine (2μg/mouse) for three times at an interval of 2 weeks. The control group was inoculated with PBS. Peripheral blood was collected at 2 weeks post each vaccination. Five weeks post the third vaccination, the mice were euthanized. Peripheral blood, bronchial alveolar lavage fluid (BALF) and splenocytes were collected for assays of antigen-specific immune responses.

An in-house enzyme-linked immunosorbent assay (ELISA) was developed to measure SARS-CoV-2 RBD specific binding antibody responses. High-binding 96-well EIA plates (Cat# 9018, Corning, USA) were coated with purified SARS-CoV-2 RBD protein (Cat# 40592- V08B, Sino Biological, China) at a final concentration of 1µg/ml in carbonate/bicarbonate coating buffer (30mM NaHCO₃,10mM Na₂CO₃, pH 9.6). Subsequently, the plates were blocked with 1×PBS containing 5% skimmed milk for 1 hour at 37°C. Next, 100μl of serially diluted mouse serum or plasma was added to each well. After 1-hour incubation at 37°C, the plates were washed with 1×PBS containing 0.05% Tween20 for 5 times. Then, 100μl of an HRP labeled goat anti-mouse IgG antibody (Cat# 115-035-003, Jackson Immuno Research, USA) diluted in 1×PBS containing 5% skimmed milk were added to each well and incubated for 1 hour at 37°C. After a second round of wash, 100μl of TMB substrate reagent (Cat# MG882, MESGEN, China) was added to each well. 15 minutes later, the color development was stopped by adding 100μl of 1M H₂SO₄ to each well and the values of optical density at OD450_nm and OD630_nm were measured using 800 TS microplate reader (Cat# 800TS, Biotek, USA). The cut-off value was defined as 2-fold of the average OD450-630 of PBS group at 1:100 dilution.

The binding antibody titers against the full-length S protein were measured using a method of competitive ELISA (14), which can help to avoid the interference of pre-existing cross-reactive antibody responses against S2. Briefly, high-binding 96-well EIA plates were coated with purified SARS-CoV-2 S protein (Cat# VISC2-S002, East Mab, China) at a final concentration of 1µg/ml in carbonate/bicarbonate coating buffer. The experiment procedure was generally similar with the aforementioned in-house ELISA assays, except that the diluted mouse serum were incubated with a synthesized peptide (P144, SFKEELDKYFKNHT) (10μg/ml) for 1 hour at 37°C before adding into the coated EIA plates.

Avidity of Ag-specific Ab was determined by avidity ELISA as reported (15–17) with minor modifications. Briefly, plates were coated as the regular ELISA assay described above. Diluted mouse sera were added into each well. After 1-hour incubation, ELISA plates were washed with washing buffer and incubated with 1.5M NaSCN or PBS for 15 minutes at room temperature and then immediately washed with washing buffer. Ab avidity index was defined as the ratio of the OD value of a sample with 1.5M NaSCN treatment versus the OD value of the same sample with PBS treatment.

Freshly isolated splenocytes or peripheral blood mononuclear cells were plated into round-bottom 96-well plates (2×10⁶ cells per well) and incubated with either R10 (RPMI1640 with 10% FBS) or R10 containing synthesized peptides encompassing the full length of S protein (0.66μg/ml for each peptide) (Synthesized by Gill Biochemistry Co., Ltd., Shanghai, China). Two hours later, brefeldin A and monensin were added to each well at final concentrations of 1μg/ml and 1μM, respectively. Another 12 hours later, the cells were washed and stained sequentially with Live/Dead dye (Fixable Viability Stain 510, cat# 564406, BD Pharmingen), surface markers (PE/Cyanine7-labeled anti-mouse CD3, cat# 100220, BioLegend; APC-labeled anti-mouse CD4, cat# 100412, BioLegend; PE-labeled anti-mouse CD8, cat# 100708, BioLegend) and intracellular markers (BV421-labeled anti-mouse IFN-γ, cat# 505830, BioLegend; FITC-labeled anti-mouse IL-2, cat# 503806, BioLegend; BV711-labeled anti-mouse TNF-α, cat# 506349, BioLegend). Stained samples were analyzed using a BD Fortessa flow cytometer and data were analyzed with the FlowJo software version 10 (TreeStar Inc., Ashland, OR, USA). The gating strategy was shown in Supplementary Figure 1 .

All statistical analyses were performed using GraphPad Prism 9 (GraphPad Software, Inc., La Jolla, CA, USA). Comparisons between two groups were conducted by the method of t-test. P<0.05 was considered as statistically significant.

C57BL/6 mice were immunized via three different inoculation strategies ( Figure 1 ): Mice in the group of site-fixed inoculation (SFI) were immunized by injection into the tibialis anterior of the same limb. Mice in the group of successively site-translocated inoculation (SSTI) were immunized by injecting into the left hind limb, the right hind limb and the left hind limb, sequentially. The mice in the group of simplified translocating inoculation (STI) were injected into the left hind limb for the 1^st and 2^nd shots, and into the right hind limb for the 3^rd shot. Peripheral blood samples were collected at 2 weeks post each immunization. Five weeks post the last vaccination, the mice were euthanized. Serum, BALF and splenocytes were collected for detection of antigen-specific immune responses ( Figure 1 ).

Our data showed that the serum levels of RBD binding IgG titers induced by the DNA and the protein subunit vaccines were not improved by either the SSTI or the STI strategies ( Figures 2A, B ). Of note, the average binding antibody responses elicited by the protein subunit vaccine for the STI and SSTI groups tended to be lower than that of the SFI group ( Figure 2B ). Besides, we found that the binding antibody titers reached the plateau after two shots of DNA vaccine, while it required at least 3 shots for the protein subunit vaccine to boost the binding antibody responses. Similarly, binding antibody titers against the full-length S protein also showed that there was no significant difference among different inoculation strategies ( Supplementary Figure 2 ). The avidities of RBD binding antibodies determined at 5 weeks post the 3^rd immunization were similar among different strategies ( Figure 2C ). The levels of RBD binding IgG in BALF were different between mice immunized with the DNA vaccine and the protein vaccine. However, no significant difference was observed among different inoculation strategies for either the DNA or the protein subunit vaccine ( Figure 2D ). The levels of RBD binding IgA in BALF were very low in mice inoculated with either the DNA or the protein vaccine ( Supplementary Figure 3 ).

To characterize the influences of different inoculation strategies on cellular response kinetics during the vaccination process, we collected mouse peripheral blood lymphocytes (PBL) through mandibular vein puncture at 2 weeks post the 2^nd immunization and 2 weeks post the 3^rd immunization, respectively. S protein specific T cell responses were measured using flowcytometry. Our data showed that the cellular immune responses elicited by the alum adjuvanted S protein were obviously weaker than those induced by the S DNA vaccine ( Figures 3 and 4 ). Both the CD8⁺ and CD4⁺ T cell responses at the two time points showed no significant differences among groups of mice immunized with the protein vaccine ( Figure 4 ). For mice immunized with the SARS-CoV-2 S DNA vaccine, the average frequency of circulating specific IFN-γ secreting CD8⁺ T cells in the SSTI group (1.310 ± 0.393, 6) was significantly higher than that of the SFI group (0.621 ± 0.177, 7) (P=0.0015) at 2 weeks post the 3^rd immunization ( Figure 3B ), whereas the STI strategy did not significantly improve the frequency of circulating IFN-γ⁺CD8⁺ T cells at this time point, suggesting that changing the inoculation site for the second shot of DNA vaccine was essential to induce optimal T cell responses. Interestingly, our data showed that specific CD8⁺ T cell responses elicited by the DNA vaccine were not significantly improved by the SSTI strategy at 2 weeks post the 2^nd shot ( Figure 3A ), implying that the SSTI improved DNA induced T cell responses in an incremental mode. Compared with the CD8⁺ T cell responses, the frequencies of S protein specific IFN-γ secreting CD4⁺ T cells in PBL were weak at these time points and showed no significant difference among different inoculation strategies ( Figures 3C, D ).

The incremental improvement of T cell response mediated by the SSTI strategy was verified by an independent experiment using a DNA vaccine encoding a CD8⁺ T cell epitope derived from HIV (18). We used the DNA vaccine immunizing mice via two different inoculation strategies: SFI and SSTI ( Supplementary Figure 4A ).Our data showed that antigen-specific CD8⁺ T cell responses in SSTI group were significantly higher than that of the SFI group (P=0.0484) at 2 weeks post the 3^rd immunization while not at 2 weeks post the 2^nd vaccination ( Supplementary Figure 4B ).

To further investigate the influences of different inoculations on the cellular immune responses elicited by the S DNA vaccine, we euthanized the mice at 5 weeks post the 3^rd immunization and measured the multifunctionality of S protein specific T cells in spleens using flow cytometry. Compared with the SFI group, the average S protein specific T cell responses (IL-2, TNF-α and IFN-γ) of both the SSTI and the STI groups tended to be higher ( Figures 5A, B ), of which the frequencies of IFN-γ⁺CD8⁺ T cells ( Figure 5A ) and IFN-γ⁺CD4⁺ T cells ( Figure 5B ) were found to be significantly improved in the SSTI group. To evaluate the magnitudes of S protein specific T cells more accurately, we calculated the integrated median fluorescence intensity (iMFI) according to previous reports (18, 19). Our data showed that the iMFI of S protein specific IFN-γ⁺CD8⁺ T cells were significantly improved only by the SSTI strategy ( Figure 5C ). While, the iMFI of TNF-α⁺CD4⁺ cells and IFN-γ⁺CD4⁺ T cells were significantly improved by both the SSTI and STI strategies ( Figure 5D ). Next, the functional features of S protein specific CD8⁺ and CD4⁺ T cells were further delineated via Boolean gating. Our data showed that the mean frequency of IFN-γ⁺IL-2^-TNF-α^- CD8⁺ T cells in the SSTI group was significantly higher than those in the STI and the SFI groups ( Figure 6A ). The mean frequencies of IFN-γ⁺IL-2^-TNF-α⁺CD8⁺ T cells in both the SSTI and the STI groups were significantly higher than that in the SFI group ( Figure 6A ). Similarly, the mean frequencies of IFN-γ⁺IL-2^-TNF-α⁺ and IFN-γ⁺IL-2⁺TNF-α⁺CD4⁺ T cells were significantly higher in the SSTI and the STI groups than those in the SFI group ( Figure 6B ).