The animal studies were performed in accordance with the Guide for the Care and Use of Laboratory Animals (National Research Council). All animal procedures were reviewed and approved by the Animal Welfare and Research Ethics Committee of Changchun BCHT Biotechnology Co. (No. BCHT- AEEI-2023-157 and BCHT- AEEI-2023-451). All animals were euthanized using euthanasia systems by inhaling CO₂ after all experiments.

Specific pathogen-free female SD rats (age, 6-7 weeks; weight, 160–200 g) were purchased from Liaoning Changsheng Biotechnology Co., Ltd. (Liaoning, China) and maintained under standard approved conditions. The rats acclimated to the lab for 3-5 days were vaccinated until weights of 220-240 g (aged 9-10 weeks) were achieved, which was available to inject by GV-01 with applicable skin thickness.

To ensure that GV-01 precisely delivered the vaccine intradermally, we mixed the vaccines with the indicator, rhodamine. Following injection using the optimized PSI parameters, the rats (n = 2) were sacrificed, and the corresponding tissues were collected and photographed for the initial proof-of-concept.

Rats were randomly divided into 10 groups ( Table 1 ). Four groups were inoculated with a series dose of mRNA-LNP (30, 50 μg) and mRNA (30, 50 μg) per rat, respectively, via the IM route. Other groups were vaccinated by GV-01 with same doses. Unvaccinated control animals were administered an equivalent volume (100 μL) of LNP buffer. Vaccination was performed on days 0 and 15. Sera were collected 2, 4 (n = 6) and 6 (n = 3) weeks after each vaccination in all groups. The spleens (n = 3) for ELISpot detection and Single-cell analysis were collected four weeks after the boost.

RBD3-Fc mRNA expressing BA.4 RBD, Delta (B.1.617.2) RBD, prototype RBD (D614G strain), and human IgG Fc containing YTE mutation was designed as illustrated in Figure 1A , and comprised 5’ Cap1 structure, 5’UTR, coding sequence, 3’UTR structure, and poly A tail. Briefly, 50 µg/mL linear plasmid (Genscript Biotech Co., China) was used as the template for an IVT reaction at 37°C for 6 h with ribonucleotides and 2000U RNase Inhibitor. The cap analog (3-OMe-GAG) was co-transcribed into the mRNA strands. The IVT reaction was quenched by adding DNase I (100 units; GMP-E127, Novoprotein, China). All mRNAs were purified using the oligo-dT affinity method for subsequent detection (89.7%).

mRNA-loaded LNPs were prepared using a microfluidic method. Briefly, the mRNA was dissolved in RNase-free citrate buffer (50 mM, pH 4.0). Lipids including ((4-hydroxybutyl) azanediyl) bis (hexane-6,1-diyl) bis (2-hexyldecanoate), DSPC, cholesterol, and methoxypoly (ethylene glycol)-ditetradecylacetamide were dissolved in anhydrous ethanol at a molar ratio of 46.8:8.9:42.7:1.6. The lipid solution was then mixed with the mRNA solution at a 1:3 ratio using a microfluidic device and further diluted with Tris buffer. Finally, the hydrodynamic size mRNA-LNP was maintained in the range of 116.4 ± 0.65 nm, with PDI of 0.087 ± 0.023 and encapsulation efficiency of 94.5%.

RBD3-Fc mRNA was transfected into HEK293T cells in 6-well plates using Golden Tran-mRNA (Golden Transfer Technology, China; mRNA220521015), according to the manufacturer’s instructions. At 48 h after transfection, the cell lysates and supernatants were collected for western blot analysis. All cell lysates and supernatants were treated with sample buffer and subjected to 12% precast protein gel electrophoresis (Yeasen, China; 36261ES10). Thereafter, the gel was transferred onto PVDF membranes and incubated with 3% no-fat milk for 1 h at room temperature (RT), followed by HRP-linked human Fc antibody (Abcam, ab97225) for 2 h at RT. The samples were then developed using an HRP Substrate (Meilunbio, China, MA0186).

The needle-free injector of GV-01, which is a gas-propelled device, is presented in Figure 2A . The GV-01 can deliver 0-1 mL of liquid in a single injection. To optimize the device parameters for ID delivery of the vaccine, we first evaluated the driving force-dependent delivery speed and depth of the mRNA-LNP buffer using a polyacrylamide gel. This gel was selected owing to its similar properties to the dermis and subcutaneous tissues in the human skin. Polyacrylamide gels are also transparent and easy to observe. The driving forces (injection pressure) of GV-01 were measured using a high-speed image acquisition system, which was used to observe and track the movement of the push bar. An X190M-64G high-speed camera (Hefei Fuhuang Junda High-tech Information Technology Co., Ltd.) with a resolution of 1280 × 1024 and a small-frame acquisition frame rate of up to 700000 fps was employed in this study. Data were transmitted and collected using network cables with personal computer (PC) terminals. The delivery depth was calculated as the vertical distance between the surface of the gel and the end of the puncture channel at which the solution diffused in all directions, forming a fan-shaped or spherical radiation surface in the gel. Based on the above parameters, injection was carried out on the backs of SD female rats using GV-01 to further optimize the injection pressure and ensure successful ID delivery of the vaccine.

After ID inoculation using GV-01, at the local injection site, a white wheal with a micropore in the center, which was caused by the high-pressure liquid stream, developed on the skin of rats. The sizes of the white wheals were recorded at different time points using a vernier caliper. After inoculation, the animals were examined daily for general signs and local skin reactions at the injection sites. Body weight and food consumption were measured daily and weekly, respectively.

SARS-CoV-2 RBD-specific total immunoglobulin G (IgG) was detected in rat serum using ELISA. The 96-well plates were coated with SARS-CoV-2 prototype RBD (Vazyme, China, CG201) (0.5 μg/well) and incubated at 4°C overnight. The plates were washed three times with 0.05% PBS-T and blocked with 1% BSA at 37°C for 1 h. Diluted serum was added, and the plates were incubated for 2 h at 37°C. After four washes with 0.05% PBS-T, HRP-conjugated goat anti-rat IgG (Absin, China, abs20031) (1:5000) was added to detect antigen-specific total IgG. 3,3′,5,5′-tetramethylbenzidine (TMB) (KPL SeraCare, 5120-0047) was used for the color reaction, and H₂SO₄ (2 M) was used to terminate the reaction. The optical density was measured at 450 nm using a microplate reader.

To analyze the immunoglobulin isotype profile of sera from the vaccinated rats, standardized ELISA protocols were performed. Briefly, 96-well plates were coated with SARS-CoV-2 prototype RBD (5 mg/mL) and incubated overnight at 4°C. After blocking with 1% BSA at 37°C for 1 h, diluted serum was added and the plates were incubated for 2 h at 37°C. Horseradish peroxidase-conjugated goat anti-rat IgG1 (Sigma, PA1-84708) (1:10000), IgG2a (Sigma, PA1-84709) (1:10000), IgG2b (Sigma, PA1-84710) (1:10000), and IgG2c (Sigma, PA1-84711) (1:10000) were added to detect antigen-specific subclass IgG. The color was developed using TMB and the reaction was terminated using H₂SO₄ (2 M). Finally, the absorbance was measured at 450 nm using a microplate reader.

Rat spleens were aseptically extracted 4 weeks after the boost (n = 3). The spleens were washed with RPMI 1640 medium (Gibco, 72400120) and lysed with an ammonium-chloride-potassium (ACK) lysis buffer (Leagene, China; CS0001). Complete medium was then used to stop the lysis reaction. The cell mixture was centrifuged at 300 × g for 5 min to obtain the lymphocytes. Cells were diluted to 2×10⁶ cells/mL and incubated with RPMI 1640 medium (negative control), RBD protein (2.5/5 µg/mL), and PMA (Dakewe, China, positive control) for 48 h. The supernatant was collected and analyzed using the Rat IFN-γ T cell ELISPOT kit (U-CyTech, CT079-PR5) and Rat IL-4 T cell ELISPOT kit (U-CyTech, CT081-PR5), according to the manufacturer’s instructions, to detect IFN-γ and IL-4 secretion. The spots were inspected and counted using an ImmunoSpot S6 Analyzer (Cellular Technology Limited, OH, USA).

The SARS-CoV-2 pseudoviruses, including D614G, BA.4/5, and B.1.617.2, with Firefly Luciferase (Fluc), were manufactured by Vazyme (China). Neutralizing antibodies prevented pseudovirus-infected cells from expressing Fluc. Briefly, vaccine-immunized serum samples (deactivated at 56°C for 30 min) were serially diluted, incubated with 2×10⁴ TCID50/mL 50 μL/well pseudovirus (1 h at 37°C, in a 5% CO₂ incubator), and co-cultured with 2 × 10⁴ HEK293-ACE2 cells for 48 h. After reacting with the luminescent substrate, luminescence values were detected. Through a comparison with the luminescence value of the pseudovirus control group, the IC50 (dilution ratio of the antibody when the pseudovirus was suppressed by 50%) was calculated, indicating the neutralizing activity of the antibody against the pseudovirus.

CD45+ splenocytes were sorted by flow cytometry. The cells were then loaded into a 10× chromium microfluidic system based on the manufacturer’s guidelines. A set of libraries was obtained from 10× loaded samples: a 3 × gene expression messenger RNA library. The Illumina NovaSeq 6000 sequencing platform and PE150 sequencing mode were used to sequence the completed library; the sequencing amounts should be more than 20 k reads/cell. The scRNA-seq data analysis was performed by NovelBio Bio-Pharm Technology Co., Ltd. using the NovelBrain cloud analysis platform.

Statistical analyses were performed using one-way analysis of variance (ANOVA) or two-way ANOVA in GraphPad Prism 10. Multiple comparisons within a group were performed using Tukey’s multiple comparison test. Differences among groups were considered significant at p < 0.05.

To record the driving forces (injection pressure) of GV-01 ( Figure 2A ), changes in the movement of the push rod were captured and tracked using a camera during the entire injection process. An analysis software was used to obtain the relationship between the displacement of the push rod and time, as well as the movement speed and jet velocity of the push rod ( Figure 2B ). As shown in Figures 2B–C , the injection pressure was positively correlated with the jet velocity and injection depth. With increasing driving force, mRNA-LNP buffer could be delivered to deeper positions. To further confirm injection parameters for ID delivery, SD female rats (age, 9-10 weeks old) were injected with an mRNA-LNP vaccine containing rhodamine B. As shown in Figures 2D–G , 90-95 psi was sufficient to deliver the above solution into the ID layer on the backs of rats, while 100 psi enabled SC administration. GV-01 also achieved IM injection into rat legs when the psi was set to 60. Compared with IM administration using needle/syringe, the liquid delivered by GV-01 was more dispersed, aligning with the findings of previous studies (5, 36).

An mRNA vaccine candidate (RBD3-Fc) against the SARS-CoV-2 variants was developed based on our previously established mRNA-LNP platform. Spike RBDs from the prototype, BA.4/5 and B1.617.2, were selected as the encoded antigens. In addition, a human IgG1 Fc fragment with a YTE mutation was added to the ORF of the mRNA following the RBD sequence to achieve a long half-life of the antigen and increase its immunogenicity (37–40) ( Figure 1A ). Western blotting was performed using an HRP-conjugated human Fc antibody to evaluate the in vitro expression profile of RBD3-Fc mRNA in HEK293T cells. As expected, cell lysates and supernatants from mRNA-transfected HEK293T cells expressed the proteins of interest. Notably, after glycosylation (41), the size of the resultant protein was larger than the expected molecular weight (~77 kD) ( Figure 1B ; Supplementary Figure 1 ). In addition, the antigen was secreted outside the cells, which may aid in the induction of humoral responses through antigen uptake and presentation by APCs at local injection site.

To evaluate the safety of the RBD3-Fc mRNA vaccine inoculated using GV-01, side effects resulting from GV-01 injection was evaluated. As expected, wheals with size ranging from 6-12 mm were observed in all groups after inoculation using GV-01 ( Figures 3A, B ). Generally, 8-12 mm wheals indicate successful ID injection. At 1.5 h after immunization, the wheals disappeared in all groups. As for other local side effects, GV-01 resulted in erythema and scab formation, which were more serious than those induced by IM inoculation in the antigen groups ( Supplementary Figures 2A, B ). Notably, the negative control group did not exhibit a similar reaction after ID inoculation, implying that the local reaction was caused by mRNA-LNPs, but not buffer. However, these adverse reactions is transient and disappeared 6 days after inoculation in the groups treated using GV-01. Besides, for either ID or IM inoculation, body weight and food consumption did not significantly differ from those of the negative control, except 50 μg mRNA-LNP via IM, in which transient weight loss occurred after each inoculation, with a significant difference observed (p < 0.05) ( Supplementary Figures 2C–F ). No remarkable systemic side effects were found in the groups vaccinated using GV-01. Overall, ID delivery of the mRNA vaccine using GV-01 was associated with acceptable safety profiles.

To assess the immunogenicity of the RBD3-Fc mRNA vaccines, rats were vaccinated as shown in Figure 3C and Table 1 . Prototype RBD-specific binding antibodies, IgG isotypes and neutralization antibodies (nAbs) were determined. As shown in Figures 4A, B , antigen-specific antibody levels after boosting in all groups were obviously higher than those after priming, suggesting the superior performance of the prime-boost strategy over single injection. In addition, as expected, at the same antigen dose, the groups inoculated with the mRNA-LNP vaccines via two routes displayed significantly higher antibody levels than those inoculated with mRNA vaccines. Higher antibody levels may result from encapsulation with LNP, which protects mRNA from ribozyme degradation during the delivery process in vivo. In addition, LNP-mediated endosome escape of mRNA may be more important in improved antigen expression and subsequent antigen presentation. Interestingly, in the IM groups inoculated with either mRNA-LNP or mRNA vaccines, antigen-specific antibody levels increased in an antigen dose-dependent manner, regardless of the prime or prime-boost injections ( Figure 4A ). However, no similar phenomenon was observed in ID groups inoculated with mRNA-LNP vaccines using GV-01, which suggested that ID delivery of 30 µg mRNA-LNP per dose may lead to saturation to some extents. This reasonable speculation was supported by the remarkable antigen dose-dependent increase in the antibody levels in groups inoculated with 5, 10, and 30 µg mRNA-LNP using GV-01 ( Supplementary Figure 3A ). The lower range of antigen doses, corresponding to an mRNA dose-dependent increase in antibody levels of the ID groups, indicated that GV-01-mediated ID delivery of the mRNA-LNP vaccine was superior to needle-mediated IM inoculation. In addition, for the mRNA vaccines, ID delivery generally induced slightly higher RBD-specific antibody levels at the same antigen dose, despite the lack of a significant difference. After 4 weeks of boost, relatively high binding antibodies ( Supplementary Figures 3B, C ) similar with the results of 2 weeks were observed, which suggested antibodies evoked by RBD3-Fc vaccines might be long-lasting.

To identify bias in the immune response induced by mRNA vaccines, we used rat serum to perform IgG isotype assays via ELISA. As shown in Figures 4C, D , all mRNA-LNP groups had high levels of antigen-specific IgG1, IgG2a, and IgG2b (titer >10⁵), except for IgG2c (titer >10²), than mRNA groups after the boost regardless of inoculation routes. IgG2b and IgG2c are reported to be Th1-related isotypes, while IgG1 and IgG2a are Th2-related isotypes in rats (15, 42, 43). Thus, the above results suggested that mRNA-LNP vaccines elicited both Th1 and Th2 immune responses. In terms of mRNA vaccines, all groups had higher levels of IgG1 than IgG2a, but low IgG2b and IgG2c titers, regardless of the inoculation route and injection technique, indicating an obvious Th2-biased immune response. These results suggested that delivery carrier, but not inoculation route or injection technique, had a more remarkable influence on immune response bias.

Neutralizing antibodies are crucial for preventing COVID-19 infection. Pseudovirus neutralization assays were performed to evaluate the neutralizing antibody levels induced by mRNA vaccines in rats inoculated via the IM or ID route. As shown in Figures 4E, F , regardless of the administration route and antigen dose, all groups inoculated with the mRNA-LNP vaccines produced nAbs against the homologous viruses D614G, B.1.617.2, and BA.4/5. In addition, for all groups, the neutralizing antibody levels of BA.4 and D614G were slightly higher than those of B.1.617.2. Interestingly, ID delivery of 30 µg mRNA-LNP led to higher neutralizing antibody levels against all three strains than that induced by IM delivery. However, no statistically significant difference was found between the two routes. When mRNA dose increased to 50 µg, IM delivery of mRNA-LNP showed improved nAbs titers against all three strains, while no obvious increase was found in group inoculated with mRNA-LNP via ID route. As a result, both routes showed comparable capacity in inducing the production of nAbs at the dose of 50 µg mRNA. Besides, as expected, naked mRNA vaccine was less effective at inducing the production of neutralizing antibodies against the above strains than mRNA-LNPs, regardless of the administration route ( Supplementary Figure 3D ). Similar phenomena were observed 4 weeks after the booster ( Supplementary Figure 3E ). Overall, ID inoculation had comparable or better performance in eliciting neutralizing antibody responses against homologous viruses after prime-boost relative to IM inoculation.

To further assess the T cell immune responses elicited by RBD3-Fc mRNA vaccines delivered via the ID and IM routes, ELISpot assays were performed to detect the secretion of IFN-γ and interleukin-4 (IL-4) in splenocytes ( Supplementary Figures 4A, B ). The delivery of 30 μg of the mRNA-LNP vaccine was more effective at inducing IFN-γ than 30 μg of the mRNA vaccine in two routes, which suggested antigen was able to evoke T cell immune responses against infections. In addition, 30 μg of the mRNA-LNP vaccine via the IM route led to a higher IFN-γ level than delivery via the ID route. In contrast, the secretion levels of IL-4 were very low in all groups, and only one of three rats administered 30 μg of mRNA-LNP via ID route had detectable IL-4 level. These results suggested that the delivery of the mRNA-LNP vaccine via either the IM or ID route elicited T cell immune responses against virus. The above conclusion was also supported by the results of antigen-specific IgG isotype assays.

To further illustrate the effect of inoculation routes on T cell immune responses induced by the mRNA-LNP vaccine, single-cell RNA sequencing assay (RNA-Seq) was performed using rat splenocytes harvested at week 6. CD45+ splenocytes were separated and captured as immune cells using flow cytometry. Louvain clustering was applied to the filtered and integrated objects and plotted in the uniform manifold approximation and projection (UMAP) space; 14 cell types were identified and manually annotated ( Supplementary Figure 4C ). After merging the two samples, differences between B cells (highly expressing the Cr2 gene) and T cells (CD4 T cells, CD8 T cells) were observed ( Supplementary Figure 4D ). The distribution patterns of these major immune cell subsets in rats were not significantly altered by the vaccination routes; however, both routes led to relatively potent immune responses against COVID-19. To analyze the transcription factors that contribute to immune responses, a single-cell regulatory network inference and clustering (SCENIC) analysis was performed ( Figure 5A ). Tcf7 regulates cell responses to IL-4 and the differentiation of γδT cells, which was in accordance with IL-4 ELISpot results. Upregulation of Tbx21 induces naïve CD4 T cells to acquire Th1 cell-like features (44) to further give the evidence of evoking T cell immunity in line with IFN-γ secretion by ELISpot assays. FoxP3 is a specific marker of regulatory T cells (Tregs) that maintained immune homeostasis in rats (45). Lef1, a 48-kD nuclear protein, is a crucial transcription factor for the proliferation and survival of B and T cells (46), which proved that systematic immune responses after inoculation by NFIS were efficiently aroused against SRSA-Cov-2. Furthermore, the number of CD8 T cells in the group inoculated via the ID route was obviously greater than that in the group inoculated via the IM route ( Figure 5B ); these cells are responsible for adaptive responses and IFN-γ secretion ( Figure 5C ). Moreover, CD8+ T cells displayed robust interactions with DC via the mannose receptor 1 (MRC1) and macrophages via MRC1 and vascular cell adhesion molecule-1 (VCAM-1) ( Figure 5D ), which may suggest vaccination by NFIS increased antigen presentation and improve the ability of CTL and macrophages to clear virus by directly killing infected cells. CD4 T cells had similar biological functions with CD8 T cells to have activity against infections ( Figure 5C ; Supplementary Figure 4E ). As shown in Supplementary Figures 4F, G , antigen-presenting cells like macrophages and DCs first presented processed peptides to CD4 T cells and made them help B cells product antibodies. After GO and KEGG pathway analyses of two groups, different genes were found to be enriched in T cell responses against COVID-19 for T cell differentiation and cytokine secretion, suggesting that systematic immune responses were elicited by both routes, especially using GV-01 ( Supplementary Figures 4H–J ).

To further demonstrate differential T cell immune responses induced by naked mRNA vaccines delivered via different routes, 14 immune cells were annotated and visualized using UMAP software and there were no obvious differences between two groups ( Supplementary Figures 5A, B ). Of note, the number of CD4 T cells induced by 30 μg mRNA delivered using GV-01 was relatively greater than that induced by needles, which was in line with the higher titer of serum antibodies. According to GO analysis ( Supplementary Figure 5C, D ), CD4 and CD8 T cells were responsible for adaptive immune responses. Besides, CD4 T cell also functioned in differentiation and TCR signaling pathway to response to antigens. Overall, the naked mRNA vaccine delivered by two routes was less effective than the mRNA-LNP vaccine in inducing robust systematic immune responses.