Bacteria strains and plasmids used in this study are listed in Supplementary Table 1 . E. coli MG1655 wild type (WT) strain was utilized as a parental strain for mentioned isogenic gene deletion and chromosomal modification. E. coli strain BL21(DE3) was used for recombinant protein expression. Standard Luria-Bertani (LB) broth with appropriate antibiotics (100 μg/ml ampicillin, 50 μg/ml kanamycin or 25 μg/ml chloramphenicol) was used for E. coli culture.

To create the lpxM deletion strain, PCR-amplified DNA fragments containing 40 bp homologous arms flanking lpxM open reading frame together with a selection antibiotic marker were purified. PCR products were then electroporated into MG1655 harboring λ red recombinase expression plasmid pKD46. Positive clones were selected by 25 μg/ml chloramphenicol and verified by DNA sequencing. Subsequently, the antibiotic selection marker flanked by loxP sites was removed using the helper plasmid p705Cre. The ΔlpxM ompA-spycatcher strain was further constructed using the similar recombineering method. An ompA(1-465 bp)-G4S linker-spycatcher-6his fusion gene was synthesized and used to replace the chromosomal ompA gene. Positive clones were selected using 50 μg/ml kanamycin and confirmed by DNA sequencing.

A gene fragment encoding SpyTag (AHIVMVDAYKPTK) was first inserted in the pET28a plasmid between NheI and BamHI restriction sites to produce the pET28a-spytag plasmid. Gene fragments of GFP and S. aureus antigen rEsxA, rSbi, and a mutated form of rSpA_KKAA (rSpA) from the Newman strain were synthesized and inserted between BamHI and SalI restriction sites to create a series of SpyTag-fused GFP or antigen expression plasmids. Positive clones were screened and verified by colony PCR and DNA sequencing. Primers used in this study are listed in Supplementary Table 2 . Correctly constructed plasmids were then prepared and transformed into BL21 (DE3) strain for protein expression and purification.

SpyTag fusion proteins were expressed in E. coli BL21 (DE3) strains and purified by nickel column. Bacteria were grown in LB broth at 37°C with constant shaking (220 r.p.m) until OD₆₀₀ of 0.6 before adding 1 mM IPTG to induce fusion protein expression. After induction for an additional 3-6 h at 37°C, cells were harvested by centrifugation. Bacterial pellets were then resuspended in 2 ml of lysis buffer (0.5 M NaCl, 50 mM pH 7.2 Tris-HCl, 40% glycerol, 20 mM imidazole, 150 mM PMSF, and 500 mM DTT) followed by sonication for cell lysis. Bacterial lysates were obtained by centrifugation at 15,000 g for 30 min and the resultant supernatant was subjected to a 1.5-ml Ni-nitrilotriacetic acid-agarose column (Qiagen) equilibrated with binding buffer (0.5 M NaCl, 50 mM pH 7.2 Tris-HCl, 40% glycerol, and 40 mM imidazole). After rounds of washing with binding buffer, proteins were eluted with elution buffer (0.5 M NaCl, 50 mM pH 7.2 Tris-HCl, 40% glycerol, and 500 mM imidazole).

Engineered bacteria were grown in 250 ml LB broth at 37°C with constant shaking (220 r.p.m) until OD₆₀₀ of 1.0. The bacteria culture was centrifuged at 4,500 × g for 20 min at 4°C. The resultant supernatant was filtered by 0.45-µm filters (Corning) and followed by ultracentrifugation at 100,000 × g for 1 h at 4°C. OMV-containing pellets were then resuspended in DPBS and filtered by 0.22-µm filters (Corning) for further use. The total protein concentration of purified OMVs was determined by BCA assay. The surface display of SpyCatcher on OMVs was verified by western blot using a His-Tag Monoclonal antibody (Proteintech). The OMV size was determined by Nano-flow cytometry (NanoFCM) according to manufacture instructions.

For TEM analysis, diluted OMVs were fixed with 4% PFA and absorbed to copper grids. Negative staining was performed with 1% aqueous uranyl acetate at room temperature and washed to remove excess aqueous uranyl acetate. Electron micrographs were collected from randomly-selected fields using Philips CM100 transmission electron microscope.

Equal amount (50 μg) of OMV-SpyCatcher and SpyTag-GFP proteins were incubated at room temperature for 30 min. Unbound SpyTag-GFP in the reaction were removed by ultrafiltration using a 0.5mL 100-KDa ultrafiltration unit (Millipore). For the first round of ultrafiltration, the reaction was supplemented to a volume of 300 μl in total by adding DPBS and then centrifuged at 14,000 g until a volume of 50 μl. The resultant reaction was resuspended to 300 μl for the second round of ultrafiltration to a volume of 50 μl. GFP fluorescence was captured by UVP ChemStudio (Analytikjena). To perform the reaction assay in living cells, ΔlpxM and ΔlpxM ompA-spycatcher bacterial cells were collected and washed in PBS. Bacteria were then adjusted to OD₆₀₀ of 0.3 and incubated with or without 5 μg Spytag-GFP at room temperature for 10 min. Samples were washed three times with ice-cold PBS and subjected to flow cytometry analysis using FITC channel (CytoFLEX, Beckman).

Six-week-old female BALB/c mice were immunized with OMV triple-linked antigens, a cocktail of three OMV single-linked antigens and antigens formulated with 10% aluminum hydroxide gel (AHG) by subcutaneous (s.c) injection. For OMV triple-linked antigens group, 50 μg OMV-SpyCatcher were incubated with three antigens (10 μg each) for 30 min. For a cocktail of three OMV single-linked antigens group, each 16.7 μg OMV-SpyCatcher were incubated with 10 μg of a single antigen for 30 min, then the three OMV single-linked antigens were pooled together to make a cocktail before injection. To prepare AHG+Antigens group, 30 μg antigens (10 μg for each) were mixed with 10% AHG. 50 μg empty OMVs purified from ΔlpxM strain and PBS were also injected as control groups. Mice were immunized three times at 2-week interval. Seven days after the last immunization, blood samples or spleens were collected for ELISA or ELISpot assay respectively. Animal immunization experiments were performed in accordance with institutional guidelines following experimental protocol review and approval by the Institutional Animal Care and Use Committee.

The IgG detection were performed by enzyme-linked immunosorbent assay (ELISA). SpyTag fused EsxA, tSbi and SpA proteins (1 µg/ml in 0.05 M carbonate-bicarbonate buffer, pH 9.6, 200 µl/well) were coated in ELISA plates (Nunc, Denmark) and incubated overnight at 4°C. The plates were blocked with phosphate-buffered saline (PBS) containing 5% nonfat milk at room temperature for 3 h and washed four times with PBS containing 0.05% Tween 20. Three-fold serially diluted mice sera were added into the wells and incubated at room temperature for 1 h. Then the plates were washed six times with PBS containing 0.05% Tween 20 and incubated with horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG at room temperature for 1 h. The color was developed using tetramethylbenzidine (TMB) solution (Sigma) and absorbance was measured using a plate reader (BioTek Microplate Reader) at 450 nm. The antibody level was determined as the area under the curve (AUC) calculated by GraphPad Prism 9 (GraphPad Software, USA).

IFN-γ-producing splenocytes from vaccinated or naive unvaccinated mice were analyzed using a Mouse IFN-γ ELISpot ^PLUS (HRP) kit (Mabtech). Spleens were minced in RPMI 1640 Medium (Thermo Fisher Scientific) and cells were filtered through a 70-μm sterile cell strainer (Corning). Red blood cells were lysed using red blood cell lysis buffer. Remaining splenocytes were washed and adjusted to a concentration of 1 × 10⁶ cells ml^-1 in complete RPMI 1640 medium. 100 μl cells/well were added to conditioned ELISpot plates and each recombinant antigen, stSpA, stSbi or stEsxA, was used as the inducer with a final concentration of 2 μg/ml. Plates were incubated at 37°C in a humidified incubator with 5% CO₂ for 24 hours. Following 2 h primary antibody incubation and 1 h secondary antibody incubation at room temperature, spots were developed using TMB substrate solution and recorded in an immunospot analyzer.

Immunized animals were challenged on day 42 by intravenous injection with 5 × 10⁷ CFU of S. aureus Newman strain. The well-being of infected mice was monitored daily for 6 days and lethal disease was recorded. The surviving mice were euthanized after day 6 post challenge as a predetermined humane endpoint. Lethal challenge experiments were approved by the Committee on the Use of Live Animals in Teaching and Research of the University of Hong Kong (approval no. CULATR 5163-19).

One way ANOVA was used to analyze the statistical significance of ELISpot assay results and GFP binding. Log rank (Mantel-Cox) analysis was used to analyze the statistical significance of the data from the lethal-challenge experiments. Analysis was performed using GraphPad Prism 9 (GraphPad Software, USA), and a P value of <0.05 was considered statistically significant.

We modified the composition of bacterial OMVs by deleting the lpxM gene which encodes lipid A acyltransferase in E. coli MG1655. The deletion of lpxM leads to under-acylation of LPS and largely attenuates the endotoxic activity of LPS which may cause adverse effects for in vivo applications of engineered OMVs (39–41). To further generate OMVs with surface-exposed SpyCatcher, we did genomic insertion in the ΔlpxM strain to replace the chromosomal ompA open reading frame with a truncated ompA (1- 465 bp) and spycatcher fusion gene ( Figure 1A ). OmpA is one of the most abundant outer membrane proteins in E. coli and subsequently in OMVs (42). It was reported that the 144-160 amino acids of OmpA were extracellular exposed (43) and fusion with a truncated version of OmpA (a.a 1-159) allowed the display of a fused protein on E. coli cell surfaces (44, 45). Therefore, the SpyCatcher is expected to be exposed on the surface of engineered OMVs when fused with the 155 a.a truncated OmpA (trOmpA) ( Figure 1A ).

We first detected the expression of the trOmpA-SpyCatcher fusion protein by western blot in engineered OMVs (OMVsc) isolated from the ΔlpxM ompA-spycatcher strain, but not in control OMVs (OMVnc) from the ΔlpxM strain ( Figure 1B ). We confirmed the OMV morphology by TEM as electron-dense particles with characteristic OMV structures ( Figure 1C ). We further determined the size of purified OMVs by Nano-flow cytometry. Figure 1D showed the size of OMVsc displayed a similar pattern as that of OMVnc with a typical diameter distribution ranging from 40 to 200 nm, which is in accordance with previous findings (46–48).

To examine whether the OMV-presented SpyCatcher based on trOmpA fusion works well in vitro, we applied SpyTag-fused GFP (stGFP) to test the ability of ‘click’ display of heterogenous antigens by our OMVsc ( Figure 2A ). OMVsc and OMVnc were separately incubated with stGFP and the ‘click’ linking of OMV and GFP was analyzed by western blot. The result indicated that only OMVsc group but not OMVnc generated trOmpA-GFP fusion proteins (~70 kDa), proving specific conjugation between OMVsc and stGFP mediated by SpyCatcher-SpyTag system ( Figure 2B ).

We further tested the specific binding of stGFP on OMVsc by fluorescence. When incubating living bacterial cells with stGFP, fluorescence signal could only be detected by flow cytometry for the strain expressing trOmpA-SpyCatcher after washing ( Supplementary Figure 1 ). We then incubated purified OMVsc with stGFP and subsequently removed the unbound stGFP by ultrafiltration. After a second round of ultrafiltration, we detected significantly higher GFP fluorescence in OMVsc+stGFP filtration group compared with either stGFP or OMVsc group, suggesting the trap of stGFP by OMVst during ultrafiltration ( Figure 2C ). Together, our results demonstrate that engineered OMVsc could utilize trOmpA-SpyCatcher to ‘click’ display SpyTag fused GFP, which makes a versatile antigen display vaccine platform.

We selected three S. aureus antigens which are conserved in a large panel of S. aureus strains to assemble with OMVsc. The three antigens, EsxA, truncated Sbi and a mutated form of truncated SpA (SpA_KKAA) (49), target different virulent or immune evasion pathways of S. aureus. We fused SpyTag with the entire EsxA (a.a 2-97), one Ig binding domain and one complement binding domain of Sbi (a.a 96-195) and two Ig binding domains of SpA (a.a 34-153) to separately generate stEsxA, stSbi and stSpA antigens ( Figure 3A ). To test whether the recombinant antigens could be displayed on OMVsc, we incubated 10 μg of OMVsc with various amounts of stEsxA, stSbi, or stSpA antigens and then detected the OMV-antigen binding using western blot. As expected, all three SpyTag fused antigens formed conjugation with OMVsc in a dose-dependent manner ( Figure 3B ).

To evaluate the immunogenicity of our click vaccine in vivo, we immunized mice subcutaneously with 3 doses of different vaccines ( Figure 4A ). In particular, we prepared two forms of click vaccines, OMV-Antigens (OMVsc triple-linked antigens) and OMV-Antigen cocktail (a cocktail of three OMVsc single-linked antigen). The click vaccines were compared with AHG+antigens (Antigens mixed with commonly used adjuvant AHG), OMVs (OMVsc alone) and PBS as a negative control.

Th1-biased cellular immunity has been shown to play essential roles in the protection against S. aureus. The release of IFN-γ is indicative of Th1-biased immune responses. Therefore, we performed IFN-γ ELISpot assays for splenocytes collected from immunized mice 7 days after the third immunization. The results showed that splenocytes from both click vaccine groups had significantly more IFN-γ secreting cells than that of AHG+antigens and other control groups ( Figures 4B, C ). Quantification of IFN-γ⁺ spots indicated that OMV-Antigens induced stronger anti-Sbi and anti-EsxA cellular immune response than OMV-Antigen cocktail ( Figure 4C ).

To investigate the stimulation of humoral immunity by our click vaccine, we also collected the serum samples from each group to detect the levels of antigen-specific immunoglobulin G (IgG) by ELISA in terms of area under the curve (AUC). The results showed that immunization with OMV-Antigen cocktail induced high levels of IgG for all antigens, while immunization with OMV-Antigens only induced a high level of IgG for SpA. The anti-Sbi and anti-EsxA IgG induced by OMV-Antigens were comparable with that induced by AHG+antigens ( Figure 4D ). Taken together, the above results suggested that the OMV-based click vaccine could successfully evoke potent S. aureus antigen-specific humoral and cellular immunity in a mouse model.

We then investigated whether immune responses elicited by the click vaccine could translate into protective effects against S. aureus lethal challenge in mice. Mice receiving different vaccinations were challenged with a lethal dose of S. aureus Newman strain ( Figure 5A ). All mice in the PBS treated group died within 72 hours post Newman challenge. However, survival rates were significantly improved by vaccination with OMV-Antigens (40% survival; P=0.0027) or OMV-Antigen cocktail (40% survival; P=0.0027). Notably, our click vaccines also conferred better protection compared with AHG+antigens vaccination (20% survival; P=0.0150) ( Figure 5B ).