Bacterial strains and plasmids used in this study are presented in Supplementary Table 1. The reference S. aureus strains RN4220 (NCTC 8325-4) were kindly provided by Prof. Baolin Sun (University of Science and Technology of China), USA300 (ATCC BAA-1556) was a gift from Dr. Min Li (Shanghai Jiao Tong University, China), Newman (NCTC 8178) was provided by Prof. Lu Yu (Jilin University, China), MW2 was a gift from Prof. Xiaoxue Ma (China Medical University), and strains ATCC29213, ATCC4330, and ATCC25923 and two clinical isolates (TJ274 and TJ275) were kindly provided by Dr. Dongmei Wang (The 983 Hospital of PLA). S. aureus XQ was isolated from the blood sample of a 16-year-old juvenile, whose onset was a skin wound then deteriorated to a lethal S. aureus sepsis with mental disorder (Rao et al., 2015; Liu et al., 2018). S. aureus strains were grown in brain heart infusion (BHI) medium (Oxoid, UK) at 37°C with shaking (200 rpm) or cultivated on BHI agar. Escherichia coli DH5α was cultivated in Luria–Bertani (LB) medium (Oxoid, UK) at 37°C with shaking or cultured on LB agar. For plasmid pBT2 (Supplementary Table 1) maintenance, cultures were supplemented with 50 μg/ml of ampicillin (AMP) and 10 μg/ml of chloramphenicol (Cm) for E. coli and S. aureus, respectively. All strains were stored in an appropriate medium with 20% glycerol at −80°C and resuscitated by streaking on agar plates. A single colony was picked and cultured with BHI or LB medium for the subsequent experiments.

MV isolation was performed with some modifications as previously described (Hong et al., 2011; Yuan et al., 2018; Chen et al., 2020; Zaborowska et al., 2021). In brief, a single colony of S. aureus RN4220 from the BHI agar plate was inoculated into 1 ml BHI medium and cultured at 37°C for 6 h. Then, 100 μl of culture was transferred to 10 ml BHI broth (1:100) and cultured at 37°C overnight with shaking. On the next day, the culture was inoculated into 900 ml BHI medium (1:100) and cultured at 37°C for 22 h to achieve the highest MV production (Yuan et al., 2018). The culture supernatant was collected after centrifugation at 6,000 × g at 4°C for 20 min to remove bacterial cells and centrifugation at 10,000 × g at 4°C for 30 min to remove cell debris and then filtered through a 0.45 μm vacuum filter (Millipore, USA), to remove possible cell debris or aggregates. The filtrate was concentrated to 240 ml with a 100 kDa cutoff hollow fiber cartridge column (GE Healthcare, USA) and then filtered through a 0.22 μm filter (Millipore, USA). The supernatant was ultracentrifuged at 200,000 × g at 4°C for 3 h.

The MV pellets were washed twice with phosphate-buffered saline (PBS, pH7.2) and resuspended in sterile PBS, and the amount of protein in the MVs was measured with Bradford Protein Assay Kit (Beyotime, China) as previously described (MacDonald and Beveridge, 2002; Tartaglia et al., 2018; Bitto et al., 2021b). Prepared MVs were stored at −80°C for use.

The protein composition in MVs of S. aureus RN4220 (^wtMVs) was identified by using 4D-label-free liquid chromatography–tandem mass spectrometry (LC-MS/MS) as previously described (Cox et al., 2014). In brief, the proteins in three biological replicate ^wtMV samples were digested with trypsin (Promega, USA), desalted on a C18 cartridge column (Empore™, Sigma, USA), and reconstituted in 40 μl of 0.1% (v/v) formic acid. The peptides were subjected to 4D-label-free LC-MS/MS analysis by Applied Protein Technology (Shanghai, China). The MS raw data for each sample were searched against the database of S. aureus (http://www.uniprot.org/) by using MaxQuant 1.5.3.17 software (Parianichnikov et al., 2020). Cluster 3.0 and Java TreeView software were used to perform hierarchical clustering analysis (Hoon et al., 2004; Saldanha, 2004). CELLO (http://cello.life.nctu.edu.tw/) was used to predict protein subcellular localization. Protein sequences were searched using InterProScan software to identify protein domain signatures from the InterPro member database Pfam (Quevillon et al., 2005). The gene ontology (GO) was analyzed using the protein sequences of the selected differentially expressed proteins. NCBI BLAST+ client software and InterProScan were used to find homolog sequences; the GO terms were mapped, and sequences were annotated using software Blast2GO (Conesa et al., 2005; Quevillon et al., 2005; Camacho et al., 2009). The results of GO annotation were plotted by R scripts. Kyoto Encyclopedia of Genes and Genomes (KEGG) was analyzed using the online database (http://geneontology.org/). KEGG orthology identifications were retrieved and subsequently mapped to pathways in KEGG. Enrichment analysis was applied based on Fisher's exact test, considering the whole quantified proteins as the background dataset. Benjamini–Hochberg correction for multiple testing was further applied to adjust derived P-values. Only functional categories and pathways with P-values under a threshold of 0.05 were considered significant. All raw data of 4D label-free LC-MS/MS identification was deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the iProX partner repository with the dataset identifier PXD039734.

The purified MVs were mixed with 5 × SDS-PAGE loading buffer at a rate of 4:1 and incubated in boiling water for 10 min. The proteins in samples (10 μl/each) were separated by 12% (m/v) sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and stained with Coomassie brilliant blue (Thermo Scientific, USA). For Western blot, the proteins after SDS-PAGE were transferred to a polyvinylidene difluoride (PVDF) membrane (GE Healthcare, USA). The PVDF membrane was blocked with PBS-T (PBS + 0.1% (v/v) Tween-20) containing 5% (m/v) skim milk (ABgent, USA) at 37°C for 60 min, followed by incubation with anti-Hld rabbit polyclonal primary antibody (1:1,000 dilution in PBS-T containing 5% (m/v) skim milk) at 4°C overnight. After five times of wash with PBS-T, the membrane was incubated with horseradish peroxidase-conjugated goat anti-rabbit secondary antibody (Abmart, China; 1:5,000 dilution) at 37°C for 1 h. After five times of wash with PBS-T, the target protein was visualized with the enhanced chemiluminescence system (Thermo Scientific) and photographed.

The site mutation S. aureus RN4220-Δhld and agrA deletion RN4220-ΔagrA mutants were generated via allelic replacement as previously described (Peng et al., 2017). All primers used in this study are presented in Supplementary Table 2. For the construction of RN4220-Δhld with normal RNAIII function, the 20th base “C” in the hld gene was changed to “G” to achieve a stop codon TGA and result in a truncated translation of Hld. In brief, the homologous left arm and homologous right arm were amplified with primer pairs delta-L/pBT2-delta-R (BamHI) and pBT2-delta-L (SalI)/delta-R, respectively, from the genomic DNA template of S. aureus RN4220 (GenBank accession no. NZ_CP101124) by polymerase chain reaction (PCR). The PCR fragments were purified and ligated to pBT2 plasmid via ClonExpress MultiS One Step Cloning Kit (Vazyme, China). The recombinant plasmid pBT2-Δhld was transformed into competent E. coli DH5α, extracted, and sequenced. The correct plasmid was then electrotransformed into S. aureus RN4220. The plasmid-carrying S. aureus was cultured at 30°C for 16 h, and the plasmid pBT2-Δhld was verified by double digestion with BamHI and SalI (TakaRa, Japan). The correct strain carrying pBT2-Δhld was inoculated into fresh BHI medium and cultivated at 42°C to induce the integration of the recombinant plasmid into the genome, which was verified by colony PCR with primer pairs delta-outer-L/pBT2-down and pBT2-up/delta-outer-R. The plasmid was eliminated via culturing a single colony at 25°C to achieve the mutant strain RN4220-Δhld, which was verified by PCR amplification with primer pair delta-outer-L/delta-outer-R and DNA sequencing. The construction of RN4220-ΔagrA used similar strategies mentioned above.

For genetic complementation, the tetracycline promoter tetP was introduced in front of the hld gene to obtain tetP/hld, which was inserted behind the eno gene on the genome of RN4220-Δhld. The homologous left arm, right arm, and hld gene were amplified by PCR with primer pairs up-pBT2-eno-tetP-L/up-tetP-R, down-eno-L/down-pBT2-eno-R, and tetP-hld-L/tetP-hld-R, respectively, from genomic DNA template of S. aureus RN4220. The tetP promotor fragment was amplified with primer pair tetP-L/tetP-R from the plasmid pYT3 (Supplementary Table 1). The four PCR fragments were purified and ligated to the pBT2 plasmid via ClonExpress MultiS One Step Cloning Kit. The resulting pBT2-tetP/hld was transformed into S. aureus RN4220-Δhld, and the genetic complementation strain RN4220-Δhld-tetP/hld was screened and identified using similar strategies for RN4220-Δhld construction. The hld overexpression strain RN4220-tetP/hld was generated after the transformation of pBT2-tetP/hld into the wild-type S. aureus RN4220.

The growth curves of S. aureus RN4220 and its derivatives RN4220-Δhld and RN4220-ΔagrA were determined. In brief, S. aureus strains of interest were cultured in BHI broth at 37°C with shaking overnight, and 0.2 ml of each strain culture was inoculated into 20 ml of fresh BHI medium. The optical density (OD) at 600 nm (OD600) was detected every hour for 15 h after inoculation. Growth curves were drawn using the OD600 values over culture times.

The size distribution of S. aureus MVs was evaluated by dynamic light scattering (DLS) analysis using Zetasizer Nano ZS90 (Malvern, UK). In brief, purified MVs were diluted in PBS to 0.1 mg/ml. Reads of 60-s durations were conducted in triplicate at 25°C. Data outputs were generated using Zetasizer (Nano, μV, APS) software v7.13 (Malvern, UK). The average number of particles at each binned center in the Experiment Summary output was adjusted by the dilution factor using GraphPad Prism 8.0.1 software, and the mean of three biological replicates was plotted as particles.

A hemolysis assay was performed as previously described (Miller et al., 2021). Approximately 10 ml of whole blood was collected from the central auricular artery of a New Zealand male rabbit, which was purchased from the Laboratory Animal Center of the Third Military Medical University, and added to 1 ml heparin sodium solution, followed by three times of wash with PBS with centrifugation at 1,000 × g for 5 min at 4°C. A concentration of 2% (v/v) rabbit red blood cells (RBCs) was prepared with PBS. In total, 40 μl of PBS containing 2, 5, 10, and 20 μl, respectively, of 0.5 mg/ml MVs derived from S. aureus RN4220 (^wtMVs) or its derivatives RN4220-Δhld (^ΔhldMVs) and RN4220-ΔagrA (^ΔagrAMVs) was added to 1 ml of 2% RBCs to the final concentration of 1, 2.5, 5, and 10 μg/ml of MVs, respectively, mixed gently, and incubated at 37°C for 30 min. After centrifuging at 8,000 × g for 10 s, the supernatant was collected, and OD540 values were determined with a spectrophotometer (Thermo Scientific). PBS-treated RBCs served as negative control, while RBC treatment with 1% (v/v) Triton X-100 represented positive control. The percentage hemolysis is calculated as follows: [(OD540 of sample−OD540 of negative control)/(D540 of positive control–OD540 of negative control)] × 100%.

Macrophage RAW264.7 cells were purchased from a Chinese company Cas9X, maintained in Roswell Park Memorial Institute (RPMI) 1640 medium (Gibco, USA), and supplemented with 10% (v/v) fetal bovine serum (Zeta Life, USA) and 1% (v/v) penicillin/streptomycin (Beyotime, China) at 37°C in a humidified 5% CO₂ atmosphere. RAW264.7 cells were seeded at 2.5 × 10⁵ cells/ml in 500 μl in 24-well plates for 12 h and stimulated with 5, 10, and 20 μg/ml of ^wtMVs, ^ΔhldMVs, or ^ΔagrAMVs for 6 h, and the cell culture supernatants were collected for further study.

The cell viability was detected by a CCK-8 cell viability kit (ABclonal, China), according to the instructions provided by the manufacturer. In brief, RAW264.7 cells were seeded at 2.5 × 10⁵ cells/ml in 100 μl RPMI1640 medium in 96-well plates for 12 h and stimulated with 10 μl of 0.05, 0.1, and 0.2 mg/ml of ^wtMVs, ^ΔhldMVs, or ^ΔagrAMVs (the final concentration of 5, 10, and 20 μg/mL of MVs, respectively) for 6 h at 37°C with 5% CO₂. In total, 10 μl of CCK-8 solution was added to each well and incubated for 2 h, and OD450 values were determined with a spectrophotometer (Thermo Scientific). The 10 μl of PBS-treated RAW264.7 served as a negative control, and the wells with culture medium and CCK-8 solution served as blank controls. The percentage of cell-survival rate is calculated as follows: [(OD450 of sample−OD450 of blank control)/(OD450 of negative control−OD450 of blank control)] × 100%.

The LDH cytotoxicity assay kit (Beyotime, China) was used to quantify the activity of LDH released from damaged cells. RAW264.7 cells were seeded at 1 × 10⁵ cells/ml in 200 μl of RPMI1640 medium in 96-well plates for 12 h, the culture solution was discarded, and the cells were washed once with sterile PBS. In total, 200 μl of fresh low serum culture medium (1% serum) and 20 μl of PBS containing 0.05, 0.1, and 0.2 mg/ml of ^wtMVs, ^ΔhldMVs, or ^ΔagrAMVs were added to each well to the final concentrations of 5, 10, and 20 μg/ml of MVs, respectively. Cell-free culture medium supplemented well served as blank control, untreated cell-seeded well served as sample control, 20 μl PBS-treated RAW264.7 served as negative control, and cells lysed with 20 μl LDH releasing reagent used as maximum enzyme activity control. After 6 h of incubation, the plate was centrifuged at 400 × g for 5 min, and 120 μl supernatant of each well was transferred to a new 96-well plate, followed by adding 60 μl of detection working solution, mixed well, and incubated at room temperature to avoid light for 30 min. OD490 and OD600 values were determined. The measured OD490 and OD600 values were each subtracted from the values of blank control. LDH activity is calculated as follows: [((OD490 − OD600 of sample) − (OD490 − OD600 of sample control))/((OD490 − OD600 of maximum enzyme activity control) − (OD490 − OD600 of sample control))] × 100%.

All animal study was conducted in accordance with the ARRIVE guidelines. In total, 6- to 8-week-old female BALB/c mice (20 ± 2 g) were purchased from Hunan Slyke Jingda Experimental Animal Co. LTD. All animal experiments were approved by the Laboratory Animal Welfare and Ethics Committee of the Army Medical University (protocol No. AMUWEC2020735).

BALB/c mice (n = 3) were intraperitoneally injected 100 μl (10 or 20 μg/ml) of ^wtMVs, ^ΔhldMVs, or ^ΔagrAMVs. PBS was used as the control. Mouse sera were collected 6 h post-injection. Then, the levels of inflammatory factors, including IL-6, IL-1β, and TNFα, were determined with an enzyme-linked immunosorbent assay (ELISA) kit, according to the manufacturer's instructions (ABclonal, China). The results were examined with an ELISA reader Multiskan Go (Thermo Scientific), and inflammatory factor levels were determined according to the standard curve.

All statistical analyses were performed using GraphPad Prism 8.0.1 software. One-way analysis of covariance (ANOVA) was used for comparing multiple groups. The results were presented as mean ± standard deviation (SD), and P-values of <0.05 were considered to be statistically significant.

An accurate proteome-wide 4D-label-free LC-MS/MS (Cox et al., 2014) was performed to quantify the proteins in MVs of S. aureus RN4220 (^wtMVs). A total of 704 proteins in the ^wtMVs were identified, and they were predicted to be membrane proteins (178, 25.3%), extracellular (147, 20.9%), and/or cytoplasmic proteins (458, 65.1% with some overlapped ones). KEGG enrichment revealed that oxidative phosphorylation was the most significant pathway involved, followed by citrate cycle, glycolysis/gluconeogenesis, pyruvate metabolism, carbon-fixation pathways in prokaryotes, and nucleotide metabolism (Supplementary Figure 1). GO analysis showed that the enriched proteins were involved in metabolic and cellular processes, catalytic activity and binding of molecular functions, and cell and cell part of cellular components (Figure 1A).

Notably, Hld was the most abundant molecule in ^wtMVs of S. aureus RN4220, which presented a label-free quantity (LFQ) intensity of 367,430,000 and accounted for 15.6 ± 7.9% of the total protein quantity identified (Figure 1B). SDS-PAGE analysis also revealed the loaded Hld in ^wtMVs, which could be characterized by specific antibodies against Hld in Western blot assay (Figure 1C; Supplementary Figure 2). Compared with the protein in bacterial cell lysate and culture supernatant, Hld was remarkably loaded in the ^wtMVs of S. aureus RN4220 and elevated in an MV-dose-dependent manner (Figure 1D; Supplementary Figure 2). To determine whether loaded Hld occurred in MVs derived from other strains, we cultured nine S. aureus strains (USA300, XQ, ATCC25923, Newman, MW2, ATCC29213, ATCC4330, TJ274, and TJ275) and individually prepared their MVs with the same conditions to RN4220. SDS-PAGE and Western blot assays revealed the loaded Hld in all MVs tested (Figure 1E; Supplementary Figure 2), suggesting that the encapsulation of Hld in S. aureus MVs is common.

The gene encoding Hld is within the RNAIII molecule (85–165 nt) (Figure 2A), a global RNA regulator involved in the S. aureus Agr quorum-sensing system (Hodille et al., 2016). To evaluate the function of loaded Hld in MVs, we generated an S. aureus mutant (RN4220-Δhld) by changing the 20th base of C to G in the hld gene to achieve a stop codon TGA and abolish the translation of Hld protein (Supplementary Figures 3A, B). An agrA gene knocked-out strain (RN4220-ΔagrA) was also constructed to inactivate RNAIII and Hld functions (Supplementary Figure 3C). Compared with wild-type RN4220, agrA deletion and site mutation in the hld gene did not affect the growth of RN4220-Δhld and RN4220-ΔagrA mutants (Supplementary Figure 3D). No Hld was detected in ^ΔhldMVs and ^ΔagrAMVs derived from RN4220-Δhld and RN4220-ΔagrA, respectively (Figure 1C). Western blot showed that the expressions of virulence factors such as alpha-hemolysin (Hla), gamma-hemolysin B subunit (HlgB), and Panton–Valentine leukocidin (PVL) element LukS in the RN4220-Δhld were similar to those in the wild-type strain (Supplementary Figures 3E, F). However, agrA deletion resulted in decreased expression of all virulence factors tested. These data indicated that the site-mutation strain RN4220-Δhld disabled Hld expression but had a functional RNAIII, which controlled the expression of many virulence factors, including Hla, HlgB, and PVL-LukS. By contrast, Hld expression and RNAIII function were absent in the RN4220-ΔagrA.

The MV production in RN4220-Δhld and RN4220-ΔagrA was substantially reduced compared with that in wild-type RN4220 (Figure 2B). For complementation, a tetracycline promoter TetP-controlled hld gene was constructed in RN4220-Δhld by inserting it behind the eno gene, which encoded the MV-associated enolase (Yuan et al., 2018). The restored MV production was observed in RN4220-Δhld-tetP/hld (Figure 2B). Moreover, Hld overexpression in wild-type strain RN4220 (RN4220-tetP/hld) significantly increased MV yield (P < 0.001; Figure 2B). Next, the sizes of ^wtMVs, ^ΔhldMVs, and ^ΔagrAMVs were measured, and the results showed that the sizes of ^wtMVs (45.4 ± 6.9 nm), ^ΔhldMVs (56.0 ± 9.3 nm), and ^ΔagrAMVs (94.5 ± 23.2 nm) enlarged gradually (Figure 2C). The ^ΔagrAMVs presented the largest particle size, which was significantly larger than those derived from RN4220-Δhld and wild-type RN4220 (P < 0.001; Figure 2D). These data suggested that Hld promoted MV production in S. aureus. Loaded Hld was associated with smaller MV particles, and other Agr-regulated factors may synergistically condense the sizes of S. aureus MVs.

RN4220-Δhld and RN4220-ΔagrA showed lower abilities to produce MVs compared with wild-type RN4220. 4D-label-free proteomics was further performed to detect differentially loaded proteins in the ^ΔhldMVs and ^ΔagrAMVs. The results showed that 684 proteins were shared in ^wtMVs, ^ΔhldMVs, and ^ΔagrAMVs (Figure 3A). Compared with proteins included in ^wtMVs (704), Hld mutation resulted in increased proteins encapsulated in ^ΔhldMVs (1,381), and RNAIII inactivation even enhanced the proteins loaded by ^ΔagrAMVs (1,698). The differentially loaded proteins in ^wtMVs, ^ΔhldMVs, and ^ΔagrAMVs were primarily predicted cytoplasmic proteins (260), extracellular components (77), and membrane-associated proteins (128, Figure 3B). Compared with ^wtMVs, ^ΔhldMVs had 156 upregulated (fold change ≥ 2) and 12 downregulated (fold change ≤ 0.5) proteins, whereas ^ΔagrAMVs exhibited 277 upregulated and 59 downregulated proteins. However, ^ΔhldMVs presented 136 upregulated and 390 downregulated proteins relative to ^ΔagrAMVs (Figure 3C). These data indicated that loaded Hld was associated with reduced protein diversity in S. aureus MVs.

The top 10 MV proteins dominant in ^ΔhldMVs and ^ΔagrAMVs were analyzed. Out of the top 10 proteins, 8 were shared by ^ΔhldMVs and ^wtMVs (Figures 1B, 3D). Among the top 10 MV proteins, Hld and immunoglobulin-binding protein Sbi were unique in ^wtMVs, whereas extracellular adherence protein and 30 kDa neutral phosphatase fragment were specific for ^ΔhldMVs. By contrast, ^ΔagrAMVs had diverse abundant proteins (Figures 1B, 3E). KEGG analysis revealed that the upregulated proteins in ^ΔhldMVs vs. ^wtMVs were involved in ABC transporters, bacterial secretion, photosynthesis, and protein export, whereas the downregulated proteins were enriched in purine metabolism, quorum sensing, biosynthesis of cofactors, and nucleotide metabolism (Supplementary Figure 4A). GO analysis indicated that the most abundant differentially loaded proteins in ^ΔhldMVs vs. ^wtMVs were involved in cellular metabolic process and localization for biological process; catalytic activity, binding, and transporter activity for molecular function; and membrane, membrane part, cell, and cell part for cellular component (Supplementary Figure 4B). These results demonstrated that the loading of Hld in ^wtMVs may inhibit the encapsulation of other proteins and affect the diversity of MV-loaded proteins.

Hld plays crucial roles in staphylococcal pathogenicity, such as causing hemolysis and allergy (Nakamura et al., 2013). The hemolytic activities of ^wtMVs, ^ΔhldMVs, and ^ΔagrAMVs were detected. Rabbit RBCs (2%) were mixed with 1, 2.5, 5, and 10 μg/ml of MVs, respectively, and incubated at 37°C for 30 min. The values of OD540 were detected, and the results showed that ^wtMVs presented higher hemolytic activity than ^ΔhldMVs and ^ΔagrAMVs (Figure 4A). The hemolytic activity of ^ΔhldMVs prepared from Hld mutant RN4220-Δhld significantly increased in an MV-dose-dependent manner (P < 0.001, Figure 4B). However, different concentrations of ^ΔagrAMVs derived from the agrA deletion strain RN4220-ΔagrA presented comparable hemolytic activity to PBS control. The content of hemolytic proteins in MVs was analyzed by proteomics, and the results showed that ^wtMVs loaded a large amount of hemolytic proteins with the most Hld, which accounted for 15.9% of the total proteins in ^wtMVs. The proportion of hemolytic proteins in ^ΔhldMVs was reduced, and Hla occupied the most in ^ΔhldMVs (6.8%). By contrast, the content of hemolytic proteins in ^ΔagrAMVs was the lowest (Supplementary Figure 5). These results indicated that Hld was an important component that contributed to the hemolytic activity of ^wtMVs, and other factors such as Hla and leukocidin-like proteins may also play important roles in the hemolytic activity of S. aureus MVs.

Macrophage RAW264.7 cells were treated with 5, 10, and 20 μg/ml of ^wtMVs, ^ΔhldMVs, or ^ΔagrAMVs for 6 h, and the cell death was evaluated. Microscopy revealed that RAW264.7 treated with 20 μg/ml of ^wtMVs presented greater cell disruption than those treated with ^ΔhldMVs (Figure 5A), whereas the morphology of RAW264.7 treated with 20 μg/ml ^ΔagrAMVs was similar to that of the PBS control group. CCK-8 assay and LDH release detection demonstrated that ^wtMV was highly toxic, even at the low concentration of 5 μg/ml (Figures 5B, C). However, the toxicity of ^ΔhldMVs was significantly reduced, and only a high concentration of ^ΔhldMVs (20 μg/ml) presented toxicity to RAW264.7 cells compared with the PBS control. No significant difference in cell survival and LDH release was observed between ^ΔagrAMVs and PBS treatment. The proteomic data showed that ^wtMVs loaded a large amount of virulence proteins (33.7%), whereas 19.36% of the total proteins in ^ΔhldMVs were counted to be virulence factors, and only 5.1% of virulence proteins were loaded in ^ΔagrAMVs (Supplementary Figure 5). These data suggested that Hld-loaded ^wtMVs were more toxic to RAW264.7 macrophages.

The levels of inflammatory factor production represent the activation of the immune system caused by bacterial virulence factors in vivo. BALB/c mice were intraperitoneally injected with different concentrations of ^wtMVs, ^ΔhldMVs, and ^ΔagrAMVs (n = 3 per group). The sera were collected at 6 h post-injection and subjected to quantification of IL-6, IL-1β, and TNFα levels through ELISA. As shown in Figure 6A, the IL-6 levels in mice treated with 20 μg/ml of ^wtMVs significantly increased compared with those in ^ΔhldMV- and ^ΔagrAMV-challenged mice (P < 0.001). However, comparable serum IL-1β and TNFα levels were presented between MV- and PBS-treated mice (Figures 6B, C). These results indicated that loaded Hld was a potent virulence factor for S. aureus MV pathogenicity.