Oligonucleotides (primers) were designed using Serial Cloner and Benchling. Densitometry was performed with Image Studio Lite Version 5.2 (Licor).

E. coli strains (BL21(DE3) derivatives, DH5α, TOP10, HK100) were routinely cultured in lysogeny broth (LB Miller; Sigma-Aldrich) supplemented with ampicillin (100 mg/L), chloramphenicol (25 mg/L) or a combination of both, when necessary. For recombinant protein expression in OMVs, cultures were supplemented with 0.1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) at OD₆₀₀ between 0.5 and 0.7 and OMVs were collected after 3 hours from induction. Fermentation was carried out using an ez-Control bioreactor (Applikon) at 30°C until OD₆₀₀ 0.5, then growth continued at 25°C, pH 6.8 ( ± 0.2), dO₂ > 30%, 280-500 rpm. At OD₆₀₀ 1.0, the culture was induced with 0.1 mM IPTG together with a feed consisting in ampicillin (50 mg/L) and/or chloramphenicol (12.5 mg/L), glycerol (15 g/L), and MgSO₄ (0.25 g/L).

Plasmid derivatives of the pET21b(+) backbone encoding FhuD2, SpA_KKAA, LukE, Hla_H35L, ClfA_Y338A were already described (9, 17), while derivatives of the pACYC backbone were constructed ex novo using PIPE-PCR (18). Briefly, the coding sequence of the target protein was PCR amplified from its respective template pET21b(+) plasmid (I-PIPE) using the primers reported in Supplementary Table 1 . These primers feature overhangs that are complementary to the 5’- and 3’-ends of a linearized “recipient” pACYC derivative. The latter plasmid was linearized by PCR (V-PIPE) using the primer couples listed in Supplementary Table 1 . The PCR products of both reactions were digested with DpnI (New England BioLabs), combined (2 μl each) and used to transform chemically competent E. coli HK100 cells. Positive clones identified by colony PCR using GoTaq^® Green Master Mix (Promega) and gene specific primers ( Supplementary Table 1 ) were subjected to sequence analysis (Eurofins) before using the resultant plasmids ( Supplementary Table 2 ) for transformation of the OMV producer strain.

The PIPE-PCR method (18) was also used to construct plasmids encoding chimera. Here, the coding sequence of the C-terminal protein of the chimera was PCR amplified from its respective template plasmid (I-PIPE) using the primers reported in Supplementary Table 1 . These primers feature overhangs that, at the 5’-end, encode a short linker sequence (translated as GGGGS) complementary to the 3’-end of a “recipient” pET21b(+) derivative carrying the Lpp leader sequence followed by the N-terminal antigen of the chimera, while at the 3’-end the overhangs carry the stop codon and are complementary to the 5’-end of the recipient plasmid. The recipient plasmids were linearized by PCR (V-PIPE) excluding the stop codon and introducing the same linker sequence stated above using the primer couples listed in Supplementary Table 1 . The PCR products were mixed together as described above and the resultant plasmids ( Supplementary Table 2 ) were used for transforming the OMV producer strain.

FhuD2, LukE, SpA_KKAA and Hla_H35L S. aureus proteins were purified as previously described (9).

For ClfA_Y338A, the gene coding for the wild-type N1N2N3 portion (amino acid 40 to 559) of the antigen was amplified from the genomic DNA of S. aureus Newman strain. After genomic amplification, the Y338A mutation was inserted using PIPE-PCR. In particular, the gene was first amplified using primer couples 32 + 37 and 36 + 33 ( Supplementary Table 1 ) and then the PCR product was used as template for the final PCR with the primer 32 + 33. The resulting gene (clfA_Y338A ) was then cloned in a pET15 plasmid (containing His₆-tag and TEV cleavage site coding sequences) downstream from a T7 inducible promoter and expressed in E. coli BL21(DE3) strain. Bacterial biomass was produced using an ez-Control bioreactor (Applikon Biotechnology) and purification of the recombinant protein was performed using an ÄKTA purifier System (General Electric Healthcare) for IMAC and size-exclusion chromatography and 10 mM Tris + 100 mM NaCl and PBS (both pH 8.0), respectively, as buffer.

OMVs were purified as previously described (9, 17, 19). Purification from cultures in bioreactor was carried out using Tangential Flow Filtration (ÄKTA flux system, GE Healthcare) with a Hollow Fiber cartridge UFB-500-C-3MA (GE Healthcare). Protein content of OMVs was quantified using DC™ Protein Assay (Bio-Rad) and the quality of OMVs was monitored by SDS-PAGE loading OMVs (normalized by μg) on Any kD™ Criterion™ TGX Stain-Free™ Protein Gel (BioRad) stained with ProBlue Safe Stain (Giotto Biotech). The size of purified OMVs was determined using NanoSight NS300 (Malvern Panalytical Ltd) as previously described (17).

OMVs (100 μg of proteins) were diluted in PBS, ice cold Triton X-114 was added to 1% final concentration and the OMV-containing solution was incubated at 4°C for 1 hour under shaking. The solution was then heated at 37°C for 10 minutes and the aqueous phase was separated from the detergent phase by centrifugation at 13,000 g for 10 minutes. Proteins in both phases were then precipitated by standard chloroform/methanol procedure, separated by SDS-PAGE electrophoresis and the protein of interest visualized by Western blot. Western blot analysis was performed as previously reported (9). Antibodies against each antigen were obtained from Genscript by immunizing rabbits with the following antigen-specific synthetic peptides: LukE: NEFVTPDGKKSAHD, SpA_KKAA: AKKLNDAQAPKADN, Hla_H35L: GTNTKDKWIDRSSE, ClfA_Y338A: IDKPVVPEQPDEPG. All synthetic peptides were conjugated to KLH protein.

Animal experiments were carried out in accordance with experimental protocols that were reviewed and approved by the Animal Ethical Committees of the University of Trento and Toscana Life Sciences (Siena, Italy) and by the Italian Ministry of Health.

For antibody titers, five-week old CD1 female mice were immunized intraperitoneally (i.p., 100 μl final volume) and intramuscularly (i.m., 50 μl final volume) three times every two weeks with 20 μg of CLSH-OMVs_Δ60 formulated either with or without Alum hydroxide (2 mg/ml). Blood was collected from anesthetized mice through cardiac puncture at day 35 and serum was obtained from blood through centrifugation at 2,000 rpm for 10 minutes.

For challenge studies, mice were immunized i.p. at day 0, 14 and 28 with either 20 μg of “empty” OMVs_Δ60 or 20 μg of CLSH-OMVs_Δ60 in the presence of 2 mg/ml Alum. Alum was included in the vaccine formulation to use conditions similar to Bexsero, a vaccine which contains OMVs and has been approved for human use (2). For the sepsis model of infection two weeks after the third immunization mice were i.p. challenged with 3×10⁸ CFUs of S. aureus Newman strain. Mice were monitored daily for a 7-day period. Animals health was evaluated using a 1 to 4 pain scale. A value of 4 was given to mice with: loss of weight >15%, very rough hair coat, impaired mobility. A score of 3 was given to mice with loss of weight approximately 15% and rough hair coat, while scores of 2 and 1 were given to mice with a loss of weight between 6% and 14% or 0% and 5%, respectively. All procedures were approved by the National Health Institution and the Ethical Committee and for human reasons animals were sacrificed at symptoms of sickness as recommended by 3Rs rules (‘‘Refinement, Reduction, Replacement’’ policy towards the use of animals for scientific procedures_ 99/167/EC, Council Decision of 25/1/99).

Experiments using the renal abscess and skin infection model were performed as previously described (13). Briefly, for the renal model mice were challenged intravenously (i.v.) 10 days after the third immunization with 1×10⁷ CFUs of S. aureus Newman strain, while in the skin infection model mice were challenged subcutaneously (s.c.) 14 days after the third immunization with 5×10⁷ CFUs of S. aureus Newman strain.

ELISA assays on mice sera collected after immunizations were performed as previously described (9) using alkaline phosphatase-conjugated secondary antibodies that recognize murine total IgG (Sigma-Aldrich), IgG2a or IgG1 (both Biolegend).

To test the possibility of designing a multivalent vaccine constituted by a single preparation of E. coli–derived OMVs engineered with four heterologous proteins, we focused our attention on SpA_KKAA, ClfA_Y338A, Hla_H35L and LukE antigens.

SpA substantially contributes to S. aureus virulence and toxicity. By binding to the Fcγ domain of antibodies, surface-exposed SpA interferes with the association of the C1q component of the complement classical pathway, and this results in the inhibition of bacterial opsonophagocytosis and killing (OPK) by phagocytic cells (23). Moreover, by cross-linking V_H3 clonal B cell receptors, secreted SpA acts as a B cell super-antigen promoting the production of all V_H3 antibodies, irrespectively of their antigen specificity (24). Finally, SpA is responsible for an anaphylactic syndrome due to its binding to the V_H3 region of IgG and IgE antibodies associated to basophils and mast cells (25). It was elegantly shown that immunization with the SpA_KKAA, a mutant no longer capable of binding Fcγ and V_H3, induces antibodies which (i) inhibit SpA binding to Ig, (ii) promote OPK in mouse, guinea pig and human blood (23, 26), and (iii) protect animals from bacteremia (27).

Clumping Factor A (ClfA) is a surface-exposed virulence factor expressed in most S. aureus isolates and whose primary function is to allow the adhesion of S. aureus to fibrinogen (28–32). Since fibrinogen is ubiquitous in the host, ClfA promotes a rapid attachment of bacteria to tissues and organs, thus playing a relevant role in the initiation of infection (33). Inhibiting the fibrinogen binding capacity of ClfA has been proposed as a way to prevent pathogenesis. Indeed, immunization of monkeys and human volunteers with the N1N2N3 portion of ClfA carrying the Y338A mutation (30) elicited high titers of antibodies which prevented S. aureus adhesion to fibrinogen (32).

S. aureus α-hemolysin (Hla) (also known as α-toxin) is a pore-forming β-barrel toxin (34, 35), which assembles in a heptameric structure forming a pore of 2-nm in diameter into the plasma membrane (35). The toxin is expressed by almost all clinical isolates (36) and its level of expression correlates with increased severity of Skin and Soft Tissue Infections (SSTIs) and pneumonia in humans (37, 38). Hla causes dermonecrotic skin injury by interacting with ADAM10, a zinc-dependent metalloprotease that cleaves E-cadherin and destabilizes the epithelial barrier upon toxin binding (39, 40). The key role of Hla in S. aureus pathogenicity is highlighted by experiments showing that the intraperitoneal injection of 10 μg of purified Hla is sufficient to kill mice (41) and the passive transfer of neutralizing monoclonal antibodies is highly protective against S. aureus infection in different animal models (42, 43). Finally, vaccination with the nontoxigenic Hla_H35L mutant (44) protects mice against S. aureus-mediated pneumonia (37), lethal sepsis, organ dissemination (45) and skin infection (13).

Human S. aureus isolates produce five additional pore-forming toxins known as leukocidins (46). Leukocidins are constituted by two subunits, the host cell targeting S component and the polymerization F component. The major target of leukotoxins are immune cells of myeloid and lymphoid lineages, suggesting that these toxins have evolved to inhibit both innate and adaptive immunity. Among the five leukocidins, LukED appears to be particularly important in that it kills virtually all immune cells, including neutrophils, monocytes, macrophages, dendritic cells, T cells, erythrocytes and NK cells (46). The toxin is encoded in the stable S. aureus pathogenicity island vSaβ53, which is present in about 70% of all clinical isolates (47). Because of their key role in pathogenesis, inactivated leukocidins are considered to be key targets for both therapeutic and prophylactic intervention (46, 48).

Based on the above, a vaccine including SpA_KKAA, Hla_H35L, ClfA_Y338A and LukE should elicit immune responses that, in the presence of innate immune responses promoted by the OMVs, should synergize and confer protection against S. aureus infections. Therefore, we selected these four antigens to attempt the construction of a multivalent vaccine based on a single OMV preparation.

Supported by preliminary data indicating that even high molecular weight heterologous antigens can be efficiently incorporated in OMVs, we created two chimeras carrying the lipoprotein leader sequence of Lpp at their 5’-ends (9) by fusing ClfA_Y338A-LukE genes and SpA_KKAA-Hla_H35L genes. The gene fusions were cloned in both pET21b(+) and pACYC, two compatible plasmids, which differ in copy number. The plasmids were used to transform the hyper-vesiculating strain E. coli BL21(DE3)Δ60 recently created in our laboratories (17). OMVs were purified from the culture supernatants and analyzed by SDS-PAGE. As shown in Figure 1A , both chimeras accumulated in the vesicular compartment at a level of 7% of total OMV proteins, as determined by densitometry scanning. Interestingly, the different copy numbers of pET21b(+) and pACYC did not substantially influence the level of accumulation in the vesicular compartment, suggesting that the rate limiting step of chimera compartmentalization in OMVs was largely determined by the translation and/or transport machineries.

pET21b(+) and pACYC plasmids have compatible origins of replication. This prompted us to test whether the co-transformation of E. coli BL21(DE3)Δ60 with two recombinant plasmids, one having the pACYC scaffold and the other the pET21b(+) scaffold, could lead to the co-expression of the two chimeras in the OMVs. E. coli BL21(DE3)Δ60 was co-transformed with the plasmid couple pET(ClfA_Y338A-LukE)/pACYC(SpA_KKAA-Hla_H35L). The recombinant strain was grown in liquid culture using a bioreactor, and after 4-hour induction of protein expression with 0.1 mM IPTG, the OMVs, named CLSH-OMVs_Δ60, were purified from the culture supernatant ( Figure 1B ). Particles sizes and protein content of CLSH-OMVs_Δ60 were analyzed by NanoSight NS300 and SDS-PAGE, respectively. CLSH-OMVs_Δ60 had a mean size of 111.6 +/- 2.1 nm, in line with the size of “empty” OMVs_Δ60 previously reported (17) ( Figure 1B ). As shown in Figure 1C , two protein bands corresponding to the molecular mass of the two chimeras were visible among total OMV proteins. Based on densitometry analysis, the two co-expressed chimeras accumulated in similar amounts, corresponding to 4.4% and 5.8% of total proteins from SpA_KKAA-Hla_H35L-OMVs_Δ60 and ClfA_Y338A-LukE-OMVs_Δ60, respectively.

The fusion of the two chimeras to the lipoprotein leader sequence of Lpp should promote their N-terminal lipidation and subsequent compartmentalization in the OMV membrane (9). To indirectly demonstrate the presence of the acyl groups at the N-terminus of the two chimeras, CLSH-OMVs_Δ60 were solubilized at 4°C with a 1% water solution of Triton X-114 and subsequently the sample was warmed to 37°C to partition Triton X-114 into two phases: a detergent-rich ‘hydrophobic’ phase and a detergent-poor ‘hydrophilic’ phase. Membrane proteins, including lipoproteins, typically partition into the hydrophobic phase (49). Aliquots from both aqueous and organic phases were analyzed by Western Blot using antibodies specific for the corresponding S. aureus antigens. As shown in Figure 1C , the two chimeras compartmentalized in the hydrophobic phase.

Next, we tested whether the immunization with CLSH-OMVs_Δ60 could elicit antigen-specific antibody responses. Mice were immunized intraperitoneally (i.p.) three times, two weeks apart, with 20 μg of CLSH-OMVs_Δ60 + Alum and we determined the antibody titers against each antigen by ELISA using plates coated with the respective purified recombinant proteins. As shown in Figure 2A , immunization elicited antibodies specific for each of the four S. aureus antigens present in the CLSH-OMVs_Δ60, while in sera from mice immunized with “empty” OMVs no antigen-specific antibodies were detected ( Supplementary Figure 1 ).

We also tested the role of Alum and the route of immunization in determining the level of antibody titers. To this aim, we repeated the i.p. immunization in the absence of Alum, and we immunized mice intramuscularly (i.m.) with 20 μg of CLSH-OMVs_Δ60 in the presence or absence of Alum. Sera from immunized mice were analyzed by ELISA ( Figure 2A ). In the presence of Alum, the route of immunization did not substantially influence the antibody titers, with the exception for SpA_KKAA antibody titers, which were approximately 3.5-fold lower when the vaccine was administered intramuscularly. As far as the influence of Alum is concerned, only a marginal difference was observed when mice were immunized i.p., while titers against Hla_H35L and LukE were enhanced by Alum in i.m.-immunized mice. Finally, IgG titers against SpA_KKAA appeared to be higher in the absence of Alum when the i.m. route was used.

Next we evaluated the IgG isotypes as an indication of the type of immune response (Th1 vs. Th2-skewed response) induced by CLSH-OMVs_Δ60 immunization. The sera from animals immunized i.p. in the presence of Alum were analyzed by ELISA using anti-mouse anti-IgG2a and anti-IgG1 antibodies as secondary antibodies and coating plates with ClfA_Y338A and SpA_KKAA, the two N-terminal antigens of the chimeras. As shown in Supplementary Figure 2 , and in line with what previously observed (8), IgG2a were abundantly present in the sera of mice immunized with CLSH-OMVs_Δ60, indicating that the immunization elicited a Th1-skewed response.

Finally, we tested whether the antigen-specific antibodies induced by immunization with CLSH-OMVs_Δ60 had functional activities. First, we analyzed the capacity of differentiated HL-60 cells to kill the S. aureus Newman strain in the presence of different dilutions of mouse sera and guinea pig complement. As shown in Figure 2B , the sera from mice immunized with CLSH-OMVs_Δ60 rescued the HL-60-mediated bacterial killing in a dose-dependent manner.

Next, we tested whether the same sera could inhibit the biological activity of ClfA, LukED and Hla. To follow ClfA inhibition, plates were coated with fibrinogen and the binding of S. aureus Newman strain was analyzed in the presence or absence of sera from CLSH-OMVs_Δ60-immunized mice. In the case of LukED, a toxin which can kill activated HL-60 cells, anti-CLSH-OMVs_Δ60 sera and HL-60 cells were incubated together with purified LukED. Finally, the inhibition of the hemolytic activity was tested by incubating the sera with purified Hla and rabbit erythrocytes. As shown in Figures 2C–E , the pool of mouse sera from the CLSH-OMVs_Δ60-immunized group inhibited the activities of all three virulence factors. In the case of LukED inhibition, “empty” OMVs_Δ60 gave some neutralization, but well below the inhibition of sera from CLSH-OMV_Δ60-immunized mice. This effect could be due to the presence of some cross-reacting antibodies elicited by OMVs_Δ60.

Finally, we followed the protective activity of CLSH-OMVs_Δ60 immunization in mice challenged with S. aureus Newman strain. In particular, three challenge models were used: the sepsis model, the skin model and the kidney abscess model. For the sepsis model, mice were immunized three times with 20 μg of CLSH-OMVs_Δ60 and 14 days after the last immunization the animals received 3×10⁸ colony forming units (CFUs) of bacteria (i.p.). Animal health was followed every day over a period of seven days using a 1 to 4 pain scale (see Material and Methods). Animals that reached the pain value of 4 were sacrificed since they reached the “near mortality point” and would have died within 24 hours. On the other hand, those animals that maintained a score lower than 4, fully recovered within 7 days. Figure 3A reports the data of mice immunized with Alum alone, “empty” OMVs_Δ60 and CLSH-OMVs_Δ60. As shown in the figure, immunization with “empty” OMVs_Δ60 conferred a certain level of protection, with 50% of the animals that did not reach pain value 4 and survived. By contrast, 100% protection was observed in the group of mice immunized with CLSH-OMVs_Δ60 and none of the animals experienced a pain value higher than 2. The protective activity of the CLSH-OMVs_Δ60 was also tested in the skin infection model and in the kidney abscess model. As shown in Figures 3B, C , CLSH-OMVs_Δ60 vaccination was highly protective in both models. In particular, differently from Alum-immunized and “empty” OMVs_Δ60-immunized animals, mice immunized with CLSH-OMVs_Δ60 and subsequently s.c. challenged with 5×10⁷ CFUs developed only a mild and transient skin abscess with no dermonecrotic lesions. The differences between vaccinated and control animals was statistically significant (p < 0.001) as evaluated by means of repeated ANOVA measures using the R base function aov(). As far as the kidney abscess model is concerned, according to which 10⁷ CFUs were given intravenously and bacteria were counted in the kidneys four days after challenge, the CFUs counts in 13 out of 16 immunized mice were below the detection limit (10² CFUs). As previously shown (9), immunization with “empty” OMVs_Δ60 also resulted in a substantial level of protection (62.5%), in line with the protective role of an innate-type immune response induced by the OMVs.

The results described above demonstrate the feasibility of co-expressing the two chimeras SpA_KKAA-Hla_H35L and ClfA_Y338A-LukE in the same OMVs preparation. For a broad applicability of the technology it would be important to prove that OMVs can be successfully decorated with other chimera combinations. Therefore, we created three additional chimeras by combining LukE and Hla_H35L with FhuD2, another virulence factor involved in S. aureus pathogenesis (50), and by fusing together the two toxins LukE and Hla_H35L. Three plasmids were generated pET(FhuD2-LukE), pACYC(FhuD2-Hla_H35L) and pET(LukE-Hla_H35L). The plasmids were used to transform E. coli BL21(DE3)Δ60 strain and one transformant colony of each transformation was used to produce OMVs. As shown in Figure 4A , the three chimeras efficiently accumulated in the vesicular compartment. Next, we co-transformed the strain with the plasmid couples pET(FhuD2-LukE)/pACYC(SpA_KKAA-Hla_H35L) and pET(ClfA_Y338A-LukE)/pACYC(FhuD2-Hla_H35L) and the successful compartmentalization of the two chimera couples in FLSH-OMVs_Δ60 and CLFH-OMVs_Δ60 was confirmed by SDS-PAGE ( Figure 4A ). Finally, we used CLFH-OMVs_Δ60 to immunize mice and we analyzed the antibody titers against all antigens. As shown in Figure 4B , the engineered OMVs elicited antibodies specific for all four chimera antigens.

On the basis of these experiments we concluded that our OMV engineering strategy has a broad applicability and can be exploited for the preparation of multi-component vaccines against one or more pathogens.