Salmonella enterica serovar Typhi strain STH2370 (S. Typhi WT) (Valenzuela et al., 2014) was obtained from the Infectious Diseases Hospital Lucio Córdova, Chile. S. enterica serovar Typhimurium ATCC14028s (S. Typhimurium WT) was obtained from the Instituto de Salud Pública, Chile. Strains were routinely grown in liquid culture using Luria Bertani medium (Bacto peptone, 10 g/L; Bacto yeast extract, 5 g/L; NaCl, 5 g/L; prepared in phosphate buffer pH 7.0) at 37°C, with aeration. When required, medium was supplemented with kanamycin (50 mg/ml), chloramphenicol (20 mg/ml), ampicillin (50 mg/ml), Xgal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside) (40 μg/ml), or agar (15 g/L).

Mutagenesis was performed using the EZ-Tn5 < R6Kγori/KAN-2 > transposome (Epicenter^®) as previously described (see detailed description in Supplementary Figure 1) (Goryshin et al., 2000). We screened 15,000 mutants to cover all the genome, and hemolytic colonies were selected for further analyses. As the next step, a “clean mutagenesis” (backcross) of the corresponding gene interrupted by the transposon (i.e., the zzz gene) was performed in a S. Typhi WT background using Red-Swap recombination (Datsenko and Wanner, 2000) yielding S. Typhi Δzzz::FRT mutants. For the experiments presented in this work, at least two different S. Typhi Δzzz clones, independently generated, were tested in each case. All mutants were corroborated by PCR amplification, as described (Datsenko and Wanner, 2000). S. Typhi Δzzz::FRT ΔhlyE::kan double mutants were constructed using the Red/Swap method (Datsenko and Wanner, 2000) iteratively.

S. Typhi hlyE-3 × FLAG was constructed as previously described (Fuentes et al., 2009). S. Typhi Δzzz::FRT hlyE-3 × FLAG was constructed by moving the hlyE-3 × FLAG (kan^R) allele over the corresponding S. Typhi Δzzz::FRT mutant as described (Toro et al., 1998). A full list with the primers used in this study is presented in the Supplementary Table 1.

To complement the S. Typhi Δzzz mutants, we amplified the corresponding zzz genes (using primers listed in Supplementary Table 1) prior to cloning them into the pCR TOPO TA^® plasmid according to the manufacturer’s instructions.

Quantitative RT-PCR was performed as described (Jofre et al., 2014). Briefly, total mRNA from the strains grown in LB broth (approximately OD₆₀₀ = 1.4), was extracted using TRIzol reagent (Invitrogen) as described by the manufacturer. RNA was precipitated with isopropanol for 10 min at room temperature, washed with ice-cold 70% v/v ethanol and resuspended in DEPC-treated water, prior to the treatment with DNase I to remove any trace of DNA. Purity of extracted RNA was determined by spectrometry. Reverse transcription and relative quantification of each mRNA was performed as previously described (Jofre et al., 2014). Experiments were performed in three biological and technical triplicates. Statistical significance of differences in the relative expression data was determined using Kruskal–Wallis with Dunn’s test as post hoc analysis to compare with the WT. A full list with the primers used in this study is presented in the Supplementary Table 1.

Strains carrying the epitope-tagged hlyE gene were grown in LB at 37°C with shaking to stationary phase (approximately OD₆₀₀ = 1.1). For the whole lysate, bacterial pellets were resuspended in 1 mL of Tris–HCl pH 8.0 and sonicated on ice during 100 s. For the supernatant fraction, proteins were previously precipitated using 1/4 volume of trichloroacetic acid, incubated at 4°C for 10 min, and centrifuged at 15,000 ×g for 10 min. The pellet was washed twice with cold acetone and finally resuspended in Tris–HCl 100 mM pH 8.0. Bradford method (Bradford, 1976) was used to calculate the volume of the whole lysate containing 60 μg of proteins for each strain to be tested (load control). An equivalent volume was used for the corresponding supernatant fraction. Proteins were resolved by 12% SDS–PAGE, transferred to poly(vinylidene difluoride) membranes and stained with Ponceau S to confirm the protein load as previously described (Stochaj et al., 2006). Western blot was performed as described, using Hsp60 as load control (Jofre et al., 2014; Ortega et al., 2016).

The same procedure described for the SDS–PAGE was used to qualitatively assess the exported proteins of S. Typhi WT and S. Typhi Δzzz mutants.

In the case of OMV proteins, OMVs obtained from either S. Typhi WT or S. Typhi Δzzz mutants (corresponding to 12 μg of proteins) were resolved by SDS–PAGE and visualized with silver stain.

First, strains expressing β-lactamase (bla gene) from a single chromosomal copy were constructed by the chromosomal insertion of pNFB9. pNFB9 is a plasmid that contains the bla gene, the attachment site for the phage Gifsy-1 (att_PG1) and the respective integrase int (Lemire et al., 2008), allowing its integration into the att_BG1 chromosomal site of S. Typhi (naturally empty) (Valenzuela et al., 2014), yielding the strain S. Typhi att_BG1::pNFB9 (Lac^- Amp^R). The same procedure was performed for all the mutants (i.e., S. Typhi Δzzz) to generate the corresponding Amp^R derivatives. Spots of S. Typhi att_BG1::pNFB9 (WT background), and S. Typhi Δzzz att_BG1::pNFB9 derivatives (Δzzz backgrounds) were plated over a lawn of S. Typhimurium ATCC14028s ΔompD::MudJ (10⁶ cfu, Lac⁺ Amp^S, reporter strain) previously spread on a LB agar plate supplemented with ampicillin and Xgal. For semi-quantitative assay to assess export of β-lactamase, supernatants from the strains described above was obtained (OD₆₀₀ = 1.1), filtered (0.45 μm), precipitated with trichloroacetic acid (as described above), serially diluted and used for the bioassay. In all cases, export of functional β-lactamase was inferred by the growth of blue satellite colonies.

Bacteria were grown in LB with shaking at 37°C to OD₆₀₀ = 0.5 or 1.1, prior to being treated during 2 h with sodium deoxycholate 0.5% or PBS (control) at 37°C. Percentage of survival was calculated as following: (cfu treated with sodium deoxycholate/cfu with PBS) × 100, relative to S. Typhi WT. To determine sensitivity to vancomycin, a Kirby-Bauer assay was performed as reported (Villagra et al., 2012), using vancomycin disks loaded with 30 μg of antibiotic.

Preparation of bacterial samples was performed as described (Azari et al., 2013) with modifications. Briefly, bacterial suspensions (OD₆₀₀ = 1.1) were centrifuged and washed three times with PBS. Aliquots were applied onto the surface of circular coverslips and allowed to dry. Bacterial cell surfaces were imaged in contact mode or optical fiber cantilevers [n-type silicon cantilevers (f 1/4 37.2 kHz; k 1/4 0.01–0.60 N/m; tip radius of 10 nm)], using a Nanonics MultiView MV1000. Analysis was performed with the Gwyddion 2.42 software.

S. Typhi WT and S. Typhi Δzzz derivatives were transformed with a religated pCR TOPO TA^® multicopy plasmid (Lac⁺), producing strains constitutively expressing β-galactosidase (since S. Typhi lacks the lacI gene) (Valenzuela et al., 2014). All the strains were tested in LB plates supplemented with ampicillin and Xgal before the assays. For the assay, bacteria were cultured in LB to approximately OD₆₀₀ 1.1, prior to obtaining the pellet and the supernatant fraction by centrifugation. β-galactosidase activity was determined as described (Fuentes et al., 2009). The percentage of β-galactosidase release was calculated as (β-galactosidase activity of supernatant)/[(β-galactosidase activity of supernatant + β-galactosidase activity of pellet)] × 100.

To isolate OMVs, we followed a previously reported protocol, with modifications (Liu et al., 2016b). Bacteria were grown in LB at 37°C with shaking (approximately to OD₆₀₀ = 1.1), prior being centrifuged 10 min at 5400 ×g at 4°C. The pellet was discarded, and the supernatant fraction was filtered (0.45 μm), ultrafiltered with Ultracel^® 100 kDa ultrafiltration disks (Amicon Bioseparations), and ultracentrifuged 3 h at 150000 ×g at 4°C (Thermo Scientific™ Sorvall™ WX, Rotor AH-629). The supernatant was discarded, and the pellet resuspended in 1 ml DPBS. OMVs were stored at -20°C until their use. We quantified OMV yield as by determining the both protein content (BCA assay) and lipid content (FM4-64 molecular probe), and normalizing by CFU/ml (McBroom et al., 2006; Deatherage et al., 2009). We determined OMV size as described (Deatherage et al., 2009). Briefly, OMV size (diameter) was measured from at least 3 TEMs of 3 independent OMV extracts per strain in Adobe Photoshop using the ruler tool. Results were presented in diameter ranges of 3 nm, where 3 represents 1 to 3 nm. All OMVs larger than 60 nm were grouped in the last category (60+ nm).

Outer membrane vesicle extracts were bound to formvar-coated slot grids, stained with 1% aqueous uranyl acetate for 1 min, and viewed with a Philips Tecnai 12 (Biotwin) transmission electron microscope.

Mass spectrometry based on Matrix Assisted Laser Desorption Ionization Time-of-Flight (MALDI-TOF) was performed in a VITEK^® MS device (bioMérieux) (wavelength: 337.1 nm, beam divergence: 3.5 × 3 mrad; rate repetition and pulse duration: 3 ns/60 Hz; Maximum output: 150 μJ/pulse). Data were processed by Acquisition Software. Experiments were performed in biological duplicate, using two independent OMV extracts.

Although the cytolysin/hemolysin hlyE is efficiently expressed in rich media at 37°C, S. Typhi WT exhibits no hemolysis when cultured on blood agar plates (Jofre et al., 2014). Taking advantage of this feature, we performed a random insertional mutagenesis based on the EZ-Tn5 < R6Kγori/KAN-2 > transposome (Supplementary Figure 1) to find S. Typhi mutants with an increased HlyE secretion. To this purpose, we screened 15,000 random mutants (approximately 3 times the total number of S. Typhi genes, to cover the whole genome) (Porwollik et al., 2004; Valenzuela et al., 2014) on blood agar plates. Twenty-three clearly hemolytic colonies were identified and used for further analyses, including the identification of the corresponding interrupted genes (Figure 1A and Table 1). To corroborate that the observed hemolytic phenotype was effectively due to the loss of function of the identified genes, we performed a “clean” deletion by Red-Swap recombination (Datsenko and Wanner, 2000) for each identified gene in a S. Typhi WT background prior to testing the hemolytic phenotype of the resulting mutants. Some “clean” deletion mutants showed no hemolysis (Table 1), strongly suggesting that the hemolytic phenotype could be due to a secondary, spontaneous mutation or to polar effects attributed to the presence of the transposon. In these cases, genes were discarded. By contrast, 9 clean mutants (collectively called S. Typhi Δzzz::FRT) conserved the hemolytic phenotype and were used for further analysis (Figure 1B and Supplementary Figure 2).

To determine that the observed hemolysis depends on the presence of hlyE, we generated S. Typhi Δzzz::FRT ΔhlyE::kan double mutants. We were unable to detect hemolysis in the double mutants, showing that hlyE was effectively the genetic determinant of this phenotype (Figure 1C and Supplementary Figure 2).

To assess whether the Δzzz deletions affected hlyE transcription, we performed qRT-PCR. As shown in Figure 1D, Δzzz deletions did not affect hlyE transcription, indicating that the hemolytic phenotype exhibited by the S. Typhi Δzzz mutants is more likely due to an increased HlyE secretion than to an increased hlyE expression. The only exception was observed with S. Typhi Δhns, where the hlyE transcript was highly abundant compared with the WT, showing that H-NS contributes to the repression of hlyE expression (Figure 1D).

To corroborate that Δzzz deletions contribute to an increased HlyE secretion, we epitope-tagged HlyE (Jofre et al., 2014), prior to analyzing the supernatant and the whole bacterial lysate by Western Blot. As shown in Figure 1E, almost all the mutants exhibited similar amounts of HlyE in the whole lysate. The increased production of HlyE, detected in the Δhns genetic background, can be attributed to the increased amount of hlyE mRNA in this same background (Figure 1D). On the other hand, the presence of HlyE in the supernatant fraction, as an indicator of HlyE secretion, was clearly increased in all the tested mutants in comparison with the WT (Figure 1E).

Altogether, these results indicated that deletions of zzz genes produced an increased HlyE liberation.

Since periplasmic proteins are prone to be secreted via OMVs (Haurat et al., 2011), increased secretion of periplasmic, functional proteins, such as β-lactamase, has been considered a typical feature of hypervesiculating mutants (Schwechheimer and Kuehn, 2015). To determine whether Δzzz deletions promote the release of periplasmic proteins, different from HlyE, we constructed a S. Typhi parental strain harboring one copy of the bla (β-lactamase) gene stably inserted into the chromosome (i.e., S. Typhi Δatt_BG1::pNFB9). We used this strain to construct the S. Typhi Δatt_BG1::pNFB9 Δzzz double mutants. To determine whether Δzzz deletions improve β-lactamase secretion, we performed a satellite-based bioassay. As shown in Figure 2A, all the S. Typhi Δatt_BG1::pNFB9 Δzzz mutants allowed the growth of blue, satellite colonies around a spot of bacteria plated over a lawn of a Lac⁺ Amp^S reporter strain. By contrast, no satellites were observed with the S. Typhi Δatt_BG1::pNFB9 strain (WT background) (Figure 2A). As a semi-quantitative approximation, dilutions of precipitated proteins obtained from the corresponding supernatants were tested as described above. As shown in Figure 2B, some Δzzz mutants seem to secrete higher amounts of β-lactamase, since it was possible to observe blue satellites even in the most diluted samples (e.g., ΔrfaE, ΔtolR, and ΔdegS), whereas other S. Typhi Δzzz mutants only allowed the growth of satellites with the most concentrated samples (e.g., ΔompA, ΔwaaC, ΔnlpI, and Δhns). On the other hand, S. Typhi Δatt_BG1::pNFB9 (WT background) supernatant was unable to allow the growth of satellites, even in the most concentrated sample. These results strongly suggest that Δzzz mutants release distinct amounts of proteins. To test this hypothesis, we compared the amount of total released proteins from S. Typhi WT and S. Typhi Δzzz mutants. We found that S. Typhi ΔdegS secreted 30% more proteins into supernatant fraction compared with the WT (Figure 2C). This result might explain the increased secretion of both HlyE and β-lactamase in this mutant. Nevertheless, we observed that S. Typhi ΔompA, ΔmrcB, ΔwaaC, ΔrfaE, ΔnlpI, ΔyibP, and ΔtolR were undistinguishable from the WT regarding the amount of total secreted proteins (Figure 2C). On the other hand, S. Typhi Δhns released approximately 25% fewer proteins (Figure 2C). These data suggest that the increased secretion of HlyE and β-lactamase cannot be merely explained by a generalized protein hypersecretion, at least in these last cases, suggesting that periplasmic proteins are prone to be secreted due to another phenomenon. These results prompted us to qualitatively assess the exported proteins by SDS–PAGE. As control, we loaded proteins obtained from the respective whole lysate. As shown in Figure 2D, S. Typhi Δzzz mutants seem to secrete a distinct pattern of proteins (including distinct abundance and diversity).

All these results show that S. Typhi Δzzz mutants can secrete a distinct pattern of proteins, including functional periplasmic proteins such as HlyE and β-lactamase.

It has been reported that S. Typhimurium mutants lacking OmpA, LppAB, Pal, TolA, or TolB exhibit an increased production of OMVs (Deatherage et al., 2009). Nevertheless, those mutants also present severe problems regarding membrane integrity, leading to high susceptibility to bile acids (e.g., sodium deoxycholate). This phenomenon was also observed in other Gram-negative bacteria (McBroom et al., 2006). In this sense, it has been proposed that mutants producing more subtle changes in envelope crosslinking can reveal more physiological conditions involved in OMVs biogenesis (i.e., “subtle mutants”). To check the envelope integrity and to determine whether the S. Typhi Δzzz mutants are more susceptible to sodium deoxycholate, bacteria grown to approximately OD₆₀₀ = 0.5 were challenged with 0.5% deoxycholate or PBS (control) for 2 h. As shown in Figure 3A, only S. Typhi ΔompA and S. Typhi ΔyibP exhibited a statistically significant increased susceptibility to sodium deoxycholate. The same experiment performed with bacteria grown to approximately OD₆₀₀ = 1.1, and including complementation of the mutations, gave similar results (Supplementary Figure 3). To further estimate the envelope integrity, we determined the relative susceptibility to vancomycin by a Kirby–Bauer assay. Gram-negative bacteria are intrinsically resistant to vancomycin because of the limit of diffusible molecules through the bacterial envelope (Pimenta et al., 1999). Thus, an increased susceptibility to vancomycin could be interpreted as an increased permeability. We found that almost all the S. Typhi mutants exhibited complete resistance to vancomycin, except for S. Typhi ΔompA and S. Typhi ΔyibP (Figure 3B). These same mutants showed increased sensitivity to sodium deoxycholate, strongly suggesting that the envelope may be compromised in these cases. The remaining mutants were undistinguishable from the WT strains (i.e., “subtle mutants”). Remarkably, when S. Typhimurium and the corresponding S. Typhimurium Δzzz mutants were evaluated, a different pattern of sensitivity was observed (Supplementary Figure 4), reinforcing the fact that data obtained from S. Typhimurium cannot be freely extrapolated to S. Typhi without an experimental approach.

To better characterize the S. Typhi Δzzz mutants regarding their envelope integrity, we assessed the release of cytoplasmic proteins. β-galactosidase release has been used as an indicator of cell disruption in many cell types, including S. enterica (Kong et al., 2008). Thus, we transformed the pCR TOPO TA^® religated, multicopy plasmid (lacZ⁺) into S. Typhi WT and S. Typhi Δzzz mutants, to generate the corresponding Lac⁺ version of these strains. Since S. Typhi lacks lacI (Parkhill et al., 2001; Valenzuela et al., 2014), the Lac⁺ phenotype is constitutive and presented no statistically differences among S. Typhi Δzzz mutants (Supplementary Figure 5). As shown in Figure 3C, the ratio of β-galactosidase activity in the supernatant versus total β-galactosidase activity indicated leakage of cytoplasmic enzymes. Although most S. Typhi Δzzz mutants seem to release more β-galactosidase than the WT, only S. Typhi ΔyibP and ΔdegS exhibited statistically significant differences. Since it has been considered that a ratio superior to 0.5 could correspond to cell lysis (Kong et al., 2008), we speculate that S. Typhi ΔyibP and ΔdegS present some leakage of cytoplasmic enzymes, where bacterial lysis is not the most important factor, and such leakage is less pronounced in the other S. Typhi Δzzz mutants. Altogether, these results show that most S. Typhi Δzzz can be considered as “subtle mutants” since they exhibit a good envelope integrity and a relatively low leakage of cytoplasmic proteins, even though the reporter gene (lacZ) is in a high gene dosage in this case.

At this point, we have evidence arguing for the role of zzz genes in OMV biogenesis: (1) all the S. Typhi Δzzz mutants export increased amounts of periplasmic proteins (i.e., HlyE and β-lactamase) (Figure 1, 2), where S. Typhi HlyE is known to be secreted via OMVs (Wai et al., 2003); (2) S. Typhi Δzzz mutants have a distinct protein secretion into the supernatant fraction (Figure 3); (3) all S. Typhi Δzzz mutants, except for S. Typhi ΔompA and S. Typhi ΔyibP, exhibited a good envelope integrity, compared with the WT (Figure 3); and (4) all zzz genes, except hns, encode functions clearly related to the bacterial envelope (Table 1). In this context, we examined bacterial surface by AFM to determine whether Δzzz deletions produce envelope changes that could be associated to OMV biogenesis. AFM allows obtaining high-resolution images of bacterial samples, measuring structures in scales from molecules to cells (Liu and Wang, 2010). As shown in Figure 4, S. Typhi WT exhibited a smooth surface. By contrast, we found that, in all the cases, S. Typhi Δzzz mutants presented some changes in bacterial surface, such as waves, protrusions, creases or structural modifications that could correspond to envelope evaginations, consistent with a potential hypervesiculator phenotype (see Supplementary Figure 6 to complement the analysis). In the case of S. Typhi ΔyibP, we found two cell configurations: filament-like (Supplementary Figure 7) and isolated (Figure 4). On the other hand, several structures, consistent with released OMVs, can be observed in S. Typhi Δzzz mutants. Similar images obtained by AFM have been previously interpreted as OMV release (Wai et al., 2003; Chatterjee and Chaudhuri, 2011; Thoma et al., 2018). In this sense, ompA has been previously reported as involved OMV biogenesis in S. Typhimurium (Deatherage et al., 2009), and zzz genes are encoding different functions related to the bacterial envelope. For these reasons, we hypothesized that S. Typhi Δzzz are producing distinct OMVs, exhibiting different structural features, such as the size distribution, and potentially distinct OMV cargo. Thus, we isolated OMVs from S. Typhi WT and Δzzz mutants, prior to observing them with transmission electronic microscopy. As shown in Figure 5A, S. Typhi Δzzz mutants produced morphologically distinct OMVs, showing different mean sizes and more variability compared with OMVs produced by the S. Typhi WT. Δzzz deletions affected OMV size and/or OMV size distribution, where S. Typhi ΔompA, ΔtolR, ΔdegS, and Δhns tended to produce bigger OMVs compared with the WT, S. Typhi ΔmrcB produced smaller OMVs, and S. Typhi ΔwaaC was prone to produce both smaller and bigger OMVs. Nevertheless, all mutants seemed to produce more variable OMVs regarding their size, when compared with WT OMVs (Figure 5B and Supplementary Figure 8, see Supplementary Table 2 for statistical analysis). In addition, we observed that the increased β-lactamase release found in S. Typhi Δatt_BG1::pNFB9 Δzzz double mutants (Figure 2) depends, at least in part, on the secretion of OMVs (Supplementary Figure 9). On the other hand, we were able to observe the direct release of OMVs from S. Typhi ΔtolR and ΔdegS, where differences in OMV size were also evident (Supplementary Figure 10).

To determine the OMV yield of S. Typhi Δzzz mutants, we performed two independent strategies. To this end, we cultured both the S. Typhi WT and the S. Typhi Δzzz in LB to stationary phase (OD₆₀₀ = 1.1), and OMVs were extracted. Then, we determined the protein content of OMVs as an indicator of abundance, as previously described, normalizing with CFU/ml (Deatherage et al., 2009). As shown in Figure 6A and Supplementary Figure 11, S. Typhi ΔtolR and S. Typhi ΔdegS clearly presented more proteins in the OMV fraction, whereas S. Typhi ΔompA, ΔmrcB, ΔrfaE, ΔyibP, and Δhns exhibited a slight but consistent increase in the OMV protein content. On the other hand, S. Typhi ΔnlpI exhibited similar results compared with S. Typhi WT and, strikingly, S. Typhi ΔwaaC seems to present less proteins in the OMV fraction. Considering that the heterogeneity in OMV size and the leakage of cytoplasmic proteins, such as in the ΔdegS mutant, could affect the protein content in the OMV fraction, we also performed a direct quantification of lipids in OMV preparation for vesicle quantification, as previously described. As well as for protein content, we also normalized our results with CFU/ml in each case (McBroom et al., 2006). As shown in Figure 6A and Supplementary Figure 11, the lipid content determination corroborated the results obtained with the protein content determination. An increase in both protein and lipid content in the OMV fraction have been previously interpreted as an increased OMV yield, and vice versa (McBroom et al., 2006; Deatherage et al., 2009; Schwechheimer et al., 2015). Interestingly, S. Typhi ΔwaaC released more HlyE and β-lactamase to the supernatant, compared with the WT, when 5 μl of a stationary culture were plated on agar plates (Figure 1, 2), but produced less OMVs per CFU. This result could be explained by the fact that S. Typhi ΔwaaC exhibited 10-fold more CFU/ml in a stationary culture, compared with the WT (Supplementary Figure 12).

Altogether, these results indicate that zzz genes participate in the OMV biogenesis, affecting OMV production. This assertion is supported by the AFM (Figure 4), where Δzzz mutants seem to show structures potentially corresponding to a hypervesiculating phenotype.

As stated above, protein cargo selection is a process that should occur during OMV biogenesis (Schwechheimer and Kuehn, 2015). Since zzz genes participate in the OMV biogenesis, we expect that cargo in OMVs derived from S. Typhi Δzzz mutants should be affected. To test this hypothesis, we performed MALDI-TOF mass spectrometry of OMVs extracts. We found that most mutants produced OMVs presenting MALDI-TOF profiles that were very similar to that obtained with the WT OMVs, except for S. Typhi ΔnlpI-derived OMVs, which exhibited slight differences (Supplementary Figure 13), and OMVs obtained from S. Typhi ΔtolR and ΔdegS (Figure 6B), which showed more evident differences. These results indicate that, at least for tolR and degS, zzz genes participate in processes that modulate OMV cargo. To observe potential differences in protein cargo in an independent way, we visualized OMV proteins with SDS–PAGE. As shown in Figure 6C, OMVs derived from S. Typhi WT exhibited few observable proteins, with a high abundance of a protein of approximately 55 kDa (plausibly corresponding to flagellin). This pattern is similar to OMV extracts derived from other WT S. enterica serovars (Liu et al., 2016a, 2017). We observed that protein cargo of OMVs derived from S. Typhi Δzzz mutants was different to that observed for the WT OMVs, corroborating the role of these genes in OMV cargo selection. Interestingly, when OMVs obtained from S. Typhi ΔmrcB were analyzed, we also observed differences in comparison with the WT OMVs. Although OMVs derived from S. Typhi WT and S. Typhi ΔmrcB presented a similar MALDI-TOF pattern (Supplementary Figure 13), it is possible that more subtle differences cannot be detected by the methods performed in this work. Accordingly, albeit many bands are shared between WT OMVs and ΔmrcB OMVs (Figure 6C), their abundance differs. The same could be extrapolated to the other Δzzz mutants. In this sense, a more profound study must be performed to determine how exactly zzz are affecting OMV cargo.

According to these results, we conclude that zzz genes participate in OMV biogenesis in S. Typhi, affecting OMV size distribution, OMV production, and, at least for tolR and degS, we conclude that zzz are involved in OMV cargo selection.