The Type VII secretion system of Staphyloccocus

Abstract

The type VII protein secretion system (T7SS) of Staphylococcus aureus is encoded at the ess locus. T7 substrate recognition and protein transport is mediated by EssC, a membrane-bound multi-domain ATPase. Four EssC sequence variants have been identified across S. aureus strains, each accompanied by a specific suite of substrate proteins. The ess genes are upregulated during persistent infection and the secretion system contributes to virulence in disease models. It also plays a key role in intra-species competition, secreting nuclease and membrane-depolarizing toxins that inhibit the growth of strains lacking neutralizing immunity proteins. A genomic survey indicates that the T7SS is widely conserved across staphylococci and is encoded in clusters that contain diverse arrays of toxin and immunity genes. The presence of genomic islands encoding multiple immunity proteins in strains such as Staphylococcus warneri that lack the T7SS strains point to a major role for the secretion system in bacterial antagonism.

1. Introduction

The production of exotoxins is a key strategy that bacterial pathogens employ during infection (50, 102). Gram-negative bacteria have evolved numerous mechanisms to transport virulence factors across their double-membraned cell envelopes. Some of these pathways secrete proteins across the cell envelope in a single step, while others are two-step, with separate translocases moving proteins across the inner and outer membranes (37, 40, 79, 94). The cell envelopes of Gram-positive bacteria are generally considered simpler, presenting a single hydrophobic barrier. The inner membrane protein translocases found in Gram-negative bacteria, Sec (general secretory pathway) and Tat (twin arginine translocation) directly secrete proteins in Gram-positive bacteria (11, 59, 95, 124). Some Gram-positive pathogens, in particular from the Streptococcus and Staphylococcus genera, have an additional ‘accessory’ Sec pathway that operates in parallel to the canonical Sec pathway, and is largely dedicated to the secretion of surface glycoproteins (16, 19). A third general pathway, termed the type VII secretion system (T7SS), is also found in many Gram-positive bacteria. Here we review the T7SS of Staphylococcus.

2. The Type7a (T7a) and Type7b (T7b) secretion systems

2.1. The T7a secretion system of Mycobacteria

The T7SS was first described in Mycobacteria (61, 101, 115). Mycobacterium tuberculosis and its close pathogenic relatives secrete two small proteins, ESAT-6 (6-kDa early secretory antigenic target) and CFP-10 (10-kDa culture filtrate protein) that are potent T-cell antigens. These proteins, subsequently re-named EsxA (ESAT-6) and EsxB (CFP-10), lack cleavable N-terminal signal peptides that are found on substrates of Sec and Tat pathways, pointing towards a different secretion route. A cluster of genes was identified at the region of difference-1 (RD-1), which was deleted in the vaccine strain of Mycobacterium bovis. Mutational analysis of strains with an intact RD-1 showed that several genes encoded at this locus were required for the secretion of EsxA and EsxB. This system was subsequently named ESX-1, for ESAT-6 system-1 (ESX-1) (21, 61, 101, 115).

Genomic inspection indicates that slow-growing Mycobacteria generally encode multiple ESX secretion systems. For example, M. tuberculosis encodes five (38), along with at least 23 EsxA/EsxB paralogs (the exact number is strain dependent; (21)). The secretion machinery itself comprises the membrane proteins EccB, EccC, EccD and EccE (Figure 1a) (15, 54, 60, 99) that assemble into a hexameric arrangement (Figure 1a (15)). At the centre of this complex is EccC, a multi-domain ATPase of the FtsK/SpoIIIE family (15, 105). Multiple lines of evidence have shown that EccC recognizes substrate proteins through a C-terminal uncleaved signal sequence (23, 31, 44, 98, 105, 115, 134). A further essential component of the secretion machinery is mycosin (MycP), a subtilisin protease (39, 90, 113, 132). Surprisingly, the essentiality of MycP for ESX secretion is not linked to its protease activity, instead it seems to stabilize the assembled ESX complex (90; 128).

Figure 1. Schematic representation of the T7a and T7b secretion systems.

Components and substrates shared between the systems are shaded blue. (a) The T7a system has been extensively characterised in Mycobacteria but is also found in Streptomyces (5, 56) and other Actinobacteria (26). Components specific to T7a are shaded purple. A cytoplasmic ATPase, EccA, is found in some but not all T7a systems. Its co-occurrence with EspG suggests it may play a role in dissociating this chaperone from substrate proteins prior to secretion (52). Substrates of T7a include proteins from the WXG100, PE/PPE and Esp families. (b) The T7b system is found in bacteria of the Firmicutes phylum, including Staphylococcus, Listeria and Bacillus (14, 25, 97, 138). EssC is related to EccC in sequence and structure, but has two fork-head associated (FHA) domains at its N-terminus that are not found in EccC (105, 121, 149). Components specific to T7b are shaded green. The small cytosolic EsaB protein has a ubiquitin fold and is essential for T7b activity (74, 127). A similar domain is found at the C-terminus of EccD in the T7a system (54, 99). Characterised substrates of the T7b system are from the WXG100 and LXG protein families. Proteins are not drawn to scale.

Three different families of T7 substrate proteins have been characterised in Mycobacteria. The canonical substrates are EsxA/EsxB (123, 131). They are members of the WXG100 family - small helical hairpin proteins of approximately 100 amino acids, with a conserved WXG motif situated at the hinge region (93, 104). These proteins are secreted as paired heterodimers, with the C-terminal signal sequence on the EsxB component (31, 44, 105, 134). A second family is the PE/PPE proteins (2, 3). This is a heavily expanded family in the slow-growing Mycobacteria, forming up to 10% of the coding capacity of M. tuberculosis (12, 38). Both proteins are extensively α-helical; the shorter PE partner forms a helical hairpin that dimerizes with a pair of long α-helices in the PPE protein to form a four-helix bundle (116). A C-terminal secretion signal is present on the PE protein (44). PE/PPE complexes interact with a chaperone, EspG (34, 52, 76). This interaction stabilizes the complex and also targets the PE/PPE pair to the cognate ESX secretion system (45, 76, 96). The third substrate family is the Esp proteins, which are substrates of the ESX-1 secretion system (49, 83). Structural analysis of EspB reveals that its N-terminal region forms a four-helix bundle that can be superimposed onto an EsxAB or PE/PPE dimer, suggesting EspB is secreted as a monomer (77, 114).

Mycobacteria and other relatives in the Corynebacteriales order have an unusual cell envelope that comprises in addition to peptidoglycan, arabinogalactan and an outer ‘mycomembrane’ containing mycolic acids (42). The mycomembrane to some extent mimics the outer membrane of Gram-negative bacteria, and recently it was shown that an important function of PE/PPE proteins is in nutrient transport across this barrier (133). The complex envelope also acts as a barrier to secretion as well as import. The core ESX machinery resides in the inner membrane and does not span the envelope, so how substrate proteins reach the cell surface is unclear (15, 54, 105). Although numbering of protein secretion systems has generally been confined to those in true Gram-negative bacteria, an argument was made that the ESX system should be named the type VII secretion system (T7SS; (1)). Despite some opposition (48) this nomenclature has now become established. Indeed, genomic analysis has revealed that components of the T7SS are encoded in some Gram-negative bacteria, although they have yet to be functionally characterised (126).

2.2. The T7b secretion system of Staphylococcus aureus

Even before the T7SS had been described in Mycobacteria, it was noted that homologs of ESAT-6/EsxA were encoded in the genomes of diverse Gram-positive bacteria. Furthermore, these genes often clustered with a gene encoding a predicted AAA+ ATPase of the FtsK/SpoIIIE family (93). In commonly-studied strains of Staphylococcus aureus, the EsxA homolog is encoded at a locus with a second WXG100 protein, EsxB. Burts et al. showed that both of these proteins were secreted, and that secretion was dependent on a functional FtsK/SpoIIIE ATPase, EssC (25). Further analysis of genes at the esxA cluster revealed that in addition to essC, esaA, esaB, essA, essB were all essential for EsxA secretion (25, 74) (Figure 1b). Studies with Bacillus subtilis confirmed that these five genes were also required for the secretion of YukE, the only WXG100 protein encoded in that organism (14, 62). Given the limited similarity between the Mycobacterial and Firmicutes T7SS, they have been designated T7a and T7b, respectively (1).

Sequence comparisons indicate that apart from the ATPase EssC/EccC, there is little similarity between the membrane components mediating T7 secretion in the T7a and T7b systems. Even the ATPases differ at their N-termini; the T7b ATPase, EssC, has two forkhead associated (FHA) domains prior to the transmembrane segments, that are lacking in EccC (121, 149) (Figure 1). FHA domains have diverse functions, sometimes interacting with phosphothreonine-containing peptides (140), although residues required for recognition and binding of phosphothreonine are not conserved among EssC proteins (121, 149). A further difference can be found with the WXG100 proteins – while EsxAB heterodimers are seen for T7a, in the T7b systems, EsxA, and EsxB (where present) instead form homodimers (104, 118, 119). Structural information is lacking on T7b complexes; there are structural details for the individual components EsaA (108), EssB (122, 148, 150) and EssC (86, 105, 149), but little information about how these proteins interact with one another. Homo-oligomeric interactions of EsaA, EssB and EssC have been detected (6, 68, 85, 122), including hexamers of EssC (86, 149). Complexes containing mixtures of these proteins have been purified in small amounts (6, 85), but at present the overall architecture of the T7b system is not known.

2.3. Structural and mechanistic analysis of T7 protein secretion

Despite the differences in machinery components, it is likely that the mechanism of protein secretion is similar between T7a and T7b systems. At least some substrates of T7b have a C-terminal signal sequence similar to that found on T7a substrates (7, 119), and substrate recognition in type 7b is also at the level of the EssC ATPase (9, 28, 67, 86). Experiments using irreversibly crosslinked forms of B. subtilis EsxA have shown that the protein is exported as a folded dimer, and it is probable that the heterodimeric EsxAB proteins from the T7a systems are also secreted as a complex (e.g. (44, 104)). Structural analysis of an EccB/EccC/EccD/EccE protomer of Mycobacterial ESX-3, and docking into the lower resolution structure of the assembled ESX-5 complex, suggests that the translocation pore is formed primarily from the transmembrane helices of hexameric EccC, with an estimated pore size of 25Å, sufficient to accommodate a pair of folded WXG100 proteins (54).

The EccC/EssC ATPase clearly lies at the heart of the secretion mechanism. The protein has an extensive C-terminal cytoplasmic region. Initially anticipated by sequence analysis to have three P-loop ATPase domains, structural studies have now revealed a fourth, that was originally assigned as a domain of unknown function (D0 on Figure 2) (54, 99). ATP binding and hydrolysis by each of these domains appears to be essential for secretion activity (54, 86, 103, 105, 149). X-ray crystallographic analysis of a fragment of EccC containing D1-D3 has shown that each of these domains has a surface cleft. For D1 and D2, the cleft is occupied by a linker from the neighboring domain, which allosterically regulates ATPase activity (105). D3, which is at the very C-terminus of the protein, has an unoccupied cleft that interacts with the C-terminal signal sequence of a WXG100 substrate protein (105, 134). Experiments in S. aureus have also implicated D3 in substrate recognition (67), but have shown that in vitro, D2 can also bind a substrate protein (86). Taken together, these data suggest a model whereby substrate binding to D3 modulates the conformation and catalytic activity of the ATPase domains, freeing up the substrate binding pockets on D2 and D1. Sequential interaction of a substrate with D2 and D1 would funnel it towards the translocation pore (Figure 2). A stalk domain that bridges transmembrane helix 2 and ATPase domain D0, contacts the N-terminal helical region of EccB, and could link conformational changes in EccC to pore opening (54) (Figure 2). The analogous partner to EccB in the T7b system is not clear but could be EssB (which displays a similar topology to EccB, although it has a much larger cytoplasmic domain) (Figure 1).

Figure 2. Model for protein secretion by the T7SS.

For clarity, only EccC is shown. Step 1. A substrate protein (shown as a WXG heterodimer) interacts with the most C-terminal ATPase (D3) of hexameric EccC. Step 2. ATP hydrolysis at D3 causes a conformational change that liberates the signal sequence binding cleft of D2 which becomes occupied by the substrate. Step 3. ATP hydrolysis at D2 results in a similar conformational change at D1 and substrate occupation of the liberated binding cleft. Step 4. ATP hydrolysis at D1 passes the substrate to the D0 domain. ATP hydrolysis at D0 results in conformational changes at the linker domain and opening of the pore for passage of substrate across the membrane. Adapted from (54).

The recent structural description of assembled T7a complexes has paved the way for a detailed understanding of protein secretion by this pathway, but several open questions remain. For example, there is growing evidence that the secretion signal is bipartite, with the positioning and sequence of the WXG motif (or variant thereof) important for substrate recognition (e.g. (20, 114, 119). At present, it is unclear how this is recognised and whether it is critical for protein secretion. Likewise, all characterised substrates have two α-helices that flank the WXG motif. Structural predictions indicate that an N-terminal helical hairpin is a common substrate feature, but it is not clear why this should be a necessary structural feature for transport by this pathway. Are all substrates of this pathway transported in a folded state? There is good evidence that the small helical hairpin complexes of EsxA/EsxB are exported as folded dimers, but some substrates are much larger and have C-terminal globular domains (e.g. (142)) that are larger than the estimated pore size of the secretion pathway. Do these globular domains unfold, and if so how? If not, does the transport channel open up further to allow the passage of larger folded proteins? And perhaps most pressing, just how structurally (and mechanistically) similar are the T7a and T7b systems?

3. T7SS diversity in Staphylococcus

3.1. T7 genetic organisation and regulation in S. aureus

The T7SS of S. aureus is encoded at the ess locus (Figure 3a). The 5’ end of the gene cluster harbours genes that encode the highly conserved core components of the T7 secretion machinery: esxA, esaA, essA, esaB and essB. Most of the essC gene is also highly conserved across S. aureus strains, but there is significant variability towards the 3’ end of essC and in the genes found downstream, which will be discussed further below.

Figure 3. The diverse genetic organisation of the S. aureus T7SS locus.

(a) Schematic of the T7 locus. The locus is flanked by genes encoding a hypothetical protein and the formate/nitrite transporter FocA. (b-e) The T7 loci of selected strains representing essC1, essC2, essC3 and essC4 variants, respectively. Genomic neighborhood comparisons were generated using FlaGs.py (107).

Genetic regulation of the ess locus is mediated through the alternative sigma factor, SigB, in a complex and indirect manner (109). Transcription of esxA is activated by the two-component system ArlRS, probably through the global regulator MgrA, and by the DNA binding protein SpoVG (41, 109). Both arlR and spoVG transcription are under positive control by SigB (71, 84). The Agr quorum sensing system also positively regulates esxA (51, 109), whereas the DNA-binding protein SarA represses esxA expression (109). Unlike arlR and spoVG, sarA is negatively regulated by SigB, meaning that the sigma factor has intricate control of T7 gene expression (17). Outside of the SigB circuit, the SaeRS two-component system, a major factor in the regulation of exoprotein production, also regulates production of the T7 components (7, 69). Clinical and laboratory strains frequently have mutations in these regulatory pathways (e.g. (70, 112)). For example, strains Newman, RN6390/NCTC8325-4 and COL, which are commonly used for the study of T7 secretion, have mutations in the SaeRS, SigB and Agr pathways, respectively (4, 57, 89). This leads to variable levels of T7 protein production across different strains (7, 30, 74).

A longitudinal study of an isogenic pair of S. aureus isolates from the lungs of a cystic fibrosis patient followed over 13 years, showed that the ess genes were highly up-regulated (approximately 10-fold) over this time (144). They are also upregulated during persistent colonisation in a mouse vaginal model (47). Other studies have shown that fatty acids present in pulmonary surfactant and serum induce expression of T7 genes (65, 82). The active component of serum is cis-unsaturated fatty acids that requires further bacterial metabolism through fatty acid kinase to exert its regulatory effect. At least part of this effect was linked to decreased membrane fluidity resulting from cis-unsaturated fatty acid incorporation (82). The fatty acid-mediated induction of ess gene expression was independent of ArlRS, SaeRS and Agr (65, 82), but in the case of pulmonary surfactant-derived fatty acids still required sigma B (65).

Despite the extensive regulation mediated by sigma B, expression of esxA is controlled solely through a sigma A-dependent promoter, mapped to 74 base pairs upstream of the start codon of esxA (74, 109). Initial work found that esxA is monocistronic in strain Newman and not co-transcribed with the other ess genes. A putative Rho-independent terminator is located just after the esxA stop codon (109). Kneuper et al. (74) extended these observations by examining transcriptional organisation of the ess cluster in several strains within the same clonal complex. Surprisingly, even though there is 100% sequence identity across the esxA region in these strains, esxA was found to be monocistronic in RN6390 and Newman, but co-expressed with the downstream ess genes in COL, USA300 and SA113. A transcriptional start site was mapped within the esxA-esaA intergenic region of RN6390 that is presumably also present in Newman, but was not found in USA300 (74). Although the terminator sequence was present for all of the strains studied, there must be some read-through in strains where an esaA promoter was not detected. Given the sequence conservation at this region, it is not clear why the esaA promoter should be cryptic in some strains, nor how termination readthrough is mediated. Studies have shown that esxA is very highly expressed relative to the other ess genes, with esxA transcript abundance approximately 100 times higher than essC across multiple strains (74, 136).

A further gene, expressed divergently to esxA on the opposite strand, is also conserved across staphylococci and encodes EssH, a non-lytic peptidoglycan hydrolase (Figure 3a) (18). EssH is essential for secretion of the T7 substrates EsxA and EsxC in USA300 (18). It should be noted that outside of the staphylococci, homologues of EssH are not encoded at the T7 clusters, or indeed elsewhere on the chromosome. However, the cysteine, histidine-dependent amidohydrolase/peptidase (CHAP) domain found within EssH is present in a number of proteins encoded by diverse Gram-positive bacteria. At present it is unclear whether any of these could substitute for the role of EssH in these organisms.

Several open questions remain regarding the regulation of T7 gene expression. Apart from sigma A, it is not clear which transcription factors directly interact with the esxA promoter region, nor how the expression of essH is controlled. It is likely, however, that there is additional regulatory complexity at this locus. For example, it was recently reported that the small non-coding RNA, RsaI, interacts with esxA mRNA and upregulates its expression and/or stabilizes its mRNA. RsaI expression is controlled by cellular glucose levels and appears to link gene expression with carbon source availability (22). It will be interesting to see how these factors interlink to fine-tune control of the T7 secretion machinery.

4. T7SS variability in S. aureus

4.1. The EssC1 variant cluster

Comparative genomic analysis revealed that there is extensive variability at the 3’ region of the ess locus (136). Sequence divergence starts approximately ¾ of the way through essC, resulting in four EssC variants across S. aureus strains (Figure 3). Each essC variant is associated with its own suite of genes directly downstream. The variant found most commonly is termed essC1 (Figure 3b) and is the most extensively studied to date, with strains Newman, USA300 and COL being among those that possess the essC1 arrangement (136). To the immediate 3’ of essC1 are a number of genes known to encode T7 substrates: esxC, esxB, esxD and esaD (8, 24, 25, 28, 74, 91). EsxB, like EsxA, is a protein of the WxG100 family. However, unlike Mycobacteria, where EsxA and EsxB heterodimerise, the two S. aureus proteins do not interact with one another (118). Moreover, esxB mRNA is at least 100 fold less abundant than esxA indicating that these two proteins are probably present at very different stoichiometries (74, 136). Instead, EsxB was shown to interact with EsxD, another small substrate protein, suggesting that they may be exported as a pair. Like EsxD, EsxC is a small secreted protein distinct to the variant 1 T7SS. EsxC and EsxD are not classified as members of the WXG100 protein family, but structural predictions indicate that both are predominantly α-helical, similar to other characterised T7SS substrates. The biological functions of EsxB, EsxC and EsxD are yet to be fully elucidated.

EsaD (also called EssD), at 614 amino acids in length, is the largest T7 substrate identified in S. aureus. It is encoded only in strains that have the essC1 variant (136). The N-terminal half of the protein is predicted to be extensively helical, whereas the C-terminal approximately 150 amino acids shares an endonuclease fold, and has been shown to have DNase activity (28, 91). EsaD is highly toxic and producing strains are protected through co-production of an antitoxin, EsaG, which interacts specifically with the nuclease domain (28, 91). EsaE, also encoded at the ess locus in variant 1 strains, interacts with the N-terminal domain of EsaD (9, 28). EsaE appears to act as a chaperone because it stabilizes the full-length form of EsaD; it also interacts with EssC1 and therefore serves to target the EsaDG complex to the secretion machinery (9, 28). In this respect it is analogous to the EspG chaperone found in some of the T7a systems, although the two proteins do not share any significant sequence similarity. Interestingly, pull-down experiments have shown that EsaE also interacts with EsxC, but not with EsxB or EsxD (9). This raises the possibility that, like the situation in Mycobacteria, some substrates are targeted to the EssC1 secretion system through interaction with a chaperone whereas others are chaperone-independent.

4.2. The EssC2 and EssC3 variant clusters

EssC1 is the most dissimilar of the four S. aureus EssC variants, diverging from the other three ATPase sequences part way through the D2 ATPase domain. The EssC2, EssC3 and EssC4 variants share 100% sequence identity throughout D2 but differ in sequence in the final ATPase domain. Furthermore, none of these other strain variants encode an EsaE-like chaperone anywhere in the genome. The essC2 and essC3 clusters are most similar to one another (136), and several genes immediately downstream of essC in these clusters also share limited sequence homology (between 30 - 50% identity; Figure 3c, d). The first two genes encode small proteins that are predicted to be predominantly α-helical. SAPIG0303/SAR0285 are members of the DUF5344 protein superfamily. Modelling suggests they are likely to have the same overall fold as EsxA and EsxB, although they lack the WXG motif. SAPIG0304/SAR0286 fall into the DUF5082 family. Nothing is known about these proteins but given their small size and predicted helical structure they are strong candidates for T7 substrate proteins.

SAPIG0305/SAR0287 are larger proteins that have an LXG domain at their N-terminus, marking them as T7 substrates. SAPIG0305, also named EsxX, has been investigated in strain ST398 and shown to be secreted by the T7SS, although its precise biological function remains obscure (43). SAPIG0306 and SAR0288 have no detectable sequence similarity but are both members of the DUF5079 family and are highly hydrophobic with six predicted transmembrane segments. All LXG proteins characterised to date are toxins, and are co-produced with a cognate antitoxin that is usually encoded immediately downstream (73, 125, 142, 146). It is therefore likely that SAPIG0306 and SAR0288 are antitoxins for the LXG proteins to which they are adjacently encoded. A close homologue of SAPIG0306, SAPIG0312, with which it shares 90% sequence identity, is also encoded at the ess locus in strain ST398 (136).

The SAPIG0307 and SAPIG0308 proteins, like SAR0289 and SAR0290, are members of the DUF5085 family. This family comprises small proteins, less than 150 amino acids in length, about which nothing is currently known. Structural prediction suggests they are unrelated to other small T7SS substrate proteins and have mixed α-helical/β-sheet secondary structure. While SAPIG0307 and SAR0289 share 50% sequence identity, the similarity between SAPIG0308 and SAR0290 is lower (28%), and neither SAPIG0307/SAPIG0308 nor SAR0289/SAR0290 have significant similarity with one another. It is not clear whether they are secreted substrates of the T7SS, but if they are, they would constitute a new exo-protein family.

4.3. The EssC4 variant cluster

EssC4 is the least common of the four EssC types, being found in only approximately 5% of sequenced S. aureus strains (136). Like essC2 and essC3, there is no WXG100-family protein encoded downstream of essC4 (Figure 3e). However, both SAEMRSA15_02480 and SAEMRSA15_02500 are small all-helical proteins that can be structurally modelled on EsxA, suggesting that they are likely substrates of the EssC4 secretion system. The encoding genes flank SAEMRSA15_02490, which codes for an LXG protein, and is therefore also a probable EssC4 substrate. Of the other genes in the essC4 3’ region, SAEMRSA15_02470 encodes a 359 amino acid protein that is predicted to have mixed α-helical/β-sheet secondary structure and is likely cytoplasmic, whereas SAEMRSA15_02510, SAEMRSA15_02530 and SAEMRSA15_02540 encode small polytopic membrane proteins from unrelated families. The other gene in this region, SAEMRSA15_02520, is a predicted lipoprotein with a cystatin-like fold, which presumably has an extracellular function.

5. T7 substrate proteins encoded outside of the ess gene cluster

All strains of S. aureus encode a conserved LXG family protein outside the ess locus (Figure 4a). This protein, named TspA, is exported by the T7SS and remains surface attached, with a C-terminal toxin domain exposed close to the cell surface. Immediately downstream of tspA is tsaI, encoding a membrane-bound antitoxin of the DUF443 protein family (125). Unlike most other LXG proteins, no WXG100-family protein is encoded in the vicinity of tspA.

Figure 4. Predicted T7SS substrates encoded outside of the ess locus.

(a) TspA, a membrane-depolarizing toxin is encoded in a gene cluster with a variable number of genes encoding DUF443 proteins. The DUF443 protein encoded immediately downstream of tspA, has been re-named TsaI and is a TspA immunity protein (125). Towards the end of the cluster is a gene that encodes just the TspA C-terminal region lacking the LxG domain. (b) The locus encoding the DUF5344-family protein. Some S. aureus strains possess an insertion at this locus encoding either (c) a TspA/DUF443 pair or (d) an orphan DUF443 protein. Pseudogenes are shown with a dashed outline. Genomic neighborhood comparisons were generated using FlaGs.py (107).

A conserved locus encoding a protein of the DUF5344 family is also present across S. aureus strains (SAOUHSC_02907; Figure 4b). This DUF family is also annotated as TIGR04197: T7SS_SACOL2603, and both SAPIG0303 and SAR0285 (encoded at the ess locus in variant 2 and variant 3 strains, respectively) are family members. TIGR04197: T7SS_SACOL2603 has been linked to the T7SS because these proteins are restricted to species that encode a T7SS, and they share remote sequence similarity to WXG100 proteins. Also encoded at this locus are two small cytoplasmic proteins, SAOUHSC_02906 and SAOUHSC_02905. SAOUHSC_02906 is a member of the DUF3958 family, about which nothing is known. Structural predictions, however, suggest that it is largely α-helical so could potentially represent a novel T7SS substrate. SAOUHSC_02905 is a DUF4176 protein and predicted to be primarily β-sheet. Intriguingly, in a few S. aureus strains there is an insertion in this region, between the DUF5958- and DUF4176-encoding genes (Figure 4c, d). In some strains, this insertion introduces two genes: one encoding a TspA homologue, TspA2, sharing 48% identity with the canonical TspA from RN6390, and the second encoding a DUF443-family protein related to TsaI (Figure 4c). A few other S. aureus strains lack tspA2 but carry the DUF443-encoding gene (Figure 4d).

6. T7SS activity in S. aureus

It has been noted that there is mutually dependent substrate secretion in the Mycobacterial T7SS, where several substrates are contingent on one another for export (e.g. (55, 145)). At present, it is unclear whether the secretion of different T7SS substrates is co-ordinated in S. aureus. It has been reported that EsxB, EsxD and EsaD are each essential for the secretion of EsxA and EsxC in USA300 (7), but this is not the case in RN6390 (28, 74). Whitney et al. (142) demonstrated that LXG proteins in Streptococcus intermedius have a cognate WXG100 protein partner that interacts specifically with the LXG domain and mediates export of the toxin. Not all LXG proteins in S. aureus are encoded with candidate WXG protein partners, although in most cases they are found in gene clusters that also encode small proteins with related folds. It is feasible that some of these may act as LXG partner proteins.

EssC is clearly a main specificity determinant for substrate recognition. It was shown that secretion of EsxC is only supported by EssC1, and not by EssC2, EssC3 or EssC4 (67). This is in accord with EsxC only being encoded in clusters alongside essC1, and implies that the EssC variable region mediates substrate selectivity. Although EsxA is a secreted substrate of the T7SS, it may also be considered a structural/functional component of the secretion system. Deletion of esxA abolishes secretion of other substrates, and overproduction of the EsaD substrate also results in much enhanced EsxA secretion (28). Interestingly, the EsxA sequence is almost invariant across S. aureus strains, and it is secreted by all four EssC subtypes (67). By implication, it would be expected that EsxA should interact with an invariant region of EssC (or potentially another T7SS component). The exact role of EsxA in T7 secretion is not known, but it may be involved in facilitating the active assembly of the machinery.

T7 secretion activity is relatively inefficient for a number of S. aureus strains when cultured in laboratory growth media, and varies considerably between S. aureus strains even within the same clonal complex. For example, production and secretion of EsxA in RN6390 and COL was observed at all cell densities, whereas the onset of EsxA secretion was delayed in USA300 and SA113. Production of EsxA in Newman was only detected at late exponential/stationary phase (74). Furthermore, in all strains examined, the vast majority of EsxA and EsxC was found in the cellular fraction, with supernatants requiring extensive concentration to detect these proteins in the secretome (30, 74).

Crosslinking and Blue native PAGE analysis of solubilised membrane fractions has indicated that in RN6390, the T7 membrane proteins do not form detectable heteromeric complexes (68). This ties in with the low secretion efficiency and may suggest that assembly of the active T7SS is transient, or that it requires some form of post-translational modification. The membrane chaperone flotillin has been implicated in promoting assembly of the S. aureus T7SS (85) and supplementation of the growth media with hemin also causes T7 hypersecretion in some strains (30). A further candidate for the post-translational regulation of T7 secretion is EsaB. This small cytoplasmic protein, which is essential for T7 activity, has a ubiquitin fold but lacks the ability to be conjugated to other proteins (14, 29, 62, 74, 127). Of note, a domain with the same fold is found at the C-terminus of the polytopic EccD component of the T7a system, where it acts as a scaffold for the cytoplasmic domains of the other T7a membrane proteins (54). It could plausibly play a similar role in the T7bSS.

7. The staphylococcal T7SS has roles in virulence and interbacterial competition

7.1. The T7SS contributes to S. aureus pathogenesis

S. aureus produces a plethora of secreted toxins and virulence factors (110, 120), and initial characterisation of the T7SS focused on establishing whether the secretion system was required for pathogenicity (24, 25). Inoculation of S. aureus into the bloodstream of mice seeds abscess formation in the peripheral tissues such as the kidney and liver (35). Mutant strains of Newman lacking the core components esxA, esaB and essC had a reduced ability to form abscesses in this infection model, and it was concluded that the T7SS supports bacterial persistence (24, 25). A similar reduction in abscess formation was noted when any of the substrate-encoding genes esxB, esxC or esaD were deleted in the Newman background (8, 24, 25).

These findings have been supported by studies using strain USA300, where a reduction in abscess numbers was observed for an esaE (essE) mutant. Further experiments measuring cytokine levels in the bloodstream of infected mice showed that production of interleukin-12, an immune modulator, was stimulated upon infection with USA300, but not to the same extent if esaD (essD) or esaE (essE) were deleted (9, 91). An essB mutant of strain ST398 also yielded smaller abscesses in a murine skin infection model and fewer kidney abscesses in a bacteremia model (9, 135). Deletion of esxX, which encodes a T7 substrate of ST398 appeared to mirror loss of essB, suggesting that EsxX may be required for abscess persistence in EssC2 strains (43).

Neutrophils are key players in the clearance of S. aureus infections (46, 87). Histology of skin abscesses formed during infection with ST398 showed that there was increased neutrophil infiltration when the T7SS was functional. In vitro experiments noted reduced neutrophil lysis by strains deleted for essB or esxX, accompanied by reduced bacterial survival of the mutant strains compared with wild-type ST398 (43, 135). Experiments using a lung epithelial line have shown that USA300, an EssC1 variant, can be readily internalized. Although no significant difference in internalization was observed when either esxA or esxB were deleted, an increase in apoptosis was seen when cells were infected with the esxA mutant strain, suggesting that EsxA or another substrate of the variant 1 T7SS acts as an anti-apoptotic factor (75).

A further virulence model that is becoming increasingly used for studies of staphylococcal infection is the zebrafish (e.g. (58, 100)). The larval stage lacks adaptive immunity and therefore provides a system to investigate interaction with the innate immune system. T7 mutants of RN6390 and COL showed a reduction in virulence in this model, and were more rapidly cleared by the larvae. A mutant of RN6390 lacking esaD was as virulent as the wild type, whereas a strain deleted for tspA was partially reduced in virulence and had increased clearance in vivo (125). Substantial recruitment of both neutrophils and macrophages to the site of infection was observed, but this was not significantly different between the wild type and any of the T7 mutant strains. Likewise, the larval pro-inflammatory markers interleukin 8 and interleukin 1 beta were induced during infection to a similar extent by the T7 mutant strains as the wild type (125).

Taken together it is clear that strains lacking a functional T7SS have a defect in virulence in murine and zebrafish models of infection, and some strains are also impaired in intracellular survival. However, the underlying mechanism/s behind these observations is not fully apparent. The inter-dependence of substrate secretion/stability in some strains is an added complication, as is the pleiotropy seen with mutations in T7 structural genes which leads to altered expression of many other genes including those for cell surface components (29, 30). Further studies are required to dissect the precise roles of the T7SS and its secreted substrates in pathogenesis.

7.2. The T7SS likely plays a major role in intraspecies competition

S. aureus is a frequent colonizer of the nasal cavity, and approximately 30% of humans are carriers (80, 141). This body site is a competitive environment, and numerous mechanisms exist among the microbiota to modulate S. aureus colonisation (66, 78, 92, 111, 147). S. aureus is also an earlier colonizer of the lungs of children with cystic fibrosis where it generally persists until adulthood (64). It is therefore likely that S. aureus employs competitive mechanisms to colonise such polymicrobial niches, and the T7SS may represent an important factor in this process. Indeed, it has been shown that the T7SS is conditionally essential for S. aureus survival during polymicrobial infection (63).

7.2.1. EsaD and TspA are T7 toxins that target closely related S. aureus strains

The first indication that the T7SS was associated with antibacterial activity came from studies on EsaD. This nuclease substrate is encoded at the T7SS locus of essC1 strains, where it is co-produced with an antitoxin, EsaG, a member of the DUF600 protein family, which protects the producer from nuclease activity prior to secretion (28, 91). It was noted that although essC2, essC3 and essC4 strains do not encode EsaD, they all encode copies of DUF600 proteins at their T7 loci (28, 136) (Figure 5). The inference was that the orphan EsaG homologues serve to protect these S. aureus strains from nuclease attack by variant 1 strains, and point to a role for the nuclease in intraspecies competition. The low T7 secretion activity of S. aureus strains when cultured in laboratory media has hampered studies to examine the role of the T7SS in interbacterial warfare. However, by overproducing EsaDG in strain COL, growth prevention of a target strain, RN6390, was observed. Growth inhibition was absolutely dependent on both an active T7SS in the attacker strain, and on functional nuclease activity. Increased protection was afforded to the target strain if it was provided with extra copies of esaG on a plasmid (28).

Figure 5. Highly variable ‘antitoxin islands’ are located at the S. aureus T7 gene cluster.

Genomic analysis reveals a diverse array of candidate antitoxin protein-encoding genes found at the ess locus of S. aureus strains; a few examples are shown. (a) essC1 variant clusters. (b) essC2 variant clusters. (c) essC3 variant clusters. (d) essC1 variant clusters. Genomic neighborhood comparisons were generated using FlaGs.py (107). Pseudogenes are highlighted with a dashed outline.

Interestingly, further DUF600 protein-encoding genes are also found just downstream of esaG in essC1 strains. For example, there are five of these in NCTC8325 (Figure 5a) (13, 136). These share 66-87% identity with EsaG, but do not appear to interact as tightly with EsaD as the cognate antitoxin. Moreover, they do not offer protection from self-killing by EsaD because even though they are present on the chromosome, esaG cannot be deleted in the presence of a functional esaD gene (28). Further analysis of EsaD indicates that it is a polymorphic toxin – the encoding gene is a hotspot for recombination within the 3’ region encoding the toxin domain (136). This leads to numerous EsaD variants across essC1 strains, and it is highly likely that accumulated EsaG homologues offer protection from EsaD variants secreted by competing strains. In support of this, it was noted that two polymorphic variants of the EsaD nuclease domain from variant 1 strains A8819 and SAU060112 showed different patterns of interaction with the six DUF600 proteins from NCTC8325 (27).

Recently it was shown that TspA is a second S. aureus T7 toxin with antibacterial activity. It has a C-terminal domain that is toxic when heterologously targeted to the periplasm of E. coli, where it depolarises the membrane. Protection from TspA activity is afforded by the membrane protein TsaI, a member of the DUF443 protein family. Across S. aureus strains, the C-terminus of TspA is polymorphic. Strains accumulate multiple DUF443 protein-encoding genes at the tspA locus, presumably as protection from TspA sequence variants that are secreted by competing strains (Figure 4b). Interestingly, many strains contain a second, truncated tspA gene at the tspA locus, that codes for a variant C-terminal domain (Figure 4b). It is plausible that through genetic recombination this may allow strains to produce different TspA variants to increase competitiveness.

The zebrafish infection model has recently been adapted to explore interbacterial competition in vivo (143). Pairs of bacteria are co-injected into the hindbrain, which is an otherwise sterile compartment, allowing their interaction to be monitored. The advantage of this system is that the T7 is clearly active when S. aureus is in this environment, and therefore avoids the requirement to overproduce candidate toxins from plasmids. Using this model, it was shown that strain COL was able to mediate killing of RN6390 in a T7SS-dependent manner, provided that the genes coding for antitoxins to EsaD and TspA were deleted from the RN6390 chromosome. Through use of attacker strains lacking either esaD or tspA it was shown that both toxins contribute to intraspecies competition (125).

7.2.2. Predicted T7-dependent toxin-antitoxin pairs

It is highly likely that there are additional antibacterial toxins across S. aureus strains, for example the EssC variant 2, 3 and 4 strains each encode an LXG-domain protein at their T7 loci. Strain variants 2 and 3 encode related LXG proteins, that share 33% sequence identity (termed EsxX in variant 2 strains (43)). In essC2 and essC3 strains, the encoding gene directly precedes a gene encoding a DUF5079 family protein that is a candidate antitoxin (SAPIG0306 and SAR0288 in variant 2 and 3 strains, respectively). Interestingly, a very close homologue of SAPGI0306 (98% sequence identity) is encoded in essC1 strains. This orphan gene, SAOUHSC_00270 in Figure 5a, is found among the cluster of genes encoding DUF600/EsaG family antitoxins in what appears to be an antitoxin ‘island’. This genomic evidence strongly supports the hypothesis that EsxX has anti-staphylococcal activity and plays a role in intraspecies competition. Likewise, a homologue of SAR0288, SAOUHSC_00254, is also encoded in essC1 strains, but this time at the 5’ end of the ess locus in what appears to be a second antitoxin island that is present in some strains (Figure 5).

These antitoxin islands are regions of high variability across S. aureus strains, containing different numbers of probable antitoxins, some examples are shown in Figure 5. In some respects, they appear analogous to the acquired interbacterial defence (AID) gene clusters found in Bacteroidales, that encode multiple antitoxins against Gram-negative type VI secreted toxins (106). Within the 3’ antitoxin island of NCTC8325 (Figure 5a), the encoded membrane protein SAOUHSC_00271, a member of the DUF5080 protein family, appears to be an orphan antitoxin, that shares 98% identity with SAPIG0309 (Figure 3). It is not clear what the likely toxin might be, but given that it is encoded immediately downstream of the DUF5085 pair SAPIG0307/SAPIG0308 it may be one or both of these proteins. SAR0291 (from variant 3 strains) is from the same DUF5080 family as SAPIG0309, and a close homologue of SAR0291 is also encoded in NCTC5325, but in the 5’ immunity island (SAOUHSC_00255; 85% sequence identity).

The LXG protein in essC4 variants, EMRSA15_002490, has yet to be characterised. It is the least common of all S. aureus toxins, being found in only a few strains primarily from clonal complex 22 (136). It is, however, encoded by other species including environmental organisms such as Sporosarcina pasteurii. The occurrence of this toxin in bacteria that are free-living non-pathogens is more likely to point to a role in bacterial competition rather than virulence.

Interbacterial competition therefore appears to be a major role of the T7SS in S. aureus. Indeed, based on growing evidence it is likely that this is also true for other Firmicutes (32, 33, 72, 73, 122, 142). This raises many interesting questions that remain to be answered. For example, which organisms can S. aureus target using its T7 toxins? Work with Streptococcus intermedius has shown that it can use T7 secretion to inhibit the growth of other Firmicutes such as Streptococcus pyogenes and Enterococcus faecalis, but not Gram-negative bacteria (142). What role does the T7SS play in S. aureus colonization of body sites and shaping the nasal microbiome? How do toxins access their sites of action in the target cell? Are they released from the surface of the producing cell like the toxin domains of the Cdi contact dependent toxins in Gram-negative bacteria (139)? And could the polymorphic toxin domains play other roles, such as kin recognition and competitive exclusion just as polymorphic toxins do in Gram-negative bacteria (10, 129, 130)?

8. T7SS of other Staphylococcus species

In contrast to the numerous T7SS studies in S. aureus, there is little information on this secretion system in other staphylococci. Genomic analysis does, however, indicate that the system is widely conserved across the species. A study of S. lugdunensis noted that the T7SS was found in two distinct organizations, one of which was closely related to S. aureus variant 2 and the other to variant 4 (81). Some strains of S. epidermidis also have the T7SS, again found in two distinct organizations related either to S. aureus variant 1 or variant 2 (36) (Figure 6a). Interestingly, while S. epidermidis strains cluster into two major lineages (the A/C lineage and the B lineage) the T7SS is only found in lineage B strains, suggesting that the lineages employ distinct mechanisms to underpin their occupancy of the skin niche (53). Neighborhood analysis indicates that extensive and variable ‘immunity islands’ are present at the 3’ end of the S. epidermidis T7 gene clusters, signifying strong selective pressure for the acquisition of immunity genes (Figure 6a). Strains of S. caprae and S. xylosus also encode the T7SS, which is closely related to S. aureus variant 2 (88, 117, 137) (Figure 6a).

Figure 6. T7-related genes are found in a wide range of Staphylococcus species.

(a) Genes encoding T7 structural components are present in a number of staphylococci including S. epidermidis and S. xylosus. Furthermore, DNA sequences that share identity with the coding regions of known toxin domains present on LXG proteins, are also found within these clusters (shaded in green with no outline). It is not clear whether these encode small toxins lacking an obvious T7-targeting motif or if these gene fragments can be subject to recombination events to fuse the toxin with an LXG domain, therefore permitting secretion. Some of these also appear to be pseudogenes (with frameshifts or premature stop codons) (b) Clusters including toxin fragments and immunity-encoding genes are found in a wide range of staphylococci, where genes encoding components of the T7 machinery are missing from the genome. * - sequence similarity to S. aureus EsaD, ** - sequence similarity to S. aureus SAPIG0305, *** - sequence similarity to HNH endonucleases, **** - sequence similarity to S. aureus SAR0287, # - sequence similarity to ribonucleases (found mostly in Bacillus strains), ## - sequence similarity to a predicted lipase. Genomic neighborhood comparisons were generated using FlaGs.py (107).

Interestingly, some staphylococcal strains, such as those listed in Figure 6b, do not encode the T7SS. Despite this, they contain genomic islands encoding multiple immunity genes to T7 toxins, along with fragments of toxin genes, again reminiscent of the Gram-negative AID gene clusters that offer protection from T6SS-mediated competition (106). It is highly likely that these genes contribute to survival of staphylococci in competitive niches, and point to a major role of the T7SS in intra-species antagonism.