A membrane-depolarising toxin substrate of the Staphylococcus aureus Type VII protein secretion system targets eukaryotes and bacteria

The type VII protein secretion system (T7SS) is conserved across Staphylococcus aureus strains and plays important roles in virulence and interbacterial competition. To date only one T7SS substrate protein, encoded in a subset of strains, has been functionally characterized. Here, using an unbiased proteomic approach, we identify TspA as a further T7SS substrate. TspA, encoded distantly from the T7SS gene cluster, is found across all S. aureus strains. Heterologous expression of TspA indicates that it has a toxic C-terminal domain that depolarizes membranes. The membrane depolarizing activity is alleviated by co-production of the TsaI immunity protein. Using a zebrafish hindbrain ventricle infection model, we demonstrate that the T7SS of strain RN6390 contributes to zebrafish mortality, and deletion of tspA leads to increased bacterial clearance in vivo. The toxin domain of TspA is highly polymorphic and S. aureus strains encode multiple tsaI homologues at the tspA locus, suggestive of additional roles in intra-species competition. In agreement, we demonstrate TspA-dependent growth inhibition of RN6390 by strain COL in the zebrafish infection model, that is alleviated by the presence of TsaI homologues. This is the first T7SS substrate protein shown to have activity against both eukaryotes and prokaryotes.


27
The type VII protein secretion system (T7SS) is conserved across Staphylococcus aureus 28 strains and plays important roles in virulence and interbacterial competition. To date only one 29 T7SS substrate protein, encoded in a subset of strains, has been functionally characterized.

30
Here, using an unbiased proteomic approach, we identify TspA as a further T7SS substrate.

31
TspA, encoded distantly from the T7SS gene cluster, is found across all S. aureus strains.

32
Heterologous expression of TspA indicates that it has a toxic C-terminal domain that 33 depolarizes membranes. The membrane depolarizing activity is alleviated by co-production of 34 the TsaI immunity protein. Using a zebrafish hindbrain ventricle infection model, we 35 demonstrate that the T7SS of strain RN6390 contributes to zebrafish mortality, and deletion 36 of tspA leads to increased bacterial clearance in vivo. The toxin domain of TspA is highly 37 polymorphic and S. aureus strains encode multiple tsaI homologues at the tspA locus, 38 suggestive of additional roles in intra-species competition. In agreement, we demonstrate 39 TspA-dependent growth inhibition of RN6390 by strain COL in the zebrafish infection model,

43
The Type VII secretion system (T7SS) has been characterised in bacteria of the actinobacteria 44 and firmicutes phyla. In pathogenic mycobacteria the ESX-1 T7SS secretes numerous proteins were found to be enriched in abundance in the essC secretome (Table S1), including 107 the heme oxygenase IsdI, which is known to be upregulated when the T7SS is inactivated 23 .  (Fig S1B). By contrast, a C-terminally Myc-tagged variant of SAOUHSC_00584 was 118 detected only in the cellular fraction ( Fig S1C). To probe the subcellular location of 119 SAOUHSC_00584-Myc, we generated cell wall, membrane and cytoplasmic fractions. Fig 2A   120 shows that the tagged protein localizes to the membrane and that it appears to be destabilized 121 by the loss of EssC. SAOUHSC_00584 was subsequently renamed TspA (Type Seven 122 dependent Protein A).

123
TspA is predicted to be 469 amino acids long and to have either one (TMHMM) or two 124 (Predictprotein.org) transmembrane domains towards its C-terminal end. To determine 125 whether it is an integral membrane protein we washed isolated membranes with urea which 126 removes peripherally bound proteins by denaturation. Fig 2B indicates

that a large fraction of
To determine whether TspA-Myc is exposed at the extracellular side we prepared 133 spheroplasts and treated them with proteinase K. Fig 2C shows that at low concentrations of 134 proteinase K, TspA-Myc was proteolytically cleaved to release a smaller fragment that also 135 cross-reacted with the anti-Myc antibody. At least part of this smaller fragment must be 136 extracellular as it was also degraded as the protease concentration was increased. An 137 approximately 37 kDa C-terminal fragment of TspA-Myc detected natively in the absence of 138 added protease was also extracellular as it was sensitive to digestion by proteinase K. The  EsxA and EsxC are secreted in the absence of TspA. We conclude that TspA is a peripheral 147 membrane protein substrate of the T7SS whose localization and/or stability at the extracellular 148 side of the membrane is dependent on EssC, and that it is not a core component of the T7SS. extracellular side to form a voltage-gated channel 27,28,29

167
To assess TspA CT activity we first used the BacLight assay to test for membrane 168 depolarization. This uses 3,3′-diethyloxacarbocyanine iodide DiOC 2 (3), a dye that exhibits 169 green florescence in dilute solution but a red shift following membrane potential-driven 170 accumulation in the bacterial cytosol. After sorting of E. coli cells by flow cytometry it was seen 171 that the majority of E. coli cells harbouring the empty vector exhibited red fluorescence, which 172 shifted to green following treatment with the uncoupler carbonyl cyanide 3-173 chlorophenylhydrazone (CCCP). A similar shift in fluorescence was also observed when E.

174
coli produced the AmiAss-TspA CT fusion (Fig 3D), indicative of loss of membrane potential.

175
We conclude that the C-terminal domain of TspA has membrane depolarising activity.

176
Pore-forming proteins are widely used as toxins to target either prokaryotic or eukaryotic 177 cells 33, 34 . To determine whether TspA forms pores we used single-cell microscopy that 178 combines the voltage-sensitive dye DiSC 3 (5) with the membrane-impermeable nucleic acid 179 stain Sytox Green 35 . E. coli cells incubated with Polymyxin B, which produces large ion-180 permeable pores in the E. coli cell envelope 36 , showed strong labelling with Sytox Green, 181 indicative of permeabilisation, coupled with very low DiSC 3 fluorescence (Fig 3E-G). By 182 contrast, cells harbouring the empty vector had high DiSC 3 fluorescence that was unaffected as evidenced by the marked reduction in DiSC 3 fluorescence, but did not detectably stain with 186 Sytox Green, even after prolonged periods of incubation ( Fig 3E-G, Fig S3). Therefore, it 187 appears that TspA acts by triggering membrane depolarisation but does so by forming ion 188 channels rather than larger, nonselective pores in the E. coli inner membrane.

189
Bacterially-produced channel-forming toxins are often co-expressed with immunity proteins 190 that protect the producing cell from self-intoxication; for example, protection from colicin Ia 191 toxicity is mediated by the membrane-bound Iia immunity protein 37 . TspA is genetically linked 192 to a repeat region of ten genes encoding predicted polytopic membrane proteins with DUF443

201
To probe whether TspA is important for S. aureus virulence, we initially developed an 202 immunocompetent murine model of S. aureus pneumonia. Previous reports have indicated 203 that 2-4 x 10 8 colony forming units (cfu) of strain Newman was a suitable infectious dose, and 204 that bacterial proliferation in lung tissue could be observed after 24 hours 38 . We found that at 205 a dose of 8 x 10 7 -2 x 10 8 of strain RN6390, the mice were asymptomatic and had almost 206 completely cleared the bacteria from lung tissue after 48 hours, whereas a dose of 8 x 10 8 -207 2 x 10 9 was lethal to all mice within 12 hours. At a dose of 3 x 10 8 , the mice developed 208 symptoms which resolved within 12 hours, and therefore using this dosage we sought to test 209 whether there was a difference in bacterial proliferation and dissemination dependent on the 210 T7SS. However, after 24 hours of infection with 3 x 10 8 cfu of RN6390 or a cognate essC strain, counts recovered from the lungs and livers of mice infected with the essC strain were 212 not significantly different than those from mice infected with the wild type ( Fig S4).

213
Given that there are likely to be roles for the T7SS in bacterial competition as well as direct 214 interaction with the host, we next developed a model where these two potentially confounding 215 factors could be investigated. The zebrafish (Danio rerio), a widely used vertebrate model for 216 development, has recently been adapted to study bacterial infection by human pathogens 39 .

217
The hindbrain ventricle offers a sterile compartment that can be used to follow bacterial x 10 3 cfu) whereas only ~55% survived a higher dose (2 x 10 4 cfu; Fig S5B). Although 28.5°C 223 is the optimum temperature for zebrafish larvae development, S. aureus has a temperature 224 optimum of 30 -37°C for growth. In agreement with this, we observed significantly increased 225 zebrafish mortality at 33°C relative to 28.5°C, at high dose infection ( Fig S5B).

226
We next assessed whether there was a role for the T7SS in zebrafish mortality. For these 227 experiments, larvae at 3 dpf were inoculated in the hindbrain ventricle with 2 x 10 4 cfu of 228 RN6390 or an isogenic strain, RN6390 ess, lacking all 12 genes (esxA through esaG) at the 229 ess locus 3 , and incubated at 33°C. We routinely observed that zebrafish mortality was 230 significantly reduced, at both 24 and 48 hr post inoculation (hpi), for fish infected with the 231 RN6390 ess strain compared to the wild type ( Fig S5C; Fig 4A,C). In agreement with this, 232 total bacterial counts of infected fish revealed that following an initial period of six hours where 233 both strains replicated in a similar manner, there was a significant decrease in recovery of the 234 ess strain compared to the wild type after 9 hours (Fig S5D), suggesting that bacteria lacking part dependent on a functional T7SS (Fig S6B), although we observed no difference in 239 bacterial burden between the wild type and essC strain at the timepoints sampled (FigS6C).

240
We conclude that the T7SS plays a role in virulence of S. aureus in this zebrafish infection 241 model.

242
In addition to TspA, a second T7SS secreted toxin, EsaD (also called EssD 9, 41 ), has been 243 identified in some S. aureus nuclease strains. EsaD was shown to inhibit growth of a 244 competitor S. aureus strain in vitro 6 , but has also been directly implicated in virulence through 245 modulation of cytokine responses and abscess formation 9, 41 . We therefore determined

256
Although TspA is clearly required for full S. aureus virulence in the zebrafish model, the 257 observed toxicity when heterologously produced in E. coli coupled with the presence of 258 immunity genes encoded downstream of tspA strongly suggests that secreted TspA may also 259 have antibacterial activity. Previously, the antibacterial activity of EsaD was demonstrated 260 using the S. aureus COL strain, and required esaD to be overproduced from a multi-copy 261 plasmid in order to observe growth inhibition of sensitive S. aureus strains 6 . We therefore first 262 tested whether overproduction of TspA in COL could inhibit the growth of strain RN6390 that 263 was deleted for tspA along with the 10 genes encoding TsaI homologues. However co-culture experiments in solid or liquid growth conditions failed to demonstrate any TspA-dependent 265 growth inhibition (data not shown).

266
Zebrafish larvae have recently been adapted to study predator-prey interactions 40 , and we 267 reasoned that since the T7SS was active in our zebrafish infection model it may also provide 268 a suitable experimental system to investigate the impact of T7-mediated bacterial competition 269 in vivo. COL as the attacker strain was co-inoculated into the hindbrain ventricle, at a 1:1 ratio, 270 with either RN6390 or an isogenic strain lacking immunity proteins for EsaD and TspA (FRU1;

271
RN6390 saouhsc00268-00278, saouhsc00585-00602). A significant reduction in recovery 272 of the target strain lacking immunity genes was observed compared to the isogenic parental 273 strain at 15 hr post infection ( Fig 5A). Conversely, there was significantly greater fish mortality 274 at 24 hr after co-inoculation of COL with the wild type RN6390 than the immunity mutant strain 275 ( Fig 5B). Since COL is almost completely avirulent at this time-point ( Fig S5) we infer that the 276 mortality arises from RN6390, and as the wild type strain survives better when co-inoculated 277 with COL than the immunity deletion strain co-inoculated with COL, this accounts for the 278 greater fish mortality.

279
To confirm that the reduced growth of the RN6390 immunity mutant strain was dependent 280 upon a functional T7SS in the attacking strain, we repeated the co-inoculation experiments 281 using a T7 mutant strain of COL (COLessC). The RN6390 immunity mutant strain showed 282 significantly higher recovery after 15 hours in the presence of the COL T7 mutant strain than

288
Finally, we investigated which of the EsaD and TspA toxins was responsible for inter-strain 289 competition by using variants of COL deleted for either tspA or esaD as the attacking strain. It was seen that in the absence of either TspA ( Fig 5F) or EsaD (Fig 5H) there was a significant 291 increase in recovery of the RN6390 saouhsc00268-00278, saouhsc00585-00602 prey 292 strain, indicating that each of these toxins has activity against the target strain. However, there 293 was a more pronounced increase in zebrafish mortality when the attacker strain lacked esaD 294 than tspA (compare Figs 6G and I), suggesting that EsaD has the more potent antibacterial 295 activity in these conditions.

298
Here we have taken an unbiased approach to discover substrates of the T7SS in S. aureus

314
TspA is a highly unusual toxin because it also has antibacterial activity, leading to growth a genetically-linked immunity protein, TsaI. It is interesting to note that the C-terminal domain 317 of TspA is polymorphic ( Fig S7) and that clusters of tsaI homologues are encoded at the tspA 318 locus, indicative of intraspecies competition with strains accumulating multiple tsaI-encoding 319 genes as a likely protective mechanism from TspA sequence variants.

325
Bacterial strains, plasmids and growth conditions. All strains and plasmids used in this 326 study are given in Tables S2 and S3. S. aureus strain RN6390 50 and its essC 3 , esaD 6 , 327 SAOUHSC_00268-00278 6 and ess (esxA-esaG) 3 derivatives along with strain COL 51 328 have been described previously. An in-frame deletion of tspA (SAOUHSC_00584) in RN6390 329 was constructed by allelic exchange using plasmid pIMAY (Table S3) Table S4 and were cloned into pIMAY and introduced according to 53 using plasmids pTH100 and pRN111, respectively. E. coli strain JM110 was 341 used for cloning purposes and MG1655 54 and its isogenic tatABCD derivative SG3000 55 was 342 used for toxicity assays.

343
All oligonucleotides used in this study are listed in Table S4, and RN6390 chromosomal DNA 344 was used as template unless otherwise stated. Plasmid pRAB11-tspA-myc encodes TspA with 345 a C-terminal Myc tag in pRAB11 56 and was constructed following amplification with primers 346 tspA cmyc fw and tspA cmyc rv. Plasmid pRAB11-02448-ha produces SAOUHSC_02448 with 347 a C-terminal HA tag from vector pRAB11 and the encoding gene was amplified using primers 348 02448 cha fw and 02448 cha rv. Plasmid pRAB11-0389-ha codes for SAOUHSC_0389 with a C-terminal Myc tag in pRAB11 and was constructed following amplification with primers 352 00406 cmyc fw and 00406 cmyc rv. In each case the amplified gene is preceded by the esxA 353 RBS (AGGAGGTTTCTAGTT), and were cloned as KpnI -SacI fragments. Plasmid pRMC2-354 ssaA-ha codes for SsaA with a C-terminal HA tag. It was constructed following amplification 355 of SsaA using primers pRMC2-ssaA-bglII-for and pRMC2-SSaA-HA-rev-EcoRI, digestion with 356 BglII and EcoRI and cloning into similarly cut pRMC2 57 .

357
Plasmid pBAD18-tspA codes for the full length TspA. The encoding gene was amplified using

478
To assess changes in membrane potential and permeabilisation, the same E. coli strains were 479 grown as described above. An 'uninduced' sample of E. coli harbouring each of pBAD18-Cm 480 (empty), pAmiAss-TspA CT and pAmiAss-TspA CT -TsaI was collected and adjusted to OD 600 of respectively. The images were captured using Metamorph 7.7 (Molecular Devices) and 491 analysed using ImageJ. For the analysis, the phase contrast images acquired in parallel to the 492 fluorescence images were used to identify cells as regions of interest, for which average 493 DiSC 3 (5) and Sytox Green fluorescence intensity was measured from the corresponding        hpi. Mean ± SEM also shown (horizontal bars). Significance testing performed by unpaired t 622 test. **p<0.01, *** p<0.001, **** p<0.0001, ns, not significant.      with mCherry. Zebrafish were injected at 3 dpf with a low (~7 x 10 3 cfu), medium (~1.5 x 10 4 678 cfu) or high dose (~2 x 10 4 cfu) of COL-mCherry, incubated at 33°C and monitored for 48 hpi.

684
Results are plotted as Kaplan-Meier survival curves and the p value between conditions was 685 determined by the log-rank Mantel Cox test. C. Enumeration of recovered bacteria at 0, 6, 24 686 or 48 hpi from zebrafish embryos infected with COL-mCherry (WT) or COL ΔessC-mCherry.

687
Circles represent individual larvae and data pooled data from 3 independent experiments.