Suppressor Mutations in Type II Secretion Mutants of Vibrio cholerae: Inactivation of the VesC Protease

Genome-wide transposon mutagenesis has identified the genes encoding the T2SS in Vibrio cholerae as essential for viability, but the reason for this is unclear. Mutants with deletions or insertions in these genes can be isolated, suggesting that they have acquired secondary mutations that suppress their growth defect.

Proteins secreted by the T2SS are produced with N-terminal signal peptides that direct them to the periplasm, where they fold and connect with the T2SS for outer membrane translocation. In the absence of a functional T2SS, these proteins accumulate in the periplasmic compartment. Inactivation of the T2SS in V. cholerae also results in reduced growth rate in rich media, cell envelope perturbations, increased sensitivity to bile and polymyxin B, and induction of the stress response regulator RpoE (9,(19)(20)(21). Growth defects have also been reported for pilD mutants of V. cholerae, which lack the prepilin peptidase shared by the T2SS and one of the type IV pilus systems (11). Additional reports of analogous phenotypes among T2SS mutants have been observed in Vibrio vulnificus, Vibrio sp. strain 60, and Aeromonas hydrophila (22)(23)(24)(25).
Other studies have suggested that the V. cholerae T2SS genes are essential (26)(27)(28)(29). For example, using a transposon-based approach to identify genes required for growth of the V. cholerae strain N16961, Judson and Mekalanos categorized epsD and epsG as essential (26). The genes epsD through epsG, epsI, epsK, and epsL and pilD were also identified as putatively essential genes by Cameron et al. and Kamp et al., as transposon insertions in these genes were not identified during genome-saturating transposon screens, presumably because bacteria containing transposon insertions in these genes could not be recovered (27,29). However, in a study that categorized V. cholerae genes as essential, domain essential (containing both essential and nonessential coding regions), or sick, transposon insertions in some of the eps genes that had been reported as essential by others resulted in viable but sick mutants (28). Collectively, these studies suggest that secondary mutations arise to suppress a potentially lethal phenotype associated with loss of function of the T2SS. To test this hypothesis, we subjected six T2SS mutants to high-throughput genome sequencing and identified secondary mutations in all mutants. The finding that two of the mutants had acquired distinct mutations in the same gene, vesC, prompted us to further interrogate this gene in 92 additional eps mutants by PCR amplification and Sanger sequencing. This process identified another 19 eps mutants with unique vesC mutations, suggesting a selective pressure to lessen the stress and potential lethal phenotype induced by T2SS mutations in V. cholerae.

RESULTS
Inactivation of the type II secretion system in V. cholerae reduces growth rates. We have previously reported that a V. cholerae mutant lacking all eps genes exhibits a growth rate reduction compared to the isogenic wild-type (WT) strain, suggesting that loss of the T2SS results in a slower growth phenotype (19). Strains containing inactivating mutations in single eps genes also grow slower. For example, the DepsG1, DepsL, and DepsM mutants that are lacking the genes encoding EpsG, EpsL, or EpsM, which directly interact within the T2SS complex (30)(31)(32)(33), show reduced growth rates compared to T2SS-competent WT isolates (Fig. 1A). The colony size of the T2SS mutants is also consistently smaller than that of WT strains. As reported previously, inactivation of these eps genes abolishes secretion as measured here by the loss of extracellular serine protease activity (Fig. 1B) (9,13,19,34,35).
Identification of secondary mutations in V. cholerae T2SS mutants. We hypothesized that introducing eps mutations selects for secondary mutations that suppress a potentially lethal phenotype associated with loss of secretion by the T2SS. Thus, we sought to identify secondary mutations among V. cholerae T2SS mutants using high-throughput genome sequencing. We chose to focus on mutants with deletions of epsG, epsL, and epsM, as these genes have been reported to be essential by some and nonessential or generating a sick phenotype when mutated by us and others. Using Illumina Hi-Seq technology, we sequenced the genomes of the V. cholerae El Tor strain TRH7000, a ctxAB::Hg R derivative of N16961, and the isogenic DepsG1, DepsL, and DepsM mutants. To identify genetic differences (including single-nucleotide polymorphisms [SNPs] and structural variants [SVs]) between the T2SS mutants and WT V. cholerae, we used referenceguided alignment using SeqMan software Lasergene (DNASTAR, Madison, WI) with the sequenced strain N16961 as a template. We then subtracted any SNPs or SVs found between TRH7000 and N16961. This list was then used to establish a list of SNPs or SVs unique to each of the sequenced T2SS mutants. Two secondary mutations were identified in the DepsG1 and DepsL mutants, while five mutations were identified in the DepsM mutant, as tabulated in T1-3  Tables 1, 2, and 3. Interestingly, the DepsG1 and DepsL mutants have acquired distinct mutations in the vesC gene (VC1649), which encodes a T2SS-secreted protease (13,36). The DepsG1 mutant contains a 7-bp insertion (frameshift mutation) at position 1467 of vesC, resulting in a premature stop codon at amino acid (aa) 492 of this 548-aa protein (491fs). The DepsL mutant harbors a point mutation altering residue 279 from a glutamine to a proline (Q279P) in VesC. In both DepsG1 and DepsL mutants, one additional gene contains a mutation besides vesC, and these are unique between the two strains. Specifically, the DepsG1 mutant contains a mutation in rfbV (VC0259), a gene required for lipopolysaccharide biogenesis (37), and the DepsL strain contains a mutation in a putative secreted glycoside hydrolase gene, VCA0254. The DepsM mutant contains secondary mutations in five genes that encode proteins annotated as membrane proteins, a metabolic enzyme, and a ribosomal protein (Tables 1, 2, and 3). While the rfbV mutation in the DepsG1 strain was confirmed by PCR amplification and Sanger sequencing of the genomic DNA that was subjected to whole-genome sequencing, it was absent in the DNA isolated from three new cultures of the original DepsG1 freezer stock. That only 72% of the reads (Table 1) carried this mutation suggests that the mutation occurred during the growth of the culture used for whole-genome sequencing. The mutation in vesC was confirmed in the original freezer stock, as were the two mutations in the DepsL mutant. To further demonstrate that a variety of secondary mutations arise when eps genes are inactivated, we subjected three additional independently isolated mutants to whole-genome sequencing. These mutants included a second epsG mutant, here called DepsG2, and two mutants, called PU3 and PU5, with transposons at two different positions in epsM (8). These mutants also carry secondary mutations (Tables 1, 2, and 3), but they differ from those of the DepsG1, DepsL, and DepsM mutants.
While the majority of secondary mutations in the mutants with deletions or insertions in epsG, epsL, or epsM may be directly or indirectly linked to the cell envelope, here we chose to follow up with the DepsG1 and DepsL strains for further characterization, because they contained different mutations in the same gene, vesC, indicating a "c." denotes nucleotide change at the indicated position; "p." denotes amino acid change at the indicated position. one possible conserved mechanism for suppressor mutations to reverse a potential lethal phenotype of V. cholerae T2SS mutants. Secondary mutations in vesC abolish protease activity. To determine whether the secondary mutations in vesC harbored by DepsG1 and DepsL mutants inactivate VesC, we cloned and overexpressed these mutant genes in a DvesABC mutant that lacks the genes for the VesA, VesB, and VesC proteases. The protease activity in the culture supernatants was determined and compared to that of supernatants from DvesABCexpressing genes for either WT VesC or an inactive variant, VesC-S225A, that has the catalytic serine in the active site replaced with alanine. Neither VesC-Q279P nor VesC-491fs restored extracellular protease activity in the DvesABC mutant, while protease activity was detected in the presence of WT VesC, indicating that these modifications in VesC abolish its activity ( Fig. 2A). Culture supernatants of the same strains were also analyzed by SDS-PAGE and silver staining (Fig. 2B). While overexpression of WT VesC and the catalytically inactive VesC-S225A resulted in visible proteins of approximately 55 kDa (lanes 3 and 6), neither VesC-Q279P nor VesC-491fs variants were discernible above background, suggesting that they are unstable and/or not secreted (lanes 4 and 5).  Additional mutations in vesC. The possibility that other V. cholerae T2SS mutants also harbor mutations in the vesC gene was investigated next. Genomic DNA from 92 additional mutants with eps gene deletions/modifications and growth defects to various degrees were isolated, and the vesC gene from each mutant was amplified by PCR and subjected to Sanger sequencing. We identified 19 additional unique mutations in vesC in mutants with deletions of epsC, epsD, epsE, epsF, epsG, and epsL (Tables 4  and 5). These included frameshift mutations due to insertions or deletions, sequence duplications, and a variety of single-nucleotide changes resulting in amino acid substitutions in the VesC protein.
Structure of VesC. To explain the effect of the VesC alterations, we obtained the three-dimensional structure of VesC. Optimized crystals that diffracted to 2.2-Å resolution were produced from the construct pro-VesC containing residues 23 to 522, which includes the propeptide but not the N-terminal signal peptide or the C-terminal GlyGly-CTERM motif (see Fig. S1 in the supplemental material). The structure of VesC was determined by the molecular replacement method assisted by Rosetta homology modeling with electron density using VesB protease (PDB entry 4LK4) and carbohydrate binding module (CBM) domains (PDB entries 1UXX and 2C9A) as search models (38). The final structure was refined to an R-factor of 0.201 and an R-free of 0.234 with good geometry ( Table 6).
The overall structure of VesC revealed that it consists of three domains: an N-terminal domain with a protease fold, the middle domain with an Ig-like domain, and the Cterminal domain with a b-sandwich fold (Fig. 3A). Well-defined electron density maps allowed the fitting of all three domains, the protease fold (35 to 272), the Ig-like fold (273 to 380), and the C-terminal domain (381 to 511), with the exception of the N-terminal residues (23-34), including the propeptide and C-terminal tail residues (512 to 522), due to their flexibility in the crystal structure. Three disulfide bonds are found in the protease domain (C60-C76 and C190-C212) and Ig-like domain (C330-C340), which are also conserved in VesB (PDB entry 4LK4).
The structural comparison of the protease domain of VesC (VesC PD ) with typical trypsins showed a high degree of structural similarity, with the best hit found in VesB PD . The structure of VesC PD can be superimposed onto the VesB PD structure (PDB entry 4LK4) with a root mean square deviation (RMSD) of 1.8 Å with 42% sequence identity for 205 residues (Fig. 3B). Consistent with the structures of VesB PD and other trypsinogens, VesC PD displays a similar positioning of its catalytic residues, and the three disordered loop regions (residues 164 to 172, 216 to 220, and 247 to 249) in the position of the active site are also observed in the crystal structure of VesC PD . In contrast, two loop regions, residues 115 to 121 and 194 to 208, in VesC PD adopt substantially different conformations, which extend upwards from the globular body of the protease domain. The Ig domain of VesC (VesC Ig ) is 22% identical and 55% similar in amino acid sequence to VesB Ig . In spite of this moderate sequence homology, the structures of VesC Ig and VesB Ig are very similar ( Fig. S1 and S3B). Additionally, the two-domain structures of VesC PD1Ig and VesB PD1Ig are relatively similar, with an RMSD of 2.3 Å for 301 equivalent Ca atoms. In the crystal structure, the interface between VesC PD and VesC Ig buries a 1,568-Å 2 solvent-accessible surface, which is ;28% larger than the interface between the homologous domains in VesB. The structure of the third C-terminal domain of VesC, which is missing in VesB, revealed a concave b-sandwich domain (Fig. 3A). A homology structure search using the DALI server showed that the human proteins meprin A subunit beta, receptor-type tyrosine protein phosphatase MU, reelin, and the CBM family 29 of NCP-1 from Piromyces equi are the closest structural homologs of VesC CBM (Table 7). Interestingly, the closest structural homologs found by secondary-structure matching (SSM) PDBeFold include distant homologs with CBMs (PDB entries 1OH3, 2ZEX, 2ZEZ, and 2XOM). Comparing VesC CBM with CBM family 29 (PDB entry 1OH3) revealed that the two structures share 73% of their secondary-structure elements. Although VesC CBM has a low level of sequence similarity to the previously determined crystal structures of CBMs, the superposition shows the conserved core b-sandwich composed of 10 b strands (Fig. 3C). These CBM families all adopt the b-sandwich scaffold, displaying a curved platform with diverse carbohydrate binding modes in the variable loops and the concave face and residues with aromatic side chains (39). The high degree of structural similarity between VesC CBM and CBM family proteins suggests that VesC binds to carbohydrates and/or target glycosylated substrates.
Structural analysis of VesC substitutions. The VesC structure offers the opportunity to explain the effects of the mutations, which were scattered throughout vesC, on the folding, stability, and/or function of this T2SS substrate. The S63R substitution identified in two independent eps mutants is located on the N-terminal lobe of VesC PD . The hydroxyl of S63 forms a buried hydrogen bond with Q239 (Fig. 4A). To evaluate the possibility of acceptable substitutions at this position, we performed sequence tolerance analysis using Rosetta backrub (40). Serine was the preferable residue tolerated at this position (Fig. S2A). The loss of the buried hydrogen bond and the bulky arginine sidechain resulting from the S63R substitution likely lead to destabilization of VesC PD . Similarly, duplication of residues 63 to 67 and deletion of 64 to 68 in this region could result in misfolding of VesC PD . The G159V substitution is located in the C-terminal lobe of VesC PD . The backbone conformation of G159, w 157.9°and c 177.9°, is favorable for glycine but not other residues. The sequence tolerance analysis showed that only glycine is acceptable at this position (Fig. S2B). The R163H variant is located downstream on the same loop as G159V. The guanidinium group of R163 makes a hydrogen bond with the carbonyl of F59 and is in van der Waals contact with the disulfide C60-C76 (Fig. 4B). The arginine residue is strongly preferred at this position based on sequence tolerance analysis (Fig. S2C).
The substitutions Y277H and Q279P are located in the VesC Ig domain, with the Q279 side chain facing the VesC PD domain (Fig. 4C). The side chain of Y277 is surrounded by hydrophobic residues, while the hydroxyl moiety makes hydrogen bonds with water molecules bridging Q279 and Y365 (Fig. 4C). The sequence tolerance analysis indicated that tyrosine and phenylalanine residues are preferred at this position (Fig. S2D). Therefore, the introduction of the polar histidine side chain could result in misfolding of the VesC Ig domain. The Q279 residue is present in the middle of the interface between VesC Ig and VesC PD (Fig. 4C and Fig. S3). Ninety-one percent (71 Å 2 out of 78 Å 2 ) of accessible surface area of Q279 is buried in the VesC Ig -VesC PD interface that includes a hydrogen bond with the main chain of V110 (Fig. 4C). The sequence tolerance analysis showed that glutamic acid and glutamine residues are preferred at position 279 (Fig. S2E). Since the side chain of proline is markedly different from that of glutamine, the Q279P alteration likely modifies the domain interaction and prevents proper folding of VesC. Structural analysis indicates that the analogous glutamine residue in VesB similarly forms part of an interface with residues in the protease domain (Fig. S3). Previous attempts to produce extracellular VesB without its Ig-fold domain in V. cholerae have been unsuccessful, supporting the suggestion that this domain plays an important role in protein stabilization and/or secretion of VesB and possibly also VesC (41).
The probable deleterious effect of leucine residue insertion after K357 in the VesC Ig domain could be explained by an effect on adjacent positions, because K357 is  (Fig. 4D). Therefore, this would effectively lead to an I359L substitution and the extension of the downstream loop by one residue. Both of these changes would likely decrease the stability of VesC Ig , because a small hydrophobic residue, valine or isoleucine, is preferred at position 359 (Fig. S2F), and all residues of the downstream loop are engaged in specific contacts (Fig. 4D). Substitutions P431L and L435P are located in the VesC CBM domain. Residue P431 is engaged in van der Waals contacts with P394 and Y505 (Fig. 4E), and sequence tolerance analysis indicated that a small residue is preferred at this position (Fig. S2G). Residue L435 is oriented toward the hydrophobic core of VesC CBM , and substitution with a smaller proline moiety may destabilize this domain. Additionally, w /c angles of L435 would be less preferable for a P residue. Sequence tolerance analysis showed a strong preference for leucine at this position (Fig. S2H). Furthermore, the S491 frameshift in VesC resulting in a premature stop at residue 492 is positioned in the middle of the b-sandwich fold of the VesC CBM structure, removing the C-terminal 56 residues, including a structurally important b-strand and the GlyGly-CTERM domain (Fig. 4F). Finally, a large D403-456 deletion would effectively eliminate VesC CBM . In summary, all observed alterations in VesC would negatively impact its stability and/or folding, preventing the production of an active VesC enzyme.
Overexpression of WT VesC affects growth of the DepsG1 mutant. To determine whether expression of vesC has a negative impact on cells that are deficient in extracellular secretion, we overexpressed WT vesC in the DepsG1 mutant. Following growth overnight, cultures were diluted in fresh media, split in half, and grown in the presence or absence of 50 mM isopropyl-D-thiogalactopyranoside (IPTG). All cultures grew without IPTG, whereas three of six cultures did not grow in the presence of IPTG (Fig. 5). In control experiments, all WT V. cholerae TRH7000 cultures grew as well with as without IPTG. Finally, no growth inhibition was observed when the proteolytically deficient VesC-S225A was overexpressed in the DepsG1 mutant, suggesting that VesC activity is the cause of toxicity when its secretion is impeded.

DISCUSSION
This study was initiated to determine why T2SS mutants of V. cholerae are viable even though several of the genes encoding the T2SS have been deemed essential (26)(27)(28)(29). We hypothesized that V. cholerae T2SS mutants acquire secondary mutations that suppress a potential lethal phenotype. Indeed, we confirmed the presence of secondary mutations in each of the six sequenced V. cholerae T2SS mutants. Two of the six eps mutants acquired distinct mutations in the same gene, VC1649, which encodes the T2SS-secreted serine protease VesC. This and the finding that additional unique vesC mutations are present in 19 out of 92 additional T2SS mutants suggests that one method by which eps mutations can be generated in V. cholerae is to inactivate one of its secreted substrates, which may otherwise cause damage when accumulating in the periplasm. Mutations in the vesC gene are not sufficient, however, to restore growth to WT levels. This was exemplified by the finding that cumulative mutations can appear during further growth of mutants such as the DepsG1 mutant, which picked up a mutation in a second gene, rfbV, when cultured for the isolation of genomic DNA used for whole-genome sequencing (Table 1). We identified vesC mutations in mutants that are functionally lacking six of the 13 T2SS components (EpsC through EpsG and EpsL). The reason we have not detected vesC mutations in strains deficient in EpsH through EpsK, EpsM, and PilD may not be related to their specific function in the T2SS, as they are also required for secretion, and it is likely due to the limited number of mutants that we have generated. Although alterations were identified in all three domains of VesC, a hot spot of changes in the protease domain was observed, with one identical substitution of residue 63 in two independently isolated mutants, a deletion of residues 64 to 68, and a duplication of 63 to 67 (Tables 4 and 5). Through careful analysis of the VesC structure, we determined that all vesC mutations presented in this study would very likely eliminate the formation of an active enzyme.
Over 20 proteins are secreted by the T2SS, yet we identified secondary mutations in the same gene, vesC, in 21 different eps mutants. Perhaps whole-genome sequencing of additional eps mutants besides the six mutants sequenced here will reveal secondary mutations in genes encoding other T2SS substrates, but the observation of multiple distinct mutations in the vesC gene is indicative of a conserved mechanism for putative eps mutant suppression. The finding that three out of six cultures of the DepsG1 mutant did not grow when WT vesC expression was induced from a plasmid with IPTG, while overexpression of vesC in the WT strain did not diminish growth, suggests that periplasmic accumulation of VesC is harmful and that VesC-inactivating alterations in T2SS mutants relieve this toxicity (Fig. 6). No growth inhibition was observed when the proteolytically deficient VesC-S225A was overexpressed in the DepsG1 mutant, suggesting that VesC activity and not amount is the cause of toxicity. The reason why WT vesC expression did not interfere with growth of all DepsG1 mutant cultures may be due to the appearance of additional suppressor mutations during culturing of the mutant containing the pVesC plasmid. Although we have only completed the sequencing of six T2SS secretion mutant genomes, we do not believe that the V. cholerae T2SS mutants carry inactivating mutations in the VesC-homolog VesB, as complementation of defective eps genes generally restores serine protease activity to 80 to 100% of WT activity (as exemplified in Fig. 1B), and VesB contributes to approximately 80% of secreted serine protease activity toward the fluorogenic Gln-Ala-Arg peptide (13). Thus, VesB and VesC both are capable of cleaving this short peptide; however, they likely differ in protein substrate specificity, as their protease domains are only 42% identical and VesC contains an additional domain. An alternative explanation for why VesC and not VesB (or VesA) is toxic when not secreted may be due to its higher level of production, as we have previously found that 3.5 times more VesC than VesB and VesA is detected in the culture supernatant of WT V. cholerae (13).
Mutations in genes that encode proteins essential for cell viability may have been selected to prevent their possible proteolysis by accumulating VesC in the periplasm of T2SS mutants. For example, BamA, an essential component of the beta-barrel complex required for outer membrane protein biogenesis, has an I501T substitution in the DepsG2 mutant. Sequence-based structural modeling of V. cholerae BamA on E. coli BamA suggests that this residue is located in one of the cell surface-exposed loops. PU5, with a transposon in epsM, carries a 124-nt frameshift insertion in VC2506, which encodes the RNA polymerase-associated protein HepA. It is an ATP-dependent helicase that contributes to recycling of RNA polymerase during stress in E. coli (42). The effect of this mutation is not understood, but it may result in modification of gene expression to accommodate cellular stress in PU5. The other transposon mutant, PU3, has a 60-nt insertion in VC2701, which encodes a homolog of the E. coli DsbD. This results in a 20amino-acid insertion in the first periplasmic domain of this inner membrane protein that contributes electrons to disulfide isomerases such as DsbC (43) and, thus, has an indirect function in disulfide isomerization of proteins in the periplasm. The second mutation is a synonymous mutation (A to G) at codon 927 that encodes a Ser in HisD, predicted to be involved in histidine metabolism. The frequency of the two Ser codons TCA and TCG in E. coli is the same, so it is not understood if the mutation has any effect on the synthesis of HisD in PU5, unless codon usage is different in V. cholerae. With five mutations, the DepsM mutant has the most secondary mutations of the sequenced strains. VC0613 encodes a predicted beta-N-acetylhexosaminidase and is part of a ChiS-regulated chitin-induced catabolic operon that contributes to amino sugar and nucleotide sugar metabolism in V. cholerae (44). It is possible that when V. cholerae is unable to secrete its two extracellular chitinases, there is no need for intracellular enzymes that participate in the downstream processing of chitin. In addition, the DepsM mutant has secondary mutations in three genes coding for inner membrane proteins. VC1718 encodes an ElyC homolog that is predicted to span the membrane twice. In E. coli, this protein is involved in cell wall precursor metabolism. The absence of ElyC results in defects in peptidoglycan synthesis at low temperatures (45). Residue Val190, which is replaced with a Met residue in the DepsM mutant, is predicted to be localized in the periplasmic domain. Another mutation is in VC0286, which encodes a gluconate permease. This protein is predicted to span the inner membrane 11 to 12 times and contributes to carbon metabolism FIG 6 Working model of the mechanism by which secondary mutations in vesC may suppress a potential lethal phenotype of T2SS mutants. (Left) Wild-type V. cholerae transports VesC across the outer membrane via the T2SS. Upon inactivation of the T2SS, VesC secretion is blocked and the protease accumulates in the periplasm, where it may be a contributing factor to cell envelope damage through nonspecific proteolysis and a possible lethal phenotype (right). During the process of genetic inactivation of the T2SS, we may select for mutations that inactivate VesC (A) and/or target VesC for degradation (B) to prevent proteolysis of essential components and/or avert irreparable cell envelope damage in the absence of a functional T2SS. (46). The mutation results in a substitution of Ile for Phe at position 100. The third mutation affecting an inner membrane protein is found in VC2323, a membrane protein that spans the membrane eight times and is homologous to the E. coli tellurite resistance protein. In V. cholerae, this protein may serve as an efflux pump that contributes to chloramphenicol resistance and intestinal colonization in infant mice (47). Finally, the fifth mutation is a 33nucleotide insertion in VC1915, which encodes the 30S ribosomal protein S1. The 11amino-acid insertion likely does not inactivate this protein, as it is essential for protein synthesis; however, it may alter the rate of protein synthesis to accommodate the stress associated with inactivation of the T2SS. The 73 eps mutants that lack mutations in vesC and have not been subjected to whole-genome sequencing may have similar mutations in genes encoding cell envelope components that are sensitive to periplasmic VesC. Other possibilities include mutations in genes that code for factors controlling expression, folding, and/or activation of VesC.
Finally, other genes with suppressor mutations may encode yet-to-be-identified T2SS substrates that could cause damage to components of the cell envelope, including the inner and outer membranes and peptidoglycan, when accumulating in the periplasm of T2SS mutants. For example, a mutation was detected in VCA0254 in addition to the vesC mutation in the DepsL mutant. This gene encodes a protein that is 21% identical to and can be modeled on 3,6-anhydro-D-galactosidase, an exolytic enzyme produced by a marine G-negative species, Zobellia galactanivorans, that targets carrageenan oligosaccharides (PDB entry 5OPQ [48]). Based on the extracellular location of the Z. galactanivorans enzyme and that both proteins are expressed with Nterminal signal peptides, we speculate that VCA0254 encodes an extracellular protein possibly secreted by the T2SS in V. cholerae. However, cloning and overexpression of VCA0254 in WT V. cholerae did not result in a detectable protein in the culture supernatant when analyzed by SDS-PAGE and silver staining (not shown).
Although many bacteria use the T2SS to support secretion of proteases, VesC is one of three unique trypsin-like serine proteases found in V. cholerae, other Vibrio species, and related marine species, including Aeromonas hydrophila, and may have a nonspecific and perhaps toxic activity inside the cell compared to other T2SS-secreted proteases. Therefore, we speculate that perhaps the T2SS genes are not essential per se but rather that the phenotype observed with T2SS mutants of V. cholerae, V. vulnificus, Vibrio sp. strain 60, and A. hydrophila is due to damage caused by particular T2SS substrates when they accumulate in the wrong location. Additional investigation into the relationship between organisms exhibiting T2SS inactivation-associated envelope defects and their corresponding suites of secreted substrates may reveal a conserved mechanism for suppression of T2SS-associated phenotypes in these organisms.

MATERIALS AND METHODS
Bacterial strains and growth conditions. Vibrio cholerae N16961 (El Tor), TRH7000 (thy Hg r [ctxA-ctxB]), and mutants thereof were grown at 37°C in LB broth, which was supplemented with 100 mg/ml thymine. Plasmid-containing strains were grown in the presence of 200 mg/ml carbenicillin, and gene expression was induced with 10 mM isopropyl-D-thiogalactopyranoside (IPTG) for epsG, epsL, and epsM or 50 mM IPTG for vesC expression.
Cloning. With the exception of the DepsM mutant, the mutants used in this study for the whole-genome sequencing were constructed previously (8,19,49). The DepsG2 mutant was constructed in the same way as the DepsG1 mutant (49). Mutants that were used for PCR amplification and sequencing of the vesC gene were also made previously (8,17,20,30,49,50).
The DepsM mutant was constructed by amplifying regions upstream and downstream of the epsM gene and introducing an internal kanamycin resistance cassette from pKD4 using the following primers: 59 CAAGTCTTCTTGGCTGCGGT 39 (forward [Fwd] primer for upstream fragment), 59 CGAAGCAGCTCCAG CCTACACTTCTCCTTACTTGGGCTTCACC 39 (reverse [Rev] primer for upstream fragment), 59 CTAAGGAG GATATTCATATGGCGTGGAGGCTGATATGA 39 (forward primer for downstream fragment), and 59 CCGAC ACGACAGTACCAAGCTGC 39 (reverse primer for downstream fragment). PCR products from the upstream and downstream regions were used as a template for another PCR using the first and last primers, which was used for chromosomal replacement as described previously (51).
Plasmids containing either WT or T2SS mutant variants of genes identified from whole-genome sequencing were constructed by amplifying the gene of interest from TRH7000 chromosomes and cloning into pMMB67EH. The primers used to amplify vesC (VC1649) are 59 GAGGAGCTCTGGGAGTTATCAGAGGTATC 39 (Fwd) and 59 GAGGCATGCTGGCTATCGATAGATCAGAC 39 (Rev). pMMB-VesC S225A was constructed from pMMB-VesC WT using PCR mutagenesis, with overlapping primers containing the point mutation 59 CGCT TGTTCTGGTGACGCCGGTGGCCCTATCTTTTTTG 39 (Fwd) and 59 CAAAAAAGATAGGGCCACCGGCGTCACCAGA ACAAGCG 39 (Rev). To introduce the S225A substitution, the VesC Fwd primer and VesC S225A Rev primer and the VesC S225A Fwd primer and the VesC Rev primer were used to amplify each half of the vesC gene and introduce the mutation, and these products were then used as the template for a third PCR with the VesC Fwd and Rev primers. All cloning was confirmed using Sanger sequencing of PCR products and plasmids.
Analysis of growth. Comparisons of growth rates of the WT TRH7000 and DepsG1, DepsL, and DepsM mutant strains were performed using a Bioscreen growth curve analyzer (Growth Curves USA). Overnight stationary-phase cultures of V. cholerae were back-diluted as described in the figure legends and inoculated into microtiter Bioscreen plates in duplicate wells per sample. The optical density (A 600 ) was measured at 15-min intervals for 20 h. Experiments were repeated in triplicate, and means are displayed. Growth rate analysis of WT TRH7000 containing the pVesC plasmid and the DepsG1 mutant containing either pVesC or pVesC-S225A was done manually in larger flasks with good aeration, starting with overnight stationary-phase cultures that were back-diluted and then split into two cultures, where one received IPTG to a final concentration of 50 mM. The optical density (A 600 ) was measured at 30-min intervals for 5 h.
Genome sequencing and analysis. Genomic DNA was isolated from V. cholerae using Wizard genomic DNA purification kits (Promega). Genomic DNA library preparation and sequencing were performed by the University of Michigan DNA Sequencing Core using Illumina HiSeq 2000. Paired-end libraries were constructed, and sequencing was performed with a read length of 100 by 100. Analysis was conducted using SeqMan software (Lasergene) for SNP and structural variant calling. Using SeqMan NGen, the TRH7000 sequence was aligned to the N16961 published reference sequence to serve as a template for analysis of T2SS mutant sequences. Variants were called using SeqMan Pro software (Lasergene), and visualization and coverage analysis were performed simultaneously. Genome sequences of T2SS mutants were compared to that of N16961 using SeqMan NGen reference-guided alignment. Variant calls that were found when comparing TRH7000 to N16961 were subtracted from the T2SS mutant calls.
Protease secretion assay. Extracellular protease activity was measured and quantified as described previously (19). Briefly, culture supernatants were separated from cells, and the fluorogenic probe, Ntert-butoxy-carbonyl-Gln-Ala-Arg-7-amido-4-methylcoumarin (Sigma-Aldrich), was added to the supernatants. Over the course of 10 min, protease activity was measured every minute using fluorescence at excitation and emission wavelengths of 385 and 440 nm, respectively. Assays were performed at least in triplicate, and values were normalized to the density of the culture (A 600 ). Means and standard deviations (SD) are displayed.
VesC expression and purification for crystallization. The gene encoding residues 23 to 522 of V. cholerae VesC was cloned into a modified pRSF-Duet1 vector (Novagen) for expression with a tobacco etch virus (TEV) protease recognition site prior to the C-terminal His 6 tag. The wild-type VesC expression was toxic for E. coli; therefore, the catalytic site residue Ser225 was replaced with Ala using the QuikChange mutagenesis protocol (Stratagene) to overcome difficulties of overexpressing VesC. To increase the expression level and solubility of the VesC (S225A) protein, a maltose-binding protein (MBP) tag followed by a TEV protease cleavage site was fused to the N terminus of VesC. The resultant plasmid containing the gene for the MBP-TEV site-VesC (S225A)-TEV site-hexahistidine tag was transformed into E. coli Rosetta2(DE3) cells. Transformed cells were grown to an A 600 of ;0.6 at 37°C in Luria broth and induced with 0.5 mM IPTG at 18°C for 4 h. The cells were harvested by centrifugation and resuspended in buffer containing 20 mM Tris-HCl, pH 8.0, 0.5 M NaCl, 2 mM FeSO 4 , 1 mM phenylmethylsulfonyl fluoride, 2 mM benzamidine, and 15 mM imidazole. Cells were lysed by sonication on ice. Soluble proteins were separated by centrifugation (30 min, 60,000 Â g, 4°C) from the cell pellet. The protein in the supernatant was purified on a nickel-nitrilotriacetic acid (Ni-NTA) column using 150 mM imidazole, treated with TEV protease, and purified further by a second Ni-NTA affinity step, followed by anion exchange chromatography using a Source 30Q column. VesC was concentrated to 3 mg/ml and flash-frozen in liquid nitrogen.
VesC crystallization, data collection, and structure solution. The initial screening was performed using several commercially available sparse matrix crystallization kits with a Phoenix crystallization robot (Art Robbins Instruments), where 200-nl-volume sitting protein drops were mixed with an equivalent volume of reservoir solution. The initial crystals grew from 0.1 M Tris-HCl, pH 7.5, 0.2 M MgCl 2 , 25% polyethylene glycol (PEG) 3350 at room temperature. The optimized crystals were obtained using a crystallization solution containing 0.1 M Tris-HCl, pH 8.5, 0.2 M CaCl 2 , 0.6 M NaCl, 25% PEG 3350. The crystals were cryoprotected in the crystallization solution supplemented with 20% PEG 400 and flash-cooled in liquid nitrogen.
A 2.2-Å native data set of VesC (S225A) crystal was collected on beamline BL14-1 at the SSRL (Stanford Synchrotron Radiation Lightsource) and processed with HKL2000 (52) and XDS (53). The first two domains of the structure were solved by molecular replacement using PHASER (54). The initial molecular replacement solution was found by using PHASER with VesB PD and VesB Ig (PDB entry 4LK4; reference 41) as a search model for the protease domain and Ig fold domain of VesC. Each domain sequence was aligned with VesC and prepared for molecular replacement searching models with CHAINSAW (55).
For the C-terminal domain, molecular replacement searches were carried out using 30 templates identified by HHsearch (56), with no molecular replacement solutions found by PHASER with a TFZ score greater than 8. However, two templates, the CBM of Clostridium thermocellum (PDB entry 1UXX) and the human receptor protein tyrosine phosphatase mu (PDB entry 2C9A), gave solutions with similar orientations with a TFZ around 7. The resulting density maps, however, were uninterpretable in these regions, and it was not possible to manually improve the models.
Rosetta-based homology modeling was then carried out (57), combining pieces from these two homologous structures and refining with the Rosetta all-atom energy function (58) augmented with a term assessing agreement to density (38). After rephasing with the resulting low-energy models, the density was readily interpretable, with Phenix autobuild (59) building most residues in the structure, with an R free of around 37%. Subsequent rounds of refinement in Rosetta and Phenix autobuild further improved the model.
The structural model was subsequently improved and completed using the program Buccaneer (60) and Coot (61) and refined with the program REFMAC5 (62) with 8 translation/libration/screw (TLS) groups identified by the TLSMD server (63). The quality of the crystal structure was analyzed using MolProbity (64). Crystallographic data collection and refinement statistics are shown in Table 6. Least-squares analysis to determine the structural similarity was carried out using LSQKAB (65) and DaliLite (66). Protein quaternary-structure analysis was performed with the PISA server (67). The sequence alignment figure was made with ESPript (68). All other figures of molecular structures were prepared using PyMOL (DeLano Scientific Research LLC).
Data availability. The sequence reads from TRH7000 and the mutants were uploaded (submission ID SUB8079245) as BioProject no. PRJNA661062 to the BioProject database at NCBI. Coordinates and structure factors for the crystal structure of VesC have been deposited with the Protein Data Bank under accession code 6BQM.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.