The Vibrio cholerae Quorum-Sensing Protein VqmA Integrates Cell Density, Environmental, and Host-Derived Cues into the Control of Virulence

Quorum sensing (QS) is a process of chemical communication that bacteria use to orchestrate collective behaviors. QS communication relies on chemical signal molecules called autoinducers. QS regulates virulence in Vibrio cholerae, the causative agent of the disease cholera. Transit into the human small intestine, the site of cholera infection, exposes V. cholerae to the host environment. In this study, we show that the combination of two stimuli encountered in the small intestine, the absence of oxygen and the presence of host-produced bile salts, impinge on V. cholerae QS function and, in turn, pathogenicity. We suggest that possessing a QS system that is responsive to multiple environmental, host, and cell density cues enables V. cholerae to fine-tune its virulence capacity in the human intestine.

Here, first we explore whether oxygen levels modulate QS in V. cholerae. We find that V. cholerae cultured under anaerobic conditions does not produce CAI-1, whereas increased DPO production does occur. In this work, we focus on DPO. We show that the VqmA-DPO complex more strongly activates target gene expression under anaerobic than aerobic conditions. One consequence of the absence/presence of oxygen is an altered reducing/oxidizing (here, redox) cellular environment. We show that oxygendependent changes in VqmA activity are governed by cysteine disulfide bonds that are responsive to the redox environment. In the absence of DPO, during aerobic growth, Apo-VqmA forms an intramolecular disulfide bond that limits VqmA activity. In contrast, DPO-bound VqmA forms an intermolecular disulfide bond that enhances VqmA activity, and indeed, this intermolecular disulfide bond was also shown to be present in a recently reported crystal structure of Holo-VqmA (28). The formation of the intermolecular bond is not affected by oxygen levels. In the small intestine, V. cholerae encounters both the absence of oxygen and the presence of bile. Bile salts inhibit formation of the intermolecular disulfide bond in VqmA. Thus, bile and DPO have opposing effects on VqmA-DPO activity. We propose that the VqmA-DPO-VqmR QS pathway allows V. cholerae to integrate QS information, host cues, and environmental stimuli into the control of genes required for transitions between the human host and the environment.

RESULTS
V. cholerae does not produce the CAI-1 QS AI under anaerobic conditions. To our knowledge, V. cholerae QS has been studied only under aerobic conditions. We know that the marine-human host life cycle demands that V. cholerae transition between environments containing widely varying oxygen levels (17,18). Moreover, QS is crucial in both V. cholerae habitats. Thus, we sought to investigate whether oxygen modulates V. cholerae QS. First, we assessed the relative levels of the three known QS AIs from V. cholerae C6706 Sm r (here wild type [WT]) following aerobic and anaerobic growth. AI activity in cell-free culture fluids was measured using a set of three bioluminescent V. cholerae strains, each of which exclusively reports on one QS AI (either AI-2, CAI-1, or DPO) when it is supplied exogenously.
Unlike V. cholerae cultured in the presence of oxygen (here ϩO 2 ), V. cholerae grown in the absence of oxygen (here ϪO 2 ) produced no CAI-1 ( Fig. 2A). Twice as much AI-2 and DPO accumulated in V. cholerae cultured ϪO 2 as in ϩO 2 ( Fig. 2B and C). We note that the dynamic ranges for the CAI-1 and DPO assay are ϳ1,000and ϳ4-fold, respectively, while that for the AI-2 assay is ϳ100,000-fold (2,11). Thus, we consider the changes in CAI-1 and DPO to be physiologically relevant, whereas that for AI-2 is likely not, so we do not consider AI-2 further in this work. Additionally, V. cholerae cultured ϪO 2 grew to a lower final cell density than when grown ϩO 2 (see Fig. S1A in the supplemental material). We controlled for the reduced cell growth that occurs under the ϪO 2 conditions; nonetheless, no CAI-1 could be detected (Fig. S1B). Beyond lacking O 2 , our culture medium lacked an alternative terminal electron acceptor. Thus, we also FIG 2 Oxygen deprivation modulates V. cholerae AI production. (A) 80%, (B) 25%, or (C) 30% cell-free culture fluids prepared from WT V. cholerae grown in the presence or absence of O 2 were provided to V. cholerae reporter strains that produce bioluminescence in response to exogenous (A) CAI-1, (B) AI-2, or (C) DPO. Data represent the average values from biological replicates (n ϭ 3), and error bars represent SD. Statistical significance was calculated using a two-tailed Student t test. Asterisks are as follows: * denotes P Ͻ 0.05, *** denotes P Ͻ 0.001, and **** denotes P Ͻ 0.0001. LOD, limit of detection; AU, arbitrary units.
V. cholerae VqmA Detects Redox Using Disulfide Bonds ® considered the possibility that V. cholerae cultured under ϪO 2 conditions was unable to respire and therefore unable to drive CAI-1 generation. However, supplementation of the V. cholerae ϪO 2 cultures with the alternative terminal electron acceptor fumarate, which is readily consumed by V. cholerae (29), did not rescue CAI-1 production (Fig. S1B). Collectively, these data suggest that production of CAI-1 and DPO by V. cholerae is affected by oxygen levels. In the remainder of this study, we focus on the functioning of the DPO-VqmA QS circuit under different conditions that are predicted to be encountered in the host. We address possible ramifications of our results concerning CAI-1 and AI-2 in Discussion.
VqmA exhibits increased activity in the absence of oxygen. Given that V. cholerae accumulated more DPO under ϪO 2 conditions than ϩO 2 conditions, we wondered whether the VqmA-DPO QS system would, in turn, display increased activity under ϪO 2 conditions compared to ϩO 2 conditions. VqmA controls the expression of the vqmR gene, encoding the small RNA VqmR (Fig. 1). Therefore, expression of a vqmR-lacZ transcriptional fusion can be used to assess VqmA activity (11). Betagalactosidase was selected as the reporter because its activity is not affected by oxygen. The vqmR-lacZ construct was integrated onto the chromosome of Δtdh V. cholerae. Tdh (threonine dehydrogenase) is required for DPO production (11). Thus, the Δtdh strain makes no DPO but activates vqmR-lacZ expression in response to exogenously supplied DPO. We measured activity following growth under ϩO 2 and ϪO 2 conditions and in the absence and presence of exogenous DPO. In the presence of O 2 , vqmR-lacZ activity increased following supplementation with DPO (Fig. S2A). Compared to the ϩO 2 conditions, vqmR-lacZ activity was higher under ϪO 2 conditions, in both the absence and presence of DPO (Fig. S2A). Increased DPO-independent vqmR-lacZ expression in the absence of O 2 could be a consequence of increased production of VqmA or increased VqmA activity. To distinguish between these possibilities, we first examined whether changes in O 2 levels alter VqmA production by quantifying VqmA-FLAG produced from the chromosome Ϫ/ϩO 2 and Ϫ/ϩDPO. Similar levels of VqmA-FLAG were produced in all cases, suggesting that a change in VqmA abundance does not underlie increased vqmR-lacZ expression under ϪO 2 conditions (Fig. S2B). We next tested O 2 -driven changes in VqmA activity. To do this, we uncoupled expression of vqmA-FLAG from its native promoter by cloning vqmA-FLAG onto a plasmid under an arabinose-inducible promoter (here, pvqmA-FLAG). We introduced the plasmid into a ΔvqmA Δtdh V. cholerae strain harboring the vqmR-lacZ chromosomal reporter, and we measured both ␤-galactosidase output and VqmA-FLAG abundance in the same samples. VqmA-FLAG levels did not change under the different conditions (Fig. 3A); however, vqmR-lacZ reporter activity normalized to cellular VqmA-FLAG levels increased in the cells exposed to DPO, and overall activity was ϳ4to 7-fold higher under ϪO 2 conditions than ϩO 2 conditions in both the presence and absence of DPO (Fig. 3B). We conclude that VqmA displays an increased capacity to activate gene expression under anaerobic conditions relative to aerobic conditions. VqmA forms intra-and intermolecular disulfide bonds in an oxygen-and DPO-dependent manner. We wondered what molecular mechanism drives the increase in VqmA activity under ϪO 2 conditions (Fig. 3). The cytoplasmic compartment of aerobically respiring V. cholerae is relatively oxidizing (30,31). Thus, decreased oxygen levels would shift the cytoplasm to a reducing environment (32). Proteins can respond to such changes via redox-responsive cysteine residues (30,33). Inspection of the VqmA amino acid sequence revealed the presence of four cysteine residues, and all are strictly conserved in VqmA homologs in other vibrio species, but not in the VqmA receptor recently discovered in a vibriophage (Fig. S3). These findings led us to consider a model in which, in addition to activation by DPO, VqmA activity is regulated by redox-responsive cysteine residues.
Cysteine residues often undergo disulfide bond formation (33)(34)(35). We assessed whether VqmA forms disulfide bonds in vivo and, if so, whether their formation is influenced by DPO and/or oxygen levels. We grew ΔvqmA Δtdh V. cholerae carrying the pvqmA-FLAG construct under ϩO 2 conditions. We subsequently divided the culture into four aliquots. One portion was untreated (ϩO 2 , ϪDPO), one portion was supplied with DPO (ϩO 2 , ϩDPO), one portion was deprived of oxygen (ϪO 2 , ϪDPO), and one portion was deprived of oxygen and supplemented with DPO (ϪO 2 , ϩDPO). We extracted protein and analyzed the VqmA-FLAG protein profiles by immunoblotting. These analyses were performed with or without the addition of the reductant ␤-mercaptoethanol (BME) to distinguish between VqmA-FLAG species that had and had not formed disulfide bonds. Previous studies have shown that the presence of an intramolecular disulfide bond leads to a protein species displaying increased gel mobility compared to the same protein lacking the bond (36,37). In contrast, intermolecular disulfide bonds produce cross-linked protein oligomers that migrate with lower mobility than the corresponding monomers (35,38,39). We first consider the results for VqmA under ϩO 2 conditions. Under nonreducing conditions (ϪBME) and in the absence of DPO, VqmA-FLAG displayed mobility consistent with an oxidized monomer (labeled O-Monomer; Fig. 4A and B, lane 1). Treatment with BME caused VqmA-FLAG to migrate more slowly, consistent with it being a reduced monomer (labeled R-Monomer; Fig. 4A and B, lane 2). Administration of DPO drove formation of an additional VqmA-FLAG species, corresponding in size to an oxidized dimer (labeled O-Dimer; Fig. 4A  To garner additional evidence for the presence of an intramolecular disulfide bond(s) in the species designated VqmA O-monomer in Fig. 4B, we treated samples prepared from cells grown under ϩO 2 and ϪDPO conditions with methoxypolyethyl-  (40). m-MAL-PEG alkylates reduced cysteine residues and, in so doing, confers an ϳ5-kDa molecular weight change for each alkylation event (schematic in Fig. 4A). VqmA contains four cysteine residues; thus, the fully reduced protein would undergo 4 m-MAL-PEG events, while if one intramolecular disulfide bond exists, only two cysteine residues could react. Following treatment, Apo-VqmA-FLAG migrated primarily as two bands, corresponding to four (ϳ40% of protein) and two (ϳ60% of protein) alkylation events (Fig. 4C, compare lane 1 to lane 2), confirming that while a portion of VqmA-FLAG is fully reduced in vivo, the majority of the protein exists as an oxidized species containing one intramolecular disulfide bond.
To determine the residues involved in VqmA disulfide linkages, we conducted two experiments. First, we employed an in vitro thiol-trapping strategy based on sequential reactions that modify accessible cysteine residues with two thiol-specific reagents ( Fig. S4A) (41). In the initial thiol-blocking step, purified 6ϫHis-VqmA treated with diamide, a thiol-specific oxidant (42), was denatured and incubated with chloroacetamide (CAA). CAA alkylates accessible cysteine residues (i.e., those not involved in disulfide bonds), blocking them from further modification. The CAA-treated sample was next treated with Tris(2-carboxyethyl)phosphine hydrochloride (TCEP), a reductant, enabling non-CAA-labeled cysteines to be reduced. These newly freed residues were labeled in the final alkylation step with N-ethylmaleimide (NEM). The sample was then analyzed by mass spectrometry. The logic is that if a particular cysteine residue was inaccessible to CAA due to disulfide bonding, it would be preferentially labeled with NEM in the subsequent NEM modification step. Thus, the NEM/CAA ratio would be Ͼ1. In contrast, if a cysteine residue was not involved in disulfide modification, it would be preferentially labeled with CAA and have a NEM/CAA ratio of Ͻ1. The VqmA C48 and C63 residues had NEM/CAA ratios of ϳ100 and ϳ10, respectively, suggesting both of these residues are involved in disulfide linkages (Fig. S4B). We were unable to obtain good coverage of the C134 and C22 residues using this technique, so we could not similarly assess them.
Our second experiment to probe disulfide bonds in VqmA relied on mutagenesis. We individually substituted an alanine residue for each cysteine residue in the pvqmA-FLAG construct, introduced the plasmids into the ΔvqmA Δtdh V. cholerae strain, and repeated the analyses described in Fig. 4C. We first consider the case of intramolecular disulfide bond formation under ϩO 2 ϪDPO conditions. In the mutant proteins, following replacement of a cysteine residue with alanine, a maximum of three residues can react with m-MAL-PEG in the fully reduced protein (Fig. 4A, top). However, if an intramolecular disulfide bond is present, then only one cysteine residue can be decorated with m-MAL-PEG. Figure 4C shows that the Apo-VqmA C22A and Apo-VqmA C134A proteins each migrated as two bands, a result consistent with portions of each protein harboring one and three m-MAL-PEG moieties ( Fig. 4C; compare lane 3 to lanes 2 and 1 and compare lane 6 to lanes 2 and 1). This result suggests that the fraction of Apo-VqmA C22A and the fraction of Apo-VqmA C134A that exhibit one m-MAL-PEG decoration harbor intramolecular disulfide bonds. Apo-VqmA C48A and Apo-VqmA C63A migrated largely as single bands at the region corresponding to three m-MAL-PEG decorations ( Fig. 4C; compare lanes 4 and 5 to lane 2), suggesting that these proteins exist as reduced species and are thus incapable of forming intramolecular disulfide bonds. Therefore, we conclude that in WT VqmA, an intramolecular disulfide bond is formed between cysteine residues 48 and 63.
Next, we consider intermolecular disulfide bond formation under ϩO 2 ϩDPO conditions, and under nonreducing conditions (i.e., ϪBME). Figure 4D shows that the Holo-VqmA C22A, Holo-VqmA C48A, and Holo-VqmA C63A proteins migrated as mixtures of oxidized monomers and oxidized dimers, while the Holo-VqmA C134A protein migrated exclusively as an oxidized monomer (compare lanes 2 to 5 to lane 1). These data suggest that in Holo-VqmA, there is a C134-C134 intermolecular disulfide linkage. We also constructed and assessed the double VqmA C63A C134A mutant under ϩO 2 ϩm-MAL-PEG and ϩO 2 ϩDPO conditions. Figure 4C shows that under aerobic conditions, all of the Apo-VqmA C63A C134A protein contains two m-MAL-PEG decorations (compare lane 7 to lane 2), confirming that the two remaining cysteine residues were accessible and that the protein is fully reduced. Figure 4D shows that under aerobic conditions Holo-VqmA C63A C134A migrates entirely as a monomer (compare lanes 6 to lane 1). Thus, Holo-VqmA C63A C134A is incapable of forming both intra-and intermolecular disulfide bonds.
Collectively, our data suggest that (i) VqmA forms disulfide bonds in vivo and in vitro; (ii) an intramolecular disulfide bond is formed between VqmA C48 and C63, and an intermolecular disulfide bond is made between C134 and C134; and (iii) disulfide bond formation is influenced by both oxygen and DPO.
VqmA activity is limited by the intramolecular disulfide bond and enhanced by the intermolecular disulfide bond. To explore the in vivo consequences of VqmA disulfide bond formation on VqmA function, we tested whether the Apo-and Holomutant VqmA proteins that are incapable of forming particular intra-and/or intermolecular disulfide bonds displayed altered abilities to activate target vqmR transcription (see schematic in Fig. S5A). We introduced the vqmA-FLAG, vqmA C48A-FLAG, and vqmA C134A-FLAG alleles onto the chromosome of a V. cholerae Δtdh strain carrying vqmR-lacZ and measured reporter activity following aerobic growth ϩ/ϪDPO. Our rationale was that WT VqmA forms both the C48-C63 intramolecular bond and the C134-C134 intermolecular bond. In contrast, the VqmA C48A variant is unable to form the C48-C63 intramolecular bond and the VqmA C134A variant is unable to form the C134-C134 intermolecular bond ( Fig. 4C and D). Therefore, by comparing the activities of these three proteins, we could assess the effect of individually eliminating each disulfide bond on VqmA activity. We likewise made a strain carrying vqmA C63A C134A-FLAG on the chromosome to examine the effect of simultaneous elimination of both disulfide bonds.
First, we consider the case of Apo-VqmA. Apo-VqmA C48A and Apo-VqmA C134A exhibited an ϳ2-fold and ϳ4-fold increase and decrease, respectively, in reporter activity, relative to WT Apo-VqmA (Fig. S5B). Apo-VqmA C63A C134A exhibited an ϳ3-fold increase in reporter activity relative to WT Apo-VqmA and an ϳ10-fold increase in reporter activity relative to the strain carrying the Apo-VqmA C134A single mutant (Fig. S5B). Thus, we conclude that the C48-C63 intramolecular disulfide bond limits transcriptional activity of Apo-VqmA.
VqmA activity is differentially modulated by the cellular redox environment. To test whether VqmA activity is responsive to cellular redox, we supplemented the strains carrying the different VqmA variants with dithiothreitol (DTT), a cell-permeating reductant. We reasoned that if the absence of oxygen generates a reducing environment that prevents the formation of a particular disulfide bond, addition of DTT would mimic this condition by promoting a reducing cellular environment, including in the presence of oxygen. To control for any potential DTT-induced changes in the levels of chromosomally expressed vqmA, we introduced plasmids harboring arabinoseinducible vqmA-FLAG, vqmA C134A-FLAG, or vqmA C63A C134A-FLAG into the V. cholerae ΔvqmA Δtdh strain carrying vqmR-lacZ and measured reporter activity ϩ/ϪDPO and ϩ/ϪDTT.
In the absence of DTT, the Apo-and Holo-plasmid-borne VqmA variants displayed reporter activities similar to when the variants were expressed from the chromosome (plasmid-borne variants are in Fig. 5A and B; compare the black bars in each panel; for chromosomal variants, see Fig. S5B and C). Treatment with DTT increased reporter activity for WT Apo-VqmA and Apo-VqmA C134A but did not alter the reporter activity in the strain carrying Apo-VqmA C63A C134A (Fig. 5A, compare black and gray bars for each strain). As a reminder, WT Apo-VqmA and Apo-VqmA C134A form the C48-C63 intramolecular disulfide bond, while the Apo-VqmA C63A C134A protein does not. Thus, these data suggest that a reducing environment interferes with formation of the C48-C63 intramolecular disulfide bond, thereby eliminating its negative effect on Apo-VqmA activity.
Regarding WT Holo-VqmA, reporter activity diminished by ϳ6-fold when DTT was present in addition to DPO (Fig. 5B; compare the first pair of black and gray bars). In contrast, DTT supplementation did not significantly affect reporter activity in the strain carrying Holo-VqmA C134A (Fig. 5B; compare the second set of black and gray bars), while ϳ3-fold-lower activity was produced by the strain carrying Holo-VqmA C63A C134A (Fig. 5B; compare the third set of black and gray bars). Since the activity of VqmA C134A was not affected by DTT supplementation, we conclude that the C134-C134 intermolecular bond does not form in a reducing environment, and without that bond, Holo-VqmA transcriptional activity is diminished. Indeed, further emphasizing this conclusion, following DTT administration, reporter activity declined 5-fold in the DPOtreated strain harboring arabinose-inducible vqmA C48A-FLAG (Fig. S5D).
VqmA DNA binding capacity is differentially modulated by its redox environment. We suspected that the in vivo redox-dependent changes in VqmA disulfide bond formation would have ramifications on VqmA DNA binding capability. To explore this notion, we examined the ability of purified VqmA proteins to bind pvqmR promoter DNA. First, to examine the role of the C48-C63 intramolecular disulfide bond, we assessed DNA binding for 6ϫHis-VqmA C134A treated with diamide (to enable disulfide bond formation) or DTT (to prevent disulfide bond formation) (see Fig. S6). Relative to the DTT-treated protein, the diamide-treated protein exhibited a 4-fold reduction in DNA binding (Fig. 5C; compare lanes 2 to 6 with lanes 7 to 11). This result suggests that formation of the C48-C63 intramolecular disulfide bond limits DNA binding.
We investigated the role of the VqmA C134-C134 intermolecular disulfide bond on DNA binding by comparing the DNA binding capabilities of 6ϫHis-VqmA and 6ϫHis-VqmA C134A. Diamide-treated 6ϫHis-VqmA was approximately twice as potent at binding DNA as was 6ϫHis-VqmA C134A (compare left halves of . Like 6ϫHis-VqmA C134A, DTT-treated 6ϫHis-VqmA was more proficient in DNA binding than diamide-treated 6ϫHis-VqmA, consistent with intramolecular disulfide bond formation limiting DNA binding (Fig. 5D). Collectively, the data in Fig. 4 and 5 suggest a model in which the transcriptional and DNA binding activities of both Apo-VqmA and Holo-VqmA are modulated by disulfide bond formation, the cytoplasmic redox environment, and the level of O 2 in the environment.
Bile salts interfere with VqmA disulfide bond formation and decrease VqmA transcriptional activation activity. Our above findings suggest that the VqmA-DPO Bound and free correspond to DNA that is and is not bound to VqmA protein, respectively. As indicated, VqmA had been treated with 10-fold molar excess of diamide or DTT. Data in panels A and B represent average values from biological replicates (n ϭ 3), and error bars represent SD. Statistical significance was calculated using a two-tailed Student t test. Asterisks are as follows: ** denotes P Ͻ 0.01, **** denotes P Ͻ 0.0001, and NS denotes P Ͼ 0.05.
V. cholerae VqmA Detects Redox Using Disulfide Bonds ® signal transduction pathway, which represses virulence factor production and biofilm formation, is most highly active under anaerobic conditions. V. cholerae encounters anaerobiosis in the human intestine. The paradox is that in the intestine, V. cholerae is virulent and makes biofilms. We thus wondered if a possible host intestinal signal(s) could modulate VqmA-DPO signaling, allowing infection to proceed under anaerobic conditions. Bile salts, present in high concentrations in the human small intestine, can alter the redox environment of bacterial cells and thereby affect disulfide bond formation in cytoplasmic proteins (43). Thus, we were inspired to investigate whether bile salts could abrogate VqmA-DPO transcriptional activation activity. We cultured the ΔvqmA Δtdh V. cholerae strain carrying the arabinose-inducible pvqmA-FLAG construct and vqmR-lacZ on the chromosome in the presence and absence of oxygen, bile, and DPO, and we measured reporter activity. Treatment with bile salts caused ϳ2-fold and ϳ10-fold decreases in vqmR-lacZ reporter activity under ϩO 2 and ϪO 2 growth, respectively (Fig. 6A, first four bars). Supplementation with bile salts also decreased vqmR-lacZ reporter activity in cultures supplied with DPO, again with the maximum effect observed under ϪO 2 growth (Fig. 6A, 5th bar onward). To test whether the presence of bile salts affects VqmA disulfide bonds, we used analyses similar to those in Fig. 4B. , and error bars represent SD. Data in panel C represent the average value from three biological replicates and two technical replicates for each sample (n ϭ 6), and error bars represent SD. Statistical significance was calculated using a two-tailed Student t test. Asterisks are as follows: ** denotes P Ͻ 0.01, *** denotes P Ͻ 0.001, **** denotes P Ͻ 0.0001, and NS denotes P Ͼ 0.05. Consistent with the vqmR-lacZ reporter activity, bile salt supplementation prevented formation of O-dimers, in both the presence and absence of O 2 , suggesting it interferes with the Holo-VqmA C134-C134 intermolecular disulfide bond (Fig. 6B). We do not know whether or not bile affects VqmA intramolecular disulfide bond formation.
Bile salt-mediated disruption of VqmA-DPO-driven signal transduction promotes V. cholerae virulence. In V. cholerae, VqmA-DPO-directed production of VqmR results in decreased expression of genes involved in biofilm formation and virulence factor expression, including vpsL and tcpA, respectively (11,15). VpsL is required to synthesize V. cholerae exopolysaccharide, an essential component of the biofilm matrix, and TcpA is a virulence factor required for V. cholerae to colonize the human small intestine (44,45). Our finding that supplementation with bile salts inhibited VqmA-DPO function and that the effect of bile salt-mediated inhibition occurred primarily in the absence of oxygen led us to predict that the repression of vpsL and tcpA expression would also be maximally disrupted following bile salt supplementation in the absence of oxygen. We measured transcript levels of vpsL and tcpA in the WT and ΔvqmA strains following exposure to 25 M DPO or bile salts, deprivation of oxygen, or combinations of the three treatments.
In WT V. cholerae, bile salt supplementation modestly increased vpsL and tcpA transcript levels (ϳ3-and ϳ2-fold, respectively) but, however, only under ϪO 2 ϪDPO conditions ( Fig. 6C and Fig. S7). These data suggest that bile salts induce genes required for biofilm formation and virulence in the absence of oxygen and that Holo-VqmA can override the effect of bile. Transcript levels for both vpsL and tcpA under ϪO 2 ϪDPO conditions were further increased in the ΔvqmA strain treated with bile salts (ϳ6-and ϳ3-fold, respectively). We interpret this result to mean that bile salts also induce an increase in vpsL and tcpA expression through a pathway that does not involve VqmA. However, because the major effect of bile salts occurs only in the ΔvqmA strain, we conclude that this additional pathway is epistatic to VqmA in the control of vpsL and tcpA.
We tested whether the above changes in gene expression translated into alterations in V. cholerae pathogenicity by assaying whether bile salts affected cytotoxicity of V. cholerae in coculture with human Caco-2 intestinal cells. We generated differentiated monolayers of Caco-2 cells and cocultured them with either WT or ΔvqmA V. cholerae in the presence and absence of bile salts. The presence of bile salts increased V. cholerae-mediated cytotoxicity to Caco-2 cells with the ΔvqmA strain driving twice as much killing as WT V. cholerae (Fig. 6D). Collectively, our data suggest that bile salts induce increased V. cholerae virulence when bacteria are deprived of oxygen, part of the bile salt effect is exerted via interference with the VqmA-DPO-VqmR QS circuit, and the presence of VqmA limits the ability of bile salts to affect target gene expression.

DISCUSSION
The diversity of environments that V. cholerae inhabits, from the ocean to marine organisms to the human stomach to the human intestine, necessitates that the bacterium rapidly perceives changes in its external environment and appropriately tailors its gene expression programs. Our current study reveals that V. cholerae alters both which AIs are produced and the functioning of the VqmA-DPO-VqmR QS circuit in response to its environment. To our knowledge, these findings represent the first dissection of QS activity in the absence of oxygen in a facultative aerobic bacterium. We find that (i) the amounts of two AIs (CAI-1 and DPO) produced are dictated by oxygen levels; (ii) a single QS protein (VqmA) is capable of integrating information from three sources (AI, oxygen, and bile salts); and (iii) two disulfide bonds in a QS receptor (VqmA) antagonize one another with respect to their effects on protein activity, thereby aiding in perception and response to changes in cellular redox.
The benefit(s) of producing and detecting multiple QS AIs has long been mysterious with respect to V. cholerae biology. Evidence suggests that each AI conveys specific information into the cell: CAI-1 measures the abundance of vibrios (kin) and AI-2 and DPO measure the level of nonvibrios (non-kin) in the vicinity (11, 13, 15). Our finding that CAI-1 is not produced in the absence of oxygen suggests that CAI-1 may also convey information about the external environment. Strains lacking the ability to synthesize CAI-1 display reduced survival in seawater and following challenge with oxidative stress (46). Thus, we propose that CAI-1 could drive the expression of genes required for the aerobic segment of the V. cholerae life cycle, and we are now testing this idea. The fact that CAI-1 is not produced under anoxia suggests that V. cholerae cannot take a census of kin in the absence of oxygen. Either kin counting is dispensable under anoxia, or perhaps, another molecule(s)/mechanism performs this function. In contrast, at a minimum based on sequencing data, thousands of bacterial species, including those found in the human microbiota, can synthesize AI-2, the AI used for interspecies communication (47). We found that AI-2 is synthesized in the absence of oxygen. Perhaps measurement of the abundance of non-kin bacteria is of paramount importance in densely populated niches containing complex bacterial consortia. Biosynthesis of both AI-2 and CAI-1 requires S-adenosylmethionine (SAM), an abundant metabolite that is crucial for methylation reactions (12,48,49). Thus, another possibility is that in V. cholerae, during periods of SAM limitation, CAI-1 production is curbed as a means of sparing SAM for other uses. Continued AI-2 production could suffice for QS-mediated cell density tracking. Moreover, when SAM is used to produce AI-2 but not CAI-1, SAM is regenerated via downstream reactions (12,50). Thus, making AI-2 from SAM would not deplete the SAM reservoir.
Oxygen is a terminal electron acceptor and therefore a critical substrate for bacterial growth. The human intestine is devoid of oxygen, and invading bacteria, such as V. cholerae, that normally inhabit the relatively oxygenated marine environment need to alter their physiology to survive. With the exception of a few studies (30,51,52), the molecular mechanisms by which V. cholerae perceives the absence of oxygen and translates this information into changes in gene expression are unexplored. Here, we demonstrate that, in the presence of O 2 , Apo-VqmA forms a C48-C63 intramolecular disulfide bond that restricts the ability of the protein to bind DNA. Formation of this bond is inhibited in the absence of oxygen or following supplementation of aerobic cells with a reductant that, analogous to the absence of oxygen, generates a reducing environment. Thus, we propose that by interacting with the cell's redox environment, VqmA provides V. cholerae a mechanism to monitor oxygen levels (Fig. 7A). In this context, we note that anaerobiosis causes an ϳ7-fold increase in Apo-VqmAdependent pvqmR-lacZ reporter activity (Fig. 3B). In contrast, DTT supplementation under aerobic conditions causes only an ϳ3-fold increase in Apo-VqmA activity while the activity of Apo-VqmA C63A C134A, which lacks both disulfide bonds, is unchanged (Fig. 5A). One interpretation of these data is that Apo-VqmA is responsive to additional oxygen-dependent stimuli that are not mimicked by DTT or by the inability to make disulfide bonds. We are currently testing this possibility.
With respect to DPO-bound VqmA, we found that Holo-VqmA forms a C134-C134 intermolecular disulfide bond. Our data expand on a recent study of VqmA in which the crystal structure revealed the C134-C134 disulfide bond in Holo-VqmA. The authors predicted that the C134-C134 bond could interfere with VqmA DNA binding (28,53). However, the role of the disulfide bond on VqmA activity was not experimentally investigated in that earlier work (28). Here, we show that the C134-C134 bond is largely absent in Apo-VqmA while, upon binding DPO, ϳ40 to 50% of Holo-VqmA undergoes intermolecular C134-C134 disulfide bond formation, and moreover, this bond promotes Holo-VqmA activation of transcription ( Fig. 4B and Fig. 5). We do not understand why the conclusion one naturally comes to from the crystal structure does not match the experimental data. We do note that the C134 residue is located in a flexible loop region. Thus, one possibility is that, in the static crystal, the loop region is spatially constrained and does not reflect dynamic structures that VqmA adopts in solution. This notion awaits experimental testing.
Our results show that formation of the C134-C134 bond is not modulated by oxygen levels but is inhibited by the reductant DTT. The absence of oxygen imposes a mildly reducing environment on cells, while the presence of DTT imposes reductive stress (32,54), suggesting that the C134-C134 intermolecular VqmA disulfide bond may allow V. cholerae to monitor reductive stress. Under in vitro conditions, the negative effect exerted by the C48-C63 intramolecular disulfide bond on DNA binding was more significant than the positive effect exerted by the C134-C134 intermolecular disulfide bond. This result contrasts with our in vivo data (Fig. 5B), in which the intermolecular disulfide bond has the most pronounced effect on WT VqmA activity. One explanation that we are now exploring is that in vivo, additional factors modulate the activity of WT VqmA containing the intermolecular disulfide bond. Collectively, we suggest a model in which cycling between multiple redox states, namely, Oxidized-Apo-VqmA, Reduced-Apo-VqmA, Oxidized-Holo-VqmA, and Reduced-Holo-VqmA, enables V. cholerae to tune its QS-controlled collective behaviors to a range of redox states (Fig. 7B). There exist examples of individual disulfide bonds restricting or enhancing the activity of transcription factors (55)(56)(57). To our knowledge, however, this is the first example in which the same protein simultaneously uses two different disulfide bonds to modulate activity.
Bile is an abundant compound in the human small intestine that is well known to alter virulence in V. cholerae and other enteric pathogens such as Salmonella enterica serovar Typhimurium and Shigella flexneri (58,59). Bile is a heterogeneous mixture of molecules, and studies have largely focused on defining the roles of individual components in bacterial physiology. Intriguingly, the individual components can drive opposing effects. In V. cholerae, bile fatty acids repress while the bile salt taurocholate induces virulence (20-22, 25, 60, 61). In our current study, we elected to use a mixture V. cholerae VqmA Detects Redox Using Disulfide Bonds ® of bile salts, reasoning that this strategy would more closely approximate what V. cholerae encounters in vivo. Our data suggest that bile salts disrupt the formation of the VqmA C134-C134 intermolecular disulfide bond. We do not know the mechanism by which this occurs. However, previous studies show that bile salts, specifically cholic acid (CHO) and deoxycholic acid (DOC), interfere with redox homeostasis in Escherichia coli by shifting the cellular environment to an oxidizing one and fostering disulfide bond formation in cytosolic proteins (43). In the context of our work, since VqmA intermolecular disulfide bond formation is disrupted, we propose that application of a bile salts mixture to V. cholerae causes reductive stress. Consistent with this idea, taurocholate binds to and inhibits DsbA, a protein required for the introduction of disulfide bonds in periplasmic proteins (24). We currently do not know whether incubation of V. cholerae with CHO and DOC, rather than a bile salts mixture, would drive phenotypes mimicking those observed in E. coli.
What advantage does V. cholerae accrue by using the regulatory program uncovered in our study? We propose that V. cholerae uses the different blends of AIs it encounters along with environmental modulation of VqmA activity to gauge its changing locations in the host. Thus, VqmA functions rather like a GPS device. In response to the information obtained about its microenvironment through VqmA, V. cholerae can appropriately tune its gene expression in space and time. We say this because, prior to entry into the small intestine (the site of cholera disease), V. cholerae will encounter oxygen limitation in the stomach. However, premature expression of virulence genes in the stomach, in the face of low pH and antimicrobial peptides, would be unproductive and, moreover, divert energy from combating host defense systems. Thus, increased VqmA activity, due to enhanced accumulation of DPO under anoxia coupled with formation of the C134-C134 intermolecular disulfide bond and suppression of formation of the activity-dampening intramolecular disulfide bond, will increase production of VqmR and, in turn, repress expression of genes involved in biofilm formation and virulence. Anoxia and the concurrent presence of bile, encountered upon entry into the host duodenum (upper region of the small intestine), may provide a spatially relevant signal to alert V. cholerae to begin to express virulence genes. In this case, bile salt-mediated inhibition of VqmA activity due to prevention of formation of the intermolecular disulfide bond will decrease VqmR production and, in turn, enhance expression of genes involved in biofilm formation and virulence. The combined use of bile salts and anoxia to decrease and increase VqmA activity, respectively, is also noteworthy because, as V. cholerae proceeds further through the intestinal tract, concentrations of bile salts decrease near the ileum (lower portion of the small intestine), where ϳ95% of bile salts are reabsorbed, and they reach a minimum in the large intestine. In contrast, anoxia is maintained throughout the intestinal tract. During successful infection, V. cholerae cell numbers increase as disease progresses. Accumulation of DPO should track with increasing cell density. Thus, it is possible that late in infection, V. cholerae resides in a high-DPO, anoxic environment lacking bile salts, conditions enabling reengagement of the VqmA-VqmR-DPO circuit, termination of virulence factor production, and expulsion from the host. In this context, we also note that DPO production by V. cholerae requires threonine as a substrate. Mucin is a major constituent of the intestinal tract and is composed of ϳ35% threonine (62). Intriguingly, both the stomach and large intestine, locations where V. cholerae typically does not reside, contain more mucus-secreting glands and mucus layers than does the small intestine (62). Thus, it is likely that enhanced access to mucus-derived threonine in the large intestine allows V. cholerae to increase DPO production. Again, high DPO levels repress virulence and biofilm formation. Thus, increased DPO production in the large intestine would foster V. cholerae departure from the host. We speculate that, in addition to oxygen and bile salts, the presence of mucus or threonine or the ability to synthesize DPO is also leveraged by V. cholerae as an additional spatial cue to optimize host dispersal timing. added to the culture medium at 4 h postinoculation, and growth was continued for an additional 1 h, at which point the cultures were divided and portions were supplemented with 25 M DPO and/or 0.5% bile salts and/or deprived of oxygen. Treatments were continued for another 1 h, after which cells were harvested by centrifugation at 13,000 rpm and the pellets were immediately frozen at Ϫ80°C until use.
Immunoblotting. Cells were resuspended in ice-cold phosphate-buffered saline (PBS) and diluted to a final OD 600 of 7 for protein produced from the chromosome or to an OD 600 of 3.5 for protein produced from a plasmid, in a volume of 20 l. The cells were lysed by addition of 5 l Bugbuster (Novagen) supplemented with 1 g lysozyme and 25 U/ml benzonase. Samples were combined with SDS-PAGE buffer in the presence or absence of BME (100 mM) and boiled for 20 min, and proteins were separated on 4 to 20% Mini-Protein TGX gels (Bio-Rad). Proteins were transferred to polyvinylidene difluoride (PVDF) membranes (Bio-Rad) for 1 h at 4°C at 100 V. Membranes were blocked overnight in PBST (1ϫ PBS, 0.03% Tween 20) supplemented with 5% milk, washed 5 times with PBST, and incubated for 40 min with 1:5,000 dilution of monoclonal anti-FLAG-peroxidase antibody (Sigma) in PBST. The membranes were subsequently washed another five times with PBST. FLAG epitope-tagged protein levels were visualized using the Amersham ECL Western blotting detection reagent (GE Healthcare). Thereafter, the antibody was removed by 2 serial incubations in stripping buffer (15 g/liter glycine, 1 g/liter SDS, 10 ml/liter Tween 20, pH 2.2) for 5 min each. The membrane was reequilibrated by 4 washes in PBST, 20 min each, and used to detect the abundance of the loading control, RNA polymerase ␣. Washes and incubations as described above were performed to enable antibody binding and removal of excess. The primary antibody, anti-E. coli RNA polymerase ␣ (Biolegend), and the secondary antibody, anti-mouse IgG horseradish peroxidase (HRP) conjugate antibody (Promega), were both used at a 1:10,000 dilution. In all cases, protein levels were quantified using Image J software.
Protein purification. pET15b plasmids encoding 6ϫHis-VqmA were mobilized into Δtdh E. coli BL21(DE3). Strains were cultured for protein production as described previously (15,16). Cells were harvested by centrifugation, and pellets were resuspended in 1/100 volume of lysis buffer (50 mM Tris, 150 mM NaCl, pH 7.5, containing 0.5 mg/ml lysozyme, 1ϫ protease inhibitor, and benzonase) for 5 min followed by the addition of an equal volume of Bugbuster reagent (Novagen). The cell lysate was clarified by centrifugation at 13,000 rpm, and protein was purified using Ni-NTA superflow resin (Qiagen), according to the manufacturer's recommendations for a centrifugation-based protocol, except that the loading and wash buffers all contained 1 to 5 mM imidazole to decrease nonspecific protein binding. The protein was eluted from the resin using 300 mM imidazole and thereafter dialyzed twice against 50 mM Tris, 150 mM NaCl, pH 7.5, using a Slide-A-Lyzer module (Thermo Fisher). When necessary, buffers were amended with 5 mM DTT. To purify Holo-VqmA, buffers were supplemented with 100 M DPO.
EMSAs. The DNA corresponding to the promoter region of vqmR, ϳ100 bp, was amplified using V. cholerae genomic DNA as a template. Where mentioned, protein was pretreated in binding buffer with 10-fold molar excess DTT or diamide in order to reduce or oxidize the protein, respectively. To initiate electromobility gel shift assays (EMSAs), 0.2 to 3.5 M protein was combined with 30 ng probe DNA in binding buffer (50 mM Tris-HCl, pH 8, 150 mM NaCl). Reactions were allowed to proceed at room temperature (RT) for 15 min. Samples were separated on a Novex 6% DNA retardation gel (Thermo) by electrophoresis in 1ϫ Tris-buffered EDTA (TBE) at 100 V. Gels were subsequently incubated with Sybr green reagent, diluted in 1ϫ TBE at RT for 25 min, washed with five successive rounds of ddH 2 O, and imaged using an ImageQuant LAS 4000 imager and the Sybr green channel setting.
Analysis of relative AI levels in conditioned growth medium following aerobic or anaerobic growth. V. cholerae strains were cultured overnight in LB medium (ϳ16 to 18 h) and diluted into fresh aerobic or anaerobic LB medium to a final OD 600 of 0.004. The cultures were incubated at 37°C for an additional ϳ6 h with shaking. The cells were removed by centrifugation at 13,000 rpm, and the spent medium was filtered through 0.2-m filters. Twenty percent (vol/vol) 5ϫ LB was added to the spent medium preparations (here designated reconditioned spent medium). Negative controls consisted of spent medium prepared from strains incapable of synthesizing CAI-1, AI-2, and DPO. Subsequently, reconditioned spent medium was combined with V. cholerae reporter strains expressing only a single QS receptor that, therefore, detect only one AI. In the case of CAI-1 and AI-2 detection, the reporter strains carried a plasmid encoding the Vibrio harveyi luxCDABE (luciferase) operon (2). For DPO detection, the reporter strain possessed a pvqmR-lux (luciferase) fusion on the chromosome in place of the native lacZ locus (16). Genotypes are provided in Table S1 in the supplemental material. The reporter strains and reconditioned spent medium preparations were combined to a final volume of 150 l in wells of 96-well plates, covered with Breathe-Easy film, and incubated at 30°C with shaking for 2.5 to 4 h. Finally, bioluminescence and OD 600 values were recorded. Relative light units (RLUs) were defined as light production (counts per minute) divided by OD 600 . Normalized RLUs were obtained by subtraction of the RLU values obtained from the negative controls.
Beta-galactosidase assays. V. cholerae strains cultured overnight in LB medium were diluted into fresh M9 medium to a final OD 600 of 0.004 and thereafter cultured to an OD 600 of ϳ0.3 (ϳ4 h). The cultures were held on ice prior to assay or subjected to further treatments, as described below. For strains requiring inducible protein expression, 0.01% arabinose was added to the culture medium at 4 h postinoculation and growth was continued for an additional 1 h. In all cases, cultures were divided into aliquots and individual portions were supplemented with 25 M DPO and/or 0.5% bile salts and/or deprived of oxygen. Cells were cultured for an additional 1 h and then held on ice prior to assay. To assess the effect of reductant, at 2 h postinoculation, the cultures were supplemented with 300 M DTT. Cells were cultured for another 2 h to allow DTT permeation, before vqmA expression was induced by the addition of arabinose. Subsequent treatments were as described above. LacZ activity assays were carried out as follows: cells were combined 1:1 (vol/vol) with Bugbuster reagent for 20 min. The assay was initiated by combining 20 l of the cell-Bugbuster mixture with 140 l of assay buffer (80% Bugbuster, 10% 10ϫ PBS, 1 mM MgSO 4 , 10 g/ml lysozyme, benzonase [0.05%, vol/vol], ␤-mercaptoethanol [0.1%, vol/vol], 67 g/ml fluorescein-di-␤-D-galactopyranoside). Changes in fluorescence were captured using the GFP channel on a Synergy Neo2 HTS multimode microplate reader. Activity units were defined as the change in fluorescence/minute/OD 600 of the culture at the point of harvest.
RNA isolation and quantitative RT-PCR. Strains cultured overnight in LB were diluted into fresh M9 minimal medium to a final OD 600 of 0.004. Next, the strains were grown to an OD 600 of ϳ0.3 (ϳ4 h postinoculation) with shaking at 37°C, at which point the cultures were divided into portions that were supplemented with 25 M DPO and/or 0.5% bile salts and/or deprived of oxygen. Treatments were continued for another 1 h, and the cells were harvested and treated for 15 min at room temperature with RNAProtect reagent, per the manufacturer's instructions. RNA was isolated using the RNeasy kit (Qiagen), and 2 g of total RNA was depleted of contaminating DNA using Turbo DNase (Applied Biosystems), using the manufacturer's recommended protocol. Five hundred nanograms of the resulting total RNA was used to construct cDNA libraries using SuperScript III reverse transcriptase (Invitrogen). qPCR was conducted using the PerfeCTa SYBR green FastMix Low ROX (Quanta Biosciences) reagent.
Caco-2 culture and coculture with bile salts and V. cholerae. The HTB37 cell line was obtained from ATCC and thereafter cultured and passaged in Eagle's minimum essential medium (EMEM) (ATCC) supplemented with 10% fetal bovine serum (FBS) (Thermo Fisher), 2 mM glutamine (Thermo Fisher), 1ϫ penicillin-streptomycin (Pen-Strep) (Thermo Fisher), and 2.5 g/ml plasmocin (Invitrogen), per ATCC recommendations. Prior to coculture, Caco-2 cells were seeded at 0.32 cm 2 into a tissue culture-treated 96-well plate in EMEM, as described above, except that the 2 mM glutamine was not included. The cells were cultured to confluence, the medium was removed by aspiration, and the cells were washed with Earle's balanced salt solution followed by the addition of EMEM containing 25 M DPO but lacking antibiotics/glutamine/FBS. At time zero, V. cholerae, grown as described below, was added, at a multiplicity of infection (MOI) of 10 and with 0.05% bile salts. Following coculture for 3.5 h, the medium and the bacteria were removed by aspiration, the wells were washed with Earle's balanced salt solution, and Caco-2 cell viability was assessed using the neutral red assay (69). For coculture with Caco-2 cells, V. cholerae WT and ΔvqmA strains were cultured overnight under static conditions in AKI medium containing 25 M DPO in tubes with an HV ratio of zero at 37°C. The next day, the cells were decanted into glass tubes and cultured with vigorous shaking for 1 h. Subsequently, the cultures were diluted with PBS to the appropriate density and added to the Caco-2 cells.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.03 MB.