Genome-wide mapping of Vibrio cholerae VpsT binding identifies a mechanism for c-di-GMP homeostasis

Abstract Many bacteria use cyclic dimeric guanosine monophosphate (c-di-GMP) to control changes in lifestyle. The molecule, synthesized by proteins having diguanylate cyclase activity, is often a signal to transition from motile to sedentary behaviour. In Vibrio cholerae, c-di-GMP can exert its effects via the transcription factors VpsT and VpsR. Together, these proteins activate genes needed for V. cholerae to form biofilms. In this work, we have mapped the genome-wide distribution of VpsT in a search for further regulatory roles. We show that VpsT binds 23 loci and recognises a degenerate DNA palindrome having the consensus 5′-W−5R−4[CG]−3Y−2W−1W+1R+2[GC]+3Y+4W+5-3′. Most genes targeted by VpsT encode functions related to motility, biofilm formation, or c-di-GMP metabolism. Most notably, VpsT activates expression of the vpvABC operon that encodes a diguanylate cyclase. This creates a positive feedback loop needed to maintain intracellular levels of c-di-GMP. Mutation of the key VpsT binding site, upstream of vpvABC, severs the loop and c-di-GMP levels fall accordingly. Hence, as well as relaying the c-di-GMP signal, VpsT impacts c-di-GMP homeostasis.


INTRODUCTION
Cholera is an acute diarrhoeal disease caused by the bacterium Vibrio cholerae (1). Globally, the current seventh pandemic is caused by strains of the O1 El Tor biotype that emerged in the Bay of Bengal, later spreading across the globe in three waves (2). Once introduced at a locality, V. cholerae strains capable of causing disease outbreaks can remain for several decades (3,4). In part, this is because V. cholerae can spend long periods in aquatic environments (5). In this niche, V. cholerae forms biofilms on the chitinous exoskeletons of crustaceans and marine microbiota (5). Hence, these abundant surfaces are exploited as a source of carbon and nitrogen (6,7). Major components of the V. cholerae biofilm matrix include vibrio polysaccharide (VPS) and three proteins: RbmA, RbmC and Bap-1 (8). Synthesis of VPS depends on the vps-I and vps-II gene clusters that are transcribed from the vpsA and vpsL promoters, respectively. The nearby rbm gene cluster encodes RbmA and RbmB whilst Bap-1 is encoded elsewhere (9). Phase variation between smooth and rugose morphotypes also plays a role (10). Notably, rugose cells form biofilms more readily, present as corrugated colonies on agar, and aggregate in liquid culture (11,12).
In many bacteria, the decision to form biofilms involves the nucleotide second messenger cyclic dimeric guanosine monophosphate (c-di-GMP) (13). Within cells, c-di-GMP is produced by diguanylate cyclase (DGC) proteins. Such enzymes are identifiable by virtue of a 'GGDEF' active site amino acid motif (14). Phosphodiesterases, with 'EAL' or 'HD-GYP' domains, degrade c-di-GMP (15). The signalling molecule is detected by factors that modulate cell behaviour. For instance, c-di-GMP controlled transcription factors can alter global patterns of gene expression (13). Similarly, the message can be relayed by riboswitches and sRNA molecules (16)(17)(18). In V. cholerae, two c-di-GMP dependent regulators, VpsR and VpsT, activate the vps-I, vps-II and rbm gene clusters (19)(20)(21)(22)(23). In rugose phase variants, VpsR and VpsT are required to maintain the phenotype (24,25). The wider role of these regulators in the direct control of gene expression is less well understood.
First described in 2004, VpsT is a LuxR-type response regulator (20,24,26). Direct interaction with c-di-GMP is required for VpsT dimerisation and DNA binding (20). At the vpsA, vpsL and rbm promoters, VpsT displaces the histone-like nucleoid structuring (H-NS) protein that otherwise silences transcription (27). Conversely, at other loci, the transcription activation mechanism is unknown (28,29). Transcriptional repression can also be mediated by VpsT. For instance, gene expression is blocked by VpsT binding across the rpoS transcription start sites (TSSs) (30). Similarly, VpsT can inhibit transcription of the divergent VC1303 and VC1304 genes (31). The DNA site bound by VpsT is called the T-box (32). Efforts to define a consensus T-box have been hampered by the paucity and variability of known VpsT targets. Initial studies revealed recognition of identical sequences by VpsT at the vpsL and vpsA promoters (27,32). However, this sequence is not found at other targets (28,29,32). In this work we sought a better understanding of VpsT, its DNA regulatory targets, and their interplay with c-di-GMP signalling.

Strains, plasmids and oligonucleotides
Strains, plasmids and oligonucleotides used in this study are listed in Supplementary Table S1. All V. cholerae strains are derivatives of E7946 (33). To generate rugose derivatives of strains we used alkaline phosphate water (1% w/v proteose peptone and 1% w/v NaCl, pH 8.5) according to the protocol of Ali et al. (34). Chromosomal deletions and mutants were made using either the pKAS32 suicide plasmid for allelic exchange or the MuGENT method (35,36). The Escherichia coli strain JCB387 was used for routine cloning (37). Plasmids were transferred into V. cholerae by either conjugation or transformation as described previously (38,39).

ChIP-seq and bioinformatics
The chromatin immunoprecipitation assay was done as in prior work (39) using strain E7946, carrying plasmid pAMNF-vpsT, or empty pAMNF as a control. Note that pAMNF-vpsT encodes 3xFLAG-VpsT that is expressed from a low level constitutively active promoter (40). Lysates were prepared from mid-log phase cultures, incubated with shaking at 37 • C. Following sonication, the VpsT-DNA complexes were immunoprecipitated with an anti-FLAG antibody (Sigma) and Protein A sepharose beads. Immunoprecipitated DNA was blunt-ended, A-tailed and ligated to barcoded adaptors before elution and de-crosslinking. ChIP-seq libraries were then amplified by PCR and purified. Library quality was assessed using an Agilent Tapestation 4200 instrument and quantity determined by qPCR using an NEBnext library quantification kit (NEB). Libraries were sequenced as described previously (40) and reads are available from ArrayExpress using accession code E-MTAB-10829. Single-end reads from two independent VpsT ChIP-seq experiments were mapped to the reference V. cholerae N16961 genome (chromosome I: NC 002505.1 and chromosome II: NC 002506.1) with Bowtie 2 using the QuasR package. MACS2 (v2.1.2) was used to identify ChIP-seq peaks where aligned reads were significantly enriched (using a q-value cut-off of 0.05) compared to a control (without anti-FLAG antibody). Briefly, MACS2 exploits the characteristic bimodal distribution of sequencing reads at sites enriched by immunoprecipitation of transcription factors. This distribution, in single-end short read Illumina sequencing experiments, arises because only the 5 ends of precipitated DNA fragments, on each strand either side of the transcription factor binding site, are sequenced. Hence, the actual position of DNA binding is in the centre of the offset peak maxima for each DNA strand (41). Finally, identified peaks were visually inspected, using the Artemis genome browser, to confirm accurate assignment of peak centres (42,43). Average read counts were normalised according to the coverage per base for each experiment and visualized using Gviz.

␤-Galactosidase activity assays
Promoter DNA was fused to lacZ in plasmid pRW50T that can be transferred from E. coli to V. cholerae by conjugation (38). Assays of ␤-galactosidase activity were done according to the Miller method (44). Bacterial cultures were grown at 37 • C with shaking in LB broth, supplemented with appropriate antibiotics, to mid-log phase. Values shown are the mean of three independent experiments and error bars show the standard deviation.

Electrophoretic mobility shift assays and DNAse I footprinting
We purified VpsT as described previously (31). Promoter DNA fragments were excised from plasmid pSR and endlabelled with ␥ 32-ATP using T4 PNK (NEB). EMSAs and DNase I footprints were done as previously described (39). Full gel images are shown in Supplementary Figure S1.

Primer extension assays
The RNA for primer extension assays was purified from V. cholerae carrying appropriate plasmids as described above. Primer extension assays were done as previously described (45). The 5 end-labelled primer D49724, which anneals downstream of the HindIII site in pRW50T, was used in all experiments. Primer extension products were analysed on denaturing 6% polyacrylamide gels, calibrated with arbitrary size standards, and visualized using a Fuji phosphor screen and Bio-Rad Molecular Imager FX. Full gel images are shown in Supplementary Figure S1.

Colony fluorescence quantification
Cultures of V. cholerae were incubated overnight at 37 • C with shaking in LB. A 2 l suspension was spotted onto LB agar plates that were incubated overnight at 30 • C. The fluorescence signal from each of the two fluorophores was measured with a ChemiDoc MP system (Bio-Rad) as previously described (46). Image Lab (BioRad) was used to measure the fluorescence intensity, with the same number of pixels recorded for each colony. The ratio of TurboRFP to the AmCyan signal was then calculated. Values shown are mean ratios from four independent experiments, as a proportion of the wild-type ratio.

Genome-wide DNA binding by VpsT in Vibrio cholerae
We used ChIP-seq to map the binding of VpsT across the V. cholerae genome. To facilitate this, vpsT was cloned in plasmid pAMNF. The resulting DNA construct, pAMNF-vpsT, encodes an N-terminal 3xFLAG-vpsT fusion that was used to transform V. cholerae strain E7946. Importantly, the amount of 3xFLAG-VpsT produced from pAMNF-vpsT (Supplementary Figure S2, lanes 4-6) is similar to the Nucleic Acids Research, 2022, Vol. 50, No. 1 151 amount of VpsT generated from the native chromosomal locus in both smooth (lanes 1-3) and rugose (lanes 7-9) cells. In subsequent ChIP-seq analysis, anti-FLAG antibodies were used to isolate regions of DNA bound to VpsT. An overview of the data are shown in Figure 1A. Genes are shown as dark blue lines (outer two tracks) and the VpsT binding signal is in cyan (inner track). Strikingly, VpsT does not bind chromosome II. Conversely, we identified 23 binding peaks for VpsT across chromosome I (Table 1). In previous work, VpsT binding at 7 different promoter regions has been shown in vitro (27)(28)(29)(30)32). We recovered five of these interactions (adjacent to vpsU, vpsA, vpsL, rbmA and rpoS) in our ChIP-seq analysis ( Figure 1B and Table 1). We did not detect VpsT binding upstream of tag or katB (28,29). Of the 23 ChIP-seq peaks for VpsT, 11 mapped to genes or operons expressed differently in response to perturbation of vpsT (Table 1) (20). Consistent with expectation, most VpsT peaks locate to non-coding DNA upstream of a gene 5 end ( Figure 1C). The function of each gene flanking a binding peak is illustrated in Figure 1D. Many encode proteins implicated in biofilm formation, c-di-GMP metabolism, or motility. We used electrophoretic mobility shift assays (EM-SAs) to test direct binding of VpsT with 8 of the ChIP-seq derived targets (Supplementary Figure S3A). Binding was confirmed in all cases. Conversely, VpsT did not bind the E. coli lacZ promoter, which was used as a negative control (Supplementary Figure S3B).

Reassessment of the T-box consensus sequence for VpsT binding
The precise sequence of the T-box, the DNA motif recognized by VpsT, is not clear. In initial DNAse I footprinting studies, Zamorano-Sánchez et al. showed binding of VpsT to a DNA palindrome at the vpsL regulatory region (32). A near identical sequence was identified at the vpsA, vpsT, rbmA and rbmE promoters (32). Direct binding of VpsT has also been demonstrated at the rpoS locus using DNAse I footprinting (30). However, the bound sequence differed to the motif previously found (30). Other direct VpsT targets also lack the palindrome identified by Zamorano-Sánchez et al. (28,29,31). Logically, our ChIP-seq analysis should provide a starting point for better understanding of the Tbox consensus. To identify motifs common to all ChIP-seq targets we used MEME to parse DNA sequences centred on each of the 23 VpsT binding peaks. The motif identified is shown in Figure 1E  To aid comparison, we numbered positions in both palindromes with respect to the centre of the two sequences (Figure 1E). Whilst the ChIP-seq derived logo is shorter, and exhibits greater degeneracy, there are similarities between the sequences. Most notably, positions −3C and +3G of both motifs are highly conserved and likely to be important for VpsT binding.

Mutagenesis of the vpsL promoter T-box
Recall that Zamorano-Sánchez et al. used DNAse I footprinting to locate VpsT binding upstream of vpsL (32).  (22). The centre of the VpsT ChIP-seq peak is marked by an asterisk. Positions in bold align to the motif found using ChIP-seq. Overlapping sequences matching the original motif of Zamorano-Sánchez and colleagues are underlined (32). To further understand the requirements for VpsT binding we made a series of point mutations. The base changes, numbered with respect to the vpsL TSS, are marked red in Figure 2A. To measure the impact of the mutations we used DNAse I footprinting ( Figure  2B). As expected, VpsT bound the vpsL promoter region poorly in the absence of c-di-GMP (compare lanes 1 and 2).

Comparison of sequences protected from DNAse I digestion by VpsT
As noted above, prior studies pinpointed VpsT binding, at targets other than vpsL, by DNAse I footprinting (30,31). For comparison, we used the same approach to locate binding of VpsT at the aer and frhC regulatory regions identified by ChIP-seq (Table 1). At the aer locus, VpsT produced two distinct regions of DNA protection that overlapped the centre of the ChIP-seq binding signal (Supplementary Figure  S4). A similar pattern of binding was observed upstream of frhC, except that three sections of DNA were bound by VpsT (Supplementary Figure S5). None of the sequences protected by VpsT, here or in past work, match the T-box sequence identified by Zamorano-Sánchez et al (32). We reasoned that the T-box motif identified by ChIP-seq might allow us to resolve these apparent discrepancies. Hence, we collated all eight sequences bound by VpsT in footprinting experiments. We then searched for the ChIP-seq derived T-box motif within these sequences ( Figure 2C). In seven cases, an obvious match to the motif was found. In one instance (frhC T-box I) a motif was found but the most highly conserved positions, usually −3C and +3G, were inverted ( Figure 2C and Supplementary Figure S5B). Note that the T-box sequence motif suggests that this may occasionally be the case ( Figure 1E). We conclude that the ChIP-seq derived logo, generated with the benefit of more VpsT target site information, better describes the DNA binding properties of VpsT.

VpsT binds two T-boxes upstream of the vibrio phase variation operon
Definition of the T-box consensus should permit easier dissection of gene regulation by VpsT. Hence, we focused our attention on the vibrio phase variation (vpv) operon that contains three genes: vpvA, vpvB and vpvC. Importantly, whilst VpvA and VpvB are hypothetical proteins of unknown function, VpvC is a diguanylate cyclase contributing to the corrugated morphology of rugose phase variants   (47). The ChIP-seq data for VpsT binding at the vpvABC locus is shown in Figure 3A. The sequence of the intergenic region is shown in Figure 3B where the centre of the ChIPseq peak is indicated by an asterisk. Note that sequences are numbered according to distance from the 3 end of the DNA sequence. To precisely identify the location of VpsT binding we again used DNAse I footprinting ( Figure 3C).
In the absence of c-di-GMP no VpsT binding was detected (lanes 1-5). However, in the presence of c-di-GMP, two sections of the regulatory region were protected from DNAse I digestion (lanes 6-10). The regions of protection, each ∼20 bp in length, overlap the centre of the ChIP-seq peak and are indicated by cyan boxed sequences in Figure 3B. Examination of the sequences identified T-box motifs (bold in Figure 3B) that we named T-box I and T-box II. Note that T-box I has the same sequence inversion at positions −3 and +3 described above for frhC T-box I.

The regulatory region of the vibrio phase variation operon contains two promoters
To begin understanding the regulatory role of VpsT we searched for TSSs, upstream of vpvABC, using primer extension assays. To do this, the DNA sequence shown in Figure 3B was cloned upstream of lacZ in plasmid pRW50T. The resulting DNA construct was transferred into V. cholerae E7946 by conjugation. These cells were then transformed with pAMNF, or pAMNF-vpsT so that we could determine the consequence of VpsT expression. Finally, RNA was extracted and an oligonucleotide complementary to lacZ was used to prime reverse transcription. An electrophoretic analysis of the resulting cDNA products is shown in Figure 3D. In the absence of ectopic VpsT expression, we could detect a single cDNA ( Figure 3D, lane 1). This primer extension product indicates a TSS located at position 151 of the cloned vpvABC regulatory region (marked orange in Figure 3B). In the presence of VpsT, we could detect an additional cDNA species ( Figure 3D, lane 2). The second extension product results from a TSS located at position 61 of the cloned vpvABC regulatory region (marked purple in Figure 3B). We named the promoters associated with each TSS vpvAP1 and vpvAP2, according to their proximity to the vpvA start codon ( Figure 3B). We note that Papenfort and co-workers previously used dRNA-seq to map TSSs genome-wide in V. cholerae (48). This analysis, done using conditions not expected to induce VpsT controlled genes, identified only the vpvAP2 TSS. We conclude that vpvP1 is a strictly VpsT activated promoter. The constitutive activity of vpvAP2 is likely a result of better matches to promoter element consensus sequences (compare underlined sequences in Figure 3B). Briefly, the better −35 element of the vpvAP2 promoter is likely to reduce the need for a transcription factor to stabilise RNA polymerase binding (49).

Activation of vpvAP1 by VpsT requires the adjacent T-box
To further understand interplay between the promoter and T-box sequences we made derivatives of the vpvABC regulatory region. The DNA fragments are illustrated schematically in Figure 4A and the different 5 ends are marked by inverted triangles in Figure 3B. The various sequences were fused to lacZ in plasmid pRW50T and ␤-galactosidase activity was determined in lysates of V. cholerae strain E7946, with or without ectopic VpsT expression. As expected, the starting regulatory fragment was able to stimulate lacZ expression and ␤-galactosidase activity doubled when VpsT was produced ( Figure 4B, wild-type). To understand the contribution of vpvAP2 promoter, and the overlapping Tbox II, we removed 60 base pairs from the 5 end of the wild-type DNA fragment. The truncated DNA sequence, named vpvAP2, stimulated lower levels of ␤-galactosidase activity but the fold induction by VpsT was unaltered (Figure 4B, vpvAP2). Hence, activation of vpvAP1 by VpsT does not require T-box II. To confirm the importance of T-box I we altered the sequence of this DNA element in the context of the truncated DNA fragment. The mutations are indicated by red text in Figure 3B and the resulting DNA fragment was named vpvAP2 Mut . Mutation of T-box I abolished activation of vpvAP1by VpsT ( Figure  4B, vpvAP2 Mut ). In a final experiment, we deleted T-box I completely and activation by VpsT was similarly ablated ( Figure 4B, vpvAP2 T ).

The vpvAP1 promoter and T-box I are needed to maintain intracellular c-di-GMP levels
Previous work has shown that diguanylate cyclase activity generates c-di-GMP (13) and that this molecule positively regulates the activity of VpsT (20). In turn, we show that VpsT activates expression of vpvABC, which encodes a diguanylate cyclase. We reasoned that this should constitute a positive feedback loop dependent on T-box I. To test   this, we made alterations to the chromosomal vpvABC regulatory region. Specifically, we deleted a 60 bp section of chromosomal DNA, containing vpvAP2 and T-box II, before making a derivative of this strain with point mutations in T-box I. The chromosomal changes are equivalent to the vpvAP2 and vpvAP2 Mut constructs described above for lacZ fusion assays. We also examined the effect of deleting vpsT or vpvABC. To assess the consequences, we used a plasmid encoded intracellular c-di-GMP biosensor (46). Briefly, the system relies on expression of two different fluorescent reporters. One reporter, AmCyan, is produced constitutively and serves as a normalization control. The second reporter, TurboRFP, is regulated by a c-di-GMP activated riboswitch. Intracellular c-di-GMP levels are inferred by calculating the AmCyan:TurboRFP ratio. These ratios are shown in Figure 5 alongside images of representative colonies. Wild-type colonies have an orange appearance due to the combined expression of AmCyan and TurboRFP. Cells lacking vpsT formed colonies with a green-blue appearance due to greatly reduced c-di-GMP levels. Surprisingly, complete deletion of the vpvABC operon had no effect. Similarly, deletion of vpvAP2 did not induce a large change in c-di-GMP levels. Conversely, mutation of the vpv-ABC T-box I, required for activation of vpvAP1 by VpsT, reduced intracellular levels of c-di-GMP by 40%.

The vpvAP2 promoter is important for V. cholerae phase variation
In a final set of experiments, we wanted to understand the contribution of the different vpvABC operon regulatory elements to phase variation. Smooth to rugose phase variation in V. cholerae can be induced by growing cells in nutrient depleted media (34). We inoculated alkaline phosphate water using a single smooth colony of each strain. After incuba-tion, cells were plated and the number of colonies with the smooth or rugose phenotype were counted. The results are shown in Table 2. As expected, deletion of vpsT significantly reduced smooth to rugose phase variation by 4-fold. Conversely, complete deletion of vpvABC had no effect. Deletion of vpvAP2 caused a significant 2-fold increase in rugose colony numbers that did not change significantly upon deletion of T-box I needed to activate the remaining vpvAP1 promoter. We conclude that vpvAP2 impacts phase variation but vpvAP1 does not.

CONCLUSIONS
We argue that the T-box is best described as a degenerate palindrome with the consensus 5 - This resembles part of a prior T-box motif, generated from a smaller number of targets, and explains observations of VpsT binding diverse sequences ( Figure 2C) (30)(31)(32). Positions −3 and +3 of the motif are key and, whilst usually C and G respectively, exhibit the complimentary sequence in some cases ( Figures 2C and 3B). In both scenarios, mutation of the key positions prevents interaction with VpsT ( Figures 2B and 4B). Globally, the DNA binding properties of VpsT are intriguing. Despite recognising a highly degenerate sequence, VpsT targets just 23 loci in our ChIP-seq assay (Table 1). Furthermore, whilst vpsT resides within V. cholerae chromosome II, all VpsT targets locate to chromosome I. This is largely consistent with prior transcriptome analysis, which identified a similar skew in VpsT regulated genes (20). The unusual distribution is unlikely to be driven solely by occurrence of the T-box sequence; incidence of the motif can be identified frequently across both chromosomes (50). Hence, the ability of VpsT to bind DNA must be influenced by additional factors. Other regulators and DNA supercoiling could both play a role. We note that VpsT appears to bind DNA differently at certain targets. For instance, at the vpsL regulatory region, DNA protection stretched ∼60 base pairs upstream of the T-box (Figure 2). This protection cannot be due to VpsT independently binding multiple sites; a single T-box point mutation was sufficient to abolish the footprint. Hence, we suggest that VpsT must cooperatively oligomerise across, or tightly wrap, the vpsL regulatory DNA. The VpsT binding pattern is different upstream of vpvABC; a precise footprint of each T-box is observed. Interestingly, VpsT activates vpsL transcription by antagonising H-NS mediated repression (27,32). This is not the case for vpvABC, that does not bind H-NS (27). Hence, we speculate that the more extensive interactions of VpsT at the vpsL locus may be necessary to displace H-NS.
Our data suggest that the vpvAP1 and vpvAP2 promoters have different roles. The latter is constitutively active whilst vpvAP1 is activated by VpsT (Figure 3) (48). The constitutive activity of vpvAP2 is consistent with this promoter After 72 h incubation, cells were plated and the number of colonies with the smooth or rugose phenotype were counted. 3 The percentage of all colonies counted that had a rugose phenotype. 4 All P values were calculated using the Chi-square test. We compared the number of smooth and rugose colonies for each derivative to the parent strain. In most cases, the wild-type E7946 was the parent. The exception was the vpvAP2 Mut strain, where the vpvAP2 strain was the parent. The null hypothesis was that no differences between strains would be observed. impacting the rate of phase variation; rugose colonies can result from mutation of vpvC, basal expression of which requires vpvAP2 ( Figure 3D) (47,48). Conversely, the VpsT activated vpvAP1 promoter did not impact phase variation (Table 2). Hence, changes in rates of phase variation due to deletion of VpsT cannot be mediated via vpvABC. Instead, it is likely that these changes are due to some other factor subject to direct or indirect control by VpsT. By contrast, the VpsT activated vpvAP1 promoter impinged on cdi-GMP homeostasis. This is likely a result of a positive feedback loop; c-di-GMP activates VpsT that subsequently upregulates expression of VpvC and so diguanylate cyclase activity ( Figure 6). Hence, mutation of T-box I, needed to activate vpvAP1, led to a 40% decrease in intracellular c-di-GMP ( Figure 5). Deletion of vpsT caused the intracellular c-di-GMP signal to decrease dramatically ( Figure 5). However, given the smaller effect of mutating T-box I, this phenotype may depend on regulatory targets in addition to vpv-ABC. Indeed, as noted above, it is likely that loss of vpsT has many indirect effects caused by changes in overall cell physiology and levels of other transcription factors ( Figure 1D).

c-di-GMP VpsT
The stability of the intracellular c-di-GMP signal following vpvABC deletion suggests other DGC proteins might play a compensatory role. Indeed, V. cholerae encode thirty one proteins with the 'GGDEF' motif and redundancy is common (15,51). If such compensatory changes do occur, they are not evident when the vpvABC operon is downregulated by either promoter or T-box mutation ( Figure 5 and Table 2). Hence, low level vpvABC expression provokes phenotypic effects reminiscent of a dominant negative allele.
We also note that, as the roles of vpvA and vpvB are less well defined, their deletion could have confounding consequences. Previous work using V. cholerae strain A1552 also examined the role of the vpvABC operon (47). Consistent with our findings, phenotypic effects were observed in rugose cells, encoding the vpvC W240R allele, but not smooth parent strain (46, and F. Yildiz personal communication).
In conclusion, VpsT both relays the c-di-GMP signal and is important for the control of intracellular c-di-GMP levels.

DATA AVAILABILITY
Original DNA sequencing data are available from ArrayExpress using the accession numbers provided above. Original gel mages are provided as supplementary material. All other data are available from the authors upon request.

SUPPLEMENTARY DATA
Supplementary Data are available at NAR Online.