Mechanism underlying the DNA-binding preferences of the Vibrio cholerae and vibriophage VP882 VqmA quorum-sensing receptors

Quorum sensing is a chemical communication process that bacteria use to coordinate group behaviors. In the global pathogen Vibrio cholerae, one quorum-sensing receptor and transcription factor, called VqmA (VqmAVc), activates expression of the vqmR gene encoding the small regulatory RNA VqmR, which represses genes involved in virulence and biofilm formation. Vibriophage VP882 encodes a VqmA homolog called VqmAPhage that activates transcription of the phage gene qtip, and Qtip launches the phage lytic program. Curiously, VqmAPhage can activate vqmR expression but VqmAVc cannot activate expression of qtip. Here, we investigate the mechanism underlying this asymmetry. We find that promoter selectivity is driven by each VqmA DNA-binding domain and key DNA sequences in the vqmR and qtip promoters are required to maintain specificity. A protein sequence-guided mutagenesis approach revealed that the residue E194 of VqmAPhage and A192, the equivalent residue in VqmAVc, in the helix-turn-helix motifs contribute to promoter-binding specificity. A genetic screen to identify VqmAPhage mutants that are incapable of binding the qtip promoter but maintain binding to the vqmR promoter delivered additional VqmAPhage residues located immediately C-terminal to the helix-turn-helix motif as required for binding the qtip promoter. Surprisingly, these residues are conserved between VqmAPhage and VqmAVc. A second, targeted genetic screen revealed a region located in the VqmAVc DNA-binding domain that is necessary to prevent VqmAVc from binding the qtip promoter, thus restricting DNA binding to the vqmR promoter. We propose that the VqmAVc helix-turn-helix motif and the C-terminal flanking residues function together to prohibit VqmAVc from binding the qtip promoter.

Introduction Quorum sensing (QS) is a cell-cell communication process that allows bacteria to coordinate collective behaviors [1]. QS relies on the production, release, and group-wide detection of extracellular signaling molecules called autoinducers (AIs). In the global pathogen Vibrio cholerae, the AI, 3,5-dimethyl-pyrazin-2-ol (DPO), together with its partner cytoplasmic QS receptor and transcription factor, VqmA (VqmA Vc ), comprises one of the QS circuits that controls group behaviors [2][3][4]. VqmA Vc , following binding to DPO, activates transcription of the vqmR gene encoding the small RNA, VqmR, which, in turn, represses the expression of genes required for biofilm formation and virulence factor production [2][3][4].
Recently, bacteria-specific viruses, called phages, have been shown to engage in densitydependent regulation of their lysis-lysogeny decisions via chemical dialogs [5,6]. Germane to our studies are phages that encode proteins resembling bacterial QS components [5,7]. Vibriophage VP882 is one such phage: It encodes the QS receptor VqmA (VqmA Phage ), a homolog of the V. cholerae QS receptor VqmA Vc [5]. VqmA Phage , like VqmA Vc , binds hostproduced DPO. DPO-bound VqmA Phage activates transcription of the phage gene qtip. Qtip is an antirepressor that sequesters the phage VP882 repressor of lysis, leading to derepression of the phage lytic program and killing of the Vibrio host at high cell density [5,8]. Thus, the DPO AI mediates both bacterial and phage lifestyle decisions. Curiously, VqmA Phage can substitute for VqmA Vc to activate the V. cholerae vqmR promoter (PvqmR) [5]. In contrast, VqmA Vc cannot substitute for VqmA Phage and recognize the phage VP882 qtip promoter (Pqtip). Presumably, the ability of VqmA Phage to bind both PvqmR and Pqtip provides phage VP882 the capacity to influence host QS and simultaneously enact its own lysis-lysogeny decision.
VqmA Phage shares~43% amino acid sequence identity with VqmA Vc , and most of the key residues required for ligand and DNA binding are conserved [5,9]. Thus, how VqmA Phage can recognize two different promoters, while VqmA Vc cannot, is not understood. Here, we define the mechanism underlying this asymmetry. We show that VqmA selectivity for target promoters is driven by the DNA-binding domain (DBD) of the respective protein. We identify 6 key nucleotides within PvqmR and Pqtip that contribute to VqmA promoter-binding selectivity, as exchanging these critical DNA sequences inverts the DNA-binding preferences of the two VqmA proteins. The 192 nd and 194 th residues in VqmA Vc and VqmA Phage , respectively, within the helix-turn-helix (HTH) motifs, contribute to promoter-binding specificity. Isolation of VqmA Phage mutants capable of activating vqmR expression but incapable of activating qtip expression revealed conserved or functionally conserved residues in VqmA Phage and VqmA Vc , indicating that VqmA Vc likely possesses an additional feature that prevents it from binding Pqtip DNA. A mosaic VqmA Vc protein containing the VqmA Phage HTH motif along with the C-terminal 25 flanking VqmA Phage residues was capable of binding Pqtip. Thus, the two corresponding regions in VqmA Vc must function in concert to prevent VqmA Vc from binding to Pqtip. Together, our analyses demonstrate how VqmA Phage , via its promiscuous DNA-binding activity, can control phage VP882 functions and drive host V. cholerae QS. Moreover, we discover why V. cholerae VqmA Vc cannot do the reverse, as its DNA binding is strictly constrained to the host V. cholerae genome.

VqmA promoter-binding selectivity is conferred by the DNA-binding domain
VqmA proteins are composed of N-terminal Per-Arnt-Sim (PAS) domains responsible for binding the DPO AI and C-terminal DBDs containing HTH motifs [10]. Both VqmA Vc and VqmA Phage bind DPO. By contrast, with respect to DNA binding, VqmA Phage binds to Pqtip and PvqmR, whereas VqmA Vc only binds to PvqmR [5]. We reasoned that this asymmetric DNA-binding pattern arises from differences in the DBDs (S1 Fig). To test this idea, we constructed chimeras in which we exchanged the VqmA Vc and VqmA Phage C-terminal domains to produce Vc N-C Phage and Phage N-C Vc proteins. We chose to make the junction at a residue near the C-terminal end of the PAS domain immediately following an amino acid stretch (GTIF) that is identical in both VqmA Vc and VqmA Phage (S1 Fig). We cloned vqmA Vc , vqmA Phage , Vc N-C Phage , and Phage N-C Vc under an arabinose-inducible promoter and transformed each construct into recombinant Δtdh E. coli harboring a PvqmR-lux or a Pqtip-lux reporter. The Tdh enzyme is required for DPO biosynthesis, therefore a Δtdh E. coli strain makes no DPO [3]. Apo-VqmA displays basal transcriptional activity in vivo [9]. Thus, while DPO enhances VqmA DNA-binding activity, it is not an absolute requirement for binding. Using Δtdh E. coli for these studies ensured that any transcriptional activity that occurred was exclusively a consequence of the DNA-binding capabilities of the chimeras and not ligand-binding-driven transcriptional activation of the chimeras. Consistent with our hypothesis, promoter activation by each chimera was determined by the protein from which the DBD originated: All four versions of VqmA activated PvqmR-lux, whereas only VqmA Phage and Vc N-C Phage activated Pqtip-lux ( Fig 1A and 1B, respectively). Next, we conjugated the four versions of VqmA into Δtdh ΔvqmA Vc V. cholerae lysogenized by a phage VP882 mutant in which the endogenous vqmA Phage was inactive (VP882 vqmA Phage ::Tn5). Thus, the only source of VqmA protein was that made from the plasmid. As expected, following arabinose-induction, only VqmA Phage and Vc N-C Phage activated qtip expression and induced host-cell lysis (Fig 1C).
We verified the above findings in vitro using electrophoretic mobility shift assays (EMSAs). Consistent with the cell-based assays, the purified VqmA Vc , VqmA Phage , Vc N-C Phage , and Phage N-C Vc proteins shifted PvqmR DNA, whereas only the VqmA Phage and Vc N-C Phage proteins shifted Pqtip DNA (Fig 1D). Assessing the ratios of bound to total DNA across varying protein concentrations allowed us to calculate the relative binding affinities (EC 50 ) of the VqmA proteins for PvqmR and Pqtip DNA (S2A Fig). Our EMSA analyses show that Phage N-C Vc , like VqmA Vc , only bound PvqmR, but with an estimated~7-fold lower affinity. Consistent with our previous findings, VqmA Phage bound Pqtip about 3-fold more strongly than it bound PvqmR [5]. By contrast, Vc N-C Phage showed a modest increase in its preference for Pqtip relative to that for PvqmR, with binding to both promoters at a level similar to that with which VqmA Phage bound Pqtip. Indeed, in agreement with our EC 50 measurements, when Pqtip and PvqmR DNA were supplied at equimolar concentrations in a competitive

PLOS GENETICS
DNA-binding preferences of the Vibrio cholerae and vibriophage VP882 VqmA quorum-sensing receptors DNA-binding assay, lower amounts of VqmA Phage and Vc N-C Phage were required to shift Pqtip DNA than to shift PvqmR DNA (S2B Fig). In conclusion and in agreement with our in vivo results, the respective DBD of each purified VqmA protein drives promoter selectively.
We next assayed the VqmA Vc and VqmA Phage DBDs lacking their PAS domains (DBD Vc and DBD Phage , respectively) for activation of PvqmR-lux and Pqtip-lux. Deletion of the PAS domains resulted in inactive proteins as neither DBD activated transcription (S3A and S3B  Fig, respectively), and likewise, EMSA analyses showed that neither DBD bound either promoter (S3C Fig). Gel filtration analyses indicated that the DBD proteins purified as monomers (S3D Fig), suggesting that the DBDs were unable to dimerize in the absence of their partner PAS domains. This result is consistent with previous findings that, in addition to sensing DPO, the VqmA Vc PAS domain is responsible for dimerization [9,11].
Transcriptional activity driven by HTH-containing proteins typically depends on dimer formation. Soluble glutathione S-transferase (GST) spontaneously forms a homodimer [12], and so GST can be employed as a substitute for native dimerization domains of proteins [13]. Thus, to examine the VqmA requirement for dimerization, we fused GST to the N-terminus of each VqmA DBD to yield recombinant GST-DBD Vc and GST-DBD Phage and we tested whether DNA-binding function was restored. Indeed, the GST-DBD proteins purified as dimers (S3D Fig). PvqmR-lux and Pqtip-lux expression analyses revealed that the DBDs, when fused to GST, regained function, with the caveat that the GST-DBD Vc exhibited 10-fold reduced activity compared to wild-type (WT) VqmA Vc ( Fig 1E). Importantly, the DNA-binding preferences mimicked those of the full-length proteins: GST-DBD Phage activated both PvqmR-lux and Pqtip-lux, whereas GST-DBD Vc only activated PvqmR-lux (Fig 1E and 1F). Companion EMSA analyses showed that GST-DBD Phage bound Pqtip~5-fold more strongly than it bound PvqmR, whereas GST-DBD Vc showed almost no binding to PvqmR and, unexpectedly, some weak binding could be detected to the Pqtip DNA ( Fig 1G). We confirmed that purified GST alone did not bind either PvqmR or Pqtip (S3E Fig). Given that the GST-DBD Vc driven activation of Pqtip-lux was undetectable in vivo (Fig 1F), we presume that the observed in vitro GST-DBD Vc binding to Pqtip DNA is a consequence of the simplified context in which the EMSA is performed. Likely, the DNA:VqmA ratio in the EMSA is far higher than in cells, which, in the case of GST-DBD Vc , fosters modest non-specific DNA binding. Taken together, our results show that VqmA promoter-binding selectivity is conferred by the DBD, and that dimerization is necessary.

VqmA DNA-binding preferences can be inverted by exchanging key DNA sequences in PvqmR and Pqtip
To study the VqmA promoter-binding asymmetry from the aspect of the DNA, our next goal was to identify the critical DNA sequence within Pqtip that prevents VqmA Vc from binding. In the phage VP882 genome, Pqtip resides between vqmA Phage and qtip and VqmA Phage activates its own and qtip expression, suggesting that VqmA Phage binding may involve both DNA strands. Similarly, VqmA Vc has been shown to interact with both strands of PvqmR [11]. Thus, in each case, both DNA strands need to be considered (Fig 2A). Previous work revealed that the critical region in PvqmR required for VqmA Vc binding is -AGGGGGGATTTCCCCCCT- [2,11]. The corresponding fragment from Pqtip, but on the opposite DNA strand, -TAGGGG GAAAAATACCCT-, possesses~56% sequence identity to this region suggesting it could be the key stretch of DNA that drives VqmA Phage promoter selection. The highest divergence in the two promoters is in the central 6 nucleotides: "-AAAATA-" in Pqtip and "-TTTCCC-" in PvqmR. We synthesized DNA probes in which we exchanged the "-AAAATA-" in Pqtip with "-TTTCCC-" from PvqmR and tested VqmA Vc and VqmA Phage binding by EMSA analysis.

PLOS GENETICS
We call these probes PvqmR � and Pqtip � , respectively. Indeed, promoter DNA-binding specificity was exchanged: VqmA Vc shifted Pqtip � , whereas it only weakly shifted PvqmR � (Fig 2B). VqmA Phage bound to PvqmR � twice as strongly as it bound to Pqtip � , showing the opposite preference for the two synthetic promoters compared to the native promoters ( Fig 2B). PvqmR � -lux and Pqtip � -lux transcriptional fusions mimicked the EMSA results: VqmA Vc only activated expression of Pqtip � -lux, whereas VqmA Phage activated expression of PvqmR � -lux and Pqtip � -lux (Fig 2C and 2D). Thus, this 6-nucleotide stretch is the key sequence that determines the DNA-binding specificity for the two VqmA proteins. Moreover, the presence of the -AAAATA-nucleotide sequence in Pqtip is sufficient to prevent VqmA Vc from activating transcription of Pqtip.
Protein sequence-guided mutagenesis reveals that residue E194 in phage VP882 VqmA Phage and the equivalent A192 residue in V. cholerae VqmA Vc contribute to specificity for Pqtip We considered two possible mechanisms that could underpin the asymmetric VqmA DNAbinding patterns: phage VP882 VqmA Phage could possess a feature that relaxes its DNA-binding specificity, and/or V. cholerae VqmA Vc could possess a feature that restricts its DNA-binding ability. To distinguish between these possibilities, we first probed which residues drive VqmA Phage interactions with Pqtip but do not contribute to interactions with PvqmR. To do this, we performed site-directed mutagenesis of VqmA Phage with the goal of identifying mutants that fail to bind Pqtip but retain binding to PvqmR. Charged residues in HTH motifs typically mediate interactions between VqmA-type transcription factors and DNA, and indeed, both VqmA HTHs are enriched in positively-charged amino acids [9,11,14]. Sequence alignment of the HTHs in VqmA Phage and VqmA Vc revealed four obvious differences in charged residues that could underlie the DNA-binding asymmetry between the two proteins (S1 Fig). We mutated those residues in VqmA Phage to the corresponding VqmA Vc residues. VqmA Vc variants failed to do so ( Fig 3D). All of the VqmA Vc variants drove the WT level of PvqmR-lux activity ( Fig 3E). VqmA Vc A192E generated low but detectable Pqtip-lux expression, while the other VqmA Vc variants did not ( Fig 3F). The VqmA Vc variants were produced at similar levels to WT VqmA Vc in V. cholerae and E. coli (S4B Fig). We conclude that, among the tested residues, only A192 plays a role in preventing VqmA Vc from binding Pqtip.

PLOS GENETICS
Our mutagenesis analyses for VqmA Vc are consistent with our analyses for VqmA Phage : The residue at the 192 nd position in V. cholerae VqmA Vc and the analogous residue at the 194 th position in phage VP882 VqmA Phage contribute to selection of Pqtip. However, given that the A192E substitution in VqmA Vc results in only partial activation of Pqtip expression, and the E194A substitution in VqmA Phage results in only partial loss of activation of Pqtip, the E194 residue in VqmA Phage cannot be the sole amino acid responsible for the preference VqmA Phage shows for Pqtip. Rather, additional residues in VqmA Phage must participate in conferring specificity.
Random mutagenesis of the VqmA Phage DBD reveals that residues G201, A202, E207, and M211 are required for VqmA Phage to bind Pqtip but are dispensable for binding PvqmR Our protein sequence-guided approach did not reveal the primary mechanism underlying promoter-binding specificity for either of the VqmA proteins. We therefore performed a genetic screen to forward our goal of identifying phage VP882 VqmA Phage mutants that fail to bind Pqtip but retain the ability to bind PvqmR. We constructed a library of random mutations in the region of vqmA Phage encoding the DBD in the context of the full-length gene, cloned them into a plasmid under an arabinose-inducible promoter, and introduced them into Δtdh ΔvqmA Vc V. cholerae harboring PvqmR-lux on the chromosome and lysogenized by phage VP882 harboring inactive vqmA Phage (vqmA Phage ::Tn5). The logic of the screen is as follows: When propagated on agar plates supplemented with arabinose, V. cholerae exconjugants harboring vqmA Phage alleles possessing reasonable Pqtip-binding activity will lyse because those VqmA Phage proteins will bind Pqtip on the phage VP882 genome and launch the phage lytic cascade (S5 Fig). Such exconjugants will die and thus be eliminated from the screen. Exconjugants that survive but carry vqmA Phage null alleles will produce no light because those VqmA Phage proteins will fail to bind PvqmR-lux, so they also can be eliminated from the screen. The vqmA Phage alleles of interest to us are those that are maintained in surviving exconjugants (because they encode proteins that cannot bind Pqtip) and produce light (because they encode proteins that can bind PvqmR-lux).
Our screen yielded the following mutants: VqmA Phage , and VqmA Phage M211K (Fig 4A). To verify that these VqmA Phage mutants were indeed defective in binding Pqtip, we individually transformed them into Δtdh E. coli carrying the Pqtip-lux reporter or the PvqmR-lux reporter and measured light production. All variants retained WT capability to activate PvqmR-lux, but they did not harbor WT capability to activate Pqtip-lux expression (>10-fold reductions in activity) (Fig 4B and 4C, respectively). Thus, any residual Pqtip binding by these mutant VqmA Phage proteins is insufficient to induce host-cell lysis in the phage VP882 lysogen ( Fig  4A). , like WT VqmA Vc , activated PvqmR-lux but failed to activate Pqtip-lux (S6A and S6B Fig, respectively). We make the following four conclusions from these findings: 1) There are at least four residues (G201, A202, E207, and M211) required for VqmA Phage to recognize Pqtip DNA. 2) Because the VqmA Phage G201D, G201R, A202V, E207K, E207V, and M211K variants exhibit WT binding to PvqmR, the substitutions at these four residues must not significantly affect PvqmR recognition. 3) Because these residues are conserved or similar between VqmA Phage and VqmA Vc , one would expect VqmA Vc to have the capacity to bind Pqtip. 4) However, because VqmA Vc in fact does not bind Pqtip, VqmA Vc likely possesses an additional feature that resides elsewhere in the protein that prevents Pqtip binding from occurring.

The restrictive element that prevents VqmA Vc from binding Pqtip is located in its HTH motif and the adjacent C-terminal region of 25 residues
To test the hypothesis that a feature in the VqmA Vc DBD restricts its DNA-binding capacity to PvqmR, we performed a genetic screen aimed at identifying VqmA Vc mutants capable of activating Pqtip-lux expression. To do this, we constructed a library of random vqmA Vc DBD alleles containing, on average, 1-2 substitutions, and we cloned them into a plasmid under an arabinose-inducible promoter. The library was transformed into the Δtdh E. coli strain harboring the Pqtip-lux reporter and transformants were propagated on plates containing arabinose. We screened~10,000 transformants for colonies that produced light indicating that they contained VqmA Vc proteins that activated Pqtip-lux. This strategy yielded no such transformants. Several possibilities could explain our result: We did not screen sufficient numbers of mutants, the mutagenesis did not yield the crucial change, or no alteration of a single residue can enable VqmA Vc binding to Pqtip.
We expanded our search for the DNA-binding restrictive element present in VqmA Vc by assessing whether a particular region in the VqmA Vc DBD constrains promoter binding to PvqmR. To do this, we constructed five VqmA Vc mosaic proteins by replacing~20-30 residues in the V. cholerae VqmA Vc DBD with the corresponding residues from the phage VP882 VqmA Phage DBD. We call these proteins VqmA Vc

�126-149
, VqmA Vc �150-170 , VqmA Vc �171-199 , VqmA Vc �200-224 , and VqmA Vc �225-246 (see S1 Fig for relevant protein segments). Each superscript denotes the VqmA Vc amino acid residues that have been replaced by the corresponding residues from VqmA Phage . In all the mosaics, either the intact VqmA Vc HTH or the intact VqmA Phage HTH was present. For reference, the VqmA Vc HTH motif consists of residues 173 to 198 and the VqmA Phage HTH spans residues 175 to 200. We tested the mosaic VqmA Vc proteins for activation of the PvqmR-lux and Pqtip-lux reporters. The DNA specificity of all the VqmA Vc mosaics mimicked WT VqmA Vc as PvqmR-lux was expressed but Pqtip-lux was not (Fig 5A and 5B, respectively). We confirmed that the mosaic VqmA Vc proteins are expressed at levels similar to WT VqmA Vc (S7 Fig). Our results suggest that the feature that prevents V. cholerae VqmA Vc from binding to Pqtip is larger than the regions delineated by any of the VqmA Vc mosaics, or it could be that multiple patches in the VqmA Vc DBD that are not contiguous in amino acid sequence are responsible.
Pinpointing non-contiguous regions that could, together, contain the VqmA Vc restrictive element is challenging. However, testing for a larger contiguous expanse that could contain the putative restrictive element is straightforward. Thus, we constructed two additional V. cholerae VqmA Vc mosaic proteins. In one construct, called VqmA Vc �150-199 , we introduced the VqmA Phage HTH along with the immediate N-terminal 25 amino acids in place of the corresponding VqmA Vc region. Second, in a construct called VqmA Vc

PLOS GENETICS
DNA-binding preferences of the Vibrio cholerae and vibriophage VP882 VqmA quorum-sensing receptors

PLOS GENETICS
DNA-binding preferences of the Vibrio cholerae and vibriophage VP882 VqmA quorum-sensing receptors WT VqmA Vc , eliminating the possibility that the observed binding to Pqtip was a consequence of overexpression (S7 Fig). We conclude that the region encompassing both the HTH motif and the C-terminal 25 residues are required to restrict the VqmA Vc DBD from binding Pqtip.

Discussion
The DPO-VqmA QS AI-receptor pair controls lifestyle transitions in the pathogen V. cholerae and in the vibriophage VP882. Here, we studied the DNA-binding function of VqmA. VqmA proteins are cytoplasmic transcription factors composed of N-terminal PAS domains responsible for binding the DPO ligand and C-terminal DBDs containing HTH motifs. Most of the key residues required for binding the DPO ligand and for binding to PvqmR DNA are conserved between the two VqmA proteins. Indeed, both VqmA Vc and VqmA Phage bind DPO and activate transcription of vqmR. By contrast, only VqmA Phage activates the phage gene qtip.
Here, we investigated this asymmetric DNA-binding pattern. Our work shows that, in both proteins, the DBD determines promoter recognition. We have previously shown that DPO binding enhances VqmA transcriptional activity [9]. This earlier work, together with our present results, suggest a model in which the PAS domain specifies DNA-binding affinity (between the apo-and holo-states), and the DBD specifies DNA-binding selectivity.
The main goal of the present work was to discover features of the VqmA proteins that confer specificity in transcriptional activity. We propose that phage VP882 VqmA Phage possesses a feature that relaxes its DNA-binding specificity and V. cholerae VqmA Vc possesses a feature that restricts its DNA-binding capability. Regarding VqmA Vc , our genetic analyses support the hypothesis that the VqmA Vc DBD harbors elements that prevent it from binding Pqtip. This hypothesis stems from our finding that residues G201, A202, E207, and M211 are crucial for VqmA Phage recognition of Pqtip. These residues are conserved between VqmA Vc and VqmA Phage . Specifically, in VqmA Vc they are: G199, A200, Q205, and L209, respectively. More broadly, sequence alignments of VqmA proteins among Vibrios reveal that the residue at the 207 th position in VqmA Phage (205 th position in VqmA Vc ) is most frequently either a Glu or a Gln [5]. Similarly, the residue at the 211 th position in VqmA Phage (209 th position in VqmA Vc ) is commonly a hydrophobic residue, like Met, Leu, Ile, or Val. Thus, E207 and M211 are not unique to VqmA Phage , but rather occur in most VqmA proteins. We propose that because the key residues for Pqtip binding are conserved in VqmA Phage , VqmA Vc , and other Vibrio VqmA proteins, VqmA Vc is likely restricted from binding Pqtip by additional features elsewhere in its DBD. Regarding VqmA Phage , the DPO-VqmA Phage structure was reported during review of this manuscript [15]. Superimposition of this new structure (7DWM) onto the DPO-VqmA Vc and DPO-VqmA Vc -PvqmR structures (6KJU and 6IDE, respectively, and [9,11,14]) reveals two insights (S8 Fig). First, the conformations of the three PAS domains are similar except for the orientations of the first 20 N-terminal residues in each protein, indicating that the PAS domains do not confer the differences in promoter DNA specificity. Second, the DPO-VqmA Phage DBDs adopt a conformation that is intermediate between that of the more open DBDs in the DPO-VqmA Vc structure and the closed DBDs in the DPO-VqmA Vc -PvqmR structure. Additionally, the interaction interface between the VqmA Phage DBDs is less extensive, and thus more relaxed than that of the VqmA Vc DBDs [15]. Likely, the more relaxed conformation exhibited by the VqmA Phage DBDs underpins its promiscuity for promoter binding with respect to PvqmR and Pqtip.
In the case of VqmA Phage , the residues G201, A202, E207, and M211 identified in our mutagenesis screen as necessary for Pqtip binding are, surprisingly, not in the HTH motif, nor do the corresponding VqmA Vc residues make direct contacts with DNA in the DPO-VqmA Vc -PvqmR crystal structure (Fig 4D). Thus, we wonder how the G201, A202, E207, and M211 residues could govern recognition of Pqtip. Our in vivo analyses showed that substitutions in VqmA Phage at these residues enable activation of vqmR expression to WT levels, whereas only residual activation of qtip expression occurs (Fig 4A-4C). Surprisingly, the purified VqmA Phage mutant proteins maintained some capability to bind Pqtip in vitro. A representative experiment using the VqmA Phage G201D protein is shown in S9A Fig. We consider several possibilities to explain our findings: First, the VqmA Phage G201, A202, E207, and M211 residues could mediate interactions with an additional bacterial factor involved in transcription. Importantly, the failure of these VqmA Phage variants to activate Pqtip expression in V. cholerae lysogens also occurred in E. coli, eliminating the possibility that these residues interact with a phage-specific or Vibrio-specific factor. Rather, these residues could be important for coordinating interactions with a conserved factor, such as RNA polymerase. If so, these mutant VqmA Phage proteins, while capable of binding promoter DNA, are incapable of activating transcription. This situation would be analogous to the positive control mutants of the lambda phage cI repressor (cI lambda ). So called pc mutants bind DNA and exhibit repressor activity, but are deficient in positive transcriptional regulation due to the inability of the mutant cI lambda proteins to productively interact with RNA polymerase [16,17]. In our case, the VqmA Phage mutants maintain the capacity to activate vqmR expression so they must successfully interact with RNA polymerase at least at PvqmR. For this reason, we consider it unlikely that these VqmA Phage mutants are analogous to lambda pc mutants.
Second, a global transcriptional regulator could be involved that is present in both V. cholerae and E. coli. One candidate is the histone-like nucleoid structuring protein (H-NS) that functions as a universal repressor of transcription [18]. In Vibrio harveyi, the QS master regulator, LuxR, displaces H-NS at promoter DNA to activate expression of QS-controlled genes [19]. Perhaps, the VqmA Phage G201, A202, E207, and M211 mutants cannot successfully compete with H-NS for binding at Pqtip in vivo, whereas in an EMSA assay, since H-NS is not present, binding to Pqtip DNA occurs. To address this possibility, we examined whether WT VqmA Phage and VqmA Phage G201D competed with H-NS for binding to Pqtip using EMSA assays. There was no difference between WT VqmA Phage and VqmA Phage G201D binding to Pqtip DNA in the presence of purified H-NS (S9C and S9D Fig). These experiments suggest that it is unlikely that H-NS competition underlies our findings. Third, the binding of the VqmA Phage G201, A202, E207, and M211 mutants to Pqtip in vitro, while demonstrating loss of activity in vivo, could be a consequence of the unnaturally high DNA: VqmA Phage stoichiometry in the EMSA, similar to what we observed for the GST-DBD Vc construct ( Fig 1G). Thus, the EMSA is not sufficiently sensitive to distinguish between the strength of DNA binding of WT VqmA Phage and the residual binding by the VqmA Phage G201, A202, E207, and M211 mutants. If this is the case, we propose that VqmA Phage G201, A202, E207, and M211 could play allosteric roles in correctly positioning the VqmA Phage HTH for proper contact with particular DNA nucleotides. Here, we compare this possibility to how site-specific recognition is accomplished by cI lambda . Genetic and biochemical studies revealed that residues outside of the cI lambda HTH motif are crucial for sitespecific DNA recognition [20][21][22][23][24]. The crystal structure of the cI lambda repressor bound to DNA shows that charged residues adjacent to those in the HTH interact with the DNA sugar phosphate backbone [25]. Additionally, the N-terminal arm of cI lambda wraps around the DNA and makes contacts on the backside of the helix [25]. It is presumed that the backbone contacts function to position the HTH residues to contact specific DNA nucleotides. Thus, while the VqmA Phage residues that we identified as important for Pqtip recognition (G201, A202, E207, and M211) do not function perfectly analogously to those in cI lambda because they do not make contact with the DNA backbone, their role in site-specific recognition could be similar. A caveat of our interpretation is that, as noted, we do not have a structure of VqmA Phage bound to Pqtip and we mapped the residues identified in our VqmA Phage mutagenesis to the DPO-VqmA Vc -PvqmR crystal structure. Therefore, it remains possible that the residues we identified here do indeed make contacts with DNA. A further possibility is that the residues we identified foster increased plasticity to the VqmA Phage DBDs, perhaps, allowing VqmA Phage to bind the longer palindrome that exists in Pqtip, which we discuss below. The recently reported DPO-VqmA Phage crystal structure [15], together with the existing DPO-VqmA Vc structures, could enable modeling to predict the roles played by particular residues in conferring a relaxed conformation to the VqmA Phage DBDs. To our knowledge, no region analogous to the one we discovered in VqmA Phage has been shown to confer promoter specificity to a transcription factor. Going forward, determining the structure of VqmA Phage bound to Pqtip DNA should reveal the mechanism enabling recognition of Pqtip and the role that these residues play, individually and collectively, in determining DNA-binding specificity.
Previous work demonstrated that VqmA Vc recognizes a key GG-N 6 -CC palindrome in PvqmR [2,11]. Our sequence alignment of PvqmR and Pqtip showed that Pqtip does not possess this palindrome. Rather, the corresponding sequence in Pqtip is GG-N 6 -TA (Fig 2A). The most obvious divergence between the two sequences is in the central six nucleotides: "-AAAATA-" in Pqtip and "-TTTCCC-" in PvqmR (Fig 2A). We hypothesized that this nucleotide stretch could be responsible for conferring the asymmetric DNA-binding patterns to the two VqmA proteins. Indeed, exchanging these nucleotides in Pqtip and PvqmR reversed the promoter binding preferences of the VqmA proteins. We verified our conclusion that this core 6 nucleotide stretch drives VqmA DNA-binding preference using our VqmA chimeric proteins ( Vc N-C Phage and Phage N-C Vc ), a representative mosaic protein (VqmA Vc �171-224 ), and a representative protein containing a point mutation (VqmA Phage G201D ) (S9A, S9B, and S10 Figs). While the present manuscript was under review, Gu et al. reported that a GG-N 9 -CC palindrome in Pqtip is the key sequence for VqmA Phage recognition [15]. According to our DNA sequence alignment, the GG-N 6 -CC palindrome required for VqmA Vc binding is only present in PvqmR, while the key GG-N 9 -CC palindrome required for VqmA Phage binding exists in both Pqtip and PvqmR (Fig 2A). Together, our results and those of Gu et. al. [15] explain, at the level of the promoter DNA, why VqmA Phage binds both Pqtip and PvqmR while VqmA Vc recognizes only PvqmR.
Genomic sequencing data have revealed the presence of many QS receptor-transcription factors encoded in phage genomes [26]. In general, however, their transcriptional outputs are uncharacterized, with the exception of VqmA Phage , which is promiscuous with respect to binding to PvqmR and Pqtip, the only two promoters tested to our knowledge. It remains possible that VqmA Phage regulates additional genes specifying bacterial and or/phage functions. Given that VqmA Phage can regulate biofilm formation through its control of V. cholerae vqmR, probing the host regulon controlled by VqmA Phage under various growth conditions could reveal unanticipated roles of QS in phage-Vibrio interactions.
Finally, we found that the VqmA Vc A192E variant exhibited modest, but detectable binding to Pqtip, whereas the VqmA Vc quadruple mutant, and the VqmA Vc �171-199 mosaic protein did not. Western blot and PvqmR-lux assays eliminated the possibility that any of the mutant proteins were not expressed or were misfolded. Rather, we infer that a particular regional conformation in the VqmA proteins is required for this key residue to function properly. Our results also show that exchanging both the VqmA Vc HTH motif and C-terminal 25 residues with the corresponding residues from VqmA Phage enables some but not WT-level binding to Pqtip. This finding supports the notion that a set of non-contiguous amino acids or a particular conformation of the VqmA Vc DBD prevents binding to Pqtip. This arrangement is perhaps not surprising given that V. cholerae would pay a significant penalty if VqmA Vc bound the phage VP882 qtip promoter, as the consequence would be the launch of the phage lytic program and death of the host cell. To our knowledge, VqmA Vc binds to only one promoter, PvqmR [3]. Thus, even in the context of the V. cholerae genome, VqmA Vc transcriptional activity is tightly constrained. It is possible that other negative ramifications stem from non-specific VqmA Vc binding in the V. cholerae genome. Distinct mechanisms are employed to restrict other QS receptor/transcription factors from promiscuously binding to DNA. For example, LuxR-type QS receptors can typically bind >100 promoters, but their solubilization, stability, and DNAbinding capabilities strictly rely on being bound to an AI whose availability is, in turn, highly regulated [27][28][29][30][31]. Therefore, precise control of gene expression is maintained in many QS circuits by confining QS receptor activity to the ligand-bound form coupled with discrete affinities of the ligand-receptor complexes for target promoters. By contrast, VqmA Vc is expressed constitutively, and its DNA-binding capabilities are not limited by the presence of an AI. Thus, exquisitely tight control over promoter DNA-binding specificity by VqmA Vc -restricting it to one and only one promoter-is apparently crucial for proper regulation of gene expression and survival.
Primers were obtained from Integrated DNA Technologies. Gibson assembly, intramolecular reclosure, and traditional cloning methods were employed for all cloning. PCR with Q5 High Fidelity Polymerase (NEB) was used to generate insert and backbone DNA. Gibson assembly relied on HiFi DNA assembly mix (NEB). All enzymes used in cloning were obtained from NEB. Mutageneses of the VqmA Phage and VqmA Vc DBDs were accomplished using the GeneMorph II EZClone Domain Mutagenesis Kit (Agilent) according to the manufacturer's instructions. Transfer of plasmids carrying vqmA genes into the V. cholerae phage VP882 lysogen employed conjugation followed by selective plating on polymyxin B, chloramphenicol, and kanamycin, based on previously described protocols [32].

Genetic screens for VqmA Phage and VqmA Vc DNA-binding mutants
E. coli carrying a library of plasmid-borne vqmA Phage mutants was mated with V. cholerae harboring a phage VP882 mutant (vqmA Phage ::Tn5) and the PvqmR-lux reporter integrated at the lacZ locus. Exconjugant V. cholerae colonies were collected and streaked onto LB agar plates supplemented with polymyxin B, chloramphenicol, kanamycin, and arabinose. PvqmR-lux activity of surviving exconjugants was assayed using an ImageQuant LAS4000 imager (GE). V. cholerae colonies that produced light were harvested for plasmid DNA preparation. Isolated plasmid DNA was subsequently transformed into E. coli strains carrying Pqtip-lux or PvqmRlux to validate activity.
A library of plasmid-borne vqmA Vc mutants was transformed into E. coli carrying the Pqtip-lux reporter. Transformants were plated on LB agar supplemented with ampicillin, kanamycin, and arabinose. Pqtip-lux activity was assayed using an ImageQuant LAS4000 imager.

Growth, lysis, and bioluminescence assays
To measure growth of V. cholerae phage VP882 lysogens or activation of the PvqmR-lux and Pqtip-lux reporters in bacterial strains, overnight cultures of V. cholerae or E. coli were backdiluted 1:1000 into LB medium supplemented with appropriate antibiotics prior to being dispensed (200 μL) into 96-well plates (Corning Costar 3904). Arabinose was added as specified. The plates were shaken at 37˚C and a Biotek Synergy Neo2 Multi-Mode reader was used to measure OD 600 and bioluminescence. For bioluminescence assays, relative light units (RLU) were calculated by dividing bioluminescence by the OD 600 after 5 h.

Protein expression, purification, and electrophoretic mobility shift assay (EMSA)
Protein expression and purification were performed as described [9,19]. EMSAs were performed as described [8] with the following modifications: Following electrophoresis, 6% DNA retardation gels were stained with SYBR Green (Thermo) and visualized using an ImageQuant LAS 4000 imager with the SYBR Green settings. Unless specified otherwise, the highest concentration of VqmA assessed was 600 nM. 25 nM PvqmR or Pqtip DNA was used in all EMSAs. The percentage of promoter DNA bound was calculated using the gel analyzer tool in ImageJ and the estimated EC 50 values were derived from EC 50 analyses in Prism.

Western blot analysis
Western blot analyses probing for abundances of 3xFLAG-tagged proteins were performed as reported [3] with the following modifications: E. coli and V. cholerae carrying N-terminal 3xFLAG-tagged VqmA Vc and N-terminal 3xFLAG-tagged VqmA Phage alleles were backdiluted 1:1000 in LB supplemented with appropriate antibiotics and harvested after 6 h and 4 h of growth at 37˚C, respectively. Cells were resuspended in Laemmli sample buffer at a final concentration of 0.006 OD/μL. Following denaturation for 15 min at 95˚C, 5 μL of each sample was subjected to SDS-PAGE gel electrophoresis. RpoA was used as the loading control (Biolegend Inc.). Signals were visualized using an ImageQuant LAS 4000 imager.

Sequence alignments
Protein and DNA sequences in FASTA format were aligned in the BioEdit Sequence Alignment Editor using the default setting under the ClustalW mode. Figs 2A and S1 were prepared via the ESPript 3.0 online server [33].