Natural silencing of quorum-sensing activity protects Vibrio parahaemolyticus from lysis by an autoinducer-detecting phage

Quorum sensing (QS) is a chemical communication process that bacteria use to track population density and orchestrate collective behaviors. QS relies on the production, accumulation, and group-wide detection of extracellular signal molecules called autoinducers. Vibriophage 882 (phage VP882), a bacterial virus, encodes a homolog of the Vibrio QS receptor-transcription factor, called VqmA, that monitors the Vibrio QS autoinducer DPO. Phage VqmA binds DPO at high host-cell density and activates transcription of the phage gene qtip. Qtip, an antirepressor, launches the phage lysis program. Phage-encoded VqmA when bound to DPO also manipulates host QS by activating transcription of the host gene vqmR. VqmR is a small RNA that controls downstream QS target genes. Here, we sequence Vibrio parahaemolyticus strain O3:K6 882, the strain from which phage VP882 was initially isolated. The chromosomal region normally encoding vqmR and vqmA harbors a deletion encompassing vqmR and a portion of the vqmA promoter, inactivating that QS system. We discover that V. parahaemolyticus strain O3:K6 882 is also defective in its other QS systems, due to a mutation in luxO, encoding the central QS transcriptional regulator LuxO. Both the vqmR-vqmA and luxO mutations lock V. parahaemolyticus strain O3:K6 882 into the low-cell density QS state. Reparation of the QS defects in V. parahaemolyticus strain O3:K6 882 promotes activation of phage VP882 lytic gene expression and LuxO is primarily responsible for this effect. Phage VP882-infected QS-competent V. parahaemolyticus strain O3:K6 882 cells lyse more rapidly and produce more viral particles than the QS-deficient parent strain. We propose that, in V. parahaemolyticus strain O3:K6 882, constitutive maintenance of the low-cell density QS state suppresses the launch of the phage VP882 lytic cascade, thereby protecting the bacterial host from phage-mediated lysis.

Introduction Vibrio parahaemolyticus is the major worldwide cause of seafood-borne bacterial gastroenteritis [1]. The first pandemic of this pathogen was caused by V. parahaemolyticus serotype O3:K6 [2]. The marine environment, which V. parahaemolyticus naturally occupies, abounds with bacterial viruses called phages, and indeed, both lytic and temperate phages have been identified in pandemic V. parahaemolyticus O3:K6 strains. Phages are central to bacterial evolution, and their existence is often linked to the emergence of toxigenic bacteria from non-toxigenic strains [3][4][5]. Phages can benefit their host bacteria by endowing them with traits that promote niche colonization and spread to new territory. Phages can also parasitize their hosts by exploiting resources for their own replication and dissemination to uninfected bacteria. One such phage is the temperate, plasmid-like vibriophage 882 (here forward called phage VP882) isolated from V. parahaemolyticus strain O3:K6 882 [6] Phage VP882 regulates its lysis-lysogeny decision by "eavesdropping" on a host-produced quorum-sensing (QS) autoinducer (AI) molecule called DPO (3,5-dimethylpyrazin-2-ol, produced by the Tdh synthase; Fig 1A and 1B and [7,8]). Phage VP882 encodes the DPO-binding receptor and transcription factor called VqmA Phage which is a homolog of the bacterial DPO QS receptor VqmA [8]. The bacterial VqmA QS pathway is considered ubiquitous within the Vibrio genus, however, to date, it has only been characterized in the species it was discovered, Vibrio cholerae [7,9]. V. cholerae produces DPO and detects it via VqmA (hereafter called VqmA Vc ). DPO-bound VqmA Vc activates transcription of vqmR (vqmR Vc ), encoding a small regulatory RNA, VqmR Vc (Fig 1A). VqmR Vc represses translation of genes required for the production of virulence factors and formation of biofilms [9,10]. When phage VP882 infects Vibrio bacteria, it introduces a second vqmA into the system, vqmA Phage . The current model for the phage-encoded pathway is that an unknown cue induces expression of vqmA Phage in the infected Vibrios. Once produced, VqmA Phage binds host-produced DPO and launches the phage lytic program by activating the expression of an antirepressor-encoding gene called qtip [8]. In addition to qtip, VqmA Phage also activates expression of host vqmR Vc , demonstrating that phage VP882 influences both its own and its host's QS pathways (Fig 1A). By contrast, the VqmA Vc protein only binds to its partner vqmR Vc promoter, not to the phage qtip promoter ( Fig 1A and [8,11]).
Regarding the phage lysis pathway, Qtip sequesters the cI VP882 phage repressor of lysis ( Fig 1B). The host SOS response also launches the phage lytic pathway via RecA-mediated proteolysis of cI VP882 (Fig 1B). Sequestration or proteolysis of cI VP882 derepresses the phage q gene. Q activates expression of the gp69-71 operon encoding the phage lytic genes, driving host-cell lysis. Presumably, tuning into the host SOS response allows phage VP882 to connect (circles), AI-2 (triangles), and CAI-1 (squares) are detected by LuxN, LuxPQ, and CqsS, respectively. At LCD, the receptors are kinases and funnel phosphate to LuxO (via a protein LuxU, not shown). LuxO~P activates transcription of genes encoding the Qrr sRNAs. The Qrr sRNAs post-transcriptionally repress opaR, the master regulator of group behaviors. At HCD (shown in the cartoon), the liganded receptors act as phosphatases, reversing the phosphorylation cascade. Consequently, production of Qrr sRNAs ceases, and opaR translation is derepressed, OpaR is produced, and the cells undertake group behaviors. In parallel, at HCD, the DPO autoinducer (hexagons) interacts with the VqmA transcription factor. The complex activates vqmR expression and the VqmR sRNA controls QS behaviors. Phage VP882 carries vqmA Phage . Binding of VqmA Phage to host-produced DPO activates expression of a gene encoding an antirepressor called Qtip, which drives the phage lytic pathway. VqmA Phage also binds the host vqmR promoter to drive host QS behaviors. (B) Two pathways control phage VP882 lysis-lysogeny transitions. Upper: DPO-bound VqmA Phage activates expression of qtip. Qtip sequesters the phage cI VP882 repressor of lysis leading to phage replication and hostcell killing. Lower: the phage cI VP882 repressor is proteolyzed in response to host SOS/DNA damage via a host RecAdependent mechanism. The consequence is phage replication and host-cell killing. https://doi.org/10.1371/journal.pgen.1010809.g001 its lifestyle transitions to host viability, whereas tuning into host QS allows phage VP882 to execute host-cell lysis at high host-cell density when the probability of infecting the next host cell is maximized [8].
Here, we report the genome sequence of V. parahaemolyticus strain O3:K6 882 (here forward called strain 882), the strain from which phage VP882 was isolated. Strikingly, strain 882 encodes vqmA (vqmA 882 ) but lacks its partner vqmR gene. Specifically, there is an 897 bp deletion relative to other V. parahaemolyticus strains that eliminates the vqmR promoter, the vqmR coding sequence, and a portion of the non-coding region upstream of vqmA 882 . We also discover that the strain 882 luxO gene (luxO 882 ) harbors a 36 bp deletion that leads it to function as a phospho-mimetic. In this report, we investigate the consequences of these mutations on the QS capabilities of strain 882 and the broader implications for host-phage biology. We show that lack of vqmR and vqmA 882 expression, and the phospho-mimetic allele of luxO 882 lock strain 882 into the LCD QS state. Repair of vqmR-vqmA 882 or luxO 882 or both vqmR-vqmA 882 and luxO 882 partially and fully, respectively, restores QS capability. Re-establishment of strain 882 QS capability activates phage VP882 gene expression, more rapid phage VP882-driven host-cell lysis, and production of higher titers of virions compared to that in the QS-incompetent parent strain. We propose that the locked LCD QS state silences the information encoded in QS AIs for both the strain 882 host and the VP882 phage. This arrangement suppresses the potential for phage VP882 to transition from lysogeny to lysis, presumably protecting the host from killing by the phage that it carries.

V. parahaemolyticus strain 882 lacks vqmR and does not express vqmA
Strain 882 is unique relative to other studied V. parahaemolyticus strains because it harbors a phage with the capacity to surveil host QS by monitoring DPO abundance. The DNA sequence of phage VP882 is deposited [6], however, no sequence of its host was reported. We sequenced the strain 882 genome and compared it to that of the closely related V. parahaemolyticus O3:K6 type-strain, RIMD2210633. Strain 882 has an 897 bp deletion that eliminates its vqmR gene and promoter, a portion of a neighboring upstream gene (a putative transcriptional regulator of a proline metabolic operon), a predicted gene of unknown function, and a portion of the non-coding region upstream of vqmA 882 (Fig 2A). To determine the incidence of the vqmR deletion, we examined the coding regions of vqmR-vqmA loci in other publicly available Vibrio DNA sequences (5460 genomes total). We disregarded 12 of the 5460 Vibrio sequences whose assembly gaps included this region. Most sequenced Vibrios encode predicted vqmR-vqmA pairs. However, 312 strains lack both vqmR and vqmA (S1A Fig), and an additional 37 strains possess either a frameshift or a nonsense mutation in vqmA, presumably disrupting its function. Regarding V. parahaemolyticus strains (1731), all possess vqmR-vqmA. In three cases, however, the vqmA gene is predicted to be non-functional (frameshift or nonsense mutation). From this analysis, we could not discern if nucleotide changes in vqmR genes affect the functions or the expression levels of the VqmR sRNAs. Interestingly, of the >5000 analyzed strains, strain 882 is the only sequenced Vibrio that lacks vqmR but retains vqmA in its genome.
Our findings suggest that in strain 882, the VqmA-VqmR QS circuit cannot function because vqmR is absent. Moreover, given that the deleted region includes the predicted vqmA 882 promoter, it was not clear if vqmA 882 is expressed. Western blot analysis of VqmA 882 -3XFLAG produced from the native locus shows that VqmA 882 is not made (Fig 2B). By contrast, V. parahaemolyticus RIMD2210633 harbors an intact vqmR-vqmA locus and VqmA RIMD -3XFLAG was produced, and at higher levels at HCD than at LCD (Fig 2B). Thus, the non-coding region deleted in strain 882 contains the vqmA 882 promoter, which eliminates vqmA 882 expression. To test this prediction, we repaired the putative vqmA 882 promoter region by inserting the analogous intergenic DNA from V. parahaemolyticus RIMD2210633. VqmA 882 production was restored despite the absence of vqmR (Fig 2B). We call this strain 882 vqmA 882 Curiously, we observed that vqmA expression is cell-density dependent in both V. parahaemolyticus RIMD2210633 and strain 882 vqmA 882 + . In V. cholerae strain C6706, the only species in which vqmA expression has been studied, vqmA Vc is expressed and VqmA Vc is made constitutively [7]. We eliminated the possibility that V. parahaemolyticus VqmA directly activates its own expression (S1B Fig). To assess the underlying mechanism for the speciesspecific difference in expression patterns, we aligned Vibrio vqmA promoters (S1C Fig), revealing two potential clades differing by length and nucleotide sequence ( Fig 2C). The first group is comprised of shorter, V. cholerae-type vqmA promoters (~100 bp, Figs 2C and S1C), and the second is comprised of longer, V. parahaemolyticus RIMD2210633-type vqmA promoters (~300 bp; Figs 2C and S1C). Exchanging V. cholerae and V. parahaemolyticus RIMD2210633 vqmA promoters eliminated production of both VqmA proteins as judged by Western blot analysis ( Fig 2D). Thus, the vqmA promoter sequence predicts whether vqmA expression is constitutive, as in V. cholerae, or cell-density dependent, as in V. parahaemolyticus RIMD2210633, and the promoter sequence only functions in its species of origin.
We wondered whether VqmA 882 , if produced by strain 882, would bind DPO and activate expression of vqmR. The vqmA 882 coding sequence is identical to those of other V. parahaemolyticus strains, including V. parahaemolyticus RIMD2210633. VqmA 882 also shares 65% and 41% amino acid identity with VqmA Vc and VqmA Phage , respectively, and, residues F67, F99, and K101, which are critical for DPO binding in VqmA Vc and VqmA Phage , are conserved in VqmA 882 (S2A Fig and [13]). We also wondered whether VqmA 882 operates analogously to VqmA Vc in that it does not bind the phage qtip promoter. To examine VqmA promoter binding capabilities, which have not previously been studied for V. parahaemolyticus, we transformed arabinose-inducible vqmA 882 into Escherichia coli harboring PvqmR RIMD or Pqtip fused to the luciferase genes (PvqmR RIMD -lux or Pqtip-lux, respectively).

V. parahaemolyticus strain 882 possesses a LCD-locked LuxO allele
Inspired by our finding that in strain 882, the VqmA-VqmR QS circuit is non-functional, we determined whether its other QS systems also harbored defects. Examination of QS genes revealed a 36 bp deletion exists in luxO (luxO 882 ) corresponding to elimination of LuxO residues 91-102 (Δ91-102, and Fig 3A). To characterize the consequence of the deletion on LuxO function, we introduced the QS-controlled luciferase operon from Vibrio harveyi (luxCDABE) into strain 882 and used light production as a heterologous QS target. In V. harveyi at HCD, luxCDABE expression is activated by LuxR, the HCD QS master regulator [12]. Across Vibrios, homologs of LuxR including OpaR from V. parahaemolyticus, also activate luxCDABE at HCD [14]. We compared light production by strain 882 to three other 882 strains that we constructed: luxO + (in which luxO 882 was repaired), luxO D61E , and luxO D61A . LuxO D61E is a LuxO~P mimetic and drives constitutive production of the Qrr sRNAs, thus locking cells into the LCD QS state [15]. Conversely, LuxO D61A is a non-phosphorylatable LuxO allele, is inactive, and therefore locks cells into the HCD QS state. As expected, the 882 strains carrying luxO D61E and luxO D61A were constitutively dark and bright, respectively ( Fig 3B and [15]), and strain 882 engineered to be luxO + produced light in a cell density-dependent manner, the hallmark of a QS-controlled process. By contrast, the parent strain 882 produced a modest amount of light (~100-1000-fold less than the 882 luxO + strain at HCD) with dramatically reduced cell densitydependent regulation (Fig 3B). The light production pattern of strain 882 was most similar to that of strain 882 carrying luxO D61E (Fig 3B), indicating that LuxO 882 mimics LuxO~P (see Fig 1A). Indeed, introduction of a reporter of qrr3 expression (Pqrr3-mRuby) verified that both luxO 882 and luxO D61E drove high levels of fluorescence, whereas luxO + and luxO D61A produced 11-fold and 73-fold less fluorescence, respectively ( Fig 3C). Moreover, because repair of luxO 882 to luxO + fully restored proper QS function, the deletion in luxO 882 is the defect that eliminates signal transduction from LuxN, LuxPQ, and CqsS (Fig 1A and [12]). Taken together, our results show that strain 882 is LCD-locked with respect to all of its known QS systems.
The LCD-behavior driven by luxO 882 could be due to increased LuxO transcriptional activity, increased LuxO protein production, or both. First, to examine LuxO activity, we engineered the D61E and D61A mutations into luxO 882 harboring the 36 bp deletion and introduced the alleles onto the strain 882 chromosome. We call these genes luxO 882 D61E and , corresponding to the Δ91-102 LuxO D61E and Δ91-102 LuxO D61A proteins, respectively. We assayed activity using the luxCDABE reporter. As expected, strain 882 with luxO 882 D61E was constitutively dark (Fig 3B) Fig 3D). Finally, mapping sequences onto the existing LuxO crystal structure revealed that the residues missing in LuxO 882 reside along interface II between the LuxO receiver domain which contains the D61 site of phosphorylation and the ATP-binding catalytic domain ( Fig 3E). Phosphorylation of LuxO is predicted to disrupt interface II, thereby enhancing LuxO transcriptional activity [16]. We predict that deletion of residues 91-102 alters interface II, and consequently, increases LuxO activity.
To investigate whether other Vibrios possess luxO mutations similar to that present in strain 882, we analyzed the luxO genes of the same 5460 sequenced Vibrios described above and found that 88 Vibrio strains, including strain 882, carry the identical Δ91-102 luxO mutation (S3B Fig). This group of Vibrios is dominated by V. parahaemolyticus strains. Five V. cholerae and three Vibrio owensii strains are also in the group. Other large, in-frame deletions are present in luxO genes in several V. parahaemolyticus strains spanning residues 67-87 and 114-134 (S3C Fig). While we did not test the functions of these LuxO proteins, we speculate that they act similarly to LuxO 882 . Regarding the frequency of the Δ91-102 luxO mutation in V. parahaemolyticus strains, several possibilities could explain our finding: (1) There is an overrepresentation of sequenced V. parahaemolyticus strains in the database (1731/5460), (2) the mutation originated from a common ancestor, and/or (3) environmental pressures on V. parahaemolyticus strains select for this particular mutation. We cannot eliminate the first possibility of sample bias. To distinguish between the latter two possibilities, we performed a core genome alignment of the 88 V. parahaemolyticus strains carrying Δ91-102 luxO (including strain 882) with V. parahaemolyticus RIMD2210633 as the wild-type reference (Fig 3F). Our analyses suggest that the incidence of the Δ91-102 luxO mutation does not reflect acquisition from a common ancestor, given that strain 882 is most similar to V. parahaemolyticus RIMD2210633 yet the two have dissimilar luxO alleles. Rather, it is more likely that V. parahaemolyticus strains encounter environmental pressures that select for mutations in QS components that maintain the strains in the LCD QS mode.

V. parahaemolyticus strain 882 QS circuits converge to regulate phage VP882 gene expression by an SOS-independent mechanism
One of our goals is to understand how QS shapes bacterial-phage partnerships. In the present context, we sought to determine the ramifications the QS defects have on interactions between strain 882 and its lysogenized prophage VP882. Toward this aim, we assessed the transcriptional changes that occur in strain 882 following repair of the vqmR-vqmA and luxO QS pathways, individually and together. We  (Fig 4A). A transcriptional reporter in which the phage gp69 lytic gene promoter was fused to lux (Pgp69-lux) verified that phage VP882 gene expression is activated by host QS. Specifically: the 882 luxO + and 882 vqmR-vqmA 882 + luxO + strains exhibited~10-20-fold higher Pgp69-lux expression than strains 882 and 882 vqmR-vqmA 882 + (Fig 4B). Light production from Pgp69-lux was nearly undetectable in 882 strains that had been cured of phage VP882, indicating that the QS state of the host affects Pgp69 activity via phage-encoded regulators (Fig 4C). Indeed, the transcriptomic analyses show that the key phage lytic genes vqmA Phage , qtip, and q were upregulated in 882 luxO + and 882 vqmR-vqmA 882 + luxO + , while expression of the cI VP882 gene remained unchanged (Fig 4A and S1 Table). There were no significant growth differences among 882 test strains cultivated under standard liquid growth conditions (S4A Fig). This finding suggests that, while host QS activates phage VP882 lytic gene expression, the level of activation is insufficient to produce global host-cell lysis. Thus, to characterize the consequences of host QS signaling on phage VP882 lifestyle transitions, we monitored two key features of phage virulence, phage reproductive success (i.e., virion production) and time to host-cell lysis following phage induction in each of our test strains. First, regarding QS effects on spontaneous virion production, consistent with the above Pgp69-lux reporter data, the QS repaired 882 luxO + and 882 vqmR-vqmA 882 + luxO + strains produced twice as many viral particles as the QS-defective parent strain 882 despite there being no detectable differences in growth under our conditions (Figs 5A, left, and S4A). Second, regarding QS effects on time to lysis, we introduced a plasmid carrying arabinoseinducible vqmA Phage into each strain to induce the phage lytic program and drive host lysis. Strains 882 luxO + and 882 vqmR-vqmA 882 + luxO + lysed 1 and 2 h earlier, respectively, than did strains 882 and 882 vqmR-vqmA 882 + (Fig 5B). Similar to the results for spontaneous induction of the phage, following expression of vqmA Phage , the 882 luxO + and 882 vqmR-vqmA 882 + luxO + strains produced twice as many viral particles as strain 882 (Fig 5A, right). We considered two possible mechanisms by which host QS could affect phage VP882 lytic activity: Host QS could increase RecA-dependent cI VP882 proteolysis or host QS could activate phage-encoded qtip expression. To test the first possibility, we introduced HALO-tagged cI VP882 (HALO-cI VP882 ) into our set of 882 lysogens and monitored cI VP882 cleavage. No differences were detected between the strains in the absence of ciprofloxacin, which measures spontaneous HALO-cI VP882 proteolysis. In the presence of ciprofloxacin, which induces HALO-cI VP882 cleavage, all strains exhibited similar elevations in proteolyzed HALO-cI VP882 ( Fig 6A). Consistent with this finding, deletion of recA in the 882 lysogens did not eliminate differences in Pgp69-lux expression or time-to-lysis between the QS-repaired and QS-defective 882 strains (Fig 6B and 6C, respectively). Thus, host QS input into phage VP882 lifestyle transitions occurs independently of the RecA pathway.
To test the possibility that host QS increases Qtip-mediated cI VP882 sequestration, we constructed phage VP882 qtip::cm, eliminating the phage VqmA Phage -Qtip pathway to lysis (S4B Fig). We assayed Pgp69-lux expression in our set of QS-repaired and QS-defective 882  Fig 2B (B, C). strains carrying phage VP882 qtip::cm or a phage we call VP882 Ctr::cm with cm integrated at a neutral locus. Interruption of qtip severely reduced host QS-dependent effects on Pgp69-lux expression (Fig 6D), indicating that host QS influences phage lysis-lysogeny transitions via a Qtip-dependent mechanism. As the only known regulator of qtip is VqmA Phage , we tested whether a fragment of the phage VP882 genome harboring vqmA Phage -qtip and the key cI VP882 and q regulatory genes is sufficient to drive differential Pgp69-lux induction in the various test strains. Similar to the 882 strains carrying full-length phage VP882, the 882 luxO + and 882 vqmR-vqmA + luxO + strains harboring only the phage VP882 regulatory fragment (called VP882*) produced >10-fold more light from Pgp69-lux than did strains 882 and 882 vqmR-vqmA + carrying VP882* (Fig 6E, left). Deletion of vqmA Phage -qtip from VP882* (VP882* ΔvqmA Phage -qtip) dramatically reduced Pgp69-lux levels in all the strains, indeed, abolishing differences between them (Fig 6E, right). We conclude that regulation of the phageencoded vqmA Phage -qtip pathway connects host QS to phage VP882 lytic gene expression.

Discussion
In this report, we characterize V. parahaemolyticus strain 882, the natural host of phage VP882. We discover that strain 882 harbors defects in all of its QS circuits, such that the strain is locked in the LCD QS mode. The consequence is suppression of phage VP882 lytic activation. Thus, phage VP882 is more virulent to QS-proficient V. parahaemolyticus 882 than to the naturally occurring QS-deficient 882 strain. We previously showed that, at HCD, detection of host-produced DPO by phage VP882 launches phage-mediated bacterial killing [8]. Our present results suggest that phage VP882 reacts to inputs from multiple QS components: DPO, VqmR-VqmA, and LuxO.
We do not yet know the mechanism by which LuxO and VqmR-VqmA funnel information into activation of phage VP882 lytic genes. We do know that VqmA Phage and Qtip are required, and that the RecA-dependent pathway is dispensable. We presume that the host QS  Beyond lysogenizing strain 882, phage VP882 has been found in two Salmonella enterica strains (accession codes: AAFWQT000000000 and AAEKWQ000000000 and [17]), Shewanella algae sp. (accession code: LTBI01000119.1), and V. parahaemolyticus (accession code: NNHH01000051.1 and [17]). Regarding the additional V. parahaemolyticus strain, curiously, host vqmR-vqmA is intact, but an 819 bp deletion encompassing vqmA Phage and qtip exists in its phage VP882. As described here, the reverse is the case in strain 882: vqmA Phage in phage VP882 is intact, but host vqmR-vqmA is disrupted. Salmonella and Shewanella do not possess vqmA and vqmR genes. Thus, no naturally occurring host that is lysogenic for phage VP882 has been identified that harbors two copies of vqmA. We know that both bacterial VqmA and VqmA Phage activate expression of host vqmR [8,11], and our work here indicates that VqmR production activates phage VP882 lytic gene expression (albeit modestly and in the presence of a functional LuxO; Fig 4A). Given these two features, we speculate that harboring two copies of vqmA is disadvantageous to the host, and bacteria that eliminate one copy of vqmA, either from the host or apparently from the phage, receive superior protection from their parasitic viral partner. An alternative possibility is that the phage benefits from this arrangement by limiting superinfection. Specifically, the QS-defective state of the host renders it insensitive to AIs produced by kin bacteria in the vicinal community, curbing QS-mediated induction of the VP882 prophage, and potential costly production of phage particles that lack hosts to infect in a fully lysogenized bacterial community.
Unexpectantly, the Δ91-102 LuxO (i.e., LuxO 882 ) variant was present in an additional 87 sequenced Vibrios. Indeed, a recent screen for mutations that affect colony sectoring and kinkilling in an environmental isolate of V. cholerae, called V. cholerae 2740-80, identified a mutation in LuxO with a similar deletion (residues 94-106 and [18]). Analogous to what we report here, this allele drove LCD QS behavior in V. cholerae [18], suggesting that in-frame deletions in LuxO confer advantages in different Vibrios and in different contexts. Mapping of LuxO 882 onto the V. angustum LuxO crystal structure revealed that the eliminated amino acids in LuxO 882 reside along interface II between the LuxO catalytic and receiver domains [16]. In the unphosphorylated state, the LuxO catalytic and receiver domains interact via interface II. By contrast, when residue D61 is phosphorylated, interface II is disrupted, fostering LuxO transcriptional activity. Likely, deletion of residues 91-102 permanently disrupts intra-protein domain interactions at interface II, locking LuxO 882 into its transcriptionally-active state. LuxO mutations, as well as mutations in homologs of OpaR, which also confer the LCD QS state, are common in Vibrios in nature. Presumably, these variations in QS system function are driven by selective pressures that deliver fitness benefits under particular conditions [19][20][21]. One possible example is that mutations that lock Vibrios into the LCD QS state prevent the costly production of QS-controlled public goods [21], and here we show protect from phage killing.
https://doi.org/10.1371/journal.pgen.1010809.g006 including in this present work, demonstrate that temperate phages can garner information from host-produced QS cues to regulate their lysis-lysogeny transitions [8,28,29]. A recent study demonstrated that in Vibrio anguillarum, QS represses φH20-like phage p41 lytic development at HCD [30]. By contrast, QS in V. parahaemolyticus drives phage VP882 lytic development at HCD, whereas the naturally locked LCD QS state of strain 882 suppresses phage VP882 viral reproduction. Likely, the ramifications of host QS signaling on temperate phage biology are specific to particular phage-host partnerships.

Cloning techniques
All primers and dsDNA (gene blocks) used for plasmid construction and qPCR, listed in S3 Table, were obtained from Integrated DNA Technologies. Gibson assembly with HiFi DNA assembly mix, intramolecular reclosure, and traditional methods were employed for cloning. PCR with iProof was used to generate insert and backbone DNA. Cloning enzymes were obtained from NEB. Plasmids used in this study are listed in S4 Table. Transfer of plasmids into V. parahaemolyticus and V. cholerae was carried out by conjugation followed by selective plating on LB plates supplemented with appropriate antibiotics.

Growth, lysis, and reporter assays
Overnight cultures of V. parahaemolyticus and E. coli were back-diluted 1:1000 and 1:100, respectively, with fresh medium containing appropriate antibiotics prior to being dispensed (200 μL) into 96 well plates (Corning Costar 3904). Arabinose and DPO were added as specified. Wells that did not receive treatment received an equal volume of water. A BioTek Synergy Neo2 Multi-Mode reader was used to measure OD 600 , fluorescence, and bioluminescence. Relative light units (RLU) and relative fluorescence units (RFU) were calculated by dividing the bioluminescence and fluorescence readings, respectively, by the OD 600 at that time.

qPCR
Overnight cultures were back-diluted 1:1000 with specified treatments. Subsequently, cultures were grown for an additional 4.5 h prior to 1:10 dilution in water. Samples were heated at 95˚C for 20 min to release viral DNA packaged in phage VP882 virions, linear phage VP882, and host genomic DNA from cells. DNA was further diluted 1:100 in water, and 1 μL was used in qPCR reactions. SYBR Green mix (Quanta) and Applied Biosystems QuantStudio 6 Flex Real-Time PCR detection system (Thermo) were used for real-time PCR. Data were analyzed by a comparative CT method in which the gp69 target gene was normalized to an internal bacterial control gene (hfq).

in vitro HALO-cI VP882 repressor cleavage and in-gel HALO detection
Assessment of HALO-cI VP882 cleavage was carried out in V. parahaemolyticus 882 strains as described [8] with modifications. Specifically, overnight cultures carrying the plasmid harboring HALO-cI VP882 were diluted 1:200 in medium and grown for 2.5 h with shaking. Cultures were divided in half and administered ciprofloxacin or water as specified. The treated cultures were incubated without shaking for an additional 1.5 h. Cells were collected by centrifugation (16,100 X g for 1 min), resuspended in a lysis buffer containing 1x BugBuster, benzonase, 300 mM NaCl, and 1 μM HALO-TMR (excitation/emission: 555/585 nm). Clarified lysates were loaded onto NEBExpress Ni spin columns (NEB), washed once with lysis buffer containing 10 mM imidazole, and eluted in lysis buffer containing 500 mM imidazole. Eluted samples were subjected to electrophoresis on 4-20% SDS-PAGE stain-free gels. Gels were imaged using an ImageQuant LAS 4000 imager under the Cy3 setting prior to being exposed to UV-light for 7 min and re-imaged under the EtBr setting. Exposure times never exceeded 30 sec.

Western blot analyses
Western blot analyses were performed as reported [11] with the following modifications: V. parahaemolyticus and V. cholerae producing C-terminal 3XFLAG-tagged VqmA were back-diluted 1:1000 and harvested at OD 600 = 0.2 and 2.0. E. coli producing 3XFLAG-tagged VqmA were back-diluted 1:100 and harvested at OD 600 = 2.0. V. parahaemolyticus producing 3XFLAG-tagged LuxO were back-diluted 1:1000 and harvested at OD 600 = 2.0. E. coli and Vibrios were resuspended in Laemmli sample buffer at final concentrations of 0.006 OD μL -1 and 0.02 OD μL -1 , respectively. 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.

DNA and RNA sequencing
All sequencing was carried out by SeqCenter (formerly Microbial Genome Sequencing Center; MIGS). For preparation of samples for genomic sequencing, overnight cultures of strain 882 were back-diluted 1:1000 and grown to stationary phase prior to harvesting. Genomic DNA was extracted using the DNeasy Blood & Tissue Kit (Qiagen) according to the manufacturer's protocol. For preparation of samples for RNA sequencing, overnight cultures of the 882 strains were back-diluted 1:1000 and grown for 4 h prior to harvesting. Total RNA was isolated using the RNeasy Mini Kit (Qiagen). RNA samples were treated with DNase using the TURBO DNA-free Kit (Thermo).

Multiple sequence alignment
Genomic DNA sequences of 5534 Vibrio strains were downloaded from the GenBank database [31]. A first pre-processing step was performed to discard duplicate genomes and genomes that contained assembly gaps in regions of interest, which yielded 5460 genomes for further analyses. A custom MATLAB (Mathworks, 2022) algorithm was used to automate the sequence-similarity-based search of vqmR, vqmA, and luxO genes in each of the 5460 genomes [32]. The DNA sequence of vqmR, and protein sequences of VqmA and LuxO from V. cholerae C6706 were used as queries. Multiple sequence alignment of DNA and protein sequences were performed and analyzed in MATLAB and visualized by NCBI Multiple Sequence Alignment Viewer (v1.22.1).

Phylogenetic analyses
Genomic sequences were annotated using Prokka (v 1.14.6) [33]. A core genome was extracted from Prokka-output GFF files and aligned using built-in MATLAB functions. A post-processing step was next applied to remove regions containing long consecutive gaps or low-quality alignment from the core genome alignment to avoid using false SNPs for phylogeny reconstruction. Pairwise p-distances were calculated based on the reduced core genome alignment and were subsequently used to build a phylogenetic tree using the UPGMA (unweighted pair group method with arithmetic mean) method. Using other distance-based methods, such as WPGMA (weighted pair group method with arithmetic mean), or a maximum parsimony method yielded similar results of phylogeny reconstruction.

Quantitation and statistical analyses
Software used to collect and analyze data generated in this study consisted of: GraphPad Prism 9 for analysis of growth and reporter-based experiments; Gen5 for collection of growth and reporter-based data; SnapGene v6 for primer design; QuantStudio for qPCR collection; LASX for acquisition of western blots; and FIJI for image analyses. Data are presented as the means ± std. The number of technical and independent biological replicates for each experiment are indicated in the figure legends.
Supporting information S1  Fig 2B. Right: representative western blot of VqmA 882 -3XFLAG produced by the E. coli in the left panel. RpoA was used as the loading control. (C) Multiple DNA sequence alignment of the intergenic regions between vqmR and vqmA for the V. cholerae clade (Clade 1) and the V. parahaemolyticus clade (Clade 2). A representative strain (as designated) was chosen for each species in each clade. Thick gray or red bars indicate, respectively, nucleotides that are identical with or different from the consensus (>50% agreement among aligned sequences). Thin gray lines indicate gaps in the sequence alignments. Scale bar indicates 50 bp. Data in B are represented as means ± std with n = 3 biological replicates (left) and representative of two independent experiments (right). qtip::cm and arabinose-inducible vqmA Phage in medium treated with water (white), arabinose (gray), or ciprofloxacin (black). Arabinose (0.2%) was used to induce vqmA Phage expression, and ciprofloxacin (500 ng mL -1 ) was used to induce host SOS. Data are represented as means ± std with n = 3 biological replicates (A, B). (TIF)