LuxT controls specific quorum-sensing-regulated behaviors in Vibrionaceae spp. via repression of qrr1, encoding a small regulatory RNA

Summary

Quorum sensing (QS) is a process of chemical communication bacteria use to transition between individual and collective behaviors. QS depends on the production, release, and synchronous response to signaling molecules called autoinducers (AIs). The marine bacterium Vibrio harveyi monitors AIs using a signal transduction pathway that relies on five small regulatory RNAs (called Qrr1-5) that post-transcriptionally control target genes. Curiously, the small RNAs largely function redundantly making it difficult to understand the necessity for five of them. Here, we identify LuxT as a transcriptional repressor of qrr1. LuxT does not regulate qrr2-5, demonstrating that qrr genes can be independently controlled to drive unique downstream QS gene expression patterns. LuxT reinforces its control over the same genes it regulates indirectly via repression of qrr1, through a second transcriptional control mechanism. Genes dually regulated by LuxT specify public goods including an aerolysin-type pore-forming toxin. Phylogenetic analyses reveal that LuxT is conserved among Vibrionaceae and sequence comparisons predict that LuxT represses qrr1 in additional species. The present findings reveal that the QS regulatory RNAs can carry out both shared and unique functions to endow bacteria with plasticity in their output behaviors.

Keywords:

Bioluminescence – DNA transcription – Gene expression – Messenger RNA – Regulator genes – Vibrio – Vibrio cholerae – Post-transcriptional gene regulation

Introduction

Bacteria can coordinate gene expression on a population-wide scale using a process of cell-cell communication called quorum sensing (QS). QS depends on the production, release, and detection of signal molecules called autoinducers (AIs). Because AIs are self-produced by the bacteria, as cell density increases, extracellular AI levels likewise increase. Bacteria respond to accumulated AIs by collectively altering gene expression, and in turn, behavior. QS-regulated processes include bioluminescence, biofilm formation, and the secretion of virulence factors [1, 2].

Vibrio harveyi is a model marine bacterium that uses QS to regulate over 600 genes [3–8]. V. harveyi produces and responds to three AIs, which act in parallel. The LuxM synthase produces AI-1 (N-(3-hydroxybutanoyl)-L-homoserine), LuxS produces AI-2 ((2S,4S)-2-methyl-2,3,3,4-tetrahydroxytetrahydrofuran-borate), and CqsA produces CAI-1 ((Z)-3-aminoundec-2-en-4-one) [3, 9–16]. The three AIs are recognized by the cognate receptors LuxN, LuxPQ, and CqsS, respectively [3, 4, 13, 14]. At low cell density (LCD, Fig 1A), when little AI is present, the unbound receptors act as kinases that transfer phosphate to the phosphorelay protein LuxU, which shuttles the phosphoryl group to the response regulator, LuxO [4, 6, 17, 18]. LuxO-P, together with the alternative sigma factor σ⁵⁴, activates expression of genes encoding five non-coding small regulatory RNAs (sRNAs), Qrr1-5, that function post-transcriptionally [6, 19, 20]. The five Qrr sRNAs promote translation of aphA and repress translation of luxR, encoding the LCD and high cell density (HCD) QS master transcriptional regulators, respectively (Fig 1A) [19, 21–26]. When the Qrr sRNAs are produced, individual behaviors are undertaken and the luciferase operon (luxCDABE), responsible for the canonical bioluminescence QS output in V. harveyi, is not expressed. At HCD (Fig 1B), when the AIs bind to their cognate receptors, the receptors’ kinase activities are inhibited, allowing their phosphatase activities to dominate. Consequently, phospho-flow through the QS circuit is reversed [27]. Dephosphorylated LuxO is inactive. Thus, Qrr1-5 are not produced, aphA translation is not activated, and luxR translation is not repressed (Fig 1B). In this state, LuxR is produced, and it controls expression of genes underpinning group behaviors. Notably, LuxR activates expression of luxCDABE, causing V. harveyi cells to make light at HCD [14].

The five V. harveyi Qrr sRNAs have high sequence identity and they are predicted to possess similar secondary structures with four stem loops [19]. Mechanistic studies of Qrr3 as the exemplar Qrr showed it regulates translation of its different target mRNAs by four mechanisms, all mediated by the chaperone Hfq; repression via catalytic degradation of the mRNA target, repression via coupled degradation of Qrr3 with the mRNA target, repression through sequestration of the mRNA target, and activation via revelation of the mRNA ribosome-binding site [25]. In addition to aphA and luxR, the Qrr sRNAs also feedback to repress luxO and luxMN translation [28, 29]. Microarray analyses following qrr overexpression revealed 16 additional Qrr-controlled target mRNAs [26, 30].

The extreme relatedness of the Qrr sRNAs, coupled with their similar QS-controlled production patterns, has made it difficult to assign any unique role to a particular Qrr sRNA. Nonetheless, among the Qrr sRNAs, Qrr1 stands out: it lacks nine nucleotides in stem loop 1 that are present in Qrr2-5 [19, 26, 30]. Due to this difference, Qrr1 does not regulate aphA and two of the other known target mRNAs [30]. Qrr2-5 regulate an identical set of target mRNAs [30]. Thus, the failure of Qrr1 to control one subset of mRNAs is the only functional difference known among the Qrr sRNAs. Also of note is the position of qrr1 in the V. harveyi genome: qrr1 is located immediately upstream of luxO, oriented in the opposite direction [19, 20]. No other qrr genes reside near known QS genes.

Predicted LuxO-P and σ⁵⁴ binding sites lie upstream of each qrr gene. The sites vary in sequence and relative position with respect to the qrr transcriptional start sites. Other than these sites, there is little sequence similarity between qrr promoter regions [19, 20]. There also exist hallmarks of transcription factor binding sites upstream of qrr genes, which differ in every case, hinting that unique factors could regulate each qrr gene [19]. Indeed, while all the Qrr sRNAs are made at LCD, they exhibit distinct production profiles. Specifically, in order of highest to lowest expression: Qrr4 > Qrr2 > Qrr3 > Qrr1 > Qrr5 [19]. The strength by which each Qrr sRNA represses luxR translation, and therefore downstream bioluminescence emission, correlates with Qrr production level: Qrr4 is the strongest repressor of light production, while Qrr1 and Qrr5 are the weakest [19]. When introduced into Escherichia coli, all five qrr sRNA genes are activated to high levels by LuxO D61E, a LuxO-P mimetic, suggesting that regulation by additional factors, that are not present in E. coli, occurs in V. harveyi [19]. Investigating the possibility that other regulators are involved in qrr control in vivo is the subject of the present work.

LuxT is a 17 kDa transcriptional regulator of the AcrR/TetR family, initially identified as a protein that binds strongly to DNA containing the region upstream of the V. harveyi luxO gene [31, 32]. An approximate 50 bp region that is bound by LuxT was discovered [31]. A follow-up report showed that LuxT activates light production in V. harveyi, the presumption being that LuxT functioned via repression of luxO [32]. At the time of this earlier study, the Qrr sRNAs had not been discovered and LuxO was assumed to be a repressor of bioluminescence. Thus, the logic of the first LuxT manuscripts were: LuxT represses luxO, and LuxO represses luciferase.

Research undertaken since the original LuxT publications has led to the current understanding of mechanisms underlying V. harveyi QS-controlled gene regulation (Fig 1). Key is that LuxO phosphorylation, not luxO expression, is regulated (Fig 1). This incongruity inspired us to reconsider the earlier findings concerning LuxT. Here, we explore the role of LuxT in V. harveyi QS with a focus on its connection to qrr1. We show that LuxT does indeed bind upstream of luxO at the site originally identified [31]. However, LuxT does not regulate luxO. While the experiments in the initial manuscripts were rigorously performed and interpreted appropriately, the authors could not have known that the gene encoding Qrr1 is located adjacent to luxO. We discover that the LuxT binding region is located within the qrr1 promoter. Indeed, we show that LuxT represses the transcription of qrr1 at LCD. LuxT does not repress qrr2-5. Relative to wild-type (WT) V. harveyi, in a ΔluxT mutant, qrr1 is expressed more highly at LCD. As a consequence, Qrr1 is available to post-transcriptionally regulate its target genes, including a gene encoding an extracellular protease (VIBHAR_RS11785), a gene encoding a pore-forming aerolysin toxin (VIBHAR_RS11620), a gene encoding a chitin deacetylase (VIBHAR_RS16980), and a gene specifying a component involved in capsular polysaccharide secretion (VIBHAR_RS25670) [30]. In addition to indirect activation of these genes via repression of qrr1, LuxT also activates transcription of these same four genes. Finally, we show that LuxT repression of qrr1 transcription is not specific to V. harveyi. LuxT also represses qrr1 in Aliivibrio fischeri, a species that, interestingly, harbors only a single Qrr sRNA: qrr1. Phylogenetic analyses show that luxT is conserved among Vibrionaceae and suggest that LuxT may repress qrr1 in other species within the Vibrionaceae family. Together, our results support a new QS model that incorporates LuxT and provides a mechanism for the unique control of one of the Qrr sRNA genes, qrr1. This newly revealed regulatory arrangement shows how Qrr1 controls downstream targets distinct from those controlled by the other Qrr sRNAs.

Results

LuxT binds upstream of luxO but does not repress luxO transcription

In the original works that identified and studied V. harveyi LuxT, DNA binding assays revealed the LuxT binding site to be a roughly 50 bp region lying 76 bp upstream of the luxO start codon [31]. (We note that in those reports, the site was designated to be 117 bp upstream of luxO, due to initial mis-annotation of the luxO start codon.) By assaying changes in light production, the authors concluded that LuxT represses luxO transcription [32]. This result is curious because our subsequent work showed that luxO is transcribed constitutively and only its phosphorylation state changes in response to QS signaling [18, 33]. Indeed, all fluctuations in LuxO levels in V. harveyi have been ascribed to intrinsic noise [33]. To confirm that LuxT binds upstream of luxO, we conducted electrophoretic mobility shift assays (EMSAs) using purified LuxT protein. Analogous to the results described by Lin et al. [31], LuxT caused a shift of a 95 bp DNA probe encompassing the luxO promoter region, whereas no significant binding to a control DNA probe occurred (Fig 2A). In the context of the 95 bp luxO promoter probe, randomizing the DNA sequence of the identified 50 bp LuxT binding region nearly eliminated LuxT binding (S1 Fig). Also consistent with the initial findings, deletion of luxT caused an ~11-fold reduction in light production by V. harveyi at LCD, indicating that LuxT is a LCD activator of luciferase (Fig 2B) [32]. At HCD (OD₆₀₀ > 1), the WT and ΔluxT V. harveyi strains exhibited similar light production profiles (Fig 2B). Therefore, LuxT activation of luciferase expression is cell-density dependent, indicating a possible role for QS.

The implication from the above findings, based on the original work, is that LuxT functions via repression of luxO. To investigate this possibility, we measured luxO transcript levels in WT and ΔluxT V. harveyi. We also measured transcript levels of luxC, the first gene in the luciferase operon. There were no detectable differences in luxO transcript levels in the WT and ΔluxT strains at either LCD or HCD (Fig 2C and 2D, respectively). Thus, LuxT does not repress luxO transcription. By contrast, and consistent with the results in Fig 2B, WT V. harveyi possessed 7-fold more luxC mRNA than did ΔluxT V. harveyi at LCD (Fig 2C) while the difference was only 2-fold at HCD (Fig 2D). Thus, LuxT activates luxCDABE expression, primarily at LCD. Finally, measurements of AphA and LuxR protein levels showed no significant differences between the WT and ΔluxT strains at either LCD or HCD (Fig 2E). Because aphA and luxR lie downstream of LuxO in the QS circuit, changes in LuxO levels necessarily drive changes in AphA and LuxR levels, albeit in opposite directions (Fig 1 and [8, 23, 28]). We conclude that LuxT has no role in regulating luxO expression. Therefore, LuxT activation of light production must occur through an alternative mechanism. We return to this point below.

LuxT represses qrr1, not luxO, transcription

As mentioned in the Introduction, at the time of the Lin et al. studies, the Qrr sRNAs that function between LuxO and QS target genes had not been discovered. Thus, Lin et al. could not have known that qrr1 lies immediately upstream and in the opposite orientation of luxO in the V. harveyi genome. In fact, qrr1 is located in closer proximity to the identified LuxT binding region than luxO. Specifically, if +1 designates the qrr1 transcriptional start site, the LuxT DNA binding region spans bases -76 to -27, suggesting that LuxT binds in the qrr1 promoter between the predicted LuxO-P and σ⁵⁴ binding sites that are essential for activation of qrr1 transcription (Figs 3A and S2) [19, 20, 31].

To test our prediction that LuxT represses qrr1 transcription, not luxO transcription, we employed two fluorescent reporters. First, we constructed a qrr1 promoter fusion containing the 193 nucleotides immediately upstream of qrr1 fused to mRuby3. Thus, the promoter fragment harbored the LuxO-P, LuxT, and σ⁵⁴ binding sites. A consensus ribosome-binding site was included to drive mRuby3 translation. Second, a luxO promoter fusion was constructed by cloning the same 193 bp DNA fragment in the opposite orientation upstream of mRuby3. Reporter fluorescence was measured in four V. harveyi strains: WT, luxO D61E, ΔluxT, and luxO D61E ΔluxT. As mentioned, V. harveyi luxO D61E encodes a LuxO-P mimetic. LuxO D61E constitutively activates qrr1-5, causing strains harboring this mutant allele to display a “LCD-locked” phenotype irrespective of the actual culture cell density [18]. The V. harveyi luxO D61E strain is a crucial tool for our studies. It enables investigation of the consequences of maximal qrr transcription when the culture cell density is high enough to allow accurate measurements of QS-controlled gene expression using reporter assays or qRT-PCR [19, 20]. The output of the P_qrr1-mRuby3 reporter was low in the WT, luxO D61E, and ΔluxT V. harveyi strains (Fig 3B). This result was expected because qrr1 exhibits only low-level expression in V. harveyi, even at LCD [19]. Eight-fold higher expression of P_qrr1-mRuby3 occurred in the luxO D61E ΔluxT V. harveyi strain (Fig 3B). Regarding the P_luxO-mRuby3 reporter, compared to the WT, the output was lower in the V. harveyi strains harboring luxO D61E (Fig 3C). This result was also expected because a negative feedback loop exists between LuxO-P and luxO [28]. What is crucial is that elimination of luxT caused no change in P_luxO-mRuby3 reporter expression compared to WT and caused no further change in the luxO D61E mutant (Fig 3C). Together, the qrr1 and luxO reporters show that LuxT does not regulate luxO. Rather, LuxT represses qrr1 transcription.

The distinct level of in vivo expression displayed by each qrr gene in V. harveyi has been interpreted to suggest that, beyond being controlled by LuxO-P, each qrr gene is controlled independently by other regulators [19]. Fig 3B shows that LuxT is one such regulator of qrr1. To investigate whether LuxT also regulates qrr2-5, levels of all five Qrr sRNAs were measured using qRT-PCR in WT, ΔluxT, luxO D61E, and luxO D61E ΔluxT V. harveyi strains. Confirming the reporter assay results, Qrr1 levels were ~4 fold higher in the luxO D61E ΔluxT strain than in the other three strains (Fig 3D). While increased levels of Qrr2-5 were detected in the luxO D61E strain compared to WT, deletion of luxT did not cause any additional changes (Fig 3D). Verification of the qRT-PCR results comes from analyses of mRuby3 transcriptional reporters to qrr2-5. All four reporters displayed higher activity in the luxO D61E V. harveyi strain than in WT, and deletion of luxT had no effect (S3 Fig). Therefore, among the qrr genes, LuxT exclusively represses qrr1.

LuxT activates luxCDABE via a mechanism that is independent of Qrr1

Our next goal was to investigate how LuxT activates expression of luxCDABE, given that the mechanism is not via repression of luxO. The Qrr sRNAs repress luxR translation, and therefore they indirectly repress luxCDABE (Fig 1). Thus, an obvious possibility is that LuxT repression of qrr1 activates luciferase. However, luxR is downstream of Qrr1 (Fig 1) and Fig 2E shows that deletion of luxT does not significantly alter LuxR levels at LCD, suggesting that LuxT does not control luciferase via a LuxR-dependent mechanism. To validate this finding, we tested whether Qrr1 is required for LuxT to activate light production. To do this, we measured bioluminescence from a V. harveyi ΔluxT mutant and compared it to that made by a ΔluxT Δqrr1 double mutant. Both strains exhibited the identical phenotype: ~10-fold reduced light production relative to WT V. harveyi and the Δqrr1 mutant (S4 Fig). Thus, LuxT activation of luciferase occurs by a mechanism that is independent of Qrr1.

We next tested the possibility that LuxT directly activates luxCDABE transcription. The luxCDABE promoter and regulatory region extend approximately 350 bp upstream of the luxC start codon [34–36]. To determine if LuxT binds within this region, we amplified six overlapping DNA fragments from -405 to +81 relative to the luxC start codon (S5A Fig). Compared to the avid binding of LuxT to the qrr1 promoter (Fig 2A), LuxT bound the luxC promoter only very weakly. Specifically, binding to all the luxC promoter-containing DNA fragments was comparable to the binding of LuxT to control (E. coli lacZ) DNA (Figs 2A and S5B–S5D) with modestly stronger binding to Probe 3 (S5C Fig). As another test for direct LuxT activation of luciferase, we introduced plasmid-borne arabinose-inducible luxT and a plasmid with IPTG-inducible luxR into recombinant E. coli carrying luxCDABE. LuxR is a direct activator of luxCDABE [21, 36, 37]. As expected, induction of luxR drove increased light production compared to the empty vector control (S6A Fig). By contrast, induction of luxT did not increase light production in the presence or absence of luxR (S6A Fig). We confirmed that luxT was expressed from the plasmid using qRT-PCR (S6B Fig). We note that induction of luxT expression in E. coli caused a modest growth defect (S6C Fig). In conclusion, we find no evidence that LuxT directly activates luxCDABE.

To further investigate the mechanism underlying LuxT activation of luciferase, we probed whether LuxT functions via other known QS components. To do this, we compared the bioluminescence profiles of the V. harveyi Δqrr1-5, ΔluxO, and luxO D61E strains to the identical strains lacking luxT (S7A–S7C Fig). We also included a test of the VIBHAR_RS03920 gene (S7D Fig), a homolog of Vibrio parahaemolyticus swrZ. In V. parahaemolyticus, SwrT, the LuxT equivalent, represses swrZ encoding a GntR family transcription factor, which in turn, represses lateral flagellar (laf) genes [38]. We considered that in V. harveyi, LuxT could repress VIBHAR_RS03920, which could repress luxCDABE. In all four cases, introduction of the luxT deletion reduced light output (S7A–S7D Fig). Thus, LuxT activates luxCDABE by a mechanism that does not require qrr1-5, luxO, or VIBHAR_RS03920. We could not perform a similar experiment to assess whether LuxT regulation of luxCDABE is LuxR-dependent because the ΔluxR mutant makes no light. However, as mentioned above, LuxR protein levels are similar in WT and ΔluxT V. harveyi (Fig 2E), and moreover, there are no significant differences in luxR or aphA transcript levels between WT and ΔluxT V. harveyi at LCD (S8 Fig). Thus, LuxT affecting luxCDABE expression via regulation of luxR does not seem a reasonable possibility. To conclude, unfortunately, we did not discover the mechanism by which LuxT activates luciferase. We do know that the mechanism is likely indirect and that the component that connects LuxT to luxCDABE is not any of the regulators in the V. harveyi QS pathway. From here forward, we focus on the consequences of LuxT regulation of qrr1. In future studies, we hope to define the mechanism by which LuxT activates light production.

LuxT controls target genes via repression of qrr1

Only low-level expression of qrr1 occurs in WT V. harveyi, including at LCD, and that feature has made it difficult to detect Qrr1-mediated regulatory effects in vivo. Based on our discovery of LuxT repression of qrr1, we hypothesize that LuxT activity could mask Qrr1 function in vivo. If so, LuxT would indirectly activate the known Qrr1-repressed mRNA targets. To test this possibility, we used qRT-PCR to compare the levels of Qrr1 mRNA targets in V. harveyi luxO D61E to that in V. harveyi luxO D61E ΔluxT. We assayed the 14 Qrr1 target genes that lie outside the QS pathway [30] as well as luxR and luxMN, Qrr1 targets that function inside the QS system [19, 26, 29]. Deletion of luxT caused a significant decrease in the mRNA levels of 9 of the 16 tested genes (S9A Fig). Thus, we suspected that LuxT activated expression of the 9 genes via repression of qrr1. To test this prediction, we compared transcript levels of the 9 genes in V. harveyi luxO D61E, V. harveyi luxO D61E ΔluxT, V. harveyi luxO D61E Δqrr1, and V. harveyi luxO D61E Δqrr1 ΔluxT. To our surprise, in all cases, the two strains lacking luxT possessed lower levels of the transcripts than did the two strains possessing luxT (S9B Fig). These data show that these target genes are controlled by LuxT in a Qrr1-independent manner.

The data in S9B Fig inspired us to expand our LuxT/Qrr1 regulatory model to include two key findings: (1) LuxT represses qrr1, encoding a sRNA that post-transcriptionally regulates target genes (Fig 3 and [30]), and (2) LuxT also activates expression of the same target genes, independently of Qrr1. Thus, we propose that LuxT functions by two mechanisms to activate expression of the 9 target genes, one transcriptionally and one post-transcriptionally: LuxT is a transcriptional activator of the target genes and LuxT additionally activates the target genes by repressing their repressor, Qrr1.

To test the above model, we focused on the four most highly LuxT-regulated target genes: VIBHAR_RS11785, VIBHAR_RS11620, VIBHAR_RS16980, and VIBHAR_RS25670. First, to examine whether LuxT indeed activates their transcription, we eliminated Qrr-dependent regulation using a V. harveyi Δqrr1-5 strain. In all four cases, transcript levels were lower in the Δqrr1-5 ΔluxT strain than in the Δqrr1-5 strain. Complementation with luxT expressed from a plasmid restored the transcript levels, confirming that LuxT activates the expression of these genes via a Qrr-independent mechanism (Fig 4A). To demonstrate that LuxT control of these genes is exerted at the level of transcription, we made lux transcriptional reporters and measured their outputs in luxA::Tn5 and luxA::Tn5 ΔluxT V. harveyi strains. Using a luxA null mutant for this analysis ensured that all light production came from the transcriptional fusions. All four reporters exhibited lower activity in the luxA::Tn5 ΔluxT strain than in the luxA::Tn5 strain (~400, 4, 48, and 7-fold lower activity for, respectively, VIBHAR_RS11785, VIBHAR_RS11620, VIBHAR_RS16980, and VIBHAR_RS25670, S10 Fig). These data confirm an aspect of our model: LuxT activates transcription of these target genes.

The other tenet of our model, that LuxT activates expression of the target genes via repression of qrr1 cannot be detected by the above qRT-PCR assay (S9B Fig). Fig 4B and 4C depicts the issue. In WT V. harveyi, transcription of qrr1 is repressed by LuxT. Therefore, deletion of qrr1 has no effect on target gene regulation (Figs 4B and S9B). In ΔluxT V. harveyi, qrr1 expression is de-repressed. However, in the absence of the LuxT activator, transcription of the target genes does not occur. Thus, although Qrr1 is present, its mRNA targets are absent, so again regulation via Qrr1 does not occur (Fig 4C).

To circumvent these issues and probe the connection of LuxT to Qrr1 in post-transcriptional regulation of the four target genes, we used a strategy in which we eliminated LuxT transcriptional control of the target genes to unmask post-transcriptional effects. To accomplish this, we constructed translational fusions to the fluorescent protein mVenus. DNA upstream of each target gene containing the site that base pairs with Qrr1 and the ribosome-binding site was cloned in frame with mVenus downstream of the tetracycline-inducible tetA promoter. Therefore, the fusions were constitutively transcribed following addition of aTc, irrespective of the presence or absence of luxT. Analogously designed translational reporters were previously shown to be repressed in E. coli following qrr1 overexpression [30]. We confirmed that the reporters are all activated by aTc and repressed following overexpression of qrr1 in V. harveyi (S11 Fig).

The translational mVenus reporter fusions were used to test the aspect of our model in which we predict that LuxT activates target genes post-transcriptionally via qrr1 repression. Reporter activities from the four target gene constructs were measured in the following LCD-locked V. harveyi strains: luxO D61E, luxO D61E ΔluxT, luxO D61E Δqrr1, and luxO D61E Δqrr1 ΔluxT. The results for all four reporters were similar (Fig 5A–5D). The luxO D61E strain exhibited higher reporter activity than the luxO D61E ΔluxT strain, presumably due to the de-repression of qrr1 that occurs in the absence of LuxT. Importantly, deletion of luxT in the luxO D61E Δqrr1 strain had no effect on reporter translation (Fig 5A–5D, compare luxO D61E Δqrr1 and luxO D61E Δqrr1 ΔluxT bars). We conclude that LuxT post-transcriptionally regulates the four tested genes in a Qrr1-dependent manner. We note that higher translation of the reporters occurred in the luxO D61E Δqrr1 strains than the luxO D61E strains containing qrr1 (Fig 5A–5D). This pattern is consistent with Qrr1 functioning as a repressor, and we interpret the result to mean that when the qrr1 gene is present, residual Qrr1 production occurs, including in the presence of LuxT. We presume that this pattern cannot be observed in the qRT-PCR analyses (S9B Fig) because Qrr1 represses translation of target genes by a sequestration mechanism that does not significantly alter mRNA levels [25, 30].

The four genes that are regulated transcriptionally by LuxT and post-transcriptionally by LuxT via Qrr1 encode a peptidase (VIBHAR_RS11785), an aerolysin toxin (VIBHAR_RS11620), a chitin disaccharide deacetylase (VIBHAR_RS16980), and a protein involved in export of capsular polysaccharide (VIBHAR_RS25670). Interestingly, all four genes encode secreted public goods or products involved in secretion of public goods (i.e., VIBHAR_RS25670), a class of components that are commonly controlled by QS. We focus on the aerolysin toxin (VIBHAR_RS11620) here to probe in vivo LuxT and Qrr1 regulation. Secreted aerolysin-like toxins form pores in eukaryotic cells, and in the case of red blood cells, cause lysis [39]. Thus, aerolysin hemolytic activity can be assessed by growing bacteria on blood agar plates and monitoring them for zones of clearance. We used this assay to test if LuxT and Qrr1 influence aerolysin secretion according to our dual-mechanism model (Fig 4B). First, the V. harveyi luxO D61E strain exhibited modest clearing, whereas no clearing occurred around the luxO D61E ΔluxT strain (Fig 5E). This result is consistent with LuxT functioning as an activator of aerolysin production. Second, compared to the luxO D61E strain, luxO D61E Δqrr1 showed increased hemolytic activity (Fig 5E). This result can be explained by Qrr1-mediated post-transcriptional repression of VIBHAR_RS11620 (Fig 5B). Finally, the luxO D61E Δqrr1 ΔluxT strain did not display hemolytic activity (Fig 5E). In agreement with our model (Fig 4B and 4C), the transcriptional effect of LuxT overrides the post-transcriptional effect of Qrr1. The hemolysis activities of the identical strains were also quantified using a liquid assay (S12 Fig). Analogous results were obtained for the four strains except that the luxO D61E strain exhibited a level of hemolytic activity similar to that of the luxO D61E Δqrr1 strain. Possibly, this discrepancy is due to the different growth conditions used for the plate and liquid hemolysis assays.

LuxT represses qrr1 in A. fischeri

In members of the Vibrionaceae family, AI structures and the types of proteins employed as receptors vary between species. However, LuxO is conserved in all sequenced vibrio species [40] and LuxT is also often present [38, 41–43] and we address this further in the next section. We wondered whether LuxT-mediated repression of qrr1 is V. harveyi specific or whether LuxT has this function in other Vibrionaceae species. To explore this question, we tested three species, Vibrio cholerae, V. parahaemolyticus, and A. fischeri in experiments analogous to those in Fig 3B. Plasmids harboring transcriptional reporter fusions to qrr1 from each representative species were introduced into WT, ΔluxT, luxO D61E, and luxO D61E ΔluxT strains of those species. As mentioned, luxT is called swrT in V. parahaemolyticus, and the LCD-locked LuxO-P mimetic in A. fischeri is luxO D55E. In V. cholerae, LuxO D61E activated the P_qrr1-luxCDABE reporter relative to WT, however elimination of luxT did not affect reporter activity in either strain (S13A Fig). Activity from the V. parahaemolyticus P_qrr1-mRuby3 reporter remained low in all four strains (S13B Fig). Thus, we do not find evidence for qrr1 repression by LuxT in V. cholerae or by SwrT in V. parahaemolyticus. We note, however, that regarding V. parahaemolyticus, we cannot rule out the presence of an additional qrr1 repressor that masks LuxT function and maintains qrr1 transcription at an especially low level.

A. fischeri is distantly related to V. harveyi and, curiously, A. fischeri only encodes a single qrr gene, qrr1, and Qrr1 post-transcriptionally represses LitR, the LuxR homolog (Fig 6A) [44]. Through additional regulatory steps, activation of LitR drives the downstream activation of luxCDABE [45]. The A. fischeri P_qrr1-mRuby3 reporter exhibited low-level expression in the WT, ΔluxT, and luxO D55E strains (Fig 6B). However higher fluorescence was emitted in the A. fischeri luxO D55E ΔluxT strain (Fig 6B). Thus, as in V. harveyi, LuxT is a repressor of qrr1 in A. fischeri.

The redundancy among the five Qrr sRNAs in V. harveyi prevents the elimination of qrr1 from driving large effects on LuxR levels (Fig 2E and [19]), and in the context of the present work, masks the consequences of deletion of luxT. Because no Qrr redundancy exists in A. fischeri, we predicted that LuxT repression of qrr1 would affect LitR levels. Indeed, compared to the A. fischeri luxO D55E strain, the luxO D55E ΔluxT strain showed a 4-fold reduction in litR transcript levels (Fig 6C). To test if this manifestation of LuxT occurs via repression of qrr1, we measured litR transcription in A. fischeri luxO D55E Δqrr1 and A. fischeri luxO D55E Δqrr1 ΔluxT. There was no significant difference in litR transcript levels showing that LuxT activates litR expression in a Qrr1-dependent manner (Fig 6C). The differences in litR transcript levels observed between the luxO D55E and luxO D55E Δqrr1 strains are likely a result of Qrr1 feedback control of luxO [28]. To determine if the observed LuxT-dependent effects on LitR likewise affect downstream expression of luciferase, we measured bioluminescence in the four A. fischeri strains. Indeed, the luxO D55E ΔluxT strain made less light than the luxO D55E strain (Fig 6D). The luxO D55E Δqrr1 and luxO D55E Δqrr1 ΔluxT strains emitted similar levels of light showing that LuxT controls light production in A. fischeri via regulation of qrr1 (Fig 6D). We conclude that LuxT is a repressor of qrr1 in A. fischeri, and because Qrr1 is the sole Qrr, LuxT has a more major role in controlling the overall QS state in A. fischeri than in V. harveyi. We discuss possible advantages of the different regulatory arrangements below.

Putative LuxT regulation of qrr1 is diversified in the Vibrionaceae family

Members of the Vibrionaceae family can be divided into two classes, those encoding a single qrr upstream of luxO, and those encoding multiple qrr loci [20, 44]. Species with multiple qrr genes always encode qrr1 upstream of luxO, suggesting that qrr1 is the ancestral gene. Our finding of LuxT repression of qrr1 in both V. harveyi and A. fischeri inspired us to investigate whether LuxT is conserved among all Vibrionaceae family members, and if so, whether LuxT possesses an evolutionary pattern that corresponds to that of the Qrr sRNAs. To compare luxT and qrr phylogenies, we scanned all Vibrionaceae sequenced genomes to identify qrr genes, expanding on previous analyses [19, 20, 44]. The majority of species within the Vibrio genus encoded multiple qrr loci, most often 4 or 5 qrr genes, like V. cholerae and V. harveyi, respectively (Fig 7A). All members of non-Vibrio genera encoded only a single qrr gene, like A. fischeri, except for Photobacterium galatheae, which had no putative qrr gene (Fig 7A). Analogous examination of the genomes for luxT homologs showed that luxT genes exist in most Vibrionaceae species possessing one and multiple qrr genes (Fig 7B). Within the Vibrio genus, species lacking apparent qrr genes also lacked luxT homologs, and the luxT genes were more similar to V. harveyi luxT in species with multiple qrr genes than were the luxT genes in species possessing only a single qrr gene.

To predict whether LuxT does or does not control qrr1 expression in a particular species, we compared the DNA sequences upstream of qrr1 in the four Vibrionaceae species tested in our experiments. The σ⁵⁴ binding sites are highly conserved among the four species (Fig 7C and [19, 20]), while the LuxT binding regions show less conservation. Thus, harboring a luxT homolog does not necessarily signify that it controls qrr1. The “GGTTAAA” upstream of qrr1 in the LuxT binding region was the most conserved sequence between the species. Consistent with our experimental results, the V. cholerae sequence in this region, i.e., “GATTTG–”, is the most dissimilar from those of the other three species (Fig 7C). This sequence divergence may underlie our finding that LuxT does not regulate qrr1 in V. cholerae (S13 Fig). In V. parahaemolyticus, this region is identical to that in V. harveyi. However, we do not observe LuxT regulation of qrr1 in V. parahaemolyticus (S13 Fig). As mentioned above, qrr1 expression in V. parahaemolyticus may be too low to detect repression by LuxT, possibly due to additional repression by another factor. To more broadly examine the conservation of LuxT binding regions, we also performed phylogenetic analysis comparing the putative LuxT binding regions in the qrr1 promoters of all Vibrionaceae family members possessing both qrr1 and luxT genes. A variety of sequences exist (Fig 7D), and we find no evidence for a correlation between the number of qrr genes and similarity in the upstream LuxT binding regions. It remains possible that the DNA binding domains of LuxT coevolve with the DNA sequences in the LuxT binding regions. Together, our results indicate that while qrr1 and luxT are broadly conserved in Vibrionaceae species, LuxT regulation of qrr1 has diversified. Going forward, we will combine experimental and bioinformatic approaches to pinpoint the precise LuxT binding site, determine its conservation between species, and define the ramifications of particular DNA sequence changes.

Discussion

To survive, bacteria must appropriately respond to fluctuating environments. For marine bacteria such as V. harveyi, successfully competing against a diversity of other microbes and adapting to dynamic microscale nutrient gradients are key [46, 47]. Sensory relays that tune gene expression via transcriptional and post-transcriptional mechanisms enable bacteria to overcome varying environmental challenges [48]. In the context of the present work, QS signal transduction allows bacteria to monitor their changing cell numbers and transition between executing individual and collective activities [49].

In vibrios, one or more Qrr sRNAs function at the core of QS signaling pathways, and thus the concentration of Qrr sRNAs present at any time dictates the QS output response in which hundreds of genes are either activated or repressed. The Qrr sRNAs, and other bacterial sRNAs, are post-transcriptional regulators. Bacterial sRNAs are thought to be especially beneficial regulators due to the low metabolic cost of their production coupled with their fast synthesis and turnover rates, the latter of which can drive rapid changes in target mRNA levels [50, 51]. Moreover, because the QS Qrr sRNAs function by multiple mechanisms (sequestration, catalytic mRNA degradation, coupled mRNA-sRNA degradation, and mRNA translational activation), they can confer distinct timing and expression levels to particular target genes providing “bespoke” QS output responses [25]. These features of sRNAs are presumed to drive dynamic patterns of gene expression that might not be achievable through the use of canonical transcription factors.

Gene duplication has led to the V. harveyi QS circuit harboring five similar Qrr sRNAs [19]. Beyond QS, in bacteria it is common for multiple sRNAs to function redundantly in a single pathway. Presumably, possessing more than one copy of a sRNA gene can increase the available sRNA pool, and in turn, confer increased control over target gene expression. In addition or alternatively, duplication may allow individual sRNA genes to diversify, in sequence and/or in expression pattern, either or both of which can enable differential regulatory effects [52]. Indeed, regarding the V. harveyi Qrr sRNAs, deletion analyses and Qrr quantitation studies have demonstrated that the pool of Qrr sRNAs available to regulate downstream target gene expression increases with increasing numbers of qrr genes. Curiously, however, at least in the laboratory and with luxR as the measured target gene, only four of the five Qrr sRNA genes are required to achieve this effect. Thus, the final qrr duplication event does not appear to enhance regulatory control [19]. Moreover, only low-level production of Qrr1 and Qrr5 have been documented, suggesting that those two sRNAs do not contribute dramatically to changes in the levels of the sRNA pool [19]. These findings, together with the knowledge that the qrr promoter regions vary, has led us to hypothesize that some or all of the qrr genes may be subject to additional control by as yet undefined regulatory components.

Here, our discovery of V. harveyi LuxT as a repressor of qrr1 provides evidence for a QS model in which individual qrr genes are uniquely regulated. While LuxT repression of qrr1 does not affect expression of the genes encoding the master QS regulators LuxR and AphA, it does alter expression of a subset of Qrr1 target genes. Separate from its role as a qrr1 repressor, we also found that LuxT controls the same set of Qrr1 target genes at the transcriptional level. A regulatory strategy in which control is exerted at two levels, via a transcriptional regulator and a post-transcriptional sRNA, occurs in other systems and is proposed to prevent leaky target gene expression and to alter target gene expression dynamics [53, 54]. In the case of LuxT, at least four genes are subject to such control, and they encode a protease, an aerolysin toxin, a chitin deacetylase, and a gene involved in capsular polysaccharide secretion. Notably, all four gene products are secreted, perhaps emphasizing the need for especially tight control of public goods production. We imagine that LuxT initially evolved to transcriptionally activate this set of target genes and later incorporated repression of V. harveyi qrr1 to reinforce activation at the post-transcriptional level. Thus, the gene duplication events that generated qrr redundancy in V. harveyi also provided the required substrate for regulation by LuxT, ultimately enabling finely tuned expression of select members of the QS regulon that rely on Qrr1, while avoiding blanket alteration of the QS response. Our discovery of LuxT repression of V. harveyi qrr1 hints that analogous regulators may exist that uniquely control qrr2-5.

The luxT gene is conserved among Vibrionaceae bacteria, but we only observe LuxT repression of qrr1 in two of four tested species, V. harveyi and A. fischeri. These two species harbor five and one qrr genes, respectively. More broadly, our phylogenetic analyses of the LuxT binding regions upstream of Vibrionaceae qrr1 genes show that this DNA sequence has diversified, and consistent with our results, may signify that LuxT represses only a subset of qrr1 genes. Further investigation is necessary to understand the regulatory logic underlying LuxT repression of qrr1 in some species but not in others. We can speculate on these different circuit arrangements. To do so, we consider the diversity of QS system components and regulatory architectures present in Vibrionaceae species. We know from our and previous phylogenetic analyses that luxO is highly conserved in vibrios, and species commonly possess from one to five qrr genes [20, 40, 44, 55]. Beyond these two core components, Vibrionaceae QS systems vary with respect to the number and structures of QS AIs, the number, subcellular locations, and signal relay mechanisms of the QS receptors, and the number and identities of the downstream target genes [1, 56, 57]. Presumably, the differences in QS system architectures represent the outcomes of distinct selective pressures experienced by particular species over evolutionary time. As species diverged, a common set of parts were mixed and matched, duplicated, and their placements in the regulatory hierarchies altered with LuxO and the Qrr sRNAs remaining as the core of the QS networks. Similar, but not identical QS systems emerged, each presumably capable of promoting ideal biology for a given species. With regard to the present work, LuxT represents one more component that evolution can insert into Vibrionaceae QS systems in different places in the various hierarchies to enable it to specialize for each species.

Lastly, LuxT is a member of the bacterial TetR family of transcriptional regulators, a widely distributed family of proteins possessing characteristic helix-turn-helix DNA-binding domains [58]. V. harveyi LuxR is a member of this same protein family. Prior to our discovery of V. harveyi LuxT as a qrr1 repressor, the functions of some LuxT homologs had been studied including in V. parahaemolyticus, A. fischeri, Vibrio vulnificus, and Vibrio alginolyticus. In V. parahaemolyticus, the LuxT homolog, SwrT, activates genes promoting lateral-flagellar-driven swarming, enabling translocation across surfaces [38, 59–61]. LuxT is a transcriptional activator of siderophore biosynthetic genes in A. fischeri [43]. In V. vulnificus and V. alginolyticus, LuxT is reported to control QS via regulation of expression of the luxR homologs [41, 42]. Additionally, the V. alginolyticus ΔluxT mutant is defective for virulence in a zebrafish infection model [42]. Whether Qrr1 acts as a LuxT-controlled intermediary in these other vibrio pathways has not been investigated. These earlier studies, together with our findings that LuxT also controls gene expression independent of Qrr1 in V. harveyi hint that LuxT is a global regulator of gene expression in Vibrionaceae. Future transcriptomic analyses will be used to identify the set of genes comprising the V. harveyi LuxT regulon and to fully define which LuxT target genes are Qrr1 dependent and which are Qrr1 independent. Similar analyses in other Vibrionaceae species could reveal which functions of LuxT are general and which are species specific. Finally, it will be of particular interest to investigate the environmental signals that control luxT expression and LuxT activity. Under standard laboratory conditions, we have not observed variation in luxT mRNA or protein levels, however, examining its activity under conditions that more closely mimic nature may reveal how luxT itself is regulated.

Materials and methods

Bacterial strains and culture conditions

V. harveyi strains were derived from V. harveyi BB120 (BAA-1116) [62]. A. fischeri strains were derivatives of A. fischeri ES114 [63]. V. cholerae strains were derived from V. cholerae C6706str2 [64], and V. parahaemolyticus strains were derived from V. parahaemolyticus BB22OP (LM5312) [65]. E. coli BW25113 was used for heterologous gene expression and E. coli S17–1 λpir was used for cloning. All strains are listed in S1 Table. Vibrio and Aliivibrio strains were grown at 30°C shaking in either Luria Marine (LM) medium or minimal Autoinducer Bioassay (AB) medium, the latter supplemented with 0.4% vitamin-free casamino acids (Difco) [4, 66]. E. coli strains were grown shaking at 37°C or at 30°C in LB medium. Antibiotics were added as follows (μg mL^-1): ampicillin, 100; chloramphenicol, 10; kanamycin, 100; polymyxin B, 50; and tetracycline, 10. Induction of genes on plasmids was accomplished by the addition of 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) (Thermo Fisher), 0.2% arabinose (Sigma), or 100 ng mL^-1 anhydrotetracycline (aTc) (Takara), as necessary.

DNA manipulation and strain construction

PCR reactions were carried out with either KOD Hot Start DNA Polymerase (Sigma) or iProof DNA Polymerase (Bio-Rad). Oligonucleotides were purchased at Integrated DNA Technologies (IDT) and are listed in S2 Table. A DNA fragment containing the randomized LuxT binding region was synthesized by IDT. Cloning was performed using isothermal DNA assembly with the Gibson Assembly Master Mix (New England Biolabs) [67]. All plasmids were validated by sequencing (Genewiz) and are listed in S3 Table. Plasmids that enable overexpression of genes are designated with a lowercase p (e.g. pqrr1). For reporter fusion constructs, a capital P designates the promoter that drives transcription (e.g. P_qrr1-mRuby3). Transcriptional reporters to luxO and to qrr1 included approximately 200 bp of promoter DNA upstream of mRuby3. Transcriptional reporters to qrr2-5, VIBHAR_RS11785, VIBHAR_RS11620, VIBHAR_RS16980, and VIBHAR_RS25670 contained approximately 300 bp of promoter DNA. A consensus ribosome binding site was included to drive translation. The putative base pairing regions between Qrr1 and the VIBHAR_RS11785, VIBHAR_RS11620, and VIBHAR_RS25670 mRNAs were excluded from those reporter constructs. Due to its location far upstream of the gene, the putative Qrr1 base pairing region for VIBHAR_RS16980 could not be excluded [30]. Translational reporters employing mVenus were designed using a previously described method and transcribed from the aTc inducible tetA promoter [30, 68]. Plasmids were introduced into E. coli by electroporation using a Bio-Rad Micro Pulser. Plasmids were introduced into Vibrio and Aliivibrio strains via conjugation with E. coli S17-1 λpir. V. harveyi, V. cholerae, and V. parahaemolyticus exconjugants were selected on agar plates with polymyxin B. A. fischeri exconjugants were selected on agar plates containing ampicillin. Chromosomal alterations in Vibrio and Aliivibrio strains were generated using the pRE112 suicide vector harboring the sacB counter-selectable marker as previously described [34, 43, 69]. Selection for the second crossover event was performed on LM agar plates containing 15% sucrose (Sigma). Mutations were validated by PCR and/or sequencing.

LuxT-6xHis protein production and purification

The DNA encoding LuxT-6xHis was cloned into the pET-15b vector and the protein was overexpressed in E. coli BL21 (DE3) using 0.4 mM IPTG at 18°C for overnight growth. Cells were pelleted at 16,100 x g for 10 min and resuspended in lysis buffer (25 mM Tris-HCl pH 8, 150 mM NaCl) supplemented with 1 mM DTT, 2 mM PMSF, and 5 μM DNase I. The cells were lysed using sonication and subjected to centrifugation at 32,000 x g for 1 h. The LuxT-6xHis protein was purified from the clarified supernatant by Ni-NTA Superflow resin (Qiagen). Following washes with lysis buffer containing 20 mM Imidazole, the protein was eluted using lysis buffer containing 300 mM Imidazole. The collected elution fraction was loaded onto a HiTrap Q column (GE Healthcare) and further purified using a linear gradient of buffer A (25 mM Tris-HCl pH 8, 1 mM DTT) to buffer B (25 mM Tris-HCl pH 8, 1 M NaCl, 1 mM DTT). Peak fractions were pooled, concentrated, and subjected to a Superdex-200 size exclusion column (GE Healthcare) in gel filtration buffer (25 mM Tris-HCl pH 8, 100 mM NaCl, 1 mM DTT). The protein was concentrated, flash frozen, and stored at -80°C.

Electrophoretic mobility shift assays (EMSAs)

Oligonucleotide primers used to amplify DNA probes are listed in S2 Table. Reaction mixtures of 10 μL volume containing 20 nM dsDNA probe and 1:2 serial dilutions of LuxT-6xHis in low salt buffer (25 mM Tris pH 8, 50 mM NaCl) were incubated at room temperature for 15 min. LuxT-6xHis dimer concentrations ranged from 10 nM (0.5x) to 320 nM (16x). After incubation, 2.5 μL of 5X loading buffer (LightShift EMSA Optimization and Control Kit, Thermo) was added to the mixtures, and the samples were loaded onto a 6% Novex TBE DNA retardation gel (Thermo) at 4°C. Gels were subjected to electrophoresis in 1x TBE buffer at 100 V for 1.75 h. Gels were stained using SYBR Green I Nucleic Acid Gel Stain (Thermo) for 30 min. After five washes with 20 mL 1x TBE, gels were imaged using an ImageQuant LAS 4000 imager under the SYBR Green setting.

Bioluminescence assays

Cells from overnight cultures of V. harveyi were pelleted by centrifugation at 21,100 x g (Eppendorf 5424) and resuspended in fresh LM medium. Flasks containing 25 mL of LM medium were inoculated with the washed cells, normalizing each culture to a starting OD₆₀₀ = 0.005. Culture flasks were incubated with shaking at 30°C. Every 45 min, bioluminescence and OD₆₀₀ were measured using a Tri-Carb 2810 TR scintillation counter and DU800 spectrophotometer, respectively. A. fischeri cultures were grown as described for V. harveyi and bioluminescence was measured using a Tri-Carb 2810 TR scintillation counter when the OD₆₀₀ = 1. To assay regulation of luxCDABE by LuxT, E. coli BW25113 harboring three plasmids, described in the legend to S6 Fig, was grown in LB medium for 16 h at 30°C. Cells from cultures were pelleted by centrifugation at 21,100 x g (Eppendorf 5424) and resuspended in PBS. Bioluminescence and OD₆₀₀ were measured as above. RNA was harvested as described below for qRT-PCR analysis of luxT overexpression. Transcriptional output from VIBHAR_RS11785, VIBHAR_RS11620, VIBHAR_RS16980, and VIBHAR_RS25670 lux fusions was measured from V. harveyi strains grown to OD₆₀₀ = 1 in LM medium using a Tri-Carb 2810 TR scintillation counter. P_qrr1-luxCDABE activity was measured in V. cholerae strains using a BioTek Synergy Neo2 Multi-Mode Reader (BioTek, Winooski, VT, USA).

Quantitative real-time PCR analyses

Cells from overnight cultures of V. harveyi or A. fischeri were pelleted by centrifugation at 21,100 x g (Eppendorf 5424) and the cells were resuspended in fresh LM medium. 25 mL LM medium was inoculated with the washed cells, normalizing each culture to a starting OD₆₀₀ = 0.005. The cultures were grown shaking at 30°C. At the desired cell densities, RNA was harvested from three independent cultures using the RNeasy mini kit (Qiagen #74106). RNA levels were normalized to 200 ng/μL and the samples were treated in two sequential reactions with DNase (Turbo DNA-free Kit, Thermo Fisher AM1907). cDNA was generated from 1 μg of RNA using Superscript III Reverse Transcriptase (Thermo Fisher, 18080093) as previously described [19]. Real-time PCR was performed using a QuantStudio 6 Flex Real-Time PCR detection system (Thermo Fisher) and PerfeCTa SYBR Green FastMix (Quantabio, 95074) as previously described [19]. In every case, 10 μL reactions were analyzed in quadruplicate technical replicates. Control reactions were performed with samples lacking reverse transcriptase and with samples lacking cDNA templates. Relative transcript levels were measured and normalized to an internal hfq control gene using a comparative ΔΔC_T method. qRT-PCR primers are listed in S2 Table.

Western blot analyses

Overnight cultures of WT and ΔluxT V. harveyi strains harboring either aphA-3xFLAG or 3xFLAG-luxR at their native loci were pelleted by centrifugation at 21,100 x g (Eppendorf 5424) and resuspended in fresh LM medium. Flasks containing 125 mL LM medium were inoculated with the washed cells, normalizing the starting OD₆₀₀ of each culture to 0.00001. When the cultures reached the desired cell densities, cells equivalent to 1 OD₆₀₀ were pelleted by centrifugation at 2,808 x g for 10 min (Eppendorf 5810 R) and the pellets were flash frozen. Next, cells were lysed by resuspension in 150 μL of buffer containing 1x BugBuster (Sigma), 1x Halt Protease Inhibitors (Thermo Fisher), 0.5% Triton X-100 (Sigma), and 50 μg/mL lysozyme (Sigma). After incubation at room temperature for 30 min, proteins were solubilized in 1x SDS-PAGE buffer for 1 h at 37°C. Samples were loaded onto 4–20% TGX Stain-Free gels (Bio-Rad, #17000435) and subjected to electrophoresis at 50 mA for 30 min. Total loaded protein in the Stain-Free gel was visualized using an ImageQuant LAS 4000 imager using the EtBr setting. A second Stain-free gel was used for Western blot and was loaded with total protein levels normalized according to band intensities on the first gel. The normalization was verified by imaging. A dominant band from this gel image serves as a loading control in Fig 2E. FLAG-tagged protein detection was performed as previously reported [70] using an Anti-FLAG M2-Peroxidase (HRP) antibody (Sigma, A8592) and bands were visualized using an ImageQuant LAS 4000 imager.

Fluorescence reporter assays

Fluorescent reporter plasmids are listed in S3 Table. The primers used to construct them are listed in S2 Table. Cells in overnight cultures of Vibrio or Aliivibrio strains harboring transcriptional or translational fluorescent reporter plasmids were pelleted by centrifugation at 21,100 x g (Eppendorf 5424) and washed in AB medium. AB medium was inoculated with the washed cells, normalizing each to OD₆₀₀ = 0.005. 150 μL of the cultures were transferred to clear-bottom 96-well plates (Corning) in quadruplicate technical replicates. 50 μL of mineral oil was added to each well to prevent evaporation. The plates were shaken at 30°C, and fluorescence and OD₆₀₀ were monitored over a 24 h period using a BioTek Synergy Neo2 Multi-Mode Reader. Relative fluorescence values represent the values when the OD₆₀₀ reached 0.3 or 0.6, as indicated in the figure legends, for each sample. The OD₆₀₀ values are the cell densities at which maximal differences between experimental and control reporter outputs could be measured.

Hemolysis assays

Cells in overnight cultures of V. harveyi were pelleted by centrifugation at 21,100 x g (Eppendorf 5424) and resuspended in fresh LM medium. Culture densities were normalized to OD₆₀₀ = 1, and 2 μL of each culture were spotted onto a TSA plate containing 5% sheep’s blood (Thermo Fisher, R060312). The plates were incubated at 30°C for 72 h and imaged above a white light. To measure hemolysis activity in liquid cultures, V. harveyi strains were grown for 24 h in AB medium. Cells were pelleted by centrifugation at 21,100 x g (Eppendorf 5424), and the clarified culture fluids were filtered through 0.22 μm filters (Sigma, SLGP033RB). Hemolysis of defibrinated sheep’s blood cells (Thomas Scientific, DSB030) was measured as previously described [71, 72]. Briefly, mixtures containing 1% blood cells in PBS and 25% of the filtered fluids were incubated for 2 h at 37°C in in a 96-well plate. 1% blood cells were incubated in ddH₂O or PBS as the positive and negative control, respectively. Following incubation, the plate was subjected to centrifugation at 1,000 x g (Eppendorf 5810 R) for 5 min at 4°C, and 100 μL of the resulting supernatants were transferred to a clean 96-well plate. Absorbance at 415 nm, indicative of blood cell lysis, was measured using a BioTek Synergy Neo2 Multi-Mode Reader.

Bioinformatic analyses

Genomic DNA sequences of 418 Vibrionaceae family members were downloaded from the GenBank database (ftp.ncbi.nlm.nih.gov/genomes/genbank/bacteria/) [73]. To identify genes encoding qrr or luxT, the chromosomes were scanned for regions similar to the template sequences of qrr or luxT. As the query for qrr genes, we used the 3’-most 31 nucleotides of V. harveyi qrr1, which are highly homologous among all the qrr genes in V. harveyi, V. cholerae, A. fischeri, and V. parahaemolyticus [19, 20, 26]. The DNA encoding the entire V. harveyi luxT gene was used as the probe to identify other luxT genes. Local sequence alignments were performed in MATLAB (Mathworks, 2020) using the Smith-Waterman (SW) algorithm [74]. The standard scoring matrix NUC44 (see ftp.ncbi.nih.gov/blast/matrices/) was used to compute similarity scores, which take into account both the length and sequence similarity of the alignment. Cutoff values for the similarity scores yielded from the SW algorithm were set to 30 for qrr genes and 100 for luxT genes. Genes identified as possible luxT homologs were verified to encode TetR family transcriptional regulators. Species lacking either qrr1 or luxT were excluded from further phylogenetic analyses.

Multiple sequence alignments were performed using T-Coffee [75]. Phylogenetic analyses and tree building were performed in MATLAB. To construct the phylogenetic tree based on the putative LuxT binding regions residing upstream of qrr1 genes (see Fig 7), using the maximum-likelihood based Jukes-Cantor model [76], we first computed the pairwise difference scores between the 30 nucleotides upstream of the σ⁵⁴ binding sites in the qrr1 promoter regions for every two species. The unweighted pair group method with arithmetic mean (UPGMA) was subsequently used to progressively build a hierarchy of species clusters [77]. In brief, each species was initially represented by one node. At each clustering step, the pair of nodes with the minimal difference score were clustered into a new node. The arithmetic means of the difference scores between this node pair and each of the other nodes were then assigned to be the difference scores between the newly clustered node and other nodes. The sequence logos were generated by WebLogo [78, 79].

Statistical methods

All statistical analyses were performed using GraphPad Prism software. Error bars correspond to standard deviations of the means of three biological replicates.

Supporting information

S1 Table [pdf]
Strains used in this study.

S2 Table [pdf]
Oligonucleotides used in this study.

S3 Table [pdf]
Plasmids used in this study.

S1 Fig [left]
LuxT binds upstream of .

S2 Fig [pdf]
The 1 locus.

S3 Fig [a]
LuxT does not repress 2-5.

S4 Fig [black]
LuxT activates independently of Qrr1.

S5 Fig [a]
LuxT does not bind the promoter.

S6 Fig [a]
LuxT does not directly activate in . .

S7 Fig [black]
LuxT activation of does not depend on known QS genes.

S8 Fig [black]
LuxT does not regulate and .

S9 Fig [a]
LuxT activates Qrr target mRNAs independently of Qrr1.

S10 Fig [rlu]
LuxT activates the transcription of target genes.

S11 Fig [pqrr1]
Qrr1 overexpression represses translational reporter constructs.

S12 Fig [ddh2o]
LuxT and Qrr1 control aerolysin production.

S13 Fig [a]
LuxT does not appear to control 1 in . or . .

S1 Data [xlsx]
Numerical data for Figs , , , , , , , , , , , , , , .

S2 Data [xlsx]
Numerical data for , , , , , , , , , , , , Figs.