LuxT Is a Global Regulator of Low-Cell-Density Behaviors, Including Type III Secretion, Siderophore Production, and Aerolysin Production, in Vibrio harveyi.

Quorum sensing (QS) is a chemical communication process in which bacteria produce, release, and detect extracellular signaling molecules called autoinducers. Via combined transcriptional and posttranscriptional regulatory mechanisms, QS allows bacteria to collectively alter gene expression on a population-wide scale. Recently, the TetR family transcriptional regulator LuxT was shown to control Vibrio harveyi qrr1, encoding the Qrr1 small RNA that functions at the core of the QS regulatory cascade. Here, we use RNA sequencing to reveal that, beyond the control of qrr1, LuxT is a global regulator of 414 V. harveyi genes, including those involved in type III secretion, siderophore production, and aerolysin toxin biosynthesis. Importantly, LuxT directly represses swrZ, encoding a GntR family transcriptional regulator, and LuxT control of type III secretion, siderophore, and aerolysin genes occurs by two mechanisms, one that is SwrZ dependent and one that is SwrZ independent. All of these target genes specify QS-controlled behaviors that are enacted when V. harveyi is at low cell density. Thus, LuxT and SwrZ function in parallel with QS to drive particular low-cell-density behaviors. Phylogenetic analyses reveal that luxT is highly conserved among Vibrionaceae, but swrZ is less well conserved. In a test case, we find that in Aliivibrio fischeri, LuxT also represses swrZ. SwrZ is a repressor of A. fischeri siderophore production genes. Thus, LuxT repression of swrZ drives the activation of A. fischeri siderophore gene expression. Our results indicate that LuxT is a major regulator among Vibrionaceae, and in the species that also possess swrZ, LuxT functions with SwrZ to control gene expression. IMPORTANCE Bacteria precisely tune gene expression patterns to successfully react to changes that occur in the environment. Defining the mechanisms that enable bacteria to thrive in diverse and fluctuating habitats, including in host organisms, is crucial for a deep understanding of the microbial world and also for the development of effective applications to promote or combat particular bacteria. In this study, we show that a regulator called LuxT controls over 400 genes in the marine bacterium Vibrio harveyi and that LuxT is highly conserved among Vibrionaceae species, ubiquitous marine bacteria that often cause disease. We characterize the mechanisms by which LuxT controls genes involved in virulence and nutrient acquisition. We show that LuxT functions in parallel with a set of regulators of the bacterial cell-to-cell communication process called quorum sensing to promote V. harveyi behaviors at low cell density.

INTRODUCTION

Bacteria have the remarkable ability to rapidly detect and adapt to environmental fluctuations. Often, bacteria employ transcriptional and posttranscriptional regulatory mechanisms to tune gene expression patterns that enhance survival under various conditions (1, 2). Such combined regulatory mechanisms are used in a process called quorum sensing (QS) to monitor and react to changes in the cell density and the species composition of the vicinal community. QS involves the production, release, accumulation, and group-wide detection of signaling molecules called autoinducers (AIs). QS fosters the synchronous execution of collective behaviors, typically ones that are unproductive for an individual bacterium to carry out alone but that become effective when enacted by the group, e.g., bioluminescence, virulence factor production, and biofilm formation (3, 4).

Vibrio harveyi, the focus of the current work, is a model QS bacterium that uses three AIs that act in parallel to control bioluminescence, type III secretion, siderophore production, and hundreds of other traits (5–8). Each AI is detected by a cognate membrane-bound receptor. At low cell density (LCD), AI concentrations are low, and the three unliganded QS receptors act as kinases that drive the phosphorylation of the response regulator LuxO. Phosphorylated LuxO (LuxO-P) activates the transcription of genes encoding five small regulatory RNAs (sRNAs) called the Qrr sRNAs that function posttranscriptionally to activate and repress the translation of the master QS regulators AphA and LuxR, respectively (9–12). Thus, at LCD, AphA is made, and LuxR is not. AphA is responsible for executing the LCD QS program (see Fig. S1A in the supplemental material). At LCD, the Qrr sRNAs also directly control 16 other target mRNAs, and they operate as feedback regulators within the QS signaling pathway (13–15). At high cell density (HCD), accumulated AIs bind to their cognate QS receptors. In the liganded state, the QS receptors act as phosphatases, and phosphate is stripped from LuxO, which inactivates it (16, 17). Thus, at HCD, Qrr sRNA production terminates, AphA is not made, and, in contrast, LuxR is produced (8, 10, 12). LuxR drives the HCD QS regulon (Fig. S1B).

Simplified V. harveyi QS pathway. LCD (A) and HCD (B) V. harveyi QS circuits are shown. Circles, triangles, and squares represent AIs. See the introduction in the main text for details. Download FIG S1, PDF file, 0.02 MB.

The five V. harveyi Qrr sRNAs (Qrr1–5) possess high sequence identity, and Qrr2–5 regulate an identical set of mRNA targets (10, 13). Qrr1 is the outlier. Because it lacks 9 nucleotides that are present in Qrr2–5, Qrr1 fails to regulate aphA and two additional mRNA targets that are controlled by Qrr2–5 (10, 11, 13). The genes encoding the Qrr sRNAs differ in their LCD expression levels: qrr4 is the most highly transcribed of the set, whereas only low-level transcription of qrr1 and qrr5 occurs (10). Recently, we discovered that a transcription factor called LuxT represses the expression of qrr1. LuxT does not regulate qrr2–5. Thus, the exclusive involvement of LuxT in qrr1 regulation provides a mechanism enabling Qrr1 to uniquely control downstream targets (18). Indeed, via the repression of qrr1, LuxT indirectly controls the translation of Qrr1 target mRNAs, including those encoding a secreted protease, an aerolysin toxin, a chitin deacetylase, and a component involved in capsule polysaccharide secretion. LuxT also activates the transcription of this same set of genes by a Qrr1-independent mechanism (18). LuxT repression of qrr1 does not significantly alter LuxR translation, and as mentioned above, Qrr1 does not regulate AphA (11, 18). These findings indicate that LuxT repression of qrr1 tunes the expression of select Qrr1-controlled mRNAs without altering the entire QS regulon.

To date, qrr1 is the only gene identified to be directly controlled by LuxT (18, 19). However, LuxT has been linked to the regulation of additional phenotypes in V. harveyi and other Vibrionaceae, including bioluminescence, siderophore production, virulence factor production, and motility (18, 20–23). These findings, together with our initial knowledge that LuxT can regulate the transcription of target genes independently of Qrr1 in V. harveyi (18), inspired us to investigate the role of LuxT beyond its control of qrr1. Here, we use RNA sequencing (RNA-Seq) to identify the LuxT regulon, revealing LuxT to be a global regulator of ∼414 genes in V. harveyi. We find that LuxT directly represses the V. harveyi swrZ gene encoding a GntR family transcriptional regulator. We use genetic and molecular analyses to show that in V. harveyi, LuxT activates genes required for type III secretion, siderophore production, and aerolysin toxin production, and activation occurs by two mechanisms. One mechanism depends on SwrZ: LuxT represses swrZ, and SwrZ represses target gene expression. The second mechanism is SwrZ independent. Finally, the LuxT-controlled traits identified here are also QS controlled and are enacted primarily at LCD. Therefore, LuxT functions in parallel with QS to establish V. harveyi LCD behaviors. Finally, we analyze luxT and swrZ conservation among Vibrionaceae bacteria, and we demonstrate that via swrZ repression, LuxT also activates genes required for siderophore production in Aliivibrio fischeri.

RESULTS

LuxT is a global regulator that directly represses swrZ transcription.

To define the set of genes regulated by LuxT, we used RNA-Seq to compare the transcriptomes of wild-type (WT) and ΔluxT V. harveyi. We considered transcripts with changes in expression of 2-fold or higher (P < 0.01) to be LuxT regulated, revealing a total of 414 genes: 243 activated and 171 repressed (Fig. 1A; see also Data Set S1 in the supplemental material). One gene, swrZ (VIBHAR_RS03920), stood out due to its dramatic repression by LuxT. swrZ was previously shown to be repressed by the LuxT homolog SwrT in Vibrio parahaemolyticus (22). Specifically, SwrT repressed the transcription of swrZ, and SwrZ repressed the laf genes encoding the lateral flagellar machinery required for V. parahaemolyticus swarming locomotion. Thus, through this cascade, repression of a repressor, SwrT activates V. parahaemolyticus swarming (22). While V. harveyi carries laf genes, their expression levels are extremely low, and V. harveyi swarming has not been documented (24). The V. harveyi laf genes were not revealed by RNA-Seq to be LuxT regulated. Nonetheless, our data demonstrate that the regulatory arrangement in which LuxT represses swrZ exists in V. harveyi: quantitative real-time PCR (qRT-PCR) verified that LuxT represses swrZ expression by ∼100-fold (Fig. 1B).

FIG 1

LuxT is a global transcriptional regulator in V. harveyi that directly represses swrZ. (A) Volcano plot depicting RNA-Seq data comparing the transcriptome of ΔluxT V. harveyi to that of WT V. harveyi. Each data point represents one V. harveyi gene. FC, fold change. A total of 243 genes were significantly activated (black) and 171 genes were significantly repressed (dark gray) by LuxT. (B) qRT-PCR measurements of swrZ transcript levels in WT (black) and ΔluxT (blue) V. harveyi. RNA was isolated from strains grown in LM medium to an OD600 of 0.1. (C) Activity of a plasmid-borne P transcriptional reporter in recombinant E. coli. pControl denotes that the second introduced plasmid is the empty parent vector, and pluxT designates that the second vector harbors arabinose-inducible luxT. Strains were grown in LB to an OD600 of 1 in the absence (black) or presence (gray) of 0.01% arabinose. (D) EMSA showing binding of LuxT-6×His to a 113-bp DNA fragment containing the swrZ promoter (left) and a 107-bp control fragment containing the V. harveyi luxC promoter (right). Previously, it was shown that LuxT does not bind the luxC promoter (18). The DNA probe (20 nM) was incubated with the indicated relative concentrations of the LuxT-6×His dimer. - indicates no protein, 1× indicates 20 nM, and 16× indicates 320 nM. (E) qRT-PCR measurements of luxT transcript levels in WT (black) and ΔswrZ (red) V. harveyi. Growth conditions are the same as those described above for panel B. (F) Activity of a plasmid-borne P transcriptional reporter in ΔluxA (black) and ΔluxA ΔluxT (blue) V. harveyi strains. pControl and pluxT denote the empty parent vector and a vector harboring IPTG-inducible luxT, respectively. Strains were grown in LM medium supplemented with 0.5 mM IPTG to an OD600 of 1. (G) Same as for panel D, for a 99-bp DNA fragment containing the luxT promoter (left) and a 107-bp control fragment containing the V. harveyi luxC promoter (right). (H) Activity of a plasmid-borne P transcriptional reporter in recombinant E. coli. pControl denotes that the second introduced plasmid is the empty parent vector, and pswrZ designates that the second vector harbors IPTG-inducible swrZ. Strains were grown in LB to an OD600 of 1 in the absence (black) or presence (gray) of 0.5 mM IPTG. (I) Model for the LuxT/SwrZ regulatory circuit. In panels C, F, and H, RLU denotes relative light units. In panels B, C, E, F, and H, error bars represent standard deviations of the means from 3 biological replicates. Unpaired two-tailed t tests with Welch’s correction were performed comparing the indicated conditions. ns, not significant (P ≥ 0.05); **, P < 0.01; ****, P < 0.0001.

The V. harveyi LuxT regulon. Genes activated (sheet 1) and repressed (sheet 2) by LuxT in V. harveyi are shown. FC, fold change in gene expression in the ΔluxT strain compared to the WT. Download Data Set S1, XLSX file, 0.1 MB.

To determine if LuxT repression of swrZ is direct, we used recombinant Escherichia coli to isolate LuxT and the swrZ promoter from other V. harveyi regulatory components. Two plasmids were introduced into E. coli. One plasmid harbored a P transcriptional reporter, and the second plasmid carried arabinose-inducible luxT. The induction of luxT expression drove the repression of P in the E. coli strain carrying the pluxT vector, whereas no repression occurred in E. coli carrying the empty control vector (Fig. 1C). We conclude that LuxT directly represses swrZ. Consistent with this supposition, purified LuxT-6×His protein bound to the swrZ promoter in an in vitro electrophoretic mobility shift assay (EMSA), whereas LuxT-6×His did not bind to control DNA (Fig. 1D). The laddering present in the P EMSA may indicate that multiple LuxT-binding sites exist in the swrZ promoter and/or that LuxT oligomerizes when binding the swrZ promoter (Fig. 1D).

Both luxT and swrZ are subject to feedback regulation, and neither luxT nor swrZ is controlled by QS.

To further explore the connections between LuxT and SwrZ in V. harveyi, we assessed whether feedback regulatory loops exist. First, regarding SwrZ regulation of luxT, we find no evidence for SwrZ-mediated feedback onto luxT because luxT transcript levels were similar in WT and ΔswrZ V. harveyi (Fig. 1E); moreover, the overexpression of swrZ in WT V. harveyi did not alter luxT transcription as measured by qRT-PCR (Fig. S2). qRT-PCR validated that swrZ was indeed overexpressed from the pswrZ plasmid (Fig. S2). Second, we investigated autoregulation of luxT. The activity of a P reporter was measured in ΔluxA and ΔluxA ΔluxT V. harveyi strains. V. harveyi is naturally bioluminescent, and luxA encodes a luciferase subunit. Thus, deletion of luxA was required to ensure that all light produced by the test strains came from the P-lux reporter. Figure 1F shows that in strains harboring an empty control vector, P expression level is 6-fold higher in the ΔluxA V. harveyi strain than in the ΔluxA ΔluxT strain. The introduction of a vector expressing luxT restored light production (Fig. 1F). Thus, LuxT activates its own transcription. Unfortunately, autoregulation of luxT could not be tested in recombinant E. coli because the P reporter was not expressed to any detectable level. However, in an in vitro EMSA, LuxT-6×His bound the luxT promoter, providing evidence for a direct autoregulatory role (Fig. 1G). Finally, we investigated direct autoregulation of swrZ. In recombinant E. coli, the induction of swrZ expression repressed a P reporter 30-fold, indicating direct feedback repression (Fig. 1H). We conclude that LuxT and SwrZ can be placed into a regulatory pathway in which LuxT directly represses swrZ. Positive and negative feedback loops exist for LuxT and SwrZ, respectively (Fig. 1I). Below, we speculate on the implications of this regulatory arrangement.

SwrZ does not regulate luxT. Shown are qRT-PCR measurements of luxT and swrZ transcript levels in WT V. harveyi harboring the following plasmids: pControl (black) denotes the empty parent vector, and pswrZ (red) denotes a vector carrying IPTG-inducible swrZ. RNA was isolated from strains grown in LM medium containing 0.5 mM IPTG to an OD600 of 1. Unpaired two-tailed t tests with Welch’s correction were performed comparing the pControl and pswrZ conditions for each gene. ns, not significant (P ≥ 0.05); ***, P < 0.001. Download FIG S2, PDF file, 0.02 MB.

Analysis of the LuxT regulon revealed that a subset of LuxT-controlled genes is also regulated by QS (8) (Data Set S1). One possible explanation for this finding is that regulatory interconnections exist between LuxT and QS. We know that LuxT does not regulate aphA or luxR, so LuxT cannot reside upstream of these two components in the QS cascade (18). Alternatively, QS could control luxT and/or swrZ expression. To test this possibility, we measured luxT and swrZ expression in WT, ΔluxO, and luxO D61E V. harveyi strains. The logic is as follows. The WT strain undergoes the normal LCD-to-HCD QS transition. The ΔluxO strain exhibits “HCD-locked” behavior because in the absence of LuxO activity, no Qrr sRNAs are produced (9, 10) (Fig. S1B). luxO D61E encodes a LuxO-P mimetic, so the strain harboring this mutation exhibits “LCD-locked” behavior in which the Qrr sRNAs are constitutively produced (10, 17) (Fig. S1A). There were no significant differences in luxT or swrZ expression in these strains (Fig. S3A). Consistent with this finding, no dramatic changes in luxT expression occurred over growth as would be expected for a QS-regulated gene (Fig. S3B). As expected from our above-described findings (Fig. 1B), swrZ transcript levels were >100-fold higher in the ΔluxT strain than in WT V. harveyi, but again, cell-density-dependent changes in swrZ expression did not occur in either strain, arguing against QS control (Fig. S3C). We conclude that the expression of luxT and swrZ is not QS regulated. Thus, we infer that QS and LuxT/SwrZ function independently and in parallel in the regulation of a subset of genes in each of their regulons.

QS does not regulate luxT or swrZ in V. harveyi. (A) qRT-PCR measurements of luxT and swrZ transcript levels in WT (black), ΔluxO (green), and luxO D61E (orange) V. harveyi strains. RNA was isolated from strains grown in AB medium to an OD600 of 1. (B) qRT-PCR measurements of luxT transcript levels over growth. RNA was isolated from WT V. harveyi grown in LM medium to different cell densities, as indicated by the OD600 values on the x axis. (C) Same as for panel B, except swrZ transcript levels were measured in WT (black) and ΔluxT (blue) V. harveyi strains. In all panels, error bars represent standard deviations of the means from 3 biological replicates. Standard deviations that are smaller than the symbols are not shown. In panel A, different letters indicate significant differences between strains (P < 0.05 by two-way analysis of variation [ANOVA] followed by Tukey’s multiple-comparison test). Download FIG S3, PDF file, 0.1 MB.

LuxT activates the expression of type III secretion system genes via repression of swrZ.

As mentioned above, LuxT regulates ∼413 genes in addition to swrZ (Fig. 1A and Data Set S1). To gain insight into the LuxT regulon, as examples, we characterize the mechanisms by which LuxT controls genes specifying three V. harveyi functions: type III secretion, siderophore production, and aerolysin production. We begin with type III secretion genes. Type III secretion systems (T3SSs) are syringe-like molecular machines that ferry toxic effector proteins across bacterial inner and outer membranes and into target eukaryotic cells (25–27). In V. harveyi and other vibrios, type III secretion is crucial for virulence (28–30). The structural proteins comprising the V. harveyi T3SS are encoded in four operons (T3SS.1, T3SS.2, T3SS.3, and T3SS.4) (Fig. 2A) (5, 8, 31). Type III secretion is QS regulated in V. harveyi (5). Specifically, T3SS genes are repressed by both AphA and LuxR. The consequence is that T3SS gene expression peaks at low- to mid-cell density when AphA levels have decreased and LuxR levels have not risen (8, 12). QS regulation of type III secretion occurs indirectly through ExsA, the master transcriptional activator of T3SS genes (Fig. 2A). Both AphA and LuxR directly repress exsA (8, 31).

FIG 2

QS and LuxT/SwrZ regulate V. harveyi T3SS gene expression. (A) Schematic of T3SS gene organization in V. harveyi. There are four major T3SS operons, as labeled. All genes shown in blue were identified by RNA-Seq as members of the LuxT regulon. The numbers shown within select genes designate the fold change differences in transcript levels between WT and ΔluxT V. harveyi, as measured by RNA-Seq (see Data Set S1 in the supplemental material). The expression of the underlined genes was measured in panels B and C. (B) qRT-PCR measurements of mRNA levels of the indicated genes in WT (black), ΔluxO (green), and luxO D61E (orange) V. harveyi strains. RNA was isolated from strains grown in AB medium to an OD600 of 1. (C) qRT-PCR measurements of mRNA levels of the indicated genes in WT (black), ΔluxT (blue), ΔswrZ (red), and ΔswrZ ΔluxT (purple) V. harveyi strains. RNA was isolated from strains grown in LM medium to an OD600 of 0.1. In panels B and C, error bars represent standard deviations of the means from 3 biological replicates. Different letters indicate significant differences between strains (P < 0.05 by two-way analysis of variation [ANOVA] followed by Tukey’s multiple-comparison test).

The genes in all four T3SS operons and exsA were revealed by RNA-Seq to be activated by LuxT (Fig. 2A and Data Set S1). To understand the dual control of T3SS genes by QS and LuxT, we engineered a set of strains and developed a companion strategy to probe the regulatory mechanisms. First, we verified LuxT involvement from the RNA-Seq analysis, and we determined if qrr1 is required for LuxT control of T3SS genes. Using qRT-PCR, we quantified the transcript levels of four T3SS genes, hyp, vopN, vscO, and exsD, representing the four T3SS operons (underlined in Fig. 2A). Expression levels were compared between WT and ΔluxT V. harveyi and between Δqrr1 and Δqrr1 ΔluxT V. harveyi. Expression levels were ∼10-, 14-, 8-, and 4-fold lower, respectively, in the ΔluxT strain than in the WT, confirming that LuxT is an activator of T3SS genes (Fig. S4A). Results nearly identical to those shown in Fig. S4A were obtained for the Δqrr1 and Δqrr1 ΔluxT strains, indicating that LuxT activates T3SS genes independently of Qrr1 (Fig. S4B).

LuxT activates T3SS, siderophore, and aerolysin genes by Qrr1-independent mechanisms in V. harveyi. Shown are qRT-PCR measurements of transcript levels of the indicated genes in WT (black) and ΔluxT (blue) V. harveyi strains (A, C, and E) or in Δqrr1 (black) and Δqrr1 ΔluxT (blue) V. harveyi strains (B, D, and F). RNA was isolated from strains grown in LM medium to an OD600 of 0.1. In all panels, error bars represent standard deviations of the means from 3 biological replicates. Unpaired two-tailed t tests with Welch’s correction were performed comparing two conditions, as indicated. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Download FIG S4, PDF file, 0.1 MB.

To verify the mechanism by which QS regulates T3SS genes, we measured the transcript levels of the four representative genes in WT, ΔluxO (HCD-locked), and luxO D61E (LCD-locked) V. harveyi strains. The expression levels of hyp, vopN, vscO, and exsD were >100-fold higher in the luxO D61E strain than in the WT and ΔluxO strains at HCD (Fig. 2B). Consistent with previously published results, high-level expression of T3SS genes occurs at LCD due to LuxR-mediated repression of them at HCD (5, 8). Specifically, compared to the WT, a ΔluxR V. harveyi strain exhibited >100-fold-higher expression levels of the four genes (Fig. S5A). AphA is known to repress T3SS.1 and T3SS.4 in V. harveyi (8). The expression levels of hyp (T3SS.1) and exsD (T3SS.4) were ∼1.6-fold higher in the luxO D61E ΔaphA strain than in the luxO D61E strain; however, the differences were not statistically significant (Fig. S5B). We presume that we did not observe a larger role for AphA here because of differences in our experimental growth conditions compared to those used previously and the use of qRT-PCR compared to microarrays (8).

V. harveyi T3SS genes are repressed at HCD by LuxR. (A) qRT-PCR measurements of transcript levels of the indicated T3SS genes in WT (black) and ΔluxR (gray) V. harveyi strains. RNA was isolated from strains grown in AB medium to an OD600 of 1. (B) Same as for panel A, except transcript levels were measured in luxO D61E (black) and luxO D61E ΔaphA (gray) V. harveyi strains. In both panels, error bars represent standard deviations of the means from 3 biological replicates. Unpaired two-tailed t tests with Welch’s correction were performed comparing two conditions, as indicated. ns, not significant (P ≥ 0.05); *, P < 0.05; **, P < 0.01. Download FIG S5, PDF file, 0.04 MB.

To assess the mechanism by which LuxT controls T3SS gene expression, we first determined whether the LuxT-mediated activation of T3SS genes requires SwrZ by measuring the transcript levels of the four representative T3SS genes in WT, ΔluxT, ΔswrZ, and ΔswrZ ΔluxT V. harveyi strains at LCD. Comparison of expression levels between the WT and ΔluxT strains validates LuxT activation of the genes. Comparison of expression levels between the ΔswrZ and ΔswrZ ΔluxT strains tests whether LuxT activation of genes requires SwrZ. As described above (Fig. S4A), the expression levels of hyp, vopN, vscO, and exsD were lower in the ΔluxT strain than in the WT (Fig. 2C). However, the elimination of luxT did not alter transcript levels in the ΔswrZ background (Fig. 2C). Thus, LuxT activates the expression of all four T3SS operons via a SwrZ-dependent mechanism. Most likely, LuxT represses swrZ, and SwrZ represses T3SS genes. We test this assumption below.

LuxT regulates exsA by SwrZ-dependent and SwrZ-independent mechanisms to activate type III secretion.

To quantify the contributions of QS and LuxT/SwrZ to the regulation of type III secretion, we used Western blot assessment of VopD, a T3SS.1 protein that accumulates in and is secreted from the cytoplasm following the activation of T3SS gene expression (5). We measured VopD levels in WT, ΔluxT, ΔswrZ, and ΔswrZ ΔluxT V. harveyi strains at HCD and in the LCD-locked (luxO D61E) background strain. No VopD was detected in the strains at HCD, consistent with LuxR repression of T3SS genes (Fig. 3A, first four lanes) (5, 31). In contrast, VopD was made in the luxO D61E strain, and deletion of luxT in the luxO D61E strain reduced VopD levels to below our detection limit (Fig. 3A, fifth and sixth lanes). These data mirror our qRT-PCR results from Fig. 2C and confirm the role of LuxT as an activator of type III secretion. To assess whether swrZ is required for LuxT activation of type III secretion, VopD levels were measured in luxO D61E ΔswrZ and luxO D61E ΔswrZ ΔluxT strains. Lower levels of VopD were present in the luxO D61E ΔswrZ ΔluxT strain than in the luxO D61E ΔswrZ strain (Fig. 3A, rightmost two lanes). These results differ from the qRT-PCR results in Fig. 2C, likely because of the different growth conditions required for the qRT-PCR and the VopD Western blot analyses. We summarize the data in Fig. 3A as follows: deletion of luxT reduces VopD production in the presence and absence of swrZ; however, a more dramatic reduction occurs when swrZ is present. To explain these data, we developed and tested the following model for LuxT/SwrZ regulation of T3SS genes.

FIG 3

LuxT activates V. harveyi type III secretion gene expression by two mechanisms. (A) Western blot of cytoplasmic VopD levels in the indicated V. harveyi strains at HCD. “LCD-locked” denotes that the parent strain for the samples in the right four lanes harbors the luxO D61E mutation. Detection of the LuxS protein serves as the loading control, as previously described (5). (B to E) Schematics depicting the outcomes for LuxT regulation of T3SS genes by SwrZ-dependent and SwrZ-independent mechanisms for WT (B), ΔluxT (C), ΔswrZ (D), and ΔswrZ ΔluxT (E) scenarios. See the text for details.

We propose that LuxT activates T3SS genes by two mechanisms: one mechanism is SwrZ dependent, and one mechanism is SwrZ independent. This model is depicted in Fig. 3B to E and is used to explain the data for the LCD-locked strains in the Western blot shown in Fig. 3A (rightmost four lanes). In WT V. harveyi (Fig. 3B), LuxT is present. It represses swrZ, and it also activates T3SS genes by a SwrZ-independent mechanism. The consequence is LuxT activation of T3SS genes. Therefore, high-level VopD production occurs (Fig. 3A, fifth lane). In ΔluxT V. harveyi (Fig. 3C), swrZ is derepressed. The consequence is that SwrZ is produced and available to repress T3SS genes. Therefore, no VopD is produced (Fig. 3A, sixth lane). In the ΔswrZ strain (Fig. 3D), the situation is essentially identical to that of the WT in which swrZ is repressed by LuxT, but in this case, there simply is no swrZ. The consequence is LuxT activation of T3SS genes. Therefore, high-level VopD production occurs (Fig. 3A, seventh lane). Finally, in the case of the ΔswrZ ΔluxT double mutant (Fig. 3E), there is no repression of T3SS genes by SwrZ, and there is no activation by LuxT. Thus, basal-level expression of T3SS genes ensues, and an intermediate amount of VopD is made (Fig. 3A, eighth lane). Key to this model is that VopD levels in the LCD-locked ΔswrZ and ΔswrZ ΔluxT strains are not identical because of the SwrZ-independent mechanism by which LuxT activates T3SS genes (Fig. 3D and E).

To test our model, we assessed whether LuxT activates T3SS gene expression by a SwrZ-independent mechanism. To do this, we introduced a pluxT overexpression vector into a ΔluxR ΔswrZ ΔluxT V. harveyi strain. The ΔluxR mutation was included to eliminate HCD repression of T3SS genes. This strategy is superior to making measurements from RNA collected at LCD from strains harboring luxR. In the latter case, residual LuxR repressor is present (Fig. S5A). In this setup, we measured hyp, vopN, vscO, and exsD transcript levels. We also measured the expression of exsA, which is located immediately upstream of T3SS.4 and, as mentioned above, encodes the master transcriptional activator of T3SS genes. Compared to the empty vector control, the overexpression of luxT caused ∼122-, 74-, 34-, 57-, and 19-fold increases in hyp, vopN, vscO, exsD, and exsA expression, respectively (Fig. 4A). These data demonstrate that LuxT activates the four T3SS operons and exsA by a mechanism that does not require SwrZ. We also assessed swrZ overexpression in this experimental setup. Repression by SwrZ was not observed (Fig. 4A), presumably because in the absence of LuxT activation, T3SS gene expression is minimal, so no detectable repression by SwrZ can occur. To test this supposition, we introduced the pswrZ overexpression construct into a ΔluxR ΔswrZ V. harveyi strain. Indeed, in the luxT+ background, the overexpression of swrZ caused an ∼10-fold reduction in hyp, vopN, and vscO transcript levels and ∼9- and 3-fold reductions in exsD and exsA levels, respectively (Fig. 4B). Thus, SwrZ is a repressor of the four T3SS operons and exsA.

FIG 4

LuxT activates V. harveyi exsA and, in turn, T3SS genes by SwrZ-dependent and SwrZ-independent mechanisms. (A) qRT-PCR measurements of transcript levels of the indicated genes in ΔluxR ΔswrZ ΔluxT V. harveyi harboring the indicated plasmids. pControl (black) denotes the empty parent vector, and pluxT (blue) and pswrZ (red) denote vectors harboring IPTG-inducible luxT and swrZ, respectively. RNA was isolated from strains grown for 16 h in AB medium supplemented with 0.5 mM IPTG. (B) Same as for panel A, for ΔluxR ΔswrZ V. harveyi harboring the indicated plasmids. (C) qRT-PCR measurements of transcript levels of the indicated genes in WT (black), ΔluxT (blue), ΔexsA (gray), and ΔexsA ΔluxT (white) V. harveyi strains. RNA was isolated from V. harveyi strains grown in LM medium to an OD600 of 0.1. (D) Model for QS and LuxT/SwrZ regulation of T3SS genes. In panels A to C, error bars represent standard deviations of the means from 3 biological replicates. In panels A and C, different letters indicate significant differences between strains (P < 0.05 by two-way ANOVA followed by Tukey’s multiple-comparison test). In panel B, unpaired two-tailed t tests with Welch’s correction were performed comparing the indicated pControl and pswrZ conditions. *, P < 0.05.

Previously, it was discovered that T3SS gene expression does not occur in the absence of exsA irrespective of the presence or absence of LuxR. Thus, ExsA is epistatic to QS in the control of T3SS genes (31). Consistent with this finding, the overexpression of exsA overrode repression by LuxR to activate T3SS gene expression at HCD (31). Based on these previously reported data and our identification of LuxT and SwrZ as regulators of T3SS genes, we wondered whether LuxT and SwrZ also exert control over T3SS genes via the regulation of exsA. To test this possibility, we measured hyp, vopN, vscO, and exsD expression in WT, ΔluxT, ΔexsA, and ΔexsA ΔluxT V. harveyi strains. Figure 4C shows that, as expected, the expression levels of all four genes were lower in the ΔluxT strain than in the WT. However, there were no significant differences in hyp, vopN, and vscO transcript levels between the ΔexsA and ΔexsA ΔluxT strains (Fig. 4C). We conclude that exsA is required for LuxT activation of T3SS.1, T3SS.2, and T3SS.3, encompassing hyp, vopN, and vscO, respectively. Because LuxT activation of hyp, vopN, and vscO requires swrZ (Fig. 2C), we can also conclude that regulation by SwrZ requires exsA. Regarding T3SS.4, there were ∼3-fold-fewer exsD transcripts present in the ΔexsA ΔluxT strain than in the ΔexsA strain. It is possible that LuxT activates the T3SS.4 operon encompassing exsD independently of ExsA, or alternatively, based on their sequential orientation in the genome (Fig. 2A), LuxT activates only the exsA promoter, and readthrough transcription occurs for T3SS.4.

Together, our results show that LuxT activates T3SS genes by two mechanisms: one is SwrZ dependent, and one is SwrZ independent. Also, LuxT functions independently of QS to control these genes. QS and LuxT/SwrZ modulate T3SS.1, T3SS.2, and T3SS.3 via control of the expression of exsA encoding a transcriptional activator of T3SS genes. Previous results showed that exsA is repressed by both AphA and LuxR, which constrains its expression to low to mid-cell densities. Our scheme for QS and LuxT/SwrZ regulation of T3SS genes is depicted in Fig. 4D.

QS and LuxT/SwrZ regulate V. harveyi siderophore production.

Similar to type III secretion, the RNA-Seq analysis revealed LuxT to be an activator of siderophore production genes in V. harveyi. Siderophores are small-molecule iron chelators that bacteria produce and secrete to scavenge extracellular iron (32). Companion siderophore uptake systems import siderophore-Fe3+ complexes, facilitating iron acquisition. V. harveyi carries the genes required to produce, secrete, and import two siderophores, amphi-enterobactin and anguibactin (33, 34), and both sets of genes are regulated by LuxT (Fig. 5A and B, respectively, and Data Set S1). V. harveyi produces siderophores at LCD, suggesting QS regulation; however, the mechanism connecting QS to siderophore genes is not defined (6, 35). Using the steps laid out above for the characterization of T3SS gene regulation, we determine how QS, LuxT, and SwrZ control siderophore production.

FIG 5

QS, LuxT, and SwrZ regulate V. harveyi siderophore production. (A and B) Schematic of siderophore gene organization in V. harveyi for amphi-enterobactin (A) and anguibactin (B). All genes labeled in blue were identified by RNA-Seq as members of the LuxT regulon. The numbers shown within or below the genes designate the fold change differences in transcript levels between WT and ΔluxT V. harveyi, as measured by RNA-Seq. The expression of the underlined genes was measured in panels D, F, and G. (C) CAS assay quantitation of siderophore levels in cell-free culture fluids isolated from the indicated V. harveyi strains. Strains were grown for 16 h in AB medium. (D) qRT-PCR measurements of transcript levels of the indicated genes in WT (black), ΔluxT (blue), ΔswrZ (red), and ΔswrZ ΔluxT (purple) V. harveyi strains. RNA was isolated from strains grown in LM medium to an OD600 of 0.1. (E) Same as for panel C. “LCD-locked” denotes that the strains harbor the luxO D61E mutation. (F) qRT-PCR measurements of transcript levels of the indicated genes in ΔluxR ΔswrZ ΔluxT V. harveyi harboring the indicated plasmids. pControl (black) denotes the empty parent vector, and pluxT (blue) and pswrZ (red) denote vectors harboring IPTG-inducible luxT and swrZ, respectively. RNA was isolated from strains grown for 16 h in AB medium supplemented with 0.5 mM IPTG. (G) Same as for panel F, for ΔluxR ΔswrZ V. harveyi harboring the indicated plasmids. (H) Model for Fur, QS, and LuxT/SwrZ regulation of V. harveyi siderophore production. In panels C to G, error bars represent standard deviations of the means from 3 biological replicates. In panels C to F, different letters indicate significant differences between strains (P < 0.05 by two-way ANOVA followed by Tukey’s multiple-comparison test). In panel G, unpaired two-tailed t tests with Welch’s correction were performed comparing the indicated pControl and pswrZ conditions. *, P < 0.05; **, P < 0.01.

First, qRT-PCR validated the RNA-Seq data. The transcript levels of aebC encoding isochorismate synthase were measured as a representative amphi-enterobactin biosynthetic gene (33). Regarding anguibactin, the expression levels of angM and fatD were measured, and they encode a nonribosomal peptide synthetase and an anguibactin-Fe3+ transporter, respectively (36, 37). The transcript levels of aebC, angM, and fatD were ∼3-fold lower in ΔluxT V. harveyi than in the WT (Fig. S4C). Analogous results were obtained for the three genes when measured in Δqrr1 and Δqrr1 ΔluxT strains, showing that LuxT activates siderophore genes independently of Qrr1 (Fig. S4D).

Second, to determine the mechanism by which QS controls siderophore production, we used a chrome azurol S (CAS) dye to quantify siderophore levels in cell-free culture fluids prepared from WT V. harveyi and QS mutant strains. Approximately 24-fold more siderophore was present in fluids from the LCD-locked luxO D61E strain than in fluids from the WT and ΔluxO strains, consistent with the known LCD production pattern (Fig. 5C) (6, 35). We used mutants defective in each siderophore biosynthesis pathway to show that the siderophore detected by the CAS assay was amphi-enterobactin, not anguibactin, as follows: AebF is required for amphi-enterobactin production (33), and the luxO D61E ΔaebF strain had almost no siderophore in its cell-free culture fluids (Fig. S6A). In contrast, the luxO D61E ΔangN strain, which lacks a gene required for anguibactin production (38), did not exhibit reduced siderophore production compared to the luxO D61E strain (Fig. S6A). V. harveyi apparently does not produce anguibactin when grown under standard laboratory conditions, as reported previously (35).

LuxR and Fur repress the production of the V. harveyi amphi-enterobactin siderophore. (A to C) CAS assay quantitation of siderophore levels in cell-free culture fluids isolated from the indicated V. harveyi strains. Strains were grown for 16 h in AB medium. (D) qRT-PCR measurements of transcript levels of the indicated genes in WT (black), Δfur (gray), and Δfur ΔluxT (blue) V. harveyi strains. Strains were grown in LM medium to an OD600 of 0.1. (E) Same as for panel D, for WT (black), ΔluxT (blue), ΔswrZ (red), and ΔswrZ ΔluxT (purple) V. harveyi strains. In all panels, error bars represent standard deviations of the means from 3 biological replicates. In panels A, D, and E, different letters indicate significant differences between strains (P < 0.05 by two-way ANOVA followed by Tukey’s multiple-comparison test). In panels B and C, unpaired two-tailed t tests with Welch’s correction were performed comparing two conditions, as indicated. ns, not significant (P ≥ 0.05); **, P < 0.01. Download FIG S6, PDF file, 0.1 MB.

With respect to V. harveyi QS, three mechanisms can explain how a gene acquires a LCD expression pattern (Fig. S1): it is activated at LCD by AphA; it is activated, directly or indirectly, at LCD by the Qrr sRNAs; or it is repressed at HCD by LuxR. Regarding AphA, siderophore levels were similarly high in fluids from the luxO D61E and luxO D61E ΔaphA V. harveyi strains, so AphA has no role in the regulation of siderophore production (Fig. S6B). Regarding LuxR, the ΔluxR strain possessed ∼16-fold more amphi-enterobactin in its culture fluids than did the WT at HCD (Fig. S6C). Thus, QS control of amphi-enterobactin production occurs via LuxR repression at HCD. Regarding the Qrr sRNAs, at HCD, somewhat less amphi-enterobactin was produced by the ΔluxR strain than by the luxO D61E strain (Fig. S6B and C), hinting that in addition to the primary repressive role of LuxR at HCD, at LCD, the Qrr sRNAs could activate siderophore biosynthesis or export. Possibilities include direct posttranscriptional activation by the Qrr sRNAs, or alternatively, another QS-regulated factor may exist that is involved in amphi-enterobactin gene regulation. We did not investigate this mechanism further.

To position LuxT and SwrZ in the siderophore regulatory pathway, we used qRT-PCR quantitation of aebC, angM, and fatD in WT, ΔluxT, ΔswrZ, and ΔswrZ ΔluxT V. harveyi strains at LCD (Fig. 5D). As shown in Fig. S4C, there were lower mRNA levels in the ΔluxT strain than in the WT, showing again that LuxT activates siderophore genes. Modestly increased expression of the three genes occurred in the ΔswrZ strain compared to the WT, indicating that SwrZ is a repressor of siderophore genes. Finally, the ΔluxT ΔswrZ double mutant expressed siderophore genes at a level intermediate between those of the ΔluxT and ΔswrZ strains. These data resemble the findings in Fig. 3A for LuxT/SwrZ regulation of T3SS genes, suggesting an identical model: LuxT activates siderophore production by both a SwrZ-dependent and a SwrZ-independent mechanism. We validate this model below.

Fur represses V. harveyi siderophore production under iron-rich conditions.

The above-described qRT-PCR measurements of siderophore transcripts showed only modest differences between strains (Fig. 5D), possibly as a consequence of the generally low siderophore gene expression levels that occur under iron-replete conditions due to repression by Fur, the major transcriptional regulator of iron transport. Fur repression is relieved under iron-limiting conditions (39). Indeed, culture fluids used for the CAS assays in Fig. 5C and Fig. S6A to C were prepared from V. harveyi grown in low-iron minimal medium, and a significant amount of the siderophore was produced. To assess the role of Fur in the regulation of V. harveyi siderophore production, aebC, angM, and fatD transcript levels were compared between WT and Δfur V. harveyi. In iron-rich medium, in the absence of fur, the expression levels of the three test genes were ∼284-, 119-, and 66-fold higher than when fur was present (Fig. S6D). Deletion of luxT in the Δfur strain caused 3-, 43-, and 52-fold decreases in aebC, angM, and fatD transcript levels, respectively (Fig. S6D). Thus, Fur is a repressor of amphi-enterobactin and anguibactin genes, which diminished our ability to observe their regulation by LuxT/SwrZ under iron-replete conditions, particularly for angM and fatD. We note that fur expression is not controlled by LuxT or SwrZ (Fig. S6E). Thus, Fur and LuxT/SwrZ regulate siderophore genes independently. The siderophore experiments in the next section were conducted following the growth of V. harveyi under iron-limited conditions.

LuxT activates V. harveyi siderophore production by SwrZ-dependent and SwrZ-independent mechanisms.

To validate the roles of and relationship between LuxT and SwrZ in the regulation of siderophore production, we quantified siderophore produced by the LCD-locked luxO D61E, luxO D61E ΔluxT, luxO D61E ΔswrZ, and luxO D61E ΔswrZ ΔluxT mutant strains using the CAS assay. Consistent with the qRT-PCR results (Fig. 5D), the luxO D61E ΔswrZ ΔluxT mutant produced siderophore levels intermediate between those of the luxO D61E ΔluxT strain and the luxO D61E ΔswrZ strain (Fig. 5E). These data support a model in which LuxT activates siderophore production by two mechanisms: one dependent on SwrZ and one independent of SwrZ. We performed complementation analyses in a V. harveyi ΔluxR ΔluxT ΔswrZ strain as a final test of our model. Even in the absence of swrZ, the overexpression of luxT caused increased expression of aebC, angM, and fatD, supporting the SwrZ-independent role for LuxT (Fig. 5F). Following the reasoning described above for T3SS genes, repression due to swrZ overexpression could be observed only in the luxT+ background (Fig. 5F and G). Our model for the regulation of siderophore production in V. harveyi is depicted in Fig. 5H: Fur represses siderophore genes under iron-replete conditions, and LuxR represses siderophore genes at HCD. LuxT activates siderophore genes by both a SwrZ-dependent and a SwrZ-independent mechanism.

QS and LuxT/SwrZ regulate the production of the aerolysin toxin.

The final V. harveyi LuxT-regulated behavior that we investigated was the production of the pore-forming toxin aerolysin. Aerolysin-like proteins, originally discovered in the bacterium Aeromonas hydrophila, are a family of secreted proteins that form transmembrane β-barrel pores in eukaryotic target cells, causing cell death (40–42). Aerolysins are virulence factors in Aeromonas spp. and Vibrio splendidus (43, 44).

We previously reported that LuxT is an activator of VIBHAR_RS11620 encoding an aerolysin toxin. For clarity, we now call VIBHAR_RS11620 aerB. We determined that LuxT activates aerB expression by two mechanisms. First, LuxT activates aerB transcription by a Qrr1-independent mechanism. Second, LuxT indirectly activates aerB translation by repressing the expression of qrr1 encoding a posttranscriptional repressor of aerB (13, 18). V. harveyi aerolysin production can be measured by monitoring the hemolysis of blood cells in liquid or on blood agar plates. Activation by LuxT is required for V. harveyi hemolytic activity in both assays. Observable repression by Qrr1 occurs only in the plate assay (18).

In agreement with our above-described findings, RNA-Seq revealed LuxT to be an activator of the aerB aerolysin toxin gene. Three additional genes upstream of aerB were also activated (Fig. 6A and Data Set S1): VIBHAR_RS11600, which we call aerR, encodes a transcriptional regulator; VIBHAR_RS11605, which we do not name, encodes a protein of unknown function; and VIBHAR_RS11610, which we call aerA, encodes an additional aerolysin family toxin. The protein sequences of AerA and AerB are similar (50% pairwise identity, 66% pairwise similarity), indicating a possible gene duplication event. The VIBHAR_RS11615 gene that resides between aerA and aerB was not identified as being regulated by LuxT (Fig. 6A).

FIG 6

QS, LuxT, and SwrZ regulate V. harveyi aerolysin production. (A) Schematic of aerolysin gene organization in V. harveyi. All genes labeled in blue were identified by RNA-Seq as members of the LuxT regulon. The numbers shown within the genes designate the fold change differences in transcript levels between WT and ΔluxT V. harveyi, as measured by RNA-Seq. The expression of the underlined genes was measured in panels C, E, and F. (B) Hemolytic activity in the indicated V. harveyi cell-free culture fluids as judged by lysis of defibrinated sheep’s blood cells. Culture fluids were collected after 24 h of growth in AB medium. (C) qRT-PCR measurements of transcript levels of the indicated genes in WT (black), ΔluxT (blue), ΔswrZ (red), and ΔswrZ ΔluxT (purple) V. harveyi strains. RNA was isolated from strains grown in LM medium to an OD600 of 0.1. (D) Same as for panel B, for the indicated strains. (E) qRT-PCR measurements of transcript levels of the indicated genes in ΔluxR ΔswrZ ΔluxT V. harveyi harboring the indicated plasmids. pControl (black) denotes the empty parent vector, and pluxT (blue) and pswrZ (red) denote vectors harboring IPTG-inducible luxT and swrZ, respectively. RNA was isolated from strains grown for 16 h in AB medium supplemented with 0.5 mM IPTG. (F) Same as for panel E, for ΔluxR ΔswrZ V. harveyi harboring the indicated plasmids. (G) Model for LuxT/SwrZ regulation of aerolysin production. In panels B to F, error bars represent standard deviations of the means from 3 biological replicates. In panels B to E, different letters indicate significant differences between strains (P < 0.05 by two-way ANOVA followed by Tukey’s multiple-comparison test). In panel F, unpaired two-tailed t tests with Welch’s correction were performed comparing the indicated pControl and pswrZ conditions. ns, not significant (P ≥ 0.05); *, P < 0.05; **, P < 0.01.

As described above, we first confirmed LuxT regulation of aerR, aerA, and aerB and determined if Qrr1 is required. Deletion of luxT in WT V. harveyi resulted in ∼5-, 28-, and 15-fold-lower expression levels of aerR, aerA, and aerB, respectively (Fig. S4E). In a strain lacking qrr1, deletion of luxT caused similar reductions in expression (Fig. S4F). These data show that Qrr1 is not required for LuxT activation of aerolysin genes. We know from our previous work that Qrr1 is a posttranscriptional regulator of aerB. Regulation by Qrr1 can be detected only when measuring aerB translation, not transcription (13, 18). For the remainder of the present work, we focus on the transcriptional, Qrr1-independent mechanism by which LuxT activates aerolysin production.

QS regulation of aerolysin production is more complicated than we anticipated; nonetheless, we lay out what we know here. In the liquid hemolysis assay, culture fluids from the LCD-locked luxO D61E strain possessed ∼17-fold more hemolysis activity than fluids from HCD WT V. harveyi (Fig. 6B). Fluids from the HCD-locked ΔluxO strain had activity similar to that of the WT strain (Fig. 6B). Thus, aerolysin production is QS controlled and produced at LCD. Deletion of aphA in the luxO D61E LCD-locked strain did not reduce hemolysis (Fig. S7A), and deletion of luxR in the WT did not increase hemolysis at HCD (Fig. S7B). These data were unexpected and show that while aerolysin is QS controlled, neither AphA nor LuxR regulates its production. We confirmed the results using blood agar plates. The results on the plates were identical to those in liquid with the exception that the ΔluxR strain caused less hemolysis on plates than did the WT (Fig. S7C). Given the LCD production pattern for aerolysin, we conclude that aerolysin genes must be activated at LCD by the QS Qrr sRNAs, either directly or indirectly. We note that this finding seemingly contradicts our previous result showing that Qrr sRNAs posttranscriptionally repress aerB (13, 18). We preliminarily predict that the Qrr sRNAs exert both positive and negative effects on aerolysin production. Activation by the Qrr sRNAs is likely indirect through an unknown regulator, while posttranscriptional repression of aerB occurs directly (13). Further investigation to characterize the mechanism of QS control of aerolysin genes is required.

Assessment of the requirements for V. harveyi aerolysin production. (A, B, and D) Hemolytic activity present in the indicated V. harveyi cell-free culture fluids as judged by lysis of defibrinated sheep’s blood cells. Culture fluids were collected after 24 h of growth in AB medium. (C, E, and F) Hemolytic activity produced by the indicated V. harveyi strains as judged by halo formation on tryptic soy agar (TSA) plates containing 5% sheep’s blood. Strains were grown for 72 h at 30°C. In panels A, B, and D, error bars represent standard deviations of the means from 3 biological replicates. In panels A and B, unpaired two-tailed t tests with Welch’s correction were performed comparing two conditions, as indicated. ns, not significant (P ≥ 0.05). In panel D, different letters indicate significant differences between strains (P < 0.05 by two-way ANOVA followed by Tukey’s multiple-comparison test). Download FIG S7, TIF file, 1.4 MB.

While the mechanism underlying QS regulation of aerolysin genes is not fully defined, what is clear is that aerR, aerA, and aerB are all regulated by LuxT (Data Set S1 and Fig. S4E). Thus, we next explored which of the three genes is required for V. harveyi hemolysis activity. We deleted each gene individually in the LCD-locked luxO D61E V. harveyi strain. Deletion of aerR and aerA eliminated hemolysis activity in the liquid and plate assays (Fig. S7D and E, respectively). In contrast, deletion of aerB did not reduce hemolytic activity, so aerB is not required (Fig. S7D and E). We speculate that AerR is an activator of aerA, and AerA is the primary secreted aerolysin toxin. Because aerB is repressed by the Qrr sRNAs, the aerB expression level may remain low at LCD.

To understand the mechanism underlying LuxT activation of aerolysin gene expression, we measured aerR, aerA, and aerB transcript levels in WT, ΔluxT, ΔswrZ, and ΔswrZ ΔluxT V. harveyi strains at LCD. The expression levels of all three genes were high in the WT and ΔswrZ strains, low in the ΔluxT strain, and intermediate in the ΔswrZ ΔluxT strain (Fig. 6C). The corresponding hemolysis patterns in the LCD-locked luxO D61E background exactly mirrored the transcription patterns both in the liquid assay (Fig. 6D) and on the blood agar plates (Fig. S7F). Exactly analogous to what we have explained above, these data indicate that LuxT activates aerolysin production by both a SwrZ-dependent and a SwrZ-independent mechanism. We verified this assumption using complementation analyses (Fig. 6E and F). The results were as expected according to our putative mechanism except that swrZ overexpression did not repress aerB (Fig. 6F). However, the qRT-PCR results in Fig. 6C show that SwrZ is a repressor of aerB. We presume that swrZ complementation did not repress aerB because aerB exhibits quite low basal expression levels. Our model for LuxT/SwrZ regulation of aerolysin production is depicted in Fig. 6G.

luxT is and swrZ is not conserved among Vibrionaceae family members.

As mentioned above, SwrT (LuxT) repression of swrZ was originally discovered in V. parahaemolyticus and shown to be relevant for the regulation of swarming motility (22). Here, we have shown that LuxT also regulates swrZ in V. harveyi, and both LuxT and SwrZ regulate type III secretion, siderophore production, and aerolysin production. LuxT also regulates genes in V. harveyi that SwrZ does not control. For example, LuxT activates luxCDABE, encoding luciferase, and SwrZ plays no role (18, 20) (Fig. S8). We assessed SwrZ involvement in the regulation of 5 additional LuxT-activated and 3 additional LuxT-repressed genes. SwrZ regulated all 5 of the LuxT-activated test genes, whereas SwrZ did not regulate the 3 LuxT-repressed test genes (Fig. S8). Based on the opposing roles of LuxT and SwrZ (Fig. 3B to E), we speculate that SwrZ may function only as a transcriptional repressor, possibly explaining why it is not required to participate in LuxT repression of gene expression. We conclude that within the LuxT regulon, a subset of genes does not employ SwrZ in regulation.

LuxT regulates genes by SwrZ-dependent and SwrZ-independent mechanisms in V. harveyi. Shown are qRT-PCR measurements of transcript levels of the indicated genes in WT (black), ΔluxT (blue), ΔswrZ (red), and ΔswrZ ΔluxT (purple) V. harveyi strains. Strains were grown in LM medium to an OD600 of 0.1. Error bars represent standard deviations of the means from 3 biological replicates. Different letters indicate significant differences between strains (P < 0.05 by two-way ANOVA followed by Tukey’s multiple-comparison test). Download FIG S8, PDF file, 0.03 MB.

In V. harveyi and V. parahaemolyticus, the luxT (swrT) and swrZ genes are carried on the two different chromosomes, so while they coregulate many genes, it is unlikely that they were inherited as a pair. Knowing their interconnected roles in V. harveyi and V. parahaemolyticus, we wondered whether LuxT and SwrZ jointly regulate functions in other Vibrionaceae species. To garner preliminary evidence for or against this possibility, we examined the conservation of both genes within the Vibrionaceae family. Among the 418 sequenced Vibrionaceae species, the luxT gene is present in all but 16 species (96%) (Fig. 7A). None of the 16 species lacking luxT carries a swrZ gene (Fig. 7A). Among the 402 Vibrionaceae species that possess luxT, 227 species (56%) also have swrZ (Fig. 7A); these genera include Aliivibrio, Photobacterium, and Vibrio (Fig. 7A). Thus, in cases in which both luxT and swrZ are present, the possibility of coregulation of target genes exists. In the species that have luxT but lack swrZ, LuxT must regulate genes independently of SwrZ. Together, these observations may explain the evolution of SwrZ-dependent and SwrZ-independent functions for LuxT in V. harveyi.

FIG 7

Cooccurrence of luxT and swrZ genes among Vibrionaceae species. (A) Highest similarity in protein sequences (BLOSUM62) in the indicated genera to V. parahaemolyticus SwrZ. Species are divided into two groups, one possessing luxT (luxT+) and one lacking luxT (luxT−). The dashed line indicates the cutoff used for the similarity score to differentiate swrZ+ from swrZ− strains. Gray indicates groups of species that lack swrZ. The numbers in parentheses indicate the number of species in each group. Boxes show the means ± standard deviations for groups with >5 species. Circles represent individual species in groups with ≤5 members. Purple, green, and orange denote Aliivibrio, Photobacterium, and Vibrio species that possess swrZ, respectively. (B) Phylogenetic tree based on the swrZ promoter regions in Vibrionaceae. Phylogenetic distances were computed based on sequences spanning positions −122 to −8, relative to the swrZ start codons. Colors are the same as those in panel A. (C) qRT-PCR measurements of swrZ transcript levels in WT (black) and ΔluxT (blue) A. fischeri. RNA was isolated from strains grown in LM medium to an OD600 of 0.1. (D) CAS assay quantitation of siderophore levels in cell-free culture fluids prepared from the indicated A. fischeri strains. Strains were grown for 16 h in AB medium. (E) Model for LuxT/SwrZ regulation of A. fischeri aerobactin siderophore production. In panels C and D, error bars represent standard deviations of the means from 3 biological replicates. In panel C, an unpaired two-tailed t test with Welch’s correction was performed comparing the indicated WT and ΔluxT conditions. **, P < 0.01. In panel D, different letters indicate significant differences between strains (P < 0.05 by two-way ANOVA followed by Tukey’s multiple-comparison test).

To probe the possibility that LuxT also represses swrZ in other species, we performed a phylogenetic analysis by comparing the swrZ promoter regions of all luxT+ species. Our prediction was that swrZ promoters would be more similar in species in which LuxT represses swrZ than in species in which LuxT does not regulate swrZ. In support of this idea, the V. harveyi and V. parahaemolyticus swrZ promoter sequences are indeed similar. Thus, when we constructed a phylogenetic tree based on swrZ promoter sequences, V. harveyi and V. parahaemolyticus reside in proximity (Fig. 7B). As an initial test of LuxT regulation of swrZ in a species beyond V. harveyi and V. parahaemolyticus, we analyzed LuxT repression of swrZ in Aliivibrio fischeri, a species that neighbors V. harveyi and V. parahaemolyticus on the swrZ promoter phylogenetic tree (Fig. 7B). Indeed, swrZ transcript levels were ∼10-fold higher in ΔluxT A. fischeri than in the WT, showing that LuxT is a repressor of swrZ in A. fischeri (Fig. 7C). We previously reported that LuxT is an indirect activator of iucABCD, encoding the aerobactin siderophore biosynthesis genes in A. fischeri (21). At that time, we did not know the identity of the regulator connecting LuxT to iucABCD. Our findings here point to SwrZ fulfilling that function. To test this idea, we used the CAS assay to measure siderophore levels in culture fluids from WT, ΔluxT, ΔswrZ, and ΔswrZ ΔluxT A. fischeri strains. Indeed, LuxT activates A. fischeri siderophore production only in the swrZ+ background, as the deletion of luxT in the ΔswrZ strain caused no change in siderophore production (Fig. 7D). Thus, LuxT activates A. fischeri siderophore production via a SwrZ-dependent mechanism. There is no evidence that a second SwrZ-independent mechanism occurs (Fig. 7D). Our model for LuxT and SwrZ regulation of A. fischeri siderophore production is depicted in Fig. 7E.

DISCUSSION

QS controls over 600 genes in V. harveyi (8). Integral to the V. harveyi QS process are five, largely redundant, Qrr sRNAs that posttranscriptionally regulate target gene expression. LuxT is a TetR family transcriptional regulator that was previously identified to repress the expression of V. harveyi qrr1 encoding the Qrr1 sRNA (18). When LuxT acts as a qrr1 repressor, it tunes the expression of select Qrr1-controlled genes that are members of the much larger QS regulon. Here, we use RNA-Seq to discover that beyond controlling qrr1 expression, LuxT is a global regulator in V. harveyi that controls 414 genes. Analogous to what has been reported for V. parahaemolyticus (22), LuxT directly represses swrZ, encoding an additional transcriptional regulator. We show that in V. harveyi, LuxT activates type III secretion, siderophore production, and aerolysin production genes independently of Qrr1 by both SwrZ-dependent and SwrZ-independent mechanisms. In the future, defining the SwrZ regulon will be a crucial step in gaining a comprehensive understanding of the individual and combined roles of LuxT and SwrZ in regulating V. harveyi gene expression.

Two feedback loops exist in the V. harveyi LuxT/SwrZ regulatory pathway: LuxT and SwrZ activate and repress, respectively, their own transcription. Generally, positive feedback is thought to amplify responses to a stimulus, whereas negative feedback buffers pathway output against fluctuations in stimuli (45, 46). Thus, the inclusion of both positive and negative feedback in a single system can alter both input sensitivity and output dynamics (46, 47). We speculate that the LuxT positive feedback loop enables V. harveyi to rapidly commit to the LuxT “on” state following the initial activation of luxT expression. In contrast, SwrZ-directed negative feedback on swrZ, which occurs when LuxT is in the “off” state, should prevent runaway swrZ expression. Moreover, LuxT and SwrZ have opposing regulatory functions, so genes activated by LuxT are repressed by SwrZ. Negative SwrZ autoregulation could, by tamping down SwrZ production, accelerate the expression of LuxT-activated genes when V. harveyi transitions to the LuxT “on” state.

Our analyses revealed that LuxT activates T3SS, siderophore, and aerolysin genes by two mechanisms, one SwrZ dependent and one SwrZ independent. We propose two advantages to this regulatory arrangement. First, this dual mechanism may promote a “switch-like” response. Because SwrZ represses genes that are activated by LuxT, when LuxT is in the “on” state, stronger target gene activation will occur if swrZ is repressed by LuxT than if not. Second, if the expression and/or activity of either or both LuxT and SwrZ responds to additional regulatory inputs, linking the two nodes could promote more nuanced control of gene expression than if LuxT and SwrZ were not connected in the circuit.

For V. harveyi to transition between LuxT “on” and LuxT “off” states, a cue(s) must activate luxT expression or increase LuxT activity. We do not know the identity of this putative cue. We do know that LuxT is not regulated by QS, and consistent with this finding, luxT expression remains steady over the V. harveyi growth curve (see Fig. S3A and B in the supplemental material). Curiously, however, genes in the LuxT-controlled regulon overlap those in the QS-controlled regulon. In particular, LuxT activates the expression of type III secretion, siderophore production, and aerolysin production genes, all of which are also QS controlled and expressed at LCD, which is the typical pattern for QS-controlled virulence genes in vibrios (48). The above-mentioned QS- and LuxT-regulated traits are all virulence factors that contribute to the success of vibrios as pathogens in host organisms (43, 49, 50). For example, in shrimp larvae, V. harveyi T3SS gene expression is 1,000-fold higher than that in laboratory culture (49). Possibly, luxT expression or LuxT activity is modulated by a host factor. Alternatively, or in addition, TetR family members frequently bind small-molecule ligands (51). It could be that a V. harveyi-, host-, or environment-derived ligand promotes heightened LuxT activity. Future studies investigating the regulation of luxT expression or LuxT activity should define the connection between the LCD state and LuxT function.

Our phylogenetic analyses revealed that luxT, but not swrZ, is conserved among Vibrionaceae, and LuxT repression of swrZ is conserved at least in V. parahaemolyticus, V. harveyi, and our A. fischeri test case (22). Going forward, it will be interesting to learn how LuxT function has diverged in species that harbor luxT but lack swrZ, such as in Vibrio cholerae C6706. RNA-Seq experiments similar to those described here could be used to begin to parse the roles that LuxT plays and, based on the findings, paint a coherent evolutionary picture of LuxT control of Vibrionaceae biology.

MATERIALS AND METHODS

Bacterial strains, culture conditions, and standard methods.

All strains are listed in Table S1A in the supplemental material. V. harveyi strains were derived from V. harveyi BB120 (ATCC BAA-1116) (52). Previously, V. harveyi BB120 was reclassified as Vibrio campbellii BB120 (53). For literary consistency, we refer to this strain as V. harveyi. A. fischeri strains were derived from A. fischeri ES114 (54). E. coli S17-1 λpir was used for cloning. E. coli MG1655 was used for heterologous gene expression. V. harveyi and A. fischeri strains were grown in either Luria marine (LM) medium or minimal autoinducer bioassay (AB) medium at 30°C with shaking (55, 56). AB medium contained 0.4% vitamin-free Casamino Acids (Difco). E. coli strains were grown in LB medium at 37°C for cloning or at 30°C for heterologous gene expression. When necessary, kanamycin, chloramphenicol, ampicillin, and polymyxin B were added at 100 μg mL−1, 10 μg mL−1, 100 μg mL−1, and 50 μg mL−1, respectively. Gene expression from the P and P promoters was induced following the addition of 0.01% arabinose and 0.5 mM isopropyl β-d-1-thiogalactopyranoside (IPTG), respectively, at the time of inoculation. Procedures for LuxT-6×His purification, EMSAs, qRT-PCR measurements, and hemolysis assays were described previously (18), as were CAS siderophore detection assays (21). Primers used to amplify DNA probes for EMSAs and for qRT-PCR measurements are listed in Table S1B. Growth conditions for qRT-PCR experiments are provided in the figure legends.

Strains, primers, and plasmids used in this study. (A) Strain list; (B) primer list; (C) plasmid list. Download Table S1, DOCX file, 0.05 MB.

DNA manipulation and strain construction.

Oligonucleotide primers were purchased from Integrated DNA Technologies (IDT) and are listed in Table S1B. PCR reactions contained either KOD Hot Start DNA polymerase (Sigma) or iProof DNA polymerase (Bio-Rad). All cloning was completed using Gibson assembly master mix (New England BioLabs) for isothermal DNA assembly (57). Plasmids were verified by sequencing (Genewiz) and are listed in Table S1C. Regarding the nomenclature of our constructs, a capital P designates the promoter driving transcription (e.g., P-lux). Plasmids that promote the overexpression of genes are designated with a lowercase p (e.g., pswrZ). The P-lux and P-lux transcriptional reporters included 114- and 531-bp promoter regions, respectively, to drive the transcription of luxCDABEG. A consensus ribosome-binding site was included in both reporters to drive translation. Plasmids were introduced into E. coli by electroporation using a Bio-Rad MicroPulser. Plasmids were conjugated into V. harveyi and A. fischeri from E. coli S17-1 λpir. V. harveyi exconjugants were selected on agar plates containing polymyxin B, and A. fischeri exconjugants were selected on agar plates containing ampicillin. Chromosomal alterations in V. harveyi and A. fischeri were introduced using the pRE112 vector harboring the sacB counterselectable marker as previously described (18, 58, 59). Mutations were verified by PCR and/or sequencing.

RNA sequencing.

Cells from cultures of WT and ΔluxT V. harveyi strains grown overnight in LM medium were pelleted by centrifugation at 21,100 × g (Eppendorf 5424 centrifuge) and resuspended in fresh LM medium. Fresh cultures containing 25 mL of LM medium were inoculated with the washed cells at an OD600 of 0.005. The cultures were incubated at 30°C with shaking, and RNA was isolated from 3 biological replicates of each strain when the cultures had reached an OD600 of 0.1 using the RNeasy minikit (catalog number 74106; Qiagen). RNA-Seq was performed at the Genomics Core Facility at Princeton University, as previously described (60). Reads were mapped to the V. harveyi BB120 (ATCC BAA-1116) genome using TopHat (61). Genes with differential expression were identified using DESeq2 (62), and those exhibiting a log2 fold change in expression >1, along with a P value <0.01, in the ΔluxT strain compared to WT V. harveyi were designated members of the LuxT regulon.

Bioluminescence assays.

E. coli harboring the P-lux reporter or V. harveyi harboring the P-lux reporter were grown overnight, pelleted by centrifugation at 21,100 × g (Eppendorf 5424 centrifuge), and resuspended in LB or LM medium, respectively. The E. coli and V. harveyi cells were inoculated into fresh LB and LM medium, respectively, with normalization to a starting OD600 of 0.005. A total of 150 μL of each culture was transferred to wells of a 96-well plate (Corning) in quadruplicate technical replicates and overlaid with 50 μL of mineral oil (Sigma). The plates were incubated with shaking at 30°C, and the bioluminescence and OD600 were measured every 15 min for 24 h using a BioTek Synergy Neo2 multimode reader (BioTek, Winooski, VT, USA). Relative light units (RLU) (bioluminescence/OD600) represent the values when each sample was at an OD600 of 1.

VopD Western blot analyses.

Cytoplasmic VopD levels were measured in V. harveyi that had been grown in AB medium supplemented with 5 mM EGTA as previously described (5). Cells equivalent to 1 OD600 unit were pelleted by centrifugation at 21,100 × g (Eppendorf 5424 centrifuge), resuspended in 100 μL of SDS-PAGE sample buffer (Bio-Rad), and boiled for 10 min. Samples were loaded onto 4–20% Mini-Protean TGX gels (Bio-Rad) and subjected to electrophoresis for 30 min at 50 mA. Following Western transfer, nitrocellulose membranes were cut in half. One portion was probed with an antibody against VopD, and the other portion was probed with an antibody against the LuxS control, as previously described (5). An anti-rabbit IgG(H+L) horseradish peroxidase (HRP) conjugate (Promega) was used as the secondary antibody. Proteins were visualized using SuperSignal West Femto maximum-sensitivity substrate (Thermo Fisher) and an ImageQuant LAS 4000 imager.

Phylogenetic analyses.

Genomic DNA sequences of 418 Vibrionaceae species were downloaded from the GenBank database (63). Genes encoding luxT among Vibrionaceae species were identified previously (18). A custom MATLAB (2020; MathWorks) search algorithm based on protein sequence similarity was used to identify genes encoding SwrZ. Briefly, protein sequences of SwrZ from V. harveyi ATCC BAA-1116 and V. parahaemolyticus RIMD 2210633 were used as queries. Both yielded similar search results. The chromosomes or contigs of species under consideration were first converted to amino acid sequences and subsequently scanned for regions similar to the query sequences. To identify regions of highest similarity between protein sequences, local sequence alignments were performed using the Smith-Waterman (SW) algorithm (64). The standard substitution matrix BLOSUM62 (https://ftp.ncbi.nih.gov/blast/matrices/) was used to compute similarity score, S, which considers both the length and sequence similarity of the alignment. A cutoff S value of >150 was used to identify putative genes encoding SwrZ. Genes identified as swrZ homologs were verified to encode GntR family transcriptional regulators. Species lacking swrZ were excluded from further phylogenetic analyses. The protocols used to perform phylogenetic analyses and tree building were described previously (18).

Data availability.

All relevant data are within the manuscript and the supplemental material. RNA-Seq data and files containing the numerical data for the main text and supplemental figures are provided at Zenodo (https://doi.org/10.5281/zenodo.5719716).