Control of the Type 3 Secretion System in Vibrio harveyi by Quorum Sensing through Repression of ExsA

ABSTRACT The type 3 secretion system (T3SS) genes of Vibrio harveyi are activated at low cell density and repressed at high cell density by quorum sensing (QS). Repression requires LuxR, the master transcriptional regulator of QS-controlled genes. Here, we determine the mechanism underlying the LuxR repression of the T3SS system. Using a fluorescence-based cell sorting approach, we isolated V. harveyi mutants that are unable to express T3SS genes at low cell density and identified two mutations in the V. harveyi exsBA operon. While LuxR directly represses the expression of exsBA, complementation and epistasis analyses reveal that it is the repression of exsA expression, but not exsB expression, that is responsible for the QS-mediated repression of T3SS genes at high cell density. The present work further defines the genes in the V. harveyi QS regulon and elucidates a mechanism demonstrating how multiple regulators can be linked in series to direct the expression of QS target genes specifically at low or high cell density.

Bacterial pathogens must properly regulate the expression of virulence traits to successfully initiate and maintain disease. A type 3 secretion system (T3SS) is one such tightly controlled trait. T3SSs form needle-like pores through the inner and outer membranes of bacteria that deliver effector proteins from the cytoplasm of bacteria directly into their target eukaryotic host cells. Effector proteins have a variety of activities, including modifying host signaling and immune responses that result in disease progression (for a review, see reference 41).
Vibrio harveyi, a bioluminescent marine bacterium that infects marine animals, possesses a T3SS. In V. harveyi, the T3SS is regulated in a cell density-dependent manner by the process known as quorum sensing (QS) (24,31,43). QS is a mechanism of chemical communication that involves the production, secretion, and detection of signal molecules called autoinducers (AIs) (42). When bacteria are at low cell density, the concentrations of AIs are low, leading to the expression of behaviors that are important for survival as individuals. As a population of QS bacteria grows, the concentration of AIs increases proportionally. At a critical AI concentration, which corresponds to a particular population density, the population switches behaviors to those that are beneficial for survival as a group. QS regulates behaviors such as conjugation, bioluminescence, biofilm formation, and virulence factor production in many bacterial species (42).
V. harveyi produces and detects three AIs to monitor cell density (3,10,11,25). Each AI is recognized by a cognate membrane-associated receptor (2,3,20,25). At low cell den-sity (i.e., in the absence of AIs), the receptors function as kinases and funnel phosphate into a phosphorelay pathway that ultimately leads to the phosphorylation of the response regulator LuxO (4,19,20). Phosphorylated LuxO (LuxOϳP) activates the expression of genes encoding five regulatory small RNAs (sRNAs), termed the Qrr sRNAs, that repress the translation of luxR mRNA (27,28,39). LuxR is the master transcriptional regulator of QS-dependent genes (22,24,31,38,43). At high cell density, accumulated AIs inhibit the kinase activities of the receptors, switching their net activity to that of phosphatase. This event reverses the phosphate flow through the pathway, leading to the dephosphorylation of LuxO, the termination of the expression of the qrr genes, and the translation of the luxR mRNA. LuxR functions as both an activator and a repressor and is responsible for switching the gene expression patterns from those that underpin individual behavior to those appropriate for group behavior. In V. harveyi, more than 60 genes have been identified to be regulated by QS, but only 15 of them are directly controlled by LuxR (24,31,33,43). This finding suggests that a hierarchy of gene regulation exists downstream of LuxR to control QS-dependent genes.
The T3SS of V. harveyi consists of four major structural operons and associated regulatory genes (24,43). Three of the operons, denoted T3SS.1, T3SS.2, and T3SS.3, lie adjacent to one another on chromosome 1 of V. harveyi BB120 (ATTC BAA-1116; VIBHAR_01723 to VIBHAR_01742). The fourth structural operon, T3SS.4, is located approximately 15 kb away (VIBHAR_01692_VIBHAR_01712). This arrangement parallels that of Vibrio parahaemolyticus T3SS1, a close relative of V. harveyi (30,47). V. parahaemolyticus also encodes a second T3SS (T3SS2) on chromosome 2 (30). Park et al. concluded that T3SS2 is not present in other Vibrio species by screening different species for homologous T3SS2 genes (32). In support of this conclusion, the recently completed V. harveyi BB120 genome reveals that none of the T3SS2 genes are present.
Multiple studies have shown that the V. harveyi promoters driving the expression of the T3SS.I-T3SS.IV operons are expressed at low cell density and repressed by QS at high cell density (24,31,43). Furthermore, LuxR, the master transcriptional regulator of QS, is required for QS-mediated repression, because T3SS genes were not repressed at high cell density in a ⌬luxR strain (24). Biochemical analyses to measure the secretion of the canonical effector protein VopD demonstrated that secretion occurs only at low cell density (24). In addition, the expression of T3SS.1, T3SS.2, and T3SS.3 are all required for the secretion of VopD, and T3SS.IV, while not yet tested, is presumed to be necessary as well. Finally, VopD secretion in V. parahaemolyticus is repressed by QS through the LuxR homolog, OpaR (24). However, the regulatory wiring linking LuxR and OpaR to the expression of the T3SS is not known.
In the present work, we characterize the mechanism by which QS, via LuxR, represses the expression of T3SS genes. Specifically we find that, at high cell density, LuxR functions indirectly to control T3SS gene expression by binding to a promoter upstream of the exsBA operon, repressing the expression of both exsB and exsA. However, it is the repression of exsA (encoding a transcriptional activator) and not exsB (encoding a putative pilot protein) that is critical for the QSmediated control of T3SS genes.
(This work was presented in part as a poster [H-120] at the 107th General Meeting of the American Society for Microbiology, Toronto, Canada, 21 to 25 May 2007.)

MATERIALS AND METHODS
Bacterial strains and culture conditions. V. harveyi strains were cultured in Luria-murine (LM) medium with shaking at 30°C. Escherichia coli S17-1pir (15) was used for cloning procedures and was grown in Luria-Bertani medium at 37°C with shaking. Antibiotics were obtained from Sigma and used at the following concentrations (in g/ml): ampicillin, 100; kanamycin, 100; and tetracycline, 10. V. harveyi exsB and exsA mutants were constructed as previously described (39,43). V. harveyi genomic DNA segments cloned into the cosmid pLAFR2 were mutated in E. coli using -red-mediated recombination (14). This process results in the complete deletion of the target gene from the start to stop codon and the insertion of the cat cassette from the plasmid pKD3. To generate unmarked deletions, FLP-mediated recombination was used, resulting in an 85-nucleotide (nt) scar (14). Mutations were transferred onto the V. harveyi chromosome using homologous recombination. Strains used in this study are listed in Table 1.
DNA manipulations. DNA manipulations were performed using standard procedures (34). T4 DNA ligase and restriction enzymes were purchased from New England Biolabs (NEB). PCRs were carried out with iProof DNA polymerase or Phire DNA polymerase (NEB). For cloning purposes, exsB and exsA were amplified by PCR from BB120 genomic DNA. The DNA for the complementation of the exsB mutant was amplified with primers containing BamHI recognition sites and subsequently cloned into the BamHI restriction site of pLAFR2. The exsA gene was amplified to contain a 5Ј EcoRI site and 3Ј BamHI site. The 3Ј primer was designed to incorporate the FLAG sequence DYKD-DDDK-stop at the terminus of the exsA open reading frame (ORF) to confirm the expression of ExsA upon the addition of isopropyl-␤-D-thiogalactopyranoside (IPTG). The ExsA-FLAG-encoding fragment was cloned into the EcoRI and BamHI sites of the overexpression vector pEVS143 (17). gfp transcriptional fusions were constructed in pCMW1 (43) by cloning into the SphI and SalI restriction sites. The exsB-lux fusion was cloned into the SpeI and BamHI sites of pBBRlux (23), while the exsA-lux fusion was cloned into the BamHI site of pBBRlux. All plasmid constructs were confirmed by sequencing. Primer sequences are available upon request.
Genetic screen to identify QS-controlled regulators of the T3SS. Tn5 mutagenesis of the V. harveyi luxR mutant JW507 was performed using the suicide vector pRL27c (26). Ten thousand kanamycin-resistant colonies were pooled. Plasmid pCMW209-cat contains a 319-bp V. harveyi genomic fragment containing the promoter region of the T3SS.IV operon in a transcriptional fusion to gfp. This fragment was isolated in a previous genetic screen as a QS-repressed promoter. In addition, the kanamycin resistance gene of the original isolate, pCMW209 (43), was replaced with a chloramphenicol acetyltransferase (cat) gene to generate pCMW209-cat. The expression of the fusion is repressed 50-fold by LuxR (43). pCMW209-cat was introduced into the above-described Tn5 mutant pools by conjugation. Exconjugant colonies were pooled and grown for 20 to 24 h at 30°C in LM medium containing chloramphenicol. This time point was selected because it provides a homogenous population of cells expressing high levels of gfp. Cells from these cultures were diluted 1:50 in LB and sorted on a FACSAria flow cytometer (Becton Dickinson), and the lowest 1% of gfp-expressing transformants was isolated. The sorted cells were plated, and 4,000 colonies were individually transferred to 96-well plates (Costar 3595; Corning) containing 200 l LM medium with chloramphenicol. The plates were agitated for 20 to 24 h at 30°C, and green fluorescent protein (GFP) production from each well was measured on a 1420 Victor2 multilabel counter (Wallac). Mutants exhibiting low gfp expression were collected and grown in fresh medium, and gfp expression was verified by flow cytometry. This strategy identified 21 candidates expressing low gfp. The pCMW209-cat plasmid was isolated from each candidate mutant and reintroduced into strain JW507 to confirm that no spontaneous mutations had occurred on the reporter plasmid to cause reductions in gfp expression.
Genomic DNA was isolated from confirmed candidate mutants using a DNeasy blood and tissue kit (Qiagen), and the DNA was digested separately with BamHI, MluI, or EcoRV. Resulting fragments were circularized by ligation and introduced into E. coli S17-1pir by electroporation. Plasmids containing the transposon and flanking V. harveyi DNA were selected on LB medium with kanamycin and sequenced. Sequence analysis was performed using Vector NTI suite 10.1.1 (InforMax Invitrogen Software Package), The Institute for Genomic Research (TIGR) V. parahaemolyticus database (http://www.tigr.org), and NCBI BLAST-X (http://www.ncbi.nlm.nih.gov/BLAST/). Fluorescence, bioluminescence, and qPCR measurements. All gene expression measurements were performed between 20 and 24 h, as this was the time point when the original screen was performed. gfp expression was measured using a BD FACSAria (Becton Dickinson) flow cytometer in overnight cultures grown for 20 to 24 h in triplicate (settings: forward scatter [log] voltage, 400; side scatter [log] voltage, 505; and fluorescein isothiocyanate [log] voltage, 626) or, as mentioned, using a 1420 Victor2 multilabel counter (Wallac). For the analysis of bioluminescence expression, strains were grown with shaking in 150 l from a 1/150 dilution of an overnight culture in Costar 3903 microtiter plates. Bioluminescence was determined on a Molecular Devices SpectraMax M5 microplate spectrophotometer system. For quantitative real-time PCR (qPCR), RNA was isolated from V. harveyi strains grown to an optical density at 600 nm (OD 600 ) of 1.0 using the RNeasy mini kit (Qiagen). RNA was quantified on a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies). Samples were treated with DNase I (Roche) in reaction mixtures containing 10ϫ PCR buffer (ABI) and 25 M MgCl 2 (ABI). cDNA was generated in 100-l reaction mixtures containing 1 g of RNA, 5ϫ first-strand buffer (Invitrogen), 100 M dithiothreitol (DTT) (Invitrogen), 10 M deoxynucleoside triphosphates (dNTPs) (ABI), random hexamers (Roche), and SuperScript II reverse transcriptase (Invitrogen). Reactions were completed in a thermocycler for 10 min at 25°C, for 50 min at 42°C, and lastly for 10 min at 72°C. To control for genomic DNA contamination, identical reactions were performed in the absence of reverse transcription enzyme. Real-time PCRs were performed on an ABI Prism 7900HT sequence detection system. Reactions were carried out in 384-well optical reaction plates (ABI) in quadruplicate with 2ϫ Sybr green mix (ABI). Ten microliters was loaded into each well using a Beckman Coulter Biomek FX machine. Real-time PCR primers were designed using Primer Express 2.0 (ABI Software). hfq mRNA was used as an endogenous loading control for the reactions. Relative mRNA levels using ⌬⌬CT relative quantification were determined by the RQ Manager, version 1.2, software program. Electrophoretic mobility shift assays (EMSAs). LuxR was purified with the IMPACT protein purification system (NEB) as previously described (43). DNA probes for gel mobility shift analyses were generated using 5-/6-carboxyfluorescein [5-(and-6)-FAM; mixed isomer] 5Ј-tagged fluorescent primers in a standard PCR and purified using the Zymoclean gel DNA recovery kit (Zymo Research). Each probe (10 nM, 5 l total volume) was combined with 1 l of 1-g/ml poly(dI-dC) and the indicated amount of LuxR (0 to 1,000 nM) in a final volume of 20 l at 30°C for 15 min. Thus, every reaction mixture contained a total of 14 l of the buffer that was used to purify LuxR (20 mM Tris [pH 7.5], 1 mM EDTA, 10 mM NaCl, and 0.1 mM DTT with 20% glycerol [43]). Mobility shifts were performed on 5% Tris-acetate-EDTA (TAE)-polyacrylamide gels and visualized using a Storm 860 imaging system (Molecular Dynamics).
Examination of operon structure of exsBA and 5-rapid amplification of cDNA ends (RACE) analysis. To determine if exsB and exsA are cotranscribed, RNA was prepared from a ⌬luxR V. harveyi mutant grown to an OD 600 of 1.0 using the RNeasy protocol (Qiagen) and treated with Turbo DNase (Invitrogen). One g of RNA was converted to cDNA using iScript (Bio-Rad). A control reaction that did not contain reverse transcriptase but was otherwise identical also was performed. The primer set exsAB-op-1A (5Ј-GGCAGAGTGCATAAGGATTT-3Ј), located on the sense strand in the 3Ј end of the exsB genes, and exsAB-op-3B (5Ј-TAGTTGGCCTAACACATCCA-3Ј), located on the antisense strand in the 5Ј end of the exsA gene, was used to amplify BB120 genomic DNA, the reverse transcription-negative control, and the cDNA samples using a standard PCR with Phire DNA polymerase (NEB) for 35 cycles. The reaction product was visualized by agarose gel electrophoresis.

RESULTS
Identifying the regulatory pathway connecting QS to expression of T3SS genes. To determine the mechanism by which QS controls T3SS gene expression in V. harveyi, we first examined whether LuxR is a direct regulator of the T3SS promoters. To do this, we performed EMSAs with purified LuxR and fluorescently labeled probes encoding the upstream regions of each of the four T3SS operons. LuxR showed no binding to these promoter regions, suggesting that LuxR controls T3SS genes by an indirect mechanism (data not shown). This result is consistent with our previous finding that LuxR does not directly bind to the promoter region upstream of the T3SS.IV operon encoded in the plasmid pCMW209 (43).
To identify regulatory elements linking LuxR to the repression of T3SS gene expression, we developed a fluorescenceactivated cell sorting (FACS)-based screen. The plasmid pCMW209-cat, encoding a gfp transcriptional fusion to the exsD promoter (this region is shown in Fig. 1A), was used as the readout for T3SS expression. pCMW209-cat is highly expressed in a ⌬luxR mutant, and gfp expression is repressed 50-fold in strains that are WT for luxR (43). We hypothesized that LuxR regulates T3SS in V. harveyi through the repression of an activator protein or the induction of a repressor protein.
To examine the first possibility, we performed mutagenesis of a ⌬luxR mutant and screened for mutants that showed the loss of the expression of pCMW209-cat.
Mutants containing random Tn5 insertions generated in a ⌬luxR V. harveyi strain were pooled, pCMW209-cat was intro-duced, and mutants exhibiting reduced gfp expression relative to that of pCMW209-cat in the parent ⌬luxR strain were isolated using FACS. The locations of the transposon insertions in 21 such mutants were identified. Fifteen of these insertion mutants were siblings, so five unique transposon mutants were identified. A single unique insertion was located in the following genes: VIBHAR_00301, a putative hydrolase; VIB-HAR_06142, ribokinase; and VIBHAR_05400 and VIB-HAR_05860, hypothetical genes. Two independent insertions, C6 and E10, were located in the gene VIBHAR_01694 ( Fig.  1A and B). VIBHAR_01694 encodes a homolog of the Pseudomonas aeruginosa exsB gene (21) (also named yscW or virG in Yersinia enterocolitica). In Y. enterocolitica, virG encodes a pilot lipoprotein required for the efficient assembly of the T3SS ring complex (9). VirG partners with YscC, a secretin, that oligomerizes to form the upper basal body of the Y. enterocolitica T3SS needle complex (9). VIBHAR_01694 (here referred to as ExsB) and VirG from Y. enterocolitica share only 16.9% amino acid identity, making it unclear if they are functionally similar.
Mutation of exsB alters T3SS gene expression. exsB is located approximately 2 kb upstream of the T3SS.IV operon of V. harveyi (Fig. 1A). Importantly, in the exsB::Tn5 mutants C6 and E10, exsD-gfp transcription was assessed in trans from the plasmid pCMW209-cat. Therefore, the reduction in gfp expression resulting from Tn5 insertions in exsB cannot be due to polar effects of the transposon on the plasmid-encoded T3SS.IV promoter. To further examine if exsB is part of the regulatory pathway connecting QS to the expression of T3SS genes, the chromosomal exsB gene in the V. harveyi ⌬luxR mutant was replaced with the cat gene encoding resistance to chloramphenicol. The expression of exsD-gfp from pCMW209 in the ⌬luxR, exsB::cat double mutant was reduced compared to that of the ⌬luxR parent, confirming the phenotypes of the transposon mutants isolated in the genetic screen (Fig. 1B). We wondered if the exsB::cat mutation specifically reduced expression from the T3SS.IV operon, or if this mutation reduced the expression of all of the T3SS operons in V. harveyi. To determine this, each T3SS operon promoter was cloned as a gfp transcriptional fusion into plasmid pCMW1. The expression of each of the fusions was measured in the ⌬luxR and ⌬luxR, exsB::cat mutants (Fig. 2). No gfp expression could be detected for the T3SS.I promoter fusion in either mutant strain, so we conclude that the expression level of the T3SS.1 promoter is below the limit of detection for gfp in our system under all conditions. However, gfp expression from the T3SS.II and T3SS.III fusions was reduced 6-and 2-fold, respectively, in the ⌬luxR, exsB::cat double mutant compared to that of the ⌬luxR parent ( Fig. 2A). This result is similar to what we observe for the effect of exsB::cat on the expression of the T3SS.IV promoter (Fig. 1B, 2A). To determine if the promoter of the T3SS.I operon also was repressed in the ⌬luxR, exsB::cat mutant, we constructed a transcriptional fusion of this promoter to the lux operon in the pBBRlux plasmid (23). Luciferase provides a much more sensitive reporter for gene expression, and indeed the transcription of the T3SS.I operon was detectable from this reporter. Similarly to the results observed for the promoters of the T3SS.II-T3SS.IV operons, the expression of the T3SS.I promoter was reduced 3-fold in an exsB::cat mutant (Fig. 2B). In each case, the fold reduction in the expression of the T3SS.I-III promoters in the exsB::cat mutant was less than what we observed for the T3SS.IV promoter. The T3SS.IV promoter has the highest levels of the expression of the four T3SS operons in V. harveyi, and thus it apparently provides a larger dynamic range of regulation. The importance of these differences in expression levels is not known. In summary, the mutation of exsB leads to an overall loss of T3SS gene expression from all four T3SS operons in V. harveyi.
QS regulates exsB expression. If exsB is part of the regulatory pathway linking QS to T3SS gene expression, we predict that exsB also is regulated by QS. To determine if this is the case, quantitative PCR was used to measure the expression of exsB in four V. harveyi strains: the wild type grown to high cell density, the ⌬luxO mutant that is locked into the high-celldensity state, and the luxO(D47E) and ⌬luxR mutants that are locked into the low-cell-density state. The ⌬luxO mutant is locked at high cell density because it does not express qrr genes, so LuxR is constitutively produced. In contrast, the luxO(D47E) allele locks the cells into the low-cell-density state because LuxOD47E mimics phospho-LuxO, leading to the constitutive activation of the qrr genes and the repression of LuxR independent of cell density. Finally, in the ⌬luxR strain, the cells are locked at low cell density because there is no LuxR to promote the high-cell-density gene expression pattern (reviewed in reference 42). Our results are shown in Fig. 3 and indeed are analogous to what we observe for the T3SS genes; the expression of exsB is 50-fold higher in the two low-celldensity-locked strains than in the wild-type and ⌬luxO highcell-density strains.
Secretion is not required for T3SS gene expression. In T3SSs of many bacteria, functional secretion is intimately tied to the expression of genes required for the assembly of the T3SS. Gen- Our previous results indicate that each of these mutations results in defective type three secretion in V. harveyi (24). To activate transcription from the promoters of the T3SS operons in these mutants, the luxO(D47E) allele was incorporated into the chromosome of each transposon mutant. As mentioned above, the luxO(D47E) allele constitutively locks the cells into the low-cell-density state and causes the maximal expression of T3SS genes. Consequently, the expression of exsD-gfp from pCMW209 is high in the luxO(D47E) strain (Fig. 4). Transposon mutations in the T3SS.1, T3SS.2, or T3SS.3 operon in combination with luxO(D47E) caused no decrease in exsD-gfp expression (Fig. 4), in contrast to what we observed for the exsB::cat mutation (Fig. 1, 2). Thus, we conclude that the loss of T3SS gene expression in the exsB mutant is not due to a secretion defect.

Mutation of exsB affects exsA.
To determine if the loss of ExsB production is solely responsible for reduced T3SS expression, we attempted to complement the exsB::cat mutation with the exsB gene and a region 500 bp upstream of its translation start site cloned into the pLAFR2 plasmid (Fig. 5A). The complementation plasmid was introduced into the ⌬luxR, exsB::cat mutant strain containing pCMW209-cat (exsD-gfp), and gfp expression was measured. As previously described, the expression of exsD-gfp is high in the ⌬luxR strain and low in the ⌬luxR, exsB::cat strain. However, the introduction of the exsB gene alone did not restore the transcription of exsD-gfp in the ⌬luxR, exsB::cat mutant (Fig. 5A). Using quantitative PCR, we confirmed that exsB mRNA was produced in the ⌬luxR, exsB::cat strain containing the complementation plasmid at levels equivalent to those in the ⌬luxR strain (data not shown). Therefore, exsB alone was unable to complement the exsB::cat mutation.
We reasoned that the failure of exsB to complement could be due to polar effects of the chromosomal exsB::cat mutation on downstream gene expression. To test this possibility, two more complementation constructs were engineered, one encoding exsB and the 601-bp intergenic region immediately downstream of the gene and one encoding exsB, this intergenic region, as well as the downstream gene exsA (Fig. 5A). Only the construct containing both exsB and exsA restored T3SS gene expression in the ⌬luxR, exsB::cat mutant strain (Fig. 5A). Therefore, the chromosomal exsB::cat mutation appears to lead to a reduction in T3SS gene expression due to polar effects on exsA because exsB and exsA exist in an operon. The orientation of these two genes is reminiscent of the exsB and exsA genes of Pseudomonas aeruginosa, which have been shown to be expressed in the exsCEBA operon (6,21,44). Although the exsA gene of V. harveyi is 44% identical to its P. aeruginosa exsA homolog, the exsB genes of these organisms are only 13% identical. Moreover, V. harveyi does not appear to encode an exsE homolog, and the exsC gene is divergently transcribed from exsB (Fig. 1A).
To determine if the V. harveyi exsB and exsA genes are cotranscribed, RNA was harvested from a ⌬luxR mutant, treated with DNase, and converted into cDNA. As a control, an equivalent amount of RNA was prepared under the same conditions, except that no reverse transcriptase was added (RTϪ). PCR was performed on genomic DNA, the RTϪ con- trol, and the cDNA using primers complementary to the 3Ј end of exsB and the 5Ј end of exsA. Amplification was observed from the genomic DNA and cDNA but not the RTϪ control, indicating that exsB and exsA are cotranscribed (Fig. 5B). exsA is regulated by QS and is required for T3SS gene expression. ExsA, a member of the AraC/XylS family of transcriptional activators, activates the expression of T3SS genes in P. aeruginosa (18,45), Y. enterocolitica (13), and V. parahaemolyticus (47). Our data suggest that QS regulates the production of the T3SS of V. harveyi through the control of exsA expression. To test this prediction, the expression of exsA was measured using quantitative PCR in the same two high-cell-density (wild-type and ⌬luxO) and low-cell-density [luxO(D47E) and ⌬luxR] strains of V. harveyi used for the analysis of exsB expression (Fig. 3). Indeed, the expression pattern for exsA is similar to that for exsB. That is, like exsB, exsA is maximally expressed at low cell density and minimally expressed at high cell density.
In P. aeruginosa, Y. enterocolitica, and V. parahaemolyticus, exsA mutants are defective in T3SS gene expression (6,7,12,40,47). To determine if this is also the case for V. harveyi, we constructed an in-frame deletion of exsA in the ⌬luxR background and measured the expression of VIBHAR_01699, the second gene in the T3SS.IV operon (Fig. 1A), using quantitative PCR. Consistent with our results described above, VIBHAR_01699 was more highly expressed in the ⌬luxR strain than in the wild-type strain, and mutation in exsB in the ⌬luxR background reduced VIBHAR_01699 expression (Fig. 6, black  bars). The deletion of exsA in the ⌬luxR background also reduced VIBHAR_01699 expression to the levels observed for the wild-type strain. Thus, ExsA is required for the activation of T3SS gene expression in V. harveyi.
ExsA is the regulator connecting QS to T3SS gene expression. The data in Fig. 5A show that ExsA is necessary for the expression of T3SS genes in the ⌬luxR, exsB::cat mutant strain. Is ExsA sufficient? To answer this question, the entire coding sequence of exsA containing a C-terminal FLAG tag was cloned under the control of the Ptac promoter and an exogenous translation start site on a complementation plasmid. This plasmid was introduced into four V. harveyi strains: the wild type (i.e., high cell density), ⌬luxR (i.e., low cell density), ⌬luxR, exsB::cat, and ⌬luxR, ⌬exsA, and T3SS gene expression was measured using quantitative PCR, again probing for VIBHAR_01699 levels. The exogenous production of ExsA increases the expression of VIBHAR_01699 in every strain examined, confirming that ExsA expression is sufficient to induce the expression of T3SS genes (Fig. 6, gray bars). Notably, induction occurs in the high-cell-density wild-type strain in which QS normally represses T3SS, indicating that the production of ExsA is epistatic to LuxR in controlling the expression of T3SS genes in V. harveyi.
The exsBA operon encodes two QS-regulated promoters. Taken together, the results described above suggest that QS, via LuxR, regulates the expression of the exsBA operon, and ExsA in turn controls T3SS gene expression. To determine if the regulation of exsBA by LuxR is direct, we analyzed the exsBA operon for transcription start sites using 5Ј-RACE analysis. One putative transcription start site was identified 370 bp upstream of the translation start site of exsA ( Fig. 1A; termed P A to denote exsA), and a second was identified 38 bp upstream of the translation start site of exsB (termed P B to denote exsB; marked by arrows in Fig. 1A).
To determine if these putative transcription start sites contain functional promoters, we constructed luciferase transcriptional fusions to each potential promoter. The exsB-lux promoter (P B ) fusion encodes the DNA sequence 535 nucleotides upstream of the exsB translation start to 17 nucleotides downstream of the translation start site. The exsA-lux promoter (P A ) fusion encodes the entire intergenic region between exsB and exsA from 109 nucleotides upstream of the exsB stop codon to 125 nucleotides downstream of the exsA translation start site. Luciferase production was measured in a V. harveyi strain that is wild type for QS but contains a Tn5 insertion in the luxA gene, the ⌬luxR mutant, and the ⌬luxR, ⌬exsA mutant. None of these three strains is bioluminescent, making them suitable hosts for the analysis of the lux transcriptional fusions. However, to ensure that the lux measured in this experiment was generated by the reporter plasmid and not the chromosomally encoded luciferase operon, luminescence from the promoterless parent vector, pBBRlux, also was determined (Fig. 7).
Luciferase production from the exsB promoter fusion oc- curred in the ⌬luxR mutant and the ⌬luxR, ⌬exsA strains; however, expression in the luxA::Tn5 mutant strain was equal to the background bioluminescence from the vector control (Fig. 7). This result indicates that the P B promoter is functional and is repressed in the high-cell-density QS state. Furthermore, the expression of the exsB-lux promoter construct also was significantly reduced in the ⌬luxR, ⌬exsA mutant compared to that of the ⌬luxR mutant (Fig. 7), suggesting that ExsA regulates its own expression through the autoactivation of the P B promoter driving the expression of the exsBA operon. Bioluminescence expression above background levels also was observed in the exsA-lux promoter reporter, although overall expression from this fusion was 10-fold lower than that from the exsB-lux promoter construct (Fig. 7). The exsA-lux promoter was strongly repressed in the luxA::Tn5 mutant compared to expression in the ⌬luxR strain; however, no significant difference in expression was observed between the ⌬luxR and ⌬luxR, ⌬exsA strains. These results show that the P A promoter is functional, although it is not expressed as highly as the P B promoter. Like the P B promoter, the P A promoter is repressed by QS. However, unlike the P B promoter, the P A promoter is not regulated by ExsA.
LuxR directly represses the P B promoter. Two lines of evidence suggest that QS primarily controls exsA expression through the transcriptional regulation of the P B promoter. First, the P B promoter is expressed 10-fold more than the P A promoter at low cell density, but it is repressed to low levels in the high-cell-density QS state. Second, the P A promoter is intact in the exsB::cat mutant, yet T3SS promoters are not expressed in this mutant, suggesting that insufficient ExsA is produced from the P A promoter to drive the transcription of the T3SS operons.
To examine the QS regulation of the P B promoter, we analyzed the surrounding nucleotide sequence for potential LuxR binding sites using a recently reported LuxR binding site prediction algorithm (33). Of note, no LuxR binding sites are predicted in the vicinity of the P A promoter. Two putative LuxR binding sites were identified near the P B promoter (Fig.   8A, underlined). One site is located at ϩ5 to ϩ22 relative to the transcription start site in the 5Ј-untranslated region of the exsBA mRNA, while the other site is located at the nucleotides encoded Ϫ43 to Ϫ63. We examined LuxR binding at these sites by performing EMSA with three regions of the exsBA promoter (Fig. 8B). The most upstream probe, containing a region of the exsBA promoter from Ϫ2843Ϫ116, does not possess a putative LuxR binding site, and, consistent with this, shows no interaction with LuxR (Fig. 8B, left). The central probe containing the region from Ϫ1723Ϫ6 that includes the predicted upstream LuxR binding site also did not bind to LuxR (Fig. 8B, middle). The most downstream probe, harboring the region from Ϫ553ϩ63 and encoding the second putative LuxR binding site, bound to LuxR (Fig. 8B, right). Therefore, we conclude that LuxR directly regulates the expression of exsBA through binding to a site located between the transcriptional and translational start sites of exsB (Fig. 8A).

DISCUSSION
In this study, we decipher the molecular mechanism by which QS controls the expression of T3SS genes in V. harveyi. At low cell density, when the concentration of AIs is low, LuxR is repressed, leading to the derepression of two promoters in the exsBA operon and the production of ExsA (and likely ExsB). ExsA, in turn, activates the expression of the operons encoding the T3SS structural genes. At high cell density, LuxR directly represses the transcription of the P B promoter located upstream of the exsBA operon, preventing the production of ExsA. Reduced levels of ExsA lead to the reduced expression of the four operons encoding the T3SS genes of V. harveyi.
QS regulation of T3SS has been reported for multiple bacteria. P. aeruginosa, like V. harveyi and V. parahaemolyticus, also represses T3SS at high cell density through a RhlR-C4homoserine lactone autoinducer-dependent QS pathway (5). On the other hand, the expression of T3SS in enteropathogenic and enterohemorrhagic E. coli is reported to be induced at high cell density by the QS molecule AI-3 (36) through the QseC/QseA two-component regulatory cascade (35,37). It is unclear why different bacterial pathogens express T3SS at different cell densities, but presumably these distinct patterns optimize the particular disease process of each pathogen. Although V. harveyi is not a human pathogen, V. harveyi does cause disease in marine organisms (1,16,46). The T3SS system of V. harveyi is proposed to be involved in these infections (24). The expression of T3SS genes at low cell density implies that only one or a few bacteria are sufficient to initiate disease by delivering effector proteins into target host cells. One can envision that, in the ocean, free-living V. harveyi cells exist at relatively low cell densities. Based on our results, under this condition, T3SS genes should be maximally expressed. Thus, the T3SS appears poised to initiate infection upon an individual bacterium encountering a suitable host.
Using purified LuxR and EMSAs, we determined that LuxR directly regulates the exsBA operon in V. harveyi at the P B promoter. This finding differs from that described for the QS regulation of the T3SS in P. aeruginosa, in which the operons encoding the structural genes for the T3SS are regulated by QS but the exsCEBA operon is not (5). LuxR acts as both an activator and a repressor; however, the parameters that govern when LuxR will activate gene expression and when LuxR will repress gene expression are not known. In the present work, we have determined that LuxR binds to a site located between the transcription start site and translation start site of exsB to repress expression. This promoter architecture resembles that described for hapR (encoding the LuxR homolog in Vibrio cholerae). In this case, HapR binding to its own promoter results in the autorepression of hapR expression (29). Thus, an emerging pattern is that the binding of LuxR-type proteins between the transcription and translation start sites of the target genes leads to repression rather than activation.
Our results identify two promoters located in the exsBA operon of V. harveyi that are regulated by QS, although it is unknown if the P A promoter is directly controlled by LuxR. Two pieces of evidence suggest that the P B promoter is the primary promoter controlling the transcription of exsBA. First, our studies indicated that expression from P B is 10-fold higher than that from P A . In addition, mutations in exsB reduced the expression of exsA even though they are located hundreds of base pairs upstream of the P A promoter. This second result suggests that any ExsA produced from the P A promoter alone is not sufficient to induce the expression of the T3SS operons. Nevertheless, P A is clearly transcribed and QS regulated, suggesting that it plays some role in T3SS gene regulation through the modulation of exsA expression. Further experimentation is required to clarify this mechanism. Finally, similar to our results showing that the V. harveyi exsBA operon is autoactivated by ExsA, the P. aeruginosa exsCEBA operon also is activated by its ExsA homolog (6).
QS in V. harveyi controls numerous genes both positively and negatively. Of the approximately 60 promoters identified to be QS regulated to date, only 15 are known to directly interact with LuxR (24,31,33,43). Thus, most of the genes in the LuxR regulon are indirectly controlled, suggesting that there exists a complex regulatory hierarchy emanating from LuxR to the control of the expression of downstream genes. Here, we describe one such multicomponent pathway that controls the expression of T3SS genes. In this case, LuxR functions as a repressor of the AraC-type transcriptional activator, ExsA, to promote T3SS expression exclusively at low cell density. In an analogous scenario, the LuxR repression of a repressor protein could promote the expression of target genes specifically at high cell density. As mentioned, LuxR also functions as a transcriptional activator. Having opposing regulatory abilities residing within one protein presumably provides V. harveyi with maximal flexibility to coordinate the expression of QScontrolled genes. Moreover, circuits involving additional regulatory layers could provide other mechanisms for the precise regulation of QS-controlled genes. These circuits likely function in a combinatorial fashion to direct the specific temporal expression of individual and collective behaviors at low and high cell density, respectively.