Output Targets and Transcriptional Regulation by a Cyclic Dimeric GMP-Responsive Circuit in the Vibrio parahaemolyticus Scr Network

Abstract

The Vibrio parahaemolyticus Scr system modulates decisions pertinent to surface colonization by affecting the cellular level of cyclic dimeric GMP (c-di-GMP). In this work, we explore the scope and mechanism of this regulation. Transcriptome comparison of ΔscrABC and wild-type strains revealed expression differences with respect to ∼100 genes. Elevated c-di-GMP repressed genes in the surface-sensing regulon, including those encoding the lateral flagellar and type III secretion systems and N-acetylglucosamine-binding protein GpbA while inducing genes encoding other cell surface molecules and capsular polysaccharide. The transcription of a few regulatory genes was also affected, and the role of one was characterized. Mutations in cpsQ suppressed the sticky phenotype of scr mutants. cpsQ encodes one of four V. parahaemolyticus homologs in the CsgD/VpsT family, members of which have been implicated in c-di-GMP signaling. Here, we demonstrate that CpsQ is a c-di-GMP-binding protein. By using a combination of mutant and reporter analyses, CpsQ was found to be the direct, positive regulator of cpsA transcription. This c-di-GMP-responsive regulatory circuit could be reconstituted in Escherichia coli, where a low level of this nucleotide diminished the stability of CpsQ. The molecular interplay of additional known cps regulators was defined by establishing that CpsS, another CsgD family member, repressed cpsR, and the transcription factor CpsR activated cpsQ. Thus, we are developing a connectivity map of the Scr decision-making network with respect to its wiring and output strategies for colonizing surfaces and interaction with hosts; in doing so, we have isolated and reproduced a c-di-GMP-sensitive regulatory module in the circuit.

INTRODUCTION

Cyclic dimeric GMP (c-di-GMP) plays a determining role in diverse adaptations of many bacteria, often by moderating switches between biofilm and planktonic lifestyles (reviewed in references 16 and 32). In Vibrio parahaemolyticus, this second messenger also participates in modulating choices; however, for this organism, we know that c-di-GMP plays a key role in influencing lifestyle decisions during growth on surfaces. In particular, it participates in the decision-making processes determining biofilm development or the profound differentiation events leading to swarming that occur during growth on surfaces. A high concentration of c-di-GMP impairs surface motility and promotes biofilm formation, whereas a smaller amount of this second messenger favors swarming. In part, the concentration of c-di-GMP that reciprocally influences swarming and capsule production is established by the membrane-bound ScrC enzyme, which contains functional GGDEF and EAL domains. Proteins with the GGDEF domain are responsible for the formation of c-di-GMP, whereas EAL or HD-GYP domain-containing proteins participate in the degradation of the second messenger (reviewed in references 12 and 18). During swarming, expression of the scrABC operon is upregulated. The enzyme ScrA produces the S signal, an autoinducer signaling molecule that is thought to interact with ScrB, which in turn promotes the c-di-GMP-degrading activity of ScrC (10, 14, 40). At the onset of swarming, c-di-GMP is present at a lower level than it is in planktonic cells or in a cell type that is locked off for swarmer cell differentiation; furthermore, the ScrABC proteins contribute to this pool, because c-di-GMP is higher in a ΔscrABC strain than the wild type (10, 13, 14).

Although we know that the amount of this second messenger in V. parahaemolyticus affects transcription of lateral flagellar (laf) and capsular polysaccharide (cps) genes, the global response to altered c-di-GMP concentrations has been unexamined until now. In this study, we investigate the output of the scr regulon with respect to the scope of genes that are regulated and the molecular mechanism by which transcription is modulated via fluctuating levels of c-di-GMP. By comparing the transcriptomes of wild-type and scrABC strains, the genome-wide consequences of altered c-di-GMP allow identification of new output members of the scr circuit. Many of these affect or have the potential to influence surface colonization, and so we found that the meaning of the scr designation can be expanded. A few transcription factors were also found to be regulated in the microarray comparison. Here, the role of one of these, named CpsQ, is described with respect to some of its target genes, its position in the complex network of cps regulation, and its responsiveness to c-di-GMP. A portion of this network could be reconstituted as a c-di-GMP-responsive module in Escherichia coli.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

The strains and plasmids used in this work are described in Table S1 of the supplemental material. All V. parahaemolyticus strains were derived from BB22TR (8) and were routinely grown at 30°C. Heart infusion (HI) medium contained 2.5% Bacto heart infusion broth (Difco) and 1.5% NaCl. HI swarm and nonswarm plates contained 1.5% and 2% Bacto agar (Difco), respectively. Congo red plates were prepared by adding 0.025% Congo red and 5 mM CaCl₂ to 2.5% HI medium (with no added NaCl) and were solidified with 2% granulated agar (Difco). Escherichia coli strains were grown at 37°C in LB medium containing 1% tryptone, 1% NaCl, and 0.5% yeast extract. Antibiotics were used at final concentrations of 10 μg/ml chloramphenicol for V. parahaemolyticus strains and 25 μg/ml for E. coli strains, 25 μg/ml gentamicin for V. parahaemolyticus strains and 15 μg/ml for E. coli strains, 50 μg/ml kanamycin, and 10 μg/ml tetracycline for both species. Expression plasmids were induced with 0.1 mM isopropyl-β-d-galactopyranoside (IPTG) unless otherwise indicated in the figure legends.

Genetic and molecular techniques.

Deletion/insertion mutations were made using a λ Red recombinase system in E. coli (6) to introduce drug resistance and/or lacZ reporter cassettes into cosmids for chromosomal allelic exchange. The ΔcpsQ1 and ΔcpsQ2 alleles removed 563 bp of internal coding sequence, with substitution of a chloramphenicol or kanamycin resistance cassette, respectively; the ΔcpsS2::Cam^r and ΔcpsS3::Kan^r alleles removed 562 bp of internal coding sequence with substitution of the drug cassettes; the ΔcpsSQ5::Cam^r allele deleted 1,554 bp to remove the coding sequence for the two genes. Kanamycin-resistant pDSW361 and derivatives were transferred to V. parahaemolyticus by electroporation; cosmids were transferred by conjugation. Conjugation and allelic replacement methods for V. parahaemolyticus have been previously described (34). Allelic replacements were confirmed by PCR.

For expression studies, cpsR (VP0514) and cpsQ (VPA1446) genes were amplified by high-fidelity PCR (Phusion polymerase; New England BioLabs, Inc.) and cloned into the IPTG-inducible expression vector pDSW361 to make pLM3503 and pLM3467, respectively. The wild-type cpsQ was also cloned into the His tag vector pCOLADuet-1 (Novagen). This clone (pLM3808) was used as a template for overlapping PCR to make a site-directed change resulting in the R134A substitution (and plasmid pLM4019). A His-tagged version of the cpsQ⁺ allele was subsequently cloned from pCOLADuet-1 by a blunt end ligation into pDSW361 to make pLM4074. Cloning was confirmed by sequencing.

The promoter probe plasmid pPROBE-gfp[AT] (28) was used to fuse the potential cpsA promoter region to gfp by cloning a PCR-amplified 442-bp region upstream of the cpsA coding region and using the EcoRI and BamHI restriction sites of pPROBE-gfp[AT]. The lacZ transcriptional fusion in cpsQ was constructed essentially as described previously (17). Briefly, PCR primers were designed and used to amplify the region containing lacZ and the antibiotic resistance cassette (∼5 kb) of a lacZ fusion plasmid, and this fragment was recombined into the cosmid containing the cpsSQ locus. This cosmid contained an internal markerless deletion of cpsS, which encodes another cps regulator, in order to obviate complications of introducing a second cps regulator. This deletion was derived from ΔcpsS3::Kan^r by using the flp recombinase (as per the methods described in reference 6). The cpsA- and cpsR-lacZ fusions have been described elsewhere (15).

Luminescence and β-galactosidase assays.

For time course experiments, V. parahaemolyticus strains were pregrown on plates and suspended to an optical density at 600 nm (OD₆₀₀) of 0.05, and 100-μl aliquots were spread onto multiple fresh HI swarm plates (with antibiotics and IPTG, when appropriate). Plates were incubated at 30°C. At specified times, cells were suspended in 5 ml of HI broth for OD₆₀₀ and relative light unit (RLU) measurements or β-galactosidase assays. All experiments were repeated at least three times with similar results. Luminescence was quantified by measuring samples in a TD20-202 luminometer (Turner Designs). Luminescence is reported as specific light units (SLU; RLU per second per ml per OD₆₀₀ unit). Exponentially growing liquid cultures of E. coli strains were diluted to an OD₆₀₀ of 0.1 into fresh medium with antibiotics and IPTG. Samples were harvested periodically for measurements. LacZ activity was measured according to the methods described by Miller (27), and cells were permeabilized using Koch's lysis solution (31). Lux and LacZ measurements were performed in triplicate; averages of these triplicates are reported, and error bars show standard deviations.

SDS-PAGE.

Cultures were harvested in Laemmli sample buffer, and proteins were fractionated using sodium dodecyl sulfate-polyacrylamide gel electrophoresis. The 12.5% acrylamide resolving gel was 7.5 cm in length, and it was stained with Coomassie blue as previously described (4, 26).

Transcriptome analysis.

The custom GeneChip representing the V. parahaemolyticus RIMD2210633 genome has been described previously (14) and was constructed by Affymetrix Inc. (Santa Clara, CA). The V. parahaemolyticus strain BB22 with which these studies were performed has not been sequenced; however, probing the Affymetrix GeneChip with genomic BB22 DNA yielded present calls for ∼98% of the predicted open reading frames. It seems reasonable to conclude that these transcriptome studies provide a good reflection of genome-wide expression patterns in BB22, although some fraction of the gene activity could not be assessed.

We compared transcript profiles of V. parahaemolyticus LM5674 with its ΔscrABC derivative, strain LM6567, grown on HI plates. Methods for the growth and harvesting of cells and RNA and cDNA preparation have been described previously (13, 14). The cDNA samples were hybridized and scanned at the University of Washington Center for Expression Arrays. Two microarray replicates per strain were used in our study and, therefore, statistical analysis of the microarray data was limited. Results are the averages of two independent biological experiments per strain, and the results were analyzed by using the Affymetrix Microarray suite version 5 and Cyber-T (http://cybert.microarray.ics.uci.edu/) (1) at a P level of <0.05. Measures of interreplicate reproducibility are provided in Table S2 of the supplemental material.

CpsQ purification and c-di-GMP extraction.

Two preparations of wild-type CpsQ and one of mutant CpsQ were purified. The His-tagged versions of wild-type and mutant CpsQ were produced using the pCOLADuet-1 expression vector in strain BL21(DE3). The wild-type protein was also produced using the pDSW341 vector in strain DH5α. In addition, the strains contained a plasmid expressing scrC⁺ in order to increase the intracellular concentration of c-di-GMP and maximize the potential yield of ligand-bound target protein. The strains were grown in Terrific broth (12 g/liter tryptone, 24 g/liter yeast extract, 4 ml/liter glycerol, 0.017 M KH₂PO₄, 0.072 M K₂HPO₄) with 50 μg/ml kanamycin to maintain the cpsQ plasmids and 125 μg/ml ampicillin to maintain the scrC⁺ plasmid. Strains were induced using 0.1 mM IPTG. The BL21 strains were induced at an OD₆₀₀ of 0.8 and grown for 16 h at 20°C; the DH5α strain was induced at an OD of 1.0 and grown for 3.5 h at 37°C. Cells were centrifuged and suspended in lysis buffer, and pellets were frozen at −80°C. The lysis buffer (pH 7.5) contained 50 mM HEPES, 550 mM NaCl, 20 mM imidazole, 5 mM β-mercaptoethanol, and 1 mM phenylmethylsulfonyl fluoride. Cells were lysed using a French pressure cell, cell debris was pelleted by centrifugation, and cleared lysates were incubated with TALON metal affinity resin and purified according to instructions from the manufacturer (Clontech, Mountain View, CA). Protein was eluted in 50 mM HEPES, 550 mM NaCl, 500 mM imidazole (pH 7.5) and concentrated using a 10K Amicon Ultra centrifugal filter (Millipore, Ireland). High salt was found to be key to the purification of the CpsQ homolog VpsT (23). Yield and purity were assessed using the Bradford method (Bio-Rad Life Sciences) and inspection by SDS-PAGE. Cultures yielded ∼140 μg purified protein/g (wet weight) of cells. Purity was judged to be similar and ∼95% for all preparations.

Purified CpsQ was extracted by two methods that were modifications of whole-cell extraction methods (14, 36). For extraction method 1 (heat-acid-organic), 30 μl of protein in elution buffer was heated at 95°C for 10 min, and samples were then placed on ice for 10 min. Ice-cold 11 N formic acid (1.5 μl) was added, and samples were vortexed and incubated on ice for 30 min. Cold methanol and acetonitrile (60 μl each) were added with vortexing, and samples were incubated for 15 min on ice. Cell debris was pelleted by centrifugation at 12,500 rpm for 20 min at 4°C. The supernatant (125 μl) was neutralized with 12 μl 20% NH₄HCO₃ (pH ∼5) and dried using a speed vacuum at room temperature. Samples were suspended in 125 μl mobile phase A buffer (described below) and centrifuged for an additional 15 min at 4°C. Extraction method 2 (heat-organic) was similar except for the elimination of the acid treatment and neutralization step. The protein samples were extracted in methanol-acetonitrile-elution buffer (40:40:20; 150-μl final volume) for 15 min on ice, then heated and treated as above (without the neutralization step). Eleven extractions were performed: two concentrations of each protein preparation were extracted (75 and 150 μg) using both extraction methods, except for the mutant (because protein was limiting). Bovine serum albumin (BSA; 150 μg) and BSA spiked with 150 nM c-di-GMP were extracted similarly by both methods to serve as controls.

Samples were analyzed by the Spectrometry Facility at Michigan State University using liquid chromatography-tandem mass spectrometry (LC-MS/MS). A standard calibration curve was generated using 7 concentrations (5 to 500 nM) of c-di-GMP (Biolog Life Science Institute) in mobile phase A. Mobile phase A was 10 mM tributamine–15 mM acetic acid in 3% aqueous methanol, pH ∼4.95. The LC-MS/MS analysis has been described elsewhere (2, 14). The complete summary report for standards and the test samples can be found in the Table S3 of the supplemental material.

Statistical analysis.

Tests for statistical significance were conducted by using Student's t test (two-tailed distribution with two-sample, equal variance calculations).

Microarray data accession number.

Transcriptome data have been submitted to the NCBI and assigned the GEO accession number GSE30508.

RESULTS

ScrABC regulates multiple genes in V. parahaemolyticus.

The transcriptome of the scrABC mutant strain LM6567 was compared to that of the wild-type strain LM5674 in order to assess the global effect of scrABC on gene expression. Because the scrABC operon is known to modulate laf expression during growth on surfaces (10), these strains were grown on plates and harvested for RNA sampling at the optimal time for laf gene induction (14). Successful capture of the RNA samples was evaluated by quantitative reverse transcription-PCR (qRT-PCR) to quantify transcription of gapA (a constitutive control), lafA, and flgB_L. The levels of lafA and flgB_L mRNAs were ∼7- and 20-fold higher in the wild-type strain than in the ΔscrABC strain, respectively (data not shown), consistent with the expected levels of transcription for these two genes. Two independently grown and prepared RNA samples for each strain were used to query the custom V. parahaemolyticus microarray Affymetrix GeneChip. A total of 106 genes were differentially expressed between the two strains when we used a 2-fold cutoff and P level of <0.05 (see Table S2 in the supplemental material).

ScrABC positively regulated 77 genes. Many were lateral flagellar genes, and they were repressed between 4- and 35-fold in the ΔscrABC strain compared to its parent (see Table S2 in the supplemental material). This was an expected result and provided one kind of validation of the transcriptome analyses, because the scrABC operon was discovered using a screen for altered expression of the lateral flagellar hook operon, which is a class 2 flagellar gene (5). It also provided new information, because genes at all levels in the flagellar transcriptional hierarchy were found to be differentially expressed. In addition, the transcriptional profile revealed that scrABC exerted control over the entire surface-sensing regulon, which includes nonflagellar as well as flagellar genes. All but 8 of the 79 genes that were repressed in the scrABC mutant belong to the specialized program of gene control that V. parahaemolyticus invokes in response to growth on surfaces (14).

To verify the regulation of some members of this group, the scrABC operon was deleted by allelic exchange in various lux reporter strains that had been isolated previously on the basis of displaying surface-induced luminescence (14, 37). Each reporter strain and its ΔscrABC derivative were grown on plates and harvested periodically to measure growth (OD₆₀₀) and luminescence over a period of 5 to 7 h of growth. Figure 1 reports the percent maximal specific light units produced by the mutant normalized to the maximal luminescence of the parental strain. Mutation of the scrABC operon in a negative-control strain containing a lux fusion in a gene (VP0829) that was not observed regulated in the microarray analysis had little effect on luminescence. In comparison, ΔscrABC caused the diminished expression of class 1 as well as class 2 lateral flagellar genes (VPA1544 and VPA0267, respectively). Class 1 genes are expressed prior to class 2 genes in a temporal hierarchy of gene control (14). The regulation of both classes of flagellar genes was similarly affected by ΔscrABC, showing ∼30-fold-diminished expression, suggesting that c-di-GMP exerts control at the highest level in the flagellar regulatory cascade. The effects of deleting scrABC on reporters in other, nonflagellar members of the surface-sensing regulon were examined. The scrABC deletion reduced expression of VPA0227 (encoding an alkaline serine protease) by 12-fold, of VP1002 (encoding a 906-amino-acid protein with a lipoprotein signal sequence) by 5-fold, and of VPA1598 (encoding an N-acetylglucosamine/chitin-binding protein) by 2,000-fold.

Fig 1.

New output targets of the Scr circuit. Gene expression was examined in pairs of ΔscrABC and scrABC⁺ luminescent reporter strains. Strains were spread on multiple HI plates, harvested hourly between 5 and 8 h, and assayed for luminescence. Values reported are the percent activity in the ΔscrABC mutant strain compared to the parental strain for each lux reporter strain at the time corresponding to maximal activity for the parental strain, which was usually 7 h. Strain pairs for each lux fusion (scr mutant/parent) were as follows: VP0829::lux ΔscrC/VP0829::lux (LM9081/LM6321); VPA0267::lux ΔscrABC/VPA0267::lux (LM6565/LM1017); VPA1544::lux ΔscrABC/VPA1544::lux (LM9503/LM8882); VPA0227::lux ΔscrABC/VPA0227::lux (LM9621/LM9376); VPA1598::lux ΔscrABC/VPA1598::lux (LM9492/LM6159); VP1002::lux ΔscrABC/VP1002::lux (LM9497/LM6161); VPA1443::lux ΔscrABC/VPA1443::lux (LM9837/LM6901). The predicted products of these genes are N-acetylglucosamine-6-phosphate deacetylase (VP0829), alkaline serine protease (VPA0227), class 2 lateral flagellar gene product (VPA0267 [FlgE]), class 1 lateral flagellar gene product (VPA1544 [FliR]), N-acetylglucosamine/chitin-binding protein (VPA1598), lipoprotein (VP1002), and type I secretion membrane fusion protein (VPA1443). Error bars indicate standard deviations of the averages of ratios obtained from at least three independent experiments. In all experiments, the reporter activity for each scr mutant was significantly different from the activity in its parental strain (P < 0.0003), with the exception of the control reporter VPA0828::lux pair.

In the opposite direction of gene control, the scrABC operon negatively regulated expression of 29 genes (see Table S2 in the supplemental material). Eleven were capsular polysaccharide genes (cpsA to cpsJ), in agreement with previous data demonstrating that scrABC represses the transcription of the biosynthetic cps genes (5, 10). Five of these genes encode hypothetical proteins. Other genes encode products that seem associated with the cell surface, including a predicted pilin, an agglutination protein, and other proteins involved in polysaccharide biosynthesis and phosphorylcholine metabolism, including LicD1 (see Table S2).

The scrABC operon also repressed the mfp (membrane fusion protein) operon, which consists of three genes known to affect cell surface properties. The first gene in the operon, VPA1445, encodes a potential secreted calcium-binding protein (prosite pattern PS00330) (33); the second gene, VPA1444, encodes a potential ABC-type transporter with both ATPase and permease components (COG4618) (39); the last gene, VPA1443, is predicted to encode a protein homologous to members of the type 1 secretion membrane fusion proteins of the HlyD family (pfam00529) (11). Mutants with transposon insertions in this operon have been studied previously. As these strains were isolated as biofilm-defective strains, these mutants form biofilms with compromised structural integrity and display altered colony morphologies and color on Congo red agar (7). To confirm a role for c-di-GMP in the regulation of the mfp operon, the ΔscrABC allele was introduced into the vpa1443::lux reporter strain. The parental strain and the scr derivative were grown on plates under conditions similar to those for growth for the microarray experiment and monitored for light production. The scrABC vpa1443::lux strain produced ∼1.8-fold more light than the vpa1443::lux parental strain (Fig. 1). Thus, we have found new V. parahaemolyticus genes negatively and positively regulated by scrABC, and many of these seem pertinent to the cell surface.

CpsQ is a new transcriptional regulator of CPS.

Among the genes whose transcription was affected by deletion of scrABC were 5 encoding transcriptional regulators—specifically, the lateral flagellar regulator LafK, the type III secretion regulator ExsA, and three others. Two, VPA1446 and VP2710, were each transcribed at levels ∼3-fold higher in the ΔscrABC strain than in the wild type. Both VPA1446 and VP2710 contain C-terminal helix-turn-helix DNA-binding domains that share homology with LuxR/GerE family transcriptional regulators (pfam00196; E-values of 6.41e−20 and 1.35e−15, respectively). VPA1446 maps between the cpsS gene (VPA1447) and the mfp operon. CpsS is another LuxR family transcriptional regulator (E-value, 1.37e−17). It was found previously to strongly repress cps gene expression (15). Due to its genetic linkage, we targeted VPA1446 as a transcription factor potentially pertinent to the regulation of cell surface characteristics. We were also intrigued by its homology to members of the CsgD family of LuxR-type helix-turn-helix DNA-binding proteins, because many have been implicated in c-di-GMP-mediated regulation (23, 42). The coding sequence for VPA1446 was cloned into an IPTG-inducible expression vector, and the plasmid was introduced into V. parahaemolyticus. Ectopic expression of VPA1446 produced a crinkly colony morphology on Congo red agar (Fig. 2A). Swarming was unaffected (data not shown). The consequence of expression of VPA1446 was further examined by using a cpsA::lacZ reporter strain (Fig. 2B). IPTG-induced expression of VPA1446 increased β-galactosidase activity by ∼4-fold. Because of its effect on expression of the capsule-encoding locus, we named the gene encoding VPA1446 cpsQ.

Fig 2.

CpsQ affects colony morphology and cpsA expression. (A) IPTG-induced expression of cpsQ induces a crinkly colony morphotype on Congo red medium. Single colonies of LM7794 (wild type/vector) and LM7796 (wild type/cpsQ⁺) were picked with toothpicks on HI Congo red medium with kanamycin and 0.1 mM IPTG. Colony morphology distinctions were apparent after 2 days growth, continued to develop with room temperature incubation, and were photographed at day 9. (B) IPTG-induced expression of cpsQ increases expression of a cpsA::lacZ reporter fusion. Strains LM7812 (cpsA::lacZ/vector) and LM7814 (cpsA::lacZ/cpsQ⁺) were spread on HI kanamycin plates with 0.1 mM IPTG. Plates were incubated for 22 h and harvested for β-galactosidase measurements. Error bars indicate standard deviations of triplicate measurements from a representative experiment. The difference was statistically significant (P < 0.0001).

CpsQ, like CpsR, is necessary for scr-dependent regulation of cpsA.

Deletion/insertion mutations in cpsQ were introduced into the chromosome of various cpsA::lacZ reporter strains. Mutants with lesions in the scrABC operon (i.e., ΔscrABC4 or ΔscrA1) have elevated expression of a cpsA::lacZ reporter fusion compared to the wild-type strain (10) (Fig. 3). Deletion of cpsQ in the ΔscrA1 background suppressed this elevated level of gene expression and resulted in β-galactosidase activity similar to that produced in the wild type. When introduced into the wild-type background, the cpsQ mutation caused little effect on colony morphology (data not shown) or cps gene expression (Fig. 3). These observations suggest that CpsQ requires high levels of c-di-GMP for activity.

Fig 3.

Loss of cpsQ, like cpsR, suppresses the scrA phenotype. cpsA::lacZ expression was assayed in strain LM5984 (wild type) and derivatives LM9690 (ΔscrABC4), LM6241(scrA1), LM6133 (cpsR1), LM9694 (cpsQ2), LM6243 (scrA1 cpsR1), and LM9695 (scrA1 cpsQ2). Cultures were spread on HI plates and incubated for 22 h before harvesting for β-galactosidase assays. Error bars indicate standard deviations from triplicate assays. ***, P < 0.0001.

The cpsQ phenotype is reminiscent of the cpsR phenotype. CpsR was discovered in a mutant screen for suppressors of the crinkly colony morphology of a scrA mutant (15). Introduction of the cpsR mutation also reduces cpsA::lacZ expression in the ΔscrA1 strain and has little effect upon introduction in a wild-type background (15) (Fig. 3). CpsR, encoded by VP0514, is homologous to members of the AAA+ family of transcriptional regulators, including VpsR of V. cholerae and Vibrio fischeri (reviewed in reference 41). Taken together, the data indicate that CpsR and CpsQ are necessary for the elevated cpsA transcription elicited by c-di-GMP.

Transcription of the mfp operon is affected by mutation of cpsQ.

Because cpsQ was closely linked to the mfp operon, and the mfp operon (like the cpsA operon) displayed enhanced transcription in scr mutants (Fig. 1), we wondered if cpsQ regulated the mfp operon in addition to the cps operon. Introduction of ΔcpsQ1 decreased the production of luminescence in the mfp::lux reporter strain compared to the parental strain (Fig. 4A). Although, the two loci are separated by ∼200 bp (Fig. 4B) and the biofilm and colony morphology phenotypes of mfp and cpsQ mutants are distinct, we considered the possibility that there may be some polarity of the cpsQ1 allele on transcription of the mfp operon. To examine this, cpsQ⁺ was provided in trans by introducing the plasmid pLM3102 (Fig. 4C). A negative-control plasmid carrying the cpsQ1 allele was also introduced (pLM3767 derived from pLM3102). The provision of cpsQ⁺ in trans was sufficient to increase mfpC::lux expression in the wild-type background (∼2.5-fold) and in the cpsQ1 background (∼20-fold). The level of expression resulting from the cpsQ1/cpsQ⁺ configuration was ∼3-fold lower than a cpsQ⁺/cpsQ1 configuration. In addition, RT-PCR analysis detected a product spanning the cpsQ-mfpA intergenic region (see Fig. S1 in the supplemental material). Thus, it seems clear that CpsQ in trans can stimulate mfp gene expression, although it may do so at least in part through effects at the cpsQ promoter.

Fig 4.

The mfp operon is regulated by CpsQ. (A) Mutation of cpsQ decreases expression of mfpC::lux. The mfpC::lux reporter strain LM6901 and its cpsQ1::Cam^r derivative, LM9836, were grown on HI plates and harvested at the indicated times for OD₆₀₀ and light measurements. Luminescence is reported as normalized light units (SLU). Error bars indicate standard deviations of triplicate light measurements. There is a statistical difference between the wild-type (wt) and cpsQ1 mutant strain at 6 and 7 h (P < 0.002). (B) Organization of the cpsS-cpsQ-mfp locus. Arrows (with gene tag numbers) indicate the direction of transcription and relative size of each gene; the lollipop indicates a predicted transcriptional terminator; sizes of the intergenic regions are indicated in bp. (C) CpsQ activates mfpC::lux in trans. A cosmid carrying the wild-type cpsS-cpsQ-mfp locus and its congenic cpsQ1 derivative were introduced into cpsQ1::Cam^r mfpC::lux and cpsQ⁺ mfpC::lux strains. Strains were grown on HI plates with tetracycline and harvested at 7 h for light measurements. Strains included LM10178 (cpsQ1/cpsQ1), L10177 (cpsQ1/cpsQ⁺), LM10181 (cpsQ⁺/cpsQ1), and LM10180 (cpsQ⁺/cpsQ1). Luminescence was measured and is reported as described for panel A. Light production was significantly different for each strain compared to the others (P < 0.0001).

CpsR regulates cpsQ transcription.

A cpsQ::lacZ transcriptional reporter was constructed on the low-copy-number cosmid carrying the cpsQ locus. The reporter cosmid was introduced into various wild-type and mutant strains. Consistent with the microarray analysis, cpsQ::lacZ expression was elevated ∼2- to 3-fold in two scr strains compared to the level observed in the wild-type strain (Fig. 5). When the reporter cosmid was introduced into an scrA1 cpsR1 double mutant strain, β-galactosidase activity was decreased ∼3-fold compared to the scrA1 single mutant strain. Thus, cpsQ transcription is elevated in scr strains, which have a high c-di-GMP level (14), and this increased expression requires CpsR.

Fig 5.

Expression of cpsQ is elevated in scrA mutants and this requires CpsR. A cpsQ::lacZ reporter that was carried on a cosmid was introduced into the following strains: wild type (wt; to make LM9771), ΔscrABC (LM9776), scrA1 (LM9772), cpsR1 (LM9773), and scrA1cpsR1 (LM9774). Cultures were spread on HI tetracycline plates, and the cells were harvested after 22 h of incubation for β-galactosidase measurements. Error bars indicate standard deviations of triplicate measurements. ***, P < 0.001.

CpsQ suppresses cpsS, and CpsS represses cpsR.

In addition to CpsR, CpsS is the other known regulator of capsule production. It too was discovered in a transposon mutagenesis screen: disruption of cpsS resulted in a super-rugose colony morphology and elevated cpsA::lacZ transcription (15). Since it was known that cpsR was epistatic to cpsS, we examined their relationship with cpsQ. Like cpsR, mutation of cpsQ was sufficient to suppress the superelevated cpsA::lacZ levels associated with the cpsS phenotype (Fig. 6A). To further explore this network of transcriptional regulators, a cpsR::lacZ transcriptional reporter was introduced into various strains (Fig. 6B). LacZ activity was elevated in both ΔcpsS and ΔcpsSQ strains compared to the wild-type background. Taken together, this evidence seems consistent with a regulatory pathway in which CpsS represses cpsR, CpsR activates cpsQ, and CpsQ activates cpsA.

Fig 6.

cpsQ is epistatic to cpsS, and CpsS regulates cpsR. (A) The cpsA::lacZ allele was recombined onto the chromosome in the indicated backgrounds to make the following reporter strains: wild type (wt; LM5984), ΔcpsS (LM9745), ΔcpsSQ (LM9748), ΔcpsS cpsR1 (LM9039), ΔcpsSQ cpsR1 (LM9041). Expression of cpsA is significantly different in the cpsS mutant compared to the wild type and in the other mutants compared to the cpsS strain (P < 0.001). (B) The cpsR::lacZ reporter carried on a cosmid was introduced into the following backgrounds to make merodiploid reporter strains: wild type (wt; LM10103), ΔcpsS (LM10101), ΔcpsSQ (LM10102). Expression of cpsA was significantly different in two mutant strains compared to the wild type (P < 0.002). Cultures were spread on HI tetracycline plates, and the cells were harvested after 18 h of incubation for β-galactosidase measurements. Error bars indicate standard deviations of triplicate measurements.

CpsQ is a direct regulator of cpsA transcription.

The epistasis experiments performed in V. parahaemolyticus suggested that CpsQ acts after CpsR to regulate transcription of the cps locus. To test whether CpsQ directly regulates cpsA gene expression, regulation was reconstituted in E. coli. The kanamycin-resistant cpsQ expression clone was induced with IPTG in an E. coli strain that also contained the cpsA::lacZ fusion on a tetracycline-resistant cosmid. This cosmid carried a large genomic insert (VPA1402 to VPA1413) containing the cpsA locus. CpsQ was found to activate cpsA::lacZ expression ∼26-fold compared to a strain carrying the vector control (Fig. 7A). To confirm this, a second, compatible plasmid was constructed carrying only the promoter region of the cpsA gene fused to the gfp gene (P_cpsA::gfp). Specifically, the intergenic region between VPA1402 and VPA1403 (cpsA) was cloned into the pPROBE-AT vector (28). CpsQ activated PcpsA::gfp about 50-fold above the level of the strain carrying the vector (Fig. 7B).

Fig 7.

CpsQ is sufficient to induce cpsA expression in E. coli. (A) CpsQ stimulates cosmid-derived cpsA::lacZ expression in E. coli. Exponentially growing E. coli strains LLM3727 (DH5α containing the tetracycline-resistant cpsA::lacZ cosmid and the kanamycin resistance expression vector) and LLM3728 (DH5α containing the cpsA::lacZ cosmid and cpsQ⁺ expression plasmid) were induced in LB medium (with tetracycline, kanamycin, and 0.1 mM IPTG) and grown for 3 h before harvesting for β-galactosidase assays. (B) CpsQ stimulates PcpsA::gfp expression in E coli. The intergenic region upstream of the cpsA coding region was cloned into the pPROBE-gfp vector. Exponentially growing E. coli strains LLM3744 (DH5α containing the ampicillin resistance PcpsA::gfp plasmid and the kanamycin resistance expression vector) and LLM3745 (DH5α containing the PcpsA::gfp plasmid and cpsQ⁺ expression plasmid) were induced in LB medium (with ampicillin, kanamycin, and 0.1 mM IPTG). Samples were harvested at 3 h for fluorescence and OD₆₀₀ measurements. Fluorescence is reported as units per ml per OD₆₀₀ unit. All samples were measured in triplicate (error bars indicate standard deviations). The differences elicited by CpsQ and vector were statistically significant (P < 0.0001 for both panels).

CpsR regulates cpsQ in E. coli and CpsQ regulates itself.

The analysis of cpsQ gene expression in V. parahaemolyticus demonstrated that CpsR regulates cpsQ. To establish whether CpsR was the direct regulator, cpsR was expressed using an IPTG-inducible plasmid in an E. coli strain carrying the cpsQ::lacZ reporter cosmid. CpsR stimulated cpsQ expression greater than 3-fold (Fig. 8A). This experiment was performed using a very low IPTG concentration (0.01 mM, compared to our usual 0.1 mM inducing concentration), because higher levels of induction of cpsR were toxic to E. coli. We also examined the ability of CpsR to affect cpsA::lacZ expression. The concentration of IPTG (0.01 mM) that was sufficient to induce cpsR and produce a positive regulatory effect on cpsQ::lacZ did not promote activation of the cpsA::lacZ promoter; whereas this concentration was sufficient to induce a similarly constructed cpsQ plasmid and allow some stimulation of cpsA::lacZ expression (Fig. 8B), albeit not as robust as the stimulation obtained using 0.1 mM (Fig. 7A). Taken together, the data support a regulatory pathway in which CpsR activates cpsQ, and CpsQ activates cpsA. In addition, IPTG-induced expression of cpsQ (using 0.1 mM IPTG) stimulated cpsQ::lacZ expression in E. coli by ∼10-fold (Fig. 8C). Thus, CpsQ regulates cpsQ gene expression.

Fig 8.

CpsR activates cpsQ expression in E. coli, and CpsQ regulates its own expression. (A) CpsR stimulates cpsQ::lacZ. E. coli strains carrying the cpsQ reporter plasmid and IPTG-inducible expression vector (LLM3975) or cpsR⁺ clone (LLM3978) were induced using 0.01 mM IPTG and grown for 7 h before harvesting for β-galactosidase assays. LLM3978 produced significantly more activity than LLM3975 (P < 0.0001). (B) CpsR does not stimulate cpsA::lacZ. E. coli strains carrying the cpsA reporter plasmid and the IPTG-inducible expression vector (LLM3727), cpsR⁺ clone (LLM3982), or cpsQ⁺ clone (LLM3728) were induced using 0.01 mM IPTG and grown for 6 h before harvesting for β-galactosidase assays. Although the positive-control CpsQ significantly increased cpsA::lacZ expression (P < 0.0001), CpsR did not stimulate the reporter. (C) CpsQ regulates its own expression. E. coli strains carrying the cpsQ reporter plasmid and IPTG-inducible expression vector (LLM3975) or cpsQ⁺ clone (LLM3977) were induced using 0.1 mM IPTG and grown for 3 h before harvesting for β-galactosidase assays. LLM3977 produced significantly more activity than LLM3975 (P < 0.0001). Exponentially growing strains containing the tetracycline resistance reporter cosmids and the kanamycin resistance expression vector or clones were diluted to an OD₆₀₀ of 0.1 into LB medium (with tetracycline, kanamycin, and IPTG) and grown for the indicated times before harvesting for β-galactosidase assays. Error bars indicate standard deviations of the mean of triplicate measurements.

c-di-GMP stimulates the ability of CpsQ to activate cpsA transcription in E. coli.

To probe the role of c-di-GMP in transcription of cpsA, cpsQ⁺ was coexpressed with either scrC⁺, scrABC⁺, or the vector control in the E. coli strain containing the cpsA::lacZ cosmid. In V. parahaemolyticus, cellular c-di-GMP levels can be altered ∼10-fold upon ectopic expression of scrC⁺ versus scrABC⁺, and a similar magnitude of c-di-GMP production is elicited by scrC⁺ expression in E. coli (10). For these experiments, each of the strains contained the cpsA::lacZ cosmid and two compatible plasmids: plasmids of type 1 were used for CpsQ expression and type 2 plasmids were used for scr gene expression. The presence of the vector used for cpsQ expression (plasmid 1) or the vector used for scr gene expression (plasmid 2) produced no effect on cpsA::lacZ expression (Fig. 9A). Neither did the expression of scrC⁺ or scrABC⁺ with the vector in the absence of cpsQ⁺ coexpression. Coexpression of scrC⁺ (resulting in high c-di-GMP) with cpsQ⁺ stimulated cpsA::lacZ expression almost 2-fold more than in the strain carrying cpsQ⁺ and the control vector. Coexpression of scrABC⁺ with cpsQ⁺ completely abolished the ability of CpsQ to activate cpsA transcription. Growth was similar in all strains. These results suggested that c-di-GMP plays a direct role in the ability of CpsQ to activate capsular polysaccharide gene expression.

Fig 9.

CpsQ-dependent regulation of cpsA is dependent on c-di-GMP in E. coli. The Scr circuit was examined in E. coli by introducing components on three plasmids. All strains contained the cpsA::lacZ reporter cosmid (tetracycline resistant). In addition they contained kanamycin resistance plasmids (type 1) encoding cpsQ⁺ (Q) or the parent vector (V) and ampicillin resistance plasmids (type 2) encoding scrC⁺ (C), scrABC⁺ (ABC), or the parent vector (V). These strains included the following: LLM3760 (vector 1 and vector 2), LLM3761 (cpsQ⁺ and vector 2), LLM3762 (vector 1 and scrC⁺), LLM3763 (cpsQ⁺ and scrC⁺), LLM3752 (cpsQ⁺ and scrABC⁺), LLM3757 (vector 1 and scrABC⁺). The scr and cpsQ genes were IPTG inducible. Strains were grown for 3 h in LB with antibiotics and 0.1 mM IPTG and harvested for β-galactosidase assays (A) and protein samples (B). The SDS-PAGE resolving gel (12%) was stained with Coomassie blue. Lanes: 1, LLM3760; 2, LLM3761; 3, LLM3763; 4, LLM3752; 5, LLM3760.

Detection of CpsQ in E. coli is dependent on c-di-GMP.

In the E. coli experiment shown in Fig. 9A, the transcription and translation of cpsQ were controlled by the exogenous, IPTG-inducible promoter and ribosome-binding sequences of the expression vector; thus, the mechanism by which c-di-GMP affected the ability of CpsQ to modulate cps gene expression appeared to occur posttranslationally. We reasoned that c-di-GMP could achieve its effect by altering CpsQ activity, as observed for V. cholerae's VpsT, or CpsQ stability, as was found for Salmonella's CsgD (23, 42). To distinguish between these possibilities, the protein profiles of CpsQ under high and low c-di-GMP conditions were examined on Coomassie-stained SDS-polycrylamide gels (Fig. 9B). The predicted molecular mass of CpsQ is 25,149 Da. A protein band of the appropriate molecular size was observed in the strains that exhibited activation of cpsA::lacZ, i.e., in strains with plasmids bearing cpsQ⁺ and vector or cpsQ⁺ and scrC⁺; however, this band was missing in strains that coexpressed cpsQ⁺ and scrABC⁺. Coexpression of scrABC (resulting in low c-di-GMP levels) with cpsQ resulted in a protein profile that resembled that of the control strain and which did not harbor the cpsQ⁺ expression plasmid. Thus, activation of cpsA::lacZ correlated with the presence of CpsQ, and the level of observed CpsQ was dependent on the c-di-GMP concentration in the cell.

CpsQ is a c-di-GMP-binding protein.

To directly examine the capacity of CpsQ to bind c-di-GMP, the protein was purified. Wild-type cpsQ⁺ and a site-directed mutant allele cpsQ(R134A) were cloned to tag each protein at its N terminus with six-His. For VpsT, the analogous mutation eliminates one of the key residues participating in c-di-GMP binding (23). The tagged form of wild-type CpsQ, but not CpsQ_R134A, was able to activate cpsA::lacZ in E. coli (Fig. 10A). An immunoblot of these cultures is provided in Fig. S2 of the supplemental material. To purify the proteins, they were produced in strains expressing scrC⁺ in order to maximize the amount of potential ligand during overproduction. Two independent batches of wild-type CpsQ and one of the mutant CpsQ protein were purified. Eleven extractions were performed (8 for the wild type and 3 for the mutant), using protein at two concentrations and by using two different procedures, one involving formic acid, then heat and organic extraction and the other with heat and organic treatment. Extracts were analyzed by liquid chromatography-tandem mass spectrometry. The mutant preparations yielded concentrations of c-di-GMP of less than 1.4 nM, whereas the wild-type protein produced 345 to 2,400 nM c-di-GMP. A summary of the LC-MS/MS quantification is provided in Table S3 of the supplemental material. The yields, expressed as nmole of c-di-GMP/μmole of protein, are reported as averages of the replicate extractions (Fig. 10B). For the mutant, acid-heat-organic extractions were performed at 2 protein concentrations, while only a single heat-organic extraction was performed, due to limiting amounts of purified CpsQ_R134A. These results demonstrated that CpsQ binds c-di-GMP, and the arginine at position 134 is critical for this binding.

Fig 10.

CpsQ is a c-di-GMP-binding protein. (A) Mutation of a residue in the predicted c-di-GMP-binding pocket produces an altered form of CpsQ (R134A) that fails to activate cpsA::lacZ. Exponentially growing E. coli strains carrying the cpsA::lacZ reporter cosmid and pCOLADuet-1 expression plasmids were induced with 25 or 200 μM IPTG to elicit cpsQ expression. Strains were harvested after 4 h to measure β-galactosidase activity. Strains with the indicated expression plasmid were as follows: LLM3816 (vector), LLM4101 [cpsQ (R134A)], and LLM4102 (cpsQ⁺). The amount of β-galactosidase in the strain with the His-tagged wild-type CpsQ was significantly different from that in strains with the vector or His-tagged CpsQ (R134A) at both IPTG concentrations (P < 1E−05). (B) Quantification of c-di-GMP released from purified, His-tagged wild-type and mutant CpsQ. Purified CpsQ protein was extracted by two different methods (heat-acid-organic and heat-organic), and the supernatants were examined by liquid chromatography-tandem mass spectrometry. Each bar represents the average of 2 independent extractions (error bars indicate standard deviations), with the exception of heat-organic extraction for Q (R134A); due to limiting amounts of protein, only one extraction of this type was performed (thus, there is no error bar). Two batches of wild-type CpsQ were purified and extracted (WT Q-1 and WT Q-2 from LLM3829 and LLM4095, respectively) and one of the mutant (Q R134A) from LLM4093. Statistical differences between the wild type and mutant were significant (P < 0.001, for comparison of the combined WT Q-1 or Q-2 extractions with the combined mutant extractions).

DISCUSSION

V. parahaemolyticus is an organism that is well adapted for growth on surfaces. At times, it moves rapidly over surfaces by using swarming motility, while under other circumstances it forms robust biofilms that can be manifested, for example, as thick pellicles at air-liquid interfaces (7, 25). This bacterium senses its physical environment by using its polar flagellum as a tactile sensor and invokes a special program of gene control when growing on a surface (14); however, it possesses additional sensory systems that help inform the organism of appropriate decisions pertinent to the lifestyle choice between surface mobility and sticking. One such system is the Scr system, encoding swarming and cell surface regulators, that modulate the level of the second messenger c-di-GMP. In this work, summarized in the network diagram in Fig. 11, we investigated two previously unexplored aspects of the Scr circuit: the scope of its output targets and the molecular mechanism by which alteration of the cellular c-di-GMP concentration elicits changes in gene expression.

Fig 11.

Model for the Scr network controlling swarming and sticking. ScrC modulates the level of c-di-GMP, and its activity is controlled by the influence of ScrA and ScrB. ScrA produces a cell-cell signaling molecule, the S-signal, which upon accumulation elicits ScrB-dependent stimulation of the phosphodiesterase activity of ScrC. Other sensors, such as ScrG, can also feed into this circuit to impact c-di-GMP. A high level of c-di-GMP promotes production of capsular polysaccharide production and biofilm formation, while a low concentration of this signaling molecule favors swarming motility. New output targets identified in this work are listed in Table S2 of the supplemental material. They include c-di-GMP-repressible members of the surface-sensing regulon, including genes encoding a protease, GpbA (a cell surface adhesin predicted to bind chitin or N-acetylglucosamine), and the type III secretion system on chromosome 1 (T3SS1). They also include c-di-GMP-inducible genes, such as the mfp operon, which encodes a type I membrane fusion transport system and its putative secreted calcium-binding substrate, that plays a role in biofilm development. Although how c-di-GMP represses the surface-sensing regulon remains to be elucidated (indicated by the question mark), elements of the positive regulatory circuit governing expression of cps and mfp expression are defined in this work. The circuit involves three transcriptional regulators. CpsQ is the direct regulator of expression of cps biosynthetic genes; it also regulates its own expression. CpsQ binds c-di-GMP. This second messenger influences the activity and stability of CpsQ. CpsR positively regulates cpsQ, and CpsS prevents expression of cpsR, although it is not known whether it works directly (indicated by dashed line).

Mutations in the scrABC operon were found previously to diminish swarming motility and reduce expression of the hook operon (5). By comparing the transcriptome of the ΔscrABC strain with that of its wild-type parent, we found that the entire lateral flagellar gene system is repressed under high c-di-GMP conditions, including expression of the gene encoding the σ⁵⁴-dependent regulator LafK. This is quite unlike the mechanism inhibiting motility in E. coli, which is the consequence of c-di-GMP interfering with the flagellar switch complex function (3, 9, 29). Furthermore, the expression profiling expanded the known scope of the c-di-GMP-repressed genes beyond the flagellar system. The majority of these (90%) are members of the surface-sensing regulon; they include genes that are induced upon contact with surfaces, and in addition to the laf system, some encode type III secretion proteins comprising the T3SS1 system, a noncoding RNA (∼225 nucleotides), and other potential virulence factors, such as a collagenase and the N-acetylglucosamine/chitin-binding protein GbpA (14).

In general, the gene expression profiles in V. parahaemolyticus conform to the consistent themes found for many bacteria. Namely, genes repressed by c-di-GMP are pertinent to motility and virulence, while gene sets that are upregulated in response to elevated c-di-GMP are relevant to the cell surface (reviewed in references 16 and 32). However, for the Scr circuit c-di-GMP signaling seems more tuned to dictating decisions affecting surface colonization strategies and specificities rather than a planktonic versus biofilm choice. Swimming motility, which is powered by the distinct polar flagellum, is largely unaffected by c-di-GMP (21). Regulation of VPA1598 encoding the N-acetylglucosamine-binding protein seems an illuminating example, because it is both a predicted adhesin and virulence factor. Its homolog GbpA in V. cholerae is a secreted attachment factor for chitin and intestinal surfaces (22). Thus, in this case, low c-di-GMP seems to dictate a preference for adherence to very particular kinds of substrate, namely, the surfaces of shellfish or the host intestine.

The expression of VPA1598 is also interesting because the ∼3-fold regulation observed in the transcriptome comparison of wild-type versus the scr mutant is similar in magnitude to that observed for other members of the surface-sensing regulon; however, when gene expression was assessed using a lux reporter, the observed degree of c-di-GMP-mediated repression seemed anomalously high. Elevated c-di-GMP in the scr mutant switched off VPA1598::lux by ∼2,000-fold, compared to the 5- to 33-fold diminishment of luminescence observed for other members of the surface-sensing regulon (Fig. 1). This discrepancy seems consistent with the observation that the coding region for VPA1598 is preceded by a predicted GEMM c-di-GMP-binding riboswitch (38). Our results imply that VPA1598 could be sensitive to dual layers of c-di-GMP control: one mediated transcriptionally as part of the repertoire of genes induced upon contact with surfaces and another modulated posttranscriptionally via a riboswitch. We note that VPA1598 is strongly induced during growth on surfaces (∼100-fold), and this results in an extremely high level of gene expression (14), so it may be an important strategy for the cell to have a second layer of control to immediately and profoundly increase or dampen production of the gene product when the cell makes a decision to change its lifestyle.

A few genes encoding transcriptional regulators were differentially regulated in the transcriptome comparison. Two, VP2710 and VPA1446 (CpsQ), belong to the CsgD family of DNA-binding proteins. Intriguingly, CsgD family regulators have been implicated in biofilm formation and c-di-GMP-mediated regulation in other bacteria; however, they have been found to do so in very different ways. Members of this family contain a C-terminal helix-turn-helix DNA-binding domain, and some possess N-terminal potential response regulator domains of the FixJ/NarL family. In Salmonella enterica serovar Typhimurium CsgD is a key regulator of biofilm formation, affecting expression of many genes, including those encoding the curli fimbriae and the diguanylate cyclase AdrA involved in cellulose production. Although CsgD has a major role in c-di-GMP signaling (20, 35), it seems a central target in the network rather than being an effector directly responsive to c-di-GMP. CsgD does not bind c-di-GMP; its activity and stability are controlled by its phosphorylation state (42). Most interestingly, it is the unphosphorylated form of CsgD that binds DNA and regulates genes pertinent to biofilm development. In comparison, the putative phosphorylation site of V. cholerae's VpsT is not important; rather, c-di-GMP binds to and influences the oligomerization and activity of VpsT (23).

V. parahaemolyticus CpsQ is a c-di-GMP-responsive protein and hence behaves in a manner similar to VpsT. It is a positive regulator of gene expression, including the operons encoding capsule and a secreted calcium-binding protein. Its transcriptional activating ability can be reconstituted in E. coli in a c-di-GMP-dependent manner. In the absence of c-di-GMP the protein is not stable. Such a property is consistent with the observation that c-di-GMP promotes oligomerization of VpsT (23); it suggests that the monomer is unstable in vivo. Interestingly, this instability of CpsQ in the absence of the nucleotide signal seems also reminiscent of the behavior of CsgD, although in the latter case it is the phosphorylation state that affects protein stability and perhaps multimerization. This property is observed generally for many members of the LuxR superfamily of DNA-binding proteins (cd06170): modification of the C-terminal domain—by small-molecule binding (e.g., autoinducer binding to LuxR orthologs) or by phosphorylation (such as of NarL) —affects multimerization and DNA-binding activity of the N-terminal domain (24, 43). Using radiolabeled nucleotide and methods similar to others (19, 30), we have tried to detect labeled c-di-GMP binding to CpsQ without success, even though positive and specific binding was observed using the control proteins, e.g., the GGDEF-EAL domain-containing ScrC and the PilZ-binding protein MrkH of Klebsiella pneumoniae (data not shown). We interpret this negative result to be in keeping with the predicted nature of the protein. With stability being conferred by c-di-GMP, its binding to the signaling molecule may be so tight that it cannot be competitively dissociated. A similar case occurs with TraR, a LuxR family member that requires its small-molecule inducer for proper folding, stability, and dimerization. Dissociation of the autoinducer from TraR is quite difficult (43).

Nevertheless, the c-di-GMP-binding pocket defined in VpsT (23) is strongly conserved in CpsQ (see Fig. S4 in the supplemental material for the sequence alignment). To directly examine second messenger binding, purified CpsQ was extracted by acid and/or a heat-organic treatment and analyzed by liquid chromatography-tandem mass spectrometry. Two different batches of wild-type CpsQ that were grown in slightly different ways with respect to temperature and time of induction were purified. In addition, a mutant form of CpsQ with an altered amino acid in the predicted c-di-GMP-binding site (R134A) was purified and extracted. CpsQ_R134A failed to release appreciable c-di-GMP. The molar yield of c-di-GMP was proportional to the amount of wild-type protein extracted and was ∼0.2 moles of c-di-GMP per mole of dimer, which we believe seems a reasonable amount given that we do not know the efficiency of ligand loading during overexpression or the efficiency of extraction. The strain producing CpsQ_R134A was no longer able to activate cpsA expression. Even though the mutant form of the protein was detectable by Western analysis, it was greatly overproduced when we used the T7 expression system, and much of the protein was insoluble (as was also true for the wild type). Thus, these experiments do not allow us to fully discriminate whether the failure to activate was due to an inability of the apo-protein to promote transcription and/or the inherent instability of the ligand-unbound form of cytoplasmic CpsQ_R134A. Nevertheless, we conclude that R134 is critical for c-di-GMP binding and that loss of binding results in loss of transcriptional activation.

Intriguingly, V. parahaemolyticus possesses four CsgD homologs: VPA1446 (CpsQ), VPA1447 (CpsS), VP2710, and VPA0358. Of these, CpsS is most similar to VpsT (Vibrio cholerae), whereas CpsQ, VPA0358, and VP2710 are more similar to each other (E values and a multiple sequence alignment are provided in Fig. S3 of the supplemental material). All of the V. parahaemolyticus homologs have a potential c-di-GMP binding motif (being more like VpsT than CsgD), although there is some degeneracy among the sequences. A multiple sequence alignment and phylogram are shown in Fig. S4. Whether these proteins all are c-di-GMP responsive will be most interesting to explore, as will their specific regulatory targets. Their potential response regulator domains are each very poorly conserved, making it unlikely that phosphorylation is a modulator of activity. What seems most curious is that CpsS acts oppositely from CpsQ and VpsT in the sense that cpsS mutations result in elevated (but CpsR-dependent) capsule gene expression (15). Presently, we have no evidence of an involvement of c-di-GMP with CpsS. Although the genetic experiments presented here indicate that CpsS-mediated repression of capsule production works through repression of cpsQ, the mechanism could be direct or indirect. Second messenger or phosphorylation-dependent dimerization may play a role with this regulator as well. Moreover, given the high degree of homology between CpsQ and CpsS and their juxtaposition in the chromosome, it may be a much more complicated scenario of regulation. Perhaps mixed oligomers form under certain conditions.

Unraveling the relationships of and roles for the four homologs in the VpsT/CsgD family will be a task of future work, as will elucidation of the molecular wiring connecting c-di-GMP and the regulation of the surface-sensing regulon. The present work shows that the Scr system influences the expression of ∼100 genes that are pertinent to moving over and colonizing surfaces. We found that c-di-GMP participates in modulating activity within a complex web of capsular polysaccharide regulators. Although only the role of scrABC was examined, we know that other Scr sensors, such as the PAS domain-coupled phosphodiesterase ScrG (21), can feed into the c-di-GMP pool that affects surface colonization. In fact, we suspect a large network of Scr sensors integrate diverse information to allow this organism to make the most appropriate choice between swarming and sticking, based on the level of the second messenger. Discovering the nature of these input signals and sensors will be one of the next challenges.