Vibrio parahaemolyticus senses surfaces via impeded rotation of its polar flagellum. We have exploited this surface-sensing mechanism to trick the organism into thinking it is on a surface when it is growing in liquid. This facilitated studies of global gene expression in a way that avoided many of the complications of surface-to-liquid comparisons, and illuminated ∼ 70 genes that respond to surface sensing per se. Almost all are surface-induced (not repressed) and encode swarming motility proteins, virulence factors or sensory enzymes involved with chemoreception and c-di-GMP signalling. Follow-up studies were performed to place the surface-responsive genes in a regulatory hierarchy. Mapping the hierarchy revealed two surprises about LafK, a transcriptional activator that until now has been considered to be the master regulator for the lateral flagellar system. First, LafK controls a more diverse set of genes than previously appreciated. Second, some laf genes are not under LafK control, which means LafK is not the master regulator after all. Additional experiments motivated by the transcriptome analyses revealed that growth on a surface lowers c-di-GMP levels and enhances cytotoxicity. Thus, we demonstrate that V. parahaemolyticus can invoke a programme of gene control upon encountering a surface and the specific identities of the surface-responsive genes are pertinent to colonization and pathogenesis.
How do bacteria respond to growth on surfaces? The lifestyle of bacteria growing in liquid clearly seems quite distinct from their life growing in biofilms or on surfaces. This has been reflected in comparisons of global gene activity or protein content between planktonic and biofilm bacteria (reviewed in An and Parsek, 2007) and swimming and swarming bacteria. Swarming is one particular type of adaptation to growth on a surface that results in differentiation to a specialized cell type that is able to move rapidly over surfaces and through viscous environments.
Comparisons of cells grown in liquid and on surfaces have revealed profound differences between swimmers and swarmers. For example, almost one-third of the Salmonella enterica serovar Typhimurium transcriptome was observed to be differentially regulated on analysing a comprehensive time series of liquid- and surface-grown bacteria (Wang et al., 2004). Moreover, a proteomic analysis of S. enterica swimmer and swarmer cells revealed changes in the abundance of more than 130 proteins (Kim and Surette, 2004). In Pseudomonas aeruginosa 417 genes, representing ∼ 7.5% of the genome, showed greater than twofold changes in expression comparing swimmers and swarmers (Overhage et al., 2008). Similarly, in Proteus mirabilis 587 genes (∼ 14% of the genome) were differentially expressed > 2-fold comparing broth-grown and swarming bacteria (Pearson et al., 2010). In total, these studies present a striking picture of how large the differences between life in liquid and life on a surface can be. They seem important because many of the observed differences have implications for virulence as well as environmental survival strategies.
However, there are many parameters impacting growth in liquid compared with growth on a surface. These include diffusion and availability of oxygen and nutrients and accumulation of toxic by-products of growth. Concentrations of cell–cell signalling molecules must differ for cells packed in a community compared with cells growing in liquid culture – as must also factors such as cell–cell contact, mobility and heterogeneity of the population. These considerations make it impossible to discern from the studies cited above which changes in gene expression reflect metabolic influences and which are the result of some intrinsic response to growth on surfaces.
We examine this issue using the marine bacterium and pathogen Vibrio parahaemolyticus. V. parahaemolyticus swarming is robust, occurring on 0.5–2.5% agar. The swarmer cells are highly flagellated and can be quite long (Belas et al., 1986). Key for these studies is that surface sensing per se is known to elicit induction of the swarming motility system. The organism possesses two, distinct flagellar systems. The polar system (Fla) is dedicated to swimming, whereas lateral system (Laf) is reserved for swarming (McCarter, 2004). There are no shared structural components. The single, sheathed polar organelle is produced continuously, and hundreds of lateral flagella are produced only during growth on surfaces and in viscous environments (reviewed in McCarter, 2001; 2004; 2005). Thus, with respect to specificity and fold regulation of gene expression, the laf system of V. parahaemolyticus is an ideal marker of surface adaptation and differentiation: the swimmer makes no lateral flagella, while the swarmer is elongated and makes numerous lateral organelles (illustrated in Fig. 1).
Surface sensing in V. parahaemolyticus occurs via the polar flagellum: conditions impeding movement of the swimming organelle inform the bacterium of its physical environment. These conditions can include growth on surfaces or in highly viscous liquid media, mutation of the polar flagellar genes, and the sodium channel blocking drug phenamil, which poisons the energy source driving polar (but not lateral) flagellar rotation (Belas et al., 1986; McCarter et al., 1988; Kawagishi et al., 1996). All polar flagellar mutants that have been isolated with defects affecting swimming motility cause expression of lateral flagellar genes in liquid, including mutants that produce no polar flagellar components and mutants that produce a paralysed polar flagellum (reviewed in McCarter, 2004). Thus, when the organism cannot swim, it induces a new mode of translocation. However, interference with the polar organelle is not sufficient to evoke the swarmer cell programme. A second signal is required, which is iron limitation (McCarter and Silverman, 1989). We suspect that imposition of dual requirements for such a costly differentiation event ensures that differentiation to the surface mobile cell type occurs under strictly appropriate conditions, i.e. when the bacterium finds itself in an environment where it cannot move (i.e. swim) and cannot grow well (i.e. partial starvation conditions).
In this work, we define a core set of genes whose expression is specifically responsive to the surface-adapted lifestyle. By using polar flagellar mutants grown in iron-deficient broth and by using iron-poor medium supplemented with the drug that specifically prevents rotation of the polar flagellar motor, we use ‘pseudosurface’ conditions to elicit swarming gene expression in liquid and analyse the global surface-responsive transcriptome in a new way. Coupling the transcriptome studies with a combination of mutant analyses, reverse transcription polymerase chain reaction (RT-PCR) and reporter gene fusions allowed us to establish the initial framework for a circuitry diagram of the surface-responsive network of gene control. The genes in this network provide information about general surface colonization strategies. Our data imply that surface sensing in V. parahaemolyticus not only induces the swarming motility system but also invokes new capacities for sensing and responding to environmental cues and for preparing for encounters with a potential host.
Overview of microarray growth conditions
In order to dissect the complex transcriptional differences occurring during growth in liquid and on surfaces, RNA profiles were captured using multiple conditions, wild-type and mutant strains, and by exploiting our understanding of the induction of the laf gene system. All strains were grown in heart infusion (HI) medium. Our benchmark swarming strain LM5674 was grown on a surface (S), which was HI medium solidified with 1.5% agar, in liquid (L), which was HI broth with no addition, and in iron-limiting liquid (LD), which was HI broth with the iron chelator 2,2′-dipyridyl. A pseudosurface condition (PS1) was achieved by growing two different polar flagellar mutants in iron-limiting broth. An alternate pseudosurface condition (PS2) was achieved by growing the wild-type strain in iron-limiting broth amended with phenamil, a specific inhibitor of the sodium-driven polar flagellar motor (Yorimitsu and Homma, 2001). The swarming-defective strain LM7789, which has a lesion in the regulatory gene encoding the σ54-dependent transcription factor LafK (Stewart and McCarter, 2003), was also grown on a surface [S(K-)]. Table 1 summarizes the microarray experiments performed for this analysis. Six conditions were used to query the chips; each of these conditions was repeated two or more times (as indicated in Table 1) with independently isolated RNA samples for each replicate.
Table 1. Microarray strains, growth conditions and cell length data.a
Cell length data were calculated by scoring at least 100 cells per sample at time of harvest for each RNA preparation and averaging the samples. The number of cells longer than 5 µm was also calculated and expressed as a percentage of the total. Standard errors of the mean or standard errors of the average percentage are given in parentheses.
Number of chips probed is also equal to number of independent RNA preparation.
Wild type (LM5674)
Wild type (LM5674)
Wild type (LM5674)
Iron-limiting broth (liquid + iron chelator Dipyridyl)
flaM::EzKAN (LM5392) motA2::TnphoA (LM4652)
Pseudosurface 1: polar flagellar mutants grown in iron-limiting broth
Wild type (LM5674)
Pseudosurface 2: wild type grown in iron-limiting broth with polar flagellar motor inhibitor phenamil
Swarm-defective lafK regulatory mutant grown on a surface
Establishing surface growth conditions for RNA harvesting
Luminous reporter strains with fusions in lateral flagellar genes (laf::lux fusions) have been used to study laf gene expression (Stewart and McCarter, 2003), and such a strain was used as the guide to establish the timing for the microarray experiments. In a typical experiment, the laf::lux reporter strain was spread on plates and harvested periodically during growth for OD600 and luminescence measurements (Fig. 2A). Strong induction of laf gene expression was detected at 6 h, i.e. luminescence increased from a baseline of ∼ 20 to ∼ 20 000 SLU (light units normalized to OD600). This time was selected as the appropriate point for capturing the swarming RNA profile. It coincided with the maximal numbers of long cells and maximal average cell length for the wild-type strain (Table 1). The amount of lateral flagellin (Laf), the structural subunit for the lateral flagellar propeller, produced by the surface-grown wild type (S) harvested at the 6 h time point is shown in the immunoblot in Fig. 2D. The lafK regulatory mutant strain LM7789, which cannot swarm, was grown and harvested similarly for the RNA purification. The flagellin profile for this mutant grown on a surface and harvested at the 6 h time point [S(K-)] is also shown in Fig. 2D: it does not synthesize lateral flagella, although it does produce polar flagellin (Fla).
Establishing pseudosurface growth conditions for RNA harvesting
Lateral flagellar gene expression was induced in liquid by using ‘pseudosurface’ conditions, achieved by impairing polar flagellar function – either by using mutant strains with structural defects in the polar flagellum (PS1) or by using the sodium channel blocking drug phenamil to inhibit rotation of the sodium-driven, polar flagellum (PS2) (McCarter et al., 1988; Kawagishi et al., 1996). Figure 2C shows a typical profile of laf induction in liquid medium containing phenamil supplemented with the highest concentration of the iron chelator sufficient to induce laf::lux gene expression without having a significant deleterious impact on growth rate (Fig. 2B). The data also demonstrate that iron restriction per se does not induce laf::lux expression. RNA was prepared from wild-type cultures grown similarly and harvested at the 200 min time point, which corresponded to the peak of laf::lux expression. The amount of lateral flagellin produced by the wild-type strain at the time of harvest is shown in Fig. 2D: lateral flagella are produced only in the pseudosurface condition (PS2 = LD plus phenamil) and not in liquid (L) or liquid plus dipyridyl (LD).
The alternative pseudosurface condition (PS1) was achieved by using polar flagellar mutant strains grown in iron-limiting broth. The transcriptional profiles of two polar flagellar mutant strains were examined in order to avoid potential allele-specific effects on the surface-sensing response. [Although it turned out that there were no appreciable differences in transcription profiles between the 2 strains.] LM5392 (flaM1) contains a defect in a polar flagellar regulatory gene; it cannot swim and fails to synthesize the polar flagellum (Fla-). LM4262 (motX1699) contains a defect in a polar flagellar motor gene; it is also completely unable to swim but produces a paralysed polar flagellum. The amount of lateral flagellin produced by the Fla- strain LM5392 grown in LD and harvested for the microarray RNA sample is shown in Fig. 2D (PS1).
Cell length data (Table 1) generally conformed to the gene expression and flagellin measurements. For example, approximately 14% of the PS1 cells and 19% of the PS2 cells were greater than 5 µm in length compared with 4–5% of the L or LD population. Although neither pseudosurface condition induced elongation as effectively as growth on a surface (∼ 46% long), we believe this is partly a consequence of not fully limiting for iron (a compromise adopted in order to match growth optimally) because increasing dipyridyl conditions resulted in higher fold induction of laf::lux reporters and greater cell elongation (data not shown). However, iron is not the only relevant factor promoting elongation as the swarming-defective lafK mutant failed to elongate as effectively as the wild-type strain when grown on a surface (∼ 15% versus 46% long respectively).
Defining the LafK regulon
To begin to analyse global gene expression during growth on a surface, we examined the LafK regulon. The comparison of RNA profiles between surface-grown lafK+ and lafK- strains was predicted to display differences in known laf genes and permit a validation test for our newly developed microarrays, with respect to both chip design and assessment of the method to capture swarmer cells. Prior studies identified 38 laf genes and established a transcriptional hierarchy of gene control by using reporter transposon (Tn5lux) mutagenesis (Stewart and McCarter, 2003). This cascade of gene control is summarized in Fig. 1C. LafK was found to be a key σ54-dependent transcriptional regulator, directing the expression of Class 2 operons encoding many of the components of the hook/basal body structure of the flagellum as well as a gene encoding a specialized σ factor FliAL (σ28) that is in turn required to direct transcription of Class 3 laf genes, including lafA (VPA1548), the structural gene encoding the flagellin subunits that comprise the flagellar propeller. In the 2003 study, one flagellar operon, VPA1540-46 encoding C-ring components (FliM and FliN) and cytoplasmic membrane export components (FliP, FliQ, FliR, FlhA and FlhB), was not placed in the hierarchy because no lux fusions were isolated in this operon. LafK is homologous to the polar flagellar regulator FlaK, which is orthologous to the master polar flagellar regulators in other members of the Vibrionaceae as well as Pseudomonas species, e.g. FlrA in Vibrio cholerae and Vibrio fischeri and FleQ in P. aeruginosa (McCarter, 2001; Kim and McCarter, 2004).
The microarray analysis showed decreased transcription of 44 genes (> 4-fold, P-value < 0.05) in the lafK strain compared with the wild type (Table 2 and Table S1). No genes appeared to be negatively regulated by LafK (using the fourfold cut-off). Many (68%) but not all of the genes identified by this comparison were the previously defined laf genes. Moreover, the observed fold differences in gene expression between the two strains were large, e.g. the change in transcription between lafK+ versus lafK- strains for lafA (VPA1548) was 984-fold. Thus, the data not only reveal that we have developed growth conditions to successfully capture the transcriptome of the swarmer cell, but also that this system has the capacity to report a robust dynamic range of gene expression.
Table 2. The LafK regulon.a
Three new potential laf genes were revealed by the transcriptome comparison. These genes were not annotated as flagellar genes; however, they each occur as the distal gene in a laf operon confirmed by RT-PCR (data not shown). Transposon mutants were isolated with insertions in VPA0275 (encoding a hypothetical protein) and VPA0260 (encoding a periplasmic lytic transglycosylase; pfam01464; E-value = 5.9e-29). They have severe swarming defects, comparable to a mutant with a lesion in the hook gene (Fig. 3A). The phenotype of a mutant with a lesion in VPA1558 (also encoding a hypothetical) is not yet known. LafK regulated one other new, motility-related gene, VPA1492, encoding a predicted, methyl-accepting chemotaxis protein (Mcp, 13-fold), and its mutant phenotype also is not known (NB: V. parahaemolyticus possesses 29 genes encoding Mcps).
The scope of LafK regulation extends beyond flagellar genes
Microarray comparisons of lafK+ and lafK- strains led to the unexpected discovery that LafK regulates several genes that are not likely to be directly involved in assembly or function of the lateral flagella. Some of these genes showed very high fold changes in gene expression. For example, VPA1295 and VPA1294, which form a potential operon, showed > 170-fold expression differences. VPA1295 encodes a small hypothetical protein (90 aa) with a predicted signal peptide. VPA1294 encodes a predicted 134 aa protein containing a SPOR domain (pfam05036; E-value = 2.2e-12). SPOR domains seem involved in peptidoglycan binding and have been found in proteins affecting sporulation and cell division. Other highly regulated genes include VP0763 (40-fold), encoding a very small predicted protein (39 aa), and an intergenic region between VP1293-1294 (80-fold), which appears to encode a small RNA whose size is at least 420 bp, the size of a product detected by RT-PCR. Another potential operon showing LafK dependence included VPA1649 and NTO1VPA1551 (> 20-fold). The deduced protein VPA1649 is a predicted metalloendoprotease; the downstream and overlapping coding region, NTO1VPA1551, encodes a predicted product (147 aa) containing the CcmA conserved domain, which is involved in cell shape determination (COG1664; E-value = 7e-17).
To explore the regulation of some of these genes, new RNA preparations were isolated from the wild type grown in liquid and the following surface-grown strains: wild type, lafK and fliAL. Figure 4 shows the products of RT-PCR reactions that were designed to amplify the new potential LafK-dependent target mRNAs, specifically those encoding the small RNA, the SPOR protein, the metalloendoprotease and the Mcp. Primers for the polar flagellar fliAP, whose expression is constitutive, were used to produce the normalization control in each reaction. Each of the queried genes showed LafK dependence, i.e. a product was found in the wild-type sample whereas little or no product was observed for the lafK mutant RNA (and in liquid-grown wild type). Furthermore, these genes were not (or poorly) transcribed in a strain with a defect in fliAL. We examined nucleotide sequences upstream of the newly identified members of the LafK regulon for potential σ28-dependent promoters using a consensus sequence derived previously on the basis of primer extension (Stewart and McCarter, 2003). In agreement with the RT-PCR results, potential signature σ28-dependent promoter sequences could be detected for these LafK/σ28-dependent genes (Table S2). A signature promoter was also found for VP0763, which showed 40-fold LafK-dependent regulation by the microarray analysis, but was not examined by RT-PCR. Thus, the transcription of these genes can be placed within Class 3 of the laf hierarchy of gene control, as they required both LafK and the laf-specific sigma factor for transcription. The RT-PCR profile of the Class 2 gene (LafK-dependent but σ28-independent) fliDL, encoding the flagellar capping protein, is shown for comparison (Fig. 4).
Four additional genes showed approximately fivefold differential expression: they encode type III secretion proteins (VP1682, VP1701), a hypothetical protein (VP0649) and an alkaline serine protease (VPA0227). We note that the evidence for their differential expression is suggestive, but weak in light of genome-wide testing, as their false discovery rates (Table S1) are higher than the other genes included in Table 2. LafK dependence for vpa0227 was examined directly by introducing a lafK deletion into a strain carrying a luminescence reporter in VPA0227. Luminescence was greatly reduced in strain LM9620 (vpa0227::luxΔlafK) compared with LM9376 (vpa0227::lux) throughout a time-course of growth on plates (Fig. 3B). Strain LM9376 showed no defect in swarming (Fig. 3A).
LafK is not at the apex of the lateral flagellar hierarchy of gene transcription – nor is the fliML operon
The transcriptome comparisons displayed LafK dependence for transcription of all known laf genes except for the fliML operon (VPA1540-46, fliMNPQflhABL). Thus while homologous to the master regulators of many flagellar systems (McCarter, 2001; Kim and McCarter, 2004), LafK is not a master lateral regulator in the sense that it is not required for all laf gene expression. Although the result was surprising, it was consistent with prior primer extension mapping of the start site of VPA1540 (fliML) transcription, which failed to illuminate a predicted σ54-dependent promoter for fliML. Transcriptional fusions in lafK and flgCL (a gene in the basal body/hook operon) were used to examine the relationship between the fliML operon and the flagellar hierarchy. Reporter plasmids were moved into the wild type or a host of mutant strains, including lafK, rpoN, and two mutants with defects in the fliML operon (flhALor fliQL mutations). [All of these mutants have similarly severe defects in swarming motility (Stewart and McCarter, 2003).] These strains were grown on plates and harvested periodically for β-galactosidase measurements. Activity is reported as the maximal activity during growth and is normalized to the wild-type strain. Unlike lafK and rpoN mutants, which failed to express flgC::lacZ, the fliQ and flhA mutants modulated but did not abrogate flgC::lacZ expression, reducing activity to ∼ 30–50% of the wild-type level (Fig. 5B). The fliM operon mutations also had only small effects on lafK::lacZ expression, reducing the β-galactosidase level ∼ 65–80% of the wild-type level (Fig. 5A). In addition, the data demonstrate that feedback regulation occurs to modulate lafK expression because lafK and rpoN mutants showed lower lafK::lacZ activity than the wild type; therefore, the lafK operon belongs to both Class 1 and Class 2 (in Fig. 1). Mutational interference within the fliML operon does not create a full functional block – or checkpoint – in the hierarchy of gene expression to preclude expression of lafK and downstream genes. These results (and data presented in Fig. 4 and Table 3) suggest that the fliML operon, like the lafK operon, belongs to Class 1. Together these two operons encode the core components of the flagellar export apparatus.
Table 3. Surface-induced genes.
Growth on surfaces versus growth in liquid
To capture the full spectrum of differences in gene expression between swimming and swarming cells, we compared the profiles obtained for exponentially growing swimming and swarming cells, i.e. cells grown in liquid (L) and on an agar surface (S). This comparison produced 338 genes whose expression was increased and 361 genes whose expression was decreased by growth on surfaces (> 2-fold and P-value < 0.05). This represents ∼ 13.5% of the genome. Of these, 235 were regulated > 4-fold (145 up and 90 down, listed in Table S3). Among the induced genes were all members of the LafK regulon (in Table 2). The fold inductions observed in the S versus L comparison were generally comparable to the fold regulation found in the lafK+ versus lafK- comparison, and these values are also given for comparison in Table 2. For example, VPA1548 (encoding the flagellin subunit) was induced 641-fold when grown on a surface compared with liquid, similar to the 984-fold induction observed for the lafK+ versus lafK- comparison. Moreover, the entire fliML operon, which was not regulated by LafK, was induced in the S versus L comparison; e.g. fliML (VPA1540), the first gene in this operon, showed 286-fold induction upon growth on a surface. Data for all of the genes in this operon are reported in Table 3. The RT-PCR verification for the expression profile of fliML is shown in Fig. 4: very little mRNA was detected for the wild type grown in liquid compared with the amount of mRNA produced for all surface-grown samples (wild type, lafK, fliAL strains).
Pseudosurface and surface comparisons allow the construction of a core set of surface-responsive genes
Comparison of growth in liquid and pseudosurface conditions eliminated many of the physiological variables occurring between L and S environments, and permitted elucidation of the gene expression changes unique to surface sensing, i.e. the polar flagellum-mediated adaptation to the surface. The degrees of laf expression induced by the PS conditions compared with the L condition were similar (generally within threefold) although not quite fully achieving the degree of laf induction observed in cells grown on plates (S). For example, the flagellin gene VPA1548 was induced 346-fold by the PS1 condition and 398-fold by the PS2 condition, compared with 641-fold for surface-grown conditions (Table 3).
Six surface-to-liquid comparisons (i.e. S versus L, S versus LD, PS1 versus L, PS1 versus LD, PS2 versus L and PS2 versus LD) were made as well as the iron-limiting liquid versus liquid control comparison (LD versus L). The fold cut-off for these comparisons was twofold and the P-values were < 0.05. We then identified a core set of surface-regulated genes that behaved similarly in these comparisons (61 induced and 5 repressed genes). The core genes and their fold regulation are listed in Tables 3 and 4, induced and repressed respectively. (Venn diagrams are provided for an overview of the commonly regulated genes in Fig. S1.) The core genes include but are not limited to swarming motility genes (and their identities are described below). Many of these genes showed very large changes in expression upon growth on a surface compared with liquid (S versus L). Moreover, there was coherence in the gene sets, i.e. for the most part if a gene was part of an operon, all of the genes in that operon showed a similar pattern of regulation. Figure 6 plots the representative expression profiles in a way that reveals the relative abundance of message and the magnitude fold change in expression. Three general profiles of gene expression were observed: LafK-dependent surface-induced, LafK-independent surface-induced, and LafK-independent surface-repressed.
Table 4. Surface-repressed genes.
In addition, Tables 3 and 4 include some genes (∼ 20 in number, designated as Tier 2) that marginally missed being included by using the core criteria, i.e. genes that showed consistent regulation in four of six comparisons. Some of the Tier 2 genes were placed in the core if the surface regulation profile was verified by an independent means (as indicated below). We emphasize that the core represents a minimum set – a robust set of genes derived from six comparisons and mostly showing > 4-fold changes in gene expression when grown on a surface. Certainly some of the Tier 2 and additional genes at the periphery of our stringency criteria may also belong to the core set of surface-responsive genes and future work must evaluate the significance of their observed regulation. We note that genes regulated specifically by PS1 or PS2 conditions for the most part include polar flagellar or phenamil-specific genes respectively.
The comparisons and cut-offs used to populate Tables 3 and 4 are not independent, i.e. each condition is used twice. As a result, it is not obvious how to define an overall false discovery rate for these tables. Nevertheless, a conservative upper bound may be established by taking the two essentially independent comparisons S versus L and PS1 versus LD and multiplying their false discovery rates. Using this as justification, we estimate that the false discovery rate among Tier 1 genes is below 0.5%, and that the false discovery rate for Tier 2 genes is below 5%. We remark that because these analyses were performed using a heterologous array (strain BB22 has not been sequenced), the core also represents the minimal list because there may be some strain-specific genes to be discovered.
Flagellar and non-flagellar genes are induced by surface sensing
The pseudosurface (PS1 and PS2) and surface (S) conditions induced all of the LafK-dependent genes identified by the lafK+ versus lafK- trancriptome comparison; representative LafK-dependent, flagellar and non-flagellar gene expression profiles are shown Fig. 6A. Genes in the fliML operon (VPA1540-1546) exemplify the second observed pattern of expression. Transcription of this operon was induced by S or PS conditions, but was not regulated by LafK. Some non-flagellar genes were regulated similarly as those in the fliML operon; representative transcription profiles are shown in Fig. 6B, including VP1002 (∼ 39-fold, S versus L), VP2370 (∼ 6-fold, S versus L), VPA1598 (∼ 111-fold, S versus L) and VPA0459 (∼ 127-fold, S versus L). These results were corroborated by RT-PCR (Fig. 4). Similar to fliML, little or no product was produced for VPA0459, VPA1598 or VP2370 using RNA prepared from the wild-type strain grown in liquid, whereas product was found for all of the surface-grown RNA samples.
Strains with reporter Tn5lux insertions in VP1002 and VPA1598 were isolated. They exhibited surface-induced luminescence, as did the LafK-regulated vpa0227::lux reporter strain (Fig. 3C). These strains produced little light when grown in L (< 16 SLU), whereas growth in S induced luminescence (> 2000, SLU, Fig. 3C). The transposon insertions caused no defect in swarming motility (Fig. 3A). VP1002 encodes a large hypothetical lipoprotein (906 aa). VPA1598 encodes a product homologous to the secreted, N-acetyl glucosamine-binding protein A of V. cholerae (VCA0811; 70% identities/82% positives; E-value = 0.0). In addition to having large fold degrees of regulation, the level of expression for some of these genes was particularly high. For example, growth on a surface induced a level of expression of VPA1598 (encoding the GlcNAc-binding protein) almost equivalent to that for a ribosomal protein gene (Fig. 6).
Genes negatively regulated by growth on a surface
Only a few genes showed lower expression profiles in surface or pseudosurface grown cells (Table 4). Transcription of one methyl accepting chemoreceptor gene, VP1892, was decreased ∼ 10-fold by growth on S or PS compared with L. Repression was observed for another set of genes (VP2876 to VP2878) that potentially form an operon involved in nucleotide signalling because they encode proteins with cyclic nucleotide binding and CBS domains (COG2905; E-value = 1.0e-180) and DnaQ-like exonuclease (DEDDh; E-value = 4e-32) domains (Fig. 6C).
Two gene sets pertinent to c-di-GMP signalling were also negatively regulated. VP1482 and VP1483 (potentially forming an operon) were repressed approximately eightfold for all surface conditions (Fig. 6C); this result was confirmed by RT-PCR (Fig. 4). This operon encodes one protein with a response regulator receiver domain (REC; E-value = 5e-18) and another protein with a PAS_4 domain (pfam08448; E-value = 0.008) and a GGDEF domain (pfam00990; E-value = 1e-43), which is the catalytic domain responsible for c-di-GMP formation. VP1881 was repressed ∼ 11-fold over all surface-to-liquid comparisons; it encodes a predicted membrane-bound c-di-GMP phosphodiesterase, containing a highly conserved EAL domain (pfam00563; E-value = 3e-34).
c-di-GMP levels during growth on a surface
These transcriptome analyses revealed positive and negative expression changes for genes encoding enzymes involved in the synthesis and degradation of the second messenger c-di-GMP. Genes encoding both a diguanylate cyclase and phosphodiesterase were repressed approximately the same fold and to similar levels of expression by growth on a surface (Table 4 and Fig. 6C), while another operon encoding a phosphodiesterase was induced by growth on surfaces and pseudosurfaces (Table 3). This operon, named scrABC (VPA1513-11), was implicated previously in the regulation of swarming. Mutants with defects in the scrABC operon are profoundly defective for swarming motility; the operon is approximately eightfold upregulated by growth on surfaces compared with liquid as measured by using a scrC::lux reporter; and ScrC was demonstrated to affect the cellular c-di-GMP pool (Boles and McCarter, 2002; Ferreira et al., 2008). The level of expression of this operon during surface growth was > 5 log2 higher than the levels observed for either of the surface-repressed c-di-GMP-pertinent genes. Although one cannot deduce c-di-GMP levels directly from the gene expression patterns, particularly since these are sensory-coupled enzymes, the fact that microarray data report a high absolute level of expression of the scrABC operon (Fig. 6B) is consistent with its strong mutant phenotype. Together, the data suggest that c-di-GMP signalling plays a role during growth on a surface.
To examine c-di-GMP, the nucleotide pools of the wild-type strain grown in liquid and on a surface according the microarray conditions were extracted for analysis by using high-performance liquid chromatography-coupled mass spectrometry. The mean observed value of the c-di-GMP concentration for liquid-grown cells (11.8 pmol mg−1 total bacterial protein) was higher than the concentration for surface-grown cells (7.64 pmol mg−1) (Fig. 7). The ΔscrABC strain was grown similarly on a surface. In a prior study, the products of the scrABC operon were demonstrated to decrease the cellular level of c-di-GMP; however, these studies did not examine the level of this nucleotide during growth on a surface. Consistent with the demonstrated role of ScrABC acting as a phosphodiesterase and playing a regulatory role during swarming, the intracellular level of c-di-GMP was higher for the ΔscrABC mutant strain (11.0 pmol mg−1) than its wild-type parent during growth on a surface (Fig. 7).
Linking surface-responsive gene expression and virulence
Some genes induced by surface and pseudosurface conditions encode products that are potential virulence factors, e.g. collagenase, alkaline serine protease, N-acetyl glucosamine-binding protein and metalloendoprotease. Collagenase production was examined for strains grown in liquid by assaying supernatants using a fluorogenic substrate. Specifically, the wild-type strain was grown in L, LD and PS1; the fla mutant in LD (=PS2), and the lafK mutant in LDP (pseudosurface condition). Little collagenase activity was detected in the supernatants for the wild-type strain grown in L or LD, whereas considerable activity was detected for all of the strains grown under PS conditions (Fig. 8A). These results provide evidence that VPA0459 encodes a collagenase that is specifically induced in a LafK-independent manner by growth on a surface.
Another class of virulence genes showing enhanced expression as a consequence of growth on surfaces encodes components of a type III secretion system that is located on chromosome 1 (T3SS1), exemplified by VP1701 in Fig. 6A. These T3SS1 genes showed increased expression as a consequence of growth on S and PS conditions and decreased expression in the lafK mutant. In fact, many linked genes encoding other T3SS components showed a similar trend in gene expression, albeit their fold or P-values fell slightly outside of our criteria (Fig. S2 plots gene expression profiles for many of these).
To examine whether growth on a surface might stimulate virulence, the capacity of V. parahaemolyticus to kill host cells was measured using a cytotoxicity assay. For these experiments, surface and liquid grown cells were used to infect Chinese hamster ovary CCL-61 (CHO) cells and host cell lysis was monitored over time by measuring release of the host enzyme lactate dehydrogenase. This assay has been used as a measure of V. parahaemolyticus T3SS1 activity during co-culture with a variety of mammalian cell lines (Park et al., 2004; Zhou et al., 2009). Pre-growth of the wild-type V. parahaemolyticus on S enhanced the degree of cytotoxicity to host cells compared with prior growth in L as did pre-growth under PS2 conditions, albeit not quite as well as growth on plates (Fig. 8B). The cytotoxicity of a lafK mutant was also assessed because this mutant displayed decreased expression of T3SS1 genes and other potential virulence genes (Table 2). The surface-adapted lafK mutant [S(K-)] showed delayed and reduced cytotoxicity compared with the surface-adapted wild-type strain [S] (Fig. 8C). Thus, we conclude that growth on a surface can elevate the potential virulence of the organism and this is in part due to regulation occurring through LafK.
For bacteria, growth on surfaces results in profound changes in gene expression compared with growth in liquid. Although many of these changes are due to the complex physiology of growth under different circumstances, some may be the result of a programmed response to the surface per se. In this work we have defined a core set of genes in V. parahaemolyticus belonging to a surface-responsive programme of gene regulation. The surface-responsive genes have been distinguished from a larger group of genes whose expression changes as a consequence of growth on a solid medium because their expression also changes when the bacterium is manipulated into sensing that it is on a surface while growing in liquid.
The surface-responsive core is relatively small in number (< 70) and contains many more positively than negatively regulated genes. Almost two-thirds of the induced genes are involved in swarming motility. We have positioned the core genes in a wiring diagram of gene control (summarized in Fig. 9). Transcription of some requires the specialized lafσ28 for expression (Class 3 genes), other genes utilize the σ54-dependent LafK regulatory protein (Class 2 genes), and others are co-regulated with lafK itself (Class 1 genes). Certain of the Class 2 genes, in particular the flagellar genes, strictly require LafK (and σ54) for expression, whereas the transcription of other genes (such as certain virulence genes) is modulated by, but not stringently dependent upon LafK. We emphasize that our core genes comprise a robust minimal set; we have confirmed their membership in various ways. Clearly there are more surface-responsive genes to be discovered, e.g. genes with small fold regulation found at the periphery of our statistical criteria as well as genes unique to our particular strain of V. parahaemolyticus since a heterologous array was probed. Below we discuss some themes that have emerged from our transcriptional profiles and follow-up studies.
As expected from the dimorphic pattern of flagellation in this organism, the majority of the surface-regulated genes are laf genes; however, there were some surprises. A few new potential motility genes were discovered. These encode non-traditional products as they have no known flagellar orthologues, although they were found to be transcribed as part of extended laf operons. Transposon insertions in two of these genes resulted in swarming deficiencies. Two chemotaxis receptor proteins were regulated approximately 10-fold – one induced and one repressed by growth on the surface. The σ54-dependent regulator LafK was disqualified from its putative role as the ‘master’ lateral flagellar regulator by the discovery that some flagellar genes were not regulated by LafK; instead, they were found co-regulated with lafK. Products of the fliML operon and the downstream genes in the lafK operon form the C-ring and export apparatus. Not being strictly under control of a flagellar regulator suggests that perhaps these genes are not dedicated solely to laf protein export.
There were a few genes showing high fold surface induction whose products may be pertinent to the morphological changes accompanying swarmer cell differentiation. For example, VPA1294 (regulated ∼ 500-fold between L and S) encodes a product predicted to be a small outer membrane protein with a peptidoglycan-binding SPOR domain. Perhaps inserting hundreds of flagellar basal bodies into the elongated swarmer cell envelope requires some remodelling or reinforcement of the cell wall. Alternately, VPA1294 could play a more direct role in swarmer cell elongation as multiple proteins with SPOR domains have recently been shown to participate in cell division in bacteria (Dai et al., 1993; Gerding et al., 2009; Moll and Thanbichler, 2009; Arends et al., 2010).
A two-gene operon showed similarly large surface induction (∼ 400-fold induction in S versus L). VPA1649 is predicted to be in the M23 peptidase (or LytM) family, members of which are zinc metalloproteases. It shows conservation of all of the active-site residues determined for the staphylococcal ALE-1 enzyme, a glycylglycine endopeptidase (Fujiwara et al., 2005). VPA1649 might also be involved in cell wall modification during swarming. Support for such an idea also derives from the identity of the product encoded by the downstream gene co-regulated with VPA1649. The second gene product contains the predicted conserved domain CcmA (COG1664) that has been implicated in curved cell morphology in a few organisms. In fact, the genes and their organization in the V. parahaemolyticus operon show striking resemblance to the three-gene shape locus of Helicobacter pylori, which encodes two M23-family metalloendopeptidases and ccmA (Sycuro et al., 2010); interestingly, the coding region of ccmA overlaps with the upstream protease gene in both organisms. H. pylori mutants with defects in this locus form curved, rather than helical rods. P. mirabilis also has a ccmA homologue, although it is not found in an operon with a peptidase and it encodes a larger, membrane-bound protein. Nevertheless, expression of the gene (PMI1153) is upregulated during swarming (sevenfold), the Ccm1 product is more abundant in swarmer cells (∼ 20-fold), and a C-terminal truncation mutation (but not a full deletion) affected cell morphology and swarming behaviour (Hay et al., 1999; Pearson et al., 2010).
Swarmers cells are often quite long (Harshey, 1994); however, the basis for this elongation is not understood in any organism. One interesting finding resulting from our studies was that the lafK mutant failed to elongate fully when grown on a surface. However we note, with some disappointment, that there were no observed changes in expression of genes known to be involved in cell division in any of our microarray comparisons. Taken together, we suspect that elongation may result from small changes in gene expression or be initiated by post-transcriptional mechanisms that could be further amplified by downstream events, e.g. the massive overproduction of the lateral flagella might interfere with assembly of a functional septal ring.
Several of the surface-regulated genes encode enzymes involved in sensing and or transducing signals. Among these are two chemotaxis receptor proteins and three surface-regulated genes encoding enzymes with the capacity to modulate levels of the second messenger c-di-GMP. The latter are multidomain proteins containing signal reception/transduction domains coupled to output enzymatic domains capable of synthesizing (DC activity) or degrading (PDE activity) c-di-GMP. Expression of one MCP increased and another decreased in response to growth on a surface. The task of deciphering their specific roles is complicated by the fact that the organism possesses a plethora of chemoreceptors (29 in number); however, we hypothesize that the fluctuations represent an altered capacity for modulating swarming behaviour in response to new or different cues. In V. parahaemolyticus, chemotaxis is required to co-ordinate swarming motility, although regulation of lateral flagellar gene expression is unaffected by chemotaxis mutations (Sar et al., 1990 and L. McCarter, unpublished).
The direction of regulation of the genes encoding the c-di-GMP-pertinent enzymes was also mixed: expression of genes encoding one PDE and one DC was repressed while that for another PDE was increased by growth on the surface. Moreover, like the panoply of chemoreceptors found in V. parahaemolyticus, the organism possesses at least 59 c-di-GMP-pertinent sensory proteins. We interpret this to indicate that particular enzymes do not solely dictate the concentration of cellular c-di-GMP; rather their influence on the cellular level depends upon their relative abundance and the signals they process. Nevertheless, we were able to analyse cellular c-di-GMP and found that the c-di-GMP pool was lower for cells growing on a surface at the onset of swarming compared with growth in liquid. A lower concentration of c-di-GMP is consistent with the fact that surface sensing induces a very active form of motility, i.e. swarming (Wolfe and Visick, 2008).
Colonization and virulence
The third theme of regulation appears to be one relevant for pathogenesis. VPA1598 encodes a potential key colonization factor. Upon induction, this gene showed remarkably high-level expression in the microarray analyses, i.e. similar to the level of expression for the flagellin gene lafA or rRNA genes. The predicted product of VPA1598 is an N-acetyl glucosamine and chitin-binding protein, and is homologous to V. cholerae GbpA, which promotes adherence to zooplankton as well as human epithelial cells (Kirn et al., 2005).
One large gene set exhibiting enhanced expression under surface and pseudosurface conditions encodes components of a T3SS. These genes are predominantly found in the pathogenicity island located on chromosome 1 (Makino et al., 2003). They were regulated between four- and sixfold in L versus S and L versus PS comparisons. In fact, this trend in expression was observed for most of the predicted T3SS1 genes, albeit the fold regulation was smaller. Furthermore, this group of genes showed diminished expression in the lafK mutant.
Other surface-induced genes that seem pertinent for the pathogenesis of the organism include an alkaline serine protease (VPA0227), a collagenase (VPA0459) and perhaps the aforementioned LytM-family metalloprotease (VPA1649). Two V. parahaemolyticus collagenases (VP1340 and VPA0459) have been characterized previously. VP1340 was believed to encode the major extracellular enzyme; however, the experiments were performed using broth-cultivated bacteria (Miyoshi et al., 2008). Our array analyses report relatively low constitutive expression values for VP1340 over all conditions examined, but the expression of VPA0459 increased > 125-fold when the cells were grown on plates. In confirmation, secreted collagenase activity greatly increased upon cultivation in PS conditions compared with L. Thus, the pathogenic potential of the cell was found strikingly different between growth in liquid and on a surface.
Although many virulence traits have been observed to be co-regulated with swarming in many bacteria, the link between swarming and virulence has not been directly demonstrated in most cases (reviewed in Kirov, 2003; Verstraeten et al., 2008). To probe whether our observed increases in virulence gene expression were biologically meaningful in V. parahaemolyticus, the cytotoxicity of surface-adapted and liquid-grown cells was compared. We found that pre-growth on a surface or pseudosurface condition increased the ability of V. parahaemolyticus to kill host tissue culture cells. Furthermore, mutation of the swarming regulator lafK diminished cytotoxicity. Thus, the surface-responsive programme of gene control seems to prepare V. parahaemolyticus for virulence.
Commonalities with other swarming bacteria
Swarming cells may play a particular role in infection because surface-associated motility can promote colonization of specialized niches, and the coupling of swarming motility and virulence gene expression may confer distinct advantages. For example, swarming motility of P. mirabilis promotes the ascending colonization of the urinary tract (Allison et al., 1994). Lateral flagella aid colonization of human intestinal cell lines by Aeromonas spp. (Kirov et al., 2004), and swarming similarly enables colonization of the distal part of the root by certain plant pathogens (Sanchez-Contreras et al., 2002). The global gene transcriptome and proteome trends observed in P. aeruginosa, P. mirabilis, S. enterica serovar Typhimurium, and now in V. parahaemolyticus clearly support the idea that surface-adapted cells are more pathogenic (Kim and Surette, 2004; Wang et al., 2004; Overhage et al., 2008; Pearson et al., 2010). Type III secretion genes were induced upon growth on surfaces in S. enterica Typhimurium, P. aeruginosa and V. parahaemolyticus (Kim and Surette, 2004; Wang et al., 2004; Overhage et al., 2008; Pearson et al., 2010). The expression of genes encoding a variety of extracellular enzymes – e.g. elastase in P. aeruginosa (fivefold), collagenase in V. parahaemolyticus (125-fold), haemolysin and urease in P. mirabilis (eight- and fivefold respectively) and a variety of proteases in all three organisms – is upregulated and probably provides indicators of host target specificity (Kim and Surette, 2004; Wang et al., 2004; Overhage et al., 2008; Pearson et al., 2010). What seems worth remarking with respect to V. parahaemolyticus is that the fold and absolute level of expression is very high for many of these potential virulence genes, i.e. there is a profound on/off switch from one mode to the other.
The physiology of growth on surfaces
Although this report focuses on the genes that respond to surface sensing mediated by the polar flagellum, we emphasize that the genes that have been subtracted will also be important to explore in future studies – these genes, which are induced by growth on a surface but not by either of the pseudosurface conditions, are indicated in Table S3. The identity of some will provide insight into the basic physiology of growth on surfaces and others may illuminate surface-specific responses that are distinct from those mediated by polar flagellar sensing.
Our data did not reveal evidence that surface sensing per se enhances expression of antibiotic resistance genes. However, there were some expression changes observed for genes encoding efflux pumps and other transport proteins in the S versus L comparison (Table S3). The differences in antibiotic sensitivity between swimmers and swarmers that have been observed by others (Kim and Surette, 2004; Overhage et al., 2008; Lai et al., 2009) might be organism specific; however, we suggest that those observed changes might also accrue from differences provoked by the physiology of growth on a surface and the complexity of comparing such differently growing cell types. We note that recent work supports the latter idea (Butler et al., 2010).
One physiological distinction between life in liquid versus an agar medium apparent from the transcriptome comparisons is iron availability. Diffusion of iron to V. parahaemolyticus cells growing in communities on surfaces seems limiting and iron is a key signal mediating swarmer cell differentiation (McCarter and Silverman, 1989). The problem of iron acquisition seems generally important for cells growing in communities on surfaces, whether it be in a biofilm (Glick et al., 2010) or on agar during swarming. The microarray analysis presented here supports this observation: expression of some iron-acquisition genes increased (e.g. the ferrienterochelin TonB-dependent receptor) and expression of some genes encoding iron-containing proteins (e.g. ThiC) decreased. A list of iron-regulated genes (defined in the L versus LD comparison) that were also regulated during growth on the agar medium is provided in Table S4. Regulation of iron-pertinent genes by growth on a surface was also observed in the S. enterica Typhimurium and P. aeruginosa microarray analyses performed comparing cells harvested from plates and liquid (Wang et al., 2004; Overhage et al., 2008). Indeed, much of the surface-induced transcriptome response of P. aeruginosa seems centred on producing siderophores for iron acquisition (Overhage et al., 2008).
Here we begin to define the surface-responsive programme of gene control in V. parahaemolyticus. The genes in this network provide information about general surface colonization and virulence programmes as well as insight into the particular strategies employed by the organism. Some general themes with respect to the kinds of genes that are co-regulated with swarming are emerging for swarming bacteria, including enhanced production of virulence factors (Kirov, 2003; Wang et al., 2004; Overhage et al., 2008; Pearson et al., 2010). For V. parahaemolyticus, we show that regulation of some of these traits is the direct consequence of the organism's perception of its physical environment rather than a secondary consequence of metabolic differences that accrue during growth on surfaces in tightly packed communities. Moreover, our data demonstrate that the surface-sensing response primes the cell to be prepared for encounters with a potential host. We also find that the panel of sensory proteins and the level of the second messenger c-di-GMP present in the cell differ between life in liquid and on a surface, suggesting that cells may be programmed to detect and respond to different cues in different environments. The ability to move over and colonize a surface has important consequences with respect to bacterial survival strategies, and our findings show that bacteria can do some things inherently differently as a direct result of their perception of their physical environment.
Bacterial strains and media
The bacterial strains and plasmids used in this work are described in Table 5. HI broth contained 25 g of heart infusion (Difco) and 15 g of NaCl per litre; HI plates contained HI broth with 20 g l−1 granulated agar (Difco). Supplements were used at the following concentrations: 50 µM 2,2′-dipyridyl (Sigma) and 40 µM phenamil methanesulphonate salt (Sigma). The isolation of Tn5 and Tn5lux insertions in V. parahaemolyticus has been described (Stewart and McCarter, 2003; Jaques and McCarter, 2006).
To examine V. parahaemolyticus grown in liquid culture, overnight cultures were diluted 1:200 in HI and grown to an OD600 of ∼ 0.6. The subculture was diluted to an OD600 of 0.025 in 25 ml of HI in a 250 ml baffled flask and grown with shaking for 200 min at which time the cells were diluted into RNAprotect Bacteria Reagent (Qiagen) to a final OD600 of 0.5. To examine V. parahaemolyticus grown on plates, cells grown overnight on HI plates were suspended and to an OD600 of 0.05. Fifty microlitres of cells were spread onto HI plates. After 6 h of growth, cells were harvested and diluted to OD600 of 1.0 by using RNAprotect that had been diluted twofold in 1 × DPBS pH 7.1 (Gibco). These harvested cells were used for RNA isolation, examined microscopically to document average cell length and analysed for flagellin production by immunoblotting.
RNA isolation, cDNA synthesis, labelling and hybridization
Cells in RNAprotect were centrifuged and lysed by using lysozyme, proteinase K and QIAzol (Qiagen) according to manufacturer's protocol (Qiagen RNAprotect Bacteria Reagent Handbook). For each RNA preparation, 4 ml of liquid-grown cells or 2 ml of plate-grown cells were extracted and the RNA purified using RNeasy minicolumns (Qiagen), yielding ∼ 50 µg of RNA. cDNA was synthesized using the Superscript II Reverse Transcriptase kit (Invitrogen) and pD(N)6 random hexamers (GE Healthcare). Approximately 36 µg of RNA was converted to ∼ 7 µg of cDNA and DNase digested to yield fragments between 50 and 200 bp. cDNA was labelled with Biotin-ddUTP (BioArray Terminal Labeling Kit; Enzo). Samples (∼ 4.5 µg) were hybridized to the chips in the DNA Core Facility at the University of Iowa according to the standard Affymetrix protocols for Escherichia coli.
Custom GeneChips (rhofispaa52026F; 11-µm feature size) were constructed by Affymetrix (Santa Clara, CA). Chip annotation conforms to original genome annotation and tag numbers were assigned VP numbers for chromosome 1 and VPA numbers for chromosome 2 (Makino et al., 2003). The GeneChip contains 5161 probe sets designed by using the V. parahaemolyticus RIMD2210633 genome. Analysis was performed using the following software available at the University of Iowa DNA facility: Affymetrix Gene Chip Operating Software (1.2.1), Affymetrix GeneChip® Genotyping Analysis Software (GTYPE) 4.0. Using the raw signal output generated by GTYPE, processing and comparisons were made by using R version 2.8.1 GUI 1.27 Tiger build 32-bit (5301) ( http://www.r-project.org/) within Bioconductor software for bioinformatics (http://www.bioconductor.org) (Gentleman et al., 2004). The complete data set was pre-processed using the GCRMA method to perform optical adjustment, background adjustment, normalization and summarization functions (Wu et al., 2004). Data have been submitted to the NCBI and assigned the GEO accession number (GSE18763).
We note that the V. parahaemolyticus strain BB22 with which these studies were performed has yet to be sequenced; however, probing the Affymetrix GeneChip with genomic BB22 DNA yielded present calls for ∼ 98% of the predicted ORFs. Of the missing genes, many encode predicted phage, transposon, hypothetical proteins and the O-antigen locus. We also note that for multiple large transposon bank screens in strain BB22 for mutants with defects in swarming or biofilms, in only rare instances did the insertions produce ambiguous mapping to the RIMD genome. Thus, it seems reasonable to conclude that these transcriptome studies provide a good reflection of genome-wide expression patterns although a small fraction of gene activity could not be assessed.
Tests for differential expression were performed using an anova model. Because P-values can be misleading in the presence of genome-wide hypothesis testing, q-values, which adjust for multiple testing, were also calculated (Storey and Tibshirani, 2003). If a gene has a q-value of 0.01, this means that if the requirements for statistical significance were lowered just enough to allow this gene to meet them, then the overall false discovery rate of the study would be 1% (Benjamini and Hochberg, 1995).
Cell length measurements
Harvested cultures (100 µl) were fixed by dilution in an equal volume of 16% paraformaldehyde (Electron Microscopy Sciences) and were photographed using an Olympus BX60 microscope equipped with a black-and-white Spot 2 cooled charged coupled device camera (Diagnostic Instruments, Sterling Heights, MI). Cell length was measured using Image-Pro software version 4.1 (Media Cybernetics, Silver Spring, MD). At least 100 cells were counted per sample.
Cultures, grown and harvested as per the specified microarray condition, were centrifuged to pellet the cells (1 ml volumes). Collagenase activity of the supernatant was measured using the EnzChek Gelatinase/Collagnase Assay Kit (Molecular Probes). The fluorescent substrate, DQ gelatin, was used at 5 µg ml−1. Supernatants were diluted 1:2 into reaction buffer with substrate (200 µl final reaction volume). The reactions were performed in triplicate in 96-well optical plates and assayed at room temperature over time (30–180 min) by measuring fluorescence intensity using a microplate reader (GENious, TECAN). The excitation and emission wavelengths were 485 and 535 nm respectively. Background fluorescence (in the reaction mixture without cells) was subtracted. Purified Clostridium histolyticum Type IV collagenase (EnzChek Collagenase Assay Kit, Invitrogen) (0.1 U) produced ∼ 7000 fluorescence units in this assay. All strains were grown and assayed in at least three independent experiments.
Cytotoxicity was measured by co-culture of V. parahaemolyticus strains with CHO cells (ATCC CCL-61) essentially as described (Dasgupta et al., 2006). For co-culture experiments, CHO cells were seeded to yield ∼ 1 × 105 cells per well at the time of infection. The V. parahaemolyticus inoculum was grown as indicated using the same conditions for the microarray analyses. Bacteria were then suspended in pre-warmed Ham's F12 at 1.5 × 106 cells ml−1 to yield an approximate multiplicity of infection (moi) of 15; the inoculum was also plated to calculate the actual moi. Co-cultures were incubated at 37°C and 5% CO2 for the indicated times. Cytotoxicity was assayed by measuring release of lactate dehydrogenase (LDH) by using the CytoTox96 non-radioactive kit (Promega Corp., Madison, WI). Six replicate wells were measured per data point. Per cent of lysis was calculated by comparing to total lysis obtained in control reactions using 0.9% Triton X-100; control wells without bacteria were used to calculate the background level of lysis. Experiments have been repeated at least three times with similar results. Tests for statistical significance were conducted by using Student's t-test (two-tailed distribution with two sample, equal variance calculations).
The extraction procedure was kindly provided by Chris Waters (pers. comm.; Waters et al., 2008). The equivalent of 16 OD600 units of cells that were grown and harvested as per the microarray conditions were centrifuged for 3 min at 15 000 g and 4°C and suspended in 0.25 ml of extraction buffer, which was 40% methanol-40% acetonitrile in 0.1 N formic acid. Samples were incubated at −20°C for 30 min and centrifuged in a microcentrifuge for 5 min at 13 000 r.p.m. at 4°C. Supernatant was harvested (0.2 ml) and placed on ice. The residual supernatant and pellet were extracted a second time with 0.125 ml of extraction buffer. The second extraction supernatant (0.1 ml) was pooled with the first and neutralized by adding 4% volume of 15% NH4CO3. Samples were stored at −80°C and centrifuged prior to analysis. Each extraction and c-di-GMP measurement was repeated for two independent growth experiments.
Ten microlitres per extract was analysed by the Mass Spectrometry Facility at Michigan State University using liquid chromatography tandem mass spectrometry (LC-MS/MS) on a Quattro Premier XE mass spectrometer (Waters) coupled with a Acquity Ultra Performance LC system (Waters). C-di-GMP was detected with electrospray ionization using multiple reaction monitoring in negative ion mode at m/Z 689.16→344.31. The MS parameters were as follows: capillary voltage 3.5 kV, cone voltage 50 V, collision energy 34 V, source temperature 120°C, desolvation temperature 350°C, cone gas flow (nitrogen) 50 l h−1, desolvation gas flow (nitrogen) 800 l h−1, collision gas flow (nitrogen) 0.15 ml min−1 and multiplier voltage of 650 V. Chromatography separation was reverse phase using a Waters BEH C18 2.1 × 50 mm column and a flow rate of 0.3 ml min−1 with the following gradient of solvent A (10 mM tributylamine + 15 mM acetic acid in 97:3 water : methanol) to solvent B (methanol): t = 0 min, A-99%:B-1%; t = 2.5 min, A-80%:B-20%; t = 7.0 min, A-35%:B-65%; t = 7.5 min, A-5%:B-95%; t = 9.01 min, A-99%:B-1%; t = 10 min, end of gradient. Chemically synthesized c-di-GMP (Axxora) was dissolved in the extraction buffer at the concentrations of 250 nM, 100 nM, 20 nM, 10 nM, 5 nM and 2 nM and used to generate a standard curve for calculation of the concentration of c-di-GMP in each extract. These analytical methods do not report the ‘absolute’ cellular concentration of c-di-GMP. This is in part because the efficiency of our extraction procedure is not known and in-source ionization fragmentation has not been considered. Moreover, as observed by Waters et al. (2008), there is an unknown compound in extracts of Vibrio species that has a retention time and molecular mass close to c-di-GMP that complicates peak integration. Nevertheless, these values do reflect reproducible and significant differences in the cellular c-di-GMP pool between the strains and conditions measured. The concentration of c-di-GMP was normalized to OD600 and to bacterial protein concentration.
In order to normalize OD600 to mg protein extracted, the protein concentration was determined by using the Bradford method (Bio-Rad Protein Assay). A culture of known optical density grown as per the microarray conditions in liquid or on a surface was centrifuged, and the cell pellet dissolved in 0.1 N NaOH by heating for 15 min at 95°C as described (Spangler et al., 2010). The protein standard was BSA. Our conversion factors for mg protein per OD600 unit were 262 (± 4.48), 193 (± 14.2) and 184 (± 22.7) mg protein per OD600 unit for LM5674 (L), LM5674 (S) and LM6567 (S) respectively; the standard error of the mean is given in parentheses. The difference between OD600 to protein conversion factors for cells grown in liquid and surface conditions was statistically significant (i.e. L versus S for LM5674, the P-value < 0.0001). The conversion factors were derived from four independent growth experiments and quadruplicate assays per experiment. Samples of matched OD600 were also compared by using a dilution series for SDS-PAGE and Coomassie staining; the estimated difference between LM5674 (L) and LM5674 (S) was less than our limit of discrimination (< 25%). Thus, the comparison using the mg protein normalization method is the most conservative estimate of the differences in c-di-GMP concentration.
The wild-type strain LM5674 was grown according to microarray growth conditions in HI broth (L) and HI plates (S) to attain swim and swarmer cells. Cells were fixed and processed for immunofluorescence microscopy as described previously (Pogliano et al., 1997). The anti-Laf antiserum #127 and the anti-Fla antiserum #129 were diluted 1:80 000 and 1:5000 respectively. Antibody-reacted cells were incubated with membrane stain FM 4–64 (0.1 mg ml−1, Invitrogen) for 5 min. Micrographs were recorded on an Olympus BX60 microscope equipped with a 100 × UPlanApo objective.
Samples corresponding to ∼ 0.01 OD600 units were separated in 10% acrylamide gels by SDS-PAGE and transferred to nitrocellulose. Membranes were incubated for 1 h with polar flagellin antiserum #129 and lateral flagellin antiserum #127, which have been described (Boles and McCarter, 2000). Following a second incubation with anti-rabbit IgG antibody conjugated to horseradish peroxidase (1:10 000), membranes were developed with SuperSignal chemiluminescent substrate (Pierce) and visualized using an LAS-1000 luminescent imager (Fujifilm Life Science).
Lateral flagellar reporter assays
Vibrio parahaemolyticus strains carrying luminescence (lux) reporter fusions were grown as indicated, generally according to microarray growth conditions, and monitored periodically for bioluminescence as described (Stewart and McCarter, 2003). Maximum light production, which occurred between the 2–3 h time point in liquid and the 6–7 h time point on plates, is reported as specific light units, which are relative total light units per minute per millilitre per OD600. Lux assays were performed in triplicate, and each experiment was performed at least twice with similar results. V. parahaemolyticus strains carrying cosmids with a lacZ reporter fusion in the lateral flagellar lafK or flgC genes were grown overnight on HI plates with 10 µg ml−1 tetracycline. Each fusion shows surface-dependent induction of lacZ gene expression. Cells were suspended to an OD600 of 0.10, and 100 µl were spread upon multiple HI tetracycline plates. Cells were periodically harvested from these plates by suspending in 5 ml of medium. β-Galactosidase measurements that were performed according to Miller (1972), except that the cells were permeabilized by using Koch's lysis solution (Putnam and Koch, 1975). Assays were performed in triplicate, and standard deviations were generally less than 10% of the mean. Each experiment was performed at least twice with similar results. Tests for statistical significance were conducted by using Student's t-test (two-tailed distribution with two sample, equal variance calculations).
Reverse transcription polymerase chain reaction was performed by using the Access RT-PCR system (Promega Corp., Madison, WI). Annealing temperature was 56°C and the reactions proceeded for 30 cycles. Control reactions, to monitor the RNA for DNA contamination, showed no products in the absence of reverse transcriptase enzyme. Each reaction contained 50 ng of RNA template and four primers. Two of these were gene-specific primers and two were control primers designed to monitor mRNA for the constitutively expressed gene polar flagellar fliA. PCR products were examined in 1.0% agarose gels. Primer sequences and product sizes are supplied in Table S5.
This work was supported by NIH Grant 5 R21 AI065526, NSF grant 0817593, and a Medical Research Initiative Grant from the Carver College of Medicine. C.G.-P. and P.J.B. were supported by the NIH Training Grant 5 T32 GM077973 ‘Statistics in Microbiology, Infectious Diseases & Bioinformatics’. We thank the expert resources provided by University of Iowa Carver College of Medicine DNA Core Facility, Chris Waters for his kind help in transferring the technology of measuring cellular c-di-GMP, and the excellent assistance that we received from the Mass Spectrometry Facility at Michigan State University.