The bacterial strains and plasmids used in this study are listed in Table 1. All V. cholerae and Escherichia coli strains were routinely grown aerobically in Luria–Bertani (LB) medium (1 % tryptone, 0.5 % yeast extract, 1 % NaCl) at 30 °C and 37 °C respectively, unless otherwise noted. Agar medium contained 1.5 % granulated agar (Difco). Concentrations of antibiotics used were as follows: ampicillin, 100 μg ml⁻¹; rifampicin, 100 μg ml⁻¹; gentamicin, 50 μg ml⁻¹.

Restriction enzymes, DNA modification enzymes and Phusion High-Fidelity DNA polymerase were purchased from New England Biolabs. PCRs were carried out with primers purchased from Bioneer USA Corporation, listed in Supplementary Table S1, available with the online version of this paper. DNA sequencing was carried out by the Sequetech Corporation.

Deletion mutants were generated according to the protocol previously published (Fong et al., 2006; Fullner & Mekalanos, 1999). The DNA sequences of the constructed deletion plasmids were verified by DNA sequencing.

For analysis of corrugated colony morphology development, cultures grown overnight at 30 °C with shaking (200 r.p.m.) were serially diluted with LB medium and 100 μl aliquots of the diluted cultures were plated onto LB agar medium. The cultures were incubated at 30 °C for 2 days. For analysis of pellicle formation, glass culture tubes (18×150 mm) containing 5 ml LB medium were inoculated with overnight-grown cultures, resulting in a 200-fold dilution. The tubes were incubated at 30 °C under non-shaking conditions for 2 days. Assays were repeated with at least two different biological replicates.

Isolation of crude VPS from wild-type and mutant strains and immunoblot analyses of the extracted VPS were carried out according to a protocol similar to those previously published (Enos-Berlage & McCarter, 2000; Fong & Yildiz, 2007; Yildiz & Schoolnik, 1999). Briefly, overnight-grown cultures on LB agar medium were harvested and resuspended in PBS. Normalization was carried out by adjusting each culture to the same OD₆₀₀, measured with a Beckman Du Series 500 spectrophotometer. Equal volumes of the cultures were pipetted into Erlenmeyer flasks and the suspension was incubated overnight at 4 °C, shaking at 100 r.p.m. Crude VPS supernatant was separated from the bacterial cells and debris by centrifugation twice at 20 000 g for 30 min and precipitated with 3 volumes of ethanol at −20 °C overnight. Crude VPS was pelleted by centrifugation at 20 000 g for 30 min, washed with 70 % ethanol, air-dried and resuspended in 500 μl water. Purified VPS from the wild-type rugose variant was used as a positive control and for quantification. Purified VPS was isolated as described above with additional purification steps published previously (Fong & Yildiz, 2007). Briefly, crude VPS pellet was resuspended in nuclease buffer (40 mM Tris/HCl pH 8.0, 10 mM MgCl₂, 2 mM CaCl₂, 0.05 % NaN₃). DNase I and RNase A were added to the VPS suspension at final concentrations of 2 units ml⁻¹ and 50 μg ml⁻¹, respectively, followed by incubation at 37 °C shaking for 8 h. Proteinase K was then added at a final concentration of 200 μg ml⁻¹, and the suspension was further incubated with shaking overnight at 37 °C. Phenol/chloroform extractions (equal volumes) were carried out, followed by precipitation with 3 volumes of ethanol, washing with 70 % ethanol, and solubilization in water. Detoxi-Gel endotoxin-removing gel (Pierce) was used to remove contaminating lipopolysaccharides (LPS) according to the manufacturer's instructions and the purified VPS suspension was dialysed with water at 4 °C overnight using a Slide-A-Lyser 3.5K MWCO dialysis cassette (Pierce). The VPS suspension was then dried using a Micro Modulyo freeze-drier (Thermo Savant), the dry weight of the VPS sample was determined and the dried purified VPS was resolubilized in water. Crude and purified VPS (3 μl) was spotted onto nitrocellulose membranes and immunoblot analyses were carried out using anti-VPS antiserum and goat anti-rabbit horseradish-peroxidase-conjugated antibody. The immunoblots were developed with the SuperSignal West Pico chemiluminescent kit (Pierce) and quantified using ImageQuant 5.2 software (Molecular Dynamics). VPS immunoblot analyses were carried out with two different biological replicates and at least three technical replicates. It should be noted that the VPS antibody used in this study is polyclonal, with undefined epitopes. As such, the immunoblot assay was used as a phenotypic screen to identify mutants that cannot produce ‘wild-type’ VPS and not for determining the effect of a given mutation on the VPS structure.

Biofilm formation assays were carried out in PVC microtitre plates (BD Falcon) with 100 μl of overnight-grown cultures diluted to an OD₆₀₀ of 0.04. The microtitre plates were incubated at 30 °C for 8 h. Crystal violet staining and ethanol solubilization were carried out as previously described (Fong et al., 2006; Yildiz et al., 2001). Absorbance of the solubilized crystal violet stain was measured at 595 nm using a VersMax Tunable Microplate Reader (Molecular Devices). The assays were repeated with at least two different biological replicates and at least five technical replicates.

V. cholerae wild-type and deletion strains were tagged with green fluorescent protein (GFP) according to the procedure previously described (Fong et al., 2006). Flow-cell experiments were carried out at room temperature as described previously (Beyhan et al., 2007). Images were acquired using a Zeiss Axiovert 200 M laser scanning microscope. 3D images of the biofilms were reconstructed using imaris software (Bitplane) and quantified using comstat (Heydorn et al., 2000). Experiments were carried out with at least two biological replicates.

The infant mouse intestinal colonization model system was used in the single-strain infection experiment. Oral administration of 10⁶ exponential-phase V. cholerae wild-type variants or deletion mutants (unable to produce biofilm matrix components) to 4–5-day-old CD-1 suckling mice was carried out. At 20 h post-infection, the mice were sacrificed and their small intestines were removed and homogenized. Intestinal homogenates were serially diluted and plated onto selective agar. The number of bacteria per intestine was quantified by colony counting and used to determine intestinal colonization of the mutant strain. One-way ANOVA (Kruskal–Wallis test) was performed to determine statistically significant differences. Individual groups were then compared using the Mann–Whitney non-parametric t-test; a P-value of ≤0.05 was considered significant (Prism 5, GraphPad Software).

A major component of V. cholerae biofilm matrix is the VPS exopolysaccharide. VPS production is mediated by proteins encoded by the vps genes, which are organized into vps-I and vps-II clusters on the large chromosome, separated by an 8.3 kb intergenic region containing six genes (Fong & Yildiz, 2007; Fong et al., 2006; Yildiz & Schoolnik, 1999). There are 18 vps genes, 12 of which are located in the vps-I cluster (vpsU, VC0916; vpsA–K, VC0917–27), while the other six are located in the vps-II cluster (vpsL-Q, VC0934–9) (Fig. 1). To determine if all the vps gene products are required for biofilm formation, we generated in-frame deletion mutants for each vps gene in the wild-type rugose genetic background (hereafter referred as RΔ). We also constructed mutant strains containing deletions of the first and/or second vps clusters [RΔvps-I (vpsA–K), RΔvps-II (vpsL–Q) and RΔvps-IΔvps-II]. We then characterized the mutants for colony morphology, pellicle formation, biofilm formation and VPS production (Table 2). It is important to note that the deletion strains do not exhibit growth defects when compared to the parental wild-type strain (data not shown).

The rugose wild-type forms corrugated colonies (Fig. 2) and a robust pellicle (Fig. 3), and has increased capacity to form a biofilm (Fig. 4a) and produce VPS (Fig. 4b, c). We termed these phenotypes ‘rugosity-associated phenotypes’. Mature biofilm formed by the rugose wild-type also exhibited the characteristic elaborate 3D structures as analysed by CSLM (Fig. 5). We first characterized strains lacking the vps clusters (RΔvps-I, RΔvps-II and RΔvps-IΔvps-II). These mutant strains formed flat and smooth colonies (Fig. 2), and exhibited a marked decrease in the production or secretion of VPS (Fig. 4b, c), and reduction in the ability to form pellicles (Fig. 3) and biofilms (Fig. 4a). CSLM analysis of biofilms of the rugose wild-type and RΔvps-IΔvps-II revealed that the latter is unable to form the elaborate 3D mature biofilm structures (Fig. 5). comstat analysis revealed that total biomass, and mean and maximum thicknesses, are lower in RΔvps-IΔvps-II compared to the rugose wild-type, particularly after 24 h (Table 3). Strains carrying in-frame deletions of either the vps-I or vps-II cluster also form biofilms resembling that of RΔvps-IΔvps-II (data not shown).

We then analysed the rugosity-associated phenotypes of the individual vps deletion mutants. Based on the predicted amino acid sequences and domains of the vps gene products and their predicted functions, we can group these VPS proteins into six classes (Table 2). Class I consists of proteins that are predicted to be required for production of nucleotide sugar precursors. VpsA is predicted to encode UDP-N-acetylglucosamine 2-epimerase while VpsB is predicted to encode UDP-N-acetyl-d-mannosaminuronic acid dehydrogenase. UDP-N-acetylglucosamine 2-epimerase catalyses the interconversion of UDP-N-acetyl-d-glucosamine to UDP-N-acetyl-d-mannosamine. The latter is then converted to UDP-N-acetyl-d-mannosaminuronic acid by UDP-N-acetyl-d-mannosamine/mannosaminuronic acid dehydrogenase (Kawamura et al., 1979, 1985). Since VpsA and VpsB are predicted to be involved in the production of nucleotide sugar precursors, it is not surprising that deletion of vpsA and vpsB resulted in strains that exhibited decreased colony corrugation (Fig. 2) and reduction in the ability to form pellicle (Fig. 3), biofilm and VPS (Fig. 4).

Class II consists of four predicted glycosyltransferases encoded by vpsD, vpsI, vpsK and vpsL. VpsD and VpsI are classified as members of the glycosyltransferase family 4 (GT4), while VpsK is classified as a member of GT26 in the Carbohydrate Active Enzymes (CAZY) database (http://www.cazy.org) (Cantarel et al., 2009). Although VpsL is not classified as part of any GT family, the Universal Protein Resource (UniProt) database (http://www.uniprot.org/) shows that VpsL contains a bacterial sugar transferase domain (PF02397), suggesting that VpsL is probably a glycosyltransferase. Glycosyltransferases catalyse the transfer of sugar moieties to specific acceptor molecules (Campbell et al., 1997; Whitfield, 2006), forming glycosidic bonds that are predicted to be involved in initiation and/or elongation of VPS subunits (repeat units). Thus, the mutants lacking glycosyltransferase activity are unlikely to produce mature VPS. As expected, strains containing in-frame deletion of these genes (RΔvpsD, RΔvpsI, RΔvpsK and RΔvpsL) exhibited flat and smooth colony morphology (Fig. 2). These mutants were also unable to form a pellicle (Fig. 3), and exhibited a significant reduction in biofilm formation and were unable to produce VPS (Fig. 4).

Class III consists of proteins encoded by vpsE, vpsH, vpsN and vpsO that are predicted to be involved in VPS polymerization and export. VpsE contains a predicted polysaccharide synthesis domain (PF01943), which is also found in Wzx-like proteins that are O-antigen translocases involved in LPS export in E. coli (Marolda et al., 2004, 2006; Whitfield, 1995). Besides the presence of the loosely conserved PF01943 domain, there is very low conservation in the protein primary sequence in these Wzx-like proteins (Marolda et al., 2006). Interestingly, PelG from Pseudomonas aeruginosa, which is involved in biofilm formation, has also been described as a Wzx-like polysaccharide transporter (Vasseur et al., 2005). Based on computational analysis, VpsE is likely to be a polysaccharide export protein. VpsH is annotated as a putative CapK protein in The Institute for Genomic Research Comprehensive Microbial Resource (TIGR-CMR) and UniProt databases. CapK in Staphylococcus aureus is reported to be a possible capsular polysaccharide polymerase (Sau et al., 1997). VpsN is predicted by the UniProt database to encode a polysaccharide export-related protein that contains a polysaccharide export domain (PF02563), also associated with the Wzx-like proteins, suggesting that VpsN may also be involved in polysaccharide export. VpsO is predicted by the UniProt database to contain a PF02706 domain associated with the E. coli chain length determinant protein Wzz (Franco et al., 1998).

Mutants in class III may produce lipid-linked subunits of VPS, but not full-length VPS. If VPS transport takes place after complete polymerization in the cytoplasm or periplasm, VPS could accumulate in the cytoplasm or periplasm. Alternatively, unpolymerized or partially polymerized VPS may be secreted. Mutants lacking functional VpsE, VpsN and VpsO produced flat and smooth colonies (Fig. 2), were unable to form a pellicle (Fig. 3), and exhibited drastic reduction in biofilm formation and VPS production (Fig. 4). Although a mutant lacking functional VpsH (RΔvpsH) exhibited smooth colony morphology, it is important to note that the colony is more compact compared to the vps-clusters deletion mutants. Furthermore, RΔvpsH formed a pellicle with altered structures (Fig. 3, top panels) and exhibited a moderate decrease in biofilm formation (Fig. 4a). In addition, RΔvpsH produced much less VPS than the rugose wild-type, but not as little as the vps-cluster deletion mutants (Fig. 4b, c), suggesting that production of VPS intermediates may still occur in RΔvpsH.

Class IV consists of proteins that are encoded by vpsC and vpsG. VpsC and VpsG are predicted to be acetyltransferases involved in the addition of acetyl groups to polysaccharide. Mutants lacking the acetyltransferases were expected to produce VPS without acetyl modifications, which are likely to have different biophysical properties from those of the wild-type. RΔvpsG exhibited a significant reduction in colony corrugation but the colonies were not completely flat and smooth (Fig. 2). RΔvpsG also formed a pellicle with altered structure (Fig. 3, top panel), and showed reduced biofilm formation and VPS production (Fig. 4). In contrast, RΔvpsC retained its rugose colony corrugation (Fig. 2), formed a pellicle with structure similar to the rugose variant (Fig. 3) and did not exhibit significant defects in biofilm formation on solid surfaces or VPS production (Fig. 4). The structural modification of alginate by O-acetylation alters its physical properties, including viscosity (Nivens et al., 2001). O-Acetyl groups are required for the formation of mature biofilm structures in alginate-producing mucoid P. aeruginosa strains, as well as for maintenance of chronic P. aeruginosa infections (Pier et al., 2001). VPS is highly viscous, and thus non-acetylated VPS may have reduced viscosity, affecting biofilm structure and stability. While the altered rugosity-associated phenotypes exhibited by RΔvpsG are expected, it is surprising that RΔvpsC formed colonies and pellicles that resemble those of the rugose parent. We also generated a strain carrying in-frame deletions of both vpsC and vpsG. This double mutant exhibited phenotypes similar to those of the single vpsG mutant (Figs 2, 3 and 4). It is possible that VpsC is not a functional acetyltransferase, or that VpsG and VpsC may be required for the modification of different portions of the VPS, or even that these different acetyltransferase may be required at different stages of biofilm formation. We are currently testing these hypotheses.

Class V consists of the phosphotyrosine-protein phosphatase encoded by vpsU, and deletion of vpsU resulted in a strain with altered colony morphology (Fig. 2), pellicle structure (Fig. 3), biofilm formation and VPS production (Fig. 4). Phosphotyrosine-protein phosphatases catalyse the dephosphorylation of tyrosine-phosphorylated proteins (Kennelly & Potts, 1999). A number of phosphotyrosine-protein phosphatases in both Gram-positive and Gram-negative bacteria have been identified (Morona et al., 2000, 2002; Soulat et al., 2002; Vincent et al., 1999), and have been shown to be involved in regulation of exopolysaccharide production (Morona et al., 2000; Vincent et al., 2000). The phosphotyrosine-protein phosphatases found in Gram-negative bacteria usually contain two conserved active-site sequence motifs (C-X₄-C-R and D-P-Y) (Kennelly & Potts, 1999; Su et al., 1994; Vincent et al., 1999), while those found in Gram-positive bacteria usually contain four conserved motifs (Aravind & Koonin, 1998; Morona et al., 2002), except for PtpA and PtpB from S. aureus, which resemble phosphotyrosine-protein phosphatases found in Gram-negative bacteria (Soulat et al., 2002). Alignment of VpsU (V. cholerae), Wzb (E. coli), PtpA (P. aeruginosa PAO1) and PtpAB (S. aureus) indeed showed that VpsU contains the two active-site motifs conserved in other similar phosphotyrosine-protein phosphatases (data not shown). The altered rugosity-associated phenotypes exhibited by RΔvpsU indicate a role of VpsU in VPS production, although the target for VpsU remains to be determined. In E. coli, Wzb (phosphotyrosine-protein phosphatase) and Wzc (protein-tyrosine kinase) function together in the production of the capsular polysaccharide colanic acid (Vincent et al., 2000). Although the UniProt database suggested that VpsO is similar to the chain length determinant protein Wzz based on the presence of a PF02706 domain, blastp search using Wzc peptide sequence identified VpsO as the top match (E-value 4.6×10⁻²²), suggesting that VpsO may be the target of VpsU. We are currently investigating this hypothesis.

Class VI consists of hypothetical proteins encoded by vpsF, vpsJ, vpsM, vpsP and vpsQ. While RΔvpsP and RΔvpsQ strains produce VPS and form colonies, pellicles and biofilms similar to those of the wild-type, other mutants lacking vpsF, vpsJ and vpsM exhibited complete loss of colony corrugation, inability to form pellicle and drastic reduction in biofilm formation and VPS production (Figs 2, 3 and 4). Since vpsF, vpsJ and vpsM are predicted to encode hypothetical proteins, identifying the roles of these hypothetical proteins in building the biofilm matrix is critical.

Collectively, phenotypic analysis of the vps mutants shows that, under our experimental conditions, most of the VPS proteins are required for wild-type rugose colony corrugation, pellicle and biofilm formation, and VPS production.

V. cholerae biofilm formation inside the host and its contribution to pathogenesis is not well understood. While some biofilm-related factors have been studied for their roles in V. cholerae intestinal colonization (Lauriano et al., 2004; Rashid et al., 2004; Watnick et al., 2001; Zhu & Mekalanos, 2003), different research groups have obtained contradictory results regarding the importance of these factors. For example, mutants lacking VpsR, a positive transcriptional regulator of biofilm matrix production, have given variable results for mouse infection, including normal and decreased colonization (Rashid et al., 2004; Tischler & Camilli, 2005). In another study, a rugose variant (resulting from flaA deletion) was found to be defective for infant mouse intestinal colonization in a competition infection, suggesting that a high level of VPS production is detrimental (Watnick et al., 2001). It is therefore evident that many factors, possibly including the presence of VPS and matrix proteins, affect in vivo fitness of V. cholerae.

To this end, we carried out experiments to investigate whether known biofilm determinants in V. cholerae contribute to pathogenesis. We initially performed competition experiments using the smooth (low-biofilm-forming) and rugose (high-biofilm-forming) wild-type variants in the commonly used infant mouse model, but did not see a significant difference in the competitive index (data not shown). We reasoned that competition assays alone may be misleading, as VPS produced by the rugose variant could complement the co-inoculated strain in trans, thus masking differences in in vivo fitness. Therefore, we performed single-strain infections in the infant mouse model. We first compared the colonization capacities of smooth and rugose wild-type variants. As shown in Fig. 6, the rugose variant colonized the intestine better (approx. 3.6-fold) than the smooth variant. These data suggest that matrix components contribute to colonization and increase in vivo fitness, since the rugose variant produces more biofilm matrix than the smooth variant. To test this hypothesis, we compared the colonization capacity of the rugose wild-type variant to those of a mutant that is incapable of producing VPS (RΔvps-IΔvps-II) and mutants that are unable to produce matrix proteins (RΔrbmA and RΔrbmCΔbap1). RbmA, RbmC and Bap1 matrix proteins modulate rugosity-associated phenotypes (Fong & Yildiz, 2007; Fong et al., 2006). RbmC and Bap1 share 46.7 % peptide sequence similarity and appear to be partially redundant. Deletion of rbmA and bap1 in the rugose genetic background resulted in strains with altered colony morphology and biofilm formation when compared to the rugose wild-type (Fong & Yildiz, 2007; Fong et al., 2006). Although deletion of rbmC (a homologue of bap1) in the rugose genetic background did not alter the colony morphology and biofilm formation significantly, the double deletion mutant RΔrbmCΔbap1 exhibited a marked decrease in biofilm formation (Fong & Yildiz, 2007).

We determined that RΔvps-IΔvps-II exhibited a severe defect in intestinal colonization, with a 9.4-fold lower c.f.u. recovery compared to the rugose variant. This result indicates that VPS production increases in vivo fitness. The mutant lacking the biofilm matrix protein RbmA also exhibited a defect in intestinal colonization, with a 4.1-fold lower c.f.u. recovery. This outcome indicates that VPS and RbmA contribute significantly to in vivo fitness of V. cholerae in these models, and that they contribute to biofilm formation in both in vivo and aquatic environments. In contrast, there was no significant difference in the colonization ability of the RΔrbmCΔbap1 double mutant when compared to the rugose variant, indicating that either RbmC and Bap1 do not significantly influence intestinal colonization, or their phenotype is not evident under the experimental conditions utilized, thus suggesting that the major function of these biofilm matrix proteins is in the formation of biofilms in aquatic environments. Collectively, these results indicate that the biofilm matrix components are important in in vitro biofilm formation, and that VPS and RbmA also contribute to in vivo fitness of V. cholerae.

The results presented in this study have revealed how the various vps gene products contribute to wild-type biofilm formation and VPS biosynthesis, and have laid the groundwork for future biochemical characterization of the VPS biosynthetic pathway. In vitro experiments carried out with the vps mutants in this study, and rbm mutants in previous studies, clearly demonstrate the importance of matrix components, both VPS and matrix proteins, in the building of a robust wild-type biofilm. In-frame deletion of 15 of the 18 vps genes resulted in strains that exhibited drastic altered rugosity-associated phenotypes (Figs 2, 3 and 4). VPS and RbmA also contribute to in vivo fitness of V. cholerae, as shown in in vivo mouse colonization studies, where mutants not able to produce VPS or RbmA were defective for intestinal colonization (Fig. 6).

Humans ingest V. cholerae biofilms as part of the pathogen's normal transmission route. In this study, we have shown that VPS and biofilm matrix proteins are critical for in vitro biofilm formation and are also important for in vivo fitness. The nature of the VPS coating and matrix proteins in the biofilms may influence the progression of the disease, and/or development of an immune response against V. cholerae. A better understanding of biofilm matrix biosynthesis and function could allow us to develop inhibitors that specifically alter the matrix properties and affect either ex vivo (aquatic) survival or in vivo pathogenesis.