Vibrio cholerae Utilizes Direct sRNA Regulation in Expression of a Biofilm Matrix Protein

Vibrio cholerae biofilms contain exopolysaccharide and three matrix proteins RbmA, RbmC and Bap1. While much is known about exopolysaccharide regulation, little is known about the mechanisms by which the matrix protein components of biofilms are regulated. VrrA is a conserved, 140-nt sRNA of V. cholerae, whose expression is controlled by sigma factor σE. In this study, we demonstrate that VrrA negatively regulates rbmC translation by pairing to the 5′ untranslated region of the rbmC transcript and that this regulation is not stringently dependent on the RNA chaperone protein Hfq. These results point to VrrA as a molecular link between the σE-regulon and biofilm formation in V. cholerae. In addition, VrrA represents the first example of direct regulation of sRNA on biofilm matrix component, by-passing global master regulators.


Introduction
Vibrio cholerae inhabits aquatic environments and when it enters the human intestine, e. g., through ingestion of contaminated food or water, it causes the severe diarrheal disease, cholera. Vibrios are shown to form biofilms on zooplankton, insects and intestines [1][2][3][4][5]. Compared to planktonic cells, bacteria within biofilms are more resistant to stress conditions, e. g., osmotic and oxidative stress, acidity, antibiotics exposure and immune clearance [6][7][8][9][10][11][12]. Biofilm structures are constructed of and maintained by biofilm matrix components [13]. In V. cholerae, formation of biofilm requires production of exopolysaccharide (VPS) and the biofilm matrix proteins RbmA, RbmC and Bap1 [14][15][16][17][18]. These matrix proteins appear to be involved at particular steps during the biofilm formation process. RbmA is involved in the initial cell-cell adhesion step and serves as a tether, forming flexible linkages between cells and the extracellular matrix [18,19]; Bap1 facilitates adherence of the developing biofilm to surfaces; and the heterogeneous mixtures of VPS, RbmC and Bap1 appear to form envelopes to encase the cell clusters [18]. Without RbmC, incorporation of VPS through the biofilms is significantly reduced, suggesting an essential role for RbmC in maintaining the mature biofilm structure [18].
To date, studies on the regulation of biofilm formation have been mainly focused on VPS synthesis. A complex regulatory network controls transcription of the vps gene in response to multiple environmental signals, such as signals from quorumsensing bacterial autoinducers [20], polyamines [21,22], nucleosides [23,24], indole [25] and nutrient scarcity [26]. Recently, glucose-specific enzyme IIA has also been shown to regulate biofilm formation through binding to a carbon storage regulator homolog MshH, demonstrating a link between the phosphoenolpyruvate phosphotransferase system and biofilm formation [27,28]. In contrast to the vast body of knowledge about VPS regulation, very little is known about regulation of the matrix proteins (RbmA, RbmC and Bap1). Fong et al [29] has demonstrated the involvement of two factors: the cyclic AMP (cAMP)-cAMP receptor protein (CRP) complex and a transcriptional regulator VpsR. While VpsR positively regulates transcription of the rbm genes, cAMP-CRP appears to negatively regulate rbm expression, both mediated by and independently of VpsR [29].
In the past decade, an increasing body of evidence has highlighted the important and complex roles of small regulatory RNAs (sRNAs) in bacterial physiology and pathogenesis [30,31]. Many sRNAs are produced in response to specific environmental signals/stresses. They act by base-pairing with target sequences, resulting in up-or down-regulating gene expression through modulating the translation or the turnover of target mRNAs (see review [32]). This mechanism of regulation often requires the RNA chaperone protein Hfq that facilitates base pairing between sRNAs and their target mRNAs [33,34]. In Vibrio, a s Edependent sRNA, VrrA, has been shown to be induced by envelope stress and to repress the outer membrane proteins OmpA and OmpT through base pairing to the 59 untranslated regions (UTR) of the corresponding mRNAs. When the OmpA level decreases, envelope stress is reduced by releasing outer membrane vesicles (OMVs) [35,36]. These OMVs further protect bacteria against environmental hazards such as UV damage [37]. Using the infant mouse model, VrrA was demonstrated to attenuate V. cholerae virulence [35], which could be partially explained by the VrrA-mediated down-regulation of TcpA, a major V. cholerae virulence factor essential for host colonization. In this study, we provide evidence that VrrA down-regulates the biofilm matrix protein RbmC by base-pairing with the 59-UTR of rbmC mRNA. Because RbmC is essential for maintaining the mature structure of biofilms, this VrrA-mediated suppression of RbmC might be an additional mechanism of biofilm regulation in V. cholerae.

VrrA down-regulates RbmC independently of Hfq
In our previous studies, VrrA was shown to down-regulate bacterial structural proteins such as OmpA, TcpA and OmpT [35,36]. When we analyzed the profile of secreted proteins by SDS-PAGE and Coomassie-brilliant-blue staining, we noticed that a protein band at <100 kDa was more abundant in the Dhfq background than in the wild-type background (Fig. 1, compare lanes 5-8 to lanes 1-4). Further, this protein appeared to be more abundant in the DhfqDvrrA strain than in the Dhfq strain, and the lower level was restored in the vrrA complemented strain (Fig. 1A,  lanes 7 and 8). The protein band, marked with asterisk in Fig. 1A lane 6, was excised from the gel, subjected to mass spectrometry analysis, and identified as the biofilm matrix protein RbmC (VC0930).
In order to detect the low levels of RbmC in the wild-type background, we performed Western blot analysis using anti-RbmC polyclonal antiserum [38]. As expected, the antiserum could detect RbmC in the wild-type strain (Fig. 1B, upper panel, lanes 1-4) while no band was detected in a DrbmC mutant (Fig. 1B, upper panel, lane 9), confirming antibody specificity. Similar to what was earlier noticed in the Dhfq background strains, the RbmC level was elevated in the absence of VrrA in the wild-type background strains and this elevated level was also reduced when the DvrrA strain was complemented with VrrA expressed from a plasmid (Fig. 1B, upper panel, lanes 1-4). A SDS-PAGE Coomassie blue staining gel was shown (Fig. 1B, lower panel) as a sample loading control. These data indicated that the VrrAmediated regulation of RbmC expression did occur in the absence of Hfq. This suggests that Hfq is not essential for RbmC repression by VrrA although it is also feasible that Hfq can enhance the repression. We also observed that in the hfq mutant the basal RbmC protein level was higher (compare lane 1 with lane 5 in Fig. 1B, upper panel). The apparent repression by Hfq was presumably not strictly dependent on VrrA and could possibly also be mediated by some other sRNA. The higher basal level of the RbmC protein in the hfq mutant could also be an indirect effect through transcriptional control by a transcriptional regulator that is affected by Hfq.
The 59 region of rbmC mRNA is responsive to VrrA regulation In order to further study the interaction between VrrA and the rbmC mRNA, we first determined the transcriptional start site of rbmC by 59 RACE analysis. After sequencing analysis as described in Material and Methods, the rbmC transcriptional start site was determined to be 125 nt upstream from the AUG start codon.
Our earlier studies on the interaction between VrrA and its targets demonstrated that VrrA represses translation initiation by base-pairing with the 59-UTR of target mRNAs (ompA, tcpA and ompT). We hypothesized that VrrA would interact similarly with the rbmC mRNA. To test this hypothesis, we used a publically available prediction program, the RNAhybrid algorithm [39], to predict possible RNA duplexes formed by VrrA and the 59 region of the rbmC mRNA. The query sequence used for rbmC mRNA included the region from the transcriptional start site to 30 nt into the rbmC coding region. As shown in Fig. 2A, RNAhybrid algorithm predicted duplex formation between the residues 91-106 of VrrA and the 28 to 225 region of rbmC mRNA (numbering of rbmC is relative to the AUG start codon). This 13bp duplex is interrupted by a bulge dividing the stretch into a 7-bp and a 6-bp duplex, with the latter masking the Shine-Dalgarno (SD) region required for translation initiation.
In order to dissect interacting base pairs, we introduced point mutations into VrrA (Fig. 2B). Plasmid pTS2 is a ColE1-based plasmid expressing vrrA from its own promoter [35] generated plasmids pTS2-M7, pTS2-M8, pTS2-M9 and pTS2-M10, respectively. Each plasmid was introduced by transformation into strain DNY7 (DvrrA) and sRNA expression from the resulting plasmids were confirmed by Northern blot analysis (Fig. 3A, upper panel). The 5S rRNA was probed as internal control (Fig. 3A, lower panel). Interestingly, the VrrA-M7 level appeared higher than other VrrA variants. To compare the potential structures of these VrrA variants, RNA folding and pattern examination were performed using the Mfold web server [40]. The predicted structure of VrrA-M7 was found to be somewhat different from the predicted structures of the other variants (Fig. 3B). A feasible explanation would be that the VrrA-M7 might be more stable than wild-type VrrA, VrrA-M8, VrrA-M9, and VrrA-M10 due to a structural alteration. Another possible explanation for the higher levels of the VrrA-M7 mutant might be that this mutation could disrupt binding and codegradation of the sRNA with another target.
Supernatant proteins of the different sRNA-expressing strains were then analyzed to compare the production of RbmC. As shown in Fig. 4A (upper panel), compared to the wild-type VrrA WT (expressed from pTS2), VrrA M7 (expressed from pTS2-M7) partially lost its ability to repress RbmC production whereas VrrA M8 (expressed from pTS2-M8) could repress RbmC production to the same extent as VrrA WT . In contrast, VrrA M9 and VrrA M10 (expressed from pTS2-M9 and pTS2-M10, respectively) completely lost their ability to repress RbmC production. A SDS-PAGE Coomassie blue staining gel (Fig. 4A, lower panel) was included as the sample loading control. These results show that C 100 U 101 U 102 in VrrA are important for regulating expression of RbmC.
We next introduced mutations in the rbmC 59-UTR (A 221 A 220 G 219 to U 221 U 220 C 219 , Fig. 2A), generating the compensatory rbmC* allele. This rbmC* allele was introduced into the chromosome of DNY7 (DvrrA) by site-directed mutagenesis. As shown in Fig. 4B, the VrrA M10 variant expressed from plasmid pTS2-M10 lost its ability to repress RbmC production ( . These data suggest that VrrA acts directly as an antisense RNA to repress rbmC mRNA in vivo. In our earlier study, VrrA mutant variants (VrrA M1 to VrrA M6 ) expressed from plasmids pTS2-M1 to pTS2-M6 ( Fig. 2B) were constructed to study the interaction between VrrA and the ompT mRNA. We showed that VrrA mutant variants covering the VrrA region from residues 69-78 was responsible to base-pair with 59 UTR of ompT mRNA [36]. In order to see whether these residues would be important for RbmC regulation as well since the residues 69-78 were closed to the interacting region, we monitored RbmC levels in the strains expressing VrrA M1 to VrrA M6 by Westen blot analysis. As shown in Fig. 5 (upper panel) none of these variants lost its ability to repress RbmC, suggesting that ompTand rbmCregulating regions in VrrA do not overlap.

VrrA modulates biofilm formation
The findings about the ability of VrrA to down-regulate RbmC levels prompted us to analyze the impact of VrrA on biofilm formation by V. cholerae. We compared the biofilm forming ability using a once-through flow cell system and analysis by confocal laser scanning microscopy (CLSM). The over-expression of VrrA from a plasmid clone in the wild-type strain markedly decreased biofilm formation at 48 h when compared to that of the same strain containing the plasmid vector (Fig. 6A, c and f). Although initial stages of biofilm formation at 2 h and 24 h were not markedly altered by vrrA gene overexpression (Fig. 6A, a and d; b and e), COMSTAT analysis of biofilms developed 48 h post inoculation revealed that total biomass, average and maximum thicknesses of the wild-type strain overexpressing vrrA were markedly decreased after 48 h compared to those of the wild-type strain harboring only the plasmid vector after 48 h although the growth rate and yield were similar between control and overexpression strains. These results show that over-expressing VrrA impairs the ability of V. cholerae to form biofilms.

Discussion
V. cholerae transits between fundamentally different habitats the aquatic environment and the human digestive tract. Such transitions require rapid acquisition and integration of environmental cues in order to coordinate adequate genetic programs and adapt to the new niche. One such adaptation program involves the switch between a planktonic, motile lifestyle and a biofilm-based sessile lifestyle. To date, numerous regulator proteins have been found to affect biofilm formation in V. cholerae, such as those described in the Introduction. Results from this study add a new class of regulators, sRNAs, as a direct regulator of a biofilm matrix component. Through down-regulation of RbmC, VrrA weakens the stability of the mature biofilm structure and might therefore facilitate dispersal of bacteria from a sessile to a planktonic life style. In addition, because expression of VrrA is controlled by sigma factor s E , VrrA serves as a molecular link between the s Eregulon and biofilm formation in V. cholerae.
Several sRNAs have been shown to be involved in biofilm formation in E. coli and Salmonella, e. g. OmrA/B [41], McaS [42,43], RprA [44] and GcvB [42]. In contrast to VrrA, these sRNAs do not target biofilm matrix components directly, instead they target biofilm master regulators such as CsgD, which in turn regulates biofilm components. This generates a hierarchical regulatory network and enables csgD mRNA to serve as a hub for complex signal integration via multiple sRNAs [45,46]. Similarly in Vibrio, sRNAs Qrr1-4 and CsrB/C/D regulate the biofilm master regulator HapR or the regulatory molecule cyclic di-GMP (through diguanylate cyclase) [47,48], and thus are indirectly involved in biofilm formation.
VrrA belongs to a growing family of sRNAs that regulate multiple targets [48,49]. VrrA uses unique pairing regions to differentially regulate different mRNA targets. Compensatory base pair change experiments revealed that residues C 100 U 101 U 102 (numbers relative to the +1 transcriptional start site) in VrrA are essential for base-pairing with rbmC mRNA, while those required for the regulation of ompT mRNA are G 73 C 74 U 75 in VrrA [36].
In addition to the target-specific regulating regions in VrrA, dependency on the chaperon protein Hfq differs among mRNA targets as well. Although deletion of hfq abolishes the interaction between VrrA and ompT mRNA, Hfq is not absolutely required for the regulation on ompA [35] or rbmC mRNAs (this study). The observation that OmpA and RbmC levels were elevated in the Dhfq strain and that VrrA could only partially repress this elevated expression suggests that additional sRNAs are involved in the regulation. The combination of target-specific regions in VrrA and differentiated requirement of Hfq allows VrrA to modulate multiple targets differentially.
According to the RNAhybrid prediction, as shown in Fig. 2, A 91 C 92 U 93 C 94 C 95 U 96 in VrrA base pairs to the potential SD sequence (AGGGAGU) of rbmC. We therefore expected to see the most drastic change in RbmC level in strains expressing VrrA M7 (substitution of A 91 C 92 U 93 C 94 C 95 U 96 with U 91 G 92 A 93 G 94 G 95 A 96 ) and VrrA M8 (substitution of A 91 C 92 U 93 with U 91 G 92 A 93 ). However, our results showed that VrrA M9 and VrrA M10 , which base pairs to the region upstream of the SD sequence, had more impact on the regulation of RbmC. This unexpected result might be due to the fact that the SD sequence was predicted based on the consensus sequence and therefore might not be the exact SD site. Future studies using e. g. toeprint analyses will hopefully identify the actual interaction site(s) between VrrA and rbmC mRNA. Nevertheless, the present results from the compensatory base pair substitution experiment demonstrate that there is a direct interaction between VrrA and rbmC at the region upstream of the putative SD sequence (Fig. 4B).
It is noteworthy that there are only a few functional homologs to VrrA in other Gram-negative bacteria. One such example is the MicA sRNA in Salmonella and E. coli [50,51]. Both MicA and VrrA are s E -dependent and are capable of down-regulating multiple outer membrane proteins by basepairing mechanisms [35,52]. Interestingly, Kint et al [45] observed that MicA in Salmonella was involved in biofilm formation, although the molecular mechanism remains unknown. Systematic searches for MicA targets using bioinformatics prediction tools have not identified yet any biofilmrelated genes. Future work will be needed to examine possible interactions between MicA and Salmonella biofilm components such as curli and fimbriae.
In summary, VrrA is the first example of an sRNA molecule that directly targets expression of a biofilm matrix component. Given the similarities between VrrA and its homologs in other Gram-negative bacteria, it is plausible that similar direct regulation exists in other bacteria as well. Because VrrA weakens the stability of the mature biofilm structure, strategies directed towards mechanisms or levels of sRNAs to disturb bacterial biofilm formation may potentially be used to combat biofilmrelated infections. Furthermore, in our earlier studies, we showed that the TcpA, one of the colonization factors of V. cholerae, was down-regulated by VrrA (Song et al. 2008). In this study, we demonstrated that the expression of one of the extracellular matrix proteins, RbmC that is important for the biofilm formation by V. cholerae was modulated by VrrA. We hypothesize that at the later stage of V. cholerae infection in the host, bacteria can move away from the epithelial surface and into the fluid-filled lumen of the intestine. During this time, the bacteria may undergo a switch from attachment to the epithelial surface to detachment. This process may be associated with up-regulation of VrrA. We suggest that this transition prepares the bacteria to leave the intestine, for survival in the environment, and for eventual transmission to a new host. This process might be orchestrated by VrrA that can modulate expression of both a colonization factor (Tcp) and attachment factor (RbmC).

Oligonucleotides
The complete list of DNA oligonucleotides used for cloning and generating probes in hybridization is provided in Table 1.

Bacterial strains and growth conditions
Strains used in this study are listed in Table 2. V. cholerae El Tor Inaba strain A1552 is referred to as the wild-type throughout this study. V. cholerae strains were grown in LB at 37uC or 30uC, as indicated. Carbenicillin was supplemented at 100 mg ml 21 when appropriate.

DNA manipulations
An in-frame deletion of rbmC in A1552 resulting in strain DHS196 was performed using the method described by Skorupski and Taylor [53]. Primer sequences are summarized in Table 1.
The rbmC* allele was introduced into the chromosome of DNY7 (DvrrA) by site-directed mutagenesis, resulting in strain DNY189. The site-directed mutagenesis experiment was performed as previously described [36], with the addition of an intermediate step using strain DNY188. Primers TIS-96 and TIS-97 were used to introduce a nucleotide change (from 221AAGGT to 221AAGCT) into DNY7, resulting in strain DNY188; primers TIS98 and TIS-99 were used to introduce nucleotide changes (from 221AAGCT to 221TTCGT) into DNY188, resulting in strain DNY189. The intermediate strain DNY188 contains an AluI restriction site (AGCT), which allows for mutant screening. Generation of GFPtagged V. cholerae wild-type strain A1552 was performed as described in the earlier studies [16]. A DNA fragment (304 bp) containing the vrrA gene including its putative promoter region was amplified from the A1552 genome and cloned into pBAD18 vector [54]at the EcoR1/Xba1 sites. The resulting plasmid pBAD/vrrA and its vector control (pBAD18) were introduced by transformation into the wild type V. cholerae strain A1552-gfp, resulting WT-gfp/ pBAD and WT-gfp/pBAD-vrrA respectively.
Plasmid pTS2 is a ColE1-based plasmid expressing wild-type VrrA from its own promoter [35]. This plasmid served as template for the construction of plasmids pTS2-M7, pTS2-M8, pTS2-M9 and pTS2-M10 that carry the nucleotide changes shown in Fig. 2B. Procedures were performed as described earlier [55], and primers used to introduce nucleotide change are summarized in Table 2.

SDS-PAGE and Western blot analysis
Protein samples were prepared from equal amounts of bacteria cells after overnight growth at 30uC. Bacteria were harvested by centrifugation at 10,0006g for 10 min at 4uC. The culture supernatant fluid was precipitated with 10% trichloroacetic acid (TCA). Briefly, 1 volume (250 ml) of 50% TCA stock was added to 4 volumes (1 ml) of protein sample. The protein-TCA mixture was kept on ice for 15 min, and subsequently the tube was centrifuged at 15,0006g for 5 min. The supernatant was removed and the protein pellet was washed with 200 ml of cold acetone. Finally, the tube was centrifuged at 15,0006g for 5 min and the resulting pellet was dissolved in sample buffer containing 10% glycerol, 0.05% bromophenol blue, 2% SDS, 5% 2-mercaptoethanol, and 10 mM Tris-HCl, pH 6.8. Proteins with known molecular masses (Fermentas) were used as molecular mass markers. SDS-PAGE and Western blotting were carried out according to the methods of Laemmli [56] and Towbin et al. [57]. HRP-conjugated donkey anti-rabbit IgG (Promega, USA) was used as secondary antibody. Detection was performed using ECL Prime Western Blotting Detection Reagent (Amersham or GE Life Sciences, USA). Pre-stained Protein Ladder (SM0679, Fermentas) was used as size standards. Gels were stained with Coomassie brilliant blue.

RNA isolation and Northern blot analysis
RNA samples were prepared as previously described [36] from bacterial cultures grown overnight (14 hr) at 37uC. The RNA was treated with DNase I and quantified on a NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies, USA). For Northern blot analysis, 10 mg RNA sample was resolved in a polyacrylamide gel and transferred to a Hybond-XL membrane (GE Healthcare, USA) by electro-blotting (1 h, 50 V, 4uC) in a tank blotter. Radiolabeled probes were used to visualize the required mRNA or sRNA. Northern blots were exposed to a phosphorimager screen and scanned on a Storm TM phosphorimager (Molecular Dynamics, USA). Quantification was performed using Quantity One software (Roche, USA). For VrrA and 5S rRNA detection, radio labeled (c-P32-ATP) oligo probe JVO-8109 and JVO-8106 was used respectively.

Biofilm analysis
Flow cell experiments were carried out according to the procedure previously described [58]. Briefly, overnight-grown cultures of gfp-tagged V. cholerae strains were diluted to an optical density at 600 nm (OD 600 ) of 0.02 in 2% LB (0.02% tryptone, 0.01% yeast extract, 1% NaCl; pH 7.5) containing 100 mg/ml of ampicillin and used to inoculated flow chambers. Flow cell experiments were carried out at room temperature with 2% LB containing ampicillin (100 mg/ml) and arabinose (0.2%, wt/vol). CLSM images of the biofilms were captured with a LSM 5 PASCAL system (Zeiss) at 488 nm excitation and 543 nm emission wavelengths. Three dimensional images of the biofilms were reconstructed using Imaris software (Bitplane) and quantified using COMSTAT (Heydorn and Molin, 2000). Flow cell experiments were carried out with at least two biological replicates.