Hfq is essential for Vibrio cholerae virulence and downregulates σE expression



Hfq is an RNA-binding protein that interacts with both small untranslated RNAs (sRNAs) and mRNAs to modulate gene expression post-transcriptionally. In Escherichia coli and Salmonella typhimurium, Hfq is required for efficient expression of the stationary phase sigma factor σS, and consequently is critical for Salmonella virulence. We have found that Hfq is also essential for the virulence of Vibrio cholerae, as strains lacking hfq fail to colonize the suckling mouse intestine. Deletion of the V. cholerae hfq does not prevent production of σS, nor does it prevent expression of TCP, V. cholerae's primary colonization factor. The expression and activity of the alternative sigma factor σE are dramatically increased in a V. cholerae hfq mutant. Comparison of the transcriptome of an hfq mutant with that of an rseA mutant, which also overexpresses σE, revealed that σE controls approximately half the genes found to be upregulated in the hfq mutant. However, increased σE does not appear to account for this strain's reduced virulence. It is likely that sRNAs, in conjunction with Hfq, are critical regulators of V. cholerae pathogenicity.


Hfq is an RNA-binding protein that was first identified as an Escherichia coli protein required for replication of the RNA phage Qβ (Franze de Fernandez et al., 1968). Subsequently, it has been shown to be involved in numerous cellular processes (Tsui et al., 1994; Muffler et al., 1997; Zhang et al., 2003). In E. coli, Hfq binds to at least 35 small, untranslated RNAs (sRNAs) (Zhang et al., 2003). It is thought that these sRNAs, in conjunction with Hfq, regulate mRNA stability and translation (Masse and Gottesman, 2002; Repoila et al., 2003); however, the targets for most of these sRNAs have yet to be identified. Hfq has also been shown to bind directly to mRNAs and thus to influence message stability, polyadenylation and ribosome binding, in part by preventing mRNA degradation by RNase E (Hajnsdorf and Regnier, 2000; Vytvytska et al., 2000; Vecerek et al., 2003). As with regulation mediated by sRNAs, the mechanistic details of these processes are largely uncharacterized.

One of the few well-studied processes influenced by Hfq is the regulation of σS production in E. coli (Repoila et al., 2003). σS (also known as RpoS) is an alternative sigma factor that mediates transcription of many genes expressed during stationary phase and under certain stressful conditions. In E. coli, σS facilitates adaptation to nutrient limitation, as well as resistance to high osmolarity, elevated H2O2, UV light and elevated temperature. Efficient translation of rpoS message depends upon Hfq, as well as upon several sRNAs; consequently, E. coli hfq mutants share many phenotypes with E. coli rpoS mutants, including reduced tolerance to prolonged starvation, oxidative stress and heat shock (Muffler et al., 1997).

Homologues of E. coli hfq have been described in several Gram-negative bacteria, including the pathogens Salmonella typhimurium, Brucella abortus and Yersinia enterocolitica (Nakao et al., 1995; Brown and Elliott, 1996; Robertson and Roop, 1999). In S. typhimurium and B. abortus, Hfq contributes to stress resistance, probably by promoting production of σS or a similar mediator of stationary phase survival, and thus appears to influence bacterial pathogenicity. Hfq also contributes to the virulence of Y. enterocolitica, but the underlying regulatory processes have not been characterized.

Vibrio cholerae, the aetiological agent of the severe diarrhoeal disease cholera, also appears to produce a homologue of Hfq, encoded by vc0347 (E-value = 4e-33, compared wit E. coli ) (Fig. 1). Several genes surrounding hfq are also conserved between V. cholerae and E. coli, including the downstream genes hflX, hflK and hflC, which have been shown in E. coli to be co-transcribed with hfq. An upstream gene, miaA, which contains two of hfq's three promoters in E. coli, is conserved as well.

Figure 1.

Structure of the hfq locus on N16961 chromosome I. The filled block shows the extent of the in frame deletion of hfq, which leaves 22 codons intact. Filled triangles show the sites of plasmid insertions that disrupt hflX, hflK and hflC. The horizontal arrow indicates the chromosomal fragment that was cloned into pMMB67EH to complement the hfq deletion. Percentage similarities were obtained from pairwise blast analyses comparing putative proteins of V. cholerae with E. coli proteins. The number of amino acids that comprise the protein from each species is also presented.

We have found that V. cholerae lacking hfq are highly attenuated in the suckling mouse model of cholera. hfq mutants were more than 1000-fold less effective than wild-type bacteria at colonizing the murine small intestine, despite exhibiting relatively normal growth in vitro. This profound colonization defect does not appear to result from insufficient expression of TCP, V. cholerae's primary colonization factor, or from insufficient production of σS. In addition, a V. cholerae hfq mutant is not hypersensitive to most stresses tested, although it does produce elevated levels of σE, which has been shown to mediate transcription in response to extracytoplasmic stress (Mecsas et al., 1993; Kovacikova and Skorupski, 2002). Overexpression of σE appears to account for the changes in expression of ≈ 50% of the genes with upregulated transcript levels in hfq V. cholerae. Thus, it appears that the pathways regulated by Hfq in V. cholerae are distinct from those known to be critical to other bacterial pathogens.


Hfq is required for V. cholerae intestinal colonization

We used suckling mice, a well-established model host for study of V. cholerae pathogenesis, to compare the intraintestinal growth of wild-type V. cholerae and V. cholerae lacking hfq. Overnight cultures of NHfq (a derivative of the El Tor V. cholerae strain N16961 with an in frame deletion of 65 of hfq's 87 codons) and NLAC (N16961ΔlacZ; Ding and Waldor, 2003) were mixed, then intragastrically inoculated into 5-day-old CD-1 mice. The next day, the number of NLAC and NHfq cfu in the small intestine of each suckling mouse was determined. The input ratio of the two strains was typically close to 1:1; however, after ≈ 22 h within the small intestine, the average ratio of NHfq:NLAC was ≈ 0.0001, and no NHfq were recovered from five of 11 mice (Fig. 2A). In contrast, the median competitive index (CI; output ratio/input ratio) for NHfq in vitro was ≈ 0.2. These data indicate that the absence of Hfq is far more significant for growth in vivo than in vitro. They suggest that a specific requirement for Hfq in vivo, rather than a general growth defect, accounts for the dramatic reduction in intestinal colonization by NHfq.

Figure 2.

Growth of NHfq and control strains in the suckling mouse intestine.
A. Competitive indices from in vivo and in vitro competition assays. NHfq, NHflX, NHflK and NhflC were all compared with NLAC (wt). Competitive index is defined as mutantoutput/wtoutput ÷ mutantinput/wtinput; thus, a competitive index of < 1.0 indicates a strain with a competitive disadvantage. Open circles denote the limit of detection for experiments in which no cfu from a mutant strain were recovered; these ratios were calculated as if 1 cfu was recovered.
B. Colonization by NHfq, NLAC and NTCP inoculated singly. The open circles represent the limit of detection (≈ 700 cfu) for experiments in which no cfu were recovered. For all experiments, the inocula were ≈ 105−106 cfu. Data for each strain come from two or three independent experiments. Horizontal bars show median values.

Based on Hfq's role in other bacterial species, we hypothesized that survival or growth of NHfq might be impaired after reaching stationary phase, and that this might contribute to NHfq's low CI in intestinal colonization assays. To test this possibility, we performed additional in vivo competition assays, using as input NHfq and NLAC grown to log phase (OD600 = 0.4–0.6) rather than stationary phase. As in the previous experiments, we found that NHfq was dramatically outcompeted by the wild-type strain in vivo (median CI < 0.0001), and that it was far more impaired in vivo than in vitro (median CI = 0.16). In this set of experiments, colonization by NHfq was undetectable in a majority of the mice (six out of nine; Fig. 2A). Thus, NHfq's inability to colonize infant mice is not dependent upon the growth phase of the strain at the time of inoculation.

Colonization by NHfq inoculated alone (without NLAC) was also assayed in suckling mice. We found that NHfq was significantly less able than NLAC to persist and multiply in the small intestine even in the absence of a competitor strain (Fig. 2B). The median number of NHfq cfu recovered was less than 1050 per mouse, more than 5000-fold lower than for NLAC (median = 5.8 × 106 cfu) and close to the limit of detection (700 cfu per mouse). No colonies were obtained from four of 10 mice tested. NHfq was found to be almost as attenuated as NTCP, which cannot produce V. cholerae's primary colonization factor, the toxin co-regulated pilus (TCP; Fig. 2B).

A variety of experiments was performed to confirm that NHfq's phenotype is the result of an Hfq deficiency, and not an unexpected polar effect upon expression of the downstream genes or a secondary mutation. First, we independently disrupted each of the three downstream genes and assessed the virulence of these new mutants using in vivo competition assays. We found that the CI for each strain was near 1.0, and thus that hflX, hflK and hflC are not required for growth in vivo (Fig. 2A). We also tested additional strains with in frame deletions within hfq. Several independently derived strains of N16961Δhfq were as attenuated as NHfq in the suckling mouse colonization assay (data not shown). We were not able to restore the ability of NHfq to colonize the infant mouse by providing hfq on a plasmid (pMHfq); however, pMHfq was poorly maintained by NHfq in the absence of selection, both in culture and during intestinal growth (data not shown). Overall, these observations indicate that Hfq is critical for colonization of the small intestine and that the primary factor underlying NHfq's attenuation in vivo is the absence of Hfq.

Hfq is not required for production of TCP

Relatively few factors have been identified that have as profound and specific an effect on V. cholerae growth in vivo as Hfq. Of these, the majority impair expression of TCP, generally by blocking expression of key transcriptional activators, such as ToxR, ToxT, TcpPH and AphAB (Miller et al., 1987; DiRita et al., 1991; Hase and Mekalanos, 1998; Kovacikova and Skorupski, 1999). These factors all control expression of the ToxR regulon, which includes both TCP and cholera toxin, another key virulence factor. To determine whether deletion of hfq disrupts the ToxR regulon in NHfq, we monitored production of TcpA (the major pilin subunit) and cholera toxin by this strain. Western blot analysis showed that NHfq actually produced slightly more TcpA than did a wild-type strain, indicating that Hfq is not required for expression of tcpA(Fig. 3A). In addition, NHfq could be infected by CTXφ, the filamentous phage whose receptor is TCP, at a frequency comparable to or greater than that for wild-type V. cholerae, demonstrating that functional pili are assembled by NHfq (data not shown). These data suggest that the failure of NHfq to colonize infant mice is not the result of insufficient production of TCP. We also found that NHfq secretes at least as much cholera toxin as does wild-type N16961 (Fig. 3A). Thus, it appears that deletion of hfq does not block activation of the ToxR regulon in N16961.

Figure 3.

Western blot analyses of TcpA, cholera toxin and the alternative sigma factors σS and σE. Cells were grown under AKI conditions (ToxR-inducing) for assays of TCP and cholera toxin and in LB for other assays.
A. Immunoblots probed with antisera to TcpA and cholera toxin. TcpA was detected in total cell extracts, and CT was assayed in culture supernatants.
B. Immunoblot of total-cell extracts from stationary phase cultures probed with antisera to E. coliσS. σS (the upper band) migrated more slowly on denaturing gels than predicted from its molecular weight (MW), as has been observed previously. The lower band, present in all lanes, is a cross-reacting protein.
C and D. Immunoblots probed with antisera to E. coliσE. Total-cell extracts were generated from stationary phase (stat) and exponential phase (log) cultures of wild-type V. cholerae (N16961) and several derivatives, including hfq (NHfq), hfq(pMHfq) (NHfqP), rpoS (NRpoS), rpoE (NRpoE) and rseA (NRseA-2).

Characterization of NHfq growth in vitro

Comparison of the results from in vivo and in vitro competition assays suggests that the contribution of Hfq to V. cholerae growth is more significant in vivo than in vitro; nonetheless, it is clear from the in vitro competition assays that Hfq does play a role in growth in vitro, at least in rich media. To understand better the results from the in vitro competition assays, we characterized growth of NHfq under a variety of conditions. We found that the growth rate of NHfq during exponential phase is slightly slower than that of N16961 (Fig. 4A and B), and that cultures of NHfq reach a lower density (OD600) at saturation than the wild-type strain, as has also been reported for E. coli hfq mutants (Tsui et al., 1994). Both these phenotypes are at least partially prevented by introduction of pMHfq into NHfq, and both are likely to contribute to NHfq's reduced CI in vitro. Interestingly, the slower growth of NHfq was not markedly exacerbated in minimal media, showing that NHfq is not an auxotroph and suggesting that the poor growth of NHfq in vivo is not simply due to a nutrient deficiency.

Figure 4.

Growth of N16961, NHfq and NHfq(pMHfq) in (A) LB and (B) M63 minimal media. Filled diamonds, N16961 (wt); open circles, NHfq; open boxes, NHfq(pMHfq).

σS production and σS-dependent responses in hfq V. cholerae

As Hfq is required for efficient production of σS in E. coli, we assessed the sensitivity of NHfq to a variety of stressful conditions known to induce a σS-dependent response in E. coli and/or V. cholerae (Yildiz and Schoolnik, 1998). NHfq did not exhibit increased sensitivity to ultraviolet irradiation (1000–5000 µJ) or oxidative stress (2 mM H2O2), and its growth rate was not appreciably reduced in hyperosmotic media (LB + 0.5 M NaCl; data not shown). In addition, NHfq did not consistently differ from wild-type V. cholerae in assays of long-term starvation survival (although it should be noted that this assay yielded highly variable results). These results suggested the possibility that, unlike in E. coli, Hfq is not needed for efficient production of σS in V. cholerae. Consequently, we performed more direct analyses of σS production in NHfq. Western blot analyses of stationary phase cultures revealed no difference between the level of σS in wild-type and NHfq V. cholerae (Fig. 3B). In addition, the activity of an rpoS::lacZ translational fusion was not reduced in NHfq (data not shown). These data provide additional evidence that Hfq is not a critical regulator of σS synthesis in V. cholerae.

Transcriptome of V. cholerae lacking Hfq

To gain insight into the range of genes with expression that is influenced by Hfq in V. cholerae, we used DNA microarrays to compare the transcriptome of NHfq with wild-type V. cholerae N16961. Probes were made corresponding to RNA transcribed during log phase growth in minimal media, as this was found to yield the most reproducible results. The transcript levels for numerous genes were found to be changed, consistent with the idea that Hfq contributes to the regulation of numerous cellular pathways. Of the roughly 4000 genes in V. cholerae, we found that expression of 224 differed reproducibly by at least 1.5× between N16961 and NHfq (Table 1 and Supplementary material, TableS1). Northern blot analysis indicated that expression of at least one of these genes (vc0177) is entirely dependent upon Hfq. Upregulated genes were almost twice as common as downregulated genes in NHfq (144 versus 80 genes), suggesting that Hfq may function more frequently as a repressor than as an activator. However, it is worth noting that many of the genes with altered expression may be indirectly, rather than directly, controlled by Hfq; thus, the predominance of genes with elevated transcript levels in NHfq could be more indicative of the roles of genes downstream of Hfq than a direct effect of Hfq itself. Consistent with Western blots showing that NHfq produces TcpA and cholera toxin, we did not detect any significant changes in expression of genes encoding the transcriptional regulators ToxR and ToxT.

Table 1. Relative expression of selected Hfq-regulated genes in NHfq, NRseA and NrpoE.
  Mutant straina
  • a

    . Ratio of transcript abundance in the mutant strain relative to the wild-type strain, N16961. Data shown were generated from cultures grown in M63 + glucose to OD600≈ 0.3. All genes selected had ratios of either > 1.5 or < 0.66 for NHfq, with a P-value of < 0.1, at both OD600≈ 0.3 and OD600≈ 0.6 (not shown). Not all genes meeting these criteria are listed. Four independent experiments were performed for NHfq for each condition, and two independent experiments were performed for NRseA and NRpoE.

  • b

    . Genes likely to be positively regulated, either directly or indirectly, by RpoE.

  • c

    . The rseA mutation disrupts rseA expression and is likely to have a polar effect on expression of rseB and rseC; thus, positive regulation by RpoE is not apparent.

Upregulated genes
 VC0034bThiol:disulphide interchange protein1.682.740.30
 VC0150bRNA polymerase sigma-32 factor2.092.880.99
 VC0475 irgA, enterobactin receptor2.280.442.40
 VC0554bProtease, insulinase family/protease, insulinase family2.953.970.67
 VC0566bProtease DO2.254.270.17
 VC0972bPorin, putative2.281.450.75
 VC1045bRNA polymerase sigma70 factor, ECF subfamily2.163.800.56
 VC1450RTX toxin activating protein3.820.720.98
 VC1451aRTX toxin RtxA, bp 1–34202.680.801.09
 VC1451bRTX toxin RtxA, bp 3421–68402.950.580.95
 VC1565bOuter membrane protein TolC, putative2.893.200.52
 VC1676Phage shock protein C2.913.296.00
 VC1677Phage shock protein B2.152.203.84
 VC1678Phage shock protein A3.672.796.06
 VC1854Porin, putative3.100.558.57
 VC1902bDisulphide bond formation protein B2.154.210.40
 VC1918bPeptidyl-prolyl cis-trans isomerase D1.691.980.64
 VC1987bOuter membrane lipoprotein Slp, putative1.962.340.61
 VC2213Outer membrane protein OmpA2.351.003.22
 VC2464cSigmaE factor regulatory protein RseC1.570.210.26
 VC2465cSigmaE factor regulatory protein RseB1.600.100.13
 VC2466cSigmaE factor negative regulatory protein RseA2.020.120.11
 VC2467bRNA polymerase sigmaE factor1.622.420.36
 VC2701bThiol:disulphide interchange protein DsbD2.543.560.55
 VC2732bGeneral secretion pathway protein E1.673.050.77
 VC2734bGeneral secretion pathway protein C2.043.390.63
 VCA0059bMajor outer membrane lipoprotein1.503.390.63
 VCA0588bPeptide ABC transporter, ATP-binding protein, putative2.114.180.43
 VCA0590bPeptide ABC transporter, permease protein, putative1.984.350.47
 VCA0591bPeptide ABC transporter, periplasmic peptide-binding protein, putative2.164.320.39
Downregulated genes    
 VC0177Hypothetical protein0.050.690.88
 VC0347Host factor-I, putative (Hfq)
 VC0622Sensory box sensor histidine kinase/response regulator0.570.810.76
 VC1050Response regulator0.640.690.78
 VC1082Response regulator0.550.530.77
 VC1084Sensory box sensor histidine kinase0.550.640.78
 VC1085Sensor histidine kinase0.420.650.71
 VC1086Response regulator0.580.750.77
 VC2201Chemotaxis protein methyltransferase CheR0.630.650.84
 VCA0895Chemotactic transducer-related protein0.500.690.81
 VCA1091Chemotaxis protein methyltransferase CheR0.520.750.69
 VCA1096Chemotaxis protein CheY0.380.500.73

Several categories of genes differed notably in prevalence among upregulated and downregulated transcripts (Table 1). For example, chemotaxis-related genes were more common among downregulated genes (seven out of 80) than among upregulated genes (three out of 144). In particular, two of V. cholerae's four forms of cheR were downregulated, as was one of four functional cheYs, whereas no genes encoding CheR or CheY were upregulated. Similarly, transcripts for sensor histidine kinases and response regulators were also notably deficient in NHfq. Six genes encoding sensor histidine kinases and/or response regulators were downregulated in NHfq, whereas none was upregulated. In contrast, eight genes encoding porins and other outer membrane proteins (OMPs) were upregulated in NHfq, whereas none was downregulated. In addition, several genes activated in response to outer membrane and/or periplasmic stresses had more abundant transcripts in the hfq mutant. Interestingly, transcripts for the alternative sigma factor σE (also known as RpoE), which has been shown to control expression of several of the upregulated genes in E. coli (Dartigalongue et al., 2001), were also elevated in NHfq.

NHfq produces increased σE

Based on the transcriptome analyses, we assessed whether σE protein and activity, as well as rpoE transcripts, are increased in NHfq. On Western blots probed with antisera generated against E. coliσE, which is 90% similar to V. choleraeσE, we found that NHfq contains elevated levels of σE during stationary phase and log phase growth in both rich and minimal media (Fig. 3C; data not shown). σE was restored to normal levels in NHfq by the hfq-complementing plasmid, pMHfq. These data suggest that Hfq either directly or indirectly regulates production of σE in V. cholerae. In addition, we found that β-galactosidase activity from an rpoE::lacZ transcriptional reporter, which is largely dependent upon σE for expression (Kovacikova and Skorupski, 2002), was increased at least twofold in the hfq strain background. The reporter yielded 64 and 148 Miller units in log phase cultures of the hfq and wild-type strains, respectively, and 154 and 369 Miller units in stationary phase cultures. These results provide additional evidence that σE activity is elevated in NHfq.

Identification of σE-dependent genes in V. cholerae

In E. coli, σE activity is regulated by an antisigma factor, RseA, which is encoded downstream of rpoE (De Las Penas et al., 1997; Missiakas et al., 1997). E. coli strains lacking rseA have increased σE activity and increased expression of σE-dependent genes. V. cholerae has an rseA homologue (59% similarity), and we have found that disruption of rseA results in increased σE levels in V. cholerae as well (Fig. 3D). We used microarray analyses to identify genes with transcripts that are more abundant in an rseA mutant (NRseA) than in wild-type V. cholerae. By comparing this list of genes with those found to be upregulated in NHfq, we could determine which changes in the NHfq transcriptome are likely to result from overexpression of σE. Before this analysis, rpoE was the only gene in V. cholerae known to be controlled by σE (Kovacikova and Skorupski, 2002).

We found that there is extensive overlap between the set of genes with increased transcript levels in NHfq and those with increased abundance in NRseA. Of 144 genes found to have at least 1.5× higher transcript levels in NHfq, 71 also had increased transcripts in NRseA (Tables 1 and S1). This similarity does not appear to result from reduced expression of Hfq in NRseA, as hfq transcripts were actually slightly elevated in NRseA. The probability of finding such similar sets of genes for NHfq and NRseA was 4.1 × 10−30. For the majority of genes present in both lists, reduced transcript levels were detected in an rpoE mutant, providing additional evidence that expression of these genes is controlled by σE. Furthermore, several genes with homologues that have been found to be regulated by σE in E. coli were identified, including rpoE itself (vc2467 ), rpoH (vc0150), RNA polymerase σ70 (vc1045), a peptidyl-prolyl isomerase (vc1918) and several thiol:disulphide interchange proteins (vc0034 and vc2701) (Dartigalongue et al., 2001). Interestingly, two genes encoding components of V. cholerae's type II system for extracellular protein secretion (eps) were also upregulated in both NHfq and NRseA (vc2732 and vc2734). Additional genes of the eps gene cluster were found to have significantly increased transcripts only in NRseA (data not shown). Finally, numerous hypothetical and conserved hypothetical proteins had increased transcript levels in both strains. The fact that 49% of the genes with increased expression in NHfq were also increased in NRseA suggests that the elevated level of σE in NHfq accounts for a significant portion of the transcriptional alterations in this strain.

Overexpression of σE does not inhibit intestinal colonization

To test whether overexpression of σE and the genes that it regulates contributes to NHfq's reduced intestinal colonization, we assessed intestinal colonization by several independently derived strains with insertion mutations in rseA, as these strains also produce excess σE. This approach seemed to be preferable to plasmid-based overexpression of σE, as plasmid loss can present a significant problem in the absence of selection, as in the intestine. We tested several strains because we found that the colony morphology of mutants with rseA insertions was variable. It appears that mutation of rseA is detrimental to V. cholerae, and that suppressor mutations that enable more rapid growth arise within the original mutants. We tested both primary mutants and some of the variants that arose from them (strains shown by Western blotting still to overproduce σE) by performing in vivo competition assays between individual rseA mutants and NLAC. These competition assays yielded variable results, presumably because of heterogeneity in development of suppressor mutations, which probably continue to arise during the experiment. However, four out of five strains with mutations in rseA had a CI of at least 1.0 in the suckling mouse model of colonization (data not shown). These data suggest that overproduction of σE does not impair colonization of the suckling mouse small intestine by V. cholerae, and that overexpression of σE by NHfq does not account for this strain's colonization deficit.


We have found that the RNA-binding protein Hfq is essential for the virulence of V. cholerae in the suckling mouse model of cholera pathogenesis. V. cholerae lacking hfq cannot colonize the murine small intestine, despite retaining the ability to produce V. cholerae's principal colonization factor, TCP. These bacteria display a slight growth defect in vitro; however, this phenotype does not appear to be sufficient to account for their severe deficiency in vivo. Hfq's role in cholera pathogenesis appears to differ from its contribution to the virulence of S. typhimurium and B. abortus. In these pathogens, as in E. coli, Hfq is required for efficient production of σS and/or for stress resistance during stationary phase. In contrast, deletion of hfq in V. cholerae does not impair production of σS, and mutations within hfq and rpoS induce distinct phenotypes. In particular, V. cholerae hfq mutants are far more attenuated in growth in the suckling mouse intestine than are rpoS mutants, which are at most minimally impaired (Yildiz and Schoolnik, 1998; Merrell et al., 2000).

Our findings suggest that Hfq's (and most probably sRNAs’) contribution to control of σS differs between E. coli and V. cholerae. In E. coli, Hfq-dependent control of σS abundance requires at least three sRNAs (OxyS, DsrA and RprA), each of which modulates σS translation in response to specific stimuli. Sequence analyses of the V. cholerae genome did not detect homologues of these RNA-encoding genes (Hershberg et al., 2003); thus, it is not surprising that alternative pathways appear to govern production of σS in V. cholerae.

The pathways regulating production of σE in V. cholerae may also differ from those of other organisms, as a connection between Hfq and σE expression has not been noted previously. In V. cholerae, rpoE is known to be autoregulated, but little is known about additional stimuli that regulate it or additional genes that are controlled by it (Kovacikova and Skorupski, 2002). In E. coli, rpoE mediates the cellular response to extracytoplasmic stresses such as overexpressed or misfolded proteins within the periplasm, and it can promote survival after exposure to EtOH or elevated temperature. It is not yet clear why σE protein and activity are elevated in V. cholerae lacking hfq. It could be that Hfq influences rpoE message translation directly, perhaps in conjunction with an sRNA. It is also theoretically possible that increased production of porins and other OMPs by the V. cholerae hfq mutant induces σE expression. However, most of the OMPs with increased transcripts in NHFq also had increased transcripts in NRseA, suggesting that their transcription is regulated by, rather than a regulator of, σE. A few OMP-encoding genes were found to be upregulated only in the hfq mutant (e.g. vc1854, vc2213 and vca1028), but double mutants lacking both Hfq and one of these OMPs still overproduced σE (not shown). Thus, it appears unlikely that increased OMP production underlies the elevated expression of σE in NHfq.

Comparison of NHfq with NRseA, which also produces excess σE, suggests that many of the genes with increased transcripts in NHfq are likely to be regulated by σE. These genes include several known to be controlled by σE in E. coli, but also a number of genes not previously linked to σE, including a putative transcriptional regulator (vc1955), genes of the eps gene cluster (vc2732 and vc2734) and a cluster encoding a putative peptide ABC transporter (vca0588, vca0590 and vca0591). Interestingly, although there is significant overlap between the upregulated genes in NHfq and NRseA, most of the genes with increased transcripts in NRseA did not have increased transcript levels in NHfq (368 out of 439). One possible explanation for this finding is that regulation of these genes is modulated not only by σE but also by Hfq, presumably acting downstream of σE in the regulatory pathways, either alone or in conjunction with an sRNA. Hfq might contribute to post-transcriptional regulation of these genes (e.g. the eps genes vc2723–vc2731 and vc2733), perhaps to facilitate differential expression of genes within a single operon. Alternatively, it is possible that RseA has a cellular role in addition to regulation of σE, and that disruption of this function accounts for some of the RseA-specific changes in gene expression.

It is unlikely that overexpression of σE contributes to the in vivo attenuation of hfq mutants. Previous studies have shown that V. cholerae lacking rpoE colonize suckling mice less efficiently than do wild-type V. cholerae (CI ≈ 0.03) (Kovacikova and Skorupski, 2002), and it seemed possible that inappropriate transcription of genes within the σE regulon might also be detrimental in vivo. However, several strains that overproduce σE as a result of insertional inactivation of rseA were not attenuated in vivo. In addition, we have noticed that a V. cholerae rpoS mutant also overproduces σE, during both stationary phase and exponential growth (data not shown), and rpoS mutants have at most a fivefold deficit in colonization (Yildiz and Schoolnik, 1998; Merrell et al., 2000). Together, these data strongly suggest that overexpression of σE-regulated genes does not account for the in vivo phenotype of NHfq.

Inadequate production of V. cholerae's primary colonization factor, TCP, also does not appear to account for NHfq's colonization deficit. In fact, NHfq appears to produce slightly more TCP than a wild-type strain. Increased expression of the ToxR regulon does not prevent colonization by numerous strains of V. cholerae that are unable to shut down this regulon at stationary phase (e.g. hapR mutants, such as N16961 and O395; Zhu et al., 2002); thus, we think it is unlikely that the small increase in TCP production in NHfq is detrimental. It is not clear whether Hfq influences TCP production directly in NHfq or whether it acts through upstream regulators such as ToxT; however, microarray analyses did not detect significant changes in toxT transcript levels. Interestingly, it has been noted recently that, in hapR+ strains of V. cholerae, in which quorum sensing allows for downregulation of virulence genes, Hfq acts indirectly as a positive regulator of virulence gene expression (B. Bassler, personal communication). Thus, Hfq appears to contribute in several distinct ways to the virulence of V. cholerae. As Hfq often acts in conjunction with sRNAs, it is likely that sRNAs will also be found to play major roles in the control of cholera pathogenesis.

Experimental procedures

Bacterial strains and growth conditions

All the V. cholerae strains used in this study are derivatives of the V. cholerae El Tor genome strain, N16961 (Heidelberg et al., 2000). NTCP is N16961 tcpA::mTn5 (Merrell et al., 2002). For microarray studies, cells were grown in M63 minimal medium supplemented with 0.2% glucose and 1 mM MgSO4 (Ausubel et al., 1995). For analysis of TCP production, cells were grown in AKI medium as described previously (Iwanaga et al., 1986); otherwise, cells were grown in LB broth at 37°C. Antibiotics were used at the following concentrations: streptomycin, 200 µg ml−1; ampicillin, 100 µg ml−1; kanamycin, 50 µg ml−1. Xgal was used in LB agar at 40 µg ml−1.

Strain and plasmid construction

NHfq, an N16961 derivative with all but the first 12 and last 10 codons of hfq (vc0347) deleted, was generated by allele exchange using plasmid pCHfq, a pCVD442 derivative, as described previously (Donnenberg and Kaper, 1991). The insert in this plasmid was constructed using splicing by overlap extension polymerase chain reaction (SOE PCR) and the primers 5′-ACTGATTTATCGAGGGATGG-3′ with 5′-CGATCCAGAAATGGGTCTTGTAGAGATTGCC-3′; and 5′-CCCATTTCTGGATCGTCCAGCAGAGAAGTCT-3′ with 5′-TTACGCAAAGTAGGATCGAG-3′ (Horton et al., 1989). The pir-dependent suicide vector pGP704 (ApR; Miller and Mekalanos, 1988) was used to inactivate insertionally hflX, hflK and hflC (vc0348–vc0350) in N16961, generating NHflX, NHflK and NHflC. Internal fragments of each gene, which were generated by PCR using primers 5′-AGCGAGCCGTACTT GTTCA-3′ and 5′-GCCTTCATCATCCCGTT-3′ for hflX, 5′-GCGTGGAATGAGCCTGGTA-3′ and 5′-CGGGTTACTTTCG GTGC-3′ for hflK, and 5′-AATCCCTAGTATCGTTCTGATC-3′ and 5′-AAATCTTCGCGGCTTCT-3′ for hflC, were cloned into pGP704, and the resulting plasmids were transferred by conjugation from SM10λpir into N16961. A similar strategy was used for disruption of rpoE (vc2467) and rseA (vc2466). The rpoE mutant, NRpoE, was generated by insertion of a pGP704 derivative containing an rpoE internal fragment amplified with 5′-AAAGTGTGCAACCTTATTTCCCG-3′ and 5′-GAACGCACCGTACCAACAGG-3′. Several rseA mutants (NRseA, NRseA-2 and others not shown) were generated by insertion of pGP704 containing rseA-linked sequences amplified with 5′-GCTCTAGATTGCGGAAGTGATGGATTGC-3′ and 5′-GCTCTAGAGACCAAACTGAGACAACCAAGCAG-3′. All mutants constructed for this study were verified by PCR and/or Southern analysis.

The entire 261 bp hfq coding sequence and 182 bp of upstream sequence were amplified from N16961 chromosomal DNA using the primers 5′-GCTGTGTATAATGTGAT CGC-3′ and 5′-GCAAAGAATTACTCTTCAGA-3′ and cloned into pMMB67EH (Furste et al., 1986) to generate pMHfq. Expression of hfq from pMHfq is driven by the plasmid-encoded promoter Ptac and perhaps by a native hfq promoter (Tsui et al., 1996). Cells containing pMHfq were grown in the presence of ampicillin (to promote plasmid maintenance) unless otherwise noted.

The rpoE::lacZ transcriptional reporter consists of the rpoE promoter region, amplified from N16961 using primers 5′-ATCACTTCCGCATCACCTGAATTGCTTACA-3′ and 5′-GCTCTAGAATTGAGCAGACATGCCACAGTG-3′, cloned into the vector pCB182 (Schneider and Beck, 1986). Its activity was assessed in lacZ derivatives of N16961 and NHfq using standard protocols (Miller, 1992). Assays were performed in duplicate.

Intestinal colonization assays

Five-day-old CD-1 mice were used for intestinal colonization assays essentially as described previously (Taylor et al., 1987). Stationary phase cells were prepared by overnight growth in LB broth at 30°C with rotation. Exponential phase cells were prepared by diluting overnight cultures 1:1000 into LB broth and incubating at 37°C with shaking until an OD600 of 0.4–0.6. After dilution (1:1000 for stationary phase cultures, 1:100 for exponential phase cultures), mutant and wild-type cells were mixed 1:1 to generate the inocula (105−106 cfu) for competition assays or used without mixing for single strain colonization assays. For in vitro competition assays, 30 µl of the input culture was inoculated into 3 ml of LB broth and grown for ≈ 22 h at 37°C with rotation.

Western blot analyses

Cell pellets from cultures grown in AKI (for TCP) or LB broth (for σE and σS) were normalized based on culture OD600, then lysed in 1× LDS sample buffer (Invitrogen) and boiled. Cell extracts and supernatants were run on 10% NuPAGE acrylamide gels (Invitrogen), then transferred to nitrocellulose and probed with polyclonal antisera. Horseradish peroxide-conjugated anti-rabbit antisera (Pierce) and SuperSignal West Pico chemiluminescent substrate kit (Pierce) were used for detection of bound primary antibodies.

Cholera toxin was measured in filter-sterilized supernatants from overnight cultures grown in AKI medium. Supernatant proteins were precipitated in 90% saturated ammonium sulphate solution and pelleted at 20 000 g for 20 min. Protein pellets were resuspended in H2O, dialysed against H2O in Slide-A-Lyzer dialysis cassettes with a 10 kDa molecular weight cutoff (Pierce) and analysed on Western blots as above.

Microarray analyses

Arrays.  Each microarray consisted of ≈ 4000 DNA oligonucleotides spotted onto UltraGAPS coated slides (Corning) using a Biorobotics MicroGrid II. Oligonucleotides were 70-mers designed to correspond to unique sequences within V. cholerae open reading frames (Illumina), plus additional control sequences. DNA was cross-linked to slides immediately before prehybridization using a Stratalinker.

Probes.  Cell pellets were isolated from cultures grown in M63 at 37°C to an OD600 of ≈ 0.3 and preserved with the RNA stabilization reagent RNAlater (Qiagen). RNA was extracted from cell pellets using the RNeasy mini kit and RNase-free DNase set (Qiagen). RNA (10 µg) was reverse transcribed in the presence of 0.5 mM dATP, dCTP and dGTP, 0.2 mM dTTP and 0.3 mM aminoallyl dUTP. After reverse transcription, RNA was hydrolysed with NaOH, and cDNAs were purified on Miniprep Spin Prep columns (Qiagen). cDNAs were coupled to monoreactive Cy3 and Cy5 (Amersham), then repurified, dried down and resuspended in hybridization buffer (25% formamide, 5× SSC, 0.1% SDS, 0.1 mg ml−1 salmon sperm DNA, 0.2 mg ml−1 yeast tRNA). Probes were denatured at 95°C before hybridization at 42°C.

Hybridization.  Slides were prehybridized at 42°C in 5× SSC, 0.1% SDS, 1% BSA, then briefly rinsed with water and isopropanol. Each array was then hybridized simultaneously to two differentially labelled cDNA probes corresponding to RNA from wild-type and mutant strains. After hybridization overnight at 42°C, slides were washed sequentially with (i) 2× SSC, 0.1% SDS; (ii) 0.1× SSC, 0.1% SDS; and (iii) 0.1× SSC.

Analysis.  Slides were scanned using a Packard Scanarray 4000 with the PMT adjusted so that fluorescence from Cy3- and Cy5-labelled probes was balanced. Scans were analysed with quantarray (Packard) to calculate spot intensity and background, and the quantarray output was subjected to further analyses using genespring (Silicon Genetics).

All experimental replicates used probes generated from independent RNA samples, and dye swaps were performed for all comparisons. For comparison of the N16961 and NHfq transcriptomes, four independently generated probe sets were made from cultures grown to both OD600≈ 0.3 and OD600≈ 0.6. Genes were counted as differentially regulated if the ratio of mutant/wild-type transcripts was either > 1.5 or < 0.66 with P < 0.1 under both conditions. Comparisons of N16961 with NRpoE and NRseA were each performed with two independent probe sets.


We thank C. Gross, A. C. Matin, S. Lynch, R. Taylor and J. Mekalanos for supplying antisera, Ron Taylor for co-ordination of array oligonucleotide production, and Lan Wei of the T-NEMC Expression Array Core for technical assistance. We thank C. Gross, S. McLeod and P. Budde for helpful discussions and/or critical reading of this manuscript, and the NEMC GRASP Center for preparation of plates and media. This work was funded by NIH grants AI42347 to M.K.W. and P30DK-34928 to the NEMC GRASP Digestive Center, and by the Howard Hughes Medical Institute.

Supplementary material

The following material is available from http://www.blackwellpublishing.com/products/journals/suppmat/mmi/mmi4142/mmi4142sm.htm

Table S1. Transcript abundance in the mutant strain relative to N16961.