The novel σ54- and σ28-dependent flagellar gene transcription hierarchy of Vibrio cholerae



The human pathogen Vibrio cholerae is a highly motile organism by virtue of a polar flagellum. Flagellar transcriptional regulatory factors have been demonstrated to contribute to V. cholerae virulence, but the role these factors play in the transcription hierarchy controlling flagellar synthesis has been unclear. The flagellar genes revealed by the V. cholerae genome sequence are located in three large clusters, with the exception of the motor genes, which are found in three additional locations. It had previously been demonstrated that the alternative sigma factor σ54 and the σ54-dependent activators FlrA and FlrC are necessary for flagellar synthesis. The V. cholerae genome sequence revealed the presence of a fliA gene, which is predicted to encode the alternative flagellar sigma factor σ28. A V. choleraeΔfliA mutant strain is non-motile, and synthesizes a truncated flagellum. Vibrio cholerae FliA complements both V. cholerae and Salmonella typhimurium fliA mutants for motility, consistent with its function as an alternative flagellar sigma factor. Analysis of lacZ transcriptional fusions of the V. cholerae flagellar promoters in both V. cholerae and S. typhimurium identified σ28-, σ54-, FlrA- and FlrC-dependent promoters, as well as promoters that were independent of all these factors. Our results support a model of V. cholerae flagellar gene transcription as a novel hierarchy composed of four classes of genes. Class I is composed solely of the gene encoding the σ54-dependent activator FlrA, which along with the σ54-holoenzyme form of RNA polymerase activates expression of Class II genes. These genes include structural components of the MS ring, switch and export apparatus, as well as the genes encoding both FliA and FlrC. FlrC, along with σ54-holoenzyme, activates expression of Class III genes, which include basal body, hook and filament genes. Finally, σ28-holoenzyme activates expression of Class IV genes, which include additional filament genes as well as motor genes. Thus, this novel V. cholerae flagellar hierarchy has incorporated elements from both the σ54-dependent Caulobacter crescentus polar flagellar hierarchy and the σ28-dependent S. typhimurium peritrichous flagellar hierarchy.


Vibrio cholerae is a Gram-negative bacterium that is motile by means of a single polar flagellum. Vibrio cholerae are marine organisms that persist in both salt and freshwater environments. This bacterium also causes the life-threatening and epidemic form of diarrhoea known as cholera, after entering the human host following the ingestion of contaminated food or water. Within the intestine, V. cholerae expresses a number of virulence factors including cholera toxin (CT) and toxin co-regulated pilus (TCP) (Lospalluto and Finkelstein, 1972; Mekalanos et al., 1983; Taylor et al., 1987).

Motility has been identified as a virulence determinant of V. cholerae. Non-motile mutants have been shown to cause less fluid accumulation in rabbit-ligated ileal loops and less disease in the rabbit RITARD model (Richardson, 1991), and non-motile mutants are also defective for adherence to isolated rabbit brush borders (Freter and O'Brien, 1981). Non-motile mutants of live attenuated V. cholerae vaccines also show reduced reactogenicity in humans (Coster et al., 1995; Kenner et al., 1995). Oddly, in the widely used infant mouse model system, non-motile mutants show no significant defect in their ability to colonize the small intestine in competition assays (Richardson, 1991; Gardel and Mekalanos, 1996). Still, genetic evidence suggests that V. cholerae motility and virulence gene expression are inversely related, namely that expression of motility genes may repress virulence genes, and vice versa (Gardel and Mekalanos, 1996). A Na+ gradient across the membrane drives V. cholerae flagellar rotation (Kojima et al., 1999), and also regulates transcription of the toxT gene (Hase and Mekalanos, 1999), which is required for CT and TCP expression, thus illuminating a possible mechanism for coupling virulence factor expression to motility. The exact connection between motility and virulence gene expression has remained elusive, in part because virtually nothing is known about regulation of V. cholerae motility.

In its natural marine environment, V. cholerae are probably found in biofilms, as these organisms readily form biofilms within the laboratory (Yildiz and Schoolnik, 1999). Motility plays an important role in the development of V. cholerae biofilms. Motility is critical for the initial stage of biofilm development, apparently by mediating approach and attachment to the abiotic surface (Watnick and Kolter, 1999). However, within a mature biofilm the cells are found within large multicellular sessile structures, and thus one would imagine that motility is not required at this later stage. Interestingly, the lack of flagellar synthesis triggers the production of the extracellular polysaccharide (VPS) that is necessary for the three-dimensional biofilm architecture, at least in some V. cholerae serogroups, suggesting that flagellar synthesis (or lack thereof) controls biofilm formation (Watnick et al., 2001).

Flagellar synthesis has been best studied in the peritrichously flagellated Salmonella typhimurium, where flagellar gene transcription is organized into a hierarchy of three classes of genes (for review see Macnab, 1996). The Class I flhDC gene products stimulate σ70-dependent transcription of Class II genes, primarily the structural components of the basal body-hook complex (Kutsukake and Iino, 1994). One of the Class II genes, fliA, encodes the flagellar alternative sigma factor σ28. The σ28-holoenzyme form of RNA polymerase transcribes Class III genes, which encode filament, motor and chemotaxis components (Kutsukake et al., 1990). σ28 is kept inactive by the antisigma factor FlgM, which is exported through the completed basal body-hook structure, thus preventing Class III gene expression until Class II synthesis and assembly (Hughes et al., 1993; Kutsukake et al., 1994).

Flagellar synthesis has also been well studied in the polarly flagellated Caulobacter crescentus, where flagellar gene expression is organized into a hierarchy of four classes of genes (for reviews see Wu and Newton, 1997; Osteras and Jenal, 2000). The Class I ctrA gene product activates transcription of Class II genes, which encode components of the MS ring-switch, and also the alternative sigma factor σ54 and a σ54-dependent activator FlbD (Quon et al., 1996). FlbD and σ54-holoenzyme activate transcription of Class III and IV genes, which encode the basal body-hook and filament components respectively (Ramakrishnan and Newton, 1990). Phosphorylation of FlbD and post-transcriptional mechanisms, rather than export of an antisigma factor, control expression of Class III and IV genes (Wingrove et al., 1993). The initiation of flagellar synthesis is tied to cell cycle cues in C. crescentus, in contrast to the environmental cues that stimulate flagellar synthesis in S. typhimurium (Komeda et al., 1975; Quon et al., 1996).

The V. cholerae polar flagellar filament is composed of five distinct flagellins, but only one of these, FlaA, is essential for motility (Klose and Mekalanos, 1998a). Flagellar synthesis requires the alternative sigma factor σ54 (Klose and Mekalanos, 1998b), which is necessary for both flaA and flrBC gene transcription. FlrB/FlrC are a two-component regulatory system (Klose and Mekalanos, 1998b). FlrB transfers a phosphate to FlrC, and phospho-FlrC activates σ54-dependent transcription of flaA (Correa et al., 2000). Another σ54-dependent activator, FlrA, activates the transcription of flrBC. The four σ54-independent flagellin genes appear to be transcribed from σ28-dependent promoters (Klose and Mekalanos, 1998a). Thus, V. cholerae flagellar synthesis involves σ54, two different σ54-dependent transcriptional activators, and apparently σ28 as well, but the co-ordination of flagellar gene expression by these various transcription factors has been unclear.

In the present study, we have elucidated the flagellar transcription hierarchy of V. cholerae. This hierarchy appears to be a novel combination of both σ54- and σ28-dependent genes, organized into four classes. The V. cholerae flagellar hierarchy, therefore, combines elements of both the σ54-dependent C. crescentus hierarchy and the σ28-dependent S. typhimurium hierarchy.


Organization of the V. cholerae flagellar genes

The complete genome sequence of V. cholerae (Heidelberg et al., 2000) revealed three chromosomal loci with clusters of flagellar genes (Fig. 1). The genes encoding the flagellar motor components (motABXY) are located at three additional loci: all these genes are on the large V. cholerae chromosome. Additional chemotaxis genes are scattered throughout the genome. All the V. cholerae flagellar open reading frames (ORFs) have sufficient homology with flagellar components of other bacteria to ascribe putative functions to each of the gene products. The organization of these genes is almost identical to the organization of the polar flagellar genes of V. parahaemolyticus (Kim and McCarter, 2000), with two notable exceptions. First, regions II and III are contiguous in the V. parahaemolyticus chromosome, but separated by ∼ 48 000 nucleotides in the V. cholerae chromosome. Second, every V. cholerae gene shares homology and location with its V. parahaemolyticus counterpart, but an additional flagellin gene is found in region I of V. parahaemolyticus that is absent in V. cholerae.

Figure 1.

The Vibrio cholerae flagellar genes. The complete genome sequence of V. cholerae revealed six regions containing flagellar genes (labelled I–VI), all located on the large chromosome (Heidelberg et al., 2000). Region I encompasses nucleotides 2332971–2350790 (, region II is found in nucleotides 2275107–2301252, region III is nucleotides 2213661–2227373, region IV is contained in nucleotides 950906–952635, region V includes nucleotides 2770510–2771142 and region VI entails nucleotides 1078369–1079250. Thick arrows denote ORFs (not drawn to scale) with corresponding gene names below. The gene with significant homology to flgM genes of other bacterial species has been labelled flgM here. The gene labelled flgK here also shares significant homology to flgK genes of other bacterial species but has unfortunately been named flgM in V. cholerae (Das et al., 1998); we believe the nomenclature here is less confusing. Predicted functions of gene products are also noted. Putative and confirmed promoters are designated by thin arrows, along with transcription factors known to be required for expression. Transcription of the flrBC promoter has previously been shown to require FlrA and σ54 (Klose and Mekalanos, 1998b), the flaA promoter is transcribed by phosphorylated FlrC and RNA polymerase containing σ54 (Correa et al., 2000), and transcription of the flaE, flaD and flaB promoters is σ28-dependent (Klose and Mekalanos, 1998a). Transcription of flrA is σ54- and σ28-independent [(Klose and Mekalanos, 1998b); denoted by ‘?’], whereas flaC is poorly transcribed in Salmonella typhimurium (Klose and Mekalanos, 1998a), leading to uncertainty about the nature of this promoter (also denoted by ‘?’). Transcription of the other 11 promoters is described in this report.

The flagellin and regulatory genes we had previously cloned and characterized, flaAC, flaEDB and flrABC (Klose and Mekalanos, 1998a,b), are located within regions I and II. Region I contains regulatory (flgM) and chemotaxis (cheVR) genes, as well as structural genes for the basal body rod, rings, hook and filament (flgABCDEFGHIJKL and flaAC). We have previously demonstrated that flaA is transcribed from a σ54- and FlrC-dependent promoter (Klose and Mekalanos, 1998a; Correa et al., 2000). Although flaC appears to have a reasonable σ28 consensus promoter, it was transcribed at low levels in S. typhimurium and was only slightly affected by a fliA mutation, and thus regulation of this promoter appeared both σ54- and σ28-independent, at least in S. typhimurium (Klose and Mekalanos, 1998a). The remaining genes in region I appear to be organized into five transcriptional units, with the flgBCDEFGHIJ, the flgKL, the cheVR and the flgMN genes apparently organized into operons, based on the close or overlapping coding sequences and lack of transcriptional terminators.

Region II contains genes encoding basal body, switch and export (fliEFGHIJKLMNOPQRS, flhB, and flaGI), filament (flaEDB) and cap (fliD) components, as well as the regulatory locus flrABC. We have previously demonstrated that flrBC is transcribed from a σ54- and FlrA-dependent promoter, whereas flrA is transcribed from a σ54- and σ28-independent promoter (Klose and Mekalanos, 1998b). We also showed that flaE, flaD and flaB are each transcribed from σ28-dependent promoters, at least in S. typhimurium (Klose and Mekalanos, 1998a). The remaining genes in region II appear to be organized into two operons, flaGfliDflaIfliS and fliEFGHIJKLMNOPQRflhB.

Region III contains genes involved in chemotaxis (cheYZABW), export (flhA), and regulation (fliA, flhFG), as well as three ORFs of unknown function. flhF and flhG gene homologues are also found in polarly flagellated Pseudomonas species, where they dictate the number (one) and location (polar) of the flagellum (Dasgupta et al., 2000; Pandza et al., 2000). The fliA gene encodes the alternative flagellar sigma factor σ28 (see below). All the region III genes appear to be organized into a single operon. The four motor genes are found in three additional chromosomal locations: region IV contains the motAB operon, region V contains the motX gene and region VI contains the motY gene.

The V. cholerae fliA gene encodes a σ28 homologue that is required for motility

Because the V. cholerae flaE, flaD and flaB genes are transcribed from σ28-dependent promoters in S. typhimurium, we had predicted the existence of a σ28 homologue in V. cholerae (Klose and Mekalanos, 1998a). Region III contains a fliA gene whose predicted gene product shares homology with S. typhimuriumσ28 (47% identity). To determine whether the V. cholerae fliA gene product is a functional homologue of S. typhimuriumσ28, the V. cholerae fliA gene was expressed from the arabinose-inducible PBAD promoter in a S. typhimurium fliA mutant strain. Expression of V. cholerae fliA restored motility to the S. typhimurium fliA strain, demonstrating that this gene encodes a σ28 homologue (Fig. 2A).

Figure 2.

Vibrio cholerae fliA encodes a σ28 homologue. Motility was measured by a swarm assay in semisolid agar media containing 0.05% arabinose.

A. The wild-type S. typhimurium strain (ATCC 14028; ‘wild type’) is motile, whereas the isogenic fliA mutant (KK105; ‘fliA’) is non-motile. The fliA mutant KK105 carrying pKEK345, which expresses V. cholerae fliA from the PBAD promoter [‘pfliA (Vc)’], is motile. A similar level of motility was seen with KK105 carrying pMC147, which expresses S. typhimurium fliA from the same promoter (data not shown). The plate was incubated for 9 h at 30°C.

B. The wild-type V. cholerae strain (O395; ‘wild type’) is motile, whereas the isogenic ΔfliA mutant KKV1113 (‘fliA’) is non-motile. The ΔfliA mutant KKV1113 carrying pKEK345, which expresses V. cholerae fliA from the PBAD promoter [‘pfliA (Vc)’], is motile. A similar level of motility was seen with KKV1113 carrying pMC147, which expresses S. typhimurium fliA from the same promoter (data not shown). The plate was incubated for 18 h at 30°C. In A and B, ‘wild type’ and ‘fliA’ strains carry the vector pBAD24 (Guzman et al., 1995).

To determine the role of σ28, a V. cholerae strain containing an in frame deletion of fliAfliA) was constructed (Experimental procedures). The V. choleraeΔfliA strain KKV1113 was non-motile, and motility could be restored by expression of either V. cholerae or S. typhimurium fliA (Fig. 2B and data not shown). Examination of the V. choleraeΔfliA strain by electron microscopy revealed that no cells had a wild-type-length flagellum; rather these cells contained either a truncated flagellum or no flagellum (Fig. 3). These results demonstrate a requirement for σ28 in V. cholerae flagellar synthesis and motility.

Figure 3.

A V. cholerae fliA mutant strain synthesizes a truncated flagellum. Bacteria were harvested mid-logarithmic growth and prepared for microscopy (Experimental procedures).

A. Wild type (O395 X5000).

B. ΔfliA (KKV1113 X5000). Bar represents 1 micron, arrows in B indicate truncated flagella.

V. cholerae fliA is required for transcription of flaBCDE

We had previously demonstrated that the V. cholerae flagellin genes flaE, flaD and flaB are transcribed from σ28-dependent promoters in S. typhimurium, whereas the flagellin gene flaC was poorly expressed in S. typhimurium, in a seemingly σ28- and σ54-independent manner (Klose and Mekalanos, 1998a). We re-examined flagellin transcription in S. typhimurium strains containing chromosomal flaEp–, flaDp–, flaBp– and flaCp–lacZ transcriptional fusions, in both fliA and fliA+ strains. As was seen before, the relatively high level of flaE, flaD and flaB transcription in S. typhimurium is dependent upon an intact fliA gene, whereas the flaC promoter is transcribed at very low levels in both fliA and fliA+S. typhimurium (Fig. 4).

Figure 4.

The flaB, flaC, flaD and flaE promoters are σ28-dependent in S. typhimurium. S. typhimurium strains KK159, KK173, KK156 and KK167 (wild type with flaBp–, flaCp–, flaDp–and flaEp–‘lacZ chromosomal fusions respectively), and KK161, KK175, KK158 and KK169 (fliA5059::Tn10dTc with flaBp–, flaCp–, flaDp– and flaEp–‘lacZ chromosomal fusions respectively)(Klose and Mekalanos, 1998a),and additionally containing either vector alone (pBAD24) or the plasmids expressing V. cholerae fliA[pKEK345; ‘pfliA(Vc)’] or S. typhimurium fliA[pMC147; ‘pfliA (St)’] were measured for β-galactosidase activity during logarithmic growth in LB supplemented with 0.05% arabinose. Results are the average of three samples.

The fliA S. typhimurium strains containing the four flagellin promoter-lacZ fusions were then complemented with either S. typhimurium or V. choleraeσ28, expressed from the same arabinose-inducible vector. Both S. typhimurium and V. choleraeσ28 complemented the fliA mutation and allowed similar levels of flaE transcription, but the V. choleraeσ28 gave noticeably higher levels of transcription than S. typhimuriumσ28 at the flaB and flaD promoters. Interestingly, in contrast to the S. typhimuriumσ28, the V. choleraeσ28 drove high levels of flaC transcription in the fliA S. typhimurium strain (18-fold higher than S. typhimuriumσ28). These results are consistent with flaB, flaC, flaD and flaE having σ28-dependent promoters, and additionally suggest that the V. choleraeσ28 recognizes the flaC promoter (and possibly the flaB and flaD promoters also) better than the S. typhimuriumσ28.

Transcription of all five flagellin genes was measured in wild-type and ΔfliA V. cholerae strains, as well as in rpoN (encodes σ54), flrA and flrC strains, utilizing plasmid-encoded flaAp–, flaBp–, flaCp–, flaDp–and flaEp–lacZ transcriptional fusions. The flaA promoter was transcribed at similar levels in both fliA and fliA+V. cholerae (Fig. 5), but its transcription demonstrated a dependence on rpoN, flrA and flrC, consistent with the previous observation that flaA transcription is σ54- and FlrC-dependent. However, the flaB, flaC, flaD and flaE promoters were transcribed at significantly reduced levels in the ΔfliA strain, compared with the wild-type strain. Taken together with the results above, this indicates that flaB, flaC, flaD and flaE are transcribed by σ28-holoenzyme in V. cholerae as well as in S. typhimurium.

Figure 5.

The flaB, flaC, flaD and flaE promoters are σ28-dependent, as well as σ54-, FlrA- and FlrC-dependent in V. cholerae. V. cholerae strains KKV598 (‘wild type’), KKV56 (‘rpoN’), KKV59 (‘flrA’), KKV98 (‘flrC’) and KKV1113 (‘fliA’) carrying plasmids pKEK80 (flaAp–lacZ), pKEK79 (flaBp–lacZ), pKEK76 (flaCp–lacZ), pKEK77 (flaDp–lacZ) and pKEK81 (flaEp–lacZ) (Klose and Mekalanos, 1998a) were assayed for β-galactosidase activity during logarithmic growth in LB. Results are the average of three samples.

The flaB, flaC, flaD and flaE promoters are also dependent on σ54 and FlrA for high levels of transcription (Fig. 5; Klose and Mekalanos, 1998b). This could be due to a requirement for σ54 and FlrA for fliA expression, which in turn is required for flaB, flaC, flaD and flaE transcription. An amino-terminal fragment of the fliA gene was fused to a promoterless luciferase (luc) gene in a suicide plasmid (pKEK383; Experimental procedures), integrated into the chromosome of wild-type, rpoN, flrA, flrC and fliA V. cholerae strains, and fliA transcription measured. fliA transcription was reduced in the rpoN and flrA V. cholerae strains (2.1 and 2.4 × 108 RLUml−1 inline image respectively), compared with the wild-type, flrC and fliA strains (7.0, 7.2, and 6.7 × 108 RLUml−1 inline image respectively). These results are consistent with fliA being transcribed from the flhA promoter, which is σ54- and FlrA-dependent (see below). The FlrC-dependence of flaB, flaC, flaD and flaE transcription (Fig. 5) is hypothesized to be due to a Class III–IV checkpoint, as described in the Discussion.

Characterization of σ54- and σ28-dependent V. cholerae flagellar promoters in S. typhimurium

Salmonella typhimurium was used as a heterologous system to determine the nature of the 11 uncharacterized V. cholerae flagellar promoters (Fig. 1). Use of a heterologous system, such as the well-characterized S. typhimurium, is critical to identify the promoters as either σ54- or σ28-dependent in the absence of any native epistatic effects or modulating factors present in V. cholerae. Chromosomal lacZ transcriptional fusions of the flgA, flgM, cheV, flgB, flgK, flaG, fliE, flhA, motA motX and motY promoters were constructed in wild-type, fliA and ntrA (rpoN; encodes σ54) S. typhimurium strains. While σ28-holoenzyme is sufficient for transcription of σ28-dependent promoters, σ54-holoenzyme requires a σ54-dependent activator to transcribe σ54-dependent promoters (Kustu et al., 1989). In high concentration, σ54-dependent activators in solution can activate transcription from any σ54-dependent promoter (i.e. without binding a specific DNA sequence) (North and Kustu, 1997). For example, we have shown previously that FlrA, when overexpressed from the PBAD promoter, can activate σ54-dependent transcription of the S. typhimurium glnA promoter, which is normally activated by NtrC (Klose and Mekalanos, 1998a). We, thus, utilized overexpressed FlrA here to identify any σ54-dependent promoter, regardless of whether FlrA is the cognate activator of this promoter in V. cholerae.

Table 1 shows the transcription of the 11 V. cholerae flagellar promoters in the various S. typhimurium strains. Transcription of the fliE, flhA, flgB, flgK, motX and flaG  promoters in S. typhimurium was consistent with these promoters being recognized by σ54-holoenzyme (Table 1A). The provision of a σ54-dependent activator (FlrA) in high concentration allowed for the transcription of these promoters in S. typhimurium, but only in wild-type and fliA strains, and not in ntrA strains. The motX promoter exhibited relatively high levels of transcription even in the absence of fliA and ntrA, suggesting the presence of additional σ54- and σ28-independent motX promoter(s). Sequences could be found in each of the six promoters that resemble the consensus σ54 recognition site (Fig. 6).

Table 1. β-Galactosidase activity of chromosomal flagellar promoter-lacZ transcriptional fusions in S.typhimuriuma
A. σ54-dependent flagellar promoter transcription in S.typhimuriumb







  • a.

    Assayed as described (Experimental procedures); results are the average of three samples reported in Miller units.

  • b.

    As described in the text, FlrA overexpressed from the PBAD promoter can activate any s54-dependent promoter, regardless of whether FlrA normally does so in V. chloerae; thus, this protin serves here to identify all s54-dependent promoters.

  • c.

    Actual background strains used (Experimental procedures) were 14028 (wild type), KK1 (ntrA) and KK105 (fliA); these strains contain the various flagellar promoter-lacZ fusions inserted into the putPA locus.

  • d.

    Plasmids used (Experimental procedures): pBAD24 (vector), pKEK94 (pBADfliA).

Wild typepBAD2471 ± 258 ± 262 ± 725 ± 3382 ± 1141 ± 1
Wild typepBADflrA515 ± 27490 ± 10211 ± 5938 ± 441272 ± 31513 ± 24
ntrA pBAD2474 ± 562 ± 264 ± 330 ± 1423 ± 229 ± 4
ntrA pBADflrA61 ± 362 ± 552 ± 332 ± 2436 ± 3134 ± 2
fliA pBAD2475 ± 253 ± 246 ± 226 ± 3434 ± 944 ± 2
fliA pBADflrA596 ± 21430 ± 19184 ± 5868 ± 281352 ± 50593 ± 30
B. σ28-dependent, and σ28-, σ54-independent flagellar promoter transcription in S. typhimurium
Relevant genotypecPlasmiddσ28-dependentσ28-and σ54-independent  
flgMp motAp motYp flgAp cheVp 
Wild typepBAD242915 ± 651113 ± 753313 ± 152141 ± 5147 ± 11 
Wild typepBADflrANDNDND167 ± 16265 ± 19 
Wild typepBADfliA6423 ± 2422882 ± 1419080 ± 188116 ± 7232 ± 35 
ntrA pBAD242907 ± 411039 ± 172232 ± 553131 ± 9286 ± 11 
ntrA pBADflrANDNDND131 ± 13315 ± 38 
ntrA pBADfliA5137 ± 382370 ± 308033 ± 452123 ± 10194 ± 6 
fliA pBAD24824 ± 82377 ± 241614 ± 19113 ± 13359 ± 6 
fliA pBADflrANDNDND109 ± 17331 ± 35 
fliA pBADfliA4677 ± 891517 ± 339839 ± 305112 ± 16306 ± 53 
Figure 6.

Alignment of putative σ28 and σ54 promoter elements in V. cholerae flagellar promoters. The consensus σ54 sequence is from reference (Barrios et al., 1999), and the consensus σ28 sequence is from reference (Kutsukake et al., 1990).

The flgM, motA and motY promoters demonstrated two characteristics of σ28-dependent transcription (Table 1B): (i) elevated expression in both wild-type and ntrA strains, but diminished transcription (∼ twofold or greater) in fliA S. typhimurium; and (ii) increased transcription in wild-type, ntrA and fliA strains when V. choleraeσ28 was expressed from a plasmid. All three of these promoters exhibit high levels of transcription even in the absence of σ28, indicating that there are probably additional σ28- and σ54-independent promoter(s) for these genes. Sequences could be found in each of these three promoters that resemble the consensus σ28 recognition site, although these promoters were less similar when compared with the four σ28-dependent flagellin promoters (Fig. 6).

Transcription of the flgA and cheV promoters is σ54- and σ28-independent in S. typhimurium (Table 1B). These two divergent promoters are comprised within the same intervening sequence (see Fig. 1). It remains unclear what holoenzyme form of RNA polymerase is required for their transcription.

Identification of FlrA-, FlrC-, σ54- and σ28-dependent flagellar transcription in V. cholerae

The 11 plasmid encoded flagellar promoter-lacZ fusions were transformed into wild-type, rpoN, flrA, flrC and fliA V. cholerae strains, to determine which of these various regulatory factors is required for their transcription. Four different patterns of transcription were observed (Table 2). First, the fliE and flhA promoters required only rpoN and flrA for wild-type levels of transcription (Table 2A). Given that these promoters were shown above to be σ54-dependent in S. typhimurium, these results are consistent with the fliE and flhA promoters being transcribed by σ54-holoenzyme and FlrA in V. cholerae. This pattern of expression is similar to that seen at the flrBC promoter, which is also transcribed by FlrA and σ54-holoenzyme (Klose and Mekalanos, 1998b). Transcription of fliA was already shown to be σ54- and FlrA-dependent (above), consistent with this gene being transcribed in an operon from the flhA promoter.

Table 2. β-Galactosidase activity of flagellar promoter-lacZ transcriptional fusions in V.choleraea.
A. σ54-dependent flagellar promoter transcription in V.choleraeb
Relevant genotypecσ54- and FlrA-dependentσ54- and FlrC-dependent
fliEp flhAp flgBp flgKp motXp flaGp
  • a.

    Assayed as described (Experimental procedures); results are the average of three samples reported in Miller units.

  • b.

    Plasmids used (Experimental procedures): pKEK327 (fliEp), pKEK329 (flhAp), pKEK332 (flgBp), pKEK331 (flgKp), pKEK412 (motXp), pKEK415 (flaGp).

  • c.

    Actual strains used (Experimental procedures): KKV598, KKV56, KKV59, KKV98 and KKV1113; all strains also contain ΔlacZ mutation.

  • d.

    Plasmids used (Experimental procedures): pKEK416 (flgMp), pKEK414 (motAp), pKEK413 (motYp), pKEK417 (flgAp) and pKEK418 (cheVp).

Wild type1102 ± 1101125 ± 12663 ± 15274 ± 71678 ± 361450 ± 76
ΔrpoN155 ± 3284 ± 3594 ± 217 ± 1544 ± 36788 ± 103
ΔflrA182 ± 14184 ± 386 ± 720 ± 1500 ± 49645 ± 30
ΔflrC929 ± 141071 ± 15 144 ± 553 ± 1555 ± 28769 ± 24
ΔfliA1291 ± 291226 ± 351332 ± 100361 ± 131408 ± 881918 ± 76
B. σ28-dependent, and σ28-, σ54-independent flagellar promoter transcription in V.choleraed
Relevant genotypecσ28-dependentσ28-and σ54-independent  
flgMp motAp motYp flgAp cheVp 
Wild type3906 ± 200924 ± 502594 ± 39824 ± 543401 ± 149 
ΔrpoN1571 ± 48386 ± 33859 ± 22925 ± 212317 ± 261 
ΔflrA1572 ± 122448 ± 21778 ± 15932 ± 132755 ± 12 
ΔflrC1137 ± 72206 ± 16927 ± 42740 ± 282582 ± 190 
ΔfliA1746 ± 21443 ± 18908 ± 129875 ± 202107 ± 88 

The second pattern of expression was observed with the flgK, flgB, flaG and motX promoters (Table 2A). These promoters require rpoN, flrA and flrC for wild-type levels of transcription, but not fliA. These promoters were shown to be σ54-dependent in S. typhimurium (above). Because σ54-holoenzyme and FlrA are required for flrC transcription, one would expect FlrC-dependent promoters to exhibit this pattern of expression in V. cholerae, as is seen at the FlrC-dependent flaA promoter (see above). Therefore, this pattern of expression is consistent with these promoters being transcribed by σ54-holoenzyme and FlrC in V. cholerae. The motX and flaG promoters exhibit relatively high levels of transcription even in rpoN strains, suggesting additional σ54-independent promoter(s) of these genes.

Transcription of the motA, motY and flgM promoters fell into a third pattern of expression. These promoters have ∼ twofold reduced levels of transcription in rpoN, flrA, flrC and fliA strains, compared with expression in the wild-type strain. While this is not a large difference, the general pattern of transcription is that predicted from σ28-dependent promoters in V. cholerae. Transcription of these promoters is σ28-dependent in S. typhimurium (above), as it is for the alternative flagellin genes. In fact, this is the pattern of expression seen at the flaB, flaC, flaD and flaE promoters in V. cholerae (see above and Klose and Mekalanos, 1998b), although the alternative flagellin promoters do not demonstrate such high levels of transcription in the absence of σ54. Because the motA, motY and flgM promoters achieved high levels of transcription even in the absence of the four transcription factors in both V. cholerae and S. typhimurium, this indicates additional σ54- and σ28-independent promoter(s) involved in their transcription.

The fourth pattern of expression was observed with the flgA and cheV promoters, which were transcribed at similar levels in all V. cholerae strains. These promoters were shown to be σ28- and σ54-independent in S. typhimurium (above), so it remains unclear which regulatory factors transcribe these genes.


The novel flagellar gene transcription hierarchy of V. cholerae

The results presented here provide evidence for a novel flagellar transcription hierarchy in V. cholerae. Flagellar synthesis in V. cholerae shares similarities with that in C. crescentus. It was previously shown that flagellar synthesis is σ54-dependent in V. cholerae, similar to flagellar synthesis in C. crescentus (Brun and Shapiro, 1992; Klose and Mekalanos, 1998b). Both organisms have single polar flagella with filaments composed of multiple flagellins, and the V. choleraeσ54-dependent flagellar activator FlrC shares homology throughout its entire sequence with the C. crescentusσ54-dependent flagellar activator FlbD. Phosphorylation of both FlrC and FlbD is required for flagellar synthesis in these organisms (Wingrove et al., 1993; Correa et al., 2000). However, the presence of a second σ54-dependent flagellar activator, FlrA, as well as the presence of the alternative flagellar sigma factor σ28, suggested that flagellar gene transcription was fundamentally different in V. cholerae compared with C. crescentus. Flagellar synthesis in S. typhimurium requires σ28 (Ohnishi et al., 1990); however, σ54 plays no role in flagellar synthesis, indicating that flagellar gene transcription also differs significantly between S. typhimurium and V. cholerae.

Analysis of transcription of the V. cholerae flagellar promoters in both V. cholerae and S. typhimurium has allowed us to propose a model for the V. cholerae flagellar transcription hierarchy (Fig. 7), composed of four classes of genes, outlined below.

Figure 7.

Proposed flagellar gene transcription hierarchy of V. cholerae. See Discussion.

Class I and II

The sole Class I gene encodes the σ54-dependent activator FlrA. It was previously shown that this gene has a σ54- and σ28-independent promoter, so the nature of the regulatory factors and environmental signals that control its expression remain unknown (Klose and Mekalanos, 1998b). Expression of the Class I activator CtrA in C. crescentus is regulated by cell cycle cues (Osley et al., 1977). Connecting flagellar synthesis to the cell cycle may be important for the synthesis of a single polar flagellum, and thus we predict that flrA is also probably regulated by cell cycle cues in V. cholerae.

FlrA, along with σ54-holoenzyme, is required for transcription of Class II flagellar genes which encode the MS ring-switch and export components. Because the flagellum is synthesized from the ‘inside’, these genes are always transcribed earliest in a flagellar hierarchy. Consistent with its placement at the top of the flagellar hierarchy, flrA (along with rpoN) is required for expression of Class III and Class IV genes also (see below). In C. crescentus, transcription of rpoN is also regulated within the flagellar hierarchy (Brun and Shapiro, 1992). Although we have not analysed transcription of rpoN in V. cholerae, it is doubtful that this gene is expressed specifically for its role in flagellar synthesis, given its involvement in multiple cellular processes in V. cholerae (Klose and Mekalanos, 1998b; Klose et al., 1998).

Sigma-54-dependent transcriptional activators possess an ATPase activity required for initiation of σ54-dependent transcription (Weiss et al., 1991), and this activity is commonly modulated by an amino terminal regulatory domain. The amino terminus of FlrA (unlike FlrC) shares little homology with response regulatory domains and is missing the conserved aspartate that is the site of phosphorylation, suggesting that it is not regulated by phosphorylation. One would predict that some mechanism is necessary for shutting off FlrA activity, to prevent excess transcription of Class II genes, or to halt flagellar synthesis completely (e.g. within biofilms). What, if any, mechanism exists to downregulate FlrA-dependent transcription remains unknown, especially considering that this protein is clearly transcriptionally active in the heterologous host S. typhimurium.

Class II genes are those that are transcribed by FlrA and σ54-holoenzyme. These include components of the MS ring-switch complex, as well as the regulatory factors FlrB, FlrC and FliA (σ28). In C. crescentus, the components of the MS ring-switch complex are also expressed as Class II genes within the transcription hierarchy (Ohta et al., 1991). Because flrC is expressed as a Class II gene in V. cholerae, one predicted mechanism of control over the Class II–III transition would be prevention of FlrC phosphorylation until MS ring-switch assembly (see below).

Class III

Class III genes are those that encode the basal body-hook, and also the ‘core’ flagellin FlaA, the flagellar cap FliD and a motor component MotX. These genes are expressed from FlrC- and σ54-dependent promoters. FlrB/FlrC are a two component regulatory pair that are essential for V. cholerae flagellar synthesis (Klose and Mekalanos, 1998b). The histidine kinase FlrB transfers a phosphate to a conserved aspartate residue in the amino terminus of FlrC, which renders FlrC capable of activating σ54-dependent transcription (Correa et al., 2000). Thus, one potential mechanism for an assembly checkpoint that prevents Class III gene expression prior to assembly of the MS ring-switch complex would be regulation of phospho-FlrC generation.

In C. crescentus, the checkpoint that prevents expression of Class III genes until MS ring-switch assembly requires the bfa gene product, which negatively regulates FlbD-dependent transcription until the MS ring-switch is complete (Mangan et al., 1995), raising the possibility that bfa controls FlbD phosphorylation. We predict the checkpoint control governing the Class II–III transition in V. cholerae is the phosphorylation of FlrC, which is probably induced upon completion of the MS ring-switch. Alteration of the transcriptional activity of FlrC by mutation of specific amino acids alters the ability of V. cholerae to colonize the intestine, suggesting that the phosphorylation state of FlrC affects not only flagellar synthesis, but virulence as well (Correa et al., 2000).

Class IV

Our previous analysis of V. cholerae flagellin transcription had suggested the presence of σ28 (Klose and Mekalanos, 1998a). As demonstrated here, the V. cholerae fliA gene encodes a σ28 homologue that is required for motility. The Class IV genes are σ28-dependent and encode the alternative flagellins, the putative antisigma factor FlgM and the motor components MotABY. Additional chemotaxis genes are found throughout the chromosome, and these may also be σ28-dependent. Interestingly, a V. cholerae fliA strain synthesizes a flagellum, albeit truncated, and yet this strain is non-motile. This is probably due to the decreased σ28-dependent expression of the motor components MotABY, rather than the inability of the truncated flagellum to propel the cell. It has been shown that a mutation in any of the motor genes results in a paralyzed flagellum (i.e. non-motile) (Gosink and Hase, 2000), whereas a strain lacking the σ28-dependent flagellins FlaBCDE is still motile (Klose and Mekalanos, 1998a).

σ28 of S. typhimurium is kept inactive until assembly of the flagellar basal body-hook, which acts as an export apparatus for the antiσ28 factor FlgM, thus relieving inhibition of σ28-dependent transcription (Hughes et al., 1993). Export of FlgM signals the transition from Class II to Class III gene expression in S. typhimurium. The putative V. cholerae FlgM shares homology with S. typhimurium FlgM (27% identity), and it is thus predicted to have a similar antiσ28 function in V. cholerae. One would, therefore, expect that σ28-dependent transcription in V. cholerae would be repressed until the assembly of the basal body-hook, and that σ28-dependent transcription would temporally follow FlrC-dependent transcription of basal body-hook genes. In fact, high-level transcription of these genes in V. cholerae requires FlrC, in addition to σ54, FlrA and σ28, even though σ28 and FlrC are expressed simultaneously as Class II genes. This suggests σ28-dependent transcription follows FlrC-dependent transcription, and for this reason, we have designated these as Class IV and depicted them as temporally distinct from Class III. However, transcription of some Class IV genes (e.g. flaB, flaD;Fig. 5), though less than wild-type levels, is higher in flrC V. cholerae strains (Klose and Mekalanos, 1998b) than in rpoN, flrA and fliA mutant strains, suggesting more complex mechanism(s) of regulation. Ongoing experiments concerning the function of V. cholerae FlgM should clarify the nature of the Class III–IV checkpoint.


The flgA and cheV promoters were both σ28 -and σ54-independent, were transcribed at relatively constant levels in the various flagellar regulatory mutant strains, and contain no identifiable σ28 or σ54 promoter elements. The same pattern of expression was seen with the flrA (Class I) gene (Klose and Mekalanos, 1998b). The flrA gene clearly belongs in Class I because expression of Classes II–IV is FlrA-dependent. Dependence of Classes II–IV expression on the flgA and cheV operons remains to be determined. For these reasons the flgA and cheV promoters remain unclassified. It is interesting that at least some chemotaxis genes (cheVR) are transcribed even in the absence of a flagellum (e.g. in rpoN and flrA strains). It is likely that the flgA promoter drives transcription of flgMN also, due to the conservation of gene arrangement between S. typhimurium and V. cholerae. In S. typhimurium, early (Class II) transcription of flgM from the flgA promoter, in addition to σ28-dependent transcription of the flgM promoter, ensures the presence of the antisigma factor to enforce the Class II–III checkpoint (Gillen and Hughes, 1993).

The flagellar hierarchy proposed here for V. cholerae is doubtless an oversimplification. Several of the σ54- and σ28-dependent promoters appear to be regulated by additional σ54- and σ28-independent mechanisms. Such mechanisms that fall outside the ‘classic’ flagellar hierarchy occur in both S. typhimurium and C. crescentus flagellar transcription systems (Macnab, 1996; Wu and Newton, 1997), and thus V. cholerae flagellar synthesis is not unique in this respect. While the complete mechanisms involved in V. cholerae flagellar transcription are probably more complex than the proposed hierarchy, it serves as a suitable scaffold to begin to study the flagellar regulatory scheme.

Implications for pathogenesis/environmental persistence

The elucidation of the V. cholerae flagellar hierarchy will assist in dissecting flagellar control over both V. cholerae virulence and environmental persistence. Motility and/or flagellar synthesis have been implicated in intestinal colonization and virulence. In fact, the transcriptional activity of FlrC affects intestinal colonization by an unknown mechanism (Correa et al., 2000); knowledge of the role of FlrC in the flagellar hierarchy should be helpful in identifying those FlrC-dependent factors that affect virulence. Additionally, flagellar synthesis exerts control over VPS expression required for biofilm formation and presumably environmental persistance (Watnick et al., 2001); dissection of this mechanism of control will be assisted by an understanding of the flagellar hierarchy.

Given the remarkable similarity of the flagellar organization between V. cholerae and V. parahaemolyticus (Kim and McCarter, 2000), it is reasonable to assume that the polar flagellar gene transcription hierarchy of V. parahaemolyticus is organized similarly to that of V. cholerae. Furthermore, the regulatory genes fleQSR which control Pseudomonas aeruginosa flagellar synthesis are homologues of V. cholerae flrABC (Arora et al., 1997; Ritchings et al., 1995), and P. aeruginosa also contains a σ28 homologue (Starnbach and Lory, 1992); thus it is likely that P. aeruginosa flagellar gene transcription occurs via a flagellar hierarchy similar to that in V. cholerae. The σ28- and σ54-dependent flagellar transcription hierarchy elucidated here may be a common mechanism for control over flagellar synthesis in a variety of bacteria with polar flagella.

Experimental procedures


Luria broth was used for both liquid and agar. Antibiotics were added when appropriate. Media was supplemented with 2 mM glutamine when appropriate for growth of rpoN (ntrA) strains. Agar plates consisting of LB with 0.3% agar were used to measure motility. LB agar, made without NaCl and supplemented with 10% sucrose, was used to select for second recombinational events during construction of chromosomal deletions/insertions with pKEK229 (contains the sacB gene, see below). Evans-blue uranine plates supplemented with 2 mM glutamine were used to purify all S. typhimurium strains free from P22 phage. LB broth was also supplemented with 0.05% arabinose for strains carrying PBAD vectors.

Bacterial strains

Escherichia coli DH5α was used for cloning manipulations. For the construction of the promoter-lacZ chromosomal fusions in S. typhimurium, E. coli strains TE1335 and TE2680 were used in intermediate steps (Elliott, 1992). Escherichia coli strain SM10λpir (Miller and Mekalanos, 1988) was used to transfer plasmids to V. cholerae by conjugation. All V. cholerae strains are isogenic with the Classical Ogawa strain O395 (Mekalanos et al., 1979), and S. typhimurium strains are isogenic with strain 14028 (American Type Culture Collection). To create the single copy, chromosomal promoter-lacZ fusions in S. typhimurium we followed the method outlined by Elliott (1992). The resultant P22 phage containing the various flagellar promoter lacZ fusions inserted into the putPA locus was transduced into three background S. typhimurium strains: KK1 (ntrA::Tn10), KK105 (fliA::Tn10dTc), or 14028 (wild type) (Klose and Mekalanos, 1998a). With 11 promoters and three background strains, this resulted in 33 different S. typhimurium strains; as a result of space constraints all these strains are not elaborated in Table 3, rather only the parental background strains are noted.

Table 3. Strains and oligonucleotides used in this study.
StrainGenotypeSource or reference
V. cholerae strains
O395wild type (classical Ogawa) Mekalanos et al. (1979)
KKV56ΔrpoN1;ΔlacZ Klose and Mekalanos (1998b)
KKV59ΔflrA1;ΔlacZ Klose and Mekalanos (1998b)
KKV98ΔflrC1;ΔlacZ Klose and Mekalanos (1998b)
KKV598O395 ΔlacZ Correa et al. (2000)
KKV1113ΔfliA1;ΔlacZThis study
S. typhimurium strains
KK1 ntrA209::Tn10 Klose and Mekalanos (1997)
KK105 fliA5059::Tn10dTc Klose and Mekalanos (1998a)
Promoter fragmentupstream oligonucleotide (EcoRI site underlined)downstream oligonucleotide (BamHI site underlined)

Construction of flagellar promoter-lacZ fusions

Oligonucleotide primer pairs for the promoter sequences of the flgA, flgM, cheV, flgB, flgK, flaG, fliE, flhA, motA, motX and motY genes were designed based on the V. cholerae genome sequence (Heidelberg et al., 2000). The oligonucleotide primers were designed to create a fragment of approximately 450 bp that encompassed regions both upstream and downstream of the predicted start codon (Table 3). This does not eliminate the possibility of additional regulatory sequences outside these fragments; however, all these fragments contain demonstrably active promoters. Restriction sites for EcoRI and BamHI were added to the upstream and downstream primers, respectively, for cloning purposes. PCR with each pair of primers was performed for 30 cycles of 45 s at 92°C, 1 min at 56°C and 1 min at 72°C with Vent DNA polymerase (New England Biolabs). One fragment was produced for each promoter sequence using genomic DNA from V. cholerae Classical O1 strain O395. Each PCR product was digested with EcoRI and BamHI, and ligated into pRS551 (Simons et al., 1987) that had been similarly digested, to form pKEK327 (fliEp), pKEK329 (flhAp), pKEK331 (flgKp), pKEK332 (flgBp), pKEK412 (motXp), pKEK413 (motYp), pKEK414 (motAp), pKEK415 (flaGp), pKEK416 (flgMp), pKEK417 (flgAp) and pKEK418 (cheVp). The plasmids pKEK80 (flaAp), pKEK79 (flaBp), pKEK76 (flaCp), pKEK77 (flaDp) and pKEK81 (flaEp) have been described previously (Klose and Mekalanos, 1998a,b). These plasmids were used as episomal lacZ transcriptional fusions in V. cholerae, as well as to construct chromosomal lacZ transcriptional fusions in S. typhimurium (above).

Construction of the fliA V. cholerae mutant strain

The V. choleraeΔfliA mutant strain KKV1113 was created by generating two PCR-amplified fragments containing 5′ and 3′ pieces of the fliA gene. The 5′ fragment was PCR-amplified with the oligonucleotides FLIADEL1 (5′-GCGGATCCAGCAAAGAACATCAAGTTCAAC-3′) and FLIADEL2 (5′-GCGAATT CGATACGCTTAACCAATACAGAG-3′). This fragment was digested with EcoRI and BamHI (sites underlined in oligonucleotides) and ligated into pWSK30 (Wang and Kushner, 1991) similarly digested to form pKEK376. The 3′fliA fragment was PCR-amplified using oligonucleotides FLIADEL3 (5′-GCGAATTCGAAATTGGTGAGGTACTTGGAG-3′) and FLIADEL4 (5′-GCGAAGCTTAGCGGCTTCGATGATTTGCTCAC-3′). This fragment was digested with EcoRI and HindIII (sites underlined in oligonucleotides) and ligated into pKEK357 that had been similarly digested to form pKEK371, which thus contains ΔfliA1, a deletion of aa 31-209 in the predicted fliA gene product.

Then, pKEK371 was digested with SalI and NotI and ligated with pKEK229 (a derivative of pCVD442; Correa et al., 2000; Donnenberg and Kaper, 1991), similarly digested to form pKEK374, which was used to recombine ΔfliA1 back onto the chromosome of V. cholerae strain KKV598 generating strain KKV1113. Allelic exchange was confirmed by PCR.

Construction of fliA–luc and PBADfliA plamids

pKEK376, which contains the fliA′ fragment (above), was digested with BamHI, made blunt ended with the Klenow fragment of DNA polymerase, then digested with SalI, and ligated with pGPL01 (Gunn and Miller, 1996) that had been digested with SmaI and SalI. This plasmid contains an R6K origin of replication and is, thus, useful for creating cointegrant transcriptional fusions in the V. cholerae chromosome. The resulting plasmid pKEK383, which contains a fliA′–luc transcriptional fusion, was recombined into the chromosomes of strains 0395, KKV56, KKV98, KKV59 and KKV1113; the resultant cointegrant strains contain the plasmid integrated into the fliA gene and are thus fliA.

The PBADfliA plasmid pKEK345 was constructed by first PCR-amplifying the fliA gene from the V. cholerae chromosome utilizing oligonucleotides FLIAMet (5′-GCGCCATGGATAAAGCGCTTACATACGATC-3′) and FLIAU1 (5′-GCTCTAGATTAGTCATTCCGAGTCCAAGAAC-3′). The resulting fragment was then digested with NcoI and XbaI (sites underlined in oligonucleotides) and ligated into pBAD24 (Guzman et al., 1995) that had been similarly digested, to form pKEK345, an in frame translational fusion under control of the PBAD promoter. Plasmid pMC147 (kindly provided by K. Hughes) expresses S. typhimurium fliA from the same promoter. Plasmid pKEK94, which expresses V. cholerae flrA from the PBAD promoter, has been described previously (Klose and Mekalanos, 1998a).

Β-Galactosidase and luciferase assays

Vibrio cholerae strains were grown in LB supplemented with 2 mM glutamine at 37°C. Salmonella typhimurium strains were grown similarly with the addition of 0.05% arabinose. The samples were assayed at an optical density of approximately 0.4 at 600 nm. The samples were permeabilized with chloroform and SDS then assayed for β-galactosidase activity as described by Miller (1992). For luciferase assays, cells were sonicated, diluted in luciferase buffer and measured for relative light units in a Berthold Lumat luminometer model LB9507 as described previously (Gunn and Miller, 1996).

Electron microscopy

Strains were grown to mid-log in LB, then centrifuged and resuspended in 0.15 M NaCl. Samples were adhered to a carbon-coated grid and stained with 1% uranyl acetate before microscopy.


We thank Kelly Hughes for plasmid pMC147. This study was supported by National Institutes of Health grant AI-43486 to K.E.K.