Positive regulation of motility and flhDC expression by the RNA-binding protein CsrA of Escherichia coli



Many species of bacteria devote considerable metabolic resources and genetic information to the ability to sense the environment and move towards or away from specific stimuli using flagella. In Escherichia coli and related species, motility is regulated by several global regulatory circuits, which converge to modulate the overall expression of the master operon for flagellum biosynthesis, flhDC. We now show that the global regulator CsrA of E. coli K-12 is necessary for motility under a variety of growth conditions, as a result of its role as an activator of flhDC expression. A chromosomally encoded flhDC′–′lacZ translational fusion was expressed at three- to fourfold higher levels in csrA wild-type strains than in isogenic csrA mutants. Purified recombinant CsrA protein stimulated the coupled transcription-translation of flhDC′–′ lacZ in S-30 extracts and bound to the 5′ segment of flhDC mRNA in RNA mobility shift assays. The steady-state level of flhDC mRNA was higher and its half-life was ≈ threefold greater in a csrA wild-type versus a csrA mutant strain. Thus, CsrA stimulates flhDC gene expression by a post-transcriptional mechanism reminiscent of its function in the repression of glycogen biosynthesis.


Motility and chemotaxis permit bacterial cells to move away from stressful microenvironments and toward nutrients, O2, light or other stimuli, and are important during the infectious cycles of a number of bacterial pathogens (reviewed by Ottemann and Miller, 1997; Givaudan and Lanois, 2000). Motility is also required for biofilm formation in Escherichia coli and Pseudomonas aeruginosa under certain growth conditions (O'Toole and Kolter, 1998; Pratt and Kolter, 1998). Interestingly, the bacterial flagellum is homologous to the type III secretory pathway that is used for the transport of virulence proteins into eukaryotic host cells (Hueck, 1998), and may itself function in secreting certain virulence factors (Macnab, 1999; Young et al., 1999).

The flagellum is among the most complex cellular machinery in bacteria and requires ≈ 2% of the total cellular energy for its biosynthesis and rotation (reviewed by Macnab, 1996; Prüß, 2000). Synthesis of the flagellum and its related components in E. coli involves 14 operons and over 50 genes. In order to co-ordinate this process, the expression of these genes is organized in a three-tier hierarchy. At the top of this hierarchy is the master operon, flhDC, which encodes the subunits of a tetrameric DNA-binding protein (FlhD2C2) that recognizes second-level (class II) flagellar promoters (Liu and Matsumura, 1994). The second-level genes encode proteins for the basal body and the hook of the flagellum, as well as for the sigma factor, FliA, which is required for the transcription of class III genes. These genes are needed for assembly of the flagellar filament, motor activity, chemotaxis and for synthesis of the anti-sigma factor, FlgM, which accumulates upon completion of the flagellum and inhibits FliA activity (Hughes et al., 1993).

The development of motility is highly responsive to environmental conditions in E. coli and Salmonella typhimurium. In addition to its well-documented role in regulating motility, FlhD controls other genes (e.g. Prüß and Matsumura, 1996; Prüßet al., 1997). Accordingly, the expression of flhDC is affected by numerous global regulatory factors. Factors required for motility and full flhDC expression include cyclic AMP (cAMP) and cAMP receptor protein (CRP; Kutsukake, 1997; Soutourina et al., 1999), histone-like nucleoid-structuring (H-NS) protein (Bertin et al., 1994), the heat shock response network proteins DnaK, DnaJ and GrpE (Shi et al., 1992), Pss and Psd, which are involved in the synthesis of phosphatidylethanolamine in cell membranes (Shi et al., 1993), and inorganic polyphosphate (Rashid et al., 2000). In contrast, motility and flhDC transcription are inhibited under conditions of high acetyl phosphate levels and high osmolarity by the phosphorylated form of OmpR (Prüß and Wolfe, 1994; Shin and Park, 1995). Activation of motility by H-NS has recently be found to involve its repression of a transcriptional repressor of flhDC, hdfR (Ko and Park, 2000). Chemotaxis, or the detection of and movement towards or away from specific chemical stimuli, is itself controlled by distinct signal transduction cascades (Eisenbach, 1996).

In the past several years, we have uncovered a novel type of global regulatory system that is present in E. coli and numerous other bacteria, the carbon storage regulatory system (Csr). The effector of Csr is a small RNA-binding protein, CsrA. Genetic and molecular studies of CsrA have shown that it represses gluconeogenesis, glycogen biosynthesis and catabolism, and activates glycolysis and acetate metabolism (Romeo and Gong, 1993; Romeo et al., 1993; Sabnis et al., 1995; Yang et al., 1996; Wei et al., 2000). In addition, CsrA regulates biofilm formation in E. coli (Romeo et al., 1993).

Although the mechanism of positive regulation by CsrA has not been examined previously, its repression of glycogen biosynthesis has been elucidated in some detail. CsrA recognizes and binds specifically to glgCAP mRNA in the vicinity of the glgC ribosome binding site, thereby accelerating the rate of decay of glg transcripts. This, in turn, decreases the intracellular levels of the glycogen biosynthetic enzymes, which decreases the rate of glycogen biosynthesis (Romeo et al., 1993; Liu et al., 1995; Yang et al., 1996; Liu and Romeo, 1997). When recombinant CsrA was purified from E. coli, it was found to be bound to a 360 nucleotide (nt) non-coding RNA molecule, CsrB, in a large ribonucleoprotein complex containing ≈ 18 CsrA protein subunits (Liu et al., 1997). CsrB RNA is not required for the repression of glg gene expression by CsrA but, instead, antagonizes CsrA activity. Likewise, each of the regulatory functions of Csr that have been examined can be ascribed directly to the RNA-binding protein CsrA (reviewed by Romeo, 1998).

The broad role of CsrA in E. coli physiology, and the observation that csrA homologues of Borrelia burgdorferi (Fraser et al., 1997) and Bacillus subtilis (Mirel and Chamberlin, 1989) reside within flagellar gene clusters, prompted us to examine the effects of CsrA on motility in E. coli. The results of this study show that the csrA gene is required for motility and flagellum biosynthesis through the post-transcriptional activation of flhDC expression. This involves the binding to and stabilization of the flhDC message by the CsrA protein. Thus, the RNA-binding protein CsrA is similar to many DNA-binding proteins in that it is capable of functioning directly as either an activator or a repressor of gene expression, depending upon the target nucleic acid that it recognizes.


A functional csrA gene is essential for motility and flagellum biosynthesis in E. coli

Growth on semi-solid tryptone agar revealed that the csrA wild-type strains BW3414 and MG1655 were motile, whereas their isogenic csrA mutants were non-motile (Fig. 1A). A minimal plasmid clone of the csrA gene in pUC19, pCSR10, complemented the motility defect. Light microscopy also confirmed that parental strains grown in tryptone broth to the late exponential phase were motile, whereas csrA mutants were non-motile (data not shown). The csrA wild-type strain BW3414 was also tested and found to be motile in LB semi-solid medium (0.35% agar) containing sodium acetate (10 mM), sodium succinate (10 mM) or in the absence of any additional carbon source, but was non-motile in LB medium containing glucose (10 mM). Its isogenic csrA mutant was non-motile under all these conditions (data not shown). Finally, a precise chromosomal deletion of the csrB gene of BW3414 yielded a strain that was fully motile (data not shown), indicating that, although CsrA is essential for motility, CsrB RNA is not required.

Figure 1.

Motility and electron microscopy (EM) of csrA wild-type and mutant strains.

A. Motility of csrA wild-type, mutant (TR1-5) and overexpressing (pCSR10) strains of E. coli on semi-solid tryptone agar (0.35%) after 14 h growth at 30°C. 1, BW3414; 2, TR1-5BW3414; 3, TR1-5BW3414[pUC19] (vector control for pCSR10); 4, TR1-5BW3414[pCSR10]; 5, MG1655; 6, TR1-5MG1655; 7; TR1-5MG1655[pUC19]; 8 TR1-5MG1655[pCSR10].

B. Negative-staining EM of BW3414 grown to late exponential phase in tryptone broth.

C. EM of TR1-5BW3414 (as in B). Note the absence of flagella in this strain.

D. Flagella were counted under direct EM examination.

The loss of motility resulting from the csrA mutation might have resulted from either the absence of flagella or an inability to use flagella (a Mot-deficient phenotype). Negative-staining electron microscopy demonstrated that flagella were uniformly absent from csrA mutant cells (Fig. 1B and C). In one experiment, we observed that BW3414 exhibited an average of one flagellum cell−1, and some cells contained two or three flagella, whereas csrA mutant cells grown under identical conditions lacked flagella completely (Fig. 1D).

Effects of csrA on the expression of the flhDC operon

The complete absence of flagella in the csrA mutant suggested that flagellum biosynthesis may be regulated by csrA. As FlhDC is the master regulator of flagellum biosynthesis, the expression of a chromosomally encoded flhDC′–′lacZ translational fusion was monitored in isogenic csrA wild-type and mutant strains. In previous studies, it was noted that csrA has no effect on the expression of the wild-type chromosomal lacZ gene, validating the use of ′lacZ reporter fusions to monitor csrA regulation (Yang et al., 1996). Specific β-galactosidase activity from a genomic flhDC′–′lacZ fusion in each strain that was tested was low early in the exponential growth phase, increased throughout the exponential phase to reach a maximum in the late exponential phase, and declined thereafter (Fig. 2). Although the temporal pattern of expression was not altered by the csrA mutation, β-galactosidase activity was considerably higher in the csrA wild-type strains and exhibited a maximal difference of ≈ fourfold during the late exponential phase in two different strain backgrounds. These results suggested that CsrA affects motility and flagellum biosynthesis by positively regulating the expression of the flhDC operon. In addition, a multicopy plasmid containing the flhDC operon, pPM61, complemented the motility defect of the csrA mutant TR1-5MG1655 (data not shown), reinforcing this notion.

Figure 2.

Effects of csrA on the expression of chromosomal flhDC′–′ lacZ translational fusions. Cultures were grown in LB medium (minus glucose) at 30°C on a gyratory shaker (250 r.p.m.). Turbidity readings of cultures (A600) are indicated by circles, and average β-galactosidase activities from two independent experiments (i.e. four values were used to calculate each average) are shown as squares. For each value shown, one standard deviation was ≤ 12% of the mean.

A and B. BW3414 and CF7789 backgrounds, respectively, and open and solid symbols represent csrA wild-type and mutant strains respectively.

CsrA protein stimulates coupled in vitro transcription–translation of flhDC

The CsrA protein only has been demonstrated to function directly as a post-transcriptional repressor (Romeo, 1998). Thus, the effects of the csrA mutation on flhDC expression might have resulted from either a direct role of CsrA as an activator or an indirect role, e.g. through the negative regulation of a flhDC repressor. In order to clarify this issue, the effects of a recombinant CsrA protein on coupled in vitro transcription–translation of flhDC were examined in S-30 extracts prepared from a CsrA-deficient strain. Control reactions monitoring glg gene expression revealed that the CsrA protein was fully active in glg repression, as shown previously (Liu and Romeo, 1997). Initial attempts to demonstrate the synthesis of the native FlhD and FlhC polypeptides were unsuccessful, because the molecular masses of these two proteins are similar to predominant vector-encoded gene products, which obscured the autoradiographic analyses. Therefore, plasmid pFDCZ6 was constructed, which contained a flhDC′–′lacZ translational gene fusion capable of encoding a 117 kDa FlhC–LacZ fusion protein, which could be resolved from vector-encoded proteins. β-Galactosidase activity expressed from this plasmid-encoded gene fusion in vivo was activated ≈ threefold by csrA (data not shown). The in vitro synthesis of the FlhC–LacZ fusion protein was also stimulated up to ≈ threefold upon the addition of recombinant CsrA (provided as the CsrA–CsrB ribonucleoprotein complex), the effects of which saturated at 0.7–1.0 µg (2.8–3.8 µM) of monomeric CsrA protein (Fig. 3A–C). The identity of the FlhC–LacZ hybrid protein encoded by pFDCZ6 was confirmed by the construction and analysis of pFDCΔZ1, in which 1443 bp of ′lacZ was deleted internally. This plasmid encodes a 63 kDa truncated fusion protein, whose synthesis was stimulated by CsrA (Fig. 3A, lanes 5 and 6). The transcription–translation studies indicated that CsrA directly activates the expression of flhDC. A csrA ompR mutant was observed to be non-motile (data not shown), demonstrating that the effect of csrA on motility is not mediated indirectly, through an effect on OmpR, a repressor of flhDC.

Figure 3.

Effects of purified CsrA–CsrB complex on the coupled transcription–translation of flhDC′–′lacZ translational fusions. Reactions (35 µl) contained 2 µg of plasmid DNA and were conducted in an S-30 extract from TR1-5BW3414 (csrA::kanR).

A. Lanes 1 and 2 depict reactions containing the plasmid vector pMLB1034; lanes 3 and 4, pFDCZ6; lanes 5 and 6, pFDCΔZ1. One microgram of CsrA protein (3.8 µM monomer or 0.21 µM CsrA–CsrB complex) was present in lanes 2, 4 and 6 only.

B. The effects of CsrA–CsrB on the transcription–translation of the flhDC′–′lacZ fusion of pFDCZ6.

C. The incorporation of [35S]-methionine into the FlhC–LacZ hybrid protein from pFDCZ6 was determined by liquid scintillation counting, as described previously (Romeo and Preiss, 1989). The basal reaction, to which no CsrA–CsrB was added, synthesized 14 fmol of the FlhC–LacZ fusion protein.

Effects of csrA on flhDC mRNA steady-state levels and stability

Initial attempts to investigate the effects of csrA on flhDC mRNA levels in vivo by Northern blot analysis were unsuccessful because of the low abundance of this transcript. Therefore, reverse transcriptase–polymerase chain reaction (RT–PCR), a more sensitive method, was chosen for the analyses. Previous studies had revealed that levels of the Krebs cycle enzyme, isocitrate dehydrogenase (encoded by icd), are not affected by csrA (Wei et al., 2000). Thus, icd mRNA was chosen as an internal control for these analyses. Preliminary experiments revealed that icd mRNA was detectable after 20 cycles of PCR amplification and exhibited a linear range of detection with respect to input RNA up to 25 cycles (data not shown). In contrast, flhDC mRNA was detected only after 25 cycles of amplification and exhibited a linear range of detection up to 35 cycles (data not shown). Figure 4 shows the analysis of flhDC and icd transcripts from mid-exponential, late exponential and early stationary growth phases of csrA wild-type and mutant strains at 30 cycles of amplification. Under this condition, flhDC mRNA was not detected in either strain growing at mid-exponential stage or in the csrA mutant in stationary phase (Fig. 4, lanes 2, 3 and 7). At the late exponential phase, flhDC transcripts were detected in both strains and were ≈ threefold more abundant in the csrA wild-type strain (Fig. 4, lanes 4 and 5). In three repetitions of this experiment, flhDC transcript levels ranged from 2.5- to fourfold more abundant in the wild-type strain relative to the csrA mutant. In contrast, icd transcript levels were not affected by csrA, as observed at 25 cycles of amplification for these same RT reactions (data not shown).

Figure 4.

Detection of flhDC mRNA by RT–PCR.

A. Total RNA was isolated from mid-exponential (lanes 2 and 3), late exponential (lanes 4 and 5) or stationary growth phase (lanes 6 and 7); 0.5 µg of RNA was reverse transcribed, and the resulting cDNA was amplified by PCR for 30 cycles. Lanes 2, 4 and 6 show results with RNA from BW3414; lanes 3, 5 and 7 show reactions with RNA from TR1-5BW3414(csrA::kanR). Lane 1 shows the direct PCR amplification of flhDC from the plasmid pPM61, using the same primers as for RT–PCR analysis. Lane 8 contains RNA size standards.

B. Densitometric analysis of the flhDC data depicted in (A).

Previous studies revealed that the differences in the steady-state levels of glgC transcripts in csrA wild-type and mutant strains are the result of the effects of CsrA on mRNA decay rates (Liu et al., 1995). To determine whether flhDC transcript stability is affected by csrA in vivo, it was quantified by RT–PCR after the addition of rifampicin to cultures (Fig. 5). The flhDC transcripts were almost threefold more stable in csrA wild-type versus mutant cells (the chemical half-lives were ≈ 3.5 and 1.4 min respectively), which should account for differences in transcript levels in csrA wild-type and mutant strains.

Figure 5.

Analysis of flhDC mRNA stability in csrA wild-type and mutant strains by RT–PCR.

A. Cultures of BW3414 and TR1-5BW3414 were grown to late exponential phase, treated with rifampicin and harvested at several time points thereafter. Total RNA was isolated, and 0.5 µg of RNA was used for cDNA synthesis in 20 µl reactions. Two microlitres of each reaction was subjected to PCR amplification for either 30 or 31 cycles, for cDNA from BW3414 or TR1-5BW3414 (csrA::kanR) respectively.

B. The flhDC PCR products were quantified by densitometry. Mean values derived from two independent experiments are shown.

Mapping of the flhDC transcript initiation site

Previous studies to identify the 5′ end of E. coli flhDC mRNA were conducted on flhDC transcripts expressed in vivo from plasmid clones (Shin and Park, 1995; Soutourina et al., 1999). The first study identified one major and two minor transcripts, none of which contained an apparent flhDC promoter sequence. The second study revealed a single major transcript. In order to map the 5′ end of the chromosomally encoded transcript, primer extension analysis was conducted with RNA isolated from plasmid-free csrA wild-type and mutant strains, using two different primers (flh1 and flh2; Table 2). Both primers revealed the same results. Only a single major product was identified, which was detectable only in the csrA wild-type strain (Fig. 6A). The 5′ end of this transcript was mapped to a G residue 197 bp upstream from the flhD initiation codon (Fig. 6A and B), which was the same site found by Soutourina et al. (1999). The apparent −10 and −35 promoter boxes for this transcript exhibit 67% and 50% identity with σ70 consensus sequences respectively. These hexamers and the transcript initiation region exhibit ≈ 90% identity with those of the Proteus mirabilis flhD gene (Furness et al., 1997; Fig. 6B).

Table 2. Oligonucleotide primers used in this studya.
NameSequence (5′ to 3′)
  1. a. All primers were purchased from Integrated DNA Technologies.

Figure 6.

Primer extension analysis of flhDC mRNA.

A. Results of transcript mapping experiments using primers flh1 or flh2 are shown in the left or right respectively. Lanes 1 and 2 of each experiment used RNA from strain BW3414 or TR1-5BW3414 (csrA::kanR) respectively. The DNA sequencing reactions were conducted with the corresponding primers and pPM61 as the plasmid template. The sequences shown are those of the coding stand.

B. Alignment of the E. coli K-12 flhDC promoter sequence with that of Proteus mirabilis.

CsrA binds specifically to flhDC transcripts

RNA gel mobility shift assays were used to determine whether CsrA protein recognizes and specifically binds to flhDC transcripts. Three 32P-radiolabelled run-off transcripts, A, B and C, were prepared and isolated (Fig. 7). Transcript A included only upstream non-coding RNA, extending from +1 to +146 nt relative to the start of transcription. Transcript B extended from +147 to +276 and included the remaining upstream non-coding RNA segment and 79 nt of the coding region. Transcript C extended from +1 to +276 nt. Each of the transcripts generated retarded complexes in the presence of the purified recombinant CsrA–CsrB complex. The relative mobility shifts observed at the lowest concentrations of CsrA were modest and could only be explained by transfer of CsrA protein from the CsrA–CsrB complex to the labelled transcripts. As increasing amounts of CsrA–CsrB were added to the binding reactions, additional shifted complexes with decreasing mobilities were observed for transcripts A and C. Transcript B apparently formed only a single binding complex under these conditions and appeared to be bound with lower affinity than A and C. Unlabelled flhDC transcripts competed for CsrA binding in each case, whereas RNA from Saccharomyces cerevisiae did not compete for binding with any of the labelled transcripts. Binding of transcript B was further tested for competition by an unlabelled icd transcript. This non-specific transcript competed only very weakly at concentrations under which cold flhD transcript B completely inhibited binding (data not shown). In the absence of CsrA–CsrB complex (lanes 1 in Fig. 7A–C), there was only one major band, representing the transcript of interest. However, preparations of transcripts B and C each contained a minor contaminating transcript with slightly lower mobility than the expected product. CsrA did not appear to bind to either of these minor transcripts.

Figure 7.

RNA mobility shift assay with flhDC run-off transcripts. A radiolabelled flhDC run-off transcript (0.5 nM), A, B or C as depicted in (D), was incubated in the presence or absence of CsrA–CsrB complex at various concentrations and analysed on a 5% native polyacrylamide gel.

A. Reactions in lanes 1–5 contained 0, 1.0, 2.5, 5.0 or 10 µM CsrA monomer respectively. Lanes 6–9 depict competition experiments conducted using 0.5 nM labelled flhDC transcript A and 5.0 µM CsrA monomer (lane 4 conditions). The reactions shown in lanes 6 and 7 were incubated in the presence of the unlabelled run-off transcript A at 100- and 400-fold excess relative to the labelled transcript respectively. Lanes 8 and 9 contained Saccharomyces cerevisiae total RNA at 100- and 400-fold excess relative to the mass of the labelled transcript.

B. Reactions in lanes 1–4 contained 0, 1.0, 5.0 or 10 µM CsrA monomer. Lanes 5–8 show competition experiments using reactions containing 0.5 nM labelled flhDC transcript B and 5.0 µM CsrA monomer (lane 3 conditions). The reactions in lanes 5 and 6 contained unlabelled run-off transcript B at 100- and 400-fold excess relative to the labelled transcript respectively. Lanes 7 and 8 contained Saccharomyces cerevisiae total RNA at 100- and 400-fold mass excess.

C. Reactions in lanes 1–5 contained 0, 0.1, 1.0, 5.0 or 10 µM CsrA monomer respectively. Lanes 6–9 depict competition experiments using reactions containing 0.5 nM labelled flhDC transcript C and 5.0 µM CsrA monomer (lane 4 conditions). The reaction mixtures in lanes 6 and 7 contained unlabelled run-off transcript C at 200- and 500-fold excess relative to the mass of the labelled transcript respectively. Lanes 8 and 9 contained S. cerevisiae total RNA at 200- and 500-fold excess.

D. Run-off transcripts used for mobility shift experiments are shown aligned with the 5′ end of the flhDC message.


Considerable evidence indicates that the decision to become motile in E. coli and related species is determined by the expression level of the master operon, flhDC. Genes encoding global regulatory factors generally exhibit abundant regulatory features, and this is true for flhDC. The present study reveals that, in addition to transcriptional repression by phosphorylated OmpR (Shin and Park, 1995) and HdfR (Ko and Park, 2000) and activation by cAMP-CRP (Soutourina et al., 1999), flhDC is regulated post-transcriptionally by CsrA-mediated transcript stabilization. Interestingly, recent studies by Claret and Hughes (2000) have revealed that the turnover of the FlhD and FlhC polypeptides of Proteus mirabilis is rapid and is controlled by a mechanism that involves Lon protease. A short half-life for FlhD (≈ 2 min) is consistent with its regulatory role in cell cycle control (Prüß and Matsumura, 1997). Finally, the heat shock proteins DnaJ, DnaK and GrpE appear to affect flhDC expression, as well as the assembly of the mature FlhDC protein, presumably via their chaperon functions (Shi et al., 1992). Thus, there is currently evidence that flhDC expression is regulated at the levels of transcription, message stability and possibly translation (discussed below), and that the assembly and turnover of the FlhDC protein are also strictly controlled.

The series of experiments described in the present study reveals that the global regulator CsrA post-transcriptionally activates motility and flagellum biosynthesis in E. coli and that the flhDC transcript is a direct target of regulation. First, csrA::kanR mutant cells were non-motile because of the absence of flagella, and this phenotype was complemented by plasmids encoding the csrA gene or flhDC. Secondly, the expression of a chromosomally encoded flhDC′–′lacZ translational fusion was decreased by ≈ fourfold in the csrA mutant. This effect on flhDC expression should be sufficient to account for the total loss of flagella in mutant cells (Kutsukake, 1997). Thirdly, flhDC message levels were ≈ threefold higher and were more stable in a csrA wild-type strain relative to its isogenic mutant. Fourthly, a recombinant CsrA protein activated the in vitro transcription–translation of a flhDC′–′lacZ translational fusion and bound specifically to run-off transcripts that contained flhDC upstream non-coding RNA. During the later stages of this study, both csrA wild-type and mutant strains were found to be motile in CFA medium (data not shown). Interestingly, flhDC′–′lacZ expression in cells grown in CFA medium was still positively regulated by CsrA and was two- to threefold higher in both wild-type and csrA mutant cells grown in CFA versus LB or tryptone broth (data not shown). The increased basal expression of flhDC was apparently sufficient for flagellum biosynthesis in the mutant. The means by which growth in CFA medium increases flhDC expression is unknown. Nevertheless, the observations from this study tend to reinforce the concept that the absolute level of flhDC expression is critical for the development of motility.

We hypothesized previously that CsrA is part of an adaptive response pathway (Romeo et al., 1993), which carries the implicit assumption that CsrA is regulated in response to environmental or physiological conditions. Although the specific factors that regulate csrA expression remain to be defined, it is interesting to note that both CsrA protein levels and csrA::lacZ expression increase during the growth curve and reach a maximum at the late exponential phase/transition to stationary phase (S. Gudapaty, K. Suzuki, X. Wang, P. Babitzke and T. Romeo, unpublished data). Although flhDC transcript levels and flhDC′–′lacZ expression were decreased in a csrA mutant, they otherwise retained normal temporal control and were optimal at the late exponential phase of the growth curve in both csrA wild-type and mutant strains. Thus, in the cell, the CsrA protein appears to be poised to bind to and stabilize flhDC transcripts as they accumulate in the late exponential phase.

The precise mechanism by which CsrA binding stabilizes flhDC mRNA remains to be defined. Studies on the regulation of protein synthesis have shown that RNA secondary structure present in the 5′ untranslated segment can dramatically influence both mRNA stability and translation initiation. Furthermore, the rate-limiting step in mRNA decay is often an initial endonucleolytic cleavage at the 5′ extremity (reviewed by Kushner, 1996; Grunberg-Manago, 1999). Both the long 5′ untranslated segment (197 nucleotides) of flhDC mRNA and the complex mobility shift results observed here suggest that the molecular details of this regulation are complex. CsrA can interact specifically with both the far upstream non-coding segment and either the distal segment of the non-coding segment or the proximal flhD coding region. Previous genetic and molecular studies revealed that the segment of the glgC message in the vicinity of the ribosome binding site is involved in negative control (Liu et al., 1995; Liu and Romeo, 1997). These observations are consistent with a role for CsrA as a translational repressor, which subsequently destabilizes the glgC transcript (reviewed by Romeo, 1998). Likewise, it is plausible that CsrA binding activates flhDC translation, although the available data could be explained by a mechanism involving direct protection of flhDC mRNA by CsrA against endonucleolytic attack. The relatively poor Shine–Dalgarno sequence for flhD (Soutourina et al., 1999) is also consistent with the idea that ribosome binding at flhD might be favoured by an accessory factor.

It is becoming increasingly clear that CsrA homologues perform central regulatory roles in host–microbe interactions. Certainly, the effect of CsrA on motility has implications for a variety of motility-dependent host interactions (reviewed by Ottemann and Miller, 1997; Young et al., 1999). The E. coli csrA gene has strong effects on adherence and biofilm formation (Romeo et al., 1993). The formation of an adherent, matrix-enclosed and protected population of bacterial cells, i.e. biofilm, within the human host presents a serious problem in numerous bacterial infections (e.g. Costerton et al., 1999). Studies by Arun Chatterjee and his coworkers have documented an important role for the highly conserved CsrA homologue, RsmA, in the virulence mechanisms of the plant pathogen Erwinia carotovora, a close relative of E. coli (Chatterjee et al., 1995; Cui et al., 1995). This bacterium uses RsmA as a repressor of genes encoding a battery of secreted lytic enzymes that are largely responsible for soft rot disease. The genes repressed by RsmA are expressed during the stationary phase of growth, similar to those that are repressed by CsrA in E. coli. In contrast to the role of CsrA in motility, the overexpression of rsmA inhibited motility of Erwinia species (Mukherjee et al., 1996). This would seem to suggest distinct physiological roles for motility in these two bacteria. However, caution is warranted in the latter case, as a mutation in rsmA was not tested, and the mechanistic basis for the effects of rsmA overexpression on motility have not yet been examined (discussed by Romeo, 1998). More recently, reports have shown that Salmonella typhimurium uses CsrA and CsrB to regulate genes involved in invasion of the intestinal mucosa, including hilA, invF, prgH and sspC (Altier et al. 2000a, b). In addition, Pseudomonas fluorescens regulates genes involved in the production of extracellular protease activity and hydrogen cyanide production via its csrA homologue, rsmA (Blumer et al., 1999). Finally, FlhD or FlhDC regulate functions in addition to motility, including the production of host-damaging enzymes such as lipase and haemolysin in Xenorhabdus nematophilus (Givaudan and Lanois, 2000) and genes involved in cell division (Prüßet al., 1997) and respiration (B. Prüß, W. Hendrickson, X. Liu and P. Matsumura, submitted for publication) in E. coli. Based on the results of the present study, CsrA may be predicted indirectly to control a variety of processes that are regulated by FlhD or FlhDC.

Experimental procedures

Bacterial stains, plasmids, media and growth conditions

Table 1 lists the strains and plasmids that were used in this study, their sources and relevant genotypes. Luria–Bertani (LB) medium (1% tryptone, 0.5% yeast extract, 1% NaCl, pH 7.4) supplemented with 0.2% of glucose was used for routine laboratory cultures. For motility studies, LB medium was prepared without glucose. Tryptone broth contained 1% tryptone and 0.5% NaCl, pH 7.4. Colonization factor antigen (CFA) medium contained 1% casamino acids, 0.15% yeast extract, 0.005% MgSO4 and 0.0005% MnCl2, pH 7.4 (Evans et al., 1977). Media were supplemented with the following compounds as required: kanamycin, 100 µg ml−1; tetracycline, 10 µg ml−1; and ampicillin, 100 µg ml−1. Cultures were inoculated with one volume of overnight culture per 500 volumes of freshly prepared medium and grown at 30°C on a gyratory shaker at 250 r.p.m.

Table 1. Bacterial strains and plasmids used in this study.
Strain, plasmidDescriptionSource or reference
  • a

    . A strain containing the prefix TR1-5 indicates that the csrA::kanR allele was introduced by P1vir transduction.

E. coli strains
 BW3414 ΔlacU169 rpoS(Am)B. Wanner
 MG1655PrototrophicM. Cashel
 CF7789MG1655 ΔlacI-Z (MluI)M. Cashel
 MC4100FΔ(argF-lac)U169 rpsL relA flhD deoC ptsF rbsR Shin and Park (1995)
 CP992Source of genomic ΦflhDC′–′lacZ Shin and Park (1995)
 TK821MC4100 ompR331::Tn10 Garrett et al. (1983)
 TKBW3414BW3414 ompR331::Tn10This study
 TR1-5a csrA::kanR Romeo et al. (1993)
 pUC19Cloning vector; AmpR Yanisch-Perron et al. (1985)
 pCSR10 csrA in pUC19 Romeo et al. (1993)
 pPM61Contains flhDC; KanR AmpR Bartlett et al. (1988)
 pMBL1034 ′lacZ fusion vector; AmpR Silhavy et al. (1984)
 pFDCZ6Φ(flhDC′–′lacZ)6(hyb) in pMBL1034This study
 pFDCΔZ1 AvaI deletion within ′lacZ of pFDCZ6This study

Motility assays

The plate assay was initiated by stabbing a colony from an overnight culture into semi-solid agar (tryptone broth solidified with 0.35% agar). The plates were kept in a humidified incubator at 30°C for 14 h. Swimming motility of liquid cultures was examined by light microscopy.

Electron microscopy

Bacterial cells were grown in tryptone broth to late exponential phase, negatively stained using 1% (v/v) phosphotungstic acid, pH 7.2, and mounted on Formvar- and carbon-coated 200-mesh nickel grids (Mukherjee et al., 1996). Samples were visualized using a Zeiss EM 910 transmission electron microscope.

Total protein and β-galactosidase assays

β-Galactosidase activity was determined using duplicate assays at each time point by measuring the conversion of ONPG to o-nitrophenol, as described previously (Romeo et al., 1990). Specific activity was calculated with respect to total cell protein (Smith et al., 1985) using bovine serum albumin (BSA) as a standard protein.

Molecular biology techniques and nucleotide sequencing

Standard procedures were used for the isolation of plasmid DNA and restriction fragments, restriction mapping, transformation, P1vir transduction of genomic mutations and fusions, molecular cloning and PCR amplification as described previously (Romeo and Preiss, 1989; Sambrook et al., 1989; Ausubel et al., 1999). Dideoxynucleotide sequencing reactions (Sanger et al., 1977) were conducted using Sequenase version 2.0 under the conditions described by the manufacturer (US Biochemicals).

Preparation of plasmid-encoded flhDC′–′lacZ translational fusions

Plasmid pFDCZ6 contained a flhDC–lacZ translational fusion in which codon 14 of flhC′ was fused in frame to codon 6 of the recombinant ′lacZ gene in pMLB1034. It was constructed by subcloning a 1.8 kb PvuII fragment from pPM61 (Bartlett et al., 1988), which included the complete upstream regulatory region of the flhDC operon, an intact flhD coding region and the N-terminus of flhC, into the SmaI site of pMLB1034 (Silhavy et al., 1984). To construct a plasmid encoding a truncated FlhC–LacZ fusion protein, pFDCΔZ1, plasmid pFDCZ6 was digested with AvaI to generate a 1.5 kb fragment of the lacZ coding region, a 0.5 kb fragment containing the distal coding region of lacZ and the upstream coding region of lacY and a 6 kb vector fragment. The 0.5 kb and 6 kb fragments were gel purified and ligated together to produce a plasmid encoding a FlhC–LacZ fusion that was missing 481 internal amino acids, which was characterized by restriction analysis.

S-30 coupled transcription–translation reactions

The effects of a recombinant CsrA protein containing a carboxy-terminal his tag (Liu et al., 1997) on the expression of flagellum biosynthesis genes were examined in S-30 extracts prepared from TR1-5BW3414 (csrA::kanR), as described previously (Romeo and Preiss, 1989; Liu and Romeo, 1997). The recombinant protein was provided as the CsrA–CsrB ribonucleoprotein complex, prepared from the cell (Liu et al., 1997). Proteins were labelled by incorporation of [35S]-methionine (1175 Ci mmol−1; NEN Life Science Products), denatured, and equal volumes of each reaction were subjected to electrophoresis on 9.5% SDS–polyacrylamide slab gels. Radiolabelled proteins were detected by fluorography using sodium salicylate (Chamberlain, 1979).

RNA isolation and analysis

RNA was purified using MasterPure RNA purification kits, according to the manufacturer's instructions (Epicentre Technologies; Watson et al., 1998). RNA was quantified by absorbance at 260 nm and 280 nm. For molecular mass estimation, purified RNA was subjected to denaturing electrophoresis on 1.2% agarose gels containing 2.2 M formaldehyde (Sambrook et al., 1989).

RT–PCR analysis of flhDC mRNA

Total RNA was isolated from cells grown at 30°C in LB medium (lacking glucose) to mid-exponential, late exponential or stationary growth phase. Subsequently, 0.5 µg of RNA was reverse transcribed using 15 U of ThermoScript RT (Gibco BRL) in a total reaction volume of 20 µl. Each reaction was incubated at 55°C for 50 min and terminated by incubation at 85°C for 5 min. One microlitre of RNase H (2 units µl−1) was added to the reaction mixture and incubated at 37°C for 20 min. Two microlitres of reverse transcription products (cDNA) were amplified by PCR using 2.5 U of HotStarTaq DNA polymerase (Qiagen) and 25 pmol each of flhDC forward and reverse primers and icd forward and reverse primers (flh-RT1, flh-RT2, icd-RT1 and icd-RT2 respectively; Table 2) in a total volume of 100 µl. Each PCR programme started with an initial heat activation step at 95°C for 15 min. A typical PCR cycle consisted of a denaturation step (94°C, 1 min), an annealing step (59°C, 1 min) and an elongation step (72°C, 1 min). Typical reactions included several full cycles (20–35), followed by an additional extension step (72°C, 10 min). The PCR products were 202 bp and 139 bp for flhDC and icd respectively. PCR products were separated by electrophoresis on 2% agarose gels containing 0.5 µg ml−1 ethidium bromide. They were subsequently visualized and photographed under UV light and quantified by densitometry using an AlphaImager 2000 documentation and analysis system (AlphaInnotech).

mRNA stability determination

Bacterial cultures were grown at 30°C in LB medium (lacking glucose) to late exponential phase and treated with rifampicin at a final concentration of 200 µg ml−1 to inhibit transcription. Samples were collected at 0, 2, 4, 6, 8 and 12 min after rifampicin treatment. The cells were harvested at 13 000 r.p.m. in a microcentrifuge and frozen in dry ice, allowing no more than 2 min to elapse between sampling and freezing. Total RNA was extracted, and flhDC and icd mRNA levels were determined by RT–PCR analysis.

Primer extension mapping

Total RNA was prepared from cultures grown at 30°C in LB (lacking glucose) to the late exponential phase of growth. Two oligonucleotide primers were used. Primer flh1 was complementary to +175 to +194; primer flh2 was complementary to +89 to +110 of flhDC mRNA (Table 2). The primers were end labelled using T4 polynucleotide kinase and [32P]-ATP (3000 Ci mmol−1, NEN Life Science Products) according to standard procedures (Ausubel et al., 1999). Approximately 0.5 ng of labelled primer was annealed to 20 µg of total RNA. cDNA was synthesized using 15 U of ThermoScript RT (Gibco BRL) in a total volume of 20 µl, and the reaction was conducted at 55°C for 50 min and terminated by incubation at 85°C for 5 min. One microlitre of RNase H (2 units µl−1) was added to the reaction mixture and incubated at 37°C for 20 min. The same two primers were used to prepare DNA sequencing ladders that served as standards for the corresponding reverse transcriptase signals, which were analysed on urea-containing sequencing gels (e.g. Romeo and Preiss, 1989).

RNA gel mobility shift assay

Three run-off flhDC transcripts, A (146 nt), B (130 nt) and C (276 nt), and an icd transcript (139 nt) were synthesized from PCR products using T7 RNA polymerase. The primers for the synthesis of these PCR products are shown in Table 2. The resulting PCR products each contained a T7 promoter, followed by: 146 bp (+1 to +146 relative to the flhDC transcription start site using flh-MS1 and flh-MS3); 130 bp (+147 to +276, using flh-MS2 and flh-MS4); 276 bp (+1 to +276, using primers flh-MS1 and flh-MS4); and 139 bp (+105 to +244 relative to the icd coding region, using primers icd-MS1 and icd-RT2) for flhD transcripts A, B and C and the icd transcript respectively. The PCR products were gel purified and used as templates for in vitro transcription. 32P-labelled transcripts were prepared according to the method described previously (Liu and Romeo, 1997). The RNA binding reaction mixtures (20 µl) included CsrA–CsrB complex at various concentrations, labelled transcript (0.5 nM), 10 mM Tris-acetate (pH 7.5), 10 mM MgCl2, 50 mM NaCl, 50 mM KCl, 10 mM dithiothreitol and 5% glycerol (Alifano et al., 1992). The reactions were incubated at room temperature for 30 min, mixed with 1 µl of loading dye (97% glycerol, 0.01% bromophenol blue, 0.01% xylene cyanol) and immediately loaded and separated on 5% polyacrylamide vertical slab gels (Liu and Romeo, 1997). The resulting gels were dried onto filter paper and subjected to autoradiography.


We thank Harlan P. Jones for advice and assistance with the RT–PCR analyses, Seshagirirao Gudapaty for advice on the Northern analyses, Lawrence Oakford for assistance with figure preparation, and Michael Cashel for providing strains. This project was supported in part by grants from the National Science Foundation (MCB-9726197) and the National Institutes of Health (GM-59969). The Zeiss EM 910 microscope was purchased with funding from the National Science Foundation (BIR-9413907).