CsrA regulates glycogen biosynthesis by preventing translation of glgC in Escherichia coli

Authors

  • Carol S. Baker,

    1. Department of Biochemistry and Molecular Biology, Pennsylvania State University, University Park, PA 16802, USA.
    Search for more papers by this author
  • Igor Morozov,

    1. Department of Biochemistry and Molecular Biology, Pennsylvania State University, University Park, PA 16802, USA.
    Search for more papers by this author
    • Present address: Plant Science and Fungal Molecular Biology Research Group, School of Biological Sciences, Donnan Laboratories, The University of Liverpool, Liverpool L69 7ZD, UK.

  • Kazushi Suzuki,

    1. Department of Molecular Biology and Immunology, University of North Texas Health Science Center, Fort Worth, TX 76107, USA.
    Search for more papers by this author
  • Tony Romeo,

    1. Department of Molecular Biology and Immunology, University of North Texas Health Science Center, Fort Worth, TX 76107, USA.
    Search for more papers by this author
  • Paul Babitzke

    1. Department of Biochemistry and Molecular Biology, Pennsylvania State University, University Park, PA 16802, USA.
    Search for more papers by this author
    • *For correspondence. E-mail pxb28@psu.edu; Tel. (+1) 814 865 0002; Fax (+1) 814 863 7024.


Summary

The carbon storage regulatory system of Escherichia coli controls the expression of genes involved in carbohydrate metabolism and cell motility. CsrA binding to glgCAP transcripts inhibits glycogen metabolism by promoting glgCAP mRNA decay. CsrB RNA functions as an antagonist of CsrA by sequestering this protein and preventing its action. In this paper, we elucidate further the mechanism of CsrA-mediated glgC regulation. Results from gel shift assays demonstrate that several molecules of CsrA can bind to each glgC transcript. RNA footprinting studies indicate that CsrA binds to the glgCAP leader transcript at two positions. One of these sites overlaps the glgC Shine–Dalgarno sequence, whereas the other CsrA target is located further upstream in an RNA hairpin. Results from toeprint and cell-free translation experiments indicate that bound CsrA prevents ribosome binding to the glgC Shine–Dalgarno sequence and that this reduces GlgC synthesis. The effect of two deletions in the upstream binding site was examined. Both of these deletions reduced, but did not eliminate, CsrA binding in vitro and CsrA-dependent regulation in vivo. Our findings establish that bound CsrA inhibits initiation of glgC translation, thereby reducing glycogen biosynthesis. This inhibition of translation probably contributes to destabilization of the glgC transcript that was observed previously.

Introduction

The ability of microorganisms to respond rapidly to changes in the environment is crucial for their survival. Bacteria have developed several regulatory strategies that ensure their survival during the transition from exponential growth into stationary phase as a result of nutrient starvation. During this transition, the carbon storage regulatory system (Csr) of Escherichia coli plays an important role in controlling gene expression. The effector of Csr is a small RNA-binding protein, CsrA. Previous studies have established that CsrA represses gluconeogenesis, glycogen biosynthesis and catabolism and activates glycolysis, cell motility and acetate metabolism (Romeo and Gong, 1993; Romeo et al., 1993; Sabnis et al., 1995; Yang et al., 1996; Wei et al., 2000; 2001). In addition, CsrA regulates biofilm formation (Romeo et al., 1993; Jackson et al., 2002). CsrB RNA is a second component of the Csr regulatory system. This 366 nucleotide (nt) RNA functions as an antagonist of CsrA action by sequestering this protein (Liu et al., 1997; Gudapaty et al., 2001). Purification of His-tagged CsrA resulted in the isolation of a large ribonucleoprotein complex containing CsrB RNA complexed with approximately 18 molecules of CsrA. Eighteen repeated sequences in CsrB RNA that are present in predicted RNA hairpins, or between hairpins, are thought to be the sites of CsrA binding (Liu et al., 1997).

Glycogen biosynthesis in E. coli occurs primarily during the transition from exponential to stationary phase growth when a nutrient other than carbon is limiting. The glgCAP operon encodes ADP-glucose pyrophosphorylase (glgC), glycogen synthase (glgA) and glycogen phosphorylase (glgP) (Romeo, 1998). Expression of the glgCAP operon is activated by cAMP and CRP. As a consequence, expression of this operon increases gradually throughout exponential growth and reaches a plateau when cultures reach stationary phase (Romeo and Preiss, 1989; Romeo et al., 1990). The stringent response mediated by ppGpp also activates the expression of the glgCAP operon (Romeo and Preiss, 1989; Romeo et al., 1990). Previous genetic and biochemical studies demonstrated that CsrA binding to the glgCAP operon transcript controls the expression of this operon by promoting glgCAP mRNA decay (Romeo et al., 1993; Liu et al., 1995; Liu and Romeo, 1997). It was also shown that the addition of CsrA to a coupled cell-free transcription–translation system reduced GlgC synthesis (Yang et al., 1996; Liu and Romeo, 1997; Liu et al., 1997). CsrA is not a general inhibitor of translation since expression of other plasmid-encoded genes was not appreciably altered at a CsrA concentration that eliminated GlgC synthesis (Yang et al., 1996; Liu and Romeo, 1997; Liu et al., 1997).

In prokaryotic mRNAs, the Shine–Dalgarno (SD) sequence basepairs with the 3′ end of the 16S rRNA present in the 30S ribosomal subunit and correctly positions the corresponding initiation codon in the ribosome (Shine and Dalgarno, 1974; Steitz and Jakes, 1975). Translational control mechanisms have been identified in prokaryotes that involve sequestration of the SD sequence by RNA secondary structure in a protein-dependent manner (Du and Babitzke, 1998) or directly by a bound protein (Winter et al., 1987; Du et al., 1997; Vytvytska et al., 2000). It has also been shown that efficient translation of a message can stabilize the transcript by inhibiting nucleolytic degradation (Yarchuk et al., 1992; Vytvytska et al., 2000). We performed experiments to elucidate further the mechanism of CsrA-dependent glgC regulation. Our findings establish that CsrA binding to the glgCAP leader transcript inhibits ribosome binding to the glgC SD sequence. As it has been shown previously that CsrA binding promotes rapid degradation of the glgC transcript (Liu et al., 1995; Liu and Romeo, 1997), our findings suggest that CsrA-mediated inhibition of glgC translation promotes rapid degradation of the glgCAP transcript. Thus, bound CsrA reduces glycogen biosynthesis by inhibiting translation initiation of glgC and by promoting more rapid mRNA decay.

Results

CsrA purification

We purified CsrA and estimated that the protein was >95% pure (data not shown). The amino acid sequence of purified CsrA was confirmed by tryptic digestion followed by mass spectrometry of the fragments. No formyl group was present on the N-terminal methionine. Gel filtration of purified CsrA suggested that CsrA is multimeric. Preliminary results from mass spectroscopy and ana-lytical ultracentrifugation experiments suggest that CsrA exists as a dimer or a trimer (data not shown). Owing to the uncertainty of the oligomeric structure of CsrA, the concentrations reported in this study are for the monomeric CsrA polypeptide.

CsrA binds specifically to glgCAP leader RNA

The major glgCAP transcript contains a 134 nt untranslated leader. Previous gel mobility shift and deletion studies localized the presumed CsrA binding site within this leader sequence (Liu et al., 1995; Liu and Romeo, 1997). To characterize further the interaction of CsrA with the glgCAP leader transcript, we performed quan-titative gel shift assays with a glg transcript containing nucleotides +46 to +179 relative to the start of transcription. CsrA binding to this transcript was detected as a distinct band in native gels between 15 and 250 nM CsrA (Fig. 1). Non-linear least-squares analysis of these data yielded an estimated Kd value of 39 nM CsrA. As the concentration of CsrA was increased further, additional shifted species were observed. Twofold increases in CsrA concentration resulted in the disappearance of one species and the appearance of a slower migrating species. This gel shift pattern suggested that multiple CsrA molecules were bound to each transcript at these higher CsrA concentrations. Furthermore, as small twofold increases in CsrA concentration gave rise to new shifted species, our results suggest that the formation of these complexes is co-operative. Although we presume that the first shifted species contained one glg leader transcript and one molecule of CsrA, and that the additional shifted species contained multiple CsrA molecules, the stoichiometry of these species has not been examined. It is also worth pointing out that the slower migrating species could result from multiple CsrA–RNA and/or CsrA–CsrA interactions. Although our footprinting studies demonstrate that CsrA can bind to two RNA segments in the glgCAP leader transcript (see below), this does not rule out the possibility that CsrA–CsrA interactions play a role in CsrA action.

Figure 1.

Gel shift analysis of CsrA complexed with glg leader transcripts. 5′ end-labelled glg leader transcripts (0.5 nM) were incubated with the concentration of CsrA indicated at the bottom of each lane. Gel shift assays were performed in the absence or presence of various competitor RNAs. The concentration of competitor RNA is shown at the bottom of each lane. The positions of free (F) and bound (B) RNA are shown. * marks the position of an RNA–RNA complex in the second gel.

The specificity of the CsrA–glg leader RNA interaction was investigated by performing competition experiments with specific (glgCAP leader, flhDC leader and CsrB RNA) and non-specific (Bacillus subtilis trp leader) unlabelled RNA competitors (Fig. 1). CsrB RNA, which contains 18 putative CsrA binding sites, was the most effective competitor followed by glgCAP leader RNA and then flhDC leader RNA. As expected, CsrA–glg RNA complex formation was not competed by B. subtilis trp leader RNA. These results confirm and extend previous findings that CsrA binds specifically to a transcript containing the untranslated glg leader (Liu and Romeo, 1997). A slower migrating species appeared in the glg leader competition experiment as the concentration of cold competitor was increased (Fig. 1, *). As this molecular species was also observed in the absence of CsrA (last lane), it is apparent that this species results from interaction between labelled glg RNA and the unlabelled RNA competitor.

CsrA binding to the glgC transcript inhibits GlgC synthesis by preventing ribosome binding

We performed a toeprint analysis to identify the position of bound CsrA in the glg leader transcript. The presence of bound CsrA should block primer extension by reverse transcriptase, resulting in a toeprint band at a position near the 3′ boundary of CsrA. We observed three clustered CsrA toeprint bands (Fig. 2). These bands correspond to UGU located 21–23 nt upstream from the glgC SD sequence (Fig. 3). An additional CsrA-dependent toeprint band was observed further upstream (*). We do not know whether this is an authentic CsrA toeprint or if it is an RNA structural toeprint. As stable RNA structures can inhibit primer extension by reverse transcriptase (Du and Babitzke, 1998), it is possible that bound CsrA promotes the formation of a structure further upstream. We carried out similar toeprint experiments to identify the position of bound 30S ribosomal subunits. We observed a prominent tRNAfMet-dependent toeprint band 15 nt downstream from the A in the AUG initiation codon (Figs 2 and 3). We also carried out toeprint experiments to determine whether CsrA could inhibit ribosome binding. When CsrA was bound to the glgC transcript before the addition of ribosomes and tRNAfMet, we observed the CsrA toeprint bands, whereas the 30S ribosome toeprint band was eliminated (Fig. 2). These results demonstrate that bound CsrA inhibits ribosome binding. When the reverse experiment was performed with ribosomes bound before the addition of CsrA, we observed ribosome and CsrA toeprints. As each transcript can only give rise to a single reverse transcriptase product, these results suggest that some of the transcripts contained bound CsrA, whereas others contained bound ribosomes. Thus, our results do not imply that CsrA and 30S ribosomal subunits can bind simultaneously to the same transcript. However, it should also be pointed out that our toeprint data do not rule out this possibility. It is conceivable that CsrA could bind just upstream of a bound ribosome. It is likely that CsrA binding to a position that includes the glgC SD sequence plays a crucial role in inhibiting ribosome binding (see below).

Figure 2.

CsrA and 30S ribosomal subunit toeprints of glgCAP leader RNA. The presence, as well as the order of addition, of CsrA and/or 30S ribosomal subunits is shown at the top of each lane. Arrows indicate bands corresponding to CsrA or 30S ribosomal subunit toeprints. An additional CsrA-dependent toeprint is indicated (*). Positions of the glgC SD sequence and the AUG initiation codon are shown. Sequencing lanes to reveal U, G, A or C residues are marked.

Figure 3.

Summary of the glg leader toeprint, structure mapping and footprint results. This figure is adapted from the data presented in Figs 2 and 5. The positions of the CsrA toeprint, as well as the glgC SD sequence and start codon (Met), are shown in bold type. Positions of cleavage by the single-strand specific probes RNase T1 (G specific) and Pb2+ in the absence of bound CsrA are indicated by filled arrowheads and filled circles respectively. The structure of the hairpin loop in the absence of bound CsrA is shown. Nucleotides in which bound CsrA increases (+) or decreases (–) cleavage are indicated. Numbering is from the start of glgCAP transcription.

Previous studies have shown that a His-tagged CsrA protein specifically inhibits coupled transcription–translation of plasmid-encoded glg genes (Liu and Romeo, 1997; Liu et al., 1997). As our toeprint results indicated that CsrA competes with ribosomes for binding to the glgC transcript, we performed RNA-directed cell-free translation experiments to determine whether CsrA could inhibit GlgC synthesis from a pre-existing mRNA template. The effect of native CsrA on translation of glgC mRNA was initially examined using a transcript that included the 134 nt leader, coding region and stop codon. These reactions produced a smear of labelled polypeptides up to the position of full-length GlgC, but no distinct GlgC signal was observed (data not shown). This result suggested that the transcript was being degraded from the 3′ end. Inclusion of a 3′-terminal hairpin (E. coli trpT) in the glgC transcript resulted in the translation of a distinct full-length GlgC polypeptide (Fig. 4A), although a smear of truncated polypeptides was still observed. The addition of CsrA protein to the reaction inhibited both full-length and truncated GlgC synthesis (Fig. 4A). Quantification of full-length GlgC indicated that 0.4 μM CsrA resulted in 57% inhibition and 3.2 μM caused complete inhibition (Fig. 4B). In conjunction with our toeprint results demonstrating that CsrA competes with ribosomes for binding to the glgC transcript (Fig. 2), these results establish that CsrA regulates glgC expression at the level of translation initiation.

Figure 4.

Effects of CsrA on RNA-directed cell-free translation of glgC mRNA. An S-30 extract was prepared from TR1-5BW3414 (csrA::kanR). Reactions were carried out with various concentrations of purified CsrA in the absence (–) or presence (+) of 2 pmol of glgC–trpT transcript.

A. GlgC translation products analysed by SDS–PAGE and fluorography. The position of full-length GlgC is indicated.

B. The level of incorporation of [35S]-methionine into the full-length GlgC polypeptide was determined by liquid scintillation counting (Romeo and Preiss, 1989). Approximately 3.8 × 103 c.p.m. is equivalent to 1 pmol of GlgC polypeptide per reaction h−1.

CsrA binds to an RNA hairpin in the glg leader transcript and to the glgC SD sequence

Eighteen repeated sequences in csrB RNA are thought to be targets for CsrA binding (Liu et al., 1997). The predicted RNA secondary structure of the majority of these elements consists of short hairpins containing GGA residues in the loop. The RNA sequence just upstream of the prominent CsrA toeprint bands (CACGGAUU) (Figs 2 and 3) is similar to the repeated CsrB sequences. Computer predictions using MFOLD (Mathews et al., 1999) indicated that the CsrA binding site identified in the toeprint analysis was also capable of forming a short hairpin with GGA in the loop (Fig. 3). We carried out RNA structure mapping experiments to determine whether this sequ-ence was contained within the predicted secondary structure (Figs 3 and 5). We used RNase T1 (cleaves after unpaired G residues) and Pb2+-mediated cleavage (cleaves single-stranded RNA) in this analysis. RNase T1 cleavage of G78, G94, G95, G99, as well as all of the G residues between G111 and the start of the glgC coding sequence, indicated that these residues were single stranded (Fig. 5). The absence of cleavage of the G residues at positions 81, 82, 83, 88, 101, 103 and 105 suggested that these nucleotides were basepaired. Strong Pb2+-mediated cleavage at positions 84–86, 97, 98, 106–115, 118–124, 130 and 133–135 indicated that these residues were unpaired (Fig. 5). Conversely, weak cleavage or the absence of cleavage of the remaining nucleotides between position 84 and the glgC coding sequence suggested that these residues were basepaired at least part of the time. Our structure mapping results are consistent with the structure shown in Fig. 3. This structure is identical to the computer prediction with the exception that U87 is paired with G105, whereas MFOLD predicted that they would be unpaired. Interestingly, the GGA sequence predicted to be in the loops of the CsrB RNA hairpins is also in the loop of the CsrA binding target in the glg leader hairpin (Fig. 5). Note that the G residues at positions 81, 82 and 83 are not part of the glg leader hairpin and appear to basepair with residues further upstream.

Figure 5.

glgCAP leader RNA structure mapping and CsrA–glgCAP leader RNA footprint. WT glgCAP leader RNA was treated with RNase T1 or lead acetate (Pb2+) in the absence or presence of the concentration of CsrA (μM) indicated at the top of the lane. Lanes corresponding to mock-treated (M) RNA as well as partial alkaline hydrolysis (OH) and RNase T1 digestion (T1) ladders are shown. The RNase T1 ladder was generated under denaturing conditions so that every G residue in the transcript could be visualized (Bevilacqua and Bevilacqua, 1998). Positions of the CsrA footprints (FP), the RNA hairpin (HP) and the glgC SD sequence and start codon (Met) are shown. Numbering is from the start of glgCAP operon transcription.

We performed a CsrA–glgC RNA footprint analysis to identify more precisely the nucleotides bound by CsrA (Figs 3 and 5 ). As the CsrA concentration was increased from 0.25 to 2 μM, we observed the protection of two RNA segments from RNase T1 and Pb2+-mediated cleavage. The protected nucleotides extend from positions A84 to U100 and from A120 to G128. Conversely, we observed enhanced cleavage between G103 and G105. The upstream protected sequence extended from just 5′ of the stem to just past the loop of the hairpin (Figs 3 and 5). The residues that were enhanced for cleavage were on the 3′ side of the base of the stem. This indicates that CsrA binds to a sequence within the hairpin structure and that the RNA structure is altered when CsrA is bound. Of particular interest was the finding that bound CsrA also prevented cleavage of the glgC SD sequence (positions 123–128). We do not understand why the CsrA binding site overlapping the glgC SD sequence was not detected in our toeprint experiments. Perhaps the affinity of CsrA to this target is not sufficiently high to prevent disruption of the complex by reverse transcriptase. As our structure mapping studies indicate that the glgC SD sequence is not sequestered in a secondary structure, our results demonstrate that CsrA binds to the single-stranded glgC SD sequence and to the sequence in the glgCAP leader hairpin further upstream. When taken together with the results of our toeprint and cell-free translation studies, our footprinting experiments establish that bound CsrA prevents ribosome binding and that this results in translation inhibition. It is also interesting to note that, although our footprinting studies identified two CsrA binding sites, our gel shift experiments suggest that more than two CsrA molecules (dimers or trimers) are present in some of the CsrA–glgCAP leader RNA complexes. Thus, it appears that both CsrA–RNA and CsrA–CsrA interactions are involved in the formation of these complexes.

Mutations in the glgCAP leader hairpin reduce CsrA binding and CsrA-dependent glgC regulation

As our footprint results indicated that CsrA binds to a glgCAP leader RNA sequence contained within a hairpin, we examined the effect of two mutations in this sequence/structure on CsrA binding. One of the mutant transcripts contained a deletion of the GGA in the loop of the hairpin (ΔGGA) (Fig. 3). This sequence is conserved in the CsrA binding site that overlaps the glgC SD sequence and in 14 of the 18 CsrB sequences that are presumed to be the sites of CsrA binding. The second mutation contained a deletion of this same GGA sequence as well as a deletion of G99 and U100 (ΔGGA–GU). We performed gel mobility shift assays to assess the effect of each mutation on the affinity of CsrA. In each case, we observed a gel shift pattern that was similar to the wild-type (WT) transcript. When the concentration of CsrA was increased, we observed a gradual transition from free to bound RNA (data not shown). Non-linear least-squares analysis of these data yielded estimated Kd values of 150 nM CsrA for ΔGGA and 330 nM for ΔGGA–GU. These results demonstrated that ΔGGA or the ΔGGA–GU mutations caused a fourfold or eightfold reduction in CsrA affinity respectively. As was the case for the WT transcript (Fig. 1), we observed additional shifted species as the concentration of CsrA was increased further. Thus, although the deletions reduced the affinity of CsrA for the glgCAP leader transcript, they did not eliminate the presumed CsrA–CsrA interactions that probably contribute to the formation of some of these complexes.

We also carried out RNase T1 footprint experiments to examine the effect of these mutations in more detail (Fig. 6). The footprint to the RNA hairpin was significantly reduced in the ΔGGA deletion and was virtually eliminated in the ΔGGA–GU transcript. In both cases, the footprint to the glgC SD sequence was retained, although the CsrA concentration required to obtain the SD footprint was twofold higher than that required for the WT transcript (Figs 5 and 6). Thus, although CsrA interaction with the upstream binding site increases the affinity for the target that overlaps the glgC SD sequence, it does not appear that this interaction is necessary.

Figure 6.

RNA structure mapping and CsrA footprints of mutant glgCAP leader transcripts. glgCAP leader RNA was treated with RNase T1 or lead acetate (Pb2+) in the absence or presence of the concentration of CsrA (μM) indicated at the top of the lane.

A. ΔGGA transcript.

B. ΔGGA–GU transcript.

Lanes corresponding to mock-treated (M) RNA as well as partial alkaline hydrolysis (OH) and RNase T1 digestion (T1) ladders are shown. The RNase T1 ladder was generated under denaturing conditions so that every G residue in the transcript could be visualized (Bevilacqua and Bevilacqua, 1998). Positions of the CsrA footprints (FP), the RNA hairpin (HP) and the glgC SD sequence and start codon (Met) are shown. Numbering is relative to the start of WT glgCAP operon transcription.

We next examined the effect of these two mutations on CsrA-dependent glgC regulation in vivo. Expression from a glgClacZ translational fusion containing the WT leader increased fivefold by the time the cells reached stationary phase (Fig. 7A). Stationary phase expression levels in the ΔGGA and the ΔGGA–GU fusion strains were twofold or sixfold higher than expression in the WT fusion strain respectively (Fig. 7A). These results indicate that the effect that each mutation had on in vitro binding correlated with the severity of the defect on regulation in vivo. Moreover, our findings indicate that CsrA binding to the upstream target is necessary for proper glgC regulation. We also compared expression of the WT and mutant fusions in CsrA-deficient strains. As expected, expression from all three fusions was similar in the absence of CsrA. In each case, expression increased about 150-fold by the time the cells entered stationary phase growth, reaching a plateau of approximately 200 units of β-galactosidase activity (Fig. 7B).

Figure 7.

Effect of glgCAP leader mutations on expression of glgClacZ translational fusions. Strains containing single-copy glgClacZ translational fusions were grown, and assays were conducted as described in Experimental procedures. Growth curves (open symbols) and β-galactosidase activity (filled symbols) are presented. Symbols corresponding to each fusion are: WT, circles; ΔGGA, squares; ΔGGA–GU, triangles. Experiments in (A) were carried out in a WT genetic background, whereas experiments in (B) were performed in a csrA null strain. Growth curves are not presented in (B) to improve the clarity of the β-galactosidase activity curves. In each case, the growth curves were essentially identical to those shown in (A).

Discussion

Our footprinting studies (Fig. 5) identified two CsrA binding sites in the glgCAP leader transcript that share some sequence conservation. One of these binding sites overlaps the glgC SD sequence, whereas the other CsrA target is located further upstream. Our toeprint results firmly establish that CsrA inhibits ribosome binding (Fig. 2), whereas our cell-free translation experiments indicate that CsrA inhibits GlgC synthesis (Fig. 4). As CsrA is not a general inhibitor of translation (Yang et al., 1996; Liu and Romeo, 1997; Liu et al., 1997), our results indicate that CsrA specifically inhibits translation initiation of glgC.

The upstream CsrA binding site is contained within a short RNA hairpin, whereas the downstream site overlaps the glgC SD sequence and is single stranded (Fig. 3). The sequence and structure of the upstream RNA hairpin is similar to several repeated elements within CsrB RNA that are thought to be the sites of CsrA interaction (Liu et al., 1997). In addition, a genetic study identified presumed CsrA binding sites that overlap the SD sequ-ences of three genes from Pseudomonas fluorescens and Pseudomonas aeruginosa that are involved in the biosynthesis of hydrogen cyanide or exoprotease production (Blumer et al., 1999). An alignment of this conserved sequence is presented in Fig. 8. The GGA residues in the loop of the E. coli glg leader hairpin are conserved in all of these sequences. Although deletion of these three residues reduced the affinity of CsrA in vitro, the deletion had only a modest effect on CsrA-dependent glgClacZ regulation in vivo (Fig. 7A). However, a more extensive deletion (ΔGGA–GU) that further decreased the affinity of CsrA in vitro, substantially reduced CsrA-mediated regulation (Fig. 7A). Although the GGA residues in the loop of the hairpin are not essential for CsrA-mediated regulation in vivo, they are required for optimal CsrA binding and regulation.

Figure 8.

Comparison of putative CsrA binding sites. The CsrA binding sites in the glgCAP leader are compared with the consensus sequence of 18 repeated sequences in E. coli CsrB RNA (Liu et al., 1997), as well as with three sequences that were identified genetically in P. fluorescens and P. aeruginosa (Blumer et al., 1999).

The upstream glg leader CsrA binding site is contained within a hairpin that consists of 7 bp and a GGAUU pentaloop (Fig. 3). However, appreciable cleavage of the CG closing basepair indicates that this basepair has a tendency to breathe (Fig. 5). We also observed weak Pb2+-mediated cleavage of the majority of the nucleotides present in the stem of this hairpin, suggesting that the entire structure is unstable. As stable RNA structures inhibit primer extension by reverse transcriptase (Du and Babitzke, 1998), the finding that this structure did not give rise to a toeprint (Fig. 2) is consistent with this interpretation. The putative CsrA binding sites in Pseudomonas are not predicted to form stable secondary structures (not shown). In addition, the presumed CsrA binding sites in E. coli CsrB RNA are only predicted to form hairpins containing from 4 to 7 bp (Liu et al., 1997). Our footprinting and RNA structure mapping results demonstrated that the base of the glgCAP leader RNA hairpin is disrupted when CsrA is bound to the transcript and that the binding site that overlaps the glgC SD sequence is single stranded in the absence of bound CsrA (Figs 3 and 5). When taken together, these observations suggest that CsrA may bind preferentially to single-stranded RNA. In the case of the upstream glgCAP leader binding site, CsrA might bind when the structure breathes. If CsrA binds to the hairpin structure, then this protein must possess two distinct RNA-binding mechanisms: one to bind to an RNA structure and another to bind to single-stranded RNA. Both of the mutant glgCAP leader transcripts (ΔGGA and ΔGGA– GU) that we examined in this study reduced the affinity of CsrA; however, in each case, the mutation altered both the primary sequence of the CsrA binding site and the structure of the hairpin. Thus, a more thorough mutational analysis is required to determine whether CsrA binds to single-stranded RNA, structured RNA or both.

Previous studies established that CsrA promotes glgCAP operon mRNA decay (Liu et al., 1995). In this study, we found that CsrA binding to the glgCAP leader blocks ribosome access to the glgC ribosome binding site (Figs 2, 3 and 5). As CsrA inhibits GlgC synthesis (Fig. 4), our results establish that CsrA regulates the translation of glgC. As inhibition of translation has been shown to increase the rate of mRNA decay in some cases (Yarchuk et al., 1992; Vytvytska et al., 2000), it is likely that CsrA-mediated translational control of glgC contributes to the accelerated decay rate of glg mRNA.

A model of CsrA-mediated glgC regulation is presented in Fig. 9. CsrB RNA functions as an antagonist of CsrA by sequestering the protein in a CsrA–CsrB com-plex (Liu et al., 1997). However, the concentration of CsrA in growing cells is sufficiently high that a substantial amount of CsrA should be available for binding to target mRNAs (Gudapaty et al., 2001). Once CsrA is bound to both of the RNA sequences in the glgCAP leader, binding of additional CsrA molecules via protein–protein inter-actions might contribute to the stability of the CsrA–glg mRNA complex. This mechanistic feature takes into consideration the multiple shifted species observed in our gel shift analysis (Fig. 1). When CsrA is bound to both RNA sequences, translation initiation is prevented. Although the terminator hairpin at the 3′ end of the transcript may serve as a protective barrier to exonucleolytic decay, the absence of translating ribosomes would allow a series of endonucleolytic cleavages, followed by 3′ to 5′ exonucleolytic degradation of the resulting fragments.

Figure 9.

Model of CsrA-mediated regulation of E. coli glgC. CsrB RNA functions as an antagonist of CsrA action. CsrA inhibits translation of glgC by binding to two positions within the glgCAP leader. The upstream site is contained within a short RNA hairpin, and CsrA probably binds to this site as it breathes (represented by dashed lines). The second site overlaps the glgC SD sequence. Once CsrA binds to both these sites, additional CsrA complexes may bind via protein–protein interactions. The absence of translation allows endonucleolytic cleavage, resulting in rapid degradation of the message. See text for details.

The Csr system is a global regulator of central carbon metabolism, cell motility and biofilm formation in E. coli (Romeo and Gong, 1993; Romeo et al., 1993; Sabnis et al., 1995; Yang et al., 1996; Wei et al., 2000; 2001; Jackson et al., 2002). In addition, CsrA homologues have increasingly been recognized for important roles in the regulation of stationary phase gene expression in other bacterial species. The CsrA homologue (RsmA) of Erwinia species regulates a variety of genes involved in soft-rot disease of higher plants (Cui et al., 1995; Cui et al., 1999; Koiv and Mae, 2001). It was reported recently that the csrA and csrB genes of Salmonella enterica serovar Typhimurium regulate genes involved in epithelial cell invasion by this species (Altier et al., 2000; Allen et al., 2001). As mentioned above, stationary phase-expressed structural genes of Pseudomonads appear to be regulated by Csr homologues (Blumer et al., 1999). Furthermore, recent evidence indicates that both of the quorum-sensing systems of P. aeruginosa, Las and Rhl, which themselves regulate several virulence factors of this species, are controlled by the Csr (Rsm) system (Pessi et al., 2001). Thus, Csr systems regulate a variety of important genes in several organisms. As csrA homologues are apparent in over 50 phylogenetically diverse species (White et al., 1996; Romeo, 1998; unpublished results), we predict that the diverse assortment of genes and important physiological functions that are regulated by Csr systems will continue to grow.

Experimental procedures

Bacterial strains and plasmids

Plasmid pCSB27 contains csrA under the control of an IPTG-inducible promoter. pAltC4 contains portions of the WT glgCAP operon leader and glgC coding regions (+46 to +179 relative to the start of transcription) (Liu and Romeo, 1997). pIM1 contains a deletion of G94, G95 and A96 (ΔGGA) from the glgCAP leader, whereas pIM2 also contains a deletion of G99 and T100 (ΔGGA–GU). pIM4 contains the flhDC leader region (+1 to +276), pIM5 contains csrB (+1 to +338), and pPB77 contains the B. subtilis trp leader (Babitzke et al., 1994). pCZ3-3 contains the WT glgCAP operon promoter, leader region and a glgClacZ translational fusion (Romeo et al., 1990). Plasmids pCSB25 and pIM3 are identical to pCZ3-3 except that they contain the ΔGGA or the ΔGGA–GU glg leaders respectively. E. coli strains used for β-galactosidase assays were constructed to create single-copy gene insertions of glgClacZ translational fusions as described previously (Boyd et al., 2000). Strains KSC365, PLB660 and KSC121 contain fusions in strain CF7789 (ΔlacI-Y) derived from plasmids pCZ3-3 (WT), pCSB25 (ΔGGA) and pIM3 (ΔGGA–GU) respectively.

CsrA purification

CsrA was overproduced by inducing the expression of csrA carried on pCSB27 by the addition of IPTG. Cell pellets were suspended in 10 mM sodium phosphate, pH 6.0, 50 mM NaCl and 10% glycerol (w/v). S100 extracts were prepared from cells that were lysed by sonication. CsrA was then precipitated with 21% ammonium sulphate. The protein pellet was dissolved in 10 mM sodium phosphate, pH 6.0, until the conductivity of the solution was equivalent to 50 mM NaCl. This protein solution was mixed with pre-equilibrated phosphocellulose (10 mM sodium phosphate, pH 5.0, 50 mM NaCl, 10% glycerol) and then packed into a column. CsrA eluted from the column between 250 and 750 mM NaCl. These fractions were combined and brought to pH 9.0 by the addition of 1 M NaOH before dialysis against 10 mM Tris-HCl, pH 9.0, 50 mM KCl and 4% glycerol. The dialysate was loaded onto a MonoQ column (Pharmacia HR 5/5) that was pre-equilibrated with 10 mM Tris-HCl, pH 9.0, and 50 mM KCl. CsrA was eluted with a linear KCl gradient. The peak fraction (200 mM KCl) was brought to pH 7.5 by the addition of 100 mM MOPS and subsequently dialysed against 10 mM Tris-HCl, pH 8.0, 50 mM KCl and 25% glycerol. CsrA purity was assessed by electrophoresis through 15% polyacrylamide–SDS gels. Protein concentrations were determined using the Bio-Rad protein assay.

Gel shift assay

Quantitative gel shift assays followed a previously published procedure (Yakhnin et al., 2000). RNA was synthesized in vitro using the Ambion MEGAscript kit and various linearized plasmid templates. Gel-purified RNA was 5′ end labelled with [γ-32P]-ATP as described previously (Yakhnin et al., 2000). RNA suspended in TE was renatured by heating to 80°C followed by slow cooling. Binding reactions (10 μl) contained 10 mM Tris-HCl, pH 7.5, 10 mM MgCl2, 100 mM KCl, 32.5 ng of yeast RNA, 7.5% glycerol, 20 mM dithiothreitol (DTT), 4 U of RNase inhibitor (Ambion), 0.5 nM WT or mutant glgCAP leader RNA, purified CsrA (various concentrations) and 0.1 μg ml−1 xylene cyanol. Competition assays also contained unlabelled RNA competitor (see text for details). Reaction mixtures were incubated for 30 min at 37°C to allow CsrA– RNA complex formation. Samples were then fractionated on native 10% polyacrylamide gels. Radioactive bands were visualized using a phosphorimager. Free and bound RNA species were quantified using IMAGEQUANT, and the apparent equilibrium binding constants (Kd) of CsrA–RNA complexes were calculated (Yakhnin et al., 2000).

Toeprint assay

Toeprint assays were performed by modifying published procedures (Hartz et al., 1988; Du et al., 1997; Du and Babitzke, 1998). glgCAP leader transcripts used in this analysis were synthesized using pAltC4 as template. Gel-purified glgCAP leader RNA (250 nM) in TE was renatured and hybridized to a 32P end-labelled DNA oligonucleotide (500 nM) complementary to the 3′ end of the transcript by heating to 80°C followed by slow cooling. Toeprint assays were carried out with 200 nM CsrA and/or 100 nM 30S ribosomal subunits and 500 nM tRNAfMet. Toeprint reactions (20 μl) contained 2 μl of the hybridization mixture, 375 μM each dNTP and 10 mM DTT in toeprint buffer (10 mM Tris-HCl, pH 7.4, 10 mM MgCl2, 60 mM NH4OAc, 6 mM 2-mercaptoethanol) (Hartz et al., 1988). Mixtures containing CsrA were incubated for 30 min at 37°C to allow CsrA–mRNA complex formation. 30S ribosomal subunit toeprint reactions were performed by incubating RNA, 30S ribosomal subunits and tRNAfMet in toeprint buffer as described previously (Hartz et al., 1988). After the addition of 0.6 U of avian myeloblastosis virus reverse transcriptase (Roche), incubation was continued at 37°C for 15 min. Reactions were terminated by the addition of 12 μl of stop solution (70 mM EDTA, 95% formamide, 0.1× TBE, 0.025% xylene cyanol, 0.025% bromophenol blue). Samples were fractionated through a 6% sequencing gel. Sequencing reactions were performed using pAltC4 as the template and the same end-labelled DNA oligonucleotide as a primer.

RNA-directed cell-free translation

A polymerase chain reaction (PCR) product containing a T7 RNA polymerase promoter, the glgC leader and coding sequence and the trpT terminator sequence was used as a template for transcription. The 28 base E. coli trpT terminator sequence was included after the glgC stop codon to inhibit exonucleolytic degradation of the mRNA in S30 extracts (Wu and Platt, 1978). The glgC–trpT transcript was generated in vitro using 1 μg of the PCR product and the Ambion MEGAscript kit. The effects of CsrA protein on cell-free translation of glgC mRNA were assessed in S-30 extracts prepared from TR1-5BW3414 (csrA::kanR) as described previously (Romeo and Preiss, 1989; Liu and Romeo, 1997), except that reaction volumes were scaled down to 17.5 μl and in vitro-generated mRNA replaced DNA in the reaction. Protein was 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–PAGE gels. Radiolabelled proteins were detected by fluorography using sodium salicylate (Chamberlain, 1979), and methionine incorporated into full-length GlgC polypeptide was quantified by liquid scintillation counting of H2O2-solubilized gel sections (Romeo and Preiss, 1989).

RNA structure mapping and footprint assays

WT and mutant 5′ end-labelled glgCAP leader transcripts used in this analysis were generated as described for the gel shift analysis. Titrations of RNase T1 (Roche) and lead acetate were performed to optimize the amount of each reagent to prevent multiple cleavages in any one transcript. Binding reactions (10 μl) containing various concentrations of CsrA and 2 nM WT or mutant glgCAP leader RNA were identical to those described for the gel shift assay. RNase T1 (0.02 U) was added to the binding reaction, and incubation was continued for 15 min at 37°C. Reactions were terminated by the addition of 5 μl of gel loading buffer II (Ambion). Pb2+-mediated cleavage of glgCAP leader RNA was achieved by the addition of 2 μl of 5 mM lead acetate to the binding reaction, and incubation was continued for 10 min at 37°C (Ciesiolka et al., 1998). Partial alkaline hydrolysis and RNase T1 digestion ladders of each transcript were pre-pared as described previously (Bevilacqua and Bevilacqua, 1998). Samples were fractionated through 6% sequencing gels.

Growth studies and β-galactosidase assays

Bacterial growth in LB at 37°C was monitored using a Klett–Summerson colorimeter (No. 52 green filter). Culture samples (4 ml) were harvested at various times, suspended in 0.5 ml of Z buffer (Miller, 1972) containing 0.2 mg ml−1 lysozyme and lysed by the freeze–thaw/deoxycholate method (Ron et al., 1966). β-Galactosidase assays were performed as described previously (Miller, 1972). The specific activity of the enzyme was determined by normalizing the activity to the protein concentration of the supernatant. Protein concentrations were determined using the Bio-Rad protein assay.

Acknowledgements

We thank Paul Lovett for providing 30S ribosomal subunits, Mike Cashel for bacterial strain CF7789, Dana Boyd for λInCh materials and advice, and Dan Jones for performing mass spectroscopy on CsrA. We also thank Alexander Yakhnin, Helen Yakhnin and Janell Schaak for technical assistance, and Charles Yanofsky for critical reading of the manuscript. This work was supported by grant GM59969 from the National Institutes of Health.

Ancillary