RNA sequence and secondary structure participate in high-affinity CsrA–RNA interaction

INTRODUCTION

The Csr (carbon storage regulation) and homologous Rsm (repressor of stationary phase metabolites) global regulatory systems of several bacterial species control numerous genes and processes post-transcriptionally. Csr (Rsm) systems are characterized by an RNA binding protein that either activates or represses expression of target mRNAs and noncoding RNAs (ncRNAs) that function as antagonists of the RNA binding protein (Romeo 1998). For example, in Pseudomonas fluorescens the Rsm system participates in controlling the production of anti-fungal secondary metabolites and exoenzymes (Blumer et al. 1999; Reimmann et al. 2005), whereas Rsm of Erwinia species regulates a variety of genes involved in soft-rot disease of higher plants (Cui et al. 1999). Csr is also known to regulate epithelial cell invasion by Salmonella enterica (Altier et al. 2000) and affect swarming and other activities of Proteus mirabilis (Liaw et al. 2003) and Legionella pneumophila (Fettes et al. 2001).

The Csr system of Escherichia coli is involved in the repression of several stationary phase processes and in the activation of some exponential phase functions (Romeo 1998). Three major components of Csr in this organism include the RNA binding protein CsrA and two noncoding RNA (ncRNA) molecules, CsrB and CsrC. CsrA represses various processes such as gluconeogenesis, glycogen metabolism, and biofilm formation (Romeo et al. 1993; Sabnis et al. 1995; Yang et al. 1996; Jackson et al. 2002; Wang et al. 2005). CsrA also activates glycolysis, acetate metabolism, and flagellum biosynthesis (Sabnis et al. 1995; Wei et al. 2000, 2001). CsrB and CsrC function as antagonists of CsrA by sequestering this protein and preventing its ability to interact with mRNA targets. The multiple imperfect repeat sequences (18 in CsrB and nine in CsrC) in these regulatory RNAs function as CsrA binding sites (Liu et al. 1997; Gudapaty et al. 2001; Weilbacher et al. 2003).

CsrA negatively regulates expression of glgC, a gene involved in glycogen biosynthesis, by binding to four sites in the untranslated leader of the glgCAP operon transcript, one of which overlaps the glgC Shine-Dalgarno (S-D) sequence (Baker et al. 2002; A.K. Dubey, T. Romeo, and P. Babitzke, unpubl.). CsrA binding to the glgCAP leader transcript inhibits GlgC synthesis by blocking ribosome binding. Presumably, CsrA-mediated inhibition of glgC translation is responsible for the accelerated rate of glgCAP mRNA decay (Liu et al. 1995). CsrA also represses translation of cstA, a carbon starvation-induced gene thought to be involved in peptide transport (Schultz and Matin 1991; Dubey et al. 2003), as well as the pgaABCD operon, a cluster of genes that are required for the synthesis of the polysaccharide adhesin poly-β-1,6-N-acetyl-D-glucosamine (PGA) that participates in biofilm formation (Wang et al. 2005). CsrA binds to four sites in the cstA transcript and to at least six sites in the pgaA leader transcript. In each case one of the CsrA binding sites overlaps the cognate S-D sequence. Translational repression of these genes proceeds by a mechanism that is similar to that of glgC (Dubey et al. 2003; Wang et al. 2005).

Considerable sequence variation exists in the known E. coli CsrA binding sites, with GGA being the most conserved element. The GGA motif is predicted to be present in the loop of short RNA hairpins in several of these binding sites, although the importance of this structural arrangement is not known (Liu et al. 1997; Baker et al. 2002; Weilbacher et al. 2003; Wang et al. 2005). Systematic evolution of ligands by exponential enrichment (SELEX) is routinely used to isolate high-affinity ligands from a pool of randomized nucleic acid sequences (Ellington and Szostak 1990; Tuerk and Gold 1990; Ulrich et al. 2002). SELEX has been used to identify RNA ligands that either bind to protein (e.g., Schneider et al. 1992; Baumann et al. 1997) or small metabolites (e.g., Lozupone et al. 2003). In this study, SELEX was used to isolate high-affinity CsrA ligands. The results presented herein establish that both the primary sequence and secondary structure of selected RNA ligands are important for high-affinity CsrA interaction.

RESULTS AND DISCUSSION

In vitro selection of high-affinity CsrA ligands

The ncRNAs CsrB and CsrC contain 18 and nine apparent CsrA binding sites, respectively. In addition, a total of 14 binding sites have been identified in the glgC, cstA, and pgaA transcripts. Despite the large number of known CsrA binding targets, identification of a reasonable binding site consensus has remained elusive. Since considerable sequence variation exists among the known CsrA binding sites, and because cooperative binding to natural RNAs is common, SELEX was carried out to identify RNA ligands containing single high-affinity CsrA binding sites. The template-primer system used by Ulrich et al. (2002) was adapted to produce RNA molecules having a stretch of 15 randomized nucleotides that were flanked by 5′ and 3′ constant regions. The use of only 15 randomized nucleotides was chosen to minimize the possibility of selecting RNAs containing more than one CsrA binding site. Because the theoretical complexity of 15 randomized positions contains just over 1 × 10⁹ different sequences, our initial randomized RNA pool contained ~1 × 10⁵ molecules of every possible sequence. Since quantitative gel mobility shift assays using native CsrA or C-terminal His-tagged CsrA (CsrA-H6) did not show any significant difference in binding affinities for target transcripts (data not shown), CsrA-H6 was used in SELEX. To remove nonspecific ligands, the RNA pool was subjected to preselection with Ni-NTA beads in the absence of CsrA-H6 protein. RNA molecules that did not bind to the Ni-NTA beads were used for the selection process (see Materials and Methods).

The progress of selection was monitored by gel mobility shift assays using native CsrA and 5′ end-labeled RNA pools from rounds 0, 2, 4, 7, and 9 (Fig. 1). The binding constants of CsrA–RNA interaction gradually decreased as the number of rounds of selection increased, indicating that subjecting the RNAs to additional rounds of selection enriched for high-affinity ligands. Initially nine rounds of selection were carried out and RT-PCR fragments from rounds 4, 7, and 9 were cloned and sequenced. Three additional rounds of selection were subsequently carried out; however, we did not identify any sequences from round 12 that differed appreciably from those selected in rounds 7 and 9 (see below).

Sequence and structural analysis of selected RNA ligands

A total of 55 selected RNA ligands were cloned and sequenced from rounds 4, 7, 9, and 12. Sequence analysis of these RNA molecules identified two ligand classes. Fifty class I RNA molecules contained a single GGA motif, whereas five class II RNAs contained two GGA sequences (Fig. 2 and data not shown). The RNA ligands from class I were subdivided into two groups. Class I-ARNAs (38 sequences) contained the GGA motif at the extreme 3′ end of the randomized sequence, while class I-B RNAs (12 sequences) contained the GGA somewhere in the middle of the randomized sequence. A sequence alignment of all 55 RNA molecules identified a SELEX-derived consensus of RUACARGGAUGU (Fig. 2). Strikingly, the ACA and GGA motifs were 100% conserved, while the GU residues were present in all but one RNA ligand. All of the class I-A CsrA binding sites utilized the AGU sequence (following GGA) from the 3′ constant region. In all of the class I-B examples, the 3′ end of the CsrA binding site was derived from randomized sequence. The presence of UGU (following GGA) in seven out of 12 class I-B RNAs led us to include UGU rather than AGU in the consensus. It is also interesting that four of the class I-B sequences (R9–8, R9–24, R9–54, and R9–56) and one class II sequence (R9–46) derived their AC residues of the conserved ACA element from the 5′ constant region (Fig. 2). In each of these examples the first A residue originated from a G-to-A substitution that presumably arose in the 5′ constant region during PCR. In the case of the class II RNAs, the sequence surrounding only one of the two GGA motifs closely resembles the consensus. When only this GGA motif is considered, four of the class II RNAs are similar to class I-A ligands, with the other being similar to class I-B.

The GGA motif of known CsrA binding sites is often found in the loop of short hairpins. Computer modeling of the 55 selected sequences using MFOLD v.3.1 (Mathews et al. 1999; Zuker 2003) revealed that 51 of the RNAs contained the GGA motif in the loop of a hairpin within the most stable predicted structure (ΔG = −17.6 to −6.3 kcal/mol) (Fig. 3 and not shown). In addition, R9–54 (class I-B, Fig. 2) contained its GGA in the loop of a hairpin within the second most stable structure (ΔG = −6.6 kcal/mol vs. ΔG = −6.2 kcal/mol, not shown). The predicted stabilities of the GGA motif-containing hairpins ranged from −14.2 kcal/mol to −0.7 kcal/mol. The GGA motifs of the remaining three ligands, R9–31 (class II), R9–42 (class II), and R9–49 (class I-B) were also predicted to be present in the loops of suboptimal structures, although in each case their predicted stabilities were substantially lower than the optimal structures (not shown). It is also interesting to note that only three of the hairpins contained more than four contiguous base pairs below the loop, suggesting that particularly stable hairpins might not be favorable for CsrA interaction. Thus, it is possible that CsrA selected all 55 ligands when the GGA motif was present in the loop of a short hairpin.

The loops of the hairpins either consisted of four (eight sequences), six (44 sequences), or eight (three sequences) nucleotides. Strikingly, with one exception, the AC residues in the 100% conserved ACA motif were predicted to form base pairs with the conserved GU residues. Moreover, in the cases in which the binding target ended in UGU, the entire ACA motif was predicted to form base pairs with these residues (Fig. 3). Finally, it is interesting to note that one class I-B ligand (R9–44) contains a GNRA (GGGA) tetraloop (Fig. 3). In this case, the first G in the tetraloop is derived from the second R in the consensus sequence. The observation that the sequence of the seven remaining four-base loops was AGGA, a sequence not known to form tetraloops, suggested that tetraloops are generally unfavorable for CsrA binding. In summary, it is apparent that both primary sequence and RNA secondary structure participate in the formation of high-affinity CsrA binding sites. More striking was the finding that primary sequence information was embedded within the stem.

A selected ligand (R9–43) competes with CsrA–pgaA mRNA interaction

Previously published results demonstrated that CsrA binds specifically to the pgaA mRNA and that this transcript contains six likely CsrA binding sites (Wang et al. 2005). Gel mobility shift assays were carried out to determine whether one of the SELEX-derived ligands (R9–43) was capable of competing for CsrA–pgaA mRNA interaction. CsrA binding to the pgaA transcript (+1 to +260 relative to the start of pgaA transcription) was detected as a distinct band in native gels between 2.5 and 40 nM CsrA and an apparent K_d of 25 nM and a cooperativity coefficient of 1.5 (Fig. 4, top panel). As the concentration of CsrA was increased further, a shifted complex of slower gel mobility was observed. Competition experiments were carried out with specific (pgaA, +1 to +260 relative to the start of transcription; R9–43) and nonspecific (Bacillus subtilis trp leader, +1 to +49 relative to the start of transcription) RNA competitors. Since the trp leader transcript contained an RNA hairpin (Sudershana et al. 1999; Du et al. 2000), use of this competitor would allow us to determine whether CsrA was capable of binding to RNA hairpins in general. Both pgaA and R9–43 RNAs were effective competitors, whereas the B. subtilis trp leader RNA did not compete for CsrA–pgaA RNA interaction (Fig. 4, bottom panel). These results indicate that CsrA interaction with the R9–43 aptamer and pgaA RNA occurs by a similar mechanism and that CsrA is not capable of interacting with RNA hairpins in general.

Binding affinity of CsrA for several selected RNA ligands

Gel mobility shift assays were performed to measure the affinity of CsrA for representative examples of selected RNA ligands (Table 1). To determine whether the identity of the purines designated by R in the RUACARGGAUGU consensus sequence influenced CsrA binding, gel shift experiments were carried out with four RNAs from class I-B that contained all possible combinations of A and G at these two positions (Table 1; R7–29, R9–9, R9–15, and R9–16). Structural predictions indicated that the GGA motif for each of these RNAs was present in a six-base loop that differed only in the identity of the second R in the consensus sequence (Fig. 3). The binding constants for these four RNAs ranged from 8 to 11 nM, indicating that CsrA does not have a significant preference for either purine at these two positions in the context of a six-nucleotide loop. The binding preference of CsrA for A or U following the GGA was also examined. Recall that the A at this position of the class I-A ligands was derived from the 3′ constant region, while seven of 12 class B ligands contain a U at this position. In this case, CsrA exhibited a slight preference for U (Table 1; R9–24 and R9–54).

The binding affinity of CsrA was also determined for each of the class II RNAs. Despite having two GGA sequences, sequence alignments suggested that only one authentic CsrA binding site was present in each transcript (Fig. 2). The binding affinity of CsrA for transcripts corresponding to clones R9–22, R9–28, R9–42, and R9–46 were between 7 and 16 nM, values that were similar to the other RNA ligands that were tested (Table 1). The finding that only one shifted complex was observed in gel mobility shift assays is consistent with each of these transcripts containing a single CsrA binding site (data not shown). The affinity of CsrA for the R9–31 transcript was considerably lower than for the other RNA ligands that were examined (Table 1). One reasonable explanation for the reduced affinity for this RNA molecule is that the GGA motif is sequestered in the stem of a predicted hairpin (Fig. 3). The low affinity of CsrA for R9–31 raises the question as to how this RNA was selected in the first place. Perhaps CsrA-H6 bound to this transcript when it was folded into a suboptimal structure in which the GGA motif is present in the loop of a hairpin (ΔG = −10.1 kcal/mol vs. ΔG = −8.1 kcal/mol) (Fig. 3). A possible explanation for the exceptionally high standard error for the binding curve observed with this transcript (Table 1) is that CsrA is only capable of binding to this transcript when the suboptimal structure is formed.

Mutational analysis of a selected RNA ligand

The relative contribution of the primary sequence and RNA secondary structure on CsrA binding was examined by altering conserved residues of a high-affinity RNA ligand. R9–43 was chosen as the progenitor wild-type binding site for this analysis because structural predictions using MFOLD indicated that none of the mutations that we intended to introduce would lead to unintended structural rearrangements (Fig. 5). The mutations that were introduced altered the primary sequence and/or the predicted secondary structure of R9–43. Gel mobility shift assays were carried out to investigate the effect of each mutation on CsrA binding (Fig. 5). The affinity of CsrA for R9–43(WT) was 7 nM (Figs. 5, 6). The C22U substitution replaced the C–G base pair with a U–G base pair (Fig. 5), resulting in a binding site that was identical to the SELEX-derived consensus. As expected, this mutation did not significantly alter the binding affinity. The A23G mutation, which altered a conserved residue and replaced the A–U base pair for a G–U base pair, led to a modest threefold reduction in binding affinity, whereas the A23U mutation, which replaced a conserved residue and disrupted the A–U base pair, led to a four- to fivefold reduction in affinity. The U32A mutation, which altered a conserved residue and disrupted the A–U base pair, resulted in a 10-fold reduction in binding affinity (Figs. 5, 6). Interestingly, the compensatory A23U:U32A double mutation did not restore binding; the affinity of CsrA for the compensatory mutant was similar to the U32A single mutant. These results indicated that the primary sequence itself, and not just its ability to form a base pair, is critical for high-affinity interaction. A similar pattern was observed for the C24U, C24G, G31C, and C24G:G31C mutations except that the double compensatory change exhibited an additive binding defect. Note that in this case, disruption of the C24–G31 base pair would likely lead to disruption of the A25–U30 closing base pair as well (Fig. 5). The A25U substitution, which replaced a conserved residue and disrupted the closing A–U base pair, led to a 15-fold reduction in binding affinity, whereas the U30A substitution did not affect binding (Figs. 5, 6). Interestingly, all of the SELEX-derived sequences in which the GGA motif was located at the extreme 3′ end of the randomized sequence (class I-A) contained an A at this position (see above). The compensatory A25U:U30A change did not restore binding and was similar to the A25U single mutant. Taken together, these data indicate that the primary sequence makes a more substantial contribution to CsrA binding than does the ability to form a hairpin with the GGA motif in the loop. The finding that CsrA can bind to natural RNA targets that are not contained with in RNA hairpins is consistent with this conclusion (Dubey et al. 2003).

In addition to changes within the stem, the affect of nucleotide substitutions within the loop was examined. Alteration of either G in the conserved GGA motif (G27C and G28C) led to a 20-fold reduction in binding affinity, whereas substitution of the A (A29C) resulted in a fivefold reduction in K_d (Fig. 5). Not surprisingly, deletion of the GGA motif resulted in a severe binding defect (>140-fold) (Figs. 5, 6). These results confirm the importance of the GGA motif in CsrA binding. Interestingly, the A26G substitution, which generated a GNRA tetraloop (GGGA), resulted in a ninefold reduction in binding affinity (Fig. 5). Recall that the SELEX-derived consensus (Fig. 2), as well as previous binding data (Table 1), suggested that either purine would be equally tolerated at this position. However, as noted above, only one of the selected RNA ligands with a four-base loop contained a G at this position, thereby generating a GNRA tetraloop. The finding that a GNRA tetraloop reduces CsrA binding provides an explanation for the scarcity of selected ligands in which the conserved GGA motif is part of a tetraloop.

Model of CsrA–RNA interaction

CsrA binding sites have been identified in the ncRNAs CsrB (18 apparent sites) and CsrC (nine apparent sites), as well as in glgC (four sites), cstA (four sites), and pgaA (six sites). The majority of the binding sites in CsrB and CsrC are predicted to contain the GGA motif in loops of short hairpins, although the CsrA binding site consensus sequence for these ncRNAs is CAGGAUG, which has a 1-nucleotide (nt) deletion relative to the SELEX-derived consensus (Liu et al. 1997; Weilbacher et al. 2003). Of the 10 binding sites in the glgC and pgaA transcripts, five are predicted to have a similar structural arrangement. However, none of the GGA motifs within the cstA transcript are predicted to be in the loops of hairpins (Baker et al. 2002; Dubey et al. 2003; Wang et al. 2005). Thus, while it is common for natural CsrA binding sites to contain both conserved sequence and structural arrangements, it is evident that the common structural arrangement is not essential for CsrA binding. It is also important to point out that not every natural CsrA binding site has a GGA motif. In four of the CsrB binding sites, the GGA is replaced with a GGG (Liu et al. 1997), while GGA is replaced with AGA in one of the pgaA binding sites (Wang et al. 2005).

Of the 55 selected CsrA ligands reported in this paper, 51 have their GGA motif in the loop of a hairpin within the most stable predicted structure. By design, all of these ligands contained a single CsrA binding site. A likely explanation for the high degree of structural conservation among the SELEX-derived CsrA targets is that having the GGA motif in the loop of a hairpin allows for high-affinity interaction. Furthermore, the finding that the AC residues of the 100% conserved ACA motif were predicted to pair with the conserved GU residues in all but one selected ligand strongly suggests that primary sequence information is embedded within the stem of each hairpin. This conclusion is supported by the mutational analysis of the R9–43 ligand, which demonstrated that the identity of the ACA and GU residues was more important for CsrA binding than the simple ability to form base pairs (Fig. 5). However, inspection of CsrB and CsrC stems does not reveal conserved primary sequence. Thus, the sites in these ncRNAs do not appear to be optimized for high-affinity binding. Perhaps the absence of the conserved stem sequences in CsrB and CsrC prevent them from competing too effectively for CsrA.

A reasonable model for how CsrA binds to RNA is as follows. CsrA initially interacts with the unpaired loop sequences when the GGA motif is present within the loop of a hairpin. Following this initial interaction, the hairpin is partially melted, leading to additional base-specific contacts. In situations where an RNA secondary structure is not present, CsrA is still able to bind but with lower affinity due to a reduction in the association rate. Consistent with this notion is the fact that of the five naturally occurring RNA targets that have been characterized, CsrA has the lowest affinity for cstA, the only known CsrA target in which none of the GGA motifs are predicted to be present in the loops of hairpins. Finally, one has to take into consideration that CsrA functions as a homodimer (Dubey et al. 2003). Thus, the cooperative interaction observed with CsrB, CsrC, cstA, and pgaA likely involves both protein– RNA and dimer–dimer interactions. However, it remains to be determined whether CsrA dimers interact with one or two binding sites. Engineering multiples of the SELEX-derived consensus CsrA binding site into a single transcript will facilitate future studies on cooperative interaction and CsrA–RNA stoichiometry.

MATERIALS AND METHODS

Purification of CsrA-H6 protein

Plasmid pCSB12 contains csrA cloned into the NdeI and BamHI sites of pET21a+ (Inovagen). This plasmid was designed to allow production of CsrA containing six additional His residues at the C terminus (CsrA-H6). pCSB12 was transformed into BL21 (DE3) pLysE (Invitrogen) to allow overproduction of CsrA-H6. The resulting strain (PLB328) was grown at 37° C in terrific broth to an OD₆₀₀ of 0.6, at which time 1 mM IPTG was added to the culture and growth was continued for 20 h. Cells were harvested by centrifugation at 4° C and the cell pellet was frozen at −20° C until further use. Cell pellets (9 g) were thawed and suspended in 45 mL lysis buffer (50 mM NaH₂PO₄, pH 8.0, 300 mM NaCl, 10 mM imidazole, 10% glycerol). Cell lysates were prepared by sonication, followed by centrifugation at 10,000g for 30 min at 4° C. The supernatant was mixed with 1 mL of prewashed Ni-NTA resin (Qiagen) for 1 h at 4° C and subsequently packed into a 1-mL column. The column was washed once with 20 mL of wash buffer (50 mM NaH₂PO₄, pH 8.0, 300 mM NaCl, 10% glycerol) containing 20 mM imidazole and twice with wash buffer containing 50 mM imidazole. CsrA-H6 was eluted with 15 mL wash buffer containing 250 mM imidazole. Column fractions were analyzed by 15% SDS-PAGE and Coomassie Blue staining. Fractions containing highly purified CsrA-H6 were pooled and dialyzed in phosphocellulose column buffer (100 mM Tris-HCl, pH 7.0, 100 mM NaCl, 10% glycerol). The dialysate was mixed with phosphocellulose pre-equilibrated with phosphocellulose column buffer and then packed into a 5-mL column. CsrA eluted from the column between 250 and 750 mM NaCl. Column fractions were analyzed by 15% SDS-PAGE and Coomassie Blue staining. Fractions containing pure CsrA-H6 were combined and dialyzed against 10 mM Tris-HCl, pH 7.0, 100 mM KCl, 10 mM MgCl₂ and 25% glycerol. The CsrA-H6 concentration was estimated using the Bio-Rad protein assay.

In vitro selection of RNA ligands

SELEX was performed by modifying published procedures (Tuerk and Gold 1990; Ulrich et al. 2002; Vo et al. 2003). The oligonucleotide used in this analysis (5′-ACCGAGTCCAGAAGCTTGTAG TAC(N15)GCCTAGATGGAGTTGAATTCTCCCTATAGTGAGT CGTATTAC-3′) contained a 15-nt randomized region (N15) flanked on both sides by constant sequences. To create the initial pool of randomized DNA templates, 10 pmol of the oligonucleotide containing the randomized sequence was used as a template for PCR amplification using 240 pmol each of primer 1 (5′- GTAATACGACTCACTATAGGGAGAATTCAACTCCATCTA- 3′) and primer 2 (5′-ACCGAGTCCAGAAGCTTGTAGT-3′). After gel purification of the PCR product, RNA was synthesized using the in vitro Megascript transcription kit (Ambion) in the presence of 0.66 μM [α³²P]ATP. The reaction mixture was then treated with five units of DNase I for 15 min to remove template DNA. Labeled transcripts were gel purified on 8% denaturing polyacrylamide gels and quantified.

RNA suspended in TE was renatured by heating to 85° C followed by slow cooling to room temperature. Binding reactions (500 μL) contained various concentrations of CsrA-H6 and RNA in binding buffer (10 mM Tris-HCl, pH 7.5, 10 mM MgCl₂, 100 mM KCl, 0.5 mg of yeast RNA, and 7.5% glycerol). After incubation for 30 min at 37°C, 40 μL of Ni-NTA slurry prewashed with binding buffer was added to the reaction mixture and further incubated for 10 min with repeated mixing. Bound RNAs were recovered by modifying a published procedure (Vo et al. 2003). The beads were washed three times with 100 μL of binding buffer and bound RNAs were eluted with 100 μL of binding buffer containing 2 M imidazole. The RNA was then extracted with phenol and phenol-chloroform and ethanol precipitated. RNAs were reverse transcribed using AMV reverse transciptase (Roche) and primer 2 and PCR amplified with primers 1 and 2. In vitro generated transcripts using the RT-PCR templates were subjected to the next round of selection. The CsrA-H6 concentration was 500 nM in the first two rounds of SELEX, 250 nM in rounds 3 and 4, 125 nM in rounds 5 through 7, 62.5 nM in rounds 8 and 9, 31 nM in round 10, 16 nM in round 11, and 8 nM in round 12. In each round the RNA concentration was five- to 10-fold higher than the concentration of CsrA-H6. RT-PCR products from rounds 4, 7, 9, and 12 were cloned and sequenced. A total of 55 clones were sequenced and this information was used to derive a consensus sequence (Fig. 2).

Gel mobility shift assays

Quantitative gel mobility shift assays were used to measure the affinity of CsrA-RNA binding by following published procedures (Baker et al. 2002; Dubey et al. 2003). For individually cloned DNA sequences, RNA was synthesized in vitro with the Ambion MEGAscript kit and linearized plasmid DNA as templates. For gel mobility shift assays with RNA pools, the transcripts from rounds 0, 2, 4, 7, and 9 were used. Gel-purified RNAs and RNA pools from various rounds were dephosphorylated with calf intestinal alkaline phosphatase and subsequently 5′ end-labeled using [γ-³²P]ATP and polynucleotide kinase. Labeled RNAs were gel purified, suspended in TE (10 mM Tris-HCl, pH 8.0, 1 mM EDTA) and renatured by heating to 85°C and slow cooling to room temperature. Binding reaction mixtures (10 μL) contained 10 mM Tris-HCl, pH 7.5, 100 mM KCl, 10 mM MgCl₂, 32.5 ng of yeast RNA, 10% glycerol, 20 mM dithiothreitol, 4 U of RNase inhibitor (Ambion), 5′ end-labeled RNAs (30 pM or 250 pM), various concentrations of purified CsrA (Baker et al. 2002), and 0.1 mg/mL xylene cyanol. Competition assays also contained unlabeled 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 through native polyacrylamide gels. Radioactive bands were visualized with a phosphorimager (Molecular Dynamics). Free and bound RNA species were quantified using ImageQuant software (Molecular Dynamics), and the apparent equilibrium binding constant (K_d) and cooperativity coefficient (n) for CsrA–RNA complex formation was calculated as described (Yakhnin et al. 2000).

Site-directed mutagenesis

Effects of mutations on CsrA binding were investigated by mutating various conserved residues of a selected RNA ligand (R9–43) using QuikChange mutagenesis (Stratagene). DNA corresponding to each mutation was cloned into plasmid pTZ18U (Stratagene). The in vitro transcription reactions, gel mobility shift assays, and determination of the binding constants were carried out as described above.