A small bacterial RNA regulates a putative ABC transporter.

A small noncoding bacterial ribonucleic acid of 62-64 nucleotides, RydC, was identified in the genomes of Escherichia coli, Salmonella, and Shigella. In vivo, RydC binds to the RNA-binding protein Hfq, and it is unstable when Hfq is absent. Mobility assays reveal that complex formation between RydC and Hfq is specific, with an apparent binding constant of approximately 300 nm. Sequence alignments combined with structural probing demonstrate that RydC folds as a pseudoknot. Hfq binds the loops crossing the deep and shallow grooves of the pseudoknotted RNA and reorganizes its overall conformation. An interaction with a polycistronic mRNA, yejABEF, which encodes a putative ABC transporter, was detected by affinity purification of immobilized RNA-Hfq complexes. In vivo, the yejABEF operon is expressed on minimal medium. Remarkably, its expression is reduced when RydC is absent, and the operon is degraded when RydC expression is stimulated. This observation correlates with the growth defects associated with a stimulation of its expression in vivo, generating a thermosensitive phenotype that affects growth on minimal media supplemented with glycerol, maltose, or ribose. We conclude that RydC regulates the yejABEF-encoded ABC permease at the mRNA level. This small RNA may contribute to optimal adaptation of some Enterobacteria to environmental conditions.

A number of 40 -400-nt 1 RNAs that generally do not encode proteins or function as transfer or ribosomal RNAs have been characterized in Escherichia coli. Because of their small sizes, they have been referred to as small (s) or noncoding (nc) RNAs. Initially, a dozen sRNAs were identified in E. coli on the basis of their high abundance or by serendipity. In the last few years, computational, microarray and cloning-based screens have led to the identification of around 50 additional sRNAs in E. coli (for recent reviews, see Refs. 1 and 2). These RNAs act mainly by pairing with other RNAs, are part of RNA-protein complexes, or adopt the structures of other nucleic acids (3).
Bacterial sRNAs base pairing with target mRNAs can have various regulatory fates: sRNAs can repress or activate translation by blocking or promoting ribosome binding to mRNAs (3). They can also destabilize or stabilize mRNAs by increasing or decreasing accessibility to RNases. Base pairing between some sRNAs and their RNA targets requires the participation of the Hfq protein, a homolog of the Sm and Sm-like eukaryotic proteins involved in mRNA splicing. Hfq binds AU-rich sequences and forms a homohexameric ring (4). It has been proposed that Hfq acts as an RNA chaperone to promote base pairing interactions between Hfq-binding sRNAs and their targets. It can protect many sRNAs and also mRNAs against RNase E digestion (5). The mechanisms by which Hfq facilitates interactions between sRNAs and their targets are, however, poorly understood.
In this report, the functional and structural identification of a novel sRNA that folds as a pseudoknot and binds Hfq in vitro and in vivo is described. The RNA was detected in the enterobacteriaceae family, in 21 sequenced strains from the three genera Escherichia, Salmonella, and Shigella. Its existence was independently detected by microarrays of Hfq-immunoprecipitated E. coli RNAs and Northern analysis (6), and it was termed RydC. We have identified by affinity chromatography an mRNA target of RydC that specifies a predicted ABC transport system. In vivo, RydC regulates the expression of the ABC permease at the mRNA level. When the expression of RydC is stimulated in vivo, the mRNA encoding the transporter gets degraded, and growth defects are observed on minimal medium with glycerol, maltose, or ribose as the carbon sources.
RNA Isolation and Northern Blots-E. coli strains were grown in either Luria-Bertani (LB) or minimal media (M9) and harvested at the indicated A 600 . The cell pellets were resuspended in Trizol (Invitrogen). Total RNA extraction was performed as suggested by the manufacturer. Total RNAs were isolated either by Trizol reagent (Invitrogen) or by acid-phenol extraction, in which the cell pellet was dissolved in 0.2 M sodium acetate, 10 mM EDTA, 1% SDS (pH 5.0), with a volume of water-saturated phenol pre-heated at 65°C. Incubation was performed for 10 min at 65°C and 0.5 volume of CHCl 3 :isoamylic alcohol (24:1). The sample was incubated for 10 additional minutes at 65°C. Total RNAs were precipitated overnight at 4°C with 2 M LiCl (final concentration), washed with 80% ethanol, and incubated with an RNase-free DNase followed by phenol extraction. For RydC and transfer-messenger RNA, Northern blots were carried out by loading 30 g of freshly extracted RNAs per lane onto an 8% PAGE in 1ϫ Tris-borate EDTA for 3 h at 350 V. For yejABEF mRNA, we used a 1% agarose gel in 2.2 M formaldehyde, in 20 mM MOPS (pH 7.0), 8 mM sodium acetate, and 1 mM EDTA (pH 8.0). For the blots performed on polyacrylamide gels, the transfer of the RNAs was achieved in 0.5ϫ Tris-borate EDTA onto charged nylon membranes (Zeta probe GT; Bio-Rad) for 3 h at 20 V at 4°C. For the blots performed on agarose gels, the transfer of the RNAs was achieved in 20ϫ SSC onto neutral nylon membranes overnight at room temperature (passive transfer). Either 100 -1000-or 200 -10000nt-long RNA markers (Novagen) were used. To monitor yejABEF expression in vivo in a wild-type strain, ⌬RydC, and pUC-RydC strains, primer P7 (5Ј-GTATCGATACCGAGGCTATAGATGGGGC-3Ј) was used for hybridization. To monitor RydC expression in vivo in a wild-type strain, ⌬RydC, and pUC-RydC strains, primer P8 (5Ј-ACCGACCCGT-GGTACAGGCG-3Ј) was used for hybridization.
Primer Extension and RACE-Total RNAs (15 g) were annealed to 5Ј-labeled primer P8 at 90°C for 30 s and 65°C for 5 min, cooled down on ice, and subjected to primer extension at 42°C for 30 min with 5 units of avian myeloblastosis virus reverse transcriptase (Qbiogene) and 0.5 mM deoxynucleotide triphosphates in a volume of 5 l. The cDNA products were separated on a 6% PAGE. RACE assays were carried out according to Ref. 10, with modifications: for the 5Ј-RACE, 10 g of total RNAs, with and without a treatment with tobacco acid pyrophosphatase, was denatured in presence of 300 pmol of RNA adapter R1 (5Ј-UGGCGGACGCGGGUUCAACUCCCGCCAGCUCCA-CCA-3Ј) (Dharmacon) at 95°C for 3 min and cooled down on ice. Ligation was performed at 16°C overnight with 10 units of T4 RNA ligase (New England Biolabs) in 50 mM Tris-HCl (pH 7.9), 10 mM MgCl 2 , 4 mM dithiothreitol, 150 M ATP, and 10% Me 2 SO. After phenol chloroform extraction and ethanol precipitation, 2.5 g of RNA was reverse-transcribed with 2 pmol of gene-specific complementary DNAs P8 or P9 (5Ј-GTACAGGCGAAGAATACGGG-3Ј) with 20 units of avian myeloblastosis virus reverse transcriptase (Qbiogene) at 42°C for 30 min in a final volume of 20 l. 1 l of the reaction was used for PCR amplification by 15 pmol of primer P8 or P9 and adapter-specific homologous primer P10 (5Ј-AACTCCCGCCAGCTCCACCA-3Ј), 1 unit of Taq DNA polymerase (New England Biolabs) in 2.5 mM MgCl 2 , and 0.2 M of each deoxynucleotide triphosphate. Products were separated on a 10% PAGE, eluted, and cloned into PCR2.1 TOPO vector (Invitrogen). Recombinant bacteria were selected by PCR, and the positive ones were sequenced by M13 reverse and universal DNA primers. For the 3Ј-RACE, calf intestinal phosphatase-treated total RNA was ligated to 5Ј-phosphorylated R1. Reverse transcription was performed with 200 pmol of primer P11 (5Ј-AGTTGAACCCGCGTCCGCCA-3Ј) complementary to R1. PCR amplification was performed with gene-specific DNA P12 (5Ј-GATGTAGACCCGTATTCTTCG-3Ј), as described for the 5Ј-RACE.
Gel-shift Assays-Hfq was overproduced from strain BL21(DE3) transformed with pTE607 plasmid (kindly provided by Dr. Hajnsdorf, Institut de Biologie Physico-Chimique, Paris, France). E. coli BL21(DE3) cells were grown in 3 liters of LB medium at 37°C to an A 600 of 0.4, induced with 1 mM isopropyl 1-thio-␤-D-galactopyranoside for 3 h, and centrifuged for 20 min at 5000 ϫ g (4°C). The bacterial pellets were dissolved in 20 mM Tris-HCl (pH 7.8), 500 mM NaCl, 10% glycerol, and 0.1% Triton X-100 in the presence of protease inhibitor mixture (Roche Applied Science) and sonicated on ice (6 rounds of 10 s each with 10-s pauses), followed by treatment with DNase I (100 units/ml, 15 min, 37°C) and a 20-min centrifugation (10,000 ϫ g at 4°C). The supernatant was filtered (0.45 ), loaded onto a FPLC (Amersham Biosciences) equipped with a Ni 2ϩ column washed with 50 mM NaH 2 PO 4 , 300 mM NaCl, and 10 mM imidazole; equilibrated at pH 6.0; and eluted in a similar buffer, except for the concentration of imidazole (300 mM). The protein was pure, as shown on a 12% SDS-PAGE (the monomeric and multimeric forms of Hfq were visible, even after boiling the sample in SDS buffer), heated for 15 min at 80°C, and centrifuged for 10 min at 13,000 ϫ g at room temperature, and the supernatant was concentrated on 5kD Amicon in 50 mM Tris-HCl (pH 7.5), 1 mM EDTA, 50 mM NH 4 Cl, 5% glycerol, and 0.1% Triton X-100. Protein concentration was determined using the Bradford assay, and the protein was stored at 4°C.
The transcripts were denatured in 50 mM Hepes (pH 6.9), 50 mM NaCl, 5 mM KCl, and 1 mM MgCl 2 for 3 min at 85°C, followed by refolding for 10 min at 30°C and chilled on ice. 0.02 to 1 pmol of labeled RNA was incubated with a 1000 -1500 molar excess of carrier yeast tRNAs, 0.2 g/l bovin serum albumin, and the indicated amounts of purified Hfq. The binding reactions were in 10 mM Tris-HCl (pH 7.5), 6 mM NaCl, 10 mM EDTA, and 5 mM dithiothreitol for 10 min at 30°C. The samples were supplemented with 10% glycerol (final concentration) and loaded on a native 4% polyacrylamide gel containing 5% glycerol. The electrophoresis was performed in 0.5ϫ Tris-borate EDTA supplemented with 0.5% glycerol at 4°C for 4 h (100 V). The results were analyzed and quantified either on a PhosphorImager (Amersham Biosciences) or directly by autoradiography.

Isolation of RNAs That Bind to a RydC-Hfq Complex-Total
RNAs were extracted at A 600 ϭ 0.4 . Both the 16S and the 23S rRNAs were removed (Single Place Magnetic Stand; Ambion). RNAs (10 g) depleted in both 16S and 23S rRNAs, corresponding to roughly 30 g of total RNAs, were 3Ј-end-labeled (as described for RydC, see above) and purified from the unincorporated [␣-32 P]pCp using MicroSpin TM G-25 columns. A complex between 1.5 pmol of refolded cold RydC and 37.5 pmol (a 25-fold molar excess) of His-tagged Hfq was formed in 10 mM Tris-HCl (pH 8.0), 50 mM NaCl, 50 mM KCl, and 10 mM MgCl 2 at 37°C for 15 min (at that ratio, all RydC was bound to Hfq). 3Ј-labeled bacterial RNAs were added to the RydC-Hfq complex, incubated for an additional 15 min at 37°C, and loaded on a 0.5 ml TALON® metal affinity column (Clontech) in 20 mM Tris-HCl (pH 7.8), 500 mM NaCl, 10% glycerol, and 0.1% Triton X-100. The elution was performed in three steps. In each, 400 l of 50 mM NaH 2 PO 4 (pH 6.0), 300 mM NaCl, and 300 mM imidazole was added to the column (first at 37°C for 30 min, and then at 37°C for 60 min) and, finally, 400 l of 50 mM NaH 2 PO 4 (pH 6.0), 300 mM NaCl, and 300 mM imidazole was added to the column, and the resin was treated with an equal volume of phenol (pH 4.5). The three fractions were collected, treated with phenol/chloroform, and ethanol-precipitated. The eluted RNAs were cloned as described for the 3Ј-RACE. The terminal 5Ј-P of the eluted RNAs was removed by the calf intestinal phosphatase. The eluted RNAs are 3Јend-labeled with [␣-32 P]pCp, thus a 5Ј-OH R1 can be ligated at their 3Ј-ends, and cDNAs were generated by primer extension with P11. R1 phosphorylated at its 5Ј-end was then ligated at the 3Ј-ends of the cDNAs, PCR-amplified, and loaded onto a 10% PAGE. The PCR fragments were eluted individually, re-amplified by PCR, cloned, and sequenced.
Structural Analyses-End-labeled and gel-purified RydC were folded in 50 mM Hepes (pH 6.9), 50 mM NaCl, 50 mM KCl, and 1 mM MgCl 2 for 3 min at 85°C and then folded for 10 min at 30°C and chilled on ice. For secondary structure analysis, the RNAs were partially digested at 30°C for 10 min with 10 Ϫ4 or 5.10 Ϫ5 unit of RNase V 1 in 50 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , and 50 mM KCl with 2.5 g of yeast tRNAs. Limited digestions with 0.2 and 1 unit of nuclease S 1 were performed at 30°C for 10 min in 25 mM sodium acetate (pH 4.5), 10 mM MgCl 2 , 50 mM KCl, and 1 mM ZnCl 2 with 2.5 g of yeast tRNAs. Partial alkaline hydrolysis was performed in 0.1 M NaOH and 2 mM EDTA at 90°C for 50 s. RNase T 1 sequencing was performed in 20 mM sodium acetate (pH 5.0), 1 mM EDTA, and 7 M urea at 55°C for 6 min in the presence of 10 g of yeast tRNAs with 2 units of RNase T 1 (final volume, 5 l). The samples were directly loaded on a 12% PAGE. Probing the structure of either 5Ј-or 3Ј-end-labeled RydC in complex with Hfq was performed after refolding the RNA as described above. Then, RydC was bound to a 50-fold molar excess of Hfq. The complex was partially digested at 30°C for 7 min with either 10 Ϫ4 or 5.10 Ϫ5 unit of RNase V 1 or 0.5 unit to 5 units of nuclease S 1 supplemented with 1 mM ZnCl 2 . There was either a 400-or a 1200-fold molar excess of yeast tRNA when probing the conformation of RydC in complex with Hfq.

Identification, Expression, and End
Mapping of a Novel sRNA-Based on sequence conservations between phylogenetically related species, a computer approach was developed (11) to identify novel sRNAs expressed from the intergenic regions of the E. coli genome. All those previously characterized (1) were identified, plus new ones. Among them, Ϫ10 and Ϫ35 promoter sequence signals flanked at their 3Ј-sides by a 50 -60-nt-long "GC-rich" region ending by a putative "Rho-independent" terminator could be predicted (Fig. 1A). We selected one of these putative sRNA-encoding genes that was identified in 21 sequenced bacterial genomes, all from the family of the enterobacteriaceae (6 E. coli, 11 Salmonella, and 4 Shigella sequences), for additional studies. A sequence alignment of those that possess sufficient sequence variations is presented in Fig. 1A. While our work was in progress, the presented putative E. coli sRNA was isolated by co-immunoprecipitation with Hfq, identified using microarray and Northern blot analysis, and named RydC (6), the name by which we will refer to it. The RydC-encoding gene, rydC, is located at 32 min on the E. coli genetic map between cybB and ydcA encoding cytochrome b 561 and a hypothetical protein of 5.9 kDa, respectively, both located on the complementary DNA strand (Fig. 1B).
Based on the alignment of seven sequences with high nucleotide identity, RydC is proposed to fold as a pseudoknot, as confirmed experimentally using structural probes (see below). Two RNA helices H1 (7 bp) and H2 (9 bp) are predicted. H2 is part of a Rho-independent terminator. Divergent rydC sequences possess three compensatory mutations to maintain the pairings within RNA helix H2. Predicted helices H1 and H2 are entangled and connected by three nt stretches L1, L2, and L3. Based on the known genes, L1 has 6 -7 "pyrimidine-rich" nt, L2 has a single conserved Cys residue, and L3 has 7-8 nt.
Primer extension analysis on total cellular RNA demonstrates that the predicted promoter is functional in vivo (Fig.  1C). Determination of the 5Ј and 3Ј boundaries of RydC was required for subsequent functional and structural analyses. They were determined using 5Ј-and 3Ј-RACE (Fig. 1D). From several DNA clones, RydC 5Ј-end mapped at two adjacent positions (Fig. 1A, boxed 5Ј-CT) 5 or 6 nucleotides downstream from the "Ϫ10 box" of the predicted promoter. RydC 3Ј-end also mapped at two adjacent positions (Fig. 1A, boxed 3Ј-TT or TG), downstream from the 3Ј-strand of H2 ending by a T 4 C stretch that forms a Rho-independent terminator. As determined from the major 5Ј-and 3Ј-end points, RydC has 62-64 nucleotides in vivo, consistent with the size estimated from our computer prediction.
RydC Forms a Complex with Hfq-Because RydC interacts with Hfq in vivo (6), we hypothesized that its quantity could be Hfq-dependent. RydC expression was compared by Northern blots between a strain deficient for Hfq (⌬hfq) and its parental strain. In contrast to the parental strain, RydC could not be detected in the derivative that does not express Hfq ( Fig. 2A). It suggests that in the absence of Hfq, RydC is unstable and rapidly degraded by RNases. Stability of sRNAs can be reduced in the absence of Hfq that protects them against RNase activity, including RNase E (1,5). Alternatively, Hfq might be required for the transcription of RydC. The binding of Hfq to RydC was tested in vitro by gel-shift assays using increasing amounts of purified Hfq and constant amounts of labeled synthetic RydC. Binding assays were performed at various salt conditions, constant pH (pH 7.5-8.0), and a large excess of bulk tRNA from yeast to reduce aspecific binding (from a 30 to a 1500 molar excess relative to RydC). The apparent binding constant between RydC and Hfq varies from 120 to 500 nM, depending on the amount of carrier tRNA (the lowest association constant corresponds to the lowest molar excess of competitor RNA). Retardation assays with increasing amounts of purified Hfq indicate that the RNA-protein complex is detected in vitro at physiological pH and salt concentrations (Fig. 2B shows a typical assay). The binding between RydC and Hfq is specific because a 4000-fold molar excess of total tRNA displaces only a minor fraction of RydC from a preformed RydC-Hfq complex, whereas a 50-fold excess of cold RydC competed labeled RydC out of the complex (Fig. 2C).
Structural Analysis of RydC and Its Interaction with Hfq-Sequence variation is too low (the sequences aligned have a high sequence identity) to fully establish the secondary structure of RydC by a phylogenetic analysis. Therefore, its conformation was analyzed further by structural probes in solution, an approach that was instrumental in establishing the secondary structures for many RNAs (12). A RydC transcript was end-labeled, and its solution conformation was probed by enzymes. RNase V 1 cleaves double-stranded RNA or stacked nucleotides, whereas nuclease S 1 cleaves single-stranded RNA. The reactivity toward these probes was monitored for each nucleotide of a 64-nt-long synthetic RNA, in the absence and presence of the protein Hfq. Four independent experiments were performed on RydC alone, and four additional ones were performed on the Hfq-RydC complex (Fig. 3, AϪC, is representative). These data are summarized on secondary structure models that they, together with the phylogenetic analysis, support (Fig. 3, D and E).
Double-stranded-specific cuts from C15 to U18 and at the predicted G12-C41 pair and the absence of nuclease S 1 cleavages at G12-U18 and A35-C41 suggest that helix H1 forms in solution (Figs. 1A and 3D). RNase V 1 cuts at U29-U31, G54, and G57 and the absence of S 1 cleavages between C25-C33 and G49-G57 support the existence of a 9-bp helix H2 (Figs. 1A and 3D). S 1 cleavages at A19-U23 and U44-U47 are consistent with loops L1 and L3 being mostly single-stranded in solution. According to the sequences, the nt content of loop L1 varies from six to seven, and that of L3 varies from seven to eight. L1 and L3 cross the deep and shallow grooves of the RNA structure that folds as a pseudoknot (13). A conserved unpaired Cys residue is in between helices H1 and H2 that forms L2. L2 is not cut by S 1 , probably because it is not accessible for cleavage. The strong S 1 cleavages between U58 and U63 suggest that the "uridine-rich" 3Ј-end of RydC is unpaired. Nucleotides C5 and G6 are cleaved by both single-stranded-and double-strandedspecific probes, suggesting that the 5Ј-end of RydC breathes in solution, as for other bacterial RNAs (14). Alternatively, the nt segment 4CCGAU8 can pair transiently with 39GUCGG43 (Fig. 3D, boxed nt), accounting for the V 1 cuts at nt C5 and G6, forming an extended helix interrupted by an internal bulge at nt G9-C14. There are no detectable degradation sites within the sequence of RydC.
The interaction between RydC and Hfq was also monitored by structural probes (Fig. 3, B, C, and E). In the presence of Hfq, nuclease S 1 cleavages disappear at U21-U24 in loop L1 and at U44-U47 in loop L3. These regions are accessible "uridine-rich"" sequences. Because Hfq binds "uridine-rich" sequences close to structured domains in target RNAs (15), it suggests that the protein binds L1 and L3, flanked by H1 and H2. Protection of L3 by Hfq is less obvious than for L1. L3 may only be a secondary binding site, a fraction of the RNA does not bind Hfq at L3 or Hfq does not bind L3 and the reactivity change is induced by an indirect structural effect. The stoichiometry of the binding of Hfq to RydC is not known and might

FIG. 2. Stabilization of RydC by Hfq in vivo and assessment of the affinity and specificity of the interaction.
A, detection of RydC expression by Northern hybridization using a labeled strand-specific probe during bacterial growth in LB medium in cells that do not express Hfq because of chromosomal gene disruption (Hfq ⌬) compared with a control strain (Hfq wt). B, native gel retardation assay of purified labeled RydC with increasing amounts of purified His-tagged Hfq. C, the interaction between Hfq and RydC is specific. Native gel retardation assay of labeled RydC and Hfq in the presence of increasing amounts of either unlabeled total E. coli tRNA or unlabeled RydC. be higher than 1:1 because the protein folds as a hexamer in solution. The pattern of some RNase V 1 cuts at stems H1 and H2 also varies: V 1 cuts at C15-C16 in H1 and at U29-U31 in H2 are weaker in the presence of the protein, probably because Hfq binds at nearby L3 and L1, respectively, therefore preventing the V 1 cuts by steric hindrance. C34, which connects H1 to H2, is cut by V 1 in the presence of Hfq, suggesting that both stems stack onto each other. In the presence of Hfq, G39-U40 gets cleaved by V 1 , and a stronger V 1 cut is observed at C41 (Fig. 3). Therefore, the conformation of RydC reorganizes in the presence of Hfq, especially the lower part of H1 that could involve alternate pairings (Fig. 3, black boxes) flanked by a G9-C14 internal bulge (two pairs, G9-C14 and U10-A13, can form within the bulge).
Affinity Purification of an mRNA Fragment That Binds a Preformed RydC-Hfq Complex-Bacterial sRNAs that interact with Hfq regulate gene expression at the transcriptional and/or translational level via mRNA-sRNA pairings (3). Affinity chromatography was used to pull putative target RNA(s) out of a preformed RydC-Hfq complex from cellular extracts. Complex was formed between purified His-tagged Hfq and synthetic RydC at a 25-fold:1-fold molar ratio; at that ratio, Ͼ95% of Hfq interacts with RydC ( Fig. 2B and "Experimental Procedures"). The RNAs that are retained by the column (the RydC-Hfq complexes were preloaded onto a nickel resin) were eluted, ligated at both ends with RNA adapters, amplified by reverse transcription-PCR, cloned, and sequenced. A ϳ350-nt fragment (part of its sequence is 5Ј-AACGGGTGTCGGCAATATCAGC-GACAGTAATTACCGT-3Ј) had a perfect match with nt 192-228 (nt 2272390 -2272426) from the coding region of the 1095nt-long yejB mRNA, a putative membrane permease (364 amino acids at position 2272199 -2273293 in E. coli K12 MG1655, extracted from the colibri data base at genolist.pasteur.fr/Colibri/). According to the size of the PCR fragment, it also contains 30 -40 nt upstream of the AUG initiation codon, which correspond to the 3Ј-end of the coding region of yejA mRNA, the putative periplasmic binding protein from the transport system. yejA and yejB belong to a predicted 5532-nt operon that encodes two additional genes: yejE is the second half of the permease, and yejF is the ATP binding component of the putative transport system. When blasting the selected sequence against all the sequenced bacterial genomes, only those from Shigella and E. coli have a nearly perfect sequence identity with yejA and yejB. Remarkably, those bacteria also encode There is a 50:1 molar ratio of the protein to the RNA, with a 1200 molar excess of carrier tRNA. Nucleotides whose reactivity is unaffected upon protein binding are represented by gray circles. Nucleotides with modified reactivity in the presence of Hfq are in capital letters. Symbols are as defined in D. Enzymatic cleavages that are reduced (Ϫ) or enhanced (ϩ) upon protein binding are indicated. Upon Hfq binding, five alternate pairings between nt located within the two black boxes are proposed.
RydC, whereas the others do not. It suggests a functional link between the selected sequence and rydC.
The mRNA Target Binds Hfq-Hfq binds some mRNAs with affinity and specificity (16,17). A purified synthetic 129-nt-long RNA (yejAB), corresponding to the nucleotide sequence of the yejA-yejB junction (91 nt from yejA and 38 nt from yejB) within the polycistronic mRNA yejABEF, was produced by in vitro transcription. Initial work was performed with a 350-nt-long RNA that was shortened to minimize nonspecific binding in the gel-shift assays. Native gel retardation assays were performed between labeled yejAB and increasing amounts of purified Hfq, in the presence of an excess of total tRNA (Fig. 4A). As for RydC, the mRNA fragment interacts with Hfq. At a ϳ170 molar excess of total tRNA, the apparent association constant of the interaction between yejAB and Hfq is ϳ3 M; at a ϳ3000 molar excess of total tRNAs, the apparent association constant is ϳ30 M. Therefore, complex formation between the mRNA fragment and recombinant Hfq is 25-50-fold weaker than that between RydC and Hfq. Complex formation between yejAB and Hfq is specific because a 1500-fold molar excess of bulk tRNA does not displace yejAB from a preformed yejAB-Hfq complex, whereas a 500-fold excess of cold yejAB prevents complex formation (Fig. 4B). Two conformations of yejAB probably coexist in solution because two bands are detected on a native gel. The one that migrates faster is predominant, and both conformers bind Hfq (Fig. 4). Alternatively, one could be a degradation product. In the presence of a 100-fold molar excess of cold RydC, there is no detectable complex formation between RydC and labeled yejAB (Fig. 4C).
RydC Regulates the Expression of the yejABEF mRNA Operon-A direct correlation between the expression levels of RydC and its target mRNA was detected in vivo. Northern blots demonstrate that the mRNA corresponding to yejABEF is expressed in wild-type cells grown in minimum media (Fig. 5). The expression of the mRNA is maximal during mid-log and stationary phases. As estimated from 200-nt to 10-kb RNA markers, the mRNA has ϳ5200 nt (Ϯ15% variation), in agreement with its predicted size of 5532 nt based on the genomic sequence. Remarkably, in a strain deficient for RydC (⌬rydC; Fig. 6B), a lower amount of mRNA is detected up to mid-log phase, compared with wild-type. Also, the mRNA operon is detected in ⌬rydC at A 600 ϭ 0.2, but not in wild-type cells. Fig.  5 is a representative of results that have been reproduced three times with different RNA extractions. In a strain that overexpresses RydC (rydCϩϩ; its expression profile during cell growth is shown in Fig. 6B) the mRNA operon is degraded, whereas the overall cellular RNAs are not (Fig. 5, bottom  panel). A DNA probe specific for 16 S rRNA was used as an internal control of loading identical amounts of total RNA per lane and also as a ϳ1500-nt RNA size marker. In rich medium (LB), the three bacterial strains do not express the yejABEF mRNA (data not shown). Interestingly, the well-characterized ABC oligopetide transporter in E. coli, oligopeptide permease (opp), has an expression that is also repressed in rich medium (LB) but is identifiable in minimum medium (18).
Growth Defects Associated with Enhancing RydC Expression-The expression profile of RydC during cell growth was monitored by Northern blots and compared with the expression pattern of transfer-messenger RNA, a highly expressed sRNA. Transfer-messenger RNA expression is optimal during the exponential phase when translation levels are high and decreases slightly during stationary phase. An E. coli strain deficient in RydC expression (⌬rydC) was constructed (see "Experimental Procedures"), and the absence of RydC was confirmed by Northern blot analysis (Fig. 6B, left panel). The ⌬rydC strain FIG. 4. In vitro binding assays between yejAB and Hfq. A, native gel retardation assay of a labeled synthetic mRNA fragment (two conformations, gray and black arrows) corresponding to the nt sequence at the junction between yejA and yejB with increasing amounts of purified Hfq (molar ratios of 10 -1000 Hfq per mRNA). B, the interaction between Hfq and yejAB is specific. In vitro competition assays with an excess of either total tRNA or unlabeled yejAB. C, RydC does not bind yejAB in the absence of Hfq, even at a 100-fold molar excess. displays no detectable growth differences in either rich or minimal medium at 30°C, 37°C, or 42°C as compared with the parental strain (data not shown). An E. coli strain harboring a multicopy plasmid containing rydC (pUC-RydC) has an increased amount of RydC (Fig. 6B, right panel). The growth of this strain on minimal medium (M9) supplemented with different carbon sources at 30°C, 37°C, or 42°C was compared with the growth of the parental strains. Whereas no differences were observed in glucose, the presence of pUC18-RydC generated a thermosensitive phenotype that drastically affected growth on a minimal medium supplemented with glycerol, maltose, or ribose at 42°C (Fig. 6C). DISCUSSION A small, noncoding, ribonucleic acid, RydC, was identified in three Enterobacteria by a systematic search for novel small bacterial RNAs. In E. coli cells, the expression of RydC during growth was monitored, and its 5Ј-and 3Ј-ends were mapped, leading to a size of 62-64 nt. RydC co-immunoprecipitates with Hfq (6), and we show here that the protein is required for its stability and activity in vivo. In vitro, RydC binds Hfq with high affinity and specificity, and the apparent binding constant of complex formation between the RNA and the protein is compatible with its presence in vivo. In solution, RydC folds as an RNA pseudoknot, and its conformation is modified upon binding of Hfq. Probing suggests that the protein binds the two connecting single-stranded loops within the pseudoknot, induces a rearrangement of the pairing in one helix, and stacks the two helices, probably to set RydC for optimal pairing with its target mRNA.
All the sRNAs that bind and require Hfq for activity act by pairing to target mRNA(s), perturbing their stability and/or translation. Hfq increases RNA unfolding or local target mRNA concentration. Affinity chromatography could detect a fragment of a cellular message that binds to immobilized RydC-Hfq complexes. It corresponds to a fragment of an operon encompassing four genes, yejA, yejB, yejE, and yejF. In E. coli cells, we have detected expression of the ϳ5.5-kb yejABEF polycistronic mRNA in minimal media. We have monitored the phylogenetic distribution of this sRNA-mediated regulation to assess its functional importance in Enterobacteria. In the genome of Salmonella paratyphi A, we could not identify RydC, and yejF is missing from the yejABEF operon, suggesting that the putative transport system is not functional in that species and that its associated sRNA is also missing. In the genome of the closely related species Salmonella paratyphi B, however, we have identified both RydC and yejABEF genes. Based on the apparent binding affinities, the RydC-Hfq complex forms and then binds to yejABEF mRNA. The binding of Hfq to RydC is probably required to fold the RNA in an active conformation and to facilitate short and perhaps noncontiguous pairing between the trans-encoded RydC and its target mRNA. Transient ternary complex formation between RydC, the mRNA, and Hfq could facilitate recognition between the two RNAs. Hfq binds and affects the stability of several bacterial mRNAs (19). Because Hfq binds both the target message and RydC, one Hfq hexamer could bind RydC and yejABEF simultaneously to increase local RNA concentration. Alternatively, one Hfq hexamer could bind RydC, a second hexamer could bind yejABEF, and the two hexamers could be brought together via interactions between their hydrophobic backs, as suggested previously (3).
The targeted operon encodes a putative ABC permease, but RydC might also regulate other targets encoded at separate locations of the chromosome. In Gram-negative bacteria, the basic units of the importers consist of an auxiliary periplasmic binding protein (the transporters that possess a periplasmic component are involved in cellular import), two membraneassociated domains, and two ATP-binding domains. Based on sequence similarity, yejABEF probably encodes a permease (20), in which yejA encodes a periplasmic binding protein, yejB and yejE encode the two membrane proteins, and yejF is the ATP-binding protein. In E. coli, the major peptide transport systems are the dipeptide permease, the oligopeptide permease, and the murein tripeptide transport from cell wall turnover (21). The expression of the operons encoding oligopeptide permease and dipeptide permease is regulated by a small RNA, gcvB, by an unknown mechanism (22).
The regulation of the yejABEF operon by RydC is achieved at the mRNA level. Putative pairings between RydC and yejABEF mRNA have been identified but require direct experimental support. Impairing RydC expression reduces the amount of the target mRNA without triggering mRNA degradation and alters the temporal expression of the operon. Therefore, endogenous expression of RydC in wild-type cells positively affects the accumulation of its target RNA, as recently described for an-FIG. 6. Growth defects associated with enhancing RydC expression. A, detection of RydC expression levels by Northern hybridization using strand-specific probes during bacterial growth of K12 C25113 cells in LB medium. Northern blots with labeled DNA complementary to either RydC (bottom panel) or transfer-messenger RNA, which was used as a control (top panel). B, RydC expression in vivo is either prevented because of chromosomal gene disruption (RydC ⌬) or increased from a multicopy plasmid that encodes RydC expressed from its natural promoter sequence (RydC ϩϩ). C, a multicopy plasmid containing rydC confers thermosensitive growth on certain media. Strain MG1655Z1 containing either pUC18 or pUC-RydC grown overnight in LB medium at 30°C was pelleted and resuspended in M63. Serial dilutions (10-fold, from left to right) were spotted (5 l) on M63 plates supplemented with either glucose, glycerol, maltose, or ribose at 30°C, 37°C, or 42°C, as indicated.
other sRNA in E. coli (23). No growth defects are observed in the absence of RydC, probably because there is still some intact yeABEF mRNA in the cells that can be translated into functional permeases. When RydC expression is stimulated, however, it leads to yejABEF mRNA degradation. High amounts of RydC destabilize the cellular message, probably by enhancing ribonuclease access and turnover, as for other bacterial sRNAs (3). The absence of the mRNA encoding the permease might trigger the thermosensitive phenotype that affects growth on minimal medium supplemented with specific carbohydrates, especially if the transporter is involved in sugar import. When monitoring the expression of yejABEF mRNA at identical absorbances (at stationary phase) in wild-type cells grown in minimal media supplemented with either glucose or ribose, those grown on ribose express higher amounts of the mRNA operon than those grown on glucose. This preliminary result suggests that the growth defect observed on ribose as the carbon source when RydC expression is stimulated is due, at least in part, to the degradation of the mRNA yejABEF. It also suggests that ribose is imported by the yejABEF-encoded permease. Alternatively, although Hfq is abundant in E. coli cells, the induction of RydC may compete with the binding of other sRNAs for the protein and impact their regulatory functions. In summary, the amount of RydC per cell is precisely set to allow the optimal expression of the yejABEF mRNA operon when the nutrients are scarce.
In some Enterobacteria, the expression of two abundant outer membrane porins is also regulated by two sRNA genes (24). Enterobacteria colonize the digestive tract of humans and animals and are attached to the epithelial cells within the digestive tract where the absorption of nutrients takes place. They are facing variations in nutrients concentrations, and a tight regulation of nutrient uptake may be required for growth and survival. 10 14 bacteria representing 400 species live in the digestive tract. The Bacteroides are the dominant flora, and the Enterobacter are only subdominant and located exclusively in the colon. Glucose absorption takes place in the stomach. En-terobacteria may require specific nutrient uptake mechanisms to capture sugars different from glucose to survive in a competitive environment.