A repeated GGA motif is critical for the activity and stability of the riboregulator RsmY of Pseudomonas fluorescens.

The riboregulator RsmY of Pseudomonas fluorescens strain CHA0 is an example of small regulatory RNAs belonging to the global Rsm/Csr regulatory systems controlling diverse cellular processes such as glycogen accumulation, motility, or formation of extracellular products in various bacteria. By binding multiple molecules of the small regulatory protein RsmA, RsmY relieves the negative effect of RsmA on the translation of several target genes involved in the biocontrol properties of strain CHA0. RsmY and functionally related riboregulators have repeated GGA motifs predicted to be exposed in single-stranded regions, notably in the loops of hairpins. The secondary structure of RsmY was corroborated by in vivo cleavage with lead acetate. RsmY mutants lacking three or five (out of six) of the GGA motifs showed reduced ability to derepress the expression of target genes in vivo and failed to bind the RsmA protein efficiently in vitro. The absence of GGA motifs in RsmY mutants resulted in reduced abundance of these transcripts and in a shorter half-life (< or = 6 min as compared with 27 min for wild type RsmY). These results suggest that both the interaction of RsmY with RsmA and the stability of RsmY strongly depend on the GGA repeats and that the ability of RsmY to interact with small regulatory proteins such as RsmA may protect this RNA from degradation.

In prokaryotes, an increasing number of small, noncoding RNAs has been described; many of them have been predicted by bioinformatic approaches in Escherichia coli, raising their number to Ͼ60 in this organism (1)(2)(3). Riboregulators (i.e. small, untranslated, regulatory RNAs) regulate gene expression at a post-transcriptional level and can be grouped into two main classes. A first class displays antisense base-pairing activity, which can regulate target mRNA translation or stability; DsrA (87 nt) 1 and RyhB (90 nt) are two such examples in E. coli. Different domains of DsrA interact with rpoS and hns mRNAs, activating rpoS but inhibiting hns translation (4), whereas RyhB negatively regulates the expression of mRNAs coding for proteins involved in iron storage (5). DsrA, RyhB, and several other antisense riboregulators require the RNA chaperone Hfq for their activity (6,7).
A second group of riboregulators includes molecules that antagonize small, regulatory, mRNA-binding proteins of the CsrA type. In E. coli, CsrA (for carbon storage regulator) can act as a negative regulator of glgCAP and cstA target mRNAs, or as a positive regulator of fhlDC mRNA (8 -10). By binding to the leader region of target mRNAs, CsrA (a dimeric protein with a 61-aa subunit) can block translation and destabilize the mRNA (8). The regulatory functions of CsrA are counteracted by the expression of two riboregulators, CsrB (360 nt) and CsrC (257 nt) (11,12). Both RNAs can bind several molecules of CsrA, thus removing this protein from its target mRNA sites and thereby relieving post-transcriptional regulation (11,12). In the biocontrol organism Pseudomonas fluorescens CHA0, a similar system controls the expression of genes related to the production of antifungal metabolites and extracellular enzymes in late growth phase (13,14). This system comprises a CsrA-homolog, RsmA (for regulation of secondary metabolism) (15), and at least two riboregulators, RsmY and RsmZ (16,17). RsmA (62-aa) represses the translation of the hcnABC (coding for hydrogen cyanide synthase), aprA (coding for the major extracellular protease), and phlACBD genes (coding for proteins involved in the biosynthesis of 2,4-diacetylphloroglucinol) (15,16,18), probably by binding to the vicinity of the ribosome binding site (15). The riboregulators of this system, RsmY (118 nt) and RsmZ (127 nt), accumulate in late exponential phase under the influence of the global regulatory system GacS/GacA and promote expression of target mRNAs by binding several molecules of RsmA protein (16,17). Regulatory cascades involving homologs of GacS/GacA and RsmA (CsrA) as well as several different small RNAs have been described in other Pseudomonas species (19 -23), E. coli (24), Salmonella enterica (25,26), Erwinia sp. (27,28), Legionella pneumophila (29), and Helicobacter pylori (30).
Whereas the protein components of the Csr/Rsm systems can be easily identified by homology searches in many Gram-negative and some Gram-positive species (31), the riboregulators of the E. coli CsrB/C type are not homologous with those of the P. fluorescens RsmY/Z type. However, despite widely different nucleotide sequences, all these riboregulators perform essentially the same function, suggesting that they share common structural elements. When their predicted secondary structures are compared, the only salient feature is the occurrence of a trinucleotide motif (GGA) in the loops of hairpins and in other single-stranded regions (10,17). At least for CsrB, the number of GGA motifs appears to correlate with the maximal number of CsrA molecules that the riboregulator can bind (11). This

Essential GGA Motifs in RsmY RNA
prompted us to investigate whether the GGA motifs of RsmY of P. fluorescens are essential for the interaction with RsmA/ CsrA-type proteins. In addition, we explored the hypothesis that base pairing between RsmY segments and target mRNAs might contribute to the biological activity of RsmY. To test these hypotheses, we constructed a series of RsmY mutants and analyzed their ability to regulate the expression of RsmArepressed genes in P. fluorescens strain CHA0.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-The bacterial strains, vectors, and oligonucleotides used in this study are listed in Table I. Strains were routinely grown in NYB (2.5% (w/v) nutrient broth, 0.5% (w/v) yeast extract) with shaking (180 rpm), or on nutrient agar (4% (w/v) blood agar base, 0.5% (w/v) yeast extract) amended with the following antibiotics, when required: ampicillin, 100 g/ml; gentamicin, 10 g/ml; kanamycin, 25 g/ml; tetracycline, 25 g/ml (125 g/ml for P. fluorescens). Routine incubation temperatures were 37°C for E. coli and 30°C for P. fluorescens (before transformation, cells were grown at 35°C). For the quantification of HCN production, P. fluorescens strains were grown under oxygen limitation in tightly closed 125-ml bottles with 60 ml of glycine minimal medium (32), containing 0.05% (w/v) Triton X-100 to avoid cell aggregation.
DNA Manipulations-Small-and large-scale plasmid preparations were done with the cetyltrimethylammonium bromide method (33) and the Jetstar kit (Genomed GmbH), respectively. Chromosomal DNA from P. fluorescens was prepared as previously described (34). DNA manipulations were carried out with standard protocols (35). DNA fragments were purified from agarose gels with the QIAquick gel extraction kit (Qiagen Inc.). DNA sequencing was performed with the Big Dye Terminator Cycle sequencing kit and an ABI-prism 373 automatic sequencer (Applied Biosystems). PCR reactions were typically carried out with 2 units of thermostable DNA polymerase (Extra-Pol II, Eurobio) in a reaction mixture containing target DNA, 200 M of each of the four dNTPs (Roche Applied Science), 20 pmol of each of two primers, 1.5 mM MgCl 2 , and 1ϫ Extra-Pol buffer in a final volume of 20 l. For the amplification reaction, an initial denaturation step of 2 min at 95°C was followed by 25-30 cycles (30 s at 95°C, 30 s at 50-65°C (depending on the GϩC content and length of the primers), and 0.5-2 min at 72°C (depending on the length of the amplicon)) and a final elongation step of 2 min at 95°C.
Construction of rsmY Mutant Alleles and Integration into the Chromosome of P. fluorescens-RsmY variants were obtained by site-directed PCR mutagenesis using the wild type rsmY locus present on pME6914 as template (17). For each RsmY variant, four oligonucleotides were used (Table I). A first PCR product (400 bp) was obtained using primer orf76 (which anneals upstream of the rsmY promoter and of a natural PstI site in pME6914) and a reverse mutagenic primer annealing to wild type rsmY in a region where mutations were desired. A second PCR product (250 bp) was obtained by using the complementary forward mutagenic primer and lacZmut (which anneals to lacZ␣ downstream of a second PstI site of pME6914). This generated two "halves" of the rsmY locus having an overlapping sequence corresponding to the length of the mutagenic primer pair. The two gel-purified "halves" were used as both templates and primers for a third PCR with only eight cycles. Then, primers orf76 and lacZmut were added, and PCR was continued for another 25 cycles generating a full-length 650-bp product mutated at specific positions within the rsmY gene. The 650-bp products were digested with PstI, and the resulting 450-bp fragments containing the native rsmY promoter were cloned into pBLS for sequence confirmation and subcloning. RsmY mutants GGA-3, hcn, apr, and ctr were obtained using pME6914 as template, whereas GGA-5 was obtained using pME6923 (containing GGA-3). The different rsmY alleles were subcloned as PstI inserts into the suicide cloning vector pME3280b (18), which delivers the mini-Tn7 with passenger DNA into the unique Tn7 attachment site of the P. fluorescens chromosome in the presence of the Tn7 transposition helper pUX-BF13. The rsmY alleles, as well as the wild type 450-bp rsmY fragment from pME6914 and the empty mini-Tn7 (as positive and negative controls, respectively), were transferred to CHA825 (rsmY rsmZ), CHA826 (rsmY rsmZ hcnAЈ-ЈlacZ), and CHA827 (rsmY rsmZ aprAЈ-ЈlacZ) to generate the P. fluorescens strains CHA833 to CHA853 (see Table I for details). The presence of single copies of the different rsmY alleles in the chromosome was confirmed by PCR using the rsmY-specific primers TRR4 and TRR7 (data not shown).
RNA Manipulations-RNA preparations from P. fluorescens strains were obtained essentially as described previously (36), with minor modifications. Briefly, 200 -500 l of cell culture was centrifuged, and cells were resuspended in 500 l of TKM buffer (10 mM Tris-HCl, 10 mM KCl, 5 mM MgCl 2 , pH 7.5). Washed cells were mixed with 75 l of lysis solution (320 mM sodium acetate at pH 4.6, 8% SDS, 16 mM EDTA). Lysed cells were mixed for 5 min with 575 l of water-saturated phenol at 65°C. After centrifugation, the supernatant was extracted once with phenol-chloroform and precipitated with 3 volumes of ethanol. The resulting RNA pellet was dissolved in diethylpyrocarbonate-treated H 2 O and kept at Ϫ80°C. RNA concentration was determined at 260 nm. Purity and integrity of RNA preparations were assessed by denaturing agarose electrophoresis and ethidium bromide staining. RNA half-lives were estimated after addition of 200 g/ml rifampicin to the cultures, RNA extraction at indicated times, and analysis on Northern blots. Northern Blots-Northern blot analysis was done as described previously (17). Hybridizations were done with digoxigenin-labeled DNA probes generated by PCR covering the entire rsmY structural gene (17).
In Vivo Lead(II) Acetate Probing of RNA-The protocol used has essentially been described previously (37). P. fluorescens cells were grown at 30°C in NYB. Overnight cultures were diluted 100ϫ in NYB and grown to stationary phase (A 600 ϭ 3.0). Immediately before treatment, lead solutions were prepared by mixing 3 volumes of 1 M lead(II) acetate (Merck) in H 2 O with 1 volume of pre-warmed 4ϫ concentrated NYB. Unless otherwise stated, 8 ml of this lead(II) acetate/NYB solution was added to 20 ml of culture to give a final lead concentration of 100 mM. Cultures were further incubated with vigorous shaking at 30°C for 7 min. Reactions were stopped by addition of a 1.5-fold molar excess of EDTA and immediately put on ice. Total RNA was isolated and RNA samples were treated with DNase I (Amersham Biosciences) to eliminate DNA.
Lead(II) acetate cleavages were identified by primer extension. The 5Ј-32 P-end-labeled primers (GW-RsmY7, GW-RsmY9, or GW-RsmY11, see Table I)  Tris-HCl, pH 7.5, 0.1% SDS) and 3.5 l of 3 M KOH. After 3 min at 90°C, and subsequently 3 h at 37°C, 6 l of 3 M acetic acid was added, and the cDNA was precipitated with ethanol. Reverse transcription products were run on 8% polyacrylamide gels containing 7 M urea. Cleavage positions were identified by comparison with sequencing reactions, generated from suitable PCR-generated DNA templates using the same 5Ј-end-labeled primers, run in parallel. The Circumvent thermal cycle dideoxy DNA sequencing kit (New England Biolabs) was used according to the manufacturer's protocol.
Purification of Histidine-tagged RsmA-RsmA-His 6 was overexpressed in E. coli DH5␣/pME6078 cells and purified from cell extracts by nickel-affinity chromatography following manufacturer's instructions (Qiagen Inc.). The RsmA eluates from the nickel-nitrilotriacetic acid column were dialyzed overnight against 10 mM Tris acetate (pH 8.0) at 4°C, and stored at Ϫ20°C. Protein content was estimated by the Bradford method with bovine serum albumin as the standard. The purity of the preparations was Ն90% as judged from SDS-Tricinepolyacrylamide electrophoresis (data not shown).
Preparation of Transcripts for Gel Shift Assays-For in vitro T7driven transcription of RsmY mutants, PCR copies were generated from pME6947 to pME6951 (Table I) with oligonucleotides FRSMYH and RRSMYP, and cloned as HindIII/PstI inserts into pTZ19R (Fermentas).
Radioactively labeled transcripts of RsmY and mutant GGA-5 were synthesized with a T7 transcription kit (Fermentas) following the manufacturer's instructions in the presence of [␣-33 P]UTP and pME6919 or pME6935 linearized with PstI. Unlabeled competitor RNAs (i.e. RsmY, RsmZ, carA leader, or RsmY mutants) were synthesized following the same protocol but with cold UTP from linearized pME6919, pME6920, pME6926, or pME6933 to pME6937, respectively. RNA was purified by phenol extraction and desalted with Sephadex G-25 minicolumns (Amersham Biosciences). RNA concentration was estimated by UV absorption at 260 nm.
Electrophoretic Mobility Shift Assay-Binding reactions contained [␣-33 P]UTP-labeled RsmY or GGA-5 RNAs, and purified RsmA-His 6 at various concentrations (17). Some assays were also carried out in the presence of various unlabeled RNA competitors (see figure legends for details). In this case, RsmA-His 6 was added last to the binding reaction containing competing RNAs. Reaction mixtures (10 l) were incubated at 30°C for 30 min to allow complex formation. Samples were fractionated on native 10% polyacrylamide gels (4 h at 10 mA), and radioactive bands were visualized by autoradiography after drying the gels.
HCN Production and Exoprotease Activity-HCN production was measured colorimetrically in samples from oxygen limited cultures as previously described (39). Exoprotease activity was determined in cellfree supernatants from overnight cultures (adapted from Ref. 40). Briefly, 75 l of 0.22-m filtered culture supernatant was incubated with 125 l of azocasein solution (2% w/v in 50 mM Tris-HCl, pH 7.5) at 30°C for 20 -24 h. Excess azocasein was precipitated with 0.6 ml of 10% cold trichloroacetic acid, and 0.6 ml of supernatant was mixed with 0.7 ml of 1 M NaOH before measuring absorption at 440 nm. One exoprotease activity unit was defined as 1 exoprotease activity unit ϭ (A 440 ϫ 1000)/(incubation time (h) ϫ sample volume (ml) ϫ A 600 ).
Swimming Motility-Spreading of E. coli strains by swimming motility was assessed on semi-solid agar plates (0.5% (w/v) yeast extract, 2.5% (w/v) nutrient broth, 0.3% (w/v) agar). Freshly isolated colonies were spotted onto swimming plates with a toothpick and incubated overnight in sealed plastic bags at 30°C.
Glycogen Assay-Glycogen content of E. coli strains was estimated by anthrone colorimetry (41). 1 ml of culture was centrifuged, and cells  were washed in 0.9% (w/v) NaCl. 100 l of 1 M NaOH was added, and the cell suspension was incubated 30 min at 55°C to promote cell lysis. Lysates were neutralized with 100 l of 1 M HCl. The content of hexoses was determined by mixing an aliquot of the lysates with anthrone reagent (Sigma; 0.2% (w/v) in 98% sulfuric acid) and incubating in a boiling water bath for 10 min. After cooling the tubes on ice, the absorption was measured at 620 nm. The standard curve was prepared with glucose (1 mg of glucose equals to 0.9 mg of glycogen). The protein content of cell lysates was determined with the Bradford method. Glycogen content was expressed as milligrams of glycogen/mg of protein.

RESULTS
Overexpression of rsmY or rsmZ in E. coli Titrates CsrA-In E. coli, glycogen metabolism and motility are regulated by the CsrA/CsrB/CsrC system (8). A csrA mutation leads to enhanced accumulation of glycogen and reduced swimming motility (8,11) (Fig. 1). Similar effects are found when csrB or csrC are overexpressed, due to titration of CsrA (11,12). In the heterologous host E. coli, overexpression of rsmY or rsmZ from P. fluorescens strain CHA0 also resulted in increased glycogen content (Fig. 1A) and in reduced surface motility on semi-solid agar plates (Fig. 1B). These results indicate that RsmY and RsmZ can titrate CsrA in E. coli despite an almost complete lack of sequence similarity to CsrB and CsrC, suggesting that the one shared feature, the repeated GGA sequence motifs, may be the key elements responsible for regulatory activity.
RsmY Secondary Structure-Folding algorithms predict RsmY to adopt a structure that contains four stem loops exposing, entirely or partially, GGA trinucleotide motifs in singlestranded portions (nt 15-17, 36 -38, 56 -58, and 82-84) of the hairpins, as well as a terminator stem loop (Fig. 2B). Two other GGA elements appear outside hairpins in predicted unpaired regions (nt 3-5 and 45-47) (Fig. 2B). This putative structural pattern is similar to that of other Csr/Rsm riboregulators, but the number of stem loops and exposed GGA elements varies. We performed in vivo cleavage of RsmY RNA with Pb 2ϩ (37) to probe its secondary structure. The cleavage pattern obtained with three different primers essentially confirms the predicted secondary structure of RsmY (Fig. 2). Stem loops 1 and 3 seem to be stable enough to protect RsmY from attack by Pb 2ϩ in vivo ( Fig. 2A), whereas the stems of the predicted stem loops 2 and 4 were partially susceptible to cleavage. These regions of the molecule may be more flexible than are stem loops 1 and 3. We also noted that the GGA sequence in the predicted loop 2 was not cleaved, which might reflect protection by RsmA (see below).

Construction of RsmY Variants and Strategy to Test
Their Activity-We constructed three types of RsmY mutants. Nucleotide substitutions were selected such as to avoid a drastic alteration of the predicted secondary structure in the resulting RsmY variants. To test the functional significance of GGA motifs for RsmY activity, 3 and 5 GGA elements were mutated in the GGA-3 and GGA-5 mutants, respectively (Fig. 3A). When the structure of the GGA-3 mutant was probed in vivo, the overall cleavages remained almost unchanged (Fig. 2C). Stem loops 1 and 3 were still present and up to three remaining unpaired GGA elements may be exposed for interaction with RsmA (nt 3-5, 56 -58, and 82-84) (Fig. 2, C and D). We noted that the base changes in the GGA motif in stem loop 2 resulted in accessibility to Pb(II) in the loop as well as the entire surrounding region of the RNA (Fig. 2, C and D). This indicates that, in wild type RsmY, this RNA motif was indeed protected by RsmA protein binding.
Two additional mutations were based on the observation that sequences in RsmY display complementarity to sequences near the translation initiation sites of two known target mRNAs, hcnA and aprA. We thus considered that base pairing to these target mRNAs (Fig. 3B) might help RsmY to release RsmA-mediated translational regulation; selected nucleotides were replaced to produce the variants hcn and apr, which would have reduced complementarity compared with wild type RsmY (Fig. 3A). Finally, a control mutant (ctr) was constructed (Fig. 3A) by changing nucleotides in a part of the molecule that should neither affect the secondary structure exposing GGA motifs nor the putative antisense domains.
The five different rsmY mutant alleles, all of which are expressed from the natural rsmY promoter, were transferred as single copies into the chromosome of P. fluorescens CHA0 derivatives lacking the rsmY and rsmZ genes (see "Experimental Procedures"). As a positive control, the wild type rsmY allele was also introduced into the rsmY rsmZ double mutant. As the expression of the hcnA and aprA genes is very low in an rsmY rsmZ background, but high in an rsmZ mutant (17) and exoprotease activity to near wild type levels (Table II). However, complementation with RsmY mutants lacking three or five GGA motifs resulted in exoproduct levels below 20% of those harboring wild type RsmY or the RsmY ctr mutant (Table  II). In particular, the strain carrying the GGA-5 variant behaved almost like the negative control (Table II). These observations were correlated with the expression of the chromosomal hcnAЈ-and aprAЈ-ЈlacZ fusions in strains complemented with the RsmY mutants (Fig. 4). These results demonstrate the requirement of the GGA elements for a full RsmY activity.
Activity of Antisense Mutants-As the two putative antisense domains within RsmY were located in different regions of the molecule (Fig. 3B), mutations affecting the presumed interaction with one of the targets, the hcnA mRNA leader, should not affect the interaction with the other one, the aprA mRNA leader. Complementation with either the hcn or the apr variant had comparable effects on the production of HCN and exoprotease (Table II) as well as on the expression of hcnAЈ-and aprAЈ-ЈlacZ fusions (Fig. 4): both hcn and apr mutants restored partially, and to a similar extent, these phenotypes. These results fail to support an antisense-related mechanism for RsmY-mediated target gene regulation but suggest that regions outside the GGA motifs can also have an impact on RsmY activity, probably because they contribute to the stability of the RNA (see below).
Differential interaction of RsmZ and the RsmY variants with RsmA-His 6 was studied in competition experiments. Unlabeled RsmY and RsmZ RNAs specifically competed with labeled RsmY for binding to RsmA-His 6 , whereas an unspecific competitor, i.e. the leader of carA mRNA, did not (Fig. 6A). Similarly, the GGA-3, hcn, apr, and ctr variants still interacted with RsmA-His 6 , whereas GGA-5 was severely impaired (Fig. 6, B and C). Collectively, these experiments show that the loss of five GGA elements results in the most drastic reduction in the ability of RsmY RNA to interact with RsmA-His 6 .
RsmY RNA Stability-The reduced biological activity of the RsmY variants observed in complementation experiments (Table II and Fig. 4) could result from reduced RsmA binding or from a shorter half-life of the RsmY variants or from both. We sampled cell cultures at a time point at which maximal phenotypic differences were observed, and analyzed the steady-state RNA level of each RsmY mutant by Northern blots. As shown in Fig. 7, the RNA levels roughly paralleled the extent of phenotypic restoration (Table II). We investigated the stability of RsmY, wild type and mutant RNAs, by determining their decay rates after blocking transcription with rifampicin. In the wild type strain CHA0, the RsmY half-life was estimated to be 27 min, a value similar to that of the RsmY ctr mutant (Fig. 8). GGA-3, whose concentration was markedly reduced in vivo (Fig. 7), was degraded at a rate corresponding to a half-life of 6 min, whereas the RsmY hcn mutant RNA was degraded with a half-life of about 17 min (Fig. 8). In the case of the GGA-5 mutant RNA, decay rates could not be measured due to low abundance of this RNA in steady state, resulting in insignificant signals on Northern blots (Fig. 7). DISCUSSION In this work we studied the structure and function of the riboregulator RsmY of P. fluorescens, as an example of regulatory RNAs belonging to the global Csr/Rsm regulatory systems of diverse bacteria. We provide evidence that RsmY activity strongly depends on repetitive unpaired GGA trinucleotides.
Overexpression of P. fluorescens rsmY, or its analog rsmZ, in E. coli caused an alteration of glycogen accumulation and swimming motility that reproduced the phenotypes observed when either csrA is inactivated or the riboregulator genes csrB or csrC are overexpressed (11,12) (Fig. 1). This suggests that RsmY and RsmZ on the one hand and CsrB and CsrC on the other hand use common mechanisms for biological activity, and that, despite general sequence dissimilarity, they can functionally replace each other.
The secondary structure of the Rsm/Csr type riboregulators may provide the basis for a common biological function, although they are divergent at the nucleotide sequence level. These RNAs bind several molecules of their protein partners (i.e. the RsmA/CsrA proteins) and have similar predicted secondary structures (8,12,17,31) (Fig. 2). We probed the structure of RsmY RNA in vivo to test the validity of the secondary structure prediction (Fig. 2). From the cleavage pattern, it seems that the GGA motif of stem loop 2 is more protected from Pb(II) attack than are the remaining GGA motifs (Fig. 2, A and  B). This may be a consequence of a high RsmY:RsmA molar ratio in stationary phase, where most of the RNA molecules should be unbound, and protection of the site(s) of highest affinity for RsmA might be favored.
RsmY, as other members of this group of riboregulators, have hairpins that expose the trinucleotide GGA in 4 loops (Fig. 2). Based on this structure, we mutated some of these repetitive elements (Fig. 3) and observed that loss of GGA motifs entailed progressively reduced biological activity (Table  II and Fig. 4). In particular, the RsmY mutant GGA-5, lacking five of these motifs, displayed a dramatic decrease in binding affinity to the RsmA protein in vitro (Figs. 5 and 6). In vivo, a comparison of wild type RsmY and GGA-3 RNAs indicates that protection of at least one GGA motif (located in stem loop 2) is lost in the mutant RNA (Fig. 2). The analysis is not conclusive when the other GGA motifs are concerned. Partial protection may occur, but different affinities of the protein to different sites could render binding difficult to detect.
As the Gac/Rsm system of P. fluorescens CHA0 possesses at least two riboregulators, RsmY and RsmZ (16,17), we explored whether such redundancy might reflect yet another, antisensebased, mechanism of regulation of known target RNAs. Based on the presence of complementarity regions to two known targets (e.g. hcnA and aprA), the corresponding putative antisense sequences of RsmY were mutated separately (Fig. 3). The resulting RsmY RNA variants, hcn and apr, were less active than was wild type RsmY, but failed to show differential activation of either hcnAЈ-or aprAЈ-ЈlacZ fusions (Table II and Fig. 4). Thus, the observed complementarity might be fortuitous and the reduced activity of the variants appears to result from reduced stability (Fig. 8).
The interaction of RsmY with its protein partner appears to influence stability of the riboregulator and hence affects its intracellular concentrations. The level of biological activity displayed by each of the RsmY mutants paralleled the amount of RNA detected in each strain (Fig. 7). The GGA-3 RNA had a FIG. 6. Competition of RsmY with RsmY mutants for binding to RsmA. Labeled RsmY (20 nM) and different cold RNA competitors were incubated with RsmA-His 6 (550 nM) before fractionation in native gels and autoradiography. A, competition of RsmY-RsmA binding by specific (RsmY and RsmZ) and unspecific competitors (leader of carA mRNA (17)). B, competition of RsmY-RsmA binding by RsmY mutants GGA-3, GGA-5 and ctr. C, competition of RsmY-RsmA binding by RsmY mutants hcn, apr, and ctr. FIG. 7. Northern blot analysis of RsmY mutants in stationary phase. RNA was extracted from the same cultures that were used to assay exoprotease activity (Table II), grown at 30°C in NYB to A 600 values of 4.2-5.2. 500 ng of RNA (as estimated by UV absorption) were loaded. The lower panel (loading control) shows the 23 S and 16 S rRNA bands obtained from the same samples fractionated in denaturing agarose gels and stained with ethidium bromide. The amount of each RsmY variant RNA relative to wild type RsmY (%; after correction with loaded RNA) is indicated below the panel.

FIG. 8. Stability of RsmY and RsmY mutant RNAs.
A, decay of RsmY, hcn, GGA-3, and ctr RNAs after blocking transcription with rifampicin as determined by Northern blotting. The amount of RNA loaded was 500 ng for strain CHA0, 1 g for hcn, 2 g for GGA-3, and 500 ng for ctr. B, RNA decay determined by densitometry and correction for RNA loading. E, RsmY; f, hcn; ࡗ, GGA-3; and q, ctr. significantly shorter half-life than had the hcn and ctr mutant RNAs (Fig. 8), although GGA-3 conserved some of the important structural features of RsmY, especially the stem loops 1 and 3 (Fig. 2, C and D), as well as its ability to bind RsmA in vitro (Fig. 6B). Results were inconclusive for the GGA-5 RNA half-life, because its concentration was below the detection limit in steady state (Fig. 7). Given that the same rsmY promoter drives the expression of all RNAs in this study, we infer that very low steady-state levels of GGA-5 reflect a half-life, which is significantly shorter than that of GGA-3. Why are these GGA-less mutant RNAs degraded so rapidly? It is likely that the impaired interaction of these RNAs with the RsmA protein results in a faster turnover, due to lack of protection. Such a phenomenon has been described in E. coli for the Hfq protein, which stabilizes spot 42 RNA, an antisense riboregulator involved in regulation of the gal operon (6). In Erwinia carotovora subspecies carotovora, the absence of the RsmA protein also results in a marked reduction of RsmB RNA halflife (46), whereas the E. coli homolog CsrA fails to show a comparable effect on the stability of CsrB RNA (47). In P. fluorescens strain CHA0, which encodes a second RsmA-like protein (RsmE) (13), the situation, may even be more complex. Preliminary experiments with an rsmA rsmE double mutant indicate that wild type RsmY and RsmZ RNAs are very unstable in this genetic background, suggesting that interaction of these riboregulators with RsmA and RsmE confers protection against cellular RNases.
What are the characteristics of RsmA/CsrA recognition sites in target mRNAs? An alignment of the known CsrA binding sites on glgC and cstA mRNAs of E. coli results in a minimal consensus YANGGANR (10). This resembles the minimal sequence element CANGGANG that seems to be recognized by RsmA in the proximity of the ribosome binding site of hcnA and aprA mRNAs of P. fluorescens (15). Both consensus recognition sequences contain a GGA element, i.e. the binding site for the RsmA/CsrA proteins proposed by the present study. However, target mRNAs, unlike the Rsm/Csr riboregulators, do not appear to contain clustered GGA motifs and therefore additional elements might be involved in the recognition of target mRNAs.