Genome-wide search reveals a novel GacA-regulated small RNA in Pseudomonas species

Prediction and screening of sRNAs in P. aeruginosa PAO1

The majority of the bacterial sRNA genes found to date is located in IgRs [3, 21] and the size of the sRNAs generally varies between 50 and 400 nt [1, 22]. Accordingly, we based our search for sRNAs in P. aeruginosa PAO1 on the criterion that sRNA genes should be located in IgRs that are larger than 50 bp. A total of 3168 IgRs fulfilled this condition. We then assumed that functional sRNA sequences should be conserved in closely related species [4, 20, 23] and we used BLASTn to search for sequences that are homologous to the 3168 IgRs of P. aeruginosa in the genomic sequences of five related pseudomonads (Pseudomonas putida KT2440, Pseudomonas syringae pv. tomato DC3000, Pseudomonas fluorescens Pf0-1, P. fluorescens SBW25 and P. fluorescens Pf-5). Only those PAO1 IgRs were selected that shared more than 65% sequence identity with a genomic sequence of at least one other pseudomonad. The alignments resulting from the BLAST program were then subjected to QRNA analysis [5]. In pairwise alignments of homologous sequences, this algorithm evaluates pairs of nucleotide substitutions and calculates their probabilities of having structural or codon usage conservation and discriminates sequence pairs without a sufficient number of substitutions. RNAs having substitutions with a bias towards codon usage are likely to code for polypeptides whereas pairs of substitutions that maintain stem-loop structures have a higher probability to reflect secondary structure conservation of non-coding RNAs. Although the sequences subjected to BLASTn analysis consisted only of intergenic, non-coding regions according to the P. aeruginosa PAO1 genome annotations, sequences whose coding probabilities calculated by QRNA were found to be high were nevertheless retained for further analysis. By applying QRNA in this way, 162 out of the initial 3168 IgRs from P. aeruginosa PAO1 were retained (see Table in Additional file 1). Among these 162 IgRs, 32 were found to contain exclusively tRNA or rRNA genes and were not considered further. In the 130 remaining regions, the presence of rho-independent terminators was assessed using TransTerm [24] set at a confidence cut-off of 93% (i.e., expecting 93% of all predictions to be correct).

Experimental observation of sRNAs in P. aeruginosa PAO1

IgRs occurring in P. aeruginosa and at least two other pseudomonads and having recognizable promoter elements (e.g., for the housekeeping sigma factor RpoD or the stress sigma factor RpoS) were preferentially chosen for Northern blot analysis. Cultures of P. aeruginosa PAO1 were grown in nutrient yeast broth (NYB) to exponential or stationary phase and total RNA was extracted. Northern blot analysis was carried out with digoxigenin (DIG)-labeled probes, each covering an entire IgR. Among 49 IgRs thus analyzed, 14 were reproducibly found to express sRNAs (Table 1). Figure 1 shows all transcripts revealed, except for four sRNAs that had been described previously: 4.5S RNA (encoded by the ffs gene in IgR 888) [2, 25], an RNase P component found in many bacteria (encoded by rnpB in IgR 2510) [2, 20], and PrrF1 and PrrF2 found in P. aeruginosa (encoded by prrF1 and prrF2 in IgR 2667) [19]. They were considered to validate the method and not investigated further. Three additional sRNAs (encoded by IgRs 491, 1698, 1887; Table 1) are among the 17 P. aeruginosa sRNAs described by Livny et al. [20]. In that study, the three sRNAs are designated P5, P16 and P20, respectively. In conclusion, we detected eight new sRNAs encoded by the remaining IgRs.

Intergenic region a	Coordinates in PAO1 genome (bp)	5' gene		3' gene		sRNA
		Length (bp)	Dir.b	Length (bp)	Dir.b	Predicted length (nt)c	Observed length (nt)d	Dir.g
491	912780–913085	1185	<	486	>	90	100, 200	>
622	1140860–1141267	1146	>	465	>		200, 300	?
645	1204782–1205770	1149	>	1767	>	300 ± 10	300	<
888 = ffs	1668833–1669085	378	>	867	<	113	90	>
1059	1996807–1997508	1659	<	1041	<		150, 200, 300	?
1466	2918212–2918965	1041	>	534	<		300	?
1559	3112151–3112876	519	>	900	>		250, 300e	?
1698 = rgsA	3318657–3318881	1134	>	777	<	120 ± 10	100, 300	>
1714	3360654–3360873	1386	<	3474	>		200	?
1887	3705161–3705888	1995	<	603	<	210 ± 10	200	<
2315	4536493–4536919	660	<	1122	<	160 ± 10	180	<
2510 = rnpB	4956029–4956732	452	<	849	<	208	200	<
2626	5196833–5197184	1464	>	576	>		200	>
2667 = prrF1,2	5283906–5284368	798	>	888	<	110, 110	100, 110 f	>>

^a Transcripts that were also found by Livny et al. [20] are shown in boldface

^b Dir., direction of flanking genes

^c Lengths of transcripts described in other studies are indicated by their experimental values; otherwise, predicted lengths are estimates ± 10 nt

^d Bands that were observed in cells grown in NYB, GGP or MME medium; estimated size ± 10 nt

^e Not tested in GGP and MME medium

^f No transcript was observed in cells grown in NYB or GGP in exponential phase; transcripts were present in stationary phase cells grown in all media

^g Dir., direction of sRNA genes. When rho-independent terminators are predicted to be present, the direction of the sRNA is annotated (> or <). When the direction is unknown, this is indicated by a question mark

All transcripts were also detected when cells had been cultivated in the iron-limited medium GGP or in the minimal medium MME (data not shown), except for 1559 sRNA, which was not tested in these alternative media. Six regions (IgRs 491, 622, 1059, 1559, 1698, 2315) produced more than one transcript (Figure 1); we only indicate the most prevalent bands in this figure. Multiple transcripts can be due to processed sRNAs, 5' leader mRNA sequences or 3' mRNA fragments related to flanking genes [26]. Three other regions (IgRs 1714, 1887, 2626) each revealed a major band, together with possible degradation products. Whenever the sRNA transcript length could be predicted with reasonable certainty (Table 1), at least one of the main transcripts observed had the expected size (Figure 1). IgR 491 specified a transcript that had previously been observed [20]; we deduce from the sequence that it originates from an RpoS-dependent promoter. The 180-nt sRNA encoded by IgR 2315, which is located between the ribC and ribD genes, is a homolog of sroG, a transcript resulting from the cleavage of a riboswitch element found upstream of ribB in E. coli [27]. The P. aeruginosa sroG homolog was slightly longer than the sroG transcript in E. coli (147 nt) [27]. The 2315 sRNA of P. aeruginosa may be involved in regulation of riboflavin biosynthesis, by analogy with the homologous element in E. coli [28, 29]. In four cases (IgRs 645, 1698, 1887, and 2315), the coordinates of the sRNA genes can be deduced either from recognizable promoter and terminator elements or by sequence comparison with an E. coli RNA (Additional file 2 for the IgRs).

GacA- and RpoS-regulated expression of the 1698 sRNA in P. aeruginosa PAO1 and P. fluorescens CHA0

The expression of the 11 sRNAs shown in Figure 1 was tested in the P. aeruginosa gacA mutant PAO6281 by Northern blotting. The 1698 sRNA was the only transcript that showed a significantly decreased expression in the gacA mutant, compared with that in the wild-type PAO1 (data not shown). This 120-nt transcript was further characterized. It was virtually absent from strain PAO1 during the exponential phase, but was produced abundantly in the stationary phase. In the gacA mutant, the expression of the 1698 sRNA was about two-fold lower than in the wild-type (Figure 2a), suggesting positive regulation by GacA. Moreover, in a mutant lacking the stress/stationary phase sigma factor RpoS, PAO1-rpoS [30], no expression of this RNA was detected (Figure 2a). We therefore named this sRNA RgsA (for r egulation by G acA and s tress).

The RgsA sRNA was also observed in P. fluorescens CHA0 (Figure 2b), a strain closely related to P. fluorescens Pf-5, whose genome has been sequenced. There is 99% identity between the two P. fluorescens strains for the IgR 1698 (Figure 3a). In strain CHA0, the band corresponding to this sRNA was faint in the wild-type CHA0, the gacA mutant CHA89 and the rpoS mutant CHA815 during the exponential phase. In the stationary phase, the transcript was observed clearly in the wild-type, whereas the expression was strongly reduced in the gacA mutant and in the rpoS mutant (Figure 2b), suggesting that again both GacA and RpoS contribute to the regulation of RgsA RNA.

As a control, we included the RsmZ sRNA of strain CHA0 in this Northern blot analysis. Confirming earlier results [9], we see that rsmZ expression depends on positive control by GacA, but is independent of RpoS (Figure 2c).

Location of the rgsA gene in pseudomonads

The rgsA gene is strongly conserved in seven related pseudomonads (Figure 3a). The promoter, the rho-independent terminator and a putative upstream regulating sequence (URS) show extensive sequence similarities. The transcription start site was tentatively deduced from the consensus -10 element (T/G CTATACT) of RpoS-dependent promoters [31]. The presence of this RpoS promoter element is in agreement with our finding that RgsA sRNA was absent from an rpoS mutant of P. aeruginosa and that a reduced amount of this sRNA was present in an rpoS mutant of P. fluorescens(Figures 2a and 2b). In a database search, we did not find any sequence that is highly similar to the entire conserved URS sequence, suggesting that it might be a binding site for a specialized rather than for a global regulator. In particular, the URS motif is not related to the palindromic upstream activating sequence which is located upstream of the rsmX, rsmY and rsmZ genes and which is a potential GacA binding site [11, 16]. We therefore suspect that the GacA effect on RgsA expression is likely to be indirect. The predicted size of RgsA (120 nt) agrees well with the length of the 1698 transcript observed in the initial Northern blots (Figure 1).

The genomic context of the rgsA gene in P. fluorescens Pf-5 and CHA0 is shown in Figure 3b. The flanking ORF PFL_1804 (coding for a regulator of the TetR family), which is located 800 bp upstream, is conserved only in P. putida KT2440. The downstream, divergently transcribed ORF PFL_1803 (a homolog of the E. coli tatD gene) is conserved in all Pseudomonas strains sequenced to date. In E. coli, the tatD product exhibits Mg-dependent DNase activity, but is not required for protein transport by the Tat pathway, contrary to what was originally expected [32].

Regulation of rgsA gene expression in P. aeruginosa PAO1 and P. fluorescens CHA0

To monitor the rgsA regulation by GacA during growth, we constructed rgsA _PAO1 -lacZ and rgsA _CHA0 -lacZ transcriptional fusions in plasmids pME7235 and pME7234, respectively. The primers used for these constructions (Figure 3b) carry a BamHI site, which enabled us to place the lacZ gene under the control of the respective rgsA promoter. At low cell population densities, the wild-type strains PAO1 and CHA0 as well as the gacA mutants PAO6281 and CHA89 exhibited similar basal levels of lacZ expression from the rgsA-lacZ constructs (Figures 4a and 4b). At high cell population densities, the gacA mutants showed an approximately two-fold reduction of β-galactosidase activity, compared with the activity in the wild-type strain. This finding is consistent with the Northern blot results (Figures 2a and 2b). The rgsA _CHA0 -lacZ fusion carried by pME7234 was also assayed in the rpoS mutant CHA815; at OD₆₀₀ ≈ 4.0, this fusion showed a 40% decrease of lacZ expression, compared with that in the wild-type CHA0 (data not shown). In conclusion, transcription of the rgsA gene is probably activated indirectly by GacA and directly by RpoS, which is assumed to bind to the -10 promoter element (Figure 3b), both in P. aeruginosa and in P. fluorescens.

Role of RgsA sRNA in the Gac/Rsm regulatory cascade of P. fluorescens CHA0

A single, typical RsmA-binding motif (ANGGA) was found in an unpaired region of the RgsA sRNA of strain CHA0 (Figure 5). The GacA-controlled sRNAs RsmX, RsmY and RsmZ and their functional homologs each contain several GGA and extended ANGGA motifs [8]. In the case of RsmY, these motifs have been shown to be essential for sequestration of RsmA and its homolog RsmE in P. fluorescens [33]. We therefore investigated whether the RgsA sRNA might also have a titrating role, by testing whether rgsA overexpression would enhance the expression of a translational hcnA'-'lacZ fusion in P. fluorescens CHA0. Expression of hcnA, the first gene of the hcn operon involved in HCN biosynthesis, is tightly controlled by the Gac/Rsm cascade in strain CHA0. In an rsmX rsmY rsmZ triple mutant, artificial overexpression of rsmX rescues hcnA'-'lacZ expression [10]. We measured the expression of a chromosomal hcnA'-'lacZ fusion in the wild-type background and in the rsmXYZ triple mutant where the activity was almost totally lost (Figure 4c). When rgsA _CHA0 was overexpressed in the wild-type strain (CHA207/pME7236), no significant difference was observed, compared with the native context (CHA207 containing the empty vector pME6032). Similarly, the rgsA _CHA0 overexpressing plasmid pME7236 was unable to elevate hcnA'-'lacZ expression in the rsmXYZ triple mutant (CHA1145/pME7236), which was indistinguishable from the strain carrying the vector alone (CHA1145/pME6032) (Figure 4c). In a control experiment, we verified rgsA _CHA0 overexpression by plasmid pME7236. Total RNA was extracted from strains carrying either pME7236 or the vector pME6032 and the amount of RgsA sRNA expressed in the cells was evaluated by Northern blotting. Overexpression of RgsA sRNA was detectable in the wild-type strain CHA0 and even more clearly in the rpoS mutant CHA815 (Figure 4d). In conclusion, the single RsmA-binding motif of RgsA was unable to derepress hcnA expression in P. fluorescens, confirming earlier evidence that multiple GGA motifs are required in sRNAs for effective sequestration of RsmA-like proteins [33].

The Gac/Rsm cascade modulates the response of P. fluorescens to stress imposed by hydrogen peroxide [14]. We constructed an rgsA deletion mutant of strain CHA0 and tested its survival after a 30-min exposure to 40 mM H₂O₂ in NYB with shaking. This experiment was carried out three times with triplicate cultures. Only 69 ± 16% of the mutant cells survived, compared with 94 ± 12% surviving wild type cells. (The initial viable count was set at 100%.) Thus, survival was significantly (P < 0.05) higher in the wild type than in the mutant. Overexpression of rgsA afforded no significant effect. These results suggest that RgsA may contribute to oxidative stress response.

Bacterial strains and growth conditions

We used P. aeruginosa PAO1 (ATCC 15692) and its derivatives PAO1-rpoS [30] and PAO6281 (gacA::Ω-Sm/Sp) [13], P. fluorescens CHA0 [36], CHA89 (gacA::Km^r) [37], CHA207 (hcnA'-'lacZ) [38], CHA815 (ΔrpoS) [9], CHA1145 (rsmXYZ, hcnA'-'lacZ) [10] and CHA1181 (ΔrgsA, this study). E. coli DH5α [39] served for gene cloning. Strains were routinely grown in nutrient yeast broth (NYB; 2.5% [wt/vol] nutrient broth, 0.5% [wt/vol] yeast extract) with shaking, or on nutrient agar (4% [wt/vol] blood agar base, 0.5% [wt/vol] yeast extract). When required, ampicillin at 100 μg ml^-1 or tetracycline at 25 μg ml^-1 (100 μg ml^-1 for pseudomonads) were added. For some RNA extractions, strains were also grown in GGP medium [40] or in MME minimal medium [41]. To monitor β-galactosidase expression qualitatively, 5-bromo-4-chloro-3-indolyl-β-D-galactoside (Xgal) was incorporated into solid media at a final concentration of 0.02%. Routine incubation temperatures were 37°C for E. coli and P. aeruginosa and 30°C for P. fluorescens. Survival of stationary phase P. fluorescens was assayed in NYB following H₂O₂ stress as described previously [14]. Statistical significance was assessed by Student's t-test.

DNA manipulations and cloning procedures

DNA manipulations were carried out according to standard procedures [39]. Plasmid isolation was performed using the cetyl-trimethyl-ammonium bromide method (CTAB) [42] for small preparations, and the JetStar kit (Genomed GmbH) for large preparations. DNA fragments were purified from agarose gels with the MinElute Gel Extraction Kit or the Qiaquick Gel Extraction Kit (Qiagen Inc.). Transformation of the strains was done by electroporation [43]. DNA sequencing was carried out by Microsynth (Microsynth AG). Polymerase chain reaction was carried out as previously described [11].

Plasmid constructions

To generate pME7234 carrying a transcriptional lacZ fusion to the rgsA gene from P. fluorescens CHA0, a 510-bp fragment containing the rgsA sRNA promoter was amplified from chromosomal DNA by PCR using primers 1698Fus1 (EcoRI) and 1698Fus2 (BamHI) listed in Additional file 3, digested by EcoRI and BamHI, and inserted into EcoRI/BamHI-cut pBLS (pBluescript II KS+ cloning vector, ColE1 replicon, Ap^R, Stratagene). After sequencing, the fragment was cloned into the shuttle vector pME6016 [44]. The resulting plasmid pME7234 carries the rgsA _CHA0 promoter controlling the expression of the lacZ gene. Similarly, we constructed pME7235 carrying a transcriptional lacZ fusion to the rgsA gene from P. aeruginosa PAO1 using primers 1698Fus3 (EcoRI) and 1698Fus4 (BamHI) (Additional file 3) and analogous cloning steps.

To overexpress rgsA from P. fluorescens CHA0, plasmid pME7236 was constructed in several steps. First, a 180-bp fragment containing the rgsA gene was amplified from chromosomal DNA using the primers 1698Sur4 (EcoRI, PstI) and 1698Sur2 (HindIII) (Additional file 3), digested with EcoRI and HindIII, and inserted into EcoRI/HindIII-cut pBLS. A 200-bp fragment was then excised from pBLS with EcoRI and KpnI and cloned into pME6032 [9]. The Shine-Dalgarno sequence carried by pME6032 was removed, as it is unsuitable for the expression of a non-coding sRNA gene, by a second round of PCR using the primers 6032K (HpaI) and 1698Sur6 (PstI) (Additional file 3), digestion by HpaI and PstI, and cloning. The resulting construct pME7236 carries the pME6032 tac promoter 13 nt upstream of the rgsA +1 site. The introduction of the PstI restriction site downstream the tac promoter added 3 nt and modified 2 nt of the RgsA sRNA at the 5' end (5'-TCTGCAGGCGGG... instead of 5'-GTCGGCGGG...).

Construction of an rgsA mutant of P. fluorescens

A deletion of the rgsA gene in the P. fluorescens CHA0 chromosome was made as follows. A 860-bp fragment containing the upstream region and the rgsA promoter was amplified by PCR with primers 1698DelE (EcoRI) and 1698Del3 (NcoI) (Additional file 3). A 750-bp fragment containing the rgsA terminator and most of the downstream PFL_1803 gene was amplified by PCR with primers 1698Del4 (NcoI) and 1698DelX (XbaI) (Additional file 3). Both fragments were digested with NcoI and ligated together; in the resulting ~1.6-kb fragment there remained 11 bp between the transcription start site and the terminator of the rgsA gene. The fragment was digested with EcoRI and XbaI and cloned into pME3087 (suicide vector; ColE1 replicon; Tc^r; [36]). E. coli DH5α was transformed with the resulting pME3087 derivative and used in a triparental mating with the helper HB101/pME497 [45]. In this way the deleted rgsA gene (ΔrgsA) was introduced into CHA0, giving CHA1181.

Computational search parameters in the screening for sRNAs

Our screening was limited to the IgRs of the P. aeruginosa PAO1 genome having a length exceeding 50 bp. The corresponding 3168 sequences were listed in FASTA format and compared by BLASTn to the genome sequences of P. putida KT2440 [46], P. syringae pv. tomato DC3000 [47], P. fluorescens SBW25 [48], P. fluorescens Pf0-1) [49], and P. fluorescens Pf-5 [50]. We fixed a cut-off at 65% sequence identity. The PAO1 sequences fulfilling this criterion, together with the homologous sequences of at least another Pseudomonas strain, were given as input data into the program QRNA [5]. Briefly, this program is able to compare each of the PAO1 sequences with a paired sequence, and tests the pattern of substitutions observed in the pairwise alignment of two homologous sequences. Three models of substitution are analyzed. Model (i) assumes that mutations occur in a position-independent manner (called OTH); model (ii) assumes that in homologous coding regions mutations often result in conservative amino acid substitutions (called COD); model (iii) assumes that the pattern of mutations significantly conserves the secondary structure in a homologous RNA (called RNA). For each alignment QRNA establishes scores according to each model and calculates to which category the aligned sequences most probably belong. For IgRs, sequences can be described as COD if after the alignment with their homologous sequences they do not have enough substitutions reflecting secondary structure conservation and present a (probably fortuitous) codon usage bias. Using QRNA we obtained a list of 162 intergenic regions that share coding (COD) or structural homology (RNA) with at least one related genome sequence (see Additional file 1).

Sequence analysis tools

FUZZNUC, a free access nucleic acid pattern search program [35], was used to search for conserved motifs in IgRs, e.g. for RpoD (σ⁷⁰) binding sites (TTGACAN₍₁₇₎TATAAT), RpoN (σ⁵⁴) binding sites (TGGCACN₍₅₎TTGCW, where W is A or T, based on Barrios et al. [51]) or for the -10 element of RpoS (σ³⁸)-binding sites (TGN_(0–2)CCATACT, according to Lacour et al. [52]). We used CLUSTALW [53] for multiple sequence alignments and TransTerm [24] to identify rho-independent terminators in IgRs. The analysis was done using the entire PAO1 chromosomal sequence as the template and the list of PAO1 ORFs (such that the program can distinguish the intragenic from the intergenic parts of the chromosomal sequence); both were obtained from the P. aeruginosa PAO1 sequencing project site [54]. The confidence threshold was set to 93%.

RNA extraction and Northern blots

RNAs used for Northern blot analysis were isolated using a hot acid phenol extraction protocol, based on the technique described by Massé et al. [55]. The original protocol was modified as follows: 5 ml of cell culture was centrifuged and resuspended in 5 ml of TKM buffer (10 mM Tris-HCl, 10 mM KCl, 5 mM MgCl₂, pH 7.5), before mixing with lysis buffer (320 mM Na acetate, 8% [wt/vol] Na dodecylsulfate, 16 mM EDTA, pH 4.6, treated with di-ethyl-pyrocarbonate [DEPC] and autoclaved). DIG-labelled probes were obtained as follows: PCR fragments were synthesized with P. aeruginosa or P. fluorescens genomic DNA as the template and pairs of primers listed in the Additional file 3. For each probe, the first primer was designed to hybridize to the proximal part of the upstream ORF (forward primer, indicated by F) and the second to hybridize to the proximal part of the downstream ORF (reverse primer, indicated by R). The fragments obtained corresponded to the IgRs to be analyzed and were then used as templates for a second PCR, which was carried out with the same pairs of primers as before, but in the presence of DIG DNA labeling mix (DNA labeling Mix, 10 × conc., Roche). RNAs (~5 μg per lane) were separated on a denaturing urea-polyacrylamide gel and analyzed by Northern blotting as previously described [11], with minor modifications: RNAs were transferred by electroblotting (20 min at 150 mA) onto a charged Nylon membrane (Hybond-N⁺, GE-Amersham) and revealed by hybridization with the DIG-labeled probes described above and exposure to a light-sensitive film (Super RX, Fujifilm). 5S rRNA served as loading control. For this purpose, a 5S-rDNA probe was synthesized with primers 5S-rRNA-1 and 5S-rRNA-2 (Additional file 3) and DIG-labeled. The membranes could be used a second time for Northern blot detection of 5S rRNA. To completely detach the probe previously hybridized, membranes were rinsed with DEPC-treated H₂O and immersed twice in 0.2 M NaOH containing 0.1% Na dodecylsulfate for 15 min. The membranes were then rinsed with 2× SSC solution [39] and used in further Northern blots.

β-Galactosidase assays

P. fluorescens cells were grown in 20 ml of NYB, in 100 ml Erlenmeyer flasks. Triton X-100 (0.1% wt/vol) was added to the cultures to avoid cell aggregation. Samples were taken during various growth phases and permeabilized with 5% toluene. β-Galactosidase activities were then measured according to the Miller method [56].