Among the 2920 chromosomal IgRs of S. meliloti 1021 28, a first set of IgRs potentially encoding sRNAs (Table 1) was compiled by: 1) selection of IgRs with annotated orphan transcriptional terminators 28; 2) selection of IgRs in the vicinity of tRNAs 2930; 3) application of sRNAPredictHT (J. Livny; unpublished data), an improved version of the program sRNApredict2 developed by Livny and co-workers 31. Using default parameters, sRNAPredictHT identified 186 sequence elements as putative sRNAs (Additional file 1). However, almost 60% of the hits corresponded to annotated 192832 or non-annotated sequence repeats. Each IgR was used to query Rfam database 33 to identify previously annotated RNA regulatory elements and then inspected for the presence of transcriptional signals (promoters and Rho-independent terminators; see below). We retained 17 chromosomal IgRs that were likely to encode sRNAs and an additional IgR encoding a putative 6S RNA homologue (Table 1).

Sinorhizobium meliloti strain 2011 34 was maintained on TY agar plates 35 with streptomycin (400 μg/ml). We chose the Rhizobium defined medium (RDM) 36 with shaking (120 rpm) at 28°C as the referential growth condition. For preparation of RNA extracts, 125-ml flasks containing 20 ml of RDM were inoculated with 0.2 ml of a saturated RDM pre-culture and incubated at 120 rpm until cell harvest. To introduce stress conditions, the RDM basal medium or growth conditions were modified as follows: high salt RDM (0.3 M NaCl), low phosphate RDM (0.1 mM phosphate, 10 mM MOPS pH 7.0), RDM with ethanol (2% v/v), RDM with SDS (0.1% w/v) and RDM with H₂O₂ (0.1 mM). High temperature stress was applied by growing cells at 37°C. For acid stress, exponential phase cells growing in 20 ml of RDM (OD₆₀₀ = 0.5) were collected by low speed centrifugation, washed with and resuspended in 20 ml of RDM containing 20 mM MES and equilibrated at pH 5.5, and incubated 90 min at 28°C with shaking before harvesting cells for RNA extraction.

Total RNA was extracted immediately after cell harvest by low speed centrifugation (1800 g, 10 min, and 20°C). The cell pellet was resuspended in Trizol^® (Invitrogen; 1.5 ml for cultures with OD₆₀₀ < 1.5 or 3.0 ml for cultures with OD₆₀₀ > 1.5) and treated 1 min at 60°C. Upon addition of 0.2 vols of chloroform and vigorous shaking during 15 secs, the RNA present in the aqueous supernatant was precipitated with 0.5 vol of isopropanol. The pellet was washed in 70% ethanol, air dried and resuspended in 20 μl of DEPC-treated deionized water. RNA samples were conserved at -130°C. The purity and integrity of RNA preparations were assessed by denaturing PAGE electrophoresis followed by silver staining 37 and the RNA concentration was estimated by UV spectrometry 38. For Northern blots, 1–3 μg RNA present in 1 μl of each sample were fractionated on denaturing polyacrylamide gels (60 × 80 × 0.75 mm containing 8.3 M urea, 8% acrylamide and 0.2% bisacrylamide in 1× TBE buffer). The lane corresponding to the molecular weight markers (low range RNA ladder; Fermentas) was cut out, stained with 5 μg ml^-1 ethidium bromide and photographed under UV light. The rest of the gel was electroblotted at 150 mA (15–25 V) onto a Hybond-N membrane in 1× TBE buffer for 20 min. Membranes were washed with 2× SSC (0.3 M NaCl and 30 mM sodium citrate) before nucleic acids were cross-linked by exposure to UV light for 5 min 38. Northern hybridizations were done with digoxigenin (DIG)-labeled DNA probes generated by PCR covering entirely or partially each IgR (Additional file 2). The IgR amplicons of detected candidate sRNAs were cloned in the pCR^®2.1-TOPO vector and sequenced to confirm the identity of the PCR products. Hybridized membranes were developed following the protocol recommended by the manufacturer (Roche Diagnostics GmbH). The detected RNA bands were quantified by densitometry with ImageJ v1.38 39 and standardized by the amount of loaded RNA visualized by silver staining.

Pre-cultures of S. meliloti strain 2011 were grown at 30°C in TY 35 or GMS 40 media. For RNA isolation, 100 ml flasks with 50 ml TY or GMS medium, supplemented with 8 μg/ml nalidixic acid, were inoculated with 200 μl of pre-culture and incubated in a rotary shaker (175 rpm) at 30°C to an OD₆₀₀ = 0.6. To induce stress, the medium and growth conditions were modified as follows. High salt stress: addition of NaCl to a final concentration of 0.4 M in GMS medium. Oxidative stress: addition of H₂O₂ to a final concentration of 10 mM in GMS medium. Cold shock stress: temperature shift of the culture in TY medium from 30°C to 20°C. Heat shock stress: temperature shift of the culture in TY medium from 30°C to 40°C. Acid or alkaline stress: cultures grown in GMS to an OD₆₀₀ = 0.6 were centrifuged and then re-suspended in GMS modified by adding HCl to pH 5.8, or by adding NaOH to pH 8.5. In all cases, cells were harvested 15 and 45 min after exposure to stress conditions.

RNA was isolated and separated into small RNA (< 200 nt) and long RNA (> 200 nt) fractions using the miRNeasy Mini Kit (Qiagen) or Ambion mirVana miRNA Isolation Kit (Ambion) according to the manufacturers' instructions. Quality of RNA was analyzed applying the Agilent RNA 6000 Pico Kit on the Agilent 2100 Bioanalyzer (Agilent Technologies). To consider both orientations, aliquots from the same fractions of small and long RNA pools were sense labelled using the mirVana miRNA Labeling Kit (Ambion) and antisense labelled as described 41. Differing from the cDNA labelling procedure 41, small RNA fractions were first tailed with PolyA polymerase (Ambion). Oligo dT and amino-allyl random hexamer primers were used for the synthesis of cDNA.

Hybridization of the small RNA fraction (Cy3-fluorescent marker) was compared to that of the long RNA fraction (Cy5-fluorescent marker). Three combinations were performed: 1. the small RNA fraction with the long RNA fraction, both of which were sense labelled, 2. the same fractions in which both were antisense labelled, and 3. a combination of the sense labelled small RNA fraction and the antisense labelled long RNA fraction. Slide processing, sample hybridization, and scanning procedures were performed as described 41 applying the Sm14kOLI microarray that carries 50 mer to 70 mer oligonucleotide probed directed against coding regions and intergenic regions 42. Analysis of microarray images was carried out applying the ImaGene 6.0 software (BioDiscoveries) 41. Lowess normalization and significance test (fdr) were performed with the EMMA software 43. The M-Value represents the logarithmic ratio between both channels. The A-Value represents the logarithm of the combined intensities of both channels. Positive M-values ≥ 2.5 represent an enrichment of small RNA fragments (< = 200 nt) and therefore were classified as sRNA candidates.

From the original 2920 chromosomal IgRs, all the annotated repetitive elements of 1021 chromosome (Sm-repeats, RIMEs and AB, C motifs) 192832 were removed and the flanking IgR segments were treated as new IgRs. 1720 chromosomal IgRs free of annotated repeats and longer than 150 nt were retained for further analysis. Certain IgRs were also removed if they gave BlastN hits with E-value < 10^-3 when queried against themselves, reducing the number of IgRs to 778. With the help of open source algorithms and web based tools, the 778 chromosomal IgRs were subjected to the following sequence analyses: prediction of Rho-independent transcription terminators and of promoter signals, sequence conservation (BlastN; 44) and secondary structure conservation (QRNA analysis) 45.

For prediction of Rho-independent transcription terminators, the web based TranstermHP server 46 was queried to generate a list of putative terminator sequences in chromosomal IgRs of strain 1021, having a stem length of 4–23 bases, a hairpin score ≤ -1.5, a tail score ≤ -2.0 and ≥ 80% of confidence. Orphan terminators (i.e., those that do not correspond to flanking CDS) were scored 3. Predicted terminators co-oriented with flanking ORFs were scored according to their relative distance to the 3'-end of the corresponding annotated gene so that a score of 2 was assigned if the terminator was farther than 200 bp, 1 if the distance was 100–200 nt, and 0 if it was closer than 100 bp.

Promoter signals were predicted with three alternative methods. A first set of putative promoters was generated with a web based neural network based routine 47 set up for bacterial sequences in both DNA strands with a minimum score of 0.8. A second set of putative promoters was compiled by querying IgRs with Fuzznuc 48 using available S. meliloti consensus sequences as input. For σ⁷⁰-dependent promoters the query was CTTGAC(N₁₇)CTATAT 49 with up to 4 mismatches allowed. For σ⁵⁴-dependent promoters the query was TGGCACG(N₄)TTGCW 50 with up to 2 mismatches allowed. For putative PhoB-binding sites the results of two queries were pooled, CTGTCAT(N₄)CTGTCAT 51 with up to 4 mismatches allowed and TGWCAM(N₄)CYKTCAK 52 with up to 2 mismatches allowed. A third group of promoters was predicted with the help of the matrix-scan tool available at the Rsat web server 53, upon introduction of available scoring matrices for S. meliloti σ⁷⁰-, σ⁵⁴- and PhoB-dependent promoters 49505152 and with default parameters. A similar scoring criterion to that used for terminators was applied to predicted promoters. Orphan promoters were scored 3. Putative promoters were rated 2 if the 5'-end of the co-oriented flanking CDS was farther than 300 bp, 1 if this distance was 200–300 bp and 0, if they were closer than 200 bp.

Similarity searches performed with BlastN were done using default parameter values. IgRs were used to query against a database of 559 complete eubacterial genomes 54 and we defined a Blast score (#BlastN) that for each input IgR sequence consists in the sum of all the hits with E-values below 10^-3. We used QRNA 45 to analyze the sequence alignments generated for each IgR and a score was derived summing all the positive hits detected (#QRNA).

Finally, the individual scores for predicted terminators (#T), promoters (#P), BlastN (#BlastN) and QRNA analysis (#QRNA) were combined to generate a Global Score (GS). If a putative promoter and a terminator lay co-oriented and separated from each other by 40–500 bp, suggesting the presence of a single and independent transcriptional unit, the IgR is scored 10 and the individual scores for promoter and terminator are no longer considered. The GS for those IgRs containing such putative elements indicative of sRNAs was calculated as (10 + #BlastN + #QRNA). For those IgRs lacking putative independent transcriptional units, the GS was calculated as (#T + #P + #BlastN + #QRNA).

At the time we initiated this study, the only chromosomal non-coding RNA gene that had been characterized in the α-proteobacterium S. meliloti was the tmRNA homolog ssrA 23. However, several findings suggested that other sRNAs might be expressed in this α-proteobacterium. The electrophoretic fractionation in denaturing polyacrylamide gels of total RNA from strain 2011 cells grown under different conditions (Additional file 3) revealed several RNA bands of < 300 nt other than the conserved and abundant 5S RNA, 4.5S RNA and tRNAs 27. Another indirect evidence of the existence of sRNAs in S. meliloti comes from the pleiotropic phenotype of the S. meliloti 2011hfq mutant (Sobrero & Valverde, unpublished). These observations suggest that the product of the hfq gene (SMc01048 = nrfA) may be required to assist diverse regulatory interactions between mRNAs and sRNAs, as reported for other bacterial species 755. We thus decided to perform a bioinformatic search of sRNA genes using the genomic information of the sequenced strain S. meliloti 1021.

Although there are reports of sRNAs transcribed from coding regions in other bacteria 5657, we focused our search in the regions between annotated ORFs (hereafter IgRs) of the S. meliloti chromosome 19. We first identified in the S. meliloti annotated database 28 chromosomal IgRs containing transcriptional terminators unlikely to be associated with flanking ORFs as well as regions of sequence conservation in the vicinity of annotated tRNA genes, which may represent horizontally transferred genetic elements 2930. This "manual" procedure resulted in the identification of a few interesting IgRs (tagged OT and tR in Table 1). Next, we applied sRNAPredictHT, an improved version of the systematic and integrative tool sRNApredict2 already used for the prediction of sRNA genes in several bacterial species 31. sRNAPredictHT identifies sRNA-encoding loci based on the co-localization of transcriptional terminators and IgR sequence conservation 31. Among the 186 candidate loci identified by sRNAPredictHT (Additional file 1), 56% were identified in IgRs containing at least one repetitive DNA element, either the annotated Rhizobium-specific intergenic mosaic elements (RIMEs) 1932, Sm-repeats 1928, AB, C palindromes 1928, or in some cases, even non-annotated repeats. Rhizobial genomes are characterized for the presence of dozens of these intergenic sequences of unknown function that typically share significant primary sequence and secondary structure conservation 58. Upon elimination of IgRs containing repeats, the sRNAPredictHT output was narrowed down to a list of 76 candidate IgRs (Additional file 4). To further reduce the number of IgRs for experimental verification, we looked for candidates associated with putative promoters. This stringent filtering yielded a list of 17 interesting IgRs (Table 1). In fact, 15 candidate IgRs have both potential 5' and 3' transcriptional signals and are conserved in related species (Table 1), suggesting that they correspond to bona fide sRNA-encoding genes. Table 1 also includes a putative homolog of the widely conserved 6S RNA (IgR#1; 33) which was not picked up by sRNAPredictHT because it lacks a typical Rho-independent terminator (Table 1). With the exception of IgR#5, all the candidates in Table 1 are conserved in at least one related α-proteobacterium. All IgRs but the aforementioned IgR#1 (6S RNA) are associated with a predicted Rho-independent terminator.

For experimental verification of most putative sRNA genes listed in Table 1, we performed Northern hybridizations and microarray analysis of RNA from S. meliloti strain 2011 that, like the sequenced strain 1021, is a streptomycin-resistant mutant derived from the isolate SU47. Although the separate and parallel continuous manipulation of these isogenic strains gave origin to subtle differences in their symbiotic behaviour and gene expression 5259, the overall high degree of sequence similarity between both strains permits the use of strain 2011 to test predictions based on 1021 sequence. As many characterized sRNAs are involved in regulatory processes induced by a variety of external stimuli 9, RNA extracts were prepared from cells grown both under standard culture conditions and under a variety of stressful conditions.

Of the 12 candidate IgRs from our initial compilation that were subjected to experimental verification by Northern analysis of S. meliloti 2011 RNA, 11 were detected (Table 1, Figure 1, Additional file 5). For the majority, the transcript size was consistent with our predictions (Table 1). In some cases (e.g., IgR#10, IgR#11 or IgR#13), multiple bands were observed. Two IgRs (#4 and #14) revealed a complex banding pattern (Table 1; Additional file 5) and further experiments are required to elucidate the origin of the detected RNA bands. Microarray analysis of strain 2011 RNA detected enrichment of RNA molecules < 200 nt corresponding to the predicted DNA regions for IgR#1, IgR#3, IgR#6, IgR#7, IgR#12, IgR#14, IgR#15, IgR#17 and IgR#18 (M-value > 2.5; Table 1). For the rest of the IgRs for which no signals were detected in Northern blot or microarray analysis, it may be that the transcript level is below our threshold of detection or that this candidate sRNA has a very specific inducing signal different from those included in our assays. This may be the case for IgR#5 with no detected bands in Northern blot (Figure 1) and a slightly lower enrichment detected in the microarray experiment (M-value = 2.15 under 45 min of saline stress; Table 2). In fact, two transcripts of different polarity (sra12a and sra12b) were reported for the same IgR in strain 1021 27. During the preparation of this manuscript, transcripts were reported in total RNA from strain 1021 for IgR#1, IgR#3, IgR#5, IgR#10, IgR#12 and IgR#13 2627. Thus, our data independently confirmed the expression of those putative sRNAs under different experimental conditions and in a different but closely related strain, so we assume that the corresponding IgRs of strain 2011 encode the sRNA homologues of 6S RNA (smrC22 = sra56), smrC9 (= sra32), sra12, sra25, smrC10 (= sra33) and smrC7 (= sra03), respectively 2627.