Identification of novel growth phase- and media-dependent small non-coding RNAs in Streptococcus pyogenes M49 using intergenic tiling arrays

Background Small non-coding RNAs (sRNAs) have attracted attention as a new class of gene regulators in both eukaryotes and bacteria. Genome-wide screening methods have been successfully applied in Gram-negative bacteria to identify sRNA regulators. Many sRNAs are well characterized, including their target mRNAs and mode of action. In comparison, little is known about sRNAs in Gram-positive pathogens. In this study, we identified novel sRNAs in the exclusively human pathogen Streptococcus pyogenes M49 (Group A Streptococcus, GAS M49), employing a whole genome intergenic tiling array approach. GAS is an important pathogen that causes diseases ranging from mild superficial infections of the skin and mucous membranes of the naso-pharynx, to severe toxic and invasive diseases. Results We identified 55 putative sRNAs in GAS M49 that were expressed during growth. Of these, 42 were novel. Some of the newly-identified sRNAs belonged to one of the common non-coding RNA families described in the Rfam database. Comparison of the results of our screen with the outcome of two recently published bioinformatics tools showed a low level of overlap between putative sRNA genes. Previously, 40 potential sRNAs have been reported to be expressed in a GAS M1T1 serotype, as detected by a whole genome intergenic tiling array approach. Our screen detected 12 putative sRNA genes that were expressed in both strains. Twenty sRNA candidates appeared to be regulated in a medium-dependent fashion, while eight sRNA genes were regulated throughout growth in chemically defined medium. Expression of candidate genes was verified by reverse transcriptase-qPCR. For a subset of sRNAs, the transcriptional start was determined by 5′ rapid amplification of cDNA ends-PCR (RACE-PCR) analysis. Conclusions In accord with the results of previous studies, we found little overlap between different screening methods, which underlines the fact that a comprehensive analysis of sRNAs expressed by a given organism requires the complementary use of different methods and the investigation of several environmental conditions. Despite a high conservation of sRNA genes within streptococci, the expression of sRNAs appears to be strain specific.


Background
In recent years, the role of small non-coding RNAs (sRNAs) in regulation of bacterial gene expression has become more evident; however, the large number of sRNAs identified in different bacterial species was unexpected [1][2][3]. Even though sRNAs were conventionally regarded as inhibitory antisense regulators, a significant number of sRNAs that activate bacterial gene expression have been characterized [4]. Furthermore, regulatory mechanisms include both the stabilization and destabilization of target transcripts [5]. Bacterial sRNAs influence the expression of genes involved in processes as diverse as stress response, sugar metabolism, and surface composition [6][7][8][9][10]. With sRNAs representing a whole new level of post-transcriptional regulation, it is no surprise that these molecules play an important role in the tightly controlled expression of virulence factors in many pathogens [11,12].
We were interested in the regulatory sRNAome of Streptococcus pyogenes (group A streptococci, GAS), a common, exclusively human pathogen that causes a variety of diseases. GAS is responsible for mild superficial infections of the skin (impetigo contagiosa) and mucosal membranes (pharyngitis and tonsillitis). Additionally, there is a high global burden of severe GAS diseases such as post-streptococcal sequelae, and severe systemic (streptococcal toxic-shock-like syndrome) or invasive infections (necrotizing fasciitis), leading to over 500,000 deaths per year [13,14]. The controlled expression of virulence factors plays a role in GAS infection, persistence in the host, and development of invasive diseases, which makes the investigation of virulence factors and their regulation a research priority. GAS expresses a large number of virulence factor genes coding for a variety of proteins, including surface components, lytic enzymes, proteinases, cytotoxins, superantigens, and immunoprotective proteins, that are controlled at least partially by the 30 stand-alone transcriptional regulators and 13 two-component systems identified to date in GAS [15]. Virulence factor expression in GAS is highly responsive to environmental conditions and greatly depends on the growth phase.
Little is known, however, about the importance of sRNAs for virulence-related gene regulation in GAS. An overview of small RNAs in streptococci is nicely presented by Le Rhun and Charpentier [16]. Several individual sRNAs have been identified in GAS [17,18], with many more predicted by bioinformatic screens [12,19]. Previous analysis of sRNA expression in a GAS M1T1 serotype using an intergenic tiling array approach identified 40 potential sRNAs, with a very low predicted overlap with candidate genes [20]. The authors concluded that sRNA expression in GAS is serotypedependent. The current work focused on sRNA expression in the skin isolate GAS M49. An intergenic tiling array identified 42 novel and 13 known sRNAs. Data from this experiment were compared to the results of the former GAS M1T1 study, and to predictions of two recently published bioinformatics tools [19,21]. Additionally, we tested the regulation of sRNA expression in correlation to growth media and growth phase. We found very little overlap between the different screening methods, which underlines the importance of using several complementary methods, as well as several environmental conditions, to attain a comprehensive analysis of bacterial sRNAomes.

Identification of sRNAs in the intergenic regions of GAS M49
Custom intergenic tiling arrays representing the genome of S. pyogenes NZ131 (NCBI accession number: NC_011375) were designed to detect the expression of potential sRNAs. A total of 17,823 50-mer probes with an overlap of 15 bp were synthesized to cover the intergenic regions, with 9,082 probes representing the positive strand and 8,741 probes covering the negative strand. Additionally, 174 probes were designed as control probes covering tRNA genes or genes coding for known sRNAs. For example, we probed for fasX [17], SR914400, and SR1754950 [20], all of which were detected in our experiments. Genedata Selector software was used to integrate genomes, tiling array probe sequences, sRNA predictions, and experimental data. Expression data were analysed in their genomic-and sequence-based contexts (Genedata AG).
Total RNA for the tiling array experiments was isolated from GAS M49 grown in chemically defined medium (CDM). Samples were taken from four biological replicates during both mid-log growth phase (OD 600 = 0.4-0.6) and stationary growth phase (OD 600 = 1.2). A signal intensity of >300 was set as a threshold. A positive signal required that a minimum of one probe specific for one strand showed an intensity above the threshold in at least three replicates. Intergenic regions featuring high intensities on both strands were manually removed following the analysis.
We identified a total of 55 putative sRNAs in GAS M49 that were expressed during growth in CDM, 42 of which were novel. Computational functional prediction revealed that a subset of the newly identified RNAs included molecules with similarities to one of the common non-coding RNA families included in the Rfam database [22]. The database covers functional categories of non-coding RNAs determined from multiple sequence alignments. Using the Rfam database, we predicted functions for 14 putative sRNAs. One of these RNAs was predicted to be the structural RNA of the bacterial signal recognition particle (SRP), and another was predicted to be the bacterial RNase P RNA. Further functional categories included three T-box leader elements, three CRISPR family members, one tmRNA, and one endoribonuclease, RNa-seP_bact_b (Table 1). Table 1 contains a summary of the information pertaining to all 55 candidate sRNAs, including the flanking genes, Rfam prediction, and conservation across other genomes. Five putative sRNA sequences overlapped with adjacent ORFs on the same strand. Functional studies will be necessary to clarify whether the corresponding sRNAs are transcribed independently. The overall GC content of all 55 sRNA candidate sequences was 38.3%. This correlates with the GC content of the whole NZ131 genome (38.0%). No specific strand prevalence and no clustering in specific genomic regions were observed for the regulatory sRNA genes. The replication-related gene orientation bias of the protein coding genes [23,24] was mirrored by the sRNA gene candidates. A circular depiction was created with the Artemis DNAPlotter tool [25] ( Figure 1) to visualize the sRNA genes in the context of the NZ131 genome. In a previous bioinformatic approach (MOSES), 20 probable candidates were predicted [19]. In our array analysis, expression of five of the predicted sRNAs was confirmed under the conditions studied (Table 1). Furthermore, we detected 12 streptococcal RNAs that were previously identified by Perez et al. [20] ( Table 1). The phylogenetic conservation of the putative sRNAs was tested by BLAST analysis, and the taxonomic classification was presented following the nomenclature of Facklam [26]. The tiling array technique with overlapping probes did not allow us to detect an accurate start site for the respective sRNA genes. Thus, we described the nucleotides represented by the active probes as the preliminary start and end (Table 1).
For a subset of six candidate sRNA genes, we determined the transcriptional start site (TSS) of the sRNA molecules using the 5 0 rapid amplification of cDNA ends-(5 0 RACE) technology (Invitrogen). The TSS of the analysed sRNAs is shown in Table 1. 5 0 RACE was conducted for two known sRNA genes, fasX and sRNASpy490822 (CRISPR1), for one candidate predicted by MOSES (MOSES4), and for three novel sRNA candidates, sRNASpy490483c, sRNASpy491311c, and sRNASpy491738. The results of the 5 0 RACE analysis are shown in Figure 2. Promoter and terminator predictions for the respective sRNA candidate genes are also included.
Comparison of the sRNA expression data using different sRNA screens The tiling array data were compared with the prediction results of two recently published bioinformatics tools, sRNAScanner [21] and MOSES [19]. As shown in Figure 3A, the overlap between the GAS M49 array data and the sRNA predictions performed with the sequence of the NZ131 genome was minimal. From 20 MOSES candidates, five were detected in the tiling array analysis, while from 137 sRNAScanner predictions, 11 showed a signal in the array. Eight putative sRNAs were selected by both programs. There was only one sRNA that was predicted by both algorithms and was also detected in the tiling array ( Figure 3A). The general accuracy of the three independent screens was supported by the fact that this mutual candidate was the known streptococcal sRNA FASX [17].
We also compared our tiling array data with a previously published sRNA microarray study in GAS M1T1 [20]. The previous study identified 40 putative sRNAs. Even though the sequences of the candidates were conserved across streptococcal genomes, serotype-specific variation of sRNA transcript abundance was observed in northern blot experiments. Screening of GAS M1T1 was conducted with cells grown in complex medium [20], whereas the expression experiments in this study were performed with GAS M49 grown in CDM. Consequently, we found only a limited overlap between the two microarray screens ( Figure 3B). Twelve sRNA candidates were detected in both strains.

Analysis of common motifs in the GAS M49 sRNA population
To identify putative functional regions within the sRNA candidates, all sequences were screened for common features by motif-based sequence analysis using MEME SUITE [27]. The occurrence of shared sequence motifs in different sRNA species could be an indication of common structural properties with functional significance in this region. Seven motifs were identified with consensus sequences with p-values < 1.0 ×10 -7 , spanning 9-27 base pairs ( Figure 4). Putative motif sequences were compared to members of the Rfam database families, and were subjected to TOMTOM [28] motif analysis using the RegTransBase prokaryotic database. Candidates sRNASpy490592 and sRNASpy491336c shared a 14 bp consensus sequence with no apparent known function (motif 4, Figure 4). For all other identified motifs, a known function could be assigned to the respective candidates. The corresponding RNAs were either predicted to be RNAs with non-regulatory functions, e.g. RNAseP (motif 5, Figure 4), or more typically, to represent cis-regulatory RNA elements, e.g. FMN riboswitch or MET box (motif 6 and motif 7, Figure 4).

Regulated expression of sRNA genes in GAS M49
Regulation of sRNA gene expression in GAS M49 under different growth conditions was studied by intergenic tiling array analyses. Total RNA was isolated from bacteria grown in CDM, BHI, or THY. Samples were collected in the exponential and stationary phases. Transcript level changes were expressed as the log 2 signal ratio between conditions, and are listed in Table 2. sRNA gene expression was considered significantly different when the log 2 ratio of the signals was ≤ −1.58 or ≥ 1.58. Twenty-four sRNA genes were regulated in a growth phase-and/or medium- dependent fashion. During growth in CDM, five genes were up-regulated in the stationary growth phase compared to the exponential growth phase, whereas three genes were down-regulated. One of the down-regulated sRNA genes was fasX. This is in accord with previous results, where a reduction in fasX transcript abundance in the stationary growth phase was detected by northern blot analysis [17]. This observation was also confirmed by qRT-PCR analyses ( Figure 5B). Comparison of sRNA gene expression during growth in THY with expression during growth in CDM revealed differential expression of 17 sRNA genes. Of these, 13 genes were down-regulated and four up-regulated in THY. Growth in BHI led to the detection of 12 mediadependent controlled sRNA genes. Nine genes were downregulated and three were up-regulated during growth in BHI compared to growth in CDM. From the 20 sRNA genes that showed media-dependent regulation, seven were regulated in both THY and BHI, and showed the same direction of regulation compared to CDM. Twelve sRNA genes were exclusively regulated in one of the two media, and only one gene was down-regulated in THY but upregulated in BHI compared to CDM. These results are in accord with the fact that both media THY and BHI are complex media, as opposed to the synthetic medium CDM, which forces the bacteria to synthesize a number of components essential for growth. Thus, in CDM, changes in bacterial metabolism are necessary for successful growth and require adaption of the bacterial transcriptome, including the sRNAome.

Validation of sRNA expression by qRT-PCR and northern blot analyses
Expression of sRNA candidates by GAS M49 was tested by gene-specific reverse transcription followed by realtime PCR analysis. Experimental expression validation was performed for the RNAs FASX and sRNASpy490822 (CRISPR1), for the sRNA scan7, which was predicted using sRNAscanner [21], and for three more candidates  identified by tiling arrays in this study (sRNASpy490380c, sRNASpy490957c, and sRNASpy491311c). The expression of the candidates in GAS M49 was verified. Moreover, we confirmed the orientation of the sRNA genes by employing single gene-specific primers for the reverse transcription reaction. Three reactions were performed in parallel: one including the forward primer, one including the reverse primer, and one without any primers. Signals were only detected in samples containing the primer complementary to the coding strand of the respective gene (data not shown). We compared sRNA expression of GAS M49 cultured in CDM medium and THY broth throughout growth ( Figure 5). 5S RNA was used as an internal control for normalization, and was expressed in comparable amounts under all conditions tested in this experiment. Expression of fasX was equivalent during growth in CDM and THY ( Figure 5A). FASX was down-regulated in the stationary phase, an observation that confirmed the array data and previously published results from northern blot analyses [17]. We did not detect strong regulation of CRISPR1, sRNASpy490957c, or sRNASpy491311c during growth in CDM ( Figure 5B). In contrast, scan7 was highly upregulated in the stationary phase ( Figure 5B). The expression of sRNASpy490380c was much higher in THY compared to CDM (almost 100-fold, data not shown). During growth in CDM, no changes in the low level expression of sRNASpy490380c were observed (data not shown).
To further verify candidate gene expression, northern blot analysis of the same putative sRNA genes was performed ( Figure 6). This method allows the determination of approximate transcript sizes. Probes specific for 5S RNA and FASX were included as controls. The apparent molecular weight of candidates CRISPR1, sRNASpy490380c, sRNASpy490483c, sRNASpy491311c, and scan7 corresponded to the length predicted by 5 0 RACE determination. The CRISPR transcript, tracrRNA, showed a band at the expected size of 176 nucleotides, as well as several smaller bands that were likely the result of RNA processing, as observed previously in GAS M1T1 [29] (Additional file 1A). For the putative sRNASpy490957c, transcript analysis by 5 0 RACE predicted a 161 nt full-length product, including the terminator region. However, the most prominent band detected by northern blot analysis migrated at approximately 80 nt. Low intensity bands were detected at approximately 90 nt and 160 nt (Additional file 1B), which might indicate post-transcriptional sRNA processing.

Discussion
Bacterial gene regulation by sRNAs has gained a lot of attention in recent years, because it plays an important role in many cellular processes, including response to environmental changes, growth, and pathogenesis. There is an intriguingly large diversity of regulatory mechanisms, including cisand trans-acting sRNAs, untranslated regions, and riboswitches. Some sRNA molecules act as repressors of translation and destabilize mRNA transcripts, but others act by activating and stabilizing target mRNAs [30][31][32]. One of the best characterized sRNAs in GAS is FASX, which is involved in virulence-related gene regulation [17,33]. Knock-out mutants of fasX show a reduced expression of secreted virulence factors such as streptokinase and streptolysin S. The mechanism for streptokinase gene (ska) expression control is the stabilization of the ska transcript [33]. Lack of FASX-ska-mRNA-interaction in the fasX deletion mutant decreased transcript levels, and consequently decreased streptokinase protein abundance.
A second example of a regulatory RNA in GAS is the untranslated mRNA of the streptococcal pleiotropic effect locus (pel), which contains sagA, the structural gene for streptolysin S. This region was described as a positive regulator of important streptococcal virulence factors, including M-protein, Sic, and SpeB [34]. Strain specificity of PEL function is indicated by the fact that emm transcription was not affected in a sagA-deficient mutant with a M6 background [35]. Similar results have been obtained in GAS M1 and M18 Tn916 sagA mutant strains [36]. Additionally, pel deletion mutant analysis of four M1T1 GAS isolates did not identify any regulatory function for the pel sRNA in this serotype [20]. Another, more recently described untranslated RNA with influence on streptococcal virulence is the 4.5S RNA, a component of the bacterial signal recognition particle (SRP) [37]. While the 4.5S RNA gene is not essential, mutation impairs bacterial growth, lowers virulence factor secretion, and reduces virulence in a mouse infection model.
Recently, several whole genome sRNA screens in Gram-positive bacteria, employing either tiling array or next generation sequencing approaches, revealed an unexpected number of potential sRNAs in several pathogenic species [38][39][40][41][42]. In this context, it is likely that GAS expresses more sRNAs responsible for virulence gene expression control. One whole-genome intergenic tiling array screen of GAS M1T1 identified approximately 40 sRNAs that were expressed during the exponential growth phase in cells cultivated in THY complex medium [20]. The GAS M49 sRNAome in the present study was determined using cells grown in CDM. From 55 putative sRNAs in GAS M49, only 12 were detected previously in the GAS M1T1 screen ( Figure 3B). This result is in accord with the concept that sRNA expression is serotype-dependent and regulated by environmental stimuli. Consequently, we detected media-and growth-phase-dependent sRNA gene regulation in the tiling array expression analysis, or by qRT-PCR of selected candidate genes. It would be interesting to monitor sRNA gene expression regulation under infectionrelevant conditions.  Clustered, regularly interspaced, short palindromic repeat (CRISPR) loci represent an adaptive RNA-based immune system that protects bacteria and archaea from horizontal transfer of phage and plasmid DNA [43]. Among the putative sRNA genes detected in GAS M49 by the tiling array approach, two sequences were categorised by the Rfam prediction program as CRISPR-related RNAs (Table 1). sRNASpy490822 and sRNASpy490827 are encoded by the system II (Nmeni/CASS4 subtype) [44] CRISPR/Cas locus, which was characterized recently by differential RNA sequencing in GAS SF370 (M1 serotype) [29]. Our data suggest that this locus is also active in GAS M49. Expression of sRNASpy490822 was confirmed by RT-PCR on the opposite strand of the CRISPR-associated genes under all conditions tested in this study. This transcript corresponds to the trans-activating CRISPR RNA (tracrRNA), which is responsible for the maturation of CRISPR RNA in concert with RNase III and the CRISPRassociated Csn1 protein [29]. A third CRISPR-related RNA detected in our expression screen, sRNASpy491206c, is encoded in the system I-C (Dvulg/CASS subtype) [44] CRISPR/Cas locus, which is also conserved in streptococcal genomes. In contrast to our array data, this locus appeared to be silent in GAS SF370, where no expression was detected in the differential RNA sequencing approach [29]. Even though the CRISPR loci are conserved throughout GAS genomes, the activity of different CRISPR subtypes appears to be serotype-specific.
In the early years of sRNA research, many bioinformatic prediction tools were developed. One of the most prominent programs was the SIPHT tool, which has been used for many bacterial species [45,46]. However, comparison of the prediction results with the actual in vivo expression of sRNAs often revealed very little overlap between the different screening methods [20,41,47]. The reasons for this discrepancy may be the limitations of the prediction programs as well as the fact that not all sRNAs are expressed under all conditions. The development of sRNA prediction software with improved properties is on-going. We compared our tiling array data with the prediction results of two recently published bioinformatics tools, sRNAScanner [21] and MOSES [19]. As depicted in Figure 3A, the overlap between the tiling array expression data and the sRNA predictions was low. From the 20 most probable candidates of the MOSES analysis, 25% were expressed in GAS M49, whereas 8% of the predicted sRNAScanner predictions were found in the array analysis. Even the overlap between the two bioinformatics data sets was low. The only sRNA that was detected in all three screens was the previously characterized sRNA FASX [17]. These results strongly suggest that a comprehensive analysis of bacterial genomes requires the combination of mathematical predictions with the collection of expression data. In the long term, testing of different conditions, especially mimicking in vivo situations by employing infection models, might lead to an increased overlap of expression detection and bioinformatics analyses.

Conclusions
We present here the identification of 55 putative sRNAs in GAS M49 by an intergenic tiling array approach. The candidate sRNA genes were expressed during growth in CDM. Forty-two of the RNAs were novel, whereas 13 RNAs have been described previously. The sequences of most of the candidates were conserved over streptococcal genomes. However, comparison of our GAS M49 sRNA expression data to another array analysis of a GAS M1 strain, and to two in silico screening methods, revealed little overlap between the different approaches. Thus, the investigation of several conditions and the combination of screening tools will be necessary to gain a comprehensive understanding of the abundance of sRNAs in GAS. The identification of novel differentially expressed sRNA genes will enhance our understanding of virulence related gene regulation in GAS. To account for specific expression patterns of putative sRNAs, infection relevant conditions combined with next generation RNA sequencing should be employed to investigate sRNA dependent regulatory networks in GAS.

RNA isolation
Total bacterial RNA from cultures grown to exponential and stationary phase of growth was isolated using the FastRNAProBlue Kit from MP Biomedicals according to manufacturer's instructions. The purified total RNA was digested with DNaseI (Ambion) to remove remaining traces of chromosomal DNA. The RNA preparation was treated with 10 U of DNase1 for 30 min at 37°C. The enzyme was subsequently heat inactivated at 72°C for 5 min.

Enrichment of small RNAs
Five micrograms of total RNA were fractionated using the Ambion FlashPAGE Fractionator, Ambion FlashPAGE Precast Gels, and the Ambion FlashPAGE Buffer Kit, following manufacturer's instructions. To collect the fraction of RNA molecules <200 nucleotides in length, the protocol was modified by increasing the running time from 12 min to up to 45 min at 75 V.

Labelling
The small RNA fraction was ethanol-precipitated overnight at −20°C. The RNA was pelleted by centrifugation, dissolved in nuclease-free water, and labelled using the Ambion mirVana miRNA Labeling Kit following the manufacturer's instructions. In brief, this kit involves two main steps; the 3 0 amine-modified tailing reaction, and labelling with NHS-esters. Poly(A) Polymerase and a mixture of unmodified and amine-modified nucleotides were used to add a 20-50 nucleotide tail to the 3 0 end of each RNA molecule in the sample. The amine-modified RNA molecules were purified and coupled to aminereactive labelled biotin moieties as NHS-esters.

Design and synthesis of microfluidic microarrays
We used a microfluidic biochip (Febit Biomed) consisting of eight independent reaction chambers, the arrays, enclosed in a cartridge for fully automated processing. Each array contains 15,625 features which are synthesized in situ inside the microchannels using the Geniom One technology (febit biomed) [49]. The 50mer probes were designed as a whole genome tiling array, covering the intergenic regions of the S. pyogenes NZ131 genome (NCBI accession number: NC_011375). The forward and reverse oriented probes were synthesized in separate arrays. Thus, two arrays per sample were used.

Microarray hybridization and detection
All hybridization and detection steps were carried out using a Geniom RT Analyzer (febit biomed). Hybridizations were performed overnight (16 hours) at 42°C. Subsequently, biotin was detected with streptavidin-phycoerythrin (SAPE). A signal amplification step was added using biotinylated anti-streptavidin antibodies (Vector Laboratories) and a second incubation with SAPE (Invitrogen). Signal detection using the appropriate filter set (Cy3) of the Geniom device employed the auto-exposure function of the Geniom software. The data discussed in this publication have been deposited in the NCBI Gene Expression Omnibus [50] and are accessible through GEO Series accession number GSE31228 (http://www.ncbi.nlm.nih. gov/geo/query/acc.cgi?acc=GSE31228).

Microarray data analysis
Raw intensities were analysed and extracted using Geniom Wizard software (febit biomed) as a tab delimited text file. The data were then converted into a matrix, with rows corresponding to the features and columns corresponding to the different samples. Data analysis was performed using GeneSpring GX (version 11) software (Agilent Technologies). The array background was calculated as the median signal intensity of all "blank-control" features on the array. Data were background corrected and then normalized using quantile normalization [51]. Following normalization, a quality control step was performed that removed all data sets with a correlation coefficient less than 0.9 compared to the corresponding biological replicates. Of the original four biological replicate data sets representing cells grown in CDM, at least three were included in the analysis. The remaining probes of the biological replicates required intensity values greater than 300 on all three arrays. Regions that showed signals on probes of both strands were manually removed following the primary analysis. The statistical significance of the determined signals was tested by unpaired student's t-test with a false discovery rate of 5%. Resulting data were combined with gene information from the flanking coding regions. Terminators and promoters were predicted by TransTermHP [52] (http://transterm.cbcb.umd.edu/tt/Streptococcus_pyogenes _NZ131.tt) and BDGP Neural Network Promoter Prediction [53], and BProm (www.SoftBerry.com), respectively. To investigate sRNA gene regulation, two biological replicates of growth experiments conducted in THY or BHI were included. Following data normalization, threefold signal intensity differences between various conditions were determined using the GeneSpring GX (version 11) software (Agilent Technologies). A motif search was conducted using MEME Suite [27], followed by motif analyses using TOMTOM [28], (http://meme.sdsc.edu/ meme/intro.html).

0 RACE
The transcriptional start sites of sRNA candidates were determined using 5 0 RACE (Invitrogen) following the manufacturer's instructions. Briefly, first strand cDNA was synthesized using gene-specific primers (Additional file 2). The original mRNA was enzymatically removed and the 3 0 end of the cDNA was tailed with dCTP by terminal deoxynucleotidyl transferase (TdT). PCR amplification was performed with nested, sequence-specific primers and an anchor primer provided by the 5 0 RACE system. Primers specific for the sRNA genes tested here are listed in Additional file 2. Following amplification, PCR products were cloned into a TOPO-TA vector (Invitrogen) and sequenced (GATC Biotech AG).

Quantitative reverse transcription PCR
Acidic phenol-extracted, DNaseI-treated total RNA was reverse transcribed to generate cDNA using the First-Strand cDNA Synthesis Kit from Invitrogen following the protocol provided by the manufacturer. For genespecific reverse transcription (RT), three reactions were performed: two strand-specific reactions with either one forward or one reverse primer, and one control reaction without any primer. Primers were designed based on the full genome sequence of S. pyogenes M49 strain NZ131 (NCBI accession number: NC011375) and are listed in Additional file 3. Three independent RT experiments were performed and all subsequent PCR reactions were performed in triplicate. Primer efficiency was tested on genomic GAS M49 DNA prior to use in RT reactions. All cDNA products were amplified by PCR with two primers specific for the respective candidate sequences. Real time PCR amplification was performed with SYBR Green (Fermentas) using an ABI PRISM 7000 Sequence Detection system (Applied Biosystems). The level of 5S RNA gene transcription was used for normalization. Relative gene expression was determined by the ΔΔCT method [54].

Northern blot analyses
Total RNA was isolated during exponential (OD 600 = 0.4), transitional (OD 600 = 0.8), and stationary (OD 600 =1.2) growth phases. RNA samples (10 μg per growth phase) were loaded onto an 8% TBE-Urea polyacrylamide gel and separated by electrophoresis. Size standards (Ultra Low Range Ladder, Fermentas) were loaded on the same gel. RNA was electroblotted onto positively charged nylon membranes (Ambion), UV cross-linked, and probed overnight with a probe complementary to a candidate sRNA. Probes were generated by PCR with the same primers as