Identification of CRISPR and riboswitch related RNAs among novel noncoding RNAs of the euryarchaeon Pyrococcus abyssi

Detection of novel ncRNAs

Annotation and expression of P. abyssi CRISPR loci

The comparative computational screen selected the four CRISPR loci annotated recently in the P. abyssi genome [40]. A careful sequence inspection of these cassettes revealed an erratic number of repeats and spacers in the P. abyssi CRISPR loci. Based on recent observations showing that new spacers are integrated mainly on the side proximal to the leader sequence of the CRISPR cassette and that degenerated direct repeats accumulate at the distal 3' end of the cassette [29], we revised the annotation of the P. abyssi CRISPRs (Figure 1A). The CRISPR 1 (encoded on the reverse strand) and CRISPR 4 (encoded on the sense strand) are composed of 23 and 29 direct repeats and spacers, respectively, with similar 30 nt direct repeat sequences. The observation that the penultimate and ultimate 3' end-direct repeats are degenerate by one to six mismatches confirms the orientation proposed by the UCSC archaeal genome browser [41], which differs from the orientation proposed by [40]. The atypically short CRISPR 3 locus contains only four spacers with degenerate direct repeat sequences related to CRISPR 1 and 4. The CRISPR 2 cassette encompassing eight spacers is distinguished by the sequence of its 29 nt direct repeats. In order to position the degenerate direct repeat (four mismatches) at the end of the CRISPR cassette, as in the majority of the CRISPR loci [27], we propose that CRISPR 2 is encoded by the reverse strand as opposed to previously reported annotations (UCSC archaeal genome browser and [40]).

No signals were detected by Northern blotting with strand-specific direct or reverse probes against sequence matching spacers 2 and 7 of CRISPR 2 and spacer 2 of CRISPR 3 suggesting that these loci are not transcribed (data not shown). This tentative conclusion was confirmed by more sensitive tests involving primer extension analysis to map 5' ends of the CRISPR precursors and to identify promoters (see below). In contrast, probes against sequences matching spacer 1 of CRISPR 1 and spacers 4 and 12 of CRISPR 4 allowed the detection of transcripts in all three growth phases (Figure 1B; Table 1). The approximately 60 nt length of CRISPR-derived RNAs, named hereafter Cr1-1, Cr4-1 and Cr4-12, are of the size reported for crRNAs corresponding to a spacer sequence and part of the direct repeat [16, 24]. Moreover, primer extensions with total RNA extracted from cells in entry-stationary phase (Additional file 3, Figure S2A) accurately identified their 5' ends (Figure 1C), which correspond to the dicing site reported for the P. furiosus endoribonuclease Cas 6 [42]. Since only single reverse transcription arrests are observed (Additional file 3, Figure S2A) even though multiple Northern blot signals are detected (Figure 1B), we propose that length heterogeneity results from partial or incomplete 3'end processing. In conclusion, the CRISPR 1 and 4 cassettes are transcribed to produce small crRNAs in P. abyssi. No signals corresponding to precursor or other intermediates were detected by Northern hybridization (Figure 1B). Nevertheless, primer extensions with oligonucleotides hybridizing to the 5' junction of their respective first direct repeat resulted in the detection of an RNA precursor and thus permitted the identification of the transcription starts of the CRISPR 1 and 4 RNA precursors (preCr1 and 4, respectively) (Figure 1C & Additional file 3Figure S2A). Similar experiments with specific primers of CRISPR 2 and 3 did not reveal any RNA precursor confirming that CRISPR 2 and 3 are not expressed. Signals were not detected in Northern hybridization with probes against the ultimate spacer 22 of CRISPR 1 and spacer 28 of CRISPR 4, which harbor degenerate repeats. These results suggest that the direct repeats are important for the processing and/or stability of the mature crRNAs. To identify transcription signals, we analyzed the 200 bp region upstream of the first direct repeat of each P. abyssi CRISPR cassette. This analysis recognized conserved AT rich motifs arguing for the presence of promoters. The CRISPR 2 leader sequence has three AT-rich and one A-rich element upstream of the first direct repeat, not corresponding to any typical P. abyssi consensus promoter sequence as defined in [38]. In contrast, the CRISPR 1 and 4 cassettes have consensus promoter sequences 10 nt upstream of the experimentally determined 5' starts of preCr1 and 4 (Figure 1C). Two additional short elements at position +30 and -100 relative to the transcription starts were detected in CRISPR 1 and 4. These sequences are also observed in other thermococcal CRISPR cassettes at the same distance from the first direct repeat (data not shown). It should be noted that the general organization of the CRISPR 3 leader is similar to those of CRISPR 1 and 4 except for the distance between the +25 A-stretch and the first direct repeat (Figure 1C). This feature might account for the absence of transcription of the CRISPR 3 locus.

Expression of unique intergenic regions conserved among thermococcal genomes

Four ncRNAs were clustered in this category (Table 1). For three of them, sRk11, sRk33 and sRk61, expression is detected at about the same intensity in all growth conditions whereas sRk28 is clearly growth phase-regulated with little or no detectable RNA in stationary phase (Figure 2A). Remarkably, the genomic contexts of sRk11, sRk28 and sRk33 are similar, all being flanked by ORFs transcribed in the same direction. Both, sRk33 and sRk61 cover predicted promoter sequences of the downstream annotated genes (Figure 2B). The respective 5' ends (Table 1) determined by primer extension experiments (Additional file 3, Figure S2B) are at an acceptable distance from the predicted promoter elements (BRE/TATA and TATA boxes) for sRk11 and sRk33, and overlap a predicted TATA box for sRk61 (Figure 2B). No such sequence is found upstream of sRk28 which might suggest that its transcription depends on the upstream gene PAB1991 since this region is predicted to form an operon that includes PAB1992. Except for sRk61, these ncRNAs are extremely well conserved in sequence and structure throughout other thermococcal genomes (Additional file 4, Figure S3A). The proposed RNA secondary structures, based on RNAfold and multiple alignment covariation analysis, highlight compensatory variations that maintain hairpin structures and, for sRk28, a potential pseudoknot structure (Figure 2C). Only sRk33 was predicted and validated in a previous comparative analysis in P. furiosus (referred to as SscA in [19]). sRk33 has a predicted secondary structure with a basal P1-stem with a high proportion of co-variations, an apical P2 stem conserved in sequence, and a 3' end sequence composed of A and U rich stretches that do not fold into a stable secondary structure (Figure 2C). Altogether, these characteristics suggest that sRk11, sRk28 and sRk33 might have conserved cellular functions in the Thermococcales.

Expression of repetitive intergenic regions conserved among thermococcal genomes

Among the three ncRNAs clustered in this category (Table 1), sRk49 shows sequence similarities with two additional regions in P. abyssi (49.2 and 49.3) and three in P. horikoshii(Additional file 5, Figure S4A). Sequence similarities preclude the design of specific probes to distinguish between sRk49 and 49.2. Northern blot signals provided evidence for a growth stage-specific transcription pattern with the most abundant species migrating about 150 nt. These transcripts could arise from either of the loci harboring an upstream TATA-like promoter (sRk49 and 49.2) (Additional file 5, Figure S4B & C). CR-RT-PCR experiments failed to detect primary transcripts, but suggested that antisense transcripts are expressed from the sRk49 operon (data not shown). Specific RNA structures or RNA modifications could have impeded the accurate amplification of the primary transcripts. Moreover, related sequences in the P. horikoshii genome are all clustered with CRISPR cassettes showing an organization comparable to that observed for the 49.2 region. All sRk49 variants are atypical in that they contain multiple short polyA stretches making RNA folding into stable helical structures implausible.

The two other ncRNAs of this category are sRk48 and sRk52 (Table 1, Figure 3). They were identified independently in our initial computational search and are part of a set of six similar genomic sequences in P. abyssi. The similarities extend to all thermococcal genomes with four, three, two and one copies in P. furiosus, P. horikoshii, T. sibiricus and T. kodokaraensis, respectively (Additional file 6, Figure S5). The sequence similarities precluded the design of specific probes for each locus. Therefore, the signals detected in the Northern blot in Figure 3A could arise from any of the six repeated genomic regions. In order to identify the transcribed regions, a CR-RT-PCR experiment was performed. Most clones corresponded to the 130 nt transcript with a majority matching sRk52 and a minority sRk48 (Figure 3B). They shared similar 5' ends and variable 3' ends. In addition, a small percentage of clones corresponded to longer variants of sRk52 (177/180 nt) mapping a transcription initiation site 22 nt downstream a predicted BRE/TATA box promoter (Additional file 5, Figure S4). The shorter 5'-monophosphorylated sRk52 transcript is likely to have arisen from maturation of the longer primary transcript. Interestingly, RNAfold and multiple sequence alignments suggest a highly conserved RNA secondary structure within the Thermococcales (Figure 4A). Many nucleotides, particularly in the P0 and P3 stems, support co-variations arguing for structural conservation. The majority of the cloned CR-RT-PCR fragments contain sequences corresponding to the P1, P2 and P3 stem-loops whereas a minority includes the P0 stem-loop in their 5' part (Figure 4A). The P1 stem-loop is unusual since its primary sequence is highly conserved throughout the Thermococcales giving this structure not only a high propensity to form, but also suggesting an important cellular function. Inspection of the P2-loop of sRk52 suggested that it forms a K-loop, which is consistent with the gelretardation of 5'end-labeled sRk52-transcripts in presence of the recombinant P. abyssi L7Ae protein (Figure 4B). The P3-loop is variable and can be extended into a stem structure for the two P. abyssi sRk48-like sequences.

Expression of repetitive intergenic regions specific to P. abyssi

Amounts of the two ncRNAs of this category, sRkB and sRkC, were highest upon entry into stationary phase (Figure 5A; Table 1). Because these loci (Figure 5B) have significant sequence similarities to four other regions dispersed in the P. abyssi genome, we performed additional experiments to determine if the other loci were transcribed. Despite the sequence similarities, we could design oligonucleotide probes specific for each repetitive locus and perform additional Northern blot probing (data not shown), CR-RT-PCR experiments (Figure 5B and 5C) and primer extension analyses (Additional file 3, Figure S2B). From these data, the sRkB and sRkC loci were shown to be the only regions transcribed. In order to identify the 5' and 3' ends of these transcripts, three distinct CR-RT-PCR experiments were carried out. One pair of primers specific of the sRkB locus (R1) and two pairs of primers specific of the sRkC locus (R2 and R3) were chosen to characterize the overall RNAs expressed at these loci. The majority of the R1 clones correspond to transcripts from the sRkB locus harboring an identical 5'end and a variable 3'end of which half ended between position 209 to 218 nt and half between position 340 to 343 nt (Figure 5B and Additional file 4, Figure S3B). The length of sRkB RNAs as determined by CR-RT-PCR, corresponds to the sizes of the transcripts detected in Figure 5A. Moreover, sequence analysis revealed two typical promoter consensus sequences 24 nt and 179 nt upstream of the experimentally determined 5' end (Figure 5B and Additional file 4Figure S3B). Interestingly the 3' extremities of all the RNA species are in the first 25 to 150 nucleotides of the PAB0571 ORF. Additional primer extension analysis with a specific probe within PAB0571 suggests that a unique RNA precursor is transcribed from this locus and subsequent processing events including 3' end trimming generate the different sRkB RNA species (data not shown). The R2 and R3 CR-RT-PCR analyses permitted the identification of three stable RNA species from the sRkC locus (Figure 5B). The two longer RNA species of roughly 137 nt (majority of the sequenced R2 clones) and 211 nt (half of sequenced R3 clones) share exactly the same 5' extremity. The 5' ends of the shorter 64/66 nt RNA species are positioned 149 nt downstream (minority of sequenced R3 clones). In addition, primary transcripts were distinguished from processed ones by omission of treatment of the RNA preparations with tobacco acid pyrophosphatase (TAP) (Figure 5C). The 5' ends of 211/18 nt-RNA species are mostly triphosphorylated. As in the case of sRkB, a promoter sequence is found at an acceptable distance (24 nt) from the 5' end (Figure 5B and Additional file 4Figure S3B). RNA species transcribed from the sRkC locus could result from 5' and 3' end processing events of the primary 211/218 nt long transcripts resulting in precise 5' monophosphorylated ends and variable 3' ends. However, it should be noted that the 64/66nt RNA species were not detected in Northern blot analysis with specific probes (data not shown) suggesting that they are in low abundance in our RNA preparations. In contrast to the sRkB locus, the sRkC locus is located in a long intergenic region. The 3' end of PAB1080.5n, the nearest ORF, is located approximately 400 bp upstream of the 5' end of sRkC (Figure 5B).

As mentioned earlier, the sRkB and sRkC loci are part of a set of six similar sequences that had no counterparts in other thermococcal genomes and are therefore specific to P. abyssi. Consistent with their lack of expression, the BRE/TATA promoter sequences are missing in the other four other P. abyssi loci (Additional file 4, Figure S3B). Strikingly, one of these loci corresponds to the 3' end of the PAB1452 open reading frame. This ORF has been identified as a 'single OrfB element' of the IS605/IS200 family of transposases [43].

Interestingly, in our initial computational characterization, sRkB and sRkC were predicted to fold into a structure with a K-turn RNA motif. To test if sRkB and sRkC form a K-turn, we performed gel retardation assays with recombinant P. abyssi L7Ae (Figure 6A). This assay reveals at least two distinct ribonucleoprotein complexes (RNPs) suggesting that more than one L7Ae protein binds to the sRkB and sRkC RNAs. To elucidate these RNA-protein interactions, footprint experiments using RNase T1 protection were performed with radioactively labeled 216 nt sRkB transcript (Figure 6B). Two sets of clustered guanosines are protected from RNase T1 digestion in the presence of L7Ae, suggesting K-turns in stems P1 (KT₁) and P4 (KT₂) (Figure 6B). Further support for the existence of KT₁ and KT₂ was obtained by mutations of the sheared GA base pairs, which are the hallmark of the K-turn. In this case, gel retardation assays showed that the shift with recombinant P. abyssi L7Ae was abolished (data not shown). Sequence alignment (Additional file 4, Figure S3B) and RNA-fold predictions support the RNA secondary structure shown in Figure 6C. Altogether, our data reveal that several transcripts expressed from the sRkB and sRkC loci have K-turn motifs that might be important for their cellular function. Another well-known RNA structural motif, a loop E [39], might form in the upper part of P5. However, since there is no specific ligand that interacts with loop E, validation of this fold requires a detailed structural analysis. It is interesting to note that the initiation codon for the PAB0571 is part of the P5 stem of sRkB suggesting a functional link with the expression of the downstream PAB0571.

Discovery of novel ncRNAs by combined bioinformatic approaches

The rationale for carrying out a search of ncRNAs in P. abyssi, the best studied Thermococcale, was that with the exception of the box H/ACA [13] and the box C/D [12] noncoding guide RNA families, few thermococcal ncRNA families have been described. Our study provides evidence for the growth-regulated expression of 12 novel ncRNAs in the P. abyssi genome consisting of three crRNAs from CRISPR loci, four ncRNAs from unique loci conserved throughout Thermococcales of which one appears to be a homologue of the P. furiosus SccA RNA [19], three ncRNAs from repetitive loci conserved throughout Thermococcales, and two ncRNAs from repetitive loci specific to P. abyssi. Since small proteins (sproteins) of less than 50 amino acids have been largely disregarded in genome annotations [44], we cannot exclude the possibility that the ncRNAs identified here encode sproteins or peptides. We therefore searched for open reading frames within the ncRNA sequences starting with AUG, GUG or UUG, and ending with UGA, UAG or UAA. For several of the ncRNAs, it was possible to find small ORFs of 25 to 40 amino acids. The only significant ORF identified corresponds to the distal portion of a putative transposase CDS (see below). We also analyzed the GC content of the ncRNAs. We found that it ranges from 33% to 66%. This large spectrum is consistent with the GC content for known ncRNAs (low for C/D guide RNAs, high for tRNAs and rRNAs). Since the computational analyses were restricted to intergenic regions, no cis-encoded antisense ncRNA complementary to ORFs, as identified in transcriptome analyses of Sulfolobus solfataricus and Methanosarcina mazei[22, 45] could be predicted. Below, experimental and genomic features of the new P. abyssi ncRNA families are discussed.

Processing of ncRNAs

In this study, transcripts of different length were identified at one locus suggesting that RNA processing occurs in P. abyssi. The sRkB, sRkC, sRk48 and sRk52 primary transcripts with triphoshorylated 5' ends are processed into shorter 5' end monophosphorylated transcripts through maturation reactions (Figure 3C and Figure 5C). For example, the 3' end heterogeneities observed for sRkB, sRkC, sRk48 and sRk52 RNAs could arise from trimming of the primary transcript by the 3' to 5' exonucleolytic activity carried by the Pyrococcus exosome [46]. The 5' monophosphorylated RNAs could result from endonucleolytic or pyrophosphohydrolytic activity. Previous examples of processing in Pyrococcus were limited to tRNA processing by RNase P [47] and crRNA maturation by the Cas 6 endonuclease [42, 48]. Recent results suggest that the 5' to 3' exonucleolytic activity of RNase J homologues may also be involved in RNA processing in Euryarchaea and Crenarchaea [49, 50].

CRISPR-derived RNAs

Our detection of crRNAs is the first demonstration of the existence of an active CRISPR defense system in P. abyssi. The CRISPR loci 1 and 4 appear to be expressed independently of growth phase in our experimental conditions. In agreement with the findings in P. furiosus[51], no antisense transcription of these CRISPR loci was detected (data not shown) excluding the possibility that double strand RNA intermediates are formed. This contrasts with the report that RNA is transcribed from the complementary spacer strand in S. acidocaldaricus[52]. This difference could reflect a specific feature of the Sulfolobales or a technical issue regarding the sensitivity of the Northern blots analysis used in the studies of P. furiosus and P. abyssi. Based on our data, we propose a new annotation for the four P. abyssi CRISPR arrays that differs from earlier proposals (UCSC archaeal genome browser and [40]). The presence of a typical consensual promoter in the leader sequences of CRISPR 1 and CRISPR 4 correlates with the detection of CRISPR-derived RNAs from these loci. It has been reported that CRISPR cassettes expressing crRNAs provide acquired immunity [28, 53, 54]. It should be noted that short regions of spacers 7 and 19 of CRISPR 1 (14 bp and 17 bp, respectively) are complementary to two coding regions of PAV1, a virus that infects P. abyssi (data not shown) [55]. No other similarities were observed between CRISPR 1 and 4 spacer sequences and known P. abyssi mobile genetic elements [56, 57]. The silent CRISPR 2 and 3 loci differ from the expressed cassettes in several ways. First, the CRISPR 2 locus has divergent direct repeat and leader sequences. The CRIPSR 3 locus is atypical by its leader sequence, the reduced number of repeats, and the sequence degeneracy of its repeats, suggesting that it is a relic of an active CRISPR cassette as mentioned in [27, 28, 58]. In general, CRISPR cassettes are physically linked to a cohort of conserved cas or cmr genes, in varying orientation and order, that encode CRISPR-associated proteins (reviewed in [24, 27, 31]). It is interesting to note that the only cas gene (Cas6, PAB1613) in the vicinity of P. abyssi CRISPR cassettes is located upstream of CRISPR 3 (Figure 1A). In the P. abyssi genome, all the other genes encoding Cas/Cmr core proteins are grouped into an operon located far from any CRISPR locus (Figure 1A). Further studies will be required to elucidate the mode of action and the dynamics of the CRISPR/Cas system in P. abyssi.

5' UTR-derived ncRNAs

Distinct RNA species are not always indicative of independently synthesized RNAs. As revealed by RACE experiments and analysis of genomic context, several ncRNAs (Table 1) seem to derive from mRNA leaders. This class of ncRNAs appears to originate from maturation of longer transcripts, as we demonstrated for sRkB and propose for sRk28, sRk33 and sRk61. Stable RNAs derived from RNA leaders were initially identified in E. coli and some appear to correspond to riboswitch elements such as RFN and THI [59]. It is now well established that 5'UTRs can encompass transcription-termination signals involving riboswitch structures [60, 61]. In Archaea, it has recently been suggested that riboswitches may also exist. Based on comparative genomics, crcB has been found in Archaea and Bacteria, and TPP and flp in Euryarchaea [60, 62]. The crcB element was found within our set of 82 candidates (Additional file 2, Table S1), but ncRNA corresponding to this element was not detected by Northern blotting under the conditions employed in this study. To date, no biological evidence showing gene regulation by riboswitches has been reported in Archaea. Nevertheless, the conserved 70 nt sRk28 shares several feature with the bacterial preQ1 riboswitch [63, 64]. PAB 1992, adjacent to sRk28, is predicted to encode an ATPase with a queosine synthesis-like protein domain (QueC). The growth-regulated sRkB RNA is transcribed from a locus adjacent to PAB0571, which encodes a putative ATPase. In bacteria, the expression of several ATPase-like proteins is controlled by cis-mechanisms involving the SAM-I riboswitch. It is interesting to note in Listeria monocytogenes that this type of riboswitch also acts in trans as a small ncRNA to regulate expression of the virulence factor PrfA [7]. Moreover, our data strongly suggest that sRkB has the propensity to form K-turn structures, which are found to be functional RNA motifs involved in the assembly of ribonucleoprotein complexes and in the orientation of pseudoknot interactions in the aptamer structures of the SAM-I and lysine riboswitches [36, 37]. Therefore, sRkB could have a specific function in the P. abyssi cell that involves the regulation of the expression of the putative ATPase encoded by PAB0571. We speculate that sRk28 and sRkB may be representative of a novel family of archaeal ncRNAs produced by transcription attenuation or RNA processing. Finally, the promoter-associated sRk33 and sRk61 could be related to the recently discovered ncRNAs associated with transcription starts sites reported in higher eukaryotes [65, 66], in yeast [67, 68] and in Salmonella enterica serovar typhi [69]. This class of ncRNAs is thought to interfere with transcription of the downstream promoter by cis-mediated occlusion or by a trans-mechanism.

Transposase-related ncRNAs

In this study, sRk48, sRk52, sRkB and sRkC ncRNAs, which are transcribed from repetitive loci, were shown to be related to a CDS annotated as an orphan OrfB element of the IS607 and IS200/605 family, [43]. For example, in the case of sRk48 and sRk52, similar sequences correspond to 3' end coding regions annotated as partial OrfB-like IS607 transposase in P. furiosus and T. sibiricus. In bacteria, it has been shown that IS10 transposition events are controlled by a complementary cis-encoded antisense ncRNA [70], but this seems unrelated to our observations since the ncRNA sequences matched the sense sequence of the 3' end of the annotated ORFs. However, it is pertinent to note that a link between ncRNAs and transposase ORFs was already observed in Sulfolobus solfataricus[16, 17] and in Salmonella [71]. Moreover, in Sulfolobus solfataricus[17] it was shown that several small RNAs linked to annotated transposase ORFs could bind the multifunctional ribosomal protein L7Ae through recognition of an RNA kink turn as is the case for sRkB, sRkC and sRk52.

Computational search

Genome sequences and related annotation files (gbk, .fna and .ptt extensions) were downloaded from the NCBI ftp database for P. abyssi, P. furiosus, P. horikoshii and T. kodakaraensis genomes. For these four genomes, a comparative analysis of all intergenic sequences was realized using RNAsim as described in [2]. RNAsim searches for conserved sequences and structured regions between different genomes by using Wu-blast 2.0 to select pairwise alignments including conserved sequences (here with W = 7, E < 0.0001) and QRNA [72] to identify in pairwise alignments base substitution patterns that could correspond to a conserved RNA secondary structure. In a final step, RNAsim combines this information to predict loci that are conserved in multiple genomes. In this analysis, all the regions encoding annotated ncRNAs except for the box H/ACA RNA genes were excluded from the set of intergenic regions (46 tRNAs, 1 rRNA operon, 2 5S rRNAs, 59 C/D guide RNAs, 1 RNase P RNA and 1 7S/SRP RNA). GC-rich regions were predicted in P. abyssi by using the same segmentation approach as previously used by Klein and colleagues to predict new ncRNAs in P. furiosus[19]. However, our approach differed in that the transition probabilities were adjusted to enrich the number of candidate regions. Additional sequence comparisons were performed on putative ncRNA candidates with Blastn (default values and W = 7) to add to this initial set other P. abyssi and archaeal homolog sequences. BRE/TATA, consensus boxes 5'_RNNANNTTTAWA_3' and 5'_RAAANNTWWWWA_3', and TATA consensus box 5'_TTWWWWA_3', K-turn and K-loop consensus motifs, inferred from the literature [34, 38, 39, 73], were identified using Patscan [74]. Regions of favorable free energy were computed by setting the free energy threshold such that all tRNAs were found in sliding windows of 150 bp (here threshold=-32.3). Only regions of free energy below the threshold and longer than 50 bp were displayed in ApolloRNA, an extension of the annotation environment Apollo [75], developed to support ncRNA analysis. Highly structured hairpins with minimal hairpin stems of 6 bp (including G-U and U-G pairs) were searched with Patscan. RNA secondary structures were proposed on the basis of multiple alignments of similar regions within the Pyrococcales using Multalin [76] that were improved manually by combining RNAfold predictions and covariations. All data including predictions, motifs and annotations were integrated and visualized in ApolloRNA.

Oligonucleotides used in this work

Additional file 7 Table S2 list the primers used for Northern blot detection, CR-RT-PCR, primer extension, and the preparation of transcription templates by PCR.

Preparation of total cellular RNA

P. abyssi strain GE5 cells were grown as described in [77] at 95°C in Vent Sulfothermophiles Medium. P. abyssi cells were stopped at three different growth phases: exponential, end of exponential and stationary phases. The exponential and stationary phase cell paste was provided by A. Hecker (Institut de Génétique et de Microbiologie, Paris Sud-Orsay). Entry into stationary phase cell paste was purchased (Reims University, France). Total RNA was prepared from P. abyssi cell paste by Trizol extraction followed by treatments with DNase RQ1 (RNA-free, Promega), proteinase K and phenol/chloroform extraction followed by ethanol precipitation.

Northern blotting analysis

Total RNA (10 μg)extracted from cells in exponential growth phase (E), entry into stationary phase (ES) or stationary phase (S), and a 5' [³²P]-end-labeled denatured PhiX174/HinfI DNA ladder were separated on a denaturing 6% polyacrylamide gel (8M urea, 1 × TBE buffer). Gels were transferred onto Hybond N+ nylon membrane using a Transphor Power Lid (Hoefer) apparatus in 0.5 × TBE buffer. The RNAs were UV cross-linked to the membrane (1200 J/cm²). Prehybridization was carried out for 30 minutes at 50°C in hybridization buffer (5 × SSC, 1 × Denhartd's solution, 1% SDS, 0.05 mg/mL sperm DNA herring). DNA oligonucleotides were designed using Primer designer or Vector NTI and 5' end labeled with [γ-32P] ATP and polynucleotide kinase. Hybridizations were carried out at 50°C for 16 h followed by two washes in 0.1 × SSC, 0.1% SDS buffer at room temperature for 10 min. The blots were analyzed by phosphorimaging (Molecular Dynamics) or autoradiography using MR or MS film (Kodak).

Primer extension and Circular RACE (CR-RT-PCR)

Primer extension analysis was performed using 10 μg of total RNA prepared from P. abyssi cells in entry into stationary phase. Total RNA was reverse transcribed at 42°C by AMV reverse transcriptase (Promega) using a 5' end labeled specific primer. CR-RT-PCR was performed with 20 μg of total RNA prepared from P. abyssi cells in entry into stationary phase, treated with (+) or without (-) 25U of Tobacco Acid Pyrophosphatase (TAP) according to manufacturer's protocol (Epicentre Biotechnologies). RNAs were extracted with phenol/chloroform then precipitated with ethanol. RNA (1 μg) +/- TAP was circularized with 40U of T4 RNA ligase (New England Biolabs), extracted with phenol/chloroform, ethanol precipitated and reverse transcribed by AMV reverse transcriptase using specific primers. After ethanol precipitation, the reverse transcripts were PCR amplified using Crimson Taq (New England Biolabs) and appropriate primers. The products were separated on a 6% native polyacrylamide gel (1% glycerol, 0.5 × TBE), treated with TAP, cloned in pCR^®II-TOPO^® with TOPO-TA Cloning^® Kit according to the manufacturer's instructions (Invitrogen) and sequenced by Eurofins MWG Operon. About 100 sequences were analyzed for each CR-RT-PCR experiment.

In vitro transcription

A portion of the intergenic region corresponding to sRk52, sRkB and sRkC, respectively, was amplified by PCR from P. abyssi genomic DNA using specific primers (Additional file 2, Table S1). PCR products served as templates for in vitro transcription using T7 RNA polymerase as previously described [34].

Electrophoretic mobility shift assay (EMSA) and enzymatic structural probing

EMSA was performed using E.coli tRNA as a non-specific competitor as previously described [34]. RNA and RNP complexes were separated on a native 6% (19:1) polyacrylamide gel containing 0.5 × TBE and 5% glycerol. Electrophoresis was performed at room temperature at 250 V in 0.5 × TBE running buffer containing 5% glycerol. The gels were dried and visualized using a Fuji-Bas 1000 phosphorImager.