To identify sRNAs in F. tularensis we first assessed the highly expressed sRNAs by polyacrylamide gel electrophoresis followed by ethidium bromide staining (Figure 1). This approach has been used successfully to find novel sRNAs in S. aureus 31. We isolated total RNA from F. tularensis LVS after growth in normal complex broth in exponential growth phase (E) or in stationary phase (S), after oxidative stress (10 mM H₂O₂) or after growth at high osmolarity (Schaedler broth + 2% NaCl). We chose to assess the RNA profile under different conditions because sRNA expression in other bacteria was found to be constitutive for some sRNAs (mostly cis-encoded sRNAs), but induced under specific conditions (such as exposure to stress) for most trans-encoded sRNAs 1. However, no major differences were observed in the pattern of RNA species that were visible after PAGE (Figure 1A). Four bands ranging in size from ~100 nt to ~400 nt were visible after staining (Figure 1A; labeled 1-4). To determine the origin of these four RNA species, we eluted the RNAs from the gel and performed linker ligation, cDNA synthesis, and PCR amplification of each eluted RNA (see Methods). Sequence analysis of cloned PCR fragments identified the RNAs, although in all cases we also found contaminating rRNA and/or tRNA sequences.

Ten clones corresponding to RNA #4, the largest RNA, were sequenced and seven of these identified the RNA as the transfer messenger RNA (tmRNA; FTL_R001). This RNA is also designated SsrA and is 421 nt in size. tmRNAs are found in all bacteria and are highly expressed. The tmRNA has properties of both a tRNA and a mRNA. It participates in trans-translation, which is a reaction that transfers the translational complex to the tmRNA and ultimately leads to degradation of the polypeptide (for review, see 32).

Based on the size of RNA #2, we presumed it was the 5S rRNA (114 nt). This was confirmed after cDNA sequencing. Next, RNA #1 was identified as the 4.5S RNA component of the signal recognition particle (FTL_R0044; 108 nt). FTL_R0044 was found in one of ten clones sequenced, whereas the other nine clones were the 5S rRNA. This is not very surprising considering the very similar size of these two transcripts and the high expression of the 5S RNA. Despite several attempts, we did not obtain any cDNA clones of RNA #3, the RNA with an apparent size of ~200 nt.

As our analysis of highly expressed RNA did not result in the identification of novel small RNA transcripts, we proceeded to construct a limited cDNA library of low molecular weight RNAs. For this, we used the same approach as described for the highly expressed sRNAs. Total RNA from bacteria grown under the different conditions described above, but depleted of ribosomal RNAs (16S and 23S rRNA), was separated on a 8% denaturing polyacrylamide gel and RNAs of approximately 100-200 nt were eluted from the gel and processed the same way as for the highly expressed RNAs. The gel was divided into smaller areas before eluting the RNAs and ten clones from each area (shown in Figure 1B) were sequenced in order to identify new sRNA species. By this method, we identified two different RNAs, one of which was identified twice from the same gel piece, whereas the remaining clones contained rRNA, tRNA or other contaminating sequences.

To further confirm that the two identified candidates are authentic sRNAs, we performed Northern blotting analysis (Figure 2) using ³²P-labeled oligonucleotides as probes. As can be seen in Figure 2, a strong band was observed for each of the sRNA candidates with sizes of approximately ~70 nt and ~110 nt, respectively. We refer to these sRNAs as FtrA and FtrB for

F

t

R

To map the ends of the RNAs we performed 5'- and 3'-RACE (Table 1). This revealed that FtrA is 111 nt and FtrB is 115 nt in length. The size found for FtrB is in agreement with the strong bands seen in the Northern blot, whereas it seems to correspond to one of the larger, but less intense, bands seen in the Northern blot of FtrA.

The ftrA gene is located between FTL_1319 and FTL_1320 (Table 1 and Figure 3) and is encoded on the coding strand. The flanking genes encode a transposase and a conserved hypothetical protein. FtrB is encoded on the coding strand between FTL_0035 and FTL_0036, encoding two hypothetical proteins. The predicted secondary structures of FtrA and FtrB, using the Quikfold program http://dinamelt.bioinfo.rpi.edu/quikfold.php 3334, are shown in Figure 3. Both FtrA and FtrB are predicted to be highly structured and contain several stem-loops.

To determine whether FtrA and FtrB are similar to known bacterial sRNAs, we performed BLAST to search the sequences of the non-coding RNA database http://ncrnadb.trna.ibch.poznan.pl/blast.html. No bacterial RNAs in the database showed sequence similarity to the two F. tularensis sRNAs. Sequences were also queried against the Rfam database http://rfam.sanger.ac.uk/, which did not result in any matches. Furthermore, a BLASTN search of the genes encoding either sRNA against the NCBI total sequence database showed that the genes are only found in Francisella and not in any other bacterial genome. This strongly indicates that the two RNAs are novel bacterial sRNAs.

To study the role of the sRNAs, we proceeded by creating mutant strains carrying a chromosomal deletion of either the ftrA or ftrB gene. Sequencing and Northern blotting were respectively used to confirm deletion of these loci and the absence of their corresponding transcripts (Figure 2). Neither mutant strain exhibited any growth defects or increased sensitivity to several stress conditions (H₂O₂, ethanol, low pH, NaCl; data not shown), showing that deleting either ftrA or ftrB does not affect the ability of F. tularensis to survive a number of stresses. Likewise, we did not observe any change in expression of either FtrA or FtrB after exposure to oxidative or osmolarity stress (data not shown).

We next tested if either gene plays a role in F. tularensis intracellular multiplication. As shown in Figure 4, both mutant strains multiplied intracellularly in J774 murine macrophage-like cells in a manner indistinguishable from the wild-type strain, indicating that neither FtrA nor FtrB is required for multiplication in macrophages in vitro.

To examine if FtrA or FtrB is required for virulence of F. tularensis, we assessed the ability of each mutant strain to induce disease in the mouse model of infection. No major difference in survival was observed between mice infected with wild-type LVS or with either of the mutant strains, showing that neither gene is required for virulence (data not shown). In addition, after infecting mice with a 1:1 mixture of mutant and wild-type bacteria we recovered equal numbers of wild-type and mutant bacteria from the spleen (data not shown). This further demonstrates that deleting ftrA or frtB does not affect F. tularensis pathogenicity in the mouse model.

It is important to recall that most sRNAs affect gene expression negatively. Deleting a specific sRNA gene and thus alleviating the repressive effect may not be of major consequence for the bacterium. Our results from in vitro characterization of the mutant strains therefore do not rule out the possibility that FtrA or FtrB may indeed control functions related to growth, stress resistance or virulence.

Most characterized sRNAs control expression of their target genes by base-pairing with a target mRNA. In addition to affecting translation, base-pairing between a sRNA and its target frequently leads to changed stability of both sRNA and mRNA. Therefore, to experimentally identify potential targets for regulation by FtrA and FtrB, we compared the transcriptomes of LVS wild-type bacteria grown in regular broth to that of either the ftrA or the ftrB mutant. This way we identified several potential targets for FtrA and FtrB (Table 2). None of the potential targets identified by microarray analysis are located in the same genomic region as the sRNAs, indicating that the sRNAs are trans-acting. The analysis identified four genes for which the transcript was found at a higher level in the ftrA mutant than in LVS. These are FTL_0045 (encoding orotidine 5'-phosphate decarboxylase, PyrF), FTL_0207 (encoding pyrrolidone carboxylate peptidase, Pcp), FTL_0902 (encoding an oxidoreductase), and FTL_1922 (encoding a YggT family protein). Interestingly, pyrF transcripts were also found at higher levels in a hfq mutant 30. Thus, deleting either hfq or ftrA affects pyrF mRNA levels in the same manner, in agreement with a regulatory mechanism in which Hfq-mediated FtrA-pyrF mRNA base-pairing leads to degradation of the target mRNA. This, however, needs to be demonstrated experimentally. The mRNA levels of five genes were changed in the ftrB mutant compared to wild-type bacteria: two were found in higher levels (FTL_0836 and FTL_1754) and three at lower levels (FTL_0324, FTL_0421 and FTL_1966) in the ftrB mutant. Of these, FTL_0421 (encoding a lipoprotein) was also found to be down-regulated in the F. tularensis hfq mutant 30. This finding could be explained by a mechanism in which RNA duplex formation leads to stabilization of the mRNA. To confirm the results obtained by microarray analysis, we chose two genes with changed expression in either mutant strain and performed quantitative RT-PCR analysis. For each mutant strain, this analysis confirmed the change in expression for both genes (see Additional File 1). Presently we do not know if FtrA and FtrB regulate these targets directly as it is also possible that the observed changes in mRNA level are due to indirect effects of deleting the sRNA.

While our cloning based screen proved effective in identifying previously unknown sRNAs, the relatively small number of clones sequenced meant that only abundant small transcripts could be found using this approach. In an effort to identify putative sRNAs in F. tularensis that may have been missed in our physical screen, we used the bioinformatic tool SIPHT 35 to conduct a genome-wide bioinfomatic screen for candidate sRNAs in F. tularensis LVS. SIPHT identifies candidate intergenic loci based on the co-localization of intergenic conservation and Rho-independent terminators. It then annotates these candidates for several features, including the similarity of their primary sequence and predicted secondary structure to intergenic sequences in other bacterial replicons, their proximity to the putative binding sites of various transcription factors, and their homology to previously confirmed or predicted sRNAs and cis-regulatory RNA elements.

A total of 24 candidate loci were identified by SIPHT, including three intergenic regions (IGRs) each containing two overlapping candidate loci, leading to a total of 21 IGRs (Table 3 and Additional File 2). Importantly, both the FtrA and FtrB sRNA were identified in the bioinformatic prediction (Table 3, candidate 19 (FtrA) and candidate 14 (FtrB)), suggesting that other candidate loci identified by SIPHT may represent real non-coding RNAs. One of these candidates was annotated by SIPHT as sharing sequence homology with SprB, a previously identified but uncharacterized sRNA in Staphylococcus aureus 31, while another was found in a similar genomic location based on its flanking genes to P26, an sRNA of unknown function in Pseudomonas aeruginosa. For fourteen of the loci identified, including FtrA and FtrB, conservation of secondary structure was predicted by QRNA 36. Putative sigma-70 promoters were identified within 400 nucleotides upstream of the predicted terminators of 9 candidates, including two upstream of ftrA, one of which in a position consistent with the larger band seen by Northern analysis and the results of the 5'-RACE.

Neither tmRNA nor 4.5S RNA were identified by SIPHT. Further analysis revealed that these loci were missed because they were not associated with predicted Rho-independent terminators. This may reflect a true paucity of Rho-independent terminators in F. tularensis or limitations in the ability of the three terminator predicting algorithms used by SIPHT to identify terminators in an AT rich genomes such as that of F. tularensis. The fact that these two sRNAs were not identified by SIPHT suggests that other F. tularensis sRNAs were likely missed in our screen.

Several characteristics of Francisella likely hinder predictions of sRNAs based on the proximity of intergenic sequence conservation and Rho-independent terminators. First, no genome sequences for close relatives of F. tularensis are available for BLAST comparisons and the relatively low GC content (32%) of the F. tularensis genome leads to spurious hits in BLAST searches that are more likely due to low sequence complexity than to actual sequence conservation. Second, as described above, it is also likely that existing terminator prediction programs, most of which had been trained on E. coli, are less effective in identifying terminators in AT-rich species such as Francisella. Indeed, in a kingdom-wide search for putative sRNAs in over 500 different strains, all Francisella sp. were among the 20 strains with the lowest density of putative intergenic Rho-independent terminators. However, while these factors likely contribute to both the decreased sensitivity and specificity of our sRNA predictions, the fact that both FtrA and FtrB were identified in our in silico screen suggests that the predictive algorithm used by SIPHT is effective in identifying novel sRNAs even in species such as Francisella that are AT-rich and have relatively few sequenced relatives.

Recently, a number of small ORFs encoding small peptides (<50 amino acids), previously non-annotated, have been identified in other bacterial species 3738. We therefore examined whether the predicted sRNA loci in F. tularensis contain ORFs encoding small peptides (Table 3). A few loci could encode small peptides of 24-52 amino acids, but most loci contained none or only very short ORFs, further supporting the hypothesis that the majority of the loci encode putative RNAs.

All strains and plasmids used in this study are listed in Table 4. F. tularensis LVS was grown in Schaedler broth supplemented with vitamin K3 (Biomérieux, France), on chocolate agar plates supplemented with IsoVitalex vitamin solution (Biomérieux, France), or in defined medium 39. Escherichia coli was grown in LB. When needed, media was supplemented with kanamycin (50 μg ml^-1 for E. coli and 10 μg ml^-1 for F. tularensis). Oligonucleotides used are described in Table 5.

Candidate sRNAs were predicted using the SIPHT program, as described previously 35. Thresholds for BLAST E value, RNAMotif score, FindTerm score, and TransTerm confidence were set to 5e-3, -6, -10, and 86, respectively. All other parameters were set to default values.

LVS was grown to mid-exponential phase or to stationary phase in Schaedler-K3 before harvesting the bacteria. For oxidative stress experiments, bacteria at OD₆₀₀ = 0.3 were exposed to 10 mM H₂O₂ for 30 min before harvested. For experiments at high osmolarity, LVS was grown in complex medium containing 2% NaCl to OD₆₀₀ = 0.3 before being harvested. Total RNA was extracted using Trizol reagent (Invitrogen) according to the protocol provided by the manufacturer and treated with Turbo DNase (Ambion). For cDNA library construction, total RNA was depleted of 16S and 23S rRNAs using the MICROBExpress kit (Ambion) according to the protocol provided by the manufacturer before loaded on gel.

20 μg of total RNA was heated and loaded onto an 8% polyacrylamide/8M urea gel with the RNA Century Plus (Ambion) as size marker. After separation, the gel was stained with ethidium bromide and a gel piece containing RNA of a certain size was cut out. RNA was eluted from the gel after crushing and incubation at 37°C in 0.5 M ammonium acetate and 1 mM EDTA (pH 8.0) for 6 hours under constant agitation. Supernatants were extracted once with chloroform before precipitation of the RNA with isopropanol in the presence of glycogen. Next, 3' adapter was ligated to the dissolved RNA with T4 RNA ligase in the presence of DMSO and RNase inhibitor for 90 min at 37°C. The RNA was purified from a 15% polyacrylamide/urea gel as described above to remove non-incorporated adapters. The eluted RNA was treated with tobacco acid phosphatase (TAP) before ligation of 5' adapter (using the same method as for 3' adapters) and subsequently non-incorporated adapters removed by passing and twice washing the RNA on a Microcon YM-30 (Millipore) column. The RNA was reverse transcribed using the 3'PRIMER and SuperScript II reverse transcriptase (Invitrogen), PCR amplified using 5'PRIMER and 3'PRIMER and ligated into pCR2.1-TOPO. All RNAs were identified using this protocol, except the tmRNA for which we used a protocol that ligates the 3'adapter before the 5'adapter and utilizes different adapters (adapters B; see Table 5).

Total RNA was extracted using Trizol according to the manufacturer's instructions (Invitrogen) and quantified on a NanoDrop ND-1000 Spectrophotometer (Thermo Scientific). After extraction, 10 μg of total RNA were loaded on a polyacrylamide gel (8% acrylamide/8M urea). After migration, the RNA was transferred to a Hybond-N+ membrane (Amersham) and crosslinked with UV light. The membrane was prehybridized in Rapid-Hyb Buffer (Amersham). Then, ³²P-labelled gene-specific probe (oligonucleotide, see Table 5) was added directly to the prehybridization buffer with the membrane and incubated for 16 hours at 42°C. After hybridization, the membrane was washed twice with 2×SSC/0.1% SDS, once with 1×SSC/0.1% SDS and twice with 0.1×SSC/0.1% SDS. Results were analyzed on a Storm 860 PhosphorImager (Molecular Dynamics) using the ImageQuant software (Molecular Dynamics).

For 5'-RACE, 15 μg of total RNA was incubated with TAP for 2 hours at 37°C after which 5' adapter was ligated with RNA ligase. RNA was purified from gel and reverse transcribed with a gene specific primer and SuperScript II RT. cDNA was then used as a template in PCR reaction with the 5'PRIMER and the gene specific primer (GSP_5'RACE) and cloned into pCR2.1-TOPO before sequencing.

3'-RACE was performed by ligating the 3' adapter to total RNA, followed by gel purification of adapter ligated RNA, reverse transcription with 3'PRIMER and finally conducting PCR with 3'PRIMER and GSP_3'RACE. The 3' end of each sRNA was determined by sequencing of cloned PCR products in pCR2.1-TOPO.

Regions of approximately 1 kb upstream and downstream of ftrA and ftrB were amplified by PCR using primer pairs DelR/DelB and DelO/DelT. The upstream and downstream fragments were purified from gel, annealed and extended in 20 cycles of PCR without primers and the product further used as template in a PCR reaction with primers DelR and DelT. The ~2 kb PCR products were cloned into pCR2.1-TOPO (Invitrogen) and mobilized as a NotI-SacI fragment into the sacB based suicide vector pMP812 40 creating pMP-ΔftrA and pMP-ΔftrB. pMP-ΔftrA and pMP-ΔftrB were introduced in LVS by electroporation and integration of the plasmid into the chromosome confirmed by PCR. Strains were then passed once in medium without selection, subsequently streaked on solid medium containing 6% sucrose and isolated colonies were tested for loss of the gene by PCR (using primer pairs DelCheck1/DelCheck4 and DelCheck2/DelCheck3). Deletion of the gene was confirmed by sequencing.

RNA used in microarray experiments was extracted using Trizol reagent combined with purification of the aqueous phase on RNeasy columns (Qiagen). cDNA labeling and microarray hybridizations were performed as described 30. Three independent experiments and RNA extractions were performed and each set of RNAs was used in one hybridization experiment. The F. tularensis microarrays (obtained from the "Pathogen Functional Genomics Resource Center", PFGRC) contain 70-mer oligonucleotides, in five copies, representing all genes of strains SchuS4 and LVS. Microarrays were scanned with a Genepix 4000B scanner (Molecular Devices). To quantify signal fluorescence intensities, TIFF images were analyzed using the Genepix Pro 6.0 software. Statistical analyses were performed using publicly available software, the R/Bioconductor package LIMMA (available from http://www.bioconductor.org). A list of statistically significant differentially expressed genes was obtained using lowess normalization (after inspection of MA plots) and applying the empirical Bayes moderated t-test. Microarray data are available at ArrayExpress.

To validate the microarray results, two ORFs were selected (for each mutant strain) for real-time quantitative reverse transcription PCR (qRT-PCR) analysis with the same RNA samples used in the microarray hybridizations. For the analyses, one microgram of RNA was reverse transcribed using random hexamers and SuperScript II reverse transcriptase (Invitrogen) according to the protocol provided by the manufacturer. Real-time PCR was performed with gene-specific primers using an ABI PRISM 7700 and FastStart SYBR master mix (Roche Diagnostics). To calculate the amount of gene-specific transcript, a standard curve was plotted for each primer set using a series of diluted genomic DNA from LVS. To compare the transcript amounts in the different strains, the amounts of each gene transcript was normalized to DNA helicase (FTL_1656) as this gene has been shown to change little in expression during growth 23. The expression of each gene was determined from three replicates in a single qRT-PCR experiment.

J774 cells were propagated in RPMI or Dulbecco's Modified Eagle's Medium (DMEM) medium containing 10% fetal calf serum. Cells were seeded at a concentration of ~2 × 10⁵ cells per well in 12-well tissue plates (Falcon) and monolayers were used 24 hours after seeding. J774 macrophage monolayers were incubated for 90 min at 37°C with the bacterial suspensions (approximate multiplicities of infection 100) to allow the bacteria to enter. After washing (time zero of the kinetic analysis), the cells were incubated in fresh culture medium containing gentamicin (10 μg ml^-1) to kill extracellular bacteria. At several time-points, cells were washed three times in PBS and processed for counting of surviving intracellular bacteria. For this, bacteria were recovered by lysis of macrophages with distilled water and the titer of viable bacteria released from the cells was determined by spreading preparations on agar plates. For each strain and time in an experiment, the assay was performed in triplicate. Each experiment was independently repeated two times and the data presented are from one experiment.

LVS and mutant strains were grown in Schaedler-K3 to exponential growth phase and diluted to the appropriate concentrations. 6 to 8-week-old female BALB/c mice (Janvier, Le Genest St Isle, France) were i.p. inoculated with 200 μl of bacterial suspension. Groups of five mice were inoculated with various doses of bacteria (approximately 10¹ to 10⁴ bacteria) and the mortality was followed for 9 days. The actual number of viable bacteria in the inoculum was determined by plating dilutions of the bacterial suspension on chocolate plates. For competitive infections, LVS and mutant bacteria were mixed in 1:1 ratio and a total of approximately 400 bacteria were used for infection of five mice. After four days, mice were sacrificed. Homogenized spleen tissue from the five mice in one experiment were mixed, diluted and spread onto chocolate agar plates. PCR to distinguish wild-type and mutant bacteria were performed on 100 colonies. Animal experiments were performed according to the INSERM guidelines for laboratory animals' husbandry.