RNase J depletion leads to massive changes in mRNA abundance in Helicobacter pylori

ABSTRACT

Degradation of RNA as an intermediate message between genes and corresponding proteins is important for rapid attenuation of gene expression and maintenance of cellular homeostasis. This process is controlled by ribonucleases that have different target specificities. In the bacterial pathogen Helicobacter pylori, an exo- and endoribonuclease RNase J is essential for growth. To explore the role of RNase J in H. pylori, we identified its putative targets at a global scale using next generation RNA sequencing. We found that strong depletion for RNase J led to a massive increase in the steady-state levels of non-rRNAs. mRNAs and RNAs antisense to open reading frames were most affected with over 80% increased more than 2-fold. Non-coding RNAs expressed in the intergenic regions were much less affected by RNase J depletion. Northern blotting of selected messenger and non-coding RNAs validated these results. Globally, our data suggest that RNase J of H. pylori is a major RNase involved in degradation of most cellular RNAs.

Introduction

The level of gene expression in bacteria is usually correlated with the steady-state level of its mRNA; which, in turn, is controlled by the transcription rate and efficiency of RNA degradation. Initially thought to be invariable between bacterial species, the machinery of RNA degradation in fact varies significantly. Thus major RNA-degrading enzymes are not conserved between 2 organisms used as paradigms of the Gram-negative (Escherichia coli) and Gram-positive (Bacillus subtilis) bacteria. Whereas the major ribonuclease in E. coli is RNase E,¹ the two most important RNases of B. subtilis are RNase J1 and RNase Y.^2,3 RNase E is a single-strand-specific endoribonuclease, initiating decay of most E. coli mRNAs by internal cleavage of phosphodiester bond. Generated RNA fragments are further degraded by a number of redundant 3′–5′-exoribonucleases.¹ RNase Y was also shown to be an endoribonuclease and is thought to initiate the mRNA degradation by internal mRNA cleavage.^3,4 RNase J enzymes are exoribonucleases that progressively remove single nucleotides from the RNA polymer. The remarkable property of RNases J is the 5′–3′ direction of the RNA degradation.^5,6 Whereas several 3′–5′-directional RNases were characterized in bacteria,⁷ RNase J is the only example of a bacterial RNase with 5′–3′-directionality. RNase J1 of B. subtiltis was also initially suggested to have an endonucleolytic activity ² with relaxed nucleotide specificity in vitro.⁸ However in vivo determinants for exo- or endonucleolytic cleavage remain unknown.

Analysis of transcriptomes of RNase J1-depleted B. subtilis strains was performed using gene microarrays ⁹ and tiling microarrays.⁴ It revealed relatively few changes in RNA abundance upon RNase J1 depletion. About 8% of mRNAs were increased or decreased in the first study probably due to the partial depletion of RNase J1.⁹ Partial depletion of RNase J1 was chosen to maintain the same growth rate upon depletion, avoiding most indirect effects due to the growth rate changes. However, under this condition, lower affinity RNA substrates have probably been missed. About 30% of transcripts had increased abundance in the other approach where RNase J1 depletion was more complete.⁴ In this case growth was slowed down significantly upon RNase J1 depletion and this approach probably allowed identification of most RNase J1 targets.

Our group investigates how mRNA stability is controlled in the gastric pathogen Helicobacter pylori, a topic that has been poorly explored. H. pylori colonises the human stomach and causes chronic gastritis, peptic ulcer disease and is associated with the development of gastric cancer. Infection by H. pylori is responsible for about 800,000 deaths in the world every year. H. pylori has no known niches outside the human stomach and in agreement with this it has a small genome and reduced redundancy of enzymatic activities in particular few transcriptional regulators. Its RNA metabolism is also governed by a reduced number of known conserved ribonucleases compared to E. coli or B. subtilis (8 in H. pylori versus 19 in either E. coli or B. subtilis ¹⁰) in correlation with the smaller size of its genome (1.6 Mb vs over 4.6 and 4.2 Mb for E. coli or B. subtilis, respectively). RNase E is absent from the H. pylori genome whereas RNases J and Y are conserved.¹¹ Remarkably, of 8 conserved RNases, RNase J and RNase P (involved in tRNA maturation) are the only essential ribonucleases of H. pylori (our unpublished data). We previously characterized H. pylori RNase J ¹¹ and found that its role is different in H. pylori compared to its homolog from B. subtilis, as it does not mature rRNA.¹² Moreover, it forms a minimal degradosome complex with the only DExD-box RNA helicase of H. pylori, RhpA.¹¹ This interaction leads to mutual activation of both enzymes, and results in efficient RNA degradation to completion. We also found that the complex is associated with translating ribosomes, which suggests that RNA degradation in H. pylori might be coupled with translation. Translation-RNA degradation coupling was recently demonstrated in yeast,^13,14 but was not yet directly shown in bacteria.

Here, we investigated the global role of RNase J in RNA metabolism of H. pylori. To find potential targets of RNase J we employed RNA deep sequencing and differential expression analysis of strains that are strongly depleted for RNase J. Exhaustive RNA deep sequencing analysis of the H. pylori transcriptome was recently performed and transcription start sites for coding and non-coding RNAs were determined using differential RNA-seq approach.¹⁵ This analysis was done for the strain 26695 that is widely used as a model strain of H. pylori. However, this strain does not support plasmid maintenance, therefore, we analyzed RNase J impact on the H. pylori transcriptome using strain B128 that does not have these constraints.¹⁶ We discovered that depletion for RNase J leads to massive increase in abundances of mRNAs and antisense RNAs (asRNA). Non-coding RNAs (ncRNAs) expressed from intergenic regions are less dependent on the level of RNase J expression than messenger or antisense RNAs. We discuss implications of RNase J depletion for H. pylori metabolism.

Results

Establishing conditions for the transcriptome analysis of H. pylori strain depleted for RNase J

We previously showed that RNase J is essential for normal growth of H. pylori and constructed two conditional mutants derived from strain B128 ¹⁶ that allowed RNase J depletion.¹¹ In these mutants, the chromosomal copy of RNase J-encoding gene (rnj) was deleted and RNase J was expressed from a plasmid under control of an IPTG-inducible promoter. Either full-length RNase J or its shorter variant lacking a non-conserved N-terminal extension of 132 amino acids, ΔN-RNase J, were expressed in strains UPH738 and UPH739, respectively (Fig. 1A and ¹¹). We previously reported that the deletion of the N-terminal extension significantly decreased the activity of the enzyme in vitro and its stability in vivo.¹¹ Therefore, in the absence of induction, depletion of the full-length RNase J was approximately 3-4-fold and H. pylori growth was only slightly impaired (Fig. 1B). In contrast, the ΔN-RNase J UPH739 mutant presented a strong depletion for both RNase J protein and activity in the absence of the inducer and, accordingly, its growth rate was greatly reduced (Fig. 1C).

Figure 1.

(A). Genetic organization of the H. pylori strain used for the depletion of RNase J. Kan – kanamycin resistance cassette replacing hp1430 gene coding for RNase J. pILL2157 plasmid was bearing either full-length RNase J or its truncated variant lacking the N-terminal 132 amino-acids. (B). Growth curves and Western blots with RNase J-specific antibodies for H. pylori strains depleted for full length RNase J. Cultures for RNA preparation were collected at the OD600 corresponding approximately to 0.5 (dotted line). Coomassie staining was used to control protein loading (panels below Western blots). (C). Same as B for strain expressing ΔN-RNase J.

To evaluate the global impact of RNase J on RNA metabolism in H. pylori and to find potential RNase J targets, we performed total RNA sequencing for strain UPH739 depleted for RNase J (ΔN-RNase J – IPTG). In this strain, strong depletion of RNase J activity allows the identification of maximum putative bona fide RNase J targets. For comparison we chose the wild type strain bearing the empty vector pPH85 ¹¹ and not the wild type to avoid possible effects of plasmid maintenance on transcriptome. We also did not choose the ΔN-RNase J + IPTG condition for comparison as amounts and activity of the truncated RNase J under that condition was lower than that of the wild type strain. Two biological replicas for each condition were analyzed. All sequencing results described above are available at the MicroScope platform at the following web site (http://www.genoscope.cns.fr/agc/microscope/transcriptomic/NGSProjectRNAseq.php).¹⁷

Deep-sequencing of total RNA of RNase J-depleted mutants, annotation and analysis of non-coding RNAs

Prior to the sequencing library construction, total RNA preparations were depleted for rRNA (see methods). Whereas more than 98% of 5S and 16S rRNAs were removed by this procedure, only about 65% of 23S rRNA was eliminated. After rRNA depletion, total RNA was used to construct cDNA libraries that were sequenced with a Solexa sequencer (Illumina). Table 1 shows for each condition, the numbers of reads aligned on the genomic sequence and plasmid of strain B128. The estimation of non-rRNA reads mapped to the genome was suggestive of an overall increase in non-rRNA levels in RNase J-depleted cells.

Table 1.

RNA sequencing results for RNase J depleted condition and the wild type.

Strain	Sequenced condition	Total read number	Aligned reads	Non-rRNA reads
WT B128+pPH85	WT + pPH85_1	85 935 037	98.91%	13.57%
WT B128+pPH85	WT + pPH85_2	88 036 875	98.71%	16.56%
UPH739	Pi-ΔN-rnj - IPTG_1	68 759 853	98.45%	23.79%
UPH739	Pi-ΔN-rnj - IPTG_2	67 516 620	98.27%	24.60%

The reads were mapped to the H. pylori genome of strain B8 (NC_014256.1),¹⁸ which is strain B128 ¹⁶ isolated after gerbil infection. Published annotation file (gff3) for strain B8 was used for the annotation of genomic objects (coding sequences, CDS; fragmented CDS, fCDS; rRNAs and tRNAs) using the MicroScope platform.¹⁷ Non-coding RNAs were not annotated in this strain therefore we annotated them manually. It should be noted that RNA degradation products intrinsically present in bacterial cells are also detected by RNA sequencing, therefore, the length of the annotated ncRNAs does not necessarily correspond to the actual transcript length but also includes cleaved extensions.

Non-coding RNAs found in intergenic regions were designated ncRNA_IGR_N_ORF; those expressed as antisense to open reading frames (ORFs) were designated asRNA_N_ORF. All of them were designated miscRNA_N in the order of their occurrence on the chromosome starting from the origin of replication upon creation of common gff file. Coordinates of non-coding RNAs were integrated in the MicroScope interface (¹⁷ and website). MicroScope tools allow fast calculation of read coverage for both coding and non-coding transcripts representing their expression levels. This facilitates differential gene expression analysis. Actual coverage plots can be visualized with MicroScope as it integrates the IGV (Integrative genomics viewer) browser.¹⁹

The first striking observation was the low level of expression of antisense RNAs compared to the corresponding ORFs. This contrasts with previously published transcriptome analysis of strain 26695 of H. pylori.¹⁵ We attribute this disparity to differences in library preparation protocols and sequencing approaches. In our analysis total RNAs were fragmented before library preparation and sequenced by Illumina/Solexa with average read length of 50 nt, whereas in Sharma et al ¹⁵ (keep super) approach RNAs were not fragmented upon library preparation and Roche/454 sequencing technology was used that allows sequencing of up to 350 nt. We believe that the sequencing approach used in our case Therefore, our approach resulted in a much more complete coverage of each genomic object. However, despite differences in relative expression of sense and antisense RNAs, the total number of asRNAs estimated with our approach was 665 (40%) per 1651 of annotated genes and their fragments (Table 2). This value was similar to the estimate number of asRNAs in Sharma et al (46%). We also detected 43 non-coding RNAs in intergenic regions, 10 of which were relatively abundant.

Table 2.

Number of transcripts differentially expressed in the RNase J-depleted strain.

Genomic Object (GO) type	GO subtype	Total N	N of GO with Log2(FC)≥1 and FDR≤0.05	N of GO with Log2(FC)≥2 and FDR≤0.05
CDS	Gene	1517	1296 (85%)	828 (55%)
fCDS	Gene fragment	134	98 (73%)	54 (40%)
misc_RNA	asRNA	665	519 (78%)	323 (49%)
misc_RNA	nc_RNA_IGR	43	26 (60%)	5 (12%)
misc_RNA	oRNA*	2	1	1
Total		2361	1939	1211

^*oRNA include RNAI and tmRNA with respective Log2(FC) of 3,31 and −0,83 (Table S2).

Conventional normalization strategy vs. massive changes in RNA abundance

We noticed that 23S rRNA depletion during library preparation was unequal between samples, and this generated variations of ribosomal to non-rRNA ratios. At the same time, we observed that RNase J depletion led to an important increase in steady-state levels of non-rRNA pool. As library preparations were normalized to the total input RNA and rRNAs were significantly reduced after their depletion, this led to a distortion of proportions of rRNAs to non-rRNAs in samples relative to each other. The standard DEseq algorithm ²⁰ used for normalization and comparison of RNA sequencing data was therefore not suited for the analysis of our results as it allows normalization of samples for which most genes are not differentially expressed. Therefore, we chose to normalize sequencing data using the DEseq algorithm applied to a subset of stable reference RNAs with validated unchanged abundances in analyzed samples.

As reference RNAs, we chose tRNAs that are highly structured and thus should not be susceptible to RNase J degradation. Indeed, Northern blotting for tRNAs-His, -Asp and -Phe confirmed virtually unchanged levels of these tRNAs under used conditions (Fig. S1). This confirmed that highly structured RNAs are resistant to RNase J-catalyzed degradation in H. pylori, which is consistent with previously observed properties of the RNase J ortholog from B. subtilis.⁵

Thus, we applied the DEseq algorithm to this subset of reference tRNAs (Table S1). It showed relatively few changes in depth ratios of selected RNAs. Calculated library size factors were used further for normalization.

Analysis of differential expression of genes and non-coding RNAs under conditions of RNase J depletion

Control of size factors used for normalization

The reference RNA coverage obtained after library normalization using calculated size factors for tRNAs were plotted to generate linear regression analyses (Fig. S2). They show that the normalization minimized biases both between replicas of the same experimental conditions (Fig S2 A, B) and experimental conditions (Fig. S2 C), as for all comparisons we obtained regression lines close to equation y = x with very weak dispersion (R˜0.95, pval < 0.05).

Differential expression analysis

To evaluate the differential expression of the H. pylori genome upon RNase J depletion, we compared normalized sequencing data for ΔN-rnj - IPTG and WT+pPH85. Strikingly, severe RNase J depletion led to a massive increase in steady-state levels of most transcripts of the H. pylori transcriptome. Table 2 shows numbers and percentage of different types of transcripts with 2- and 4-fold increase in abundances (Log2(FC)≥1 and Log2(FC)≥2; FDR≤0.05) (FC = fold change, FDR = false discovery rate). More than 80% mRNAs and are increased 2-fold upon RNase J depletion, whereas only a minority (7) of these transcripts are significantly unchanged or decreased (FC≤1.5; FDR≤0.05). These include mRNAs of hypothetical proteins HPB8_1195, HPB8_1111, HPB8_926 (FC<1), HPB8_1321 (FC<1.5) two fragments of hypothetical proteins HPB8_477 and the adjacent HELP3_0459 (FC<1) and one asRNA (HPB8_misc_RNA_4, antisense to a fCDS of a hypothetical protein HPB8_11; FC<1). It is of note that 8% of CDSs and 20% of fCDS and asRNA had the FDR above 0.05 and therefore were not taken into account in this analysis. These might include unchanged or decreased transcripts. Interestingly, a much smaller proportion of intergenic ncRNAs was increased 2-fold (60%, Table 2) and 7% of these RNAs were decreased. The analysis using a cut-off of 4 (Log2(FC)≥2 and FDR≤0.05) showed that about half of mRNAs and only 12% of intergenic ncRNAs have a 4-fold increase in abundances upon RNase J depletion.

Validation of direct mRNA targets by Northern blotting

The observed increase in the level of most transcripts upon RNase J depletion may either be due to direct RNA stabilization as a result of reduced RNA decay or to an indirect positive effect on transcription. To distinguish between these two possibilities, RNA decay after transcription arrest with RNA polymerase inhibitor rifampicin was followed by Northern blotting with specific probes. We performed this analysis for selected coding and non-coding RNAs under conditions used for sequencing. Conditions of partial RNase J depletion for the strain UPH739 grown with inducer (Pi-ΔN-rnj+IPTG) and strain UPH738 expressing full length RNase J (Pi-rnj±IPTG) were added for comparison.

We first chose to explore whether genes encoding major H. pylori virulence factors such as urease and CagA are direct RNase J targets. Urease allows H. pylori to resist to high acidity that it encounters in the stomach by hydrolyzing urea and generating the buffering compounds CO₂ and NH₃. Urease consists of two structural proteins (UreA and UreB) encoded by the 2 genes of the first urease operon (Fig. 2A). This operon is followed by a second one composed of the ureI gene encoding a urea channel and 4 genes that are involved in incorporation of catalytic nickel ions at the urease active center.^21,22 Regulation of stability of ureIE bicistronic transcript involves RNase J as depletion of RNase J leads to significant stabilization of the transcript (¹¹ and Fig. 2B). We investigated whether other genes of the urease operon are also post-transcriptionally regulated by RNase J. ureA and ureB are transcribed as one transcriptional unit present mostly as a 2.7 kb mRNA (Fig. 2). Differential expression analysis for this transcript (Pi-ΔN-rnj-IPTG versus WT+pPH85 conditions) was inconclusive due to high FDR value (Table S2). Northern blotting showed that this transcript did not change in abundance and half-life upon depletion of the full length RNase J. In contrast, its half-life was significantly increased after rifampicin treatment of ΔN-RNase J-depleted cells (Fig. 2B). This suggests that RNase J is indeed involved in the decay of this transcript. Other transcripts of urease operon have increased abundance at least 4-fold upon RNase J depletion with low FDR values (Table S2). This was largely confirmed by Northern blotting (Fig. 2B). Moreover, rifampicin treatment of RNase J-depleted cells led to a significant increase in half-lives of all mRNAs of accessory genes of urease operon (Fig. 2B). These results strongly suggest that RNase J regulates stability of all urease transcripts. Interestingly, the H. pylori strains expressing ΔN-RNase J (Fig. 2B, last 2 panels) showed the accumulation of short RNA fragments that were also more stable upon RNase J depletion.

Figure 2.

(A). Genetic organization of the urease operon and corresponding RNAs detected by Northern blotting. Sizes of ORFs in base pairs are presented above gene arrows. Riboprobe positions are indicated as arrows above genes. (B). Northern blots revealing transcripts of urease operon after rifampicin treatment of RNase J-and ΔN-RNase J depleted cells. Approximate transcript half lives are indicated under each Northern blot panel. rRNAs were stained with Methylene blue before probe hybridization and were used as control of RNA loading (lower panels under each Northern blot). Images representing ureIE Northern blots are those published previously ¹¹ and are used here for complete analysis of urease operons. ureG riboprobe had a cross-hybridization with a non-specific possibly rRNA (upper band at the ureG Northern blot).

Cytotoxin-associated gene (CagA) is another important virulence factor of H. pylori. It is encoded by the Cag-pathogenicity island and its injection into the host gastric epithelial cells upon H. pylori colonization is associated with cell hyperproliferation and gastric adenocarcinoma, implicating this molecule as a bacterial oncoprotein.²³ We obtained a non-conclusive result of the differential expression analysis for cagA (Table S2, see high FDR). Northern blotting reveals a non-significant change in the abundance of a full-length 3.6 kb transcript (transcript I in Fig. 3) between the wild type and the strain with depleted ΔN-RNase J that was used for the differential expression analysis. However, smaller fragments of cagA mRNA were detected with riboprobe hybridizing to the 5’-end of cagA transcript. These fragments are significantly increased in abundance and stabilized upon rifampicin treatment (transcripts II-IV Fig. 3B). They may correspond to products of either a secondary cleavage by another RNase (i.e. RNase Y) or transcriptional arrest that normally are degraded by RNase J. As for transcripts of urease operon, these shorter transcripts were particularly abundant in the cells that expressed truncated RNase J probably due to a better depletion for RNase J activity compared to the cells depleted for the full-length protein. Interestingly, the full-length transcript of cagA seems to increase in abundance upon partial depletion of full-length RNase J despite little change in its half-life. Altogether, these results suggest that RNase J is also involved in the degradation of the cagA transcript and its fragments.

Figure 3.

(A). Transcriptional organization of cagA. Position of riboprobe used to detect cagA mRNA and its fragments (lines under the gene arrow) is shown as an arrow. Approximate sizes of the detected full-length mRNA and its fragments are indicated. (B). Northern blots and approximate transcript half-lives of cagA mRNA and its fragments.

Effect of RNase J on the expression of non-coding RNAs

To test whether non-coding RNAs are targets of RNase J, we analyzed the most abundant ncRNAs by Northern blotting. Two were ncRNAs located in the intergenic regions that were annotated as ncRNA_IGR_189-190 (HPB8_misc_RNA_82) and ncRNA_IGR_1019-1020 (HPB8_misc_RNA_448). The three most abundant asRNAs are asRNA_386 (HPB8_misc_RNA_175), asRNA_671 (HPB8_misc_RNA_313) and asRNA_1162 (HPB8_misc_RNA_507; named ncHP5490 in Sharma et al ¹⁵ and RepG in Pernitzsch et al ²⁴). It is noteworthy that our classification of the ncRNA in asRNA and intergenic ncRNAs is conditional and that the asRNA_1162 (RepG) was reported to act in trans as a regulatory sRNA targeting a chemotaxis receptor mRNA of TlpB.²⁴

Each of these transcripts either decreases or remains unchanged upon depletion of truncated RNase J (Fig. 4). Rifampicin treatment for up to 60 min has no impact on their abundance in either of the tested condition suggesting the high stability of each of these transcripts and their independence of RNase J. ncRNA_IGR_1019-1020 seems to be processed by RNase J as ΔN-RNase J depletion led to the accumulation of a one-nucleotide longer form of the ncRNA. Altogether our results suggest that RNase J is not involved in the degradation of the analyzed ncRNAs.

Figure 4.

Northern blots showing non-coding RNA in intergenic regions of hpB8_1019/hpB8_ 1020 and hpB8_0189/hpB8_0190 (ncRNA_IGR_1019_1020 and ncRNA_IGR_189_190 respectively) and antisense to hpB8_0671, hpB8_0386 and hpB8_1162 (asRNA_671, asRNA_386 and asRNA_1162 respectively) after rifampicin treatment of RNase J and ΔN-RNase J-depleted cells. The membranes were stained with methylene blue prior to probe hybridization to reveal 5S rRNA to control loading.

Discussion

Since major cellular RNases are often essential for growth, identification of their targets is challenging and requires construction of conditional knockdown mutants with regulated expression. The level of their expression, however, has to be strongly reduced as even small quantities of these enzymes are sufficient to catalyze cleavage of the most specific RNA substrates, and under these conditions the accumulation of RNAs will not be detected. Genetic tools available in H. pylori are limited, and to deplete the level of RNase J, an essential and abundant protein in H. pylori, we used an available plasmid pILL2157 allowing the regulated expression of proteins under control of a strong H. pylori promoter.^11,25 However, RNase J expression and the growth rate of the constructed mutant were reduced only partially in the absence of inducer (Fig. 1A). Under such condition only low-affinity RNase J substrates could be accumulated and detected by transcriptome analysis. Therefore, to identify the maximum of RNase J-dependent transcripts we used the same plasmid pILL2157 in which a truncated RNase J mutant was expressed under control of IPTG-inducible promoter.¹¹ We previously showed that the truncated RNase J is less active in vitro than the full-length protein, is weakly expressed even in the presence of the inducer and is undetectable in its absence.¹¹ Therefore, in the absence of the inducer the level of RNase J activity available for RNA degradation is dramatically decreased and accordingly bacterial growth rate is significantly slowed down (Fig. 1C).

Specific consequences of growth rate differences on gene expression levels in bacteria are possible and to avoid potential changes that are due to the growth rate differences it is preferable to compare conditions when growth rates are similar.²⁶ However, when an RNase is essential for growth, partial depletion of its level leads to determination of only a minority of its most specific targets.⁹ It has been shown that upon decrease in growth rate, RNA decay rates remain unchanged in model bacteria whereas transcription rates slow down.²⁶ Therefore, in such case we may underestimate RNA stabilization upon RNase depletion due to the growth rate decrease and transcription attenuation. Interestingly, unicellular eukaryotes like yeast can regulate RNA decay rates as a function of transcription rates and vise versa thus buffering mRNA levels.²⁷ No similar effects are known in bacteria.

In the present study, we found that strong depletion for RNase J leads to a massive increase in the steady-state level of the non-rRNA pool in H. pylori. About 80% of mRNAs are increased 2-fold and over 50% are 4-fold more abundant than in the wild type (Table 2). Such a rise in total mRNAs levels upon RNase J depletion is striking as only 30% mRNAs were reported to be increased 2-fold in B. subtilis upon 30-fold depletion of RNase J1 ⁴ and 20% mRNAs were 4 times more abundant in an RNase J mutant of S. aureus.²⁸ Interestingly, we observed that under our test conditions, the majority of genes are expressed in H. pylori (see full data in Table S2, normalized read counts column and Sharma et al ¹⁵). This is probably related to the low functional redundancy of H. pylori, a bacterium with a restricted habitat, the human stomach. It was reported that only half of the genes are usually expressed in rich medium in B. subtilis, a bacterium that can encounter widely varying environments.²⁹ Therefore, if we extrapolate the 30% increase in B. subtilis transcriptome to a situation in which all the genes are expressed, we obtain 60% genes dependent on RNase J1 level. This extrapolated number would be comparable to the transcriptome increase upon RNase J depletion that we report here.

Important increase in global RNA abundance upon RNase J depletion may reflect both direct and indirect effects of RNase J on RNA levels. The most probable indirect effect is the increase in expression levels of the transcriptional machinery. All subunits of RNA polymerase and both sigma factors (rpoD and rpoN genes, see Table S2) have increased abundances upon RNase J depletion. This may lead to the increase in the corresponding protein levels and, consequently, globally enhanced transcription. This does not exclude, however, possible additional transcriptome stabilization due to the lack of RNase J activity. It is likely that both direct and indirect effects take place, and this is probably what was observed in our case. This may certainly explain such dramatic increase in non-rRNA levels.

Analysis of RNA decay after transcription inhibition by rifampicin helps to discriminate between direct and indirect effects of RNase J depletion. We addressed this question by analyzing a selection of messenger and non-coding RNAs by Northern blotting. We clearly showed that the role of RNase J was direct for the analyzed mRNAs and their fragments as RNase J depletion led to their dramatic stabilization. This suggests that, even if indirect transcriptional effect may have taken place, further elimination of messengers is still slowed down when RNase J is downregulated.

We previously showed that RNase J is associated with translating ribosomes, which suggested that mRNA might be degraded co-translationally.¹¹ The observed global increase in mRNA levels upon RNase J depletion corroborates with this model and suggests a major role of this RNase in the degradation of translated RNAs.

Interestingly, estimated half-lives of analyzed mRNA were rather long even in the wild type bacteria, about 23-40 minutes. Such long half-lives have already been observed for some of H. pylori mRNAs (i.e., 36 min for napA mRNA in Barnard et al ³⁰ or urease operon transcripts at neutral or low pH in Akada et al ³¹). This differs dramatically from the E. coli or B. subtilis mRNA average half-lives (3-8 min for majority of transcripts both in E. coli and B. subtilis ^32,33). This observation positions H. pylori apart from the conventional RNA decay programs studied in these two bacterial paradigms of RNA decay.

Northern blotting of selected messenger and non-coding RNAs also validated the results of RNA sequencing and differential expression analysis. It globally confirmed the major role of RNase J in the degradation of transcripts of urease operon and cagA and a minor role in the degradation of ncRNA. It also revealed the accumulation of the full-length transcripts and probably their degradation intermediates for most analyzed mRNA upon RNase J depletion.

We found that the amounts of antisense RNAs are increased upon RNase J depletion to the same extent as mRNAs (Table 2). This suggests that RNase J does not specifically degrade antisense transcripts in H. pylori as does the nuclear exosome in yeast.³⁴ Importantly, we observed generally low expression levels of asRNAs compared to the corresponding mRNA levels, which contrasts to the previously observed relative expression levels of sense and antisense RNA.¹⁵ Differences in library preparation protocols and sequencing approaches may explain the observed discrepancies. The origin and function of antisense transcripts in H. pylori still have to be determined.

Our analysis shows that H. pylori RNase J does not catalyze degradation of highly structured non-coding RNAs. This may be due to the RNA secondary or tertiary structure prohibitive for RNase J degradation but also impeded access of RNase J to such non-translated RNAs. We took advantage of this property of RNase J to normalize sequencing libraries using highly structured tRNAs that are independent of RNase J. This approach significantly reduced distortions in rRNA and non-rRNA ratios due to variations in rRNA depletion between samples. Differential expression analysis that we performed with the limited set of reference RNAs for normalization removed the experimental biases and allowed us to reveal the global role of RNase J in the H. pylori RNA metabolism.

The observation that ncRNAs are less targeted by RNase J than mRNAs in H. pylori suggests that RNase J is not involved in sRNA-mediated regulation of gene expression as is RNase E in E. coli.³⁵ RNase Y is the best candidate for such a role and this hypothesis needs further investigation. However, the mechanisms of sRNA-mediated gene regulation may be radically different in H. pylori compared to E. coli.³⁶

Given the number of potential targets of RNase J in H. pylori, one can wonder how the targets are discriminated. The importance of tri- or mono- phosphate or a NAD modification ³⁷ at the 5’-end of mRNA for the RNase J-catalyzed degradation needs to be evaluated in H. pylori. A homolog of the pyrophosphatase RppH that cleaves 2 phosphate residues at the 5’-end of mRNA thus initiating mRNA degradation in E. coli and B. subtilis ^38,39 is present in one copy in the genome of H. pylori and is not essential for H. pylori growth (C. Sharma personal communication). Our unpublished observations suggest that RNase J of H. pylori may be independent of the 5’-end RNA status and thus of RppH activity. It is possible that RNase J of H. pylori has higher RNA-binding affinity and thus higher capacity for the internal entry than its RNase J1 homolog of B. subtilis. The role of the N-terminal extension should be investigated further in this regard.

The strong dependence of the H. pylori transcriptome on RNase J activity probably reflects particularities of the H. pylori restricted habitat, human stomach, where alimentary affluence interchanges with a harsh acidic shock. Such rapid environmental changes would necessitate fast adaptation of gene expression programs. It was previously observed that some mRNA have increased half-lives upon acid exposure of H. pylori.^30,31,40 If there were advantage of slowing down the RNA metabolism in response to acidic shock, decrease of RNase J protein level would be too slow given rather high abundance of this protein under normal conditions (our unpublished observations). However, its activity may be rapidly regulated in response to dynamic changes in environmental conditions. The impact of acidity and metal ions, for example, nickel, an important determinant of urease activity, on RNase J activity could shed light on the particularities of RNA degradation in H. pylori.

To conclude, we investigated the role of RNase J in RNA metabolism of the important human pathogen H. pylori by transcriptome analysis of RNase J-depleted cells. It revealed a pivotal role of RNase J for the RNA degradation in H. pylori as over 80% of mRNAs and asRNAs had abundances increased more than 2-fold upon RNase J depletion. ncRNAs in the intergenic regions were much less dependent on RNase J. Together, our data probably reflect the link between RNA translation and degradation, taking into account association of RNase J with translating ribosomes. Dynamic changes in the enzyme activity depending on environmental conditions may be the answer to the question how the bulk degradation of RNA in H. pylori is regulated in vivo.

Materials and methods

Strains and growth conditions

H. pylori strain B128 was used to construct mutants with controlled RNase J expression as it allows stable maintenance of the expression plasmid pILL2157 and its derivatives.^11,25 H. pylori strains (WT B128, UPH738 - Pi-rnj, UPH739 - Pi-ΔN-rnj and WT B128+pPH85) and growth conditions were as described previously.¹¹

For the preparation of total RNAs, bacterial cultures were pelleted at the exponential growth phase at OD around 0.5 (Fig. 1, dotted line) to ensure similar growth state for each condition. RNA sequencing was performed with two biological replicas of the strain UPH739 (Pi-ΔN-rnj) grown without IPTG and the other two replicas for the strain WT+pPH85 (B128 carrying the empty vector pPH85 ¹¹). Culture pellets were frozen at −80°C and used for total RNA preparation as described previously.¹¹

Sequencing library preparation

Each sample of total RNAs was treated with TURBO DNA-free™ (Ambion) before cDNA library construction to eliminate traces of contaminating genomic DNA. Enriched non-rRNAs preparations were obtained from 7 µg of total RNAs using rRNA modified capture hybridization approach from MICROBExpress™ kit (Ambion), according to manufacturer's instructions. Enriched non-rRNAs were then fragmented by using Fragmentation kit from Ambion, and purified on RNeasy MinElute columns (Qiagen). For strand-specific high-throughput sequencing, directional cDNA libraries were prepared from enriched fragmented RNAs using the TruSeq Small RNA Sample Prep Kit (Illumina). cDNA fragments of about 150 bp were purified from each library, checked on Bioanalyser (Agilent) and sequenced in single-end mode for 50 bp, in duplexes per channel, using an Illumina HiSeq2000 instrument according to manufacturer's instructions (Illumina).

Mapping, normalization and differential expression analysis

The complete transcriptomic high-throughput sequencing data were analyzed with the TAMARA bioinformatic pipeline (Cruveiller S., unpublished) currently implemented within the MicroScope platform.¹⁷ The pipeline is a “Master” shell script that launches various parts of the analysis (i.e. a collection of Shell/Perl/R scripts) and checks that all tasks are completed without error. We first assessed RNA sequencing data quality by including options, such as read-trimming or the use of merging/split paired-end reads. Reads were then mapped onto the contigs of the H. pylori strain B8 genome sequence with the SSAHA2 package.⁴¹ After reads were mapped on the target genome, we minimized the false positive discovery rate by using SAMtools (v.0.1.8; ⁴²) to extract only reliable alignments from SAM-formatted files. The number of reads matching each genomic object harbored by the reference genome was then calculated with the Bioconductor-GenomicFeatures package.⁴³ If reads matched several genomic objects, the count number was weighted so as to keep the total number of reads constant. Finally, the Bioconductor-DESeq package ²⁰ was used with custom parameters (i.e., custom size factor during normalization step; see « Conventional normalization etc » section for details) for the analysis of raw count data and to determine whether expression levels significantly differed between conditions.

Northern blotting

Northern blotting was performed using riboprobes synthesized by in vitro transcription as described previously.¹¹ Primers used for riboprobe synthesis are listed in Table S3. If stated in the figure legend, 0.05% methylene blue solution in 300 mM sodium acetate pH 5 was used to stain rRNAs on the membrane after the RNA transfer to control loading.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgments

We are thankful to Dr C. M. Sharma for useful discussions and comments and Dr C. Condon for critical reading of the manuscript. We thank Andrea Sirianni for help in the annotation of the B128 genome. We are also grateful for funding from the Agence National de la Recherche and Institut Pasteur [ANR 09 BLAN 0287 01, PyloRNA to H.D.R. and F.D.], from Institut Pasteur for the Bourse Roux research fellowship (to Y.R.) and from the French research ministry, PhD fellowship to E.G.