Ribonucleases J1 and J2 of Bacillus subtilis are evolutionarily conserved enzymes combining an endoribonucleolytic and a 5′−3′ exoribonucleolytic activity in a single polypeptide. Their endoribonucleolytic cleavage specificity resembles that of RNase E, a key player in the processing and degradation of RNA in Escherichia coli. The biological significance of the paralogous RNases J1 and J2 in Bacillus subtilis is still unknown. Based on the premise that cleavage of an mRNA might alter its stability and hence its abundance, we have analysed the transcriptomes and proteomes of single and double mutant strains. The absence or decrease of both RNases J1 and J2 together profoundly alters the expression level of hundreds of genes. By contrast, the effect on global gene expression is minimal in single mutant strains, suggesting that the two nucleases have largely overlapping substrate specificities. Half-life measurements of individual mRNAs show that RNases J1/J2 can alter gene expression by modulating transcript stability. The absence/decrease of RNases J1 and J2 results in similar numbers of transcripts whose abundance is either increased or decreased, suggesting a complex role of these ribonucleases in both degradative and regulatory processing events that have an important impact on gene expression.
The degradation and maturation of messenger RNA is an important element in the control of gene expression in all organisms. In contrast to Escherichia coli, our understanding of mRNA metabolism in Bacillus subtilis and Gram-positive microorganisms in general is still quite limited. This is notably true concerning the initiating events in mRNA decay as many of these bacteria including B. subtilis have no authentic RNase E homologue. A considerable amount of data now clearly indicate that the 5′ end is a critical determinant for mRNA stability in B. subtilis, probably even more than in E. coli. For example, a 5′-proximal secondary structure or events such as ribosome stalling and regulatory protein binding are prone to greatly stabilize extended regions of transcripts downstream of these ‘roadblocks’ (Bechhofer and Zen, 1989; Sandler and Weisblum, 1989; Glatz et al., 1996; Hambraeus et al., 2002; Sharp and Bechhofer, 2005).
We have previously identified two paralogous endoribonucleases, RNase J1 (rnjA/ykqC) and J2 (rnjB/ymfA) that share functional homologies with RNase E but no sequence similarity (Even et al., 2005). In fact, RNases J1/J2 and RNase E can cleave at the same site in the leader region of the B. subtilis thrS mRNA, causing an increase in thrS mRNA stability (Condon et al., 1996; 1997; Even et al., 2005). This maturation site resembles what is considered to be a ‘classical’ RNase E target site, i.e. a single stranded AU-rich sequence next to a secondary structure. Moreover, all three enzymes show sensitivity to the phosphorylation state of the 5′ end of the mRNA (Even et al., 2005). However, in the case of RNase J1, it is the recently discovered additional 5′−3′ exoribonucleolytic activity that is stimulated by a 5′-monophosphorylated substrate (Mathy et al., 2007) while its endoribonucleolytic activity is probably independent of the substrate 5′ end (Even et al., 2005; Li de la Sierra-Gallay et al., 2008). By contrast, a 5′-P end increases the catalytic rate constant for the endonucleolytic activity of RNase E (Bouvet and Belasco, 1992; Mackie, 1998; Tock et al., 2000; Jiang and Belasco, 2004).
Both activities of B. subtilis RNase J1 have also been shown to be important for the processing of two stable RNAs, 16S rRNA (Mathy et al., 2007) and scRNA (Yao et al., 2007).
The resolution of the Thermus thermophilusRNase J 3D structure and the in vitro study of B. subtilis RNase J1 mutant proteins revealed that both activities of RNase J are carried out by the same catalytic site (Li de la Sierra-Gallay et al., 2008). Moreover, we have shown that the exonucleolytic activity of native RNase J1 strictly discriminates against primary 5′-triphosphorylated transcripts (Li de la Sierra-Gallay et al., 2008). The identification of a mononucleotide binding pocket in the vicinity of the active site provides a rationale for this discrimination (Li de la Sierra-Gallay et al., 2008). However, the pocket which specifically co-ordinates the monophosphate can also accommodate a 5′-OH group. This is consistent with a recent finding that the B. subtilis glmS riboswitch, which generates a 5′ OH following metabolite-induced self-cleavage, targets the downstream transcript for RNase J1 dependent degradation (Collins et al., 2007). The colocalization of both RNase J activities in a single active site suggests that RNase J might be able to switch from endo- to exonucleolytic mode on the same RNA, a potentially useful capacity for the degradation of transcripts with structured 3′ ends that are resistant to 3′−5′ exoribonucleases. Studies on 3′-protected leader transcripts in B. subtilis (Choonee et al., 2007; Deikus et al., 2008; N. Choonee and H.P., unpublished results) are in favour of this hypothesis.
Among the two RNase J proteins in B. subtilis, only RNase J1 (rnjA) is an essential enzyme (Kobayashi et al., 2003). We have previously shown that in an rnjA-rnjB double mutant (where expression of the essential rnjA gene was reduced) global mRNA stability was only slightly increased (Even et al., 2005). By contrast, inactivation of RNase E increased bulk mRNA stability twofold to threefold (Babitzke and Kushner, 1991) which is reflected by an increase in the abundance of more than 40% of all E. coli mRNAs (Lee et al., 2002). In the only case of RNase J1/J2 mediated endonucleolytic cleavage of an mRNA studied to date (B. subtilis thrS/thrZ mRNA), cleavage by RNases J1/J2 has a stabilizing effect (Even et al., 2005). The 5′−3′ exoribonucleolytic activity of RNase J would be expected to play a predominantly destabilizing role in mRNA metabolism. At present, the only known example where a structural mRNA is destabilized by the exonucleolytic activity of RNase J1 is the glmS mRNA mentioned above (Collins et al., 2007). This raises the possibility that RNase J may be implicated in more complex regulatory processing/maturation of specific mRNAs. If this was the case, one could expect to observe a differential response on the stability of individual mRNAs following RNase J inactivation. It is this critical aspect of RNase J function that is tested here. Our experiments are based on the premise that cleavage of an mRNA by RNase J alters its stability and hence affects its concentration in the cell. In order to obtain a global view we analysed mRNA abundance at single-gene resolution in RNase J1/J2 single and double mutant strains using DNA microarrays and confirmed the results by analysing the corresponding proteomes. This approach also allowed us to examine the individual contribution of each enzyme and to reveal potentially overlapping substrate specificities of RNases J1 and J2. Finally, individual case studies demonstrate that RNases J1/J2 are effectively involved in the degradation of specific mRNAs.
Construction and verification of B. subtilis strains with altered rnjA and rnjB expression profiles
To study the effects of the absence or scarcity of RNases J1 and J2 on global gene expression, we constructed three strains. In the first, the entire rnjB gene was replaced with a spectinomycin cassette (strain SSB355). The absence of rnjB-specific transcripts was checked by Northern analysis (Fig. 1A, lane 6). The absence of RNase J2 did not affect growth of B. subtilis in complex or defined media nor did it alter the expression of the rnjA gene encoding RNase J1 (Fig. 1A, lanes 1 and 2).
Due to the essential nature of RNase J1 the rnjA gene could not be inactivated. Instead, expression of the chromosomal rnjA gene was brought under the control of the inducible xylose promoter by Campbell type integration of plasmid pHMJ25 as described in Experimental procedures (strain SSB356, Fig. 1B). Finally, we constructed a strain (SSB357) harbouring the rnjB deletion and the Pxyl–rnjA inducible construct.
The expression of RNase J1 from the Pxyl–rnjA construct was quantified by Western analysis in the presence and absence of xylose. The antibody used recognizes both nucleases. As RNases J1 and J2 (Mr = 61.5 and 61.2 kDa respectively) could not be effectively resolved on SDS-PAGE, the experiment was carried out with the double mutant which is completely devoid of RNase J2. In the absence of inducer, RNase J1 was barely detectable (Fig. 1C, lane 4) and the cells eventually cease growth. Even under inducing conditions (100 mM xylose) the intracellular level of RNase J1 was only about 11% (11.4%) of that observed in the wild-type strain (Fig. 1C, lanes 2 and 3). In apparent contradiction, the ninefold reduction in RNase J1 protein amount was accompagnied by a 5.3-fold increase in rnjA mRNA (Fig. 1A, compare lane 1 with lanes 3 and 4). This discrepancy could be explained by a low efficiency of translation of the rnjA gene in the mutant strain. In the wild-type strain the rnjA mRNA was co-transcribed with a small upstream open reading frame of 207 bp (ykzG). This co-transcription is reflected by the presence of a larger transcript in the wild-type compared with the Pxyl–rnjA strains representing the bicistronic ykzG-rnjA mRNA (Fig. 1A, compare lane 1 with lanes 3 and 4). The weak rnjA ribosome binding site is located within the ykzG ORF and preliminary experiments with various lacZ fusion constructs showed that translational coupling between ykzG and rnjA is important for the efficient translation of the rnjA messenger (data not shown). The rnjA/rnjB Northern analysis of the single mutant strains also indicated that under the conditions used RNases J1 and J2 did not cross-regulate their expression at least at the transcriptional level. In fact, the abundance of the individual mRNAs was not significantly affected by the absence or scarcity of the respective paralogous RNase (Fig. 1A, compare lanes 1 and 2 with lanes 5 and 7).
Influence of RNases J1 and J2 on individual mRNA steady state levels on a global scale
We compared the transcriptomes of the wild-type and the RNase J1/J2 mutant strains described above at single gene resolution. This analysis is based on the premise that most endoribonucleolytic cleavages have an effect on the stability of RNA and hence on the steady-state levels of transcripts. As RNase J1 is an essential protein, a major concern was to find conditions where its expression is significantly reduced, while ensuring a growth rate comparable to that of the wild-type strain at the same time. This issue was important for the double mutant strain SSB357. All the strains were thus cultivated in the presence of inducer (100 mM xylose). Under these conditions, we obtained ninefold reduced RNase J1 levels (Fig. 1C, lanes 2 and 3), whereas the growth defect of the double mutant strain was kept to a minimum (81 min doubling time versus 45 min for the wild-type strain). Under the same conditions, the Pxyl–rnjA single mutant (SSB356) grew with a doubling time comparable to that of the wild-type strain (53 versus 45 min respectively).
To identify transcripts whose cellular levels are affected by RNase J1 and J2, we used whole-genome DNA microarrays containing 4022 B. subtilis ORFs. These experiments compared abundance of cDNAs which were differentially labelled with fluorescence dyes and hybridized competitively to the DNA probes arrayed on glass slides. The cDNAs were synthesized using RNA samples prepared from the wild-type, the isogenic rnjA and rnjB single mutant and the rnjA/rnjB double mutant strains described above. The level of RNase J1 was determined in the same cultures used for RNA isolation (Fig. 1C).
The amount of gene-specific mRNA was considered to be different between the wild-type and the respective mutant when the change was at least twofold and the statistical criteria described in Experimental procedures were fulfilled. A compilation of the respective genes is provided in Tables S1–S6. As shown in the Venn diagram in Fig. 2, the rnjA/rnjB double mutant (SSB357) had a much more global impact on mRNA abundance than the single mutant strains SSB355 (rnjB) and SSB356 (rnjA). In the absence of RNase J2, the levels of 44 transcripts were altered while a ninefold reduction in the RNase J1 level affected the abundance of 79 mRNAs. By contrast, a combination of the two genotypes (SSB357) resulted in a change of the steady-state levels of more than 650 transcripts. It is interesting to note that these changes were equally distributed between mRNAs whose abundance was increased (328 genes) and decreased (330 genes). This was also true for the rnjA and to a lesser extent the rnjB single mutant (46 versus 33 and 12 versus 32 genes respectively). Most of the mRNAs whose levels were increased in the rnjB (SSB355) single mutant were also more abundant in the double mutant (SSB357). This was the case for only 17 of the 46 transcripts increased in the rnjA (SSB356) single mutant (Fig. 2). By contrast, with a few exceptions, all mRNAs exhibiting decreased abundance in the single mutants were unaltered in the double mutant. The great majority of these mRNAs represent transcripts of the SP-beta prophage (Fig. 2, Tables S1, S2 and S4).
In the double mutant strain, significant numbers of mRNAs whose abundance was increased could be assigned to the SigD regulon, to sporulation-specific regulons and to the ComK regulon involved in the development of competence (Table S5). The increase in the steady-state mRNA levels of many competence related genes in the double mutant was actually more than 10-fold with a maximum of 38-fold observed for the comGA gene. This is likely due to a considerable increase in comS and comK mRNA levels (ninefold and fivefold respectively) observed under these conditions. The regulon controlled by the alternative sigma factor SigD comprises about 80 genes involved in flagellar synthesis, motility, chemotaxis and autolysis (Helmann and Moran, 2002; Serizawa et al., 2004). SigD controlled genes were significantly over-represented among the genes upregulated in the rnjB deletion strain (8 of 12 genes) and in the double mutant strain (61 of 328 genes). In fact, the mRNA levels of nearly 80% of the SigD regulon were increased in the rnjA/rnjB double mutant, probably as a direct result of the 2.5-fold increase in sigD mRNA observed under these conditions. Among the genes exhibiting an increase in transcript abundance in the double mutant strain, we also found 14 of the 33 PhoP regulon members as well as 16 of the 56 genes belonging to the SigW regulon (Table S5). The PhoP–PhoR two-component regulatory system controls the expression of genes whose products are involved in phosphate metabolism (Allenby et al., 2005). The alternative, extracytoplasmic function (ECF) sigma factor SigW is activated by intramembrane proteolysis of the membrane-bound anti-sigma factor RsiW in response to antimicrobial compounds and other stresses affecting the cell envelope (Wiegert et al., 2001; Cao et al., 2002; Pietiäinen et al., 2005; Butcher and Helmann, 2006). Interestingly, the prsW (ypdC) gene encoding the protease responsible for site-1 cleavage of RsiW (Ellermeier and Losick, 2006; Heinrich and Wiegert, 2006) was also upregulated in the rnjA/rnjB double mutant (Table S5).
Among the mRNAs with decreased steady-state levels in the double mutant, a significant number encoded ribosomal proteins and other factors belonging to the translational and transcriptional apparatus (Table S6). It is possible that many of these changes are not directly related to the absence of RNases J1 and J2 but rather to the decrease in the growth rate of the double mutant strain (81 min versus 45 min doubling time). However, the rpmA transcript coding for ribosomal protein L27 is more abundant under these conditions. Whereas most of the amino acid biosynthesis genes were not altered in their expression, two-amino-acid biosynthesis operons were found to be downregulated, the trp and the ilv–leu operons as well as a few genes implicated in proline, histidine and tyrosine biosynthesis (Table S6). A detailed transcriptional analysis of the ilv–leu operon revealed that the mRNA levels are actually affected by both, growth rate and the action of RNases J1/J2 (see below). One of the most pronounced effects concerns the S-box regulon (Table S6). Of the 11 transcriptional units belonging to this global transcription termination control system for methionine and cysteine biosynthesis genes (Grundy and Henkin, 1998), eight were downregulated in the double mutant strain. The S-box genes are regulated by a riboswitch sensitive to the intracellular concentration of S-adenosine methionine (Epshtein et al., 2003; McDaniel et al., 2003; Winkler et al., 2003). It is thus tempting to speculate that RNases J1/J2 may play a direct role in this regulation as they do in the T-box anti-termination system of thrS/Z (Even et al., 2005) (see Discussion). The transcript levels of three T-box genes (cysE, glyS and tyrS) were also found to be decreased in the double mutant (Table S6). Other functional categories to which a significant number of mRNAs with decreased abundance in the double mutant can be assigned include heat-shock genes belonging to the HrcA regulon which includes the dnaK and the groESL operons. In the dnaK operon, a processing event leading to differential segmental mRNA stability has previously been described (Homuth et al., 1999).
Effect of RNases J1 and J2 on global protein synthesis
The changes in the steady-state levels of hundreds of mRNAs observed in the transcriptome of the rnjA/rnjB double mutant strain (SSB357) prompted us to investigate whether these changes are reflected on the protein level. We prepared extracts from the wild-type and the single and double rnjA/rnjB mutant strains grown under the same conditions as those used for the transcriptome analysis described above. The extracts were separated by two-dimensional (2-D) PAGE and analysed as described in Experimental procedures.
The global protein synthesis pattern was almost unaffected in the rnjA and rnjB single mutants (Fig. 3A and B). By contrast, the proteome of the double mutant was extensively altered compared with the wild-type strain (Fig. 3C). From about 1000 proteins visible on the 2-D gels in the pI range 4–7, the amount of more than 200 proteins was significantly changed in the double mutant. These changes concerned roughly equal numbers of upregulated and downregulated proteins. The transcriptome and proteome data thus give the same overall picture indicating that RNases J1 and J2 must have largely overlapping substrate specificities.
We identified 46 spots by mass spectrometry from the rnjA/rnjB double mutant proteome representing 39 different proteins (5 proteins formed multiple spots) significantly over- or under-expressed in two independent experiments. The 39 identified proteins comprised 19 upregulated and 20 downregulated proteins (Fig. 3C). For 27 of the 39 proteins we found a corresponding change in the transcriptome, i.e. at least twofold different mRNA levels between the wild-type and the double mutant. In the case of one of the strongest upregulated proteins (> 10-fold), Hag, we observed a fourfold increase in mRNA abundance in the double mutant strain (Table S5). Among the downregulated proteins we identified three polypeptides encoded by the leuA, C and D genes implicated in the biosynthesis of leucine (Fig. 3C). These genes are part of the ilv–leu operon which we analysed in more detail (see below).
Detailed transcriptional analysis of mRNA species up- or downregulated in rnjA/rnjB mutant strains
The increase in mRNA abundance of certain transcripts is caused by an increase in mRNA stability. As detected by the transcriptome analysis, the mRNA amounts of hundreds of genes are increased in the rnjA/rnjB double mutant compared with the wild-type (Fig. 3 and Table S5). This increase in mRNA abundances is assumed to be caused by the absence or reduction of the 5′−3′ exo- as well as the endoribonucleolytic activities associated with RNases J1 and J2. Consequently, these mRNA species should exhibit increased stabilities in the double mutant compared with the wild-type. In order to validate this premise, we analysed the mRNA stabilities of four representative genes chosen on two simple criteria: (i) likelihood of being monocistronic, for easier analysis of the data and (ii) showing a clearly higher mRNA amount in the transcriptome of the rnjA/rnjB mutant strain. These genes included cspC, encoding a major cold shock protein (Schindler et al., 1999), spoVG, encoding a protein interfering with assymmetric septation (Matsuno and Sonenshein, 1999), tagD, encoding the glycerol-3-phosphate cytidylyltransferase involved in teichonic acid biosynthesis (Bhavsar et al., 2001) and yweA, encoding a protein of unknown function.
First, we performed a Northern analysis to confirm the transcriptome data. When comparing the abundances of the transcripts in the double mutant and the wild-type strain, the ratios obtained by Northern and microarray analyses were very similar (Fig. 4A).
We then determined the half-lives of the different mRNA species in the mutant and wild-type context by Northern analyses after addition of rifampicin. Representative results are shown in Fig. 4B.
In all four cases, the mRNA half-life was clearly extended in the double mutant strain. Whereas the half-life was doubled in the case of cspC, it was nearly threefold and sixfold longer for spoVG and yweA respectively. In order to facilitate the comparison between wild-type and mutant strains, the images in Fig. 4B were adjusted to give identical signals at time zero. However, the calculation of half-lives was based on signals obtained at all time points including those that are not visible in the presented exposure (i.e. cspC). In the case of tagD, mRNA half-life could not be precisely determined in the wild-type, because there was no more mRNA detectable 2 min after rifampicin addition. In the double mutant, the half-life increased significantly to 2.5 min. These results strongly suggested that RNases J1/J2 are directly involved in the degradation of these four mRNAs.
RNases J1/J2 are implicated in the processing of the polycistronic ilv–leu operon. The transcriptome and proteome data both showed that the four promoter distal genes of the ilv–leu operon (leuA-D) were specifically downregulated in the rnjA/rnjB double mutant (Fig. 3C and Table S6). And because the transcriptional organization and post-transcriptional regulation of this amino acid biosynthetic operon have been extensively studied (Mäder et al., 2004), we decided to analyse the effects of RNase J1/J2 activity on the ilv–leu operon in more detail.
One of the key features of the regulation of this operon is the endoribonucleolytic processing of the heptacistronic 8.5 kb primary transcript close to the end of the second cistron (ilvH, Fig. 5A). According to the published model of the post-transcriptional regulation of the operon, this cleavage generates three processing products displaying differential mRNA stabilities: a barely detectable 2.7 kb upstream product, and two downstream products of 5.8 and 1.2 kb. The latter was probably generated by 3′−5′ exoribonucleolytic degradation of the 5.8 kb transcript halted at an internal stem-loop structure downstream of ilvC (Fig. 5A (Mäder et al., 2004)). Compared with the full-length 8.5 kb primary transcript, the mRNA segment upstream of the cleavage site was less stable while the downstream mRNAs (1.2 and 5.8 kb respectively), most likely due to presence of a secondary structure at their 5′ ends, were significantly more stable (Mäder et al., 2004).
In the wild-type strain, the four LeuABCD proteins wereare essentially specified by the 5.8 kb processing product (Fig. 5A). Their specific downregulation in the rnjA/rnjB double mutant pointed to an involvement of the RNases J1/J2 in the crucial mRNA processing event within the ilv–leu operon. To verify this assumption, we first compared the ilv–leu mRNA profile of the wild-type strain with that of rnjA/rnjB single and double mutant strains using an ilvC-specific RNA probe. This probe could not detect the 2.7 kb upstream transcript (Fig. 5A) and a band of similar size most prominent in the mutant strains (Fig. 5B) likely consisted of ilvC containing degradation intermediates whose electrophoretic mobility is disturbed by the comigration with large quantities of 23 S rRNA. The Northern analysis showed that the ratio of the 8.5 kb primary transcript to the 5.8 kb processing product is clearly altered in the mutant strains (Fig. 5B). Quantification of the obtained signals revealed a complete inversion of the ratios of the relative abundances of the 8.5 and 5.8 kb mRNAs in the double mutant when compared with the wild-type strain: 4.2:1 (double mutant) versus 1:4.7 (wild-type). A similar trend was observed in the rnjA (SSB356) but not in the rnjB (SSB355) single mutant. The amount of the 5.8 kb mRNA in the double mutant was determined to be 20-fold lower when compared with the wild-type, whereas the ratio between the 8.5 kb primary transcripts differs only marginally between the two strains (ratio wild-type/double mutant: 1.3, see Discussion). Thus, the abundance of the 5.8 kb processing product was strongly reduced in the double mutant (SSB357) compared with the other strains, suggesting that RNase J1 (and probably to a lesser extent RNase J2) were responsible for the majority of the processing of the original 8.5 kb primary transcript. If this holds true, the half-life of the 8.5 kb mRNA should be increased in the double mutant compared with the wild-type strain, whereas the half-life of the 5.8 kb processing product should remain unchanged. To verify this prediction, we performed Northern analyses after addition of rifampicin to determine the half-lives of the particular mRNA species. One representative result is shown in Fig. 5C. The half-life of the 8.5 kb primary transcript was indeed significantly increased in the double mutant compared with the wild-type strain (3.6 min versus 1.5 min respectively), while the half-life of the 5.8 kb processing product was identical in both strains (3.3 min, Fig. 5C) corroborating the involvement of RNases J1/J2 in ilv–leu processing.
The transcriptome and proteome data presented here clearly demonstrate that B. subtilis RNases J1 and J2 affect gene expression on a global scale. The small number of mRNAs whose abundance is altered in the single mutants as compared with the double mutant indicates that RNases J1 and J2 have largely overlapping substrate specificities and work cooperatively. This conclusion is also corroborated by our previous in vitro experiments that showed that purified RNases J1 and J2, respectively, can cleave the thrS and thrZ leader mRNA with equal specificity and efficiency (Even et al., 2005). The essential nature of RNase J1 may reside in its capacity to specifically process/degrade one or a few RNA substrates that are vital for the cell. However, no obvious candidate genes emerge from the list of transcripts whose abundance is specifically altered in the strain with reduced RNase J1 levels (SSB356, Tables S3 and S4). Surprisingly, there exists a significant number of mRNAs whose abundance is altered in one or both single mutants but not, as might be expected, in the rnjA/rnjB double mutant. At present we have no obvious explanation for this observation. One possibility would be that the remaining RNase J1 or J2 in the single mutants specifically processes an mRNA whose translation product plays a role in the altered expression of the genes detected in the transcriptome.
The use of an inducible rnjA gene obviously has some inherent drawbacks that need to be taken into consideration when interpreting the data. The most important concern was the fact that we had to reconcile a significant reduction in RNase J1 activity with maintaining a reasonable growth rate compared with the wild-type strain. Interestingly, a ninefold reduction in RNase J1 levels caused only a very light growth defect in the rnjA single mutant (53 versus 45 min doubling time for the wild-type strain) but had a more severe impact on growth of the rnjA/rnjB double mutant strain (81 versus 45 min). This indicates that B. subtilis growth requires a cellular RNase J1 concentration at least 10–15% of normal, a level that is no longer sufficient to maintain wild-type growth in an rnjB background. Thus, more RNase J1 is needed in the double mutant to compensate for the loss of RNase J2. This may be due to a titration of RNase J1 by substrates normally taken charge of by RNase J2 thereby not leaving enough RNase J1 to take care of its ‘essential’ substrate(s). Alternatively, the cumulative lack of RNase J1/J2 activity on a multitude of substrates rather than an essential target may account for the reduced growth rate of the double mutant strain.
In E. coli, 10–20% of wild-type RNase E levels are also sufficient for ensuring normal growth (Jain et al., 2002). Under these conditions, mRNAs known to decay in a RNase E-dependent fashion start to show increased half-lives (Jain et al., 2002; Ow et al., 2002), but the full impact on mRNA decay was only observed in strains with temperature sensitive or truncated alleles of the rne gene (Lopez et al., 1999; Lee et al., 2002; Ow and Kushner, 2002). As in our analysis we used a partial depletion of RNase J1 it is likely that the absolute number of transcripts affected by the rnjA/rnjB double mutant is actually higher than we observed. Many of the transcripts increased in the double mutant belong to large regulons involved in competence (ComK regulon) and motility/chemotaxis (SigD regulon). Their respective regulators ComS/ComK and SigD were also upregulated, implying that the increase of many mRNAs belonging to these regulons might not be directly due to the action of RNases J1/J2. However, many sigD dependent genes are also upregulated in the single mutant strains (e.g. 8 out of 12 in the rnjB mutant) where sigD levels were not elevated. Only the analysis of individual mRNAs can provide a definite answer as to whether the effect of RNase J is direct or indirect. We chose four monocistronic mRNAs for further scrutiny and found in all cases that their half-lives were increased in the absence of RNases J1/J2. This is the first direct evidence that these ribonucleases can decrease mRNA stability. A detailed case-by-case study will be needed to evaluate the contribution of the exo- and/or endonucleolytic activities of RNase J1/J2 in the degradation of these mRNAs.
Among the upregulated genes in the rnjA/rnjB double mutant we identified spo0E (Table S5), encoding a specific phosphatase for phosphorylated Spo0A, a master regulator of sporulation initiation (Ohlsen et al., 1994). We indeed observed that the double mutant is asporogenous (data not shown) and the increased expression of Spo0E provides a good rationale for such a phenotype. Our data also corroborate a previous observation that a reduction of RNase J1 alone reduces sporulation efficiency, a defect that could be corrected by the disruption of spo0E (Ogura et al., 2006).
The decrease of the abundance of hundreds of transcripts in the double mutant is more difficult to interpret. It implies that if the absence of RNases J1/J2 were the principal cause for this decrease then these messages had to be stabilized by RNase J in the wild-type strain. Certain specific examples are known where an increase in mRNA stability is directly caused by RNase J1/J2 cleavage, for example in the regulation of expression of the thrS/thrZ genes controlled by the T-box anti-termination system (Condon et al., 1996; Even et al., 2005). The fact that the thrS mRNA was not among the transcripts with decreased abundance is not unexpected as processing of the thrS leader occurs efficiently only under inducing (when threonine is limiting) conditions (Condon et al., 1996). As the rnjA/rnjB mutant strain grew significantly more slowly than the single mutant and wild-type strains the downregulation of many mRNAs might actually have been caused by the decreased growth rate. At first glance this seems the obvious explanation for the 30 ribosomal protein-encoding transcripts (Table S6). However, the rpmA mRNA encoding ribosomal protein L27 is upregulated under these conditions and the remaining 19 ribosomal protein transcripts are not affected. Other genes prone to be downregulated by the reduced growth rate of the rnjA/rnjB double mutant include those involved in amino acid biosynthesis. However, we observed no general decrease in transcripts from this class of genes. Among those that were reduced are two well-studied biosynthetic operons, trpE-A and ilv–leu (Table S6), known to undergo complex RNA processing events.
The dual activity of RNase J and its sensitivity towards a variety of structural configurations of an RNA provide for a complex pattern of cause and effect. For example, the infC–rpmI–rplT operon encoding translation initiation factor IF3 and the ribosomal proteins L35 and L20, respectively, is downregulated in the rnjA/rnjB mutant strain (Table S6). We have recently shown that the expression of this operon is auto-controlled by the L20 protein via a transcription attenuation mechanism (Choonee et al., 2007). This mechanism could well explain why this operon is downregulated when growth rate is reduced. However, the leader mRNA is also cleaved by RNases J1/J2 (Choonee et al., 2007). The precise role of the ribonucleases in this regulation is still under investigation but this observation incites to be careful in attributing the downregulation of certain mRNAs in the double mutant solely to a decrease in growth rate.
Processing of the polycistronic ilv–leu mRNA is the first example where RNases J1/J2 cleave within an open reading frame (ilvH) suggesting a potential linkage between processing and translation. In further support of this notion, RNase J1 has been found to be associated with the ribosome (Even et al., 2005; Hunt et al., 2006). The strongly reduced processing of the ilv–leu messenger in the double mutant resulted in a smaller than expected increase of the 8.5 kb primary transcript with respect to the wild-type strain (Fig. 5B). However, as expression of the ilv–leu operon is known to be growth rate regulated (Tojo et al., 2005), this is most likely due to a lower synthesis rate of the primary transcript in the double mutant as a consequence of its lower growth rate compared with the wild-type (81 versus 45 min). In accordance, in the rnjA single mutant, where the growth rate is less affected (53 versus 45 min), the 5.8 kb processing product is also significantly reduced and in this strain a concomitant increase of the 8.5 kb primary transcript could be observed (Fig. 5B, lane 3). Consequently, the total amount of leuA-D-specific mRNA is similar to the wild-type, which corresponds well with the transcriptome and proteome data. The specific downregulation of leuA, leuB, leuC and leuD, but not of ilvB, ilvH and ilvC in the double mutant revealed by transcriptome and proteome analyses is therefore caused by the combination of reduced endoribonucleolytic processing of the 8.5 kb primary transcript and a growth rate-dependent downregulation of ilv–leu expression. The amount of the 1.2 kb processing product encoding ilvC would also be predicted to decrease in the double mutant; however, the observed effect is weaker than expected because the stability of this mRNA is increased in the double mutant from 9 to 16.7 min (Fig. 5C), pointing to an involvement of the RNases J1/J2 in the degradation of the monocistronic ilvC mRNA. We propose that RNases J1/J2 actually fine-tune the expression of the ilv–leu operon through a processing event that generates matured transcripts with differential stabilities. It should be noted that the action of RNases J1/J2 on the ilv–leu mRNA is discussed here considering previously published data that processing of the primary transcript is endonucleolytic (Mäder et al., 2004). While this might be the most likely scenario, we cannot rule out the possibility that the conversion of the primary transcript to the 5.8 kb mRNA is achieved by 5′−3′-exonucleolytic degradation of the primary transcript halted at a secondary structure present at the 5′ end of the 5.8 kb messenger. Further experiments will be necessary to distinguish between these two possibilities.
A similar regulatory processing has also been reported for the glycolytic gapA operon. Cleavage and differential stability of the matured mRNAs account for a 100-fold difference in expression between the CggR repressor protein and the glyceraldehyde-3-phosphate dehydrogenase encoded by gapA (Meinken et al., 2003). Even though gapA was downregulated in the rnjA/rnjB double mutant strain (Table S6), we have not been able to reproducibly show that RNases J1/J2 are responsible for this cleavage.
Another functional category whose mRNAs were less abundant in the double mutant included transcripts from the S-box regulon. Genes of this family involved in methionine and cysteine metabolism are regulated by a riboswitch that causes premature transcription termination in response to elevated levels of S-adenosyl methionine (SAM) (Epshtein et al., 2003; McDaniel et al., 2003; Winkler et al., 2003). Eight of the 11 transcriptional units of this regulon were downregulated up to eightfold in the double mutant. As the metK transcript encoding SAM synthetase is also downregulated (2.4-fold), this global effect is most likely not caused by an increase of the SAM level in the cell. A preliminary Northern analysis of certain S-box genes in an rnjA/rnjB double mutant revealed an altered transcription profile compared with a wild-type strain (data not shown) hinting at a possible direct involvement of RNases J1/J2 in the expression of genes of this regulon.
The large number of transcripts affected in the double mutant strain is reminiscent of E. coli RNase E, which plays a global role in the initiation of mRNA decay. In fact, we have previously shown that RNases J1/J2 share some functional and even structural homologies with E. coli RNase E (Li de la Sierra-Gallay et al., 2008). However, the dual activity of RNase J provides for an unprecedented range of action for a ribonuclease. This versatility allows for RNase J to be used extensively for regulatory processing/maturation of RNAs but at the same time makes it more difficult to interpret the role of this nuclease in RNA metabolism. Nevertheless, the large number of mRNAs whose abundance is altered in the double mutant strain also provides a reservoir of new potential substrates that can be used to study diverse functions of RNase J.
Bacterial strains and growth conditions
The prototrophic B. subtilis 168 strain 1A2 (BGSC) was used for the construction of mutants. In strain SSB355 (rnjB::spc), the rnjB (ymfA) gene was deleted by double cross-over recombination with plasmid pHMJ13. For strain SSB356 (Pxyl::rnjA Cmr), the rnjA (ykqC) gene was brought under the control of the Pxyl promoter by Campbell-type integration of plasmid pHMJ25. Strain SSB357 (rnjB::spc, Pxyl::rnjA Cmr) was constructed by transforming strain SSB356 with chromosomal DNA from strain SSB355 and selection for spectinomycin resistance. For all experiments strains were grown at 37°C in a defined medium (Stülke et al., 1993) with 0.75% fructose (w/v) as carbon source. For strains carrying the rnjA gene under control of the xylose promoter, 100 mM xylose was added to the medium. The cultures were inoculated with exponentially growing cells and harvested in the exponential growth phase after reaching an OD500 of 0.5.
Escherichia coli JM109 served as the host for plasmid constructions and was grown at 37°C in LB medium. When required, antibiotics were added at the following concentrations: ampicillin (200 μg ml−1), spectinomycin (100 μg ml−1), and chloramphenicol (5 µg ml−1).
pHMX2. A poly-linker was inserted at the BamHI site of the integrative plasmid pX2 (Mogk et al., 1997) using oligonucleotides HP906 (GATCCGGTACCACGCGTATGCATGAGCTCCCGGGTGTACA) and HP907 (GATCTGTACACCCGGGAGCTCATGCATACGCGTGGTACCG). The following unique restriction sites are available downstream of the Pxyl promoter: BamHI-KpnI-MluI-NsiI-SacI-SmaI-BsrGI.
pHMJ13. A 737 bp fragment upstream and a 756 bp fragment downstream of rnjB (ymfA) were amplified by PCR using oligonucleotide pairs HP894/HP895 (AGTTCGGAATTCGTATTACAATAAACCTTCTCA, TCGTACAACTGCAGCAAAATCTATATCCTCCTAGTC) and HP896/HP897 (TCAGAACTACTCGAGTGACTGACTAAAGACCGGAGCTG, ATGACTCATCTAGACCGTATTTAACTGCATCCTTACC) respectively. The upstream fragment was cut with EcoRI and PstI and the downstream fragment with XhoI and XbaI. A third fragment containing the spectinomycin cassette from plasmid pDG1727 (Guerout-Fleury et al., 1995) was cut with PstI and XhoI. These three fragments were cloned into plasmid pJRD184 (Heusterspreute and Thi, 1985) cut with EcoRI and XbaI in a four-partner ligation.
pHMJ25. A 379 bp fragment containing rnjA sequences from position −25 to +354 with respect to the ATG initiation codon was amplified by PCR. The BamHI and MluI sites included in the primers were used to ligate this fragment into the respective sites of plasmid pHMX2 downstream of the xylose promoter.
Cell harvesting and preparation of total RNA were performed as described previously (Eymann et al., 2002). RNA quality was checked using the Bioanalyzer 2100 (Agilent Technologies, Berlin, Germany) according to the manufacturer's instructions. The RNA samples were obtained from three (SSB355 and SSB356) or four (SSB357) independent cultures and were used for independent cDNA synthesis and DNA array hybridization. Generation of the Cy3/Cy5-labelled cDNAs and hybridization to B. subtilis whole-genome DNA microarrays (Eurogentec) were performed as described by Jürgen et al. (2005). The slides were scanned with a ScanArray Express scanner (PerkinElmer Life and Analytical Sciences, Rodgau-Jügesheim, Germany). Quantification of the signal and background intensities of the individual spots was carried out using the ScanArray Express image analysis software.
Data were analysed using the GeneSpring software (Agilent Technologies). Raw signal intensities were first transformed by intensity dependent LOWESS normalization. The normalized array data were subjected to a statistical analysis using Cyber-T, a program based on a t-test combined with a Bayesian statistical framework (Baldi and Long, 2001). The software is accessible through a Web interface at http://cybert.microarray.ics.uci.edu. The mRNA abundance was considered to be significantly different between the wild-type and the respective mutant strain if (i) the Cyber-T Bayesian P-value was < 0.001 and (ii) the averaged fold change was at least 2. All differentially expressed genes are listed in Tables S1–S6. The potential and known functions of the encoded proteins were predominantly inferred from the SubtiList database (http://genolist.pasteur.fr/SubtiList/).
Northern blot analysis
Northern blot analysis was carried out as described previously (Homuth et al., 1997). Chemiluminescence was detected with the Lumi-Imager (Roche Diagnostics), and chemiluminographs were quantified using the Lumi-Analyse software package (Roche Diagnostics). For mRNA half-life determination, Rifampicin was added to exponentially growing B. subtilis cells to prevent transcription initiation. Samples for RNA preparation were removed before and at different times after rifampicin addition, and preparation of total RNA was carried out as described previously (Homuth et al., 1997). These RNA samples were hybridized in Northern experiments with gene specific RNA probes. Quantification of the resulting luminographs allowed the calculation of specific half-lives. For each gene, at least three independent biological replicates were analysed for both strains (wild-type and SSB357). Transcript sizes were determined by comparison with an RNA size marker (Invitrogen). The digoxygenin-labelled specific RNA probes were synthesized by in vitro transcription using T7 RNA polymerase and specific PCR products as templates. PCR synthesis of the templates was performed using the following pairs of oligonucleotides: for the rnjA probe, rnjA-for (5′-ATGAAATTTGTAAAAAATGA-3′) and rnjA-rev (5′-CTAATACGACTCACTATAGGGAGATTCACTGTTTGTGCTGTCTG-3′); for rnjB, rnjB-for (5′-ATGCATCCAGAAAACGAAAT-3′) and rnjB-rev (5′-CTAATACGACTCACTATAGGGAGACTGAATCCGGTTAATATTGG-3′); for ilvC, ilvC-5′ (5′-TATAACGGTGATATCAAAGA-3′) and ilvC-3′ (5′-CTAATACGACTCACTATAGGGAGAGCCGCAAAGAACTGCTTGCT-3′); for cspC, cspC-for (5′-ATGGAACAAGGTACAGTTAA-3′) and cspC-rev (5′-CTAATACGACTCACTATAGGGAGAAGCTTTTTGAACGTTAGCAG-3′); for tagD, tagD-for (5′-ATGAAAAAAGTTATCACATA-3′) and tagD-rev (5′-CTAATACGACTCACTATAGGGAGATAAACCAGCAATTTCCTCTT-3′); for spoVG, spoVG-for (5′-GTGGAAGTTACTGACGTAAG-3′) and spoVG-rev (5′-CTAATACGACTCACTATAGGGAGAAGAAGCTCCAGCTTCTTCGA-3′); and for yweA, yweA-for (5′-ATGCTAAAAAGAACTTCATT-3′) and yweA-rev (5′-CTAATACGACTCACTATAGGGAGAACGGGGATCAATCACCAGCT-3′). The underlined sequences indicate the T7 promoter region.
Preparation of protein extracts and two-dimensional protein gel electrophoresis
To isolate the cytoplasmic protein fraction, cells were re-suspended in TE buffer with 1 mM PMSF and mechanically disrupted using the RiboLyser (Thermo Electron Corporation GmbH, Dreieich, Germany) for 45 s at 6.5 m s−1. The samples were centrifuged at 16 000 g for 15 min at 4°C to remove the cell debris. Subsequently, samples were centrifuged at 20 000 g for 30 min at 4°C, and protein concentrations were determined according to the method of Bradford (1976) with a commercially available dye reagent (Protein Assay from Bio-Rad).
Immobilized pH gradient (IPG) strips with a pH range from 4 to 7 (GE Healthcare, Freiburg, Germany) were re-hydrated for 24 h in a solution containing 8 M urea, 2 M thiourea, 2% CHAPS, 28 mM DTT, 1.3% Pharmalytes (pH 3–10) and 400 μg of crude protein extract. Subsequently, the IPG strips were subjected to isoelectric focusing on a Multiphor II unit (GE Healthcare) with the voltage profile described previously (Büttner et al., 2001). In the second dimension, proteins were separated on 12.5% SDS polyacrylamide gels using the Dodecan electrophoresis system (Bio-Rad). The gels were fixed with 40% ethanol, 10% acetic acid and stained with colloidal Coomassie brillant blue G-250. After scanning, the 2-D gel images were analysed with the Delta 2-D software (Decodon, Greifswald, Germany). Protein extracts from two independent cultures of each strain were separated, and only changes in the protein pattern appearing on both parallel gels were considered. Protein identification by MALDI-TOF-TOF mass spectrometry was performed as described previously (Eymann et al., 2004).
Western blot analysis
For Western blot analysis, 10 μg of protein extract was separated by one-dimensional SDS-PAGE (12.5%). After electrophoretic transfer of the proteins, the nitrocellulose membrane was blocked with 5% skimmed milk powder in Tris/NaCl buffer (0.5 M Tris/HCl, pH 7.6, 0.15 M NaCl). Subsequently, the membrane was incubated with polyclonal anti-RNase J2 antibodies. After washing, alkaline phosphatase-conjugated anti-rabbit IgG (Dianova, Hamburg, Germany) was added, and the blot was developed with fluorescent ECF substrate (GE Healthcare). The blot was scanned using a Storm 860 PhosphoImager (Molecular Dynamics Sunnyvale, California). Quantification of the bands was carried out with the ImageQuant software (Molecular Dynamics).
We thank Jackie Plumbridge for discussions and critical reading of the manuscript and Dirk Albrecht and Birgit Voigt for protein identification. This work was supported by funds from the Bundesministerium für Bildung und Forschung (03ZIK012), the CNRS (UPR 9073), MRE (Contract 92C0315), Université Paris VII (Contract DRED) and PRFMMIP from the Ministère de l'Education Nationale.