Endonuclease cleavage of messenger RNA in Bacillus subtilis

Authors


Summary

A deletion derivative of the ermC gene was constructed that expresses a 254-nucleotide mRNA. The small size of this mRNA facilitated the detection of processing products that did not differ greatly in size from the full-length transcript. In the presence of erythromycin, which induces ribosome stalling near the 5 end of ermC mRNA, the 254-nucleotide mRNA was cleaved endonucleolytically at the site of ribosome stalling. Only the downstream product of this cleavage was detectable; the upstream product was apparently too unstable to be detected. The downstream cleavage product accumulated at times after rifampicin addition, suggesting that the stalled ribosome at the 5 end conferred stability to this RNA fragment. Neither Bs-RNase III nor RNase M5, the two known narrow-specificity endoribonucleases of Bacillus subtilis, was responsible for this cleavage. These results indicate the presence in B. subtilis of another specific endoribonuclease, which may be ribosome associated.

Introduction

The rate of mRNA decay is an important element in the control of gene expression. For bacterial species other than Escherichia coli, very little is known about the mecha-nism of mRNA decay. To study mRNA decay in Bacillus subtilis, a model Gram-positive organism, we have used the 910-nucleotide (nt) ermC mRNA, which codes for resistance to erythromycin (Em) (Bechhofer, 1993). The ermC message codes for a ribosomal RNA methyltransferase (‘methylase’) that methylates a particular residue of ribosomal RNA, rendering the ribosome resistant to Em binding. Expression of ermC is inducible by the addition of subinhibitory concentrations of Em, and this regulation occurs at the translational level (Weisblum, 1983; Dubnau, 1984). In the absence of Em, the ermC leader region is in a highly structured form (see Fig. 1A) that inhibits translation of the methylase coding sequence. In the presence of Em, an Em-bound ribosome stalls while translating the leader peptide, resulting in opening of the leader structure and high-level methylase translation.

Figure 1.

RNA (254 nt) encoded by pYH196.

A. Schematic diagram of wild-type ermC mRNA. Stem–loop structures representing the leader region, ermC stem and transcriptional terminator are shown. Translational signals for the leader peptide and methylase coding sequences are indicated. RBS, ribosome binding site. The site of Em-induced ribosome stalling is at codons 5–9 of the leader peptide coding sequence. Location of HpaI sites used to create pYH196, the ermC deletion derivative, are indicated by arrows with Δ symbol.

B. ermC mRNA encoded by pYH196. The 5′-proximal 80 nt of wild-type ermC sequence are retained, followed by the 3′-terminal 175 nt of ermC. DNA and RNA probes complementary to sequences of the 254 nt RNA are shown below the schematic.

C. Northern blot analysis of pYH196 ermC RNA isolated from a wild-type strain (wt) and a pnpA deletion strain (pnpA). The blot is overexposed to demonstrate the absence of detectable decay intermediates in the wild-type strain. The probe was the 5′ oligonucleotide probe. The extent of RNA fragments is indicated schematically on the right. The arrow on the left points to a prominent decay intermediate, whose 3′ ends map just downstream of the ermC stem.

In addition to induction of translation by Em, ermC mRNA stability is increased in the presence of Em (Bechhofer and Dubnau, 1987). We have shown that Em-induced ribosome stalling confers stability to diverse downstream RNA sequences (DiMari and Bechhofer, 1993). Based on the ermC model, we have argued that the initiation of decay in B. subtilis is determined by the accessibility of the 5′ end. Initiation of decay is by a ‘5′-binding endonuclease’ that gains access to the message by binding to the 5′ end and then cleaves endonucleolytically at a target site that may be near the 5′ end (Bechhofer and Zen, 1989). This is similar to recent models for initiation of mRNA decay in E. coli by RNase E (Coburn and Mackie, 1999), which is thought to be the major decay-initiating activity in E. coli and which requires a free 5′ end for binding to the message (Mackie, 1998). A sequence homologue to RNase E is not identifiable in the B. subtilis genome, although indirect evidence for an RNase E-like activity in B. subtilis has been obtained (Condon et al., 1997).

Earlier studies showed that RNA turnover in B. subtilis was primarily phosphorolytic, rather than hydrolytic as in E. coli (Duffy et al., 1972). The basis for this difference was shown to result from the absence in B. subtilis of a hydrolytic 3′-to-5′ exonuclease activity like that of E. coli RNase II (Deutscher and Reuven, 1991). Rather, the phosphorolytic 3′-to-5′ exonuclease activity of polynucleotide phosphorylase (PNPase) was the major degradative activity in B. subtilis extracts. Subsequently, pnpA, the PNPase-encoding gene of B. subtilis, was identified (Luttinger et al., 1996; Mitra and Bechhofer, 1996). Although a strain that was deleted for the pnpA gene displayed several phenotypic traits (Luttinger et al., 1996; Wang and Bechhofer, 1996), the viability and relative vigour of this strain suggested the existence of other, yet to be identified exoribonucleases that participate in mRNA turnover.

Decay of bacterial mRNA in wild-type strains is usually observed as an all-or-nothing process. That is, in Northern blot analysis of mRNA decay after inhibition of transcription by rifampicin, the signal for the full-length transcript is seen at early times and then gradually disappears, without any detectable decay intermediates. Examples of B. subtilis transcripts that are processed by endonucleolytic cleavage have been reported (Ebbole and Zalkin, 1988; Condon et al., 1996; Vázquez-Cruz and Olmedo-Alvarez, 1997; Homuth et al., 1999). The endonuclease activity or activities responsible for such cleavages has not been determined. In the B. subtilis pnpA deletion strain, ermC mRNA decay intermediates could be detected (Bechhofer and Wang, 1998). These RNA fragments all started at the transcriptional start site and ended at various downstream sites. However, a decay intermediate that represents the product of endonucleolytic cleavage (i.e. has a 5′ end other than the transcriptional start site) was not observed.

In this report, the combination of a small deletion de-rivative of ermC mRNA and induced ribosome stalling allowed the detection of an endonuclease cleavage product.

Results

Construction and analysis of a 254 nt RNA encoded by an ermC deletion derivative

To facilitate analysis of processing of the 910 nt ermC mRNA in B. subtilis, deletions of the ermC gene were constructed. The ermC gene has a unique HpaI site, as shown in Fig. 1A. Plasmids were constructed that contained an additional HpaI site inserted at various 5′-proximal locations. Deletion of the internal HpaI fragment generated ermC derivatives encoding considerably smaller transcripts. One of these chosen for further study was plasmid pYH196, which encodes a 254 nt RNA with an in frame fusion of the leader peptide and methylase coding sequences (Fig. 1B). The 254 nt RNA has a 12 nt sequence at the 5′-terminus (placed there to facilitate the construction of insertion mutations) that replaces the +1 A residue of the native ermC mRNA. The 254 nt RNA has the same transcription terminator sequence as native ermC mRNA and contains a 62-amino-acid coding se-quence, the translation of which is dependent on ribosome binding at RBS1. Thus, the 254 nt RNA encoded by pYH196 is a mini-messenger RNA, which serves as a model for B. subtilis mRNA processing.

Northern blot analysis was performed on steady-state RNA isolated from wild-type and pnpA strains carrying pYH196 (Fig. 1C). The probe was a 5′ end-labelled oligonucleotide probe, which is complementary to nts 21–38 of the ermC transcriptional unit (Fig. 1B, ‘5′ oligo probe’). The results show that 5′-proximal RNA fragments are readily detectable in the pnpA strain but not in the wild type. We assume that the fragments seen in the pnpA strain are normally degraded rapidly by PNPase in the wild-type strain. The probe detected a particularly strong band of about 150 nt (arrow in Fig. 1C). Using riboprobes complementary to downstream regions of the transcript, as well as Northern blot analysis from high-resolution polyacrylamide gels (Farr et al., 1999), the 3′ ends of these RNAs were found to be at a site of predicted secondary structure located 70 nt downstream of the ermC HpaI site (data not shown). In previous work, we identified this structure as one that is susceptible to decay by PNPase but that constitutes a block to 3′-to-5′ exonucleolytic decay specified by a ribonuclease other than PNPase (Farr et al., 1999). The predicted secondary structure that gives rise to this decay intermediate is referred to as the ‘ermC stem’.

Decay of the 254 nt ermC RNA encoded by pYH196, in the pnpA strain, was followed by Northern blot analysis of samples isolated at times after rifampicin addition, in the absence or presence of Em (Fig. 2). For the +Em samples, Em was added 15 min before the addition of rifampicin. The probe was a uniformly labelled antisense RNA, riboprobe A, which was complementary to the 3′-proximal region of the 254 nt RNA (Fig. 1B). In the absence of Em, the 254 nt RNA decayed with a half-life of 5.5 min. In the presence of Em, the full-length RNA had a half-life of 12.7 min, and an additional band was detected that was 45 nt shorter than the full-length RNA. This band was detectable also upon Em addition to the wild-type strain carrying pYH196 (not shown). The intensity of the shorter band increased over time as the intensity of the full-length band decreased. Plotting of relative number of counts in these two bands showed an inverse relationship between the decrease in full-length ermC RNA and the increase in the shorter RNA (Fig. 2B). Thus, it appears that full-length pYH196 RNA undergoes a processing event to give a new RNA that accumulates upon further incubation in the presence of rifampicin. This newly discovered RNA fragment was designated the ‘processed ermC RNA’.

Figure 2.

Detection of processed ermC RNA in the presence of Em.

A. Northern blot analysis of RNA samples taken at times indicated above each lane (min) after rifampicin addition. The probe was riboprobe A. Sizes (in nt) of the detected RNAs are indicated on the left. Schematic diagrams of the detected RNAs are on the right.

B. Plot of relative counts for the full-length RNA (closed squares) and the processed ermC RNA (open squares). Data were obtained using a Molecular Dynamics PhosphorImager.

To confirm that the processed ermC RNA was a product of the full-length RNA and not the result of transcription from an alternative ermC promoter that was somehow activated in the presence of Em, the pYH196 ermC transcriptional unit was placed under the control of the pspac promoter, which is IPTG inducible. This construct was integrated in the B. subtilis chromosomal amyE locus. The presence of the gene in single copy allows transcription to be shut off almost completely in the absence of IPTG. Figure 3A is a Northern blot analysis of IPTG-induced pYH196 ermC RNA expression. Em was added 15 min before the addition of IPTG. At the time of IPTG addition (time zero), the full-length and processed ermC RNAs were barely detectable (not visible in the exposure shown in Fig. 3A). At 1 min after IPTG addition, the full-length RNA was easily detected, and the amount of this RNA increased linearly until about 4 min after induction. At 2 min after the addition of IPTG, there was a 13-fold increase in the amount of the full-length RNA (relative to the zero time point) but only a 2.2-fold increase in the amount of the shorter RNA. The lag in the appearance of the processed ermC RNA suggests that it is not a new transcript induced by Em, but is rather the result of processing of the full-length form.

Figure 3.

A. Northern blot analysis of RNA isolated from a pnpA strain containing the 254 nt RNA under pspac promoter control. Above each lane are times after IPTG addition. The probe was riboprobe A. Schematic diagrams of the detected RNAs are on the right.

B. Northern blot analysis of RNA isolated from a pnpA strain carrying the RBS1 mutant ermC gene. The probe was riboprobe C. 5′ end-labelled DNA fragments of TaqI-digested plasmid pSE420 (Brosius, 1992) were used as marker (lane M). Fragment sizes (in nucleotides) are indicated on the left.

The effect of adding Em could be due to ribosome stalling or some other consequence of the presence of Em. To test this, a mutant of pYH196 was made that had an altered RBS1 sequence (AGGAGG to AGATCT), which was inactive for ribosome binding. For this mutant, no difference was observed in the pattern of RNAs detected with or without the presence of Em (Fig. 3B). Thus, ribosome binding at RBS1 and subsequent stalling in the presence of Em are required to generate the processed ermC RNA.

Reverse transcriptase analysis was performed to determine the 5′ end of the processed ermC RNA. The results (Fig. 4A) showed that an RNA that was 45 nt shorter was detected in the presence of Em but not in its absence. The 5′ end of this RNA is at the fifth codon of the ermC leader peptide coding sequence (Fig. 4B), which is the site of Em-induced ribosome stalling (Mayford and Weisblum, 1989). Of note is that, in this and other reverse transcriptase mapping experiments, a single major reverse transcriptase product for the processed ermC RNA was identified, rather than a ladder of bands.

Figure 4.

Reverse transcriptase mapping of the 5′ end of processed ermC RNA.

A. RNA isolated from a pnpA strain grown in the absence (–) or presence (+) of Em. The identity of the mapped 5′ ends is indicated on the right. A sequence ladder (lanes A, C, G and T) was run as a size standard.

B. Nucleotide sequence of the 5′ end of the leader peptide coding sequence, with encoded amino acids. The site of ribosome stalling (codons 5–9) is indicated by asterisks. The endonuclease cleavage site, as determined from the reverse transcriptase analysis, is at codon 5 (the exact 5′ end may be ±1 nt).

C. Reverse transcriptase analysis of RNA isolated from pnpA strains carrying pYH196 or the pYH196 derivative that was deleted for the first 10 nt of the transcriptional unit, in the absence (–) or presence (+) of Em.

As mentioned above, pYH196-encoded ermC RNA contained 12 nt at the 5′ end that are not present in the native ermC mRNA. To test whether processing was dependent on these 12 nt, a pYH196 derivative was constructed in which 10 of these 12 nt were deleted, giving a transcriptional start site that was only 1 nt upstream of the native start. Reverse transcriptase analysis showed that, for this RNA, the same processing was observed in the presence of Em (Fig. 4C). Thus, an RNA with the native ermC 5′-terminal sequence also undergoes processing that can be observed only in the presence of ribosome stalling. Additionally, this experiment demonstrated that the processing site is specific and not a function of the distance from the 5′ end.

Exonucleolytic versus endonucleolytic processing

Two possibilities can be considered for the ribonucleolytic process that gives rise to the processed ermC RNA. It is possible that, in the absence of Em, pYH196 ermC RNA is degraded rapidly and processively by a 5′-to-3′ exoribonuclease. Although decay intermediates that are missing 5′-terminal sequences are not normally detectable, stalling of a ribosome in the ermC leader peptide sequence could result in a block to 5′-to-3′ exonucleolytic decay, allowing the detection of an RNA decay intermediate. This possibility would be contrary to biochemical results to date and current genomic sequence data, both of which yield no evidence for a 5′-to-3′ exoribonuclease in bacteria (Deutscher and Li, 2000; Zuo and Deutscher, 2001). The other explanation for the processed ermC RNA is that it is caused by endonucleolytic cleavage at the +45 site. Cleavage at this site results in two RNA fragments: an upstream fragment of 45 nt, and a downstream fragment of 209 nt. The upstream 45 nt fragment was not detected even on overexposure of Northern blot analyses of strains carrying pYH196. This could result from either rapid degradation of the 45 nt fragment, which has an unprotected 3′ end, or perhaps the size of this fragment was too small for efficient transfer and detection by riboprobe C.

To increase the chance of detecting the upstream cleavage product, the deleted ermC gene from pYH196 was cloned on a plasmid pE194 copy number mutant that replicates at 200 copies per cell (Villafane et al., 1987). This is about a 14-fold higher level than the copy number of pYH196, which contains a wild-type pE194 replicon that replicates at 14 copies per cell. Short and long exposures of a Northern blot of RNA isolated from a wild-type strain carrying the high-copy pYH196 derivative are shown in Fig. 5. (The wild-type strain was used because Northern blot analysis of RNA isolated from a pnpA strain is complicated by the detection of 5′-terminal decay products. However, similar results to those shown in Fig. 5 were also obtained with RNA isolated from the pnpA strain.) From the short exposure, the processed ermC RNA was observed only in the presence of Em (Fig. 5A). A longer exposure showed an Em-dependent accumulation of a group of fragments that ran just above the 77 nt marker (Fig. 5B). From experiments using riboprobes B and C, as well as the 5′ oligo probe (Fig. 1B), this group of small RNAs was mapped to the 5′ end, i.e. their 5′ ends were co-terminal with the 5′ end of the 254 nt RNA, as shown schematically in Fig. 5B. In an earlier report, we demonstrated that such fragments are the result of a block to 3′-to-5′ exonuclease decay of the full-length RNA, which results from the stalled ribosome (Bechhofer and Wang, 1998). The precise sizes of these RNAs were determined by running RNA samples on a sequencing gel and performing Northern blot analysis (Fig. 5D). Two sets of RNAs were detected, one 77–78 nt in length and the other a ladder of RNAs 87–94 nt in length. We deduce that these minor RNAs are the result of a ribosome-mediated block to 3′-to-5′ exonucleolytic decay of the full-length 254 nt RNA. Thus, they are only observed in the presence of Em.

Figure 5.

Short and long exposures of Northern blot analysis of RNA isolated from a wild-type strain carrying the high-copy-number pYH196 derivative.

A. Short exposure, showing the processed ermC RNA only in the presence of Em. RNA was isolated at the time of rifampicin addition (zero) and 10 min thereafter. The probe was riboprobe BC.

B. Longer exposure of the Northern blot shown in (A). A schematic of the RNAs between 77 and 94 nt in length, observed only in the presence of Em, is on the right. These RNAs begin at the 5′-terminus and end downstream of the stalled ribosome. The exact sizes were mapped by Northern blot analysis of total RNA (from the 0 and 10 min time points in the presence of Em) run on a high-resolution denaturing polyacrylamide gel alongside a sequencing ladder, as shown in (D).

C. Longest exposure of the Northern blot shown in (A), revealing the presence of small RNAs running at <46 nt. The schematic on the right indicates that these RNAs are likely to be small fragments generated by a combination of endonuclease cleavage and exonuclease digestion, which are protected by the stalled ribosome.

In the longest exposure, shown in Fig. 5C, a group of very small fragments that runs just below the 46 bp DNA marker was also seen, and these accumulated in the presence of Em. Although the size of these fragments is consistent with their being upstream products of endonucleolytic cleavage (expected to be 45 nt), these fragments were not detected with the 5′ end-specific oligonucleotide probe (data not shown), indicating that they were not 5′-terminal fragments. More likely, these RNAs are fragments that result from a combination of endonucleolytic cleavage at +45 and 3′ exonucleolytic decay that is blocked by the stalled ribosome (as shown schematically in Fig. 5C).

To test whether the upstream cleavage fragment could be detected if it was larger, high-copy-number derivatives of pYH196 were constructed that contained additional 5′-terminal sequences (derived from the E. coli lacZ gene sequence and chosen for the absence of any predicted secondary structure). Northern blot analysis was performed on ermC RNAs with insertions of 27 and 73 nt at the 5′ end, which were isolated in the presence of Em. The results in Fig. 6A show that the same processed ermC RNA was present in the insertion constructs. Note that, in Fig. 6A, the full-length RNAs containing inserts of 27 and 73 nt were less stable than pYH196 RNA. We presume that this is because the addition of 5′-terminal sequences places the 5′ end further away from the stalled ribosome, thus lessening its 5′ stabilizing effect, as we have found earlier (Bechhofer and Zen, 1989). Upon longer exposure of this blot, an upstream endonuclease cleavage fragment of the appropriate size could not be detected (Fig. 6B). Indeed, the constructs with insertions of 27 and 73 nt gave the same <46 nt RNA fragments as were detected for pYH196, indicating that they are not the product of endonucleolytic cleavage (which would be 27 and 73 nt larger in the constructs with inserts). Apparently, the upstream cleavage fragment cannot be detected because it is extremely unstable.

Figure 6.

Northern blot analysis of pYH196 derivatives with 5′-proximal insertions of 27 and 73 nt. The probe was riboprobe C.

A. Shorter exposure shows processed ermC RNA (209 nt). RNA was isolated from cultures grown for 15 min in the presence of Em, at the time of rifampicin addition (zero) and 10 min thereafter. The expected sizes of full-length RNAs are indicated on the left.

B. Longer exposure of the same Northern blot, with schematic diagrams of small RNAs on the right.

[For the RNAs isolated from strains containing the 27 nt or 73 nt insertions, we might have expected the addition of Em to allow the detection of RNA fragments that were 27 and 73 nt larger than the group of RNAs measuring 77–94 nt in length. However, these were not observed (Fig. 6B). Most probably, the addition of 27 and 73 nt, which displaces the ribosome stall site away from the 5′ end, resulted in extreme instability of such RNA fragments – as we noted above for the full-length RNA.]

Ribosome-dependent versus ribosome-independent endonucleolytic cleavage

The observed endonuclease cleavage could be a result of ribosome stalling, or the endonuclease cleavage could be constitutive but only observable when ribosomes stall. In other words, it was possible that stalling of the ribosome itself triggered endonuclease cleavage (see Discussion). Thus, the 209 nt cleavage product is not observed in the absence of Em because endonuclease cleavage only takes place when ribosome stalling occurs. Alternatively, endonuclease cleavage could be constitutive. However, only in the presence of Em-induced ribosome stalling can the downstream cleavage product be detected, because of the stabilizing effect of the stalled ribosome at its 5′ end.

Additional mutants of the pYH196-encoded ermC RNA were made to try and distinguish between these possibilities. As shown in Fig. 1, only a portion of the ermC leader sequence is present on the 254 nt RNA. Derivatives were constructed that contained increasingly longer segments of the leader sequence. It was demonstrated earlier that Em-induced ribosome stalling is dependent on the presence of the first nine codons of the leader peptide coding sequence and does not require the rest of the leader sequence (Hue and Bechhofer, 1991). Thus, if the observed endonuclease cleavage was triggered by ribosome stalling, we would expect to see the same level of downstream cleavage product, independent of whether a short or long portion of the ermC leader sequence was present. However, if endonuclease cleavage was not a function of ribosome stalling but was specific for a certain sequence/structure (and ribosome stalling only enabled detection of the downstream cleavage product), then we might expect that changing the sequence/structure of the leader region near the cleavage site would affect the level of cleavage.

Leader regions of pYH196 and two derivatives with additional ermC leader sequences (pYH263 and pYH264) are shown schematically in Fig. 7A–C. In each case, ribosome stalling takes place at codons 5–9. From the Northern blot analysis, it is apparent that much less of the cleavage product was observed for pYH263- and pYH264-encoded RNAs. Note that, in the presence of Em, the amount of pYH263 and pYH264 full-length RNA decreased only slightly between 0 and 10 min after rifampicin addition. This indicates that the addition of Em was causing ribosomes to stall at the ribosome stall site, providing stability to the full-length RNA. Despite ribosome stalling, there was a reduced amount of cleavage product. This suggests that the cleavage itself, or the stability of the cleavage product, is affected by elements other than the stalled ribosome.

Figure 7.

A–C. Northern blot analysis of RNA encoded by pYH196 derivatives with longer portions of the ermC leader region. The probe was riboprobe BC. The extent of ermC leader region present is indicated by the schematic diagrams to the left of each Northern blot.

D. Reverse transcriptase analysis of 5′ ends from wild-type ermC mRNA and pYH196 ermC mRNA isolated from strains grown in the absence (–) and presence (+) of Em. The identity of the mapped 5′ ends is indicated on the right.

Owing to the size of the wild-type ermC mRNA, it is difficult to distinguish full-length from processed ermC RNA by Northern blotting. (The constructs examined in the Northern blots above were all deleted for much of the ermC coding sequence.) In order to determine whether detectable processing was occurring in wild-type ermC mRNA, reverse transcriptase analysis was performed. The result is shown in Fig. 7D. Although the addition of Em to the strain carrying pYH196 resulted in the appearance of abundant reverse transcriptase product corresponding to the 5′ end of processed ermC RNA, the addition of Em to the strain carrying wild-type ermC had no such effect. Thus, although ribosome stalling was presumably occurring on the wild-type ermC mRNA, little, if any, processing was detected.

Involvement of known B. subtilis endoribonucleases

Two narrow-specificity endoribonucleases have been identified in B. subtilis: Bs-RNase III, encoded by the rncS gene (Oguro et al., 1996; Wang and Bechhofer, 1997); and RNase M5, encoded by the rnmV gene (Condon et al., 2001). Although we have shown that rncS is an essential gene, we have isolated an rncS null strain that apparently contains a suppressor mutation (Herskovitz and Bechhofer, 2000). To test whether the observed cleavage at the +45 site resulted from Bs-RNase III, the rncS null strain was transformed with pYH196, and RNA was analysed by Northern blotting. Cleavage of the 254 nt RNA was observed as in the wild-type strain (Fig. 8, left).

Figure 8.

Northern blot analysis of the 254 nt RNA in the rncS null strain and the rnmV-disrupted strain. The probe was riboprobe BC.

To test whether the observed cleavage at the +45 site resulted from RNase M5, an rnmV-disrupted strain (Condon et al., 2001) was tested. The results in Fig. 8 (right) show that the same 209 nt cleavage product was detected. The amount of cleavage was much reduced in the rnmV strain (about threefold less than in the rncS strain). We attribute this to the fact that the rnmV strain was constructed by the integration of a pMUTIN plasmid (Vagner et al., 1998), which carries a constitutively expressed EmR gene. Therefore, there is a high basal level of methylated ribosomes, and the addition of Em in the experiment would lead to less ribosome stalling than in a strain that contains either no methylase-encoding gene or an inducible ermC gene. We conclude that a different, as yet unknown, endoribonuclease is responsible for cleavage at the +45 site.

Discussion

In the presence of Em-induced ribosome stalling, the 254 nt ermC deletion derivative is processed to give a 209 nt RNA. This is consistent with endonucleolytic cleavage at the site of ribosome stalling. The alternative explanation for this 209 nt product is a ribosome-mediated block to a 5′-to-3′ exoribonuclease. As mentioned above, there is no evidence for a 5′-to-3′ exoribonuclease in bacteria. In addition, the pattern of the reverse transcriptase analyses shown in Figs 4 and 7 is inconsistent with a block to exonuclease activity. RNA fragments that result from a block to exonuclease processivity would be expected to show a ‘ladder’ pattern. This was observed previously with a block to PNPase processivity mediated by a strong stem–loop structure (Farr et al., 1999). Similarly here, a ladder of small (77–94 nt) RNA fragments that were detected only upon the addition of Em (Figs 5B and 6B) were shown to be the result of a block to 3′-to-5′ exonuclease activity by the stalled ribosome. On the contrary, the reverse transcriptase patterns showed a unique 5′ end, from which we conclude that the 209 nt RNA is the result of endonucleolytic cleavage.

The upstream product of this cleavage could not be detected, even when the RNA was transcribed from a high-copy-number plasmid and probed with highly labelled riboprobes. The upstream fragment from the 254 nt RNA would be only 45 nt in length and would contain a 3′ end without any predicted secondary structure. Thus, it is quite likely that this fragment is extremely unstable and is degraded too rapidly to be detected. We have tried several times in the past to use S1 and RNase protection analysis to monitor ermC RNA processing. These attempts have been uniformly unsuccessful, and we believe that this is because of the extremely low GC content (26%) of ermC mRNA.

In an effort to detect the upstream cleavage product, its size was increased by inserting various sequences near the 5′ end of the deleted ermC gene carried on pYH196 (Fig. 5). These insertions did not affect the site of endonuclease cleavage, but an upstream product of the predicted size still could not be detected, even in the pnpA strain. The inability to detect this fragment is consistent with the general observation of an all-or-nothing pattern of mRNA decay. We hypothesize that the downstream cleavage product is easily detectable because it is protected by a stalled ribosome at the 5′ end and a transcription terminator structure at the 3′ end.

A major question raised by observation of the processed ermC RNA is whether cleavage at the +45 site is a result of ribosome stalling or is independent of ribosome stalling but is only detectable when the product is stabilized by the stalled ribosome. Earlier results from Sandler and Weisblum (1989) regarding the ermA transcript in B. subtilis are germane in this regard. The ermA leader region contains two leader peptide coding sequences, which have no sequence similarity to the ermC leader peptide coding sequence. The addition of Em results in ribosome stalling in both the ermA leader peptide coding sequences, as well as the appearance of shorter ermA RNAs. Using S1 nuclease analysis, Sandler and Weisblum (1989) mapped the 5′ ends of the pro-cessed ermA RNAs, and these were at codon 5 of the first leader peptide coding sequence and codons 5 and 7 of the second leader peptide coding sequence. Thus, the addition of Em resulted in RNA processing at sites of ribosome stalling that correspond to what we observed with ermC, despite the lack of sequence similarity. This indicates that processing is a result of ribosome stalling.

Cleavage of an mRNA on which ribosome stalling has occurred is a plausible mechanism for regulating translation. One could conjecture that a stalled ribosome is an indication of insufficient resources (e.g. amino acids, charged tRNAs) or an unsuitable template (fragmented or damaged RNA), signalling the cell to inactivate the mRNA functionally. Endonuclease cleavage at the site of ribosome stalling, followed by rapid degradation of the upstream fragment, would prevent further translational initiation, as the ribosome binding site would have been eliminated. The endoribonuclease activity that cleaves mRNA in response to a stalled ribosome could be specified by a soluble factor, a ribosome-associated protein or a ribosomal protein itself. Indeed, it has been reported that ribosomal protein S16 of E. coli has DNA endonuclease activity (Oberto et al., 1996). A model for ribosome-mediated RNA cleavage has been proposed for the processing of fimbrial messenger RNA in E. coli, encoded by the daa operon (Loomis and Moseley, 1998; Loomis et al., 2001). Cleavage of daa operon mRNA is dependent on the translation of a peptide sequence located in the daaP gene and, according to a recently published model, the cleavage may result from a ribosome-associated activity (Loomis et al., 2001).

Our results with RNAs that contain longer portions of the ermC leader region suggest that, even if cleavage is ribosome directed, the sequence/structure of the cleavage site affects the level of cleavage. We found that the abundance of endonuclease cleavage was affected by sequences downstream of the ribosome stalling site (Fig. 7A–C). Furthermore, by reverse transcriptase analysis, we have been unable to detect cleavage of the wild-type (910 nt) ermC mRNA in response to Em addition (Fig. 7D). If we assume that ribosome stalling is maximally efficient when the wild-type leader sequence is present, and if cleavage was ribosome mediated, we would expect to see abundant processing of the wild-type ermC mRNA. However, efficient cleavage occurs only in the context of the pYH196 5′-proximal sequence, and the ability to cleave this RNA is adversely affected by the addition of downstream sequences. Alternatively, the amount of cleavage for the constructs with longer ermC leader regions could be similar in all cases, but stabilization of the otherwise unstable downstream cleavage product (which is a function of ribosome–mRNA interactions) is affected by the presence of longer ermC leader region sequences.

The fact that ribosome stalling at the 5′ end of the processed ermC RNA stabilizes this fragment, and even allows it to accumulate at later times after rifampicin addition, indicates that further cleavage of the processed ermC RNA is dependent on access from the 5′ end. This is reminiscent of recent reports concerning the 5′ end dependence of E. coli RNase E endonuclease cleavage (Mackie, 1998; Jiang et al., 2000), which is thought to be a major decay-initiating step. Protection of downstream RNA by a paused ribosome has also been observed in E. coli (Björnsson and Isaksson, 1996).

The narrow-specificity endoribonucleases Bs-RNase III and RNase M5 were shown not to be required for cleavage (Fig. 8), and it is unlikely that the previously characterized broad-specificity endoribonucleases of B. subtilis (Mathur et al., 1993) are responsible for the specific endonuclease cleavage described here. Thus, the nature of the endonuclease activity responsible for processed ermC RNA is not known. To identify the B. subtilis endonuclease activity responsible for the observed in vivo cleavage, it will be necessary to establish in vitro conditions in which similar endonuclease cleavage can be recapitulated.

Experimental procedures

Bacterial strains

The wild-type B. subtilis host was BG1, which is trpC2 thr-5. The pnpA mutant host was BG119, a derivative of BG1 in which an internal portion of the pnpA gene was replaced with a kanamycin resistance cassette (Wang and Bechhofer, 1996). The rncS and rnmV strains are described in the text. The preparation of B. subtilis growth media and competent B. subtilis cultures took place as described previously (Dubnau and Davidoff-Abelson, 1971). E. coli strain DH5α (Grant et al., 1990) was the host for plasmid constructions.

Plasmids

The starting ermC plasmid was pYH40 (Hue et al., 1995), which contains a deletion of a portion of the ermC leader region. Additional HpaI sites were introduced individually into the ermC gene carried on pYH40, using oligonucleotide-directed mutagenesis (Kunkel et al., 1991). One such deriva-tive was chosen, which had the upstream HpaI site located 80 bp from the transcriptional start site. An HpaI deletion was made, giving plasmid pYH196. The 12 nt insert that is present at the 5′ end of pYH40 contains NcoI, BglII and HindIII cloning sites. The 27 nt insertion was made by annealing complementary olignonucleotides that contained 4-base overhangs for insertion into pYH196 DNA that had been linearized with NcoI and HindIII. A subsequent insertion to give the 73 nt insert used the same method: a 46 bp fragment was inserted into a unique BamHI site that was part of the 27 nt insert.

To create strains with the deleted ermC gene present in high copy number, plasmid pBD404, which is a cop-1 version of pBD144 (Bechhofer and Wang, 1998), was used. The cop-1 mutation results in a copy number of around 200 copies per cell (Villafane et al., 1987). Versions of the deleted ermC gene, carried on pYH196 derivatives, were amplified by polymerase chain reaction (PCR) and digested to give fragments with XbaI and StuI ends. These were cloned into similarly digested pBD404.

The IPTG-inducible ermC gene was inserted into the chromosomal amyE locus by cloning into pDR67 (Ireton et al., 1993), which contains a chloramphenicol resistance marker, amyE front and back segments to allow for integration by double cross-over, a pspac promoter upstream of a multicloning site and a lacI repressor gene.

The ermC gene carried on plasmids pYH263 and pYH264 (Fig. 7) was constructed by amplifying portions of the wild-type ermC leader region, using a downstream oligonucleotide primer that was complementary to a sequence upstream of RBS2 (in the case of pYH263) or a sequence at the base of the wild-type leader stem (in the case of pYH264). These oligonucleotides contained an HpaI recognition site, allowing for replacement of the shorter leader sequence in pYH231. Plasmid pYH231 is identical to plasmid pYH196, except that a MunI recognition site present downstream of the chloramphenicol resistance gene in pYH196 was changed to an EcoRI recognition site.

RNA analysis

Northern blot analysis, using 6% denaturing polyacrylamide gels, was performed as described previously (Wang and Bechhofer, 1997). Electroblotting from high-resolution gels was done as described previously (Farr et al., 1999). The extent of riboprobes A, B and C is shown in Fig. 1B. For some experiments, a riboprobe consisting of sequences B+C was used (‘riboprobe BC’). Uniformly labelled riboprobes were synthesized by T7 RNA polymerase transcription of gelpurified DNA fragments that were generated by PCR amplification of various portions of the pYH196 ermC deletion gene. The upstream primer in the PCRs contained the 17 nt T7 RNA polymerase promoter sequence at its 5′ end. Reverse transcriptase analysis was performed with 5–10 μg of total RNA, using Gibco BRL Superscript II RNase H reverse transcriptase.

Acknowledgements

This work was supported by Public Health Service grant GM-48804 from the National Institutes of Health.

Ancillary