|Home | About | Journals | Submit | Contact Us | Français|
The Bacillus subtilis rpsO gene specifies a small (388-nucleotide), monocistronic mRNA that encodes ribosomal protein S15. We showed earlier that rpsO mRNA decay intermediates accumulated to a high level in a strain lacking polynucleotide phosphorylase. Here, we used inducibly expressed derivatives of rpsO, encoding smaller RNAs that had the complex 5′ region deleted, to study aspects of mRNA processing in B. subtilis. An IPTG (isopropyl-β-d-thiogalactopyranoside)-inducible rpsO transcript that contained lac sequences at the 5′ end, called lac-rpsO RNA, was shown to undergo processing to result in an RNA that was 24 nucleotides shorter than full length. Such processing was dependent on the presence of an accessible 5′ terminus; a lac-rpsO RNA that contained a strong stem-loop at the 5′ end was not processed and was extremely stable. Interestingly, this stability depended also on ribosome binding to a nearby Shine-Dalgarno sequence but was independent of downstream translation. Either RNase J1 or RNase J2 was capable of processing lac-rpsO RNA, demonstrating for the first time a particular in vivo processing event that could be catalyzed by both enzymes. Decay intermediates were detected in the pnpA strain only for a lac-rpsO RNA that was untranslated. Analysis of processing of an untranslated lac-rpsO RNA in the pnpA strain shortly after induction of transcription suggested that endonuclease cleavage at 3′-proximal sites was an early step in turnover of mRNA.
The rate of decay of an mRNA is an important component in determining the level of gene expression. Studies of the mechanism of mRNA decay in Escherichia coli have progressed based on a detailed knowledge of the ribonucleases involved in the process and the construction of RNase mutant strains. A similar level of understanding of the mechanism of mRNA decay has not been achieved for the model gram-positive organism Bacillus subtilis. Earlier studies of a number of B. subtilis mRNAs, some of which were constitutively or inducibly stable, revealed that mRNA decay in B. subtilis initiates from the 5′ end (9) and that polynucleotide phosphorylase (PNPase) (encoded by the pnpA gene) plays a major role in 3′-to-5′ exonucleolytic degradation (15, 30, 42). More recently, the role of RNase J1 and RNase J2 ribonucleases (18) has become paramount. Of the two, only RNase J1 is essential, and reduced expression of RNase J1 results in an increase in global mRNA half-life, suggesting a general role for RNase J1 in initiation of decay. RNase J1 has been shown to be involved in decay and processing of a number of specific RNAs (3, 7, 14, 18, 44), and a recent study demonstrated the effect of reduced RNase J1 and/or lack of RNase J2 on hundreds of B. subtilis mRNAs (26). While the RNase J enzymes were initially purified on the basis of their endoribonuclease activity, it was shown subsequently that RNase J1 also has 5′-to-3′ exoribonuclease activity (27), which is inhibited by a 5′ triphosphate end (14, 25).
A model for mRNA turnover in B. subtilis involves RNase J1 in two possible pathways (1). For exonucleolytic decay from the 5′ end, the 5′ triphosphate is converted to a monophosphate by a pyrophosphatase activity (not yet identified for B. subtilis) similar to what has been found for E. coli (6, 12). For the endonucleolytic pathway, an mRNA bearing a 5′ triphosphate end can serve as a substrate, and endonuclease cleavage occurs at a downstream RNase J1 recognition site. The upstream product of cleavage is rapidly degraded by 3′-to-5′ exonuclease activity, primarily PNPase, while the downstream product, which has a 5′ monophosphate end, either is acted on by RNase J1 5′-to-3′ exonuclease activity or serves as a substrate for additional endonuclease cleavages.
Other endonucleases of B. subtilis that have been characterized to some extent include RNase III, RNase M5, RNase P, RNase Z, and Mini-III (8, 37), which are involved in stable RNA processing (10, 21, 23, 32, 37), and EndoA, which is part of a toxin-antitoxin system (31). None of these endoribonucleases has been shown to be involved in decay of an endogenous mRNA.
Previously, we used 5′-proximal oligonucleotide probes to analyze the steady-state pattern of decay intermediates from seven small, monocistronic mRNAs, comparing the pattern detected in a wild-type strain versus that detected in a PNPase-deficient strain (30). In all cases, decay intermediates were barely detectable in the wild-type strain, but prominent decay intermediates were detected in the pnpA strain, suggesting that PNPase was the primary 3′-to-5′ exoribonuclease responsible for turnover of RNA fragments. We suggested that, in the wild-type strain, endonuclease cleavage(s) in the downstream portion of the message produces an RNA fragment with an unprotected 3′ end. This is rapidly acted upon by PNPase, which is able to degrade processively through the secondary structures present in the body of the mRNA. In the wild-type strain, the combination of endonuclease cleavage and processive 3′-to-5′ degradation prevents the accumulation of decay intermediates. In the pnpA strain, however, the remaining 3′ exonucleases are blocked at the 3′ side of structured RNA sequences, resulting in an accumulation of decay intermediates.
One of the mRNAs studied was rpsO mRNA, a 388-nucleotide (nt) mRNA that encodes ribosomal protein S15 (Fig. (Fig.1A).1A). Translation of E. coli rpsO mRNA is negatively regulated by binding of S15 protein to the 5′ end of its own mRNA, which results in trapping of the ribosome at its loading site (28, 34-36). Since the pseudoknot structure that is involved in E. coli rpsO mRNA regulation is predicted to be present in B. subtilis rpsO mRNA as well (41), we assume that regulation of rpsO expression in B. subtilis occurs by a similar mechanism. Thus, at the point in the growth curve at which our RNA experiments are performed we expect a significant portion of rpsO mRNA to be untranslated.
In the current study, we examined further the cis and trans elements involved in mRNA processing in B. subtilis, using RNAs encoded by rpsO deletion constructs that were inducibly expressed.
The B. subtilis host strain was BG1 (trpC2 thr-5), which is designated the wild type. B. subtilis RNase mutant strains used in this study were derivatives of BG1. These were constructed in earlier work by us and by others either by replacement of RNase gene coding sequences (CDS) with an antibiotic resistance marker or by placing expression of the RNase gene under IPTG (isopropyl-β-d-thiogalactopyranoside) control (Table (Table1).1). The E. coli host for lac-rpsO constructs was DH5α (22). Transformation of wild-type and RNase mutant strains with lac-rpsO plasmid constructs was done as described previously (16).
For construction of lac-rpsO-encoding plasmids, rpsO transcript sequences were amplified by PCR, using a 5′ primer that contained an XbaI site and a 3′ primer that contained an SphI site. rpsO sequences in lac-rpsO constructs were from nt 122 of the rpsO transcript, which is the middle nucleotide of codon 12 of the rpsO CDS, to 40 nt downstream of the rpsO transcription terminator. (The ATG sequence immediately adjacent to the XbaI site in lac-rpsO was included in the upstream PCR primer used to make this construct. The rpsO codon 12 is actually ATC.) Amplified DNA was cloned between the XbaI and SphI sites of pDR66 (24), which put transcription of rpsO sequences under the control of the Pspac promoter. All constructs were confirmed by DNA sequencing. Plasmid DNAs were linearized with NruI and used to transform BG1 by integration at the amyE locus. To make the lac-rpsO constructs with a mutated Shine-Dalgarno (SD) sequence, with a mutated lac operator, or with 5′ additions, a mutagenic PCR primer containing the sequence that includes the HindIII site (AAGCTT) next to the SD sequence (Fig. (Fig.1B)1B) was used in conjunction with an upstream primer that allowed amplification of a DNA fragment that could be digested with EcoRI and HindIII for insertion into pDR66.
RNA was isolated by hot phenol extraction from B. subtilis cultures grown in minimal medium containing Spizizen salts with 0.5% glucose, 0.1% Casamino Acids, 0.001% yeast extract, 50 μg/ml tryptophan and threonine, and 1 mM MgSO4, as described previously (13). Strains were grown to the late logarithmic growth stage (100 Klett units, using a no. 54 green filter). For RNase conditional mutant strains, strains were grown overnight in 2× YT (1% yeast extract, 2% tryptone, 1% NaCl) with selection for chromosomal and plasmid markers and with 1 mM IPTG. Cells were precipitated, washed twice in 2× YT, and used to inoculate cultures to an optical density at 600 nm of 0.1, with or without 1 mM IPTG present. Cultures were grown to an optical density at 600 nm of 0.6, at which point RNA isolation was performed. Northern blot analysis of RNA separated on 6% denaturing polyacrylamide gels was done as described previously (19). 5′-end-labeled oligonucleotide probes were prepared using T4 polynucleotide kinase (New England Biolabs) and [γ-32P]ATP. To control for RNA loading, membranes were stripped and probed for 5S rRNA, as described previously (40).
Primer extension analysis on 20 μg total RNA was done with Superscript III reverse transcriptase (RT) (Invitrogen) according to the manufacturer's instructions. The products of reverse transcription (see Fig. Fig.4)4) were separated on a 6% denaturing polyacrylamide gel.
Quantitation of radioactivity in bands on Northern blots was done with a Storm 860 PhosphorImager instrument (Molecular Dynamics) or a Typhoon Trio variable mode imager (GE Healthcare). RNA half-lives were determined by a linear regression analysis of percent RNA remaining versus time. Half-life data were derived only from experiments for which the R2 value was greater than 0.9. Comparison of the amounts of processing in wild-type and RNase mutant strains was done with a two-sample t test to derive P values. The corrected value for the half-life of the processed product of lac-rpsO RNA, which is decaying at the same time as it is being generated from the full-length RNA, was based on an approach described previously (17). Theoretical curves were fitted to the experimental data using the Solver feature of Microsoft Excel, and half-life values were obtained from the best fit.
The predicted free energy of formation of the secondary structure was determined at the Zuker website (http://www.bioinfo.rpi.edu/applications/mfold/cgi-bin/rna-form1.cgi), using the default conditions of 37°C, 0.1 mM RNA concentration, and 1 M NaCl.
We constructed a derivative of rpsO that is transcribed from the Pspac promoter (IPTG controlled) and that lacks the 5′-proximal rpsO sequences (nt 1 to 121) that are predicted to form a pseudoknot structure (41). This construct had several advantages. First, we wanted to examine decay of a small rpsO mRNA derivative that did not contain the 5′-proximal pseudoknot structure, which is involved in autoregulation by binding of S15 protein. The presence of this 5′ region results in a mixture of translated and untranslated RNAs, which is a complicating factor in the analysis of decay. Second, we planned to express rpsO RNA derivatives with altered sequences and therefore needed a tag to differentiate between endogenous rpsO mRNA (which is required for normal growth and cannot be altered) and the RNA encoded by our constructs. This tag was provided by the lac sequence at the 5′ end of the Pspac-driven transcript (see below). Conversely, deletion of the 5′ portion of rpsO in our constructs allowed us to probe for endogenous rpsO mRNA with a 5′-proximal probe, without detecting RNA encoded by our constructs. Third, an inducibly expressed rpsO RNA derivative could be used to follow accumulation of decay intermediates with time.
The deletion construct, which was integrated at the amyE locus, contained nt 122 to 388 of the rpsO transcript, fused to the Pspac promoter such that a 5′-proximal ribosome binding site was present, followed by a start codon and a 78-codon open reading frame. We call the RNA encoded by this construct lac-rpsO RNA. The construct included the strong secondary structure (nt 143 to 172 in the full-length rpsO mRNA) (Fig. (Fig.1B),1B), which we hypothesize causes accumulation of a prominent rpsO mRNA decay intermediate (the “180-nt RNA”) by blocking 3′ exonuclease activity present in the PNPase deletion strain (30). At least two other strong secondary structures are predicted to form near the transcription terminator (Fig. (Fig.1A),1A), and we have mapped 3′ ends of decay intermediates in the pnpA strain to the downstream side of these structures (30). A schematic diagram of lac-rpsO is shown in Fig. Fig.2A.2A. The diagram shows only two of several predicted strong stem-loop structures: one at the beginning of the lac-rpsO CDS that is a block to 3′ exonuclease processivity in the pnpA strain and one that serves as the transcription terminator. The sizes of the prominent decay intermediates from rpsO mRNA (180 nt) and from the lac-rpsO construct (102 nt) expected to be observed in the pnpA strain are indicated to the right of the upstream stem-loop in the figure.
The sequence at the 5′ end of lac-rpsO RNA is shown in Fig. Fig.1B,1B, starting from the +1 nucleotide of Pspac transcription, past the XbaI site that was used to make the construct. The lac portion of the sequence includes the sequence that constitutes the lac operator in the DNA, the SD sequence, and the start codon from which the rpsO portion of the transcript is read in frame. Also shown in Fig. Fig.1B1B is the extent of complementarity of the “upstream” and “downstream” lac oligonucleotide probes that were used to detect Pspac-driven lac-rpsO transcripts.
We analyzed the steady-state pattern of lac-rpsO RNA. RNA was isolated from wild-type and pnpA strains containing the lac-rpsO construct, grown in the presence of IPTG. Northern blot analysis was performed first using the 5′ rpsO-specific probe (Fig. (Fig.1A,1A, “5′ oligo probe”), which hybridizes to endogenous rpsO mRNA but not to lac-rpsO RNA, and then, after stripping, using the downstream lac probe (Fig. (Fig.1B),1B), which hybridizes to lac-rpsO RNA but not to endogenous rpsO mRNA. For the rpsO probe, the expected difference between the wild-type and pnpA strains was observed (Fig. (Fig.3A,3A, left): only full-length rpsO mRNA was detected in the wild-type strain, while decay intermediates were detected in the pnpA strain, including the prominent 180-nt RNA as well as larger decay intermediates that are close in size to the full-length RNA. For the downstream lac probe, the full-length lac-rpsO RNA, which is about 80 nt smaller than endogenous rpsO mRNA, was detected in the wild-type strain, as was an additional band that was slightly smaller than full length (Fig. (Fig.3A,3A, right). The same two bands were detected in the pnpA strain, but no decay intermediates were detected.
To resolve the nature of the doublet detected in the strain carrying lac-rpsO, a Northern blot analysis was performed on RNA that had been run for a longer time (compare migrations of marker fragments in Fig. Fig.3A3A versus B), and the blot was probed in succession with the upstream and downstream lac probes (Fig. (Fig.3B).3B). Only the upper band, representing full-length RNA, was detected by the upstream lac probe, whereas both bands were detected by the downstream lac probe. These results indicated that the 5′ end of the lower band was a short distance downstream from the expected start site for transcription. These two RNAs could be the result of alternate transcription start sites, or the shorter RNA could be a processed product of the longer RNA. To distinguish between these possibilities, a time course experiment of IPTG-induced transcription was performed (Fig. (Fig.3C).3C). The data showed that the full-length lac-rpsO RNA was detectable on this exposure as early as 1 min after the addition of IPTG but that the shorter RNA was detectable only after 4 min, indicating a precursor-product relationship. We concluded that the initial transcript is either cleaved endonucleolytically a short distance from the 5′ end or degraded exonucleolytically up to a point a short distance from the 5′ end.
To map the 5′ end of the processed fragment, primer extension with RT was performed (Fig. (Fig.4).4). The primer used was complementary to nt 161 to 181 of the rpsO sequence (Fig. (Fig.1A,1A, “RT primer”). In the absence of IPTG (i.e., lac-rpsO was not transcribed) (Fig. (Fig.4,4, lane 1), the expected full-length rpsO RT extension product (181 nt) was observed, as were several smaller bands, which are likely the result of inhibition of RT processivity by the presence of the 5′-proximal pseudoknot structure (41). In the presence of IPTG (i.e., the lac-rpsO construct was transcribed) (Fig. (Fig.4,4, lane 2), two new RT products were detected: one that was the expected size for an RT runoff product at the 5′ end of lac-rpsO RNA (103 nt), and a group of bands that was about 23 to 25 nt shorter than the RT runoff product. Thus, the 5′ end of the processed lac-rpsO RNA fragment was mapped to a location that was 4 nt upstream of the SD sequence (Fig. (Fig.1B),1B), but the 5′ end was not precise, as there were additional bands 1 nt longer and 1 nt shorter. Herein, the processed lac-rpsO RNA fragment is referred to as “+24lac-rpsO RNA.”
The 5′ end of the newly generated +24lac-rpsO RNA fragment is adjacent to the SD sequence, and competition with ribosome binding could delay further processing/decay of the downstream RNA product. We determined the stability of the two bands detected with the downstream lac probe. As shown in Fig. Fig.3D,3D, the half-life of the full-length lac-rpsO RNA was approximately 6.1 min, whereas the half-life of the +24lac-rpsO RNA was about 11.3 min (see Materials and Methods for the calculation of the +24lac-rpsO RNA half-life).
As shown in Fig. 3A and B, we could not detect lac-rpsO decay intermediates in the pnpA strain. We hypothesized that the absence of the 5′ regulatory region resulted in constitutive translation of lac-rpsO and that decay intermediates from native rpsO mRNA in the pnpA strain (30) were derived from untranslated molecules (see Discussion). Thus, interfering with translation of lac-rpsO should allow detection of lac-rpsO RNA decay intermediates in the pnpA strain. A derivative of lac-rpsO in which the SD sequence was mutated from AAGGAGG to AACCTCC, called lac-rpsO(SDX), was made (Fig. (Fig.2B).2B). lac-rpsO(SDX) RNA was isolated from wild-type and pnpA strains and analyzed by Northern blotting. The results shown in Fig. Fig.3E3E demonstrate that the lack of ribosome binding had two consequences. First, only a single lac-rpsO band was detected in the wild-type strain. This was confirmed in the primer extension assay (Fig. (Fig.4,4, lanes 3 and 4), where the 5′ end of a processed product was undetectable in the strain containing the lac-rpsO(SDX) construct (Fig. (Fig.4,4, compare lanes 2 and 4). In the Discussion, we explain possible reasons why the absence of ribosome binding may preclude detection of the +24lac-rpsO RNA. Second, while we found (Fig. 3A and B) that the steady-state patterns of lac-rpsO RNA were similar in the wild-type and pnpA strains, the patterns for lac-rpsO(SDX) were different in the wild-type and pnpA strains (Fig. (Fig.3E).3E). In the pnpA strain, the expected 102-nt decay intermediate was observed (Fig. (Fig.3E),3E), along with other prominent decay intermediates that had 3′ ends much farther downstream.
According to numerous studies of B. subtilis mRNA decay, it could be predicted that the nature of the 5′ end of lac-rpsO RNA would affect its stability, presumably by interfering with access of a decay-initiating RNase. To test the effect of the 5′-terminal secondary structure, constructs with a 26-nt sequence added to the 5′ end of lac-rpsO RNA were made. Three constructs were made (Fig. 2C to E): (i) lac-rpsO(5′-SS), with a strongly structured 5′ end that had a predicted free energy of −10.0 kcal mol−1; (ii) lac-rpsO(5′-MS), with a moderately structured 5′ end that had a predicted free energy of −5.7 kcal mol−1; and (iii) lac-rpsO(5′-US), with an unstructured 5′ end. The predicted intramolecular base pairing in the secondary structures for the 5′ ends of lac-rpsO(5′-SS) and lac-rpsO(5′-MS) is shown in Fig. Fig.5A.5A. The moderately structured and strongly structured sequences were patterned, respectively, after the wild-type and mutant forms of mdr (or bmr3) mRNA (29). The strongly structured sequence was shown to increase mdr mRNA stability at least fourfold. Northern blot analysis of RNA encoded by these constructs revealed striking changes in their steady-state concentrations relative to that of lac-rpsO RNA without a 5′ addition (Fig. (Fig.5B).5B). The lac-rpsO(5′-SS) RNA, with the most stable 5′-end structure, accumulated to a concentration that was about 33-fold greater than that of lac-rpsO RNA, while the concentration of lac-rpsO(5′-MS) RNA, which had a moderately stable 5′-end structure, was about eightfold greater than that of lac-rpsO RNA. The lac-rpsO(5′-US) RNA concentration was about half that of lac-rpsO RNA itself. From this experiment, it was difficult to tell whether the lower band seen in the case of RNAs encoded by lac-rpsO, lac-rpsO(5′-MS), and lac-rpsO(5′-US), which is the result of processing of full-length RNA, was present in significant amounts in the lac-rpsO(5′-SS) strain.
RNA half-lives were determined from Northern blot analysis (Fig. (Fig.5C)5C) of a gel that was run for a longer time than the one showed in Fig. Fig.5B,5B, in order to distinguish the upper and lower bands in the lanes containing abundant lac-rpsO RNA. The stability of these RNAs was consistent with the amount of RNA detected at steady state. The half-life of lac-rpsO RNA without any 5′ extension was 5.9 min (average from two experiments), and the half-life of the lac-rpsO(5′-US) RNA was 3.6 min. lac-rpsO(5′-SS) RNA showed essentially no decrease in concentration for the course of the experiment, and lac-rpsO(5′-MS) RNA had a half-life of 18.1 min. Notably, the lac-rpsO(5′-SS) RNA showed little, if any, evidence of a lower band, i.e., the presence of the 5′-terminal secondary structure appeared to completely inhibit processing at the 24-nt site.
To assess the effect of ribosome binding on the stability of RNAs with 5′-terminal structures, constructs that had the 5′ strong structure or the 5′ unstructured sequence in the context of a mutated SD sequence were made. These were designated lac-rpsO(5′-SS,SDX) and lac-rpsO(5′-US,SDX), respectively (Fig. 2F and G). Northern blot analyses of decay of these constructs and lac-rpsO(SDX) are shown in Fig. Fig.5D.5D. Only a single band was observed in all cases, similar to our earlier finding that mutating the SD site resulted in a loss of the downstream processing fragment (Fig. (Fig.3E,3E, wild-type strain). For lac-rpsO(SDX) RNA, the half-life was 2.1 min (average from three experiments), which was significantly less stable than that for lac-rpsO RNA itself (half-life of 5.9 min). The SD sequence mutation had a profound effect on the stability of RNA containing the 5′-terminal strong structure: the half-life of lac-rpsO(5′-SS,SDX) RNA was now 1.9 min, similar to that of lac-rpsO(SDX) and far less stable than that of the lac-rpsO(5′-SS) RNA, which did not decay at all in the course of the 20-min experiment (Fig. (Fig.5C).5C). For lac-rpsO(5′-US,SDX), the half-life was determined to be around 1.2 min, substantially less than the 3.6-min half-life for lac-rpsO(5′-US). These results demonstrated that 5′-terminal secondary structure alone was not enough to confer stability on the full-length RNA and that a proximal ribosome binding event was also required. When these RNAs were isolated from the pnpA strain (data not shown), we could detect the expected 128-nt decay intermediate (Fig. 2F and G).
To differentiate between the effects on stability of ribosome binding and translation of the lac-rpsO CDS, a final set of constructs that had a stop codon at codon 3, named lac-rpsO(stop), lac-rpsO(5′-SS,stop), and lac-rpsO(5′-US,stop), was made (Fig. 2H to J). Northern blot analysis was performed on these RNAs, and the observed effect of the premature stop codon on mRNA decay was remarkable (Fig. (Fig.5E).5E). For lac-rpsO(stop), the full-length RNA decayed with a half-life of 8.8 min (average from two experiments), and 75% of the RNA detected at time zero was in the form of an extremely stable lower band, presumably the +24lac-rpsO RNA. This amount was far larger than that for lac-rpsO itself, where the +24lac-rpsO RNA constituted on average ~20% of the total RNA detected (cf. Fig. Fig.5C).5C). For lac-rpsO(5′-SS,stop), the full-length RNA was completely stable for the time of the experiment, and the +24lac-rpsO RNA was present in low abundance. The results for the lac-rpsO(5′-US,stop) construct were similar to those for lac-rpsO(stop); full-length RNA decayed with a half-life of 5.9 min, but the +24lac-rpsO RNA, which constituted 80% of the RNA at time zero, was extremely stable. Taken together, the results shown in Fig. Fig.55 are consistent with a model in which 5′-end-dependent initiation of mRNA decay is blocked by the 5′-terminal secondary structure. Stability is conferred by such a structure even in the absence of translation of the body of the message, but nearby ribosome binding is required for stabilization.
In the pnpA strain carrying the lac-rpsO RNA construct with the SD mutation, prominent decay intermediates, including the expected 102-nt band, were detected at steady state (Fig. (Fig.3E).3E). We took advantage of the inducibility of lac-rpsO(SDX) to examine the accumulation of decay intermediates immediately after the onset of transcription. Specifically, we were interested in determining whether there was a temporal pattern to the appearance of intermediates, which could be informative as to the order of sites that are cleaved endonucleolytically in the decay process. Figure Figure66 shows a Northern blot analysis of a time course of IPTG-induced lac-rpsO(SDX) expression. This blot was probed with the downstream lac probe (Fig. (Fig.1B),1B), but an identical pattern was observed when this blot was stripped and probed with the upstream lac probe (see Discussion). Thus, RNA fragments that were detected contained the 5′ end of the lac-rpsO(SDX) RNA, and the heterogeneous sizes of the detected decay intermediates represented different 3′ ends. At the earliest time point after IPTG addition (time zero), only the full-length lac-rpsO(SDX) RNA (306 nt) was detected. At 1 min, two faint bands of ~280 and ~260 nt, representing decay intermediates with 3′ ends that were close to the native 3′ end, were observed. We have noted previously that a relatively strong secondary structure is predicted for rpsO sequences whose 3′ end corresponds to nt 272 of lac-rpsO RNA (30). At 2 min, these bands increased in intensity, and smaller bands that were ~230 nt became visible. Again, we have noted previously a predicted strong secondary structure in the rpsO sequence whose 3′ end corresponds to nt 230 of lac-rpsO RNA (30). At 3 min, fainter bands between ~150 and 200 nt were observed, as was the 102-nt band whose 3′ end was just downstream of the strong secondary structure shown in Fig. Fig.1B.1B. The latter band increased in intensity up to the 10-min time point.
We wished to determine which RNase was responsible for processing to result in the +24lac-rpsO RNA. Deletion mutants for nonessential endonucleases and conditional mutants for essential endonucleases, which are dependent on IPTG for growth, were obtained for known B. subtilis endonucleases (Table (Table1).1). Since the conditional mutant strains were to be grown in the presence and absence of IPTG, we needed to be able to express lac-rpsO RNA, which is under the control of the IPTG-inducible Pspac promoter, even in the absence of IPTG. For this, an A→C change was introduced into the lac operator at the eighth position (Fig. (Fig.1B),1B), which has been shown to interfere with lac repressor binding such that expression occurs in the absence of inducer (2). Control experiments showed that the resulting lac-rpsO construct expressed lac-rpsO RNA equally well in the presence and absence of IPTG (data not shown). The conditional RNase mutant strains also carried plasmid pMAP65 (33), which provided extra copies of the lac repressor to ensure full shutoff of transcription of endoribonuclease genes in the absence of IPTG.
The data in Table Table11 are the percentages of full-length lac-rpsO and +24lac-rpsO RNAs for each mutant, as determined by Northern blot analysis (examples of which are shown in Fig. Fig.7A).7A). Except for the RNase P mutant, there was no statistically significant effect of the absence of, or lower levels of, any of the ribonucleases. For RNase P, we observed an opposite effect: the relative amount of processed product was increased significantly when RNase P was depleted. A half-life experiment using the RNase P mutant strain grown with various levels of IPTG showed that the increase in the amount of +24lac-rpsO RNA was the result of an increase in RNA half-life, rather than an effect on processing (data not shown). The increased stability of +24lac-rpsO RNA was proportional to the increased doubling time for growth in the presence of 1 mM IPTG, 0.05 mM IPTG, and no IPTG.
Since RNases J1 and J2 have been shown to cleave at the same sites in vitro (18), we reasoned that we might not see an effect with a strain that only had reduced levels of RNase J1 or that only had RNase J2 deleted. Therefore, a strain that contained the lac-rpsO construct in the context of a conditionally expressed RNase J1 gene and a deleted RNase J2 gene was made. When this strain was grown in the absence of IPTG, a significant difference in the pattern of Northern blotting was observed (Fig. (Fig.7B).7B). Instead of a clearly defined lower band, there was a short smear underneath the upper band. This pattern was consistent in two repeats of this Northern blot analysis (not shown), and we have observed this type of smear previously with ΔermC mRNA in an RNase J1 conditional mutant (43). To confirm that the lower band was not present in the double mutant grown in the absence of IPTG, a Northern blot analysis was performed with RNA separated on a high-resolution gel. The results shown in Fig. Fig.7C7C demonstrate that there was little evidence of cleavage products in the RNase J double mutant. We conclude that the site located 24 nt from the 5′ end of lac-rpsO RNA is recognized for cleavage by both RNase J1 and RNase J2.
We showed that processing of lac-rpsO RNA to yield a decay intermediate that was shortened at the 5′ end by ~24 nt could be accomplished by either RNase J1 or RNase J2. This processing was not detected for native rpsO mRNA, and so it is likely a consequence of the lac sequence and ribosome binding site of lac-rpsO RNA and is not a factor in decay of native rpsO mRNA. Endonucleolytic cleavage at a particular site by both RNase J1 and RNase J2 was demonstrated previously in vitro (18), and recent data from a transcriptome analysis indicated that the stability of hundreds of mRNAs was affected only when both RNase J1 expression was lowered and RNase J2 was absent (26). However, we believe our results are the first demonstration in vivo that a specific transcript can be processed in the same way by either enzyme. These results highlight the need to do experiments also with the double mutant strain if no effect of RNase J1 depletion alone or RNase J2 knockout alone is observed.
Either of the two activities of RNase J could be involved in processing lac-rpsO RNA. (i) The +24 lac-rpsO RNA could be generated by 5′-to-3′ exonuclease processing up to the site of ribosome binding. RNA processing that involves a block to RNase J1 processivity by a bound ribosome has been observed in vitro for an RNA containing the STAB-SD sequence (27) and most recently in vivo for hbs mRNA (11). We note that the SD sequence of lac-rpsO RNA has 12 consecutive bases that are complementary to successive bases near the 3′ end of B. subtilis 16S rRNA (shown in Fig. Fig.1B),1B), potentially quite a strong interaction that could effectively block 5′ exonuclease processivity. Consistent with this model, the +24lac-rpsO RNA was not detected for lac-rpsO(SDX) RNA (Fig. (Fig.3E3E and and5D),5D), which had no SD sequence and presumably no ribosome binding. (ii) The +24lac-rpsO RNA could be generated by endonucleolytic cleavage at that site. The absence of this RNA from the lac-rpsO(SDX) construct could be explained either because cleavage is dependent on ribosome binding, as we have shown for the endonucleolytic cleavage in ΔermC mRNA (43), or because the +24lac-rpsO RNA may be detectable only if a ribosome is bound to its 5′ end to stabilize it. A third alternative is that both RNase J activities are involved, with an endonuclease cleavage close to the 5′ end being followed by 5′ exonucleolytic processing up to the ribosome binding site, similar to what has been described for processing of hbs mRNA (11).
Support for the endonuclease model for generation of the +24lac-rpsO RNA comes from the Northern blot analysis of decay intermediates from lac-rpsO(SDX) RNA in the pnpA strain (Fig. (Fig.6).6). Precisely the same pattern and band intensity was observed whether the downstream lac probe or the upstream lac probe (Fig. (Fig.1B)1B) was used. Since a transcriptome analysis of the pnpA strain did not detect any difference in expression of ykqC (RNase J1) or ymfA (RNase J2) (13), we may assume that, in the pnpA strain, RNase J functions as it does in the wild-type strain. If efficient RNase J attack on lac-rpsO(SDX) RNA were occurring exonucleolytically from the 5′ end, we should have observed a difference in the patterns of decay intermediates when using the two probes. If, however, the +24lac-rpsO RNA were the result of RNase J endonuclease cleavage that is dependent on ribosome binding, then the absence of ribosome binding on the lac-rpsO(SDX) RNA construct would result in no cleavage at the 24-nt site, and decay intermediates could easily be detected with the 5′-terminal probe. It is also possible that RNase J attack at the 5′ end is a slow step relative to more-rapid endonuclease cleavage in the body of the message. If so, then the observed pattern of decay intermediates (Fig. (Fig.6)6) could be explained by rapid downstream cleavage, followed by 3′-to-5′ exonuclease activity to result in the decay intermediates, which are acted upon by RNase J (either exonucleolytically or endonucleolytically) only slowly from the 5′ end. Analysis of RNA isolated at steady state would not be sensitive to the disparate rates of these processing steps.
Together with the observation that extreme stability is conferred by the 5′-terminal structure (Fig. (Fig.5C),5C), our results suggest the following model for processing of lac-rpsO RNA. RNase J recognizes the 5′ end of the RNA, and the interaction of RNase J with ribosomes directs cleavage to a nearby endonuclease target. The full-length RNA is converted by cleavage at this site to an RNA that is 24 nt shorter and that has a 5′ monophosphate end. Continued ribosome binding at the 5′ end of +24 lac-rpsO RNA slows its decay somewhat, conferring a longer-than-average 10-min half-life (Fig. (Fig.3D).3D). We do not know yet which RNase(s) is required for decay initiation of the body of lac-rpsO RNA.
lac-rpsO RNA decay intermediates could be detected in the pnpA strain only when the RNA was not translated, i.e., lac-rpsO(SDX) RNA (Fig. (Fig.3E)3E) and lac-rpsO(stop) RNA (not shown). One explanation for this is that the potential RNA secondary structure may be “ironed out” by ribosomes in the case of lac-rpsO RNA that is translated. Without translation, as is the case for lac-rpsO(SDX) RNA, secondary structures form and decay intermediates arise from downstream endonuclease cleavage, followed by 3′-to-5′ exonuclease decay up to the base of stem-loop structures. Such secondary structures are not a barrier to PNPase in the wild-type strain. An alternative explanation is that initiation of decay occurs by different mechanisms depending on translational status. For translated lac-rpsO RNA, decay occurs exonucleolytically from the 5′ end, and no decay intermediates are detected. For untranslated lac-rpsO RNA, decay occurs by endonuclease cleavage, followed by 3′ exonucleolytic decay, which is blocked by the secondary structure in the strain lacking PNPase.
Our previous observation of abundant decay intermediates in the pnpA strain from several monocistronic mRNAs (30) may also have been coming only from untranslated transcripts. This would resolve an apparent difficulty that we raised earlier. Abundant decay intermediates with intact 5′ ends and 3′ ends that map upstream of the stop codon would trap translating ribosomes that translated to the end of the fragment. The activity of the tmRNA system, which releases ribosomes sitting at the 3′ end of a broken mRNA, was found to be only threefold higher in the pnpA strain than in the wild type (30). This level of activity may not be high enough to alleviate the problem of ribosome pool depletion in the pnpA strain. If this is the case, we wondered why the pnpA strain was relatively healthy. The current results provide an answer. If decay intermediates arise mostly from untranslated mRNAs, then the abundant decay intermediates seen in the pnpA strain are RNA fragments that, in fact, do not have ribosomes trapped at their 3′ end.
From the results shown in Fig. Fig.6,6, it appears that the concentration of the larger decay products (>200 nt long) reaches a steady-state level while the smallest decay intermediate (102 nt) continues to accumulate. This is consistent with cleavage in the 3′-terminal region occurring early on, followed by additional endonuclease cleavages at sites going in the upstream direction and/or 3′-to-5′ exonuclease degradation. Studies of the directionality of RNase E cleavage for E. coli suggested as well that there is a 3′-to-5′ directionality in the search for endonuclease recognition sites (5, 20).
The results with lac-rpsO containing 5′ additions (Fig. (Fig.5)5) were indicative of a strict 5′-end dependence for the initiation of decay. Surprisingly, however, the stability conferred by the 5′-terminal secondary structure was also dependent on the presence of a functional downstream SD sequence. While the lac-rpsO(5′-SS) RNA was completely stable over the course of a 20-min experiment (Fig. (Fig.5C),5C), mutating the SD sequence of the lac-rpsO(5′-SS) construct to lac-rpsO(5′-SS,SDX) resulted in an RNA with a half-life of about 2 min, which was threefold less stable than even that of lac-rpsO mRNA itself (Fig. (Fig.5D).5D). This cannot be attributed to a requirement for ribosome flow to protect the body of the message, since a change of codon 3 to a stop codon in the construct with the strong secondary structure at the 5′ end [lac-rpsO(5′-SS,stop)] had no effect on stability of the full-length RNA (Fig. (Fig.5E).5E). The instability of lac-rpsO(5′-SS,SDX) RNA shows that mRNA stabilization by a 5′-terminal structure cannot be explained simply by an inhibition of RNase J binding at the 5′ end. Similarly, we reported earlier that the stabilizing effect of a 5′-terminal secondary structure on ΔermC mRNA was abolished when the downstream start codon was mutated (39).
In the cases of lac-rpsO(stop) and lac-rpsO(5′-US,stop) RNA, the presence of the stop codon led to extreme stabilization not of the full-length lac-rpsO RNA but of the +24lac-rpsO RNA (Fig. (Fig.5E).5E). We propose that the relative stability of the +24lac-rpsO RNA (Fig. (Fig.3D)3D) could be due to competition between ribosome binding and further attack at the 5′ end of the processed RNA fragment. In the case of the construct with an early stop codon, we speculate that the stop codon at this position results in a longer dwell time of the ribosome after translation initiation, which causes even more of an inhibition of the 5′-end access required for turnover of the downstream fragment. More work is needed to understand the relationship between ternary-complex formation at a ribosome binding site and 5′-end-dependent initiation of decay.
This work was supported by Public Health Service grant GM-48804 from the National Institutes of Health.
We thank Gintaras Deikus for help with the analysis of processed RNA half-life and Ciaran Condon for helpful comments on an early draft of the manuscript.
Published ahead of print on 24 July 2009.