In Escherichia coli, 5′-terminal stem–loops form major impediments to mRNA decay, yet conditions that determine their effectiveness or the use of alternative decay pathway(s) are unclear. A synthetic 5′-terminal hairpin stabilizes the rpsT mRNA sixfold. This stabilization is dependent on efficient translational initiation and ribosome transit through at least two-thirds of the coding sequence past a major RNase E cleavage site in the rpsT mRNA. Insertion of a 12–15 residue ‘ectopic’ RNase E cleavage site from either the rne leader or 9S pre-rRNA into the 5′-non-coding region of the rpsT mRNA significantly reduces the stabilizing effect of the terminal stem–loop, dependent on RNase E. A similar insertion into the rpsT coding sequence is partially destabilizing. These findings demonstrate that RNase E can bypass an interaction with the 5′-terminus, and exploit an alternative ‘internal entry’ pathway. We propose a model for degradation of the rpsT mRNA, which explains the hierarchy of protection afforded by different 5′-termini, the use of internal entry for bypass of barriers to decay, ‘ectopic sites’ and the role of translating ribosomes.
The functional association of mRNA with ribosomes necessarily implicates translation in modulating mRNA longevity and its accessibility to RNase E. Although protection of mRNA by ribosomes is not universal, numerous observations have correlated translational efficiency with mRNA stability, presumably through the masking of initiating cleavage sites by bound ribosomes (Yarchuk et al., 1991; Petersen, 1993; Iost and Dreyfus, 1995; Braun et al., 1998). Somewhat surprisingly, translation also appears to be required to stabilize mRNAs which contain intrinsic protective features. For example, in the ompA mRNA, a threshold level of ribosome passage through the protein-coding region is necessary for its stability (Arnold et al., 1998). Furthermore, lacZ mRNA containing an artificial 5′-stem–loop is no longer protected against RNase E-initiated degradation in the absence of translation (Lopez and Dreyfus, 1996; Joyce and Dreyfus, 1998). This raises a conundrum: if initiation of mRNA decay by RNase E occurs via a rate-limiting 5′-end-dependent mechanism, how does translation influence the stability of mRNAs, and particularly those with 5′-terminal stem–loop structures?
The small, monocistronic rpsT mRNA whose properties are well characterized has provided many insights into the mechanisms of mRNA degradation (Coburn and Mackie, 1999; Goodrich and Steege, 1999). In particular, cleavage by RNase E between residues 300 and 301 in the RNA sequence (between codons 55 and 56) can initiate its decay (Mackie, 1991). Moreover, this mRNA demonstrates 5′-end-dependent decay in vivo (Mackie, 2000). In this report, we have placed a stable hairpin structure at the extreme 5′-terminus of the rpsT mRNA and have investigated how this barrier to decay can be overcome by RNase E in vivo. We show that the 5′-protective barrier is ineffective without efficient translational initiation or ribosome translocation past the rate-limiting RNase E site. Most importantly, we demonstrate that short, 12–15 nt insertions based on known RNase E cleavage sites (‘ectopic sites’), can largely or fully overcome the otherwise efficient 5′-protective RNA structure. This provides clear evidence for two modes of initial substrate recognition by RNase E: one involving 5′-end recognition (Mackie, 1998; 2000), and a second which we term ‘bypass’ or ‘internal entry’.
Regulated expression and stability of chimeric rpsT mRNA
To facilitate the analysis of sequence or structural changes on the kinetics of rpsT mRNA decay, recombinant rpsT mRNA was expressed from a modified low-copy-number vector based on pBAD28 which contains the arabinose-inducible promoter, PBAD (Guzman et al., 1995). To avoid the inclusion of vector-encoded sequences in the expressed mRNA, pBAD28 was altered by site-directed mutagenesis to introduce a unique KpnI restriction site at the exact point of PBAD transcriptional initiation, creating pKEB106 (see Experimental procedures). Residues 91–447 of the endogenous rpsT gene, beginning at the site of initiation of the second (P2) of two tandem promoters, and including the native, rho-independent transcriptional terminator at the 3′-end, were cloned into pKEB106 to generate pKEB107, the parental plasmid for all further gene constructions. The rpsT mRNA expressed from PBAD in pKEB107 (Fig. 1A) was undistinguishable in size from the 356 nt chromosomally encoded rpsT P2 mRNA (data not shown; Mackie and Parsons, 1983). Recombinant mRNAs transcribed from the PBAD promoter will be designated according to their template; thus, rpsT(107) mRNA is the rpsT mRNA encoded by pKEB107.To confirm that expression of the cloned rpsT gene in pKEB107 was under catabolite repression by glucose and inducible by arabinose (Schleif, 1996), RNA was isolated from MG1693 (wild type) containing pKEB107 grown in the presence of glucose or arabinose. Plasmid-encoded rpsT(107) mRNA was not detectable in MG1693/pKEB107 grown in glucose; in contrast, it was induced two- to fivefold relative to the endogenous rpsT P2 mRNA after growth of cells in 0.05% arabinose for 60 min (Fig. 2A; compare w and P2 in lanes 1 and 2). The kinetics of decay of the rpsT mRNAs in MG1693/pKEB107 was evaluated by Northern blot analysis of RNA samples isolated after blocking transcriptional initiation with rifampicin. The half-lives of the coincident P2 and rpsT(107) mRNAs were 2 min (Fig. 2D), identical to that of the chromosomally encoded P2 rpsT mRNA in glucose-grown MG1693/pKEB107 or in untransformed MG1693 (Table 1, entries 1 versus 2; Mackie, 1987). In addition, the rpsT P1 mRNA transcribed from the endogenous, upstream promoter decayed with a half-life of 1.5 min, identical to published values (Mackie, 1987). Maintenance of wild-type half-lives suggests that transient expression of plasmid-encoded rpsT(107) mRNA does not alter the kinetics of rpsT mRNA decay during the experiment.
Table 1. . Decay rates of recombinant rpsT mRNA in Escherichia coli.
. Sequence of ‘ectopic sites’: rne→, 5′-UAACCCAUUUUGCCC-3′; rne←, 5′-UAGGGCAAAAUGGGU-3′; 9Sa→, 5′-UACAGAAUUUUGCGA-3′; and 9Sa←, 5′-UAUCGCAAAAUUCUG-3′. The first column refers to sites in the 5′-UTR; the second to sites in the coding region.
. UUG, wild-type translational initiation sequence; AUG, mutation increasing translational initiation; −3, −4, mutation decreasing translation initiation; stop, position of premature stop codon (nucleotide position based on Mackie, 1992).
. Values represent the average of at least four measurements.
. Empty pBAD28 vector.
. Half-life of the chromosomally encoded rpsT P2 mRNA (Mackie, 1987).
We created a stable 5′-terminal secondary structure in the rpsT mRNA (cf. Bouvet and Belasco, 1992) by inserting complementary oligodeoxynucleotides into the unique KpnI site in pKEB107 (see Fig. 1A and Experimental procedures). The resulting chimeric rpsT gene in pKEB110 should encode an mRNA with a 14 bp stem–loop closed by a GNRA tetra–loop at its extreme 5′-terminus (overall ΔG ≤ 30 kCal mol−1; Fig. 1B). Formation of the desired secondary structure was confirmed by structure mapping (see Fig. 4 below, and data not shown). Northern blot analysis of RNA from MG1693/pKEB110 revealed an rpsT mRNA of intermediate size (rpsT(110) mRNA; 388 nt [w]) expressed from the plasmid (Fig. 2B), and, at reduced levels, the two chromosomally encoded rpsT transcripts of 447 nt and 356 nt (P1 and P2, respectively). The half-life of the rpsT(110) mRNA was 12 min (Fig. 2D; Table 1, entry 3), corresponding to a sixfold stabilization relative to rpsT(107) mRNA (Fig. 2D; Table 1, entry 2). This value is consistent with the extent of stabilization obtained with similar heterologous fusions of stem–loop structures to other mRNAs (Emory and Belasco, 1990; Bouvet and Belasco, 1992; Hansen et al., 1994). Unexpectedly, the rpsT(110) mRNA decayed with biphasic kinetics, with a second, more rapid decay phase commencing 8–9 min after addition of rifampicin (see Discussion). Henceforth, the half-lives of the mRNAs related to the rpsT(110) mRNA will be determined from the initial rates (i.e. the first phase) of disappearance of the mRNA.
We determined the effect of several mutations affecting RNA metabolism on the stability of the rpsT(110) mRNA (see Experimental procedures) to confirm that this mRNA retains the same dependencies as the endogenous rpsT mRNAs. In strain SK5665, which is temperature-sensitive for RNase E activity, the endogenous rpsT P1 mRNA from SK5665 decayed with a half-life of 6 min, fourfold slower than the wild-type rate, demonstrating the sensitivity of the endogenous rpsT mRNA to RNase E (Fig. 2C; Mackie, 1991). The rpsT(110) mRNA decayed with a half-life of 13 min (Fig. 2C), a value not significantly different from the rate in MG1693 (Table 1, entry 3). However, unlike the situation in MG1693, the kinetics of decay of rpsT(110) mRNA in SK5665 were clearly monophasic, with no break in the decay curve even after 21 min of exposure to rifampicin. The kinetics of decay of rpsT(110) in isogenic strains with mutations in pnp, rnb, pcnB or rng, encoding PNPase, RNase II, poly(A) polymerase, and RNase G, respectively, were not altered from that observed in MG1693 (data not shown). Therefore, none of these enzymes can play a role in initiating the decay of the rpsT(110) mRNA.
Degradation of 5′-protected rpsT mRNA is sensitive to translational efficiency
As the rpsT(110) mRNA is relatively resistant to decay, we tested whether its stability is still influenced by translational efficiency, a major determinant of the lifetime of the endogenous rpsT mRNAs (Wirth et al., 1982; Parsons et al., 1988). Previously characterized mutations that vary the efficiency of translational initiation in the rpsT mRNA were introduced into pKEB110 (Fig. 1B). Altering the context three and four residues 5′ to the UUG initiation codon (GA → CU; pKEB110 → pKEB143) decreases translational efficiency sixfold, whereas changing the UUG initiation codon to AUG (pKEB110 → pKEB144) increases translational efficiency 2.5-fold (Parsons et al., 1988). The half-life of the rpsT(143) mRNA was 1–2 min, effectively abolishing any protective effect of the 5′-stem–loop structure (Fig. 3A and E; Table 1, entry 12). In contrast, the half-life of the rpsT(144) mRNA was greater than 25 min (Fig. 3B and E; Table 1, entry 13). Interestingly, the rapidly decaying rpsT(143) mRNA displayed uniphasic rather than biphasic decay kinetics, whereas the onset of the rapid second phase of decay of the very stable rpsT(144) mRNA was delayed by 2–3 min in comparison to rpsT(110), presumably reflecting the increase in translational efficiency (see Discussion).
To determine whether translation initiation per se, or ribosome transit through the coding region governs the stability of rpsT(110) and its derivatives, we introduced premature termination codons into pKEB110 (Fig. 1A; see also Experimental procedures). A stop codon placed at residue 285 (codon 52) in pKEB136 occurs 16 nt 5′ to the major RNase E cleavage site at residues 300/301 (between codons 55 and 56). In pKEB147, the stop codon was introduced at residue 337 (codon 69), 36 nt 3′ to the same major cleavage site. Neither nucleotide change was predicted to alter the secondary structure of the rpsT mRNA (Mackie, 1992). The half-life of the rpsT(136) mRNA with the proximal stop codon was 2 min (Fig. 3C and E; Table 1, entry 14), a sixfold decrease relative to the parental rpsT(110) mRNA. However, the half-life of the rpsT(147) mRNA (distal stop codon) was 7 min, a more modest reduction (Fig. 3D and E; Table 1, entry 15). Thus, the stability of stem–loop-protected rpsT mRNA is enhanced by passage of ribosomes beyond the major internal RNase E cleavage site in the rpsT mRNA; merely loading ribosomes does not suffice.
‘Ectopic’ RNase E cleavage sequences destabilize 5′-terminal stem–loop-protected rpsT mRNA
We postulated that a synthetic sequence containing a known or potential RNase E cleavage site might be sufficient to destabilize a stem–loop-protected rpsT mRNA if inserted at a site available for recognition and cleavage. As prototypes for such ‘ectopic sites’, we used 15 nt sequences encompassing one of two well characterized RNase E cleavage sites: a defined cleavage site within the rne leader (Jain and Belasco, 1995; ‘the rne recognition sequence’) or the 9S ribosomal RNA ‘a’ site (Ghora and Apirion, 1978). Each site was inserted as complementary oligodeoxynucleotides into the NdeI site in the 5′-UTR of rpsT in pKEB110 (Fig. 1B), in both native or reverse-complement orientations. We thereby generated two plasmid pairs: pKEB127 (rne–native) and pKEB128 (rne–reverse-complement); and pKEB152 (9Sa–native) and pKEB154 (9Sa–reverse-complement). Structure mapping was used to evaluate the reactivity of residues in the 5′-UTR of rpsT mRNAs containing ‘ectopic sites’ to control for possible structural rearrangements. To eliminate interference from endogenous rpsT mRNA, strain MRA10, deleted for chromosomal rpsT(Rydén-Aulin et al., 1993), was transformed with plasmids pKEB127, 128, 152 or 154, and RNA was isolated from arabinose-induced cultures. The resistance to cleavage of G-residues in the 5′-terminal stem–loop confirmed the formation of the predicted structure in rpsT(127), rpsT(128), rpsT(152), and rpsT(154) mRNAs (Fig. 4A, lanes 6 and 8 and Fig. 4B, lanes 6 and 8 respectively). The data also revealed that all G residues within the ‘ectopic sites’ were accessible to RNase T1 (Fig. 4A and B: box marked ‘insert’ in lanes 5–8). However, G residues in the ribosomal binding site (5′-GGGAG) of rpsT(127) mRNA (Fig. 4A, lane 6, box marked RBS) were less susceptible to T1 cleavage than the same residues in rpsT(128), rpsT(152) and rpsT(154) mRNAs. We conclude that the 5′-UTRs of rpsT(128), rpsT(152) and rpsT(154) mRNAs distal to the 5′-stem–loop are single-stranded, as is the equivalent region in the endogenous rpsT P2 mRNA (Mackie, 1992). In contrast, the 15 nt rne insertion in rpsT(127) mRNA subtly alters the secondary structure of the 5′-UTR to reduce the single-strandedness of residues encompassing its Shine–Dalgarno sequence.
The effect of the presence of ‘ectopic sites’ on the stability of the appropriate mRNA was measured by Northern blotting. The half-life of the rpsT(127) mRNA (rne–native) was 1 min, a 12-fold reduction compared with the parental rpsT(110) mRNA (Fig. 5A and E; Table 1, entry 4). In contrast, rpsT(128) mRNA (rne–reverse-complement) decayed with a half-life of 13 min, essentially identical to that of the parental rpsT(110) mRNA (Fig. 5B and E; Table 1, entries 5 versus 3). In the pair of strains containing plasmids with the 9Sa site, the rpsT(152) mRNA (native orientation) decayed with a half-life of 3 min (Fig. 5C and E; Table 1, entry 6), representing a moderate (fourfold) destabilization of the rpsT(152) relative to its parent, rpsT(110) mRNA. Somewhat surprisingly, the half-life for rpsT(154) mRNA (9Sa–reverse-complement) was 1 min (Fig. 5D and E; Table 1, entry 7). Inspection of the reverse-complement 9Sa sequence (5′-UAUCGCAAAA UUCUG) reveals two AU dinucleotides commonly found in consensus RNase E cleavage sites (Ehretsmann et al., 1992; McDowall et al., 1994). Moreover, RNase E can cleave antisense 9S rRNA in vitro at this sequence with high efficiency (Cormack and Mackie, 1992).
We attempted to detect if the inserted ‘ectopic sites’ are cleaved by RNase E by mapping the anticipated degradative intermediates using primer extension (see Experimental procedures). Analysis of rpsT(152) mRNA (9Sa–native orientation) revealed a 5′-terminus within the 9Sa site corresponding to the exact position of cleavage by RNase E in 9S RNA (AGA↓AUUUUG; Fig. 6, lane 5, arrow A; Ghora and Apirion, 1978). Likewise, examination of rpsT(154) mRNA showed an extension product corresponding to an endonucleolytic cleavage 5′ to an AU dinucleotide in the reverse-complement 9Sa insertion sequence (Fig. 6, lane 6, arrow B). As expected, in addition to some minor products, several previously characterized RNase E cleavage sites within the rpsT 5′-UTR (Mackie, 1991) were also detected in both rpsT(152) and rpsT(154) mRNAs (Fig. 6, lanes 5 and 6; ‘E’ arrows). We also examined rpsT mRNAs containing the rne leader site by similar methods. A minor primer extension product terminating at a position 5′ to an AU dinucleotide within the site in rpsT(127) mRNA was weakly detectable (data not shown). In contrast, no unique products terminating within the rne site in the reverse-complement orientation (rpsT(128) mRNA) could be detected (not shown).
As an additional test of whether ‘ectopic sites’ mediate cleavage by RNase E, we measured the kinetics of decay of rpsT(152) and rpsT(154) mRNA in SK5665 (rne-1). After thermal inactivation of RNase E in SK5665, the half-lives of both the rpsT(152) and rpsT(154) mRNAs were approximately 9 min, representing a three- to sixfold stabilization (data not shown). Taken together, these data demonstrate that insertion of synthetic RNase E recognition sites into the 5′-UTR of the rpsT mRNA can overcome the stability conferred by a 5′-terminal stem–loop. This effect requires RNase E activity and correlates with observable cleavages within the ‘ectopic sites’.
mRNA destabilization by an ‘ectopic site’ in the rpsT open reading frame (ORF)
We tested whether an ‘ectopic site’ can destabilize the rpsT mRNA when placed within its coding sequence rather than in the 5′-UTR. The rne‘ectopic site’ was cloned into pKEB110 in both native (pKEB138) and reverse-complement (pKEB142) orientations, 145 residues downstream of the start codon into a single-stranded region of the rpsT mRNA (Fig. 1A; see Experimental procedures). At this position, it is well removed from all known RNase E cleavage sites, and does not disrupt the rpsT translational reading frame. The half-life of rpsT(138) mRNA containing this site in the native orientation was 3 min, a 3.5-fold reduction relative to rpsT(110) mRNA (Fig. 7A and E; Table 1, entries 8 versus 3). In contrast, the half-life of the rpsT(142) mRNA (reverse-complement orientation) was 10 min, essentially identical to that of the parental mRNA (Fig. 7B and E; Table 1, entries 9 and 3).
We also evaluated whether the action of an ‘ectopic site’ in the coding region was sensitive to translational efficiency. We introduced either the GA → CU or UUG → AUG mutations into pKEB138 to generate pKEB145 and pKEB146 respectively (Fig. 1B). The half-life of the rpsT(145) mRNA containing the former mutations that decrease translation was 1.5 min, half that of its parent (Fig. 7C and E; Table 1, entry 10). Conversely, the rpsT(146) mRNA exhibited a half-life of 15 min, a fivefold stabilization of the mRNA in comparison to the parental rpsT(138) mRNA (Fig. 7D and E; Table 1, entry 11). Thus, when placed within the ORF, an ‘ectopic site’ displays moderate efficiency and orientation-dependence. Moreover, the efficacy of such an insertion is highly sensitive to translational efficiency.
Ectopic RNase E cleavage sites can bypass stabilizing 5′-stem–loops
The 5′-terminal hairpin in the ompA mRNA in several microorganisms is a major determinant of its atypical longevity (Belasco et al., 1986; Melefors and von Gabain, 1988; Bouvet and Belasco, 1992; Emory et al., 1992; Hansen et al., 1994). The protection afforded by this 5′-stem–loop is not absolute, however, as a number of factors, including growth rate (Emory and Belasco, 1990) or a heterologous insertion (Meyer et al., 1996), can modify this mRNA's stability. Our data show that the rpsT mRNA, whose structure and mode of decay are well characterized (Coburn and Mackie, 1999), can be also stabilized by a prosthetic stem–loop at its extreme 5′-terminus. However, we show here that this protective feature can be overcome by highly effective, relatively short, destabilizing elements (‘ectopic sites’). The efficacy of these sites depends on RNase E activity and is sensitive to a number of modifying factors, most notably translation. These results document for the first time the possibility of increasing dramatically the sensitivity of an mRNA to RNase E by introducing small, well defined, destabilizing elements.
Our work defines several key features of such ‘ectopic sites’. First, they can be as short as 12 nt, if not smaller. Second, although we did not undertake an exhaustive survey, each of the three effective destabilizers was based on a known site of RNase E cleavage and thus probably functions as a site of RNase E cleavage, rather than recognition. Although the products of cleavage at ‘ectopic sites’ should be ephemeral given their 5′-mono-phosphorylated termini (Spickler et al., 2001), our data show that the 9Sa sites function as direct targets for RNase E cleavage in vivo (as they do in vitro; Cormack and Mackie, 1992) and do not operate through simple expansion of or structural changes in the target RNA. Moreover, the destabilizing effect of ‘ectopic sites’ depends on an active rne allele. Interestingly, the 9Sa site destabilizes its host mRNA to different extents depending on its orientation, presumably due to differential recognition of primary sequences at the site of cleavage by RNase E (Cormack and Mackie, 1992). In the rpsT(127) mRNA, however, insertion of the rne leader site induced a subtle reorganization of RNA structure that probably reduces the efficiency of translation. In this case, we cannot distinguish whether the ‘ectopic site’ acts directly or indirectly to overcome the protective effect of the 5′-stem–loop (see below). Care is required, therefore, in the design of such elements to avoid creating alternative secondary structures that might preclude endonucleolytic cleavage or affect translation. Third, the efficiency of an ‘ectopic’ cleavage site depends on its position within the targeted mRNA, with positioning in the 5′-UTR being more effective for a given sequence than in the coding region. We were unable to test the effect of ‘ectopic sites’ in the 3′-UTR of the rpsT mRNA, which is highly structured. Nonetheless, precedent suggests that ‘ectopic sites’ ought to function well in such a location as a 287 nt fragment from the 3′-UTR of the cat gene can be inserted into the 3′-UTR of the ompA gene to produce a dominant, orientation-dependent destabilization of the ompA-cat mRNA (Meyer et al., 1996). Finally, perhaps the most critical and complex determinant of an ‘ectopic site’ is the efficiency with which the host mRNA is translated (see below). In particular, efficient translational initiation can virtually eliminate the effectiveness of an internal ‘ectopic site’ (e.g. compare rpsT(138) mRNA with rpsT(146) in Table 1).
The internal entry model
How 5′-terminal hairpin loops protect mRNAs from RNase E or an alternative decay pathway has not been clearly established (Arnold et al., 1998). We believe that 5′-stem–loops block the 5′-end-dependent pathway (Mackie, 1998; 2000) and channel the rpsT mRNA into an inherently less efficient internal entry pathway. In the first pathway, an accessible, triphosphorylated, 5′-terminal residue is thought to facilitate an initial interaction between RNase E and an mRNA substrate (Mackie, 1998; Tock et al., 2000; Jiang et al., 2000; Spickler et al., 2001). This interaction would weakly tether RNase E to the RNA before cleavage. The enzyme–RNA complex would then rearrange by looping to permit endonucleolytic cleavage at a favourable downstream site (Fig. 8A). In stem–loop-protected or circular mRNAs, the 5′-terminus of the mRNA is sequestered, precluding interaction with the 5′-triphosphate. As such RNAs still decay slowly, dependent on RNase E (Bouvet and Belasco, 1992; Hansen et al., 1994; Mackie, 1998; 2000), RNase E must bypass the 5′-end of the protected mRNA to interact directly with its substrate to catalyse the initiating cleavage (Fig. 8B). The efficiency of this ‘internal entry’ process is determined by the intrinsic susceptibility of the rate-limiting cleavage site and by external variables such as translation (see below). ‘Ectopic sites’ clearly favour use of the internal entry pathway. They can be visualized as special cases of resident RNase E cleavage sites that function with high efficiency so that the stimulatory effect of a free 5′-terminus is unnecessary.
The interplay between translation and decay of the rpsT mRNA
The extent to which the rpsT mRNA can be stabilized by the protective 5′-stem–loop clearly depends on the efficiency of translational initiation. Generally similar outcomes have been obtained in other mRNAs. A 5′-hairpin loop is unable to protect the lacZ mRNA from rapid decay in the complete absence of translation (Joyce and Dreyfus, 1998). Likewise, translation is required for the longevity of the ompA mRNA. In this regard, competition between 30S subunits and the growth-rate regulated protein, Hfq, for binding to the 5′-UTR of the ompA mRNA leads to reduced translational initiation and significantly decreased stability (Vytvytska et al., 1998; 2000). Moreover, our data show that passage of ribosomes beyond an efficient site for RNase E cleavage (i.e. residues 300/301 within the rpsT coding region) is also required for the protective effect of the 5′-terminal stem–loop (cf. Nilsson et al., 1987).
The simplest model to explain the protective effect of translating ribosomes in either the 5′-end-dependent or internal entry pathways is steric masking of cleavage sites resulting in competition between translation and RNase E. Ribosomes can protect at least 30 residues of an mRNA (Steitz, 1969). As RNase E is also large, especially when incorporated into the RNA degradosome (Py et al., 1996), ribosomes would readily interfere with cleavage site recognition by RNase E acting in either a looping or internal entry mode (compare panels A and B, respectively, in Fig. 8). Masking can explain the negative correlation between the effectiveness of a ‘ectopic site’ and translational efficiency as ribosomes and RNase E effectively compete with each other. It also rationalizes the differential effect of premature termination codons (Fig. 3C and D). For these reasons, we believe that masking of cleavage sites better explains the correlation between translational initiation and mRNA stability in rpsT mRNAs containing 5′-stem–loops than a model invoking competition between initiation of translation and recognition of the 5′-end of an RNA by RNase E (Rapaport and Mackie, 1994). Further support for steric masking of RNase E sites has been found in the rpsO mRNA (Braun et al., 1998). Increasing the distance between the termination codon and the rate-limiting M2 cleavage site in the 3′-UTR of this mRNA destabilizes it. As the M2 site occurs 10 residues 3′ to the termination codon, the normal delay in the release step would effectively mask the site (Braun et al., 1998). A more complex alternative to steric masking postulates that RNase E initially binds non-specifically to an mRNA either during looping (in a 5′-end-dependent mode) or direct entry. It would then migrate to the first cleavage site. Translating ribosomes would compete with the migration step rather than the actual initiating cleavage. The principal argument against this model is that RNase E would have to migrate through both single and double-stranded RNA without unwinding or cleaving either.
Although the 5′-terminal stem–loop stabilizes rpsT(110) mRNA significantly, moderate expression of this mRNA leads to reduced steady state levels of the two chromosomally encoded rpsT mRNAs (e.g. Fig. 2B). In addition, the rpsT(110) mRNA decays with biphasic decay kinetics that are lost when RNase E is inactivated. Both observations probably reflect autogenous translational repression of rpsT mRNAs exerted by accumulated S20 protein (Wirth et al., 1982), which is further exacerbated by inhibition of accumulation of 16S rRNA by rifampicin. During rifampicin treatment, the ratio of S20 to its ligands increases resulting in severe translational repression. This would decrease translational initiation, resulting in loss of ribosomal masking, and thus would enhance access of RNase E to the repressed rpsT mRNA. This would explain the reduced levels of the chromosomal rpsT P1 and P2 mRNAs in many of our experiments. Such behaviour is similar to that observed for the alpha operon and other ribosomal protein mRNAs (Singer and Nomura, 1985; Cole and Nomura, 1986). In addition, the ratio of RNase E to its substrates will increase as the rifampicin treatment persists. The time of onset of the second rapid phase of decay of the plasmid-borne rpsT mRNA should reflect both factors. In support, we observed that the presence of an AUG codon (rpsT(144) mRNA) that reduces translational repression (Parsons et al., 1988) prolongs the first phase of decay. Similarly, in the absence of active RNase E (in SK5665), the second phase of decay is abolished. This interpretation of how translational control affects the stability of the rpsT mRNA differs from previous results in which gene dosage was much higher and where expression of rpsT mRNAs could not be regulated as tightly as in the present work (Mackie, 1987).
Ectopic sites and the hierarchical model for RNase E action
Escherichia coli K12 strains MG1693 (thyA715, rph-1; Arraiano et al., 1988), and MRA10 (MG1655, rpsT147; Rydén-Aulin et al., 1993) were routinely grown at 37°C in Luria–Bertani (LB) medium supplemented with glucose (0.2%), MgSO4 (1 mM), thymidine (50 µg ml−1), ampicillin (50 µg ml−1) and chloramphenicol (25 µg ml−1), as needed. Induction of transcription from PBAD (see below) was achieved by filtering mid-exponential phase cells through 1.2 µ filters (Millipore), washing with unsupplemented medium, and resuspending in supplemented LB containing 0.05% arabinose in place of glucose, followed by further growth for 60 min. Cultures of strain SK5665 (rne-1, thyA715, rph-1; Arraiano et al., 1988) were grown at 30°C, induced as described above, then shifted to 44°C for 15 min before the addition of rifampicin to 300 µg ml−1.
Restriction endonucleases, T4 DNA ligase, and polynucleotide kinase were obtained from either New England Biolabs (Beverly, MA, USA) or Life Technologies (Gaithersburg, MD, USA). Oligonucleotides were synthesized by either the NAPS Unit, University of British Columbia, or by Life Technologies. Manipulations of DNA followed standard procedures (Sambrook et al., 1989). All recombinant plasmids were verified by restriction analysis and partial DNA sequencing. The vector pBAD28 (Guzman et al., 1995) was obtained from Dr J. Beckwith (Harvard Medical School, Boston, MA), USA. It was linearized with KpnI and re-ligated after treatment with mung bean nuclease to eliminate the unique KpnI restriction site from the polylinker, generating pKEB105. A new unique KpnI site was introduced into the position of transcriptional initiation of PBAD in pKEB105 by site-directed mutagenesis (Stratagene QuikChange™, La Jolla, CA, USA) using the complementary oligonucleotides 5′-GCCCAAAAAAACGG GTACCGAGAAACAGTAGAGAG-3′ and 5′-CTCTCTACTGTT TCTCGGTACCCGTTTTTTTGGGC-3′ (mutant sequences are underlined and the KpnI site italicized), creating pKEB106. Residues 92–447 of the rpsT gene from pGM79 (Mackie, 1991) were amplified using Taq DNA polymerase and the primers 5′-GGGGTACCTTTGAATTGTCCATATGG AACACATTTGGG-3′ (the KpnI and NdeI sites are italicized) and 5′-GCTCTAGAGCATCACAAAAGCAGCAGGC-3′ (XbaI site italicized). The product was ligated into KpnI- and XbaI-digested pKEB106 to generate pKEB107. Complementary oligonucleotides 5′-CATCGCCACCGGGAGACCGGTGGCG ATGGTAC-3′ and 5′-CATCGCCACCGGTCTCCCGGTGGCG ATGGTAC-3′ were 5′-phosphorylated, annealed and ligated into the KpnI site of pKEB107 to generate pKEB110 encoding an rpsT mRNA with a stem–loop at its extreme 5′-terminus (see Fig. 1B).
Complementary oligonucleotide pairs 5′-TAACCCATTT TGCCC and 5′-TAGGGCAAAATGGGT-3′ or 5′-TACAGAA TTTTGCGA-3′ and 5′-TATCGCAAAATTCTG-3′ encoding the RNase E cleavage site at residues 46–56 in the rne gene (Jain and Belasco, 1995) or nucleotides encompassing the 9S rRNA ‘a’ cleavage site (Ghora and Apirion, 1978), respectively, were ligated into the unique NdeI site of pKEB110. Plasmids harbouring the sequence in the native and reverse-complement orientations (pKEB127, pKEB128 and pKEB152, pKEB154, respectively; see Table 1) were selected. The rne cleavage site (see above) was inserted into the rpsT coding region of pKEB110 through polymerase chain reaction (PCR) amplification of two partial rpsT sequence fragments. Sequences 5′ to the position of insertion were amplified from pKEB110 DNA using oligonucleotides 5′-GATTAGCGGATCCTACCTGACGC-3′ (BamHI site italicized) and 5′-ACTCCGCTCGAGAGCAGCTTTGTCGC CAGCTTCG-3′ (XhoI site italicized). The 3′-fragment encompassing the insertion sequence (the XhoI site is italicized and the rne insertion sequence is underlined) was amplified using as forward primers either 5′- TGAGGCCTCGAGACCATTT TGCCCAAAGCATTTAACGAAATGCAACCG-3′ for the native orientation, or 5′-TGAGGCCTCGAGGGGCAAAATGGTA AAGCATTTAACGAAATGCAACCG-3′ for the reverse-complement orientation, coupled with 5′-GGTCGACTCTA GAGCATCAC-3′ (XbaI site italicized) as reverse primer. The resultant DNA fragments were digested and ligated into pBAD28 digested with BamHI and XbaI. Mutations in the translational initiation region of the rpsT gene were introduced by site-directed mutagenesis and complementary oligonucleotides 5′- GGGAGTTGGACCATGGCTAATATCAAATC-3′ and 5′-GATTTGATATTAGCCATGGTCCAACTCCC-3′ (altered residue(s) underlined), for increased translational efficiency, or 5′-CACATTTGGGAGTTGCTCCTTGGCTAATATC-3′ and 5′-GATATTAGCCAAGGAGCAACTCCCAAATGTG-3′, for decreased translational efficiency (Parsons et al., 1988). A stop codon was introduced into the rpsT gene of pKEB110 at codon 69 to make pKEB147 by site-directed mutagenesis using oligonucleotides 5′-CCTCTGATCCACTAAAACAAAG CTGCACGTC-3′ and 5′-GACGTGCAGCTTTGTTTTAGTG GATCAGACC-3′ (mutant residue underlined). Codon 51 was mutated to a stop codon by PCR amplification of two rpsT partial sequences to create pKEB136 (details provided upon request).
RNA isolation and Northern blot analysis
Early log phase cultures grown as described above were poisoned by the addition of rifampicin to 300 µg ml−1. Aliquots (25 ml) were removed at various times and added to 12.5 ml ice-cold 12.5 mM NaN3 containing 200 µg ml−1 of chloramphenicol. Bacteria were collected by centrifugation and RNA isolated as described previously (Method I in Mackie, 1989). Purified RNA was quantified spectrophotometrically. For RNA blots, 5 µg was separated on 5.5% polyacrylamide gels containing 8 M urea. RNAs were transferred electrophoretically to Hybond-NX (Amersham Pharmacia Biotech), fixed by UV crosslinking, and probed to detect the rpsT mRNAs with a radiolabelled RNA complementary to rpsT residues 91–441. Hybridization was performed at 55°C in 50% formamide, 5x SSC, 5x Denhardt's solution (Sambrook et al., 1989), 40 mM Na2HPO4 (pH 7.0), 0.25 mg ml−1 of yeast RNA, 0.06 mg ml−1 of salmon sperm DNA, and 0.1% SDS. Membranes were washed (4 × 15 min) with 2× SSC and 0.1% SDS at 55°C. Signals were visualized and quantified using a Molecular Dynamics PhosphorImager. RNA loading was periodically verified by reprobing for 5S rRNA. Half-lives were deduced from the initial slope of a semilogarithmic plot of the fraction of mRNA remaining versus time after rifampicin addition.
Primer extension analysis
First, 10 µg of RNA isolated from MRA10 transformed with either pKEB127, 128, 152 or 154 (see Table 1) was concentrated by precipitation with ethanol and 100 µg ml−1 of glycogen. RNA was dissolved in 10 µl of 50 mM Tris-HCl (pH 8.3), 75 mM KCl, 10 mM DTT and 2 pmol primer (PE-mer; 5′-[32P]-CGGCTTGCGTTGTGCTTACGAG-3′) complementary to rpsT residues 182–203 (see Fig. 1A; Mackie, 1992) and incubated at 65°C for 5 min. Synthesis of cDNAs was initiated by addition of (final concentrations) 50 mM Tris-HCl (pH 8.3), 75 mM KCl, 3 mM MgCl2, 10 mM DTT, 5 mM dNTPs and 100 units of M-MLV Reverse Transcriptase in a final volume of 20 µl followed by incubation at 37°C for 45 min. Extension products were resolved alongside sequence ladders, using the same primer, on 6% sequencing gels containing 8 M urea and visualized by autoradiography.
We thank Drs S.R. Kushner, University of Georgia, Athens, GA, and L. Isaksson, Stockholm University, Stockholm, Sweden, for providing strains. Drs M. Dreyfus, P. Régnier and members of the Mackie laboratory offered valuable comments. This work was funded by an operating grant to G.A.M. from CIHR. K.E.B. received a fellowship from NSERC and the UBC Killam Endowment.