In Escherichia coli, RNA degradation is mediated by endonucleolytic processes, frequently mediated by RNase E, and also by a poly(A)-dependent mechanism. The dominant pathway of decay of the rpsO transcripts is initiated by an RNase E cleavage occurring at a preferential site named M2. We demonstrate that mutations which prevent this cleavage slow down degradation by RNase E. All these mutations reduce the single-stranded character of nucleotides surrounding the cleavage site. Moreover, we identify two other cleavage sites which probably account for the slow RNase E-mediated degradation of the mutated mRNAs. Failure to stabilize the rpsO transcript by appending a 5′ hairpin indicates that RNase E is not recruited by the 5′ end of mRNA. The fact that nucleotide substitutions which prevent cleavage at M2 facilitate the poly(A)-dependent degradation of the rpsO transcripts suggest an interplay between the two mechanisms of decay. In the discussion, we speculate that a structural feature located in the vicinity of M2 could be an internal degradosome entry site promoting both RNase E cleavages and poly(A)-dependent degradation of the rpsO mRNA. We also discuss the role of poly(A)-dependent decay in mRNA metabolism.
The stability of mRNA is one of the parameters which determine the efficiency of gene expression. Mechanisms of mRNA decay in bacteria have been the subject of recent reviews (Coburn and Mackie, 1999; Grunberg-Manago, 1999; Régnier and Arraiano, 2000; Steege, 2000; Arraiano and Maquat, 2003). It is widely admitted that mRNA turnover in bacteria is mediated by a combination of endo- and exoribonucleases whose activities are modulated by structural features of the mRNA. Five of the 15 ribonucleases identified in Escherichia coli have been clearly implicated in mRNA decay (Nicholson, 1997). Current models postulate that RNase E, which is believed to be a key enzyme in this process, catalyses rate-limiting endonucleolytic cleavages and generates RNA fragments that are then degraded by the 3′ to 5′ exoribonucleases, polynucleotide phosphorylase (PNPase) and RNase II. The endonucleolytic activity of RNase E and the exonucleolytic activity of PNPase are thought to be co-ordinated by their association in a multienzyme complex known as the degradosome (Miczak et al., 1996; Py et al., 1996).
Secondary structures such as transcription terminators and REP sequences protect the 3′ end of mRNAs from degradation by PNPase and RNase II (Régnier and Hajnsdorf, 1991; Higgins et al., 1993). It has recently been demonstrated that poly(A) tails, added post-transcriptionally by poly(A) polymerase I (PAP I) at the 3′ end of RNA, promote degradation of such structured molecules (Xu et al., 1993; Hajnsdorf et al., 1995; O’Hara et al., 1995; Coburn and Mackie, 1998; Blum et al., 1999; Dreyfus and Régnier, 2002). The current model postulates that poly(A) tails and stretches of 3′-terminal unpaired nucleotides provide binding sites where PNPase, free or associated with the degradosome, can initiate the processive exonucleolytic degradation of RNAs harbouring 3′ secondary structures. The degradosome contains an RNA-helicase, RhlB, which facilitates the exonucleolytic degradation of hairpin structures by PNPase present in degradosome (Vanzo et al., 1998). In contrast, the poly(A)-dependent degradation of mRNA is impeded by another exoribonuclease, RNase II, which removes the poly(A) extensions but fails to degrade stably folded RNAs (Coburn and Mackie, 1998; Marujo et al., 2000). Finally, some RNAs also seem to be the target of RNase G, the RNase E homologue (Umitsuki et al., 2001) and other RNase E-independent mechanisms that are regulated by 5′ stem loops (Arnold et al., 1998).
The rpsO mRNA is a model monocistronic mRNA whose degradation has been shown to be carried out by either an RNase E-dependent endonucleolytic process or by a mechanism which requires polyadenylation (Hajnsdorf et al., 1996). Under normal conditions, the stability of the rpsO transcript is controlled by RNase E, which catalyses the first endonucleolytic step of decay at a specific site referred to as M2(Fig. 1) (Régnier and Hajnsdorf, 1991). However, the extent of stabilization that occurs when RNase E is inactivated or becomes less abundant in the cell is less than expected if rpsO mRNA decay depended only on RNase E ( Hajnsdorf et al., 1994a; Jain et al., 2002; Ow et al., 2002). It has been shown that, under these conditions, an alternative pathway requiring polyadenylation is responsible for the degradation of this transcript (Hajnsdorf et al., 1995). Moreover, the fact that inactivation of poly(A) polymerase, which accounts for the synthesis of most, if not all, the poly(A) tails in E. coli, does not stabilize the rpsO mRNA when RNase E is active suggests that the poly(A)-dependent mechanism of decay is effective primarily on molecules that are not attacked by RNase E.
In agreement with this hypothesis, we demonstrate here that rpsO transcripts lacking the M2 site, where RNase E normally initiates degradation, are better targets for the poly(A)-dependent machinery of decay. As expected, these mutant transcripts that are not processed at M2 are less efficiently degraded by the RNase E-dependent pathway. Moreover, we show that this latter pathway is 5′ end independent and probably involves an internal entry site located in the vicinity of the preferential M2 cleavage site, just upstream of the hairpin of the Rho-independent transcription terminator. We finally propose to explain these data that the degradosome bound at this site controls both cleavage at RNase E sites and the poly(A)-dependent degradation of the mRNA.
Construction of rpsO genes lacking the M2 RNase E site
As a continuation of our previous studies on mRNA decay, we constructed a plasmid-borne rpsO gene lacking the M2 RNase E site to investigate further the role of this preferential cleavage site in the decay of the rpsO transcript. We modified the M2 site of the rpsO gene of plasmid pΔS15(M2+), deleted of 111 nucleotides just downstream of the translation start codon (Fig. 1) (Braun et al., 1998). All mRNAs containing this deletion will be referred to as rpsOΔ mRNAs throughout the manuscript. Moreover, all these transcripts lack the M3 RNase E site located at the beginning of the coding sequence (Hajnsdorf and Régnier, 1999) and harbour a premature UAA stop codon, 67 nucleotides upstream of the M2 cleavage site, which prevents translating ribosomes from interfering with RNase E cleavage (Braun et al., 1998).
In two derivatives of pΔS15(M2+), the A+U-rich M2 cleavage site was replaced by G+C-rich sequences not expected to be cleaved by RNase E (Ehretsmann et al., 1992; Nilsson et al., 1996). pΔS15(M2–1) and pΔS15(M2–2) contain the sequences GGGGCC and GCGGCG instead of the GAG↓UUU cleavage site respectively (the arrow shows the location of the RNase E cleavage) (Fig. 1B). A third plasmid named pΔS15(M2Δ) was obtained by deleting the AGUUU sequence. The replication of these plasmids is regulated independently of RNA polyadenylation.
Transcripts carrying mutations are not processed at M2
To make sure that these mutations impaired RNase E processing, we first verified that the P1-M2 fragment resulting from cleavage of M2-containing rpsOΔ mRNAs (Régnier and Hajnsdorf, 1991; Hajnsdorf et al., 1994a; Braun et al., 1996), was no longer produced in bacteria transformed with the mutagenized plasmids. For this purpose, we introduced the original pΔS15(M2+) plasmid, harbouring the M2 site, and the mutagenized derivatives, pΔS15(M2–1), pΔS15(M2–2) and pΔS15(M2Δ) into strain CMA303, which carries mutations in both 3′ to 5′ exonucleases, PNPase (pnp7) and RNase II (rnb500ts) and in PAP I (ΔpcnB), referred to hereafter as PNP–RII–PAP–. This strain was chosen because it facilitates detection of mRNA fragments resulting from endonucleolytic cleavages on Northern blots (Hajnsdorf and Régnier, 1999). As expected, inactivation of thermosensitive RNase II of the PNP–RII–PAP– strain transformed with pΔS15(M2+) at 44°C led to the accumulation of a 279 nucleotide P1-M2 fragment resulting from processing of the rpsOΔ transcripts (Braun et al., 1998) (Fig. 2). Failure to detect this cleavage product in bacteria transformed with either pΔS15(M2–1), pΔS15(M2–2) or pΔS15(M2Δ) indicated that disruption of the M2 site reduces the efficiency of the RNase E cleavage (Fig. 2). The inhibition of processing of the pΔS15(M2–2) mRNA was confirmed by primer extension experiments showing that the 5′ extremity of the downstream product of cleavage at M2 is far less abundant in cells transformed with pΔS15(M2–2) than in cells transformed with pΔS15(M2+) (data not shown). Moreover, the accumulation of mRNA fragments containing the mutated M2 sites in cells transformed with pΔS15(M2–1), pΔS15(M2–2) and pΔS15(M2Δ) (see below) and in vitro processing experiments (M. Folichon, unpubl. data) confirmed that the mutated M2 sites are either not cleaved or are very inefficiently cleaved by RNase E (see below). Chemical probing of the secondary structure of the rpsOΔ transcripts show that mutated M2 sites become more double stranded and thus less likely to be bound and/or cleaved by RNase E (Coburn and Mackie, 1999) (Supplementary material, Figs S1 and S2).
rpsOΔ mRNAs lacking the M2 site are still degraded by an RNase E-dependent mechanism
We then examined whether removal of the M2 site where RNase E initiates decay stabilizes the rpsO transcript (Régnier and Hajnsdorf, 1991; Hajnsdorf et al., 1994a). Intriguingly, relative to the parental rpsOΔ mRNA, the half-lives of the M2–1 and M2–2 mRNAs were only slightly increased (1.7-fold) or unchanged respectively (Table 1, column 1). Because decay of the rpsO mRNA depends on two pathways (Hajnsdorf et al., 1995), half-lives of transcripts were also measured in a PAP I deficient strain, referred to as PAP–RE+, in order to determine the sensitivity of the rpsOΔ mRNAs to RNase E-mediated decay independently of any contribution from poly(A)-dependent degradation (Table 1, column 2). Under these conditions, the M2-less transcripts were 3-, 3.5- and 6.5-fold more stable than the pΔS15(M2+) mRNAs containing M2. This demonstrates that preventing cleavage at M2 slows down degradation of the rpsOΔ transcript mediated by RNase E. However, the fact that half-lives of the pΔS15(M2–1), pΔS15(M2–2) and pΔS15(M2Δ) mRNAs lacking the M2 site increased at least fivefold upon RNase E inactivation implies that they are still degraded by an RNase E-dependent pathway. Indeed, half-lives of these transcripts which range from 3 to 6 min in the PAP–RE+ strain are longer than 20 min in PAP–RE– cells (Table 1, compare columns 2 and 4). Consistently, in the presence of PAP I, the pΔS15(M2–1) and pΔS15(M2–2) rpsOΔ mRNAs were also more stable (about 1.8- and 2.3-times respectively) in the PAP+RE– strain lacking RNase E than in isogenic PAP+RE+ cells containing RNase E (Table 1, columns 3 and 1). All these data lead to the conclusion that rpsOΔ transcripts which are not processed at the M2 site can still be degraded, albeit at a slower rate, by RNase E. On the other hand, the slow decay rate of the pΔS15(M2Δ) mRNAs in PAP+RE+ and PAP–RE+ strains, compared with that of other M2-containing and M2-less rpsOΔmRNAs (Table 1), indicates that the deletion of five nucleotides surrounding M2 disrupts structural feature(s) involved in the degradation of the molecule that are not affected by nucleotide substitutions (see below).
Table 1. Sensitivity of mRNAs harbouring or lacking the M2 site to RNase E and poly(A)-dependent decay.
. The strains used are IBPC694 (PAP+RE+), MG1693 (PAP+RE+), IBPC681 (PAP+RE–), IBPC690 (PAP–RE+), SK7988 (PAP–RE+) and CMA304 (PAP–RE–).
. When only one or two independent measurements were performed, we assumed that the relative error in the half-lives of mRNA of plasmid origin was the same as the percentage error calculated for the transcript of chromosome origin from all the experiments (at least eight) performed with the same cells transformed with plasmids.
Half-lives were measured in cells grown at 30°C and shifted to 44°C to inactivate RNase E with simultaneous addition of 500 µg ml−1 rifampicin. Aliquots were withdrawn at t0 (time of rifampicin addition) and 1, 2, 4, 8, 12, 16 and 20 min later. The relative amounts of the monocistronic rpsO transcripts of plasmid and chromosome origin remaining in the cell were estimated by Northern blot and plotted as a function of time. Half-lives were determined by linear regression analysis and errors were calculated from standard deviations estimated from at least three independent measurements.
Inactivation of the M2 site changes the contribution of the poly(A)-dependent pathway to the degradation of the rpsOΔ transcripts
Decay rates in different strains deficient for the RNase E-dependent decay (PAP+RE–), poly(A)-dependent decay (PAP–RE+) or both (PAP–RE–) also show that rpsOΔ transcripts harbouring the M2–1 and M2–2 nucleotide substitutions which reduce the efficiency of degradation by RNase E are better substrates for the poly(A)-dependent machinery. Indeed, the fact that transcripts originating from pΔS15(M2+) and from the chromosome were degraded at the same rate in the presence (half-lives = 1.1 and 1.4 min respectively) and in the absence (half-lives = 0.9 and 1.1 min respectively) of PAP I (Table 1, columns 1 and 2) while they are dramatically stabilized (7.3- and 5.6-fold respectively) upon RNase E inactivation (Table 1, columns 1 and 3) proves that mRNAs containing M2 are primarily degraded by the RNase E-dependent pathway. In contrast, the pΔS15(M2–1) and pΔS15(M2–2) transcripts that are slightly stabilized upon inactivation of RNase E (1.8- and 2.3-fold respectively) (Table 1, columns 1 and 3) and PAP I (1.5- and 2.6-fold respectively) (Table 1, columns 1 and 2) are presumably equally sensitive to both pathways. Moreover, the half-lives of the pΔS15(M2+), pΔS15(M2–1), pΔS15(M2–2) mRNAs in a PAP+RE– strain transferred to the non-permissive temperature for RNase E (8.0, 3.5 and 2.8 min respectively) suggest that the M2–1 and M2–2 mutations promote degradation of the rpsOΔ mRNA by the poly(A)-dependent pathway (Table 1, column 3). Consistent with this hypothesis, compensation of the lower rate of degradation by RNase E by an improved efficacy of the poly(A)-dependent pathway presumably explains why the M2-less transcripts containing the M2–1 and M2–2 mutations were degraded at nearly the same rate as the M2-containing mRNAs in the PAP+RE+ strain (Table 1, column 1).
Contrasting with the nucleotide substitutions of pΔS15(M2–1) and pΔS15(M2–2), which changed the contributions of the RNase E and poly(A)-dependent pathways to the process of decay without affecting significantly the overall rate of degradation, the deletion of five nucleotides encompassing the M2 site increases the global stability of the rpsOΔ transcripts. In this case, the stability of the pΔS15(M2Δ) mRNA compared with that of the pΔS15(M2+) mRNA in a PAP–RE+ strain (5.9 versus 0.9 min respectively) (see Table 1) shows that the deletion of the M2 site dramatically reduces the efficiency of degradation of the rpsOΔ transcript by the RNase E-dependent pathway. In contrast, half-lives of these two transcripts in a PAP+RE– strain (5.8 versus 8.0 min respectively) (see Table 1) indicates that this deletion hardly affects the rate of degradation of the rpsOΔ transcript by the poly(A)-dependent pathway. Consistent with these observations, failure to compensate for the slower rate of RNase E-mediated decay by poly(A)-dependent degradation probably explains why the pΔS15(M2Δ) transcript lacking the M2 RNase E site is more stable than the pΔS15(M2+) mRNA, in the PAP+RE+ strain (half-lives shown in Table 1 are 1.1 and 6.0 min respectively) (see Table 1).
rpsOΔ mRNAs mutated at M2 are processed at other sites
Because the M2-less mRNAs are still targets of RNase E-mediated decay, we performed Northern blots and primer extensions to identify processing intermediates generated by RNase E in transcripts whose primary processing site was inactivated. As expected, new mRNA fragments presumably resulting from endonucleolytic events are seen in cells transformed with pΔS15(M2–1), pΔS15(M2–2) and pΔS15(M2Δ) that do not accumulate in cells containing pΔS15(M2+) (Fig. 3A). Characterization of the two RNA fragments referred to as M5-t1 and M6-t1 was performed with RNAs extracted from PAP I– bacteria shifted to 44°C because their intracellular concentration is higher under these conditions (Fig. 3A). Detection of the 135–140 nucleotide long M5-t1 fragment originating from pΔS15(M2–2) on a Northern blot with a short oligoprobe complementary to the 3′-terminal region of rpsO (Fig. 3B), suggests that it extends from a processing site named M5 to the 3′ end of the primary transcript. The 5′ ends of the M5-t1 and M6-t1 fragments were mapped by primer extension performed with RTprimer1, which anneals close to the 3′ end of the rpsO transcript (Fig. 1). In agreement with the conclusion above, a 5′ end presumably resulting from the cleavage at M5 was identified 137 nucleotides upstream of the 3′ end of the primary transcript, in cells transformed with pΔS15(M2–1), pΔS15(M2–2) or pΔS15(M2Δ) but not in cells containing rpsO mRNAs whose M2 site is in the wild-type form (Fig. 3C). This experiment locates the M5 site in the sequence UGG↓ UUU where the cleavage site is shown by an arrow (Fig. 1). Moreover, another strong reverse transcriptase stop mapping to the sequence AAA↓GAU, 171 nucleotides upstream of the 3′ end of the transcript in bacteria transformed with pΔS15(M2Δ) (Fig. 3C), may correspond to the 5′ end of the 170 nucleotide long M6-t1 fragment detected in the same cells (Fig. 3A). Again, this fragment is presumably the downstream product of cleavage of the pΔS15(M2Δ) mRNA at a site named M6 (Fig. 1). Consistent with this hypothesis, this 5′ end was hardly or not detected in cells transformed with pΔS15(M2–1), pΔS15(M2–2) or pΔS15(M2+) (Fig. 3C) which do not contain detectable amounts of M6-t1 (Fig. 3A). The relative abundance of 5′ ends mapping at M5 and M6 in bacteria transformed with pΔS15(M2Δ) (Fig. 3C) suggests that M5-t1 is not detected in these cells (Fig. 3A) because transcripts originating from the plasmid are efficiently cleaved at M6 (Fig. 3C). These data demonstrate that rpsOΔ mRNAs lacking M2 are cleaved at secondary sites named M5 and M6 located 137 and 171 nucleotides upstream of the 3′ end of the primary transcript, respectively, and suggest that the M5-t1 and M6-t1 downstream products resulting from these cleavages accumulate because their mutated M2 sites located at 35 nucleotides from the 3′ end of the fragment are not cleaved by RNase E. Moreover, the failure to detect these RNA fragments in PAP I+ strains (Fig. 3A) confirm that they are protected from exoribonucleases by the hairpin of the Rho-independent terminator (Hajnsdorf and Régnier, 1999).
It is also worth pointing out that several RNA fragments detected in PAP– cells transformed with pΔS15(M2–1), pΔS15(M2–2) or pΔS15(M2Δ) that are not discussed above (Fig. 3A) are probably also stable downstream products resulting from endonucleolytic cleavage of transcripts lacking the primary M2 site.
RNase E cleaves the rpsOΔ mRNAs at M5 and M6
Previous experiments showing that the wild-type rpsO transcripts are cleaved in the vicinity of M5 and M6 by RNase E in vitro (Braun et al., 1996) suggest that the cleavages of rpsOΔ mRNAs described above are catalysed by this ribonuclease. Accordingly, the in vitro processing experiment of Fig. 4, where RNA 5′ ends were detected by primer extension, demonstrates that all in vitro synthesized rpsOΔ mRNAs either harbouring or lacking the M2 site are cleaved at M5 and M6 sites by RNase E of a degradosome preparation. As expected, the transcripts were not cleaved at these sites when thermosensitive RNase E of degradosomes prepared from the mutant strain rne3071 was inactivated at 44°C (data not shown). Furthermore, reduction of the amounts of M5-t1 and M6-t1 fragments in cells whose thermosensitive RNase E has been inactivated confirmed that cleavages at M5 and M6 are also catalysed by RNase E in vivo (data not shown). This conclusion is reinforced by the observation that the efficacy of these cleavages in vitro parallels the amounts of the downstream cleavage products detected in vivo. Indeed, transcripts harbouring the M2Δ deletion that were cleaved at M6in vivo (Fig. 3C) were much more efficiently cleaved at M6 by purified degradosome than rpsOΔ mRNAs harbouring the M2–1 mutation (Fig. 4) that are primarily cleaved at M5in vivo (Fig. 3C). Chemical probing of RNA structures showed that these differences of processing efficiencies at M5 and M6 do not result from modifications of secondary structure in the vicinity of these sites (Supplementary material, Fig. S1).
Finally, it is also worth pointing out that M2-less mRNAs were cleaved by RNase E in vitro in the untranslated leader of the rpsO transcript, at a site located at position 59, in the sequence GA↓AUU. This cleavage is referred to as M7 in Fig. 4. Products resulting from this cleavage have not yet been characterized in vivo.
The 5′ end of the rpsO transcript is not the major RNase E entry site
The crucial role of M2 in the degradation of the rpsO transcript led us to postulate that direct binding of RNase E or degradosome at this site could initiate decay independently from an interaction with the 5′ end of the mRNA (Bouvet and Belasco, 1992; Mackie, 1998; Baker and Mackie, 2003). In order to test this hypothesis, we fused to the 5′ end of the rpsOΔ mRNA, a stable hairpin (hp*) which is expected to stabilize the transcript if recruitment of RNase E or the degradosome took place at the 5′ end of the molecule (Fig. 1). The data shown in Fig. 5 and Table 2 show that the pΔS15(M2+)hp* mRNA was not significantly stabilized by the hairpin. Furthermore, addition of the 5′-hp* hairpin did not affect either the stability of the pΔS15(M2–1), pΔS15(M2–2) or pΔS15(M2Δ) mRNAs that are cleaved by RNase E at M5 or/and M6 (Table 2 and Fig. 5). These experiments indicate that RNase E or the degradosome directly bind an internal entry site which bypasses an interaction with the 5′ end of the rpsO transcript.
Table 2. Effect of a 5′ hairpin on the stability of the rpsOΔ transcript.
Half-lives of mRNAs lacking hairpins are taken from Table 1. Half-lives of transcripts harbouring hairpins were measured in strain MG1693 grown at 30°C and transferred to 44°C as in Table 1. Errors on half-lives of mRNAs harbouring the hairpin were deduced from standard deviations of the slopes calculated by linear regression analysis.
1.1 ± 0.2
1.2 ± 0.2
1.9 ± 0.2
1.7 ± 0.17
1.2 ± 0.2
1.52 ± 0.16
6.0 ± 0.5
6.15 ± 0.09
We have proposed previously that stability of the rpsO transcripts is controlled in vivo by an RNase E cleavage occurring at a preferential site named M2 (Régnier and Hajnsdorf, 1991; Hajnsdorf et al., 1994a). Consistent with this hypothesis, we show here that mutations which prevent processing at M2 reduce the rate of degradation of these transcripts by the RNase E-dependent pathway. Moreover, we identify two other sites, where M2-less transcripts are still cleaved by RNase E in vivo and show that nucleotide substitutions in the M2 site which reduce the sensitivity of the rpsOΔ transcript to RNase E, promote its degradation by an alternative poly(A)-dependent mechanism.
Most importantly, lack of stabilization by a 5′ hairpin indicates that cleavage of the rpsO transcripts by RNase E is not facilitated by an interaction with the 5′ end of the mRNA. Rather, the dramatic effect of a deletion encompassing the M2 site on RNase E-dependent decay strongly indicates that RNase E or the degradosome primarily uses M2 or structural features in its vicinity as an entry site. The fact that decay of M2-less transcripts that are cleaved at M5 and/or M6 is also 5′-independent indicates either that RNase E is directly recruited by the sites of cleavage or that the entry site, that could be in the vicinity of M2, is physically separated from the cleavage sites (Mackie, 1998; Diwa and Belasco, 2002; Baker and Mackie, 2003; Dreyfus and Joyce, 2003). This site, might be functionally similar to direct entry sites or to the 5′ terminal and internal motifs which stimulate RNase E processing (Mackie, 1998; Diwa and Belasco, 2002; Baker and Mackie, 2003). It must be pointed out, however, that the twofold stabilization of rpsOΔ transcripts observed in a different genetic background upon addition of the same 5′-terminal hairpin indicates that an interaction with the 5′ end of the mRNA may also slightly promote its degradation by RNase E (data not shown).
Interestingly, we observed above that substitution mutations in M2 which prevent RNase E processing modify the contributions of the RNase E and poly(A)-dependent pathways to the degradation of the rpsOΔ transcript without affecting significantly its overall rate of decay. One can imagine that the interplay between the two pathways is mediated by RNase E bound at the M2 site, just upstream of the hairpin of the transcription terminator which could both cleave the mRNA at M2 and interfere with binding and activity of enzymes mediating polyadenylation or degradation of polyadenylated mRNA at the base of this hairpin (Fig. 6A). For example, a direct contact between RNase E and PAP I (Raynal and Carpousis, 1999) and/or interactions between the RNA binding domains of RNase E and poly(A) tails (McDowall and Cohen, 1996; Leroy et al., 2002) could explain why mutations which inactivate the M2 cleavage site facilitate poly(A)-dependent degradation. On the other hand, it is also conceivable that modifications of mRNA conformation resulting from mutations in the vicinity of M2 (Supplementary material, Fig. S2) could affect both RNase E processing and polyadenylation (Feng and Cohen, 2000). However, the correlation between the inhibition of the RNase E-dependent pathway and the facilitation of poly(A)-dependent decay may also reflect the fact that the global rate of the two pathways is controlled by a common limiting step that might be the binding of degradosome to the M2 entry site (Fig. 6B). This hypothesis based on the capability of the degradosome to carry out both RNase E (Lopez et al., 1999; Leroy et al., 2002) and poly(A)-dependent degradation (Py et al., 1996; Carpousis et al., 1999; Coburn et al., 1999) of mRNAs is supported by the finding that the rpsO mRNA is dramatically stabilized in cells lacking the degradosome (P. Marujo, unpubl. data) (Ow et al., 2000). The degradosome bound at M2 could either cleave the mRNA at RNase E sites or engage PNPase of the complex in exonucleolytic degradation of the mRNA. In the case of M2-containing transcripts, the degradosome bound to the M2 entry site are probably so rapidly engaged in endonucleolytic processing at M2 that polyadenylation is not required for their degradation (Fig. 6B). In contrast, when transcripts have substitution mutations in M2, cleavages at the secondary sites, M5 and M6, occur at slower rates which probably allows PNPase to attack oligo(A) tails of transcripts that are not cleaved by RNase E. The overall result is that M2-less transcripts are degraded at a similar rate to those containing M2 provided RNase E and poly(A) polymerase are both active (Fig. 6B). On the other hand, the stabilization of the rpsOΔ transcripts deleted of five nucleotides could be due to a reduction of the affinity of the degradosome for the deleted M2 site. However, visual examination did not detect obvious structural or sequence homologies between the region of the rpsO transcript surrounding the M2 cleavage site that probably contains the presumptive entry site and the proposed internal RNase E/degradosome binding site essential for feedback regulation of rne expression (Diwa and Belasco, 2002).
Interestingly, the model above is reminiscent of the mechanism of degradation of RNA I of colE1 plasmids. It has been proposed, in this case, that the degradosome, bound simultaneously at the RNase E processing site and at the 3′ end of RNA I, co-ordinates the RNase E cleavage which initiates the degradation of RNA I and the poly(A)-dependent degradation of the RNase E processed molecule (Cohen, 1995). It is also worth pointing out that the assumption that both pathways of decay are carried out by the degradosome bound at M2 implies that the thermosensitive RNase E encoded by rne1 can form degradosomes able to interact with the M2 site and to promote rapid poly(A)-dependent degradation of the rpsOΔ mRNAs at 44°C. Indeed, pΔS15(M2–1) and pΔS15(M2–2) mRNAs are efficiently degraded in a PAP+RE– strain under these conditions (Table 1). This observation is consistent with earlier data showing that inactivated thermolabile RNase E and an inactive Rne polypeptide support formation of defective degradosomes able to exonucleolytically degrade structured RNA (Py et al., 1996; Coburn et al., 1999).
We identify here two new sites referred to as M5 and M6 where RNase E cleaves rpsO transcripts in vivo in addition to the M2 and M3 sites previously characterized (Hajnsdorf and Régnier, 1999). The finding that M2-containing and M2-less rpsO transcripts are all cleaved at M5 and M6in vitro but that only the M5-t1 and M6-t1 downstream cleavage products lacking the M2 site are detected in vivo leads us to conclude that M2-containing mRNAs are presumably also cleaved at M5 and M6in vivo but that products of cleavage are not detected because they are rapidly processed at M2.
We clearly demonstrate here that there is a hierarchy of RNase E processing sites and that M2 is the preferential site of cleavage where the ribonuclease initiates decay. A corollary is that M3, M5 and M6 are minor sites which may ensure, together with exoribonucleases, the rapid degradation of truncated mRNAs devoid of the 3′ stabilizing hairpin of the transcription terminator (Hajnsdorf and Régnier, 1999). Importantly, the preferential M2 site seems to ensure pre-eminence of RNase E over poly(A)-dependent degradation whereas mRNA harbouring only minor RNase E sites can be degraded simultaneously by poly(A)-and RNase E-dependent pathways.
Interestingly, the data presented here reinforce the idea that polyadenylation is mostly used in the cell to prevent accumulation of inactive RNAs as demonstrated in the case of misfolded tRNA (Li et al., 2002) and for tightly folded mRNA fragments generated by endonucleolytic cleavages (Coburn and Mackie, 1998; Goodrich and Steege, 1999; Hajnsdorf and Régnier, 1999; Régnier and Marujo, 2003). In the case of mRNA, poly(A)-dependent decay may ensure a basal level of degradation which prevents accumulation of messages not required for protein synthesis when alteration of preferential RNase E cleavage sites or deficiency of RNase E or degradosome activity impairs the normal pathway of decay.
Strains and growth conditions
Strains MG1693 (thyA715 rph1), SK5665 (thyA715 rph1 rne1) (Arraiano et al., 1988), SK7988 (thyA715 rph1 ΔpcnB) (O’Hara et al., 1995), IBPC 694 (thyA715 rph1 pRS415), IBPC 690 (thyA715 rph1 pcnB80 pRS415) (Hajnsdorf et al., 1995), IBPC681 (thyA715 rph1 rne1), CMA304 (thyA715 rph1 rne1 ΔpcnB) and CMA303 (thyA715 rph1 rnb500 pnp7 ΔpcnB pDK39) were transformed with plasmids pΔS15(M2+), originally named pΔS15AUG(67) (Braun et al., 1998), or derivatives described in Fig. 1, and grown in Luria–Bertani (LB) medium supplemented with thymine (50 µg ml−1) and spectinomycin (50 µg ml−1). Strains CMA304 and CMA303 were obtained by P1 transduction of the ΔpcnB allele from SK7988 into SK5665 and SK5726 (thyA715 rph1 rnb500 pnp7 pDK39) (Arraiano et al., 1988) respectively. Strain IBPC681 was obtained by curing IBPC670 (Hajnsdorf et al., 1995) from the pRS415 plasmid. pRS415 is a ColE1 replicon (Simons and Kleckner, 1988) harbouring lacZ that was used to follow transduction of the pcnB80 allele.
Construction of plasmids
Uracil containing single-stranded pΔS15(M2+) DNA (Fig. 1), a low-copy-number PAP I independent derivative of pCL1921, previously reported as pΔS15AUG(67) (Braun et al., 1998) was used for site-directed mutagenesis by the Kunkel method (Kunkel et al., 1987). Plasmids pΔS15(M2–1) and pΔS15(M2–2) were created with oligonucleotides 5′-CCCC TTTTCTGGGCCCCGCAAGAATTAGCG and 5′-CCCCTTTT CTGCGCCGCGCAAGAATTAGCG respectively. The mutated nucleotides relevant to this study are underlined. Plasmid pΔS15(M2Δ) was created by PCR-directed mutagenesis of pΔS15(M2+) with oligonucleotides 5′-GGGTCTGCGTCGCT AA TTCT TGCGCAGA A A A GGGGGCCTGAG TGG and 5′-CCACTCAGGCCCCCT TT TCTGCGCAAGAA T TAGCGACG CAGACCC.
In order to insert a hairpin at the 5′ end of the rpsO transcripts, restriction sites cleaved by SmaI and AatII were created just downstream of the transcription initiation site of plasmid pΔS15AUG (20) (Braun et al., 1998) (Fig. 1). For this purpose, the uracil containing single-stranded DNA template was mutagenized with the 5′-TTT ACGCGACGT CACC CGGG AGTATTCTACTCGTAGCG oligonucleotide. Nucleotide substitutions which create the SmaI and AatII restriction sites shown in italic are underlined. The resulting plasmid is referred to as pΔS15AUG (20)M (Fig. 1). Then, an oligonucleotide linker resulting from hybridization of the oligonucleotides 5′-GTTCCGCCACCGGCAGCTGCCGGTGGCGGAA CTTAACGT and 5′-TAAGTTCCGCCACCGGCAGCTGCCG GTGGCGGAAC coding for the hairpin structure (Fig. 1) was inserted between the SmaI and AatII sites of pΔS15AUG (20)M to create pΔS15AUG (20)hp*. The HindIII–PstI fragment of pF1(+) (Fig. 1) containing the 5′ part of rpsO was replaced by the corresponding fragment of pΔS15AUG (20)hp* containing the hairpin and the resulting plasmid was mutagenized by PCR to create pΔS15(M2+)hp*, pΔS15(M2–1)hp* and pΔS15(M2–2)hp*. pΔS15(M2+)hp* was constructed by inserting an early termination codon 67 nucleotides upstream of M2 as in pΔS15(M2+) with the 5′-CTGCTCGA CT ACCTG TAACGTAAAGACGT AGCACGT T ACACC and 5′-GGTGTAACGTGCTACGTCTTTACGTTAC AGGTAGTCGAGCAG pair of oligonucleotides. Then, pΔS15(M2+)hp* was mutagenized by the same method to give pΔS15(M2–1)hp* and pΔS15(M2–2)hp* with the 5′-GCGTCGCT AA T TCTTGCGGGGCCCAG AAAAGGGGGCC TGAGTGGC and 5′-GCCACTCAGGCCCCCT T T TCTGGG CCCCGCAAGAAT TAGCGACGC pair of oligonucleotides for pΔS15(M2–1)hp* and with 5′-GCGTCGCTAATTCTTGC GCGGCGCAGAAAAGGGGGCCTGAGTGGC and 5′-GCC ACTC A GG CCCCCT TTTCTGCGCCGCGCA A G AA T T AG C GACGC oligonucleotides for pΔS15(M2–2)hp*. Plasmid pΔS15(M2Δ)hp* was constructed by the ligation of two DNA fragments: the 5655 bp PstI–HindIII fragment derived from pΔS15AUG(20)hp* (containing the hp* structure) and the 514 bp HindIII–PstI fragment derived from pΔS15(M2Δ) (containing the UAA67 and the deleted M2 site). All constructs were verified by sequence analysis.
RNA preparation and analysis
RNA preparations, Northern blots and primer extension experiments were performed essentially as previously described (Hajnsdorf et al., 1994b). Quantification of autoradiograms was performed with a PhosphorImager (Molecular Dynamics). Northern blots were probed with uniformly labelled antisense RNAs or the 5′ end labelled oligonucleotide shown on Fig. 1. The template for synthesis of the rpsO probe, covering the rpsOΔ gene nearly completely (Fig. 1), was amplified from plasmid pΔS15(M2+) with oligonucleotides 5′-TTAACGTCGCGTAAATTGTTTAACAC and 5′-TAATACGACTCACTATAGGGAAAAAAGGGGCCACT CAGG. Underlined nucleotides correspond to T7 RNA polymerase promoter. Oligoprobe (Fig. 1C) is an oligonucleotide, 5′-GAAAAAAGGGGCCACTCAGG, complementary to the last 20 nucleotides of the rpsO transcript.
In vitro transcripts for RNase E digestions were transcribed from templates amplified by PCR from plasmids pΔS15(M2+), pΔS15(M2–1), pΔS15(M2–2) and pΔS15(M2Δ) with oligonucleotides 5′-TAATACGACTCACTATAGGGCCGCTTAACGTC GCG containing the T7 RNA polymerase promoter (underlined) and 5′-TGAATTGCTGCCGTCAGCTTGAAAAAAGGG GCCACTCAGG. Its transcription by T7 RNA polymerase generates RNA beginning at the start of transcription of rpsO and terminating 20 nucleotides downstream of the transcription termination site. In vitro synthesized transcripts were processed by purified degradosome in the absence of phosphate as described (Braun et al., 1996). Purified degradosome was a gift of A. J. Carpousis (Carpousis et al., 1994). Extremities generated by RNase E and 5′ ends of in vivo rpsO mRNAs were mapped by primer extension with the 5′ end labelled 5′-AAGAATTAGCGACGCAGACCCAGGC RTprimer1 complementary to the full-length rpsO transcript between nucleotides 355 and 380 (Fig. 1) (Braun et al., 1996).
We are indebted to C. Condon, E. Hajnsdorf, V. Kaberdin, A. J. Carpousis and J. Plumbridge for suggestions and careful reading of the manuscript. A. J. Carpousis is also acknowledged for the gift of purified degradosome and Carole Pennetier for technical assistance. We would also like to thank the referees for their valuable advice. This work was supported by the Centre National de la Recherche Scientifique (UPR9073), Paris7-Denis Diderot University, the Programme de Recherche Fondamentale en Microbiologie et Maladies Infectieuses et Parasitaires of the Ministère de l’Education Nationale de la Recherche et de la Technologie, Fundacão para a Ciência e Tecnologia (F.C.T., Portugal) and the Programme of Scientific and Technical Cooperation of Ambassade de France and Instituto de Cooperação Científica e Tecnológica Internacional. P.M. is recipient of a grant from F.C.T. (Portugal).