|Home | About | Journals | Submit | Contact Us | Français|
RNase E autoregulates its production in E. coli by governing the decay rate of rne (RNase E) mRNA. It does so by a mechanism that is dependent in part on hp2, a cis-acting stem-loop within the rne 5′ untranslated region. In principal, hp2 could function either as a cleavage site for RNase E or as a binding site for that protein or an ancillary factor. Here we show that the effector region at the top of hp2 is cleaved poorly by RNase E yet binds the catalytic domain of that ribonuclease with a sequence specificity reflecting its efficacy in feedback regulation. These findings suggest that hp2 controls RNase E synthesis by binding to RNase E and expediting cleavage elsewhere within the rne transcript.
Messenger RNA degradation plays a key and universal role in the post-transcriptional control of gene expression. By governing the number of times that a message can be translated by ribosomes, it has a direct impact on levels of protein synthesis. In both prokaryotic and eukaryotic organisms, mRNA lifetimes can differ by up to two orders of magnitude, with proportionate effects on protein production.
In E. coli, mRNA decay involves the sequential action of endonucleases and 3′ exonucleases and is often triggered by pyrophosphate removal from the 5′ end (Condon, 2007; Celesnik et al., 2007; Deana et al., 2008). The endonuclease that is most important for message turnover in that species is RNase E, which is thought to participate in the degradation of most E. coli mRNAs as well as in rRNA and tRNA maturation (Apirion, 1978; Ono & Kuwano, 1979; Mudd et al., 1990; Babitzke & Kushner, 1991; Melefors & von Gabain, 1991; Taraseviciene et al., 1991; Ow & Kushner, 2002; Li & Deutscher, 2002). RNase E cuts RNA within single-stranded regions that are AU-rich, and it is particularly effective at cleaving degradation or processing intermediates that are monophosphorylated at the 5′ end due to pyrophosphate removal or endonucleolytic cleavage (McDowall et al., 1994; Mackie, 1998). In view of its many crucial biological functions, it is not surprising that RNase E is an essential E. coli enzyme and that its overproduction or underproduction can impede cell growth (Apirion, 1978; Claverie-Martin et al., 1991; Jain et al., 2002).
To ensure a steady supply of this protein, E. coli and related bacteria have evolved a homeostatic mechanism for tightly regulating its synthesis by modulating the decay rate of rne (RNase E) mRNA in response to changes in cellular RNase E activity (Jain & Belasco, 1995). Feedback regulation by RNase E is mediated in cis by the rne 5′ untranslated region (UTR), which can confer this property onto heterologous messages to which it is fused (Jain & Belasco, 1995). The rne 5′ UTR is composed of six structural domains (Figure 1), two of which, the simple stem-loop hp2 and the branched stem-loop hp3, contribute to feedback regulation (Diwa et al., 2000). Mutational and phylogenetic studies have attributed the influence of rne hp2 to its uppermost region, which comprises a conserved internal loop and an apical hairpin loop separated by two base pairs (Diwa & Belasco, 2002). However, little is known about the mechanism by which this stem-loop directs feedback regulation by RNase E.
Here we show that, although the effector region at the top of rne hp2 is a poor target for cleavage by RNase E, it nonetheless binds to that enzyme with a specificity that reflects its regulatory activity. These results suggest that the upper portion of hp2 helps to mediate feedback regulation by binding directly to RNase E and facilitating cleavage elsewhere in the rne transcript.
In principle, rne hp2 might mediate feedback regulation of RNase E synthesis by serving as a cleavage site for that endonuclease. To test that possibility, we examined the propensity of N-RNase E (the catalytically active amino-terminal domain of RNase E, comprising amino acid residues #1–498 (McDowall & Cohen, 1996; Jiang et al., 2000)) to cleave hp2 in vitro. A 5′-end-labeled segment of the rne 5′ UTR comprising hp1, ss1, and hp2 (rne1-110) was synthesized by in vitro transcription and treated with purified N-RNase E for various time periods. Gel electrophoresis and autoradiography revealed multiple digestion products, almost all of which resulted from cleavage within single-stranded regions of hp1 or ss1 (Figure 2A). By contrast, cleavage within hp2 was barely detectable. What little cleavage occurred there was primarily in the lower portion of the stem outside the effector region that mediates the regulatory function of hp2.
We next compared the cleavage rate of rne1-110 to that of a hp2 analog (hp2*) whose upper half – an internal loop and hairpin loop joined by a 2-bp stem – was identical to the effector region at the top of hp2 and whose lower half comprised a fully base-paired stem (Figure 3A). Because it retains the portion of hp2 that is important for function, hp2* is just as effective as hp2 at mediating RNase E feedback regulation of an rne-lacZ reporter, as determined by assaying β-galactosidase activity in isogenic E. coli strains containing low versus high RNase E activity (Table 1). In vitro, N-RNase E cleaved hp2* much more slowly than rne1-110 when the two RNAs were synthesized separately and then combined (Figure 2B).
Together, these findings indicate that the effector region of hp2 is an intrinsically poor target for RNase E cleavage compared to other parts of the rne 5′ UTR. This conclusion raises the possibility that the upper portion of rne hp2 could mediate the degradation of rne mRNA not by serving as a cleavage site but by binding to either RNase E itself or an ancillary protein and facilitating RNase E cleavage elsewhere within the message. To distinguish between the latter two possibilities, we investigated whether the effector region of hp2 binds directly to RNase E.
We first tested whether the catalytic amino-terminal domain of RNase E could protect hp2* from digestion by a nonspecific ribonuclease. Radiolabeled hp2* that had been synthesized by in vitro transcription was treated with RNase A in the presence or absence of N-RNase E (Figure 4). This experiment revealed that N-RNase E can markedly slow RNase A digestion of hp2*. The finding that N-RNase E can protect hp2* from nucleolytic attack suggests a direct physical interaction between them.
The interaction implied by the ribonuclease protection assay was examined more rigorously by photocrosslinking studies with a hp2* variant containing a nucleotide analog (4-thiouridine) that could be specifically activated by long-wave ultraviolet light. In vitro transcription of a hp2* template in the presence of 4-thioUTP yielded hp2*HI, an RNA identical to hp2* but for the incorporation of 4-thiouridine into both the hairpin loop and the internal loop (Figure 3A), two regions crucial for hp2 function in vivo (Diwa & Belasco, 2002). Irradiation of radiolabeled hp2*HI with long-wave UV light in the presence of N-RNase E yielded two prominent crosslinked products upon gel electrophoresis (Figure 5A). Crosslinking of full-length RNase E to hp2*HI was also observed (Figure 5C). By contrast, little crosslinking to N-RNase E was detected when an unrelated thiouridine-containing stem-loop was tested as a control.
To assess the functional significance of the observed interaction, we next investigated its specificity by determining whether the nucleotide sequence needed for hp2 binding correlated with that required for efficient autoregulation of RNase E synthesis. Previous mutational and phylogenetic analyses have identified the sequence characteristics of rne hp2 that are important for its ability to mediate feedback regulation in E. coli (Figure 3B) (Diwa & Belasco, 2002). On the basis of those studies, four hp2* variants were designed. Three (A6C/G7A, A6U/G7A and C1A/G3C) bore double mutations of key nucleotides in the internal loop; those mutations were expected to impair feedback regulation. Another double mutant (A2G/G3A) was expected to retain regulatory function. All four mutants were predicted to preserve the overall secondary structure of hp2*, as judged by computational analysis.
The regulatory function of each of the four double mutants was first tested in E. coli in the context of an rne-lacZ reporter in which hp2 had been replaced by hp2* or variants thereof. As expected, the hp2* variants bearing pairs of mutations predicted to impair hp2 function (A6C/G7A, A6U/G7A and C1A/G3C) proved defective in feedback regulation, whereas the substitutions predicted to be innocuous (A2G/G3A) were functionally well tolerated, with little or no loss of regulatory potential (Table 1).
The N-RNase E crosslinking efficiency of hp2*HI and its variants was then compared. The five radiolabeled RNAs were prepared by in vitro transcription and gel-purified, and equimolar amounts of each were combined with N-RNase E and exposed to long-wave UV light. As before, two crosslinked products were detected upon gel electrophoresis (Figure 6). Unlike the intensity of the faster migrating (lower) band, which was similar for all five RNAs, the intensity of the slower migrating (upper) band was significantly reduced for the three variants impaired for feedback regulation but undiminished for the fully functional hp2* variant. Thus, the slower migrating band appeared to represent a sequence-specific and functionally productive interaction of hp2* with N-RNase E, whereas the faster migrating one likely represented a nonproductive interaction.
In principle, the two crosslinked products of hp2*HI might result from crosslinking of N-RNase E to one or the other thiouridine nucleotide, each of which lies in a region of the hp2 effector domain known to influence its regulatory activity (Diwa & Belasco, 2002) and therefore likely to contact the protein. To clarify their origin, the crosslinked products obtained with hp2*HI were compared to those obtained with radiolabeled hp2* derivatives containing a single 4-thiouridine residue in either the hairpin loop (hp2*H) or the internal loop (hp2*I). The major crosslinked product of hp2*H co-migrated with the low-mobility hp2*HI product, whereas that of hp2*I co-migrated with the high-mobility hp2*HI product (Figure 5B). We conclude that the slower migrating crosslinked product arises primarily from the covalent attachment of N-RNase E to a thiouridine residue in the hairpin loop of hp2*.
To confirm the regulatory productivity of the binding event represented by the low-mobility product, the relative affinity of N-RNase E for hp2* and each of the four sequence variants was examined by comparing the ability of these RNAs to compete with hp2*H for binding. Unlabeled hp2* and mutants thereof were prepared by in vitro transcription in the absence of 4-thioUTP and gel purified. A ten-fold molar excess of each was then tested for its ability to inhibit crosslinking of radiolabeled hp2*H to N-RNase E. Both hp2* and its functionally active A2G/G3A derivative acted as potent crosslinking inhibitors, whereas the three hp2* mutants defective in feedback regulation (A6C/G7A, A6U/G7A and C1A/G3C) had little inhibitory effect (Figure 7). These results, which are in complete agreement with the comparative crosslinking experiments shown in Figure 6, corroborate the conclusion that hp2* binds directly to N-RNase E with a sequence specificity that correlates well with regulatory activity.
We next used protein mutagenesis to identify RNase E residues important for crosslinking of hp2*H. Thirteen of the most highly conserved residues in the catalytic domain of RNase E were changed to alanine, and the influence of those mutations on the crosslinking efficiency of N-RNase E was determined (Figure 8). Most had little or no impact (≤2-fold), but three (F57A, R169A, and N323A) had a large (≥4-fold) inhibitory effect. Consistent with their deleterious influence on hp2* binding, these three mutations reduced the feedback regulatory activity of full-length RNase E in E. coli by a factor of 4–12, as measured with an rne-lacZ reporter (Table 2). Each may impair autoregulation not only by acting directly or indirectly to hamper hp2 binding but also by diminishing the catalytic activity of RNase E (Diwa et al., 2002; Callaghan et al., 2005; Jourdan & McDowall, 2008).
Our data indicate that, despite being an intrinsically poor target for RNase E cleavage, the effector region at the top of rne hp2 can bind to the catalytic domain of that enzyme with a sequence specificity matching that for its role in the regulation of RNase E synthesis. These findings suggest that this effector region contributes to feedback regulation by interacting directly with RNase E and facilitating cleavage elsewhere in the rne transcript. Such binding may enable RNase E to circumvent the impediment to decay imposed by the presence of a triphosphate and stem-loop (hp1) at the 5′ end of the rne transcript, which should block access to the rapid 5′-monophosphate-dependent pathway for RNA degradation (Mackie, 1998; Celesnik et al., 2007; Deana et al., 2008).
In principle, hp2 binding might expedite cleavage of rne mRNA by increasing the affinity of RNase E for that message, by altering the conformation of the enzyme and allosterically enhancing its catalytic activity, and/or by inhibiting translation of the message and thereby exposing it to ribonucleolytic attack. That hp2* can apparently bind in two ways, only one of which is productive, suggests that its ability to stimulate cleavage requires a particular mode of binding, perhaps in a pocket on the enzyme surface that is distinct from the active site. If so, this interaction and its consequences may bear some resemblance to the manner in which monophosphorylated 5′ ends bind to the catalytic domain of RNase E and faciliate RNA cleavage downstream (Callaghan et al., 2005). The role of the rne hp3, the branched stem-loop adjacent to hp2, is not yet clear.
The regulatory influence of hp2 resulting from its sequence-specific interaction with RNase E helps to explain why the rne transcript is significantly more sensitive than most other primary transcripts to changes in cellular RNase E activity. This exceptional property of rne mRNA is key to the selective homeostatic control of that important ribonuclease.
Plasmid pNRNE7000 encodes N-RNase E (the catalytic domain of E. coli RNase E, corresponding to amino acid residues 1–498) bearing both a hexahistidine affinity tag and a c-Myc epitope tag at the carboxyl terminus (Jiang & Belasco, 2004). Plasmid pRNE120, a derivative of plasmid pRNE101 (Jain & Belasco, 1995), encodes full-length RNase E bearing tandem hexahistidine and c-Myc tags at the carboxyl terminus. Plasmid pRNE2000, a derivative of plasmid pMPM-K1, encodes full-length RNase E bearing tandem hexahistidine and c-Myc tags at the amino terminus (Jiang et al., 2000).
Plasmids pEZ-hp2* and pEZΔhp2 are derivatives of the rne-lacZ reporter plasmid pEZ8-hp2-botsyn (Diwa & Belasco, 2002) in which hp2 has been replaced either by hp2* or by an unrelated stem-loop, respectively. Plasmids pEZ-hp2*-A6C/G7A, pEZ-hp2*-A6U/G7A, pEZ-hp2*-C1A/G3C, and pEZ-hp2*-A2G/G3A are pEZ-hp2* derivatives with two point mutations in hp2*.
N-RNase E (comprising residues 1–498 of E. coli RNase E fused to C-terminal hexahistidine and c-Myc tags) and mutants thereof were purified from E. coli strain BL21(DE3) containing plasmid pRNE7000 or a derivative, as previously described (Jiang et al., 2000). Full-length RNase E bearing C-terminal hexahistidine and Myc tags was harvested from E. coli strain CJ1832 (Jain et al., 2002) containing plasmid pRNE120 after growth to late-log phase (A600 = 0.7–0.8) in the absence of IPTG. The cells were pelleted, resuspended in buffer P (20 mM HEPES [pH 7.5], 0.25% Genapol X-080, 0.1 mM PMSF, and Complete EDTA-free protease inhibitor [Roche]), and lysed with a French press. The lysate was cleared by centrifugation (6000 × g for 25 min), and His-tagged RNase E was purified by affinity chromatography on Talon beads (Clontech) eluted with imidazole (150 mM in buffer P). The eluted protein was dialyzed overnight against buffer P and stored at −80°C in the presence of glycerol (10%). Protein concentrations were determined by comparative Coomassie blue staining of bands in SDS-polyacrylamide gels.
Template DNA for in vitro transcription was produced by PCR amplification and transcribed with an Ampliscribe T7-Flash transcription kit (Epicentre) according to the manufacturer’s instructions, in the presence of SuperaseIn (0.8 U/μl; Ambion). To synthesize 5′-end-labeled hp2*, the reaction mixture contained GTP (2 mM), ATP (6 mM), UTP (6 mM), CTP (6 mM), and [γ-32P] GTP (2 μCi/μl). To produce internally radiolabeled hp2* and mutants thereof, transcription was carried out in the presence of GTP (6 mM), ATP (6 mM), UTP (6 mM), CTP (0.16 mM), and [α-32P] CTP (80 μCi/ml). For internally radiolabeled transcripts bearing 4-thiouridine, the transcription reaction contained 4-thioUTP (0.80 mM; TriLink Biotechnologies) instead of UTP. Unlabeled RNA was synthesized in a reaction mixture containing GTP, ATP, UTP, and CTP (6 mM each). DNA templates for synthesizing 5′-end-labeled rne1-110 and rne1-48 were generated by PCR amplification of the rne 5′ UTR with a forward primer (5′-CGGAATTCAAATTAATACGACTCACTATAGGTTTCCGTGTCCATCC-3′) and either of two reverse primers (5′-AACTGCCTGAAAGATCAATACG-3′ or 5′-GGGTTATTCCGTAAAATTTCTTG-3′, respectively) and then transcribed in a reaction mixture containing GTP (2 mM), ATP (6 mM), UTP (6 mM), CTP (6 mM), and [γ-32P] GTP (80μCi/ml). Purified hp2*H and hp2*I RNA were obtained from TriLink Biotechnologies and 5′-end-labeled by treatment with T4 polynucleotide kinase (New England Biolabs) and [γ-32P] ATP.
The resulting RNAs were purified by electrophoresis on a 10% polyacrylamide-6.7 M urea gel, extracted into RNA buffer (10 mM Tris [pH 7.5], 100 mM NaCl), ethanol precipitated, and redissolved in RNA buffer. Concentrations of radiolabeled RNAs were calculated on the basis of their radioactivity as measured in a liquid scintillation counter, while concentrations of unlabeled RNAs were estimated by gel electrophoresis and ethidium bromide staining, using an RNA oligonucleotide of the same length and a known concentration as a standard for comparison.
5′-end-labeled rne1-110 RNA (660 nM, [γ-32P]), in the presence or absence of 5′-end-labeled hp2* RNA, was incubated with N-RNase E (2.5–16 μM) in 10 mM HEPES [pH 7.5], 10 mM NaCl, and 8 mM MgCl2 at 24°C. Reaction samples were quenched at time intervals with loading buffer (90% formamide, 20 mM EDTA, 0.05% bromophenol blue, 0.05% xylene cyanol), heated to 95°C, and analyzed by electrophoresis on a 12% polyacrylamide-6.7 M urea gel beside a pair of RNA size standards (rne1-110 and rne1-48). Sites of cleavage were estimated from a semilogarithmic plot of fragment length versus electrophoretic mobility.
Internally radiolabeled hp2* RNA (10 nM, [α-32P]) was incubated for 30 sec on ice with N-RNase E (0 or 2.5 μM) in 10 mM Tris [pH 7.5], 100 mM NaCl, and 8 mM MgCl2 before adding RNase A (5 ng/μl) and raising the temperature to 25°C. Reaction samples were quenched at time intervals with loading buffer (90% formamide, 20 mM EDTA, 0.05% bromophenol blue, 0.05% xylene cyanol), heated to 95°C, and analyzed by electrophoresis on a 12% polyacrylamide-6.7 M urea gel.
UV crosslinking was performed on ice with N-RNase E or RNase E (250 nM) and a radiolabeled, 4-thiouridine-containing RNA (500 nM) in 15 mM Tris (pH 7.5), 5 mM MgCl2, 50 mM KCl, and 5% glycerol. Crosslinking was induced by irradiation for 20 min with a hand-held long-wave UV lamp (Blak-Ray lamp, UVP Inc.) placed directly on the top of the microcentrifuge tubes that contained the samples. Competitive inhibition of crosslinking was examined by adding a ten-fold molar excess of an unlabeled competitor RNA just before adding the radiolabeled, 4-thiouridine-containing RNA. Protein and RNA concentrations were doubled when examining the effect of alanine substitutions on the crosslinking efficiency of N-RNase E. Crosslinking products were detected and quantified by using a Molecular Dynamics Storm 820 Phosphorimager.
Assays of feedback regulation of rne-lacZ reporters bearing variants of rne hp2 (ezΔ114–337 and derivatives thereof) were performed at 37°C with log-phase cultures of the isogenic E. coli strains CH1827+pRNE101 (high RNase E activity) and CH1828 (low RNase E activity), as previously described (Diwa et al., 2000). The effect of alanine substitutions on feedback regulation by RNase E was examined in E. coli strain CJ1828 (which has a chromosomal rne-lacZ reporter and low intrinsic RNase E activity) (Jain & Belasco, 1995) containing plasmid pRNE2000 and mutants thereof or plasmid pMPM-K1 (negative control) and corrected for slight differences in the cellular concentration of the RNase E variants, as described (Jiang et al., 2000).
We are grateful to Evgeny Nudler and Vitaly Epshtein for their advice and assistance. This research was supported by a grant to J. G. B. from the National Institutes of Health (GM35769).