Stabilization of Circular rpsT mRNA Demonstrates the 5′-End Dependence of RNase E Action in Vivo *

RNase E is the major intracellular endonuclease in Escherichia coli. Its ability to cleave susceptible substrates in vitro depends on both the cleavage site itself and the availability of an unstructured 5′ terminus. To test whether RNase E activity is 5′-end-dependent in vivo in the presence of all the components of the RNA degradative machinery, a known substrate, the rpsT mRNA, has been embedded in a permuted group I intron to permit its efficient, precise circularization in E. coli. Circular rpsTmRNAs are 4–6-fold more stable in vivo than their linear counterparts. Even partial inactivation of RNase E activity further enhances this stability 6-fold. However, the stabilization of circular rpsT mRNAs depends strongly on their efficient translation. These results show unambiguously the importance of an accessible 5′-end in controlling mRNA stability in vivoand support a two-step (“looping”) model for RNase E action in which the first step is end recognition and the second is actual cleavage.

The degradation of mRNAs is an important, if incompletely understood, aspect of the regulation of gene expression. In Escherichia coli, the initiating step in the decay process is usually mediated by RNase E (1-3), a 5Ј-end-dependent endoribonuclease (4). In vitro, RNase E activity is conferred by a multi-enzyme complex, the degradosome (5,6). In small RNAs, such as the ColE1 replication regulator RNA 1, and the rpsO or rpsT mRNAs encoding ribosomal proteins S15 and S20, respectively, a single RNase E cleavage is capable of inactivating the mRNA and rendering it susceptible to complete destruction to mononucleotides (reviewed in Ref. 3). In larger mRNAs this initiating endonucleolytic event can trigger a 5Ј 3 3Ј "wave" of subsequent endonucleolytic cleavages, which rapidly inactivate the entire mRNA (7). The 3Ј termini generated by RNase E cleavages are scavenged by 3Ј-5Ј-exoribonucleases (1,3).
Two features in the 5Ј-extremity of an mRNA, secondary structure and the triphosphate terminus, can control the sus-ceptibility of the entire mRNA toward RNase E. The 5Ј-terminal stem-loop structure of the ompA mRNA is largely responsible for the atypical stability of this mRNA and can confer stability to heterologous mRNAs to which it is grafted (8,9). Stabilization is abolished by as few as three single-stranded residues at the extreme 5Ј-end of an RNA (10). These effects of 5Ј-terminal secondary structure on mRNA stability are mediated directly though Rne (4,11). Evidence for the critical role of the free 5Ј-end of RNA and its phosphorylation state in mRNA turnover arises from several experiments. Most notably, circular derivatives of the well characterized RNase E substrates, rpsT mRNA or 9 S RNA, are highly resistant to cleavage in vitro by Rne or degradosomes (4). An RNase E cleavage at a site 5 residues from the 5Ј-end of RNA 1 destabilizes the 103residue 3Ј-cleavage product in vivo. However, an artificial RNA identical to the initial cleavage product but containing a triphosphorylated terminus is significantly more stable (12). Likewise, oligonucleotide-directed cleavage of the rpsT mRNA by RNase H shows that the 5Ј-segment containing a triphosphorylated terminus is quite stable, even in relatively crude extracts. In contrast, the 3Ј-segment that would be monophosphorylated undergoes RNase E cleavage at an accelerated rate compared with the unmodified substrate (11,13). Finally, either Rne protein alone or purified degradosomes preferentially cleave monophosphorylated substrates 20 -30-fold more rapidly than their triphosphorylated counterparts (4). Together, these results imply that the vectorial nature of mRNA decay is a reflection of the inherent preference of Rne for 5Ј-monophosphorylated substrates.
To address whether Rne or degradosomes display 5Ј-end dependence in vivo where a circular RNA would be exposed to all the components of the degradative machinery, I have constructed chimeric RNAs in which a portion of the rpsT mRNA is embedded in a permuted group I intron, a construction that should permit its precise, autocatalytic circularization (14,15). In this work I show that circular rpsT mRNAs form efficiently and are 4 -6-fold more stable than their linear counterparts. Circular RNAs are, however, cleaved slowly by RNase E in a 5Ј-end-independent fashion, which is highly sensitive to ongoing translation. or 37°C (JM109), shifted to 39.5°C as required, induced with 0.5 mM IPTG for 20 min, and when necessary treated with rifampicin (Ͼ160 g/ml) to inhibit transcription. Portions (2.0 ml) of the culture were harvested at intervals thereafter, and RNA was extracted using method II (18). Yields were quantified by A 260 , and the quality of the RNA was assessed by the intactness of rRNA. For RT-PCR, 2 g of total RNA (in some cases pretreated with RNase H and oligonucleotide 481 (13)) was denatured at 90°C and then mixed with a buffer containing 25 mM Tris-HCl, pH 8.3, 37.5 mM KCl, 1.5 mM MgCl 2 , 0.5 mM dNTPs, 10 mM dithiothreitol, 2.5 M oligonucleotide 992 (5Ј-CTGAATG-GCAAGCTTCTTAGC; complementary to residues 151-169 in the rpsT sequence (13)). Following addition of 100 units of Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.) and incubation for 30 min at 42°C, a portion of the products corresponding to 0.25 g of RNA template was amplified with 5 units of Taq DNA polymerase (Life Technologies, Inc.) for 30 cycles in 10 mM Tris-HCl, pH 8.4, 50 mM KCl, 1.5 mM MgCl 2 , 100 M dNTPs, and 1 g of primer 991 (5Ј-GCTAAAGGTCGGATCCACAAA; complementary to residues 319 -340 in the rpsT sequence (13)) and primer 992. For Northern blotting, samples of total RNA were dissolved in a buffer containing 90% formamide, heated to 100°C, and separated on a 6% polyacrylamide gel containing 8 M urea. RNAs were transferred electrophoretically to Hybond-N (Amersham Pharmacia Biotech), fixed, and probed to detect the rpsT mRNAs with either a complementary RNA or 5Ј-32 P-labeled oligonucleotide 993 (5Ј-GGTAGACCTGAGATCTTC). Hybridization with oligonucleotides was performed at 37°C in a buffer containing 0.75 M NaCl, 75 mM sodium citrate (pH 7), 0.1% polyvinylpyrrolidone, 0.1% Ficoll, 0.1% bovine serum albumin (fraction V), 20 mM Pipes (pH 6.4), 25 g/ml salmon sperm DNA, 62.5 g/ml yeast RNA, 0.1% SDS. Membranes were washed with 0.3 M NaCl, 0.03 M sodium citrate (pH 7), 0.1% SDS at 38 -39°C.

Synthesis and Characterization of Circular rpsT RNAs-
Based on the finding that the separate "halves" of a group I intron can self-assemble to form an active ribozyme (19) (Fig. 1a), permuted group I introns have been designed to drive the circularization of "passenger" sequences inserted between the halves of the intron (14,15). Sequences from the E. coli rpsT gene (13) were inserted into the polylinker in pRR1 (17), flanked by the appropriate group I intron sequences (Fig. 1b) and verified by DNA sequence analysis. A set of constructions, whose prototype is pGM110, contains rpsT sequences from residue 93 to 413, including the entire 5Ј-untranslated region, the coding sequence, and part of the 3Ј-untranslated region but lacking the transcriptional terminator (Fig. 1b). The plasmids pGM113 and pGM116 (see below) differ from pGM110 only by point mutation. The insert in pGM111 extends from residue 132 adjacent to the initiation codon (altered from UUG to AUG) to residue 413 and lacks the 5Ј-untranslated region completely (Fig. 1b). All constructions contain a major internal RNase E cleavage site at rpsT residues 300 -301 as well as other minor sites (20).
A Northern blot of total cellular RNA extracted from cultures of JM109/pGM110 or JM109/pGM111 with or without induction showed that expression of putative circular rpsT RNAs of 336 or 297 nucleotides, respectively, was induced by IPTG to appreciable levels. These RNAs migrated at roughly half the mobility of the linear rpsT mRNAs (Fig. 2a, compare lanes 2-3  and 4 -5). The same blot was reprobed with a complementary oligonucleotide spanning the circular junction, which clearly identified the major species in lanes 3 and 5 as circular (Fig. 2a,  lanes 6 -10). A permutation analysis (4) demonstrated that the putative circular RNAs could be converted to linearly permuted, full-length monomers after treatment with RNase H and an appropriate oligonucleotide (data not shown). Further substantiation that circular rpsT RNAs are formed in vivo was obtained by using "inverse" RT-PCR (see "Experimental Procedures"). Primers were selected such that only a circular RNA template can give rise to an amplified cDNA product (Fig. 3a). Synthesis of the expected 189-bp double-stranded DNA product from total RNA isolated from JM109/pGM110 (Fig. 3b, lanes 2, 3, and 6) was dependent on reverse transcription in the first step (compare lanes 6 and 7) and was not affected by prior treatment of the RNA template with DNase I (data not shown). Yields of the anticipated product of 149 bp from total RNA extracted from JM109/pGM111 were low (Fig. 3b, lane 4) but improved when the RNA template was initially linearized by oligonucleotide-directed RNase H cleavage at a location outside the sequences to be amplified (Fig. 3b, compare lanes 4 and 5). The amplification products of "inverse" RT-PCR were purified, cloned into pUC19, and sequenced. Both sequences agreed completely with those predicted from the sequence of pRR1 and the group I intron-rpsT boundaries in pGM110 and pGM111 (data not shown). Finally, a point mutation, G 3 C at position J7/8-5 (21), known to reduce the rate of self-splicing by group I introns by Ͼ10 3 , was introduced into the 3Ј-"half " of the group I sequence in pGM110 generating pGM116. This mutation abolishes formation of rpsT/circ-336 RNA (not shown). Taken together, the Northern blots, the permutation analyses, and the sequences of the RT-PCR products demonstrate unambiguously that expression of the rpsT sequences embedded in pGM110 or pGM111 yields circular rpsT RNAs, which accumulate to readily detectable levels, roughly 4 -5-fold higher than those of the endogenous rpsT mRNAs (Fig. 2a, lanes 1-5).
Stability of Circular rpsT mRNAs-Measurements of RNA half-life were performed using standard methods (1) (see "Ex- FIG. 1. a, principle of circularization. In step 1, the 5Ј and 3Ј "halves" of a group I intron (open boxes) can assemble into an active conformation in the presence of an intervening "passenger" sequence (solid box) (14,15,19). Attack of guanosine on the 5Ј exon-intron junction leads to "splicing" of the ends of the exon and its circularization in step 2. The vertical arrows show the 5Ј and 3Ј splice sites in permuted order. b, constructions that circularize the rpsT mRNA. The upper line shows a schematic of the vector pRR1 (17) whereas the two lower lines illustrate the portions of the rpsT gene (solid boxes) inserted between the NcoI and BglII sites in the vector in pGM110 and pGM111. Numbers give the sequence coordinates of the rpsT gene (13).
perimental Procedures"). The rate of decay of rpsT RNAs extracted from strain JM109/pGM110 was measured initially from blots probed with a complementary rpsT RNA (data not shown). The circular rpsT/circ-336 RNA disappeared with a half-life of 290 s. In contrast, two linear RNAs larger than 500 nucleotides containing rpsT sequences, presumably incompletely processed species, disappeared with half-lives of 50 and 33 s (data not shown). A typical Northern blot of RNA extracted from cultures of JM109/pGM111 probed with oligonucleotide 993, specific for circular rpsT RNAs, is shown in Fig. 2b. Two species of RNA are detectable, the major upper band representing intact circular RNA (rpsT/circ-297) and the lower fainter band, marked with an X in the left margin, which indicates circles nicked during isolation. The half-life of the intact circular species (rpsT/297-circ) was 560 s in this experiment. A similar blot with the same samples was probed with oligonucleotide 995, complementary to sequences near the 3Ј-end of the T4 td group I intron in the vector (Fig. 2c). A smear of discrete bands is visible in the zero time sample, extending from about 400 to over 1000 nucleotides (Fig. 2c, lane 0). Virtually all the detectable species disappear within 3-4 min. Half-lives were measured for three discrete non-circular species shown by triangles in the left margin in Fig. 2c, yielding values of 49 -57 s. Thus either circular rpsT mRNA species is considerably more stable than its linear precursors.
Although significantly stabilized, the circular rpsT RNAs described above are, nonetheless, metabolically labile. To determine whether RNase E is responsible for initiating the degradation of circular rpsT RNAs in vivo, the stabilities of the rpsT/circ-336 RNAs in either MG1693/pGM110 (Fig. 2d) or SK5665 (rne1)/pGM110 (Fig. 2e) were measured after cultures grown at the permissive temperature of 29°C were induced with IPTG and shifted to 39.5°C for 20 min to achieve partial inactivation of the rne1 gene product prior to RNA extraction. The half-life of the rpsT/circ-336 species in strain SK5665/ pGM110 is 1800 Ϯ 155 s (Fig. 2e). This represents a 6-fold increase relative to the same RNA in the wild type strain under identical conditions (296 Ϯ 33 s) (Fig. 2d).
Circular RNAs, including the rpsT/circ RNAs, are translated in vivo (Ref. 17 and additional data not shown). To assess a role for translation in the stability of circular RNAs, half-lives were determined for several circular rpsT RNAs containing mutations affecting translational efficiency (Table I). For comparison, halflives are also given for a set of linear, chimeric rpsT mRNAs  993 (b, d, and e). RNAs in c were probed with oligonucleotide 995 (5Ј-GGTCGTTAATCTTACCCC) complementary to group I intron sequences. Triangles in the right margin denote circular rpsT species. The triangles in the left margin of c refer to the RNAs mentioned in the text. X denotes nicked species.
FIG. 3. Inverse PCR detection of circular rpsT RNA. a, principle of the method. Reverse transcription is primed by oligonucleotide 992, and the product is amplified by oligonucleotides 991 and 992 (see "Experimental Procedures"). The position and orientation of the primers are such that circular RNA (left) but not linear RNA (right) can serve as an amplifiable template. b, analysis of inverse RT-PCR products. Template RNA was extracted from JM109/pGM110 (lanes 2, 3, 6, and 7; predicted product of 189 bp) or JM109/pGM111 (lanes 4 and 5; predicted product of 149 bp) and used for RT-PCR (see "Experimental Procedures"). In lanes 3 and 5, the template RNA was linearized outside the region of amplification with RNase H and oligonucleotide 481 (13) prior to RT-PCR. In lane 7, reverse transcriptase was omitted in the first step.
(rpsT/614) containing similar or identical rpsT sequences embedded between 40 residues of lac operon sequence in their 5Ј-leader and a portion of the 3Ј-end of the rrnB operon (16). Such transcripts also effectively mimic the linear precursors to the circular rpsT RNAs. The data in Table I show that a circular rpsT RNA is 4 -6-fold more stable than a linear mRNA containing the same rpsT sequences regardless of translational efficiency. Circular rpsT/circ-336 (UUG) RNA, which spans residues 93-412 (Fig. 1b), exhibits a half-life 5-fold greater than rpsT/614 (UUG) mRNA spanning the same rpsT sequences (Table I, line 1). A UUG to AUG change at the initiation codon increases expression of the S20 protein up to 6-fold in vivo (16). The half-life of the rpsT/circ-297 (AUG) RNA is 4-fold greater than its linear counterpart, rpsT/614 (AUG) ( Table I, line 3). A double mutation in the rpsT leader (16) (G129C, A130U; "Ϫ3, Ϫ4"), which reduces S20 expression substantially, was introduced into pGM110 to form pGM113. The half-life of the rpsT/circ-336 RNA encoded by pGM113 is almost 2-fold lower than the nearly identical circular RNA encoded by pGM110 (compare lines 1 and 2 in Table I) but is still much higher than the corresponding rpsT/614 mRNA containing the Ϫ3, Ϫ4 mutation (Table I, line 2). These results show that the stabilities of circular mRNAs are subject to almost the same degree of translational modulation as their linear counterparts and more interestingly that circularization protects the rpsT mRNA most effectively when translation is highly efficient (compare rpsT/297-circ (AUG) to rpsT/336-circ (Ϫ3, Ϫ4) in Table I).
The resistance of a circular RNA to exoribonucleases is obvious (14); it is thus surprising, if not counterintuitive, that a circular RNA is significantly more resistant to RNase E, a single-stranded specific endoribonuclease of limited sequence preference (2,3,22), when the same RNA in linear form is an excellent substrate (4,13,20). This finding is, however, fully consistent with in vitro data (4). Moreover, the increased stability of the circular rpsT/circ-336 (UUG) mRNA when RNase E activity is reduced shows that other ribonuclease activities in E. coli (3), including RNase III, RNase G/CafA, and RNase I/I*/M, are unable to compensate efficiently. These findings clearly demonstrate that RNase E is a 5Ј-end-dependent endoribonuclease in vivo as well as in vitro (4,23,24) and that the frequently observed 5Ј 3 3Ј vectorial character of mRNA decay in vivo (1,3,4,7) is due to the end dependence of RNase E (4,23).
The initial attack of RNase E on circular RNAs in vivo is 5-fold less efficient than on linear RNAs of similar primary sequence. This can be rationalized by a two-site or looping model (3). In this model, RNase E in the degradosome would first contact linear substrate mRNAs at their extreme 5Ј-ends. This 5Ј-contact would stabilize the enzyme-substrate complex and would facilitate, possibly in a first order rearrangement ("looping"), recognition of internal cleavage sites that may be partially hindered by adjacent secondary structures. In circular RNAs, there would be no 5Ј-contact, and cleavage site recognition itself would become second order and much less efficient unless the cleavage site were particularly accessible or exposed. Thus the rate of attack of RNase E on a circular RNA, rather than on a triphosphorylated RNA as suggested elsewhere (23), defines its basal rate of activity. As shown here, linear RNA with an unobstructed 5Ј-triphosphate terminus is attacked 5-fold faster than a circular RNA, showing that a triphosphate terminus can assist substrate recognition. A 5Јmonophosphate terminus provides a further 20 -25-fold stimulation of the initial rate of RNA cleavage by RNase E in vitro (4,23) and by a significant but undefined amount in vivo (12). One clear implication of these data for the two-step model is that translation affects the second (rearrangement) step rather than the initial recognition of the 5Ј-end of the substrate.
It is likely that the degradation of many mRNAs will follow the behavior of the rpsT mRNA and exhibit 5Ј-end dependence. Any exceptions that prove to be insensitive to the state of the 5Ј terminus will likely be explained by the presence of readily accessible RNase E sites in the body of the mRNA (permitting an increase in the basal rate of RNase E cleavage) or by inefficient translation. in vivo Half-lives of RNAs were determined as described under "Experimental Procedures." RNAs were extracted from nearly isogenic strains differing only in the plasmid-encoded rpsT sequences. Linear rpsT/614 mRNAs encoded by pCD6, pGP11, and pGP12 initiate at a tac promoter and contain from 5Ј to 3Ј: 40 residues of lac leader mRNA, rpsT sequences to residue 412, and 254 residues from the 3Ј-end of the rrnB operon (16). The rpsT leader extends from residue 92 to 132 in pCD6 and pGP12 and from residue 93 to 132 in pGM110 or pGM113 (Fig. 1b), followed by the UUG initiation codon. The Ϫ3, Ϫ4 mutation in pGM113 and pGP12 reduces translational efficiency without affecting the Shine-Dalgarno sequence or the initiation codon (16). The initiation codon is altered to AUG and the entire 5Ј-untranslated leader is deleted in pGM111 (Fig. 1b) and pGP11. These constructs all contain the entire rpsT coding sequence (residues 133-393) and 3Ј sequences extending to residue 412 or 413 (see also Fig. 1b).