RNase E Cleavage in the atpE Leader Region of atpE/lnterferon-@ Hybrid Transcripts in Escherichia coli Causes Enhanced Rates of mRNA Decay*

Chimeric transcripts containing the ribosome bind- ing site of the Escherichia coli atpE gene and variants of the human structural interferon-@ gene are subject to RNase E processing in the 5”untranslated atpE part of the transcripts. The absence of processing at two sites in the atpE leader-sequence caused by the RNase E deficiency in E. coli host N3431 leads to a considerable stabilization of the mRNA moiety. RNase E has originally been described as a processing enzyme for non-mRNAs such as precursor 5 S rRNA and RNA1, but cleavage mRNA substrates have also been reported. RNase E processing of the atpE gene leader sequence-containing transcripts leads to an increased rate of mRNA breakdown. The two RNase E-depend- ent processing sites in the atpE part of the mRNA transcripts exhibit some similarity to the other known RNase E processing sites. The influence of RNase E cleavage upon post-transcriptional regulation such as RNA stability and the efficiency of translational initiation is discussed.

In addition to transcriptional and translational efficiency protein synthesis is controlled by mRNA stability. In this respect mRNA decay rates may vary by as much as 50 among transcripts in a single bacterial cell (Pedersen et al., 1978;Nilsson et al., 1984). In Escherichia coli mRNA stability seems controlled by RNA sequence/structure relationships as well at the 5', the 3'end, and in addition, also in the interior of (polycistronic) transcripts (Schmeissner et al., 1984;Gorski et al., 1985;Hayashi and Hayashi, 1985;Panayotatos and Truong, 1985;Wong and Chang, 1986;Cannistraro et al., 1986;Newbury et al., 1987;Portier et al., 1987;Cho andYanofski, 1988 Chen et al., 1988). Only few of the known RNases seem to be involved in mRNA decay which include 3'exonucleases RNase I1 and polynucleotide phosphorylase (Donovan and Kushner, 1986) and endoribonuclease RNase I11 (Dunn and Studier, 1973;Schmeissner et al., 1984;Portier et al., 1987). A conditional lethal mutation in the urns gene exhibits a considerable stabilization of the bulk mRNA in E. coli at the nonpermissive temperature (Ono and Kuwano, 1979). The investigation of rate-limiting steps in mRNA decay has led to the characterization of several enzymes involved in the regulation of mRNA degradation (e.g., Nilsson et al., 1988;Cannistraro and Kennell, 1989). In this respect it also has been demonstrated that RNase E, in addition to the process-* The costs of publication of this article were defrayed in part by t,he payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ing of non-mRNA species like RNAl (Tomcsany and Apirion, 1985) and 5 S rRNA (Apirion, 1978;Ghora and Apirion, 1978), also acts upon T4 gene 32 mRNA thereby converting it into a smaller and considerably more stable mRNA species (Mudd et al., 1988). Recently it has been demonstrated that an E. coli RNase E-deficient background substantially stabilizes bacteriophage T4 mRNAs implying a general role for this endonuclease in mRNA turnover (Mudd et al., 1990).
In a previous study we investigated RNA sequence/structure influences in the translational initiation region of the E. coli atpE gene fused to variants of the human 1FN'-fi gene upon efficiency of translational initiation and mRNA stability (Gross et al., 1990). We observed a correlation between translational efficiency and mRNA stability. In the present investigation we examined the nature of two major endonucleolytic processing sites in the atpE leader sequence upon the decay rates of the hybrid transcripts. Processing in the leader sequence is mediated by endonuclease RNase E cleaving twice the translational initiation region from the E. coli atpE gene.
RNase E processing results in enhanced mRNA decay rates of the hybrid transcripts providing further evidence for a role of RNase E in mRNA degradation in E. coli.

EXPERIMENTAL PROCEDURES
Vector Constructions-The expression vector used in this study is pILA-501, which belongs to the family of pILA vectors (Schauder et al., 1987). Bacteriophage X promoters p H and p,. are regulated by the temperature-sensitive cIts857-coded repressor. Induction is initiated by a temperature shift from 30 to 42 "C. E. coli cells harboring these expression vectors are induced a t ODs,5o = 0.5. The expression vector and the atpE/IFN-P gene hybrid transcripts are as detailed in Fig. 1. Construction and features of atp-IFN,, a variant with the primary nucleotide sequence of the natural human IFN-8 gene fused to the untranslated atpE leader sequence as well as atp-synIFN8 and atp-synIFN9 variants with the nucleotide sequence of the human IFN-P gene adapted to the optimal E. coli codon usage are as described by Gross et al. (1990). The hybrid transcripts differ in translational initiation rates leading to contrasting IFN-P synthesis in E. coli: atp-synIFN, represents a variant of the synthetic IFN-/3 gene with a nucleotide composition in the translational initiation region leading to low level of ribosome binding and a concomitant level of protein synthesis yielding < 1% IFN-8 of total E. coli protein. In contrast, variants atp-IFN, and atp-synIFN, exhibit high rates of translational initiation and IFN-8 synthesis in E. coli of -18% and 30% of total cellular E. coli protein, respectively.
RNA Kinetics-E. coli cells harboring the expression vectors were grown up to OD,,o = 0.5 and induced a t 42 "C for 30 min. Rifampicin (150 pg/ml) was added. Cells were taken at the indicated time intervals, quickly pelleted, and frozen at -80 "C in a dry ice bath. RNA was isolated by the guanidinium-hot phenol method as described by Sambrook et al. (1989). Total E. coli RNA (5 pg) was separated ' The abbreviations used are: IFN, interferon; p,, and p~, early leftward and rightward bacteriophage X major promoters, respectively. The untranslated leader region from the atpE gene was fused to the structural human ZFN-p gene as described (Gross et al., 1990). The variant of the natural human ZFN-fl gene sequence atp-ZFN1 was introduced into expression vector pILA-501 (Schauder et al., 1987) and exhibits high rates of translational initiation. The variants of the synthetic human ZFN-p gene with E. coli codon usage either have reduced or high rates of translational initiation (atp-synZFNs or atp-synZFN9, respectively). Transcription gives rise to two transcripts from 1,500 and 1,200 nucleotides originating from bacteriophage X promoterspR andp,. The nucleotide sequence of the translational initiation region is indicated. Major processing sites in this region are denoted by arrows. Bold arrows mark efficient cleavage by endonucleases. The RNase E-dependent cleavage sites at positions 153 and 171 are presented. The position of the initiating 30 S ribosomal subunit is indicated as determined in Gross et al. (1990). Truncated mRNA should only be an ineffective substrate for efficient translational initiation reducing translational efficiency and thereby enhancing the rate of mRNA decay. electrophoretically in a 2.2 M formaldehyde, 1.5% agarose gel and transferred to nitrocellulose. Hybridization was carried out with a nick-translated 32P-labeled ZFN-fl gene-specific DNA probe. The halflife of mRNA decay was calculated on the basis of the 32P label present in the intact mRNA band cut from the Northern blot and determined in an scintillation counter. The level of steady-state IFN-/3 mRNA was determined in the same way by the radioactivity present in the intact IFN-@ mRNA band 30 min after induction of transcription.
Primer Extension Analysis-Total cellular E. coli RNA (10 pg) or 2 pg of RNA synthesized in uitro by E. coli RNA polymerase from the expression vector (as described by Stuber et al., 1984) was subjected to primer extension analysis by reverse transcription of the human IFN-8 mRNA initiated with Moloney murine leukemia virus reverse transcriptase (200 units) at position +91 downstream of the AUG initiation codon with the 32P-labeled oligonucleotide 5' TCATCCTGTCCTTGAGGC. Primer extension analysis was carried out as described by Hartz et al. (1988). Reverse transcription with mRNAs of the synthetic ZFN-@ gene variant and its derivatives was initiated from position +67 with the oligonucleotide 5' GAC-GACCGTTCAGCT. Elongation products were separated on a 6% acrylamide gel containing 7.5 M urea.

RESULTS
mRNA Stability Is Affected by RNase E Cleavage-In a preceding study we investigated sequence and conformational changes in the translational initiation region upon the efficiency of translational initiation of two ZFN-p genes with identical amino acid but differing nucleic acid sequence. One variant represents the natural human IFN-8 gene, and the other one was assembled chemically on the basis of the optimal E. coli codon usage. The latter shows a 76% primary sequence identity to the natural human ZFN-8 gene. Mutagenesis in the translational initiation region of the natural as well as the synthetic ZFN-6 gene was used to delineate the requirements for optimal translational initiation. Single nucleotide exchanges in this region which altered the mRNA configuration resulted in dramatic changes of IFN-p protein synthesis, rates of translational initiation, and mRNA decay. mRNA stability was correlated with translational efficiency (Gross et al., 1990). Two major endonucleolytic cleavage sites in the untranslated atpE part of the transcripts were characterized. Deletions in this particular region of the atpE leader sequence impair the efficiency of the ribosome binding site (McCarthy et al., 1985). Therefore, it is likely that processing in the ribosome binding site ( Fig. 1; positions 153 and 171) may interfere with efficient ribosome loading and in addition, could be responsible for the observed correlation of translational rates and the rate of mRNA degradation. If this reasoning is correct, the observed processing could constitute rate-limiting steps for mRNA decay in this system. A role of ribosome binding upon mRNA decay rates can also be studied with variants exhibiting high or low rates of translational initiation in RNase-deficient E. coli strains which affect these cleavages.
In this respect we analyzed several E. coli strains with defined RNase deficiencies. E. coli strains with genetic deficiencies in RNases including RNases D, E, I, 11, I11 and polynucleotide phosphorylase were used in this study. Only E. coli strain N3431, characterized by a conditional temperature-sensitive RNase E deficiency, extended the half-life of mRNA breakdown considerably ( Fig. 2 and Table I). The experimental system is detailed in Fig. 1. The chimeric IFNp genes are under the transcriptional control of the two bacteriophage X p~ and pL promoters which give rise to two transcripts from -1,500 and -1,200 nucleotides, respectively. E. coli N3431 carries a temperature-sensitive mutation inactivating RNase E at 42 "C (Goldblum and Apirion, 1981) which, in this system, is also the temperature of transcriptional induction. The isogenic revertant N3433 shows in RNase E Cleavage in the atpE Leader Region general the rate of mRNA breakdown as observed for E. coli strain DH1 which has routinely been used in the expression studies ( Fig. 2 and Table I) mRNA half-life of a natural IFN-/? gene variant directing a high rate of translational initiation (atp-IFN1) and two synthetic IFN-p gene variants exhibiting low or high rates of translational initiation (atp-synIFNu and atp-synIFN9, respectively) (Fig. 2 and Table I). The half-life of atp-IFN1 mRNAs transcribed from X promoters pl, as well as pH resulting in the two transcripts of 1,500 and 1,200 nucleotides are stabilized by a factor of 3.5 or 4, respectively, in E. coli mestrain N3431 at the nonpermissive temperature ( Fig. 2 and Table I). The isogenic revertant E. coli N3433 exhibits a t 42 "C comparable half-lives with E. coli strain DH1 (Fig. 2 and Tab. I). Similarly, the mRNAs half-lives containing the synthetic IFN-p gene are extended in RNase E-deficient E. coli strain N3431 by a factor of 2.9 in case of the p~ transcript and a factor of 3.7 for the p L transcript of the atp-synIFN9 variant. In the m estrain these values for atp-synIFNu are -20 and -16, respectively, because of the short mRNA halflife in me+ E. coli strains (< 0.5 min; Table I and Fig. 2).

RNase E Cleavage Takes Place in the Ribosome Binding
Site from the E. coli atpE Gene-The major endonucleolytic processing sites in the translational initiation region of atp-IFN1 are shown in Fig. 3. Minor cuts take place a t positions 138 and 204 and are not caused by RNase E. Another strong signal a t position 126 is probably caused by a RNA configuration interfering with the action of reverse transcriptase since this signal could also be detected with in vitro transcribed mRNA (Fig. 3). The cleavage sites at positions 153 and 171 are recognized in me+ E. coli strains DH1 and N3433. They remain uncleaved in E. coli N3431 at the nonpermissive temperature of 42 "C. Its isogenic revertant E. coli N3433 exhibits these cleavage sites and should therefore be product of RNase E processing.
In addition, the entire atp-IFN1 mRNA moiety was screened for additional RNase E-sensitive sites. They were not observed in either atp-IFN1 or in the synthetic IFN-fi gene variants atp-synIFNu., (data not shown). Therefore, RNase E cleavage at positions 153 and 171 should ultimately be responsible for the differential stability observed for hybrid mRNAs containing the untranslated atpE leader sequence in meand me+ E. coli hosts, respectively.  (Gross et al., 1990) have heen used to investigate mRNA level and half-life as detailed under "Experimental Procedures" and represent the mean values in a t least two independent experiments. IFN-p gene transcripts originate from the two bacteriophage A promoters p/t and p/:E. coli RNase E-deficient strain N3441 (Tomcsiny and Apirion, 1985) exhibits a reduced rate of mRNA decay in comparison with its isogenic revertant N3433 and the E. coli strain DHI. The relative steady-state IFN-8 mRNA level in E. coli has heen determined as described under "Experimental Procedures." The values are the means of duplicates from the same sample and were correlated by hybridization standards for the natural and the synthetic IFN-8 gene. Similar relationships were found in other experiments. The relative steady-state level is given as percent relative to the Ap!, 1,500.nucleotide transcript of the afp-synIFNsl variant which exhibits the highest steady-state mRNA level in E. coli, 100 (*).  annealed with an internal primer, and primer-mediated reverse transcription was initiated 91 nucleotides downstream from the AUG initiation codon. Elongation products were separated on 6% ureapolyacrylamide gel. In the RNase E-deficient E. coli strain N3431 the processing sites a t positions 153 and 171 are missing in comparison with E. coli N3433, which is an isogenic revertant of N3431 as well as in E. coli strain DH1. c, control reverse transcription with mRNA was synthesized in uitro with E. coli RNA polymerase as described under "Experimental Procedures." 2 pg of RNA was used for primermediated reverse transcription by Moloney murine leukemia virus reverse transcriptase. Extension inhibition a t position 126 takes place also with in uitro synthesized RNA and seems to be an effect of RNA configuration interfering with DNA synthesis. A, C, G, and T are specific sequencing reactions with expression vector pILA-501 harboring the atp-IFN1 variant and Moloney murine leukemia virus reverse transcriptase mediated from the same internal primer.

DISCUSSION
This paper demonstrates that processing of atpE leader containing mRNA is dependent on the action of RNase E. An RNase E recognition sequence motif ACAGt'AUUUG has been postulated by Tomcsany and Apirion (1985), which together with other RNase E processing sites (Mudd et al., 1988) and the two in the atpE leader sequence reported here seems more as a U-rich sequence flanked by purines: P u U1-4 Pu. Such a degenerate sequence cannot alone define the obviously high cleavage specificity observed for this endoribonuclease. So it seems highly likely that in addition to sequence information RNA configuration confers additive information for processing. However, the region in the atpE leader sequence in which processing takes place is relatively unstructured and does not exhibit a significant stability (-0.7 kcal/mol). Other RNase E processing sites also do not readily denote the secondary structure requirements for efficient RNase E processing (Tomcsiny and Apirion, 1985;Mudd et al., 1988). RNase E cleavage in the upstream region of the E. coli atpE gene results in a significant destabilization of the processed mRNA moiety. One mechanistic model of mRNA decay invokes endonucleolytic cleavages in mRNA as entry sites for the 3' + 5' exonucleases which degrade the resulting mRNA fragments in 3' + 5' direction (reviewed by Belasco and Higgins, 1988). Indeed, mRNA stability of the hybrid transcripts is influenced by 3'-terminal hairpin structures as exhibited e.g. by different terminators of transcription apparently by the capability of the 3' + 5' exonucleases to overcome this RNA configuration (Gross and Hollatz, 1988). In addition, the RNA located 5' upstream of the RNase E cleavage sites could not be detected by Northern analyses and hardly by primer extension analyses, suggesting a high turnover rate for this particular region (data not shown). This should because RNase E processing creates new 3' ends which serve as entry sites for the 3' + 5' exonucleases. However, another model should be invoked explaining the enhanced decay rates for the RNA located downstream of the processing sites. Ribosome loading in the translational initiation region could interfere with the interaction of endonuclease RNase E with its substrate. A high rate of translational initiation would therefore result in an extended mRN.4 half-life and in additional rounds of translation. Interference of the initiating ribosomal 30 S subunits with the action of RNase E could explain that RNase E processing is more efficient a t position 153 than at position 176, the latter being in closer proximity to the essential features of translational initiation: the Shine-Dalgarno region, the initiator codon AUG and, therefore, to the initiating 30 S ribosomal subunit, although the differential cleavage efficiency could also be the result of RNA sequence/ structure relationships at the two cleavage sites. Once RNase E cleavage in the translational initiation region has occurred, ribosome loading could be seriously flawed, leading to a ribosome-depleted mRNA stretch which should exhibit a higher sensitivity toward further endonucleolytic cleavages and an unidirectional net wave of 5' + 3' mRNA degradation as described by Cannistraro et al. (1986) and Belasco and Higgins (1988). However, another fact adds some complexity to this model. The absence of processing in E. coli N3431 does only marginally increase the steady-state level of atp-synZFNn mRNA in the E. coli cell. The latter variant exhibits a low rate of translational initiation and a short mRNA half-life of < 0.5 min in me+ E. coli strains. Although its mRNA is substantially stabilized in an RNase E-deficient strain and about equals the half-life of variants with high rates of translational initiation such as atp-IFN, or atp-synlFN9 (Table I) this does only result in an increase of the mRNA level by a factor of -3-4, which is 1-2 orders of magnitude below the steadystate mRNA level observed with variants exhibiting high rates of translational initiation. This could be caused by the transcriptional polarity exerted by E. coli RNA polymerase, which has been demonstrated to take place in a number of cases in which the rate of translational initiation is low (Stanssens et al., 1986;Folley and Yarus, 1989). Otherwise, it seems likely that the initiating 30 S ribosomal subunit rearranges the mRNA configuration (perhaps by local unwinding) so that RNase E is able to recognize and process the transcripts only after interaction with the initiating ribosome. Ribosome binding, i.e. efficient translational initiation, could therefore exert several effects, enabling the access for RNase E processing by local RNA rearrangement and then protection of mRNA by interfering with mRNA/RNase E interaction. Variants exhibiting a reduced capacity for translational initiation have only a minor fraction of the mRNA population accessible for structural rearrangement induced by 30 S ribosomal subunit binding and, eventually, for RNase E processing. Only this fraction should then be stabilized in an E. coli RNase Edeficient strain while the translationally inactive mRNA pool is degraded quickly and efficiently by a mode different from the RNase E-dependent way. In this respect the studies of Chevrier-Miller et al. (1990) should be emphasized; they showed that transcriptional and translational uncoupling leads to differential mRNA half-lives of lac2 mRNA. Their model, according to which the instability of poorly translated message results from inefficient ribosome loading which unwinds RNA but does not protect it rather than from a complete absence of translation, is consistent with these observations. In addition, it has been observed that certain RNAs that have been isolated from an RNase E-deficient E. coli strain are not processed in vitro by partially purified RNase E (Pragai and Apirion, 1982;Gurevitz et al., 1983), a fact that could be caused by the absence of protein factor(s) rearranging the local RNA configuration in E. coli as conceived above, although the authors favor a model of a processing enzyme complex in which RNase E is an integral component and in which a RNase E mutation obstructs the activity of additional nucleases.
If RNase E cleavage exerts rate-limiting steps for the stability of actively translated mRNA species this also means that untranslated mRNA is degraded by a mode and by factors that are different and independent from RNase E and which remain to be elucidated.