Key Words RNAse B, RNAse III, PNPase, RNase II, poly(A) polymerase, mRNA stability, autoregulation
* Abstract The stability of mRNA in prokaryotes depends on multiple factors and it has not yet been possible to describe the process of mRNA degradation in terms of a unique pathway. However, important advances have been made in the past 10 years with the characterization of the cis-acting RNA elements and the trans-acting cellular proteins that control mRNA decay. The trans-acting proteins are mainly four nucleases, two endo- (RNase E and RNase III) and two exonucleases (PNPase and RNase II), and poly(A) polymerase. RNase E and PNPase are found in a multienzyme complex called the degradosome. In addition to the host nucleases, phage T4 encodes a specific endonuclease called RegB. The cis-acting elements that protect mRNA from degradation are stable stem-loops at the 5' end of the transcript and terminators or REP sequences at their 3' end. The rate-limiting step in mRNA decay is usually an initial endonucleolytic cleavage that often occurs at the 5' extremity. This initial step is followed by directional 3' to 5' degradation by the two exonucleases. Several examples, reviewed here, indicate that mRNA degradation is an important step at which gene expression can be controlled. This regulation can be either global, as in the case of growth rate-dependent control, or specific, in response to changes in the environmental conditions.
In addition to the efficiency of transcription and/or translation, the level of expression of a gene can be affected by the stability of its messenger RNA (mRNA). Thus cells can rapidly adapt their patterns of protein synthesis in response to changing environmental conditions. Although prokaryotic rnRNAs are generally unstable relative to those of eukaryotes, their decay rates can differ considerably within a single cell. In Escherichia coli, mRNA half-life can range from a fraction of a minute to half an hour (for reviews, see 14, 41, 139).
Despite extensive study of the mechanisms of transcription and translation, the factors controlling mRNA stability are not well understood. For instance, it is still not clear what governs the disparate rate at which mRNAs are degraded, nor have studies on the degradation of many individual mRNAs yielded a universal pathway for the decay process. There are many reasons for this situation. Many mRNAs are subject to alternative decay processes, and the main nucleases involved in mRNA degradation have only a loose primary sequence or secondary structure specificity. The stability of mRNA also depends on growth conditions, environmental signals, and the efficiency of translation. It is difficult to assess the contribution of each of these elements to mRNA chemical stability, which, anyway, is not necessarily equivalent to functional stability . Nevertheless, important advances have been made over the past ten years in deciphering the pathways of mRNA degradation in E. coli; specifically, the cis-acting RNA elements that control decay rates and the trans-acting cellular proteins that carry out the degradation process have been characterized.
The decay of many mRNAs is initiated by a primary endonucleolytic cleavage, often by RNase E or, less commonly, by RNase III [6,51]. This cleavage is followed by exonucleolytic degradation at the new 3' ends. Two enzymes, polynucleotide phosphorylase (PNPase) and RNase II, are involved in this process. In general, net degradation occurs with a 5' to 3' polarity; at first sight, this is surprising since no exonuclease degrading RNAs in the 5' to 3' direction has been found to date in prokaryotes. This 5' to 3' polarity may be due to RNase E specificity, which apparently requires a free 5' end to perform the first downstream endonucleolytic cut. This first cleavage liberates an upstream fragment that is exonucleolitically degraded and a downstream fragment with a new 5' end that is recognized by RNase E and further cleaved. This processive endonucleolytic cleavage followed by a 3' to 5' exonucleolytic degradation process explains the net 5' to 3' polarity. The observation that RNA secondary structures located in the 5' region of a message can stabilize diverse downstream RNA sequences concurs with this hypothesis. However, a particular 5' stabilizing structure does not always confer the same increase in stability to the heterologous RNA to which it is fused, which indicates that the 5' structure is not the only determinant of stability. It has also been shown that 3' stem and loop structures act as barriers to the exonucleases. Thus, a message with a 5' stabilized region and a 3' stem-loop should be substantially protected from cellular decay enzymes.
Another notable advance has been the discovery of a multienzyme complex, called the degradosome, that contains, in addition to the endonuclease RNase E, the exonuclease PNPase, a glycolytic enzyme, and different enzymes utilizing ATP as cofactor. If the degradosome, whose existence has only been biochemically proven, is involved in mRNA degradation in vivo, mRNA decay may be linked to intermediary metabolism.
Finally, the evidence that most, if not all, prokaryotic mRNAs are polyadenylated at their 3' extremity prompted an investigation of the role of these poly(A) stretches. These seem to destabilize mRNAs, in contrast to what happens in eukaryotes.
In this review, we discuss first the different nucleases involved in mRNA degradation and the regulation of their expression, followed by the role of mRNA degradation and/or maturation in controlling cellular and phage gene expression.
RNases INVOLVED IN mRNA DEGRADATION
We discuss only those enzymes whose role in mRNA degradation is well documented. However, some observations indicate that RNase P, which is normally responsible for the maturation of the 5' end of tRNAs, might be involved in the degradation of some mRNAs (3). Other unknown enzymes may also participate in mRNA decay [9, 19, 36, 67, 191].
The rne Gene An important advance in understanding the mechanism of E. coli mRNA degradation was the identification of RNase E as a key enzyme controlling mRNA decay. RNase E, encoded by the rne gene, was originally discovered in E. coli as a ribosomal RNA processing enzyme that converts a precursor 9S to immature 5S rRNA [8, 127]. Subsequently, RNase E was shown, both genetically and biochemically, to be necessary for the decay of bacteriophage T4 and specific bacterial mRNAs [16, 107, 132, 134, 161]. It also cleaves RNA I, the antisense regulator of replication of ColE1-type plasmids [43, 98, 183]. At the same time that Apirion and colleagues identified the E. coli me gene , a thermosensitive mutant with impaired ability to degrade bulk mRNA at nonpermissive temperature was discovered and designated ams, for altered mRNA stability [96, 145]. Later, the rne and ams mutations were found to be allelic and to map at 24.6 min on the linkage map of E. coli [38, 132]. The connection of mRNA decay to a ribosomal RNA processing gene [10, 123, 132, 181] paved the way for molecular biological studies on mRNA degradation. The rne gene appears to be essential based on the temperature-sensitive phenotype of several mutated alleles [8, 145]. The ams/rne gene was recloned and sequenced as hmpl, a gene encoding a protein involved in cell wall invagination during division . The ams/rne/hmpl gene appears to be monocistronic and specifies a transcript of 3.6kb . The promoter is located 361 nucleotides upstream of the AUG initiation codon . The stop codon is closely followed by an inverted repeat and 6 Us, typical of a Rho-independent terminator .
The RNase E Protein RNase E seems to be a homodimer [110, 185] of 1061 amino acid subunits. The calculated molecular mass (118 kDa) differs significantly from the 180-kDA mass estimated from SDS polyacrylamide gels . The abnormal mobility is thought to be due to a proline-rich central region . The RNase E protein can be separated into three domains, the amino terminal domain (amino acids 1-528), the central domain (amino acids 500-752), and the carboxy terminal domain (amino acids 734-1061) [43, 185]. The RNA-binding domain (arginine-rich) is located between the amino acids 580 and 700, i.e. in the central domain. The N-terminal domain contains an S1 motif showing homology with each of the six stretches in the middle and C-terminal part of ribosomal protein S1 [27, 46]. The catalytic function is in the N-terminal domain and is ascribed to residues 1 to 498 (119). Therefore, the sites of RNA binding and catalytic activity do not overlap. There is some sequence homology between RNase E and myosin, both of which can be identified by antibodies raised against the other .
RNase E has only been isolated from E. coli but a homolog of the E. coli me gene has been found in Haemophilus influenzae . Proteins cross-reacting with antibodies to E. coli RNase E and some rne homologs have been characterized in Rhodobacter capsulatus, Streptomyces levidans  and halophilic archae . The N-terminal half of RNase E is conserved in Synechocystis sp. . RNase E-like activity has been suggested in B. subtilis , although no obvious RNase E homolog has been identified in the entire chromosome sequence (95). A homolog of Rnase E, CafA protein (renamed Rnase G) has been characterized in E. coli. Rnase G and Rnase E participate jointly in the sequential maturation of the 5' end of 16S rRNA [97a].
Cleavage Specificity The cleavage specificity of RNase E is not very strict. A consensus pentanucleotide (A/G)AUU(A/U) was proposed to determine susceptibility to RNase E cleavage . However, in studying S20 mRNA cleavage specificity, Mackie concluded that RNase E has no sequence constraints other than cleaving substrates 5' to AU dinucleotides in single-stranded segments [107, 109]. The cleavage specificity of RNase E was further investigated by introducing random mutations in the decanucleotide region at the 5' end of RNA I and studying their effect on the position of the cleavage [99, 120]. The primary sequence of RNA I was found to affect both the position and number of RNase E cleavages occurring in the mutated region. The site of cleavage bears no simple relationship to a particular nucleotide order. These results are not consistent with either the notion that RNase E cleavages are determined by a simple consensus sequence or the contrary view that RNase E is a virtually nonspecific single-stranded e ndonuclease with a preference cutting 5' to an AU dinucleotide. Oligonucleotides with 10-13 residues devoid of secondary structure and containing the RNase E cleavage sequence of RNA I are cut at the same site but 20 times more efficiently than RNA I. Secondary structure probably limits cleavage at potentially susceptible sites. The fact that RNase E cuts the 10-13 oligomers so efficiently provides a promising tool to study the sequence specificity of the enzyme.
Recent results  show that RNase E has inherent vectorial properties, with its activity dependent on the free 5' end of its substrates. To demonstrate the need for the 5' end, circular RNA (derivative of the mRNA encoding ribosomal protein S20) was prepared. This derivative is considerably more resistant to cleavage by purified RNase E or RNase E in the degradosome in vitro (see below), than are linear molecules with a 5' monophosphate end. The circular RNA, after linearization, was fully susceptible to digestion. RNA does not need to be covalently closed as a circle to acquire resistance to RNase E. Antisense oligo-deoxynucleotides complementary to the 5' end of linear 5' monophosphate substrates significantly reduce the latter's susceptibility to attack by RNase E. Mackie's results show that RNase E must recognize both an internal cleavage site and a 5' terminal unpaired nucleoside monophosphate residue. Natural substrates with terminal 5' triphosphate groups are also poorly cleaved by RNase E in vitro . 5' monophosphorylated substrates are strongly preferred. Assuming that the protection due to 5' triphosphates and a paired 5' end are additive, these results may explain the stabilization of natural mRNAs by 5' stem-loop structures in vivo. The length of the stem is important but not the sequence of such stem-loops. The distance of such a structure from the 5' end of the message is crucial. The presence of five unpaired nucleotides upstream of the hairpin can negate the contribution of the hairpin to the stability of mRNA [9,21, 34, 53, 54,71, 110]. Although the stabilization by stem-loops at the 5' end of mRNAs has been ascribed to reduced RNase E cleavage, endonucleases other than RNase E may also be inhibited by these structures .
RNase E, in some cases, can show an exonucleolytic activity. RNase E can shorten 3' poly(A) and poly(U) tails artificially added to RNA I , leaving 6 adenylates for poly(A) or 1 uridylate for poly(U). As for the endonucleolytic activity, the exonucleolytic activity is associated to the amino terminal domain of RNase E. The addition of a 3' terminal phosphate group prevents both removal of the poly(A) tail and endonucleolytic cleavage. However, a 3' terminal phosphate has no effect on the endonucleolytic cleavage of transcripts devoid of poly(A) or poly(U) tails.
Negative Feedback Overexpression of RNase E from high-copy-number plasmids carrying the me gene is deleterious to E. coil cells and can lead to plasmid loss and to the acquisition of me mutations . E. coli has devised a strategy for tightly regulating the production of RNase E. The enzyme represses its own synthesis by increasing the decay rate of its mRNA [80, 133]. This autogenous control is directly based on the catalytic activity of RNase E. The half-life of the me transcript varies by more than a factor of 14, from less than 40 s in strains carrying a multicopy plasmid bearing the me gene to over 8 mm in an ams1 mutant strain. The steady-state level of full-length me mRNA increases fivefold in amsl cells, compared with wild-type cells, and falls by almost a factor of 7 in cells with a multicopy plasmid carrying me. The cis-acting domain of the me transcript responsible for the feedback was identified by deletion mapping. It is located within a fragment encompassing the 5' untranslated region and the first 28 codons of the me structural gene. This region is sufficient to confer RNase E control on a heterologous transcript to which it is fused . The region of the RNase E protein that causes repression of me gene expression was also identified by deletion analysis. Deletions that remove the carboxyterminal 247, 408, or 563 amino acid residues of RNase E only partially impair its ability to feedback or to regulate. The repression ratio is reduced from 30-fold (with full-length RNase E), to 12-fold with [delta]247 and [delta]408, and to sixfold with [delta]563. All three mutants permit growth of the amsI strain at 42[degrees]C. In contrast, deletion of the C-terminal 591 or 601 residues, or of the N-terminal 29-314 residues, abolishes control and eliminates the ability of RNase E to restore viability of an amsI strain at nonpermissive temperature. This result shows that the two phenomena, feedback regulation and cell viability, are correlated .
RNase III was first detected in E. coli as an endoribonuclease that cleaves double-stranded RNA molecules. The cleavage can be either a single-stranded nick or double-stranded break in the RNA, depending in part upon the degree of base-pairing in the region of the cleavage site . Although RNase III can cleave double-stranded homopolymers, not all extended double-stranded regions are targets for RNase III in natural RNAs. Specificity is …