stabilization of translationally active mrna by prokaryotic rep

14
Cell, Vol. 48, 297-310, January 30, 1967, Copyright 0 1967 by Cell Press Stabilization of Translationally Active mRNA by Prokaryotic REP Sequences Sarah F. Newbury: Noel H. Smith:+ E. Clare Robinson, Ian D. Hiles, and Christopher F. Higgins Molecular Genetics Laboratory Department of Biochemistry University of Dundee Dundee DDl 4HN Scotland Summary The REP sequence is a highly conserved inverted re- peat that is present in about 25% of all E. coli tran- scription units. We show that the REP sequence can stabilize upstream RNA, Independently of any other sequences, by protection from r-5’ exonuclease at- tack. The REP sequence is frequently responsible for the differential stability of different segments of mRNA within an operon. We demonstrate that REP- stabilized mRNA can be translated in vivo and that cloning the REP sequence downstream of a gene can increase protein synthesis. This provides direct evi- dence that alterations in mRNA stablllty can play a role in determining bacterial gene expression. The impli- cations of these findings for the mechanisms of mRNA degradation and for the role of RNA stability in the regulation of gene expression are discussed. Introduction Between 500 and 1000 copies of the repetitive extragenic palindromic (REP) sequence are present on the E. coli chromosome. This sequence is found in about 25% of all transcription units and may occupy as much as 1% of the total genome (Higgins et al., 1982a; Stern et al., 1984). The main features of the REP sequence have been de- scribed previously (Stern et al., 1984; Gilson et al., 1984). Briefly, the REP sequence is a highly conserved inverted repeat with potential for forming stable stem-loop struc- tures in mRNA. There is no obvious correlation between the presence of the REP sequence in an operon and the gene product(s) of that operon: REP sequences have been identified in transcription units encoding biosyn- thetic, degradative, structural, and regulatory proteins. In operons containing REP sequences for which the extent of transcription has been accurately determined, the REP sequences are invariably transcribed and are located ei- ther in intergenic regions of multicistronic operons or in the 3’untranslated region upstream of the terminator. The observation that REP sequences are frequently located in intercistronic regions led to the suggestion that they might serve an attenuator-like role, regulating the relative ex- pression of genes within operons (Higgins et al., 1982a; *The order of these authors is arbitrary. t Present address: Department of Biology, University of Rochester, Rochester, New York 14627. Valentin-Hansen et al., 1984). However, the subsequent identification of REP sequences at the 3’end of transcrip- tion units argues against such a role, and we have since demonstrated that at least one example of the REP se- quence does not function as a transcription terminator and has little or no effect on the expression of downstream genes (Stern et al., 1984). Thus, a specific role for the REP sequence in the regulation of gene expression is not im- mediately apparent. Other nonregulatory roles for the REP sequence have been proposed, such as the media- tion of chromosomal rearrangements or in the organiza- tion of the chromosomal DNA (Higgins et al., 1982a; Gil- son et al., 1984; Stern et al., 1984). However, no direct evidence for such functions has yet been obtained. The observation that most, if not all, REP sequences are transcribed and the fact that they can potentially form stable stem-loop structures in mRNA implies that any function is likely to be in RNA rather than DNA. This view is strengthened by the finding that the potential to form secondary structures is highly conserved between REP sequences (discussed in Stern et al., 1984; Higgins and Smith, 1986). For example, there are certain bases in the consensus REP sequence that can be either one of two different nucleotides. Significantly, whichever nucleotide is present in any given copy of the REP sequence, the cor- responding nucleotide in the complementary arm of the inverted repeat is that required to maintain base-pairing potential. Similarly, most examples of the REP sequence that deviate from the consensus sequence involve two compensatory base changes which maintain base- pairing potential. In this paper we show that the REP se- quence does indeed play a role at the RNA level, serving to stabilize upstream mRNA by protecting it from exo- nucleolytic attack. The factors determining the stability of bacterial mRNA are rather poorly understood. This is principally because, in prokaryotic cells, transcription and translation are cou- pled. The presence of ribosomes on an RNA molecule not only influences its stability, but also make it difficult to sep- arate experimentally the processes of RNA synthesis, translation, and degradation. In addition, the half-lives of most bacterial mRNA molecules are very short, in the range of 1 to 2 minutes. The chemical decay of mRNA to small oligonucleotides and mononucleotides is principally the result of exonu- clease, rather than endonuclease, activity. Several exo- nucleases have been characterized in E. coli, two of which, RNAase II and polynucleotide phosphorylase, are believed to be the principal degradative enzymes (Kaplan and Apirion, 1974; Har-El et al., 1979; Deutscher, 1985). These enzymes are both 3’5’exonucleases. No 5’ -3’exo- nuclease has yet been identified in E. coli. Many factors can influence the rate at which these enzymes degrade a given species of mRNA. mRNA secondary structure, in- cluding the stem-loops of rho-independent terminators, can impede the progress of exonucleases and therefore increase mRNA stability (Gupta et al., 1977; Mott et al.,

Upload: camilo-ernesto-araujo-barabas

Post on 23-Dec-2015

223 views

Category:

Documents


0 download

DESCRIPTION

RNA

TRANSCRIPT

Cell, Vol. 48, 297-310, January 30, 1967, Copyright 0 1967 by Cell Press

Stabilization of Translationally Active mRNA by Prokaryotic REP Sequences

Sarah F. Newbury: Noel H. Smith:+ E. Clare Robinson, Ian D. Hiles, and Christopher F. Higgins Molecular Genetics Laboratory Department of Biochemistry University of Dundee Dundee DDl 4HN Scotland

Summary

The REP sequence is a highly conserved inverted re- peat that is present in about 25% of all E. coli tran- scription units. We show that the REP sequence can stabilize upstream RNA, Independently of any other sequences, by protection from r-5’ exonuclease at- tack. The REP sequence is frequently responsible for the differential stability of different segments of mRNA within an operon. We demonstrate that REP- stabilized mRNA can be translated in vivo and that cloning the REP sequence downstream of a gene can increase protein synthesis. This provides direct evi- dence that alterations in mRNA stablllty can play a role in determining bacterial gene expression. The impli- cations of these findings for the mechanisms of mRNA degradation and for the role of RNA stability in the regulation of gene expression are discussed.

Introduction

Between 500 and 1000 copies of the repetitive extragenic palindromic (REP) sequence are present on the E. coli chromosome. This sequence is found in about 25% of all transcription units and may occupy as much as 1% of the total genome (Higgins et al., 1982a; Stern et al., 1984). The main features of the REP sequence have been de- scribed previously (Stern et al., 1984; Gilson et al., 1984). Briefly, the REP sequence is a highly conserved inverted repeat with potential for forming stable stem-loop struc- tures in mRNA. There is no obvious correlation between the presence of the REP sequence in an operon and the gene product(s) of that operon: REP sequences have been identified in transcription units encoding biosyn- thetic, degradative, structural, and regulatory proteins. In operons containing REP sequences for which the extent of transcription has been accurately determined, the REP sequences are invariably transcribed and are located ei- ther in intergenic regions of multicistronic operons or in the 3’untranslated region upstream of the terminator. The observation that REP sequences are frequently located in intercistronic regions led to the suggestion that they might serve an attenuator-like role, regulating the relative ex- pression of genes within operons (Higgins et al., 1982a;

*The order of these authors is arbitrary. t Present address: Department of Biology, University of Rochester, Rochester, New York 14627.

Valentin-Hansen et al., 1984). However, the subsequent identification of REP sequences at the 3’ end of transcrip- tion units argues against such a role, and we have since demonstrated that at least one example of the REP se- quence does not function as a transcription terminator and has little or no effect on the expression of downstream genes (Stern et al., 1984). Thus, a specific role for the REP sequence in the regulation of gene expression is not im- mediately apparent. Other nonregulatory roles for the REP sequence have been proposed, such as the media- tion of chromosomal rearrangements or in the organiza- tion of the chromosomal DNA (Higgins et al., 1982a; Gil- son et al., 1984; Stern et al., 1984). However, no direct evidence for such functions has yet been obtained.

The observation that most, if not all, REP sequences are transcribed and the fact that they can potentially form stable stem-loop structures in mRNA implies that any function is likely to be in RNA rather than DNA. This view is strengthened by the finding that the potential to form secondary structures is highly conserved between REP sequences (discussed in Stern et al., 1984; Higgins and Smith, 1986). For example, there are certain bases in the consensus REP sequence that can be either one of two different nucleotides. Significantly, whichever nucleotide is present in any given copy of the REP sequence, the cor- responding nucleotide in the complementary arm of the inverted repeat is that required to maintain base-pairing potential. Similarly, most examples of the REP sequence that deviate from the consensus sequence involve two compensatory base changes which maintain base- pairing potential. In this paper we show that the REP se- quence does indeed play a role at the RNA level, serving to stabilize upstream mRNA by protecting it from exo- nucleolytic attack.

The factors determining the stability of bacterial mRNA are rather poorly understood. This is principally because, in prokaryotic cells, transcription and translation are cou- pled. The presence of r ibosomes on an RNA molecule not only influences its stability, but also make it difficult to sep- arate experimentally the processes of RNA synthesis, translation, and degradation. In addition, the half-lives of most bacterial mRNA molecules are very short, in the range of 1 to 2 minutes.

The chemical decay of mRNA to small oligonucleotides and mononucleotides is principally the result of exonu- clease, rather than endonuclease, activity. Several exo- nucleases have been characterized in E. coli, two of which, RNAase II and polynucleotide phosphorylase, are believed to be the principal degradative enzymes (Kaplan and Apirion, 1974; Har-El et al., 1979; Deutscher, 1985). These enzymes are both 3’5’exonucleases. No 5’-3’exo- nuclease has yet been identified in E. coli. Many factors can influence the rate at which these enzymes degrade a given species of mRNA. mRNA secondary structure, in- cluding the stem-loops of rho-independent terminators, can impede the progress of exonucleases and therefore increase mRNA stability (Gupta et al., 1977; Mott et al.,

Cdl 299

HIM

IL. I I

Figure 1. mRNA Encoded by the Histidine Transport Operon of S. typhimurium

The Northern blot shows RNA isolated from strains TAW and TA3908. These two strains are isogenic except for a 141 bp deletion within the his&nisC! intergenic region of TASQOE. The construction and sequence of this deletion have been described previously (Stern et al.. 1994). Both strains also harbor the promoter-up mutation dhuA7, which ele- vates transcription of the operon several fold (Higgins and Ames, 1982). The blot was probed using the 1940 bp Kpnl fragment that in- cludes sequences upstream and downstream of the REP sequence. Using a strain deleted for the entire histidine transport operon, the sig- nal was shown to be specific for the his operon; no cross-hybridization with RNA from the ergTgene, which is 70% identical to the his./ gene (Higgins and Ames, 1991) was detected at the stringencies used (data not shown). The full-length (9910 nucleotides) and REP-stabilized (950 nucleotides) mRNA species are indicated. In addition to these tran- scripts, there is a “smear” of nascent transcripts, consisting of inter- mediates in synthesis and degradation, which is typical of many prokaryotic operons. This smear is enhanced at those points at which the extremely large amounts of rRNA also migrate; these rRNA- induced artifactual bands are present in most of the Northern blots shown in this paper and cannot readily be eliminated. It should be pointed out that, although in this particular gel the full-length mRNA seems to be enhanced slightly by the REP sequence, this is not a re- producible effect. In the diagram of the histidine transport operon, the direction and extent of transcription, from the promoter(P) to the termi- nator (T), are indicated. The DNA fragments used as hybridization probes in Northern blots are indicated and discussed further in the text. The characterization and complete nucleotide sequence of this operon are described elsewhere (Higgins et al., 1982b).

1985). The presence or absence of r ibosomes and the rate of translation can apparently affect the half-life of mRNA molecules (e.g., Gupta and Schlessinger, 1978; Schneider et al., 1978; Har-El et al., 1979; Graham et al., 1982) al- though the recent demonstration that a reduction in the

rate of translation can cause premature termination of transcription (Stanssens et al., 1988) complicates the in- terpretation of such data. Endonuclease cleavage may precede mRNA decay, exposing free 3’ ends, which are substrates for exonucleases. Such a mechanism has been demonstrated for phage lambda, in which cleavage of the major leftward transcript by RNAase III renders the upstream int mRNA more susceptible to degradation (Guarneros et al., 1982; Rosenberg and Schmeissner, 1982). Various chromosomally encoded RNA species are also substrates for RNAase Ill (Barry et al., 1980; Takata et al., 1985) or for other as yet unidentified enzymes (e.g., Burton et al., 1983; Reed and Altman, 1983). Polycistronic mRNA molecules may also be cleaved to monocistronic units by endonucleases prior to bulk degradation (Achord and Kennell, 1974; Schlessinger et al., 1977; Lim and Ken- nell, 1979). However, a defined role for endonucleolytic processing of chromosomally encoded mRNA has yet to be established.

One critical question that remains essentially unan- swered is whether or not bacterial gene expression can be controlled at the level of RNA degradation. To address this question, two aspects of mRNA degradation must be dis- tinguished: translational or functional inactivation, which renders an mRNA molecule unsuitable for further transla- tion, and chemical decay of mRNA to small oligonucleo- tides and mononucleotides. Translational inactivation has been separated experimentally from chemical decay (e.g., Yamamoto and Imamoto, 1975; Chanda et al., 1985) and probably involves specific endonucleolytic cleavage at the 5’ end of an mRNA molecule, possibly by RNAase III (Shen et al., 1981, 1982; Cannistraro and Kennell, 1985a). Theoretically, if mRNA can be stabilized in a translation- ally active form, such that the steady-state concentration of that mRNA species in the cell is increased, this will lead to an increase in gene expression. In this paper we dem- onstrate a role for the REP sequence in the stabilization of specific intermediates in mRNA degradation. At least in certain cases, the stabilized mRNA is translationally ac- tive and the stabilization of mRNA by the REP sequence can increase gene expression. These data have a number of general implications for the mechanisms of mRNA degradation and provide direct evidence that regulation of mRNA stability can play an important role in the regulation of gene expression.

Results

Rep Sequences Cause Accumulation of Upstream RNA The histidine transport operon of Salmonella typhimurium consists of four genes, hisJ, hi.@, hisM, and hisP (Higgins et al., 198213; Figure 1). Two copies of the REP sequence, in inverted orientation with respect to each other, occupy most of the hisJ-hid2 intergenic region. Figure 1 shows a Northern blot of RNA isolated from strain TAm and from its derivative, TA3808, which has most of the hisJ-hisQ intergenic region, including the REP sequences, deleted. This deletion has been sequenced and described previ- ously (Stern et al., 1984). The blot was probed using a

Regulation of Gene Expression by RNA Stabilization 299

A

8 Figure 2. mRNA Stabilization of Plasmid-

r;? Encoded RNA by the REP Sequence

5E 'PRE' 'POST' (A) A Northern blot of mANA encoded by

0 ‘I ,3 0123456 c, 0123456 pKG1800 and PWJOI is shown, probed with the 1582 bp BstEll-Pvul fragment from pWJ151. (6) A Northern blot of RNA encoded by pWJO1. RNA samples were isolated prior to rifampicin

4 treatment (0) and at 1 to 8 min after adding rifampicin as indicated. The blot was probed with the 408 bp BstEll-Bamlil fragment of

-3100- pWJ151, which contains only sequences up- stream of the REP sequence. (G) A Northern blot of the same RNA samples as in (8) except probed using the 958 bp BamHI-Pvul frag ment of pWJ151, containing only sequences

650- downstream of the REP sequence. In all blots,

-650- full-length transcripts, running from the pro- moter (Psd) to heterogeneous terminators (T), are 2900-3100 nucleotides long. Enhance- ments of the smear of nascent transcripts by rRNA can be seen; a 850 nucleotide transcript, specific to pWJO1, is also indicated. The plas- mids used are all shown diagrammatically as if linearized at the unique EcoRl site and have been described previously (Stern et al., 1984). pWJO1 is identical to pKG1800 except for a 218 bp insert between the promoter and the ga/K gene, which includes the REP sequences and was derived from the his&his0 intergenic re- gion. pWJ151. used for isolating DNA hybridiza- tion probes, is identical to pWJO1 except that the REP sequences were inserted using BamHl linkers. The DNA probes used are indicated. E,

E a s PVY EcoRI; Bst, BstEll; 8, BarnHI; Pvu, Pvul; S, Smal; SR. hybrid Smal-Rsal site.

I tb

1940 bp Kpnl fragment of histidine operon DNA that in- cludes sequences both upstream and downstream of the REP sequence (Figure 1). An RNA species of the size predicted for full-length RNA (3310 nucleotides), extend- ing from the promoter to immediately 3’of the distal gene hisP is detected in both strains. This RNA is slightly shorter from TM808 due to the 141 bp hi&-hisQ inter- genie deletion. In addition to full-length RNA and the smear of nascent transcripts (discussed in Figure l), this blot shows a clear difference between the two strains; the wild type (TA271) accumulates a 950 nucleotide RNA spe- cies that is entirely absent from the REP-deletion strain (TA3808). This 950 nucleotide RNA species is of an appro- priate length to extend from the his promoter to the REP sequence (including the entire hisJcoding sequence) and was shown to include sequences upstream, but not down- stream, of the REP sequence using other hybridization probes (for example, the 230 bp and 955 bp Hindlll-Kpnl fragments and the 755 bp Hindlll fragment; Figure 1). Thus, at least in this operon, the REP sequence causes accumulation of upstream (5’) RNA.

REP Sequences Causes Upstream RNA to Accumulate wherever They Ale Present To ascertain whether or not the REP-dependent accumu- lation of upstream RNA is specific to the histidine trans-

port operon, we examined the RNA encoded by several plasmids into which REP sequences have been cloned. An example is shown in Figure 2A, which depicts a North- ern blot of RNA encoded by plasmid pKG1800 and its derivative, pWJO1, which has the REP sequences cloned between the gal promoter and the ga/K gene. The blot was probed with a 1582 bp DNA fragment from the plasmid that includes sequences both upstream and downstream of the REP sequences. A discrete, full-length mRNA band is not seen in this blot, as there is no terminator following the galK gene and termination occurs at heterogenous points about 2.9-3.1 kb from the promoter, within pBR322 vector sequences. As expected, these full-length mRNA species are slightly larger in pWJO1 due to the 218 bp REP insert. In addition, pWJO1, but not pKG1800, accumulates large amounts of a 850 nucleotide RNA species. This is the expected size for RNA extending from the gal pro- moter to the REP sequence and was confirmed as such by the use of different hybridization probes (the 408 bp BstEll-BamHI and the 958 bp BamHI-Pvul fragments; Figures 28 and 2C). It should be noted that the gal pro- moter in these plasmids is from an operon that does not normally include a REP sequence. We have also exam- ined several other plasmid constructs, and in every case REP-dependent accumulation of upstream RNA is ob- sewed. These include plasmids in which transcription ini-

Cell 300

Figure 3. Construction of cat-gall< Plasmids (A) pWJ61 and pWJ62 are identical except that pWJ61 has a 216 bp Rsal fragment from the histidine transport operon containing the REP sequences inserted using Bamlil linkers between the cat and ga/K genes. The two plasmids were constructed by cloning the 779 bp Sal1 ‘cat cartridge” from pCMl (Close and Rodriguez, 1962) into the unique Sall sites of pDR720 and pWJ53, respectively. pDR720 and pWJ53 have been described elsewhere (Russell and Bennett, 1982; Stern et al., 1984). S, Salk 8, BarnHI. (B) Northern blot of RNA encoded by pWJ61 and pWJ62 using the 779 bp Sal1 cat cartridge as probe.

tiates from various different promoters, such as trp, MC, and lac (including the cat-ga/Kplasmid pWJ62; Figure 3). In addition, we have shown that the REP sequence iso- lated from an entirely different operon and species (the ma/E-ma/F intergenic region of the maIS operon of E. coli; see below) also causes accumulation of upstream RNA, both in its normal chromosomal location (see Figure 5) or when cloned onto multicopy plasmids (data not shown). Thus, wherever the REP sequences are present, they cause accumulation of upstream RNA. This accumulation is not promoter- or regulon-specific and does not seem to require any specific sequence in the transcription unit other than the REP sequence itself.

Quantitation of Accumulated mRNA To quantitate the REP-dependent accumulation of RNA, a filter hybridization procedure was used in which radiola- beled RNA that hybridizes to specific single-stranded DNA probes is retained on nitrocellulose filters. To enable upstream and downstream RNA to be compared, an op- eron was constructed (plasmid pWJ61) with two genes, car (chloramphenicol acetyltransferase) and ga/K (galacto- kinase), under control of the trp promoter. Plasmid pWJ62 is a derivative of pWJ61 with the REP sequence cloned be- tween car and ga/K. The construction of these plasmids is shown in Figure 3A. Figure 36, a Northern blot, shows that, as expected, the REP sequence in pWJ62 causes ac- cumulation of upstream (car) RNA. Using a probe from the car gene (see Experimental Procedures for details of probes), about 3 times more upstream RNA was present in a strain harboring pWJ62 than the same strain with pWJ61 (11,240 and 3320 cpm retained by the filters, respectively; see also Table 1). In contrast, the levels of downstream, ga/K mRNA were similar for the two plas- mids. Thus, cloning the REP sequences into this plasmid

results in approximately a 3-fold increase in upstream RNA while having no significant effect on the level of downstream RNA.

The REP Sequence Stabilizes Upstream RNA The REP-dependent accumulation of upstream RNA im- plies either that the synthesis of RNA is enhanced or that its degradation is reduced. We have demonstrated previ- ously that the REP sequence does not enhance transcrip- tion (Stern et al., 1964); the results above show that the REP sequence only affects the levels of upstream RNA and not downstream RNA, and therefore cannot be caus- ing a general increase in transcription. To demonstrate that the REP sequence does indeed stabilize upstream RNA, we measured the rate of RNA degradation in two different systems and by two different methods. In these experiments rifampicin was used to inhibit further initia- tion of transcription.

The filter hybridization method described above was used to assess the stability of car mRNA in plasmids pWJ62 and pWJ61, with and without the REP sequence, respectively. Cultures of cell8 harboring these plasmids in midexponential growth were labeled with tritiated uridine, rifampicin was added, and samples were removed for RNA isolation at appropriate time intervals. Table 1 shows the amount of car mRNA remaining at intervals after rifam- picin addition, as measured by filter hybridization. For this particular experiment, the ratio of car RNA from pWJ62 and pWJ61 is about 5:1, slightly greater than the ratio of 3:l established above. In repeated experiments, the ratio always varied between 3- and 5-fold but was always con- sistent for any single preparation of labeled mRNA. The results in Table 1 show clearly that the REP sequence reduces the rate of car mRNA degradation.

Because of the experimental difficulties inherent in

Regulation of Gene Expression by RNA Stabilization 301

Table I, Stabilization of mRNA by the REP Sequence

Amount of cat

Time after Rifampicin Addition (Min)

RNA (Counts Retained)

pWJ61 pWJ62

% RNA Remaining

pWJ61 pWJ62

0 1388 6058 100 100 2 581 4719 42 78 5 147 1975 10 33

10 16 363 2 7

Cells harboring pWJ61 or pWJ82 in midexponential growth phase were pulse-labeled with tritiated uridine, mRNA was isolated, and the amount of cat RNA was determined by filter hybridization as described in Ex- perimental Procedures.

isolating tritiated mRNA, it was not possible to use this method to measure decay at shorter time intervals after rifampicin addition. Thus, to get a more accurate measure of half-lives and a comparison between two independent methods, we adopted a rapid dot-blot procedure, which gave results essentially identical to those obtained by filter hybridization. Figure 4 shows the results of such an ex- periment on plasmids pWJO1 and pKG1800, with and with- out the REP sequences, respectively (Figure 2). Rifampicin was added to exponentially growing cells, and, at speci- fied time intervals, RNA was prepared, fixed to nitrocellu- lose filters, and hybridized using a large excess of s*P- labeled DNA probe. Two probes were used, one specific for upstream sequences and one for downstream se- quences. The hybridized dots were cut from the filters, and the amount of bound probe was determined by scintil- lation counting (Figure 4). Using the downstream probe, there was little difference in the rate of decay of mRNA from the two plasmids (Figure 48). However, using the up- stream probe (Figure 4A), it is apparent that, for this partic- ular construction, the REP sequences increase the half- life of upstream RNA by about 3-fold.

This stabilization is also confirmed by Northern blots of RNA encoded by pWJO1, isolated at various times after rif- ampicin treatment (Figure 28). The REP-stabilized RNA species is detected for at least 6 min after rifampicin treat- ment. In contrast, full-length mRNA, whether detected with an upstream (Figure 28) or downstream (Figure 2C) probe, cannot be detected at 6 min after rifampicin treat- ment (even on long exposures of the autoradiogram) and has an estimated half-life of less than 90 sec.

To demonstrate further that the REP sequence stabi- lizes upstream RNA, we examined the rates of degrada- tion of chromosomally encoded ma/ mRNA. The ma/B (maltose) locus of E. coli was selected for these and sub- sequent experiments, as ma/ RNA is relatively abun- dant and, unlike the histidine transport operon, this op- eron is readily inducible. The ma/B locus consists of two divergent operons required for maltose transport (Hengge and Boos, 1983; Figure 5); the malE-maIF-malG operon has two REP sequences in opposite orientations located in the ma/E-ma/F intercistronic region. Cells of E. coli MC4100 in midexponential growth were treated with rifam- picin to inhibit initiation of RNA synthesis. After appropri- ate time intervals, samples were taken and the RNA was

240 360 IriE I :SECS)

120 240 360

TIME (SECS)

Figure 4. Measurement of RNA Stability by Dot-Blot

Rifampicin was added to growing cells containing plasmid pWJO1 (+REP) (0) or plasmid pKG1800 (-REP) (m). RNA was isolated and dot-blotted onto nitrocellulose as described in Experimental Proce- dures. The filters were hybridized with the appropriate DNA probes, the dots were cut from the filter, and the amount of bound DNA was deter- mined by scintillation counting. The probes used were the 406 bp BstEll-BamHI fragment from pWJ151, containing only sequences up- stream of the REP sequence (A) and the 958 bp BamHI-Pvull frag- ment from pWJ151, containing downstream sequences(B). The probes and the two plasmids are shown in Figure 2.

examined by Northern blot (Figure 5). As expected from the results obtained with other operons (above), at time zero there is considerable accumulation of a 1300 nucleo- tide RNA species that hybridizes only to DNA probes con- taining sequences upstream of the REP sequences; very much less full-length mRNA (approximately 3800 nucleo- tides) is detected. As for other operons, the 1300 nucleo- tide RNA species is REP-dependent (Newbury and Hig- gins, unpublished data). Figure 5 shows that this 1300 nucleotide RNA species is exceptionally stable and can be detected up to 20 min after rifampicin treatment. Den- sitometer tracings of the autoradiogram in Figure 5, and of other similar autoradiograms using different probes (see legend to Figure 5) show that the half-life of this mRNA species is 7 to 8 min-rather more stable than most mRNA species encoded by inducible prokaryotic operons. In contrast, full-length mRNA is completely de- graded after 6 min of rifampicin treatment; no full-length RNA can be detected at 6 min, even after long exposures of the autoradiogram, which enhance the intensity of this band to that of the REP-dependent RNA species. The half- life of full-length mRNA can be estimated from densitome- ter tracings to be about 2 min. Thus, it is clear that, in

Cdl 302

65

43 malEFG- L.:

26

2.0

l-6

Figure 5. REP Stabilization of ma/E mRNA

The ma/EFG operon is shown diagrammatically. The ma/K-/am6 op- eron is transcribed divergently from a central regulatory region (Clem- ent and Hofnung, 1981; Duplayet al., 1984; Dassaand Hofnung, 1985; Froshauer and Beckwith, 1984; Gilson et al., 1982). The hybridization probes used for Northern blots are indicated and were isolated from plasmid pEJ1 (E. Davis and I? J. F. Henderson, unpublished data). The Northern blot shows RNA isolated from cells at various time intervals (in minutes) after rifampicin addition. The full-length, 3800 nucleotide ma/EFG mRNA, extending from the promoter(P) to the terminator(T), is indicated, as is the 1300 nucleotide REP-stabilized ma/E RNA. The blot shown was probed using the 1223 bp Hinfl DNA fragment, which spans ma/E. When the same blot was probed using the 1957 bp Smal-Stul fragment the full-length mRNA was still detected, but the 1300 nuclaotide ma/E species showed no hybridization (data not shown). No hybridization was seen to either band when RNA was iso- lated from cells not induced for the maltose regulon.

several different situations, the REP sequence can in- crease the half-life of upstream RNA by several fold.

REP-Stabilized RNA Is an Intermediate in mRNA Degradation We have previously shown by Sl nuclease mapping that the S’endpoint of the REP-dependent RNA species in the histidine transport operon is precisely at the 3’base of the potential stem-loop structure (Stern et al., 1984). Similar

endpoints have been found for REP-stabilized RNA from plasmids pWJO1 and pWJ62 (unpublished data). RNA species with such an endpoint could arise as a result of transcription termination during RNA synthesis, as prod- ucts of endonucleolytic processing or as intermediates in the chemical decay of RNA. We have shown elsewhere, both in vitro and by in vivo studies using gene and operon fusions, that the REP sequence does not normally func- tion as a transcription terminator (Stern et al., 1984). Thus, REP-stabilized RNA must be derived from full-length RNA by endonucleolytic processing or, alternatively, as an in- termediate in the 3’5 degradation of mRNA by exo- nucleases. We have also shown that the REP sequence is not an RNAase Ill processing site, at least in vitro (Stern et al., 1984), and have been unable to detect endonucleo- lytic cleavage at the REP sequence by E. coli cell extracts using RNA synthesized in vitro (data not shown). Although this is negative evidence, endonucleolytic cleavage of other substrates (e.g., by RNAase Ill) was detected by such procedures. To demonstrate more definitively that the REP sequence is not a substrate for endonuclease cleavage, a plasmid (pWJ82) containing two copies of the REP sequence was constructed (Figure 6). If the REP se- quence is simply a barrier to 3’-Vexonucleases, we would expect RNA species B and C to accumulate, as indeed they do (Figure 6). If the REP sequence is an endonucleo- lytic cleavage site, an additional RNA species, A, should be detected. However, an RNA species of this length can- not be detected, demonstrating that the REP-stabilized RNA species does not arise as a result of endonuclease cleavage and consequently implying that the 3’ endpoint arises as a consequence of protection 3’-5’ exonuclease attack.

RNA Stabilized by REP Sequences Is Translationally Active It is important to assess whether or not the RNA degrada- tion intermediates stabilized by the REP sequence can be translated. To address this question we adopted the fol- lowing strategy. Cells were treated with rifampicin to in- hibit RNA synthesis. At specified time intervals after rifam- picin treatment, aliquots were taken and pulse-labeled for 45 set using [35S]methionine. Specific proteins were im- munoprecipitated and separated by SDS-polyacrylamide gel electrophoresis. This procedure enabled us to deter- mine how long a specific mRNA species remains transla- tionally active after synthesis of new mRNA is inhibited.

We have shown above that the REP sequence in the ma/E-ma/F intergenic region stabilizes a 1300 nucleotide RNA species which, potentially, can encode the MalE pro- tein. The data in Figure 7 show that synthesis of the MalE protein continues for at least 12 min after rifampicin treat- ment, and densitometer scans of this and other autoradio- grams show that the half-life for translational inactivation of MalE mRNA is 6-8 min. This half-life is very similar to the half-life of REP-stabilized ma/E RNA and is rather longer than that of the full-length RNA. Indeed, there is an excellent correlation between the amount of REP- stabilized ma/E mRNA present after rifampicin treatment (measured by Northern blots; Figure 5) and the amount of

Regulation of Gene Elxpression by RNA Stabilization 303

A A -1mo- B -6504

2500 - 3300 c

p pa1

REP REP

E ES1 rl s PVU B El

pWJ02 L II Ilk 9elK 1 I I blr I +t I I 44. I I

1 kb

Figure 6. RNA from pWJ82, Containing Two REP Sequences

(A) pWJ82 was constructed by inserting a second copy of the REP sequence (as a 218 bp Rsal fragment from the histidine transport operon) into the unique Accl site of pWJO1 (Figure 2) using BamHl linkers. The sizes and extents of the potential RNA species A, B, C, and D, are shown. See text for further details. (6) The Northern blot shows RNA from pWJ82 probed with the 1582 bp BstEll-Pvul fragment from pWJO1 (shown in Figure 2) which includes sequences upstream and downstream of the REP sequence. The locations of bands B, C, and D, are indicated, as is the position that band A would migrate to if it were present. Band X is a result of hybridization to intermediates in synthesis and degradation which are artifactually accumulated due to overloading of the gel with rRNA (see Figure 1) and is not band A; it is seen in parallel tracks of RNA from pWJO1 and is detected by the 406 bp BstEll-BamHI probe, which cannot detect band A (data not shown).

MalE protein synthesized. This strongly suggests that the MalE protein is primarily synthesized from REP-stabilized RNA rather than from full-length RNA. To demonstrate this more definitively, it is important to assess the translational half-life of full-length mRNA by measuring the synthesis of either the MalF or the MalG protein. As both these pro- teins are synthesized in very low amounts and antibodies are not available to us, we instead measured f%galactosi- dase synthesis from a ma/F-IacZ fusion. This fusion was constructed as described in Experimental Procedures. Figure 7 shows that the functional half-life of full-length mRNA, as measured by f3-galactosidase synthesis from a ma/F-lacilfusion, is only 2 to 3 min, which is similar to the half-life of full-length mRNA and much less than the half- life for functional inactivation of ma/E RNA, which is about 6-8 min. This result strongly implies that the MalE protein is primarily synthesized from REP-stabilized RNA rather than from full-length mRNA and that, consequently, at least a proportion of the REP-stabilized RNA molecules are translationally active.

Induction of the Maltose Operon As further evidence that the MalE protein is primarily syn- thesized from REP-stabilized RNA rather than from full- length RNA, we examined the synthesis of MalE protein and ma/E mRNA following induction of the maltose op- eron. Figure 8A shows a Northern blot of maltose RNA at varying time intervals after induction of the operon. Within 4 min of induction, full-length mRNA reaches its steady- state level. However, the REP-stabilized RNA species con- tinues accumulating for at least 20 min after induction. Figure 8B shows the capacity of cells to synthesize MalE protein at similar time intervals after induction. Quite clearly, the rate of synthesis of MalE protein increases for

0 2 4 6 B.

8 1012

MalE + w-rr*.-- /

Figure 7. Translation of REP-Stabilized mRNA

Cells of strain CH1431 (ma/F-/acZfusion) in exponential growth in M63 amino acids medium (plus maltose as carbon source to induce the ma/ operon) were treated with rifampicin. After the indicated time intervals (in minutes), samples were pulse-labeled and the labeled MalE and LacZ proteins were immunoprecipitated, separated by electrophoresis on a 10% gel (MalE) or a 7% gel (Laci!), and detected by autoradio- w&v.

at least 20 min after induction and closely parallels the ac- cumulation of REP-stabilized RNA. If MalE were encoded by full-length mRNA, maximum rates of MalE synthesis would be reached after only 4 min of induction when full- length mRNA reaches its steady-state level. This is the case for 6-galactosidase synthesized from the ma/F-/acZ fusion (data not shown). Thus it seems likely that most, if not all, MalE protein is translated from the REP-stabilized RNA species rather than from full-length RNA. To ascer- tain this, deletions of the REP sequences from this operon must be constructed; these experiments are in progress.

Cell 304

0 0 2 4 6 8 10 12

DNA

-3800

-1300

.MalE

Figure 6. Induction of the Maltose Operon

Cells of MC4100 were grown in M63 amino acids medium with glucose as carbon source. At time 0 the maltose operon was induced by adding maltose (0.4%) and CAMP (5 mM). Samples were taken at the indicated times after induction (in minutes) and RNA was isolated and analysed by Northern blot using the 1223 bp Hinfl fragment from the ma/E gene (Figure 5) as probe (A), or samples were pulse-labeled, and the labeled MalE protein was immunoprecipitated. separated by electrophoresis, and detected by autoradiography (8).

His J- TA 271

HisJ- VjV.”

Figure 9. Effect of Deleting the REP Sequence on HisJ Synthesis

Cells of strain TA271 (+REP) or TA3606 (with the REP deleted from the his operon) were grown to midexponential phase in M63 amino acids with glucose as carbon source. Rifampicin was added and at the indi- cated time intervals after rifampicin addition (in minutes), samples were taken and pulse-labeled. The HisJ protein was immunoprecipitated, separated by electrophoresis on a 10% gel, and detected by autora- diography.

Translation of REP-Stabilized RNA from Other Operons Previous results with the histidine transport operon of S. typhimurium showed that deletion of the REP sequence from this operon caused only a 50% reduction in synthe- sis of the upstream HisJ protein (Stern et al., 1984). Al- though this seems at odds with the data above, which show that, in this operon, deletion of the REP sequence causes a considerable reduction in the accumulation of upstream (hi.%/) RNA, the two observations can be recon- ciled in one of two ways. Either, for the his operon, a high proportion of the REP-stabilized RNA is translationally in- active, or the accumulation of HisJ protein is limited by a factor other than the availability of mRNA, for example by feedback inhibition of translation. To distinguish between these possibilities, the functional activity of his,/ mRNA was assessed by pulse-labeling and immunoprecipitation following rifampicin treatment, as described above for MalE. Figure 9 shows that the functional half-life of hisJ mRNA is reduced when the REP sequence is deleted (comparing TA271 with TA3808), indicating that a propor- tion of the REP-stabilized RNA must be translationally ac- tive. However, the amount of HisJ synthesized by strains with the REP sequence (TA271) is the sum of protein syn- thesized from full-length RNA plus the protein synthe- sized from REP-stabilized RNA. The amount of protein synthesized can be estimated from densitometer scans of these and other autoradiograms. When the amount of HisJ synthesized from full-length mRNA (TA3808) is sub- tracted from the amount synthesized from full-length plus REP-stabilized RNA (TA271) this gives a measure of the amount of protein whose synthesis is directed by the REP- stabilized species alone. The REP-stabilized RNA ac- counts for most of the HisJ synthesized after 4 min of rifampicin treatment. However, during steady-state growth (i.e., time 0, before rifampicin addition) the REP-stabilized RNA accounts for no more than 50% of the total HisJ pro- tein synthesized. This is consistent with data showing that the REP sequence only increases HisJ synthesis about P-fold (Stern et al., 1984). As Northern blots (Figure 1) indi-

Regulation of Gene Expression by RNA Stabilization 305

0 0 2 4 6 8 10 12

pWJ62 (+REP)

0 0 2 4 6 8 10 12

CAT pWJ61 (-REP)

Figure 10. Translation of REP-Stabilized cat RNA

Cells of MC4100 containing pWJ61 (-REP) or pWJ62 (+REP) at midex- ponential growth were treated with rifampicin. After the indicated time intervals, samples were pulse-labeled and the CAT protein was im- munoprecipitated, separated by electrophoresis, and identified by autoradiography.

cate that there is at least 10 times more hi.%/ RNA in strains with the REP sequence present, it is clear that, for this particular operon, only a small proportion of the REP- stabilized mRNA can be functionally active.

Stabilization of RNA by REP Sequences Can Enhance Protein Synthesis We have shown above that cloning the REP sequence downstream of the cat gene (plasmids pWJ61 and pWJ62; Figure 3) results in a 3 to 5-fold increase in upstream cat mRNA as a result of stabilization. To establish whether this increase in RNA also results in an increase in protein syn- thesis, CAT activity was assayed and found to be 3.9 times greater in cells harboring pWJ62 than in the same cells with pWJS1. Thus, cloning the REP sequence down- stream of a gene can increase protein synthesis. In this case, and unlike the case of histidine operon, the increase in upstream mRNA is similar to the increase in CAT syn- thesis, implying that in this operon most of the stabilized mRNA is functionally active. To show this more directly, synthesis of the CAT protein from these plasmids was also assayed by pulse-labeling and immunoprecipitation fol- lowing rifampicin treatment (Figure 10). Again, it can be seen that the REP sequence increases the functional half- life of upstream RNA. Significantly, the differences in functional half-life of CAT mRNA seen between these two plasmids is very similar to the differences in mRNA half- lives measured above (Table 1). These results show un- ambiguously not only that the REP sequence stabilizes upstream RNA, but also that this stabilization can signifi- cantly increase the levels of upstream protein synthesis.

Discussion

The REP sequence comprises 1% of the bacterial ge- nome and is present in about 25% of all transcription units. It is therefore of considerable importance to identify any effects this sequence may have on these transcription units and to establish its biological role.

We have shown here that the REP sequence can stabi- lize mRNA and, consequently, increase the intracellular concentration of upstream RNA by more than an order of magnitude. The REP sequence will stabilize any RNA that is placed upstream, apparently independently of the pro- moter from which transcription initiates or of any other se- quence. Thus, the ability to stabilize upstream RNA seems to be inherent to the REP sequence itself. It is well established that different mRNA molecules have different half-lives, but the factors that determine this are poorly un- derstood. Recently, Belasco and coworkers (Belasco et al., 1966) have shown that discrete segments of various transcripts can influence the degradation of cotranscribed RNA, although the mechanisms by which this is achieved are not understood and probably vary from determinant to determinant. In addition, there is no obvious sequence similarity between these determinants. The REP se- quences represent a determinant of RNA stability that is common to very many operons, and presumably such stabilization is achieved by a common mechanism. It is worth noting that the REP sequences are unique; no other class of highly conserved, repetitive sequences is pres- ent on the E. coli chromosome (M. Hofnung, personal communication).

The REP sequence stabilizes RNA by protection from 3-5’ exonuclease attack. The finding has a number of general implications for the processes of mRNA degrada- tion in bacteria. Several years ago it was hypothesized that mRNA degradation proceeds primarily by 3’-5’ exo- nuclease attack (Apirion, 1973). The 3’ endpoints that serve as substrates for these exonucleases could either be the 3’ends of full-length mRNA or be produced by en- donucleolytic cleavage of transcripts (e.g., at intercistronic regions [Achord and Kennell, 1974; Lim and Kennell, 19791). This view is supported by the rather negative evi- dence that no exonuclease with S-3’activity has been de- tected in E. coli, despite suggestions that there may nevertheless be a role for 5’-3’ decay (e.g., Cannistraro and Kennell, 1965a). The data presented here provide strong evidence that the principal mode of RNA degrada- tion is by 3’5 exonucleolytic attack. Indeed, if degrada- tion by 5’-3’exonucleases were involved to any significant extent in the bulk chemical decay of mRNA, then REP se- quences could not stabilize upstream RNA in a transla- tionally active form. Because the REP sequences stabi- lize any upstream mRNA, apparently irrespective of its sequence, it follows that 5’-3’decay, if it occurs at all, must be specific to a limited number of operons or be restricted to trimming the 5’end of RNA molecules (Cannistraro and Kennell, 1965b). In addition, the fact that we can detect full-length mRNA from a number of polycistronic operons implies that, if endonucleolytic cleavage at intercistronic regions is important, it can only occur after transcription of the operon is complete.

Given that the REP sequence protects upstream RNA from 3’-5’exonucleases, it is possible to assess the impor- tance of such stabilization for different operons and under different conditions. The concentration of REP-stabilized RNA at steady state, [Ra], can be described, to a first ap- proximation, by the equation:

Cell 306

[Ral = (h[h] - kdt,

where k, and k2 are rate constants for the 3’6’ degrada- tion of RNA and for overcoming the REP barrier, respec- tively, and [Rb] is the concentration of full-length mRNA. We make the reasonable assumptions that the 3’5’ degradation of RNA (k,) is dependent on the concentra- tion of 3’endpoints [Rb] and is therefore a first order reac- tion, while endonucleases overcoming the REP barrier (ks) is a concentration-independent, zero order reaction. That this equation does reflect the in vivo situation and that the assumptions are essentially correct is shown in Figure 8A; despite the fact that full-length RNA [Rb] reaches steady-state concentrations after 4 min of induc- tion, the concentration of REP-stabilized RNA [Ra] in- creases with time. It follows from the above equation that [Ra] will only accumulate if kl[Rb] > kp. Thus, the ac- cumulation of REP-stabilized RNA will depend upon the concentration of full-length RNA [Rb] or, assuming exo- nuclease activity is constant, upon the rate of transcrip- tion. Thus, the ratio of REP-stabilized RNA to full-length RNA will depend upon the rate of transcription of an op- eron. If REP is located intercistronically, which it fre- quently is, and assuming that translational inactivation of a message (if any) is constant, the ratio of gene products in such an operon will depend upon the rate of transcrip- tion of that operon. This provides an elegant potential mechanism for altering the ratio of gene products within a multicistronic operon as transcription of that operon is modulated under different conditions or during induction of expression.

The above considerations allow us to assess the amount of REP-stabilized RNA that will accumulate under any given conditions. However, the intracellular concen- tration of mRNA does not necessarily reflect the amount of protein made. For certain specific genes there may be autoregulation of translation by the gene product (e.g., ribosomal proteins or alanyl-tRNA synthetase; Nomura et al., 1984; Putney and Shimmel, 1981). Leaving aside feed- back control, which is limited to a few specific operons, mRNA can also be irreversibly inactivated; available evi- dence indicates that this translational or functional inacti- vation is by processing at the 5’ end of the mRNA (Shen et al., 1981; Cannistraro and Kennel, 1985b). Whatever the mechanism, this process is likely to be more or less impor- tant for different mRNA species. We have demonstrated here that REP-stabilized RNA can be translationally ac- tive, a finding of considerable general importance. First, this result demonstrates that translational inactivation does not necessarily precede chemical degradation, and therefore the suggestion that functional inactivation (en- donuclease cleavage) is a necessary first step in mRNA decay must be incorrect. It follows that, for many mRNA molecules, exonuclease attack from the S’end is probably the first step in degradation, only rendering an mRNA molecule nonfunctional once the coding sequence is reached. Second, our finding predicts that, potentially, in vivo gene expression could be regulated by modulating the rate of mRNA degradation.

For any given operon, the importance of such pro-

cesses will depend upon the relative rates of degradation (rate constant, k,) and inactivation (rate constant, k3). Thus, for the histidine transport operon, the REP se- quence causes considerable accumulation of upstream RNA, but this does not result in a corresponding increase in HisJ protein (about a 2-fold increase is observed) presumably because the bulk of this RNA is translation- ally inactive (i.e., k3 >> k2). In contrast, for the artificially constructed cat-galK operon, the REP sequence causes a 3-fold increase in upstream RNA, which leads to an equivalent increase in upstream protein synthesis, imply- ing that most, if not all, of the stabilized RNA is translation- ally active. Evidence that mutations in the REP sequence distal to the gIyA gene reduce expression of glyA (Plamann and Stauffer, 1985) is consistent with the possi- bility that REP-mediated stabilization does play an in vivo role. Finally, and most importantly, the stabilization of ma/E mRNA by the REP sequence appears to play a phys- iological role. After induction of the operon, the REP se- quence results in a massive accumulation of upstream ma/E RNA and much of this RNA appears to be translat- able. Thus, at least in this case it seems likely that differ- ential stability of mRNA will determine the relative expres- sion of genes in a multicistronic operon, at least in the period immediately following induction. Whether or not RNA stability is also important when the operon is at steady state depends upon whether translational inactiva- tion of the mRNA remains constant and can only be as- sessed by deletion of the REP sequence from this operon; these experiments are in progress.

It has previously been demonstrated that mRNA half- life can vary in response to changes in environmental con- ditions, such as growth rate (Nilsson et al., 1984) and that mRNA half-lives can correlate with levels of protein syn- thesis (Belasco et al., 1985). While such data are clearly compatible with the view that control of mRNA stability may regulate gene expression, in none of these cases has the primary event been established. Thus, it is not clear whether modulation of RNA stability leads to a change in protein synthesis or, alternatively, whether the rate of translation is altered by some other means and this in- directly alters RNA stability. In the latter case, of course, altered mRNA stability plays no role in the regulation of gene expression but is simply a secondary consequence of other controlling events. This present study provides clear evidence that, at least in certain cases, RNA can be stabilized in a translationally active form, and thus by al- tering RNA stability it is possible to alter gene expression. We have been able to increase protein synthesis by stabilizing a specific RNA species and to show that, in at least one case, the differential expression of genes within an operon is to some extent dependent on differential RNA stability.

The demonstration that the REP sequence can stabilize mRNA, and by virtue of this stabilization can influence gene expression, poses the question, is this the primary role of the REP sequence? Is the high degree of sequence conservation between REP sequences maintained solely because of this specific role? It seems to us that, while we have clearly identified a function for the REP sequence,

Regulation of Gene Expression by RNA Stabilization 307

this function is unlikely to provide sufficient selective ad- vantage to account for all the highly conserved copies of the REP sequence. The REP sequence is not the only se- quence that can stabilize RNA. Any large stem-loop struc- ture can have similar effects. For example, the trp termina- tor can protect upstream RNA against exonuclease attack (Mott et al., 1985), and we have shown that a large stem-loop structure located between the opp4 and oppB genes of S. typhimurium stabilizes upstream RNA and is probably at least partly responsible for the higher level of OppA than OppB synthesized from this polycistronic tran- script (Hiles et al., submitted). All these stem-loops presumably function by protecting the 3’end of transcripts from 3’-5’ exonucleases. If, as seems to be the case, any large stem-loop structure can stabilize RNA, then the large number of copies of the REP sequence on the chro- mosome, and their primary sequence conservation, would not be anticipated. It is of course possible that, un- like other stabilizing stem-loops, REPS have an additional activity. For example, a protein may bind to the sequence under certain conditions, or it may be a substrate for an endonuclease, providing a means of modulating RNA sta- bility and gene expression. However, several lines of evi- dence argue against this. First, the REP sequence has lit- tle effect on gene expression in some operons. Second, we have been completely unable to detect proteins that bind to the REP sequence DNA (unpublished results), cleavage by endonucleases, or changes in differential gene expression in operons carrying REP sequences un- der different growth conditions (although it could always be argued that these experiments have not been carried out under conditions in which such effects would be seen). Third, such a variety of operons contain the REP sequence that it is difficult to envisage a set of growth con- ditions under which a change in expression of all of these genes would be required.

Finally, and most convincingly, there are now several examples of operons that have been sequenced in both E. coli ant! S. typhimurium. In three such cases the loca- tion of REP sequences is different. Thus, the REP se- quence that follows the meN gene of E. coli is completely absent from this location in S. typhimurium (Saint-Girons et al., 1984; Urbanowski and Stauffer, 1985). Similarly, the REP sequence following the rpoD gene of E. coli is absent from S. typhimurium despite extensive (91%) homology between the surrounding sequences (Erickson et al., 1985), and the gInA-n&B intergenic region of E. coli con- tains a REP sequence that is absent from the equivalent intercistronic regions of both S. typhimurium and Kleb- siella pneumoniae (Ueno-Nishio et al., 1984; MacFarlane and Merrick, 1985), despite the fact that there is no known difference in the regulation of this operon in these three species. It should be pointed out that, within a species, the location of the REP sequence seems to be constant. We are aware of three operons from E. coli containing REP sequences that have been sequenced twice by different groups, and in each case the REP sequence is present and identical in sequence.

Thus, we suggest that stabilization of RNA is not the pri-

mary role of REP sequences but that, in certain operons,

this property of REP sequences has been recruited by the cell to serve a physiological function. Other possible functions for the REP sequence have been suggested, in- cluding the generation of chromosomal duplications/rear- rangements or the organization of the chromosome as binding sites for the Hu histone-like proteins. However, no good evidence for involvement in such processes has been forthcoming and for various reasons (discussed elsewhere; Higgins and Smith, 1986, and unpublished data), we think these roles unlikely. It seems highly plausi- ble that REP sequences are prokaryotic equivalents of “selfish:’ repetitive DNA and that they became dispersed around the genome by a selfish process, possibly RNA- mediated gene conversion, rather than because they confer any specific selective advantage per se; this is dis- cussed at length elsewhere and is being tested experi- mentally (unpublished data). Subsequently, secondary properties of these dispersed sequences, such as their potential to stabilize RNA, have been recruited by the cell for specific functions. It is easy to imagine that certain copies of the REP sequence may have evolved as tran- scription terminators, having acquired a following run of T’s. Thus, it seems quite plausible that in other situations, REP sequences may serve other biologically important functions.

Experimental Procedures

Bacterial Growth All S. typhimurium strains were isogenic derivatives of LT2 except for the lesions indicated. All E. coli strains were derivatives of MC4100 (Casadaban, 1976). Ceils were grown with aeration at 37% except for Mu-kc lysogens, which were grown at 30°C. LB (Roth, 1970) and M63 (Miller, 1972) were used as complete and minimal media unless other- wise stated.

Enzyme Assays Cells for CAT or GalK assays were grown to midexponential in M63 medium supplemented with 2% fructose as carbon source and, when appropriate. the inducers fucose (5 x 10e4 M; gal promoter) or indole acrylic acid (5 pglml; trp promoter). GalK was assayed by the conver- sion of galactose to galactose-l-phosphate as described (McKenney et al., 1981). CATactivity was assayed as the release of coenzyme A from acetyl CoA by its reaction with dithionitrobenzoic acid as described by Close and Rodriguez (1962). Kalactosidase activity was assayed ac- cording to Miller (1972). When expression from multicopy plasmids was measured, any variations in plasmid copy number were measured as described below and included in the calculations.

Pulse-Labeling and Immunoprecipitallon Cells (0.5 ml) were grown to midexponential growth (O.D.,o = 0.4) in M63 minimal medium containing all protein amino acids except methionine at 40 @ml and the appropriate carbon source at 0.4%. When required, rifampicin was added to a final concentration of 200 Kg/ml. The cells were pulsed for 45 set with [35SJmethionine (5 PCi; 1.3 Cilmmol) followed by a chase with unlabeled methionine at 100 vglml for 45 sec. The cells were rapidly sedimented by centrifugation and snap-frozen. Cell pellets were resuspended in 50 ~1 of STE (1% SDS, 10 mM Tris-HCI (pH 6.0), 1 mM EDTA) and boiled for 2 min. After cooling lo room temperature. 450 pl of KI buffer (50 mM Tris-HCI (pli 6.0), 2% Tritoin X-100. 150 mM NaCI, 1 mM EDTA) was added, and anv debris was removed by microcentrifugation. Chilled KI buffer (300 ~1) was added to 200 ~1 of supernatant together with the antibody, and precipitation was allowed lo proceed overnight at 4OC. Staphylococcus aureus protein A (20 ~1) was added, the mix was incubated on ice for 20 min, and the precipitate was sedimented by microcentrifugation for 1 min The pellet was washed twice in cold KI buffer and once in 10

Cell 308

mM Tris-HCI (pH 8.0) then resuspended in 30 ul of 2x Laemmli sam- ple buffer (Laemmli, 1970). The samples were boiled for 2 min and cen- trifuged for 5 min. Ten microliters of supernatant was loaded onto a 10% SDS-polyacrylamide gel. Slab gels (0.8 mm) were run as de- scribed elsewhere (Laemmli, 1970; Ames, 1974).

DNA Tachnlques Restriction endonucleases and DNA-modifying enzymes were pur- chased from Amersham or Bethesda Research Laboratories and were used according to the manufacturers instructions. DNA was prepared by the alkaline procedure (Birnboim and Daly, 1979) and, when neces- sary, was purified by cesium chloride density gradient centrifugation. Labeling DNA with =P by nick translation, other DNA manipulations, and agarose gel electrophoresis was as described by Maniatis et al. (1982).

RNA Techniques RNA for Northern blots was isolated by hot SDS-phenol extraction (Peck and Wang, 1985). Samples were separated on formaldehyde- agarose gels, transferred to nitrocellulose, and hybridized to radiola- beled DNA probes in 50% formamide as described by Maniatis et al. (1982). Sizes of mRNA species were estimated by comparison with rRNA and denatured DNA markers. RNA for quantitation by fast- blotting was isolated as follows. Cells (1 ml) at exponential growth were sedimented and snap-frozen at -70%. The pellets were resuspended in 100 ul of 20% sucrose, 100 mM Tris-HCI (pH 8.0), IO mM EDTA, and 3 mg/ml lysozyme, and the cells were broken by three cycles of freeze-thawing. Formaldehyde (100 ul), SDS (30 ul at 10%) and 20 x SSC (200 ul; White and Bancroft, 1982) were added, and the mixture was heated at 8oPC for 15 min. Debris was then removed by centrifuga- tion for 2 min, and the supernatant was diluted 4-fold with 4x SSC and bound to nitrocellulose using a Bio-Rad dot-blotter. The filters were hy bridized to labeled probe as for Northern blots.

Filter Hybridizations Filter hybridizations were carried out by an adaptation of published procedures (Gegenheim and Apirion, 1978; Hauser and Hatfield, 1984). RNA was labeled in vivo by pulsing exponentially growing cells for 3 min with [3H]uridine (10 mCi/ml; 38 Cilmmol). An equal volume of ethanol prechilled to -70X was added, the cells were pelleted by centrifugation, and RNA was extracted with hot SDS-phenol (Peck and Wang, 1985). RNA from 5 ml of cells was finally resuspended in 150 ul of hybridization buffer. Hybridizations were carried out in 50% for- mamide as described for Northern blots. Single-stranded Ml3 DNA probe (8 pg) was bound onto a nitrocellulose filter disc, prehybridized, and hybridized as for Northern blots, using tritiated RNA from 5 ml of cells as probe. The filters were then washed in 0.3 M NaCI, 30 mM so- dium citrate, and 10 mM MgClp, incubated in the same solution con- taining RNAase A (10 @ml). and washed in 2x SSC and 0.1% SDS as for Northern blots. The PHjRNA retained by the filters was deter- mined by scintillation counting. The values obtained were adjusted for background using filters with wild-type Ml3 DNA bound and for the size of the DNA probe, the proportion of uridine residues in the hybridizing RNA sequence, and plasmid copy number. It should be noted that, because the efficiency of RNA labeling varied somewhat between different preparations, the absolute number of counts re- tained by the filters varied between experiments. However, within a sin- gle experiment, values were always highly reproducible, and the ratios of RNA transcribed from different plasmids were, of course, consistent among .3xperiments.

Determination of Plasmld Copy Number DNA was extracted from I.5 ml of cells grown to the appropriate optical density using the alkaline procedure (Birnboim and Daly, 1979). The DNA was treated with RNAase A (100 ug/ml; 15 min), precipitated with ethanol, resuspended in 50 ul of water, denatured in alkali, and bound to nitrocellulose filters as described for colony hybridizations (Grun- stein and Hogness, 1975). The filters were hybridized with an excess of =P-labeled DNA in aqueous solution at 65OC as described by Mani- atis et al. (1982). After drying, the filters were cut up and the radioactive probe bound to each spot was determined by scintillation counting.

Construction of a maIF-/acZ Oparon Fusion A random pool of 10,000 independent Mud11681 insertions into the chromosome of strain MC4100 (Casadaban, 1978) was constructed as described elsewhere (Castilho et al., 1984). Mud11681 is a mini-Mu derivative, which upon insertion into a gene in the correct orientation places /acZ under control of that genes promoter. From this pool, Mal- derivatives were selected on MacConkey maltose plates. Those inser- tions in the ma/B locus (rather than any of the other ma/ genes) were identified by cotransduction with a TnlO insertion in /amB and 8-galac- tosidase activity was shown to be maltose-inducible. Cotransduction with a lam8::TnlO was also used to confirm that each strain contained a single Mu insertion and that this insertion was responsible for the Mal- phenotype. The ma/B locus consists of five genes. Fusions to /amB and malK(by virtue of their polar effect) were eliminated by their resistance to phage lambda, and fusions to ma/E, by their failure to synthesize detectable levels of the MalE protein. Any remaining fu- sions were assumed to be a result of insertions in either ma/F or ma/G, and the point of insertion was identified by Southern blot, determining the chromosomal restriction fragment that is altered by the insertion (as described in Hiles et al., submitted). One ma/F-/acZ fusion isolated in this way (CHl431) was used for all experiments.

Acknowledgments

We thank many colleagues, particularly Giovanna Ferro-Luzzi Ames and Terry Platt, for stimulating discussions. We are grateful to Giovanna Ferro-Luzzi Ames, Hannes Brass, Peter Henderson, and Bill Shaw for providing bacterial strains, plasmids, and/or antibodies. This work was supported by grants from the Medical Research Council to C. F. H. and an MRC studentship to I. D. H. C. F. H. is a Lister Institute Research Fellow.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “‘advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received September 10, 1986; revised October 30, 1986

References

Achord. D., and Kennell, D. (1974). Metabolism of messenger RNAfrom the gal operon of Escherichia co/i. J. Mol. Biol. 90, 581-599.

Ames, G. F.-L. (1974). Resolution of bacterial proteins by polyacryl- amide gel electrophoresis on slabs. J. Biol. Chem. 249, 634-644.

Apirion, D. (1973). Degradation of RNA in Escherichia co/i. A hypothe- sis. Mol. Gen. Genet. 722, 313-322.

Barry, G., Squires, C., and Squires, C. L. (1980). Attenuation and pro- cessing of RNA from the r&/L-rpo8CC transcription unit of Escherichia co/i. Proc. Natl. Acad. Sci. USA 77; 3331-3335.

Belasco, J. G., Beatty, J. T., Adams, C. W., von Gabain, A., and Cohen, S. N. (1985). Differential expression of photosynthesis genes in R. cap- sulata results from segmental differences in stability within the polycis- tronic rxcd transcript. Cell 40, 17I-181.

Belasco. J. G., Nilsson, G., von Gabain, A., and Cohen, S. N. (1986). The stability of E. coli gene transcripts is dependent on determinants localized to specific mRNA segments. Cell 46. 245-251. Birnboim, H. C., and Daly, J. (1979). A rapid alkaline extraction proce- dure for screening recombinant plasmid DNA. Nucl. Acids Res. 7, 1513-1523.

Burton, Z. F., Gross, C. A., Watanabe, K. K., and Burgess, R. R. (1983). The operon that encodes the sigma subunit of RNA polymerase also encodes ribosomal protein S21 and DNA primase in E. coli K12. Cell 32, 335-349. Cannistraro. V. J., and Kennell, D. (1985a). Evidence that the Send of lac mRNA starts to decay as soon as it is synthesised. J. Bacterial. 761, 620-822.

Cannistraro, V. J., and Kennell, D. (1965b). The 5’ ends of Eschericbia co/i lac mRNA. J. Mol. Biol. 182, 241-248. Casadaban, M. J. (1976). Transposition and fusion of the /ac genes to

Regulation of Gene Expression by RNA Stabilization 309

selected promoters in E. co/i using bacteriophage lambda and Mu. J. Mol. Biol. 104, 541-555.

Castilho, 6. A., Olfson, P., and Casadaban, M. J. (1984). Plasmid inser- tion mutagenesis and lac gene fusion with mini-Mu bacteriophage transposons. J. Bacterial. 158, 456-495.

Chanda, P. K., Ono, M., Kuwano, M., and Kung, H.-F (1965). Cloning, sequence analysis and expression of alteration of mRNA stability gene (Ams) of E. co/i. J. Bacterial. 167. 446-449.

Clement, J. M., and Hofnung, M. (1981). Gene sequence of the recep- tor, an outer membrane protein of E. coli K12. Cell 27, 507-514.

Close, T. J., and Rodriguez, R. L. (1962). Construction and character- ization of the chloramphenicol-resistance gene cartridge: a new ap- proach to the transcriptional mapping of extrachromosomal elements. Gene 20, 305-316.

Dassa, E., and Hofnung, M. (1965). Sequence of gene ma/G in E. co/i K12. Homologies between integral membrane components from bind- ing protein-dependent transport systems. EMBO J. 4, 2287-2293.

Deutscher, M. P. (1965). E. coli RNases: making sense of alphabet soup. Cell 40, 731-732.

Duplay, P., Bedouelle, H., Fowler, A., Zabin, I., Saurin, W., and Hof- nung, M. (1964). Sequences of the ma/E gene and of its product, the maltose-binding protein of fscherichia co/i K12. J. Biol. Chem. 259, 10606-10613.

Erickson, B. D.. Burton, Z. F., Watanabe, K. K., and Burgess, R. R. (1965). Nucleotide sequence of the rpsU-dnaG-rpoD operon from Sa/mone//a typhimurium and a comparison of this sequence with the homologous operon of fscherichia co/i. Gene 40, 67-76.

Froshauer, S., and Beckwith, J. (1964). The nucleotide sequence of the gene for ma/F protein, an inner membrane component of the maltose transport system of Escherichiacoli. J. Biol. Chem. 259, 10696-10903.

Gegenheimer, P., and Apirion, D. (1976). Processing of rRNA by RNAase P: spacer tRNAs are linked to 16s rRNA in an RNAase P RNAase Ill mutant strain of E. colt Cell 15, 527-539.

Gilson, E., Nikaido, H., and Hofnung, M. (1962). Sequence of the ma/K gene in E. co/i K12. Nucl. Acids Res. 70, 7449-7456.

Gilson, E., Clement, J.-M., Brutlag, D., and Hofnung, M. (1984). A fam- ily of dispersed repetitive extragenic palindromic DNA sequences in E. co/i. EMBO J. 3. 1417-1422.

Graham. M. Y., Tal, M., and Schlessinger, D. (1962). lac transcription in fscherichia co/i cells treated with chloramphenicol. J. Bacterial. 751, 251-261

Grunstern, M., and Hogness, D. S. (1975). Colony hybridization: a method for the isolation of cloned DNA’s that contain a specific gene, Proc. Natl. Acad. Sci. USA 72, 3961-3965.

Guameros, G., Montanez, C., Hernandez, T, and Court, D. (1962). Posttranscriptional control of bacteriophage I i&gene expression from a site distal to the gene. Proc. Natl. Acad. Sci. USA 79, 236-242.

Gupta, R. S., and Schlessinger, D. (1976). Coupling of rates of transcrip tion, translation and messenger ribonucleic acid degradation in streptomycin-dependent mutants of fscherichia co/i. J. Bacterial. 725, 64-93.

Gupta, R. S., Kasai, T., and Schlessinger, D. (1977). Purification and some novel properties of RNasell. J. Biol. Chem. 252, 8945-6951.

Har-El, R., Silberstein, A., Kuhn, J., and Tal, M. (1979). Synthesis and degradation of rat mRNA in E. colidepleted of 30s ribosomal subunits. Mol. Gen. Genet. 173. 135-144.

Hauser, C. A., and Hatfield, G. W. (1964). Attenuation of the i/v8 operon by amino acids reflecting substrates or products of the i/vB gene prod- uct. Proc. Natl. Acad. Sci. USA 87, 76-79.

Hengga. R., and Boos, W. (1963). Maltose and lactose transport in fscherichia co/i. Examples of two different types of concentrative transport systems. Biochim. Biophys. Acta 737, 443-478.

Higgins, C. F., and Ames, G. F-L. (1961). Two periplasmic transport pro- teins which interact with a common membrane receptor show exten- sive homology: complete nucleotide sequences. Proc. Natl. Acad. Sci. USA 78, 6036-6042.

Higgins. C. F, and Ames, G. F.-L. (1962). Regulatory regions of two

transport operons under nitrogen control: nucleotide sequences. Proc. Natl. Acad. Sci. USA 79, 1063-1067.

Higgins, C. F., and Smith, N. H. (1966). Messenger RNA processing. degradation and the control of gene expression. In Regulation of Gene Expression, Society for General Microbiology Symposium 39, I. R. Booth and C. F. Higgins, eds. (Cambridge: Cambridge University Press), pp. 179-196.

Higgins, C. F., Ames, G. F.-L., Barnes, W. M., Clement, J. M., and Hof- nung, M. (1962a). A novel intercistronic regulatory element of prokaryotic operons. Nature 298, 760-762.

Higgins, G. F., Haag, P D., Nikaido, K., Ardeshir, F., Garcia, G., and Ames, G. F.-L. (1982b). Complete nucleotide sequence and identifica- tion of membrane components of the histidine transport operon of S. typhimurium. Nature 298, 723-727.

Kaplan, R., and Apirion, D. (1974). The involvement of ribonuclease I, ribonuclease II and polynucleotide phosphorylase in the degradation of stable ribonucleic acid during carbon starvation in E. co/i. J. Biol. Chem. 249, 149-151.

Laemmli, U. K. (1970). Cleavage of structural proteins during the as- sembly of the head of bacteriophage T4. Nature 227, 660-665.

Lim, L. W., and Kennell, D. (1979). Models for decay of fscherichia co/i lac messenger RNA and evidence for inactivating cleavages between its messages. J. Mol. Biol. 135, 369-390.

MacFarlane, S. A., and Merrick, M. (1965). The nucleotide sequence of the nitrogen regulation gene nfr0 and the gInA-ntr6C intergenic re- gion of K/ebsiel/a pneumoniae. Nucl. Acids Res. 13, 7591-7606.

Maniatis, T., Fritsch, E. F., and Sambrook, J. (1962). Molecular Cloning: A Laboratory Manual. (Cold Spring Harbor, New York: Cold Spring Har- bor Laboratory).

McKanney, K., Shimatake, H., Court, D., Schmeissner, U., Brady, C., and Rosenberg, M. (1961). A system to study promoter and terminator signals recognized by fscherichia co/i RNA polymerase. In Gene Am- plification and Analysis, Vol. II, J. C. Chirikjian and T. S. Papa, eds. (New York: ElseviedNorth Holland), pp. 363-415.

Miller, J. H. (1972). Experiments in Molecular Genetics. (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory).

Mott, J. E., Galloway, J. L., and Platt, T. (1965). Maturation of fsche- richia co/i tryptophan operon mRNA: evidence for 3’ exonucleolytic processing after rho-dependent termination. EMBO J. 4, 1867-1691.

Nilsson, G., Belasco, J. G., Cohen, S. N., and von Gabain, A. (1964). Growth-rate dependent regulation of mRNA stability in fscherichia co/i. Nature 312, 75-77.

Nomura, M., Gourse, R., and Baughman, G. (1964). Regulation of the synthesis of r ibosomes and ribosomal components. Ann. Rev. Bio- them. 53, 75-117.

Peck, L. J., and Wang, J. C. (1965). Transcriptional block caused by a negative supercoiling induced structural change in an alternating CG sequence. Cell 40, 129-137.

Pfamann. M. D., and Stauffer, G. V. (1985). Characterization of a cis- acting regulatory mutation that maps at the distal end of the fsche- richia co/i g/yA gene. J. Bacterial. 167, 650-654.

Putney, S. D., and Shimmel, I? (1961). An aminoacyl tRNA synthetase binds to a specific DNA sequence and regulates its gene transcription, Nature 291, 632-635.

Reed, R. E., and Atman, S. (1963). Repeated sequences and open reading frames in the 3’flanking region of the gene for the RNA subunit of fschedchia co/i ribonuclease P. Proc. Natl. Acad. Sci. USA 80, 5359-5363.

Rosenberg, M., and Schmeissner, U. (1962). Regulation of gene ex- pression by transcription termination and RNA processing. In Interac- tion of Translational and Transcriptional Controls in the Regulation of Gene Expression, M. Grunberg-Manago and 8. Safer, eds. (Amster- dam: ElsevierlNorth Holland).

Roth, J. R. (1970). Genetic techniques in studies of bacterial metabo- lism. Meth. Enzymol. 77a, 3-35.

Russell, D. T., and Bennett, G. N. (1962). Construction and analysis of

Cell 310

in ViVO activity of f. co/i promoter hybrids and promoter mutants that alter the -35 to -10 spacing. Gene 20, 231-243.

SaintGirons, I., Duchange, N., Cohen, G. N., and Zakin, M. M. (1984). Structure and autoregulation of the rnefJ regulatory gene in E. co/i. J. Biol. Chem. 259, 14282-14288.

Schlessinger, D., Jacobs, K. A., Gupta, R. S., Kano, Y., and Imamoto, F (1977). Decay of individual Eschedchia co/ifrp messenger molecules is sequentially ordered. J. Mol. Biol. 710, 421-439.

Schneider, E., Blundell, M., and Kennell, D. (1978). Translation and mRNA decay. Mol. Gen. Genet. 760, 121-129.

Shen, V., Cynamon. M., Daugherty, B., Kung, H.-F, and Schlessinger, D. (1981). Functional inactivation of /ac a-peptide mRNA by a factor that purifies with Escherichia co/i RNaselll. J. Biol. Chem. 256, 1896-1902.

Shen, V., Imamoto, F., and Schlessinger, D. (1982). RNaselll cleavage of Eschedchia co/i beta-galactosidase and tryptophan operon mRNA. J. Bacterial. 750, 1489-1494.

Stansaens, P, Remaut, E., and Fiers, W. (1986). Inefficient translation causes premature transcription termination in the /acZ gene. Cell 44, 71 I-7I8.

Stern, M. J., Ames, G. F-L., Smith, N. H.. Robinson, E. C., and Higgins, C. F. (1984). Repetitive extragenic palindromic sequences: a major component of the bacterial genome. Cell 37, 1015-1026.

Takata. R., Mukai, T. and Hori, K. (1985). Attenuation and processing of RNA from the prsO-pnp transcription unit of Escherichia co/i. Nucl. Acids Res. 13, 7289-7297

Ueno-Nishio, S., Mango, S., Reitzer, L., and Magasanik, B. (1984). Identification and regulation of the g/nL promoter of the complex glnALG operon of f. co/i. J. Bacterial. 760, 379-384.

Urbanowski, M. L., and Stauffer, G. V. (1985). Nucleotide sequence and biochemical characterization of the meti gene from S. typhimu- rim LT2. Nucl. Acids Res. 13, 673685.

Valentin-Hansen, P, Hammer-Jespersen, K., Boetius, F., and Svend- son, I. (1964). Structure and function of the intercistronic regulatory deoC-de& element of Escherichia co/i K-12. EMBO J. 3, 179-183.

White, B., and Bancroft, F (1982). Cytoplasmic dot hybridization. J. Biol. Chem. 257, 85696572.

Yamamoto, T., and Imamoto. F. (1975). Differential stability of frp mes- senger RNA synthesised originating at the trp promoter and PL pro- moter of lambda trp phage. J. Mol. Biol. 92, 289-305.