Rho-dependent transcriptional polarity in the ilvGMEDA operon of wild-type Escherichia coli K12.

It has been generally accepted that transcriptional polarity in prokaryotic systems is due to an uncoupling of translation and transcription which unmasks latent rho-dependent termination sites in a polycistronic messenger RNA. In this report, we identify and characterize rho-dependent termination sites responsible for transcriptional polarity in the ilvGMEDA operon of wild-type Escherichia coli K12. The ilvG gene in the wild-type E. coli K12 ilvGMEDA operon contains a frameshift site which results in termination of translation in the middle of the gene. Mutations have been characterized which restore the reading frame of this gene. In addition to allowing full-length expression of the ilvG product, these mutations cause a 3-4-fold elevation in the expression of the operon distal genes. This transcriptional polarity effect on operon distal genes also has been shown to be relieved by rho suppressor mutations. We have used in vitro transcription experiments to identify rho-dependent transcriptional termination sites downstream of the frameshift site in the ilvG gene. Three tandem rho-dependent sites have been located in the ilv'GM' gene region using transcription reactions containing linear or supercoiled plasmid DNA templates. Accumulatively, these rho-dependent termination sites account for about 80% in vitro transcription termination, which is in agreement with the in vivo measurements of transcriptional polarity on operon distal gene expression. These transcriptional experiments provide in vitro confirmation for the latent rho-dependent termination site model of transcriptional polarity.

In prokaryotic systems, the premature termination of translation of a polycistronic messenger RNA can reduce transcription of operon distal genes. This phenomenon, transcriptional polarity, is the result of any mutation that causes termination of translation within a proximal gene of an operon (for review, see Refs. 1-4). Mutations in the transcriptional termination factor rho have been shown to alleviate this polar effect on distal gene expression (5)(6)(7)(8). In order to explain transcriptional polarity, it has been proposed that rho-dependent transcriptional termination sites exist in operons, but these sites are not recognized by rho factor when transcription and * This work was supported in part by Grant GM24330 from the National Institutes of Health. 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.
The nucleotide sequence(s) reported in thispaper has been submitted to the GenBankTM/EMBL Data Bank with accession number(s) 503454.
$ To whom correspondence should be addressed.
translation are tightly coupled. It is further presumed that the uncoupling of transcription and translation by premature translational termination unmasks these previously latent rho-dependent termination sites and that termination at these sites explains transcriptional polarity (1)(2)(3).
An example of transcriptional polarity is seen in the ilu-GMEDA operon of wild-type Escherichia coli K12. The first two genes of this operon, iluG and iluM, encode the large and small subunits, respectively, of acetohydroxy-acid synthase 11. This enzyme, which catalyzes the first step of the parallel pathway for the biosynthesis of the branched-chain amino acids, isoleucine, valine and leucine, is one of three isozymes present in E. coli (9,10). One of the differences distinguishing these isozymes is the sensitivity to end-product inhibition by valine. Acetohydroxy-acid synthases I and 111 are inhibited by valine, whereas the acetohydroxy-acid synthase I1 isozyme is not. Lawther et al. (11) demonstrated that the iluG gene in wild-type E. coli K12 contains a frameshift site which results in termination of translation approximately in the middle of the gene. This explains why E. coli K12 is growth-inhibited by exogenous valine, that is, the remaining two active acetohydroxy-acid synthase isozymes are both inhibited by valine, which blocks the synthesis of isoleucine. A secondary explanation for growth inhibition in the presence of valine is provided by the action of the sulfonylurea herbicide sulfometuron methyl, which inhibits the growth of bacterial cells by inhibiting acetohydroxy-acid synthase enzymatic activity (12). This inhibition results in a toxic accumulation of the acetohydroxy-acid synthase enzyme substrate a-ketobutyrate (13). By analogy, the end-product inhibition by exogenous valine of the two acetohydroxy-acid synthase isozymes (acetohydroxy-acid synthases I and 111) expressed in E. coli K12 could result in a toxic accumulation of this pathway intermediate.
Several E. coli K12 mutations have been isolated which are resistant to growth inhibition by exogenous valine. In each case which has been characterized, there is a 1-base pair deletion or a 2-base pair insertion within a 10-base pair region preceding the frameshift site in the iluG gene (11,14). Each mutation allows for the full-length expression of the ilvG product and therefore functional acetohydroxy-acid synthase I1 activity. In the valine-resistant strains, the specific activities of transaminase B and threonine deaminase, the products of the operon distal iluE and iluA genes, respectively, are elevated 3-4-fold compared to wild-type E. coli K12 (15, 16).
Isogenic strains containing rho mutations also show an increase in distal gene expression as compared to the wild-type E. coli K12 strain (15,16). Therefore, both types of mutations appear to relieve classical transcriptional polarity effects on distal gene expression in the ilvGMEDA operon.
In this report, additional aspects of transcriptional polarity in the ilvGMEDA operon are examined. In vivo analyses of the effects of the valine-resistant ilvG468 mutation and rho221 suppressor mutation in isogenic strains show that whereas either mutation alone increases the expression of the distal ilvE gene product, the two mutations together do not have an additive effect. This is consistent with both mutations relieving a common transcriptional polarity event. I n vitro transcription experiments using DNA templates containing the tuc promoter fused to a restriction fragment downstream of the iluC frameshift polarity suppression site identify three tandem rho-dependent termination sites encoded in this region. These sites, like other rho-dependent termination sites, are sensitive to different salt concentrations and require sequences upstream of the site of transcription termination. The transcription experiments provide in vitro confirmation for the latent rho-dependent termination site model of transcriptional polarity.

EXPERIMENTAL PROCEDURES
Materials-Radiochemicals were purchased from Amersham Corp. All enzymes were from Boehringer Mannheim or New England Biolabs. Rho factor was a gift from Terry Platt (Department of Biochemistry, University of Rochester Medical Center, Rochester, NY).
Plasmid Constructions-An 807-bp' PuuII-Sal1 restriction endonuclease fragment containing the ilu'GM gene region ( Fig. 1) was isolated from plasmid pAH29 (11) and inserted into the SmaZ to SalI sites of plasmid pUC8 (18). From this plasmid construct, a 825-bp EcoRI-Hind111 fragment containing the ilu'GM' insert and the flanking polylinker region of plasmid pUC8 was isolated and inserted into the EcoRI to HindIII sites of plasmid pKK223-3 (19). These two restriction sites in plasmid pKK223-3 are located between the tac promoter and the rrnB T1T2 termination sites. The resulting plasmid pJSPl contains the truncated ilu'G gene region of the insert adjacent to the tac promoter and the iluM region adjacent to the rrnB T1T2 terminators.
Plasmids pJSP30 and pJSP31 were constructed in a similar fashion. A 503-bp AluI-Sal1 restriction endonuclease fragment containing the ilu'GM gene region ( Fig. 1) was isolated from plasmid pAH29 and inserted into the SalI site of plasmid pUC8. From plasmid constructs with both orientations of the insert, 531-bp EcoRI-Hind111 restriction endonuclease fragments containing the ilv'GM' insert and the flanking polylinker region of plasmid pUC8 were inserted into the EcoRI to HindIII sites of plasmid pKK223-3. The resulting plasmid pJSP30 contains the truncated ilv'G gene region of the insert adjacent to the tac promoter and the iluM' region adjacent to the rrnB T1T2 terminators. Plasmid pJSP31 contains the insert in the opposite orientation.
Transaminase B Assay-Transaminase B assays were performed by the method of Duggan and Wechsler (20). Bacterial cultures were grown in M63 minimal medium (21) to 100 klett units and collected by centrifugation. The cell pellet from a 20-ml culture was resuspended in a 1.0-ml solution of 100 mM KPO, (pH 7.5) and disrupted by sonication. After centrifugation, a sample of the clear cell lysate was added to a 1.0-ml final volume assay mixture containing 200 pmol of Tris-HC1 (pH 7.8), 0.1 pmol of pyridoxal phosphate, 15 pmol of a-ketogluturate, and 50 pmol of valine and incubated at 37 "C for 15 min. The reaction was terminated by the addition of 1.0 ml of 0.3% 2,4-dinitrophenylhydrazine in 2 N HCl. The sample was incubated for 15 min at room temperature, 2.0 ml of toluene was added, and the mixture was vortexed for 3 min. The aqueous and organic phases were separated by centrifugation, and 1.0 ml of the toluene layer was removed and mixed with 5.0 ml of 10% Na2C03. Following centrifugation, 2.0 ml of the Na2C03 layer was removed and mixed with 2.0 ml of 1.5 N NaOH. The absorbance was measured at 540 nm following a 10-min incubation at room temperature. The protein ' The abbreviation used is: bp, base pair. concentration of the cell extracts was determined by the method of Lowry et al. (22).
Purification of DNA-Cesium chloride band-purified plasmid DNA was prepared by standard methods (23). Linear DNA restriction fragments were separated by electrophoresis on 5% polyacrylamide gels (19:l ratio of acrylamide to N,N"bisacrylylcystamine) buffered with 50 mM Tris borate (pH 8.3) and 1 mM EDTA (TBE). The polyacrylamide gels were stained with ethidium bromide; and restriction fragment bands were dissected from the gels, dissolved in 20% P-mercaptoethanol, and purified on DE52 columns.
Zn Vitro Transcriptions-Purified supercoiled plasmid DNA or linear restriction fragment DNA was used as a template in the in vitro transcription reactions. A 9-pl solution containing 0.1 pmol of DNA template, 40 mM Tris acetate (pH 7.9), 4 mM magnesium acetate, 0.1 mM dithiothreitol, 0.1 mM EDTA, 200 p~ ATP, 20 p~ GTP, 10 pCi of [LX-~'P]GTP (800 Ci/mmol), and 50 units/ml RNA polymerase (0.9 pmol) was incubated for several minutes at 37 "C. The KC1 concentration was 50 mM unless otherwise noted. Where indicated, 0.1 pg of rho factor (0.4 pmol) was added per reaction. The transcription reactions were initiated with the addition of 200 p~ CTP and 200 p M UTP. After a 20-min incubation at 37 "C, a phenol/ chloroform extraction was performed, and 15 pg of E. coli tRNA was added to the sample, followed by ethanol precipitation. The dried pellets were dissolved in a solution of 4 M urea, 0.05% sodium dodecyl sulfate, 0.125% bromphenol blue, and 0.125% xylene cyano1 blue; heated for 1-2 min at 95 "C, and analyzed by electrophoresis on a 6% polyacrylamide denaturing gel containing 8 M urea buffered with TBE. The gel was fixed in a solution of 10% methanol and 10% acetic acid, dried under vacuum, and exposed to Kodak XRP-5 film. Plasmid pUC19 (24) digested with restriction endonuclease Sau3Al was endfilled at the recessed 3'-end with [a-32P]dGTP using the Klenow fragment of E. coli DNA polymerase I (23). The 32P-labeled DNA fragments were used as markers on the polyacrylamide/urea gels. These markers were calibrated with RNA transcripts of known length to account for electrophoretic differences between RNA and DNA molecules. The percentage of transcription termination at each site was quantitated by dissecting transcript bands from the polyacrylamide/urea gels and measuring radioactivity by liquid scintillation spectroscopy. These measurements were normalized for the guanine content of each transcript.
Computer Analysis of RNA Secondary Structure-Analysis of RNA secondary structure of the ilu'GM gene region was conducted using the FOLD program (25) contained in the University of Wisconsin Genetics Computer Group program library (26). To estimate the temporal nature of RNA folding in the nascent transcript, overlapping nucleotide sequences of 300 and 100 nucleotides in length were examined for secondary structure of minimum free energy for an RNA molecule. Using a composite of these data, we predicted an RNA secondary structure for the nucleotide sequences in the 807-bp PuuII-Sal1 restriction endonuclease fragment ( Fig. 1).

RESULTS
The ilvG gene in the ilvGMEDA operon of wild-type E. coli K12 contains a frameshift site which results in termination of the encoded product approximately in the middle of the gene (11,14). Several mutants have been characterized which produce a full-length iluG product. One example is the strain T31-4-590, which contains the ilvG468 mutation, a single adenine deletion upstream of the frameshift site ( Fig. 1) (14).
In addition to restoring acetohydroxy-acid synthase I1 activity, this mutation results in a 5.4-fold increase in the operon distal iluE gene expression compared to the isogenic wild-type E. coli K12 strain T31-4-4 (Table I). This polar effect on downstream gene expression is also alleviated by rho suppressor mutations. The rho221 mutation encodes a defective rho protein with a larger molecular weight than wild-type as judged by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (7). A wild-type E. coli K12 strain containing the rho221 mutation shows a 5.2-fold elevation in ilvE expression compared to the isogenic wild-type strain (Table I). Although E. coli strains containing either mutation alone increase the iluE-encoded transaminase B activity above that in the wildtype isogenic strain, the two mutations together in strain IH-Transcriptional Polarity in the ilvGMEDA Operon  57 do not increase the distal gene expression of the iluGMEDA operon greater than that which is due to only one mutation ( Table I). This is consistent with both mutations relieving a common transcriptional polarity event.
T o characterize directly the effects of rho factor in transcriptional polarity, we performed in vitro transcription reactions using DNA templates containing the strong tac promoter fused to restriction fragments downstream of the iluG frameshift site. An 807-bp PuuII-Sal1 restriction fragment containing the ilu'GM' gene region (Fig. 1) was inserted into the polylinker region of plasmid pKK223-3 between the tac promoter and the rrnB T I T 2 termination sites. The resulting plasmid pJSPl contains this restriction fragment orientated such that the truncated ilu'G gene region of the insert is adjacent to the tac promoter. A 1080-bp BamHI-Hind111 restriction fragment containing this transcriptional fusion was purified from plasmid pJSP1. The BarnHI restriction site is located about 180 bp upstream of the tac promoter, and the HindIII site is located between the ilu'GM' gene insert and the rrnB T1T2 termination sites. Since many rho-dependent termination sites have been shown to be salt-sensitive in in vitro transcription reactions (2,27,28), we used a 25-150 mM KC1 concentration range in transcription reactions containing this linear BamHI-Hind111 restriction fragment template. As shown in the autoradiogram in Fig. 2 (lanes 2-8), the titration in KC1 concentration affected the number and intensity of the rho-dependent terminated transcripts. In the presence of rho factor a t 50 mM KC1 (lane 3 ) , the 865-nucleotide readthrough transcript indicated by the arrowhead (transcript synthesized from the tac promoter to the end of the linear template) is faint, whereas a t least three smaller transcripts are visible. These transcripts are not seen in transcription reactions in the absence of rho factor (lanes I and 8). As the KC1 concentration is increased, the smaller transcripts are no longer visible, whereas the rho-dependent terminated transcripts labeled 2-3 are retained (lanes 4 and 5). At a concentration of 150 mM KCl, the rho-dependent transcripts are diminished, and only the read-through transcript is visible (lane 7).
In vitro transcription reactions using the supercoiled plasmid pJSPl as template were also carried out in the presence and absence of rho factor (Fig. 2, lanes 9 and IO, and Fig. 3,   lanes 3 and 4 ) . In this supercoiled template reaction, RNA polymerase initiates transcription at the tac promoter, and any transcripts which read through the ilv'CM' insert terminate transcription at the downstream rho-independent rrnB T1T2 sites. An autoradiogram of products from the in vitro transcription reactions performed a t 50 mM KC1 in the presence of rho factor shows three areas of rho-dependent terminated transcripts (Fig. 2, lane 9, and Fig. 3, lane 4 ) . These three areas of transcripts co-migrate with the rhodependent terminated transcripts seen in reactions using the linear restriction fragment as template (Fig. 2). In the absence of rho factor, a 1110-nucleotide transcript indicated by the arrowhead is visible (Fig. 2, lane 10, and Fig. 3, lane 3 ) . This transcript is a result of rho-independent termination at the rrnB T1 site. A transcript about 175 nucleotides larger than the T1 terminated transcript is also detected (Fig. 3, lane 3 ) . This larger transcript band, which has much less intensity, is a result of transcription termination at the rrnB T2 site (29). In these in vitro transcription experiments using the tac promoter, we find that greater than 90% of the rho-independent transcription termination occurs at the rrnB T1 site. I n vitro transcription reactions containing the parental plasmid pKK223-3 show only a single rho-independent terminated transcript (Fig. 3, lanes I and 2). Therefore, a t least three tandem rho-dependent termination sites, as defined by in vitro transcription experiments, are located in the ilv'GM' gene region of the ilvCMEDA operon of E. coli K12.
The nucleotide sequence containing the three tandem rho-  in lanes 2,4,6 and 8. Read-through transcript bands which result from transcription from the tac promoter terminating at the rrnR T1 site are indicated by arrowheads along with the corresponding transcript site in nucleotides. Transcript bands resulting from transcriptional termination at the major rho-dependent sites, 1-3, are indicated. dependent termination sites is shown in Fig. 4. The centers of the more efficient termination sites 1 and 2 are located in the ilvM gene, whereas the weakest site 3 is centered in the distal portion of the ilvC gene. It is noted that termination sites 2 and 3 migrate as doublets on the polyacrylamide/urea gel, and the sites of termination indicated in Fig. 4 are located in the approximate center of the transcript doublet.
The efficiency of the four termination sites in the in uitro transcription reaction containing the supercoiled plasmid pJSPl template in the presence of rho factor was quantitated by dissecting the transcript bands from the polyacrylamide/ urea gels and measuring radioactivity by liquid scintillation spectroscopy. About 23% of the transcripts read through the rho-dependent termination sites. This is in agreement with the transaminase B assay of transcriptional polarity, where there was a 5-fold difference between the wild-type T31-4-4 strain and the isogenic strain containing the rho221 suppressor mutation ( Table I). The percentages of transcripts terminated at the three tandem sites show a gradient effect, with the greater termination at the downstream sites. These termination percentages are as follows: site 1, 37%; site 2, 24%; and site 3, 16%.
T o determine how much upstream sequence is required for The iluC468 mutation is a deletion of an adenosine residue a t position 1250 (11,14). This deletion results in the translation of the iluC product to position 1917. Several other mutations have been isolated which allow the full-length expression of the iluC product. For example, iluC268 is an adenosine deletion a t position 1245, iluG671 is a ribosylthymine deletion of position 1248, iluC209.5 is a ribosylthymine deletion a t position 1254, and iluC2096 is a TA insertion between positions 1251 and 1252 (14). The nucleotide sequence of the three tandem rho-dependent termination sites is underlined with the site number listed below. Nucleotide sequences containing a single mismatch with the consensus sequence CAATCAA are indicated in brackets. The iluG termination codon of the full-length iluC product is indicated with three asterisks. The overlapping iluM initiation codon is underlined. Upstream of the rho-dependent transcription termination sites a t positions 1654-1761 are 10 cytidine residues spaced by 11 or 12 nucleotides. Each spaced cytidine residue is indicated by an asterisk below the nucleotide. Regions lacking predicted secondary structure are underlined with dotted lines. rho-dependent termination at the three tandem sites, a 503bp AluI-Sal1 restriction fragment containing the ilv'GM region ( Fig. 1) was inserted into plasmid pKK223-3 between the tac promoter and the rrnI3 T1T2 termination sites. The resulting plasmid pJSP30 contains the restriction fragment orientated such that the truncated ilv'G gene region of the insert is adjacent to the tac promoter, whereas plasmid pJSP31 contains the restriction fragment insert in the opposite orientation, with the truncated ilvM' gene region adjacent to the tac promoter. I n vitro transcription reactions were performed using these supercoiled plasmid templates in the presence and absence of rho factor. An autoradiogram of the products resulting from the transcription reactions containing plasmid pJSP30 shows only two rho-dependent terminated transcripts (Fig. 3, lanes 5 and 6). These transcripts correspond to downstream termination sites 1 and 2. Thus, it would appear that greater than 290 nucleotides is required upstream of termination site 3 for rho factor to recognize this site of termination, whereas sequences 360 and 400 nucleotides upstream of terminations sites 2 and 1, respectively, are sufficient for rho-dependent activity. I n vitro transcription reactions containing plasmid pJSP31, which contains the ilv'GM' insert in the opposite orientation, show transcripts which terminated only at the rho-independent rrnB T1T2 termination sites (Fig. 3, lanes 7 and 8 ) . The major 820nucleotide T1 terminated transcript is indicated by the arrowhead. This is consistent with rho factor requiring recognition sequences upstream of the termination sites for activity.

DISCUSSION
We have used in vitro transcription experiments to understand the molecular basis of transcriptional polarity downstream of the ilvG frameshift site in the ilvGMEDA operon of wild type E. coli K12. Three tandem rho-dependent termination sites have been detected in transcription reactions containing the supercoiled plasmid pJSPl template (Fig. 2). These sites in the 3'-end of the ilvG gene and the 5'-end of the ilvM gene are located between 615 and 740 bp downstream of the ilvG frameshift site (Figs. 1 and 4). These sites show a gradient effect in the percentage of i n vitro rho-dependent termination, with downstream termination site 1 accounting for the greatest percentage (37%) of transcription termination. Accumulatively, these rho-dependent termination sites account for about 80% of in vitro transcription termination. This is in agreement with the in vivo measurements of transcriptional polarity on operon distal gene expression, where the iluE-encoded transaminase B activity increased about 5fold in strains containing either the rho221 suppressor mutation or the ilvG468 mutation compared to the isogenic wildtype E. coli K12 strain (Table I).
Currently, it, is not certain to what extent the relative contribution of the in vitro rho-dependent termination sites in the ilu'GM gene region reflects in vivo termination. I n vitro transcription reactions using a linear DNA template containing the strong tac promoter fused to this ilv'GM' gene region showed that the number and intensity of the rhodependent terminated transcripts were differentially affected by increasing concentrations of KC1. At 50 mM KC1, maximal rho activity was observed (Fig. 2). As the concentration of salt was increased (75-100 mM), the smaller molecular weight rho-dependent terminated transcripts were diminished; whereas at 125 mM KC1, the rho-dependent terminated transcripts 1-3 were similarly affected. Previous investigators (2,27,28) have noted that the in vitro termination activity of rho factor is inhibited at high salt concentrations on many linear DNA templates, with a maximal activity generally observed at approximately 50 mM KCl. A proposed explanation for these observations is that RNA secondary structure, which is inhibitory to rho action, is stabilized by high ionic concentrations (2). It is interesting to note, however, that in vitro transcription reactions using the supercoiled plasmid pJSP1 at 50 mM KC1 showed three tandem rho-dependent terminated transcripts (Figs. 2 and 3). The lower molecular weight rho-dependent terminated transcripts seen in transcription reactions using the linear DNA template at this concentration of KC1 were not visible. The transcripts encoded from linear or supercoiled DNA template are identical; and thus, RNA secondary structure would not be expected to differ. An alternative explanation for the effects of increased salt concentrations on rho-dependent transcription termination may reflect a difference in energy required to release the nascent transcript. Perhaps, greater energy is required to release rho-dependent terminated products from supercoiled DNA templates than from linear templates. This is consistent with the observation that products from in vitro transcription reactions containing the supercoiled plasmid pJSPl in the presence of rho factor are not affected by modest changes (50-100 mM) in KC1 concentration. In any case, we interpret these results to indicate that products from i n vitro transcription reactions containing the supercoiled plasmid template may reflect more closely in vivo rho-dependent termination activity.
The nucleotide sequences important for rho-directed termination are located upstream of the site of transcription termination. Deletion analyses of DNA sequences upstream of the rho-dependent trp t' termination site show that sequences 50-100 nucleotides upstream of this site are sufficient for in vivo and in vitro rho activity, regardless of the sequence content of the downstream termination site (30). In the four tandem XtRl rho-dependent terminators, RNA sequences 90-130 nucleotides upstream of each termination site are critical for rho recognition (31). In this study, nucleotide sequences encoded upstream of the A h 1 restriction site (Fig. 1) are critical for in vitro rho-dependent termination at site 3 (Fig.   4). This indicates that sequences considerably further upstream of a rho-dependent termination site than previously seen (>290 nucleotides) are required for rho recognition at termination site 3. This sequence is located downstream of the ilvG termination codon in wild-type E. coli K12 and would be free of translating ribosomes and available for rho factor recognition. Termination sites 1 and 2 are located in the ilvM coding region, about 75 and 25 nucleotides, respectively, downstream of the ilvM initiation codon. By analogy to the trp t' and XtRl rho-dependent termination examples, both sites would require RNA sequences encoded in the upstream ilvG gene for rho activity (Fig. 4). Again, these sequences in strains containing the ilvG frameshift site would be free of translating ribosomes. It is possible that a ribosome initiating translation at the ilvM start codon could interfere with rhodirected termination at these major in vitro termination sites. However, it has been proposed that the iluM gene is translationally coupled to the ilvG gene (32,331. The ilvG termination codon overlaps the ilvM initiation codon by four nucleotides in E. coli K12 (Fig. 4). This overlap is thought to ensure the equimolar expression of the large and small acetohydroxyacid synthase I1 subunits, encoded by the ilvG and ilvM genes, respectively. Therefore, premature translation termination in the ilvG gene is predicted to inhibit ilvM translation.
An intensive effort has been made to distinguish a consensus rho factor recognition sequence upstream of the transcription termination site (4,34,35). One hypothesis suggested that a stem-loop structure and the sequence CAAUCAA up-in the ilvGMEDA Operon 15261 stream of the AtRl termination sites could signal rho-dependent transcription termination (34). However, in vitro transcription reactions containing a synthetic DNA template encoding this sequence failed to show detectable rho activity (34). More recent models (4,35) have included the following characteristics for rho-binding sites on nascent RNA transcripts: 1) a region located within a few hundred nucleotides upstream of the rho-dependent termination site, 2) a region spanning 70-80 nucleotides in length, and 3) a sequence void of stable secondary structure. The importance of singlestranded RNA regions in rho recognition has recently been demonstrated by Chen et al. (36). DNA oligonucleotides complementary to two regions upstream of the termination site were shown to impede rho directed termination of the XtRl transcript.
The three tandem rho-dependent termination sites in the ilu'GM' gene region share some of these structural features.
As mentioned, the consensus sequence CAAUCAA is found adjacent to several rho-dependent termination sites (34). Interestingly, this sequence with a single mismatch at the fifth or sixth position is found adjacent to each of the three termination sites (Fig. 4). These are the only examples of this sequence with complete identity or a single mismatch found in the 807-bp PuuII-Sal1 restriction fragment which contains the ilv'GM' gene region. The sequence of the ilv'GM' region has also been examined for RNA secondary structure. Several regions have been identified which contain no dyad symmetry. These include sequences relative to the in vivo ilvGMEDA  . 4) (33). Finally, it has been noted that certain rho-dependent termination regions, including a region putatively important in rho-dependent transcriptional polarity in the his operon in Salmonella typhimurium, contain a regular spacing of cytidines (12 f 1) (4). It is postulated that these sequences could be involved in nucleating rho binding to its recognition site. In the distal portion of the ilvG gene, upstream of the three tandem rho-dependent termination sites, there is a region which contains 10 cytidine residues spaced by 11 or 12 nucleotides. This spacing of cytidines is indicated by asterisk in Fig. 4. This 108-bp region of cytidine symmetry has 18.5% guanosine composition and overlaps a region containing no predicted secondary structure. Additionally, there are regions upstream of the AluI restriction site, which is shown by the deletion experiment to be important for rhodependent termination at site 3, that contain spaced cytidines. Six cytidines spaced at regular intervals (12 f 1) are identified between positions 1415-1476 and 1499-1557 (33).
In conclusion, we have used in vitro transcription experiments to characterize rho-dependent termination downstream of the ilvG frameshift site in the ilvGMEDA operon of wildtype E. coli K12. Three tandem rho-dependent termination sites have been identified using a supercoiled plasmid template. These results provide confirmatory evidence for the latent rho-dependent termination site model of transcriptional polarity in the ilvGMEDA operon of E. coli K12, that is, premature translation termination in the ilvG operon unmasks downstream rho-dependent termination sites, which results in decreased expression of the operon distal genes.