Transcription termination at the thr attenuator. Evidence that the adenine residues upstream of the stem and loop structure are not required for termination.

The Escherichia coli thr operon attenuator has a structure similar to other Rho-independent terminators. The DNA sequence immediately 5′ to the termination site is dG+dC-rich and contains a region of dyad symmetry that, when transcribed into RNA, encodes a hairpin structure in the transcript. It also contains a stretch of 9 consecutive dA-dT residues immediately distal to the region of dyad symmetry which encode uridine residues at the 3′ end of the terminated transcript. In addition, the thr attenuator has a stretch of 6 dA-dT residues immediately upstream of the region of dyad symmetry which encode 6 adenines. These adenines could potentially pair with the distal uridines to form a hairpin structure extended by as much as 6 A-U base pairs. In this report we have examined the role of the upstream adenines in transcription termination. We used templates that specify mismatches or create new base pairs in the potential A-U secondary structure of the transcript as well as templates that delete segments of the A residues upstream of the hairpin. We conclude that A-U pairing is not required for efficient transcription termination at the thr attenuator. This conclusion is likely to apply to other Rho-independent terminators that contain hairpin-proximal dA-dT residues.

The Escherichia coli thr operon attenuator has a structure similar to other Rho-independent terminators. The DNA sequence immediately 5 to the termination site is dG؉dC-rich and contains a region of dyad symmetry that, when transcribed into RNA, encodes a hairpin structure in the transcript. It also contains a stretch of 9 consecutive dA-dT residues immediately distal to the region of dyad symmetry which encode uridine residues at the 3 end of the terminated transcript. In addition, the thr attenuator has a stretch of 6 dA-dT residues immediately upstream of the region of dyad symmetry which encode 6 adenines. These adenines could potentially pair with the distal uridines to form a hairpin structure extended by as much as 6 A-U base pairs. In this report we have examined the role of the upstream adenines in transcription termination. We used templates that specify mismatches or create new base pairs in the potential A-U secondary structure of the transcript as well as templates that delete segments of the A residues upstream of the hairpin. We conclude that A-U pairing is not required for efficient transcription termination at the thr attenuator. This conclusion is likely to apply to other Rho-independent terminators that contain hairpin-proximal dA-dT residues.
The control of gene expression in bacteria often occurs at the level of transcription termination. Transcription terminators are found upstream of operons, between genes in an operon, and at the ends of operons (for reviews, see Refs. [1][2][3]. In general, transcription terminators have been divided into two classes: Rho-dependent or Rho-independent, depending upon their requirement for Rho factor in vitro (2).
Rho-independent terminators terminate transcription in vitro in the absence of Rho protein or other factors, and have two common structural characteristics (2). The first is a dGϩdC-rich region of dyad symmetry that encodes a stem-loop or hairpin structure in the nascent mRNA. The second feature is a dAϩdT-rich region of 4 -9 base pairs immediately distal to the region of dyad symmetry within which the transcript terminates. Thus, when RNA polymerase transcribes through a Rho-independent terminator, the transcript forms a hairpin followed by 3Ј uridine residues. Evidence from several labora-tories has shown that both of these features are required for efficient termination (2,3).
The thr attenuator is a Rho-independent terminator that has a dGϩdC-rich region of dyad symmetry that encodes a hairpin containing a stem of 8 base pairs and a loop of 6 bases. The hairpin is followed by a tract of 9 consecutive dA-dT residues and termination predominately occurs at the 7th or 8th uridine in the transcript (4). The importance of the uridine residues was established by Lynn et al. (5), who made deletions that varied the length of the dA-dT tract. They found that the deletion of 1 or 3 dA-dT residues had no effect on transcription termination in vivo or in vitro. The deletion of 4, 5, or 6 dA-dT residues showed a linear decrease in termination efficiency. When 7 or 8 dA-dT residues were deleted, termination was abolished.
The thr attenuator also has a dA-dT tract of 6 bases immediately upstream of the region of dyad symmetry (4). The dA-dT tract encodes 6 A residues in the transcript that could potentially pair with 6 of the hairpin-distal U residues to extend the length and stability of the RNA hairpin. Approximately onethird to one-half of the Rho-independent terminators characterized to date also contain hairpin-proximal dA-dT residues as part of an AANAA sequence (6), which suggests that pairing of A and U residues could play a functional role in transcription termination. In this report, we describe a systematic analysis of the role of the hairpin-proximal A residues in transcription termination at the thr attenuator.

MATERIALS AND METHODS
In Vitro Mutagenesis-The procedure described by Kunkel (7,8) was used for preparation of uracil-containing single-stranded M13mp10-thr DNA templates (9). DNA was isolated from the phage particles by phenol extraction and ethanol precipitation. The "gapped duplex" template formation and primer extension procedure was carried out by the method of Bauer et al. (10). The uracil-containing single-stranded M13mp10-thr derivatives and HaeIII-digested M13mp10 w RF DNA were mixed in 30 l of hybridization buffer (40 mM Tris-HCl (pH 7.5), 100 mM NaCl, 20 mM MgCl 2 , 2 mM ␤-mercaptoethanol) and boiled for 3 min. The mixture was allowed to cool to 85°C before addition of 40 pmol of phosphorylated mutagenic primer. The hybridization mixture was slowly cooled to room temperature. Next, 70 l of primer extension buffer (20 mM Tris-HCl (pH 7.5), 11 mM MgCl 2 , 1 mM ␤-mercaptoethanol, 0.83 mM each 2Ј-deoxyribonucleoside triphosphate and 0.4 mM ATP), which contained 0.1 unit of T4 DNA ligase and 0.5 unit of DNA polymerase (Klenow fragment), was added. The reaction mixture was first incubated on ice for 15 min and shifted to room temperature for 5 min. The reaction mixture was then incubated at 30°C for another 2 h and the reaction stopped by adding 3 l of 0.5 M EDTA. The mixture was used to transform competent JM105 cells (10) and white plaques on LB plates (tryptone (10 g/liter), yeast extract (5 g/liter), and NaCl (10 g/liter)) containing 40 g/ml 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal) were screened. Constructs containing the desired mutation were identified by direct DNA sequence analyses.
Construction of thrA-lacZ Protein Fusion Vectors-The construction of the thrA-lacZ protein fusions is shown in Fig. 1A. Restriction fragments containing the Escherichia coli thr operon regulatory region were generated by digesting the appropriate M13mp10-thr derivatives with HincII and EcoRI endonucleases. The HincII site is upstream of the thr regulatory region and the EcoRI site is in the M13 vector. The 5Јprotruding ends of the DNA fragments were removed by treating the DNA with mung bean nuclease. The reaction products were subjected to electrophoresis on 1% agarose gels. The desired 557-bp 1 DNA fragments containing the whole thr regulatory region, part of the thrA gene and vector sequences were electroeluted from the agarose gel. These blunt-ended DNA fragments were ligated into the unique SmaI site of the pMC1396 vector. Plasmid pMC1396, which is a pBR322 derivative and carries a truncated lacZ segment, was constructed by Casadaban et al. (11). A fragment containing a promoter and an in-frame fusion of the thrA gene and vector sequence to the 5Ј-end of the lacZ coding sequence is required to produce a functional hybrid ␤-galactosidase. The ligation mixtures were transformed into strain MC4100 and spread on LB plates containing ampicillin (50 g/ml) and X-gal (40 g/ml). Colonies containing the desired constructions were identified by their resistance to ampicillin and their blue color on plates containing X-gal.
Construction of thrA-lacZ Fusions with Deletions of Region 1 and 2-The construction of thrA-lacZ fusions that lack the sequence from ϩ60 to ϩ107 of the thr leader region was performed as follows (Fig. 1B). The DNA of M13mp9-AT45GG contains dA to dG and dT to dG substitutions at positions ϩ45 and ϩ46, respectively, which creates a sequence that is recognized by endonuclease HpaII. 2 The 108-bp BstEII-HpaII fragment from M13mp9-AT45GG containing the sequence from -65 to ϩ43 of the thr operon leader region was isolated and purified from a 5% polyacrylamide gel (9).
The DNA sequence from ϩ44 to ϩ110 of the thr leader region was replaced by two synthetic oligodeoxyribonucleotides, one 21-mer 5Ј-CGCATGGAACGCATTAGCTGA-3Ј and the other 23-mer 5Ј-CGCGT-CAGCTAATGCGTTCCATG-3Ј. These two oligomers were annealed at 80°C for 1 min and slowly cooled to room temperature. The hybridized oligomers contained HpaII and MluI half-sites at the ends. The HpaII sites of the BstEII-HpaII fragment of the M13mp9-AT45GG and the annealed deoxyribonucleotides were ligated. The resulting 129-bp fragment was purified and ligated with M13mp11-WT thr vector that had been digested with BstEII and MluI endonucleases. The resultant recombinant clone that contained 2 extra base pairs at the ligation junction and deleted the sequences necessary for the formation of the 1:2-hairpin structure of the thr leader region was isolated and named M13mp11-WT thr ⌬1:2. The HincII-EcoRI fragment of M13mp11-WT thr ⌬1:2 was then subcloned into the pMC1403 vector to construct pMC1403-WT thr ⌬1:2 by following the procedure described above. Plasmids pMC1403 and pMC1396 are analogous to each other, except that pMC1403 contains a unique EcoRI site (11).
To construct the pMC1403 attenuator mutant derivatives with the same deletion in the 1:2 region, the MluI-BamHI fragments from the corresponding pMC1396 derivatives (Fig. 1A) were isolated and ligated with the EcoRI-MluI fragment of pMC1403-WT thr ⌬1:2. The ligation products were then subcloned between the EcoRI and BamHI sites of the pMC1403 vector (see Fig. 1C). These constructions produced plasmids that are isogenic with pMC1403-WT thr ⌬1:2.
Construction of Plasmids Containing Deletions of the dA-dT Tracts-M13mp8 derivatives (5) carrying a nested set of deletions with 1, 3, 6, or 8 dT residues distal to the region of dyad symmetry of the thr attenuator were used as sources of DNA for the constructs. The terminators were isolated as 42-49-base pair RsaI-XbaI fragments. Approximately 100 g of plasmid DNA containing the various dA-dT tract deletions (5) were digested with RsaI and XbaI, and the fragments were separated by electrophoresis in a 10% polyacrlyamide gel. After staining with ethidium bromide and visualization with ultraviolet light, the appropriate segments of gel containing the bands were excised and soaked in 0.5 ml of 0.5 ϫ TE buffer (5 mM Tris-HCl (pH 7.9) and 0.5 mM EDTA) at room temperature overnight. The gel particles were removed by centrifugation and DNA fragments were purified on DE52 columns (5).
The pUC18 and pUC19 DNA (Fig. 2) were digested with XbaI and HincII. The RsaI-XbaI fragments were ligated into pUC18 and pUC19 using T4 DNA ligase. The ligation mixtures were transformed into competent KK2186 cells [⌬ (lac-pro), supE, thi, strA, sbcB-15, endA/FЈ traD36, lacI Q Z⌬15 (12) by the CaCl 2 method (13). The desired clones were identified as white colonies on LB plates supplemented with X-gal, and the sequences of the constructs were verified by direct DNA sequence analysis.
DNA Sequence Analysis-The double-stranded DNA templates used for sequencing were prepared by the method of Kraft et al. (14). Plasmid DNAs of the pMC1396 and pMC1403 derivatives were denatured by sodium hydroxide, and a synthetic 17-mer oligodeoxyribonucleotide 5Ј-ATAGCGCACAGACAGAT-3Ј was used as a primer. This primer hybridizes to the region from positions ϩ3 to ϩ19 of the thr operon leader sequence. The resultant DNA was sequenced by the dideoxy method of Sanger et al. (15) with a modified T7 DNA polymerase, Sequenase™ from U. S. Biochemical Corp. Plasmid DNA from the constructs containing the deletions was sequenced using a Taq Track kit from Promega.
In Vitro Transcription-In vitro transcription reactions were performed with purified RNA polymerase essentially as described by Yang and Gardner (9). The final concentrations of the components of the transcription reaction mixtures were the following: 20 mM Tris acetate (pH 7.9), 0.1 mM Na 2 EDTA, 0.1 mM dithiothreitol, 150 mM KCl, 150 M GTP, 150 M ATP, 150 M UTP, 50 -80 M CTP, and 10 -20 Ci of [␣-32 P]CTP, 1-2 pmol of DNA template, and 1-2 pmol of E. coli RNA polymerase. In some experiments, GTP was replaced by ITP at a concentration of 150 M. Templates carrying the dA-dT tract deletions contain Plac. Reactions using these templates contained 1.25 M cAMP and 0.34 mg of CRP protein.
Transcription reactions (50 l) were terminated after 10 -20 min at 37°C by the addition of an equal volume of phenol, and carrier tRNA was added to a final concentration of 0.5 mg/ml. After the phenol extraction, the aqueous phase was adjusted to 0.3 M sodium acetate and the samples were ethanol-precipitated, desalted, dried in vacuo, and resuspended in formamide dyes (80% formamide, 1 ϫ TBE buffer, 0.1% xylene cyanol, and 0.1% bromphenol blue). The products of the transcription reactions were analyzed on 8% polyacrylamide, 8 M urea sequencing gels. After autoradiography, regions of the gel corresponding to the terminated and read-through transcripts (9) were cut out of the gel and counted in a Beckman LSI801 scintillation counter. The molar ratios of the terminated and read-through transcripts were calculated by correcting for the lengths and base compositions of the transcripts (9).
In Vivo Recombination and Single Prophage Selection-In vivo recombination of thrA-lacZ fusions onto RZ-5 (16) was performed as follows. Strains containing the thrA-lacZ fusion plasmids were grown in LB medium overnight. Cells (0.1 ml) were subcultured into 5 ml of LBMM (tryptone (10 g/liter), yeast extract (5 g/liter), NaCl (10 g/liter), maltose (20 mg/liter), and 10 mM MgSO 4 ) and shaken at 37°C for 2 h. One ml of the cell suspension was infected with 10 l of RZ-5 at a multiplicity of infection of 2-5. The cultures were then grown at 37°C until lysis occurred. Chloroform (50 l) was added, and the cultures were incubated for 15 min. The lysates were collected as the supernatant fraction after centrifugation at 5,000 rpm for 10 min. The phage lysates were diluted 1:100 and 1:1000 into Ca ϩ2 buffer (10 mM Tris-HCl (pH 7.9), 10 mM MgSO 4 , and 5 mM CaCl 2 ), and 0.1 ml of each dilution was added to an equal volume of overnight MC4100 cells. The mixtures were then spread on T-plates (tryptone (10 g/liter), NaCl (5 g/liter), and Bacto-agar (12.5 g/liter)) containing 25 g/ml ampicillin and 40 g/ml X-gal. The transduced MC4100 strains were identified by their resistance to ampicillin and formation of blue colonies.
All the lysogens were verified by their sensitivity to vir and resistance to lysis by h80del9C (17). Lysogenic strains containing only one copy of a prophage were identified by a modified method of Sly and Rabideau (18). Determination of the copy number of the prophage was based on the susceptibility of the lysogen to different dilutions of a cI 90c17 lysate. MC4100 strains with a single prophage were used in assays for ␤-galactosidase activity.
␤-Galactosidase Assays-The ␤-galactosidase assays were performed as described by Miller (17). Bacterial strains were grown in minimal medium (17) supplemented with 0.2% glucose. Each sample was assayed in triplicate, and the specific activities were calculated and averaged. The standard deviations were less than 10%.
Enzymes and Chemicals-Restriction endonucleases, mung bean nuclease, and T4 DNA ligase were purchased from Life Technologies, Inc. and used according to the manufacturer's specifications. Oligonucleotides were synthesized by the University of Illinois Biotechnology Center. [␣-32 P]CTP (Ͼ400 Ci/mmol) was purchased from Amersham Corp. E. coli RNA polymerase and CRP protein were gifts from E. Morgan and J. Harmon, respectively.

RESULTS AND DISCUSSION
Transcription Termination with Templates Containing Substitution Mutations-Like other Rho-independent terminators, the E. coli thr attenuator has a dGϩdC-rich region of dyad symmetry that encodes an RNA hairpin followed by 9 consecutive dA-dT residues that encode 7 or 8 uridine residues at the 3Ј end of the transcript. In addition, there are dA-dT residues that encode adenosines in the transcript immediately preced- ing the region of dyad symmetry. The 3Ј uridine residues are complementary to the stretch of adenosine residues, and the stability and length of the helix would be enhanced by extending the base pairs from the GϩC-rich hairpin to include the 6 A-U base pairs. Such base pairing could be important in promoting transcription termination at terminators that contain this feature by a variety of mechanisms. For example, the increased stability of the hairpin provided by the A-U base pairs could be required for termination or the extended helix could be required as a structural component of the termination signal by interacting with the enzyme. In templates that contain deletions of the upstream dA-dT residues, the stability and length of the hairpin would be reduced resulting in less efficient transcription termination. Alternatively, the formation of the extended helix in the transcript could compete for uridine residues that are paired with the template and disrupt U-dA interactions at the 3Ј end of the transcript to promote termination and dissociation of the enzyme. Thus, when the length of the A-U helix is less than 6 base pairs, the length of the RNA-DNA duplex would increase and possibly reduce transcription termination.
Several other Rho-independent terminators including the phe (19), leu (20), and ilvB (21) attenuators also contain upstream adenosines as part of a conserved AANAA sequence that can also potentially form base pairs with the uridines. In addition, the bidirectional tonB/P14 terminator and ␣ operon terminator, t L17 , contain stretches of adenosines upstream of the hairpin (22,23). If A-U base pairing is important for efficient transcription termination at these terminators, a simple prediction is that disruption of the base pairs, by introducing substitutions upstream of the hairpin or in the hairpin-distal region of the template, should reduce the efficiency of transcription termination. In addition, compensating mutations that restore base pairing should direct efficient transcription termination.
We constructed a series of mutant templates that contained single or double mutations in the thr attenuator and measured their transcription termination efficiencies in vitro and in vivo. The mutants specified potential base pair mismatches or new base pairs in the stretch of A and U residues of the transcript (Fig. 3B). Most of these variants, each containing single or double base changes, were made by oligonucleotide-directed site-specific mutagenesis. L131G is a substitution mutant with an A to G change in the run of A residues immediately 5Ј to the attenuator. Mutants L158C, L160A, and L160G disrupt the U tract by substituting C, A, or G for U at positions ϩ158 and ϩ160, respectively. Mutants L153A (24) and L160C 3 were isolated by a genetic selection that involved isolating mutants that decrease transcription termination at the thr attenuator in vivo. Variants L131U/L160A and L131G/L160C contain different base substitutions at the same positions: A to U or G at position ϩ131 and U to A or C at position ϩ160, respectively. These two variants disrupt the sequences of both the A and U stretches, but maintain the potential complementarity at the base of the stem. Variant L153A/L160C, which contains two single base changes at positions ϩ153 and ϩ160, has disruptions in both the GϩC-rich region and the U tract of the thr attenuator.
The BstEII-SstI restriction fragments bearing the thr operon leader region and different attenuator mutants were isolated from the M13mp10-thr constructs and used for in vitro transcription studies (Fig. 3A). The transcribed RNA was subjected to electrophoresis on 8% polyacrylamide, 8 M urea gels and subjected to autoradiography. The terminated (164 bases) and read-through (300 bases) transcripts were isolated, and the amount of radioactivity in each gel species was measured. The in vitro transcription termination efficiencies of the attenuator variants are presented in Fig. 4 and Table I.
The results showed that variant L153A/L160C, which contains C-A mismatches in both the dGϩdC-rich region and the U stretch of the attenuator, significantly decreased the termination frequency. Variants with substitutions at position ϩ160 only (L160C, L160G, and L160A) showed termination frequencies that were only slightly lower than wild type. All of the remaining variants bearing base substitution(s) that disrupted the runs of A or U residues terminated efficiently.
It has been shown previously (9) that substitution of ITP for GTP caused a decrease in termination frequencies for several other thr attenuator variants. Inclusion of ITP decreased termination at the mutant terminators dramatically but had only a slight effect on termination at the wild type site (9). If ITP caused a similar decrease in the termination frequencies of the mutants constructed in this study, it would be possible to determine the effects of base changes that either disrupted or restored the putative complementarity in the runs of A and U residues of the thr attenuator.
The termination frequencies obtained from ITP-substituted 3 J. Gardner, unpublished results.

FIG. 3. In vitro transcription template and the RNA secondary structure of the thr attenuator and its variants.
A, a BstEII-SstI fragment carrying the thr regulatory region was used for in vitro transcription experiments shown in Fig. 4. The terminated and readthrough transcripts are 164 and 300 bases, respectively. B, the secondary structure of the thr attenuator RNA is presented in the conformation that maximizes base pairing interactions. The positions of the nucleotides in the transcript are numbered starting from the transcription initiation site of the thr leader RNA, which is designated as ϩ1. The mutational changes in the thr attenuator that were generated by in vitro mutagenesis are indicated as bold letters. Variant L160C was originally isolated by a genetic method that showed relief of transcription termination at the thr attenuator in vivo. in vitro transcription experiments are shown in Fig. 4 and Table I. As observed previously, incorporation of ITP in the transcripts showed only a slight effect when the wild-type template was used with a termination value of 65% (9). Variants L160C, L160G, and L160A, which contain disruptions in the U tract by substituting C, G, or A, respectively for U at position ϩ160, showed drastically decreased termination frequencies. Since these three variants have the same GϩC-rich sequences and are expected to form hairpin structures identical to the wild type with one potential mismatch in the A-U region, it is likely that the decrease in the termination values for these variants is caused by effects of the substituted bases at the U stretch, possibly by affecting RNA-DNA template interactions or by direct effects of the sequence changes themselves. This interpretation was further supported by comparing the results from two attenuator variants, L131G/L160C and L131U/ L160A. These two variants were specifically constructed so that the putative base pairing at the base of the stem of the thr attenuator were restored; i.e. I-C and U-A base pairs in L131G/ L160C and L131U/L160A, respectively. If the effect exhibited by the single mutants (L160C or L160A) arose from disruption of the A-U base pairings at the base of the thr attenuator, the two variants would be expected to terminate as efficiently as the wild type. The results show that both of the variants retained termination frequencies similar to the variants with the same change in the U stretch only (L160C or L160A). Taken together, the in vitro transcription studies with ITP suggested that single base changes in the run of A residues have no effect on termination.
To determine the effects of the variants in vivo, the same mutants were subcloned into the vector pMC1403 to construct in-frame thrA-lacZ protein fusions (Fig. 1, B and C). These constructs lack the leader region that encodes the upstream secondary RNA structures that are involved in regulating the formation of the thr terminator structure encoded by the attenuator. These plasmids were constructed to avoid possible complications introduced by the upstream sequences. The fusions were then crossed onto RZ5 by homologous recombination and single-copy lysogens were constructed. The level of ␤-galactosidase expression should only reflect the termination efficiencies of these attenuator variants in vivo. Table II shows the results of ␤-galactosidase assays performed on cells grown in minimal medium. The results show that variant L131G exhibited the same level of ␤-galactosidase activity as wild type. This result could be explained by the formation of a G-U base pair in the helix. However, variants which only disrupted the U tract showed 3.7-4.0-fold (L160C and L158C) and 7.5-7.9-fold (L160A and L160G) higher ␤-galactosidase activities than that of wild type. Furthermore, the ␤-galactosidase activities of the two variants bearing base changes that could potentially form base pairs in the transcript at both the runs of A and U were 3.1 times (L131G/L160C) and 5.9 times (L131U/L160A) the wild-type value. The results again suggested that the restoration of complementarity at the base of stem cannot compensate for the effects of a single substitution mutation in the U stretch. In vivo ␤-galactosidase assays of the variants that contain an intact thr leader region also gave similar results (data not shown).
In Vitro Transcription with Templates Containing Deletions of the Hairpin-proximal and Hairpin-distal dA-dT Tracts-A second prediction of the model proposing that A-U base pairing is important for transcription termination is that deletions of the adenosine residues upstream of the hairpin would decrease the length of the helix and, consequently, decrease the efficiency of transcription termination. We constructed templates that encoded varying numbers of adenosine residues without   changing the sequence of the hairpin and downstream uridine residues. RsaI-XbaI fragments (42-49 bp) were isolated from several deletion mutants previously constructed by Lynn et al. (5) that contained a series of successive deletions of the hairpindistal dA-dT tract. These fragments contain the threonine attenuator plus approximately 15 bp upstream of it. The fragments were cloned into pUC18 and pUC19 at the HincII and XbaI sites (Fig. 2). The attenuator was in the wild-type orientation, with respect to the direction of transcription from the lac promoter in pUC19 and in the inverse orientation in pUC18. The number of dA-dT residues distal to the region of dyad symmetry varied in the pUCl9 clones and the number of dA-dT residues upstream of the region of dyad symmetry varied in the pUC18 clones (Fig. 2). Since the dGϩdC-rich region is an inverted repeat, the RNA hairpin encoded by it had the same sequence in either orientation. The sequences between the inverted repeats that encode the loop of the RNA hairpin and the sequences that encode RNA upstream of the dA-dT residues preceding the hairpin and nontranscribed sequences downstream of the hairpin-distal dA-dT residues were different in the two orientations. However, these sequences were constant for each series of constructs. A nomenclature was established to distinguish the various clones. This nomenclature was based on the number of upstream and distal dA-dT residues and the orientation: (ϩ) for wild type and (Ϫ) for the inverse. Thus, A6-T6(ϩ) had 6 dA-dT residues upstream of the hairpin and 6 dA-dT residues downstream of the hairpin and was in the same orientation as in the wild-type thr operon. A8-T6(Ϫ) had 8 dA-dT residues upstream of the hairpin and 6 hairpin distal dA-dT residues and was in the opposite orientation as in the wild-type thr operon. PvuII fragments varying in size from 364 to 371 bp were purified from the clones and used as templates in in vitro transcription reactions (Fig. 5). Using templates with the (ϩ) orientation, the termination efficiency decreased from 61% for A6-T8(ϩ) to 7% for A6-Tl(ϩ) as the number of uridine residues in the transcript decreased from 8 to 1 as observed previously (5). In contrast, with templates of the (Ϫ) orientation, in which the number of adenosine residues upstream of the hairpin were varied, termination was efficient and varied from 64% to 78%. Since transcription termination is efficient even on a template that contains a deletion that removes 5 of the 6 adenosine residues upstream of the hairpin, these results also indicate that A-U base pairing is not important for efficient transcription termination.
In summary, we have constructed variants of the thr attenuator to determine if the tract of 6 dA-dT residues upstream of the sequence encoding the dGϩdC-rich hairpin are important for transcription termination. Both the in vitro and in vivo results indicate that an intact dA-dT tract is not essential for efficient transcription termination. These results argue against a model that proposes that pairing between the adenosine residues upstream of the hairpin and downstream uridine residues is necessary for transcription termination. In addition, the dA-dT tract upstream of the hairpin does not contribute essential structural or sequence information because deletion of 5 of the 6 A residues from the transcript does not affect the efficiency of termination. The results are compatible with current models that propose that termination is a multi-step process involving active participation of the RNA polymerase (3,25,26) or that termination is controlled by the relative stabilities of DNA-DNA, RNA hairpin, and DNA-RNA interactions at the termination site in the transcription bubble (2,27).
It is interesting to note that a study by Wright et al. (23) with the t L17 terminator has shown that a deletion of the tract of 7 adenosines upstream of the dGϩdC-rich element increases the frequency of read-through at the terminator by a factor of 10 in vivo. They concluded that pairing between the adenosines in the A-tract and the uridines in the distal U tract is important for termination. We have no obvious explanation for the differences observed between the thr and t L17 terminators. As discussed by Wright et al. (23) it is possible that, as a consequence of the deletion of the A tract, the formation of an alternative stem-and-loop structure could occur. It would contain 4 instead of 5 G-C base pairs in the stem, and the loop would be 6 rather than 4 bases. It is also possible that the length of the helix in the stem of the RNA could be important in determining the requirement for pairing between the A tract and the distal uridines. The thr and t L17 stems are 8 and 5 base pairs in length, respectively. Perhaps the t L17 terminator requires A-U base pairs in addition to the G-C base pairs to extend the length of the stem helix in order to act as an efficient terminator. Additional systematic studies will be required to determine if either of these explanations is correct.