The pyrBI attenuator of Escherichia coli is an intrinsic transcription terminator composed of DNA with a hyphenated dyad symmetry and an adjacent 8 bp T:A tract (T-tract). These elements specify a G+C-rich terminator hairpin followed by a run of eight uridine residues (U-tract) in the RNA transcript. In this study, we examined the effects on in vivo transcription termination of systematic base substitutions in the T/U-tract of the pyrBI attenuator. We found that these substitutions diminished transcription termination efficiency to varying extents, depending on the nature and position of the substitution. In general, substitutions closer to the dyad symmetry/terminator hairpin exhibited the most significant effects. Additionally, we examined the effects on in vivo transcription termination of mutations that insert from 1 to 4 bases between the terminator hairpin and U-tract specified by the pyrBI attenuator. Our results show an inverse relationship between termination efficiency and the number of bases inserted. The effects of the substitution and insertion mutations on termination efficiency at the pyrBI attenuator were also measured in vitro, which corroborated the in vivo results. Our results are discussed in terms of the current models for intrinsic transcription termination and estimating termination efficiencies at intrinsic terminators of other bacteria.
Proper termination of transcription is vital to the cell. In bacteria, termination occurs at the ends of operons to prevent interference with downstream transcription (Richardson and Greenblatt, 1996). Termination also occurs within operons, usually in a regulated fashion, to control the expression of downstream genes (Landick et al., 1996; Henkin and Yanofsky, 2002) and, in some instances, the expression of upstream genes (Gottesman et al., 1982). There are two general types of bacterial transcription terminators that cause the irreversible release of a nascent RNA transcript from its transcription elongation complex. These terminators are called Rho-dependent and intrinsic; the former requires the transcription termination factor Rho, while the latter requires only a short region of DNA and its specified RNA to cause the spontaneous release of the nascent transcript. The basic features of the mechanisms of Rho-dependent and intrinsic termination have been established (Komissarova et al., 2002; Nudler and Gottesman, 2002; Richardson, 2002; Skordalakes and Berger, 2003).
Intrinsic terminators are found in many different bacteria and have been studied extensively in Escherichia coli (Lesnik et al., 2001; de Hoon et al., 2005). Typical intrinsic terminators contain two essential DNA elements: a hyphenated dyad symmetry and an adjacent approximately 8 bp T:A or mostly T:A segment called the thymidine (or T-) tract. These elements specify a G+C-rich hairpin, called the terminator hairpin, and an immediately downstream uridine (U) or U-rich tract in the RNA transcript. Termination typically occurs after addition of a U residue located eight bases downstream of the terminator hairpin. For some intrinsic terminators, particularly those with an atypically short (i.e. ≤ 4 residue) T-tract, the DNA sequence immediately downstream of the termination site can influence termination efficiency (Telesnitsky and Chamberlin, 1989; Reynolds and Chamberlin, 1992). However, even in this case, termination usually occurs eight bases downstream of the terminator hairpin.
The site of termination appears to be determined by the structure of the transcription elongation complex (Darst, 2001; Vassylyev et al., 2007). In this complex, approximately 14 bases at the 3′ end of the nascent RNA lie inside RNA polymerase (Komissarova and Kashlev, 1998; Korzheva et al., 2000). The 3′ end of this 14-base segment is engaged in an 8 to 9 bp hybrid with single-stranded template DNA, while the other 5 to 6 bases occupy the RNA exit channel of the enzyme (Nudler et al., 1997; Komissarova et al., 2002; Kashkina et al., 2006; Vassylyev et al., 2007). The elongation complex is stabilized primarily by the RNA/DNA hybrid and, to a lesser extent, by nucleic acid–RNA polymerase interactions (Komissarova et al., 2002). According to the current model for intrinsic termination, release of the nascent transcript requires the following events. Transcription proceeds until the entire U-tract is included in the RNA/DNA hybrid, which induces a pause in transcript elongation (Gusarov and Nudler, 1999). The terminator hairpin begins to form as the sequence of the downstream segment of the stem starts to emerge from the RNA polymerase exit channel. Formation of the terminator hairpin continues as it invades the RNA polymerase exit channel and incorporates the bases inside. Completion of the terminator hairpin disrupts at least three upstream base pairs of the RNA/DNA hybrid (Komissarova et al., 2002), resulting in a shortened RNA/DNA hybrid composed, at least mostly, of extremely weak rU:dA base pairs (Martin and Tinoco, 1980). This weak hybrid, and perhaps the loss of interactions between the nascent RNA and RNA polymerase, permits the spontaneous release of the RNA from the other components of the elongation complex. In addition, the release of the nascent RNA may be facilitated by the forward translocation of RNA polymerase, the rewinding of the DNA in the transcription bubble, and allosteric changes in RNA polymerase (Ryder and Roberts, 2003; Toulokhonov and Landick, 2003; Santangelo and Roberts, 2004).
Although intrinsic terminators possess the same essential elements, their sequences are different, even in the case of the T-tract (Lesnik et al., 2001). Numerous intrinsic terminators specify a continuous run of 7 to 9 U residues in the U-tract, while many others specify U-rich tracts in which 1 or 2 non-U residues are present. Less often, terminators specify 3 or 4 non-U residues in the U-rich tract. The location of non-U residues in U-rich tracts is not random, apparently reflecting the unique roles in termination for certain positions within the U-tract (Lesnik et al., 2001). These observations suggest that the sequence of a U-tract or U-rich tract of a particular intrinsic terminator reflects a termination efficiency that is appropriate for a specific cellular function. Presumably, non-U residues in U-rich tracts reduce termination efficiency, but the nature of these effects has not been examined systematically. In this report, we describe such an examination using an archetypical intrinsic terminator, the pyrBI attenuator of E. coli.
The pyrBI operon encodes the two subunits of the pyrimidine biosynthetic enzyme aspartate transcarbamoylase. Expression of this operon is regulated over a 50-fold range through UTP-sensitive transcription attenuation and independently over a 7-fold range by UTP-sensitive reiterative transcription at the pyrBI promoter (Turnbough et al., 1983; Liu and Turnbough, 1989; Liu et al., 1994). The pyrBI attenuator, which contains a homogeneous 8 bp T-tract, is an essential element in the attenuation control mechanism. This attenuator is located in the pyrBI leader region, within an open reading frame for a 44-amino-acid leader polypeptide. Within this open reading frame and upstream of the attenuator, the leader region also contains several DNA regions specifying clusters of U residues in the leader transcript. These regions function as UTP-sensitive transcription pause sites (Donahue and Turnbough, 1994). According to the model for attenuation control, low intracellular levels of UTP cause RNA polymerase to pause during the synthesis of the U-rich regions in the pyrBI leader transcript. This pausing provides time for a ribosome to initiate translation of the leader open reading frame and translate up to the stalled RNA polymerase. When RNA polymerase eventually escapes the pause region and transcribes the attenuator, formation of the terminator hairpin is physically blocked by the adjacent translating ribosome (Roland et al., 1988). In the absence of this hairpin, RNA polymerase does not terminate transcription and transcribes through the pyrBI genes. In contrast, when the level of UTP is high, RNA polymerase transcribes the leader region without pausing. In this case, there is insufficient time for a ribosome to establish tight coupling with RNA polymerase before the formation of the terminator hairpin. The result is transcription termination before the pyrBI genes.
In this study, we use attenuation control of pyrBI expression in E. coli as a model system to examine the effects of introducing non-T residues into the T-tract of a strong intrinsic terminator. We constructed strains that carry mutant pyrBI attenuators containing all possible base changes at positions 2 through 8 of the T-tract, and then measured the effects of these changes on termination efficiency in vivo. Our results show that every change reduces termination efficiency, and that the magnitude of this effect depends on the base used in the substitution and its position within the T-tract. We constructed another set of strains with mutations that cause the insertion of 1 to 4 base(s) between the attenuator-specified RNA hairpin and the U-tract. The effects of these insertions demonstrate the importance of the relative positions of the hairpin and U-tract for efficient termination. In addition, we analysed the effects of the same base changes and insertions on termination efficiency at the pyrBI attenuator in vitro, which corroborated our in vivo studies. Finally, we discuss our results in terms of the current model for intrinsic transcription termination and their application to estimating termination efficiencies at intrinsic terminators of other bacteria.
Construction and analysis of mutant pyrBI attenuators
To examine the effects on intrinsic transcription termination of mutations within the T-tract of the pyrBI attenuator, we used site-directed mutagenesis to construct mutant pyrBI attenuators containing base substitutions or insertions. The mutant attenuators were introduced into pyrB::lacZ fusions, which were individually integrated in single copy into the chromosome of strain CLT42 (car-94ΔlacZ). This strain is a pyrimidine auxotroph because the car-94 mutation inactivates the first enzyme in the pyrimidine nucleotide biosynthetic pathway. The pyrB::lacZ fusions were constructed by joining a DNA fragment containing the pyrBI promoter–leader region through the first four codons of the pyrB gene (Fig. 1) to the ninth codon of the lacZ gene. This in-frame fusion permits the level of pyrB::lacZ expression to be monitored by measuring the fusion-encoded β-galactosidase activity. In each construct, a T to G substitution was included in the pyrBI initially transcribed region, at position +3 counting from the transcription start site (Fig. 1). This substitution disrupts a run of three T residues required for reiterative transcription and thereby eliminates UTP-mediated regulation involving this reaction (Liu et al., 1994).
The level of readthrough transcription past mutant pyrBI attenuators, as determined by the level of pyrB::lacZ expression, was measured by growing the mutant strains in a glucose-minimal medium containing uracil as the pyrimidine source. For comparison, an isogenic fusion strain with a wild-type pyrBI attenuator was included in the analysis. Growth on uracil provides a condition of pyrimidine excess and high intracellular levels of UTP. This condition allows the pyrBI attenuator to function maximally, which is with 98% efficiency in the case of the wild-type attenuator (Turnbough et al., 1983; Levin et al., 1989; Liu and Turnbough, 1989). In control experiments to check for mutational effects unrelated to termination at the pyrBI attenuator, the fusion strains were grown in the same minimal medium but with UMP as the pyrimidine source. Growth on UMP, which is only slowly used by cells growing in this medium, causes pyrimidine limitation and low intracellular levels of UTP. This condition results in complete readthrough transcription past both the mutant and wild-type pyrBI attenuators (Liu and Turnbough, 1989).
Effects on transcription termination of single-base substitutions in the T-tract of the pyrBI attenuator
The effects on intrinsic transcription termination of base substitutions within the T-tract of the pyrBI attenuator were determined by using a set of 21 pyrB::lacZ fusion strains containing a separate T to G, C or A substitution at positions 2 through 8 in the T-tract. Equivalent changes are reflected in the U-tracts specified by these attenuators. The numbering system used to identify positions in the T-tract and U-tract is shown with RNA in Fig. 2. Mutations are labelled with the substituted base followed by the position number, e.g. G2 indicates a T (or U) to G change at position 2 in the T-tract (or U-tract). The first T of the T-tract was not changed in these experiments because it specifies a U that could base pair and extend the terminator hairpin (Fig. 2). However, based on studies of other intrinsic terminators, it is unlikely that this A:U bp would contribute to termination efficiency (Yang et al., 1995). The 21 mutant fusion strains, plus a wild-type fusion strain for comparison, were grown in minimal media supplemented with uracil. The intracellular levels of β-galactosidase activity, which reflect transcription through the pyrBI attenuator, were measured.
T to G substitutions. Compared with the wild-type fusion strain, the β-galactosidase activity was increased in each T to G mutant strain (Fig. 3), indicating that each mutation reduced the ability of the pyrBI attenuator to terminate transcription. The magnitude of the increase in β-galactosidase activity was highly variable, ranging from 23 to 90 U mg−1 (i.e. 2.5- to 10-fold), compared with 9 U mg−1 in the wild-type fusion strain. In general, the level of β-galactosidase activity was higher (i.e. termination efficiency was lower) in strains with substitutions closer to the dyad symmetry/terminator hairpin. One exception to this pattern was that the β-galactosidase activity in the strain with the G8 attenuator was higher than that in the strains with the G6 and G7 attenuators. This result suggests that a mutation at position 8, the predominant termination site, has different consequences compared with substitutions earlier in the T-tract. Another exception was that the strain with the G2 attenuator exhibited a slightly, but reproducibly, lower level of β-galactosidase activity than the strain with the G3 attenuator. This minor deviation may be explained by the fact that the G2 mutation permits an additional base pair in the terminator hairpin (Fig. 2), which may compensate in part for the adverse effect of this mutation on termination.
T to A substitutions. The results with the T to A substitutions were generally similar to those obtained with the T to G substitutions. Each T to A change resulted in a significant increase in β-galactosidase activity, which varied from 22 to 70 U mg−1 (i.e. 2.2- to 7-fold compared with 10 U mg−1 for the wild type) (Fig. 4). The increases in β-galactosidase activities observed with several of the T to A substitutions (i.e. at positions 2, 3 and 4) were substantially less than those observed with the corresponding T to G substitutions, suggesting that the latter were more deleterious to termination. The β-galactosidase activities in the T to A substitution strains also exhibited a general pattern that indicated that substitutions closer to the dyad symmetry/terminator hairpin produce weaker attenuators. The position 8 substitution was again an exception to this pattern, with the strain carrying the A8 substitution exhibiting a β-galactosidase activity that was higher than all but one (i.e. A2) of the other T to A mutant strains. Another outlier in the pattern was the strain carrying the A5 attenuator, which exhibited a higher level of β-galactosidase activity (61 U mg−1) than that observed with the strains carrying the A3 (55 U mg−1) and A4 (53 U mg−1) attenuators. This deviation appears to be real because it is greater than our experimental error and it was also observed when these attenuators were examined in vitro (see below). The reason for the deviation remains to be determined.
T to C substitutions. The results with the T to C substitutions contained clear similarities and differences in comparison to the other two sets of substitutions. Again, each T to C change resulted in a significant increase in β-galactosidase activity, which varied from 26 to 47 U mg−1 (i.e. 2.6- to 4.7-fold compared with 10 U mg−1 for the wild type) (Fig. 5). However, the increases for the substitutions at C2 through C5 were considerably less that those observed with the corresponding T to G or A substitutions, suggesting that the T to C substitutions were tolerated better. The β-galactosidase activities in the T to C substitution strains also exhibited a pattern that indicated that substitutions closer to the dyad symmetry/terminator hairpin produced weaker attenuators, but this pattern was not as striking as those observed with the T to G or A substitutions. In addition, the C8 substitution could be included in this pattern, unlike the G8 and A8 substitutions. This result, and those above, indicate that the order in which bases are tolerated as the last base in the terminated transcript is U > C > G > A.
Control experiments. To determine whether any base substitution had an effect on pyrB::lacZ expression in addition to its effect on termination at the pyrBI attenuator, we grew the 21 mutant fusion strains plus the wild-type fusion strain in minimal medium with UMP as the pyrimidine source. This growth condition effectively inactivates the pyrBI attenuator. Within experimental error, the β-galactosidase levels in all strains were the same (data not shown), indicating that the effects of the substitution mutations were restricted to those on termination at the pyrBI attenuator.
Termination efficiencies at the pyrBI attenuators containing single-base substitutions in the T-tract
To express the effects of the 21 substitution mutations in terms of termination efficiencies instead of levels of pyrB::lacZ expression, we constructed another fusion strain (i.e. CLT5240) in which the last seven positions of the T-tract were changed to CACGCGA. In the absence of the T-tract, termination at the mutant pyrBI attenuator is abolished (Liu and Turnbough, 1989). Therefore, the level of β-galactosidase activity in strain CLT5240 grown in minimal medium with uracil corresponds to 100% readthrough transcription of the pyrBI leader region. This level was determined to be 362 U mg−1 (Table 1). In comparison, the level of β-galactosidase activity in the wild-type fusion grown under the same conditions was 9.5 U mg−1. Using these values, the calculated efficiency of transcription termination at the wild-type pyrBI attenuator is 97%, which is essentially the same as previous determinations. We also showed that the levels of β-galactosidase activity in strain CLT5240 and the wild-type strain were essentially the same when these strains were grown in minimal medium with UMP (Table 1). This experiment demonstrates that the mutation in strain CLT5240 only affects termination at the pyrBI attenuator. The level of β-galactosidase activity in strain CLT5240 grown on UMP is approximately 1.5-fold higher than that when this strain is grown on uracil. This difference, which is independent of attenuation control, is due to uncharacterized regulation related to slower growth on UMP (Liu et al., 1994).
Table 1. Effects of inactivating the pyrBI attenuator on pyrB::lacZ expression.a
Doubling times were 48 and 64 min when the fusion strains were grown with uracil or UMP as the pyrimidine source respectively.
Mean of two experiments with variation of < 5%.
T7 to CACGCGA in T-tract
We then made similar calculations to convert the levels of β-galactosidase activity in the 21 mutant strains to termination efficiencies at the mutant pyrBI attenuators (Fig. 6). The results recapitulate the patterns described above, but they also provide a more meaningful view of the effects of the mutations on intrinsic termination. The results show that the mutation with the largest effect, which is the G3 substitution, reduces termination efficiency to 75%. Thus, no single-base change in the T/U-tract comes close to inactivating the attenuator. However, most of the substitutions result in a several-fold increase in readthrough transcription at the pyrBI attenuator, effects that would likely have a significant physiological impact.
Effects on transcription termination of separating the terminator hairpin and U-tract of the pyrBI attenuator
Typical intrinsic terminators specify an RNA in which the terminator hairpin is immediately followed by a U-tract. To measure the effects on termination of physically separating these two required elements, we employed a set of five pyrB::lacZ fusion strains containing different insertion mutations in the T-tract of the pyrBI attenuator. One mutation, designated +C, results in the insertion of a single C residue between positions U1 and U2 of the U-tract. The other four mutations, designated +CT, +C2T, +C3T and +C4T, result in the insertion of 1 to 4 C residues plus an adjacent downstream U residue between positions U1 and U2 of the U-tract. For the latter mutations, the U insertion was included to maintain a continuous run of eight U residues in this region of the transcript. All five mutations were designed to avoid altering the length of the terminator hairpin. The effects of the mutations were examined as described for the base substitutions.
Compared with the wild-type fusion strain, the β-galactosidase activity was increased in each mutant strain, indicating reduced ability of the pyrBI attenuator to terminate transcription (Fig. 7A). However, the magnitude of the increase in β-galactosidase activity was highly variable, and the pattern of these increases was surprisingly irregular. To investigate this unexpected pattern, we measured pyrB::lacZ expression in the same strains grown on UMP (Fig. 7B). The results (i.e. different levels of β-galactosidase activity) showed that the insertion mutations had an effect on pyrB::lacZ expression other than altering termination at the pyrBI attenuator. The pattern of the β-galactosidase activities in UMP-grown cells suggested that this ‘second effect’ was related to the reading frame of the leader polypeptide, which we confirmed in separate experiments (data not shown). The reason for this second effect remains unclear. However, to determine the effects of the insertion mutations on termination efficiency, it was necessary to normalize the β-galactosidase activities in cells grown on uracil to those in cells grown on UMP. The normalized values were then used to calculate termination efficiencies as described above for the substitution mutations.
The normalized termination efficiencies, particularly those of the +CnT insertions, exhibit a clear pattern in which reductions in percent termination are proportional to the length of the insertion (Fig. 8). However, even with the longest insertion (i.e. +C4T), termination still occurred with 37% efficiency. This result indicates that short insertions, at least those with C residues, are tolerated reasonably well by the attenuator. The results also show that the efficiency of transcription termination at the +C attenuator (90%) is slightly lower than that at the +CT attenuator (92%). This result suggests that maintaining a continuous run of eight U residues after the C (or Cn) insertion has a positive, albeit small, effect on termination efficiency. This small difference was recapitulated in transcription termination assays performed in vitro (see below).
In vitro analysis of transcription termination at the mutant pyrBI attenuators
To corroborate our in vivo studies, we employed a standard, single-round in vitro transcription assay to measure the efficiency of transcription termination at each of our mutant pyrBI attenuators. The templates in this assay were approximately 300 bp, blunt-ended, double-stranded DNA fragments containing either a mutant or (as a control) wild-type pyrBI promoter–leader region. All templates also contained the T to G substitution at position +3 (Fig. 1), which eliminates reiterative transcription at the pyrBI promoter. Nascent transcripts were 32P-labelled at their 5′ ends and separated by gel electrophoresis. Termination products were detected as 134- and 135-nucleotide-long transcripts, which result from termination at positions 7 and 8 in the T- (or U-) tract respectively (Donahue and Turnbough, 1994). Readthrough transcripts (i.e. transcripts that were not terminated at the attenuator) were detected as ≥ 170-nucleotide-long transcripts. Termination efficiency was expressed as a percentage of total transcripts that terminate at the attenuator, i.e. (terminated transcripts/terminated transcripts + readthrough transcripts) × 100.
Single-base substitutions. The negative effects on transcription termination of the T to G substitutions measured with the in vitro assay generally paralleled the effects measured in cells (Fig. 9A). This included the slightly larger effect on termination efficiency of the G3 substitution compared with the G2 substitution. However, the effects of the substitutions in the in vitro assay were often greater than those observed in cells. The difference between the two assays gradually decreased as the distance between the substitution and the terminator hairpin increased. For example, the negative effects on termination efficiency in vitro for substitutions G2 and G3 were 2.5-fold greater than those measured in vivo, but for substitutions G6 through G8, the differences between the two assays essentially disappeared.
The negative effects on transcription termination of the T to A or C substitutions measured in vitro also mimicked, in general, those observed in cells. For the T to A substitutions (Fig. 9B), this included a pattern in which the effect of the A5 mutation on termination efficiency was greater than that of the A3 and A4 mutations. With some of the T to A substitutions, namely A3 through A5, the mutational effects in the in vitro assay were greater than those observed in vivo. In this case, the effects were approximately twofold greater in vitro. For the T to C substitutions (Fig. 9C), the differences between the two assays were marginal, except for the C2 substitution. The effect of the C2 substitution on termination efficiency measured in vitro was twice that determined in vivo.
Insertion mutations. As with the base substitutions, the negative effects on transcription termination of the insertion mutations measured in vitro mimicked those observed in cells (Fig. 10). This pattern included slightly lower termination efficiency at the +C attenuator (87%) than that at the +CT attenuator (89%). The magnitudes of the effects of the +C and +CT insertions were essentially the same when measured in vitro or in vivo. In contrast, the effects of the +C2T, +C3T, and +C4T insertions measured in vitro were substantially greater than those determined in vivo. For the +C2T and +C3T insertions, the reductions in termination efficiencies were twofold greater when measured in vitro. The +C4T insertion nearly eliminated termination activity in vitro, which was only 9% compared with 37% in vivo.
Control experiments. The in vitro transcription assays reported in this study were performed with soluble RNA polymerase. To demonstrate that putative terminated transcripts were actually released from the transcription elongation complexes, we performed duplicate assays with RNA polymerase that had been immobilized by attachment to agarose beads (data not shown). In this case, bound transcription elongation complexes could be separated by centrifugation from released transcripts (Wang et al., 1995; Kashlev et al., 1996). We found that for each mutant attenuator, analysis of the released transcripts yielded transcription termination efficiencies comparable to those obtained using soluble RNA polymerase. In addition, we were unable to detect transcripts associated with immobilized RNA polymerase (Wang et al., 1995). Thus, transcripts terminated at the mutant attenuators were in fact released from RNA polymerase.
Position and sequence effects of base substitutions
The results presented in this study demonstrate that any single-base substitution within the U-tract specified by the pyrBI attenuator reduces termination efficiency in vivo, causing a 2- to 10-fold increase in readthrough transcription at the mutant attenuator. These results highlight the importance of the long, continuous run of attenuator-specified U residues for maximally efficient termination at the pyrBI attenuator, and presumably at other intrinsic terminators that specify homogeneous U-tracts. Although each substitution results in reduced termination efficiency, the magnitude of the reduction is dependent on the position and sequence of the substitution. With respect to position, substitutions closer to the dyad symmetry/terminator hairpin, in general, have a greater adverse effect on termination efficiency. This pattern indicates that the U residues close to the hairpin, specifically at positions 2 through 5, play an especially important role in termination. Such a role is in fact predicted by the current model for intrinsic termination, which proposes that disruption of weak rU:dA base pairing in the upstream half of the RNA/DNA hybrid facilitates the formation of the terminator hairpin (Gusarov and Nudler, 1999). In addition, the disruption of ≥ 3 upstream rU:dA bps results in a shortened and more unstable RNA/DNA hybrid, which permits the release of the transcript from the DNA template and RNA polymerase (Komissarova et al., 2002).
Our results also show a pattern in the effects of substitutions at positions 6, 7 and 8. Substitutions at position 6 reduce termination efficiency slightly more than observed with substitutions at position 7; however, the effects of substitutions at both positions are relatively small. Apparently, non-U residues at positions 6 and 7 do not significantly stabilize the short RNA/DNA hybrid remaining after formation of the terminator hairpin. Such a modest effect may explain the observation that most non-U residues that occur naturally in intrinsic terminators are found at positions 6 and 7 (Lesnik et al., 2001). Finally, substitutions at position 8, at least for the T to G or A mutations, disrupt termination more than do substitutions at positions 6 and 7. Presumably, this pattern indicates that the final residue of the transcript plays a uniquely critical role in transcription termination (Landick, 1997).
Our results also show that the particular base used in the substitution can influence, in a predictable manner, the magnitude of the negative effect on termination efficiency. For example, at positions 2 through 5, the order for the magnitude of the adverse effect is G > A > C. At positions 6 and 7, on the other hand, each base substitution has a similar negative effect on termination. At position 8, the order in which the substitutions reduce termination efficiency is A > G > C. Remarkably, these rankings do not show a correlation between base pairing strength and termination efficiency, which indicates that the adverse effects of the substitutions involve more than stabilizing the RNA/DNA hybrid. One obvious possibility is that there are sequence-specific interactions between RNA polymerase and the transcript or DNA template that are important for termination. This specificity may reflect the need to avoid steric hindrance in some instances. Interestingly, the smallest base, C, is the best-tolerated non-U residue at every sequence-sensitive position in the T-tract.
Separating the terminator hairpin and U-tract
Our analysis of mutant pyrBI attenuators specifying Cn insertions between the terminator hairpin and wild-type U-tract demonstrates that these two components must be immediately adjacent for maximally efficient termination in vivo. The termination efficiency decreases significantly with each residue added to the insertion. This observation is entirely consistent with the current model for intrinsic transcription termination, in which there is a requirement for weak base pairing in the upstream half of the RNA/DNA hybrid. However, none of the Cn insertions abolished termination, with the longest insertion examined (i.e. +C4T) reducing termination efficiency to 37%. Clearly the mechanism for intrinsic transcription termination can tolerate separation between the terminator hairpin and U-tract to some extent. Based on the pattern of the negative effects of the Cn insertions on termination efficiency, it is likely that increasing the length of the Cn insertion by 1 or at most 2 additional residues would eliminate termination. In addition, according to the sequence-specific effects of the substitution mutations discussed above, it is likely that insertions containing G or A residues would inhibit termination much more than the Cn insertions.
In vivo versus in vitro effects
Using an in vitro transcription assay with purified components, we also measured the efficiency of termination at all of the mutant pyrBI attenuators described above. The effects of the mutations determined in vitro closely parallel those observed in vivo, which provides strong support for the conclusions presented in this study. However, there was one major difference in the results: the in vivo effects of some of the attenuator mutations were significantly smaller than those observed in vitro. The mutant attenuators clearly exhibiting this difference included those containing base substitutions at positions 2 through 5 and those with +C2T, +C3T and +C4T insertions. It is probably not a coincidence that these mutations have the largest inhibitory effects on termination and apparently define key elements of terminator sequence and structure. Presumably, the smaller in vivo effects indicate that there are factors in cells, which are absent in our in vitro assay, that enhance termination efficiency. These factors may include proteins such as NusA (Bermudez-Cruz et al., 1999) and solution conditions such as higher salt (e.g. potassium glutamate) (Reynolds and Chamberlin, 1992; Cayley and Record, 2003), to name just a few candidates.
The mechanism of intrinsic transcription termination described in this study appears to be essentially the same in many Gram-negative and Gram-positive bacteria (de Hoon et al., 2005). Therefore, it seems a certainty that these bacteria use non-T residues in the T-tracts of their intrinsic transcription terminators and attenuators to adjust termination efficiencies to accommodate the metabolic needs of the cell. To better understand this process, it is necessary to develop ‘rules’ that define the effects of these non-T residues on termination. The present study provides at least a preliminary set of rules that can be used to make predictions about termination efficiencies at intrinsic terminators. In addition, these rules can be used to engineer intrinsic terminators to function in a desired manner. As more intrinsic terminators are analysed in vivo, it will be interesting to see how well theseterminators follow the rules described in this study, and how these rules need to be adjusted.
Escherichia coli K-12 strain CLT42 [F–car-94Δ(argF-lac)U169 rpsL150 thiA1 relA1 deoC1 ptsF25 flbB5301 rbsR] (Roland et al., 1985) was the parent strain for the construction of the lambda lysogens used in this study. These strains were constructed by inserting a recombinant lambda bacteriophage, which carries either a wild-type or mutant version of a pyrB::lacZ gene fusion, into the lambda attachment site of the chromosome of strain CLT42. The wild-type fusion contains the wild-type pyrBI promoter–leader region plus the first four codons of the pyrB gene fused to the ninth codon of lacZ. The mutant fusions are the same as the wild-type fusion except for a base substitution or insertion mutation in the T-tract of the pyrBI attenuator. For a control fusion strain (i.e. CLT5240) that is unable to terminate transcription at the pyrBI attenuator, the last seven positions of the T-tract were changed to CACGCGA (non-template strand sequence). In addition, the wild-type, mutant, and control fusions contain a T to G substitution (non-template strand sequence) at position +3 of the pyrBI initially transcribed region, which prevents reiterative transcription at the pyrBI promoter (Liu et al., 1994).
Site-directed mutations were introduced into the T-tract of the pyrBI attenuator by using a polymerase chain reaction (PCR)-based procedure that was essentially the same as that described by Barettino et al. (1994). The sequence of each mutant promoter–leader region was determined to verify the desired mutation and to ensure that spurious mutations were not introduced by PCR.
Construction of gene fusions
DNA fragments were prepared that contained approximately 300 bp segments of the wild-type or mutant pyrBI promoter–leader regions extending through codon 4 of the pyrB gene. These fragments also contained an upstream EcoRI restriction site and a downstream BamHI restriction site. These fragments were digested with EcoRI and BamHI, and then inserted between the EcoRI and BamHI sites of the fusion vector pMLB1034 (Roland et al., 1985). This vector contains the E. coli lacZ gene without its promoter, ribosome binding site, and first eight codons. The insertion into this vector created an in-frame pyrB::lacZ fusion. All fusion constructions were confirmed by DNA sequence analysis.
Transfer of gene fusions from plasmids to the E. coli chromosome
Wild-type and mutant pyrB::lacZ fusions carried on derivatives of plasmid pMLB1034 were individually recombined onto phage lambda RZ5, and a single copy of each recombinant phage was integrated into the chromosome of strain CLT42 as described previously (Roland et al., 1988). The presence of a single prophage at the chromosomal lambda attachment site was confirmed by PCR analysis (Powell et al., 1994).
Media and culture methods
Cells for enzyme assays were grown in N–C– medium (Alper and Ames, 1978) with 10 mM NH4Cl, 0.4% glucose, 0.015 mM thiamine hydrochloride, 1 mM arginine, and either 1 mM uracil or 0.25 mM UMP. Cultures (30 ml in 250 ml flasks) were grown at 37°C with shaking, and at an A650 of 0.5 (during exponential growth), 20 ml samples was taken. Cells were collected by centrifugation at 4°C, washed with ice-cold 50 mM sodium phosphate buffer (pH 7.0), and stored at −70°C.
Frozen cells were resuspended in 5 ml of ice-cold 50 mM sodium phosphate buffer (pH 7.0) and disrupted by sonic oscillation at 0°C. Extracts were centrifuged at 17 000 g for 30 min at 4°C to remove cell debris, and the supernatants were used for β-galactosidase and protein assays as previously described (Roland et al., 1988). The data reported are the mean of two independent experiments with variation of < 5%.
The DNA templates used for in vitro transcription were approximately 300 bp PvuII restriction fragments excised from recombinant plasmids carrying the pyrBI promoter–leader region with a wild-type or mutant pyrBI attenuator. Templates were purified by agarose gel electrophoresis and electroelution (Qi et al., 1996). The templates included the pyrBI sequence from −130 to +170, with +1 being the pyrBI transcription start site located 6 bp downstream of the −10 region (Donahue and Turnbough, 1990). This start site, which is used in the in vitro assay, is 2 bp upstream of the in vivo start site (Fig. 1). Transcripts initiated at either start site terminate identically at the pyrBI attenuator. All templates contained the T to G substitution in the pyrBI promoter region described above, which eliminates reiterative transcription.
In vitro transcription
Purified RNA polymerase holoenzyme containing σ70 was prepared as previously described (Turnbough et al., 1983). Single-round in vitro transcription assays were used to measure termination efficiencies at the wild-type and mutant pyrBI attenuators. The assay included a preliminary reaction to allow the formation of a heparin-resistant initiation complex containing the 5′ end-labelled nascent transcript ACAA (Donahue and Turnbough, 1994). The 8 μl reaction mixture contained 20 mM Tris-acetate (pH 7.9), 10 mM magnesium acetate, 150 mM potassium glutamate, 0.2 mM Na2EDTA, 0.1 mM dithiothreitol, 0.25 mM CTP, 5 μM [γ-32P]-ATP (240 Ci mmol−1, NEN), 12.5 nM DNA template, and 125 nM RNA polymerase. The reaction mixture was incubated for 5 min at 37°C. Extension of the ACAA transcript was started by the addition of 2 μl of 5× NTP-heparin solution [20 mM Tris-acetate (pH 7.9), 10 mM magnesium acetate, 150 mM potassium glutamate, 0.2 mM Na2EDTA, 0.1 mM dithiothreitol, 1 mM each ATP, GTP and UTP, and 0.5 mg ml−1 heparin], which had been prewarmed to 37°C. Final concentration for each NTP was 0.2 mM. The complete reaction mixture was incubated for 15 min at 37°C. The presence of heparin in this mixture prevented additional rounds of transcription initiation. Each reaction was stopped by the addition of 0.1 ml of ice-cold stop solution [0.3 M sodium acetate (pH 5.2) and 1 mM Na2EDTA]. RNA transcripts were precipitated from the sample by adding 2 vols of ethanol and chilling at −70°C for 1 h. The RNA was collected by centrifugation at 16 000 g for 15 min at 4°C, washed with 0.5 ml of cold 70% ethanol, dried in vacuo, and dissolved in a solution containing 5 M urea and 0.025% each bromophenol blue and xylene cyanol. Transcripts were separated by electrophoresis in a 10% polyacrylamide (29:1 acrylamide:bisacrylamide)/TBE [50 mM Tris-borate (pH 8.3) and 1 mM Na2EDTA] sequencing gel containing 7 M urea. Electrophoresis was carried out at 2000 V for approximately 3 h until the xylene cyanol had run 30 cm into the gel. Transcripts were visualized by autoradiography, and transcript levels were measured by using a Molecular Dynamics PhosphorImager and subtracting background counts.
In this assay, transcripts that read through the attenuator and were extended to the end of the DNA template were approximately 170 nucleotides long. Some of these transcripts were extended further by template switching, which we demonstrated by several methods (Nudler et al., 1996). Therefore, the level of readthrough transcripts was measured by adding the levels of all transcripts that were ≥ 170 nucleotides long. We demonstrated that this method was equivalent to measuring readthrough transcripts after RNase H-mediated cleavage or genetically engineered termination of pyrBI transcripts between the pyrBI attenuator and the end of the primary DNA template. The latter procedures produce readthrough transcripts of equal length (data not shown).
This work was supported by NIH Grant GM29466 (to C.L.T.) and Hungarian Fund OTKA T 048793 (to K.S.). We thank Evvie Allison for editing the manuscript.