Kinetics of RNA Polymerase Initiation and Pausing at the Lambda Late Gene Promoterin vivo

J. Mol. Biol. (1995) 254, 808–814

Kinetics of RNA Polymerase Initiation and Pausing at the Lambda Late Gene Promoter in vivo Mark Kainz and Jeffrey W. Roberts* Section of Biochemistry Molecular, and Cell Biology Cornell University, Ithaca NY 14853, USA

We have measured the kinetics of transcription initiation and pausing by Escherichia coli RNA polymerase (RNAP) at the bacteriophage lambda late promoter, pR' , in growing cells. RNAP initiating transcription from pR' pauses after transcribing 16 or 17 nucleotides, and escape from this pause could in theory be the rate-limiting step in promoter function. We tested this hypothesis by analyzing pausing and non-pausing variants of both the pR' promoter segment and a more active mutant version of pR' ; we measured reporter gene expression and used KMnO4 footprinting to measure directly occupancy of the promoter and pause sites in growing cells. We find that RNAP paused at +16/+17 does not limit expression of pR' . However, RNAP paused at +16/+17 does limit expression from the more active promoter by impeding formation of open complex. Therefore, the activity of the late gene regulatory protein Q to suppress the early pause, in addition to its antitermination activity, is unlikely to be important in phage gene expression. 7 1995 Academic Press Limited

*Corresponding author

Keywords: transcription pause; RNA polymerase; promoter activity; antiterminator

Introduction In several bacterial and eukaryotic operons or genes, RNA polymerase (RNAP) initiates transcription and then pauses at a specific site close to the promoter, where it may be a target of regulatory events. In addition to the l late operon, examples include amino acid biosynthetic operons of Escherichia coli, for which a transcription pause synchronizes RNA synthesis with translation by ribosomes that are the regulatory agents (Landick & Yanofsky, 1987); the hsp70 gene of Drosophila melanogaster, in which RNA polymerase II pauses after synthesizing approximately 25 nucleotides of RNA and is the target of heat shock transcription factor (Rougvie & Lis, 1988); and possibly many other eukaryotic genes (Giardina et al., 1992). In the phage l late gene operon, E. coli RNAP pauses at positions +16 and +17 relative to the transcription start site; this paused enzyme appears to be the essential target for modification of RNAP by the phage antiterminator protein Q (Grayhack et al., 1985). In addition to modifying RNAP to a Present address: M. Kainz, Department of Bacteriology, University of Wisconsin, Madison, WI 53706, USA. Abbreviation used: RNAP, RNA polymerase. 0022–2836/95/500808–07 $12.00/0

termination-resistant form, Q protein chases RNAP from the +16/+17 pause (Grayhack et al., 1985). Footprinting analyses with KMnO4 both in vivo and in vitro revealed that RNAP paused at +16/+17 melts the DNA helix from −1 to +17 (Figure 1); the open complex is melted from −11 to +3, as diagrammed in Figure 1 (Kainz & Roberts, 1992). It would seem likely from the overlap that RNAP could occupy either the promoter in open complex or the +16/+17 pause, but not both. This has not been shown, and the surprising ability of DNA polymerase and RNA polymerase to co-exist on DNA (Liu et al., 1993) suggests caution in dismissing overlapping occupancy; nevertheless, RNAP in the +16/+17 pause would at least freeze RNAP in the adjacent open promoter complex and thereby block promoter function. The question arises whether the +16/+17 pause limits the rate of promoter function, and thus whether Q, by chasing RNAP from the pause, acts to increase the rate of transcription from the promoter region in addition to its function in transcription antitermination. Using KMnO4 probing to measure occupancy of the promoter and pause site, we show that the pause does not limit the rate of function of the wild-type pR' promoter. However, a mutant promoter of sixfold higher intrinsic activity is limited by the pause. 7 1995 Academic Press Limited

809

Pausing and Promoter Function

Figure 1. Structure of paused elongation and open promoter complexes associated with the l late gene promoter pR' . The structures are based on KMnO4 reactivity of pyrimidines as determined by Kainz & Roberts (1992); black bars indicate KMnO4-reactivity in the template strand used in this paper to identify +16/+17 paused complexes and open promoter complexes.

Results Effect of the +16/+17 pause on promoter activity To determine if suppression of the +16/+17 pause by lQ protein could affect promoter activity, we asked if the pause limits expression from the promoter in the absence of Q. We tested both the natural late promoter pR' and a derivative of pR' that is several times as active in the absence of Q protein: trp/pR' /−17, in which lambda sequences between −17 and −31 are replaced by slightly modified sequence from the equivalent positions of the E. coli tryptophan operon promoter. The derivative is one of a series of promoter variants designed to identify sequences required for Q function by substituting completely different bases to the left of −10 (except in the −35 hexamer). The trp promoter sequences were chosen to avoid using a completely arbitrary sequence that might have unanticipated effects. RNAP initiating at trp/pR' /−17 pauses at +16/+17 as expected, because all sequences necessary for pausing are still present. As is true for pR' , a T to G mutant at +6 of the non-template strand fails to pause in vivo (see below) and in vitro (M.K. and J.W.R., unpublished experiments). Figure 2 shows the effect of RNAP pausing at +16/+17 on expression from pR' and from trp/pR' / −17, measured as the activity of a downstream galactokinase reporter gene in cells transformed with plasmids carrying either original or non-pausing versions of pR' and trp/pR' /−17 promoter segments. Transcription from pR' is not significantly

affected by RNAP pausing at +16/+17, yielding three to four units of galactokinase activity from both pausing and non-pausing (+6C) versions of the pR' plasmid. However, RNAP pausing at +16/+17 does not limit transcription from trp/pR' /−17, by fourfold: the trp/pR' /−17 pause competent promoter gives five units of galactokinase activity, whereas the same promoter directing transcription of the non-pausing (+6G) mutant template gives nearly 20 units of activity. We found a similar limitation of promoter function by RNAP pausing at +16/+17 downstream of another trp/pR' fusion, trp/pR' /−31 which has an inherent strength comparable to that of trp/pR' /−17 (data not shown). Effect of RNAP paused at +16/+17 on formation of open promoter complex KMnO4 reacts with thymine residues specifically in melted DNA and thus identifies DNA opened in the transcription bubble. We used in vivo KMnO4 footprinting to measure directly occupancy of the promoter and the +16/+17 pause site by RNAP on the four different promoter-pause constructs. Cells transformed with plasmids containing either pR' or trp/pR' /−17, directing transcription of either a pause competent or non-pausing mutant template, growing in mid-log phase, were pulsed for ten seconds with KMnO4 before and at intervals after addition of rifampicin. Plasmid DNA was isolated and quantified, and equal amounts were probed by primer extension on the template DNA strand. The initial sample reveals the equilibrium distribution of RNAP on promoter and pause site. Addition of

810

Figure 2. RNA polymerase paused at +16/+17 limits function of the trp/pR' /−17 promoter but not pR' . E. coli HB101 transformed with plasmids (pXY306, pXY306(+6C), pTL17, pTL17(+6G)) in which the galK gene is expressed under the control of +16/+17 pause competent or non-pausing (+6C, +6G) versions of the l pR' and trp/pR' /−17 promoters were assayed for galactokinase activity. Units are normalized to culture density.

rifampicin traps RNAP in open complex by blocking escape from the promoter, and thus prevents further entry of RNAP into the +16/+17 pause, but allows elongation complexes at the pause (which are not affected by the drug) to escape; thus the time course after rifampicin addition shows the rate at which RNAP fills the promoter to form open complex and the rate of escape of RNAP from the pause. To quantify the result, we measured radioactivity with a Betascope and normalized to the saturation level of the open complex after rifampicin treatment. By assuming that all promoters fill completely in these conditions, we could calculate the fraction of promoter DNA in open complex, compare the fractional occupancy in different cultures, and calculate the relative amount of paused complex in different cultures. (However, we have no reference by which to measure the absolute fraction of DNA that is occupied by RNAP in the pause.) Figure 3(a) shows KMnO4 reactive pyrimidines in the template strand of pR' pause competent or non-pausing (+6C) plasmids. On pR' (lanes 1 to 8) two major signals are present: RNAP in open complex on the late promoter, identified as KMnO4 reactivity at −11T (Kainz & Roberts, 1992), and RNAP paused at +16/+17, identified by KMnO4 reactivity at +1T and +2T (Kainz & Roberts, 1992). During the time course after rifampicin treatment the signal at −11 increases as RNAP accumulates on pR' : occupancy of the promoter open complex

Pausing and Promoter Function

increases from the steady state level of approximately 20% of saturation, to full saturation in 060 seconds (Figures 3(a), 4(a)). The signal of the paused elongation complex at +16/+17 decreases over the time course (Figure 3(a)) with a halftime of ten seconds (Figure 4(c)). The point mutation at +6 (+6C) greatly reduces the +16/+17 pause (Yang, 1988; Kainz & Roberts, 1992); correspondingly, only open complex signal (−11T) and no significant +16/+17 pause signal (+1, +2) was detected with pR' (+6C) (Figure 3(a), lanes 9 to 14). Steady state promoter occupancy on the pR' (+6C) plasmid was nearly 50% of saturation (Figure 3(a), lane 9; Figure 4(a)), more than twice that of the pause competent pR' plasmid (Figure 4(a); also compare lanes 1 and 9 of Figure 3(a)). However, despite its higher steady state occupancy, pR' (+6C), like pR' , required 060 seconds to become saturated with RNAP in open complex (Figures 3(a), 4(a)). This result is consistent with the effect of the pause on galactokinase reporter gene expression from these promoters (Figure 2): the ability of RNAP to pause at +16/+17 does not significantly limit promoter function of pR' , measured either as the expression of a reporter gene or as the rate at which the promoter can fill. Figure 3(b) shows KMnO4 reactive pyrimidines in the template strand of trp/pR' /−17 pause competent or non-pausing (+6G) plasmids. The kinetics of open complex accumulation and +16/ +17 pause decay after rifampicin treatment on the pause competent trp/pR' /−17 plasmid were similar to those on native pR' (Figure 3(b); compare lanes 1 to 6 with lanes 1 to 8 of Figure 3(a); see also Figure 4(a)). The steady state promoter occupancy of trp/pR' /−17, like pR' , was approximately 20% of saturation, and it filled at about the same rate as pR' , reaching saturation with RNAP in open complex in 060 seconds (Figure 4(a)). The halflife of the +16/+17 paused complex on the trp/pR' /−17 promoter segment was 15 seconds, slightly greater than for the pR' segment (Figures 3(b), 4(c)). The +6G mutation of the trp/pR' /−17(+6G) promoter segment, like the +6C mutation of the pR' segment, severely reduced the +16/+17 pause (Figure 3(b), lanes 7 to 12). Also like pR' (+6C), the steady state occupancy of trp/pR' /−17(+6G) was nearly 50% of saturation (Figure 3(b), lane 7; Figure 4(a)). However, the promoter of trp/pR' / −17(+6G) filled fourfold faster than pR' (+6C) or either of the pause competent templates, reaching saturation with RNAP in open complex in only 10 to 15 seconds (Figure 3(b), lanes 7 to 12; Figure 4(a)). Thus the function of the trp/pR' /−17 promoter is limited by RNAP paused at +16/+17: the promoter with the non-pausing +6G mutation becomes saturated with RNAP in open complex four times as fast as the pause-competent promoter segment, and expression of the galactokinase reporter gene from the non-pausing template is four to fivefold higher than expression from trp/pR' /−17 (Figure 2).

811

Pausing and Promoter Function

Figure 3. RNA polymerase paused at +16/+17 limits the rate of filling of trp/pR' /−17 but not pR' open complex. E. coli HB101 transformed with plasmids described in Figure 2 were treated with rifampicin (rif), and culture aliquots were removed and treated for ten seconds with KMnO4 at intervals during the time course; plasmid was extracted, and KMnO4-modified pyrimidines in the transcribed strand of the promoter region were identified by primer extension. Reactivity at −11 identifies RNAP in open promoter complex, and reactivity at +1 and +2 identifies RNAP paused at +16/+17. (a) Accumulation of RNAP in open complex on pause competent pR' (lanes 1 to 8) and non-pausing pR' (+6C) (lanes 9 to 14) promoter segments. (b) Accumulation of RNAP in open complex on pause competent trp/pR' /−17 (lanes 1 to 6) and non-pausing trp/pR' / −17(+6G) (lanes 7 to 12) promoter segments.

(a)

(b)

To obtain a value for the rate at which each promoter fills, we plotted the fraction of promoter not in open complex (i.e. either free or occupied by RNAP in pause) from Figure 4(a) versus time after rifampicin addition. The amount of unoccupied promoter before addition of rifampicin is set at 1.0, and values at later times are given as fractions of the zero time value. Figure 4(b) shows that the promoter of the non-pausing trp/pR' /−17(+6G) segment fills much faster than trp/pR' /−17, with halftime of 3 versus 16 seconds, and also several times faster than either pR' or pR' (+6C), which fill with halftimes of 19 and 12 seconds, respectively, similar to that of trp/pR' /−17.

Discussion These experiments provide a semi-quantitative view of the traffic of RNA polymerase in vivo through the lambda late gene promoter and early pause site. It is possible from the data to propose a consistent model of the rate-limiting steps in expression from these promoters and their absolute rates. It is clear that the pause does not significantly limit expression from pR' in the absence of Q protein. This is revealed both by the level of galactokinase expression, which is similar with and without the pause, and by the rate at which the open complex

812

Pausing and Promoter Function

(a)

(b)

(c) Figure 4. Kinetics of RNAP accumulation in open complex and escape from the +16/+17 pause. Radioactivity in KMnO4-reactive signals of open promoter complex and +16/+17 paused complex was quantified, normalized to the signal of the saturated open promoter complex, and plotted as a function of time of rifampicin treatment. (a) Open promoter complex formation of pR' (squares) and trp/pR' /−17 (circles) promoter segments that are +16/+17 pause competent (filled) or non-pausing (open). (b) Data of (a) replotted to determine the half-time of open complex formation. Each curve is normalized to the amount of open complex present at zero time. (c) +16/+17 pause escape on pR' (squares) and trp/pR' /−17 (circles) templates. The inset is data from the first 30 seconds replotted to determine pause half-time.

forms. We measure half times for open complex filling of 19 and 12 seconds with and without pause, a difference perhaps of marginal significance, although in the direction expected if the pause slightly retards promoter filling. Since the halftime of pause decay is ten seconds, similar to the rate of filling, little effect of the pause on expression would

be expected. Since most mutations that affect promoter function occur in the −10 and −35 elements or upstream, we have made the assumption that mutations at +6 that prevent pausing do not affect the inherent promoter strength. If the +6C mutation did slightly weaken the promoter, for example, then pausing would have somewhat more

813

Pausing and Promoter Function

effect on the wild-type promoter than we have inferred. An important implication of these results is that the purpose of the pause is to provide the substrate that Q modifies, and not to limit the rate of transcription in the absence of Q; if this is true, it makes sense that the promoter strength would roughly match that of the pause. In contrast to the wild-type promoter, expression of the more active trp/pR' /−17 promoter is limited significantly by the +16/+17 pause. Promoter trp/pR' /−17 directs synthesis of four times as much galactokinase without the pause, and open complex formation is about five times as fast without the pause (half-time of 3 versus 16 seconds). The measured halftime of the pause on trp/pR' /−17 DNA is 15 seconds, about the same as the rate of open complex formation on trp/pR' /−17 when the pause is present (16 seconds); this is consistent with the pause limiting availability of the trp/pR' /−17 promoter to RNAP. It is further consistent that for the pR' and trp/pR' /−17 promoter segments the relative amount of paused complex, the last structure on the pathway, is 0.6 (Figure 4(c)), the same as the relative rate of expression of the pR' and trp/pR' /−17 promoters (three versus five units of galactokinase). The determination of pause half-times in vivo requires the assumption that no RNA polymerase escapes the promoter after rifampicin addition. Some RNA polymerase does escape open complexes in vitro for a short time after rifampicin addition; if such escape occurred in vivo, and if it were different for the pR' and trp/pR' /−17 promoters, the pause half-lives for these promoters might be less than measured, and could be identical. However, such escape should have no important effect on the major conclusions of the paper. The significance of the measured level of open complex in cultures growing at equilibrium (pulsed with KMnO4 before rifampicin addition) is less obvious. The steady state level of open complex of the non-pausing versions of both promoter segments is equal, about 45% of saturation. The inference that open complex of trp/pR' /−17 forms four to five times as fast as that for pR' , consistent with the measured relative rate of expression of the promoters, implies that RNAP exits from open complex at trp/pR' /−17 about fivefold faster than from pR' . This conclusion seems to be at odds with the measured concentration of open complex for the pausing versions of the promoters, both about 20 to 25% of saturation. If steady state exists, and if trp/pR' /−17 expresses 1.6-fold faster than pR' , then trp/pR' /−17 should have 1.6/5 or 30% as much open complex, whereas we measure a nearly equal amount. One resolution of this apparent inconsistency might be the existence of more than one type of open complex, which have been found in different ionic conditions in vitro (Suh et al., 1993). For example, RNAP might bind to the promoter beside a paused complex and open the DNA at −11,

yet be incapable of moving because the DNA is blocked; the results would be explained if most of the open complex on trp/pR' /−17 were in this category. A different analysis might detect two classes of enzyme in open complex. We cannot infer directly how much paused complex is present. However, a detailed look at the kinetics of open complex formation for pR' and trp/pR' /−17 (Figure 4 (a), (b)) suggests a conjecture. Accumulation of open complex on trp/pR' /−17 shows a more rapid rise in the first ten seconds, and then settles to a half-time of 16 seconds, about the same as the pause; pR' does not show this initial rapid rise. If the rise represents the ability of trp/pR' /−17 to fill a population of free DNA five fold faster than does pR' , then extrapolation suggests this population is about 15% of the DNA. In this case the distribution for trp/pR' / −17 would be 20% open complex (or variant complex), 15% free DNA, and 65% pause. Since we know that the relative difference in levels of paused species on the trp/pR' /−17 pR' promoter segments is 1.6, the distribution for pR' would be 40% pause, 20% open complex (or variant), and 40% free DNA. Presumably pR' has evolved a rate of function suitable to its role in phage growth, and the fact that the Q antiterminator decreases pausing at +16/+17 is not significant to the rate of promoter function (although it likely is significant to understanding the mechanism of antitermination). However it is clear that the rate of function of a stronger promoter would be increased by suppression of pausing just after the promoter.

Materials and Methods Bacterial strains and plasmids E. coli strain HB101 (hsd S20 (rB mB ) recA13 ara-14 proA2 lacY1 galK2 rpsL20 (Smr ) xyl5 mtl1 supE44 (l− ) was used in all experiments and to propagate all plasmids.

Plasmids (1) pXY306 contains the phage l late promoter (pR' ), including 0500 bp of upstream sequence and the first 49 transcribed bp of the late operon (Yang et al., 1987). (2) pXY306(+6C) is like pXY306 except that it carries a T to C mutation at position +6 of the late operon sequence that abolishes detectable pausing by RNAP at +16/+17 (Yang, 1988; Kainz & Roberts, 1992). (3) pTL17 is similar to pXY306, except that in place of pR' it carries promoter trp/pR' /−17, one of a set of variants made to define the Q-utilization (qut) site. pTL17 has DNA from −31 to −17 of the E. coli trp operon promoter (slightly amended) replacing the corresponding sequence of pR' . The plasmid also lacks about 450 bp of the original pXY306 sequence to the left of −35, so that upstream DNA is different from pR' . As described above, trp/pR' /−17 is more active than pR' , due to the consensus −35 sequence and/or effects of the different upstream sequence. The trp/pR' /−17 sequence from −36 to +1, with original l

814 sequence underlined, and −35 and −10 hexamers in bold, is: CTTGACAATAAATCATCGTATGGGTAAATTTGACTCA (E. Bartlett, M.K., H. Zhou, S. Wirth, and J.W.R., unpublished). (4) pTL17(+6G) carries promoter trp/pR' /−17(+6G), which is like trp/pR' /−17 except that it carries the mutation T to G at +6 which abolishes detectable pausing by RNAP at +16/+17. All plasmids carry the intrinsic (rho-independent) terminator, to , and the galactokinase gene oriented such that expression of galactokinase requires transcription from the test promoter to continue past the terminator. The terminator is not 100% efficient, so that expression of galactokinase in the absence of an antiterminator directly measures function of the test promoter. We assumed that RNA polymerase initiated at all four promoter variants recognizes the terminator with equal efficiency, despite evidence from this laboratory and others that changes in the early transcribed sequence can affect termination efficiency (Goliger et al., 1989; Telesnitsky & Chamberlin, 1989). The agreement between footprinting and kinetic data supports this assumption; furthermore, the most critical comparisons are between transcribed segments of identical sequence. Galactokinase assays Cells bearing original and +6 mutant versions of pXY306 and pTL17 were grown in M-9 glucose with 0.5% (w/v) Casamino acids and 100 mg/ml ampicillin at 37°C to an A600 0.4 to 0.5, and harvested for galactokinase assays. Toluene-treated cells were assayed for galactokinase activity as described by McKenny et al. (1981). Assays were repeated in duplicate from parallel cultures. Galactokinase activity did not vary by more than 15% between parallel cultures in any experiment.

In vivo KMnO4 footprinting Cultures were grown as described for galactokinase assays to an A600 of 0.4 to 0.5, treated with rifampicin (200 mg/ml), and maintained at 37°C. Immediately prior to addition of rifampicin or during a time course of rifampicin treatment 10 ml culture aliquots were treated with 10 mM KMnO4 for ten seconds. The KMnO4 treatment was quenched by transferring cells to ice cold STE buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM EDTA). Plasmid was purified and quantified as described (Kainz & Roberts, 1992). KMnO4 reactive pyrimidines in the template strand of the promoter region were mapped by primer extension on equal amounts of plasmid DNA from KMnO4treated cells, which was quantified by slot-blotting (Kainz & Roberts, 1992). Radioactivity of primer extension products identifying RNAP in open complex and paused at +16/+17 was measured by analyzing dried gels with a Betascope 603 blot analyzer (Betagen).

Pausing and Promoter Function

Acknowledgements We thank S. Wirth, H. Zhou, and E. Bartlett for plasmid constructions, J. McDowell for helpful discussion, and R. Gourse, J. Helmann, and members of the laboratory for criticizing the manuscript. Supported by grant GM 21941 from the NIH.

References Giardina, C., Perez-Riba, M. & Lis, J. (1992). Promoter melting and TFIID complexes on Drosophila genes in vivo. Genes Dev. 6, 2190–2200. Goliger, J. A., Yang, X., Guo, H.-C. & Roberts, J. W. (1989). Early transcribed sequences affect termination efficiency of Escherichia coli RNA polymerase. J. Mol. Biol. 205, 331–341. Grayhack, E. J., Yang, X., Lau, L. F. & Roberts, J. W. (1985). Phage lambda gene Q antiterminator recognizes RNA polymerase near the promoter and accelerates it through a pause site. Cell, 42, 259–269. Kainz, M. & Roberts, J. W. (1992). Structure of transcription elongation complexes in vivo. Science, 225, 838–841. Landick, R. & Yanofsky, C. (1987). Transcription attenuation. In Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology (Neidhardt, F., Ingraham, J., Low, B., Magasanik, B., Schaechter, M. & Umbarger, H., eds), pp. 1276–1301, Am. Soc. Microbiol., Washington, DC. Liu, B., Wong, M. L., Tinker, R. L., Geiduschek, E. P. & Alberts, B. M. (1993). The DNA replication fork can pass RNA polymerase without displacing the nascent transcript. Nature, 366, 33–39. McKenney, K., Shimatake, H., Court, D., Schmeissner, U., Brady, C. & Rosenberg, M. (1981). A system to study promoter and terminator signals recognized by Escherichia coli RNA polymerase. In Gene Amplification and Analysis of Nucleic Acids by Enzymatic Methods (Chirikjian, J. G. & Papas, T. S., eds), vol. 2, pp. 383–415, Elsevier, North Holland. Rougvie, A. & Lis, J. (1988). The RNA polymerase II molecule at the 5' end of the uninduced hsp70 gene of D. melanogaster is transcriptionally engaged. Cell, 54, 795–804. Suh, W., Ross, W. & Record, M. (1993). Two open complexes and a requirement for Mg2+ to open the l PR transcription start site. Science, 259, 358– 361. Telesnitsky, A. P. W. & Chamberlin, M. J. (1989). Sequences linked to prokaryotic promoters can affect the efficiency of downstream termination sites. J. Mol. Biol. 205, 315–330. Yang, X. (1988). Transcription antitermination mediated by lambdoid phage Q proteins. PhD dissertation. Cornell University. Yang, X., Hart, C. M., Grayhack, E. J. & Roberts, J. W. (1987). Transcription antitermination by phage l gene Q protein requires a DNA segment spanning the RNA start site. Genes Dev. 1, 217–226.

Edited by K. Yamamoto (Received 3 February 1995; accepted 11 August 1995)