Abortive initiation of transcription at a hybrid promoter. An analysis of the sliding clamp activator of bacteriophage T4 late transcription, and a comparison of the sigma70 and T4 gp55 promoter recognition proteins.

Bacteriophage T4 late promoters are transcribed by an RNA polymerase holoenzyme comprising the Escherichia coli core, E, the phage gene 55-encoded promoter recognition subunit, gp55, and the gene 33-encoded co-activator, gp33. Transcriptional initiation is activated by the T4 gene 45-encoded sliding clamp, which is loaded on to DNA at enhancer-like sites by its clamp-loader. Correct initiation of transcription at late promoters in basal mode requires only RNA polymerase core and gp55 (E.gp55). Dinucleotide-primed abortive initiation of basal and activated T4 late transcription has been compared. Only the trinucleotide non-productive transcript is made at a high rate; all other short transcripts are made at rates of less than one molecule per productive transcript. Gp45 increases abortive trinucleotide synthesis along with productive transcription, although the proportion of productive transcripts is also elevated. Nevertheless, this increase accounts for only a small part of the activation of T4 late transcription that is generated by its activator and co-activator. The pattern of production of short transcripts differs subtly between basal and enhanced transcription, indicating that linking the RNA polymerase with its sliding clamp activator only generates minor changes in the transition from abortive to productive RNA chain elongation. The T4 late promoter is converted to a strong sigma70 promoter by inserting an appropriate -35 promoter element. A direct comparison at such a hybrid promoter shows sigma70 and gp55 generating qualitatively and quantitative different patterns of abortive initiation at the same start site.

Bacteriophage T4 late promoters are transcribed by an RNA polymerase holoenzyme comprising the Escherichia coli core, E, the phage gene 55-encoded promoter recognition subunit, gp55, and the gene 33-encoded coactivator, gp33. Transcriptional initiation is activated by the T4 gene 45-encoded sliding clamp, which is loaded on to DNA at enhancer-like sites by its clamploader. Correct initiation of transcription at late promoters in basal mode requires only RNA polymerase core and gp55 (E⅐gp55). Dinucleotide-primed abortive initiation of basal and activated T4 late transcription has been compared. Only the trinucleotide non-productive transcript is made at a high rate; all other short transcripts are made at rates of less than one molecule per productive transcript. Gp45 increases abortive trinucleotide synthesis along with productive transcription, although the proportion of productive transcripts is also elevated. Nevertheless, this increase accounts for only a small part of the activation of T4 late transcription that is generated by its activator and co-activator. The pattern of production of short transcripts differs subtly between basal and enhanced transcription, indicating that linking the RNA polymerase with its sliding clamp activator only generates minor changes in the transition from abortive to productive RNA chain elongation. The T4 late promoter is converted to a strong 70 promoter by inserting an appropriate ؊35 promoter element. A direct comparison at such a hybrid promoter shows 70 and gp55 generating qualitatively and quantitative different patterns of abortive initiation at the same start site.
DNA-dependent RNA polymerases are obliged to carry out prodigious feats of processive polymerization when they generate long transcripts, because release of the nascent RNA chain at any step of elongation is essentially irreversible. In contrast, the first steps of RNA chain elongation by these enzymes are tentative. Short transcripts are rapidly generated and released in a repetitive process in which the enzyme does not leave the transcriptional start site (1)(2)(3). It appears therefore that the generation of these abortive transcripts is part of the process of undocking polymerase from the promoter. The distribution of very short transcripts generated at different Escherichia coli promoters differs, in some instances strikingly (4), reflecting large-scale differences in the ability to clear different promoters that are caused by start site-proximal DNA sequence.
Undocking from the promoter takes place in two fundamentally different ways. For the bacterial polymerase autonomously recognizing and opening a promoter, leaving that site is also autonomous. The eukaryotic nuclear RNA polymerases are recruited to their promoters by homologous transcription initiation machineries through protein-to-protein contacts, and undocking from the promoter breaks those contacts. When it departs CAP-activated promoters, E. coli RNA polymerase undoubtedly also severs connections, in that instance between its ␣-subunit and the activator, at an early step of promoter clearance. Many other activators of bacterial transcription must generate similar circumstances. It is important to know whether these events confer characteristic signatures on the pattern of generation of abortive transcripts. In the case of transcriptional activators, it is also necessary to examine whether activation changes the relative yield of productive and abortive transcripts.
The experiments that are reported here concern bacteriophage T4 late promoters, whose activation is generated by the gp45 sliding clamp of the T4 DNA polymerase holoenzyme. The unique mechanism of this activation requires loading of gp45 onto DNA at entry sites such as primer-template junctions or single-stranded DNA gaps by its clamp loader, the gp44⅐62 complex. Once loaded onto DNA, the donut-shaped gp45 trimer slides along DNA by one-dimensional diffusion. The ligands of gp45 include two phage T4-encoded subunits of the T4 late RNA polymerase holoenzyme: the late promoter-recognition protein gp55 and the gp33 co-activator of late transcription (5)(6)(7). It is through these interactions that gp45 becomes part of the transcription initiation complex at the open T4 late promoter. In the T4-infected E. coli cell, transcription of the late genes requires gp55, gp33, and gp45. Replication-dependence and replication-coupling of late transcription in vivo is thought to be related to the requirement for continuously reloading gp45 onto DNA (8).
Correct and productive initiation of transcription at T4 late promoters is also generated in vitro by E. coli RNA polymerase core enzyme, E, and gp55 (9). In the experiments that are described below, we compare this basal transcription, in which RNA polymerase operates autonomously, with activated transcription, in which RNA polymerase is constrained by its sliding clamp, in regard to abortive initiation of transcription.
The -family subunits of bacterial RNA polymerases can also play an important role in determining how the transcription complex leaves the promoter. For example, a start-proximal downstream binding site on the non-transcribed DNA strand for region 2.4 of E. coli 70 generates an extended promoterproximal pausing event that is an essential component of the regulation of bacteriophage late transcription (10,11). It is conceivable, therefore, that individual proteins generate characteristic patterns of abortive transcription. Examining this issue requires a promoter at which two proteins can direct initiation of transcription at the identical nucleotide. We have modified a T4 late promoter for this purpose and report a comparison of abortive initiation generated with E. coli 70 and T4 gp55.

EXPERIMENTAL PROCEDURES
Proteins and Chemicals-DNA exonuclease III, Taq DNA polymerase, restriction enzymes, DNase I, proteinase K, nucleotides, as well as RNA and DNA oligonucleotides were purchased. T4 AsiA protein was generously provided by M. Ouhammouch and E. N. Brody. The purification of T4 gp32, gp45, and gp44⅐62 complex from overproducing bacteria was described and referenced previously (12,13). E. coli strain RL721, which encodes an RNA polymerase ␤Ј subunit with a hexahistidine affinity tag at its normal C terminus, was provided by R. Landick and K. Severinov. We also thank J. Abelson for a gift of tri-and tetraribonucleotides.
DNA-Plasmid pRT510-Cϩ18 (15), a 3283-bp 1 derivative of pDH310 (16), contains a derivative of the T4 gene 23 late transcription unit further modified to yield a transcript with a 17-nt, C-less 5Ј end. In addition to the T4 late Ϫ10 consensus sequence, TATAAATA, this transcription unit also has the Ϫ35 consensus sequence, TTGACA, of E. coli 70 . The "No Ϫ10" variant version of pRT510-Cϩ18 has the TATA-AATA Ϫ10 sequence replaced with GAACTGAC. Relevant partial sequences of both constructs are shown in Fig. 1. DNA for transcription was made by polymerase chain reaction to produce 880-bp doublestranded DNA (bp Ϫ374 to ϩ406, relative to the transcriptional start site as ϩ1), and treated with exonuclease III to generate 100 to 200 nt 5Ј overhanging ends, on average, as described (8). DNA was subsequently cut with the blunt end-generating SmaI nuclease, and the required DNA fragment containing the T4 late promoter was recovered using an immobilized short oligonucleotide complementary to its 5Јoverhanging single-stranded region.
Abortive Transcription-Transcription assays were performed at 25°C in a transcription buffer (240 mM potassium acetate, 10 mM magnesium acetate, 33 mM Tris acetate, pH 7.8, 1 mM dithiothreitol, 0.1 mM EDTA, 100 g/ml bovine serum albumin) containing 5 nM DNA, 50 g/ml gp32, 200 M dATP, 23 nM RNAP core, 92 nM gp55, and, when appropriate, 276 nM gp45 (trimer), 298 nM gp44⅐62 complex, and 92 nM gp33, in a 10-l volume. DNA was first mixed with dATP and gp32 and stored on ice before combining with a mixture of the other proteins. After preincubation for 3 to 10 min (in different experiments), transcription was initiated by adding NTPs (final concentrations: 100 M GpA, 5 M ATP, 5 M GTP, and 5 M [␣-32 P]UTP) to the DNA-protein mixture. "Chase" RNA chain elongation was done with all four NTPs at 1 mM final concentration of each nucleotide, or with 3Ј O-methyl-CTP (Pharmacia; regrettably no longer available commercially). Transcription was stopped by heating at 90 -95°C for 5 min; samples were subsequently treated with 1.5 units of shrimp alkaline phosphatase for 50 min at 37°C, transferred to dry urea (0.3 mg/l of sample), heated in boiling water for 3 min, then loaded onto 25% polyacrylamide denaturing gels (22:3 acrylamide:bisacrylamide, 7.8 M urea, 89 mM TBE) for electrophoresis. A Phosphoimager was used for quantification. Background correction for weak bands in the vicinity of strong bands followed Matsuzaki et al. (17).

RESULTS
The Pattern of Transcript Accumulation-The T4 late transcription unit from pRT510-Cϩ18 (5), which has been used for these experiments, has the consensus T4 late TATA box Ϫ10 site, TATAAATA, and the transcriptional start region of T4 gene 23 (which encodes the major phage head protein), reconstructed to generate transcripts with a 17-nt C-less 5Ј end (Fig.  1a). When open complexes at this promoter are presented with the dinucleotide primer GpA, ATP, UTP, and GTP, they generate nascent 17-mer transcripts that are arrested before CCC . . . (bp [18][19][20] and can be elongated subsequently with the missing nucleotide. Linear DNA was prepared for activated transcription by digestion with exonuclease III to create a primer-template junction downstream of the promoter that serves as the DNAloading site for the T4 gp45 sliding clamp transcriptional activator. DNA loading of gp45 by the T4 gp44⅐62 complex requires dATP (or ATP) hydrolysis and was done in the presence of an excess of T4 gp32 single-stranded DNA-binding protein. The T4 late promoter is recognized by the E. coli RNA polymerase core enzyme, E, in combination with the T4 late promoter-specific -family initiation protein gp55. Transcription in basal mode was done with E⅐gp55 only, and was also examined under conditions of partial repression by gp33. Gp45-activated (enhanced mode) transcription required gp33 and gp44⅐62 complex in addition to gp55. Fig. 1b shows the transcripts generated in an enhanced mode transcription experiment. DNA was incubated with all proteins at 25°C for 3 min, then GpA, ATP, GTP, and UTP were added for 15 min. The [ 32 P]UMP-labeled oligonucleotides generated during this time were treated with alkaline phosphatase to convert [␣-32 P]UTP to 32 P i before being resolved on a 25% polyacrylamide gel. Initial phosphodiester bond formation was dependent on priming by the dinucleotide GpA when ATP, UTP, and GTP were kept at 5 M, as expected (data not shown). The distribution of products was dominated by two bands, B and N, with smaller quantities of other components. Band N is the nascent 17-mer transcript; it was extended by 1 nt in the presence of 3Ј O-methyl-CTP. (It was also slowly extended in the absence of 3Ј O-methyl-CTP by error incorporation of UMP, yielding band NЈ, barely apparent in Fig. 1b.) The identities of other bands were established in two ways: 1) extraction of material from the gel and resolution by thin layer chromatography with non-radioactive markers (GpApU, GpApUpA, etc.; data not shown); 2) abortive transcription primed with the same series of (longer) oligoribonucleotides and one [␣-32 P]NTP, or with GMP and [␣-32 P]ATP. These experiments established that band B contains the trinucleotide GpApU and pentamer GpApUpApU, band A the tetramer GpApUpA, band C the hexamer GpApUpApUpG. Bands D and E constitute the 7-mer GpApUpApUpGpA, and 8-mer GpApU-pApUpGpApA, respectively (data not shown). The ladder series of 9-mer to 17-mer bands can be counted up from band F to band N. Unlabeled GpA migrates between bands C and D. Further proof that the principal material of band B is GpAp*U (the asterisk indicating the 32 P label) was provided by showing that treatment with RNase T1 converted it to material comigrating on a thin layer plate with Ap*U; ligation of pAp to material from band B with RNA ligase and subsequent treatment with shrimp alkaline phosphatase converted it to a compound co-migrating with unlabeled GpApUpA (data not shown).
Evidence that band B also contains GpAp*UpAp*U penta-mer was provided by the observation that the principal Ulabeled transcript generated with GpApUpA primer and labeled UTP co-migrates with band B (data not shown). This pentanucleotide was identified as a minor product in band B from an experiment like Fig. 1b by eluting material and further separation by thin layer chromatography. By quantifying the distribution of radioactivity, we determined that the 5-mer product constituted ϳ0.5% of the 3-mer product in band B (on a molar basis). The arrested transcription products also included an ϳ24-nt transcript (band O, Fig. 1b) whose provenance has not been determined except to show that it is arrested before C, since it was extended by 1 nt in the presence of 3Ј O-methyl-CTP. The Source of Minor Transcripts-Since the transcripts in bands D to M are all present in small quantities relative to the arrested 17-mer transcript, we examined whether they were in fact generated by initiation at the T4 late promoter with E⅐gp55. This was done in two ways, first, by comparing RNA synthesis by RNA polymerase core, E, and by E⅐gp55 (Fig. 1c), and second, by comparing transcription by the E⅐gp55 enzyme of DNA, respectively mutated in and wild type for the late promoter Ϫ10 TATAAATA box (Fig. 1d). The outcome of these experiments indicated that transcripts in bands F, G, and L predominantly do not come from the late promoter. The 15-mer product in band L was seen to be a transcript of TATA box-less DNA (Fig. 1d) and it was made by RNA polymerase core purified through two cycles of chromatography on Bio-Rex to remove 70 (Fig. 1c). This 15-mer transcript was extended by 1 nucleotide in the presence of 3Ј O-methyl-CTP and must therefore have been arrested for lack of CTP in the original reaction mixture. A search of the DNA sequence suggests that the 15-mer could be a nonspecific transcript of single-stranded DNA that was generated by exoIII treatment to create the gp45-loading site. Perfect duplex DNA did not yield the 15-mer transcript (data not shown), which is consistent with that conjecture.
The analysis shown in Fig. 1d similarly suggested that the relatively rare 9-mer and 10-mer transcripts (bands F and G) and at least a fraction of the 11-mer band H transcript do not come from the intended late transcription unit, because they were generated at comparable levels in transcription of "No Ϫ10" DNA, and were arrested for lack of CTP (Fig. 1d). Band E comprised two partly resolved 8-mer transcripts; only the slower-migrating component is specific to the late promoter. Whereas the 9-mer (band F) in basal mode transcription conditions was clearly nonspecific, the extremely weak 9-mer gen- erated under enhanced mode conditions had a distinctly higher mobility, was not extended in the presence of 3Ј O-methyl-CTP (Fig. 1c), and was specific to the late promoter (data not shown). Presumably enhancement was required for this specific product to reach the threshold of detection.
Transcripts That Are Not Elongated-The experiments summarized in Fig. 2 examine what happens to the distribution of radioactivity in GpA-initiated transcripts upon addition of a vast excess (1 mM) of all four unlabeled ribonucleoside triphosphates. Radioactivity that was not chased into longer transcripts resided in bands B, C, D, and E (3-, 6-, 7-, and 8-mer, respectively). For enhanced transcription, the band E 8-mer transcript that was not elongated in the chase came from the late promoter, and the elongating transcript (Fig. 2, asterisk) did not.
The most prominent feature of these non-chased 6-, 7-, and 8-mer products of enhanced transcription is that they were produced in small quantities relative to the 17-mer arrested transcript; only the 3-mer GpAp*U was made at a very high rate. The effect of nucleotide concentration on the production of these transcripts was briefly examined. Under the standard conditions of the experiment (i.e. with 5 M each of GTP, ATP, and [␣-32 P]UTP) varying GpA concentration between 50 and 400 M generated less than 25% variation in radioactivity incorporated into band B (data not shown). Increasing all concentrations of substrates in proportion, from 100 M GpA with 5 M each of the three ribonucleoside triphosphates to 800 M GpA with 40 M each of the three triphosphates, did not significantly change the distribution of radioactivity in short transcripts (data not shown).
Comparing Enhanced and Unenhanced Transcription-The next experiments examine whether attachment of the T4 late RNA polymerase to the sliding clamp transcriptional activator changes the distribution of transcripts, especially of non-chaseable (abortive and/or dead-end) transcripts. Only minor differences in the products generated during the transcription pulse were noted (Fig. 3a). In basal as in activated transcription, the distribution of shorter transcripts was dominated by band B. Only the 6-mer (band C) was produced in quantities comparable with 17-mer. The proportion of 3-mer to 17-mer was approximately 2-fold higher for basal than for activated transcription (Fig. 3a). A comparison of the size and abundance distributions among the non-chaseable transcripts for basal and enhanced transcription (Fig. 3b) shows, somewhat surprisingly, a greater proportion of 6-mer, 7-mer, and 8-mer relative to 3-mer transcripts in enhanced transcription. On a per-17mer basis, the yields of all these non-chased products were low: 0.22, 0.19, and 0.06 molecules for the 6-mer, 7-mer, and 8-mer, respectively, in unenhanced transcription and 0.84, 0.50, and 0.14, respectively, for enhanced transcription (after correcting for products generated by core enzyme alone; average of four determinations). Further examination of the data in Figs. 1c, 2, and 3 shows that the fraction of 6-mer to 9-mer transcripts that can be chased into longer products was 4 -6-fold greater for basal transcription than for enhanced transcription. Transcripts of this length that are arrested before C do not come from the late promoter in pRT510-Cϩ18. The significance of this result is that gp45 also increases specificity on this DNA template, probably by activating transcription at the T4 promoter more efficiently than at nonspecific sites.
The time course of accumulation of 17-mer and band B transcripts is compared in Fig. 4. Activation of transcription by gp45 clearly increased the yield of the abortive 3-mer and productive 17-mer transcripts alike, but also increased the relative proportion of 17-mer to 3-mer products. If every open promoter complex were converted to ternary transcription complexes arrested at bp 17, the production of GpAp*U should eventually cease. The fact that this did not happen indicates that some RNA polymerase molecules did not leave the promoter, either in enhanced or unenhanced transcription.
Comparing RNA Polymerase Core Enzymes-Short oligonucleotides are generated in transcription not only by abortive or arrested chain elongation, but also by hydrolytic retraction of RNA chains. In E. coli RNA polymerase, hydrolytic retraction requires GreA or GreB protein (18,19). To establish whether the abortive and other short transcripts show indications of contribution from RNA chain retraction, we compared our purified E. coli RNA polymerase with enzyme reconstituted from its ␣, ␤, and ␤Ј subunits (generously made available by A. Goldfarb, E. Nudler, and K. Severinov). We noted only one minor difference in enhanced transcription by these two enzymes: a late promoter-specific 8-mer non-chaseable transcript that was generated by the purified E. coli RNA polymerase at a low rate (Fig. 4b), was not generated by the reconstituted enzyme (data not shown).
The RNA polymerase ␣ subunit is modified after phage T4 infection in its C-proximal domain, at Arg-265 by ADP-ribosylation. This C-terminal domain is far removed from the catalytic site (20) and it does not participate in activation of T4 late transcription by gp45 (21). T4-modified RNA polymerase also contains a T4-encoded subunit, RpbA (22), which has been shown to confer salt sensitivity on ternary complexes formed with T4-modified RNA polymerase (23) and could conceivably affect the initiation phase of transcription. Our T4-modified RNA polymerase core, prepared from T4-infected cells, is contaminated with residual amounts of 70 . Because the T4 late promoter in pRT510-Cϩ18 also contains a Ϫ35 consensus TT-GACA element (Fig. 1a), it is also transcribed by E⅐ 70 RNA polymerase (5). To suppress such 70 -mediated transcription, the T4-modified RNA polymerase was supplemented with the 70 antagonist AsiA (24 -26) along with all the components required for T4 late transcription. Total and abortive transcripts generated by the unmodified and phage T4-modified RNA polymerase core enzymes in gp55-dependent enhanced late transcription did not differ significantly (data not shown).
Comparing -Proteins: 70 and gp55-The insertion of a consensus Ϫ35 site at the T4 late promoter in pRT510-Cϩ18 (Fig. 1a), makes this a strong promoter for the E. coli E⅐ 70 holoenzyme. We compared 70 -and gp55-generated abortive and productive transcripts in the experiment that is summarized in Fig. 5. Transcription in the presence of GpA, ATP, GTP, and UTP yielded the 17-mer transcript, indicating initiation at the same site with E⅐gp55 and E⅐ 70 . Nevertheless, E⅐ 70 yielded short and non-chased transcripts in an entirely different pattern. Much less band B transcript was produced relative to 17-mer by E⅐ 70 than in enhanced transcription by E⅐gp55, and the yield of band D transcript, the 6-mer, relative to band B transcript (predominantly 3-mer) was much higher for E⅐ 70 than for E⅐gp55. Control transcription of "No Ϫ10" DNA (Fig. 5, two bottom traces) indicated that these E⅐ 70 transcripts came almost exclusively from the intended site.

DISCUSSION
One non-productive transcript, the trinucleotide, is made at an exceptionally high rate when the T4 late holoenzyme, E⅐gp55, initiates transcription at the T4 late promoter (with the dinucleotide GpA) (Fig. 1). Addition of that first nucleotide to GpA is clearly non-processive, since production of all other transcripts is very low compared with the yield of GpApU (Figs.  1 and 4). Adding three more nucleotides drastically changes FIG. 3. Relative yields of short and productive transcripts in basal and enhanced transcription by E⅐gp55. a, the pulse (RNA polymerase and DNA preincubated for 3 min; RNA synthesis with GpA, ATP, GTP and UTP for 6 min). b, the chase (continued synthesis for 5 min after addition of all four NTPs to 1 mM each). In each distribution, incorporation by core enzyme (E) alone is subtracted. In each panel, the amount of radioactivity in band B (designated 3, 5 for its 3-mer and 5-mer content) is also normalized between basal and enhanced transcription so as to emphasize the size distribution. The pulse and chase data are from the same experiment.
processivity: an estimate is arrived at by assuming that all 6-, 7-, and 8-mer transcripts are abortive rather than arrested, the yields of these products relative to 17-mer signifying a processivity greater than ϳ0.7 for addition of the 7th and 8th nucleotides. The rarity of 4-mer and 5-mer transcripts (relative to 6-mer) also implies an especially low processivity confined to the first step of chain elongation. It is interesting to compare these findings with the situation encountered in transcription of the SUP4 tRNA Tyr gene by yeast RNA polymerase III, which is brought to its promoter by the DNA-bound transcription factor IIIB. The yield of all the abortive pol III transcripts, including the initially formed pppApA, is low, implying a relatively high processivity (greater than ϳ0.5) even for the earliest step of RNA chain elongation, 2 and showing that translocation of the polymerase or of its transcription bubble along the DNA template (27) is not a prerequisite of processive RNA chain elongation.
Transcriptional activation by the gp45 sliding clamp clearly increases abortive production of GpApU along with productive 17-mer synthesis (Figs. 1c and 4). One might have considered, a priori, that transcriptional activation by gp45 could operate primarily to relieve a block on the transition from abortive to productive transcription. That is evidently not the case; although the proportion of 17-mer transcript to abortive GpApU is increased by the activator and co-activator, the change is quantitative (up to ϳ3-fold in Figs. 3 and 4) rather than absolute, and clearly cannot account for more than a small part of the very large transcriptional activation achievable in vitro (28,29). A contrasting example of activator action is provided by recent experiments on regulation of T4 middle genes (30). T4 middle promoters are under positive control of the MotA and AsiA proteins. MotA binds to its "Mot box" DNA site 30 bp upstream of the transcriptional start. AsiA, the "anti-sigma" protein first identified by Stevens (26) binds to 70 , blocks T4 early transcription by E⅐ 70 , and is the essential co-activator of MotA for transcription at T4 middle promoters (reviewed in Ref. 24). MotA and AsiA generate their activation of middle transcription partly by recruiting RNA polymerase to the promoter and partly by facilitating promoter escape. In the absence of AsiA, there is substantial abortive initiation at a middle promoter, but almost no production of full-length transcripts (30). The contrasting results presented here specify that facilitating the escape from the T4 late promoter can only be a small part of transcriptional activation by gp45 and gp33. The sliding clamp, which interacts directly with gp55 and gp33 (6 -8) appears to be a device designed to bring RNA polymerase to the promoter. Whether it also facilitates subsequent steps leading to promoter opening is the subject of ongoing work.
We also examined the effect of gp33, which suppresses basal transcription by E⅐gp55 (12), on the generation of short transcripts. Gp33 suppressed GpApU synthesis along with production of 17-mer RNA (data not shown). Evidently, the suppression of basal transcription by gp33 is also not primarily generated by blocking the transition from abortive to productive transcription.
All other short transcripts are made by E⅐gp55 at a rate of less than one molecule per productive transcript (Fig. 3) during the time interval in which all the productive transcripts are generated (Fig. 4). A proportion of these other (longer) short transcripts are not extended upon provision of all four ribonucleoside triphosphates. Because they are relatively rare, we have not determined whether they are truly abortive (released from polymerase and DNA) or arrested (remaining in a ternary transcription complex but failing to elongate promptly despite the presence of the appropriate substrate). The patterns of production of these short non-chaseable transcripts show subtle differences between enhanced and basal transcription. There appears to be greater production of 6-mer and 7-mer transcript per 17-mer transcript in enhanced transcription at the T4 late promoter relative to basal transcription, perhaps reflecting a gp45 interaction with RNA polymerase in transition from the start site-confined holoenzyme to the elongation complex. It is interesting to note a somewhat similar effect at 2 P. Bhargava and G. A. Kassavetis, manuscript in preparation. middle promoters: in addition to their key effects on promoter clearance, MotA and AsiA also generate 8-, 10-, 11-, and 12-mer transcripts as a sort of promoter-leaving signature of their transcriptional activation (30).
If all open promoter complexes were capable of leaving the transcriptional start site and moving down the DNA template, production of the abortive GpApU transcript should cease once all 17-mer nascent transcripts have accumulated. This is not quite what happens, although there is an initial burst of production of trinucleotide. The ultimate continuation of trinucleotide production suggests that a fraction of polymerase molecules do not or cannot leave the promoter. The experiments that are presented here are not well suited for measuring this proportion accurately, but we estimate it as not greater than 15-20%. A considerably higher proportion of similarly incompetent RNA polymerase molecules were noted in preparations of E. coli 70 -holoenzyme (31).
E. coli RNA polymerase generates short oligonucleotides hydrolytically by a process that is controlled by the GreA and GreB proteins. We have shown that this action does not make a substantial contribution to the generation of short oligonucleotides by E⅐gp55 at the T4 late promoter, noting only the 8-nt non-chaseable transcript as possibly being generated by GreBmediated hydrolysis. The post-infection modification of RNA polymerase core (in its ␣ subunit) and attachment of the RpbA subunit had no effect on the generation of abortive transcripts.
The distribution and abundance of abortive transcripts, and the ability of polymerase to undock and form a stable, productive elongation complex varies markedly between promoters, due not only to the influence of initially transcribed sequence, but also responding to differences of polymerase affinity for the promoter (32). Differences of abortive initiation between E⅐ 54 at its cognate glnAp2 promoter and E⅐ 70 at the lac UV5 and T7A1 promoters suggested to Tintut and co-workers (33) that differences of affinity of polymerase core for different -subunits may also influence the abortive phase of transcription. These comparisons involve situations in which the effects of promoter sequence, initially transcribed sequence, and polymerase constitution are convoluted. The construction of the hybrid pRT510 promoter allowed us to make a head-to-head comparison of 70 and gp55 driving abortive initiation at exactly the same site. To our knowledge, this is the first such direct comparison between proteins. Gp55 and 70 generate different signatures in abortive initiation, E⅐ 70 holoenzyme producing a smaller proportion of non-chased short transcripts relative to the productive 17-mer than did E⅐gp55. Moreover, the non-chased transcripts made by E⅐ 70 were differentiated by a relatively high yield of 6-mer relative to the 3-mer (and 5-mer) band B (Fig. 5). There is a clear correlation here with previously observed differences of 70 -directed and gp55-directed opening of this promoter (5): the reactivity of DNA in the vicinity of the transcriptional start to KMnO 4 at the promoter is differently distributed when E⅐ 70 and E⅐gp55 form open complexes at the hybrid promoter in pRT510. The difference from 70 is due to gp55 itself rather than gp33 and gp45 (compare Figs. 5 and 2), also consistent with the fact that basal and gp45-activated open promoter complexes of E⅐gp55 are not distinguishable by permanganate footprinting (5). That 70 could influence abortive initiation is consistent with the loca-tion of its segment 3.2 in the vicinity of the transcriptional start site (34,35). Since gp55 is much smaller than 70 and highly diverged from it in amino acid sequence, it may not reach to every site that is occupied by 70 and may make different contacts with the core enzyme.