ABSTRACT
Transcription initiation is accompanied with iterative synthesis and release of short transcripts. The molar ratio of enzyme to template was found to be critical for the amounts and distribution of the abortive products synthesized by Escherichia coli RNA polymerase from several promoters. At a high ratio abortive synthesis of 4-8 nt were enhanced at the [lambda]PR promoter. Removing excess RNA polymerase just before initiation, achieved by washing immobilized transcription complexes, prevented this enhancement. At this high ratio synthesis of an unexpected 6 nt transcript was enhanced when the enzyme stalled at position +32, but not when it stalled at position +73. This transcript had misincorporations at its fifth and sixth positions, probably due to slippage. Hydroxyl radical footprinting of the complex stalled at +32 in the presence of excess enzyme showed that more than one molecule of RNA polymerase was tandemly bound to the same DNA. These results suggest that: (i) when RNA polymerase molecules are tandemly transcribing the same DNA, transient collisions enhance abortive synthesis by the trailing molecule; (ii) when the leading polymerase stalled in the initially transcribed region blocks progression of the trailing polymerase, the latter can commit misincorporations, probably due to slippage synthesis.
Physical interaction between proteins plays an essential role in signal transduction within the transcription machinery. Studies on prokaryotic transcription have mainly focused on contact of activators or repressors with RNA polymerase (for a review 1 ). However, possible contributions of interactions between RNA polymerase molecules engaged in RNA synthesis within the same transcription unit have not been extensively examined. At most 70 molecules of RNA polymerase simultaneously occupy a single ribosomal RNA transcription unit of 5.5 kb (2 ). This corresponds to 79 bp as an average minimum center-to-center distance between adjacent RNA synthesizing enzymes in this most intensively transcribed of all known transcription units in Escherichia coli. At first sight, this degree of physical separation seems large enough to argue against interaction between successive RNA polymerase molecules. However, the elongation speed is not homogeneous within a transcription unit. The rate of incorporation of the second to fourth nucleotides from the T7A1 promoter, 6-20 nt/s (3 ), is much slower than the averaged elongation rate of 30-135 (see 2 ). [lambda] late transcription pauses at the 16/17 positions (4 ). If there is a pause or a slowdown in elongation near the initially transcribed region, interactions between two RNA polymerase molecules become possible. A characteristic transcription in this region is abortive synthesis (5 -7 ); an iterative cycle in which short transcripts are synthesized and released. This leads us to the idea that the interaction can affect abortive synthesis.
Abortive synthesis has been reported in both prokaryotic and eukaryotic transcription (8 -13 ), but its biological significance has not been clarified. Most studies have been performed using E.coli RNA polymerase in vitro and the phenomenon has been observed on most of the promoters that have been examined (14 ). Little is known about the factors determining the amounts and distributions of abortive products. Here we demonstrate that one important factor is the ratio of RNA polymerase to promoter.
We recently found that a significant fraction of transcription complexes at the [lambda] PR unit is trapped in the abortive cycle and never escapes from it. This suggests the concept that the yield of long RNA depends on the fraction of transcription complexes which is untrapped in this non-productive cycle (15 ). Here we have examined the concentration effect of RNA polymerase on abortive transcription from several promoters. A high molar ratio of RNA polymerase to template DNA increased abortive products as compared with the full-length product. This suggests that two RNA polymerase molecules might interfere with each other within the same transcription unit, thereby resulting in enhanced abortive synthesis. Washing away excess RNA polymerase just before the start of RNA synthesis abolished this enhancement. Moreover, when the leading RNA polymerase molecule was stalled close enough to block the trailing one at the promoter, a short abortive RNA bearing a high degree of misincorporation was specifically synthesized.
Nucleoside triphosphates were purchased from Yamasa (Tokyo). [[gamma]-32P]GTP (4500 Ci/mmol) was purchased from ICN Biomedicals (Costa Mesa). Methylene and imido analogs of ATP were obtained from Sigma, thio analogs were from Boehringer Mannheim and 3'-dATP was from Pharmacia. Oxirane acrylic beads were from Eupergit One Micron (Rhöm, Darmstadt) and all other materials were as described in Fujioka et al. (16 ).
The 160 bp fragment harboring the T7A1 promoter was prepared by PCR from the linearized plasmid pAR1345 (17 ). The 205 bp fragment containing the lacUV5 promoter was prepared as described previously (15 ). Four DNA templates containing the [lambda] PR promoter were used (Fig. 1 ); template 1 encoded the original [lambda] PR mRNA sequence, templates 2 and 3 contained 32 nt A-less leader cassettes and template 4 had a 73 nt A-less leader cassette. Template 1, a 540 bp fragment, was prepared by HindIII and BglII digestion of p[lambda]PR1 (16 ). Plasmid p[lambda]PR1AL32 was constructed from p[lambda]PR1 by insertion of an oligo-duplex (the individual strands were 34 and 38 nt in length) between the BglII (blunted) and Nsp(7524)I sites. The plasmid was linearized by digestion with HindIII and the ends were biotinylated as previously described (16 ). Further digestion by NruI produced template 2, a 1130 bp fragment with a biotinylated end (the HindIII end). Plasmid p[lambda]PR1AL73 was constructed by polymerase chain reactions using p[lambda]PR1AL32 as template. A region extending upstream of the A-less leader sequence to the NdeI site (2039 bp) was first amplified with a pair of primers, 5'-GACTAGCGGCCGCCGTGTTCTCTGGCGGTTGTC-3' and 5'-CACCGCATATGGTGCACTCT-3'. The remaining region of p[lambda]PR1AL32, from the NdeI site to the downstream part of the A-less leader sequence (2584 bp) was amplified with another pair, 5'-ATCAGGCGGCCGCAAGCGGCGAGCCAGACAACC-3' and 5'-TGCACCATATGCGGTGTGAA-3'. The two amplified fragments have NdeI sites at one end and NotI sites in the A-less region at the other end. After digestion by NdeI and NotI, the two fragments were ligated together to make p[lambda]PR1AL73, which contained two tandemly repeated 32 nt A-less leader sequences separated by the NotI site, making up a 73 nt A-less leader sequence. Plasmid p[lambda]PR1AL73 was linearized by digestion with SalI and the end was biotinylated. Subsequent digestion by HindIIIproduced template 4, which is 877 bp long.Template 3 was prepared for footprinting by amplifying template 2 with primers 5'-CCGTGCGTCCTCAAGCTGCTC-3' and 5'-GGGCGTAGAGGATCTGCGCCC-3'. The primer of the non-transcribed strand was labeled by T4 polynucleotide kinase at its 5'-end.
All of the DNA fragments were purified by agarose or polyacrylamide gel electrophoresis and elution. Preparation of avidin-immobilized resins and immobilization of DNA fragments was according to Fujioka et al. (16 ). The concentration of free DNA was determined by UV absorption at 260 nm, assuming a coefficient of E1% = 200/cm. Immobilized DNA was released by phenol extraction and its concentration was determined by comparing the densities of ethidium bromide stained bands with those of DNA standards.
RNA polymerase was prepared by the methods of Burgess and Jendrisak (18 ) and Gonzalez et al. (19 ). [alpha] subunit was prepared according to Igarashi and Ishihama (20 ) and 30% excess [alpha] subunit was added to core enzyme in all experiments.
Before adding substrates, 0.048 or 0.48 pmol RNA polymerase were preincubated at 37oC with 0.12 pmol template DNA for 10 min in T buffer (50 mM Tris-HCl, pH 7.9, 100 mM KCl, 10 mM MgCl, 1 mM DTT, 0.1 mg/ml BSA or partially hydrolyzed casein) unless otherwise noted. Transcription was then performed for 20 min in an 8 [mu]l reaction mixture containing 5 [mu]M 100-400 Ci/mmol [[gamma]-32P]GTP or [[gamma]-32P]ATP and 100 [mu]M each of other NTPs.
When necessary, before initiation preincubated mixtures containing immobilized DNA and RNA polymerase were washed twice with 0.2 ml T buffer at 37oC by centrifugation at 15 000 r.p.m. for 3 min to remove unbound RNA polymerase. All transcription reactions were stopped by adding EDTA (final concentration 30 mM). After each reaction, released transcripts were, where necessary, separated from transcription complexes by centrifugation at 15 000 r.p.m. for 3 min. The precipitate containing the transcription complexes was washed once with 0.2 ml T buffer containing 30 mM EDTA to prevent cross-contamination by released transcripts. All reactions were extracted with phenol. Ethanol precipitation was omitted to preclude the loss of short transcripts. Transcripts were analyzed on 20% polyacrylamide-7 M urea gels. The autoradiograms were visualized and analyzed with a Bio-Imaging Analyzer (BAS2000, Fuji Film, Tokyo).
The transcripts were eluted from the gel and partially digested with base-specific RNases (Pharmacia) for sequencing. RNA (200-400 c.p.m.) with 2 [mu]g carrier tRNA (Boehringer Mannheim) was digested for 20 min at 55oC in 50 mM sodium citrate, pH 5.0, 1 mM EDTA, 7 M urea by 1.5 U T1 or 6 U PhyM enzymes or in 50 mM sodium citrate, pH 3.5, 1 mM EDTA, 7 M urea by 1.5 U U2 enzyme or in 50 mM sodium citrate, pH 5.0, 1 mM EDTA by 3 U B.cereus enzyme. The amount of enzyme used was doubled for 32 nt RNA because of the larger number of digestion sites. Alkaline digestion was performed with 50 mM sodium carbonate buffer, pH 10.0, for 60 min at 90oC.
RNA polymerase in storage buffer (T buffer containing 50% glycerol) was dialyzed for 1 h against T buffer just before use to remove glycerol, which can interfere with the hydroxyl radical reaction. Open complex (binary complex) was formed for 5 min at 37oC in 0.2 ml T buffer containing 2 pmol 5'-end-labeled free template 3 (2 * 105 c.p.m.) and 20 pmol RNA polymerase. Elongation complex (ternary complex) was formed by a further 5 min incubation with GTP (5 [mu]M), UTP and CTP (100 [mu]M each). After 2 [mu]g tRNA were added as a carrier, the complexes were directly subjected to hydroxyl radical reaction for 7 min at 37oC. Other conditions were as described by Tullius and Dombroski (21 ). No further purification of the complexes was performed, because this time consuming process might destroy unstable complexes. DNA was analyzed on 6% polyacrylamide-7 M urea gels maintained at 55oC.
Next we determined whether the excess RNA polymerase affects transcription prior to or following initiation of transcription. Transcription was assessed in three situations, as diagramed in Figure 3 C. In case a the RNA polymerase to DNA ratio was low and the immobilized complexes were washed twice (prewash) to remove free RNA polymerase before substrates were added. In case b the RNA polymerase to DNA ratio was high and the mixture was prewashed. In case c the RNA polymerase to DNA ratio was again high but there was no prewash. In case a each DNA molecule on average should be occupied by less than one molecule of polymerase, whereas in case c following addition of substrates each DNA molecule should retain two or more polymerase molecules. In case b the prewash would remove all but open promoter complexes, so about one molecule of RNA polymerase should occupy each transcription unit.
Figure 3 D shows the transcripts synthesized in the three cases. Excess RNA polymerase enhanced synthesis of 4-8 nt transcripts (case c), as mentioned above. However, removal of excess RNA polymerase (case b) resulted in a distribution of transcripts similar to that in case a; the 9 nt RNA was the major product and enhancement of 4-8 nt transcripts was not observed. It was, therefore, following transcription initiation that the excess RNA polymerase caused enhancement of abortive synthesis.
Artefacts due to the prewash or to template immobilization were negligible, since under the DNA excess condition the prewash and immobilization did not alter the amounts nor distribution of transcripts as compared with those observed with the free template (compare lane 1 in Fig. 3 D with lane 1 in Fig. 3 A).
Templates 2 and 3 contained an A-less leader sequence following the transcription start site (Fig. 1 ). RNA polymerase could therefore be forced to stall at the +32 site by excluding ATP from the substrates. Transcription units were tested for the presence of multiple copies of RNA polymerase using hydroxyl radical footprinting (Fig. 4 ). A binary complex (DNA/RNA polymerase) was formed on 240 bp template 3 by adding an excess amount of RNA polymerase. In this complex the region of DNA from -55 to +15 was protected (lane 3). This is consistent with previous reports for the unmodified [lambda] PR (23 ) and the T7A1 promoter (24 ).
Next we asked whether the tandem occupancy of RNA polymerase near the promoter, which was imposed by stalling, caused any changes in abortive synthesis (Fig. 5 A). On template 2, which contains a 32 nt A-less leader sequence, excess RNA polymerase enhanced the abortive synthesis of 4-8 nt transcripts not only in the absence (lane 1) but also in the presence of stalling (lane 2).
The two results of ATP elimination are stalling at +32 and enhancement of misincorporation. We then tested which directly caused the enhancement, stalling or ATP elimination. We substituted various ATP analogs for ATP and examined enhancement (Fig. 6 A). The unincorporatable analogs enhanced synthesis of the misincorporated transcript, as in the case of ATP elimination, while all the incorporatable analogs tested did not. There is a perfect correlation between stalling at +32 and enhanced synthesis of the transcript produced by misincorporation. This suggests that the direct cause of the misincorporation is stalling per se, not ATP elimination.
The short transcripts whose synthesis was enhanced in the presence of excess RNA polymerase were mostly released (Fig. 3 B). However, small amounts of these short transcripts remained in transcription complexes, presumably because of formation of ternary complexes which stably retain but cannot elongate short transcripts (15 ,25 ). We were able to detect the short transcripts retained in these complexes by performing high specific activity [[gamma]-32P]GTP labeling, which allowed increased sensitivity. We examined the fate of the trailing RNA polymerase retaining short transcripts, including the 6 nt anomalous transcript, after the stalled leading polymerase had been removed.
Figure 7 shows the retained transcripts synthesized on immobilized template 2, according to the scheme illustrated. After the first reaction of 1 min in the absence of ATP to produce a stalled complex (lane 1), a 1 min chase reaction was performed without ATP (lane 2) or with ATP added so as to liberate the leading polymerase from its stalled position (lane 3). Lane 4 is a control showing that there was no incorporation of [[gamma]-32P]GTP during the chase period. The released transcripts were removed by two quick but thorough washes and the retained RNA was loaded on the gel.
We have shown that the presence of RNA polymerase in molar excess over promoter on DNA enhanced the abortive synthesis of 4-8 nt transcripts and that removal of the excess RNA polymerase before transcription initiation prevented this enhancement (Fig. 3 ). The results indicate that the presence of two or more molecules of RNA polymerase on a single molecule of the DNA template was the cause of enhanced synthesis of the abortive products.
In general, such a concentration effect could be due to (i) an interaction between two or more transcription complexes within the same transcription unit or (ii) an interaction between a transcription complex and RNA polymerase non-specifically bound at DNA sites near the promoter or (iii) oligomerization of the holoenzyme leading to altered activities. Our working concentration of holoenzyme, however, was too low to allow formation of oligomers (26 ) and the `concentration' effect was unchanged when DNA and polymerase were diluted at the same stoichiometry (data not shown). The second explanation is also unlikely, because heparin, a strong inhibitor of non-specific complex formation, could not perfectly remove this effect (Fig. 2 ). Furthermore, we observed the same distribution of transcripts on templates 2 and 3 (data not shown), which have a 5-fold difference in the numbers of non-specific sites. The first explanation is very likely the correct one.
Under stalling conditions the size of the region protected from hydroxyl radical attack is consistent with the presence of two RNA polymerase molecules in close proximity within a single transcription unit. It is reasonable to expect that two enzyme molecules could start transcription successively and that the trailing molecule would not translocate a long distance if the preceding RNA polymerase had been stalled at +32.
Notably, stalling at position +32 did not drastically alter the amounts of abortive transcripts under RNA polymerase excess conditions (compare lane 1 with lane 2 in Fig. 5 A). Rather, the results presented here suggest that a transient collision (or a transient indirect interference) with a non-stalled leading RNA polymerase is sufficient to make the trailing molecule produce shorter abortive transcripts in large amounts. A potential complication that is specific to the [lambda] PR transcription unit is the presence of the divergent promoter [lambda] PRM, which may have some effect (27 ). We, however, observed no effects on transcription in our conditions when we inactivated the PRM promoter by two point mutations (Sen and Shimamoto, unpublished result). Therefore, the [lambda] PRM promoter did not affect the results obtained here.
There were small but significant effects specific to the situation in which the leading RNA polymerase was stalled at the +32 position; one was enhancement of misincorporation, probably due to slippage synthesis. In addition, we observed slightly enhanced synthesis of 4-5 nt transcripts (lanes 1 and 2 in Figs 5 A and 6 A). This enhanced synthesis by stalling was not observed if the stalling position was moved downstream to +73 (Fig. 6 B). The common behavior of 4-5 nt transcripts and the putative slippage products can be explained by the physical interference model; only transcripts shorter than 6 nt and the putative slippage products are synthesized by the trailing RNA polymerase when there is a block imposed by the leading RNA polymerase stalled at +32. There should be no block in that region if the leading one stalls at +73.
Many examples of slippage synthesis have been reported and in most cases the transcripts were subsequently elongated to full length (28 -32 ). However, as in this study, the slippage product from the codBA promoter is aborted (33 ). Thus the fate of slippage transcripts seems to depend on template sequences. We have shown here an example of slippage which was stimulated probably by an interaction between transcription complexes.
The anomalous transcript from the T7A1 promoter (Fig. 2 A) is a misincorporation at the fourth position (34 ). Although it is not a slippage product, the interference between polymerase molecules often seems to spoil the fidelity of the trailing transcription complex.
The yields of long transcripts are strongly dependent on the promoter. Even in the presence of heparin and excess DNA, most likely in the single-round condition, the same amount of enzyme produced various amounts of long transcripts (Fig. 2 ). The amounts of transcripts longer than 12 nt from the [lambda] PR and lacUV5 promoters were only 40 and 2% respectively of those from T7A1. Since small amounts of long transcripts were synthesized during further incubation (data not shown), a significant fraction of the enzyme had been inactivated in terms of long RNA synthesis from these two promoters in 20 min.
We previously showed that a fraction of transcription complexes forms non-productive complexes, named moribund complexes, during transcription on template 2 (15 ). The moribund complex is trapped in the abortive cycle and subsequently converts into a very stable complex which is unable to elongate its transcripts, like the dead-end complex (35 ). The presence of more than one transcription complex is in agreement with the finding that isolated transcription complexes retaining short and long transcripts give different types of protection against footprinting (36 ).
In the present study removal of the stalled leading RNA polymerase did not reduce nor allow elongation of the short transcripts (Fig. 7 ). This indicates that the trailing RNA polymerase were irreversibly converted into non-productive moribund complexes before removal of the block. Thus interference by the leading RNA polymerase may convert a productive trailing complex into a moribund one.
If we assume that this conversion into moribund complexes is also caused by transient interference between RNA polymerase molecules being normally engaged in transcription, this explains why abortive synthesis on the template with a 32 nt A-less leader sequence is not significantly affected by elimination of ATP (Fig. 5 A) and why putative slippage synthesis is detected in small amounts even in the absence of the stalling of the leading enzyme (Fig. 6 A). Moreover, conversion into moribund complexes may occur in a similar fashion on the template with a 73 nt A-less sequence and the majority of the promoter will be occupied by the non-productive complexes. This block of the promoter may decrease the amount of trailing RNA polymerase reaching the stalled leading polymerase and prevent formation of triple or quadruple enzyme molecules at the stalled position, as shown in Figure 6 B, although there is enough room for three or four enzyme molecules.
When a given RNA polymerase molecule stalls or slows down at a suitable distance downstream of a promoter, a trailing molecule may be converted into a non-productive complex. This polymerase could then idle at the promoter, preventing formation of a queue of RNA polymerase molecules. This mechanism could act as a negative feedback, counteracting any tendency towards excessive initiation by blocking the successive entry of RNA polymerase into a certain class of promoters and thereby contribute to intracellular homeostasis. The presence of a block at a promoter has been suggested by premature termination near the downstream promoter in vitro (37 ), being consistent with idling of a moribund complex.
The block of promoters by moribund complexes could be lethal. Therefore, either this block occurs at a limited set of promoters or certain factors remove this block in a cell. It is suggestive that GreA and B RNA cleavage factors (38 ,39 ) inhibit abortive synthesis (40 ,41 ) and that these factors remove arrested complexes, which may be the final fate of moribund complexes (15 ).
We are grateful to Dr Richard Hayward (University of Edinburgh), Dr Jun-ichi Tomizawa, Dr V.James Hernandez (NICHD/National Institute of Health) and Dr David Friedman (University of Michigan) for their critical reading of the manuscripts and for their insightful discussions. We would like to thank Dr Hiroji Aiba (Nagoya University) and Dr Akira Ishihama (National Institute of Genetics) for valuable suggestions. This work was supported in part by grants from the Ministry of Education, Science and Culture of Japan to T.K. and N.S., JSPS Research Fellowships for Young Scientists to T.K. and by a grant from the Agency of Science and Technology of Japan to N.S.
REFERENCES