Strong Natural Pausing by RNA Polymerase II within 10 Bases of Transcription Start May Result in Repeated Slippage and Reextension of the Nascent RNA

Reagents.

Ultrapure NTPs were obtained from Pharmacia Biotechnology, ApC and CpU dinucleotides were obtained from Sigma, ³²P-labeled NTPs (800 Ci/mmole) were obtained from NEN, Deep Vent DNA polymerase and restriction enzymes were obtained from New England Biolabs, RNase T₁ was obtained from Boehringer Mannheim, and streptavidin-coated magnetic beads were obtained from Promega Biotech.

Plasmids.

All plasmids used in this study are based on the adenovirus major late (AdML) promoter. Plasmid pML20-40 is identical to the previously- described pML20-42 construct (34) with the exception of a single nucleotide change of T to G at position +16 on the nontemplate strand (Fig. 1). Plasmids pML20-40(4G), pML20-40(6G), and pML20-40(9G) were constructed by replacing the BssHII-StuI fragment of pML20-40 with a synthetic DNA fragment carrying a G instead of a C at position +4, +6, or +9 on the nontemplate strand (Fig. 1). All constructs were verified by DNA sequencing.

Template preparation for in vitro transcription reactions.

Template DNAs were made by PCR using the forward primer 5′-GGCATCAAGGAAGGTGATTG-3′ (biotinylated at the 5′ end) and the reverse primer 5′-CAGTGCCAAGCTTGCATG-3 (a HindIII site is underlined). The 190-bp product thus obtained has the transcription start site 96 bp downstream of the biotinylated end. After purification using the Concert cleanup kit (Bethesda Research Laboratories) and digestion with HindIII to reduce transcription arrest before the end of the template (12), the DNA was extracted with phenol-chloroform and precipitated with ethanol.

In vitro transcription reaction on attached templates.

Preinitiation complexes were assembled using HeLa cell nuclear extract and biotinylated template DNA immobilized on streptavidin-coated magnetic beads as described previously (29, 34). Initial complex assembly was carried out at either 25 or 30°C for 25 min. Bead-attached complexes were washed twice in 150 μl of BC100 (20 mM Tris [pH 7.9], 8 mM MgCl₂, 100 mM KCl, 1 mM dithiothreitol, 20% glycerol, 0.2 mM EDTA), resuspended in BC100, and stored on ice until used. Transcription reaction mixtures contained 1 mM dinucleotide primer (ApC, CpA, or CpU), dATP (usually 20 μM) to satisfy the energy requirement for transcript initiation, and other nucleotides at the concentrations indicated in the figure legends. The labeling nucleotides had specific activities of 800 Ci/mmol unless otherwise indicated. Reaction times are given in the legends. When ATP was used to initiate transcription instead of a dinucleotide, it was present at 100 μM and dATP was absent. Only 5 μM dATP was used for the reactions in Fig. 8B and whenever transcription was primed with CpU. Reactions were stopped by adding an equal volume of phenol-chloroform-isoamyl alcohol (25:24:1). The aqueous phase from the initial extraction was reextracted with chloroform-isoamyl alcohol (24:1). Reaction mixtures involving synthesis of 15-nucleotide or longer RNAs were usually ethanol precipitated before analysis. For shorter RNAs, the aqueous phase was lyophilized, resuspended in water, and treated with calf intestinal phosphatase (0.1 U/μl for 10 min at 30°C). RNase T₁ digestions were performed in 10 μl at 37°C for 10 min with 1 or 2 U of T₁ and 0.25 μg of yeast tRNA as carrier. RNAs were resuspended either in 50% formamide or 8 M urea before resolution on 25% acrylamide-3% bisacrylamide gels containing 7 M urea. Gels were imaged, and individual bands were quantified using a PhosphorImager and ImageQuant software (Molecular Dynamics). RNA length markers were made by partial resection of the transcript of a particular halted complex in the presence of 2 mM sodium pyrophosphate at 37°C for 2, 4, or 8 min.

RNA polymerase II exhibits distinct kinetic pauses in the early stage of RNA synthesis.

The template we chose for our study, called pML20-40 (Fig. 1), is based on the AdML promoter. In our earlier work we had used similar templates to purify and characterize transcription complexes stalled at various locations from 13 nt downstream of the transcription start (23, 33, 34). Our most recent studies revealed unexpectedly strong pausing when Sarkosyl-rinsed transcription complexes with RNAs as short as 15 nt were allowed to resume transcript elongation (29; see also reference 18). We were therefore interested in determining what other pauses might occur during the very first steps in transcription on the AdML promoter. We assembled preinitiation complexes with HeLa nuclear extracts on bead-attached pML20-40 templates, rinsed them gently with transcription buffer (without Sarkosyl) and then initiated RNA synthesis with 1 mM ApC primer and 200 μM each NTP. This concentration of NTPs supports transcript elongation at 1 to 2 nt/s at 30°C with promoter-distal RNA polymerase II transcription complexes (29). The transcripts were uniformly labeled by including both [α-³²P]CTP and [α-³²P]UTP in the reaction mixtures. Total RNA was isolated from 5 to 300-s time points and resolved on a high-percentage polyacrylamide gel. Since all of the NTPs were present at the same relatively high concentration, one might have expected to observe a zone of transcripts moving uniformly up the gel. However, as shown in Fig. 2A. most of the RNA appeared at a series of very discrete pause sites, most prominently at +7, +18/19, and roughly +25. A less pronounced pause occurred at +13. We repeated the experiment in Fig. 2A with limiting NTP concentrations (1 mM ApC, 2 μM [α-³²P]UTP and 20 μM each CTP and GTP); in this case ATP was left out to stall transcription at +20 (Fig. 2B). With the NTP levels reduced by either 10- or 100-fold relative to the reaction in Fig. 2A, the residence time at the pause sites in Fig. 2B was considerably longer but the locations of the major pauses were unchanged. These results strongly suggest that RNA polymerase II is undergoing a series of relatively slow conformational transitions during the synthesis of the first 25 bases of the transcript. We will return to this point in Discussion.

Transcript slippage by RNA polymerase II.

A surprising outcome of these experiments can be seen in Fig. 2B, particularly in the 150- and 300-s lanes. In these reactions, no RNAs longer than 20 nt should have been made. However, a pair of bands reproducibly appeared above the 20-mer, whose length we verified with size markers (see Fig. 3A). The RNAs longer than 20 nt should not have resulted simply from readthrough of the A-stop at +20, since we saw no such readthrough on other, comparable templates at similar nucleotide concentrations (data not shown here, but see for example Fig. 4). When we ran RNAs from reactions like those in Fig. 2B along with length markers, it was apparent that the unexpected RNAs were 2, 4, and 6 nt longer than the 20-mer (see asterisk bands in Fig. 3A) whereas readthrough due to very low levels of contaminating ATP should have resulted in production of low levels of 21-nt and even lower levels of 22-nt RNAs (Fig. 1).

Note that transcripts from the pML20-40 template begin 5′-ACUCUCU (Fig. 1). This suggested to us that the unanticipated 22-nt RNA in Fig. 2B could result from upstream slippage of the initial RNA-DNA hybrid by 2 nt followed by re-pairing of the RNA with the template and resumption of RNA synthesis. The 24-nt RNA could result from two cycles of this process, and so forth. This hypothesis is diagrammed in Fig. 3B. We propose position +7 as the most likely location for slippage to begin, for two reasons. First, there is a strong natural pause at this location during transcription (Fig. 2), and second, this position marks the end of the three CU repeat elements and thus offers the maximum number of bases hybridized at the 3′ end (four) if slippage takes place.

If hybrid slipping and reextension occurred as diagrammed in Fig. 3B, cleavage of the transcripts in Fig. 3A with RNase T₁ should lead to the production of CUUU from the 3′ ends of all of the RNAs, with bands of 16, 18, 20, and 22 nt generated from the 5′ ends. The results of T1 digestion (Fig. 3A) were consistent with this expectation. Also, if our slippage model is correct, changes in the template which interrupt the CU repeats should prevent stable extended hybrid formation after slippage and thus block the slippage-reextension reaction. We generated three variants of our initial template, which we designated pML20-40(4G), pML20-40(6G), and pML20-40(9G). As shown in Fig. 1, each template differed from the original by a single C-to-G change on the nontemplate strand. None of these changes alter the strength of the RNA-DNA hybrid compared to the original template. We confirmed (data not shown) that all of these templates support the identical pausing pattern during early RNA synthesis, including the prominent pause at +7, which we observed with the pML20-40 template. Note that when the polymerase pauses at +7 on either the 4G or 6G template, the RNA-DNA hybrid cannot slip 2 nt upstream while retaining 4 bp at the 3′ end. However, slippage from the 7-mer should be unaffected on the 9G template.

When we tested these templates under reaction conditions like those used in the experiment in Fig. 3, we obtained the expected results. RNAs longer than 20 nt were produced as efficiently on the 9G template as on pML20-40, but these RNAs were obtained at a much lower level on the 4G and 6G templates (Fig. 4). RNase T₁ digestion of the RNA produced on the 9G template gave, in addition to the expected 4-, 7-, and 9-nt products, RNAs of 11 and 13 nt, consistent with two rounds of slippage and re-pairing as diagramed in Fig. 3B.

We observed a strong pause at +7 even under free-running transcription conditions with 200 μM NTPs (Fig. 2A). However, the detailed characterization of the slipping reaction was done with more limiting NTP concentrations, particularly for the ³²P-labeled NTP. To explore the dependence of transcript slipping on transcription conditions, we transcribed the pML20-40 and 9G templates with ApC primers and different concentrations of CTP, UTP, and GTP, ranging from the labeling NTP at 1 μM to all NTPs at 200 μM. We resolved the RNAs on high-percentage acrylamide gels as in the experiment in Fig. 4 and determined the relative yields of 20-, 22-, 24-, and 26-nt RNA. We also performed ApC-primed transcriptions on pML20-40 with CTP and UTP at 815 μM. In this last case, the expected product was 15 nt long and the slippage products were 17- and 19-mers. The results are shown in Table 1. Not surprisingly, the amounts of slippage-reextension products were larger when transcription was performed at low NTP levels. However, slipped RNAs represented 16% of the total transcript with all NTPs at 200 μM, and this percentage was essentially the same (18%) when 815 μM NTPs were used. Thus, transcript slippage is not an artifact of low NTP concentrations. As expected, slippage-reextension transcripts were made with equal efficiency on the pML20-40 and 9G templates when reaction conditions were identical.

Effect of the initiating nucleotide and site of initiation on transcript slippage.

A number of earlier studies have reported slippage and resumption of RNA synthesis during transcript initiation by Escherichia coli RNA polymerase (2, 35). It was suggested that the lack of slippage for certain transcripts reflected a stabilization of the ternary complex by interaction of the 5′ end of the transcript with an upstream RNA binding site. We therefore compared the extent of slippage on the pML20-40 template in the ApC-initiated reaction (priming at +1/+2 [Fig. 1]) with that seen with CpA priming (−1/+1). The results are shown in Fig. 5. In lanes 1 and 2, RNA was generated with ApC priming, CTP, UTP, and GTP, as in lane 7 of Fig. 4. Slippage products of 22, 24, and (very faint) 26 nt were obtained (Fig. 5, lane 1). These RNAs were shortened by 4 nt, as expected, after treatment with RNase T₁ (lane 2). When a similar reaction was done without GTP, the G-stop band at 15 nt and the slippage bands at 17, 19, and (faintly) 21 nt were also observed (lane 3). These latter bands all corresponded to stops before the incorporation of the G at +16, since they were unaffected by RNase T₁ (lanes 3 and 4). Interestingly, when G-less reactions were carried out using a CpA primer, much less slippage was observed (lane 5). Since we have hypothesized that slippage with the ApC primer depends on the strong pause at +7, at the end of the three CU repeats, it was important to establish that this pause also occurred with CpA-mediated initiation. Kinetic assays (as in Fig. 2B) demonstrated that for CpA initiations, a pause of similar duration occurred at the same template location as seen with ApC priming (data not shown). That is, an 8-nt RNA transiently accumulated in the CpA-primed reactions.

We also assayed transcription of the pML20-40 template without dinucleotide primers, using ATP, CTP, and UTP to support RNA synthesis. This resulted in detectable slippage products, but the level of slippage was lower than we observed with ApC priming. However, when we performed kinetic experiments like those in Fig. 2 with ATP-initiated reactions, we found that the pause at +7 was significantly shorter (data not shown). We therefore attribute the lower level of slippage in ATP-initiated reactions, relative to similar reactions primed with ApC, to the shorter residence time at the critical +7 position. On other templates which permitted pause times to be manipulated, high levels of slippage after initiation with ATP could be observed (see Fig. 7), but we were unable under any experimental conditions to obtain high levels of slippage in CpA-primed reactions (see Fig. 7).

Transcript slippage may occur upstream of +7 if pausing is induced at the upstream location.

As noted above, our results are consistent with a model in which the transcription complex paused at +7 is the primary precursor of transcript slippage. Another potential location for transcript slippage would be +5, after the synthesis of ACUCU. Upstream slippage in this case would leave 2 bases hybridized at the 3′ end of the RNA (Fig. 1). Since the 6G template supported very little transcript slippage when GTP was present in the reaction mixture (Fig. 4), it is possible that only 2 bp is insufficient to stabilize the RNA-DNA hybrid during slippage and re-pairing. Alternatively, since transcription does not pause significantly after synthesis of the 5-mer on the 6G template (data not shown), there may be insufficient time for slippage to take place. The sequence of the 6G template allows us to discriminate between these two explanations. We compared the level of slippage-reextension products in ApC-primed transcriptions on 6G, either in the continuous presence of CTP, UTP, and GTP or with only CTP and UTP (which should stall transcription at +5), followed by the addition of GTP 3 min later. The results are shown in Fig. 6. As expected, the plus-G reaction led to only low levels of RNA longer than 20 nt (lane 7). The level of slippage products was considerably increased if the addition of GTP was delayed by 3 min (lane 4). However, this reaction still produced much lower levels of slippage-reextension RNAs than did a comparable reaction on the original pML20-40 template (lane 2). Thus, 2 bp of hybrid at the 3′ end are sufficient to allow re-pairing of the transcript and continued transcription after upstream slippage. However, reextension is more efficient if 4 bp of hybrid can be formed. It is important to note that CpA-initiated transcripts showed no evidence of upstream slippage in comparable transcription reactions on the 6G template, including those in which elongation was paused for 3 min at +5 (lanes 8 to 10). Therefore, pausing and the ability to pair at least 2 bases at the 3′ end after slippage are necessary but not sufficient conditions for efficient upstream slippage, re-pairing, and continued transcription to take place.

The fact that only 2 bp at the 3′ end of the RNA is needed after upstream slippage for transcription to resume may explain the low level of slippage products we see with the 6G template (Fig. 4). Even though there is no discernable pause at +5, the residence time at this position may be long enough for some slippage to take place. Upstream slippage by 2 bases from the normal pause site at +7 seems less likely for 6G transcripts, since this would result in an attempt to pair two G residues. On the other hand, slipping upstream from the pause at +7 may account for the low level of slippage we observed with the 4G template. In this case, translocation upstream by 2 bases would result in the pairing of UCU in the RNA with ACA in the DNA. Only 2 bp are formed, but the middle position in this case would not result in a potential purine-purine clash.

The results in Fig. 6 indicated that only two of the CU repeats need be present at the 3′ end of the transcript at the point of pausing for slippage to take place. This finding allowed us to test another dinucleotide primer, CpU, for its ability to support slippage. CpU can prime transcription on our templates from two locations, −3/−2 and +2/+3 (Fig. 1). Preinitiation complexes on the 6G template incubated with CpU, CTP, and radiolabeled UTP should give rise to only one labeled product, the CUCU tetramer initiated from the +2/+3 position (Fig. 1). Note that this RNA could slip upstream by 2 nt and re-pair in a manner analogous to that just described for ApC initiation in the absence of GTP on the 6G template. As shown in Fig. 7A, CpU initiation from +2/+3 yielded not only the expected 4-nt RNA but also RNAs that are 2, 4, 6, 8, and, at a low level, even 10 nt longer. None of these RNAs represented readthrough of the G stop, since none of them were sensitive to RNase T₁ (lanes 7 and 8). High levels of the slippage-reextension products did not accumulate until the reaction had proceeded for at least 40 s. Most of the slippage products remained in ternary complex after 4 min because they could be chased into longer RNAs (lane 9). Thus, the slippage reaction not only occurs with a +2/+3 primer but is even more efficient than with the +1/+2 primer, ApC.

The ability to stall the RNA polymerase indefinitely at +5 on the 6G template allowed us to assay for slippage with all of the initiation substrates under comparable reaction conditions (Figs. 7B and C). In Fig. 7B, each 2-min reaction mixture contained CTP and radiolabeled UTP in addition to the initiating nucleotide or dinucleotide. Transcripts attributable to multiple rounds of slippage were observed with both ApC and CpU priming, but not with CpA priming, as shown above. (Note that the CpA-primed 6-mer is the normal G-stop transcript and not a slippage product.) Interestingly, when the RNA polymerase was forced to pause at +5 for an extended period, products consistent with at least three rounds of slippage could also be observed in ATP-initiated reactions (lanes 7 and 8). All of the RNAs marked with asterisks in lanes 1 to 8 of Fig. 7B should represent slippage products and not readthrough of the G stop, since these RNAs were not sensitive to RNase T₁. If readthrough had taken place, the resulting transcripts should have accumulated at +15, before the next G stop. Low levels of such RNAs were observed in the ATP and ApC-initiated reactions (lanes 3 and 7); these transcripts were sensitive to RNase T₁ (lanes 4 and 8). As a second positive control for activity of the T₁, a parallel ApC-primed reaction was run which did include GTP. These transcripts gave the expected fragments on treatment with T₁ (lanes 9 and 10).

In light of the efficient slippage supported by the CpU primer, we decided to test for slippage in CpU-initiated reactions with high levels of NTPs. For these experiments, we transcribed the pML20-40 template with 1 mM CpU and 815 μM CTP and UTP. With these substrates, transcription should proceed freely through the region in which slippage could occur and stall at +15 before the G stop (Fig. 1). As controls, we also initiated transcription on pML20-40 with ApC or CpA primers, or with ATP, in the presence of 815 μM CTP and UTP. The results are shown in Fig. 7C. The CpU primer supported synthesis of the expected 14-mer in addition to equivalent or greater amounts of 16- and 18-nt RNAs which are presumably slippage products, since they were not sensitive to RNase T₁. Since high levels of NTPs were used in this assay, it is not surprising that substantial readthrough of the G stop was also observed; these readthrough RNAs were cleaved by T₁. In the CpU-primed reaction in Fig. 7C, the 14-mer (nonslipped) RNA was 34% of the nonreadthrough transcript, while the RNA produced by one round of slippage (16-mer) represented 44% of the total and the 18-mer transcript from two rounds of slippage represented 22%. Thus, remarkably, when NTP levels are nearly physiological, the majority of CpU-primed initiations at the pML20-40 promoter go through at least one cycle of slippage. RNAs produced by slippage represented 18% of the total nonreadthrough transcript for the ApC-primed reaction (8% 17-mers, from one round of slippage, and 10% 19-mers, from two rounds [Table 1]) and 10% of the total for the ATP-initiated reactions (about 5% each for one-round and two-round products). However, we failed to detect slippage products for the CpA-primed reaction, consistent with our other studies at lower NTP levels.

Roles of RNA-DNA hybrid strength and helicase activity in transcript slippage.

Slippage requires the RNA-DNA hybrid to be transiently disrupted in order to allow re-pairing at a more upstream location. This predicts that strengthening the RNA-DNA hybrid would inhibit the slippage reaction. To test this, we substituted 5Br-UTP for UTP in ApC-primed transcription reactions with the pML20-40 template. As shown in Fig. 8A, the slippage products longer than 20 nt observed with UTP were completely absent (even at long exposure) when 5Br-UTP was substituted. The presence of 5Br-U also reduced the pause time at +7 compared to that in the UTP reaction, but the 40 to 60-s residence time at this location was sufficient to support easily detectable slippage in other experiments (data not shown) using UTP.

The location from which slippage initiates, as far downstream as +7, raises the question whether the TFIIH initiation factor might be involved in this process, since it has been established that inhibiting the XPB helicase activity of TFIIH with ATPγS blocks promoter escape (5–7) and inhibits the accumulation of transcripts longer than 9 or 10 nt. Earlier work (38) indicated that TFIIH should still be present in our early elongation complexes, which are substrates for slippage. We performed a series of transcription reactions on the 6G template with an ApC primer, CTP, and radiolabeled UTP, plus 5 μM dATP to support promoter unwinding. After 40 s, the expected 5-mer had accumulated, plus very low levels of slippage products (Fig. 8B, lane 1). When transcription was continued for 2 min without any additions (lanes 2 and 3), 7-, 9-, and 11-nt RNAs accumulated, as expected from the results in Fig. 7B. Also as expected from the results in Fig. 4, continuation of the 40-s reaction in the presence of GTP (lanes 6 and 7) led primarily to the accumulation of 20-nt RNA with only low levels of slippage. When ATPγS was added at 30 s along with GTP, the accumulation of the 7-nt RNA was not strongly affected but very little RNA longer than 7 nt was made (lanes 8 and 9), consistent with earlier reports (6, 7). Significantly, slippage products accumulated in the presence of ATPγS to essentially the same levels as they did in the absence of the inhibitor (compare lanes 2 and 4), while synthesis of a low level of 15-mer readthrough RNA was blocked. Thus, the XPB helicase is not required for transcript slippage.