INTRODUCTION

Vaccinia virus RNA polymerase is a multisubunit enzyme devoted exclusively to the synthesis of mRNA (1). Specificity for transcription of vaccinia early genes is conferred upon the RNA polymerase by an early transcription factor (ETF),¹ a heterodimer of 82- and 70-kDa subunits that binds to the early promoter and recruits RNA polymerase to the template (2, 3, 4, 5, 6). An essential 94-kDa polypeptide associated with the RNA polymerase (variously named RAP94, H4, or rpo94) is required for early promoter-specific transcription (7, 8, 9, 10, 11). rpo94 is believed to act as molecular bridge between the RNA polymerase and ETF bound at the promoter. Termination of early transcription occurs at heterogeneous sites downstream of a simple termination signal TTTTTNT in the nontemplate DNA strand (12). The signal is appreciated at the RNA level as UUUUUNU (13). A separate vaccinia termination factor (VTF) is required to transduce the UUUUUNU signal to the elongating polymerase (14). VTF is identical to the vaccinia capping enzyme, a heterodimer of 95- and 33-kDa subunits required for 5 capping and methylation of the nascent RNA chain (14, 15, 16).

VTF-dependent termination is a dynamic, energy-requiring process. The termination event is coupled to the hydrolysis of ATP (17). Termination site choice and the overall efficiency of termination are determined by a kinetic balance between the rate of signaling and the rate of polymerase movement (17). Slowing elongation rate (e.g. by analog-induced pausing) results in termination at template positions closer to the UUUUUNU element. In contrast, slowing the rate of signal transduction (e.g. by lowering the ATP concentration) shifts the distribution of termination sites further away from the termination signal (17).

Mechanistic studies of the termination reaction are complicated by the ongoing process of chain elongation, i.e. the target for the termination factor is constantly in flux as the polymerase changes template position. Under these circumstances, two events are required to achieve bona fide termination: (i) polymerase must elect not to incorporate the next NTP and (ii) the nascent RNA must be released from the ternary complex. Our goal in the present study was to focus strictly on the transcript release step of the termination reaction by removing elongation as a variable. Our strategy was to purify homogeneous populations of RNA-labeled ternary complexes paused at a unique template position located upstream (5) of a termination signal, then to ``walk'' the RNA polymerase to a defined template position downstream of the termination signal, and then test for the release of the RNA from the ternary complex in response to VTF/capping enzyme. Purification of the elongation complexes was simplified by the use of early promoter-containing DNA templates bound to a solid support (17). With this approach, we have now demonstrated that factor-dependent transcript release occurs in the absence of concomitant elongation and that the release reaction is ATP-dependent. Additional experiments establish that the UUUUUNU sequence must be situated beyond a threshold distance from the active site of the polymerase in order to signal transcript release.

EXPERIMENTAL PROCEDURES

The pBS-based G21 plasmid containing a vaccinia early promoter fused to a G-less cassette has been described (18). Construction of the G21(TER29)A78 plasmid involved replacement of the sequence between the BamHI and XbaI sites of the G21 plasmid with a synthetic A-less cassette by using standard molecular cloning techniques. The insert included a TTTTTTTTT termination signal (refer to Fig. 1). A PvuII fragment of G21(TER29)A78 containing the vaccinia transcription unit was then inserted into pUC19. Construction of plasmid G21(TER59)A78 plasmid entailed replacement of the segment between the NgoMI and XbaI sites of the pUC19-G21(TER29)A78 plasmid with a different synthetic A-less cassette (see Fig. 1). The G21(TER29)A78 and G21(TER59)A78 plasmids were linearized with Acc65-1, which cleaved the DNA template upstream of the vaccinia early promoter region. Biotinylated dATP was incorporated at the 3 ends using Klenow DNA polymerase. The biotinylated DNAs were then digested with PvuII, and the restriction fragment containing the transcription cassette was isolated by preparative agarose gel electrophoresis. Purified DNA fragments were attached to streptavidin-coated magnetic beads (Dynabeads M280; Dynal) by incubating the DNA with beads in 0.1 ml of TE buffer (10 mM Tris HCl (pH 8.0), 1 mM EDTA) for 30 min at room temperature. The beads were then concentrated using a horseshoe magnet and washed with 0.1 ml of TE. After three cycles of washing, the immobilized templates were resuspended in TE and stored at 4 °C for use in in vitro transcription reactions.

Vaccinia RNA polymerase was extracted from virion cores with deoxycholate (14), and a transcriptionally active preparation containing the ETF was purified by sequential DEAE-5PW, SP-5PW, and heparin-agarose column chromotography steps using a Waters 650 chromatography system and chromatography columns purchased from Waters. Transcription activity during purification was assayed as described elsewhere (24) in reactions programmed by a SmaI-cut pSB24 template, which contains a vaccinia early promoter fused to a 382-nucleotide G-less cassette (24). Capping enzyme was purified from virion cores as described previously (14). The phosphocellulose fraction was used in the present study; the molar concentration of active capping enzyme was determined by enzyme-GMP complex formation (14).

Ternary transcription complexes were formed in standard reaction mixtures containing (per 20-µl reaction volume) 20 mM Tris-HCl (pH 8.0), 6 mM MgCl₂, 2 mM dithiothreitol, 1 mM ATP, 0.1 mM UTP, 1 µM [-³²P]CTP (1000 Ci/mmol), 0.1 mM 3-OMeGTP, bead-linked DNA (~100 fmol), and vaccinia RNA polymerase. Reaction mixtures were incubated at 30 °C for 10 min, then concentrated by microcentrifugation for 15 s. The beads were held in place by application of an external horseshoe magnet while the supernatant was removed and replaced with 0.1 ml of 20 mM Tris-HCl (pH 8.0), 2 mM dithiothreitol. The beads were resuspended and subjected to two further cycles of concentration and washing; after the third wash, the beads were resuspended in a small volume of the wash buffer, and aliquots were distributed into individual reaction tubes to achieve approximately the same concentration of template as that used in the pulse-labeling phase. Elongation reactions were performed as specified in the figure legends.

RESULTS

Transcription in vitro by vaccinia RNA polymerase was programmed by linear templates linked to streptavidin-coated paramagnetic beads (Fig. 1). The prototype G21(TER29)A78 transcription unit consisted of a synthetic early promoter fused to a 20-nucleotide G-less cassette, which was flanked by a run of three G residues at positions +21 to +23. Downstream of the G-less cassette was inserted a 57-nucleotide A-less cassette flanked at its 3 end by a run of four A residues at positions +78 to + 81. Placed within the A-less cassette was a termination signal, TTTTTTTTT, spanning positions +29 to +37 (Fig. 1). Pulse-labeling transcription reactions contained ATP, UTP, [-³²P]CTP and 3-OMeGTP. The reaction products (template-engaged ternary complexes containing radiolabeled nascent RNA) were recovered by centrifugation and concentration of the beads with an externally applied magnet, followed by washing the beads with buffer lacking nucleotides and magnesium (17). The major pulse-labeled nascent chain was a 3-OMeGMP-arrested 21-mer, as expected (Fig. 2, lane 1). Minor 22-, 23-, and 24-mer RNA species were also detected; these are 3 co-terminal transcripts that initiated from upstream template positions 1C, 2U, and 3U) (18).

The integrity of the isolated ternary complexes was verified by their ability to resume elongation of the pulse-labeled RNA upon provision of unlabeled NTPs and magnesium. Omission of ATP from the elongation reaction and inclusion of the chain-terminating nucleotide 3-dATP (cordycepin triphosphate) allowed us to ``walk'' the ternary complexes through the A-less cassette from G21 to A78 (Fig. 2, lane 2). All polymerase molecules that elongated past G21 were arrested at A78; we detected no read through past the +78A position under these reaction conditions.

Elongation of the nascent chains beyond the arrest site at G21 depended upon removal of the blocking 3-OMeGMP moiety by a hydrolytic activity intrinsic to the vaccinia RNA polymerase elongation complex (19). The sequence of the RNA chain at the site of pulse arrest was 5-CTAG_OMe. The distribution of RNAs after the elongation in the presence of 3-dATP was very instructive with respect to the step increment of nascent chain cleavage. In the experiment in Fig. 2, most of the OMeG21 transcripts were elongated to the next templated A position at +78. However, a minor fraction was converted to a shorter dA20 species. The dA20 RNA was formed by incorporation of 3-dAMP into chains that had been shortened by 2 nucleotides to T20. The fact that very few of the G21 chains were trapped at A20 argues strongly that the predominant initial RNA cleavage event was a single nucleotide step to A20, followed by elongation to A78. We cannot tell whether chains trapped at A20 were generated by two sequential one-nucleotide cleavages or by initial removal of a dinucleotide. However, the results of the cordycepin trap confirm an earlier suggestion (19) that the vaccinia ternary complex cleaves in mononucleotide increments. The viral enzyme is clearly different in this respect from cellular RNA polymerases II and III, which shorten nascent chains primarily in dinucleotide increments when polymerase II or polymerase III elongation complexes are arrested by nucleotide omission (20, 21, 22).

The use of bead-bound DNA templates provided a convenient method to assay transcript release by centrifugal/magnetic separation of template-engaged RNA products (bead-bound) from released transcripts (recovered in the supernatant). The labeled RNAs that had been walked to A78 were recovered in the template-bound fraction (Fig. 2, lane 2) and very little A78 RNA was free in the supernatant (Fig. 2, lane 3). Addition of VTF/capping enzyme to the arrested A78 elongation complexes resulted a release of 60-70% of the total RNA from the beads into the supernatant (Fig. 2, lanes 4 and 5). Thus, complexes that had already synthesized the UUUUUNU signal could be induced to release the RNA chain in the absence of ongoing elongation. Note that the addition of VTF/capping enzyme to the A78 complexes resulted in capping of the 5 end of the A78 RNA. The mixed population of guanylylated and unguanylylated A78 chains migrated as a doublet (Fig. 2, lanes 4 and 5). Note also that the trapped dA20 transcripts remained stably associated with the template and were not released in response to VTF/capping enzyme (Fig. 2, lanes 2 and 4).

A tenet of the kinetic coupling model for VTF-dependent termination is that a slowing chain elongation rate will cause termination to occur at sites closer to the RNA signal; and, indeed, this has been observed (17). Yet there appeared to be some constraint on the system such that termination was limited to sites ~30 nucleotides or more downstream of the first U of the UUUUUNU signal (17). It was not clear whether this phenomenon represented a physical or a kinetic constraint. We were able to address the question using the pulse-walk-release assay by manipulating the physical distance between the termination signal and the downstream site of elongation arrest. In the G21(TER59)A78 transcription unit, the TTTTTTTTT terminator element was shifted distally such that the first T was at position +59 (Fig. 1).

When pulse-labeled OMeG21 chains were walked to A78 on the G21(TER59)A78 template, the transcripts remained template-engaged (Fig. 2, lanes 7 and 8). However, these ternary complexes were refractory to VTF-mediated transcript release, i.e. all RNA remained template-associated after incubation with VTF/capping enzyme (Fig. 2, lanes 9 and 10). The conversion of the template-engaged A78 transcripts into a doublet of capped and uncapped RNAs provided assurance that capping enzyme had access to the nascent RNA. Note that the size of the RNA chain is not a variable in this experiment and that the sequence of the 11 nucleotides at the 3 end of the nascent chain is the same in the VTF-responsive TER29 and VTF-refractory TER59 complexes. Because the salient variable is distance from the signal to the 3 end of the chain (50 nucleotides in the case of TER29 versus 20 nucleotides for TER59), and because elongation rate is not an issue here, we surmise that there is a threshold distance between the terminator and the termination site that is dictated by physical constraints on the access of VTF/capping enzyme to the UUUUUNU sequence.

We propose that this physical constraint arises because the UUUUUNU signal is shielded from trans-acting proteins so long as it remains within the RNA binding pocket of the polymerase elongation complex. The dimensions of the RNA binding site were defined previously by RNase protection analysis of the nascent chains engaged within OMeGMP-arrested ternary complexes (23). An 18-nucleotide region of the transcript extending back from the 3 growing point of the chain was protected from RNase digestion; the size of the protected fragment did not change as the polymerase elongation complex moved away from the promoter. This suggests that the termination signal ought to be shielded on the TER59 template, but not on TER29. However, because, the earlier RNase protection analysis involved transcription complexes paused within a different template sequence context from that of the A78 complexes, we felt it was important to determine the dimensions of the polymerase-RNA interface for the A78 complexes on the G21/A78 templates.

To accomplish this, we prepared ternary complexes containing A78 RNA labeled uniquely at the 3 end. Transcription was initiated on the G21(TER29)A78 template in the presence of unlabeled ATP, CTP, UTP, and 3-OMeGTP; the bead-bound OMeG21 ternary complexes were purified and then walked to A78 in reactions containing GTP, CTP, UTP, and 3-[-³²P]dATP. The labeled cordycepin monophosphate moiety was incorporated exclusively at the 3 end of the A78 chain. The walked complexes were digested with increasing amounts of RNase A (Fig. 3). A ladder of digestion products was generated at low concentrations of RNase A (Fig. 3A; 0.1-0.2 µg/ml). The sizes of the individual cleavage products reflected the distance of the RNase cleavage sites from the 3 end of the nascent chain. Purine gaps within the pyrimidine cleavage ladder provided landmarks for aligning the 3 end-labeled digestion products within the predicted RNA sequence (Fig. 3). The digestion products were shortened with increasing RNase (Fig. 3A). A limit size was achieved at 0.5 µg/ml nuclease and this did not change substantially at 1-2 µg/ml RNase. The protection of 21-25 nucleotides at 2 µg/ml RNase suggested an upper limit of the size of the RNA binding site. Increasing RNase to 10 µg/ml trimmed the protected transcripts to 16-18 nucleotides, but still did not fully degrade the transcripts (Fig. 3A). Control digests were performed using 3-[³²P]dAMP-labeled A78 RNA that was synthesized in the vaccinia in vitro system, then recovered by phenol extraction and ethanol precipitation (Fig. 3B). This free RNA was digested to completion at RNase levels to which the 3 segment of A78 RNA in the ternary complex was resistant (Fig. 3B).

An RNA binding site on the A78 ternary complex protects the 3 end-labeled nascent RNA from nuclease digestion.

A readable ladder of G-specific cleavage products was generated when 3-dAMP-labeled A78 ternary complexes were probed with RNase T1. Increasing the amount of input nuclease resulted in shortening of the cleavage products to a limit size of 17 nucleotides (Fig. 3C). Shorter digestion products were not detected (not shown). We conclude from these experiments that the ternary complex strongly protects a 16-21-nucleotide 3 segment of the nascent A78 chain from nuclease digestion in trans. Whereas the upstream RNA segment that includes the UUUUUUUUU termination signal was fully accessible in the G21(TER29)A78 transcription complex (this segment is demarcated by the vertical bar in Fig. 3A), the termination signal would be protected in the context of the TER59 complex. To verify this, RNase A digests were also performed on 3-[³²P]dAMP-labeled A78 RNA contained within the transcription complexes formed on the G21(TER59)A78 template. Transcripts 18-23 nucleotides long were protected (not shown).

Incorporation of bromo-UMP or iodo-UMP in lieu of UMP during synthesis of the UUUUUNU signal blocks transcription termination (13, 24, 25). This effect is unique to substituted uracil bases and is not observed when bromo-CMP or iodo-CMP replace CMP in the nascent chain (13). It was hypothesized that recognition of the uracil moieties, either by VTF or another component of the elongation complex, is essential for termination signal transduction. To eliminate elongation effects as a variable, we tested the ability of VTF to induce the release of BrUMP-substituted A78 RNA. Pulse-labeled G21 complexes were walked to A78 on the G21(TER29)A78 template in chase reactions containing either UTP or bromo-UTP (Fig. 4). (The BrUMP-substituted A78 RNA migrated more slowly than the UMP-containing transcript during gel electrophoresis.) The arrested complexes were then challenged with VTF and the bound and free RNAs were separated. VTF stimulated release of the control A78 RNA. (In this experiment, a low background of free RNA was seen in the absence of added VTF. The background free RNA varied in different experiments, but did not exceed 10% of the total RNA.) In contrast, the bromouridine-substituted A78 transcript was completely refractory to VTF-mediated release from the ternary complex (Fig. 4).

VTF-dependent termination requires significantly higher concentrations of ATP than are necessary to support RNA chain elongation by the vaccinia polymerase (17). We used the pulse-walk-release approach to determine whether ATP plays a role in transcript release. It was established previously that deoxyadenosine nucleotides could substitute for ATP in transcription termination (17). Hence, we assessed the ATP requirement for RNA release simply by adjusting the concentration of 3-dATP after the complexes had been walked to A78. (We could not use ATP in these experiments for the obvious reason that their inclusion in the reaction would permit resumption of elongation beyond the arrest site at +78A. The same applies to nonhydrolyzable analogs of ATP which are used as substrates for chain elongation by the vaccinia polymerase).

Control experiments confirmed that 10 µM 3-dATP was sufficient to arrest RNA polymerase at A78. Yet, when reactions containing 10 µM 3-dATP were challenged with VTF, the extent of RNA release was low (10% of total RNA) (Fig. 5). RNA release increased to 45% as 3-dATP concentration was raised to 50 µM and increased further to ~60% at 100 µM. The release reaction plateaued at ~60-65% at 0.1-1 mM 3-dATP (Fig. 5). We conclude that RNA release is ATP-dependent in the absence of concomitant elongation.

3-dATP dependence of transcript release.

It is instructive to compare these findings with previous studies of termination by G21 transcription complexes that were pulse-labeled and then chased through a TTTTTTTT termination signal. Hagler et al. (17) found that optimal termination efficiency was achieved at 0.5 mM ATP and that lowering the ATP concentration to 0.1 mM caused a shift in termination sites to more distal template positions. At 50 µM ATP, termination was barely detectable. From the present data, it appears that lower concentrations of adenosine nucleotide are needed to release transcripts from arrested transcription complexes than to elicit termination by elongating complexes. This is in keeping with the kinetic coupling model that posits ATP concentration as a determinant of signaling rate (17). In the case of transcription termination by elongating RNA polymerase, there is a window of opportunity during which VTF must transduce the signal, or else the polymerase will run off the end of the linear template. Hence signaling must be relatively rapid for the investigator to detect the termination event in this assay. In the pulse-walk-release assay, the polymerase is arrested at +78A, thereby providing an unrestricted temporal window for VTF to effect transcript release.

Even under optimal conditions, not all of the A78 transcripts were released upon incubation with VTF/capping enzyme. Did this reflect an equilibrium between VTF-responsive and VTF-unresponsive states of the ternary complexes or were the residual template-engaged transcripts inherently refractory to VTF-mediated release? We addressed the question as shown in Fig. 6. Pulse-labeled transcripts were walked to A78 and then incubated with VTF; this resulted in the release of 60-70% of the transcripts into the supernatant (Fig. 6, first round). The pelleted beads were resuspended in transcription buffer containing nucleotides and 3-dATP. This material was then incubated with or without VTF/capping enzyme and the bound and free RNAs were recovered. We found that RNAs that were not released during the first incubation with VTF were refractory to repeated challenge with VTF. Thus, these ternary complexes appeared to be inherently VTF-resistant. It was not simply the case that these were catalytically inert ``dead-end'' complexes, because the majority of the RNAs could be elongated to the end to the linear template when provided with ATP to overcome the elongation block at +78A (data not shown).

DISCUSSION

DNA templates containing tandem G-less and A-less cassettes have been exploited to study factor-dependent release of the nascent RNA chain from purified elongation complexes of vaccinia RNA polymerase halted at unique template positions downstream of a TTTTTNT transcription termination signal. The strategy of focusing on RNA release allows direct assessment of the signaling phase of the termination reaction, by eliminating ongoing elongation as a confounding variable. The principal findings are: (i) transcript release depends on an accessible UUUUUNU signal in the nascent RNA, (ii) the signal is inaccessible when it is contained within the nascent RNA binding site on the polymerase elongation complex, (iii) bromouridine base substitution within the signal interferes with factor-dependent release, and (iv) release depends on an ATP cofactor. The aforementioned properties and requirements for transcript release by halted polymerase complexes are consistent with all that is known about the composite termination reaction occurring during vaccinia early mRNA synthesis.

Several novel insights emerged from the experiments presented above that supplement our understanding of the relationship between elongation and signaling. For example, we provide evidence that a physical constraint to termination site-choice is imposed by the need to extrude the UUUUUNU signal out of the RNA binding site on the polymerase elongation complex. This finding has clear mechanistic implications, i.e. that the signal itself must be accessible to some component of the termination-competent transcription complex. The most plausible scenario is that UUUUUNU is bound directly by VTF/capping enzyme and that this is a prerequisite for transcript release. It is certainly not the case that the UUUUUNU signal must be extruded in order to for VTF to be recruited to the elongation complex, as we have shown that this occurs via signal-independent binding of VTF to the nascent RNA once a critical chain length is extruded from the polymerase (17). The length of the extruded chain on the A78 complexes is about 55 nucleotides (taking a conservative value of 23 nucleotides for the size of the RNA region protected by the polymerase), which we have shown is long enough to recruit VTF (17). Thus, VTF-unresponsiveness of the TER59 complex can be ascribed to lack of signaling, rather than lack of VTF binding. The same is true of the TER29 complex containing bromouridine-substituted RNA, i.e. the signal is exposed, but cannot be read by VTF. Bromouridine-substitution does not affect binding of VTF/capping enzyme to the ternary complex; indeed, VTF/capping enzyme can be efficiently UV cross-linked to bromouridine-substituted RNA within arrested elongation complexes (17). The implication is that binding of VTF to the unperturbed UUUUUNU signal elicits a conformational change in the elongation complex that precipitates transcript release. Defining the nature of the conformational change, and its connection to ATP hydrolysis, is the next mechanistic challenge.

An interesting and potentially instructive finding was that a fraction of the arrested complexes was refractory to VTF-induced RNA release and remained unresponsive after repeat challenge with VTF. We considered two models to account for this. One model posits that the vaccinia elongation complex can adopt VTF-responsive and VTF-unresponsive conformations. Conformational fluctuations of the elongation complex have been described in other polymerase systems (26, 27). Such conformational differences determine responsiveness to factor-induced RNA cleavage (27, 28). In the case of the vaccinia transcription complex, one might postulate that the polymerase active site must be appropriately positioned relative to the 3 end of the chain in order to release the chain in response to VTF, or that subtle differences in RNA-protein contacts within the RNA-binding site of the polymerase dictate VTF responsiveness.

A second model holds that VTF-responsive and VTF-refractory complexes differ not in conformation but in their protein composition. According to this view, VTF-dependent termination might require one or more additional protein cofactors that are present on some, but not all, of the A78 transcription complexes. (Alternatively, a polypeptide constituent might be differentially modified in a subfraction of the elongation complexes.) This would be consistent with the notion that VTF has either a partner or a direct target on the elongation complex besides the UUUUUNU signal. Candidates for this partner/target role would necessarily be polypeptides that are not strictly essential for chain elongation, otherwise the VTF-unresponsive complexes would not have been able to walk down the template to the end of the A-less cassette. The idea that a second component might cooperate with VTF to bring about transcription termination owes much to recent studies of the role of NusG during rho-dependent termination by Escherichia coli RNA polymerase (29, 30, 31). We discussed previously (17) the mechanistic parallels between VTF and rho, both of which transduce RNA signals to the elongating polymerase in an NTP-hydrolysis dependent, kinetically coupled reaction. Identification of a NusG-like function in the vaccinia system would consolidate the notion that RNA-based termination mechanisms are fundamentally conserved in bacteria and eukaryotes. In the future, we will seek to test the second model by analyzing the polypeptide composition of the A78 complexes. This will entail the preparation of immunological probes against all known transcription components, an effort that is currently under way.