Hydrolytic Cleavage of Nascent RNA in RNA Polymerase I11 Ternary Transcription Complexes*

Highly purified yeast RNA polymerase I11 ternary complexes were found to possess a hydrolytic chain retracting activity that cleaves nascent RNA from its 3“OH end. Most of the shortened transcripts were capable of resuming RNA chain elongation, indicating that they remain stably associated with the enzyme-DNA complex. Analysis of the products of cleavage indicated that retraction primarily occurred in dinucleotide increments, but that mononucleotides were also excised at lower frequency. The ribonuclease activity was totally dependent on the presence of a divalent cation and was stimulated by the addition of non-cognate ribonucleotides. The inclusion of ATP in the reaction enhanced both the rate and extent of transcript cleavage. Evidence suggesting that the hydrolytic activity is intrinsic to RNA polymerase I11 and factor-independent is also presented. Transcript cleavage by RNA polymerase 111 ternary complexes appears to be more closely related to the intrinsic nucleolytic activity of vaccinia virus RNA polymerase ternary complexes than to TFIIS-dependent cleavage that has been described for RNA polymerase II ternary complexes. Elongation of RNA chains in transcription is catalyzed by RNA polymerase within a highly processive enzyme-RNA-DNA ternary complex. The substrates for elongation are ribonucleoside triphosphates that are sequentially transferred to the 3”OH end of the nascent RNA through the formation of 3’-5’-phosphodiester linkages We find I11 on a SUP4 tRNA% is of cleaving nascent in both mono- and dinucleotide steps. Nucleolytic cleavage is stimulated by the presence of non-cognate triphosphates. We also present evidence that a dissociable, TFIIS-like factor is not involved with nucleolytic RNA chain retraction by pol

Highly purified yeast RNA polymerase I11 ternary complexes were found to possess a hydrolytic chain retracting activity that cleaves nascent RNA from its 3"OH end. Most of the shortened transcripts were capable of resuming RNA chain elongation, indicating that they remain stably associated with the enzyme-DNA complex. Analysis of the products of cleavage indicated that retraction primarily occurred in dinucleotide increments, but that mononucleotides were also excised at lower frequency. The ribonuclease activity was totally dependent on the presence of a divalent cation and was stimulated by the addition of non-cognate ribonucleotides. The inclusion of ATP in the reaction enhanced both the rate and extent of transcript cleavage. Evidence suggesting that the hydrolytic activity is intrinsic to RNA polymerase I11 and factor-independent is also presented. Transcript cleavage by RNA polymerase 111 ternary complexes appears to be more closely related to the intrinsic nucleolytic activity of vaccinia virus RNA polymerase ternary complexes than to TFIIS-dependent cleavage that has been described for RNA polymerase II ternary complexes.
Elongation of RNA chains in transcription is catalyzed by RNA polymerase within a highly processive enzyme-RNA-DNA ternary complex. The substrates for elongation are ribonucleoside triphosphates that are sequentially transferred to the 3"OH end of the nascent RNA through the formation of 3'-5'phosphodiester linkages and the release of pyrophosphate. The ternary complex also mediates the reverse reaction, pyrophosphorolysis, yielding back the substrates of elongation and a shortened RNA chain.
It has recently been demonstrated that a number of different RNA polymerases in ternary complexes have a second RNA chain retracting activity (reviewed in Ref. 1). In this process, retraction of the RNA polymerase along the DNA template occurs through hydrolytic cleavage of the growing end of the nascent transcript, generating short RNA products. It has been postulated that this hydrolytic retraction provides a mechanism for overcoming elongation arrest and that the process in which the paused complex retracts and subsequently resumes elongation facilitates multiple approaches to the block in transcription (2). Eukaryotic RNA polymerases must transcribe DNA that is packaged into chromatin, whose components, prin-min at 20-21 "C. The reaction was terminated by the addition of Na3EDTA to 10 m followed by gel filtration on a 1-ml Sepharose CL2B column equilibrated in transcription buffer without MgCI,. Ternary complexes were harvested by collecting 2-drop fractions (-30 pl). The first three fractions containing radioactive label were pooled, diluted in column buffer as required, and split into aliquots. Transcript truncation was performed at 20 "C in 15 pl final volume. Reactions were initiated by the addition of MgCl, to 7 m and terminated by the addition of Na3EDTA to 10 m~, followed by addition of 8 pl of 98% deionized formamide and heating at 100 "C for 3 min. Samples were analyzed on 20% (19:l or 39:l acrylamidehisacrylamide) polyacrylamide gels containing 8 M urea with 50 m~ Tris borate and 1 m~ Na,EDTA, as running buffer. The size of the C31 transcript was confirmed by the two-nucleotide extension generated upon addition of GTP to isolated C31 ternary complexes (see Figs. 1 and 4). Shortened transcripts generated by Mg2+induced cleavage were identified by counting bands shorter than the C31 initial complex on over-exposed autoradiograms. This indexing was confirmed for transcripts shortened to nt A19 by observing the effect of the presence of ATP and UTP individually on band intensity. ATP enhanced retraction signals at A26, A24, and A19 and UTP enhanced signals at U28 and U25. When an analysis of the small products of transcript cleavage was desired, samples were analyzed on 28% polyacrylamide (25:3 acrylamidehisacrylamide), 7 M urea sequencing gels with 89 m~ Tris borate, 2 m~ NazEDTA as running buffer (20). Autoradiograms were made from wet gels with or without an intensifying screen.
Thin Layer Chromatography (TLC) of Reaction Products-Transcript retraction reactions were carried out as described above. Samples were analyzed (30) by directly spotting aliquots (10 p1) onto a polyethyleneimine-cellulose TLC plate (Sigma), which was developed in either 0.5 or 1 M LiCl until the solvent front had migrated approximately 15 cm. The TLC plates were subjected to autoradiography with an intensifying screen.
Mono-and Dinucleotide Standards-The CMP standard was produced by mixing 1 pl of 0.066 p ,~ [cx-~*P]CTP (6000 Ci/mmol) with 3 pl of 1 N HCI and incubating at 100 "C for 10 min. The reaction was incubated on ice for a further 10 min before the addition of 3 pl of 1 N NaOH.

RESULTS
Purified pol I11 is capable of initiating transcription site specifically and efficiently at 3'-overhanging DNA ends generated by restriction endonuclease cleavage (28). We have used this property of pol I11 to examine whether it contains an intrinsic 3' 4 5'-ribonuclease activity, as previously observed with vaccinia virus RNA polymerase (31), or whether a separable elongation factor like TFIIS for pol I1 or GreA and GreB for E. coli RNA polymerase is required for efficient RNA chain retraction (2,22,25).
Ternary (pol 1II.DNA.RNA) transcription complexes contain-  1. Formation of C31 ternary complexes. The 5"flanking sequence of the SUP4 tRNA* gene (plasmid pLN4031) is shown extending from the 3"overhang that is generated by cleavage with PstI to the B'-proximal box A promoter element. Transcription was primed with GpG (underlined) to initiate at the single strand-double strand junction and elongated with ATP, [CX-~~PICTP, and UTP to position C31. ing a 31-nt nascent transcript were formed by priming transcriptional initiation with the dinucleotide GpG at a PstI-generated 5'-end located 14 bp upstream of the natural transcriptional start site of the SUP4 tRNAm gene and allowing RNA chain elongation to occur with a ribonucleotide mixture lacking GTP (Fig. 1). Incorporation of c~-~~P-labeled CTP in this elongation mix facilitated uniform labeling of the 31-nt transcript at positions C5, C7, C11, C14, C17, C27, C29, and C31. The products generated by this initiation protocol are shown in lane 1 of Fig. 2, panel a. Besides the predicted C31 RNA product2 several smaller RNA products were generated. Nearly all of these shorter RNA products were abortive (released from the complex) since they were separated from the linearized plasmid DNA, containing the intact ternary complexes, upon addition of EDTA and passage through a Sepharose CL2B column in buffer lacking M$+ (lane 2). These >lO-nt long abortive RNA products were unexpected since ternary complexes containing RNA chains of 10, 12, and 17 nt were previously found to be highly stable (32,26,33, and data not shown). Nevertheless, Sepharose CL2B chromatography allowed the generation of a nearly homogeneous C31 substrate suitable for examination of nucleolytic transcript shortening. A small proportion of complexes with a 32-nt nascent RNA chain was also generated (presumably due to trace deamination of ATP to ITP) and co-chromatographed with plasmid DNA on Sepharose. A majority of the C31 ternary complexes remained active through column isolation as evidenced by their ability to be elongated to full-length, terminating at the natural termination signal -120 bp downstream upon simultaneous addition of MgC12 and all four NTPs (lane 3 ) . Addition of MgC12 to 7 m~ in the absence of NTPs resulted in shortening of the C31 RNA transcript, generating a 29-nt RNA species within 5 s of incubation (lane 4 ) . After a 5-min incubation, 27-and 26-nt RNA were also detectable (lane 8). After 30 min of incubation in the presence of M$+, the 27-nt RNA had become the predominant shortened species. Longer exposures (not shown) of these autoradiograms also identified some smaller (c20 nucleotides) RNA. RNAs shorter than C31 were also generated by chasing stalled ternary complexes to full-length (lane 3 ). This suggests that instead of resuming elongation a subpopulation of C31 complexes retracted in the presence of M$+ and all four NTPs. Moreover, the fact that these transcripts were considerably shorter than 27 nt implies that ribonucleotides affected the backward reaction (and is examined further below). Nevertheless, the data suggested the action of a 3' + 5"nuclease analogous to the hydrolytic transcript cleavage previously observed with E. coli, pol 11, and vaccinia virus RNA polymerases (2, 19 22, 25, 31). No ribonucleolytic activity was detectable upon addition of pol 111, MgC12, and the DNA template to labeled C31 RNA not associated in a ternary complex (data not shown).
When all four NTPs were added after 5 and 30 min of M einduced retraction the 29-nt RNA containing ternary transcription complex was competent to resume RNA chain elongation, By convention, stalled complexes are identified according to the nucleotide at the 3' end and the length of the transcript.
The residual amount of the C31 complex that remained after 30 min of incubation in the presence of M$+ (lane 10) was very similar to the amount of apparently inactive C31 complexes that were unable to chase after ternary complex isolation (lane 31, and also had not chased after 30 min of incubation with M e (lane 11 ). In a separate experiment addition of ATP, CTP, In experiments not shown, transcript shortening by the pol 111 ternary complex was observed over a broad range ( 1-50 mM) of M e concentrations. Furthermore, other divalent cations could substitute for M e in supporting this cleavage reaction: retraction in the presence of 7 mM Mn2+ resulted in a similar distribution of shortened RNAs to that observed with M e after 5 min of incubation (that is to C29 and C27 in comparable amounts), while Zn2+ or Co2+ addition to 7 mM yielded transcripts predominantly shortened by two nucleotides but not significantly more. A U30 transcript was also observed upon Co2+ addition. Little or no transcript truncation was induced by Ca2+.
Although transcript cleavage occurred in the absence of exogenously added pyrophosphate, i t remained conceivable that RNA chain retraction was pyrophosphorolytic due to low levels of endogenous pyrophosphate bound to, and co-purifying with, DNA-bound polymerase. This does not appear to be the case, however, as RNA shortening observed upon addition of Mg2' to isolated C31 complexes (Fig. 3, lane 2) was not increased by 1 mM pyrophosphate (lane 6). A time course of RNA chain retraction also showed no significant increase in the rate of retraction upon inclusion of pyrophosphate (data not shown). In addition, RNA shortening in the presence of M e was not inhibited by a large excess of inorganic pyrophosphatase (lane 5). Thus, these results strongly indicate that the retraction process occurred by a mechanism other than pyrophosphorolysis. The product analysis shown below confirms this assessment.
Divalent metal ion-induced transcript cleavage by vaccinia virus ternary complexes has been demonstrated to be stimulated by CTP (31). We therefore determined the influence of nucleotides on RNA cleavage within the pol I11 ternary complex (Fig. 4). In these experiments nucleotides were added to a final concentration of 1 m~. The presence ofATP had a marked effect upon transcript truncation, increasing the extent of retraction, with A26 and A24 RNA becoming the predominant species  (panel a, lane 4). W also influenced the extent of retraction and the distribution of transcripts, generating mainly RNA species U28, A26, and U25 (lane 5). By comparison, CTP addition decreased the appearance of shorter transcripts, and RNA shorter than C27 was not detectable (lane 6). This result is consistent with retraction occurring in mononucleotide increments in which removal of the 3'-CMP residue would be accompanied by reincorporation of CMP. However, the presence of low levels of W, due to deamination of CTP, also would generate this result if retraction occurred in dinucleotide increments followed by reincorporation of both CMP and UMP. GTP addition allowed the stalled ternary complex to extend the RNA chain to position G33 as predicted by the sequence (lane 7).
Addition of each nucleotide in the absence of Mg2' had no detectable effect on chain retraction (lanes 8-11 1. In contrast to the disruption of ternary complexes that occurred upon Mg2"induced retraction to C27 and A26, complexes that retracted to A24 (and C29) upon addition of M e and ATP (Fig. 4b, lane 1 ) remained predominantly intact; addition of ATP, CTP, and UTP extended the transcript back to C31 (lane 2 ) and addition of all four NTPs generated full-length product (lune 3). ATP-enhanced retraction, nevertheless, resulted in mostly inactive RNA a t positions U28, C27, and A26 (compare lanes 2 and 3 with 1 ). If the presence of (the cognate substrate) ATP stabilized complexes retracted to A24 (against reiterative retraction occurring in increments of one to three nucleotides to U21, A22, andA23), it clearly was unable to do so at position A26 (where ATP is the only cognate nucleotide when retraction occurs in mononucleotide increments to U25). Thus, when retraction of two or seven nt occurred, the transcript remained associated with pol 111, whereas retraction of three, four, or five nucleotides led to the formation of inactive complexes. Evidently it is not simply the case that pol I11 releases the transcript after retracing its path along DNA by more than two nucleotides. Indeed, pol I11 may undergo conformational changes during retraction that results in the formation of either stable or metastable ternary complexes depending upon its position on the template.
Investigation of the kinetics of retraction showed that ATP increased the rate of RNA cleavage, both in regard to the extent of truncation and the proportion of the C31 complex converted to shortened RNA (Fig. 5; compare 0.25 min 2 ATP). Surprisingly, the addition of pyrophosphate to 1 m~ substantially reversed the stimulatory influence of ATP, such that the rate of transcript shortening was reduced and the distribution of retracted transcripts reverted toward that observed in the presence of Mg2' alone. This pyrophosphate-generated effect was not observed when inorganic pyrophosphatase was also included in the reaction mixture, supporting the contention that the phenomenon was due to pyrophosphate (data not shown). generated weak enhancement of retraction. This weakly enhanced retraction by non-hydrolyzable nucleotides may not necessarily signify that ATP hydrolysis was not required since the same experiments showed that ADP, AMP-PCP, AMP-PNP, and AMP-CPP were all contaminated with ITP or GTP (as evidenced by trace extension of the original C31 complex to a larger size). A 1% contamination of ATP would have approximated the enhanced level of retraction observed. The requirement of ATP hydrolysis in enhanced retraction therefore does not appear likely, but it is not excluded at this time.
RNA cleavage by pol I1 and by E. coli RNA polymerase ternary complexes has been demonstrated to be greatly stimulated by dissociable elongation factors (reviewed in Ref. 1). Although the preparation of pol I11 used for these experiments was highly purified (27), we could not rule out the possibility that a dissociable factor was responsible for, or stimulated, transcript cleavage. We therefore formed ternary complexes as previously described but added Sarkosyl to 0.3% prior to loading on the Sepharose CL2B column. A similar method has been employed by others to remove elongation factor TFIIS from pol I1 ternary complexes (22). A major fraction of the ternary complexes isolated by this procedure (Fig. 6, lane 1) remained active and were capable of elongating to full-length (lane 2 ) . Upon presentation with M e , transcript cleavage was detectable after 15 s (lane 3) and was more extensive than that observed for complexes that had not undergone Sarkosyl treatment (Fig. 2, lane 5). A similar experiment in which Sarkosyl was added to 0.3% after column isolation of C31 ternary complexes for 2 min resulted in considerable inactivation of these complexes such that only a small fraction were competent to resume elongation to full-length. We did note, however, that in the presence of 0.3% Sarkosyl equivalent proportions of complexes were capable of retraction and elongation (data not shown).
a-Amanitin, an inhibitor of RNA chain elongation, has been shown to inhibit RNA chain retraction of RNA polymerase I1 (34,22,19). Although a-amanitin is not an effective inhibitor of RNA chain elongation for yeast pol 111, tagetitoxin is (35). Addition of tagetitoxin to 8000 unitdml abolished production of full-length transcripts upon addition of nucleotides to columnisolated C31 complexes, but did not inhibit Mg2"induced retraction (data not shown). We did note that tagetitoxin appeared to change the register of retraction at this high concentration, such that Mg2"induced bands corresponding to U30 and U28 were observed in place of C29 and C27. However, product analysis, similar to that described below, demonstrated that addition of tagetitoxin resulted in the misincorporation of a pyrimidine nucleotide prior to Me-induced retraction. Likewise, addition of calf intestinal alkaline phosphatase to a reaction mixture containing tagetitoxin, prior to the addition of MgC12, but not after, eliminated the tagetitoxin effect (presumably due to the hydrolysis of the contaminating NTP) (data not shown).
We used polyethyleneimine thin layer chromatography (30) to characterize the products of the retraction process. Isolated C31 complexes uniformly labeled with CTP were exposed to M e to induce transcript truncation, and a portion of the reaction mixture was spotted onto a TLC plate that was developed in 1 M LiCl (Fig. 7a). The failure to find labeled CTP as the reaction product argues against cleavage via pyrophosphorolysis (lane 4 ) . Products co-migrating with CMP and dinucleotide markers were observed instead. We conclude that transcript cleavage in the pol I11 ternary complex occurs via hydrolytic cleavage rather than by pyrophosphorolysis. The use of 0.5 M LiCl as the solvent allowed greater resolution of mononucleoside monophosphates from dinucleoside diphosphates, predicted to be the hydrolysis products generated by incremental cleavage by one or two nt from the 3'-end of the RNA (note that pUpC and pCpU migrate identically in this system; see sequence in Fig. 1). Analysis of the reaction products in this way revealed that transcript truncation mostly yielded dinucleotides and, to a lesser extent, CMP (Fig. 7b, lane 4; 3'-UMP should migrate faster). Although inorganic phosphate also migrates between CTP and CMP in this system, other data indicate it is not a major hydrolysis product (see below). The inclusion of ATP in the reaction that was shown above to enhance cleavage, led to a decrease in the proportional yield of CMP (lane 5). We conclude that RNA chain retraction by pol I11 ternary complexes can proceed in single nucleotide and in double nucleotide steps. We sought to characterize the dinucleotide products further using high resolution polyacrylamide gel electrophoretic analysis. It is possible to resolve different species of dinucleotides on 28% polyacrylamide gels, the hierarchy of mobility being YpY > YpRlRpY > RpR (20). Thus, while it is possible to distinguish between dinucleotides with different base compositions it is not possible to resolve them by sequence (20). Portions of the reaction mixtures described in Fig. 7, a and b, were examined using this method, and the results are shown in Fig. 7c. Consistent with the polyethyleneimine-TLC analysis, the major products of retraction induced by M e were found to be CMP and a dinucleotide that co-migrates with pUpC/pCpU (lane 4 ) . As also observed by TLC analysis, ATP reduced the formation of CMP product. A putative dinucleotide species with a slightly lower mobility than pUpC/pCpU was also detectable, but this was present at a much lower level than the major dinucleotide product. Based upon its mobility we do not believe that this represents a trinucleotide. It is possible that this product is the result of pol I11 retraction past C27, which is seen at a low level in these experiments, or less probably from low level incorporation of IMP at position 32 in the starting reaction with subsequent generation of pCpI.
We also investigated the kinetics of product formation, demonstrating that the pUpC/pCpU dinucleotide is the major product 15 s after the initiation of transcript truncation (Fig. 8, lane  3). A small amount of CMP was also observed at this time. The other slowly migrating dinucleotide product formed more slowly, becoming detectable after 2.5 min of RNA chain retraction. If this slowly migrating product is pCpI, its rate of excision as the first product to be formed must be slow. Removal of 5"phosphate from these products with alkaline phosphatase was used to distinguish between dinucleotides with external and internal label. Since these transcripts were labeled with [(Y-~~PICTP, pUpC but not pCpU should have label that is phosphatase-resistant. Removal of the 5"phosphate from dinucleotide diphosphates has been previously demonstrated to result in a large decrease in mobility of the dinucleotide in this gel system (20,21). Upon treatment with alkaline phosphatase, most of the dinucleotide-incorporated label remained, but was contained in a much slower migrating compound (lane 11 ). When C31 complexes were boiled prior to the addition of MgClz and alkaline phosphatase no RNA cleavage products were generated (compare lane 12 with lane 2 ) indicating that the new product generated in lane 11 was not the result of an endonuclease contaminant in the alkaline phosphatase. These results are consistent with the conclusion that most of the dinucleotide product of 3' +. 5' RNA chain retraction contains an internal label and is therefore pUpC. Consistent with this conclusion, when transcripts were labeled with [cY-~~P]UTP most of the label in the dinucleotide retraction product was lost upon treatment with phosphatase (data not shown). This result is also consistent with the assumption that the observed dinucleotide cleavage products are 5'-rather than 3'-phosphorylated,

RNA Polymerase
Ill-associated Riboexonuclease 2305 since the initial dinucleotide product would be monophosphorylated which is not observed. Based upon our findings, we believe that while retraction can occur in either mono-or dinucleotide steps, the primary pathway for retraction of the pol I11 C31 ternary complex on the SUP4 tRNA template starts with the hydrolytic cleavage of two pUpC dinucleotides from the 3'-end of the transcript.

DISCUSSION
The potential to cleave nascent RNA by a hydrolytic mechanism has recently emerged as a property of several prokaryotic and eukaryotic RNA polymerase ternary complexes. The identification of such an activity associated with pol I11 ternary complexes reinforces the prediction that this RNA chain retracting activity may be universal. Several strategies for controlling this activity appear to operate in different systems. Efficient transcript cleavage by pol I1 is essentially dependent upon the dissociable elongation factor TFIIS, although at least part of the activity may reside within the core polymerase, since retraction, like elongation, is sensitive to a-amanitin (2,18,19,22,34). The experiments described in this work were carried out with highly purified pol I11 that was competent to catalyze transcript cleavage in the absence of any added factor. This in itself does not necessarily exclude the action of a dissociable protein. The E. coli elongation factors GreA and GreB remained undetected for some time due to their ability to contaminate even the most highly purified RNA polymerase preparations and their efficient ability to cycle substoichiometrically between ternary complexes (24,25). Nevertheless, other evidence suggests that this hydrolytic transcript cleavage activity may be intrinsic to pol 111, in that ternary complexes retained their ribonuclease activity even after Sarkosyl treatment and Sepharose chromatography had been employed to remove dissociable proteins from DNA-bound polymerase in ternary complexes. Thus, if pol I11 does employ an elongation factor it is by implication very tightly associated. In view of these considerations, RNA chain retraction by pol I11 seems much more closely related to the factor-independent activity associated with vaccinia RNA polymerase (31). Here one of the core RNA polymerase subunits, rp030, has been found to have significant homology with TFIIS and thus it has been proposed that this subunit performs a TFIIS-like function. There are more subunits in yeast pol I11 (16 at last count; 36) than in either pol I1 or pol I; conceivably one of these subunits could be related to an elongation factor. Although no statistically significant crossrelationships of amino acid sequence with TFIIS (derived from yeast, human, mouse, and fruit fly), GreA, GreB, or vaccinia virus RNA polymerase rpo30 subunit have been noted in 13 of these pol I11 subunits (analysis not shown), the amino acid sequence of three candidate subunits with approximate molecular masses of 37, 25, and ll kDa are not yet known (36).
There is a direct correlation between TFIIS-dependent RNA chain retraction and the ability of TFIIS to overcome blocks to RNA chain elongation that result from both intrinsic DNA sequences constituting arrest sites and obstacles generated by high affinity protein-DNA complexes (2, 8). Although pol IIIspecific transcription units are almost uniformly short (<200 bp), these sequences contain the binding sites of TFIIIC (box A and box B on tRNA (class 11) genes) and TFIIIA (box C on 5 S (class I) genes), such that TFIIIC-and TFIIIA-DNA complexes present potential obstacles to rapid elongation by pol 111. The delay to RNA chain elongation contributed by TFIIIC bound at its high affinity box B sequence elements has recently been estimated to be -0.2 s in a crude cell-free system (37), and the delay imposed by bound TFIIIC on RNA chain elongation by highly purified pol I11 was comparably insignificant (28). RNA chain retraction may play a role in facilitating transcription through bound TFIIIC (a protein equivalent in size to pol III), and the intrinsic nucleolytic activity present in highly purified pol 111 ternary complexes is consistent with the rapidity with which purified pol I11 surmounts the TFIIIC-DNA complex obstacle. The rate of retraction observed in this study, in which slightly more than half the active ternary complexes have retracted two nt by 5 s (Fig. 21, is apparently incompatible with chain retraction playing a significant role in limiting the TFIIIC-dependent delay of elongation to 0.2 s. However, the presence of NTPs clearly stimulates the rate of retraction by pol I11 (Figs. 4a and 51, and one might expect that the rate of retraction also has a sequence positional parameter. The appearance of aborted, <16-nt transcription products upon simultaneous addition of MgClz and all four NTPs (Fig. 2) suggests that retraction is rapid under these conditions, but this attribute has not yet been examined explicitly. Recently, it has been shown that TFIIS increases the efficiency of transcription by pol I1 through nucleosomal templates, suggesting that it may play a more general role in transcription than merely relieving elongation arrest at intrinsic termination sites (4). It is therefore tempting to suggest that the intrinsic hydrolytic cleavage activity provides pol I11 with a similar facility. On the contrary, pol I11 is incapable of transcribing nucleosomal templates in vitro (5), and thus the hydrolytic activity does not enable pol I11 to read-through chromatin. All the well defined yeast pol I11 genes have a requirement for TFIIIC in viuo, and it has been suggested that besides assembling TFIIIB into a promoter complex TFIIIC may also prevent tRNA genes from being assembled into chromatin in vivo (38). Indeed transcriptionally inactive tRNA genes can be incorporated into nucleosomes while active genes are not (39). Other evidence suggests that TFIIIC prevents repression of transcription by nucleosomes. The U6 gene can be transcribed in vitro with purified components in the absence of TFIIIC (40),3 but transcription in a crude extract and in vivo absolutely requires the box B element that is the primary determinant of TFIIIC binding (42). TFIIIC can activate U6 genes after chromatin assembly (43). Thus, pol I11 assembly factors associated with internal promoter elements may prevent nucleosomes from becoming impediments to RNA chain elongation.
Various small RNA cleavage products have been reported for retraction by pol 11. Wang and Hawley (19) observed exclusive production of mononucleotides while others (18, 20, 21) have reported the formation predominantly of dinucleotides and also of some mono-, tri-, and larger oligonucleotides. By comparison, RNA chain retraction by pol I11 yielded some mononucleotides, but predominantly dinucleotides. The product analysis coupled with the distribution of shortened RNA indicated that the primary pathway for ternary complex retraction from the C31 position on a SUP4 tRNA% template is the excision of two pCpU dinucleotides, but that other, undefined retraction pathways occur at a lower frequency. It is likely that the products of this reaction are, to some extent, dependent on the template and the position of the pause, as has been observed for TFIISmediated retraction by pol I1 (20,21).
Transcript cleavage in pol I11 ternary complexes appears to occur over a similar time frame to that observed for pol I1 and the vaccinia virus RNA polymerase. Retraction proceeds much more slowly than does elongation; this presumably ensures that transcript cleavage does not occur until the ternary complex has paused. Retraction by pol I11 on the tRNA template was less extensive than retraction in other systems. It is not clear whether this is a generalized feature of pol I11 retraction or a characteristic of the template employed. Indeed when the submitted. C. A. P. Joazeiro, G. A. Kassavetis, and E. P. Geiduschek, manuscript reaction was not supplemented with ATP, Me-dependent transcripts shorter than 27 nt were not readily detectable even though transcript cleavage could extend backward past C27 to A18 without further loss of label. However, C29 and C27 transcripts did diminish with time implying that retraction past them does occur. Therefore, transcript truncation may occur rapidly and via poorly defined pathways further back than C27.
One surprising feature of RNA truncation by pol I11 was the high proportion of shortened transcripts that were incapable of resuming elongation. This has not been observed in other systems, where the vast majority of cleaved transcripts can chase (22,25,31). The formation of inert RNA seems to be linked with transcript release or with a decrease of stability within the ternary complex. The formation of these metastable complexes was observed to occur at particular positions on the template, implying some influence of template sequence on ternary complex stability. This phenomenon may also explain why, during the initial transcription reaction to form C31 complexes, abortive transcripts greater than 10 nt were apparently produced. It would be predicted that complexes stalled at C31 and other positions would retract such that some ternary complexes would re-extend while others would become stagnant and thus not survive column chromatography.
Nucleotides clearly influenced retraction by pol 111; this has also been observed for vaccinia virus RNApolymerase (311, and it is conceivable that this may be a feature of factor-independent retraction. Nucleotide precursors were found not to enhance TFIIS-mediated retraction by pol I1 (18). The precise molecular basis for the influence of ribonucleotide precursors is not presently certain, but we have noted that a requirement for ATP hydrolysis does not appear likely. Most probably NTPenhanced retraction results from an allosteric or steric effect of the nucleotide on pol 111. The observation that ATP stimulated retraction more than did UTP, and that CTP did not stimulate at all (Fig. k ) , suggests that the presence of the non-cognate nucleotide precursor in the active site may trigger retraction. Alternatively, (relatively slow) misincorporation of the noncognate nucleotide may initiate (relatively fast) retraction. If pyrophosphorolysis and nucleolytic retraction compete for the removal of the misincorporated nucleotide, then one might expect pyrophosphate to inhibit non-cognate NTP-enhanced retraction as observed in Fig. 5. Indeed, recent evidence indicates that GreA rapidly and preferentially mediates cleavage of misincorporated nucleotides within E. coli RNA polymerase ternary complexes (44) and thereby provides a conceivable proofreading mechanism. A similar proofreading role for TFIISmediated retraction has also been proposed (2,18,19,31). Whether misincorporated nucleotides are preferentially excised by nucleolytic action within pol 11, pol 111, and vaccinia virus RNA polymerase ternary complex remains to be determined.
Chain elongation by both eukaryotic and prokaryotic RNA polymerases is unsteady due to pausing at diverse intrinsic sites along the transcription unit (reviewed in 11,41). The half-times for escape of RNA polymerase from pause sites ranges from less than 1 s to multiple min. Pausing has been interpreted with the implicit assumption that slow escape is synonymous with slow incorporation of the next nt beyond the pause site. It is conceivable that nucleolytic retraction also needs to be taken into consideration when interpreting features of RNA chain elongation and transcriptional pausing. For example, a weak site of pausing may appear strong because it is the backward motion end point of RNA chain retraction that is initiated at a downstream site. The resulting cycling of RNA polymerase between two sites would generate high occupancy at each end and give the appearance of slow release of the polymerase from a pause site. Kinetic modeling of RNA chain elongation in terms of a series of directly measured, pseudofirst order rate constants fails to generate the observed distribution of chain lengths during elongation by yeast pol 111, implying that elongation is mechanistically complex (37). Nucleolytic retraction may be one source of this complexity.