Substrate Specificity of the RNase Activity of Yeast RNA Polymerase III*

Using yeast RNA polymerase III ternary complexes stalled at various positions on the template, we have analyzed the cleavage products that are retained and released by the transcription complexes. The retained 5′ products result from cleavage at uridine residues during retraction, whereas the yield of mononucleotides and dinucleotides released indicates that multiple cuts occur near the 3′ end. Comparison of the cleavage patterns of uridine-containing and 5-bromouridine-containing transcripts suggests that RNA within an RNA-DNA hybrid duplex is the substrate for the 3′-5′ exonuclease. During transcription of the SUP4 tRNATyr gene, RNA polymerase III produces not only full-length pre-tRNATyr but also short oligonucleotides, indicating that exonuclease digestion and transcription are concurrent processes. To explore the possibility that these oligonucleotides are released by the action of the RNA polymerase III nuclease at previously observed uridine-rich pause sites, we tested modified templates lacking the arrest sites present in the SUP4 tRNATyr gene. Comparative studies of cleavage during transcription for these templates show a direct correlation between the number of natural pause sites and the yield of 3′ products made. At the natural arrest sites and the terminator, RNA polymerase III carries out multiple cleavage resynthesis steps, producing short oligoribonucleotides with uridine residues at the 3′ terminus.

Using yeast RNA polymerase III ternary complexes stalled at various positions on the template, we have analyzed the cleavage products that are retained and released by the transcription complexes. The retained 5 products result from cleavage at uridine residues during retraction, whereas the yield of mononucleotides and dinucleotides released indicates that multiple cuts occur near the 3 end. Comparison of the cleavage patterns of uridine-containing and 5-bromouridine-containing transcripts suggests that RNA within an RNA-DNA hybrid duplex is the substrate for the 3-5 exonuclease. During transcription of the SUP4 tRNA Tyr gene, RNA polymerase III produces not only full-length pre-tRNA Tyr but also short oligonucleotides, indicating that exonuclease digestion and transcription are concurrent processes. To explore the possibility that these oligonucleotides are released by the action of the RNA polymerase III nuclease at previously observed uridinerich pause sites, we tested modified templates lacking the arrest sites present in the SUP4 tRNA Tyr gene. Comparative studies of cleavage during transcription for these templates show a direct correlation between the number of natural pause sites and the yield of 3 products made. At the natural arrest sites and the terminator, RNA polymerase III carries out multiple cleavage resynthesis steps, producing short oligoribonucleotides with uridine residues at the 3 terminus.
A hydrolytic activity that effectively reverses the course of gene transcription has been found for Escherichia coli RNA polymerase, eukaryotic RNA polymerases I, II, and III, and the vaccinia virus RNA polymerase (1)(2)(3)(4)(5)(6)(7)(8)(9). In the first three cases a separate protein co-factor serves to activate the nuclease, whereas for RNA polymerase III (Pol III) 1 and the vaccinia polymerase no such co-factor is known. In attributing a functional role to the transcription factor IIS-stimulated nuclease activity of Pol II, major emphasis has been placed upon the role that nucleolytic retraction plays in overcoming the arrest of elongation at certain template positions (10 -12) that are rich in adenosine residues in the template DNA strand (13)(14)(15)(16). In the only previous report of an intrinsic nuclease in RNA polymerase III, Whitehall, Bardeleben, and Kassavetis (8) noted the excision of dinucleotide cleavage products when transcription was halted within the sequence 5Ј-UCUC in a yeast tRNA Tyr gene.
In this paper we have extended the study of pol III nuclease activity to elongation complexes stalled within a number of different sequence contexts. For several of these we characterized both products of nuclease cutting, the released 3Ј oligonucleotides as well as the shortened 5Ј transcript. An examination of the respective 3Ј product sizes relative to that of the RNA transcript underscores a surprising dilemma. More than one oligonucleotide fragment must have been produced per surviving 5Ј fragment to satisfy conservation of mass. That is, during conversion of the initial RNA into the first large product observed, the nuclease must have catalyzed multiple cleavage events.
The pattern of exonuclease degradation is also distinctive as regards the sequences within which the 3Ј termini of shortened transcripts occur. For most of the artificially arrested Pol III elongation complexes that we examined, the cleavage occurs just 5Ј to an internal uridylate residue. Because the bond between a 3Ј-terminal uridylate and the penultimate nucleotide is not efficiently cleaved, we infer that in growing RNAs having an oligouridine stretch at the 3Ј end, the last residue is not base paired to the DNA template. This suggests that at arrest sites, the 3Ј transcript end becomes detached from the template due to the exceptional instability of (rU⅐dA) hybrids (17). Nucleolytic removal of several residues from the RNA 3Ј terminus can effectively "reset" elongation by restoration of the DNA-RNA heteroduplex.
Although elongation arrest, 3Ј end misalignment, and retraction with restoration of alignment provide an attractive scenario for polymerase-nuclease function, our data suggest that the system possesses one additional element. The RNase H-like activity associated with yeast Pol III chooses to cut nascent RNA just 5Ј of unstable RNA-DNA base pairs. A base pair even more unstable than those we have studied would result from base misincorporation. This line of reasoning suggests that Pol III nuclease may have a proofreading function similar to the well known 3Ј-5Ј exonuclease of many DNA polymerases (18). Resulting circular double-stranded plasmid DNAs were used as a templates for polymerase chain reaction amplification of ϳ400 -500 bp linear double-stranded fragments carrying original, mutant SUP4 tRNA Tyr or chimeric Pol III genes. Oligonucleotides complementary to sequences ϳ200 bp upstream of the starting site and ϳ100 bp (or in the case of chimeric gene ϳ200 bp) downstream of the termination site for transcription were used as primers. Because forward primer was synthesized using a biotin derivative as the first nucleotide at the 5Ј end, the resulting polymerase chain reaction products had 5Ј-terminal biotin on the noncoding strand.

Reagents and Enzymes-[␥-
Linear double-stranded DNA fragments obtained as described above were attached to the magnetic beads through streptavidin-biotin interactions. These DNA templates immobilized on magnetic beads were used in all transcription experiments. The average concentration of template was about 300 pmol/ml of beads.
Preparation of Purified Ternary Complexes-Yeast P-11 extract prepared as in Ref. 23 was used as a source of RNA Polymerase III and its transcription factors. Ternary complexes stalled at position ϩ17 were formed in a 0.18-ml reaction mixture that contained 0.06 ml of P-11 extract, 18 pM linear double-stranded template attached to the magnetic beads, 0.5 mM ATP and CTP, 0.3 M [␣-32 P]UTP in transcription buffer (20 mM HEPES-KOH (7.9), 100 mM KCl, 7 mM MgCl 2 , 3 mM dithiothreitol, 6% glycerol) for 30 min at 25°C. The reaction was stopped by the addition of EDTA to 20 mM. Ternary complexes were magnetically concentrated and washed three times with excessive volume of washing buffer (20 mM HEPES-KOH (7.9), 100 mM KCl, 3 mM dithiothreitol, 0.25 mg/ml BSA, 10% glycerol) prior to their use in cleavage assays or for further transcription in reaction mixtures lacking one or more of the nucleoside triphosphates.
Cleavage and Transcription Assays-For the analysis of 5Ј-proximal products, transcript cleavage was performed at 25°C in 20-l reactions of transcription buffer. After incubation for various times, reactions were terminated by the addition of an equal volume of stop-mixture (1 mg/ml Proteinase K, 1% SDS, 20 mM EDTA, 0.25 mg/ml carrier RNA) followed by 65°C incubation for 20 min, magnetic separation of RNA products from magnetic beads, ethanol precipitation, and drying. Samples were dissolved in 10 l of loading buffer (100 mM NaAc (5.5), 8 M urea, 0.025% xylene cyanole dye) and loaded on 15%-polyacrylamide (29:1 acrylamide/bisacrylamide) 7 M urea gels. The polyacrylamide gel electrophoretic separation was performed with TBE (45 mM Tris borate (8.0), 1 mM EDTA) as running buffer.
For the analysis of short cleavage products, the 3Ј end of RNA was labeled by using [␣-32 P]GTP as a substrate with the addition of other appropriate NTPs, when necessary. After purification of ternary complexes, cleavage was carried out in 6-l reactions that were stopped by the addition of EDTA to 20 mM final concentration, followed by the addition of 3.6 l of 98% formamide and heating at 95°C for 30 min. After removal of the magnetic beads, samples were analyzed by 20% polyacrylamide gel electrophoresis as described above.
Cleavage during Transcription-For the analysis of cleavage concurrent with transcription, 17-mer ternary complexes were formed in the presence of "cold" ATP, CTP, and UTP, magnetically separated, and washed as described above. Transcription was then restarted by the addition of 0.2 mM CTP, ATP, and GTP and 2 M [␣-32 P]UTP, and 0.5 mg/ml heparin to prevent reinitiation. Total volume of each reaction was 6 l. After transcription was allowed to proceed for various times, reactions were quenched by the addition of EDTA to 20 mM concentration and 1:3 v/v of 98% formamide. After heating at 95°C for 30 min and magnetic separation, samples were analyzed by the autoradiography of dried 20% gels.
Preparation of Poly(ribo-A) 17 Mono-and Oligonucleotide Standards-The labeled pG and pU were prepared according to the procedure described in Ref. 8. Labeled oligonucleotides were obtained according to Ref. 5 by phosphorylation of appropriate 5Ј-nonphosphorylated oligonucleotides with T4 polynucleotide kinase.
RNase Analysis of 3Ј Cleavage Products-To obtain sufficient amounts of the short 3Ј products formed during transcription of the SUP4 gene we carried out the transcription reaction according to the procedure described above (see "Cleavage during Transcription") in a scaled up variation. Elongation with unlabeled 17-mer ternary complexes was restarted in the presence of 0.2 mM CTP, ATP, GTP and 14 M [␣-32 P]UTP and heparin (0.5 mg/ml; molecular weight, ϳ6000) in a total reaction volume of 90 l. To separate short products from fulllength transcripts and heparin, we used Microcon 3 microconcentration, which allowed us to get rid of molecules longer than 10 nucleotides in length or larger than ϳ3000 in molecular weight. Then 3Ј products were desalted by G-10 gel filtration using spin columns pre-equilibrated with water. Digestion with cytidine-specific RNase CL3 was performed by incubation in 20 mM HEPES-KOH (pH 7.9) in the presence of 8 M urea at 50°C for 1 h. Pyrimidine-specific reactions with RNase A and guanosine-specific digestions with RNase T 1 were carried out in 50 mM Tris-HCl (pH 7.5) at 25°C for 1 h. Adenosine-specific reactions with RNase U 2 were performed in 8 mM sodium citrate buffer (pH 3.5) in the presence of 1.25 M urea at 50°C for 1 h. Calf intestinal phosphatase was added, when necessary, and dephosphorylation was carried out at 25°C for 5 min. Reactions were stopped by adding 1:3 v/v of 98% formamide and heating at 95°C for 30 min. Samples were loaded on sequencing size 20% polyacrylamide gels with appropriate marker ribooligonucleotides. Electrophoretic separation was carried out at 800 V, and after xylene cyanole had entered the gel, the voltage was being increased to 1200 V. This allowed us to obtain undistorted picture of short mono-and dinucleotide products with mobility that depended on length and sequence composition.
We identified short 3Ј products by their mobility relative to the mobility of corresponding ribooligonucleotide markers and by digestion with sequence-specific RNases (see Fig. 9). The main product had the mobility of pPypPy (lane 5), the major purine-containing products were pG*pU and pA*pU, with minor ones *pUpG/pPypPypPy and *pUpA (compare lane 5 with lanes 9 and 11). Quantitative PhosphorImager analysis of various RNase digests of the 3Ј products, with or without dephosphorylation, allowed us to evaluate the ratios between different pPypPy products that could not be resolved electrophoretically. Because two lower bands of dephosphorylated products disappeared when they were treated with pyrimidine-specific RNase A (see Fig. 9, compare lanes 6 and 8), we concluded that they correspond to U*pU and C*pU products, which are present in equal amounts. According to the number of radioactive phosphate groups, we assumed that the intensity of *pU*pU product should be two times higher than that of pC*pU. Because we observed a ϳ10% difference between the total 32 P content in all pPypPy dinucleotide and the intensities of *pU*pU and pC*pU calculated from the amounts of U*pU and C*pU, we concluded that ϳ45% of pPypPy product consists of *pU*pU, ϳ45% consists of pC*pU, and ϳ10% consists of *pUpC. We did not observe any noticeable digestion with the cytidine-specific RNase CL3 (see Fig. 9, compare lanes 1 and 3). We attribute this to the inability of this enzyme to work efficiently with substrates as short as dinucleotides. Dephosphorylation in the presence of guanosine-specific RNase T 1 and adenosine-specific RNase U 2 showed that the upper dephosphorylated band corresponds to G*pU and A*pU products (see Fig. 9, compare lane 6 with lanes 10 and 12). We explain the appearance of an additional unidentified band during dephosphorylation in the presence of RNase U 2 (see Fig. 9, lane 12, band marked with an asterisk) by the known ability of this enzyme to catalyze oligonucleotide synthesis (25).

RESULTS
In their studies of Pol III elongation complexes made with highly purified Pol III and a 3Ј extended template, Whitehall et al. (8) showed that stalled polymerase can efficiently degrade the nascent transcript. From their experiments with a ternary complex having UCUC at the 3Ј end, these authors concluded that the hydrolytic activity cuts primarily in dinucleotide increments. They noticed that mononucleotides were also produced but at a lower frequency. To examine this process in detail, we studied cleavage reactions in Pol III transcription complexes stalled within many different sequence contexts. All complexes were formed in a promoter-dependent system that required Pol III factors for initiation. The transcripts were elongated up to the first G of the SUP4 gene (see Fig. 1), and then the 17-mer ternary complexes were isolated magnetically, extensively washed, and "walked" to other positions of interest on the template. To generate additional types of 3Ј-terminal sequences, we mutagenized the SUP4 gene triply at positions ϩ15, ϩ16, and ϩ17, thereby changing the TCT sequence in SUP4 to AAA in mutant template SUP4-AAA and to TTT in mutant SUP4-TTT (see Fig. 1).
Cleavage in the 3Ј-5Ј Direction in Stalled Ternary Complexes Occurs in a Sequence-specific Manner-To detect the cleavage products that remain in ternary complex (5Ј products) we labeled the 5Ј-proximal 17-mer with 32 P. When the washed 17mer complex was incubated with 7 mM Mg 2ϩ , both the rate and the pattern of cleavage were similar to those reported previously (8) (Fig. 2A). In this experiment the sequence of the transcript 3Ј end was the same as that made and cleaved by Whitehall et al. (8). Fig. 2B shows the cleavage of 17-mer formed on the SUP4-AAA template. The observed pattern of 5Ј products indicates cleavage at positions ϩ16 and ϩ14, with ϩ14 as the preferred cut site. At first glance, this would appear to confirm the previously proposed dinucleotide mode of cleavage. However, the patterns of nuclease cleavage of 19-and 22-mer arrested products made on the SUP4 and SUP4-AAA templates (Figs. 3, A and B, and 4, A and B) suggest that preferential cleavage occurs at certain sites. Comparison of the gel patterns with the SUP4 and SUP4-AAA sequences shows that selective cleavage occurs at the 5Ј-phosphate groups of uridine residues, beginning with the uridine nearest the 3Ј end and then proceeding processively in the 3Ј-5Ј direction. When a cluster of at least three consecutive A residues is present, Pol III can also cut 5Ј to the central adenosine residue.
RNA in a Duplex with Template DNA Is a Substrate for the RNA Polymerase III Nuclease Activity-We used the SUP4-TTT template to examine further the basis for the preferential cleavage at uridine residues. SUP4-TTT transcripts limited to length 17 were made either with UTP, CTP, and ATP or with BUTP, CTP, and ATP. The mode of exonuclease degradation for the two stalled ternary complexes is shown in Fig. 5 (A and  B). Cleavage of the bromouridine transcript occurs processively beginning next to the 3Ј-terminal bromouridine residue (position ϩ17), whereas for uridine-containing RNA it occurs mainly  at position ϩ16 rather than ϩ17. This difference is clearly seen after 1 and 5 min of incubation in the presence of Mg 2ϩ . Only a small fraction of the 3Ј-terminal uridine residues are recognized by the nuclease as a cleavage substrate, whereas bromouridine-terminal residues are cleaved more effectively. We attribute this difference in nuclease action between uridine RNA and bromouridine RNA to a greater extent of duplex formation by the 3Ј-terminal bromouridine than by the 3Јterminal uridine of uridine RNA. Such a difference is expected from the known enhancement in base pairing between dA⅐rBU in comparison with dA⅐rU (26,27). This conclusion is further supported by differences in the ability of uridine-and bromouridine-containing ternary complexes to restart elongation. After 5 min of incubation with Mg 2ϩ , all bromouridine-ternary complexes were chased quantitatively into full-length RNA upon the addition of NTP substrates, whereas in the case of uridine-containing complexes, only products resulting from degradation (ϩ16 and ϩ15 ternary complexes) were elongated. The residual ϩ17 complex was totally incompetent for elongation. These considerations suggest to us that the substrate for the Pol III nuclease activity is an RNA involved in duplex structure with template DNA.
For transcripts that contain internal uridine stretches, substitution of uridine by bromouridine has the effect opposite to that noted above. The results of cleavage of uridine-and bromouridine-containing 19-mer RNAs formed on SUP4-TTT are shown in Fig. 5 (C and D). In the case of uridine-containing 19-mer, RNA cleavage occurs at each uridine residue starting from that nearest the 3Ј end and proceeds sequentially (Fig.  5C). The major cleavage sites observed were at positions ϩ17, ϩ16, and ϩ15. In the case of bromouridine-containing 19-mer, similar sequential cleavage at positions ϩ17, ϩ16, and ϩ15 was observed but to a significantly lesser extent at the last two positions than for the uridine-containing transcript (Fig. 5D). The major cleavage site was at position ϩ17.
Hence, when a UUUU sequence was located at the very 3Ј end of transcript (17-mer complex), it was not cleaved effectively, whereas for the 19-mer complex that has two guanosine residues at the 3Ј end to stabilize the heteroduplex structure, this same oligouridine stretch was recognized as a nuclease substrate and cleaved at each uridylate. The opposite effect was observed for the (BU) 4 stretch. When located at the 3Ј end, it was readily cleaved, whereas when it is followed by two guanosines (in the 19-mer) the nuclease cut only at the 3Ј-most bromouridylate (position ϩ17). These data imply that the nuclease activity is sensitive to the secondary structure of RNA-DNA heteroduplex rather than to the chemical structure of the ribonucleotide bases in selecting cleavage sites.
Identification of the 3Ј Cleavage Products-To analyze the small products cleaved from the 3Ј terminus of nascent RNA we made 19-, 22-, 24-, and 27-mer ternary complexes in which RNA was labeled with [␣-32 P]GTP. If the 5Ј products of nuclease action upon 19-and 22-mer stalled transcripts were generated by a single nuclease cut, we would have expected to observe pUpC*pG*pG as the 3Ј product of 19-mer cleavage and pUpA*pG and pUpC*pG*pG as products of 22-mer degradation. Instead of these, significantly shorter 3Ј products were formed, suggesting cutting at additional positions (Fig. 6). To further analyze the 3Ј products formed concurrently with the 5Ј products, we performed cleavage experiments with 24-and 27-mer ternary complexes. In these the uridine residues near- est the 3Ј terminus were five and eight nucleotides away, respectively. Results on the cleavage of 24-and 27-mer ternary complexes are shown in Fig. 6 (C and D). Although a 5Ј product was observed that corresponds to the expected cleavage at U20, 3Ј products longer than two nucleotides were not detected. Hence, for all ternary complexes investigated we observed 5Ј products corresponding to uridine-specific cleavage but multiple 3Ј products that were shorter than expected.
The data on 3Ј products formed during cleavage of 19-and 22-mer are summarized in Table I. According to their mobility in the 20% polyacrylamide gel the 3Ј products formed by 19mer cleavage are identified as *pG, pC*pG, and *pG*pG (Fig.  7A, lane 2). In the case of 22-mer cleavage *pG, pC*pG, pA*pG, and *pG*pG 3Ј products were produced (panel A, lane 7). When transcripts were synthesized in the presence of [␣-32 P]UTP *pU, *pUpC, and *pUpA 3Ј cleavage products were observed in the case of 19-mer (panel B, lane 2) and *pU, *pUpC, *pUpA, trace amounts of *pUpApG, *pUpC*pU, and/or pC*pUpC 3Ј products were formed during 22-mer cleavage (panel B, lane 6) To examine the possibility that the apparent contradiction between observed 5Ј and 3Ј products results from the action of an exogeneous contaminating RNase activity, we analyzed the effect of two potential nuclease inhibitors. These were UpA and poly(A) [17][18][19][20][21] , having 5Ј-phosphate and 3Ј-OH groups at the termini (Fig. 7A, lanes 4 and 3). The pattern of products was unaffected by the presence of exogenous RNA (compare with the cleavage in the presence of only Mg 2ϩ , Fig. 7A, lane 2). To test for the presence of a 5Ј-phosphate group, the 3Ј products were also treated with calf intestinal phosphatase (Fig. 7A,  lane 5). All the low molecular weight products had a 5Ј-phosphate group. Possible contamination by exogenous RNase was also tested by carrying out the cleavage reactions in the presence of labeled tetramer *pUpCpGpG. Because the amount of tetramer was the same in the presence or the absence of ternary transcription complex (compare lanes 2 and 3 and lanes 5 and 6 in Fig. 7B), we conclude that no contaminating exogenous RNase was present. These results show that the breakdown of RNA to short products occurs only within the ternary complex.
While Transcribing the SUP4 tRNA Tyr Gene RNA Polymerase III Produces Both Pre-tRNAs and Short Oligoribonucleotides-To test for a possible role of the nuclease function in overall Pol III enzymatic activity we carried out experiments to detect all products formed during single-round transcription of the SUP4 tRNA Tyr gene.
After the initiation of transcription and formation of arrested 17-mer complexes, all dissociable proteins as well as possible abortive initiation transcription products were removed by extensive washing of the bead-bound ternary complexes. Elongation was then restarted in the presence of heparin to prevent reinitiation. After transcription had gone on for various times, reactions were stopped and samples were loaded directly onto a 20% polyacrylamide gel. If during elongation without interruption cleavage events occur as they do for artificially arrested complexes, we would have expected to see short oligoribonucleotide products formed. In fact (Fig. 8B, lanes 9 -12), large amounts of dinucleotides were formed concurrently with pre-tRNA molecules during transcription of the SUP4 tRNA Tyr gene. These consisted mainly of pUpU, pCpU, pApU, and pGpU (see "RNase Analysis of 3Ј Cleavage Products" under "Materials and Methods" and Fig. 9). In addition, small amounts of pUpC, pUpA, and pUpG were formed. Each product possessed a 5Ј-phosphate group removable by phosphatase treatment. Because those products accumulate progressively with time and have the same structure as the cleavage products made by arrested complexes, we conclude that these oligoribonucleotides are produced by RNA polymerase III molecules traversing the gene.
RNA Polymerase III Carries out Multiple Cleavage and Resynthesis Steps at Intrinsic Arrest Sites and Terminators-To test for a possible dependence of 3Ј nucleolytic cleavage upon the arrest of ternary transcription complexes at natural pausing sites, we made qualitative and quantitative comparisons of the 3Ј products formed on three different templates varying in the number of pause sites. These were a chimeric Pol III template (Table II), containing its only uridine residue at position ϩ20; the ϪA36A37-SUP4 deletion mutant (22), which has only one UUU sequence before its terminator; and the SUP4 tRNA-Tyr gene, which has three UUU sequences within the transcribed region. When transcribed these three genes exhibit, as expected, no, one, and three strong pause sites, respectively. The sequences of these templates are listed in Table II.
Because the ϪA36A37-SUP4 template terminates transcription at the T 6 sequence (positions ϩ47-52) created by deletion (Fig. 8A, lanes 5-8), there remains only one pause site, which gives the doublet bands at sites ϩ33 and ϩ34. As expected, the chimeric template gives no paused products, only terminated  ones (Fig. 8A, lanes 1-4). The formation of a pPypPy cleavage product on the ϪA36A37 template was about three times higher than for the chimeric template, where the distribution of products was reflective of cleavage at the terminator only. In addition, a considerable amount of pApU product was formed during transcription of ϪA36A37 but not the chimeric template. Because of the similarity of transcript length and terminator sequence between these two templates, we attribute the greater diversity and higher level of 3Ј products seen with ϪA36A37-SUP4 (Fig. 8, compare lanes 1-4 and 5-8) to pausing and restarting at the ϩ33-34 site. Further comparison with the SUP4 template (Fig. 8, lanes 9 -12), which has several more arrest sites, shows a remarkable increase in amounts of pApU, pUpU, and pGpU products consistent with the presence of additional strong pause sites at the positions ϩ52-54, ϩ69 -70, and ϩ82-83 that have the sequences AUU, CUU, and GUU, respectively. PhosphorImager analysis of the ratios between the amount of full-length RNA accumulated and 3Ј products released per uridine residue reveals a direct correlation between the number of pause sites present and the quantity of nucleolytic products made. Because the molar ratio between pGpU and full-length product was not less than ϳ90 for any of the templates investigated, we conclude that during post-initiation transcription, RNA polymerase III carries out multiple cleavage and resynthesis steps at arrest sites and terminators. DISCUSSION The RNA polymerase-related nucleases previously described in detail exhibit a dual nature (1-8, 28). The ribonuclease activities associated with E. coli RNA polymerase and with eukaryotic Pol II both degrade nascent RNA sequentially from the 3Ј terminus, releasing, in many instances, mono-, di-, and trinucleotide split products (1,4,5,28). In that sense, their associated nuclease activities behave as processive 3Ј-5Ј exonucleases. However, depending both upon the type of elongation block imposed and the elongation factor that is present, both nucleases (1,4,6) can produce longer (7-14 nucleotides) 3Ј products.
For retraction by RNA polymerases, the processive cleavage that occurs starting from the 3Ј terminus is indicative of exonucleolytic degradation, whereas the ability to generate long 3Ј products under other conditions implies an endonucleolytic mode of action.
The Substrate for Ribonuclease Activity-Our study of the Pol III nuclease has concentrated mainly upon the roles that RNA sequence and RNA-DNA secondary structure play in determining cleavage site selectivity. At each arrest site we studied, newly formed shortened transcripts were successfully elongated up to the full-length product, independently of the length of RNA digested away. The resumption of transcription by these complexes suggests that the transcript region just 5Ј to the site of cleavage is held in an RNA-DNA heteroduplex. This raises a question as to the structure of the nuclease substrate on the 3Ј side of the cleavage site. To answer that question we compared the pattern of cleavage for a 17-mer transcript having four uridine residues at the 3Ј end with that of bromouridine-containing 17-mer transcript. Although the bromouridine transcript was cleaved processively starting at the 3Ј-most bromouridine, cleavage of the uridine-containing transcript oc-  GCGTTCGACTCGCCCCCGGGAGATTTTTTTGTTTTTTTTGTC curred mainly 5Ј to the penultimate uridine and to a lower overall extent. The exceptionally weak heteroduplex formed by rU⅐dA base pairs (17) makes it likely that the 3Ј-terminal uridine residue of uridine-containing 17-mer exists mostly in an unpaired state. Because the terminal uridylate is not recognized as a substrate for cleavage, we conclude that the nuclease cleaves only RNA that is heteroduplexed to template DNA. In bromouridine-containing transcripts, stronger dA⅐rBU H-bonding ensures that the 3Ј-terminal residue will be heteroduplexed to DNA. Consequently, in these complexes it is recognized as a cleavage substrate. The possibility that the observed difference is merely due to different rates of cleavage at uridine and bromouridine residues is excluded by the fact that in the degradation of 19-mers having internal U 4 or (BU) 4 sequences followed by two guanosines that stabilize the secondary structure, uridine to bromouridine substitution has the opposite effect. In the case of the uridine-containing 19-mer, the nuclease scans backwards from two well paired 3Ј-terminal guanosines to the neighboring (base paired) U17 residue, initially cutting just 5Ј of this uridine and then effectively cutting at each residue in the uridine stretch (at positions ϩ16, ϩ15, and ϩ14). At the same time, the efficiency of cleavage of bromouridine-containing 19-mer is considerably lower, and propagation in the reverse direction is limited to the closest bromouridine at position ϩ17. This might indicate that the nuclease cuts more efficiently in heteroduplex regions of low stability. Besides rU⅐dA sites, this includes sequences such as (rA) 3 ⅐(dT) 3 .
The Relation of 5Ј Products to 3Ј Products-For most of the ternary complexes we investigated, the 5Ј products correspond to uridine-specific cleavage only, whereas the 3Ј products indicate that multiple cuts occur within 3Ј-terminal sequences that lack uridylates. Several possible explanations for this come to mind. The first involves a single mode of nuclease action, whereby rapid cleavage occurs in sequential short steps. To explain our data by this mechanism requires that there be a means whereby sequential nuclease action on the remaining 5Ј fragment is drastically slowed immediately after the removal of a mono-or dinucleotide that bears 5Ј-UMP. Alternatively, the apparent contradiction between the observed patterns of released 3Ј products and retained 5Ј products might be explained by bidirectional movement of the nuclease along its RNA-DNA hybrid substrate. Initially, the nuclease active site translocates in a 3Ј to 5Ј sense, scanning for the first internal uridine residue. Once a cut has been made 5Ј to uridine, the distal RNA fragment that is still in a hybrid with DNA undergoes further cutting, requiring 5Ј-3Ј translocation of the nuclease active center with a concerted reattachment of the newly formed transcript terminus to the polymerization site of the enzyme.
Our attempts to trap hypothetical 5Ј and 3Ј intermediates by varying the temperature from 0 to 30°C and by sampling the reaction at times as short as 1 s were not successful. To resolve between the possibilities mentioned above, new methods of analysis will be needed.
Oligonucleotides Made at the 3Ј Terminus during Elongation-We have observed that RNA polymerase III carries on multiple cleavage resynthesis steps at the arrest sites and terminator, producing a large amount of short cleavage products, predominantly with uridines at their 5Ј ends (pNpU). The direct correlation of the amount and composition of 3Ј products with the number and the sequences of the arrest sites further supports our conclusion that cleavage events are mostly associated with slow progress of RNA polymerase III through oligouridine tracts. When elongation is impeded, the nuclease activity of RNA Pol III removes the frayed 3Ј end of RNA to restore proper alignment of the transcript 3Ј terminus so that elongation may resume. Analogous to this situation is the one that exists when a misincorporated base is present at the end of a growing transcript. In that case, misalignment of the RNA 3Ј end is caused by noncomplementarity of the RNA and DNA bases that are in opposition. By resecting such mispaired termini, the Pol III nuclease can act as a part of proofreading apparatus.
Similar suggestions for human RNA Polymerase II were made previously by Reines and co-workers (2, 10 -12), who observed that an increase of transcription efficiency through natural pause sites and DNA-bound protein roadblocks was caused by transcription factor IIS activation of multi-round cleavage and resynthesis. In a recent study of the possible editing role of the nuclease activity of RNA Polymerases, Erie et al. (29) showed that the presence of GreA elongation factor reduces the misincorporation of UTP by E. coli RNA polymerase (Ͼ50%) by preferential rapid cleavage of terminally misincorporated transcripts. This and several other recent observations (28,30) that point to possible proofreading by RNA polymerases led to a reconsideration of the traditional viewpoint (31) that transcription is an unedited process.