The RNA Polymerase II Elongation Complex- Factor-Dependent Transcription Elongation Involves Nascent RNA Cleavage

Regulation of transcription elongation is an important mechanism in controlling eukaryotic gene expression. SII is an RNA polymerase II-binding protein that stimulates transcription elongation and also activates nascent transcript cleavage by RNA polymerase II in elongation complexes in vitro (Reines, D. (1992) J. Biol. Chem. 267, 3795-3800). Here we show that SII-dependent in vitro transcription through an arrest site in a human gene is preceded by nascent transcript cleavage. RNA cleavage appeared to be an obligatory step in the SII activation process. Recombinant SII activated cleavage while a truncated derivative lacking polymerase binding activity did not. Cleavage was not restricted to an elongation complex arrested at this particular site, showing that nascent RNA hydrolysis is a general property of RNA polymerase II elongation complexes. These data support a model whereby SII stimulates elongation via a ribonuclease activity of the elongation complex.

Regulation of transcription elongation is an important mechanism in controlling eukaryotic gene expression. SI1 is an RNA polymerase 11-binding protein that stimulates transcription elongation and also activates nascent transcript cleavage by RNA polymerase I1 in elongation complexes in vitro (Reines, D. (1992) J. Biol. Chern. 267,3795-3800). Here we show that SIIdependent in vitro transcription through an arrest site in a human gene is preceded by nascent transcript cleavage. RNA cleavage appeared to be an obligatory step in the SI1 activation process. Recombinant SI1 activated cleavage while a truncated derivative lacking polymerase binding activity did not. Cleavage was not restricted to an elongation complex arrested at this particular site, showing that nascent RNA hydrolysis is a general property of RNA polymerase I1 elongation complexes. These data support a model whereby SI1 stimulates elongation via a ribonuclease activity of the elongation complex.
Expression of many genes in eukaryotes requires transcription of extremely large stretches of DNA by RNA polymerase 11. During this transcription, RNA polymerase I1 encounters signals which block full-length primary-transcript synthesis (reviewed by Spencer and Groudine, 1990; Kerppola and Kane, 1991). Transcription in vitro using partially purified or defined components has resulted in the identification and characterization of RNA polymerase I1 transcription arrest sites (Spencer and Groudine, 1990). In some cases, arrest occurs at a bend in the helical axis of template DNA (Kerppola and Kane, 1990). Examples include sites within the first intron of the human c-myc gene, the first intron of the human histone H3.3 gene, and an early transcription unit of SV40. It has been suggested that at least part of the information that halts transcription resides in a bend-inducing structural element and that the bent template may provide a steric block to translocation of the enzyme (Kerppola and Kane, 1990). I n uitro, the RNA polymerase I1 elongation complex is extremely stable, as might be expected in order for RNA polymerase I1 to transcribe very large genes. However, since transcription can become efficiently blocked at numerous sites both in vivo and in vitro, it is important to understand how an arrested transcription complex can become reacti-* This work was supported by a grant (to D. R.) from the University Research Committee of Emory University, American Cancer Society Grant IRG-182, and National Institutes of Health Grant GM-46331. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be liereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

3361; Fax: 404-727-2738.
1 To whom correspondence should be addressed. Tel.: 404-727-vated for transcription. The transcription of megabase-sized genes and the regulation of growth controlling genes probably relies upon accessory transcription elongation factors to govern the efficiency of primary transcript synthesis.
SI1 is a 35-38 kDa polypeptide that allows RNA polymerase I1 to readthrough a pause in the adenovirus major late transcription unit (Reinberg and Roeder, 1987). This pause is important in regulating the viral infectious cycle Chen-Kiang, 1984, Mok et al., 1984). SI1 also enables RNA polymerase I1 to readthrough a transcription arrest signal (called site Ia) in the first intron of a human histone gene (Reines et al., 1989;SivaRaman et al., 1990). This site may serve to regulate the gene's expression pattern during the cell cycle.' The means by which SI1 activates transcription elongation is unknown.
Two properties of SI1 may provide clues to its mechanism of stimulating elongation. SI1 binds RNA polymerase I1 (Sawadogo et ul., 1980;Horikoshi et ul., 1984;Reinberg and Roeder, 1987;Rappaport et al., 1988). SI1 can also activate a previously unrecognized ribonuclease activity of elongation complex-associated RNA polymerase I1 (Reines, 1992).' In the presence of SII, nascent transcript cleavage shortens the substrate RNA from its 3'-end by a few nucleotides, although more extensive digestion has also been observed.
To determine whether SII's RNA polymerase binding activity and its nuclease activating activity are involved in efficient transcription through intragenic regions, we have examined a discrete population of purified elongation complexes arrested at a well defined site within a human histone gene. We show that transcript cleavage precedes SII-dependent transcript elongation and that an elongation complex containing a shortened RNA is an intermediate in SI1 activation of RNA chain elongation. A truncated form of SI1 that is incapable of binding RNA polymerase I1 cannot activate RNA shortening. These findings support a model in which 15516 SI1 activation of a RNA polymerase 11-associated nuclease is integral to the transcriptional readthrough process. Furthermore, transcript cleavage is not restricted to elongation complexes arrested at this SII-sensitive site in the human histone gene; therefore, nascent RNA cleavage activity is a general property of the RNA polymerase I1 elongation complex.

MATERIALS AND METHODS
Proteins and Reagents-RNA polymerase I1 and SII-free general transcription initiation factors were purified from rat liver as described (Conaway et al., 1987;Reines, 1991a). Yeast inorganic pyrophosphatase was obtained from Sigma. Acetylated bovine serum albumin, T7 RNA polymerase, and EcoRI were supplied by Promega Biotec. RNAguard, NdeI, PuuII, and unlabeled, fast protein liquid chromatography-purified nucleotides were purchased from Pharmacia LKB Biotechnology Inc. [cI-~'P]CTP was purchased from Amersham Corp. Protein concentration was determined with protein assay dye reagent (Bio-Rad) according to the supplier's directions using bovine serum albumin (Sigma, Fraction V) as a standard. DNA templates (pAdTerm-2, Reines et al., 1987;pDNAdML, Conaway and Conaway, 1988) were purified by centrifugation on CsC1-ethidium bromide gradients and linearized by cleavage at their single NdeI site.
I n Vitro Synthesis of SII Protein-An EcoRI fragment containing the region encoding mouse SI1 was removed from pSII-3 (Hirashima et al., 1988) and inserted into pGEM2 such that transcription from the T7 promoter yields sense RNA. The resulting plasmid, pGEM-SII, was linearized with PuuII (for full-length SII) or EcoRV (for ASII) and transcribed in uitro with T7 RNA polymerase. RNA was translated in the presence of [35S]~-methionine (>lo00 Ci/mmol, Du Pont-New England Nuclear) in a wheat germ translation extract HEPES3-NaOH, pH 7.9, 100 mM KCl, 1 mM EDTA, 1 mM dithio-(Promega Biotec). Translation reactions were dialyzed into 20 mM threitol, 20% (v/v) glycerol and chromatographed on phosphocellulose as described for bovine brain SI1 (Reines, 1991a). The amount of in uitro translated SI1 protein was estimated 1) by measuring the amount of methionine that was incorporated into a form precipitable in 10% (w/v) trichloroacetic acid, 2) by using 1000 Ci/mmol as the specific activity of labeling methionine, and 3) with knowledge of the cDNA sequence (Hirashima et al., 1988) of SI1 which assigns 12 methionine residues to full-length SII.
RNA Polymerase II Binding Assays-RNA polymerase I1 (0.5 pg of DEAE-Sephadex A-25 fraction; Reines, 1991a) was mixed with 15 pg of protein A-Sepharose-purified anti-RNA polymerase I1 IgG (8WG16; Thompson et al., 1989) in binding buffer (22 mM Tris-HC1, pH 7.9,3 mM HEPES-NaOH, pH 7.9, 70 mM KCI, 3% (v/v) glycerol, 0.6 mM EDTA, 2 mM dithiothreitol, 225 pg/ml acetylated-bovine serum albumin). Six pl of fixed Staphylococcus aureus (Bethesda Research Laboratories), washed in binding buffer, were added to bring the final volume to 25 pl. Reactions were incubated at 28 "C for 10 min. Precipitates were collected by centrifugation for 2 min in a microcentrifuge and were resuspended in 20 p1 of binding buffer. 35S-Labeled SI1 or 35S-labeled AS11 were added and incubated for 15 min at 28 "C. Soluble and precipitable proteins were separated by centrifugation for 2 min in a microcentrifuge and were subjected to electrophoresis on 15% polyacrylamide gels (Laemmli, 1970) and fluorographed using sodium salicylate as a fluor (Chamberlain, 1979). RNA Polymerase II Elongation Complex Assembly and Isolation-RNA polymerase I1 elongation complexes arrested at a specific site (Ia) within a human histone gene were assembled from partially purified rat liver RNA polymerase I1 and general initiation factors as described (Reines et al., 1989). Site Ia is a transcription arrest site that, in the absence of elongation factor SII, halts 40-50% of RNA polymerase I1 molecules. The template employed was pAdTerm-2 cleaved with NdeI. RNA was pulse labeled with 20 p~ ATP, 20 p~ UTP and -0.6 p M [cI-~'P]CTP (>400 Ci/mmol) in the absence of GTP. This results in synthesis of a 14-nucleotide RNA labeled with [32P]CMP at positions 2,4,6,9,10, and 12 of the RNA chain. Heparin (10 pg/ml) was added along with unlabeled CTP, GTP, ATP, and UTP. This population of template-engaged nucleoprotein complexes are referred to as elongation complexes. One reaction equivalent is 60 pl. Active elongation complexes were precipitated with an anti-RNA monoclonal antibody (D44, Eilat et al., 1982;Reines, 1991b). in reaction buffer (20 mM Tris, 3 mM HEPES, pH 7.9, 62 mM KCl, 2.2% polyvinyl alcohol, 3% (v/v) glycerol, 2 mM dithiothreitol, 0.5 mM EDTA, and 0.3 mg/ml acetylated-bovine serum albumin), were added and the complexes were collected by centrifugation in a microcentrifuge for 2 min. These immunoprecipitated elongation complexes were washed by two additional rounds of centrifugation and resuspension in reaction buffer. This preparation is referred to as washed elongation complexes. Various nucleotides, proteins, and salts were added to the washed complexes as indicated. SI1 was partially purified from bovine brain as described (Reines, 1991a). The purification of rat liver SI1 employed sequential chromatography on phosphocellulose, AcA 34, hydroxylapatite, and T S K 4 resins and will be reported elsewhere? Reactions were stopped with a SDS-containing buffer (0.2 M Tris-HCl, pH 7.5, 25 mM EDTA, 0.3 M NaC1, 2% (w/v) SDS), and RNA was isolated as described (Reines, 1992).

Recombinant SII Synthesized in Vitro
Stimulates the Ribonuclease Function of a n RNA Polymerase II Elongation Complex-To confirm that the addition of SI1 is sufficient to activate the ribonuclease activity of RNA polymerase 11, and to test whether RNA polymerase 11-binding by SI1 is required for this activation, SI1 and a truncated form of SI1 (ASII) were prepared from a mouse cDNA clone (Hirashima et al., 1988) by i n vitro transcription and translation in a wheat germ extract. The truncated derivative lacked its carboxylterminal 107 amino acids, in which residues important for RNA polymerase I1 binding reside (Horikoshi et al., 1990;Agarwal et al., 1991).
[35S]SII and [35S]ASII were separated from the bulk of the wheat germ translation extract proteins by chromatography on phosphocellulose. These i n vitro synthesized proteins were analyzed for RNA polymerase I1 binding activity by testing their ability to coimmunoprecipitate with RNA polymerase 11. A monoclonal antibody against the carboxyl-terminal domain (CTD) of the large subunit of RNA polymerase I1 (8WG16, Thompson et al., 1989) was used to isolate CTD-containing enzyme from a preparation of partially purified rat liver RNA polymerase 11. The immunoprecipitate containing 8WG16 and RNA polymerase I1 was resuspended with "S-labeled SI1 or AS11 (Fig. 1A). Efficient precipitation of the full-length protein required both the anti-RNA polymerase I1 antibody (Fig. lA, lanes 3-6) and RNA polymerase II.4 AS11 was not significantly precipitable above background when present at the same concentration at which SI1 was efficiently precipitated (Fig. lA, lanes 7-10). The relatively high background of precipitable SI1 was reproducible, detected when protein A-Sepharose was the immunosorbent, and could not be reduced after further washing of the imm~noprecipitate.~ These data independently confirm previous findings in which the region of SI1 important for RNA polymerase I1 binding was identified (Horikoshi et al., 1990;Agarwal et al., 1991). Furthermore, binding of 8WG16 to the CTD of RNA polymerase did not prevent the subsequent binding of SI1 to the enzyme.
Recombinant SI1 proteins were also tested for their ability to activate SII-dependent nascent transcript cleavage ( FIG. 1. Activity of in vitro synthesized, recombinant mouse SI1 and truncated SI1 (ASII). Panel A, RNA polymerase I1 binding assay. "S-Labeled, phosphocellulose-purified SI1 and AS11 were resuspended a t a concentration of 4 p~ with 8WG16-immunoprecipitated rat liver RNA polymerase I1 a t 28 "C for 15 min. Precipitable ( P ) and soluble ( S ) fractions were separated and run on a SDSpolyacrylamide gel. In control samples (lanes 5, 6, 9, and IO) anti-RNA polymerase I1 IgG was omitted. The amount of translation product put into each reaction ( i n ) is displayed in lanes I and 2 for SI1 and ASH, respectively. Panel B, nascent transcript cleavage assay. Two pg of partially purified bovine brain SI1 (Br, lane I ) , in vitro synthesized mouse SI1 (lanes 2 and 3), in vitro synthesized mouse AS11 (lanes 4 and 5 ) , or buffer (-, lane 6 ) were mixed with washed elongation complexes and MgC12 and incubated for 30 min a t 28 "C. The amount of in vitro synthesized SI1 or ASH added to each reaction is indicated by I x , 2 x , 5 x , and I O X where X is approximately 50 amol. In the absence of nucleotides, the size of a cleaved nascent RNA is inversely proportional to the concentration of SII.4 Hence, a shorter product was obtained in the presence of bovine brain SI1 (lane I) compared to that seen with recombinant mouse SI1 (lanes 2 and 3) since more SI1 activity was present in the former than the latter. Runoff RNA (RO) and RNA resulting from transcription arrest at sire Ia ( l a ) are indicated. Marker RNAs of 540, 420, 380 and 260 nucleotides (top to bottom) are indicated with arrowheads.
active elongation complexes can be cleaved after exposure to partially purified bovine brain SI1 (Reines, 1992). Only RNA in an elongation complex and not free RNA was cleaved. The nuclease activity was inhibited by a-amanitin (1 pg/ml). The simplest interpretation of those findings was that RNA polymerase I1 itself contains this catalytic function.
Elongation complexes containing ["'PIRNA were assembled by transcription in vitro of a region of the human H3.3 histone gene that arrests transcription with high efficiency (Reines et al., 1987(Reines et al., , 1989. Arrested elongation complexes were isolated and washed free of nucleotides by immunoprecipitation using an anti-RNA monoclonal antibody (Reines, 1991a(Reines, , 1991b. Only full-length recombinant SI1 (Fig. lB,  lanes 2 and 3 ) was capable of activating nascent transcript cleavage. Truncated SI1 (Fig. 1B, lanes 4 and 5 ) was ineffective in stimulating transcript shortening, even when present at a molar concentration 10-fold higher than that of the fulllength protein. Wheat germ translation extracts alone, or extracts programmed with control RNA, were also ineffective in activating the nuclease activity." These data demonstrated that SI1 was responsible for activating the nuclease activity of the RNA polymerase I1 elongation complex and that it need not be synthesized in mammalian cells to acquire its function as an activator of RNA cleavage. It also provides evidence that SII's polymerase binding activity is important for the activation of RNA cleavage.
Transcript Shortening Precedes SII-mediated Elongation Through Site la-RNA cleavage by the elongation complex is rapid; all precursor RNA is shortened in less than 1 min a t 28 "C (Reines, 1992). In the absence of nucleotides, the halflife of a shortened RNA can be as long as 1 h (Reines, 1992). It was important then, to determine if the nuclease activity was operative under conditions where SI1 stimulates transcriptional readthrough, i.e. in the presence of SI1 and nucleotides. Since in this experiment we did not wish to deplete the complexes of nucleotides, they were not immunoprecipitated with anti-RNA antibody. A time course of SII-dependent elongation through this region showed that shortened transcripts were detected after 20 s (Fig. 2). The half-life of the smallest detectable intermediate was less than 25 s. Compared to the more extensively cleaved RNAs seen when no nucleotides were present (Reines, 1992 and Fig. 5, B and C), the nascent RNA in this experiment underwent only limited shortening. Hence, the size and half-life of cleavage products derived from a single RNA differed in the presence and absence of nucleotides (Figs. 2 and 3, this work and Reines, 1992). The brief half-life of cleaved RNAs seen here most likely resulted from their rapid re-extension in the presence of nucleotides, implying that transcript shortening preceded SI1 activation of elongation. Furthermore, these data prove that transcript cleavage is neither an artifact of nucleotide depletion nor of elongation complex immunoprecipitation.
The accumulation of runoff transcripts after 15 min suggested that the cleaved RNAs served as precursor to fulllength RNAs (Fig. 2). To better resolve chain cleavage from chain elongation, we sought an approach that would allow us to more readily detect the products of RNA chain extension through site Ia. Thus, we studied SII-activated readthrough in the presence of a limited set of nucleotides. Immunoprecipitated elongation complexes were washed free of unincorporated nucleotides to reduce the NTP concentrations to less than 50 nM each (Reines, 1992). Washed complexes were then supplied with SI1 and 800 p~ each of GTP, CTP, and UTP. The 3' termini of RNAs ending a t site Ia map to 4 consecutive T residues on the non-transcribed DNA strand (Reines et al., 1987) as indicated in Fig. 3A. The template sequence down- Six reaction equivalents of washed elongation complexes were incubated with bovine brain SI1 (12 pg) and 800 p~ each of ATP, GTP, CTP, and UTP. After 20, 45,65, 95, and 900 s a t 28 "C, one reaction equivalent was removed, and RNA was isolated for electrophoresis. Two shortened intermediates are indicated by an asterisk. The identity of the other RNAs are as indicated in the legend to Fig. 1B. stream from site Ia is such that in the presence of SII, UTP, GTP, and CTP transcript Ia (-205 nucleotides) can be extended by only 13-16 nucleotides (Fig. 3A); thus, elongation through site Ia should be readily observed. The kinetics of elongation under these conditions clearly demonstrated that nascent RNA was first cleaved and only subsequently elongated past site Ia (Fig. 3 B ) . The prolonged half-life of these cleaved RNAs was probably due to a requirement for the insertion of an A residue into the shortened RNA between the cleavage product terminus (see legend to Fig. 3A) and site Ia (Fig. 3A). Apparently, enough residual ATP was present in the reaction to support this insertion. Alternatively, this elongation could represent the incorporation of a non-cognate nucleotide into RNA. In any case, the simplest interpretation is that transcript cleavage preceded and could be resolved from RNA extension during SII-activated transcript elongation.

S ' -U U U U U A A A A G A G G G A C G U~U U C C C U U U U U U G G~G -3 ' la
These data suggested that RNA chains elongated by RNA polymerase I1 in an SII-dependent manner must pass through an intermediate that was shorter than the original transcript. This was demonstrated by an experiment using the chainterminating nucleotide 3"O-methyl-GTP. Washed elongation complexes were permitted to elongate their RNA chains in the presence of SII, 3'-O-methyl-GTP, UTP, and CTP (Fig.   3C, lane 3 ) . If SI1 allowed the direct extension of transcript Ia without going through a shortened intermediate, transcript Ia would be extended by 12-15 nucleotides before the chain-"-1 2 3

FIG. 3. Transcript shortening precedes SII-mediated transcription elongation.
Panel A, RNA sequence around site Ia within the first intron of the human histone gene. The position of the 3'ends formed when RNA polymerase I1 stops transcription are underlined and are collectively referred to as site Ia (Reines et al., 1987). The first A residue downstream of site Ia is boxed (0). The tentative site of cleavage which generates the major product (Reines, 1992) seen when washed elongation complexes were treated with SI1 (Fig.  3C, lane 1 ) is indicated by an arrow.4 Panel B, transcription through site Ia after the addition of SII, GTP, UTP, and CTP. Washed elongation complexes were incubated with bovine brain SI1 and 800 p~ each of UTP, CTP, and GTP. After 0, 0.5, 1,4,8, and 15 min at 28 "C, aliquots were withdrawn and labeled RNA was isolated for electrophoresis. The RNA that extends to the first A residue downstream from site Ia is indicated (0). The migration position of other RNAs are as indicated in the legend to Fig. 1B. Panel C, transcription through site Ia in the presence of 3'-O-methyl-GTP. Washed elongation complexes were incubated with 800 p~ each of UTP and CTP, bovine brain SII, and either 800 p~ GTP (lane 2) or 830 pM 3'-0methyl-GTP (lane 3), for 30 min at 28 "C. RNA was isolated and analyzed by electrophoresis and autoradiography. The migration position of transcript Ia, run in an adjacent lane, is shown ( l a ) . Other RNAs are identified as described in the legend to Fig. 3B. terminating analog was inserted (Fig. 3A). If, however, chain shortening was an obligatory intermediate in the SI1 stimulation reaction, re-extension of the cleavage product would be prevented since after cleavage the chain-terminating G-residue would be inserted before the transcript could be significantly elongated. This would result in a limited extension product shorter than transcript Ia. Indeed, after 30 min at 28 "C only shortened transcripts were observed (Fig. 3C). This was consistent with the idea that incorporation of 3'-0methyl-GMP into the transcript prevented the shortened RNA from serving as a precursor for chain elongation through site Ia. We emphasize that this reaction contained elongation factor SI1 and all nucleotides necessary for direct extension of transcript la. Thus, it appears that RNA chain shortening is a prerequisite to transcription through site Ia since preventing the extension of the cleaved RNA did not reveal an alternative means of elongating RNA chains, for example via the direct extension of Ia-RNA.
Transcription Elongation SII-dependent Chain Shortening Is Not Pyrophosphorolysis-Pyrophosphorolysis, reversal of the nucleic acid biosynthetic reaction, can be carried out by DNA polymerases (Deutscher and Kornberg, 1969;Atkinson et al., 1969) and bacterial and bacteriophage RNA polymerases (Maitra and Hunvitz, 1967;Krakow and Fronk, 1969;Kassavetis et al., 1986;Arndt and Chamberlin, 1990) in the presence of millimolar levels of pyrophosphate (PPJ. Since RNA shortening by RNA polymerase I1 was dependent upon added SII, it seemed unlikely that cleavage resulted from pyrophosphorolysis (Reines, 1992). Nevertheless, it was important to rule out the possibility that SI1 acted as a PPi generator and to show that SII-dependent RNA shortening was distinct from pyrophosphorolysis. After 30 min in the presence of PPi and MgC12, RNA polymerase I1 elongation complexes shortened a small but significant fraction of RNA chains (Fig. 4 A , lane  3 ) . Since the ?' P label was present at the 5'-end of these RNA chains, nucleotide residues were removed from their 3"ends. This was an SII-independent, PPi-dependent process that yielded a product distinct from that seen in the SII-dependent, PPi-independent cleavage of nascent RNA chains (Fig. 4A,   compare lanes 1 and 3 ) . The simplest interpretation was that this reaction represented pyrophosphorolysis. Previous work has shown that no significant RNA cleavage took place when washed complexes were incubated with MgC12, in the absence of both PP, and SI1 (Reines, 1992). Washed elongation complexes were incubated for 30 min a t 28 "C with 7 mM MgCI2, and bovine brain SI1 (2 pg), or 1.8 mM sodium pyrophosphate as i'ldicated above the figure. RNA was isolated and analyzed by electrophoresis and autoradiography. Panel B, effect of pyrophosphatase on transcript cleavage. Washed elongation complexes were mixed with 7 mM MgC12 and either bovine brain SI1 (2 pg, lanes 1 and 2) or 1.5 mM sodium pyrophosphate (lanes 3 and 4 ) in the presence (+) or absence (-) of 10 units of yeast inorganic pyrophosphatase. (One unit of inorganic pyrophosphatase liberates 1.0 pmol of inorganic orthophosphate/min.) The reactions were then incubated a t 28 "C for 30 min before RNA was isolated for electrophoresis. Three pyrophosphatase-sensitive RNAs are indicated by asterisks.
To demonstrate further that we were observing two distinct processes, pyrophosphorolysis and SII-dependent transcript cleavage, we inhibited pyrophosphorolysis by the addition of pyrophosphatase. Pyrophosphatase hydrolyzes inorganic PPi to orthophosphate and thus, prevents pyrophosphorolysis (Cooperman, 1982;Tabor and Richardson, 1990). Washed elongation complexes were allowed to undergo SII-dependent or PPi-dependent RNA shortening in the presence of pyrophosphatase (Fig. 4B). Whereas the extent of PPi-dependent transcript shortening was dramatically reduced in the presence of inorganic pyrophosphatase (Fig. 4B, lanes 3 and 41, SII-dependent RNA cleavage was unaffected by the enzyme (lanes I and 2). We take this as evidence that washed RNA polymerase I1 elongation complexes can undergo pyrophosphorolysis and that it is chemically distinct from the more efficient and extensive RNA cleavage activated by SII.
SII Can Activate the RNA Cleavage Reaction in Other RNA Polymerase I I Elongation Complexes-To this point we have observed nascent RNA cleavage in only two elongation complexes, the prominent one arrested a t site Ia and a minor, less well-characterized site downstream (+325) in this template (pAdTerm-2, Reines, 1992 and Figs. 2 and 3). We tested the ability of an additional elongation complex to cleave its nascent RNA in response to SII. Washed elongation complexes containing RNA polymerase arrested at site Ia were allowed to extend their transcripts in the presence of added SII, UTP, GTP, and CTP as described above (Fig. 3). The majority of polymerase molecules became arrested a t a downstream location where the first A residue was required. Small amounts of readthrough were seen which result in the extension of this RNA by two nucleotides to the next required A residue. These complexes were washed by precipitation and resuspension with anti-RNA antibody and challenged with rat liver SI1 for varying periods of time (Fig. 5A). By 10 min, chain shortening was apparent (Fig. 5A, lane 6). Thus, an SII-activated RNAcleavage activity seems to be a general trait of an RNA polymerase I1 elongation complex. (Rat liver SI1 was used here since relatively large amounts of highly purified protein can be obtained in a concentrated form during the isolation of rat liver initiation factors. Thus far, purified bovine, human, and rat SI1 appear functionally indi~tinguishable.~) In some experiments run-off RNAs appeared shortened. This raised the possibility that some RNA polymerase molecules remained in functional elongation complexes for a significant time after they had reached the end of a duplex DNA template. If so, it would provide an opportunity to study RNA cleavage in more elongation complexes arrested a t a site distinct from Ia. Like the complexes shown in Fig. 5A, these elongation complexes also represent RNA polymerases that were arrested for a reason other than their dependence upon SI1 for transcription elongation. Labeled runoff RNA was synthesized in vitro from two different linear templates (Fig.   5, B and C) and subjected to immunoprecipitation with anti-RNA monoclonal antibody. The precipitate was washed free of nucleotides and incubated with rat liver SI1 for varying times. Indeed "runoff" RNAs became progressively shorter with increasing incubation times. A small fraction of the RNAs were shortened by as many as 100 nucleotides judging by their mobility versus marker RNAs of known size (Fig.  5B). Since the addition of nucleotides resulted in efficient transcript extension (Fig. 5, B and C, chase lanes), cleaved RNA remained in functional elongation complexes. Only a portion of the runoff RNA, which was resolved as a closely spaced doublet (Fig. 5C), was capable of being shortened, perhaps indicating that only a fraction of the RNA remained associated with the template and RNA polymerase I1 in an  5 (lane 4 ) , 2 (lane5), or 10 min (lane 6 ) a t 28 "C with approximately 5 X absorbance units (280 nm) of rat liver SI1 (TSK-phenyl 5-P W fraction). Samples were prepared for electrophoresis and autoradiography as described under "Materials and Methods." Panels B and C, runoff transcript cleavage. The plasmids pAdTerm-2 ( B ) or pDNAdML (C) were linearized with NdeI and transcribed in uitro with rat liver RNA polymerase I1 and general initiation factors. Reactions containing runoff RNAs (530 nt and 250 nt) pulse-labeled with [a-"'P]CMP were subjected to immunoprecipitation with anti-RNA monoclonal antibody. Precipitates were washed and incubated with approximately 5 X absorbance units (280 nm) of rat liver SI1 (TSK-phenyl 5-PW fraction) and incubated at 28 "C. After the indicated number of minutes, aliquots were withdrawn and RNA was isolated for electrophoresis. An additional sample (chase) from each reaction was withdrawn (at 45 min, panel B, or a t 60 min, panel C) and adjusted to 800 PM in all four nucleotides, incubated for an additional 15 min a t 28 "C ,and RNA was prepared for electrophoresis. The migration position of marker RNAs of 540, 420, and 380 nucleotides (from top to bottom) are indicated by arrowheads in panel R. nt. nucleotides. elongation complex. However, a significant number of RNA polymerase I1 molecules that synthesized runoff RNA did not appear to dissociate from these DNA termini. Collectively, the data of Fig. 5 show that many different RNA polymerase I1 elongation complexes possess the ability to cleave their nascent RNA chains in response to elongation factor SII. Therefore, elongation complexes other than those stopped a t either SII-dependent transcription arrest sites or sites coincident with a bend in the template can perform this function.

DISCUSSION
We have examined the relationship between SII-activated nascent transcript cleavage and transcription elongation by RNA polymerase I1 elongation complexes. The most important observation presented here is that nascent RNA cleavage precedes, and appears to be an obligatory intermediate in, SII-mediated transcriptional readthrough of a specific transcription arrest site. The association of nascent RNA cleavage with SII-regulated transcription elongation strongly suggests that the nuclease is causally involved in the readthrough process. The available data, including a-amanitin experiments, suggest that RNA polymerase I1 harbors the catalytic residues involved in the nuclease activity. This is not surprising since DNA polymerases, reverse transcriptases, RNA polymerase I (Huet et al., 1976), and bacterial RNA polymerase (Surratt et al., 1991) possess intrinsic nuclease activities. It has recently been reported that amino acids in the catalytic center of barnase, a bacterial ribonuclease, are shared with two other bacterial ribonucleases and the second largest subunits of Drosophila and yeast RNA polymerase I1 (Shirai and Go, 1991). Alternatively, SI1 may contain all or part of the catalytic site. If so it must interact with RNA polymerase I1 in a manner that is inhibited, perhaps allosterically, by aamanitin.
The transcription arrest site studied here lies in a bent region of DNA (Kerppola and Kane, 1990). Evidence suggests that such a DNA helix represents a population in which bent isomers exist in equilibrium with relatively unbent or linear isomers (Koo et al., 1986). A model of SII-assisted readthrough at this site which accommodates this feature of the template has been put forth (Reines, 1992). It states that RNA polymerase I1 arrested at the site Ia bend prevents isomerization of bent DNA to an elongation-permissive, unbent form. SII, acting as an allosteric ligand, binds the elongation complex and activates a latent nuclease in RNA polymerase 11. As a result of shortening the 3'-end of the growing RNA chain, polymerase moves relative to the template allowing renewed interconversion between bent and unbent DNA, followed by the resumption of transcript elongation. The findings presented here fulfill a prediction of this model: namely, that nascent RNA cleavage precedes SII-dependent transcript elongation. Our experiments with the chain terminator, 3'-0methyl-GTP show that even in the presence of SII, and all nucleotides necessary for direct transcription from site Ia, elongation does not take place (Fig. 3C). Instead, we observe a new transcription intermediate: an RNA polymerase I1 elongation complex that bears a cleaved nascent transcript. Although it is possible that RNA cleavage may not accompany readthrough at all SII-dependent arrest sites, it is likely that the short half-life of the cleavage product seen in the presence of nucleotides probably precluded detection of this event in previous studies of SII-mediated readthrough (Reinberg and Roeder, 1987;Rappaport et al., 1987;SivaRaman et al., 1990;Bengal et al., 1991). It will be interesting to learn if other elongation factors also activate nascent transcript cleavage.
Elongation complexes located a t template positions other than site Ia were also responsive to SII-activated RNA cleavage. It should be noted that the relative propensity for nascent transcript cleavage differs markedly between complexes. In washed elongation complexes, cleaved RNAs derived from transcript Ia were observed seconds after exposure to SI1 (Reines, 1992) whereas other complexes cleaved their RNAs slowly requiring many minutes or hours to obtain significant cleavage (Fig. 5 ) . These findings show that a DNA bend is not required for SI1 activation and suggests that in vivo SI1 could stimulate transcription past other kinds of impediments to elongation. These could include nucleosomes or other DNA-bound proteins. If RNA polymerase "retreats" from a blockage site through RNA shortening, it may be able to Transcription Elongation sample repeatedly regions of DNA which are only readthrough permissive part of the time, thereby increasing the probability of efficient transcript elongation. This mechanism may explain how the extremely stable RNA polymerase I1 elongation complex can be assisted in finishing gene transcription through numerous different types of barriers to transcript elongation.
T o our knowledge this is the first demonstration of pyrophosphorolysis catalyzed by a specific eukaryotic transcription complex. Furthermore, SII-activated RNA cleavage by the complex is distinct from pyrophosphorolysis in that the former does not require PPi addition and is insensitive to pyrophosphatase. Thus, SI1 does not appear to activate the cleavage reaction by serving as a source of PPI for pyrophosphorolysis. Accordingly, we think that SII-activated RNA cleavage is hydrolytic. The process is initiated presumably when SI1 associates with the arrested elongation complex. This is consistent with our finding that SI1 defective in polymerase binding is inactive in nuclease activation.
Recent cloning of a yeast protein (DSTa) with recA-like, DNA strand transfer activity has shown that it is highly homologous, if not identical, to elongation factor SI1 (Clark et al., 1991;Kipling and Kearsey, 1991). It was suggested that the strand transfer function of yeast DSTa/SII may facilitate transcription elongation by promoting displacement of a RNA-DNA hybrid within the transcription bubble and favoring renaturation of the DNA duplex (Kipling and Kearsey, 1991). Perhaps by doing so, SI1 also potentiates nascent RNA cleavage by making the 3'-end of the RNA accessible to the nuclease domain of RNA polymerase I1 where it can be hydrolyzed.