Amanitin greatly reduces the rate of transcription by RNA polymerase II ternary complexes but fails to inhibit some transcript cleavage modes.

The toxin alpha-amanitin is frequently employed to completely block RNA synthesis by RNA polymerase II. However, we find that polymerase II ternary transcription complexes stalled by the absence of NTPs resume RNA synthesis when NTPs and amanitin are added. Chain elongation with amanitin can continue for hours at approximately 1% of the normal rate. Amanitin also greatly slows pyrophosphorolysis by elongation-competent complexes. Complexes which are arrested (that is, which have paused in transcription for long periods in the presence of excess NTPs) are essentially incapable of resuming transcription in the presence of alpha-amanitin. Complexes traversing sequences that can provoke arrest are much more likely to stop transcription in the presence of the toxin. The substitution of IMP for GMP at the 3' end of the nascent RNA greatly increases the sensitivity of stalled transcription complexes to amanitin. Neither arrested nor stalled complexes display detectable SII-mediated transcript cleavage following amanitin treatment. However, arrested complexes possess a low level, intrinsic transcript cleavage activity which is completely amanitin-resistant; furthermore, pyrophosphorolytic transcript cleavage in arrested complexes is not affected by amanitin.

The toxin ␣-amanitin is frequently employed to completely block RNA synthesis by RNA polymerase II. However, we find that polymerase II ternary transcription complexes stalled by the absence of NTPs resume RNA synthesis when NTPs and amanitin are added. Chain elongation with amanitin can continue for hours at approximately 1% of the normal rate. Amanitin also greatly slows pyrophosphorolysis by elongation-competent complexes. Complexes which are arrested (that is, which have paused in transcription for long periods in the presence of excess NTPs) are essentially incapable of resuming transcription in the presence of ␣-amanitin. Complexes traversing sequences that can provoke arrest are much more likely to stop transcription in the presence of the toxin. The substitution of IMP for GMP at the 3 end of the nascent RNA greatly increases the sensitivity of stalled transcription complexes to amanitin. Neither arrested nor stalled complexes display detectable SII-mediated transcript cleavage following amanitin treatment. However, arrested complexes possess a low level, intrinsic transcript cleavage activity which is completely amanitin-resistant; furthermore, pyrophosphorolytic transcript cleavage in arrested complexes is not affected by amanitin.
The mushroom toxin ␣-amanitin, a bicyclic octapeptide, has long been used as a specific inhibitor of RNA polymerase II (1)(2)(3). Calf thymus polymerase II has been shown to bind ␣-amanitin very tightly with a stoichiometry of 1:1, a K d of 10 Ϫ9 M and a complex-dissociation half-time of about 100 h at 0°C (4). Using genetic and biochemical techniques, the amanitin binding site has been localized to the largest subunit of RNA polymerase II (5,6). While the mechanism of amanitin's action has not been demonstrated in detail, it is known that amanitinblocked transcription complexes can resume RNA synthesis after irradiation with 314-nm light, which selectively destroys the toxin (7). Thus, it seems unlikely that amanitin acts by permanently disabling part of the polymerase, for example by cleaving one of the subunits. It has also been shown that the toxin does not change affinity of the polymerase for nucleotides (4).
Both our laboratory (8,9) and others (10) had observed that promoter-initiated RNA polymerase II ternary elongation complexes can form one or more phosphodiester bonds after amanitin treatment. The combination of these results and the recent finding that the RNA polymerase II ternary complex can catalyze phosphodiester bond cleavage as well as bond formation (11,12) prompted us to perform a detailed reinvestigation of the effects of amanitin on RNA polymerase II elongation complexes. We report here that RNA polymerase II ternary complexes are generally able to continue RNA synthesis in the presence of ␣-amanitin, albeit at greatly reduced rates. Interestingly, both intrinsic cleavage activity and pyrophosphorolytic cleavage are completely amanitin resistant in arrested complexes. Given the possibility that arrest may result from a retreat of the active site of RNA polymerase away from the 3Ј end of the nascent RNA (13), these observations suggest that ␣-amanitin inhibits RNA polymerase II by disrupting the interaction of the enzyme with the 3Ј end of the nascent transcript. Our findings also lend further support to a model of transcriptional arrest in which an equilibrium exists between catalytically active and inactive states.

MATERIALS AND METHODS
Ribonucleoside triphosphates were obtained from Pharmacia Biotech Inc., except for ITP which was purchased from Sigma. We used ultrapure (fast protein liquid chromatography-purified) NTPs for transcription reactions with preinitiation complexes and standard purity NTPs for chase reactions. Labeled ribonucleotides, either [␣-32 P]CTP or [␣-32 P]UTP at 800 Ci/mmol, were purchased from DuPont NEN, Bio-Gel A1.5 m was acquired from Bio-Rad, and ␣-amanitin was purchased either from Boehringer Mannheim or Sigma.
Plasmids-All plasmids used in this study were based on pML5A, which contains the adenovirus 2 major late promoter cloned into pUC18. Plasmids pML5A (14), pML5-4NR (15), and pML20-U158 and pML20-U160 (16) have been described in detail. The pML20-U158 plasmid was referred to as pML20-G155 in Izban and Luse (16); the construction of the pML20 precursor for pML20-U158/U160 was described in Izban and Luse (11). We constructed pML5-MUT3 from pML5A by substituting a synthesized fragment having the sequence 5Ј-GATCCTTTTTTCTCCATTTTA (nontemplate strand) for the 30-nt 1 BamHI-HindIII fragment which begins at ϩ39 downstream of transcription start. The pML16 series plasmids were all built from a common precursor, pML16LNK, which was derived from pML5A by replacing a BssHII-BamHI fragment, spanning from Ϫ13 to ϩ38 relative to the major late promoter, with a synthesized oligonucleotide. The synthesized fragment bore the original sequence between Ϫ13 and ϩ15 but changed the remaining nontemplate strand sequence from 5Ј-GCT-GTCTGCGTGGGCCTGCTAAG to 5Ј-CCTTTCCCGGGCGAGCTCGG-GCCCTTG. The new sequence contains unique XmaI and ApaI sites. The pML16220 template was assembled from pML16LNK by replacing the XmaI-ApaI segment with a 228-nt XmaI-ApaI fragment containing the U-free cassette from pGR220 (17) (a gift from C. Kane). Thus, the nontemplate strand of pML16220 has no G residues from ϩ1 to ϩ23 and no U residues for the next 135 bases. The pML16C27 and pML16T27 plasmids were built by inserting modified XmaI/ApaI fragments from pGR220 into pML16LNK. These fragments were generated using the polymerase chain reaction. The 3Ј-end primer, 5Ј-GGGAA-CAAAAGCTGGGTACCGGGCCC, overlapped the ApaI site (underlined) and extended 25 nt downstream into the parent vector. The 5Ј-end primer, 5Ј-GGATCCCCCGGG(C/T)AGAAAAAGCAAACCG, was degenerate at the designated position. It overlapped the XmaI site (underlined) and extended 7 nt upstream into the parent vector. All pML16 series templates were sequenced for verification.
Elongation Factors-Human recombinant SII (rSII) purified as described previously (18) was either generously furnished by Robert Landick (Washington University, St. Louis) or else made within our laboratory using the pET11d-RAP38 expression vector kindly supplied by Zachary Burton (Michigan State University, East Lansing).
Assembly and Purification of Stalled and Arrested Elongation Complexes-The complete procedure has already been described (11,15). To form preinitiation complexes, DNA templates were incubated in HeLa cell nuclear extract, after which a gel filtration step was performed to partially purify the complexes and separate out residual NTPs. For most experiments, the initial transcription was done in the absence of GTP, since the nontemplate strand in all of our plasmids lacks G residues over at least the initial 15 bases. The initiating and labeling nucleotides are given in each figure legend. For CTP labeling, transcription reactions contained either 100 M ATP or 1-2 mM ApC (dinucleotide-primed reactions also contained 10 M dATP), along with 10 M UTP and 0.5 or 1 M [␣-32 P]CTP; after 5 min at 25°C, unlabeled CTP was added to give a final CTP concentration of 10 M and incubation was continued for another 5 min. For UTP labeling either 120 M ATP or 2 mM CpA plus 10 M dATP was added along with 10 M CTP and 0.5 or 1 M [␣-32 P]UTP; after 5 min at 25°C, unlabeled UTP was added to 10 M and incubation was continued for another 5 min. Complexes stalled at ϩ36 on template pML5-MUT3 were generated using 2 mM CpA, 10 M dATP, 1.0 M [␣-32 P]UTP and 10 M CTP and GTP, followed by 10 M UTP chase. The initial, labeled complexes were purified by a procedure we have called Sarkosyl rinsing (see Izban and Luse (15) for a complete description). This involves the addition of 1% Sarkosyl followed by gel filtration on Bio-Gel A1.5m in 30 mM Tris-HCl, pH 7.9, 10 mM ␤-glycerophosphate, 62.5 mM KCl, 0.5 mM EDTA, and 1 mM dithiothreitol. Sarkosyl-rinsed complexes lack free transcription factors and free NTPs.
Complexes stalled at ϩ155 on templates pML20-U158 and pML20-U160 were produced exactly as described elsewhere (16). Complexes arrested at ϩ194 on pML5-4NR were assembled somewhat differently in each of the three experiments where they were used. The arrested complexes shown in Figs. 3 and 7A were generated as described previously (16) except as noted in the figure legends. To generate uniformly labeled complexes for Fig. 7B, stalled complexes C15 and U18 were assembled as described previously (11) except that initiation was performed with 0.25 M each of ␣-32 P-labeled and nonlabeled CTP and 120 M ATP in place of ApC and dATP. Uniformly labeled arrested complexes were then produced exactly as described previously (16).
Chase Reactions-Elongations in excess NTPs were all performed at 37°C for the times specified in the figures. The reactions included MgCl 2 at 8 mM and the four nucleoside triphosphates at 1 mM unless otherwise noted. When a time course was run, all aliquots were withdrawn from a common pool. Chase reactions in the presence of elongation factor SII were set up with that factor at 1.5 g/ml. Chases in the presence of ␣-amanitin (at 1 g/ml) were preceded by incubating the complexes with ␣-amanitin at 37°C for at least 3 min before adding other reagents. RNAs were purified as described elsewhere (15) except that Proteinase K treatment was for 0.6 -1.25 h at room temperature and RNAs were boiled for 2.5 min before loading. Samples were resolved on denaturing polyacrylamide gels consisting of 7% (19:1), 10% (29:1), 15% (29:1), 20% (19:1), or 28% (25:3) acrylamide/bisacrylamide and visualized either by autoradiography or by use of a PhosphorImager (Molecular Dynamics) as described previously (15,16). Exact length markers for Fig. 1 were generated with sets of NTP-limiting transcription reactions (not shown) synthesized on the same template (see Izban et al. (19)). The reader will note, in Fig. 1A, that RNAs synthesized on pML16220 do not comigrate with RNAs made on pML16C27 or pML16T27. This difference was completely reproducible and presumably results from the single base change at ϩ27 on the nontemplate strand. All three of these templates were carefully sequenced to confirm that no changes occurred in creating pML16C/T27 from pML16220 except at position ϩ27.
Transcript Cleavage Reactions-Cleavage reactions with SII were done essentially as described previously (11,16), using rSII at the concentrations and for the incubation times indicated in the figure legends. Pyrophosphate-facilitated cleavage reactions, in the presence or absence of ␣-amanitin, were all performed with 2 mM sodium pyrophosphate at 37°C for the times indicated. Incubations with Mg 2ϩ alone contained 8 mM MgCl 2 and continued at 37°C for the times indicated.

RESULTS
Experimental System-We study transcript elongation by RNA polymerase II in vitro with partially purified ternary transcription complexes. The initial transcribed regions of our templates are designed so that in the presence of a subset of the NTPs, newly initiated RNA polymerases will pause between 15 and 25 nt downstream of transcription start. These complexes are sufficiently stable to allow purification by transient exposure to the detergent Sarkosyl during gel filtration, a procedure we call Sarkosyl rinsing (15). The large majority of the Sarkosyl-rinsed complexes will resume transcription when NTPs are added. These complexes lack the TFIIF and SII elongation factors. However, when these factors and NTPs are added in saturating amounts, the purified complexes will elongate their nascent RNAs at about 1500 nt/min at 37°C (20), which is essentially the chain elongation rate observed in the cell nucleus (21). We use the term "stalled" for complexes which have stopped transcription because of the absence of NTPs but which remain competent to resume RNA synthesis rapidly when NTPs are restored. Stalled complexes are named according to the length of the nascent RNA and the last base incorporated; thus, a complex with a 23-nt RNA ending in C would be a C23 complex.
␣-Amanitin Slows but Does Not Absolutely Block Transcript Elongation-In the course of recent experiments we observed an example of amanitin-resistant chain elongation by RNA polymerase II which was much more extensive than those reported previously (8 -10). The reaction which sparked our interest was performed on the pML16220 template, which has no T residues on the nontemplate strand from ϩ21 through ϩ155 (see Fig. 1A). Transcriptions performed in the absence of GTP gave complexes paused at ϩ23 (C23 complexes). These complexes were Sarkosyl-rinsed and a portion were treated with amanitin at 1 g/ml for 3 min at 37°C. (This preincubation protocol was used for all amanitin-containing reactions in this study.) We found that amanitin-treated C23 complexes made an average of 11 or 12 bonds in 5 min at 37°C when incubated with all four NTPs at 1 mM; some complexes synthesized as many as 30 bonds in 20 min under these conditions (Fig. 1A, lanes 1-5). Most of the RNAs made by noninhibited control complexes were too long to resolve on the gel shown in Fig. 1A. RNA synthesis in the presence of the toxin can continue for at least 2 h at 37° (Fig. 1B, lane 6).
The pattern of products obtained on the pML16220 template in the presence of ␣-amanitin was reproducible in many experiments using different batches of amanitin and nuclear extract. Amanitin was reported to have a very slow off-rate from RNA polymerase II at 37°C (1.2 ϫ 10 Ϫ4 /s; see Cochet-Meilhac and Chambon (4)), but those experiments were done under different conditions from those we employed. We were concerned that the catalytic activity of the polymerase in the presence of the toxin at 37°C might reflect cycling of the drug between solution and polymerase, rather than low activity of the polymerase when amanitin is bound. To address this, we repeated the experiment shown in lanes 1-5 of Fig. 1A, except that after the addition of ␣-amanitin another round of gel filtration was performed on the C23 complexes to remove free amanitin. We found that in a 20-min chase the majority of these C23 complexes behaved identically to those in lane 5 of Fig. 1A; however, about one-third of the complexes transcribed much more FIG. 1. RNA polymerase II complexes still elongate in the presence of ␣-amanitin but at a reduced rate. A, C23 complexes on the pML16220 template (lanes 1-5) or on pML16220 variants having a T (pML16T27; lanes 6 -10) or a C (pML16C27; lanes [11][12][13][14][15] at position ϩ27 on the nontemplate strand were pretreated with ␣-amanitin as indicated and chased with 1 mM NTPs for the times specified. The initial transcription reaction contained 2 mM ApC and 0.5 M [␣-32 P]CTP. B, C23 complexes on the pML16220 template were amanitin-treated and chased with 1 mM NTPs as indicated. The initial transcription reaction contained 100 M ATP and 1 M [␣-32 P]UTP. C, C23 complexes prepared as in B were chased with or without amanitin using the times and NTP concentrations indicated. For all panels, the RNA products were purified and resolved on 20% polyacrylamide gels as described under "Materials and Methods." Pertinent transcript lengths generated on pML16220 are presented together with template sequence on the left of panel A and on the right of panel B; lengths of various transcripts produced on pML16T27 and pML16C27 are presented together with the template sequence on the right of panel A. rapidly, at the same rate as uninhibited controls (data not shown). This is consistent with the rate of release of amanitin from polymerase II measured in earlier studies (4). These results indicate that the ability of RNA polymerase to make 10 or more bonds in a 5-min incubation in the presence of amanitin cannot be explained by rapid binding and release of the toxin.
The results in Fig. 1 showed that chain elongation proceeded efficiently through the initial 12 bases downstream of ϩ23 but slowed as the polymerase passed through ϩ36 to ϩ43 (Fig. 1A, lanes [3][4][5], and again when it passed through ϩ54 to ϩ61 (Fig.  1B, lanes 3-6). This reduction in rate appeared to correlate with the requirement for incorporation of C residues into the growing RNA. To begin to address this point more directly, we performed a number of experiments. First, we created two variants of pML16220 in which the A on the nontemplate strand at ϩ27 was replaced with a T or a C residue. Transcript elongation was delayed after incorporating a C at ϩ27 (Fig. 1A, lane 13), while elongation was delayed both before and after incorporating a U at this position (Fig. 1A, lane 8). This suggested that amanitin is differentially affecting the ability of pyrimidines and purines to be incorporated into RNA. However, this result could also mean that RNA polymerase II normally incorporates pyrimidines more slowly, but such behavior is not easily detected because of the rapid rate of transcript elongation in the absence of amanitin. To test this idea, we again chased C23 complexes on the pML16220 template in the presence and absence of amanitin. In this case, the rate of transcript elongation in the reactions lacking amanitin was considerably slowed by reducing the concentration of the chase NTPs to 25, 50 or 100 M. The results (Fig. 1C) show that most of the pauses which occur in the presence of amanitin over a 2-h time course are also observed over the first 10 s to 1 min in reactions performed at suboptimal NTP levels without amanitin (compare lanes 5 and 12, and lanes 6, 11, and 13). Not all pause sites were seen under both conditions, and the relative level of pausing did vary at some sites, for example at position ϩ60 (compare lanes 5 and 9).
The ability of RNA polymerase II to cross long stretches of T residues on the nontemplate strand was also investigated, which required the use of a different template. We constructed a plasmid in which the sequence of the nontemplate strand from ϩ44 downstream reads: 5Ј. . .TTTTTTCTCCATTT-TA. . .3Ј. U36 complexes assembled on this template were Sarkosyl-rinsed and then chased with NTPs at 37°C. Although most complexes in a toxin-free control reaction cleared both T-runs within 5 min (Fig. 2, lane 3), essentially all amanitintreated complexes remained near the end of the first T-run even after 2 h (Fig. 2, lane 9). Thus, amanitin appears to slow transcription very effectively during the synthesis of U-rich segments of the transcript, producing an effect similar to transcriptional arrest. This result was not unexpected, since we had previously shown (16) that stalling RNA polymerase transcription complexes after the addition of more than 3 consecutive U residues to the nascent RNA can lead to arrest. This point will be explored further under "Discussion." Amanitin could slow chain elongation either by affecting bond formation itself or by retarding the ability of the active site to translocate along the template. If translocation were the primary target of amanitin, and if complexes stalled by NTP starvation were already poised to add the next base (that is, if translocation of the active site had already taken place), then addition of the initial base in the presence of amanitin would proceed at the normal rate. We tested this idea with C23 complexes produced on the template used in Fig. 1B. While the majority of uninhibited control complexes left the starting position and added several nucleotides within 15 s, amanitin-treated complexes made no bonds in 15 s. The majority of the treated complexes had made no bonds even after 60 s (data not shown). This is consistent with the idea that amanitin blocks bond formation, but it is also possible that stalled complexes must first translocate the active site in order to add the next NTP, in which case translocation could be the step affected by amanitin.
␣-Amanitin Exacerbates the Arrested Condition-Many laboratories have observed (11,(22)(23)(24)(25) that transcription through certain DNA sequences causes a fraction of the RNA polymerase II ternary complexes to pause without termination. These arrested complexes resume RNA synthesis very slowly (i.e. from 10s of minutes to hours) in the presence of NTPs alone; however, the transcriptional competence of arrested complexes is rapidly recovered by treatment with elongation factor SII (11,22,24,25). We have shown that on the pML5-4NR template arrest occurs 194 bases downstream of the transcription start, within a stretch of T residues on the nontemplate strand (16). This observation is reproduced in Fig. 3. Complexes were arrested at ϩ194, and NTPs were removed by gel filtration (lane 1). These U194 complexes resumed elongation rapidly in the presence of SII and NTPs (lane 5; see also Izban and Luse (16)). In the absence of SII, a much longer time was needed to clear the arrest site (lanes 6 -8). Note that the chases in lanes 6 -8 were performed with ATP, CTP, and UTP only; we presume that the slow leakage through G-stops at ϩ207 and beyond resulted from GTP contamination in the other nucleotides. When the arrested complexes were challenged with NTPs after amanitin treatment, essentially no resumption of transcription from position ϩ194 was seen, even after 2 h (lane 11). Thus, in contrast to complexes stalled by NTP starvation, arrested complexes are inactivated for further chain elongation by ␣-amanitin, at least over a time course of several hours. Again, this result was anticipated; since the rate of escape from arrest by resumption of bond formation is normally very slow and bond formation rates are drastically reduced with amanitin, there should be essentially no detectable escape from arrest in the presence of the toxin over the reaction times we employed.
While the arrested complexes cannot continue RNA synthe- sis in the presence of amanitin, the results in Fig. 3 strongly suggest that transcript cleavage in these complexes is not sensitive to the toxin (compare lanes 3 and 4, or lanes 7 and 8, or 10 and 11). We will consider the question of amanitin's effect on transcript cleavage in more detail in a later section.
␣-Amanitin Increases the Likelihood That Complexes Prone to Arrest Will Fall into That Condition-Our laboratory previously showed that complexes stalled after adding a poly(U) segment (U tail) to the end of the nascent RNA behave progressively more like arrested complexes as the U tail is lengthened (16). Complexes stalled with a 3Ј end consisting of 3 U residues resumed elongation after a 2-min incubation with excess NTPs. However, 40% of complexes with an otherwise identical nascent RNA having 5 U residues at the 3Ј end did not resume RNA synthesis when chased for 2 min (16). We used the same templates employed in the earlier study to investigate the effect of ␣-amanitin as a function of the length of the U tail. Sarkosyl-rinsed complexes were chased to the end of a U-free cassette at ϩ155 on templates in which either the next 3 (pML20-U158) or the next 5 (pML20-U160) residues on the nontemplate strand are Ts (see Fig. 4). The G155 complexes were then gel-filtered and challenged with various combinations of NTPs, with or without ␣-amanitin. Most complexes on both templates resumed elongation from ϩ155 when chased with all NTPs (Fig. 4, lanes 3 and 13), or when chased to the G-stop at ϩ159 (lane 4) or ϩ160 (lane 8). The large majority of C159 complexes on the U158 template resumed transcription after adding G and incubating for 5 min at 37°C (lane 6), but less than half of those starting from ϩ160 on the U160 template did so (lane 10), as expected from previous studies (16). It is important to note that the sequence immediately downstream of ϩ159 on the U158 template and ϩ160 on the U160 template is the identical DNA segment, containing only purines on the nontemplate strand, which is present downstream of position ϩ23 on the template used in Fig. 1. Thus, one would expect that C159 complexes on the U158 template and U160 complexes on the U160 template should chase effectively in 5 min in the presence of ␣-amanitin. However, most of the C159 complexes and nearly all of the U160 complexes failed to resume elongation after amanitin treatment (compare lanes 4 and 5, and lanes 8 and 9). If the complexes were advanced further along the template before amanitin addition, such that the 3Ј ends of the nascent RNAs were no longer U-rich (complexes G163 on the U158 template and G164 on the U160 template), elongation in the presence of amanitin was once again efficient (lanes 6 and 7, and 10 and 11). These results suggest that the effect of amanitin depends not only on the sequence of bases to be added to the RNA but also on the sequence at the 3Ј end of the nascent transcript. It is worth noting that in the presence of ␣-amanitin RNA polymerase II synthesized the polypurine segment of RNA between, for example, ϩ160 and ϩ172 on the U158 template about as rapidly as it synthesized the nearly identical RNA (between ϩ24 and ϩ35) on the template in Fig. 1. Thus, the ability of transcription to proceed at a greatly reduced rate in the presence of amanitin is not strongly affected by the distance downstream of transcription start.
The Sequence Composition at the Nascent Transcript's 3Ј End Is a Crucial Determinant of ␣-Amanitin Inhibition-To further explore the role of the transcript in amanitin inhibition, we performed an experiment in which we could compare the amanitin response of transcription complexes that differed only in the 3Ј ends of their nascent transcripts. The first G residues downstream of transcription start on the nontemplate strand of the pML16220 template (see Fig. 1) occur at positions ϩ24 through ϩ26. Thus, incubation of C23 complexes with either GTP or ITP should generate G26 or I26 complexes. The different mobilities of the G26 and I26 transcripts confirmed that IMP was successfully incorporated in place of GMP (Fig. 5, compare lanes 5 and 6). As expected from the results shown in Fig. 1, the G26 complexes showed substantial chain elongation in 5 min in the presence of ␣-amanitin (lane 11). However, the I26 complexes, most of which chased in the absence of amanitin (lane 8), were inactive in a 5-min elongation reaction in the presence of the toxin (lane 9). This was not the result of a block of ITP incorporation by amanitin, since the first two bases added to the ϩ26 complexes are A residues (see Fig. 1A). Note also that the initial chase from ϩ23 to ϩ26 could be completed with ITP in the presence of amanitin (lane 3). Thus, transcription complexes which are identical except for the three residues at the 3Ј end of the nascent RNA can have very different responses to ␣-amanitin.
␣-Amanitin Greatly Reduces the Rate of Pyrophosphorolysis in Stalled Complexes-Stalled RNA polymerase II transcription complexes incubated with pyrophosphate liberate NTPs by sequential cleavage of NMPs from the 3Ј ends of the nascent RNAs (13,26). Stalled and arrested ternary complexes can also cleave their nascent transcripts without the addition of pyrophosphate or other factors (11,12,27). We believe this represents an intrinsic activity of the RNA polymerase and not residual contamination with SII; this point will be considered in detail under "Discussion." Complexes stalled at ϩ20 on the pML20 template were incubated for 2, 15, or 60 min either with Mg 2ϩ alone or with Mg 2ϩ and 2 mM pyrophosphate; for each condition reactions were performed with or without ␣-amanitin (Fig. 6). After 2 min, substantial pyrophosphorolysis was ob- served, compared with the control, which received only Mg 2ϩ (lanes 2 and 3), and this reaction was nearly completely amanitin-sensitive (compare lanes 4 and 5). After 15 min, however, it was clear that some pyrophosphorolytic cleavage did take place above the Mg 2ϩ -only background in the presence of amanitin. Quantitation of the remaining 20-mer indicated that only 46% as much uncleaved 20-base transcript remained in lane 9 compared to lane 8. Thus, as we observed with the forward reaction, pyrophosphorolysis in stalled complexes was strongly but not completely inhibited by amanitin. Cleavage in the Mg 2ϩonly case was also reduced by amanitin. After 2 min, cleavage was completely blocked (compare lanes 3 and 4); after 15 min, some cleavage had taken place in the presence of amanitin (lane 8) but the amount of uncleaved transcript in lane 7 (no amanitin) was only 51% of the amount in lane 8, indicating that transcript cleavage in the Mg 2ϩ -only reaction was not completely amanitin-sensitive. Note in lanes 11 and 12, where cleavage with Mg 2ϩ had continued for 1 h, that amanitin reduced the total amount of cleavage (the ratio of 20-mer in lanes 11 and 12 was 0.44), and it also strongly reduced the production of the 19-mer (compare also lanes 7 and 8 from the 15-min reaction). However, the production of cleavage products shorter than 19 was actually greater in the presence of amanitin after 1 h, even though total cleavage, as judged by the amount of 20-mer remaining, was reduced. We are not certain of the reason for this. It suggests that amanitin can inhibit only the initial spontaneous cleavage in a stalled complex; once this cut is made, subsequent cleavages are resistant. It is possible that the 19-mer is a metastable species. If cleavage in the presence of amanitin must bypass the 19-mer, subsequent cleavage events may be easier.
Finally, SII-mediated transcript cleavage in amanitintreated stalled elongation complexes was completely blocked by the toxin in all cases (data not shown), in agreement with earlier results (11).
␣-Amanitin Inhibits Neither the Intrinsic Cleavage Activity nor Pyrophosphate-mediated Transcript Cleavage by Arrested Ternary Complexes-We have shown that complexes arrested at ϩ194 on the pML5-4NR template cleave 7-17 nt from the 3Ј ends of their nascent RNAs in the presence of SII (13,16). The same set of 7-17 nt RNAs are released at a much slower rate when the U194 complexes are incubated with Mg 2ϩ alone (13). Pyrophosphate treatment of U194 complexes also results in the relatively rapid release of the 7-17-nt RNAs; in this case the liberated fragments have 5Ј-triphosphate termini (13). As expected from previous reports (11,12), SII-mediated transcript cleavage in U194 complexes was completely blocked by ␣-amanitin (data not shown). The results in Fig. 3, however, indicated that amanitin had no effect on the intrinsic transcript cleavage activity of arrested complexes. A 2-h incubation of ϩ194 complexes with Mg 2ϩ gave the same level of truncation products in the presence or absence of amanitin (Fig. 3, lanes 3  and 4). The major shortened transcripts in these lanes appeared to correspond to the major 10-and 14-base cleavages obtained within minutes in SII-mediated truncation reactions of the 194-nt transcript (data not shown for this figure; see Fig.  7B and Rudd et al. (13) and Izban and Luse (16)). As expected (11,12), transcript cleavage in arrested complexes led to reacquisition of elongation competence, so that intrinsic cleavage in the presence of amanitin, ATP, CTP, and UTP resulted in elongation up to the first G-stop upstream of the arrest site, at position ϩ186 (Fig. 3, lanes 10 and 11).
To confirm that amanitin has no effect on the endogenous cleavage reaction, we decided to examine directly the fragments released from the 3Ј ends of the nascent RNAs. We prepared U194 complexes whose nascent RNAs were uniformly labeled with [ 32 P]UTP and incubated them for 60 min in Mg 2ϩ , with or without ␣-amanitin. RNAs liberated in this reaction were resolved on the gel shown in Fig. 7B. For reference, lane 7 contains RNAs produced by SII-mediated transcript cleavage. As expected (13), lower levels of these same RNAs were obtained in Mg 2ϩ -only incubations (lane 3). U194 complexes treated with amanitin gave the same level of cleavage products as the noninhibited complexes (compare lanes 3 and 4). Thus, factor-independent transcript cleavage in arrested complexes is amanitin-resistant.
Pyrophosphorolysis in arrested U194 complexes occurred in the presence of ␣-amanitin (Fig. 7A, compare lanes 2 and 3). However, while the major cleavage at 14 nt from the 3Ј end appeared to occur to about the same extent with or without the toxin, other aspects of the cleavage pattern differed between the reactions. Since the initial cleavage produces an elonga- tion-competent complex which performs subsequent pyrophosphorolysis very slowly (Fig. 6), the difference might reflect multiple pyrophosphorolytic cleavages in the absence of the toxin versus a single cleavage in its presence. We resolved this point by examining the short RNA fragments released by incubation of uniformly labeled U194 complexes with pyrophosphate. These RNAs were found to be nearly identical (excepting only a single RNA of about 5 nt) regardless of the presence of amanitin in the incubation (Fig. 7B, compare lanes 5 and 6). Thus, endonucleolytic transcript cleavage by arrested complexes in the presence of pyrophosphate is not sensitive to ␣-amanitin. DISCUSSION We have found that the mushroom toxin ␣-amanitin substantially reduces the rate of transcription by elongation competent RNA polymerase II ternary complexes; however, elongation was not completely blocked. Most amanitin-treated complexes can continue elongation for hours, but complexes which are arrested are essentially unable to resume transcription in the presence of amanitin. While amanitin greatly retards pyrophosphorolysis by elongation-competent complexes, it has no effect on either the intrinsic or pyrophosphate-mediated endonucleolytic transcript cleavage activities of arrested complexes. After this study was submitted, Chafin et al. (28) reported that Drosophila RNA polymerase II initiated from poly(dC)-tailed templates also elongates RNA chains and performs pyrophosphorolysis at greatly reduced but detectable rates in the presence of ␣-amanitin.
It had been observed previously that promoter-initiated RNA polymerase II elongation complexes treated with amanitin can add several nucleotides to their nascent chains (8,9). Recently, Gu et al. (10) reported that a specifically initiated RNA polymerase II ternary complex stalled at ϩ218 or ϩ220 could continue transcription for about 8 bases in the presence of amanitin. An examination of the sequence of the RNA-like strand of the template used by Gu et al. downstream of ϩ218/220 shows that all but one of the next 9 or 11 bases are purines, followed by three pyrimidines (10). Thus, it seems likely that Gu et al. observed the same effect which we document here. We speculate that the relative rarity of sufficiently long purine runs in random sequence DNA accounts for the lack of previous reports on the incomplete inhibition of transcript elongation by amanitin (but see Job et al. (29), discussed below).
We have reported that RNA polymerase II initiating at the adenovirus 2 major late promoter cannot add even a single nucleotide to a dinucleotide primer in the presence of amanitin (30). In those experiments, production of a low level of trinucleotide was observed with amanitin, but this same level was also seen even in the absence of template, or with dinucleotides that could not prime RNA synthesis at the adenovirus promoter. We interpreted these results to mean that the amanitinresistant trimer was generated by activities other than RNA polymerase II. However, in light of our current results we cannot exclude the possibility that a very low level of transcription initiation can take place in the presence of amanitin. RNA polymerase II preinitiation complexes assembled from nuclear extracts are unstable in the presence of ATP, which is required for transcription initiation under these conditions (30, 31).  8 and 9). The G26 complexes were identically chased with or without ␣-amanitin, as indicated (lanes 10 and 11). Transcripts were resolved on a 20% polyacrylamide gel. Pertinent RNA lengths are provided in the right margin, with 26(I) and 26(G) denoting 26-nt products containing I or G residues, respectively, at their 3Ј end.
Thus, amanitin-treated preinitiation complexes might be inactivated before any bonds could be formed.
The effects of amanitin on transcription of homopolymeric templates by pure RNA polymerase II have also been studied. Two groups reported that ␣-amanitin permits formation of the first phosphodiester bond but absolutely blocks the synthesis of all subsequent bonds (32,33). It is difficult to compare these results to our own findings since the RNA polymerases in these systems did not pass through initiation at a promoter. One study on transcription of homopolymeric templates detected extensive amanitin-resistant transcription by polymerase II. Job et al. (29) reported that with poly(dC) or poly(dC)⅐poly(dG) as template and GTP as substrate, about 30% of RNA synthesis by wheat germ RNA polymerase II was resistant even to 100 g/ml ␣-amanitin. Slippage of the nascent transcript on the template clearly played a part in these results (29), but it is interesting that the only template-substrate combination which allowed significant amanitin-resistant transcription involved the synthesis of a polypurine transcript.
We found that the sequence of the region to be transcribed was a major factor in the ability of RNA polymerase II to extend the nascent RNA, in the presence or absence of amanitin ( Fig.  1, A-C). However, the results of substituting ITP for GTP in only the last three bases of the nascent RNA (Fig. 5) argue that the sequence of the transcript is also important in elongation competence. In order to rationalize these observations it is useful to briefly review current ideas on the mechanism of transcriptional arrest. Several groups have noted that arrest by RNA polymerase II most often occurs immediately after the synthesis of a U-rich RNA (20, 22, 23, 34 -38). Sequences flanking the template region encoding the poly(U) segment also play a very important role in arrest (23,34,39,40). We had noted that DNA with 5 consecutive T residues on the nontemplate strand does not provide a barrier to the polymerase during transcription with excess NTPs. However, if the polymerase was forced to pause after the incorporation of the 5 U residues, because of the absence of the next NTP required for elongation, then nearly half of the complexes could not resume transcription after a 5 min incubation with excess NTPs (16). Thus, while the incorporation of many consecutive U residues does not necessarily force arrest, polymerases crossing T-rich sections of the nontemplate strand are in danger of arrest. Our results suggested that the length of time which the polymerase spends with a transcript containing a U-rich 3Ј end is crucial to the arrest process. Very recently, the importance of "dwell time" at potential arrest sites was directly assessed by changing the overall rate of transcription with TFIIF or ammonium ions; in both cases, more rapid transcription was inversely correlated with arrest (41).
Once arrest has occurred, rapid resumption of transcription cannot take place without cleavage of the nascent RNA well upstream (from 5 or 6 to as many as 17 bases) of the initial site of bond formation (16,42). Although transcript cleavage occurs spontaneously in both stalled and arrested complexes, it is greatly stimulated by the SII elongation factor (11)(12)(13). The source of this spontaneous cleavage is still somewhat controversial. We have argued that this activity is intrinsic to the RNA polymerase itself. While we cannot absolutely eliminate the possibility of SII contamination in our partially purified complexes, the following points argue strongly against it. First, if the cleavage were caused by a very low level of residual SII, which was not removed by gel filtration in the presence of Sarkosyl, one would expect that a second round of gel filtration under the same conditions would remove almost all of the residual activity. However, when we did such an experiment, we did not see any reduction in cleavage levels after the com- plexes had been subjected to a second gel filtration step (data not shown). Second, the factor-independent cleavage activity in Fig. 6 makes its initial cut only one nucleotide from the 3Ј end; however, the addition of SII to stalled complexes leads to cleavage primarily in dinucleotide increments (26). The action of amanitin on arrested complexes also argues against contamination. We can see no stimulation of cleavage by added SII in the presence of amanitin, and yet the spontaneous cleavage activity is completely amanitin-resistant. This might be explained if amanitin could only inhibit the binding of SII; in this model, residual SII, which is already bound, would not be inhibited. However, when we tested this idea by adding SII to complexes and then followed with amanitin, we still saw absolutely no cleavage above the control (data not shown). Finally, it is important to recall that Escherichia coli RNA polymerase shows spontaneous transcript cleavage activity even when it is prepared from cells which lack functional genes for both the GreA and GreB transcript cleavage factors (43). RNA polymerase III, which has no known elongation factors, also exhibits spontaneous cleavage of transcripts in stalled complexes in the presence of Mg 2ϩ (44).
We had hypothesized that arrest might result from loss of contact between the active site of the polymerase and the 3Ј end of the transcript (16). Transcript cleavage was seen as a mechanism to generate a new 3Ј end that is accessible to the active site. The subsequent demonstration that pyrophosphate can also stimulate cleavage at the same sites as SII suggested that the active site itself might be the cleavage agent (13). Arrest could then reflect the translocation of the RNA polymerase's catalytic center upstream along the nascent RNA. It is plausible that U-rich regions are the least avidly bound by the active site, making transcription complexes with U-rich 3Ј ends the most prone to arrest. Elongation competence would be restored by SII-stimulated cleavage at upstream locations on the transcript with which the active site stably associates.
In the context of this model (see also Gu and Reines (41)) the importance of dwell time at potential arrest sites is easy to envision. If upstream translocation of the active site is much slower than the usual rate of bond formation, arrest will be very unlikely unless the polymerase can be paused for some time after synthesis of the crucial U-rich 3Ј end. Thus, in Fig.  4, a C159 complex whose transcript ends . . .GUUUC-3Ј is mostly active when chased, but when amanitin is added to greatly increase dwell time (by lowering the rate of initial bond formation), most of the C159 complexes are inactive upon chase (compare lanes 4 -6). A U160 complex, with a more U-rich 3Ј end (. . .GUUUUU-3Ј), is only partially active in the absence of amanitin and nearly inactive in the presence of the toxin (compare lanes 8 -10 of Fig. 4). We had suggested (16) that the active site probably partitions between elongation-competent and elongation-incompetent locations in arrested complexes, since arrested complexes show a very slow but easily detected rate of resumption of transcription (Fig. 3, lanes 6 -8). Bond formation at 37°C and 1 mM NTPs occurs on average about 5 times/s (20), so the active site would need to be in the elongation competent configuration for only a very short period to allow some complexes to escape from arrest. However, bond formation is much slower in the presence of amanitin, which would account for the inability of amanitin-treated, arrested complexes to resume elongation (Fig. 3).
What conclusions can we draw concerning the mechanism of inhibition of transcription by amanitin, given both our present results and the large body of earlier work on the toxin? Previous studies with homopolymeric templates (29,32,33) had suggested that translocation and not bond formation is blocked by amanitin, since the initial bond can be formed when aman-itin is present but transcription cannot continue. As we noted above, we cannot discriminate between these models from our own results. Johnson and Chamberlin (45) showed that binary complexes of yeast RNA polymerase II and RNA could not only cleave the RNA but could also add nucleotides to the newlycreated 3Ј ends. This template-independent bond addition was characterized as partially sensitive to amanitin. It is difficult to envision how amanitin functions to inhibit translocation along the template when it also affects bond addition in a templateindependent reaction. From our own work, we can say that amanitin does not "tie down" the active site, since amanitintreated arrested complexes can cleave their nascent RNAs at locations far upstream of the original polymerization site in the presence of amanitin. However, this result could still be obtained if amanitin blocks downstream, but not upstream, translocation of the active site.
Perhaps the most interesting aspect of our results is the fact that amanitin does not inhibit several of the catalytic activities of the RNA polymerase. It is striking that amanitin inhibition occurs only when the active site is near the 3Ј end of the transcript, with the exception of SII-mediated cleavage in arrested complexes. Spontaneous cleavage and pyrophosphorolysis in arrested complexes are both completely insensitive to amanitin. This suggests that amanitin must work through the 3Ј end of the transcript, or alternatively, that amanitin binds near a location normally occupied by the 3Ј end of the RNA. Such an idea is consistent with the findings of Johnson and Chamberlin (45), who showed that amanitin does not inhibit the initial SII-mediated cleavage reaction in binary complexes, when the active site presumably occupies an internal position on the transcript. However, amanitin does block any further cleavage in these complexes. Note that after the initial cleavage in binary complexes the active site must be at the 3Ј end of the RNA, since bond formation can occur after the first cleavage (45). Mutations have recently been described in the largest subunits of E. coli (46) and Bacillus subtilis (47) RNA polymerases which confer resistance to streptolydigin. These mutations occur in region F (48), a segment which shows considerable sequence similarity among the largest subunits of both eukaryotic and prokaryotic RNA polymerases. Region F is also the location of amanitin resistance mutations in RNA polymerase II (see Bartolomei and Corden (49) and references therein), which is not unexpected since streptolydigin's effect on prokaryotic RNA polymerase parallels the effect of amanitin on RNA polymerase II. Significantly, it has been proposed that region F might form part of the binding site for the 3Ј end of the nascent RNA (48).
The fact that amanitin blocks SII-mediated transcript cleavage in arrested ternary complexes would seem to violate the idea that amanitin can act only near the 3Ј end of the transcript. However, it is possible that amanitin simply blocks access of SII to the upstream sites at which transcript cleavage takes place, rather than blocking the cleavage reaction directly. This is again consistent with the binary complex results (45). Binary complexes are probably less sterically confined than ternary complexes, and as just noted the initial SII-mediated cleavage in binary complexes is not amanitin-sensitive.