RNA Polymerase II Trigger Loop Mobility

Rpb9 is a conserved RNA polymerase II (pol II) subunit, the absence of which confers alterations to pol II enzymatic properties and transcription fidelity. It has been suggested previously that Rpb9 affects mobility of the trigger loop (TL), a structural element of Rpb1 that moves in and out of the active site with each elongation cycle. However, a biochemical mechanism for this effect has not been defined. We find that the mushroom toxin α-amanitin, which inhibits TL mobility, suppresses the effect of Rpb9 on NTP misincorporation, consistent with a role for Rpb9 in this process. Furthermore, we have identified missense alleles of RPB9 in yeast that suppress the severe growth defect caused by rpb1-G730D, a substitution within Rpb1 α-helix 21 (α21). These alleles suggest a model in which Rpb9 indirectly affects TL mobility by anchoring the position of α21, with which the TL directly interacts during opening and closing. Amino acid substitutions in Rpb9 or Rpb1 that disrupt proposed anchoring interactions resulted in phenotypes shared by rpb9Δ strains, including increased elongation rate in vitro. Combinations of rpb9Δ with the fast rpb1 alleles that we identified did not result in significantly faster in vitro misincorporation rates than those resulting from rpb9Δ alone, and this epistasis is consistent with the idea that defects caused by the rpb1 alleles are related mechanistically to the defects caused by rpb9Δ. We conclude that Rpb9 supports intra-pol II interactions that modulate TL function and thus pol II enzymatic properties.

Rpb9 is a small (122 amino acids in Saccharomyces cerevisiae) subunit of RNA polymerase II (pol II) 2 that is highly conserved among eukaryotes (1,2). It is composed of two zinc ribbon domains characterized by the zinc-chelating motif (CX 2 CX n CX 2 C). Although Rpb9 is not essential for growth in yeast, rpb9 null mutants have a number of phenotypes, including slow growth (2), hypersensitivity to several drugs (2)(3)(4)(5), upstream shifts in transcription start site (2,6), defects in tran-scription-coupled repair that result in enhanced sensitivity to UV light (7), and decreased transcriptional fidelity in vivo (4,8) and in vitro (9,10).
The fidelity defect of pol II that is lacking Rpb9 (pol II⌬9) has been examined in several studies. Decreased selectivity for correct (complementary to the DNA template) versus incorrect NTP substrates has been observed (10). However, our prior experiments have not provided evidence for a selectivity defect, and, in our hands, the decreased fidelity of pol II⌬9 appears to be the result of inefficient proofreading caused by an increased rate of RNA extension from a mismatched 3Ј-end combined with a decreased efficiency of TFIIS-mediated mismatch removal (9). A role for Rpb9 in fidelity is consistent with reported roles for the homologous subunits in RNA polymerases I and III (A12.2 and C11, respectively) (11,12) as well as the related archaeal protein TFS (13,14). However, the role that Rpb9 plays in pol II fidelity is probably distinct from these homologs, all of which function as hybrid proteins combining functions of both Rpb9 and TFIIS.
Rpb9 is located on the leading edge of pol II at a position that is not in immediate proximity to the active site (15), and the absence of Rpb9 has little or no effect on the binding affinity of TFIIS with pol II (16). This leaves open the question of how Rpb9 exerts its effects on elongation and cleavage. Using rapid quench-flow kinetic techniques, it has been demonstrated that the absence of Rpb9 promotes the sequestration of NTP substrates in the active site before formation of a new phosphodiester bond (10). This behavior was similar to that observed with pol II containing Rpb1-E1103G (pol II-(Rpb1-E1103G)) (17), which contains a substitution within the trigger loop (TL), a mobile structural element that closes on the pol II active site and interacts with the substrate NTP (reviewed in Refs. 18 -21). Due to the proximity of Rpb9 to the TL, it was proposed that the C-terminal domain of Rpb9 interacts directly with the open pol II TL and thereby stabilizes its open conformation (10). However, in this same study, direct mutational analysis of an amino acid proposed to interact with the open TL (Rpb9-K93) showed no fidelity (or any other) defect.
Although deletion of RPB9 and particular substitutions within the TL can confer similar phenotypes and biochemical defects, no mechanism through which Rpb9 and the TL could interact has been defined. In this study, we provide additional genetic and biochemical evidence that Rpb9 affects pol II activity through the TL but does so indirectly through stabilization of Rpb1 ␣-helices 20 and 21 (␣20 and ␣21) and a small loop between them that we term the Rpb1 anchor loop (AL). We describe a plausible biochemical model for how this could occur and confirm predictions suggested by this model related to phenotypes of specific alleles of RPB9 and RPB1.

Effect of Rpb9 on Pol II Activity in the Presence of ␣-
Amanitin-Although previous work had shown that the absence of Rpb9 promotes NTP sequestration in the pol II active site, the proposed mechanism, in which sequestration is increased through altered mobility of the TL, was without direct experimental support (10). We decided to test the general idea that Rpb9 affects TL mobility using an experimental approach that exploits the mushroom toxin ␣-amanitin, which has been shown to inhibit pol II by locking the TL in a position where it cannot participate directly in active site chemistry (22,23). Elongation by pol II is strongly inhibited in the presence of saturating concentrations of ␣-amanitin, but it is not completely prevented (24,25). We reasoned that if Rpb9 affects TL mobility, then, by immobilizing the TL with ␣-amanitin, the effects of Rpb9 on pol II activity would be markedly altered.
We have previously reported that the absence of Rpb9 increases the rate of pol II-catalyzed RNA synthesis in a singlenucleotide misincorporation assay (9). We chose to examine the effects of ␣-amanitin with this assay rather than alternatives that would assess the faster rate of incorporation of correct, template-matched NTPs (23) (see below) because of the relatively low affinity of yeast pol II for ␣-amanitin (26). We feared that the rapid rate of elongation with template-matched NTPs might compete with the "on" rate of ␣-amanitin binding such that, even at saturation, a small amount of free pol II in equilibrium with ␣-amanitin-bound pol II could, during its brief period of freedom from the inhibitor, sometimes elongate, and this would be more likely to occur in a variant with a faster elongation rate. Indeed, observations consistent with this effect for yeast TL mutations that increase elongation rate have been reported (23).
To assess the effect of ␣-amanitin on misincorporation, purified pol II or pol II⌬9 was immobilized on Ni 2ϩ -NTA beads via a His 6 tag on Rpb3, and elongation complexes (ECs) were assembled with DNA and RNA oligonucleotides in which the 10-nt RNA was completely complementary to the template DNA and was 5Ј-end-labeled with 32 P (Fig. 1A). Rather than supplying CTP, the nucleotide specified by the template for extension of the labeled RNA oligonucleotide, a nearly saturating (9) amount of ATP (1 mM) was added, and the time course of A-for-C misincorporation was followed in the presence or absence of ␣-amanitin (Fig. 1, B and C). The fraction extended to 11 nt was plotted against time, and the data were fit to a single exponential equation (Fig. 1, D and E). The faint band at position 12 at some of the later time points in Fig. 1C probably reflects extension of the mismatched 3Ј-end by ATP, which is the template-specified nucleotide at this position. We have previously shown that such mismatch extension can occur at a rate consistent with its appearance here and that it is 2-3-fold faster for pol II⌬9 than for pol II (9). Although small in size, we combined signal in this band with the band at position 11 because both reflect misincorporation. We confirmed that the extension to 11 nt was the result of misincorporation and not the result of contaminating CTP because the 11-nt product was not rapidly extended by the addition of CTP, GTP, and UTP (data not shown), consistent with a mismatched 3Ј-end (9,27). In the absence of ␣-amanitin, the misincorporation rate for pol II FIGURE 1. Effect of ␣-amanitin on A-for-C misincorporation. A, schematic representation of assembled EC. Template and non-template DNA oligonucleotides depict the transcription bubble. A RNA oligonucleotide labeled at its 5Ј-end with 32 P is between the DNA strands. The small box indicates the template position for extension by the next correct NTP, and the shaded area represents the approximate region associated with pol II. B and C, time course of A-for-C misincorporation. The EC depicted in A was assembled with purified pol II (B) or pol II⌬9 (C) and incubated with ATP (1 mM). Aliquots were removed at the indicated times, and the products were separated by denaturing polyacrylamide gel electrophoresis and visualized on a phosphor imager. Samples in lanes 8 -13 of both B and C contained ␣-amanitin (160 g/ml), which was added 15 min before the addition of ATP. D and E, quantitation of extension time course. The fraction of ECs extended to 11 nt was plotted versus time and fit to a single exponential equation (open circles, pol II; closed circles, pol II⌬9). The y axis was normalized to the individual reaction end points to emphasize relative rates. Error bars on each data point indicate the S.D. for three independent experiments. The values of k obs are based on a fit to a single exponential equation performed as described under "Experimental Procedures." Error for k obs indicates the 95% confidence interval for the nonlinear regression fit of the experimental data to the first order exponential curve.
(k obs ϭ 0.069 Ϯ 0.006 min Ϫ1 ) was a little more than half that observed for pol II⌬9 (k obs ϭ 0.122 Ϯ 0.017 min Ϫ1 ), consistent with our previously reported observations (9), but in the presence of ␣-amanitin (160 g/ml), the rates for pol II (k obs ϭ 0.029 Ϯ 0.004 min Ϫ1 ) and pol II⌬9 (k obs ϭ 0.025 Ϯ 0.002 min Ϫ1 ) were essentially identical. This concentration of ␣-amanitin was saturating, because concentrations as high as 640 g/ml had no additional effect on the rate. Overall, the misincorporation reaction was less sensitive to ␣-amanitin than the incorporation of correct NTPs (data not shown), consistent with published observations (23), but immobilization of the TL by ␣-amanitin completely eliminated the effect of Rpb9 on the rate. This observation is consistent with the idea that the effect of Rpb9 on the rate of misincorporation is mediated, at least in part, by affecting the TL in some manner.
rpb9 Alleles That Suppress rpb1-G730D (sit1-7)-Our interest in the molecular mechanism(s) by which Rpb9 could alter TL activity fortuitously merged with a more general approach that we had pursued to identify mutations that affect Rpb9 function. An allele of RPB1 (rpb1-G730D) was originally isolated as sit1-7 (suppressor of initiation of transcription) based on its ability to support low level transcription from the HIS4 gene in the absence of normally required activating factors (28). Subsequently, it was independently isolated as a suppressor of the menadione hypersensitivity of rpb9⌬ strains (29). A severe growth defect is conferred by rpb1-G730D in the presence of RPB9, but nearly wild-type growth can be restored in rpb1-G730D strains if they also contain certain alleles of RPB9, including null alleles or alleles encoding proteins in which Rpb9 residues 48 -122, 65-70, or 89 -95 are absent (29). Previous experiments had shown that these deletions do not prevent binding of Rpb9 to pol II (7,30); thus, it is likely that some kind of negative interaction between RPB9 and rpb1-G730D is mediated through the C-terminal half of the 122-residue Rpb9.
Pol II-(Rpb1-G730D) has been shown to be catalytically slow, with an elongation rate only about 10% that of wild-type pol II (31). To examine the interaction of RPB9 and rpb1-G730D more closely, we took advantage of yeast strain CKY283 (23), in which the only copy of RPB1 is provided on a centromeric plasmid (Table 1). We introduced a RPB9 ORF knock-out into this strain (32,33) to create dopY249 (Table 1) and then replaced RPB1 with rpb1-G730D via a plasmid shuffle (34). ECs were assembled with pol II that had been immobilized from whole cell extract (WCE) of these strains via a TAP tag on Rpb3, and A-for-C misincorporation assays were performed. Rates of misincorporation for pol II derived from WCE from wild-type and rpb9⌬ strains (Fig. 2) were very similar to those obtained with highly purified pol II and pol II⌬9 (Fig. 1), with the absence of Rpb9 resulting in an increase in misincorporation rate of about 2-fold. The more convenient analysis of pol II variants from WCE thus provides a reliable measure of relative enzyme activity, as we have reported previously (9). When the rpb1-G730D allele was shuffled into the RPB9 strain, the misincorporation rate decreased, consistent with the previously reported decrease in the elongation rate with template-matched NTPs (31). However, in the rpb9⌬ background, we observed mutual suppression of the fast rate characteristic of pol II⌬9 and the slow rate associated with pol II-(Rpb1-G730D), and the result was a rate near that of wild-type polymerase (Fig. 2). This suggested that specific rpb9 alleles that relieve the negative genetic interaction with rpb1-G730D, such as those that delete Rpb9 residues 48 -122, 65-70, or 89 -95, might also affect the pol II elongation rate, and we therefore decided to screen additional rpb9 alleles for their ability to relieve the negative interaction between RPB9 and rpb1-G730D.
Our initial results were consistent with the results of Koyama et al. (29) (Fig. 3A). The severe growth defect in rpb1-G730D RPB9 strains was partially relieved when RPB9 was deleted, and the rpb1-G730D rpb9⌬ strain retained only the growth defect caused by rpb9⌬. Providing RPB9 on a single-copy plasmid restored the negative interaction with rpb1-G730D, as did an allele that changed Rpb9 amino acids Arg-91, Arg-92, Lys-93, and Asp-94 to alanine (rpb9-(91-94)A). (The Rpb9 amino acid numbering system used throughout this paper does not take into account the 8-amino acid N-terminal FLAG tag present in all of our constructs.) In contrast, an allele in which Rpb9 amino acids 88 -97 are deleted (rpb9-(88 -97)⌬) or an allele encoding only the N-terminal 59 amino acids of Rpb9 (rpb9-(1-59)) did not restore this negative interaction. Rpb9 encoded by all of the tested alleles appeared to be expressed and capable of associating with pol II because all of them restored nearly wild-type growth to an RPB1 rpb9⌬ strain, and in the rpb1-G730D rpb9⌬ background, both rpb9-  and rpb9-(88 -97)⌬ grew significantly better than when the vector alone was present. These results strongly suggest that Rpb9 residues 88 -90 and/or 95-97 play a major role in the negative interaction between RPB9 and rpb1-G730D.
Additional rpb9 alleles were constructed to further dissect this region of the protein. rpb9-Q87A, rpb9-S96A, rpb9-(88 -90)A, and rpb9-(95-97)A all restored wild-type growth to a rpb9⌬ strain, but only rpb9-(95-97)A also completely avoided the negative interaction with rpb1-G730D (Fig. 3B). The absence of the negative interaction does not appear to be the result of an inability of Rpb9-(95-97)A to bind the polymerase, because growth of the rpb1-G730D rpb9-(95-97)A strain was much more robust than that of the rpb1-G730D rpb9⌬ strain. This is consistent with published work in which amino acid substitutions near this region of Rpb9 do not affect pol II bind-  JULY 8, 2016 • VOLUME 291 • NUMBER 28

Rpb9 and Pol II Trigger Loop Mobility
ing (30), but we address this issue more directly below. When the rpb9⌬ and rpb1-G730D alleles were present together, we observed complete mutual suppression of the fast and slow misincorporation phenotypes that they confer when present alone. However, other phenotypes associated with rpb9⌬, such as slow growth (Fig. 3A) and hypersensitivity to mycophenolic acid (MPA) (data not shown) are not completely suppressed by rpb1-G730D, suggesting an additional function of Rpb9 that is not related to in vitro speed.
In Vitro Activity of Pol II-Rpb9-(95-97)A-The phenotype of rpb9-(95-97)A led us to consider how the catalytic properties of pol II are affected when Rpb9 is replaced by Rpb9-(95-97)A. In these assays, we took advantage of a method in which the rate of elongation is estimated as the average rate at which the polymerase completes transcription across a DNA template (23). ECs were assembled as shown in Fig. 4A, all four NTPs were added simultaneously, and transcription was monitored over time ( Fig. 4, B and C). After gel electrophoresis of the RNA products, the half-time for the synthesis of products 61 nt in length was estimated using the empirical observation that the time course of the synthesis of these products closely followed a single exponential equation (Fig. 4C), and the average rate of elongation was determined by dividing the actual number of nt added to the 10-nt primer at completion (51 nt) by this calculated half-time. Pausing of the polymerase clearly plays a role in the time course, but this assay has proven to be a useful method for comparing catalytic activity of pol II variants (23). At NTP concentrations of 1 mM each, which was near saturation (17) (data not shown), the average rates of elongation in the absence of Rpb9 (13.2 Ϯ 3.2 nt/s) or in the presence of Rpb9-(95-97)A (12.4 Ϯ 2.6 nt/s) were essentially identical and roughly twice that observed when Rpb9 was present (5.8 Ϯ 2.4 nt/s). As described in greater detail below, the A-for-C misincorporation rate of pol II-Rpb9-(95-97)A is also fast, yet, like that of pol II⌬9, the misincorporation rate is essentially identical to that of wild-type pol II in the presence of ␣-amanitin (see Fig. 8), which suggests that TL function may be affected by the amino acid changes in Rpb9-(95-97)A.
Although the in vivo phenotypes of both RPB1 rpb9-(95-97)A and rpb1-G730D rpb9-(95-97)A strains appeared to preclude a major defect in association of Rpb9-(95-97)A with the rest of the polymerase, we directly addressed this possibility using a pull-down assay. dopY127, a rpb9⌬ strain with a His 6 tag on Rpb3, was transformed with a centromeric plasmid containing either RPB9 or rpb9-(95-97)A, each of which was FLAG epitope-tagged at its N terminus. Pol II was immobilized from WCEs derived from these strains using Ni 2ϩ -NTA beads, and the beads were washed just as they would be for in vitro transcription assays, such as those described in Fig. 4. Bead-bound proteins were dissolved in SDS, fractionated by SDS-PAGE, and probed with antibodies against the N-terminal FLAG tag of Rpb9 as described under "Experimental Procedures." A single protein band of ϳ15 kDa, consistent with the expected size of Rpb9, was detected in lanes derived from extracts from cells with the RPB9 or rpb9-(95-97)A plasmid but not the control extract made from rpb9⌬ cells carrying only the empty vector. The membrane was then gently washed and probed with an antibody against the Rpb1 C-terminal domain, which revealed a large protein, consistent with the size of Rpb1 (191 kDa), in each sample while retaining the signal from the FLAG-tagged Rpb9 (Fig. 5). Quantitation yielded a ratio of Rpb9 to Rpb1 that is very similar for wild-type Rpb9 and Rpb9-(95-97)A, indicating that there is no major defect for binding with the Rpb9 variant. This is not surprising because the area critical for Rpb9 binding appears to reside in residues nearer to the linker between the Zn1 and Zn2 zinc ribbon domains of Rpb9 (30). These results were not compromised by the presence of the 8-amino acid FLAG tag because all of the RPB9 alleles shown in Figs. 2 and 3 contain it.
A Model for Interaction between Rpb9 and the TL-Our initial results with ␣-amanitin were consistent with the idea that Rpb9 could affect TL mobility. However, although a nearly atomic resolution crystal structure of pol II with the TL in the closed position is available (PDB entry 2E2H (35)), the study of TL mobility has been complicated by the lack of a comparable structure in the open conformation. Portions of the TL are not visible in most crystal structures in which the TL is not locked in its closed position interacting with substrate NTP (e.g. PDB entries 1Y1W (36) and 2E2J (35)), suggesting that the open TL may be composed of multiple conformations of similar stability.
Structures of several open TL conformations have, however, been published (36,37), and one of these (PDB entry 1Y1V (36)), in which the open TL is stabilized by the presence of TFIIS, served as a starting point for our analysis. In this structure, Rpb9 amino acids 95-97, as well as nearby residues, are in position to interact with the ends of Rpb1 ␣20 and ␣21, which form part of the "jaw" domain of pol II, as well as the small loop that connects them (Fig. 6A), which we have termed the Rpb1  Fig. 1A were assembled from WCE derived from the four strains shown, which are based on CKY283 and dopY249 (Table 1). A-for-C misincorporation assays were performed as described in the legend to Fig. 1. Three independent experiments were performed, and the error bars for k obs are 95% confidence intervals as described in the legend to Fig. 1.

AL. Major features include interactions between Rpb1-Arg 711
and Rpb9 residues Gln 87 and Met 97 , as well as insertion of the side chain of Rpb9-Met 97 into a hydrophobic region between ␣20 and ␣21, created in part by Rpb1-Phe 714 . In addition, Rpb9-Thr 95 has potential hydrogen bonding with the Rpb9 backbone at amino acids Arg 92 and Gln 90 , which may be important in maintaining the correct orientation of Rpb9-Gln 87 and Rpb9-Met 97 , and Rpb9-Ser 112 forms a hydrogen bond with the backbone carbonyl of AL residue Gln 698 . Importantly, the open TL in 1Y1V is positioned to interact with the N-terminal tip of ␣21 and the AL in a manner that could stabilize the open conformation (Fig. 6B). Notable features of this interaction include a series of hydrogen bonds between Rpb1-Thr 709 , Rpb1-Glu 712 , and the backbone amide nitrogen of TL residue Ala 1090 , as well as van der Waals interactions of AL residue Met 708 within a hydrophobic region between TL residues Val 1089 and Ala 1090 .
Two additional open TL conformations have been reported (37), both of which showed elongation complexes with a full transcription bubble that was stabilized by the presence of TFIIF, although TFIIF was not clearly visible in the crystal structures. One of these structures (PDB entry 5C4X) was obtained with a complete, 12-subunit pol II, whereas the other (PDB entry 5C4J) was obtained with a 10-subunit polymerase missing Rpb4 and Rpb7. In both of these structures, the interactions between Rpb9 and Rpb1 ␣20-AL-␣21 are very similar to those depicted in Fig. 6A. The only notable difference is that in one of them (5C4X), the orientation of Rpb9-Thr 95 is such that Rpb1-Arg 711 makes a hydrogen bond with this residue as well as with Rpb9-Gln 87 . The TL in both of these structures is less open than that in 1Y1V (Figs. 6, C and D), but all three of the open positions share a common feature in which the TL has direct interaction with ␣21.
The model that emerges from this analysis is that Rpb9 may anchor the ends of ␣20, ␣21, and the AL in positions that are favorable for interaction with the open TL. Stabilization of the open conformation provides time for binding of NTP substrate or proofreading of errors without TL interference. This is similar to a model proposed by Walmacq et al. (10), but it does not involve a previously proposed direct interaction between Rpb9 and the TL. Instead, our model requires two interaction interfaces, one between Rpb9 and Rpb1 ␣20-AL-␣21 and a second between the AL or ␣21 and the TL. It should be noted that there are additional interactions that could also contribute to stabilization of the TL in all three of the stable open positions, but Fig.  6 shows only those to which Rpb9 probably contributes. We note that Rpb1 residue Gly 730 (shown in red in Fig. 6) is also in ␣21, and our model provides a plausible explanation for the negative genetic interaction between RPB9 and rpb1-G730D, which is described under "Discussion." If the molecular interactions are as suggested in Fig. 6, mutations that alter them should result in in vivo phenotypes that mimic at least some of the phenotypes of rpb9⌬ strains, and some previously published observations are consistent with this idea. For example, rpb1-S713P, in which the proline would disrupt the N terminus of ␣21 and potentially alter interactions with both Rpb9 and the TL, causes an upstream shift in the transcription start site of ADH1 (38), a trait shared by rpb9⌬ strains (2) and one that is common to mutations that increase the pol II rate of elongation (39). In addition, a screen for suppressors of the downstream shift in the transcription initiation site caused by sua7-1, an allele of the gene that encodes TFIIB, identified an allele of RPB9 that encodes only amino acids 1-106 (6), disrupting potential Rpb9 interactions, such as the hydrogen bond between Rpb9-Ser 112 and the AL (Fig. 6A). Although the TL does not appear to make any direct contact with ␣20, the two helices are maintained in a parallel orientation through a series of hydrophobic interactions along their helical interface, and the positioning of ␣20 probably affects ␣21 as well.
Phenotypes of Mutations That Alter Potential Interaction Interfaces-To test this model, we introduced mutations that are likely to disrupt potential interactions between Rpb9, ␣20-AL-␣21, and the TL. In addition to the previously described rpb9-(95-97)A, we constructed rpb1 alleles (707-709)A and R711A, and rpb1-S713P (38) was also available. In addition, an ongoing genetic screen for mutations within the TL identified alleles rpb1-S1091C, rpb1-S1091W, and rpb1-K1092E, 3 and they were also chosen for analysis.
Mutations were characterized using three phenotypes of rpb9⌬ strains, decreased growth, hypersensitivity to mycophenolic acid (MPA S ), and suppression of galactose sensitivity conferred by the gal10⌬56 background (Gal R ). MPA is a potent inhibitor of IMP dehydrogenase, an enzyme important in purine biosynthesis. It has been widely assumed that MPA hypersensitivity in yeast is indicative of a defect in transcription elongation that is exacerbated when nucleotide concentrations are low. However, it has been shown that hypersensitivity is almost always a much more specific defect that results from the failure of low concentrations of GTP to cause induction of the yeast IMD2 gene (40), which encodes a relatively MPA-resistant form of IMP dehydrogenase (41). Induction is caused by a shift to a downstream transcription initiation site that is required to generate the long, stable transcript that includes the IMD2 ORF (42). Thus, strains with mutations that prevent the efficient shift from the upstream start site (which begins with GTP) to the downstream site (which begins with ATP) when GTP concentrations are low become MPA-hypersensitive. Recent evidence suggests that one (but not the only) determi-3 C. Kaplan, unpublished observation. The y axis was normalized to the individual reaction end points to emphasize relative rates. Error bars on each data point indicate the S.D. for three independent experiments. A half-time for each reaction was estimated as described under "Experimental Procedures" based on the empirical observation that the reactions proceeded with a close fit to a single exponential equation. Elongation rates and 95% confidence intervals were determined as described under "Experimental Procedures." nant of transcription initiation site selection in yeast is the catalytic activity of pol II: the faster the catalytic rate, the more hypersensitive to MPA (39). The Gal R phenotype assesses transcriptional interference between GAL10 and GAL7 (43). These genes are located in tandem on chromosome II and encode UDP-glucose-4-epimerase and galactose-1-phosphate uridylyltransferase, respectively, which are required for galactose to enter the glycolytic pathway. A small deletion in gal10⌬56 compromises its major polyadenylation signal, and the resulting read-through transcription interferes with transcription initiation at GAL7. This leads to a build-up of galactose 1-phosphate, one of the substrates of the uridylyltransferase encoded by GAL7, which is toxic to yeast. Toxicity is modulated by certain mutations in pol II (including rpb9⌬; see Fig. 7B) as well as mutations in pol II elongation factors and mRNA processing factors (43)(44)(45)(46)(47). No specific molecular defect can necessarily be inferred from altered galactose sensitivity, but it is a readout that is directly related to transcription and is characteristic of rpb9⌬ strains.
Although only rpb1-(707-709)A had a significant growth defect, many of the mutations on either side of the potential interaction interfaces between Rpb9, ␣20-AL-␣21, and the TL resulted in detectable MPA sensitivity (Fig. 7A). These ranged from the very slight MPA S phenotype caused by rpb1-R711A to the previously observed (5,48) severe phenotype of rpb9⌬ strains. Not surprisingly, given its fast elongation rate, rpb9-(95-97)A also conferred an easily detectable MPA S phenotype, as did rpb1-S713P. Some MPA S phenotypes were substitutionspecific for certain residues; TL allele rpb1-S1091C conferred moderate MPA hypersensitivity, whereas no hypersensitivity was detected with rpb1-S1091W. Similarly, rpb9-(95-97)A and rpb1 alleles (707-709)A, R711A, S1091W, and S713P all have some degree of Gal R phenotype. These results are consistent with the prediction that mutations at each of the proposed interaction interfaces should result in phenotypes shared by rpb9⌬.
To assess possible biochemical phenotypes of these pol II alleles, we examined in vitro elongation and misincorporation rates. Pol II immobilized from WCE of rpb1-(707-709)A, rpb1-R711, and rpb1-S1091C strains all share the fast overall elongation rate (Fig. 8A) and fast misincorporation rate (Fig. 8B) of pol II⌬9 and pol II- (Rpb9-(95-97)A). In addition, just as with pol II⌬9 and pol II- (Rpb9-(95-97)A), the presence of saturating amounts of ␣-amanitin decreased the misincorporation rate of all of the polymerases with mutations outside of the TL to near that observed with wild-type pol II (Fig. 8B), consistent with the idea that all of these mutations may be affecting a TL-related function.
Epistasis of Anchor Loop and ␣21 Mutations with rpb9⌬-TL mutations that affect the rate of elongation tend to be additive when combined (39). Fast, gain-of-function alleles combined with slow, loss-of-function alleles usually have an elongation rate between the two extremes, similar to what we observed with the combination of rpb9⌬ and rpb1-G730D (Fig. 2), whereas combinations of two fast or two slow TL alleles are often lethal. In contrast, when we combined rpb9⌬ with the fast rpb1 alleles that we identified, the combinations did not result in polymerases that were significantly faster than that caused by rpb9⌬ alone in the A-for-C misincorporation assay (Fig. 9). This epistasis supports a model where the defects imposed by rpb1 alleles R711A, (707-709)A, and S1091C are related mechanistically to the defects caused by rpb9⌬ and rpb9-(95-97)A. This epistasis is not observed with all fast TL variants, because the combination of rpb9⌬ and rpb1-E1103G, a gain-of-function mutation (23) located near one of the hinges of the TL, is lethal (10), an observation that we have independently confirmed (data not shown).

Discussion
The rate of transcription elongation in eukaryotes is fine tuned to the rates of other processes, such as mRNA capping, transcriptional proofreading, splicing, and termination (38), all of which must operate together to generate a functional RNA product. Our experiments, as well as previously published studies (9,10), indicate that the small pol II subunit Rpb9 plays a role in maintaining elongation at a speed that keeps all of these processes in synchrony. The ␣-amanitin sensitivity of the effect of Rpb9 on misincorporation (Fig. 1) as well as the observed increase in NTP sequestration in the pol II active site in the absence of Rpb9 (10) are consistent with the idea that Rpb9 affects pol II activity, at least in part, by delaying TL closure and thereby slowing elongation. However, a plausible biochemical mechanism by which Rpb9 could affect TL mobility has not been described. Our experiments provide genetic and biochemical evidence for an indirect mechanism by which Rpb9 affects TL mobility by positioning Rpb1 elements ␣20 and ␣21 as well as the AL that connects them into positions that allow the open TL to interact directly with ␣21 or the AL.
In addition to the crystal structures shown in Fig. 6, a molecular dynamics study of TL mobility (49) has identified numerous potential interactions between the TL and ␣21 that may be important. In this study, the open TL conformation was FIGURE 5. Association of Rpb9-(95-97)A with pol II. Pol II was immobilized with Ni 2ϩ -NTA beads from dopY127 strains carrying a centromeric plasmid with the RPB9 alleles shown. Bead-bound proteins were dissolved in SDS and separated by SDS-PAGE. Rpb9 and Rpb1 were identified as described under "Experimental Procedures" by Western blotting with anti-FLAG antibody and anti-Rpb1 C-terminal domain antibody, respectively. The Rpb9/Rpb1 ratio was normalized to that of cells containing wild-type RPB9. Lane 1 contains molecular weight markers as indicated. All images were derived from the same exposure of the same blot. JULY 8, 2016 • VOLUME 291 • NUMBER 28 assigned starting with a crystal structure in which the TL conformation was not completely defined (PDB entry 2E2J (35) A role for Rpb9 in establishing a correct orientation for Rpb1 ␣21 might at least partially explain some other phenotypes of rpb9⌬ yeast strains. For example, the ␣21 mutation rpb1-G730D has a strong negative interaction with RPB9 (29) (see Fig. 2). Gly 730 is in a small hydrophobic pocket that can accommodate neither the size nor the charge of aspartate, and it is likely that the G730D substitution alters the position of the C-terminal end of ␣21 and perhaps the orientation of the entire helix. It is easy to imagine that such a reorientation could restrict TL movement, resulting in the slow rate of elongation characteristic of pol II-(Rpb1-G730D). It is also possible that removal of the anchoring interactions with Rpb9 (as in strains containing rpb9⌬ or rpb9-(95-97)A) could allow ␣21 to adopt a position that is less restrictive to TL movement, consistent with suppression of rpb1-G730D by rpb9⌬ and rpb9-(95-97)A (Figs.  2 and 3).

Rpb9 and Pol II Trigger Loop Mobility
Rpb1 ␣21 also plays a role in TFIIS-mediated RNA cleavage of stalled or backtracked elongation complexes. In order to FIGURE 6. Model for indirect interaction of Rpb9 with the pol II TL. A, interactions between Rpb9 and Rpb1 ␣20-AL-␣21. Rpb9 is depicted in orange and Rpb1 in gray. Hydrogen bonds and van der Waals interactions are depicted in cyan and magenta, respectively. Rpb1-G730 is shown in red. The AL between Rpb1 ␣20 and ␣21 is composed of amino acids Ala 697 -Met 708 . The image is based on PDB entry 1Y1V (36), in which the open TL is stabilized by TFIIS (not shown). B, interactions between the TL and ␣21 in PDB 1Y1V. Rpb9-M97 and Rpb1-R711 are shown to provide the orientation of this image compared with that in A. Colors are identical to those in A. C, interactions between the TL and ␣21 in PDB entry 5C4X (37). This structure was obtained with a complete, 12-subunit pol II in which a transcription bubble is stabilized by TFIIF. The image is oriented in a position nearly identical to B. Colors are as described in A. D, interactions between the TL and ␣21 in PDB entry 5C4J (37). This structure was obtained with pol II that is missing Rpb4 and Rpb7 and, like 5C4X, contains a transcription bubble stabilized by TFIIF. The image is oriented in a position nearly identical to B. Colors are as described in A. ␣20 in 5C4X and 5C4J is slightly extended compared with 1Y1V, and the AL is composed of Rpb1 amino acids Asn 700 -Met 708 . This figure was prepared with UCSF Chimera version 1.10 (60).
facilitate cleavage, a long arm of TFIIS is inserted into the active site (36). This arm contains two acidic residues (Asp 290 and Glu 291 in S. cerevisiae) that chelate a Mg 2ϩ ion and reprogram the pol II active site for cleavage (50). The pathway of the TFIIS arm into the active site is immediately adjacent and parallel to ␣21, with salt bridge or hydrogen bond interactions between TFIIS and ␣21 at Rpb1 amino acids Arg 720 , Asp 727 , Arg 731 , and Glu 734 (36) (PDB entry 1Y1V). For the Thermus aquaticus RNA polymerase, it appears that the TL and Gre, a protein that stimulates cleavage of backtracked ECs in a manner very similar to that of TFIIS (51), compete for position in the active site (52). Thus, it seems plausible that the effects of Rpb9 on TFIIS-mediated cleavage (4,9,30,53,54), which occur without affecting the affinity of TFIIS for pol II (16), may be due, at least in part, to positioning of ␣21, which is important for both the TL and TFIIS to access the active site.
Our work strongly supports the idea that Rpb9 indirectly affects TL mobility. Our results also hint at the multifunctionality of Rpb9, with C-terminal functions relating to the TL and N-terminal functions supporting additional roles that may be FIGURE 7. Growth phenotypes of mutations at potential interaction interfaces. Strains based on CKY283 or dopY249 with the genotype shown were transformed with plasmids derived from pRS415 (SC-Leu) or pRS413 (SC-His) carrying the indicated alleles of RPB9 or RPB1. A, MPA hypersensitivity. Strains were grown to saturation in appropriate media, and serial 10-fold dilutions were spotted onto solid media containing the indicated concentrations of MPA. Growth after incubation for 2 days at 30°C is shown. B, Gal R phenotype. Strains were grown to saturation in media selective for plasmid retention, and serial 10-fold dilutions were spotted onto solid YP Raf or YP Raf/Gal medium. Growth after incubation for 3 days (YP Raf) or 4 days (YP Raf/Gal) at 30°C is shown. more directly related to the effect of Rpb9 on cell growth or the genetic interactions between RPB9 and genes encoding components of certain chromatin-modifying complexes (48,54). Future structural and functional work along the lines of our approaches here will illuminate the molecular nature of these possible Rpb9 functions.

Experimental Procedures
Yeast Strains and Media-Yeast strains used in this study are described in Table 1. Strain BJ5464 His-Bio (generously pro-vided by Mikhail Kashlev (55)) contains a His 6 tag on Rpb3, which allows pol II to be immobilized from WCE with Ni 2ϩ -NTA beads. CKY283 (23) contains a TAP tag on Rpb3, which allows pol II to be immobilized from WCE with IgG-agarose beads. The RPB9 ORF was deleted from BJ5464 His-Bio and CKY283 (rpb9⌬::KANMX4) by PCR-mediated one-step gene disruption (32,33) to generate strains dopY127 and dopY249, respectively. Yeast media and propagation used standard procedures (56).
Construction of rpb9 and rpb1 Alleles-Missense alleles of RPB9 were constructed in the context of various pRS4XX plasmids (33,57) by PCR-mediated site-directed mutagenesis (58). The rpb9  allele, which encodes only the N-terminal 59 amino acids of Rpb9, was constructed by PCR-mediated sitedirected deletion of DNA encoding amino acids 60 -122. The RPB9 sequences in each plasmid included DNA 285 bp upstream and 635 bp downstream of the RPB9 ORF. All of the alleles of RPB9 included an N-terminal FLAG epitope tag (DYKDDDDK) inserted between amino acids 1 and 2. All constructs were confirmed by DNA sequencing.
RPB1 alleles R711A and (707-709)A were constructed by site-directed mutagenesis (58), and alleles S1091C, S1091W, and K1092E were isolated as part of an ongoing genetic screen for mutations in the TL. 3 RPB1 allele S713P has been described previously (38). All constructs were confirmed by DNA sequencing.
Preparation of Whole Cell Extract-WCE was prepared as described by Kireeva et al. (55). Protein concentrations in WCE were 30 -45 mg/ml as estimated by absorbance at 280 nm.
Purification of Pol II and Pol II⌬9 -Pol II and pol II⌬9 were purified essentially as described by Kireeva et al. (55) with modifications described by Sydow et al. (59).
Elongation Complex Assembly and in Vitro Transcription-RNA and DNA oligomers were purchased from Integrated DNA Technologies (Coralville, IA). RNA oligonucleotides FIGURE 8. Effects of mutations at potential interaction interfaces on pol II activity. A, elongation rate. Elongation rates were determined as in Fig. 4 with ECs assembled from WCE prepared from CKY283 (for RPB1 alleles) or dopY249 (for RPB9 alleles) strains carrying centromeric plasmids containing the RPB1 or RPB9 alleles shown. Three separate experiments were performed for each WCE. Elongation rates and error bars, which represent 95% confidence intervals, were determined as described under "Experimental Procedures." B, A-for-C misincorporation. Misincorporation rates were determined as in Fig. 1 with ECs assembled from WCE derived from the strains described in A. Darker shaded bars indicate reactions that contained ␣-amanitin (160 g/ml). Error bars for k obs are 95% confidence intervals as described in the legend to Fig. 1.   FIGURE 9. Epistasis of rpb1 alleles (707-709)A, R711A, and S1091C with rpb9⌬. A-for-C misincorporation rates were determined as in Fig. 1 with ECs assembled from WCE derived from CKY283 (RPB9) strains (light bars) or dopY249 (rpb9⌬) strains (dark bars) carrying centromeric plasmids containing the RPB1 alleles shown. Error bars for k obs are 95% confidence intervals as described in the legend to Fig. 1. (16.7 pmol) were labeled with [␥-32 P]ATP (16.7 pmol) (3000 Ci/mmol, PerkinElmer Life Sciences catalog no. BLU002A) using T4 polynucleotide kinase (Promega catalog no. M4101) in transcription buffer TB(40) (20 mM Tris-HCl (pH 7.9), 40 mM KCl, 5 mM MgCl 2 , 2 mM 2-mercaptoethanol). RNA oligonucleotides were then annealed with the appropriate DNA template strand (16.7 pmol) as described by Kireeva et al. (55). Unlabeled NTPs were obtained from GE Healthcare.
ECs were assembled essentially as described (9,55). For assays using purified polymerase, pol II or pol II⌬9 (250 ng) was added to 25 l of Ni 2ϩ -NTA beads, which had been equilibrated in transcription buffer. For assays using WCEs, 25-50 l of extract (ϳ1.5 mg of total protein) were added to 25 l of Ni 2ϩ -NTA beads (or human IgG-agarose beads for CKY283derived strains carrying the Rpb3 TAP tag). Purified pol II or pol II from WCEs was then allowed to bind the beads for 30 min at room temperature in an orbital shaker (1000 rpm). After the initial incubation, beads were washed with 1 ml of buffer identical to TB(40) except with 1 mM KCl (TB(1000)) for 10 min, followed by three 1-ml washes with TB (40). The preannealed RNA/DNA hybrid (1 pmol) was then added to the beads in a total volume of 100 -150 l and incubated for 10 min, followed by the addition of the non-template strand (10 pmol), which was allowed to incubate for an additional 10 min. ECs were then washed with 1 ml of TB(1000) for 10 min, followed by three washes with 1 ml of TB (40). The final suspension of washed ECs was in 100 -200 l of TB (40).
Transcription was initiated by the addition of substrate NTPs at concentrations described in the figure legends, and aliquots (10 l) of the reaction mixture were removed at various time intervals and added to 10 l of 2ϫ loading buffer (10 M urea, 50 mM EDTA (pH 7.9), 0.005% bromphenol blue, 0.005% xylene cyanol), which stopped the reaction. The resulting extension products were resolved in 13.5% (for longer template as in Figs. 4 and 8A) or 20% polyacrylamide gels as described (9,23). For assays using ␣-amanitin, assembled ECs were incubated with ␣-amanitin (Calbiochem) for 15 min before the addition of NTPs. Quantification of elongation products was performed using a Bio-Rad PharosFX Plus phosphor imager. In order to account for possible sampling inconsistencies that could be caused by non-uniform distribution of beads in the reaction mixture, signals from bands of interest were normalized to the total signal from all bands in the same lane of the gel. For experiments with the longer template (Figs. 4 and 8A), the half-time of the reaction was determined based on the observed progress of the reaction, which empirically was found to closely fit that of a simple first-order reaction (see Fig. 4C). Plots, curve fits, and statistics were prepared using KaleidaGraph (version 4.1; Synergy Software) or Prism (version 6.0; GraphPad Software) graphing software. Errors shown for values of k obs (misincorporation assays) represent 95% confidence intervals derived from the curve fits. Errors in elongation rates (longer template) represent 95% confidence intervals calculated from curve fit-derived errors in reaction half-times.
Western Blotting Analysis-WCEs (300 l) prepared from a dopY127-derived strain carrying an empty vector (pRS413) or the same vector containing DNA encoding either Rpb9 or Rpb9-(95-97)A were incubated with 20 l of Ni 2ϩ -NTA beads for 30 min and washed as described for preparation of ECs. After the washes, most of the final wash buffer was removed to reduce the volume to ϳ20 l, and 20 l of 2ϫ Laemmli buffer (120 mM Tris-HCl (pH 6.8), 20% glycerol, 4% SDS, 0.005% bromphenol blue, and 5% 2-mercaptoethanol) was added. Samples were boiled for 2 min and loaded onto a 4 -20% SDS-polyacrylamide gradient gel along with fluorescently labeled molecular weight size standards. The resolved proteins were electrophoretically transferred to a PVDF membrane overnight at room temperature. The membrane was then blocked with TBS containing 3% nonfat dry milk for 1 h, followed by an overnight incubation at 4°C with mouse anti-FLAG antibody (Sigma catalog no. F1804) (diluted 1:1000 in TBS containing 3% nonfat dry milk). The membrane was then rinsed several times in water, followed by five washes with TBS. Anti-mouse IgG (Sigma catalog no. A9044) (diluted 1:30,000 in TBS containing 3% nonfat dry milk) was then added to the blot and incubated for 30 min at room temperature, followed by five washes with TBS. The blot was visualized by incubation with Supersignal West Femto (Thermo Fisher catalog no. 34095) for 5 min, with the signal detected by a chemiluminescent imager (Bio-Rad ChemiDoc XRS). The membrane was then washed with TBS, reblocked, and probed with anti-Rpb1 C-terminal domain mouse antibody 8WG16 (Covance catalog no. MMS-126R-500) (diluted 1:500 in TBS containing 3% nonfat dry milk), washed, treated with secondary antibody, and imaged as described above.
Growth Phenotypes of Yeast Strains-Strains were grown to saturation at 30°C in appropriate selection media, and aliquots were transferred to one row of a 96-well plate. Six 10-fold serial dilutions (in sterile water) were prepared for each strain, spotted onto the indicated media using a pin spotter (Dan-Kar), and grown for the indicated number of days.
Author Contributions-B. C. K. conducted most of the experiments, analyzed the results, and wrote the paper with D. O. P. K. C. K. purified pol II and pol II⌬9 and conducted and analyzed the experiment in Fig. 1. C. D. K. constructed various TL mutant strains, analyzed results, and provided valuable input during the writing of the manuscript. D. O. P. provided the initial concept for the experiments, analyzed the results, and wrote the paper with B. C. K. All authors reviewed the results and approved the final version of the manuscript.