Downstream sequence-dependent RNA cleavage and pausing by RNA polymerase I

The sequence of the DNA template has long been thought to influence the rate of transcription by DNA-dependent RNA polymerases, but the influence of DNA sequence on transcription elongation properties of eukaryotic RNA polymerase I (Pol I) from Saccharomyces cerevisiae has not been defined. In this study, we observe changes in dinucleotide production, transcription elongation complex stability, and Pol I pausing in vitro in response to downstream DNA. In vitro studies demonstrate that AT-rich downstream DNA enhances pausing by Pol I and inhibits Pol I nucleolytic cleavage activity. Analysis of Pol I native elongating transcript sequencing data in Saccharomyces cerevisiae suggests that these downstream sequence elements influence Pol I in vivo. Native elongating transcript sequencing studies reveal that Pol I occupancy increases as downstream AT content increases and decreases as downstream GC content increases. Collectively, these data demonstrate that the downstream DNA sequence directly impacts the kinetics of transcription elongation prior to the sequence entering the active site of Pol I both in vivo and in vitro.

The sequence of the DNA template has long been thought to influence the rate of transcription by DNA-dependent RNA polymerases, but the influence of DNA sequence on transcription elongation properties of eukaryotic RNA polymerase I (Pol I) from Saccharomyces cerevisiae has not been defined. In this study, we observe changes in dinucleotide production, transcription elongation complex stability, and Pol I pausing in vitro in response to downstream DNA. In vitro studies demonstrate that AT-rich downstream DNA enhances pausing by Pol I and inhibits Pol I nucleolytic cleavage activity. Analysis of Pol I native elongating transcript sequencing data in Saccharomyces cerevisiae suggests that these downstream sequence elements influence Pol I in vivo. Native elongating transcript sequencing studies reveal that Pol I occupancy increases as downstream AT content increases and decreases as downstream GC content increases. Collectively, these data demonstrate that the downstream DNA sequence directly impacts the kinetics of transcription elongation prior to the sequence entering the active site of Pol I both in vivo and in vitro.
DNA-dependent RNA polymerases transcribe DNA to produce RNA and, thus, are responsible for all genetic expression in living organisms. It has long been appreciated that the rate of transcription elongation for all three nuclear eukaryotic RNA polymerases (Pols) 2 is dynamic in nature (1). The dynamic nature of polymerase transcription is critical for correct splicing of mRNA genes by Pol II (1) and for correct folding and cotranscriptional modification of Pol I-synthesized rRNA (2). In vitro and in vivo studies have shown that many alterations in transcription elongation rate are protein-mediated and controlled by both activating and repressing transcription elongation factors and histone modifications (1,(3)(4)(5). It is less clear, however, whether the template DNA sequence itself is a driver of the in vivo diversity of the transcription rate. In vitro studies suggest that RNAP from Escherichia coli exhibits a heterogeneous transcription elongation rate in the absence of transcription elongation factors (6,7). Further in vitro studies characterizing transcription termination have shown that Pol transcription through A-rich template DNA tracts results in destabilization of the transcription elongation complex through accumulation of U:dA base pairs in the RNA:DNA hybrid. This destabilization is required for transcription termination (8,9). A study published recently by our laboratory demonstrated that specific DNA template sequences affect Pol I transcription elongation in a nucleotide concentration-dependent manner in vitro and suggested that Pol I is sensitive to DNA sequence elements in vivo (9). Specifically, native elongating transcript sequencing (NET-seq) adapted for Saccharomyces cerevisiae (yeast) Pol I elucidated that Pol I occupancy on ribosomal DNA is heterogeneous, which could imply that DNA sequence impacts pausing and Pol I elongation rate in vivo (9). However, in the presence of other transcription factors and chromatin modifications, it is difficult to say whether the heterogeneity of Pol I occupancy observed in vivo is a direct result of DNA sequence or whether it is mediated by other factors. This study also did not identify consensus DNA sequences that induced pausing in vivo (9).
There has been work done regarding the impact of upstream DNA sequence on both prokaryotic Pol and eukaryotic RNA polymerase II activity (10,11) as well as on the influence of the identity of the incorporated nucleotide during nucleotide addition (12,13), and some studies have also been done to probe the downstream sequence effects on transcription elongation complex properties (14 -24). There have been no previous studies, however, analyzing downstream sequence effects on RNA polymerase I. As ribosome biogenesis is tightly coupled to the rate of transcription elongation (2,25,26) and implicated in a number of diseases (27)(28)(29), it is critical to gain a complete understanding of the mechanisms that control transcription by RNA polymerase I.
In this study, we characterize the influence of downstream DNA sequence on the transcription elongation properties of Pol I from yeast both in vitro and in vivo. In vitro analyses using a scaffold-based DNA template reveal that yeast Pol I pauses at a higher frequency in the context of increased AT downstream sequence, similar to previous observations with Pol II (15). Interestingly, we observed that AT-rich downstream sequences also repress RNA cleavage by Pol I. Through mutational analy-sis, two distinct regions of Pol I that are structurally close to the downstream DNA were altered to probe whether they sensitized the polymerase to downstream DNA sequence. Loss of sensitivity to distal AT-rich regions was observed when the gene that encodes the nonessential subunit responsible for proofreading and RNA cleavage, subunit A12.2 (A12), of Pol I was deleted. Further, NET-seq revealed that this downstream DNA sequence effect is conserved in vivo. Together, these findings demonstrate that Pol I physically interrogates the downstream DNA sequence (at least in part through the A12 subunit) and that these sequence elements directly influence rRNA synthesis both in vivo and in vitro.

Single-nucleotide addition experiments with an AT-rich downstream sequence alter Pol I pause escape and dinucleotide production
Our laboratory has previously published detailed characterizations of Pol I transcription elongation using a synthetic, 56-nucleotide-long DNA transcription template containing high-GC downstream DNA (30 -32). In these previously published experiments, we built elongation complexes by incubating Pol I with a preannealed RNA:DNA template hybrid (Fig. 1A). We then added the nontemplate strand to form the elongation complex bubble (Fig. 1B). We added radioactive [␣-32 P]CTP (Fig. 1C) to produce a radiolabeled elongation complex (Fig.  1D). A GC dinucleotide was produced during labeling, but this reaction was quenched with excess EDTA (to chelate the Mg 2ϩ cofactor and inhibit Pol I RNA elongation and cleavage) prior to elongation complex mixing with ATP. Because this reaction was quenched prior to mixing with ATP, the signal of the GC species was constant throughout all time courses (Fig. 2B). After the radiolabeled elongation complex was synthesized, saturating [ATP] was added to extend the radiolabeled 10-mer RNA by one nucleotide to produce an 11-mer RNA species over time under rapid mixing conditions in a chemical quenched-flow instrument (Fig. 2B). The 11-mer species was subsequently cleaved by the intrinsic cleavage activity of Pol I to produce an unlabeled 9-mer species and a radiolabeled dinucleotide species (denoted as CA) (Fig. 2B). Heparin was included in reactions to ensure single-turnover conditions. To test the impact of downstream DNA sequence on nucleotide addition by Pol I, we conducted single-nucleotide addition experiments using two DNA templates: one containing a high downstream GC sequence (high GC) and a second with high downstream AT (high AT) content ( Fig. 2A).
In this study, the upstream DNA and RNA remained identical between the high GC and high AT templates, and the nucleotide added to the RNA chain (AMP) was also identical. For the high AT template, the GC downstream nucleotides ( Fig. 2A, green) were switched to AT nucleotides ( Fig. 2A, yellow). The sole difference between the previously reported experiments described above and the high AT template data reported here is the downstream DNA context in which the single AMP nucle- , and then nontemplate DNA is added to form the elongation complex bubble (B). Importantly, the actual size of the bubble has not been determined for Pol I with these templates, thus we do not depict annealing of the upstream or downstream regions of the DNA. However, we have established previously that the nontemplate strand does impact transcription elongation by Pol I; thus, it is likely annealed to the template DNA, we simply cannot say how far upstream and downstream annealing occurs (31). C and D, radiolabeled [␣-32 P]CTP is added to solution (C) to form a radiolabeled 10-mer RNA species (D).

Downstream DNA influences Pol I
otide is added. Two clear differences were observed when transcription was executed with the high AT DNA template compared with the high GC template. First, there was a decrease in the radioactive signal amplitude of the 11-mer species relative to the 10-mer species for the high AT template (Fig. 2C) compared with the high GC template (Fig. 2B). This was a result of persistent abundance of 10-mer species signal over the length of the time course for the high AT template (Fig.  2C) relative to the high GC template (Fig. 2B). Second, there was a decrease in the radioactive signal amplitude of CA dinucleotide produced from cleavage of the 11-mer species with the AT template ( Fig. 2C) compared with the high GC Template (Fig. 2B).
Model-independent fitting of the high AT and high GC time courses displayed a decreased observed rate constant (11-mer k obs2 and CA dinucleotide k obs1 ) for dinucleotide production for the high AT template compared with the high GC template (Fig. 2, D and E, respectively). Interestingly, the rate constant describing nucleotide addition (k obs1 ) was not statistically significantly different for the high AT template (Fig. 2D, gold) compared with the high GC template (Fig. 2D, black), suggesting that the persistent 10-mer species and the extended 11-mer species may be resultant of two distinct populations of polymerases. Importantly, it is unclear from these studies whether the polymerases that did not extend the 10-mer RNA are actively engaged in the elongation complex or whether they are terminated or arrested on the DNA. Collectively, these observations suggest that downstream DNA interacts with the polymerase and alters nucleotide addition and cleavage activity by RNA polymerase I. was measured with polymerases on templates with high downstream GC content and high downstream AT content. B and C, 28% denaturing urea PAGE gels reveal an increase in 10-mer species throughout the high AT time course (C) compared with high GC (B), concurrent with a decrease in CA dinucleotide production throughout the high AT time course (C) compared with high GC (B). D and E, quantification of high GC (black) and high AT (gold) time courses for both the 11-mer (D) and the CA dinucleotide species (E) (n3, error bars represent standard deviation of the mean). Data were fit to the sum of two exponentials for the 11-mer species and a single exponential for the CA dinucleotide species.

Downstream DNA influences Pol I Distal and proximal GC-rich downstream sequences have unique effects on dinucleotide production and 10-mer abundance, respectively
To probe which region of the downstream AT-rich sequence is responsible for perturbing Pol I activity in the high AT template used in Fig. 2, we altered the ϩ2 to ϩ3 nucleotides to AT in one template (which will be referred to as proximal AT, dark green) and the ϩ5 to ϩ7 GC-rich sequence to AT in a second template (denoted as distal AT, light green) (Fig. 3A). The distal AT template increased the radioactive signal abundance of the 10-mer species compared with the high GC template 10-mer species over the length of the time courses (compare Fig. 3B with Fig. 2B). The distal AT template also caused a slight reduction in the radioactive signal amplitude of the dinucleotide species compared with the high GC template (compare Fig. 3B with Fig. 2B). The proximal AT template had the same relative was measured with polymerases on templates with proximal and distal high downstream AT content. B and C, 28% denaturing urea PAGE gels reveal an increase in 10-mer species throughout the distal AT time course (B), whereas a decrease in CA dinucleotide production is noted in the proximal AT time course (C). D and E, quantification of Distal AT (D and F, light green) and Proximal AT (E and G, dark green) time courses compared with High GC (black) time courses for both the 11-mer (D and E) and the CA dinucleotide species (F and G) (n 3, error bars represent standard deviation of the mean). Data were fit to the sum of two exponentials for the 11-mer species (D and E) and a single exponential for the CA dinucleotide species (F and G).

Downstream DNA influences Pol I
amount of 10-mer compared with the high GC template but resulted in a decrease in abundance of the cleaved CA dinucleotide species (compare Fig. 3C with Fig. 2B). Model-independent fitting of the dinucleotide CA species to a single exponential function revealed that the observed rate constants for the proximal AT template were significantly slower than those of the high GC template (Fig. 2G). Further, although the amplitude of cleaved product is lower for the distal AT template, the rate constant describing cleavage is within error of the high GC template (Fig. 3F, k obs1 ).
Collectively, these observations suggest that the ϩ2 to ϩ3 AT sequence is primarily responsible for impairing nucleolytic cleavage by Pol I, whereas the ϩ5 to ϩ7 AT sequence drives the observed amplitude suppression of the 11-mer species. This initially seemed surprising, as one might expect the ϩ2 to ϩ3 region to have a greater direct effect on stability of the transcription elongation complex (if the polymerase falls out of the elongation complex) or backtracking and arrest of the polymerase (if the polymerase is still engaged but inactive) than the ϩ5 to ϩ7 region, as it is closer to the active site. These observations are, however, consistent with previously observed downstream sequence effects of downstream AT-rich regions on human Pol II (15) and E. coli RNAP (16). Structural studies of the downstream DNA transcribed by Pol I suggest that the ϩ5 to ϩ7 region of the DNA is potentially annealed (33), which implies that one or more domains of the polymerase may be interacting with double-stranded downstream DNA in a sequence-dependent manner.

The high AT and proximal GC templates have a small destabilizing effect on elongation complex stability compared with the distal GC and high GC templates
Although backtracking and arrest are known to be linked to a decrease in elongation complex stability (9), it is important to distinguish whether the persistent 10-mer species in high AT and distal AT template time courses is the result of either halted or disengaged polymerases. In these elongation complex stability experiments, radiolabeled elongation complexes are assembled (exactly as for single-nucleotide addition assays described previously) and mixed with RNase A at 40 mM KCl (the same concentration of salt as used in single-nucleotide addition assays) (34). Elongation complexes that fell apart released free RNA that was cleaved by RNase A to produce a 7-mer species, whereas intact elongation complexes protected the RNA from RNase A activity and resulted in an uncleaved 10-mer RNA species (Fig. 4A).
We executed elongation complex stability experiments with all four DNA templates (high GC, high AT, proximal AT, and distal AT) (Fig. 4B). A slight loss of elongation complex stability was observed for all templates on the timescale where singlenucleotide addition assays are executed, as noted by RNase-dependent cleavage of the 10-mer RNA to a 7-mer species (Fig.  4B). A similar trend was observed for the decreased elongation complex stability and persistence of the 10-mer species, as the elongation complexes with high AT (Fig. 4C, gold) and distal AT (Fig. 4E, light green) templates appeared to be the least stable, although most time points were still within error of elongation complexes bound to the high GC template (Fig. 4, C-E, black). Importantly, the difference in elongation complex stability observed between the high AT, distal AT, and high GC templates cannot account for the large population of 10-mer species observed throughout single-nucleotide transcription time courses with high AT and distal AT templates (Figs. 2C and 3B). For example, elongation complexes with the high AT template were ϳ10% less stable than with the high GC template (Fig. 4C), but the amplitude change for the high AT template compared with the high GC template was ϳ35% (Fig. 2D). These elongation complex stability data suggest that, although a small fraction of the increased 10-mer species can be accounted for by elongation complex dissociation, the majority of the population of polymerases remain engaged and are either arrested or stably paused on the DNA.

Mutations of potential downstream interacting regions suggest that both the trigger loop and the A12 subunit of Pol I contribute to downstream sequence effects
We tested the AT-rich template on previously described Pol I mutants that have been implicated in Pol I backtracking and arrest (the A12.2 subunit) (35) and that have been proposed to contact the downstream DNA in conserved regions of Pol II (the trigger loop domain of the A190 subunit) (15,36). The A12.2 subunit directly contributes to transcription elongation complex stability and backtrack recovery (32, 34,35). We tested whether Pol I lacking the entire A12 subunit would be more or less likely to arrest as a 10-mer species in single-nucleotide transcription elongation experiments. In these single-nucleotide transcription experiments on the AT-rich template, there was no observed suppression of the 11-mer versus 10-mer ratio with the ⌬A12 mutant polymerase (Fig. 5A), in contrast to observations made with WT Pol I (Fig. 2D). These data suggest that the A12 subunit may be necessary for the polymerase to sense the AT content of the downstream DNA and pause/arrest. Whether this sensory role is through direct contact with the DNA or through structural reorganization of the leading edge of the polymerase remains to be determined.
Structural data for Pol I and Pol II suggest that the trigger loop may also interact with the downstream DNA sequence (15,33). To test whether the trigger loop of Pol I sensitizes the polymerase to downstream DNA sequence, we executed singlenucleotide addition transcription experiments with a Pol I trigger loop mutant to measure whether putative changes in trigger loop flexibility altered observed AT-dependent downstream sequence effects (Fig. 5B). These experiments were done with Pol I trigger loop mutant rpa190-E1224G (30,37). This mutant has been shown to decrease the elongation rate in Pol I, and it has been hypothesized that this point mutation results in altered flexibility and dynamics of the trigger loop (30,36,37). In single-nucleotide addition time courses with E1224G Pol I on the high AT template, there was an increase in 10-mer species over the length of the time course (resulting in suppression of 11-mer species), but there was no significant change in the amplitude of CA dinucleotide produced (Fig. 2E). The lack of RNA cleavage suppression in single-nucleotide transcription by E1224G Pol I on the AT-rich template suggests that a polymerase with altered trigger loop flexibility may be insensitive to the ϩ2 to ϩ3 AT-rich region of the downstream DNA; thus, the Downstream DNA influences Pol I The radiolabeled elongation complex is mixed with RNase A and allowed to incubate over time. Intact elongation complexes result in a 10-mer species, whereas disengaged elongation complexes result in free RNA that is rapidly cleaved by RNase A to a 7-mer species. B, representative time courses for high GC, high AT, proximal AT, and distal AT templates. C-E, quantification of elongation stability time courses for high AT (C, gold), proximal AT (D, dark green), distal AT (E, light green), and High GC (C-E, black) templates. Error bars represent standard deviation of the mean; n ϭ 3 (C-E).

Analysis of NET-seq data reveals conservation of AT-rich downstream DNA effects in vivo
Although in vitro studies provide a detailed understanding of the potential impacts of downstream DNA sequence on polymerase activity, they do not assess in vivo impacts of these sequences. Our laboratory previously established NET-seq as a tool for quantifying relative Pol I occupancy in vivo in yeast (9). In brief, Pol I elongation complexes were immunoprecipitated from flash-frozen yeast cultures, and the nascent rRNA transcripts were isolated. The 3Ј ends of these transcripts corresponded to the active site of their elongation complexes with single-nucleotide accuracy. The nascent RNAs were purified and used to generate libraries for next-generation sequencing.
The reads from these libraries provided a cumulative snapshot of Pol I occupancy throughout the 35S ribosomal DNA gene (Fig. 6A). In the context of our in vitro findings, increased Pol I occupancy can be interpreted as increased dwell time before addition of subsequent nucleotides (consistent with amplitude suppression of the 11-mer species in our in vitro studies) or sequence-dependent backtracking and cleavage (consistent with dinucleotide production). NET-seq cannot differentiate between these two mechanistic explanations.
To determine whether there is a relationship between Pol I occupancy and downstream sequence, we first observed the downstream sequence content of the highest-occupancy positions. We then determined the top 2.5% of positions by occupancy (n ϭ 172) for one replicate and compiled the sequences into a sequence logo (Fig. 6B). Within these positions, we observed a similar predominance of A and T residues in the last

Downstream DNA influences Pol I
encoded nucleotide position compared with the previous NETseq study. Positions with G residue enrichment at the last encoded nucleotide also represent the minority within the top 2.5% position cohort, although not to the degree observed previously. The NET-seq libraries generated for this study have 10-fold greater read depth compared with the previous study, which may account for this difference (9). Strikingly, we observed a noted predominance of A and T residues directly downstream of these high-occupancy positions, suggesting that downstream A/T enrichment affects Pol I in vivo just as it does in vitro (Fig. 6). This enrichment was observed in the previous study, although it was not further investigated. This sequence enrichment was observed in all three replicates (Fig. S1). Next we sought to determine whether the exact sequences that were assayed in vitro above similarly affect Pol I in vivo. Unfortunately, there are not enough of these sequences within Pol I's limited transcriptome to support statistical analysis of the full high AT sequence (ϩ2 to ϩ3 concurrent with ϩ5 to ϩ7 high AT). Instead, we observed the effects of proximal (positions ϩ2 to ϩ3) and distal (positions ϩ5 to ϩ7) downstream enrichment of T and G residues on Pol I occupancy. We previously observed a systematic difference in occupancy between residues in gene regions (Fig. 6A, cyan) and spacer regions (Fig. 6A, salmon). To avoid bias between occupancy and gene region, the occupancy values in each region were normalized to the sum of occupancy within that region for the following analysis. We observed increased occupancy in positions with proximal downstream T residues relative to positions with proximal downstream G residues (Fig. 7A). Likewise, we observed increased occupancy in positions with distal T residue enrichment relative to those with distal G enrichment (Fig. 7B). Crucially, the difference between the distal enrichment groups was greater than between the proximal enrichment groups. These data suggest that distal T residue enrichment downstream affects Pol I in vivo just as it does in vitro, increasing dwell time prior to subsequent nucleotide addition steps. These trends held for all three replicates of our NET-seq data (Fig. S2).
To determine whether these phenomena occur gene-wide (as opposed to the smaller subsets shown in Fig. 7), the positions in the 35S ribosomal DNA gene were segregated by relative enrichment of each nucleotide in the ϩ1 to ϩ7 base-pair window downstream. Positions with high (four or greater) A (Fig. 8A) and T (Fig. 8D) enrichment in the 7-bp downstream window show significantly higher Pol I occupancy compared with positions with no A or T residues downstream, respectively. We observed the opposite trend for C and G nucleotides. Positions with any C (Fig. 8B) or G (Fig. 8C) nucleotides in their downstream DNA showed significantly decreased occupancy compared with positions with none (with the exception of the highest level of G enrichment). These trends were also observed in the other two replicates of our NET-seq data (Figs. S3 and S4). Collectively, these data show increased occupancy at positions with high A/T enrichment directly downstream and decreased occupancy at positions with high C/G enrichment directly downstream. These observations indicate that Pol I occupancy is affected by downstream sequence content in vivo just as it is in vitro. Although the difference between polymerase pausing and increased dwell time because of polymerase backtracking and cleavage cannot be distinguished in these data, either interpretation (or some combination of both) is in agreement with our in vitro findings.

Downstream sequence impacts transcription elongation properties and elongation complex stability of RNA polymerase I
Significant effects on transcription complex pausing and dinucleotide production by Pol I occur in response to changes Each box represents the interquartile range for each enrichment group, with the horizontal line indicating the median value. The notched portion of each boxplot corresponds to the 95% confidence intervals for the median value. The lower whisker extends to the smallest observation greater than the lower quartile value minus the interquartile range multiplied by 1.5. The upper whisker extends to the largest observation smaller than the upper quartile value plus the interquartile range multiplied by 1.5 (51). Pol I occupancies in each region (see Fig. 5A, diagram) were normalized to the sum of occupancy in that region. Groups were compared by Wilcoxon rank-sum test. The number of observations in each enrichment group is listed above each boxplot, along with the p value for the Wilcoxon rank-sum test (****, p Ͻ 5 ϫ 10 Ϫ5 ).

Downstream DNA influences Pol I
in downstream sequence both in vivo and in vitro. These effects (pausing/arrest and decreased RNA cleavage) correlate with sequences in distinct regions of the downstream sequence. G/C-to-A/T mutations revealed unique molecular consequences when introduced just downstream of the active site (proximal ϩ2 to ϩ3) versus distal mutation (ϩ5 to ϩ7). Downstream AT-dependent pausing has been observed previously in studies using human Pol II (15) and E. coli RNAP (17, 24) but never before for Pol I. Further, as Pol I carries its intrinsic cleavage factor as a bona fide subunit (the A12.2 subunit), this is the first record of AT-rich downstream DNA influence on nucleolytic cleavage. Collectively, these data suggest that downstream DNA has a significant impact on all stages of the transcription elongation cycle and that the polymerase itself interrogates the downstream DNA sequence.

Dependence on downstream sequence appears to be conserved among polymerases
Although in vivo studies have not yet revealed whether Pol II has similar responses to the downstream DNA sequence, it is clear that Pol I and II have similar responses to the content of downstream DNA in vitro. Previous literature has demonstrated that E. coli RNAP is likewise sensitive to downstream DNA sequence (17). A recent study of E. coli RNAP demonstrated that the DNA template sequence can result in elemental pausing where a subpopulation of polymerases exist in a halftranslocated, backtracked state (80%), whereas a separate subpopulation of polymerases were capable of rapidly bypassing the pause state (20%) (24). These findings are strikingly similar to our observations here of a stably paused/arrested population of Pol I on the high AT template. These previously published E. coli RNAP data are further supported by structural data that demonstrated that elemental paused elongation complexes can (but do not always) result in a distinct tilted rearrangement of the RNA:DNA hybrid (24,38). Our laboratory has demonstrated previously that, on the high GC template in single-nucleotide addition experiments, the chemical step is rate-limiting, substantiated by the observed coupling of [ATP] to the chemical step and by Pol I's sensitivity to a slowly hydrolysable nucleotide analog (Sp-ATP␣S) (30, 31). However, model-de-  (51). Each enrichment group is compared with the zero-residue enrichment group (reference group, REF) by Wilcoxon rank-sum test. The number of observations in each enrichment group is listed above each boxplot, along with the p value for the Wilcoxon rank-sum test. ns, not significant; *, p Ͻ 5 ϫ 10 Ϫ2 ; **, p Ͻ 5 ϫ 10 Ϫ3 ; ***, p Ͻ 5 ϫ 10 Ϫ4 ; ****, p Ͻ 5 ϫ 10 Ϫ5 .

Downstream DNA influences Pol I
pendent fitting of single-nucleotide transcription experiments requires inclusion of a transient state of polymerases (ϳ20%) that slowly (relative to ATP binding and the chemical step) transitions into an active state (31,32). This Pol I population exists prior to nucleotide binding, and in our experimental strategy, we were insensitive to the translocation state of this slow population of polymerases. Further structural studies of the Pol I elongation complex on templates with varying downstream sequences could further elucidate the physical nature of the observed slow population of kinetically paused polymerases.
Previous work by Landick and co-workers (15) showed that that AT-rich downstream DNA results in an increase in stable pause and arrest by human Pol II compared with DNA with greater GC content. In that study, Palangat et al. (15) found that Pol II is likewise specifically sensitive to downstream AT sequences based on the proximity of the AT-rich DNA to the active site. They noted that changes in DNA flexibility and torsion were not responsible for these observations. They further hypothesized that interactions with the trigger loop domain of Pol II were likely responsible to sensitivities in downstream DNA (15). Although we did not observe changes in polymerase pausing with AT-rich downstream DNA with the E1224G Pol I trigger loop mutant, it is possible that the phenotype induced by E1224G on Pol I activity is not severe enough to elicit a change in downstream DNA sensitivity.

Potential interactions between the Pol I trigger loop, the A12 subunit, and downstream DNA
Landick and co-workers (15) altered downstream AT content for human Pol II and observed effects on elongation complex stability, pausing, and arrest with AT-rich downstream DNA. They analyzed structural data for human Pol II and hypothesized that the downstream DNA may be sensed by the trigger loop domain (15). Our laboratory observed that alterations in trigger loop residues in Pol I have an opposite effect when compared with their Pol II counterparts. Nevertheless, we used this well-described Pol I trigger loop mutant to interrogate whether alteration of trigger loop flexibility would result in changes in sensitivity to AT-rich downstream DNA. We observed that E1224G Pol I on the high AT template did exhibit some recovery of dinucleotide cleavage, but the amplitude of the 11-mer species was still suppressed. This observation suggests that the trigger loop of Pol I may be responsible for partially sensitizing the polymerase to the downstream DNA sequence or that the dynamic nature of this domain is considerably different in Pol I compared with Pol II.
Because of the A12 subunit's role in backtracking, termination, and cleavage (35,39,40), it stood to reason that A12 could contribute to interactions between the polymerase and the downstream DNA. When we compared ⌬A12 Pol I transcription elongation time courses on the high GC and high AT sequences, we observed subtle changes in transcription elongation kinetics. A12 contributes to a number of enzymatic properties of Pol I, including nucleotide addition, dinucleotide cleavage, and backtracking (32, 34,35). Rpa12 is also thought to stabilize the association of subunits A49 and A34.5 to Pol I (40). Because the A12 subunit is involved in such a wide array of activities, it is impossible to dissect whether these effects are a result of direct contact with A12 or due to indirect effects on the conformational dynamics of other subunits. However, these data do suggest that A12 is at least in part responsible for sensitization of the polymerase to the downstream DNA. Further characterization of the roles of A34.5/A49 and the C terminus of the A12 subunit will identify individual contributions of these subunits to Pol I pausing and backtracking. Future structural studies may identify the physical nature of the interaction between A12 and the downstream DNA.

Implications of polymerase sensitivity to downstream DNA
It is now clear that polymerases from prokaryotes to eukaryotes display sensitivity to downstream DNA sequence and that this sensitivity has impacts on the entirety of the transcription cycle. It has been known for decades that transcription termination is controlled by DNA/RNA sequence (9). It is reasonable that transcription elongation may be similarly controlled by DNA/RNA sequence. The role of downstream DNA sequence in transcription elongation is important because the elongation rate directly affects overall gene expression by E. coli RNAP, splicing by eukaryotic Pol II, and ribosome biogenesis by eukaryotic Pol I (2). Here we demonstrate that downstream DNA sequence does, in fact, impact all stages of the transcription elongation cycle. Further, sensitivities to particular regions of downstream DNA appear to be conserved between eukaryotic Pol I and Pol II (15). This sensitivity is conserved in vivo, suggesting that downstream DNA sequence can directly impact Pol I-mediated rRNA biogenesis. This sensitivity to downstream DNA could function as a "second code" and directly govern the efficiency of RNA processing, thus having largescale impacts on global gene expression.

Protein purification
rpa190-E1224G Pol I, rpa12⌬ Pol I, and WT Pol I were purified exactly as described previously (34,37,41). Briefly, 50 liters of yeast cells were grown to late log phase (when cell growth begins to slow). Cells were harvested by centrifugation and frozen. Cells were thawed and lysed in breakage buffer (400 mM ammonium sulfate, 100 mM Tris-sulfate (pH 7.9) at 4°C, 20% (v/v) glycerol, 5 mM MgCl 2 5 mM imidazole, 1 mM PMSF, and 0.1% (v/v) Tween 20), and endogenously His-tagged proteins were sequentially purified using a Ni Sepharose 6 Fast Flow column (GE Healthcare) and a HiTrap Heparin HP column (GE Healthcare) followed by a MonoQ 5/50 GL column (GE Healthcare). Proteins were aliquoted and stored in storage buffer (monoQ elution buffer supplemented with 40% glycerol) at Ϫ80°C.

Single-nucleotide in vitro transcription
Single nucleotide addition assays for Pol I and the Pol I mutants rpa190-E1224G and ⌬A12 were performed as de-Downstream DNA influences Pol I scribed previously (31). Briefly, Pol I elongation complexes were preformed in a synthetic elongation complex with 54 nM template DNA prehybridized to 163 nM RNA and 163 nM nontemplate DNA in 40 mM KCl, 20 mM Tris-OAc (pH 7.9 at 25°C), 0.2 mg/ml BSA, and 2 mM DTT. Elongation complexes were radioactively labeled for 10 min with 5 nM [␣-32 P]CTP in 100 M Mg(OAc) 2 , 40 mM KCl, and 20 mM Tris-OAc (pH 7.9 at 25°C). Labeling was quenched by mixing with 1.1 mM EDTA-K 3 and 5 M CTP in 40 mM KCl and 20 mM Tris-OAc (pH 7.9 at 25°C). Labeled elongation complexes were then mixed in a chemically quenched flow over a series of time points with a 2ϫ concentration ATP substrate mixture containing 2 mM ATP, 0.1 mg/ml heparin, and 18 mM Mg(OAc) 2 in 1ϫ concentration 0.2 mg/ml BSA, 2 mM DTT, 40 mM KCl, and 20 mM Tris-OAc (pH 7.9 at 25°C).

NET-seq library preparation
Three NET-seq libraries were prepared with an S. cerevisiae strain bearing a C-terminal triple HA (3ϫHA, His 7 ) tag on the A135 subunit of Pol I as described previously (9). NET-seq read trimming and alignment to the S. cerevisiae genome were performed as described previously (9). Unix Scripts are available upon request.

Net-seq analysis
For Fig. 6A, Pol I occupancy on the negative strand in the 35S ribosomal DNA gene was normalized to the sum of counts on the positive strand of the 5S gene for each replicate and multiplied by 1,000. The median occupancy value from the three replicates for each position is displayed in Fig. 6A, top panel. The heat map in Fig. 6A (directly below the median occupancy plot) was created by determining the enrichment of A/T residues in five-residue increments throughout the 35S ribosomal DNA gene. These windows were then color-coded according to relative A/T enrichment: blue for 0 A/T residues and red for 5 A/T residues.
For Fig. 8, Pol I occupancy on the negative strand in each region (ETS 1, 18S, ITS 1, 5.8S, ITS 2, 25S, and the first 91 bp of ETS 2) of the 35 S gene was normalized to the sum of counts in that region and multiplied by 1,000,000 to account for previously observed region-specific variations in median occupancy (9). Positions were segregated according to enrichment for each nucleotide in the 7-bp window directly downstream (positions ϩ1 to ϩ7). The upper and lower 95% confidence intervals for the median value (corresponding to the upper and lower bounds of the notched portion in each boxplot) were equal to the median value Ϯ (1.58 ϫ IQR/(͌n), where IQR is the interquartile range and n is the number of observations in each group. Each level of nucleotide enrichment was compared via Wilcoxon rank-sum test to the lowest level of nucleotide enrichment (0 residues). Pol I occupancies were log 2 -compressed (using the formula y ϭ log 2 (x ϩ 1)) for ease of y axis visualization. Displayed data (Fig. 8)