Genetic analysis of RNA polymerase I unveils new role of the Rpal2 subunit during transcription

Most transcriptional activity of exponentially growing cells is carried out by RNA Polymerase I (Pol I), which produces a large rRNA precursor. The Pol I transcription cycle is achieved through complex structural rearrangements of the enzyme, revealed by recent structural studies. In the yeast S. cerevisiae the Pol 1 subunit Rpa49, particularly its C-terminal tandem winged helix domain (Rpa49Ct), is required supports both initiation and elongation of the transcription cycle. Here, we characterized novel extragenic suppressors of the growth defect caused by the absence of Rpa49. We identified suppressor mutations on the two largest subunits of Pol I, Rpa190 and Rpa135, as well as Rpa12. Suppressor mutants RPA135-F301S and RPA12-S6L restored normal rRNA synthesis and increased Pol I density on rDNA genes in the absence of Rpa49Ct. Most mutated residues cluster at an interface formed by the jaw in Rpa190, the lobe in Rpa135, and subunit Rpa12 when mapped on the structure of Pol I. Our genetic data in S. cerevisiae suggest a new role for Rpa12 at the jaw/lobe interface during transcription cycle.


Introduction
The nuclear genome of eukaryotic cells is transcribed by three RNA polymerases (Chambon, 1975). RNA polymerase II (Pol II) transcribes most of the genome and is responsible for all messenger RNA production. RNA polymerases III and I are specialized in the synthesis of a limited number of transcripts. RNA polymerase III (Pol III) produces small structured RNAs, including tRNAs and the 5S ribosomal RNA. RNA polymerase I (Pol I) produces a single transcript, the large polycistronic precursor (the 35S pre-rRNA in yeast), processed by multiple successive steps into the mature rRNAs (25S, 18S, and 5.8S in yeast). Despite producing a single transcript, Pol I is by far the most active eukaryotic RNA polymerase, responsible for up to 60% of the total transcriptional activity in exponentially growing cells (Warner, 1999). Pol I rRNA synthesis constitutes the first step of ribosome biogenesis and is a rate limiting process for cell growth. The strongly transcribed rRNA genes can be visualized using the DNA spread method developed by Miller et al, 1969, in which the 35S rRNA genes (rDNA) adopt a "Christmas tree" conformation, with up to 120 polymerases per transcribed gene (Miller & Beatty, 1969). Altered cell proliferation, often associated with a modified rate of rRNA synthesis via the deregulation of Pol I activity, has been associated with various types of cancer (Drygin et al, 2010).
The full subunit composition and structural data are now available for the three nuclear RNA polymerases of the budding yeast Saccharomyces cerevisiae (Fernández-Tornero et al, 2013;Engel et al, 2013); (Hoffmann et al, 2015;Cramer et al, 2001). Pol I contains a core of shared or homologous subunits that are largely conserved in eukaryotes and archaea, as for the other two nuclear RNA polymerases (Werner, 2008). The two largest subunits (Rpa190 and Rpa135) carry the catalytic site. Rpb5, Rpb6, Rpb8, Rpb10, and Rpb12 are shared with Pol II and Pol III, whereas Rpc40 and Rpc19 are only shared with Pol III. This nine-subunit core is associated with the stalk, a structure formed in Pol I by the heterodimeric complex Rpa43/Rpa14, which is involved in docking the essential Rrn3 initiation transcription factor to the enzyme (Peyroche et al, 2000;Yamamoto et al, 1996;Blattner et al, 2011). The Pol I-Rrn3 complex interacts with promoter bound factors, the core factor (CF), forming the initially transcribing complex (ITC) (Engel et al, 2017;Sadian et al, 2017;Han et al, 2017;Keener et al, 1998). Additionally, Pol I and Pol III contain subunits that are functionally and structurally related to Pol II-specific basal transcription factors, called the "Built-in Transcription Factors" (Kuhn et al, 2007;Geiger et al, 2010). Their presence in Pol I and Pol III results in a higher number of subunits, from 12 subunits in Pol II, to 14 and 17 for Pol I and Pol III, respectively, and correlates with substantial transcript production from a few genes (Werner, 2008). The heterodimer formed by Rpa34 and the N-terminal domain of Rpa49 (Rpa49Nt) in Pol I (i.e. Rpc53 and Rpc37 in Pol III) is related to the basal transcription factor TFIIF, and stimulates endogenous transcript cleavage activity (Geiger et al, 2010;Landrieux et al, 2006;Wu et al, 2011). Rpc34 in Pol III and the Rpa49 C-terminal domain (Rpa49Ct) bear a tandem winged helix motif similar to TFIIE (Geiger et al, 2010;Landrieux et al, 2006). Rpa49Ct binds DNA and is involved in initiation and elongation (Pilsl et al, 2016;Geiger et al, 2010). Finally, Rpa12 (Pol I) and Rpc11 (Pol III) C-terminal domains are both directly involved in stimulating endogenous transcript cleavage activity, similar to that of TFIIS for Pol II (Van Mullem et al, 2002;Kuhn et al, 2007).
Yeast genetic studies of Pol III and Pol I "Built-in Transcription Factors" have revealed striking differences, despite their clear similarities. Each Pol III subunit is essential for cell growth, but none of the Pol I "Built-in Transcription Factors" is needed for cell survival.
Deletion of Rpa34 or invalidation of Rpa49Nt, by removing the TFIIF-like heterodimer, has a very mild growth effect in vivo (Gadal et al, 1997;Beckouet et al, 2008;Geiger et al, 2010).
In contrast, full or C-terminal deletion of Rpa49 leads to a strong growth defect at all temperatures, which is more severe below 25°C (Liljelund et al, 1992;Beckouet et al, 2008;Geiger et al, 2010). Full deletion of Rpa12 leads to a strong growth defect at 25 and 30°C, and is lethal at higher temperatures (Nogi et al, 1993). Deletion of the C-terminal extension of Rpa12 abolishes stimulation of intrinsic cleavage, without any detectable growth defect (Van Mullem et al, 2002;Kuhn et al, 2007). Finally, yeast strains carrying the triple deletion of RPA49, RPA34 and RPA12 are viable, but accumulate the growth defects associated with each of the single mutants (Gadal et al, 1997).
Pol I is functional in the absence of Rpa49, but shows well-documented initiation and elongation defects, both in vivo and in vitro (Liljelund et al, 1992;Beckouet et al, 2008;Albert et al, 2011;Pilsl et al, 2016;Gadal et al, 2002;Huet et al, 1975). Restoration of active rRNA synthesis, in the absence of Rpa49, has been used to identify factors involved in initiation and elongation, such as Hmo1 and Spt5, or other Pol I subunits, such as Rpa43 (Gadal et al, 2002;Beckouet et al, 2008;Viktorovskaya et al, 2011). We made use of the spontaneous occurrence of extragenic suppressors of the growth defect of the rpa49 null mutant (Liljelund et al, 1992) to better understand the role of this subunit and reveal new interactions between the various subunits of RNA pol I during transcription. We showed that the suppressing phenotype was caused by specific point mutations in the two largest Pol I subunits, Rpa190 and Rpa135. We identified a small area at the interface between the jaw and lobe modules of the two largest subunits, in which the Rpa12 linker domain is inserted, by mapping the mutations onto the structure of Pol I. Rpa12 turned out to be essential for suppression. Moreover, Pol I carrying a single point mutation in Rpa12 was able to correct the growth defect and initiate transcription in the absence of Rpa49. Overall, our genetic data, gathered in the absence of Rpa49 or Rpa49Ct domain, suggest that the jaw/lobe interface, including the N-terminus of Rpa12, is important for the stabilization of Pol I that allows efficient transcription.

Isolation of extragenic suppressor mutants of the growth defect in absence of Rpa49
We characterized extragenic suppressors of the RPA49 deletion to better understand how cell growth is achieved in the absence of Rpa49. RPA49 full-deletion mutants show a strong growth defect at 30°C and are unable to grow at 25°C. However, spontaneous suppressors have been previously observed (Liljelund et al, 1992). We reproduced this observation and quantified the frequency of occurrence of individual clones able to grow at 25°C. There was a low frequency of colony occurrence, comparable with the spontaneous mutation rate of a single control gene (CAN1; < 5.10 -6 ). We isolated more suppressors by irradiating the cells with UV light. UV irradiation, resulting in a survival rate of approximately 50%, increased the frequency of suppressor mutations by approximately 10 fold. We identified clones that grew at 25°C after three days and selected individual colonies, called SGR for suppressor of growth defect of RPA49 deletion, with various growth rates. We ranked SGR from 1 to 186 based on their growth rates at 25°C; SGR1 had a growth rate comparable to the wild-type (WT) condition ( Figure 1A). We crossed the 186 SGR clones with a strain of the opposite mating type bearing only the deletion of RPA49 to obtain diploid cells homozygous for the RPA49 deletion and heterozygous for each suppressor. The restoration of growth of the diploids at 25°C showed that all suppressor phenotypes obtained were fully or partially dominant. We focused on the most efficient suppressor clones, SGR1 and SGR2, and performed tetrad analysis to follow segregation of the observed suppression phenotype. Each suppressor phenotype was linked to a single locus in the genome and neither SGR mutant had a strong growth defect (SGR1 in Figure 1A). We used global genomic mapping of SGR1 and SGR2, derived from "genetic interaction mapping" (GIM) methods (Decourty et al, 2008) (Materials and Methods; Suppl. Figure 1), and found a genomic linkage close to genes encoding the two largest Pol I subunits: RPA135 for SGR1 and RPA190 for SGR2 (Suppl. Figure 1). Sequencing of the genomic DNA revealed that SGR1 bears a double mutation, whereas SGR2 bears a single one (RPA135-I218T/R379K and RPA190-A1557V alleles, respectively). Furthermore, we identified an additional mutant, SGR3, in RPA135, RPA135-R305L. The heterogeneity of the growth induced by strong UV mutagenesis prevented suppressor cloning from the 183 remaining SGR clones.
We next used the dominant phenotype of these suppressors to isolate more alleles which suppress the deletion phenotype of Rpa49 in RPA190 and RPA135. We constructed a library of randomly generated mutants (see Materials and Methods) by propagating plasmids bearing WT RPA135 or RPA190 in a mutagenic E.coli strain. After phenotypic selection of rpa49∆ mutants bearing a mutagenized Rpa190 or Rpa135 subunit at 25°C, each plasmid bearing a suppressor allele was extracted, sequenced, and re-transformed into yeast to confirm the suppressor phenotype. We thus isolated nine novel alleles of Rpa190 and 13 of Rpa135 that were able to restore growth of rpa49 deletion mutant at 25°C (supplementary Table 1). We evaluated the suppression strength based on growth restoration at 25°C relative to WT growth, as for the original SGR strains. Suppressor alleles more effective than SGR1, 2, or 3 were identified using mutagenesis of Rpa190 or Rpa135 (suppl. Table 1). In conclusion, we identified novel alleles of the two largest Pol I subunits as extragenic suppressors of the rpa49∆-associated growth defect.

Rpa190 and Rpa135 mutant alleles can bypass the need of Rpa49 for optimal growth
The growth of the strains bearing one of six alleles (RPA190-E1274K, RPA190-C1493R, RPA190-L1262P, RPA135-R379G, RPA135-Y252H, and RPA135-F301S) was evaluated by a 10-fold dilution test ( Figure 1B), showing good suppression by all in the absence of Rpa49.
Previous genetic studies have isolated other genetic backgrounds that alleviate the growth defect of rpa49∆ at 25°C, such as rpa43-35,326 (Beckouet et al, 2008), decreased rDNA copy number (Albert et al, 2011), Hmo1 over-expression (Gadal et al, 2002), or Spt5 truncations (Viktorovskaya et al, 2011). For all these mutants, rRNA synthesis was only partially restored in the absence of Rpa49 and significant transcription defects remained. Here, we focused on the RPA135-F301S allele, the most effective growth suppressor of the Rpa49 deletion: the rpa49∆ RPA135-F301S double mutant grew almost as well at 25°C as the WT strain ( Figure 1B). In the absence of Rpa49, Rpa34 does not associate with transcribing Pol I (Beckouet et al, 2008). The rpa49∆Ct allele does not have the TFIIE-like module, including the tandem wing helix motif (corresponding to residues 187-415). In strains bearing the rpa49∆Ct allele, the TFIIF module (Rpa34 and Rpa49Nt) remains associated with the polymerase (Liljelund et al, 1992;Beckouet et al, 2008). However, yeast bearing rpa49∆Ct or rpa49 full-deletion have a similar growth defect (Liljelund et al, 1992;Beckouet et al, 2008).
We sought further insight into the effect of the suppressors by integrating the RPA135-F301S point mutation into the endogenous gene in three genetic backgrounds: WT, rpa49∆ (full deletion), or rpa49∆Ct. The growth rate was determined in each of these yeast strains at 30°C, in the presence or absence of RPA135-F301S. The suppressor allele RPA135-F301S had no effect on growth in the WT strain (doubling time of 102 min). The doubling time was 180 min for the rpa49∆ strain and RPA135-F301S suppression restored growth to a doubling time of 135 min. We observed similar suppression on the rpa49∆Ct background (data not shown).

Rpa49Ct in vivo
We used yeast mutant cells with a low and stable number of rDNA repeats to better associate the growth phenotype of the RPA135-F301S allele to nascent rRNA synthesis and Pol I density on transcribed genes in vivo. Almost all rRNA genes in such strains are in the active state and transcribe their rRNA genes with a very high Pol I loading rate (Cioci et al, 2003;Albert et al, 2011). Accordingly, this genetic background is better suited to study variations in the number of polymerase molecules per rRNA gene, as the cells of these strains have a fixed number of rRNA copies which are all in the open chromatin state. We generated three strains on this background (bearing rpa49∆Ct,  and determined the doubling time ( Figure 2A) and de novo synthesis of rRNA ( Figure 2B).
The presence of the RPA135-F301S allele in Pol I effectively compensated the growth defect caused by the absence of Rpa49. Labeling of the nascent rRNA was performed using a 2-min pulse with 3 H adenine. We performed the labeling in three independent cultures because of heterogeneity in the cell cultures of the rpa49∆Ct mutant. RNA synthesis was reduced approximately five-fold for rpa49ΔCter, even under permissive conditions (30°C) (compare Rpa49 is involved in initiation and elongation. We evaluated Pol I density on transcribed genes by performing Miller spreads, the only technique that currently allows the counting of individual Pol I molecules on a single rRNA gene (Miller & Beatty, 1969;Albert et al, 2011).
We previously showed that full deletion of rpa49 resulted in a three-fold decrease of Pol I density per gene in the same genetic background (Miller & Beatty, 1969;Albert et al, 2011). Overall, these results show that the presence of the RPA135-F301S allele in a strain lacking Rpa49Ct restores rRNA synthesis to WT levels and increases Pol I density on rRNA genes, indicative of productive transcription initiation and rRNA synthesis.

Most suppressors are clustered in a hotspot of Pol I
Structural data of Pol I are now available, showing Pol I in an inactive form (Fernández-Tornero et al, 2013;Engel et al, 2013), complexes of Pol I and Rrn3 (Pilsl et al, 2016;Engel et al, 2016;Torreira et al, 2017), complexes of Pol I with other initiation factors (Engel et al, 2017;Han et al, 2017;Sadian et al, 2017), and in elongating forms (Neyer et al, 2016;Tafur et al, 2016), respectively. We mapped Rpa135 and Rpa190 residues that suppressed the growth defect of the mutant strain rpa49∆Ct onto the structure of WT Pol I in which the structure of Rpa49 was determined (Han et al, 2017) ( Figure 3A). Most of the suppressor mutations which provided growth recovery (see supplementary table 1) appeared to be clustered at a specific interface between the two largest subunits, Rpa190 and Rpa135 ( figure   3B), between the lobe (Rpa135 -salmon) and the jaw (Rpa190 -blue). This small region is characterized by the presence of a beta-strand in the structure of Rpa12 (Rpa12yellow: residues 46-51, Figure 3B), following four beta-strands of Rpa190, thus forming a shared five-strand anti-parallel beta-sheet. The Rpa12 beta-strand also faces the Rpa135 lobe domain (residue 252 to 315 of Rpa135), in which six independent mutations were found, including RPA135-F301S. Notably, Rpa135 contains a pattern of three amino acids "DSF" (D299, S300, F301), which are highly conserved among eukaryotic species. Notably, we found suppressor mutations for each of these three amino acids (see supplementary table 1). The side chains of the mutated residues are apparently involved in electrostatic interactions that stabilize this jaw/lobe interface ( Figure 3C). Substituted residues resulted in destabilization of this interface (see supplementary table 1), suggesting a specific rearrangement of the interface lobe/jaw in each mutant.

Point mutations in Rpa12 suppress the growth defect when Rpa49 is missing
We then tested whether mutated alleles of RPA12 itself could behave as suppressors by generating a library of randomly mutagenized RPA12 (see Materials and Methods). Two dominant alleles (RPA12-S6L and RPA12-T49A) efficiently suppressed the growth defect of rpa49∆ and of rpa49∆Ct ( Figure 4A, not shown here for rpa49∆Ct). However, these RPA12 suppressor alleles did not fully restore WT growth when expressed in the rpa49∆ mutant (compare the size of the colonies in Figure 4A.). We integrated the RPA12-S6L point mutation into the genome at its native locus in a low rDNA-copy-number genetic background, with or without rpa49Ct deletion, and analyzed the restoration of rRNA synthesis by pulse labeling ( Figure 4B) and Pol I loading by Miller spreading ( Figure 4C). rRNA synthesis and Pol I loading were largely restored when RPA12-S6L was expressed in strain rpa49∆Ct, similar to that for the RPA135-F301S mutant. The 3D structure of Pol I shows that RPA12-S6L and RPA12-T49A obtained by random mutagenesis are specifically located in the "hotspot" at the jaw/lobe interface. Threonine 49 of Rpa12 is located on the beta-strand (Rpa12 aa 46-51) ( Figure 3B), facing residues D299, S300, and F301 of Rpa135, and Rpa190-E1274 ( Figure 3B). This residue also appears to be highly conserved among eukaryotic species. The second mutation, RPA12-S6L, is located in the N-terminal domain of Rpa12. In conclusion, all point mutations in Rpa190, Rpa135, and Rpa12 detected in the hotspot domain of the jaw/lobe interface can substitute the requirement for Rpa49Ct in vivo.

Rpa14, Rpa34, and the expander/DNA mimicking loop of Rpa190, are not involved in the suppression phenotype
The structural determination of Pol I revealed the presence of an extended loop inside the DNA-binding cleft folded in a "expander/DNA mimicking loop" conformation when Pol I is in an inactive, dimeric form (Engel et al, 2013;Fernández-Tornero et al, 2013). This element is not found in Pol II or III. These residues are inserted between the four beta-strands of Rpa190, corresponding to the mutation hotspot. A small deletion of this Rpa190 domain (1361-1390) results in a slight slow-growth phenotype (Fernández-Tornero et al, 2013). We determined whether the expander/DNA mimicking loop was responsible for the suppression phenotype. We generated a novel allele, rpa190∆loop (deletion of residues 1342-1411), which had no associated growth defect ( Figure 5A). We were unable to generate a viable double mutant when combining this mutation with the rpa49 full deletion. Thus, the DNAmimicking loop is required for Pol I activity in the absence of Rpa49. We next tested whether deletion of this loop influences suppression by the RPA135-F301S allele. Note that the rpa190∆loop combined with RPA135-F301S has no phenotype. There was no difference in the growth of the rpa49∆ RPA135-F301S double mutant and that of the triple mutant rpa49∆ RPA135-F301S rpa190∆loop ( Figure 5A). Thus, the expander/DNA mimicking loop of Rpa190 is not required for suppression, but is required for the viability of the rpa49 deletion mutant.
Rpa34 forming a heterodimer with Rpa49Nt, and Rpa14 being essential in absence of Rpa49, we also introduced RPA135-F301S into yeast strains lacking theses Pol I subunits ( Figure 5B and C). The growth of RPA135-F301S/rpa14∆ and RPA135-F301S/rpa34∆ double mutants was not different from that of the single mutants. However, RPA135-F301S suppressed the growth defect of the viable double mutant, rpa34∆ rpa49∆, lacking both Rpa49 and Rpa34, the heterodimer partner of Rpa49 ( Figure 5B). The double deletion mutant lacking both Rpa49 and Rpa14 was not viable, similar to the rpa190∆loop rpa49∆ double mutant (Gadal et al, 1997). Introduction of the suppressor RPA135-F301S, by genetic crossing, resulted in a triple mutant (rpa14∆ rpa49∆ RPA135-F301S) that could grow, but much more slowly ( Figure 5C).

Rpa12 is required for suppression of the rpa49∆-associated growth defect
We next evaluated whether specific regions of Rpa12 may be involved in the suppression of the rpa49∆-associated growth defect ( Figure 6A). The C-terminal region of Rpa12 (TFIISlike) is inserted towards the active center of Pol I to stimulate intrinsic cleavage activity but is displaced during productive initiation and elongation steps. C-terminal deletion of Rpa12 Rpa12∆Ct allele is unable to stimulate cleavage activity in vitro (Kuhn et al, 2007). Full deletion of RPA12 resulted in a slight growth defect at 24°C, which was stronger at 30°C (Nogi et al, 1993) (Figure 6A-lane 3). In contrast, rpa49 deletion resulted in a growth defect at 30°C, which was stronger at 24°C (Liljelund et al, 1992) (Figure 6A-lane 4). Combining rpa12∆Ct with rpa49∆ resulted in a mildly synergistic phenotype, with a stronger growth defect at both 24°C and 30°C ( Figure 6Alane 5). The double mutant lacking both full Rpa12 and Rpa49 subunits was viable, but had a major growth defect (Gadal et al, 1997) ( Figure 6Alane 6).
Random mutagenesis of RPA12 did not lead to the isolation of RPA12 alleles containing suppressor mutations in the C-terminal domain. We explored whether the C-terminal extension of Rpa12 is necessary for suppression of the rpa49∆ phenotype by introducing the Rpa12 C-terminal truncation into the strain bearing both rpa49∆ and suppressor alleles RPA12-S6L or RPA12-T49A. Thus, RPA12-S6L and RPA12-S6L-∆Ct resulted in similar suppression of the rpa49∆ growth defect ( Figure 6B). RPA12-T49A behaved similarly to RPA12-S6L (data not shown).
We then introduced the RPA135-F301S allele and assessed the suppression phenotype of the strain lacking Rpa49 at 25°C, with or without the entire Rpa12 subunit ( Figure 6C). We constructed a strain with RPA12 under the control of the regulatable pGAL promoter. The growth defect of rpa49∆ was completely suppressed by the RPA135-F301S allele when RPA12 was expressed ( Figure 6C, left panel), whereas suppression mediated by RPA135-F301S was completely abolished in its absence ( Figure 6C, middle panel). Residual growth of the rpa49 rpa12 double mutant was detectable after 10 days ( Figure 6B, right panel) and was the same with or without the RPA135-F301S allele.
In conclusion, we show that Rpa12, particularly its N-terminal portion, is required for the suppression of the growth defect mediated by RPA135-F301S when Rpa49 is depleted.

Rpa12 is required for efficient promoter-dependent transcriptional activity in vitro
The suppressor mutants suggest a functional interaction between Rpa49 and Rpa12. The absence of Rpa49 or its C-terminal extension strongly reduced Pol I loading on rDNA genes (Albert et al, 2011;Beckouet et al, 2008); Figure 2C), indicating an initiation defect in vivo.
The in vitro results of promoter-dependent transcription assays in the absence of Rpa49, particularly of Rpa49Ct, are consistent with the in vivo data, in which their absence results in an almost initiation-inactive form of the polymerase, whereas nonspecific RNA synthesis is not as strongly affected (Pilsl et al, 2016). Miller spread experiments have been previously performed for mutant strains lacking Rpa12 (Prescott et al, 2004), but Pol I occupancy of the rDNA genes was not analyzed, and is difficult to interpret due to the known instability of Rpa190 in the absence of Rpa12 (Van Mullem et al, 2002). However, it is possible to test whether subunit Rpa12 supports transcription initiation in vitro. We thus performed promoterdependent transcription and non-specific transcription assays comparing Pol I, lacking Rpa12, to WT Pol I (Figure 7).
In vitro, transcription on tailed templates was significantly reduced in the absence of Rpa12 when the amount of Pol I was held constant (approximately 14% using 4 nM Pol relative to that of WT Pol I). In promoter-dependent transcription the reduction in transcription efficiency was much more pronounced (approximately 1.2% of that of WT Pol I) indicating that Rpa12 is involved in the initiation step.

Discussion
Here, we investigated how the growth of cells can improve in the absence of Rpa49. We showed that altering a very specific area of Pol I resulted in a functional Pol I molecule lacking this subunit. We propose that the mutated area, which includes a part of subunit Rpa12, undergoes a conformational change that supports the initiation of transcription.

Suppressor mutants are not at the Rrn3-Pol I stalk interface
Our previous studies suggested the specific involvement of Rpa49 in the association and dissociation of Rrn3 from the Pol I stalk (Albert et al, 2011;Beckouet et al, 2008)). Here, we show that genetically modified polymerases lacking Rpa49 or Rpa49Ct, with a single modified residue in Rpa190, Rpa135, or Rpa12, at a position diametrically opposed to the position that binds to Rrn3, can initiate transcription and that strains harboring them grow normally. We propose that there is a second important interface which is involved in Pol I recruitment to the rDNA gene, in addition to the interface between Rrn3 and the stalk.

The role of Rpa12 in suppression
The Rpa12 subunit is involved in stimulating the intrinsic cleavage activity of Pol I through a TFIIS-like domain at its C-terminus. Purified Pol I with Rpa12 lacking the C-terminal domain has no cleavage activity (Kuhn et al, 2007). at the surface of Pol I to the mobile C-terminal region (TFIIS-like) and is therefore indirectly required for cleavage. In vitro, purified Pol I that lacks Rpa12 has less activity than WT Pol I in promoter-dependent transcription assays. Mutations in other Pol I domains, such as deletions in Rpa34, Rpa14, or the Rpa190-DNA mimicking loop, did not influence suppression of the rpa49 deletion growth defect by the RPA135-F301S allele. In contrast, the Rpa12 linker was absolutely required for efficient suppression. Accordingly, the RPA135-F301S allele was unable to restore efficient growth when Rpa12 was absent. Thus, Rpa12 and RPA135-F301S cooperate to support initiation if Rpa49 is absent.

Modification of the jaw/lobe interface facilitates DNA cleft closure
RNA polymerase I undergoes major conformational changes during the transcription cycle, mainly affecting the width of the DNA-binding cleft (Fernández-Tornero, 2018). During the initiation of transcription, the cleft aperture narrows from a semi-open configuration, as seen in cryo-EM structures of the enzyme bound to Rrn3 Pilsl et al, 2016;Torreira et al, 2017), to a fully closed conformation observed in transcribing complexes (Engel et al, 2017;Han et al, 2017). This allows gripping of the transcription bubble inside the cleft ( Figure 8A). Following Rrn3 release, DNA binding is further secured by the Rpa49linker, which crosses the cleft from the lobe to the clamp, passing over the downstream DNA, and the Rpa49Ct domain, which binds the upstream DNA in the vicinity of the clamp (Han et al, 2017;Tafur et al, 2016).
Cleft closure is achieved by the relative movement of two structural units, located on opposite sides of the cleft, which pivot using five hinges (Fernández-Tornero et al, 2013). The unit consisting of the shelf and clamp modules is apparently rigid, whereas that consisting of the core and lobe modules, which is in the vicinity of the mutated residues of our study, undergoes internal rearrangements (Movie 1). The most prominent reorganization within this latter unit affects the Rpa190 jaw domain, the outer rim of which shifts away from the DNA by approximately 3.7 Å, using the lobe/jaw interface as a hinge. This movement also involves the linker region of subunit Rpa12, which contains a beta-strand (residues 46-50) that completes a four-stranded beta-sheet in the Rpa190 jaw domain. As a result, a short alphahelix within the Rpa12 linker region shifts its position by approximately 3.0 Å.

Possible roles of Rpa12 in regulating DNA cleft closure
Rearrangements in the jaw of Rpa190 and Rpa12 linker regions are likely essential to allow pivoting of the shelf-clamp unit against the core-lobe unit. Without such motion, cleft closure would be impossible (Movie 1). We propose that RPA135-F301S or RPA12-S6L favor DNA capture by increasing the flexibility of the lobe/jaw/Rpa12 interface of Pol I relative to that of the WT polymerase. Structural analysis suggests that the Rpa49Ct and its linker domains are involved in securing the closed cleft conformation (Tafur et al, 2016;Han et al, 2017). Cleft closure is likely destabilized in the rpa49∆Ct mutant ( Figure 8B). We propose that the Rpa12 linker domain contributes to the hinge in the lobe/jaw interface. Rpa12 may be involved in cleft-closure, which is necessary to guide the DNA towards the active center, or it may stabilize DNA bound Pol I. Our mutated Pol I probably captures DNA more efficiently than WT polymerase. This structural rearrangement is strongly suggested by our genetic data and the requirement of Rpa12 for promoter-dependent in vitro initiation. There is indirect evidence that similar domains are involved in the initiation of transcription of Pol II and Pol III. Rpb9 in Pol II is similar to the Rpa12-N-terminal module, but lacks the TFIIS domain.
Rpb9 is required for proper start site selection and transcription fidelity (Hull et al, 1995;Walmacq et al, 2009). Furthermore, mutations in the lobe domain of Rpb2, adjacent to Rpb9, alter both Pol II-TFIIF binding and the transcription start site (Chen et al, 2007). For Pol III, TFIIS has been shown to stimulate transcription initiation in vitro and in vivo (Ghavi-Helm et al, 2008). However, motion at the lobe/jaw interface during cleft closure is only detectable in Pol I.

Plasmids and yeast strain constructions
The oligonucleotides used in this study are listed in supplementary Table 4. Plasmids and details of the cloning steps are described in supplementary Table 3. Randomly mutagenized RPA190 and RPA135 libraries were obtained by transformation and amplification of pVV190 and pNOY80, respectively, into XL1-red strains according to the manufacturer's guidelines (XL1-Red Competent Cells, from Agilent Technologies). Yeast strains are listed in supplementary Table 2, and were constructed by meiotic crossing and DNA transformation (Schiestl & Gietz, 1989) (Sambrook et al, 1989). The yeast media and genetic techniques were described previously (Sherman et al, 1986).
yCNOD226-1a was obtained from yCNOD223-2a by switching the KAN-MX to the NAT-MX marker under the control of the MF(ALPHA)2/YGL089C promoter (alphaNAT-MX4), which allowed selection of MATα haploid cells (Decourty et al, 2008). OGT9-6a is an offspring of yCNOD226-1a crossed with BY4741. OGT8-11a is an offspring of BY4742 crossed with Y1196. Strains OGT9-6a and OGT8-11a were plated on rich media and UV irradiated (5W/m 2 during 5 second), resulting in 50% survival. LH514D and LH11D are suppressor clones of the growth defect selected from UV-irradiated OGT9-6a grown at 25°C. AH29R is a suppressor clone of the growth defect selected from UV-irradiated OGT8-11a grown at 25°C. Genetic interaction mapping (GIM) analysis of the RPA49 deletion mutant was performed as described previously (Decourty et al, 2008). Microarray data were normalized using MATLAB (MathWorks, Inc., Natick, MA) as previously described (Albert et al, 2011).

In vivo labeling and RNA extraction and analysis
Metabolic labeling of pre-rRNA was performed as previously described (Hermann-Le Denmat et al, 1994) with the following modifications. Strains were pre-grown in synthetic glucose-containing medium lacking adenine at 30°C to an OD600 of 0.8 at. One milliliter cultures were labeled with 50 µCi [8-3 H] adenine (NET06300 PerkinElmer) for 2 min. Cells were collected by centrifugation and the pellets frozen in liquid nitrogen. RNA was then extracted as previously described (Beltrame & Tollervey, 1992) and precipitated with ethanol.
For high molecular weight RNA analysis, 20% of the RNA was glyoxal denatured and resolved on a 1.2% agarose gel. Low molecular weight RNAs were resolved on 8% polyacrylamide/8.3 M urea gels.

Miller spread experiments and analysis
Chromatin spreading was mainly performed as described previously with minor modifications (Osheim et al, 2009). Carbon-coated grids were rendered hydrophilic by glow discharge instead of ethanol treatment. Negatively stained chromatin was obtained by short incubation with heavy metal followed by quick drying of the sample. Images were obtained using a JEOL JEM-1400 HC electron microscope (40 to 120 kV) with an Orius camera (11Mpixels).
The position of the RNA polymerase I molecules and the rDNA fiber were determined by visual inspection of micrographs using Image J (http://rsb.info.nih.gov/ij/). Digital images were processed by software programs Image J and Adobe Photoshop® (v. CS6).
Ethanol (700 µl) p.a. was added and the tubes mixed. Nucleic acids were precipitated at -20°C overnight or for 30 min at -80°C. The samples were centrifuged for 10 min at 12,000g and the supernatant removed. The precipitate was washed with 0.15 ml 70% ethanol. After centrifugation, the supernatant was removed and the pellets dried at 95°C for 2 min. RNA in the pellet was dissolved in 12 µl 80% formamide, 0.1 M TRIS-Borate-EDTA (TBE), 0.02% bromophenol blue, and 0.02% xylene cyanol. Samples were heated for 2 min with vigorous shaking at 95°C and briefly centrifuged. After loading on a 6% polyacrylamide gel containing 7M urea and 1X TBE, RNAs were separated by applying 25 watts for 30-40 min. The gel was rinsed in water for 10 min and dried for 30 min at 80°C using a vacuum dryer.
Radiolabelled transcripts were visualised using a PhosphoImager.

Suppl movie legends 1
Supplemental Movie 1. Conformational changes in Pol I upon initiation. The movie starts with the closed cleft conformation (PDB 5W66 (Han et al, 2017)), in which melted DNA occupies the cleft and then changes to the intermediate cleft conformation observed in monomeric Pol I (PDB 5M3M (Neyer et al, 2016)). Relevant structural regions have been variously colored and labeled, and residues mutated in this report are shown in red.