Rio1 promotes rDNA stability and downregulates RNA polymerase I to ensure rDNA segregation

The conserved protein kinase Rio1 localizes to the cytoplasm and nucleus of eukaryotic cells. While the roles of Rio1 in the cytoplasm are well characterized, its nuclear function remains unknown. Here we show that nuclear Rio1 promotes rDNA array stability and segregation in Saccharomyces cerevisiae. During rDNA replication in S phase, Rio1 downregulates RNA polymerase I (PolI) and recruits the histone deacetylase Sir2. Both interventions ensure rDNA copy-number homeostasis and prevent the formation of extrachromosomal rDNA circles, which are linked to accelerated ageing in yeast. During anaphase, Rio1 downregulates PolI by targeting its subunit Rpa43, causing PolI to dissociate from the rDNA. By stimulating the processing of PolI-generated transcripts at the rDNA, Rio1 allows for rDNA condensation and segregation in late anaphase. These events finalize the genome transmission process. We identify Rio1 as an essential nucleolar housekeeper that integrates rDNA replication and segregation with ribosome biogenesis. The protein kinase Rio1 is known to promote 40S ribosome formation in the cytoplasm. Using budding yeast, the authors here show that Rio1 also acts in the nucleus, downregulates rDNA transcription by Pol I, and activates the processing of its transcripts to ensure rDNA stability and segregation.


Results
Rio1 localizes dynamically to rDNA during the cell cycle. Previous work showed that substituting conserved catalytic residue D244E in the Rio1 kinase domain provoked plasmid loss in S. cerevisiae 21 , suggesting an involvement of Rio1 in DNA replication and/or segregation. To examine this possibility, we imaged Rio1, labeled with green fluorescent protein (Rio1-GFP), in exponentially growing S. cerevisiae cells. We identified the protein both in the cytoplasm (consistent with its documented involvement in 20S pre-rRNA maturation 15 and pre-40S ribosome trans-factor recycling 12,17,18 ), and in the nucleus (Fig. 1a). As the intranuclear localization of Rio1 could not be easily determined, we isolated the nuclei from exponentially grown yeast, crosslinked and then spread them on glass slides. Immunofluorescence (IF) microscopy with an anti-Rio1 antibody revealed that Rio1 was highly enriched at the nucleolus (colocalized with nucleolar marker Nop1; Fig. 1b), the subnuclear compartment that is organized around the rDNA array ( Supplementary Fig. 1). However, Rio1 signal intensities were heterogeneous across the nuclei suggesting a cell cycle-dependent localization. To probe this observation further, 6myc-Rio1 cells were arrested in G1 with a-factor and then synchronously released into the cell cycle. Samples were taken during cell cycle progression, the nuclei were isolated, spread and analysed by antimyc IF imaging (Fig. 1c,d). We observed that Rio1 was highly enriched at the nucleolus in G1. Rio1 signals then decreased by half through S phase. In metaphase, the nucleoli contained very low amounts of Rio1, whereas at anaphase onset Rio1 became actively re-recruited to the nucleolus, reaching levels similar to those measured in G1 (Fig. 1c,d and Supplementary Fig. 2a). Low amounts of Rio1 were detectable in the nucleus (beyond the nucleolus), especially in S phase (Fig. 1d). Using chromatin immunoprecipitation (ChIP) analysis, we next probed to which sites Rio1 localized at the rDNA. 6myc-Rio1 cells were arrested in G1, S phase, metaphase and anaphase (via use of a-factor, hydroxyurea, nocodazole and the cdc15-2 allele, respectively). The rDNA sequences that were enriched with immunoprecipitated Rio1 were identified by real-time quantitative PCR (RT-qPCR) analysis using probes covering the rDNA unit. While the 6myc-Rio1 signals were distributed homogeneously across the rDNA in G1 and metaphase cells, we found the kinase to be enriched at the 35S rDNA promoter and gene sequence (probes 4 and 5) in S phase and anaphase (coloured lines in Fig. 1e). However, the averaged values obtained with the five probes (black bars in Fig. 1e) confirmed the Rio1 localization dynamics observed by imaging of the spread nuclei.
A previous report indicated that Rio1, overexpressed from the P GAL10 promoter, becomes degraded in S phase, suggesting Rio1 stability is cell cycle regulated 22 . However, our western blot analyses of yeast endogenously expressing Rio1 showed that its protein levels do not change through the cell cycle (Fig. 1f). Hence, the observed dynamic localization of Rio1 may reflect changes in its affinity for nucleolar factors and/or its active import-export from the nucleus. Rio1 shuttling between the nucleus and cytoplasm was evidenced previously by its intranuclear accumulation in an exportin mutant 16 .
Besides its localization, a role for Rio1 at the rDNA was indicated by its recent co-purification with the phosphatase Cdc14 and its inhibitor Cfi1 (ref. 23), which together with the histone deacetylase Sir2 form the rDNA-silencing complex RENT 24 (Fig. 1g). Our own yeast two-hybrid screens that used Rio1 as the bait ( Supplementary Fig. 2b) identified as interactors the rDNA helicase Sgs1 (confirmed by co-immunoprecipitation from yeast whole-cell extract; Supplementary Fig. 2c), the rDNAsilencing protein Tof2 and Rio1 itself ( Supplementary Fig. 2b). These findings extend the current Rio1 protein-interaction map that mostly comprises ribosome biogenesis factors 25 (Fig. 1g).
Distinct sets of nucleolar proteins recruit Rio1. While the 150unit repeat nature of the rDNA array satisfies yeast's huge demand for ribosomes 26 , its configuration makes rDNA highly vulnerable to genetic instability during rDNA replication 27,28 . Although the replication fork moves bi-directionally from the rARS, the leftward-moving replisome is halted at the replication fork barrier site (RFB) to prevent it from colliding with PolI, transcribing the 35S rDNA sequence in rightward direction. Such a collision would generate incomplete 35S rRNA transcripts and produce double-strand breaks in the 35S unit. However, the replisome held at the RFB may collapse, resulting in the exposure of single-strand rDNA, progressing into a double-strand break. Double-strand DNA breaks are repaired by homologous recombination. For a correct repair to occur, the sister rDNA loci must be aligned. If not, homologous recombination will result in an expansion or contraction of the rDNA array and in the formation of ERCs 27,28 , anomalies that have been linked to a shortened lifespan in yeast 29,30 . Various proteins contribute to sister rDNA alignment and faithful recombination, as described in Supplementary Fig. 1. They include Fob1, Tof2, RENT (Cfi1, Cdc14, Sir2), Sgs1, and the monopolin, cohesin and condensin complexes.
To determine the basis for the cell cycle-stage-dependent localization of Rio1 to the rDNA, we probed Rio1 recruitment by IF analysis of spread nuclei isolated from mutants lacking one of the above rDNA factors. Cells were released from G1 or S phase (a sir2D mutant does not respond to a-factor) and then tracked through the cell cycle by analysis of spindle morphology (anti-Tub1 IF) and DNA content (fluorescence-activated cell sorting, FACS) ( Supplementary Fig. 3). In short, we found that Rio1 localization to rDNA in interphase depended on Fob1, Sgs1, Sir2 and Cdc14, while its anaphase recruitment required Fob1, Sgs1, monopolin and condensin (Fig. 2a,b and Supplementary Fig. 4a). Rio1 thus localizes to the rDNA at different cell cycle stages via distinct rDNA factors (summarized in Fig. 2c). The observed changes in Rio1 recruitment in the mutant backgrounds were not due to alterations in Rio1 expression or stability, as western blot analyses showed that Rio1 protein levels remained constant through the cell cycle in all of the mutants tested ( Supplementary  Fig. 4b). Noteworthy, the phosphatase Cdc14 is kept inactive in the nucleolus from G1 through metaphase by its inhibitor Cfi1 (ref. 31) and becomes released (activated) at anaphase entry. However, when we inactivated and delocalized Cdc14 in G1   (1)(2)(3)(4)(5). IGS, intergenic spacer region; RFB, replication fork barrier; rARS, origin of replication; 5S and 35S, genes encoding 5S and 35S rRNA, respectively. The upper plot shows the data obtained with each probe per cell cycle stage (each data point corresponds to the probe indicated above). Error bars, s.d.'s. n ¼ 3. The lower plot shows the averages of the values obtained with the five probes. (f) Western blot of endogenous 6myc-Rio1 protein levels (anti-myc) through the cell cycle (Pgk1 ¼ loading control). (g) The Rio1 interactome. Rio1 interactors reported in the literature [°] 25 and identified in this study [*]). NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7643 ARTICLE (cdc14-3 mutant at the non-permissive temperature), we observed a marked reduction in Rio1 levels at interphase nucleoli and, in parallel, a distribution of the kinase throughout the nucleus (Fig. 2b). A constitutive release of Cdc14 instigated by removing Cfi1 (cfi1D) confirmed this observation. Hence, albeit anchored in the nucleolus, Cdc14 serves at least to recruit Rio1 to interphase rDNA.
Characterization of a Rio1 nuclear depletion mutant. To study Rio1 activity at the rDNA, we wished to inducibly remove the kinase from the nucleus only. Based on the observation that Rio1 shuttles between the nucleus and the cytoplasm 16 , we figured that conditionally removing its C-terminal tail containing a putative nuclear localization signal (NLS; Supplementary Fig. 5a) would prevent truncated Rio1 from entering the nucleus. Rio1 present in the nucleus at the moment of truncation should still be able to exit from the nucleus as its putative nuclear exit signal localizes at the outer edge of the RIO domain ( Fig. 3a and Supplementary  Fig. 5b). To allow for conditional truncation, we cloned the Tobacco Etch Virus (TEV) Protease with a C-terminal NLS under control of the galactose-inducible P GAL10 promoter on a highcopy plasmid (next named pP GAL10 -TEV Protease; Supplementary   Fig. 5c). Next, we introduced a TEV Protease cleavage site just upstream of the NLS in 6myc-Rio1 (next referred to as 6myc-Rio1 TEV ; Fig. 3a). To confirm the TEV Protease-induced cleavage of 6myc-Rio1 TEV , we grew the 6myc-rio1 TEV strains carrying pP GAL10 -TEV Protease or the empty pP GAL10 vector (negative control) in 2% raffinose medium and then treated the cells for 1 h with 2% galactose. TEV Protease expression led to the removal of 6myc-Rio1's C-terminal 78 residues (Fig. 3b), as indicated by anti-myc western blot analysis. Within the hour, truncated 6myc-Rio1 TEV had also become depleted from the nucleus as shown by IF microscopy of spread nuclei (Fig. 3c) and whole cells (Fig. 3d). Whereas 6myc-Rio1 TEV was no longer detected in the nucleus, its signals could still be observed and quantified in the cytoplasm (Fig. 3d). When plated on 2% galactose medium, the mutant strain died, indicating that the C-terminal region of Rio1 is essential for viability, in agreement with previous observations 12 . Importantly, overexpressing the TEV Protease in wild-type yeast did not affect viability (Fig. 4a).
To determine whether truncated Rio1 remains a functioning kinase, we measured the kinase activity of full-length and cleaved 6myc-Rio1 TEV . We synchronized the 6myc-rio1 TEV pP GAL10 -TEV Protease and 6myc-rio1 TEV pP GAL10 strains in G1 and induced TEV Protease expression with 2% galactose (    the cells arrested in G1. Next, we released the cells in 2% galactose medium and took samples 1 h before, upon and 3 h after the release. Western blot analysis confirmed the C-terminal truncation of 6myc-Rio1 TEV in the strain expressing TEV Protease (Fig. 4b). Full-length and truncated 6myc-Rio1 TEV were then immunopurified (anti-myc beads) and their activity assayed on dephosphorylated casein 21 . The measured kinase activities were identical (Fig. 4b), confirming an earlier report that recombinant Rio1 lacking a similar C-terminal region is still active as a kinase in vitro 22 . To exclude a possible contribution from casein kinases CK1 and CK2, shown to co-purify with full-length Rio1 (refs 12,22) and with Rio1 lacking its C-terminal 46 residues 9 , the assay was repeated in the presence of CK1 and CK2 inhibitors. While the overall kinase activity decreased, we measured no differences between the kinase activities of fulllength and truncated 6myc-Rio1 TEV ( Supplementary Fig. 6a). The C terminus of Rio1 was previously shown to be necessary for cytoplasmic Rio1 to bind to the pre-40S ribosome and for promoting 20S pre-rRNA maturation 12 . As such, we examined whether truncated 6myc-Rio1 TEV could still mediate these functions within the 3-h time frame during which we phenotype the mutant cells (see further down). Specifically, we The RIO domain, the putative nuclear export signal (NES), the putative NLS and the TEV Protease cleavage site (introduced between E406 and E407) are indicated. The model of full-length 6myc-Rio1 TEV bound to ATP and manganese was generated with Phyre2 (http://www.sbg.bio.ic.ac.uk/ phyre2) based on the structure of A. fulgidus Rio1 (PDB 1ZTH) 45 . (b) Ectopic expression of the TEV Protease truncates the 6myc-Rio1 TEV protein at its C-terminus. Top: To-scale representation and western blot (anti-myc) of full-length and truncated 6myc-Rio1 TEV proteins in 6myc-rio1 TEV pP GAL10 (blue) and 6myc-rio1 TEV pP GAL10 -TEV Protease (red) cells before and after P GAL10 induction with 2% galactose (Pgk1, loading control). (c) Intranuclear 6myc-Rio1 TEV levels in cells lacking (blue) or expressing (red) TEV Protease, quantified from IF images (anti-myc) of spread nuclei. Both strains were grown in 2% raffinose medium and arrested in G1. P GAL10 was induced (1 h, 2% galactose), and the cells then released into the cell cycle (2% galactose medium, 150 min    In vitro kinase activity analysis of full-length or truncated 6myc-Rio1 TEV isolated from yeast lacking or expressing TEV Protease, respectively. The reaction contained [g-32 P]ATP and dephosphorylated casein, as the substrate. Left; top: [g-32 P]-phosphorylated casein visualized radiographically; bottom: western blots (anti-myc) of full-length or truncated 6myc-Rio1 TEV . The indicated times (min) are relative to the release from G1. Right: 6myc-Rio1 TEV kinase activity normalized to the activity measured 1 h before P GAL10 induction. n ¼ 3. Error bars, s.d.'s. (c) Ribosome profiles of 6myc-rio1 TEV pP GAL10 (blue) and 6myc-rio1 TEV pP GAL10 -TEV Protease (red) strains. Samples were taken at the indicated time points and the cell extracts fractionated by 5-45% sucrose gradient ultracentrifugation. The 254-nm absorption profiles measured along the gradients are shown. (d) Cells depleted of nuclear Rio1 activity are not affected in protein neo-synthesis. 6myc-rio1 TEV pP GAL10 (blue) and 6myc-rio1 TEV pP GAL10 -TEV Protease cells (red) were grown in 2% raffinose medium and then treated with 2% galactose. The cells were provided (20 min) with [methyl-3 H]-Lmethionine and its incorporation in nascent proteins quantified (scintillation counts) at the indicated time points. The numbers represent average values. n ¼ 3. Error bars ¼ s.d's. The extracted proteins were run in parallel on a Nu-PAGE gradient gel and then Coomassie stained, confirming that equal protein amounts were analysed for [methyl-3 H]-L-methionine incorporation. (e) Northern blots of total RNA isolated from 6myc-rio1 TEV pP GAL10 (blue) and 6myc-rio1 TEV pP GAL10 -TEV Protease cells (red) (from experiment in d). Positions of the D-A 2 (green) and 18S (purple) northern probes hybridizing with 20S pre-rRNA and 18S rRNA are indicated in the sketch on the left. probed ribosome biogenesis, protein translation and 20S pre-rRNA to 18S rRNA maturation. Whole-cell extracts from exponentially grown 6myc-rio1 TEV pP GAL10 -TEV Protease and 6myc-rio1 TEV pP GAL10 strains were collected at different time points (0-3 h) after P GAL10 induction. The extracts were fractionated by 5-45% sucrose gradient ultracentrifugation and their nucleic acid profiles were recorded at 254 nm. We observed a subtle decrease in 40S subunits levels (and concomitant increase in 60S levels) at the latest time points in the cells depleted of nuclear Rio1 (Fig. 4c). The ribosome (80S) and polysome profiles of both strains were basically identical suggesting similar mRNA translation capacities. We confirmed this conclusion as pulse labelling of the mutant and control cultures with [methyl-3 H]-Lmethionine revealed no difference in the amount of radiolabelled methionine that was incorporated in the neo-synthesized proteins (Fig. 4d). Next, we probed cytoplasmic 20S pre-rRNA maturation at pre-40S particles via northern blot analysis. In good agreement with our slight decrease in 40S levels, we observed a mild 20S pre-rRNA to 18S rRNA processing defect (Fig. 4e). Because our data show that during the short time frame of one cell cycle (3 h), the cytoplasmic activity of truncated 6myc-Rio1 TEV is not markedly affected, we used this mutant to study how activities at the rDNA are affected in yeast depleted of nuclear Rio1.
Rio1 promotes rDNA stability and segregation. To phenotype the Rio1 nuclear exclusion mutant, we first analysed its progression through the cell cycle. Compared with the control cells, yeast depleted of nuclear Rio1 consistently showed a 15-20-min delay in cell cycle commencement at START (G1/S transition; orange triangles in Fig. 5a), whereas progression through the subsequent cell cycle stages was not affected. Importantly, overproduction of the TEV Protease in a wild-type strain did not delay cell cycle initiation ( Supplementary Fig. 6b). IF microscopy of spread nuclei revealed nucleolar fragmentation in the absence of Rio1 (Fig. 5b), an anomaly that is indicative of rDNA instability. To probe this phenotype further, we grew the 6myc-rio1 TEV pP GAL10 -TEV Protease and 6myc-rio1 TEV pP GAL10 strains for eight divisions in 2% galactose medium. Electrophoretic analysis of their DNA content followed by Southern blot analysis with a 25S rDNA probe 32 revealed ERCs and an expanded rDNA array, as in a sir2D mutant (positive control; Fig. 5c), known to contain 200-300 rDNA units 33 . Homologous recombinations underlaid these rDNA-instability phenotypes as GFP-labelled recombination mediator Rad52 formed fluorescent foci at the nucleolar periphery, the area where rDNA recombination takes place 34 (Fig. 5d). Despite the rDNA hyper-recombination events, progression through S phase was not delayed (Fig. 5a) and the DNA damage checkpoint kinase Rad53 not activated/ phosphorylated 35 (Fig. 5e). Importantly, the cells depleted of nuclear Rio1 were DNA checkpoint proficient as they arrested in S phase with phosphorylated Rad53 upon treatment with hydroxyurea ( Fig. 5f and Supplementary Fig. 6c).
Through IF microscopy analysis of isolated nuclei, we next probed which rDNA regulators required nuclear Rio1 activity for their localization through the cell cycle. In short, cells lacking nuclear Rio1 had reduced levels of Sir2 at rDNA in interphase and metaphase, and of condensin in anaphase. The inability of these proteins to localize to the nucleolus led to their diffusion throughout the nucleus (Fig. 5g, and Supplementary Figs 7 and 8). Reduced Sir2-mediated silencing of rDNA transcription 36,37 , required for sister rDNA alignment, may explain the observed rDNA-instability and array-expansion phenotype in yeast lacking nuclear Rio1 activity. As for condensin, this complex becomes highly enriched at anaphase rDNA to compact the array. Indeed, the large rDNA locus must be condensed before chromosome XII can move through the bud neck into the daughter cell 2 . In addition, condensin recruits topoisomerase II that resolves the remnant catenates lingering between the sister rDNA loci [2][3][4]7 . Both condensin-driven activities promote rDNA segregation in late anaphase. Indeed, live-cell microscopy revealed that yeast depleted of nuclear Rio1 was severely affected in its ability to segregate its GFP-marked rDNA loci (Fig. 6a). In contrast, centromere segregation was not affected ( Supplementary  Fig. 9a). Noteworthy, in cells lacking nuclear Rio1, the anaphase spindle (identified by mCherry-Tub1) extended till 11 mm (versus 7 mm in the control strain; Fig. 6a), likely to try and segregate the rDNA loci. This phenotype is consistent with a defect in rDNA condensation 3 .
Rio1 downregulates PolI and stimulates pre-rRNA processing. For rDNA to condensate and segregate, yeast must turn down PolI activity and locally resolve its transcripts before condensin can be loaded 1,4 . The Cdc14 phosphatase was previously shown to downregulate PolI in anaphase 4 . We wondered whether also Rio1 reduces PolI activity since our ChIP-qPCR analyses had identified high levels of 6myc-Rio1 at the 35S promoter and gene sequence in anaphase cells (cdc15-2) (probes 4 and 5; Fig. 1e). As such, we arrested the 6myc-Rio1 TEV cells in anaphase (cdc15-2) following depletion of nuclear Rio1 activity at the metaphaseanaphase transition. RT-qPCR analysis of cDNA with a 5 0 -external transcribed spacer (5 0 ETS) probe (probe 4 in Fig. 6b) revealed a threefold increase in 35S rRNA levels, as compared with the control anaphase cells. This result was confirmed by northern blot hybridization with a þ 1-A 0 probe showing a 2.5-fold increase in 35S rRNA concentrations (Fig. 6b), suggesting that Rio1 downregulates PolI activity in anaphase. To corroborate this conclusion, we localized PolI by ChIP-qPCR analysis of its subunit Rpa43 in anaphase-arrested cdc15-2 cells (37°C) lacking or containing nuclear Rio1 activity. In the anaphase control cells, Rpa43 levels at the 35S promoter and gene sequence were low but increased by five-and eightfold, respectively, in the absence of nuclear Rio1 activity (red bars, Fig. 6c). Combined, our PolI transcription and localization data indicate that Rio1 downregulates PolI activity in anaphase. The Cdc14 phosphatase reduces PolI activity by dissociating PolI from the anaphase rDNA 4 . Our Rpa43 ChIP data show an accumulation of PolI at the 35S unit in the absence of nuclear Rio1 activity, suggesting that Rio1 represses PolI transcription in a similar fashion.
Since both the Cdc14 phosphatase 4 and Rio1 kinase downregulate PolI in anaphase, we next probed their relative contributions to this process. Whereas anaphase-arrested cells lacking Rio1 activity were characterized by a threefold increase in 35S rRNA concentrations (Fig. 5b, probe 4), anaphase cells lacking Cdc14 activity (cdc14-3) harboured only a 1.5-fold increase in 35S transcript levels as compared with cdc15-2 control cells (Fig. 6d), confirming previous findings 4 . Removing both Cdc14 and Rio1 activities led to a fourfold increase, indicating that both enzymes repress in parallel PolI transcription in anaphase.
As Rio1 localizes to rDNA also before anaphase (Fig. 1c-e) and as PolI transcribes rDNA from G1 through metaphase 4 , we wondered whether Rio1 also modulates PolI transcription before anaphase. To answer this question, we arrested 6myc-rio1 TEV pP GAL10 -TEV Protease and 6myc-rio1 TEV pP GAL10 control cells in G1, depleted the cells of nuclear Rio1 activity and synchronously released them into the cell cycle. Variations in 35S transcript levels measured with the 5 0 ETS RT-qPCR probe were rather small in the control cells (low in S phase, slight increase in metaphase and slight decrease in anaphase; Fig. 7a and Supplementary Fig. 9b). In contrast, marked changes were observed in the cells depleted of nuclear Rio1. Primary transcript levels were sevenfold elevated in G1, then decreased through S phase (still 3.5-fold above the levels measured in the S phase control cells) and increased threefold in anaphase. This transcript-concentration profile is consistent with Rio1 localizing to the rDNA (Fig. 1c-e) to reduce PolI activity through the cell cycle.
To ensure rDNA segregation, yeast must not only reduce PolI activity but also locally resolve the PolI-generated transcripts. During our earlier northern blot analysis of 20S pre-rRNA processing in exponential cells depleted for 3 h of nuclear Rio1 activity (D-A 2 probe in Fig. 4e), we had noticed that the cells accumulated 32S rRNA transcripts ( Supplementary Fig. 9d), indicating a defect in 32S cleavage at A 2 by the nucleolar SSU processome and associated factors 5,38-40 . As such, we probed rRNA transcript processing in our anaphase-arrested cells (cdc15-2) depleted of Rio1 activity. Northern hybridization of total rRNA with an A 2 -A 3 probe revealed a 2.5-fold increase in 35S and 32S pre-RNA levels (Fig. 7b), indicating a derepressed PolI activity and a defect in nascent transcript processing at A 2 .
To determine whether the accumulation of pre-rRNA impaired rDNA segregation in yeast depleted of nuclear Rio1, we tracked rDNA-GFP through anaphase in 6myc-rio1 TEV pP GAL10 and 6myc-rio1 TEV pP GAL10 -TEV Protease cells that inducibly expressed the Aspergillus oryzae ribonuclease RntA 4,41 from P GAL1 (Fig. 7c). We observed that the RntAmediated degradation of the accumulated rRNA transcripts ( Supplementary Fig. 9e) promoted rDNA segregation in yeast depleted of nuclear Rio1 (Fig. 7c). As such, rDNA transmission occurs only when Rio1, together with Cdc14, downregulates PolI activity and then stimulates the co-transcriptional processing of nascent transcripts in the nucleolus.    4 . To examine whether Rio1 also dislodges PolI by targeting Rpa43, we first probed their physical interaction and found that 6myc-Rio1 and Rpa43-3HA efficiently co-immunoprecipitated from exponentially growing cells (Fig. 8a). Next, PhosTag western blot analysis of Rpa43-3HA in 6myc-rio1 TEV pP GAL10 -TEV Protease and 6myc-rio1 TEV pP GAL10 cells identified a slowmigrating, anaphase-specific Rpa43-3HA phospho species that was absent in the cells lacking nuclear Rio1 activity (Fig. 8b). To further associate this anaphase-specific Rpa43 phospho species with Rio1 kinase activity, we analysed Rpa43 phosphorylation in cells expressing either a dominant-negative kinase-dead rio1 allele (P GAL1 -rio1 D244A ) or wild-type RIO1 (P GAL1 -RIO1) (ref. 12). PhosTag western blot analysis showed that the cells expressing rio1 D244A lacked the slow-migrating Rpa43-3HA phospho species in anaphase (Fig. 8c). To examine whether Rio1 directly phosphorylates Rpa43, we performed an in vitro kinase assay with recombinant His6-Rio1 and Rpa43-His6-HA purified from Escherichia coli (Supplementary Fig. 9f) and found that Rpa43 was phosphorylated only in the presence of Rio1 (Fig. 8d). Interestingly, Rio1 autophosphorylation 12,22 was threefold (3.1 ± 0.6) higher in the presence of Rpa43, indicating that Rio1 binding to Rpa43 may have stimulated its own phosphorylation, which in turn may have promoted the phosphorylation of Rpa43 (the protein loading controls for the kinase assay are shown in Supplementary Fig. 9g). Finally, we sought to identify the Rpa43 residues targeted by Rio1 using mass spectrometry. After repeating the in vitro kinase assay in the presence and absence of unlabelled ATP, nanoLC-MS/MS analysis 44  . Tyrosine 73 is one of the few evolutionary conserved residues present in Rpa43 ( Fig. 8f and Supplementary Fig. 10) and resides-from a regulatory point of view-at a highly strategic position. Indeed, within Rpa43, Y73 lies in a short N-terminal b-strand proximal to six C-terminal b-strands and one a-helix that are enriched with phosphorylated residues, as identified by mass spectrometric analysis of Rpa43 isolated from yeast [53][54][55] . Y73 flanks K72 and lies near G78 and L87G88Y89, putative interaction points that may mediate Rpa43-Rpa14 heterodimer formation within PolI (refs 56,57). In addition, Y73 lies in close proximity to the negatively charged residues emanating from the C-terminus of Rpa135 (Fig. 5g). As such, Rio1 kinase activity could induce a charge-based repulsion and destablization of protein interactions within the polymerase complex. The N-terminal domain of Rpa43 that harbours Y73 also binds to the PolI-recruiting transcription factor Rrn3 (ref. 53). Taken together, our cell biological and biochemical data indicate that Rio1 phosphorylates Rpa43 in anaphase to remove PolI from the rDNA. Whether this clearance occurs via disruption of the Rpa43-Rrn3 interaction or via the dissolution of intra-PolI contacts will be determined in future research.

Discussion
We have identified Rio1 as a cell cycle-driven regulator of rDNA transcription, pre-rRNA processing, rDNA stability and segregation (Fig. 9). Yeast SSU processome mutants arrest at cell cycle entry (START) because pre-rRNA does not become processed and ribosomes are not being synthesized 58 .
Our cell cycle experiments revealed that G1 cells lacking nuclear Rio1 activity accumulate primary 35S transcripts but also pre-rRNA species not yet processed at A 2 (RT-qPCR analysis with a probe covering the A 2 site; Supplementary Fig. 9c). These anomalies may help to explain the delay at START of our mutant. In S phase, Rio1 acts at multiple levels. By tuning down transcription by PolI, Rio1 may reduce the frequency of collisions between the replisome and PolI within the 35S sequence, hence minimizing the production of double-strand breaks and incomplete transcripts 59 . Excessive PolI activity may also lead to extreme supercoiling and aberrant chromatin structures that stimulate DNA breakage and recombination. During rDNA replication, the RFB-bound proteins prevent the replisome from entering the 35S rDNA sequence and promote a correct alignment of the newly replicated sister rDNA arrays. Their recruitment of Rio1 indicates an involvement of the kinase in regulating these RFB activities. Rio1 also localizes the histone deacetylase Sir2 to rDNA chromatin. Sir2 promotes rDNA The 6myc-rio1 TEV pP GAL10 and 6myc-rio1 TEV pP GAL10 -TEV Protease strains, carrying the cdc14-3 allele and marked with a 256xtetOBTetR-GFP array flanking RDN1 (rDNA-GFP) were arrested in G1 (2% raffinose medium, 23°C). Cells were then released into the cell cycle (2% raffinose medium, 37°C) and arrested in early anaphase by inactivation of Cdc14-3. 2% Galactose was then added to induce P GAL10 while the Cdc14-3 protein was re-activated in parallel (downshift to 23°C). rDNA-GFP movement and segregation of the sister rDNA loci was tracked through anaphase by live-cell fluorescence microscopy. Left plots: single-cell rDNA segregation profiles. Projected movements of the sister rDNA arrays are shown in dark and light blue or in dark and light red colours. Right plots: percentage of cells with segregated sister rDNA-GFP loci, and the maximum length of their late-anaphase spindles. Error bars, s.d.'s. n ¼ Indicated. (b) Sketch of 35S pre-RNA processing into 20S and 27SA 2 pre-rRNA. The graphs show the quantification of primary 35S rRNA levels measured by RT-qPCR (5 0 ETS probe 4) or by northern blot analysis ( þ 1-A 0 probe) of cDNA or total RNA, respectively. 6myc-rio1 TEV cdc15-2 cells carrying pP GAL10 (blue) or pP GAL10 -TEV Protease (red) were released from a metaphase arrest (nocodazole) under P GAL10 -inducing conditions (37°C). Error bars, s.d.'s. n ¼ 2. The northern blot with the þ 1-A 0 probe is shown underneath the graph. (c) ChIP-qPCR based measurement of Rpa43 levels at the rDNA of cdc15-2 6myc-rio1 TEV pP GAL10 (blue) and cdc15-2 6myc-rio1 TEV pP GAL10 -TEV Protease cells (red) arrested in anaphase following release from a metaphase arrest (37°C, 2% galactose medium). Rpa43 levels across the rDNA unit were measured with probes 2-5. (d) RT-qPCR based quantification of primary 35S rRNA levels (5 0 ETS probe 4) in 6myc-rio1 TEV cdc14-3 carrying pP GAL10 (blue) or pP GAL10 -TEV Protease (red) arrested in anaphase (37°C) after release from a metaphase arrest (nocodazole) under P GAL10 -inducing conditions. Reported values are normalized to the 35S levels measured for cdc15-2 6myc-rio1 TEV pP GAL10 cells, as indicated with a dashed line (Fig. 6b). Error bars, s.d.'s. n ¼ 3.
stability by localizing both to the RFB and also to the cryptic E-Pro promoter. By repressing PolII-mediated transcription from E-Pro, Sir2 prevents the synthesis of non-coding E-Pro transcripts that displace the cohesin complexes and cause the sister arrays to misalign 36 . Double-strand rDNA breaks repaired by recombination between misaligned sister arrays results in rDNA repeat-number expansion or reduction, and in the formation of ERCs 27,28 . Both ERC accumulation and rDNA repeat-number instability have been linked to senescence (shortened lifespan) 29,30 , suggesting that Rio1 may determine yeast life expectancy. When rDNA repeat numbers fall below wild-type level, reduced repression of E-Pro may promote unequal recombination and increase repeat numbers. As such, Rio1 could act as a monitor of rDNA copy number and as an array manager by dictating Sir2 levels and activity.
Once cells have endured rDNA replication, repression of PolI activity may no longer be necessary in metaphase. As such, Rio1 dissociating from the metaphase nucleolus may lead to a temporal increase in rDNA transcription before PolI becomes downregulated in anaphase. Rio1, together with the conserved Cdc14 phosphatase, reduce PolI activity by targeting Rpa43, a subunit of the Rpa14-Rpa43 dimer, that lies at the outer edge of PolI and mediates PolI recruitment by Rrn3 (refs 42,43). The concurrent action by Rio1 and Cdc14 may establish a local phosphothreshold that commands the dissociation of PolI from 35S at anaphase entry. (a) RT-qPCR analysis of 35S cDNA (5 0 ETS probe 4) from 6myc-rio1 TEV pP GAL10 (blue) and 6myc-rio1 TEV pP GAL10 -TEV Protease (red) cells synchronously released from G1 into the cell cycle (2% galactose medium). P GAL10 was induced 1 h before releasing the G1 arrested cells. Error bars, s.d.'s. n ¼ 3. (b) Left: northern blot analysis with an A 2 -A 3 probe of total RNA isolated from 6myc-rio1 TEV cdc14-3 cells carrying pP GAL10 (blue) or pP GAL10 -TEV Protease (red). The cells were arrested in anaphase after release from a metaphase arrest (nocodazole, 23°C) under P GAL10 -inducing conditions (37°C). Right: quantified northern blot signals, and 35S and 32S pre-rRNA levels relative to those of 27SA 2 pre-rRNA. (c) 6myc-rio1 TEV pP GAL10 and 6myc-rio1 TEV pP GAL10 -TEV Protease cells marked with a 256xtetOBTetR-GFP array at RDN1, containing the pP GAL1 -A. oryzae RNTA or the pP GAL1 control vector. The strains were arrested in metaphase (nocodazole) and released into anaphase under P GAL1 -and P GAL10 -inducing conditions (2% galactose medium). rDNA-GFP segregation was tracked through anaphase by live-cell fluorescence microscopy (n ¼ 30 cells per time point).  By downregulating PolI and by stimulating PolI transcript processing at the A 2 site (as mediated by the SSU processome), Rio1 promotes condensin binding and rDNA transmission. Indeed, without Rio1 activity, the long right arm of chromosome XII harbouring the rDNA array does not segregate into the daughter cells, resulting in chromosome loss. We also observed that condensin itself recruits Rio1 to anaphase rDNA, suggesting that Rio1 might also promote condensin activity. The monopolin complex, which tethers the nucleolus to the nuclear membrane via the CLIP (Chromosome Linkage Inner nuclear membrane Protein) complex 48 , also recruits Rio1 to anaphase rDNA. It is tempting to speculate that Rio1 targeting monopolin may facilitate rDNA segregation by inducing rDNA detachment from the nuclear envelope.
Rio1 and its human orthologue RIOK1 promote the endonucleolytic cleavage of 20S pre-rRNA at pre-40S ribosomes, and stimulate the recycling of trans-acting factors that catalyse 40S maturation 12,[18][19][20] . While these events occur in the cytoplasm, a small pool of RIOK1 also localizes to the human nucleus 19,20 . Treating murine cells with toyocamycin, an inhibitor of Rio1 activity in vitro 61 , impeded nucleolar pre-rRNA processing 62 , indicating that RIOK1 might be involved in early pre-rRNA cleavage, as we observed for Rio1 in yeast.
In contrast to yeast, vertebrate cells contain five rDNA clusters on different chromosomes. In prophase, these nucleoli disassemble and their transcription becomes repressed to allow the stripped rDNA to segregate with the rest of the genome. Nucleolar transcription is downregulated by the CDK1-cyclin B kinase, whose phosphorylation of promoter selectivity factor SL1 prevents the formation of the PolI pre-initiation complex [63][64][65] . Similar to yeast Rio1, RIOK1 might just as well contribute to the transcriptional repression of PolI.
In conclusion, our study reveals the first nuclear functions of the Rio1 kinase. By downregulating PolI-mediated rDNA transcription and by promoting the processing of its transcripts, Rio1 ensures both a timely commencement and conclusion of the cell cycle (rDNA segregation permits exit from mitosis). Rio1 safeguards rDNA stability during DNA replication and integrates early nucleolar and late cytoplasmic events during ribosome biogenesis. As such, Rio1 activity allows yeast to actively grow and divide while ensuring the integrity and faithful transmission of its genome.

Methods
Yeast strains. Yeast strains (W303-1A background; Supplementary Table 1) were made by mating, tetrad dissection and spore selection, by introducing integrative or episomal plasmids or via transformation and homologous recombination of PCRgenerated deletion or epitope cassettes. None of the epitopes affected yeast fitness or the activity of the tagged protein. The plasmids used in this study are listed in Supplementary Table 2.
To generate the 6myc-rio1 TEV strain, 21 nucleotides (5 0 -GAAAACCTGTATT TTCAGGGC-3 0 ) encoding the TEV Protease cleavage site (ENLYFQG) were introduced by PCR between the nucleotides encoding Rio1 residues E406 and E407 in an integrative URA3-based plasmid harbouring 6myc-RIO1 under control of its endogenous promoter, P RIO1 . After linearization in P RIO1 by the endonuclease SgraI, the plasmid was integrated at the endogenous P RIO1 in the genome of parent strain PDW001. Following incubation on complete minimal medium containing 1 g l À 1 5fluoroorotic acid (Fluka), we PCR-identified the colonies in which wild-type RIO1 had been recombined out, leaving the 6myc-rio1 allele expressed from P RIO1 and harbouring the TEV Protease cleavage site (named 6myc-rio1 TEV ) as the only source of Rio1. The episomal high-copy 2 m pP GAL10 -HA-TEV Protease-NLS plasmid was made by subcloning the HA-TEV Protease-NLS sequence from pYeF1-TEV (ref. 66) into the HIS3-marked P GAL10 expression plasmid pYeHFc2H (pP GAL10 , gift from C. Cullin) as well as into pYeHFc2H provided with geneticin-resistance cassette KanMX4. After transforming the 6myc-rio1 TEV strain, we obtained strain 6myc-rio1 TEV pP GAL10 -TEV Protease and the 6myc-rio1 TEV pP GAL10 control strains.
Cell cycle studies. Cells were grown at 23°C in minimal synthetic drop-out media or in YEPA medium (1% yeast extract, 2% peptone and 0.3 mM adenine) containing 2% raffinose, 2% galactose or 2% glucose. To arrest yeast in G1, S phase or metaphase, cultures were treated (2.5-3 h) with 5 mg ml À 1 a-factor (GeneScript), 10 mg ml À 1 hydroxyurea (Sigma-Aldrich) or 15 mg ml À 1 nocodazole (Sigma-Aldrich), respectively. To release the cells into the cell cycle, we filtered the yeast cultures, washed and then resuspended the cells in drug-free medium. To arrest the cells in anaphase, strains carrying either the temperature-sensitive cdc15-2 or cdc14-3 allele were first arrested in G1 or metaphase (23°C) and then released at the non-permissive temperature (37°C).
To deplete Rio1 activity from the nucleus, both the 6myc-rio1 TEV pP GAL10 -TEV Protease strain and the 6myc-rio1 TEV pP GAL10 negative-control strain were grown in 2% raffinose YEPA medium (raffinose does not affect P GAL10 ) containing 220 mg ml À 1 geneticin (G418, Life Technologies). While arrested or during exponential growth, the cells were treated with 2% galactose to induce P GAL10 (for 1 h during the arrest, for 3 h during exponential growth). Arrested cells were then released in medium containing 2% galactose, 2% raffinose and 220 mg ml À 1 G418.
For rDNA and ERC analysis, total DNA was extracted from cells grown for 15 h (eight divisions) in 2% galactose medium, run on a 0.6% TBE-agarose gel, transferred for Southern blot analysis and hybridized with a 25S rDNA probe 32 . The oligomers used to produce the probe are listed in Supplementary Table 3.
Progression through the cell cycle was tracked by indirect IF microscopy analysis of spindle morphology (see further down) and by FACS analysis of DNA content. Regarding the latter, cells were fixed in 70% ethanol, incubated first with 1 mg ml À 1 RNAse A (Sigma-Aldrich), then with 5 mg ml À 1 pepsin (Sigma-Aldrich) and ultimately resuspended in 1 mg ml À 1 propidium iodide (Sigma-Aldrich). After mild sonication, cell fluorescence was analysed with a FACScan (Becton Dickinson) sorter and DNA content profiles generated with CellQuest software.
IF imaging of yeast cells was performed with 1 ml of cell culture. The cells were centrifuged and crosslinked overnight in 1 ml of 3.7% formaldehyde. The cell walls were then digested (30 min) with zymolyase (100 mg ml À 1 , Amsbio), washed with 1.2 M sorbitol and 100 mM phospho-citrate, pH 5.9, and bound to a multiwall poly-L-lysine-coated glass slide (Sigma-Aldrich). The slide was treated with DAPI and hybridized with rat monoclonal anti-Tub1 ( Figure 9 | Rio1 downregulates RNA PolI to promote rDNA stability and segregation. Working model of Rio1-mediated regulation of rDNA transcription, pre-rRNA processing, rDNA stability and segregation, as identified in this study. In G1, Rio1 downregulates 35S rDNA transcription by PolI and promotes pre-rRNA processing at A 2 by the SSU processome and associated factors, allowing for a timely commencement of the cell cycle at START. In S phase, the Rio1-mediated downregulation of PolI and recruitment of the histone deacetylase Sir2 to rDNA chromatin, promote rDNA stability. As rDNA replication progresses, Rio1 concentrations at S phase rDNA decrease, resulting in metaphase nuclei containing low levels of Rio1. At anaphase onset, Rio1 becomes rerecruited to the nucleolus to remove PolI from the rDNA by targeting PolI subunit Rpa43. PolI delocalization and promotion of pre-rRNA processing at the rDNA allow for condensin enrichment, resulting in rDNA compaction and segregation. Red lines: inhibitory activity, green lines: activating activity, solid lines: direct activity, dashed lines: direct or indirect activity. IF imaging of rDNA or centromeres in fixed cell samples was done as described above and performed with strains whose RDN1 or CEN4 sequence was flanked with a 256xtetOBTetR-GFP array. Images were acquired with a DeltaVision ELITE microscope (Applied Precision) carrying an Olympus IX71 UPlanSApo objective lens (numerical aperture 1.40) and a CoolSnap HQ2 CCD Camera (Photometrics). Fifteen Z-stacks were acquired every 0.4 mm, deconvoluted (SoftWoRx) and projected with maximum intensity.
Live-cell imaging of spindle (mCherry-Tub1) elongation and RDN1-GFP segregation was performed with microfluidic plates (CellASIC) and a DeltaVision microscope setup as described above via 15 Z-stacks (0.4 mm each). The images were deconvoluted (SoftWoRx), projected with maximum intensity and analysed with ImageJ1.43u (National Institutes of Health, N.I.H.).
RT-qPCR analysis of rRNA. Total RNA was isolated 67 from cells collected through the cell cycle or enriched at a specific cell cycle stage. RNA (1 mg) was retrotranscribed into cDNA with Random Primers (Life Technologies) and ImProm-II Reverse Transcriptase (Promega). cDNA (5 ng) was analysed by RT-qPCR analysis (7500 Fast Real-Time PCR, Life Technologies) using TaqMan probes (Life Technologies; Supplementary Table 3) against the 5 0 ETS sequence of 35S pre-rRNA (probe 4 in the figures) or against the A 2 -processing site of 35S, 33S and 32S pre-rRNA (probe 6). ACT1 cDNA (housekeeping control) was used as the internal reference. Analyses were done in triplicate.
ChIP-qPCR analysis. For each analysis, 50 ml of cell culture were treated with 1% formaldehyde (2 h). Crosslinked cells were washed with Tris-buffered saline (50 mM Tris-Cl, pH 7.5, 150 mM NaCl) and resuspended in lysis buffer (50 mM HEPES, pH 7.5, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100 and 0.1% sodium deoxycholate). The cells were broken with glass beads and the whole-cell extracts were sonicated on ice with a Branson Sonifier 250 (5 Â 30 s, 2 min on ice) to reduce the genome in B500 bp fragments. Anti-myc sepharose beads and rabbit polyclonal anti-Rpa43 antibody (gift from M. Riva) conjugated to protein A-associated agarose beads (Pierce) were incubated (4°C, overnight) with the extracts to isolate 6myc-Rio1 and Rpa43. The beads were then washed twice with lysis buffer, once with washing buffer (100 mM Tris-Cl, pH 8.0, 1 mM EDTA 1% Triton X-100, 0.1% sodium deoxycholate, 0.5% NP-40 and 250 mM LiCl) and once with TE buffer. The beads were resuspended (10 min, 65°C) in elution buffer (1% SDS, 50 mM Tris-Cl, pH 8.0, 10 mM EDTA) and the eluted material then separated from the beads and heated (65°C, overnight) to reverse the crosslinks. An amount of 25 ml of total chromatin solution were similarly heat treated ( ¼ Input). After treatment with 20 mg ml À 1 proteinase K (Roche), the material was treated with 10 mg ml À 1 RNase (Sigma-Aldrich) for 30 min at 37°C, and the DNA was extracted with phenol and chloroform. Following precipitation, the DNA was resuspended in 40 ml water. Input and immunoprecipitated samples were analysed by RT-qPCR (probes are indicated in the Figures 1a, 6b and 6c and in Supplementary Table 3).
Yeast two-hybrid interaction screens. Yeast two-hybrid screens were performed by Hybrigenics Services (www.hybrigenics-services.com). The RIO1 coding sequence (YOR119C) was cloned as a C-terminal fusion to LEXA or the GAL4 encoding DNA-binding domain, or as an N-terminal fusion to LEXA or GAL4. The constructs were screened by mating and covered the complexity of the S. cerevisiae genomic library 5-to 14-fold. Prey fragments of positive clones were amplified by PCR, sequenced and the interacting sequences were identified in the GenBank database (NCBI).
Ribosome, protein translation and northern blot analysis. For ribosome profiling, 6myc-rio1 TEV pP GAL10 and 6myc-rio1 TEV pP GAL10 -TEV Protease cells were grown exponentially in 2% raffinose medium to an OD 600 ¼ 0.5. Galactose (2%) was added and samples were taken at the indicated time points and treated on ice (5 min) with 100 mg ml À 1 cycloheximide to stabilize the polysomes. The cells were washed with extraction buffer (20 mM Tris-Cl, pH 7.5, 50 mM KCl, 10 mM MgCl 2 and 100 mg ml À 1 cycloheximide). The cells were broken with glass beads and the cleared extract (OD 254 ¼ 5.0) was loaded on a 10.5 ml 5-45% sucrose gradient in extraction buffer lacking cycloheximide. Following centrifugation (16 h; 21,000 r.p.m.; SW41 Beckman rotor), gradients were collected at a 1 ml min À 1 flow rate and the ultraviolet profile recorded at 254 nm by a ultraviolet detector (linked to BioLogic LB fractionator), visualized with LP Data View software (Bio-Rad) and exported to Excel (Microsoft Office).
For global translation analysis, cells were grown as described above. The cells were collected at an OD 600 ¼ 0.5 and washed with complete synthetic medium lacking methionine but containing 2% raffinose or 2% galactose. The cells were then resuspended in 1 ml of the corresponding medium and pulse-labelled with 100 mCi of 5 0 ,6 0 [ 3 H]-L-methionine for 20 min at 23°C. Total proteins (TCA insoluble fraction) were extracted and washed twice with cold acetone before resolubilization. The amount of newly synthesized proteins was quantified using a scintillation counter. The solubilized proteins were also separated on a Nu-PAGE 4-10% gradient gel and Coomassie stained to evidence equal loading.
For northern hybridization analysis, total RNA was isolated from the whole-cell extracts submitted to ribosome profiling (see above). Denaturing agarose gel electrophoresis, transferring of the RNA onto a positive membrane and northern hybridization with 32 P-labelled oligomers (Supplementary Table 3) were performed as described 12 .
NanoLC-MS/MS and data analysis. The nano HPLC system was comprised of an UltiMate 3000 HPLC RSLC nano system (Thermo Fisher Scientific) coupled to a Q Exactive mass spectrometer (Thermo Fisher Scientific), equipped with a Proxeon nanospray source (Thermo Fisher Scientific). Following proteolytic treatment of Rpa43-His6-HA, peptides were loaded onto a trap column (PepMap C18, 5 mm Â 300 mm ID, 5 mm particles, 100 Å pores; Thermo Fisher Scientific) at a flow rate of 25 ml min À 1 using 0.1% TFA as mobile phase. After 10 min, the trap column was switched in line with the analytical column (PepMap C18, 500 mm Â 75 mm ID, 3 mm, 100 Å; Thermo Fisher Scientific). The peptides were eluted using a flow rate of 230 nl min À 1 , and a binary 2-h gradient and a 165-min gradient. The gradient starts with the mobile phases: 98% A (water/formic acid, 99.9/0.1, v/v) and 2% B (water/acetonitrile/formic acid, 19.92/80/0.08, v/v/v) increases to 35% B over the next 120 min, followed by a gradient in 5 min to 90% B, stays there for 5 min and decreases in 5 min back to the gradient 98% A and 2% B for equilibration at 30°C. The Q Exactive mass spectrometer was operated in datadependent mode, using a full scan (m/z range 380-1,650, nominal resolution of 70,000; target value 3E6) followed by MS/MS scans of the 12 most abundant ions. MS/MS spectra were acquired using normalized collision energy 27%, isolation width of 2 and the target value was set to 1E5. Precursor ions selected for fragmentation (charge state 2 and higher) were put on a dynamic exclusion list for 10 s. In addition, the underfill ratio was set to 20% resulting in an intensity threshold of 4E4. The peptide match feature and the exclude isotopes feature were enabled.
For peptide identification, the RAW files were loaded into Proteome Discoverer (version 1.4.0.288, Thermo Scientific). The created MS/MS spectra were searched using Mascot 2.2.07 (Matrix Science, Sequest, Thermo Scientific) 68 and MSAmanda 69 against the S. cerevisiae genome database 25 . The following search parameters were used: b-methylthiolation on cysteine was set as a fixed modification, mono-and dioxidation on methionine, acetylation on lysine and protein-N terminus, deamidation on asparagine and glutamine, mono-and dimethylation on lysine, and arginine and phosphorylation on serine, threonine and tyrosine were set as variable modifications. Monoisotopic masses were searched within unrestricted protein masses for tryptic and chymotryptic peptides. The peptide mass tolerance was set to ±6 ppm and the fragment mass tolerance to ± 30 mmu. The maximal number of missed cleavages was set to 2. The result was filtered to 1% false discovery rate using the Percolator algorithm integrated in Proteome Discoverer (Thermo Scientific). The localization of the phosphorylation sites within the peptides was performed with ptmRS, based on the algorithm of phosphoRS 70 . The ion series of the Rpa43 QHLNPLVMKYNNK phospho-peptide are shown in Supplementary Table 4.