Yra1-bound RNA–DNA hybrids cause orientation-independent transcription– replication collisions and telomere instability

Yra1-bound RNA–DNA hybrids cause orientation-independent transcription– replication collisions and telomere instability María García-Rubio, 1,3 Paula Aguilera, 2,3 Juan Lafuente-Barquero, 1 José F. Ruiz, 1 Marie-Noelle Simon, 2 Vincent Geli, 2 Ana G. Rondón, 1 and Andrés Aguilera 1 1 Andalusian Center of Molecular Biology and Regenerative Medicine (CABIMER), Universidad de Sevilla-Consejo Superior de Investigaciones Científicas (CSIC)-Universidad Pablo de Olavide, 41092 Seville, Spain; 2 Marseille Cancer Research Center (CRCM), U1068, Institut National de la Santé et de la Recherche Médicale (INSERM), UMR7258, Centre National de la Recherche Scientifique (CNRS), Aix Marseille University, Institut Paoli-Calmettes, Equipe Labellisée Ligue, 13273 Marseille, France R loops are an important source of genome instability, largely due to their negative impact on replication progression. Yra1/ALY is an abundant RNA-binding factor conserved from yeast to humans and required for mRNA export, but its excess causes lethality and genome instability. Here, we show that, in addition to ssDNA and ssRNA, Yra1 binds RNA–DNA hybrids in vitro and, when artificially overexpressed, can be recruited to chromatin in an RNA– DNA hybrid-dependent manner, stabilizing R loops and converting them into replication obstacles in vivo. Importantly, an excess of Yra1 increases R-loop-mediated genome instability caused by transcription–replication collisions regardless of whether they are codirectional or head-on. It also induces telomere shortening in telomerasenegative cells and accelerates senescence, consistent with a defect in telomere replication. Our results indicate that RNA–DNA hybrids form transiently in cells regardless of replication and, after stabilization by excess Yra1, compromise genome integrity, in agreement with a two-step model of R-loop-mediated genome instability. This work opens new perspectives to understand transcription-associated genome instability in repair-deficient cells, including tumoral cells. [Keywords: Yra1; R loop; transcription–replication collision; telomeres] Supplemental material is available for this article. Received December 28, 2017; revised version accepted May 16, 2018. RNA–DNA hybrids are produced cotranscriptionally when the nascent transcript threads back hybridizing with the template DNA, leading together with the displaced nontemplate ssDNA to a structure termed the R loop. Hybrids have a tendency to accumulate preferentially at highly transcribed protein-coding genes, peaking at promoters and terminators, rDNA, tRNA-coding genes, Ty elements, centromeres, and telomeres (Ginno et al. 2012; Chan et al. 2014; El Hage et al. 2014; Wahba et al. 2016). RNA–DNA hybrids may benefit cell physiology, as shown in some cases of transcription initiation and termination, mitochondrial DNA replication, or immunoglobulin class switching (Aguilera and García-Muse 2012). However, RNA–DNA hybrids may also have a strong impact on genome instability, as shown in cells defective in specific mRNP assembly factors such as the THO complex or the SRSF1 RNA-binding protein, topoisomerase I, RNA–DNA helicases, or RNase H, a ribonuclease that specifically degrades the RNA moiety of RNA– DNA hybrids (Santos-Pereira and Aguilera 2015; Sollier and Cimprich 2015). Accumulating evidence indicates that most of this genetic instability is due to the ability of R loops to stall the progression of the replication fork, leading to its collapse (García-Muse and Aguilera 2016). Thus, RNA–DNA helicases are required for the replication of highly transcribed regions (Boubakri et al. 2010), and R-loop accumulation impairs replication fork progression from bacteria to human cell lines (Wellinger et al. 2006; Tuduri et al. 2009; Gan et al. 2011; Castellano- Pozo et al. 2012; Hamperl et al. 2017). Alternatively, R loops may generate genomic instability if the displaced 3 These authors contributed equally to this work. Corresponding authors: [email protected], [email protected] Article published online ahead of print. Article and publication date are online at http://www.genesdev.org/cgi/doi/10.1101/gad.311274.117. © 2018 García-Rubio et al. This article is distributed exclusively by Cold Spring Harbor Laboratory Press for the first six months after the full-issue publication date (see http://genesdev.cshlp.org/site/misc/terms.xhtml). After six months, it is available under a Creative Commons License (Attribution-NonCommercial 4.0 International), as described at http://creativecommons.org/licenses/by-nc/4.0/. GENES &DEVELOPMENT 32:965–977 Published by Cold Spring Harbor Laboratory Press; ISSN 0890-9369/18; www.genesdev.org 965 Cold Spring Harbor Laboratory Press on September 30, 2020 - Published by genesdev.cshlp.orgDownloaded from ssDNA is recognized and processed by flap endonucleases (Sollier et al. 2014). Despite reports showing that R loops alter replication, causing genome instability, the mechanism is still unclear. However, recent evidence indicates that R loops are not deleterious per se but require a second step, such as a local chromatin compaction, to compromise genome integrity (Castellano-Pozo et al. 2012; García-Pichardo et al. 2017). RNA–DNA hybrids also have a physiological role at telomeres. Telomeres are transcribed into a long non-coding RNA (lncRNA) called TERRA that recruits telomeric proteins, contributes to heterochromatin formation (Maicher et al. 2014), and prevents activation of the DNA damage response (Flynn et al. 2011). A small proportion of TERRA RNA hybridizes with the DNA, forming telomeric RNA–DNA hybrids. This is restrained by the actions of RNase H and the THO complex (Balk et al. 2013; Pfeiffer et al. 2013; Arora et al. 2014; Yu et al. 2014) and favored by telomere shortening. TERRA RNA–DNA hybrids promote homologous recombination (Graf et al. 2017) between telomeric repeats, enabling alternative lengthening of telomeres (ALT), a mechanism used by telomerase-deficient tumor cells to prevent telomere shortening (Balk et al. 2013; Arora et al. 2014; Yu et al. 2014). A number of transcription and RNA processing factors, such as THO or SRSF1, control cotranscriptional R-loop formation (Santos-Pereira and Aguilera 2015). Yra1 is an RNA-binding protein conserved in metazoans (ALY/ REF) that acts as an adaptor for mRNA export factors (Kohler and Hurt 2007). It is cotranscriptionally loaded onto RNA by directly interacting with the C-terminal domain (CTD) of RNA polymerase II (RNAPII) (MacKellar and Greenleaf 2011) or with other mRNP assembly factors (Johnson et al. 2011; Ma et al. 2013; Ren et al. 2017). Interestingly, Yra1 stoichiometry istightly regulated in the cell via a mechanism relying on Yra1 inhibition of YRA1 premRNA splicing (Rodríguez-Navarro et al. 2002; Preker and Guthrie 2006; Dong et al. 2007). Thus, removal of the YRA1 intron from the gene bypasses this autoregulatory circuit, causing Yra1 overexpression with a strong negative impact on mRNA export and cell viability (Rodríguez-Navarro et al. 2002; Preker and Guthrie 2006). Interestingly, we reported recently that high Yra1 intracellular levels alter genome dynamics by accumulating DNA damage and causing transcription-associated spontaneous recombination that is suppressed by RNase H overexpression (Gavaldá et al. 2016). Aiming to understand how Yra1 overexpression and transcription compromise genome integrity, we demonstrate here that excess Yra1 is recruited to chromatin in an RNA–DNA hybrid-dependent manner, thereby stabilizing R loops and converting them into genome integrity threats. Importantly, an excess of Yra1 increases R-loop- mediated genome instability regardless of transcription– replication orientation. Consistent with the fact that Yra1 also binds to telomeres, where R loops accumulate, we show that Yra1 overexpression causes an increase of RNA–DNA hybrids at telomeres, telomere alterations, and accelerated senescence in telomerase-deficient cells. Our results demonstrate not only that excess of Yra1/ ALY binds and stabilizes R loops but that R loops are transiently formed in cells regardless of replication, although they need to be stabilized to compromise genome integrity. Results Excess Yra1 binds to chromatin in an RNA–DNA hybrid-dependent manner Since YRA1 overexpression increases genomic instability in an RNase H-sensitive manner (Gavaldá et al. 2016), we wondered whether overexpressed Yra1 stabilized naturally formed RNA–DNA hybrids. For this, we analyzed whether localization of overexpressed Yra1 to transcribed genes was dependent on RNA–DNA hybrids. We performed an Yra1 chromatin immunoprecipitation (ChIP) in wild-type cells overexpressing or not overexpressing YRA1 using an HA-tagged Yra1 protein expressed from either an intron-deficient (HA-YRA1Δi) or an introncontaining (HA-YRA1) version of YRA1, respectively (Gavaldá et al. 2016). To determine whether putative Yra1 binding to chromatin was dependent on the presence of RNA–DNA hybrids, we overexpressed RNase H, an enzyme that specifically degrades the RNA moiety of RNA– DNA hybrids. Under YRA1-overexpressing conditions, wefoundaclearincrease of Yra1recruitment to theendogenous GCN4 and PDR5 genes previously reported to accumulate RNA–DNA hybrids (García-Benítez et al. 2017) and to the YRA1 intron region. Overexpressing RNase H in the cell, we restored Yra1 basal levels in all of the regions analyzed (Fig. 1A). These observations argue that Yra1 is recruited to RNA–DNA hybrids when artificially overexpressed in addition to their natural putative RNA. Yra1 increases RNA–DNA hybrid accumulation in vivo Next, we determined whether RNA–DNA hybrids were increased, presumably due to stabilization, when YRA1 was overexpressed. We performed a DNA–RNA immunoprecipitation (DRIP) with the S9.6 antibody in wild-type cells overexpressing (HA-YRA1Δi) or not overexpressing (HA-YRA1) YRA1 to detect the hybrids. We focused on the genomic regions where we showed a hybrid-depen- dent Yra1 localization: GCN4,PDR5,and YRA1 introns. We observed a S9.6 signal in the three regions analyzed that was significantly reduced by RNase H treatment (Fig. 1B). The RNA–DNA hybrids detected in GCN4, PDR5, and YRA1 introns were significantly increased when YRA1 was overexpressed (Fig. 1B). Next, we examined whether Yra1 recruitment was enriched at regions naturally forming RNA–DNA hybrids genome-wide when overexpressed. We performed S9.6 DRIP-seq (DRIP combined with high-throughput sequencing) analysis in wild-type cells and compared the data with our previously published ChIP–chip data on Yra1 recruitment (HA-Yra1 and HA-Yra1Δi) (Gavaldá et al. 2016). Our DRIP-seq data are consistent with previously published S9.6 ChIP-seq (ChIP combined with high- García-Rubio et al. 966 GENES &DEVELOPMENT Cold Spring Harbor Laboratory Press on September 30, 2020 - Published by genesdev.cshlp.orgDownloaded from throughput sequencing) data (El Hage et al. 2014), with a significant overlap in the mitochondrial DNA profile and nuclear peaks and genes (Fig. 2A; Supplemental Fig. S1A,B). Comparison with the distribution of Yra1 at wild-type levels (HA-Yra1) or overexpressed (HA-Yra1Δi) revealed a significant correlation between hybrids and Yra1 signals. In our study, the majority of the DRIP signal mapped on ORFs, tRNA genes, and mobile elements, consistent with other reports (El Hage et al. 2014), even though we detected fewer peaks (Supplemental Fig. S1C). Overexpressed Yra1 localized at 1923 genes (Fig. 2B; Supplemental Fig. S1D), with a significant overlap between Yra1-bound genes in cells overexpressing Yra1 and the RNA–DNA hybrid-accumulating genes in wild-type cells (25% or 53% when compared with our new DRIP- seq or the previously published ChIP-seq data) (Fig. 2B; Supplemental Fig. S1D). In summary, the results suggest that overexpressed Yra1 binds to R-loop-accumulating regions genome-wide. Certainly, Yra1 still binds to regions without detectable DRIP signal, as expected from an RNA-binding protein involved in mRNA export. We finally investigated the effect of Yra1-mediated hybrid stabilization in cell fitness. For this, we overexpressed YRA1 in mutants that accumulate R loops, such as hpr1Δ, mft1Δ,ortho2Δmutants lacking the THO complex involved in mRNP biogenesis or the double-mutant top1Δ top2-1 that accumulates negative supercoiling favoring R loops (Supplemental Fig. S2). As can be seen, the results showing a decrease in cell viability suggest that the increase in persistent RNA–DNA hybrids in YRA1-overex- pressing cells could contribute to cell death. An mRNA export defect is not sufficient to increase R loops Since overexpression of Yra1 impairs mRNA export (Rodríguez-Navarro et al. 2002), we wanted to confirm that the nuclear mRNA accumulation resulting from an mRNA export defect was not sufficient to increase R loops. We analyzed three different nucleoporin mutants (nup42Δand nup60Δ, known to be affected in mRNA export, and nup100Δ, not affected) (Bonnet and Palancade 2014) and the positive control (mlp1Δ, a nuclear pore mutant known to accumulate R loops) (García-Benítez et al. 2017). R loops were inferred using a genetic method based on the hyperrecombination ability of the human activation-induced cytidine deaminase (hAID), an enzyme that modifies cytidines in the ssDNA moiety of the R loop, as shown previously (García-Benítez et al. 2017; García- Pichardo et al. 2017). Recombination in nup42Δand nup100Δwas similar to the wild type, whereas in nup60Δit was increased (4.3×), but this increase was not suppressed by RNase H overexpression, consistent with its known sensitivity to HU and MMS (Niño et al. 2016) and in contrast to the positive control mlp1Δ(Supplemental Fig. S3). Therefore, we conclude that accumulation of mRNA in the nucleus due to an RNA export defect does not induce R loops per se. Yra1 binds RNA–DNA hybrids in vitro Since Yra1 localizes to RNA–DNA hybrid-enriched regions and is a well-characterized RNA-binding protein, we reasoned that Yra1 might directly bind to RNA– DNA hybrids. To test this idea, we isolated a recombinant His 6 -tagged version of Yra1 from bacteria, confirming it by Western blot (Supplemental Fig. S4). Next, we performed a gel mobility shift assay incubating increasing amounts of Yra1 with a 25-base-pair (bp) RNA–DNA hybrid that was formed by annealing an RNA oligonucleotide to the complementary radioactively labeled DNA. In parallel, we assayed Yra1 binding to the same radioactive end-labeled RNA or DNA oligonucleotide that forms the hybrid and to a dsRNA constructed by annealing the same end-labeled RNA and a complementary RNA. We observed a shift in RNA migration (confirming the previously described Yra1 ability to bind RNA) but also in ssDNA, dsRNA, and, more importantly, RNA–DNA hybrid migration (Fig. 3A). Since the RNA–DNA hybrid and the ssDNA migrate differently (Fig. 3A, lanes 6,11), we concluded that all labeled ssDNA is present in the form of an RNA–DNA hybrid, as only one band is observed in the sample without Yra1 (Fig. 3A, lane 11). A B Figure 1. YRA1 overexpression causes RNA–DNA hybrid accumulation. (A) ChIP analysis of Yra1 using anti-HA antibody in a wild-type strain overexpressing (green, HA-YRAΔi) or not overexpressing (blue, HA-YRA1) YRA1 and RNaseH (RNH + or RNH − )at the GCN4 and PDR5 genes or the YRA1 intron. n>3.(B) DNA– RNA immunoprecipitation (DRIP) with the S9.6 antibody in wild-type asynchronous cultures overexpressing (HA-YRAΔi; green) or not overexpressing (HA-YRA1; blue) YRA1 at the GCN4 and PDR5 genes or the YRA1 intron. n> 3. Samples were treated (+) or not (−) in vitro with RNase H (RNH) prior to the immunoprecipitation. Means and SEM are plotted in all panels. (∗)P< 0.05, two-tailed Student’st-test. R-loop–replication conflicts enhanced by Yra1 GENES &DEVELOPMENT 967 Cold Spring Harbor Laboratory Press on September 30, 2020 - Published by genesdev.cshlp.orgDownloaded from Therefore, Yra1 is able to bind in vitro not only ssRNA but also ssDNA, dsRNA, and a RNA–DNA duplex. To confirm these interactions, we competed Yra1 binding with different cold nucleic acid species. First, we incubated purified Yra1 with labeled dsRNA and increasing amounts of cold ssRNA or RNA–DNA hybrid. Both nucleic acids reverted Yra1 interaction with dsRNA at similar concentrations (5 µM) (Fig. 3B). Next, we challenged the Yra1–hybrid interaction with increasing amounts of cold hybrid, ssRNA, or ssDNA. In agreement with its binding to RNA–DNA hybrids, cold ssRNA, ssDNA, or RNA– DNA hybrids competed for Yra1 binding, since RNA– DNA hybrids are slightly better competitors than ssRNA and ssDNA (Fig. 3C). Altogether, these results indicate that Yra1 is able to bind RNA–DNA hybrids, supporting the conclusion that when overexpressed in vivo, Yra1 can bind R loops in chromatin, contributing to their stabilization. R-loop-mediated genome instability is linked to head-on transcription–replication We showed previously that transcription causes hyperrecombination when occurring in a head-on orientation with respect to replication but not when occurring codirectionally (Prado and Aguilera 2005). Since RNA–DNA hybrids are an obstacle for replication fork progression (Wellinger et al. 2006; Gan et al. 2011), we examined whether they could explain the orientation-dependent transcription–replication conflicts responsible for genome instability. We used the previously reported plasmids pGAL-OUT and pGAL-IN (Prado and Aguilera A B Figure 2. Genome-wide correlation between YRA1 overexpression and RNA–DNA hybrids. (A) Genomic view of Yra1 recruitment in a wild-type strain overexpressing (HA-YRAΔi) or not overexpressing (HA-YRA1)YRA1 and RNA–DNA hybrid distribution. Fragments of chromosome I (chr1; top) and chromosome IV (chrIV; bottom) are plotted with the signal log 2 ratio values of the ChIP–chip from Gavaldá et al. (2016), the log FE of immunoprecipitation over input for the wild-type (W303) DRIP-seq, and the relative enrichment of reads over the background level of sequencing in a wild-type strain (El Hage et al. 2014). Blue (HA-YRA1 IP), red (HA-YRA1Δi), and green (DRIP-seq or ChIP-seq) histograms represent the significant clusters. SGD features are represented below as blue bars. Profiles were represented using the University of California at Santa Cruz Genome Browser. (B) Venn diagrams showing the overlap between gene sets with significant Yra1 binding in the different ChIP–chip experiments from Gavaldá et al. (2016) (HA-YRA1 IP [blue] and HA-YRA1Δi [red]) and significant RNA–DNA hybrid accumulation (DRIP-seq from this study or ChIP-seq from El Hage et al. [2014]; green). P(x) was calculated with the hypergeometric distribution formula. García-Rubio et al. 968 GENES &DEVELOPMENT Cold Spring Harbor Laboratory Press on September 30, 2020 - Published by genesdev.cshlp.orgDownloaded from 2005) that contain leu2 truncated repeats transcribed from an inducible GAL1 promoter in either a codirectional (OUT) or head-on (IN) orientation with respect to replication driven from the early replication origin ARSH4 (Fig. 4A), respectively. DNA damage driven by the transcription–replication collision would be repaired by recombination between the leu2 direct repeats, generating a wild-type LEU2 gene. Consequently, we could measure genome instability as the frequency of Leu2 + recombinant colonies. We observed that the frequency of recombination in the absence of transcription was low and similar in both systems (Fig. 4B), consistent with previous results (Prado and Aguilera 2005). However, upon transcription, induction recombination was highly increased (7.7-fold) in the IN system with head-on transcription–replication, whereas only a twofold increase in recombination was observed in the codirectional OUT system (Fig. 4B). When similar experiments were performed in both systems after RNase H overexpression, the transcription-dependent hyperrecombination observed in the head-on IN system was suppressed, whereas the recombination frequencies in the codirectional OUT system did not change (Fig. 4B). These results argue that RNA–DNA hybrids are an important source of genome instability in systems undergoing head-on transcription–replication conflicts but not in those undergoing codirectional conflicts. This interpretation would be consistent with the recent observation in human cells using similarly designed plasmid-based constructs in which transcription is driven from a bacterial T7 promoter and replication is driven from an Epstein- Barr virus replication origin (Hamperl et al. 2017) as well with R-loop-dependent replication impairment observed in the Bacillus subtillis genome (Lang et al. 2017). However, none of these results provide any answer to whether R loops are formed only in head-on transcription–replication or after replication. Yra1-stabilized R loops induce instability regardless of transcription–replication orientation If Yra1 binds and stabilizes transient RNA–DNA hybrids present in the genome, we might expect that Yra1 will increase transcription–replication conflicts mediated by these structures. To test this hypothesis, we measured recombination in the codirectional and head-on systems either overexpressing (YRA1Δi) or not overexpressing (YRA1) YRA1. Yra1 overexpression enhanced recombination 7.2 times in the head-on IN system, an increase that was majorly suppressed by RNase H overexpression (Fig. 4C). Importantly, an 8.7-fold increase was also observed in the codirectional OUT system, which was also suppressed by RNase H (Fig. 4C). This result indicates that RNA–DNA hybrids also form under codirectional transcription–replication, but such hybrids are not stable A C BFigure 3. YRA1 binds to RNA–DNA hybrids in vitro. (A) Electrophoretic mobility shift assay (EMSA) of His 6 -Yra1 and either ssRNA, ssDNA, or RNA–DNA hybrids formed by annealing the ssRNA and ssDNA probes used in the single experiments or dsRNA. (B) EMSA of His 6 -Yra1 and dsRNA competed with increasing amounts of cold ssRNA or RNA–DNA hybrids (0.6–40 µM). (C) EMSA of His 6 -Yra1 and RNA–DNA hybrids competed with increasing amounts of cold RNA–DNA hybrids (0.25–8 µM) or ssRNA or ssDNA (0.25–16 µM). R-loop–replication conflicts enhanced by Yra1 GENES &DEVELOPMENT 969 Cold Spring Harbor Laboratory Press on September 30, 2020 - Published by genesdev.cshlp.orgDownloaded from enough to drive high genome instability. However, under Yra1 overexpression, hybrids are stabilized, resulting in a significant increase in recombination, presumably by constituting a stable block to replication fork progression regardless of transcription–replication orientation. Therefore, our hypothesis predicts that RNA–DNA hybrids should be present in both plasmids. Consequently, we performed a DRIP analysis using the S9.6 antibody in samples with or without RNase H treatment. In the analysis, we included the previously described hybrid-accu- mulating PDR5 gene as an internal positive control and PRE1 as a negative control. The results confirmed the presence of RNA–DNA hybrids in the plasmid-borne LEU2 gene of the recombined head-on IN system, in agreement with the in vivo data (Fig. 5A). Notably, we also detected hybrids in the recombined OUT system where LEU2 transcription is codirectional to replication fork progression (Fig. 5A). Therefore, we conclude that RNA–DNA hybrids are formed during transcription in both codirectional and head-on constructs but significantly induce genomic instability, measured as recombination, when replication and transcription are head-on and not when they are codirectional. However, when the hybrid is stabilized via binding to Yra1, replication would stall regardless of the transcription orientation, leading to a similar increase in recombination (sevenfold to eightfold above the wild-type levels). These results clearly support that R loops are formed transiently regardless of replication and not as a consequence of transcription–replication conflicts. Although unlikely, it might be possible that circular plasmids impose a specific topological constraint different from chromosomes that could enhance R-loop accumulation and Yra1 binding. To confirm that this was not the case and that R loops occurred in linear chromosomes, we constructed a transcription–replication collision system in chromosome III and analyzed RNA–DNA hybrids under the same conditions tested in the plasmid construct. To do so, we integrated the LEU2 gene under the inducible GAL1 promoter in a head-on orientation with respect to replication driven from the early-firing ARS315 origin (Fig. 5B), since this orientation was the one with the potential to cause DNA opening that could favor R loops. DRIP analyses of cells cultured in galactose-containing medium (transcription ON) versus glucose-containing medium (transcription off) revealed that RNA–DNA hybrids appeared in the LEU2 gene only when transcribed (Fig. 5B). In the PDR5 gene used as a positive control, hybrids were observed in both conditions, consistent with the fact that PDR5 is constitutively transcribed in both media (Fig. 5B). Importantly, the RNA– DNA hybrids formed in the chromosomal transcription– replication collision system increased when YRA1 was overexpressed (Fig. 5B), in agreement with the increase in recombination observed in the plasmid (Fig. 4C). Cumulative evidence with concomitant studies on RNA–DNA hybrids, γH2AX, and 53BP1 foci and recombination, comet, and DNA-combing assays suggests that R-loop-dependent γH2AX is the result of DNA damage (García-Rubio et al. 2015; Schwab et al. 2015; Salas- Armenteros et al. 2017). To assess whether the RNA– DNA hybrids detected in the LEU2 gene cause DNA damage, we measured H2A-P (the equivalent in yeast to γH2AX) levels by ChIP under conditions of both active and inactive transcription of the GAL1::LEU2 fusion, using telomeric repeats as a positive control (Kitada et al. 2011). An increase in H2A-P was observed when transcription of LEU2 was active, but not when it was inactive, as well as in the telomeric controls (Fig. 5C). Therefore, head-on transcription of the chromosomal LEU2 gene also generates DNA breaks. Notably, such breaks were suppressed by RNase H overexpression (Fig. 5C), confirming that they were R-loop-dependent. Most important, such H2A-P foci were significantly increased when Yra1 was overexpressed, and this increase was suppressed by B A C Figure 4. R-loop-mediated transcription–replication conflicts are orientation-independent in YRA1-overexpressing cells. (A) Schemes of the centromeric plasmids harboring the recombination systems in head-on (IN) and codirectional (OUT) orientation. The arrows indicate the RNAPII-driven transcription orientation of the leu2 repeats from the GAL1 promoter and the direction of replication forks initiated at ARSH4.(B) The effect of RNaseH1 overexpression on the recombination of the head-on (IN) and codirectional (OUT) systems in wild-type cells. Cells grown in either glucose (TRX − ) or galactose (TRX + ) and overexpressing (RNH + ) or not overexpressing (RNH − ) RNase H. The fold increase over no transcription and no RNase H expression is indicated at the right.(C) The effect of RNH1 overexpression on the recombination of transcribed IN or OUT plasmids in wild-type cells overexpressing (YRA1Δi) or not overexpressing (YRA1)YRA1 and RNase H (RNH + or RNH − ). The fold increase over no overexpression of YRA1 and Rnase H is indicated at the right. Means and SEM are plotted in all panels. Asterisks indicate statistically significant differences between the strains indicated, according to Student’st-tests. (∗)Ρ< 0.05; (∗∗)Ρ< 0.005; (∗∗∗)Ρ< 0.0005. García-Rubio et al. 970 GENES &DEVELOPMENT Cold Spring Harbor Laboratory Press on September 30, 2020 - Published by genesdev.cshlp.orgDownloaded from RNH1 overexpression. The results, in addition to validating our studies in plasmid-borne constructs, indicate that the genomic instability derived from transcription–replication collisions is mediated by RNA–DNA hybrids and not by any specific topological constraints that could potentially accumulate in circular plasmids and that the negative impact of those collisions is enhanced by Yra1 overexpression. An excess of Yra1 produces telomere shortening and premature senescence in telomerase-deficient cells We showed recently that an excess of Yra1 causes a cell senescence-like phenotype and a slight telomere shortening in telomerase-positive cells and is enriched at Y′telomeric regions (Gavaldá et al. 2016). Since Yra1 is a highly efficient RNA-binding protein (Fig. 3; Strasser and Hurt 2000) and since telomerase activity in wild-type cells reliesonanRNAmoleculethatisusedasatemplatefortelo- mere synthesis, it was possible that an excess of Yra1 sequestered the telomerase RNA molecule TLC1, leading to a telomerase-deficient phenotype instead of acting directly on the telomere. To test this possibility, we determined the effects of Yra1 overexpression in telomeraseminus cells (est2Δ). Heterozygous EST2/est2Δdiploids harboring the pGAL::Yra1Δiplasmid were sporulated in glucose medium. Next, haploid est2ΔpGAL::Yra1Δispore clones were selected in glucose medium and further propagated in galactose liquid medium via serial dilutions (Hardy et al. 2014). The senescence profiles, kinetics of telomere shortening, and the type of survivors formed in multiple est2Δand est2ΔpGAL::Yra1Δiclones grown in galactose were analyzed (Fig. 6). We showed that est2Δ pGAL::Yra1Δiclones (overexpressing Yra1) exhibited a rapid and deep premature senescence compared with est2Δclones (Fig. 6A). Strikingly, telomere length analysis at the first time point of the senescence assay indicated that telomeres were shorter in the est2ΔpGAL::Yra1Δi clones compared with est2Δclones (Fig. 6B–D). This result suggests an abrupt telomere shortening upon overexpression of Yra1 different from the progressive telomere erosion normally observed in the absence of telomerase. Consistently, survivors appeared earlier in est2Δcells overexpressing YRA1 (Fig. 6C). As predicted, RNA–DNA hybrids were increased in est2Δcells (Fig. 6E) as shown in the single TelVI-R telomere and consistent with previous results indicating that shortening of telomeres in est2Δcells triggers TERRA transcription and RNA–DNA hybrid accumulation (Graf et al. 2017). Importantly, such hybrids were significantly increased under Yra1 overexpression, consistent with a putative role of Yra1 overexpression in RNA–DNA hybrid stabilization (Fig. 6E). As expected, in est2Δcells after the onset of senescence, the liquid culture was dominated mainly by type II survivors because of their growth advantage over type I survivors. In contrast, est2ΔpGAL::Yra1Δicultures appeared to produce few type II survivors and a higher proportion of type I survivors. Quantification of the occurrence of type II recombination events at a single telomere, TELVI- R, further confirmed that overexpression of YRA1 impairs A B C Figure 5. Effect of Yra1 overexpression on transcription–replication collision-mediated genome instability. (A) DRIP with the S9.6 antibody in a wild-type strain carrying the recombined head-on (IN r ) and codirectional (OUT r ) plasmid. Samples from asynchronous cultures grown in galactose were treated (+) or not (−) in vitro with RNase H (RNH). The regions assayed were the LEU2 gene of the plasmid and the chromosomal PDR5 and PRE1 (negative control) genes. n>3.(B) DRIP assay in wild-type strain with the GAL1::LEU2 (URA3) system integrated in chromosome III (Chr III) close to ARS315 in head-on orientation. Nucleic acids from cells grown in either glucose (Trx − ) or galactose (Trx + ) were treated with RNase H (RNH + ) prior to the immunoprecipitation with S9.6 antibody. Samples from cells grown in galactose (Trx + )-overexpressing (GAL::YRA1Δi) YRA1 are shown in green. The regions assayed were the LEU2 gene at the head-on OUT r integrated system and the chromosomal PDR5 gene. n>3. (C) ChIP analysis of H2A-P in the GAL1::LEU2 (URA3) construct integrated in chromosome III (Chr III) in head-on orientationwith respect to replication. Cells transformed with GAL1::RNH1 (RNH + ) or with the empty vector (RNH − ) were grown in glucose (Trx − ) or galactose (Trx + ). n> 3. Cells overexpressing (GAL:: YRA1Δi) YRA1 are shown in green. The regions assayed were the LEU2 gene at the head-on IN r integrated system and the telomere (TELVI-R). n> 3. Means and SEM are plotted in all panels. Asterisks indicate statistically significant differences between the strains indicated, according to Student’st-tests. (∗)Ρ< 0.05; (∗∗)Ρ< 0.005; (∗∗∗)Ρ< 0.0005. R-loop–replication conflicts enhanced by Yra1 GENES &DEVELOPMENT 971 Cold Spring Harbor Laboratory Press on September 30, 2020 - Published by genesdev.cshlp.orgDownloaded from type II recombination (Fig. 7). These results unequivocally indicate that it is the high accumulation of overexpressed Yra1 at telomeres (Gavaldá et al. 2016) rather than an inhibition of telomerase activity due to the sequestering of the RNA template that causes a quick and profound shortening of telomeres and premature senescence. We also noticed an increase of Y′amplification upon Yra1 overexpression, although this increase did not reach a statistical significance (Supplemental Fig. S5). Finally, we tried to determine the impact of RNH1 overexpression on senescence and telomere dynamics in est2Δ cells overexpressing Yra1. However, RNase H overexpression leads to averysick phenotype of telomerase-deficient est2Δcells overexpressing Yra1 (Supplemental Fig. S6). These cells entered crisis quickly with a very low viability compared with est2Δcells overexpressing only Yra1. Unfortunately, it is not possible to know whether this behavior is the result of two independent types of stresses (RNase H and Yra1 overexpression) or whether the two processes are interdependent, and any interpretation about the type of survivors was hampered by the very low cell density at crisis in cells co-overexpressing Rnh1 and Yra1. Indeed, since survivors result from clonal events, the large heterogeneity might simply reflect differences among their very low cell densities. Therefore, we cannot conclude whether RNase H overexpression counteracts Yra1 overexpression or potentiates its effect on telomere dynamics, both possibilities being consistent in any case with the increased presence of RNA–DNA hybrids at telomeres (Fig. 6E). Discussion After demonstrating that the RNA-binding factor Yra1 can bind RNA–DNA hybrids in vitro and in vivo, we showed that dynamic DNA–RNA hybrids are bound by overexpressed Yra1 and threaten genome integrity by promoting transcription–replication collisions. This is a general effect occurring all over the genome. Hyperrecombination caused by transcription–replication collisions is R-loop-dependent and specific to head-on orientation, not being observed for codirectional transcription. Importantly, overexpression of Yra1 causes a similar and significant sevenfold to eightfold increase in R-loop-dependent recombination in systems undergoing either head-on or codirectional transcription–replication collisions. This suggests that RNA–DNA hybrids are cotranscriptionally formed independently of replication, but only when the replication fork approaches an R loop from downstream from the RNAP does it compromise genome integrity. Thus, the replication fork progressing in the same direction as transcription may easily resolve an R loop, unless Yra1 is bound to it. The ability of Yra1 to bind R loops and block replication dynamics is supported by the fact that Yra1 overexpression shortens telomeres by a telomerase- BA C DE Figure 6. Premature senescence and accelerated shortening of telomeres promoted by Yra1 overexpression. (A) Mean growth curves over time of wild type (n= 2), GAL::Yra1Δi(n= 2), est2Δ(n= 10), and est2ΔGAL:: Yra1Δi(n= 10). Each clone corresponds to a spore isolated fromtheheterozygousdiploidstrain, propagatedin liquid culture through daily serial dilutions every 2 d. OD 600 was measured every 2 d to estimate the cell density. Population doublings (PDs) were estimated from the initial spores. (B,top) Schematic representation of wild-type, type I, and type II telomeres. All telomeres contain one X element in the subtelomeric region and from zero to four long (L) or short (S) Y′subtelomeric sequences. Type I survivors show an amplification of Y′sequence and interstitial TG 1–3 repeats. Type II survivors display an elongation of TG 1–3 terminal repeats. Positions of the XhoI sites are shown. (C)TG 1–3 -probed Southern blot analysis of the telomere length of representative clones of est2Δand est2ΔGAL::Yra1Δiduring senescence. (D) Southern blot analysis and mean telomere length at the first time point. n= 10. (E) DRIP with the S9.6 antibody in wild-type and est2Δstrains. Nucleic acidsweretreatedwithRNase H(RNH + )prior tothe immunoprecipitation with S9.6 antibody. Samples from cells grown in galactose (Trx + )-overexpressing (GAL::YRA1Δi) YRA1 are shown in green. The region assayed was the telomere (TELVI-R). n> 3. Means and SEM are plotted in all panels. Asterisks indicate statistically significant differences between the samples indicated, according to Student’st-tests. (∗)Ρ< 0.05;(∗∗)Ρ< 0.005; (∗∗∗)Ρ< 0.0005. García-Rubio et al. 972 GENES &DEVELOPMENT Cold Spring Harbor Laboratory Press on September 30, 2020 - Published by genesdev.cshlp.orgDownloaded from independent mechanism presumably linked to R loops generated by TERRA RNA, which is yet to be understood. These data indicate that R loops form transiently in the cell regardless of replication and generate genome instability mainly in head-on orientation or when they are stabilized by proteins such as Yra1 by impeding the progression of replication. We showed here that the Yra1 RNA-binding factor can bind and presumably stabilize RNA–DNA hybrids in vitro and in vivo when overexpressed, although this would not be its role when present at wild-type levels in the cell. Interestingly, Arabidopsis AtNDX, a homeodomain-con- taining protein, stabilizes R loops by binding to the displaced ssDNA (Sun et al. 2013). In contrast to AtNDX, Yra1 is an RNA-binding protein that presumably affects R-loop dynamics by binding to the RNA–DNA hybrids, likely impeding the action of helicases and RNase H and leading to the accumulation of stable or more persistent R loops. Indeed, Yra1 inhibits loading of Dbp2, an RNA helicase with in vitro RNA–DNA-unwinding activity (Ma et al. 2013, 2016). However, the in vitro ssDNA-bind- ing ability observed for Yra1 (Fig. 3) suggests that Yra1 would also bind the displaced ssDNA of the R loops, potentially strengthening the stabilization and lengthening the persistency of the three-stranded R-loop structure in vivo. Although Yra1 overexpression impairs mRNA export (Rodríguez-Navarro et al. 2002), we showed that a defect in mRNA export is not sufficient to produce either R-loop accumulation or R-loop-dependent hyperrecombination (Supplemental Fig. S3) and that Yra1 binding to transient RNA–DNA hybrids and other RNA structures may also contribute to its deleterious effect (Supplemental Fig. S2). Indeed, RNase H expression reduces not only recombination and DNA damage but also recruitment of Yra1 to chromatin at transcribed genes (Fig. 1A). Thus, the ability of Yra1, in cells overexpressing it, to bind and fix R loops and compromise genome integrity might explain, at least in part, its specific regulation to avoid its overexpression. How transcription induces genome instability is not fully understood yet, but research conducted in the last two decades points to conflicts with replication as a major source of transcription-associated instability (García- Rubio et al. 2003; Prado and Aguilera 2005; Mirkin and Mirkin 2007; Boubakri et al. 2010). The transcription-de- pendent recombination results obtained in the codirectional and head-on systems when Yra1 is overexpressed reveal that R loops are a key player in transcription–replication conflicts (Fig. 4C). This is in agreement with previous work showing that R-loop-accumulating mutants hinder replication in yeast, Caenorhabditis elegans,or human cell lines (Wellinger et al. 2006; Gan et al. 2011; Castellano-Pozo et al. 2012) and that, in bacteria, the RNA–DNA helicase DinG is essential for replication of highly transcribed regions (Boubakri et al. 2010). Interestingly, we showed that R loops by themselves may not be strong inducers of genome instability; instead, they rely on a subsequent step to compromise genome integrity (García-Pichardo et al. 2017). One such step is linked to the modification of chromatin via histone H3 Ser10 phosphorylation (H3S10-P), a mark of chromatin condensation, which reveals that a local chromatin compaction triggered by an R loop may be behind its negative impact on replication progression and genome integrity (Castellano-Pozo et al. 2012; García-Pichardo et al. 2017). Thus, naturally formed R loops are presumably transient, with no consequence on genome integrity. However, here we demonstrate that by artificially stabilizing transient R loops by binding to overexpressed Yra1, R loops accumulate at high levels and compromise genome integrity. We showed previously that transcription induces hyperrecombination only when it occurs concomitantly with replication and in head-on orientation (Prado and Aguilera 2005). Now we show that the hyperrecombination induced in the head-on orientation is largely dependent on R loops (Fig. 4) but not codirectional collisions. These results are consistent with two recent reports in bacteria and human cells showing that codirectional transcription–replication conflicts have little impact on the stability of the construct-containing plasmid or on bacteria viability, while head-on orientation is unstable and causes A B Figure 7. Analysis of type II recombination events at TELVI-R. (A) The clones used in Figure 6 were analyzed by single-telomere Southern blot at the first time point of senescence and after the appearance of survivors. Each band corresponds to a single type II recombination event in the cell population independently of its intensity. (B) The frequency of type II recombination events at TELVI-R. Details are as in Figure 6D. R-loop–replication conflicts enhanced by Yra1 GENES &DEVELOPMENT 973 Cold Spring Harbor Laboratory Press on September 30, 2020 - Published by genesdev.cshlp.orgDownloaded from