Transcription is a major contributor to genome instability. A main cause of transcription‐associated instability relies on the capacity of transcription to stall replication. However, we know little of the possible role, if any, of the RNA polymerase (RNAP) in this process. Here, we analyzed 4 specific yeast RNAPII mutants that show different phenotypes of genetic instability including hyper‐recombination, DNA damage sensitivity and/or a strong dependency on double‐strand break repair functions for viability. Three specific alleles of the RNAPII core, rpb1‐1, rpb1‐S751F and rpb9∆, cause a defect in replication fork progression, compensated for by additional origin firing, as the main action responsible for instability. The transcription elongation defects of rpb1‐S751F and rpb9∆ plus our observation that rpb1‐1 causes RNAPII retention on chromatin suggest that RNAPII could participate in facilitating fork progression upon a transcription‐replication encounter. Our results imply that the RNAPII or ancillary factors actively help prevent transcription‐associated genome instability.
Collisions between RNA transcription and DNA replication machineries impair replication fork (RF) progression and cause DNA breaks, demanding homologous recombination for repair and RF restart. Identification and characterization of specific yeast RNA polymerase II mutations that enhance DNA RF stalling and trigger hyper‐recombination suggests that increased RNAP II retention at transcription sites may contribute to genomic instability and that eukaryotic RNAP II is more than a passive entity in the management of transcription–replication encounters.
Specific RNAP II mutants in Saccharomyces cerevisiae, rpb1‐1, rpb1S71F and rbp9Δ show an increase in genetic instability as detected by hyper‐recombination, DNA damage sensitivity, and dependency on DSB repair functions for viability.
Impaired RF progression in these RNAP II mutants can be compensated by additional origin firing.
Under replication stress, checkpoint‐activating DNA lesions accumulate in these mutant cells and are repaired only with considerable delay.
RNase H1 overexpression does not suppress genome instability in rpb1 mutants, indicating that an increase in R‐loops is not a main determinant of RF stalling.
The causative rpb1‐1 mutation increases retention of RNAP II at the site of transcription.
Genome integrity is essential for cell cycle progression, development and differentiation. In addition to the DNA lesions occurring spontaneously or induced by external genotoxic agents, cellular processes such as DNA replication, repair, recombination or transcription could affect the stability of the genome (Aguilera & Gómez‐González, 2008). Consequently, cells have developed a complex network to coordinate these processes and guarantee genome integrity and cell proliferation. Deficiencies in this coordination result in a variety of diseases ranging from severe genetic disorders to cancer predisposition and accelerated aging. Conflicts between replication and transcription can cause an increase in DNA breaks as a consequence of replication fork (RF) stalling and collapse leading to recombination and chromosome rearrangements in a process termed transcription‐associated recombination (TAR) (Gaillard et al, 2013b). The fact that replication is the major source of recombinogenic DNA breaks and that recombination with the sister chromatid represents the major double‐strand break (DSB) repair pathway during the S/G2 phase of the cell cycle supports this model (Aguilera & García‐Muse, 2012). Consistently, TAR is mainly seen when transcription occurs in the S‐phase and has been related to RF progression impairment caused by transcription (Prado & Aguilera, 2005; Aguilera & Gómez‐González, 2008; Merrikh et al, 2012). Collisions between the replisome and the transcription machinery could lead to DNA breaks that would rely on recombinational repair to allow RF restart (Branzei & Foiani, 2010; Labib & De Piccoli, 2011).
An increasing number of studies have tried to understand the role of replication and recombination factors in transcription‐associated genome instability both in prokaryotes and in eukaryotes (Kim & Jinks‐Robertson, 2012; Gaillard et al, 2013b). The way by which bacteria deal with collisions between the replication and transcription apparatuses has been extensively studied, since co‐directional collisions between the replisome and the RNAP are inevitable. This is due to the fact that the bacterial genome contains only one replicon; the rate of replication is much faster than that of transcription, and replication is not limited to one defined cell cycle phase. Escherichia coli contains three helicases, Rep, UvrD and DinG, which might promote replication of DNA bound to proteins such as the RNAP. Destabilization of transcription complexes can suppress the viability defect observed in repΔC33 ΔuvrD cells, indicating that the Rep–DnaB interaction facilitates resolution of transcription–replication conflicts (Atkinson et al, 2010). Moreover, DinG, Rep and UvrD are essential for efficient replication across highly transcribed regions in vivo (Baharoglu et al, 2010; Boubakri et al, 2010). Backtracking of RNAP at an obstacle has also been implicated in genome instability mediated by transcription–replication collisions (Dutta et al, 2011). RNAP mutants that reduce the frequency of RF stalling have been described in bacteria (McGlynn & Lloyd, 2000; Trautinger & Lloyd, 2002). It has been suggested that such mutants form less stable complexes with template DNA, thus decreasing the probability of collisions with replisomes.
In eukaryotes, several factors have been described as being involved in TAR, which in a number of cases proved to be a result from collisions between transcription and replication. These include RNA‐binding proteins and others with a role in transcription elongation, mRNP biogenesis, processing and/or export (Aguilera & García‐Muse, 2013). As in bacteria, DNA helicases are involved in replication through obstacles that need to be bypassed to allow replication resumption. In yeast, one such helicase is Rrm3 (Ivessa et al, 2003), which is important in solving transcription–replication collisions (Prado & Aguilera, 2005; Azvolinsky et al, 2009). However, other elements can contribute to TAR, including the formation of R‐loops, structures formed by RNA–DNA hybrids and the displaced DNA strand (Aguilera & García‐Muse, 2012), or the local negative supercoiling transiently occurring behind the RNAP (Gaillard et al, 2013b). In such cases, the formation of single‐stranded DNA (ssDNA), more susceptible to damage, would be facilitated (Schmidt et al, 2006; Kim et al, 2007). R‐loops seem to occur naturally, but their accumulation is greatly enhanced in cells deficient in different steps of RNA processing and mRNP biogenesis and export. These deficiencies cause high levels of genome instability as detected by hyper‐recombination, DNA damage accumulation or chromosome loss from yeast to humans (Jimeno et al, 2002; Huertas & Aguilera 2003; Li & Manley, 2005; Paulsen et al, 2009; Domínguez‐Sanchez et al, 2011; Mischo et al, 2011; Wahba et al, 2011; Castellano‐Pozo et al, 2012; Stirling et al, 2012). R‐loops impair transcription elongation and RF progression (Wellinger et al, 2006; Tous & Aguilera, 2007; Gan et al, 2011) likely due to the ability of the R‐loop and/or a putatively associated RNA polymerase to become a roadblock for transcription and replication (Drolet et al, 1995; Huertas & Aguilera, 2003).
Despite the increasing evidence that transcription may be one of the main obstacles to RF progression as a source of genome instability (Gaillard et al, 2013b), little is known about the factors implicated. Here, we asked whether the transcription machinery itself has a direct role in genome instability. We searched for specific mutations of different RNAPII subunits in Saccharomyces cerevisiae that caused genome instability and a strong dependency on DSB repair functions for viability. We analyzed three mutations rpb1‐1, rpb1S751F and rpb9∆ with these phenotypes. Our results demonstrate that RNAPII by itself can become an obstacle to replication and can help solve or prevent the RF stalling and collapse responsible for genome instability.
Genetic interactions between RNAPII and DNA repair mutants
To investigate whether the RNAPII itself might play a role in the origin of genome instability, we analyzed the sensitivity of a selected number of known RNAPII mutants (Supplementary Table S1) to several DNA damaging drugs. In the presence of hydroxyurea (HU), which inhibits the ribonucleotide reductase causing a depletion of dNTPs that impair replication progression, and methyl‐methanesulfonate (MMS), an alkylating agent that can indirectly cause DNA breaks, the selected RNAPII mutants showed different degrees of growth defects (Supplementary Fig S1A). Next, we asked whether or not there were genetic interactions between the RNAPII mutants and those of homologous recombination (HR) such as rad52 and mre11 (Fig 1 and Supplementary Fig S1B). The strongest interaction was observed in the tetrad analysis of the crosses of rad52Δ and mre11Δ with rpb1‐S751F, where we did not find viable double mutants (Fig 1B), consistent with the sick phenotype used to isolate this mutant in a rad52 background (F. Malagón, pers. comm.) (Strathern et al, 2013). However, rpb1‐S751F was tolerated in both rad51∆ and pol32∆ backgrounds. As we have previously shown that Rad51 and Pol32 control two overlapping outcomes of Rad52‐dependent HR events (Moriel‐Carretero & Aguilera, 2010; Muñoz‐Galván et al, 2013b), we tested whether rpb1‐S751F rad51∆ pol32∆ triple mutants were viable or not. Indeed, they were not, suggesting that these mutants accumulate replication‐born DNA breaks that cannot be repaired in the absence of Rad51 and Pol32.
Also, rpb1‐1 and rpb2‐10 strains showed strong growth defects in the absence of HR functions. The rpb9Δ mutant was highly sensitive to HU and in the absence of Mre11 presented a severe growth defect in MMS. Additionally, the four mutants, in combination with the replicative clamp loader RFC1 allele cdc44‐8, essential in recombinational DSB repair (Holmes & Haber, 1999), showed a reduced‐growth phenotype in low concentrations of HU and MMS. An extended analysis of rpb1‐1 in combination with mutations in other HR repair genes such as rad50∆, xrs2∆, rad51∆, rad54∆, mus81∆, pol32∆, sgs1∆ or dia2∆ confirmed the need of DSB repair for viability of rpb1‐1 (Supplementary Fig S1C). Taking into account these results, we selected rpb1‐1, rpb1‐S751F, rpb2‐10 and rpb9Δ for further analyses.
RNAPII mutants show increased DSBs and genetic instability
The need of HR functions for viability of rpb mutants under replication stress suggests that they may accumulate DSBs. We tested this possibility by Western blot against the phosphorylated form of histone H2A (H2A‐P), a molecular marker for DSBs, in cells with and without HU treatment. In rpb1‐1 cells, an H2A‐P signal was detected in the absence of HU (Fig 2A), confirming that breaks occur spontaneously at high frequency. Importantly, this signal was exacerbated in HU. In the other mutants, the H2A‐P signal was stronger than in the WT in the presence of HU, but not detectable in its absence. This tendency of all four rpb mutants to accumulate DSBs was confirmed by analyzing the accumulation of Rad52 foci, a marker for DSB repair centers (Lisby et al, 2001). RNAPII mutations increased the percentage of cells with Rad52 foci, and these were further increased in thermo‐sensitive mutants rpb1‐1, rpb1‐S751F and rpb9Δ at 37°C (Fig 2B). Therefore, DNA breaks accumulate in these RNAPII mutants and are processed into a recombinogenic intermediate.
Next, we asked whether recombination was enhanced in the mutants. We used the plasmid‐based pTINV system, carrying inverted repeats of truncated leu2 fragments, and the chromosomal leu2‐k::ADE2‐URA3::leu2‐k direct‐repeat system (Gómez‐González et al, 2011b) (Fig 2). A significant increase of Leu+ recombinants could be seen in rpb1‐1 cells with respect to WT in both systems (Fig 2C and D), whereas the other mutants show recombination levels similar to WT. We then used a number of plasmid‐based recombination systems to study TAR. These were the L‐lacZ and GL‐lacZ systems carrying 0.6‐kb leu2 direct repeats flanking lacZ under the LEU2 or the GAL1 promoter, respectively. Recombination can be analyzed in these systems under conditions of low (GAL1p in 2% glucose), medium (LEU2p) and high transcription (GAL1p 2% galactose) (Gómez‐González et al, 2011b). As can be seen in Fig 2E, the higher the levels of transcription, the stronger the increase in recombination in rpb1‐S751F. rpb9Δ showed a clear hyper‐recombination phenotype that increased with transcription, which is particularly interesting since both mutants have defects in transcription elongation (Hemming et al, 2000; Strathern et al, 2013). No difference was observed with the L/GL‐lacZ systems in rpb1‐1 mutant. This could be due to a lower efficiency of HR leading to detectable recombination products, which are different for each assay (Gómez‐González et al, 2011b). Interestingly, although the recombinational behavior of rpb1 mutants and rpb9∆ was heterogeneous, they all cause a hyper‐recombination phenotype that was mainly transcription‐dependent. In addition, all mutants, with the exception of rpb2‐10, showed a significant increase in plasmid loss when lacZ was transcribed, as determined with pGAL‐LacZ (Fig 2F). In summary, despite the heterogeneity of phenotypes, genetic instability increased in rpb1‐1, rpb1‐S751F and rpb9∆, whereas rpb2‐10 was poorly or not affected.
Among the mutants studied, viability of rpb1‐1 and rpb1‐S751F shows the strongest requirements for HR functions, in particular under replication stress. Then, we analyzed the effect of mre11∆, which abrogates the early steps of DSB repair, on the genome instability phenotype of rpb1‐1. Inverted‐repeat recombination increased in the rpb1‐1 mre11Δ double mutant above the single mutants, whereas Rad52 foci accumulated at lower levels with respect to the single mre11∆ cells (Fig 3A and B). This suggests that unrepaired DNA breaks are channeled into a single‐strand annealing pathway of recombination in the absence of an active MRX complex, avoiding Rad52 foci accumulation. As expected, this pathway was less effective provided that the Rad52‐dependent HR pathway was also impaired. Consistently, rpb1‐1 mre11Δ double mutants were highly sensitive to HU (Fig 3C).
These results are consistent with the idea that rpb1 mutants generate more DNA breaks in a transcription‐dependent manner. So we wondered whether this was also dependent on R‐loops as it occurs in the hpr1∆ mutant, in which R‐loops are reduced by RNH1 overexpression (Huertas & Aguilera, 2003), but this was not the case. As can be seen in Fig 3D, rpb1‐1 and rpb1‐S751F levels of Rad52‐foci were not reduced by RNH1 overexpression.
DNA replication is impaired in rpb1 and rpb9 mutants
Next, we analyzed whether DNA replication was affected in these mutants by different means. First, we synchronized cells in G1 with α‐factor and monitored growth at different times after G1 release under replicative stress. A clear arrest in G1/early S‐phase was observed in rpb1‐1, rpb1‐S751F and rpb9∆, whereas only a slight delay was observed in rpb2‐10 (Fig 4A). In the absence of HU, a similar tendency of S‐phase delay was observed (Supplementary Fig S2).
Consequently, we directly analyzed the effect of the rpb mutations in RF progression by 2D gel electrophoresis. A scheme of the migration pattern of replication intermediates is shown (Fig 4B). Replication through the SPF1 (YEL031w) gene of chromosome V lying close to the early replication origin ARS508 (Gómez‐González et al, 2009) in G1‐synchronized cells released into S‐phase under low replication stress (40 mM HU) revealed a clear delay in RF progression in rpb1‐1, rpb1‐S751F and rpb9∆ cells. Y‐arcs persisted much longer in the mutants than in WT cells (Fig 4C). In addition, replication initiation was significantly delayed; whereas the Y‐arc is fully visible in WT cells after 20 min, it was absent in the three mutants. However, the rpb2‐10 mutant was not affected either in initiation or in RF progression, since the 2D electrophoresis profile was the same as that of WT (Fig 4C). The results, consistent with the FACS analysis, evidence that RF progression is impaired in the mutants showing genome instability. Interestingly, 2D gel electrophoresis of rpb1‐1 reveals an accumulation of asymmetric X molecules not observed either in WT cells or in the other mutants tested. These structures could in principle be a consequence of two convergent forks unable to terminate properly, but this seems unlikely given the distance of the fork coming from the other side. The other possibility was that formation of X‐structures depended on HR factors such as Rad51. FACS analysis revealed that the double‐mutant rpb1‐1 rad51∆ shows a profile similar to that of rpb1‐1, indicating that rpb1‐1 was epistatic to rad51∆ (Supplementary Fig S3A). The 2D gel electrophoresis confirmed this epistatic relationship and revealed that the X‐structures disappear in the double‐mutant rpb1‐1 rad51∆. Therefore, the replication defect observed in rpb1‐1 leads to the accumulation of Rad51‐dependent recombination intermediates (Supplementary Fig S3B).
Next, we assayed whether this was general to other regions all over the genome, for which we focused the rest of the study on the rpb1 alleles, which altered the largest and essential subunit of RNAPII. First, we analyzed the kinetics of bromodeoxy‐uridine (BrdU) incorporation using conveniently modified strains harboring the human TK gene via chromatin immunoprecipitation (ChIP) using anti‐BrdU antibody. Two regions were analyzed in which transcription of the ORF was convergent to replication originated from its closest ARS. As can be seen in Fig 5A, BrdU incorporation not only was observed at later times than in WT cells, suggesting a delayed initiation as observed by 2D electrophoresis, but the speed of incorporation was slower, as can be deduced from a much lower slope of the incorporation profile of the SPF1 gene. This reduction in the speed was also observed at YFL034W, located near a late replication origin (Fig 5A). We could detect replication impairment both under natural conditions in the absence of replication stress and with low concentration of HU. We could see that the effect was more severe in rpb1‐S751F cells than in the rpb1‐1 mutant.
To assay whether RF progression was impaired all over the genome, we analyzed RF progression by DNA combing under low replication stress. This technique offers the advantage of analyzing replication all over the genome rather than in a specific region. Representative DNA fibers are shown in Supplementary Fig S4. Consistent with both the 2D gel electrophoresis and BrdU incorporation, RF velocity was significantly reduced in both rpb1‐1 and rpb1‐S751F cells (Fig 5B). Importantly, the distance between active origins was reduced from 161 to 105–106 kb on average from the WT to the two rpb1 strains, suggesting that slower RFs are compensated for by additional origin firing and confirming that the major effect of the rpb1 mutations analyzed is in RF progression (Fig 5B).
Altogether, our results indicate that the rpb1 and rpb9 mutants studied are impaired in replication, which can clearly be seen at regions actively transcribed.
The DNA damage response is delayed in rpb1 mutants in the presence of genotoxic agents
DNA replication impairment is a major source of DNA lesions that can compromise genome integrity. Under these circumstances, the cellular DNA damage response (DDR) becomes essential. To test whether the delay in replication recovery implied either a failure in DDR activation or a longer time needed by the cells to repair the damage, we analyzed the kinetics of Rad53 phosphorylation after HU treatment (Fig 6A). Neither rpb1‐1 nor rpb1‐S751F showed spontaneous detectable levels of Rad53‐P, implying that the damage accumulated is not sufficient or its nature is not appropriate to activate the DDR. Interestingly, both rpb1 mutants were able to respond to external damage by activating the Rad53 checkpoint, which can be visualized by high levels of Rad53‐P after HU treatment. In G1‐synchronized WT cultures treated with 100 mM HU for 1 h and subsequently released from G1 in HU‐free media Rad53, phosphorylation was not visible after 40 min of release. However, Rad53 remained phosphorylated after 100–120 min in both rpb1 mutants. Therefore, mutant cells take longer to remove the HU‐induced damage, maintaining Rad53 activated for a more extended length of time, as can be seen by quantification of Rad53‐P (Fig 6A).
Next, we tested the ability of the rpb1 mutants to survive in combination with different rad53 mutations such as the null rad53∆ (together with sml1∆), rad53‐21, which cannot delay the cell cycle in the presence of damage (Desany et al, 1998), or rad53‐1, unable to avoid late origin activation under replication stress (Santocanale & Diffley, 1998). Whereas rpb1‐1 was lethal in combination with rad53∆, rpb1‐S751F showed a synthetic growth phenotype (Supplementary Fig S5A). In a rad53‐21 background, rpb1‐1 growth was heavily impaired, whereas the effect was fairly mild in rad53‐1. The effect on rpb1‐S751F was pretty mild. HU enhanced the phenotypes, but the differences between the mutants were similar as without HU (Supplementary Fig S5B). However, both rpb1‐1 and rpb1‐S751F in combination with rad53‐1 were arrested in the G1/early S‐phase (Supplementary Fig S5C). The results indicate that rpb1‐1 requires a proper DDR for viability whereas this demand is less for rpb1‐S751F, consistent with an accumulation of recombinogenic DNA breaks and X‐structures.
Finally, we assayed whether altered DDR of the rpb1 mutants could affect the capacity of replication restart after MMS, as a way to explain the general impairment of RF progression in these mutants. This was assayed by pulsed‐field gel electrophoresis (PFGE), provided that only fully replicated chromosomes are able to enter the gel, whereas non‐fully replicated chromosomes remain in the well as a consequence of the RF‐containing branched chromosomes (Moriel‐Carretero & Aguilera, 2010). Both rpb1‐1 and rpb1‐S751F took longer to resume replication after MMS exposure (Fig 6B). Altogether, our results suggest that rpb1 mutants show a replication impairment likely due to the incapacity of the RF to restart replication at WT efficiency after encountering obstacles.
Global gene expression analyses of the rpb mutants
Next, we wondered whether mutations in RNAPII might affect expression of multiple genes. We performed microarray expression analysis in rpb1‐1 and rpb1S751F mutants and in rpb2‐10, used as a control. The global expression profiles of rpb1‐1 and rpb1‐S751F were different from each other and different from the rpb2‐10 mutant (Fig 7A), suggesting that a secondary effect on global gene transcription does not explain the similar phenotypes of genome instability and replication impairment of rpb1‐1 and rpb1‐S751F. Basic statistical analyses of the average length, G+C content and expression levels of up‐ and down‐regulated genes revealed that both up‐ and down‐regulated genes were shorter in rpb mutants than the genome average. Moreover, up‐regulated genes were significantly richer in G+C and with lower expression in the 3 mutants analyzed (P < 0.001) (Supplementary Fig S6A). Instead, down‐regulated genes were G+C poorer in rpb1‐S751F, G+C richer in rpb2‐10 and with lower expression levels in rpb1‐1 and rpb2‐10 than in WT cells.
We focus the analysis on the data intersection of rpb1‐1 and rpb1‐S751F mutants. From all RNAPII‐transcribed genes, we selected those whose expression levels in the mutants were significantly different from the WT (P ≤ 0.05). A twofold change cutoff relative to the WT yielded 645 genes in rpb1‐1 and 638 in rpb1‐S751F (Fig 7B). From these, 77 genes were down‐regulated in both rpb1 mutants. In this subgroup, expression was reduced in several genes with a potential role in DDR (Supplementary Table S2), but the effect was not common to all of them and was too low as to explain the common phenotypes reported in this study for rpb1 mutants. To confirm this, we selected 4 genes whose expression was decreased in both rpb1 mutants: HTL1, a component of the RSC chromatin remodeling complex whose expression decreased twofold and sixfold in rpb1‐1 and rpb1‐S751F, respectively; NAT4, an N‐α‐acetyl‐transferase responsible for acetylation of the N‐terminal residues of histones H4 and H2A (Song et al, 2003), whose expression was reduced 4.7‐fold in rpb1‐1 and 2.4 in rpb1‐S751F; IRC4, whose null mutant had been reported to increase the levels of spontaneous Rad52 foci (Alvaro et al, 2007) and whose expression was reduced 14‐fold in rpb1‐1 and 2.6‐fold in rpb1‐S751F; and SUS1, a component of the SAGA histone acetylase and TREX‐2 complexes, involved in transcription activation, elongation and mRNA export (Rodríguez‐Navarro et al, 2004; González‐Aguilera et al, 2008), whose expression decreased 2.3‐ and 2.7‐fold in each mutant. We found that both nat4Δ and htl1Δ recombination and Rad52 foci levels were similar to the WT strain (Supplementary Fig S7A) whereas neither IRC4 nor SUS1 (sus1Δi) overexpression rescues the DNA damage sensitivity observed in rpb1 mutants (Supplementary Fig S7B and C).
In addition, we analyzed RNR3, the large subunit of ribonucleotide‐diphosphate reductase, which is strongly induced by DNA damage (Domkin et al, 2002) and whose expression was enhanced 2.8‐fold in rpb1‐S751F. RNR3 overexpression enhances growth of WT and both rpb1 mutants analyzed in HU or MMS (Supplementary Fig S7D).
These results exclude the possibility that a change in gene expression of a number of specific genes could be the major cause of replication impairment and genome instability in rpb1‐1 and rpb1‐S751F. Indeed, Gene Ontology analyses did not reveal any relevant functional class of genes involved in the maintenance of genome integrity that was equally affected by rpb1‐1 and rpb1‐S751F (Supplementary Table S3). Altogether, the results suggest that there is neither a common profile of gene expression nor genes whose gene expression alterations could reasonably explain the genome instability phenotypes of the rpb1 mutants by themselves. This supports the idea that a major part of the effect of rpb1‐1 and rpb1‐S751F on genome instability is due to a direct role of the RNAPII itself and is consistent with the fact that hyper‐recombination is, indeed, observed at transcribed DNA regions.
RNAPII retention at chromatin in rpb1‐1 mutant
Next, we wondered whether RNAPII accumulated at higher levels in the transcribed chromatin in the different RNAPII mutants analyzed, since this could reflect the incapacity of RNAPII to be released efficiently from the transcription site as a possible cause of genome instability. We analyzed the profile of RNAPII distribution at the actively transcribed PMA1 gene by ChIP (Fig 7C). An increase of RNAPII retention in the transcribed region of PMA1 with respect to the WT could be seen in rpb1‐1 cells. This retention is not specific of the PMA1 gene; it is also observed in SPF1 and PDC1 genes in rpb1‐1 respect to WT (Supplementary Fig S8). Despite the fact that the four mutants studied were affected in transcription (Fig 7D and Supplementary Fig S8B), only rpb1‐1 showed an increased retention of RNAPII on genes.
Genome‐wide occupancy of Rrm3 in rpb1‐1 cells as a marker of replication impairment
Since the rpb1 mutants with transcription‐dependent genome instability share a defective RF progression (Figs 4 and 5), we asked how one representative rpb mutant was affected in replication throughout the whole genome. Therefore, we selected rpb1‐1 for Rrm3‐FLAG ChIP‐chip analyses. Rrm3 is required for the progression of the RF through obstacles in the DNA, and its accumulation at specific DNA sites has been used to identify RF pauses (Ivessa et al, 2000, 2003). ChIP‐chips performed in asynchronous cultures showed an increase of Rrm3 all over the genome in rpb1‐1 compared to WT (Fig 8 and Supplementary Fig S9). The hits of Rrm3 occupancy were clearly enhanced in telomeres in rpb1‐1 (Fig 8A and Supplementary Fig S9), implying that natural RF pause sites are stronger or more frequent at chromosome ends in rpb1‐1 cells. Finally, Rrm3 binds to less replication origins in rpb1‐1, consistent with the delay in replication initiation as detected by 2D gel electrophoresis and BrdU incorporation (Supplementary Table S4).
There was a high coincidence between the ORFs enriched in Rrm3 in rpb1‐1 and WT cells although the difference between the two strains is significant (Supplementary Table S4). The subset of ORFs with high Rrm3 occupancy in rpb1‐1 is richer in G+C content and shorter in length than the average WT cells (P < 0.001) (Supplementary Fig S10A), which are coincident with the features of genes with higher mRNA levels in yeast (Marin et al, 2003). Rrm3 accumulates all over the length of genes in both WT and rpb1‐1 cells. The binding profiles show a tendency to increase toward the 3′‐end (Supplementary Fig S10B), similar to transcription‐dependent hyper‐recombinant mutants defective in mRNP biogenesis (Gómez‐González et al, 2011a).
In addition, we took advantage of microarray expression data and ChIP‐chip analyses to study the correlation between replication impairment and gene expression. As can be seen in Supplementary Fig S11A, Rrm3 enrichment correlates with gene expression levels, even though these refer to asynchronous cultures and not to S‐phase transcription, supporting the idea that replication obstacles occur preferentially at highly transcribed genes. This correlation is still more evident when we compare genes up‐ or down‐regulated in rpb1‐1 (Supplementary Fig S11B).
Next, we analyzed the results taking into account the average of the significant signals in ORFs and ARSs all over the genome. For this, we divided the sequence of each ORF of the yeast genome into 10 equivalent segments (segments 2–10) plus two additional segments of the same size upstream (5′) (segment 1) and downstream (3′) of each ORF (segment 12). In a similar way, each ARS sequence was divided into 11 equivalent segments. Then, we calculated the average signal log2 ratio for Rrm3 hits mapping on each segment of all ORFs or ARSs, respectively, and these values were plotted. Interestingly, we see an increment of Rrm3 retention throughout ORFs and ARSs in rpb1‐1 respect to WT (Fig 8B and C). Since a major occupancy of Rrm3 at ORFs may be a consequence of transcription–replication collisions, our data are consistent with the idea that RNAPII itself participates in preventing such collisions.
We provide genetic evidence that specific RNAPII mutants, rpb1‐1, rpb1‐S71F and rbp9∆, show an increase in genetic instability as detected by hyper‐recombination, DNA damage sensitivity, plasmid loss and/or a dependency on DSB repair functions for viability. These RNAPII mutants display RF progression impairment that could be compensated by additional origin firing. DNA damages accumulate in these mutants under replicative stress that activate the Rad53‐mediated checkpoint, but that are repaired with considerable delay. Genome‐wide Rrm3 occupancy analysis and experiments showing that the rpb1‐1 mutant specifically causes a major retention of RNAPII at the site of transcription suggest that RNAPII can participate actively in facilitating the progression of colliding RFs, presumably by contributing to its own release from the site. Altogether, the data indicate that RNAPII contributes to maintain genome stability by distinct manners and without involving high R‐loop accumulation.
rpb transcription elongation mutants have a differential effect on genome integrity
From the two specific mutations of the largest RNAPII subunit Rpb1 that cause genome instability, the rpb1‐1 mutation (G4622A) maps in the H region of Rpb1, a highly conserved amino acid region, which is important for the selection of the transcription start site (Nonet et al, 1987; Scafe et al, 1990). The rpb1‐1 mutation is temperature sensitive and inactivates RNAPII in vivo, causing a quick shutdown of mRNA synthesis after a shift to the non‐permissive temperature (Nonet et al, 1987; Schroeder et al, 2000), even though the mutant form of RNAPII could still be seen at the transcription sites (Kim et al, 2010). No appreciable decrease of Rpb1 levels has been detected after this temperature shift (Tardiff et al, 2007). Also, suppressors of rpb1‐1 map in the conserved segment I of RPB2, suggesting an interaction between region H of RPB1 and region I of RPB2 (Martin et al, 1990).
rpb1‐S751F maps in the F region, involved in transcription elongation (Braberg et al, 2013). The mutant Rpb1‐S751F protein, instead, undergoes elevated transcriptional slippage (Strathern et al, 2013). Transcriptional slippage occurs when RNAPs exhibit higher inherent error rates during elongation usually due to misincorporation of the wrong nucleotide in combination with failure of the intrinsic correction mechanisms like that mediated by the exoribonuclease activity of RNAPII. Indeed, rpb1‐S751F has reduced transcription levels compared to WT. The effect of the Rpb1‐S751F bulky substitution on elongation is consistent with the close proximity of Ser751 to the active center of RNAPII (Strathern et al, 2013). Both rpb1‐1 and rpb1‐S751F are MMS sensitive, but the sensitivity of the latter is clearly stronger.
The rpb2‐10 mutation maps in the H region linked to nucleotide binding. The mutant is sensitive to 6‐azauracil (6‐AU) and impaired in transcription elongation (Lennon et al, 1998), but genetic stability is poorly affected. Interestingly, rpb2‐4 also maps in the same region and causes sensitivity to 6AU, but its defect in transcription elongation is weaker (Powell & Reines, 1996) whereas rpb2‐7, which maps in the A region, is the most sensitive to 6AU and does not present transcription elongation impairment either (Powell & Reines, 1996). None of these mutants have a clear genome instability phenotype. The differential DDR phenotypes of the mutants suggest that, although transcription elongation impairment can be associated with genome instability, there is no strict correlation with the DDR defect. Indeed, analysis of transcription factor mutants affecting elongation is consistent with this conclusion (Rondon et al, 2003; Rondón et al, 2004; Luna et al, 2005). Despite the different mutants affected in transcription elongation that cause genome instability, our data suggest that the elongation defect is not sufficient to trigger genome instability.
The Rpb9 subunit is not essential for mRNA synthesis in yeast. During transcription initiation, Rpb9 modulates the selection of the transcription start site, recruiting TFIIE and interacting with TFIIF. During the elongation process, Rpb9, together with TFIIS, is required to stimulate the nascent transcript cleavage activity intrinsic to RNAPII (Van Mullem et al, 2002). The rpb9Δ mutant shows a strong sensitivity to 6‐AU or mycophenolic acid (MPA) that can be suppressed overexpressing TFIIS (Hemming et al, 2000) and a transcription elongation defect (Tous et al, 2011). Rpb9 is also involved in transcription‐coupled repair (TCR), which has led to the suggestion that the transcription elongation function of Rpb9 likely plays a role in backtracking the RNAPII stalled at a lesion, by coordinating RNAPII with TFIIS (Li et al, 2006). In addition, Rpb9 functions in ubiquitylation and degradation of RNAPII in response to UV‐induced DNA damage (Chen et al, 2007). Although the TCR defect of rpb9Δ might in principle contributes to its genetic instability, the fact that rad26Δ TCR‐defective mutants do not show hyper‐recombination (González‐Barrera et al, 2002) argues against such a possibility. Also, rpb9Δ is highly sensitive to HU, whereas this is not the case for rad26∆ strains (Gaillard et al, 2009).
Replication fork progression impairment as the cause of genome instability
Our data reveal that the rpb1‐1, rpb1‐S751F and rpb9∆ mutations increase genome instability, whereas this is not the case for rpb2‐10. Interestingly, 2D gel electrophoresis shows that in rpb1‐1, rpb1‐S751F and rpb9Δ, long Y‐shaped molecules accumulated toward the end of the descending simple Y‐arc as a result of the movement of the RFs away from the restriction fragment containing ARS508 (Fig 4). This is consistent with a slowdown of RFs and supports that replication failures occurring in the three mutants analyzed are responsible for genome instability. The same phenotype was previously described in hpr1Δ mutants (Gómez‐González et al, 2009), implying that under replicative and transcription stress cells respond with the pausing or stalling of replication. Consistently, recombinogenic intermediates can be detected in rpb1‐1 in 2D gels, whereas no RF progression impairment was observed in rpb2‐10 mutants (Fig 4).
We believe that the rpb1 and rpb9 mutations analyzed induce instability by creating transcription intermediates that impair RF progression (Aguilera & García‐Muse, 2012). The need of the recombinational DSB repair functions MRX, Rad52 and Rfc1 for proper viability of rpb1‐1 and rpb1S751F mutants is consistent with the idea that the replication impairment leads to DSBs that need to be repaired via homologous recombination; otherwise, cells die. Such breaks promote two different types of recombination events that allow the restart of RFs, one Rad51‐ and the other Pol32‐dependent, consistent with previous results in rad3‐102 (Moriel‐Carretero & Aguilera, 2010) and histone deacetylase mutants (Muñoz‐Galván et al, 2013a) that also impair RF progression causing genome instability.
Replication origins fire more often in rpb1‐1 and rpb1‐S751F mutants than in WT cells, as determined by DNA combing (Fig 5). This may be a mechanism to counteract the slower RF progression in these mutants.
RNAPII function contributes to the prevention of transcription–replication conflicts
The molecular mechanisms by which a defective RNAPII or transcription elongation could lead to RF progression impairment and genome instability seem to be multiple. A lower capacity of the RNAPII to be released from the DNA after a premature transcription termination due to an elongation failure or after finding an obstacle, such as a DNA lesion, can increase the probability of a collision with the RF or its collapse. This explanation would fit for rpb1‐1, in which RNAPII accumulates at higher levels than in WT cells at the ORFs analyzed, but not for rpb1S751F and rpb9∆, which show the same RNAPII occupancy as WT cells. Therefore, an elongation failure, regardless of whether or not linked to a less efficient release of the RNAPII from the DNA, can increase genome instability. In this sense, analysis of the Rrm3 occupancy along the genome of rpb1‐1 showing no large increase in the number of hits with respect to WT cells suggests that the mutant form of the RNAPII does not enhance the probability of collisions or RF stalling. It is possible that the main effect of the mutated RNAPII is a reduced ability to resolve RF stalls, increasing the probability of their collapse. This could explain why Rrm3 accumulates preferentially in transcribed DNA regions (Fig 8) as occurs in WT cells (Azvolinsky et al, 2009) and that RF progression in RNAPII mutants is strongly impaired as determined by 2D electrophoresis, BrdU incorporation and DNA combing. This effect differs from that of other transcription mutants with a clear transcription‐dependent genome instability phenotype as is the case of THO mutants, impaired in transcription elongation and RNA export, in which Rrm3 is accumulated at higher levels and at more ORFs than in WT cells (Gómez‐González et al, 2011a). Indeed, transcription‐associated genome instability can often be associated with the accumulation of R‐loops (Aguilera & García‐Muse, 2012) as it is the case of THO mutants (Huertas & Aguilera, 2003). However, since RNase H1 overexpression does not suppress genome instability in rpb1 mutations, we rule out that an increase in R‐loops is a major determinant of RF progression impairment observed in the mutants. Indeed, the lack of effect of null mutations or overexpression of genes deregulated in rpb1‐1 and rpb1S751S strains (HTL1, NAT4, IRC4, SUS1 and RNR3) on DDR and genome instability is consistent with the idea that major effect of the rpb mutations is direct. In this sense, it is worth noting that RNR3 overexpression diminished the sensitivity of the rpb mutants to replicative stress, as occurs in wild‐type cells, consistent with the general conclusion that these mutations generate transcription‐mediated RF progression failures as a source of instability.
Our results suggest that the RNAPII helps in minimizing the effects of natural transcription–replication collisions on genome integrity. This could occur by either facilitating the release of the RNAPII from chromatin after a putative collision with the RF or by preventing RF collapse, a function that could be defective in rpb1‐1 mutants. The high RNAPII retention at ORFs could increase the probability of replication collisions. In addition, since selected elongation factors associate with RNAPII in a transcription‐dependent manner in rpb1‐1 mutants (Tardiff et al, 2007), rpb1‐1 could also affect recruitment or turnover of factors involved in transcription or chromatin remodeling that could promote conflicts between the transcription and replication machineries. In this sense, it is possible that specific chromatin modifications associated with RNAPII elongation (Selth et al, 2010) or the topological constrains serve as a signal to prevent the replisome from entering a transcribed region and colliding with the RNAPII, or that a defective elongation causes a TCR failure leading to DNA lesions that perturb replication and can become DSBs (Belotserkovskii et al, 2013; Gaillard & Aguilera, 2013a). Deregulation of gene gating to the nuclear pore can also contribute to the instability of transcribed DNA (Bermejo et al, 2012). However, our identification of different RNAPII mutants with impact on RF progression indicates that eukaryotic RNAPII has evolved to minimize the possible impact of transcription–replication conflicts in genome instability, with important implications for the genetic origin of cancer.
Materials and Methods
Strains and plasmids
All experiments were performed at 30°C with few exceptions indicated in figure legends.
Analysis of recombination, Rad52‐YFP foci and plasmid loss
Recombination frequencies were obtained as the median values of three fluctuation tests performed with six independent yeast colonies one from each transformant analyzed. The frequency of plasmid loss was calculated as the median frequency of six independent cultures grown on non‐selective rich medium for 23 generations as previously described (Chavez & Aguilera, 1997). Spontaneous Rad52‐YFP foci from S‐G2 mid‐log‐phase cells bearing the plasmid pWJ1344 were visualized with a Leica DC 350F microscope (Lisby et al, 2001).
Phosphorylated H2A and β‐actin were detected by Western blot of TCA‐extracted proteins separated in 15% PAGE using the ab15083 (Abcam) and ab8226 (Abcam) antibodies, respectively. For signal detection, SuperSignal® West Pico system was used (Pierce). Detection of Rad53 and β‐actin was accomplished by Western blot analysis of TCA‐precipitated proteins separated in 4–20% Criterion TGX gradient PAGE (Bio‐Rad). Antibodies sc‐20169 (Rad53; Santa Cruz Biotechnology) and ab8224 (β‐actin; Abcam) were used. For quantification, secondary antibodies conjugated to IRDye 680CW or 800CW (LI‐COR) were used; the blot was scanned in an Odyssey IR scanner and analyzed with Image Studio 2.0 software (LI‐COR). For phosphorylation signal quantification, the ratio between Rad53‐P versus total Rad53 protein signal was calculated.
DNA combing was performed as described (Bianco et al, 2012) in mid‐log phase cell TK‐containing strains synchronized in G1 with α‐factor and released in the presence of 40 mM HU and 200 μg/ml BrdU. DNA fibers were extracted in agarose plugs after 60 and 120 min with BrdU as previously described (Duch et al, 2013).
Pulsed‐field and 2D gel electrophoresis
RNAPII and BrdU chromatin immunoprecipitation
RNAPII ChIPs were performed as described (Gonzalez‐Aguilera et al, 2011) using the monoclonal anti‐Rpb1‐CTD antibody 8WG16 (Berkeley Antibody Company, Richmond, CA, USA) and Dynabeads protein A (Invitrogen). The intergenic region at positions 9,716–9,863 of chromosome V was used as a negative control. BrdU ChIPs were performed in TK‐containing strains synchronized in G1 with α‐factor and released in the presence of 200 μg/ml BrdU (Sigma) using anti‐BrdU antibody (Medical & Biological Laboratories). As negative control we used the region between coordinates 240,032 and 243,611 of chromosome V. Primers used are indicated in Supplementary Table S7.
Microarray gene expression analysis
Global gene expression by microarray analysis was performed using total RNA isolated from yeast cells grown on YPD‐rich medium at 30°C to mid‐log phase by mechanical disruption using the RNeasy Midi kit (Qiagen). Microarray data analysis was performed in triplicate using the Affymetrix platform. RNA quality was confirmed with the Bioanalyzer H (Agilent technology). Synthesis, labeling and hybridization of cRNA to GeneChip® Yeast Genome 2.0 Arrays covering 5,841 genes of S. cerevisiae were performed following Affymetrix protocols. Probe signals were captured and processed with GeneChip Operating Software 1.4.0.036 (Affymetrix), and the resulting CEL files were reprocessed using the Robust Multichip Average (RMA) normalization (Irizarry et al, 2003). Fold change (log2) values (M) and their FDR‐adjusted P‐values were calculated with LIMMA (linear models for microarray analysis) (Smyth, 2004) using the affylmGUI interface (Wettenhall et al, 2006). Statistical analysis was performed using R language and the packages freely available from the ‘Bioconductor Project’ (http://www.bioconductor.org). Fold change cutoffs were analyzed at 95% confidence levels (FDR‐adjusted P‐values, 0.05). The expression profile of mutants was compared with that of its isogenic wild‐type strain cultured in the same conditions. Genes showing at least twofold expression change were considered as altered and selected for further bioinformatics analysis. The expression data can be accessed at Gene Expression Omnibus (GS55223). GO annotations were obtained with FatiGO, available in Babelomics (http://babelomics.bioinfo.cipf.es). Hits with P‐value < 0.05 were selected.
Saccharomyces cerevisiae high‐density oligonucleotide tiling microarrays from Affymetrix that allow an analysis of yeast chromosomes at a 300‐bp resolution, each of the 300‐bp region being covered by at least 60 probes, were used. ChIP‐chip of asynchronously growing cells was carried out basically as described (Bermejo et al, 2009; Gómez‐González et al, 2011a,b). Data were analyzed using a modified version of the Affymetrix Tiling Array Suite software (TAS), which produces a signal and the change P‐value per probe position taking into account the probes localized within a given bandwidth around the inspected probe. The ChIP‐chip data can be accessed at Gene Expression Omnibus (GS55184). Chromosomal distribution of the signals was analyzed by binding clusters, defined as ranges within the chromosome under the following conditions: estimated signal (IP/SUP‐binding ratio) positive in the whole range; P‐value < 0.01, minimum run of 100 bp and maximum gap of 250 bp. Results were visualized with the University of California Santa Cruz Genome Browser (http://genome.ucsc.edu/). Mapping of binding clusters into Stanford Genome Database genomic features (www.yeastgenome.org) was performed using specific Perl scripts. To visualize the distribution of binding sites along features, ORFs and ARS were divided into equivalent segments from the start and end coordinates as previously described (Gómez‐González et al, 2011a,b; Santos‐Pereira et al, 2013).
Cell cycle synchronization, flow cytometry analysis (FACSCalibur; Becton‐Dickinson), Northern and other molecular assays were performed following standard procedures.
IF‐A performed most of the experiments; JL‐B and MLG‐R performed part of the characterization of rpb1‐S751F. IF‐A and AA designed the experiments and wrote the paper.
Conflict of interest
The authors declare that they have no conflict of interest.
Source Data for Supplementary Figure S3B
Source Data for Supplementary Figure S8B
We would like to thank F. Malagón, S. Rodríguez‐Navarro, J. Strathern and R.A. Young for kindly providing reagents; F. Malagón for pointing us that rpb1S751F showed a synthetic growth defect with rad52; P. Domínguez (Microscopy Unit, CABIMER) for technical assistance in confocal microscopy; E. Andújar and M. Pérez (Genomics Unit, CABIMER) for technical support with microarray and ChIP‐chip experiments; Antonio Marín for providing bioinformatics tools for analysis of genome‐wide data; U. Galindo for technical assistance; and D. Haun for style supervision. Research was funded by grants from the Spanish Ministry of Economy and Competitiveness (Consolider 2010 CSD2007‐0015 and BFU2010‐16372), the Junta de Andalucía (CVI4567) and the European Union (FEDER). IF‐A and JL‐B were recipient of predoctoral training grants from the Spanish Ministry of Economy and Competitiveness and the Instituto Carlos III, respectively.
FundingSpanish Ministry of Economy and Competitiveness CSD2007‐0015 BFU2010‐16372
- © 2014 The Authors