Chromatin structure is modulated during deoxyribonucleic acid excision repair, but how this is achieved is unclear. Loss of the yeast Ino80 chromatin-remodeling complex (Ino80-C) moderately sensitizes cells to ultraviolet (UV) light. In this paper, we show that INO80 acts in the same genetic pathway as nucleotide excision repair (NER) and that the Ino80-C contributes to efficient UV photoproduct removal in a region of high nucleosome occupancy. Moreover, Ino80 interacts with the early NER damage recognition complex Rad4–Rad23 and is recruited to chromatin by Rad4 in a UV damage–dependent manner. Using a modified chromatin immunoprecipitation assay, we find that chromatin disruption during UV lesion repair is normal, whereas the restoration of nucleosome structure is defective in ino80 mutant cells. Collectively, our work suggests that Ino80 is recruited to sites of UV lesion repair through interactions with the NER apparatus and is required for the restoration of chromatin structure after repair.
DNA lesions that induce helical distortion are repaired by the versatile nucleotide excision repair (NER) apparatus, which has been well characterized through biochemical reconstitution studies (Aboussekhra et al., 1995; Guzder et al., 1995; Mu et al., 1995; Riedl et al., 2003; Staresincic et al., 2009). Knowledge of how NER occurs in the complex chromatin environment of the nucleus is limited. Chromatin is disrupted to permit efficient NER, and chromatin structure is restored after repair (the access–repair–restore model; Smerdon, 1991; Green and Almouzni, 2002; Dinant et al., 2008). Histone chaperones (Caf1 and Asf1) are required for the restoration of chromatin structure after NER (Mello et al., 2002; Polo et al., 2006). Less is known about how chromatin access is achieved during NER. Human switch/sugar nonfermentable and Drosophila melanogaster ATPase-remodeling factors stimulate NER reactions performed in vitro on nucleosomal templates (Ura et al., 2001; Hara and Sancar, 2003). The yeast Snf5/6-remodeling proteins contribute to efficient cellular NER (Gong et al., 2006), and histone acetyltransferases modulate in vivo rates of NER at certain genomic locations (Teng et al., 2008). Finally, ubiquitination of the histones H3 and H4 by the CUL4–DDB1–ROC1 complex regulates the recruitment of xeroderma pigmentosum group C to DNA damage in mammalian cells (Wang et al., 2006).
Chromatin remodeling during DNA double-strand break (DSB) repair has been reviewed in detail previously (for reviews see Downs et al., 2007; Osley et al., 2007; van Attikum and Gasser, 2009), and we will briefly summarize this work, focusing on the Ino80 chromatin-remodeling complex (Ino80-C). The Ino80-C is an ATPase capable of nucleosome sliding in vitro (Shen et al., 2000) and is recruited to the DSBs in a γ-H2A–dependent fashion, perhaps via its Arp4 and Nhp10 subunits (Downs et al., 2004; Morrison et al., 2004; van Attikum et al., 2004). The Ino80-C might displace nucleosomes in the vicinity of a DSB (Tsukuda et al., 2005; van Attikum et al., 2007; Chen et al., 2008), and Arp8 (a subunit of the Ino80-C) has been shown to influence the rate of loading of Rad51 at breaks, possibly through a role in nucleosome displacement, independent of H2A phosphorylation (Tsukuda et al., 2005). Most recently, several groups have implicated the Ino80-C in replication restart after replicative stress and in damage tolerance pathways during replication (Papamichos-Chronakis and Peterson, 2008; Shimada et al., 2008; Falbo et al., 2009). Here, we report a role for the Ino80-C during chromatin restoration associated with UV lesion repair in yeast.
Results and discussion
Cells lacking Ino80 are UV sensitive but globally repair photoproducts normally
Formal killing curves confirmed a moderate UV sensitivity for ino80 cells, as previously reported (Fig. 1 A; Shen et al., 2000). A strain co-deleted for ino80 and rad14 was no more sensitive than the rad14 strain, suggesting that ino80 is epistatic to NER factors (survival of ino80 single disruptant at 20 J/m2 = 54% and survival of the wild-type strain at 20 J/m2 = 84%; Fig. 1 B). Dot blot assays were used to monitor the removal of UV photoproducts from cellular DNA. A wild-type strain removes cyclobutane pyrimidine dimers (CPDs) almost completely over 3 h (Fig. 1, C and D). A rad14 NER mutant was, as expected, completely defective in this process. Quantification of the blots for ino80 cells revealed no significant defect in the removal of CPDs (Fig. 1 D), and consistent results were obtained by probing with an anti–6-4 photoproduct antibody (not depicted). Therefore, despite being UV sensitive and epistatic to rad14, ino80 mutants have no major global defect in the removal of UV photoproducts.
The UV sensitivity of ino80 strains might be caused by a checkpoint defect. Therefore, we monitored the accumulation of cells with large buds after irradiation, as wild-type yeast cells arrest at late S/G2 phase after UV (Weinert and Hartwell, 1993). Wild-type and ino80 cells exhibited an increase in budded cells within the first hour after irradiation, and both started to recover by 3 h, suggesting that ino80 cells invoke a normal UV checkpoint and recovery response (Fig. 1 E). Consistently, phosphorylation of the checkpoint kinase Rad53 after UV irradiation occurred with similar kinetics in both wild-type and ino80 cells (Fig. 1 F). Therefore, ino80 cells repair UV products on a genome-wide level normally, suggesting the role of Ino80 in contributing to survival after UV repair could relate to repair in specific genomic contexts.
Ino80-C has a damage-inducible interaction with Rad4–Rad23
We next explored the molecular relationship between NER factors and Ino80. We performed reciprocal coimmunoprecipitations using chromosomally FLAG-tagged (C terminal) Ino80 with core NER factors, including Rad23, Rad14, and Rad1–Rad10. Extracts from undamaged cells and from cells treated with 100 J/m2 UVC were examined. In one case, we could observe a robust UV-inducible interaction between Rad23 and Ino80-FLAG (Fig. 2, A and B). The interaction was identified in reciprocal experiments, immunoprecipitating with anti-FLAG and blotting with Rad23 and vice versa, and is absent in the isogenic untagged control strain (Fig. 2, A and B). The interaction between Ino80-FLAG and Rad23 is not mediated by DNA because the samples shown in Fig. 2 were all extensively treated with DNase, and the interaction is resistant to washing with buffers containing 0.5 M sodium chloride (Fig. 2 B). Rad23 has an obligate interacting partner during lesion recognition in NER, Rad4. However, Rad23 has Rad4-independent functions and is more abundant than Rad4 (Dantuma et al., 2009). We therefore asked whether Ino80 interacts with Rad4. After immunoprecipitation of Ino80-FLAG and immunoblotting for Rad4, a UV-inducible interaction between Ino80-FLAG and Rad4 was also detected (Fig. 2 C).
The damage-inducible interaction between Ino80-FLAG and Rad4–Rad23 suggests that Ino80 might be targeted to UV-damaged chromatin in a Rad4–Rad23-dependent manner. We tested this by purifying chromatin from cells after irradiation. In wild-type cells, Ino80-FLAG is recruited to chromatin after UV irradiation within 30 min (Fig. 2 D). Strikingly, in a rad4 strain, very little chromatin recruitment of Ino80-FLAG is observed, suggesting a key role for the interaction between Ino80 and Rad4 in recruiting Ino80 to UV-damaged chromatin (Fig. 2 D). We also determined whether the association of Rad4 with damaged chromatin is perturbed in the absence of Ino80 (Fig. 2 E). Rad4 was recruited to chromatin with similar kinetics in wild-type and ino80 cells, suggesting that Ino80 is not required for efficient damage recognition by Rad4 at a global level.
Ino80 is required for efficient UV lesion repair at HMLα
The global repair assay in Fig. 1 C cannot reveal subtle differences in repair rates at specific genomic loci. In particular, there could be a requirement for the Ino80-C for efficient repair in genomic regions with high nucleosomal occupancy. We therefore measured the rate of photoproduct repair at two loci with very different levels of nucleosome occupancy. In our a-mating–type strains, the HMLα locus is repressed, and 14 nucleosomes are bound at well-defined sites (Fig. 3 A; Weiss and Simpson, 1998). In contrast, the MATa locus is actively transcribed, and this is accompanied by a very low level of nucleosome occupancy (Fig. 3 A; Ravindra et al., 1999). We measured the rate of CPD removal at specific regions within MATa and HMLα (marked in Fig. 3 A) using a sensitive quantitative PCR–based assay (Fig. 3, B and C). For the chromatinized region within the HMLα locus, we observed that CPDs were efficiently removed over 3 h in the wild-type cells. Similar results were obtained for the nucleosome-free MATa locus, although repair was slightly more rapid, as expected. In ino80 cells, repair was very slightly delayed for the MATa region compared with wild-type cells. Moreover, a modest but significant reduction in repair was observed in ino80 cells at the chromatinized HMLα region, which was most apparent at the later time points (2 and 3 h). This suggests that Ino80 contributes to efficient repair at a chromatinized locus.
Ino80 is involved in nucleosome restoration at HMLα after UV
To examine chromatin dynamics at HMLα during UV lesion repair, we developed a novel chromatin immunoprecipitation (ChIP) methodology that permits analysis of protein occupancy at specified genomic locations in the presence of UV photoproducts (Fig. S1 A). This was necessary because UV photoproducts efficiently block the thermostable polymerases used in the PCR step of ChIP, such that fragments containing damage will be lost from analysis. In brief, after irradiation and a period of repair, cells are processed as for conventional ChIP. After immunoprecipitation and reversal of the DNA–protein cross-links, the DNA is treated with a mixture of CPD and 6-4 photoproduct photolyases and photoproducts reversed. This is essential to permit equivalent amplification of all the immunoprecipitated DNA in the sample, including fragments in which repair has yet to occur or be completed (Fig. S1, B and C).
Using this ChIP assay, we examined the nucleosomal region of HMLα previously analyzed in our repair experiments (Fig. 3). Antibodies recognizing the C terminus of histone H3, which bind all histone H3 present in nucleosomes regardless of their modification status or whether they also contain histone variants, were used. In a wild-type strain, histone H3 loss at a well-defined nucleosome within HMLα occurs almost immediately after irradiation, during the few minutes when irradiation and processing occurs, and thereafter occupancy is gradually restored over the following 3 h (Fig. 3 D). This time course is consistent with the kinetics of photoproduct repair in this strain at this locus (Fig. 3 C), where the majority of photoproducts are repaired in the first hour. ChIP performed after UV damage in cells disrupted for ino80 indicated a similar magnitude of initial reduction in histone H3 occupancy compared with wild-type cells (Fig. 3 D). However, recovery of histone H3 occupancy over the 3-h repair period is absent. This suggests that ino80 cells have a severely impaired nucleosome restoration capacity despite only exhibiting a relatively mild repair defect at this locus. Note that the initial nucleosomal occupancy in this region of HMLα is equivalent for both our wild-type strain and ino80 mutants in the undamaged state (unpublished data).
It was possible that the nucleosome loss observed after UV at HMLα was the result of replication fork blockage and nucleosome disassembly associated with replication-associated repair and tolerance. We therefore repeated this work in wild-type and ino80 cells held in the G1 phase of the cell cycle with α-mating factor for the duration of treatment and recovery time. The kinetics of H3 loss at HMLα and the repair kinetics of lesions in this region are not significantly different from those observed in asynchronous culture (Fig. 4, A and B), indicating that the majority of the nucleosome loss we observed is a consequence of a response to UV damage that is not associated with DNA replication.
We also examined histone H3 occupancy at HMLα after UV treatment in cac1 disruptants. Cac1 is a histone chaperone with a clearly established role in NER-associated chromatin restoration (Green and Almouzni, 2002). A cac1 strain exhibited a strong decrease in H3 occupancy immediately after UV (Fig. S2 A) and exhibited little restoration of H3 signal over the following 3 h, similar to the ino80 strain. Interestingly, however, Cac1 is not participating in the same restoration reaction as the Ino80-C during UV damage repair, as an arp8 cac1 double mutant is not epistatic to its cognate single mutants for UV sensitivity (Fig. S2 B). Moreover, Cac1 is not required for the efficient recruitment of Ino80-FLAG to chromatin after UV damage (Fig. S2 C). This might reflect the fact that Cac1 is a histone H3 and H4 chaperone, whereas the Ino80-C has previously been shown to impact the dynamics of histone H2A and its variants, in which both might be required for efficient chromatin restoration during UV damage repair. Regardless, the defect in H3 restoration seen in the cac1 strain is consistent with several previous studies (Mello et al., 2002; Green and Almouzni, 2003; Polo et al., 2006) and supports our assertion that we are detecting NER-associated nucleosome remodeling using the modified ChIP assay.
Rad4 is required for efficient repair-associated remodeling at HMLα and UV-induced recruitment of Ino80
The Rad4 (and mammalian xeroderma pigmentosum group C) protein plays a key role in initiating chromatin remodeling associated with UV repair (Baxter and Smerdon, 1998; Gong et al., 2006). Consistently, we observed an initial delay in reduction of H3 occupancy in rad4 cells compared with the wild-type strain (Fig. 4 C), although by 1 h a reduction in histone H3 occupancy was observed. In contrast, a rad14 strain did not exhibit any delay in initial remodeling, whereas a delay in histone restoration was apparent (Fig. 4 C). As expected, no repair was observed at HMLα in the course of these experiments in rad4 or rad14 strains (unpublished data). Therefore, damage recognition by Rad4 is needed to trigger rapid remodeling, but Rad14 is not essential to trigger this response. Moreover, NER must proceed to permit chromatin restoration because this step is eliminated in both rad4 and rad14 cells.
We next determined whether Ino80 is enriched at HMLα after UV and whether this enrichment is Rad4 dependent (Fig. 4 D). In wild-type cells, Ino80-FLAG is recruited to HMLα, maximally at 1 h after irradiation, and is in decline within 2–3 h. This is consistent with the Ino80-dependent histone H3 restoration kinetics observed in wild-type cells, which is near complete at 3 h after irradiation (Fig. 3 D), and also the global UV recruitment kinetics of Ino80 (Fig. 2 D). In rad4 cells, enrichment of Ino80-FLAG was strongly reduced after irradiation (Fig. 4 D), demonstrating that Rad4 is required to target Ino80 to HMLα after UV irradiation (Fig. 2 D). However, because repair is absent in rad4 cells and repair is a prerequisite for chromatin restoration (Fig. 4 C), it could be argued that the failure to recruit Ino80-FLAG to HMLα is not directly a result of the absence of Rad4 but is caused by a lack of repair. To address this, we also determined Ino80-FLAG recruitment to HMLα in the absence of Rad14, which is recruited after damage recognition by Rad4 and does not interact with Ino80-FLAG. In rad14 cells, we observed robust recruitment of Ino80-FLAG to HMLα after UV (Fig. 4 D), despite the fact that repair is absent and chromatin restoration is defective (Fig. 4 C). Moreover, a global chromatin-binding experiment confirmed that Ino80-FLAG chromatin recruitment is efficient in rad14 cells (Fig. S3), in contrast to rad4 cells (Fig. 2 D). Collectively, these data show that the Rad4 damage recognition factor is required for the recruitment of Ino80-FLAG to UV-damaged chromatin regardless of whether repair by NER then proceeds.
Cells lacking Ino80 are UV sensitive but exhibit normal global UV photoproduct repair kinetics. Examination of a highly chromatinized region of the yeast genome, however, revealed a modest, but significant, reduction in the repair of lesions in ino80 cells. Moreover, Ino80 interacts with Rad4–Rad23 and is recruited to chromatin after UV damage in a Rad4-dependent manner. The Ino80-C is capable of ATP-dependent nucleosome sliding, which suggests a role in nucleosome reorganization during repair. Initial experiments of a role for Ino80 in nucleosome remodeling at DSBs supported a role in nucleosome displacement (Tsukuda et al., 2005), but a recent study argued that because displacement is linked to DSB resection and because resection is delayed in ino80 mutants, the precise role of the Ino80-C in DSB processing remains unclear (Chen et al., 2008). Our work revealed a wild type–like reduction in nucleosome occupancy at the HMLα after UV. Strikingly though, during repair at the HMLα locus, ino80 cells fail to restore histone occupancy.
It is interesting that defective chromatin restoration is associated with a modest decrease in repair rate at this highly chromatinized locus. However, because there will be multiple repair reactions occurring within any region of DNA in close succession or simultaneously, a compromised ability to restore chromatin during the completion of NER might interfere with efficient repair within adjacent regions. In fact, there are precedents for this because defective chromatin assembly on nascent DNA (behind the replication fork) negatively impacts fork progression in S phase (Ransom et al., 2010). Moreover, although it is well established that NER is accompanied by significant changes in chromatin structure, it remains unknown whether these changes are the result of nucleosome sliding, eviction, or a combination of both (Green and Almouzni, 2002). Which is favored will profoundly affect the chromatin landscape during NER. Alternatively, it is possible that the repair defect we observe in ino80 cells is caused by a direct role for Ino80 in facilitating the NER process itself in chromatin, unrelated to nucleosome displacement or restoration activities. However, because the repair defect at HMLα in ino80 mutants is modest, but the chromatin restoration defect severe, our data clearly identify a key role for Ino80 in chromatin restoration regardless of its possible minor contribution to the core NER reaction in chromatin.
A role for the Ino80-C in nucleosomal restoration has recently been reported during promoter remodeling in response to stress (Klopf et al., 2009). Our data are also consistent with the Ino80-C playing a role in restoring chromatin structure, but in the context of NER. The contribution of the Ino80-C to this process could involve a direct role in nucleosome deposition. Alternatively, the Ino80-C might act as a nucleosome acceptor during remodeling for repair, transiently sequestering displaced nucleosomes or stabilizing remodeled nucleosomes during repair and cooperating with other assembly factors to restore chromatin structure once repair is completed. Finally, because the Ino80-C is well conserved in mammalian cells, this work has clear relevance to the human DNA damage response (Jin et al., 2005).
Materials and methods
Yeast strains and plasmids
Yeast strains are described in Table I.
Dot blot analysis of photoproduct formation and repair
The method of McCready et al. (1993) was used in this study. Antibodies against UV photoproducts used in this study were anti-CPD (Affitech) and anti–6-4 (Stratech) at a dilution of 1:1,000.
Quantitative PCR photoproduct repair assay
The quantitative PCR photoproduct repair assay was based on the method of Kalinowski et al. (1992) but with the following modifications. Each quantitative PCR reaction contained 30 ng of genomic DNA and 10 pmol of primers: HMLα1 (5′-TTTACTTCGAAGCCTGCT-3′), HMLα2 (5′-ATCTTTCTTGAGTGGTCG-3′), MATa1 (5′-AGAGGTCCGCTAATTCTG-3′), MATa2 (5′-CTTACAGAGGACACCGGT-3′), and SensiMix from the SYBR kit (Bioline). Samples were analyzed on the Research Rotor Gene 3000 (Corbett Life Science).
Strains with chromosomally FLAG-tagged INO80 were grown to logarithmic phase, UV damaged with 100 J/m2 UVC or mock irradiated, and allowed to repair for 45 min in rich media (YEPD [yeast extract, peptone, dextrose]). Whole-cell extracts were then immunoprecipitated with anti-Rad23, and protein was analyzed on 4–12% Bis-Tris Gels (Invitrogen) blotted with an anti-FLAG mouse polyclonal antibody (1:1,000 dilution; Abcam). FLAG immunoprecipitations were performed using M2 anti-FLAG resin (Sigma-Aldrich), and elution was performed with 3× anti-FLAG peptide (Sigma-Aldrich). Gels were immunoblotted with polyclonal anti-Rad23 (1:1,000) or anti-Rad4 (1:5,000). Where DNase I treatment was used, extracts were treated with 5 U for 1 h at 37°C.
Chromatin-binding assays were performed essentially as described previously (Liang and Stillman, 1997). In brief, 109 cells were harvested, spheroplasted using Zymolase 20T at 30°C for 15 min, and then lysed in 100 mM KCl, 50 mM Hepes-KOH, pH 7.5, 2.5 mM MgCl2, 50 mM NaF, 0.25% of final volume Triton X-100 supplemented with protease inhibitor tablets (Roche), and 1 mM PMSF. Chromatin, nuclear matrix, and cell debris were spun down at 4,000 rpm for 5 min. The crude pellet was treated with 0.6 U of limited microccocal nuclease digestion in the aforementioned buffer for 3 min at 37°C and spun down at 4,000 rpm for 10 min. The solubilized polynucleosomes were then pelleted at high speed (50,000 rpm) for 1 h at 4°C.
Formaldehyde (TAAB Laboratories) was added to cells to a final concentration of 1%, and incubation for 30 min at room temperature followed. Glycine was added to a final concentration of 125 mM, and cells were incubated for 5 min. Cell pellets were resuspended in lysis buffer (50 mM Hepes-NaOH, pH 7.5, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, and 0.1% sodium deoxycholate) containing protease inhibitor solution (Complete; Roche) and 1 mM PMSF together with acid-washed glass beads and cells lysed using a bead beater. Debris was removed by centrifugation for 5 min at 2,000 rpm. Chromatin was sheared by sonication at 40% output for 5 s eight times before the lysate was clarified by centrifugation and twice at 13,000 rpm for 20 min. 2 µg antihistone H3 ChIP-grade antibody (Abcam) was added to 500 µg of protein extract and rotated overnight at 4°C. Cell extracts were also frozen for processing as input controls. 40 µl of 50% slurry protein G Sepharose was added to the protein–antibody mixture and rotated for 1 h. The beads were collected by centrifugation at 13,000 rpm for 15 s and washed with 1.5 ml of lysis buffer. The beads were washed again with 1.5 ml of lysis buffer containing 500 mM NaCl followed by 1.5 ml radioimmunoprecipitation assay and 1.5 ml TE (Tris-EDTA), pH 8.0. The beads were resuspended in 100 µl TE, pH 8.0, and incubated with 20 µg RNase A (Sigma-Aldrich) for 30 min at 37°C. 1 ml TE, pH 8.0, was added, and beads were collected by centrifugation. Protein–DNA complexes were eluted by incubating twice with 250 µl of elution buffer (1% SDS and 0.1-M NaHCO3). 20 µl of 5-M NaCl was added to the eluate and input samples, which were incubated overnight at 65°C. 1 ml ethanol was added to each sample, and samples were incubated overnight at −20°C before being centrifuged for 15 min at 13,000 rpm at 4°C. The pellet was washed with 70% ethanol, resuspended in 100 µl TE, pH 8.0, and treated with 20 µg proteinase K at 50°C for 30 min. The DNA was extracted once with phenol:chloroform:isoamyl alcohol followed by extraction of the organic phase with TE. The DNA was precipitated and then resuspended in 50 µl of TE, pH 8.0. 50 µl of extracted DNA was incubated with reactivation buffer (20 mM Tris-HCl, pH 7.8, 50 mM NaCl, 1 mM EDTA, and 1 mM DTT [Sigma-Aldrich]), 5 mM DTT, 5 µg D. melanogaster 6-4 photolyase (purified according to Kim et al., 1994; a gift from A. Sancar, University of North Carolina at Chapel Hill, Chapel Hill, NC), and 50 ng Escherichia coli CPD photolyase (R&D Systems). The reactions were incubated under a blacklight bulb and conventional lightbulb for the 6-4 and CPD enzymes, respectively, for 1 h at room temperature. After incubation, the reactions were purified using a purification kit (QIAquick PCR; QIAGEN). The column was eluted using 100 µl of deionized H2O. Quantitative PCR reactions contained 13 µl of SensiMix SYBR kit, 10 pmol of the forward and reverse primers (sequences as used in the Quantitative PCR photoproduct repair assay section at HMLα), 5 µl of immunoprecipitated DNA, and 5 µl of 0.02-ng pUC19. Reactions were performed on the Research Rotor Gene 3000, once using mastermix containing primers to specific regions of HMLα and once using M13 primers to pUC19. Threshold cycle values for immunoprecipitated DNA were normalized to both input and no antibody controls.
Immunoblots and protein analysis
The anti-Rad4 antibody was a gift from S. Reed (Cardiff University, Cardiff, Wales, UK) and was used at a dilution of 1:5,000. Polyclonal anti-Rad10, anti-Rad23, and anti-Rad53 antibodies were obtained from Santa Cruz Biotechnology, Inc., and anti-Rad14 was obtained from Abcam. Each was used at a dilution of 1:1,000.
Online supplemental material
Fig. S1 illustrates the modified ChIP assay schematically and the reasons for its development. Fig. S2 confirms that Cac1 is required for chromatin restoration after UV damage but probably contributes to a different reaction than Ino80. Fig. S3 shows that Rad14 is not required for Ino80-FLAG recruitment to UV-damaged chromatin.
We are grateful to Ian Hickson and Leonard Wu for helpful comments on the manuscript. We thank Simon Reed, Xuetong Shen (University of Texas MD Anderson Cancer Center, Smithville, TX), and Aziz Sancar for sharing yeast strains, plasmids, and antibodies.
This work was supported by Cancer Research UK.
S. Sarkar and R. Kiely contributed equally to this paper.
R. Kiely’s present address is Dept. of Biochemistry, University of Oxford, Oxford OX1 3QU, England, UK.