Genome-wide CRISPR screens identify the EXO1-CAF-1 pathway suppressing R-loop-associated DNA damage
Alexandra Nusawardhana, Anastasia Hale, Joshua Straka, Claudia M Nicolae, George-Lucian Moldovan
TL;DR
This study identifies a DNA repair pathway involving EXO1 and CAF-1 that suppresses DNA damage caused by R-loops, revealing their roles in genomic stability.
Contribution
The study reveals a novel EXO1-CAF-1 pathway for suppressing R-loop-associated DNA damage and identifies synthetic lethality between these proteins.
Findings
EXO1 is critical for suppressing cisplatin-induced DNA damage in wildtype cells.
EXO1 and CAF-1 independently suppress R-loops and work synergistically to prevent DNA damage.
Loss of both EXO1 and CAF-1 leads to R-loop accumulation and increased DNA damage even without DNA damage treatment.
Abstract
DNA repair is critical for cellular homeostasis under both normal conditions as well as in response to genotoxic agents such as chemotherapeutics. EXO1 is a 5′–3′ exonuclease with multiple roles in DNA biology. To better understand these roles, we employed CRISPR loss-of-function genome-wide screening to identify genes required for proliferation and cisplatin sensitivity in EXO1-deficient cells. We uncovered differential regulators of cisplatin sensitivity between wildtype (WT) and EXO1-deficient cells. By analyzing the genetic networks that these regulators belong to, we found that DNA repair was the main biological process suppressing cisplatin sensitivity in WT cells, but this was not the case in EXO1-deficient cells, indicating that EXO1 is critical for the repair of cisplatin-induced DNA damage. Moreover, synthetic lethality screens identified a genetic interaction between EXO1 and…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6- —NIH10.13039/100000002
- —Four Diamonds Transformative Patient-Oriented Cancer Research
- —National Institutes of Health10.13039/100000002
- —Penn State University10.13039/100008321
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsDNA Repair Mechanisms · PARP inhibition in cancer therapy · Microtubule and mitosis dynamics
Introduction
Despite recent advances in targeted cancer therapies, platinum-based chemotherapies (cisplatin or carboplatin) still represent the main treatment for many types of cancers [1]. While effective, platinum chemotherapy has major side effects which lower the quality of life of patients, including the loss of hematopoietic stem cells resulting in increased infection risk, as well as intestine dysfunction resulting in gastrointestinal distress and weight loss. In addition, cisplatin treatment can also cause heart and kidney dysfunction. Moreover, tumors frequently become resistant to cisplatin.
Mechanistically, cisplatin acts by creating intrastrand and interstrand DNA adducts between adjacent guanines [2–5]. These adducts block the progression of DNA polymerases, arresting replication forks [6–8], and ultimately kill rapidly dividing cancer cells, presumably by overwhelming the cellular machinery that handles stalled forks restart. Unless efficiently restarted, stalled replication forks can be nucleolytically processed resulting in generation of DNA double-stranded breaks (DSBs), referred to as fork collapse. This phenomenon of increased fork stalling and generation of aberrant fork structures, referred to as replication stress [6, 7], can ultimately cause cell death.
To avoid fork breakage, cells are equipped with mechanisms that stabilize and restart the fork, thus promoting genomic stability. Since they help tolerate chemotherapy-induced DNA lesions, these mechanisms also promote tumor chemoresistance. Translesion synthesis (TLS) is a mechanism that employs specialized, low fidelity polymerases able to replicate across DNA lesions [9, 10]. Upon fork arrest at DNA lesions, the ubiquitination of the polymerase co-factor PCNA signals a switch from the replicative polymerases to TLS polymerases to bypass DNA adducts allowing the progression of DNA synthesis in a continuous manner. The TLS polymerase Polη was previously shown to bypass cisplatin–DNA adducts [11–13]. If not bypassed, stalled replication forks can be restarted downstream of the lesion by the PrimPol primase–polymerase [14, 15], forming a single-stranded DNA (ssDNA) gap at the site of the adduct. This gap can be repaired by BRCA-mediated recombination with the sister chromatid [16]. The BRCA pathways can also repair collapsed forks through homologous recombination (HR) DNA repair [17]. Finally, the Fanconi anemia (FA) pathway was shown to coordinate the removal of interstand crosslinks and thus promote the repair of cisplatin DNA lesions [18].
EXO1 is an exonuclease with 5′–3′ activity which participates in multiple DNA repair processes, generally involving exonucleolytic processing of DNA lesions [19]. EXO1 catalyzes long-range DNA resection at double-stranded DNA breaks (DSBs) [20] to generate long ssDNA overhangs. Through this process, EXO1 is critical for BRCA-mediated HR repair of DSBs. In addition, EXO1 degrades nascent DNA strands containing mismatched bases during DNA mismatch repair (MMR) [21]. EXO1 also processes single stranded DNA (ssDNA) gaps as well as reversed replication forks, driving genomic instability in BRCA-mutant cells [16, 22–37]. In addition, noncatalytic roles of EXO1 in genomic stability have been described, such as the recruitment of BLM and FANCM to resolve RNA–DNA hybrids (R-loops) [38]. These findings highlight the multifaceted roles of EXO1 in DNA metabolism. The exact regulation of EXO1 in these processes, and the implications for genome stability and chemotherapy response in various cancer genetic backgrounds are unclear.
It is hoped that targeted cancer therapies can replace chemotherapy for the treatment of many tumors, especially those with specific, predisposing genetic mutations. For example, PARP1 inhibitors can specifically kill BRCA-mutant cancer cells [39, 40]. Thus, identifying synthetic lethal interactions can not only improve our understanding of biological mechanisms, but also provide new targets for targeted cancer therapy.
Genome-wide CRISPR genetic screens are powerful tools for identifying clinically relevant genetic interactions, such as synthetic lethality interactions, as well as genetic biomarkers of drug response [41, 42]. Here, we employed CRISPR loss-of-function genome-wide screening to identify genes required for proliferation and cisplatin sensitivity of EXO1-deficient cells. We uncovered differential regulators of cisplatin sensitivity between wildtype (WT) and EXO1-deficient cells. By analyzing the genetic networks that these regulators belong to, we found that DNA repair was the main biological process suppressing cisplatin sensitivity in WT cells, but this was not the case in EXO1-deficient cells, indicating that EXO1 is critical for the repair of cisplatin-induced DNA damage. Moreover, synthetic lethality screens identified a genetic interaction between EXO1 and the histone chaperone CAF-1. We show that, even in the absence of DNA damage treatment, concomitant depletion of EXO1 and CAF-1 results in synergistic R-loop accumulation and DSB formation, thus explaining the genetic interaction observed. This highlights EXO1 as a critical genome stability factor.
Materials and methods
Cell culture and protein techniques
HeLa cells (obtained from ATCC) were grown in Dulbecco’s modified Eagle’s media. To knock-out EXO1 in HeLa cells, a commercially available CRISPR/Cas9 KO plasmid (Santa Cruz Biotechnology sc-402356) was used, as we previously described [23]. Transfected cells were FACS-sorted into 96-well plates using a BD FACSAria II instrument. Resulting colonies were screened by western blot. For the complementation of EXO1-knockout cells with WT or D173A EXO1 variants, pLV[Exp]-Puro-CMV > hEXO1 lentiviral constructs (VectorBuilder) were used. For RNAseH1 overexpression, the pLV[Exp]-Puro-CMV > hRNASEH1 lentiviral construct (VectorBuilder) was used. Infected cells were selected by puromycin.
Gene knockdown was performed using Lipofectamine RNAiMAX. AllStars Negative Control siRNA (Qiagen 1027281) was used as control. CHAF1A knockdown was obtained using SilencerSelect siRNA (ID: s19499, Thermo Fisher).
For sgRNA depletion, the commercially available dual guide RNA (gRNA) CRISPR lentiviral vector pLV[2CRISPR]-hCas9:T2A:Puro-U6>[gRNA#1]-macU6>[gRNA#2] (VectorBuilder) was used. The following gRNA sequences were used:
CHAF1A: GATCCCCAGTGTCTTCTCGC and ATGTTCGCCACGGAGCTGCC;
Scramble: GTGTAGTTCGACCATTCGTG and GTTCAGGATCACGTTACCGC.
Denatured whole cell extracts were prepared by boiling cells in 100 mM Tris, 4% sodium dodecyl sulphate, 0.5 M β-mercaptoethanol. Antibodies used for western blot, at 1:500 dilution, were:
EXO1 (Novus NBP2-16391);
CHAF1A (Cell Signaling Technology 5480s);
RNAseH1 (Invitrogen PA5-PA598119);
Vinculin (Santa Cruz Biotechnology sc-73614):
GAPDH (Santa Cruz Biotechnology sc-47724).
CRISPR screens
For CRISPR knockout screens, the Brunello Human CRISPR knockout pooled lentiviral library (Addgene 73179) was used [43]. This library encompasses 76 411 gRNAs that target 19 114 genes. A total of 55 million cells from each cell lines (HeLa-WT, HeLa-EXO1^KO#1^, HeLa-EXO1^KO#3^) were infected with this library at a multiplicity of infection of 0.4 to achieve 250-fold coverage and selected for 4 days with 0.6 μg/ml puromycin. Twenty million library-infected cells (to maintain 250-fold coverage) were passaged for 11 days in the presence of 0.2 μM cisplatin and then collected. Cisplatin was prepared freshly each time, by solubilizing it directly in media, right before being added to cells. Genomic DNA was isolated using the DNeasy Blood and Tissue Kit (Qiagen 69504) and employed for PCR using Illumina adapters to identify the gRNA representation in each sample. Then, 10 μg of gDNA was used in each PCR reaction along with 20 μl 5× HiFi Reaction Buffer, 4 μl of P5 primer, 4 μl of P7 primer, 3 μl of Radiant HiFi Ultra Polymerase (Stellar Scientific), and water. The P5 and P7 primers were determined using the user guide provided with the CRISPR libraries (https://media.addgene.org/cms/filer_public/61/16/611619f4-0926-4a07-b5c7-e286a8ecf7f5/broadgpp-sequencing-protocol.pdf). The PCR cycled as follows: 98°C for 2 min before cycling, then 98°C for 10 s, 60°C for 15 s, and 72°C for 45 s, for 30 cycles, and finally 72°C for 5 min. After PCR purification, the final product was Sanger sequenced to confirm that the guide region is present, followed by qPCR to determine the exact amount of PCR product present. The purified PCR products were then sequenced with Illumina HiSeq 2500 single read for 50 cycles. Between 10 million and 44 million reads were obtained for each sample. Next, the sequencing results were analyzed bioinformatically using the MAGeCK algorithm, which takes into consideration raw gRNA read counts to test if individual guides vary significantly between the conditions [44]. The MAGeCK software and instructions on running it were obtained from https://sourceforge.net/p/mageck/wiki/libraries/. Finally, functional analysis clustering and Gene Ontology pathways analyses of the enriched among the top hits was performed using DAVID [45, 46].
Functional assays
Neutral and BrdU alkaline comet assays were performed as previously described [37] using the Comet Assay Kit (Trevigen, 4250-050). For the BrdU alkaline comet assay, cells were incubated with 100 μM BrdU as indicated. Slides were stained with anti-BrdU (BD 347580) antibodies and secondary AF568-conjugated antibodies (Invitrogen A-11031) and imaged on a Nikon microscope operating the NIS Elements V1.10.00 software. Olive tail moment was analyzed using CometScore 2.0. Immunofluorescence was performed as previously described [47] using a γH2AX antibody (MilliporeSigma JBW301). Slides were imaged on a confocal microscope (Leica SP5) and analyzed using ImageJ 1.53a software. For clonogenic survival assays, 500 cells were seeded in 6-well plates and treated with cisplatin, or left untreated, as indicated. Media was changed after 3 days. Colonies were stained after 10–14 days. Colonies were washed with PBS, fixed with a solution of 10% methanol and 10% acetic acid, and stained with 1% crystal violet (Aqua Solutions).
Proximity ligation-based assays
For proximity ligation assays (PLAs), cells were seeded into eight-chamber slides and 24 h later, were permeabilized with 0.5% Triton for 10 min at 4°C, washed with PBS, fixed at room temperature with 3% paraformaldehyde in PBS for 10 min, washed again in PBS and then blocked in Duolink blocking solution (Millipore Sigma DUO82007) for 1 h at 37°C, and incubated overnight at 4°C with primary antibodies. Antibodies used were: S9.6 (Kerafast ENH001), double-stranded DNA (dsDNA; Novus NBP3-07302), γH2AX (Abcam ab2893), EXO1 (Novus NBP2-16391), and CHAF1A (Cell Signaling Technology 5480s). Samples were then subjected to a proximity ligation reaction using the Duolink kit (Millipore Sigma DUO92008) according to the manufacturer’s instructions. Slides were imaged using a confocal microscope (Leica SP5) and images were analyzed using ImageJ 1.53a software.
Statistics analyses
For immunofluorescence, PLA, and clonogenic assays the t-test (two-tailed, unpaired) was used. For comet assays the Mann–Whitney statistical test (two-tailed) was performed. Statistical analyses were performed using GraphPad Prism 10. Statistical significance is indicated for each graph (ns = not significant, for *P *>.05; * for P ≤.05; ** for P ≤.01; *** for P ≤.001, **** for P ≤.0001). The random probabilities of identical genes within the top hits with MAGeCK score lower than 0.005 were calculated by multiplying the individual probabilities of each set: [(number of genes in set 1/total number of genes in the library) × (number of genes in set 2/total number of genes in the library)].
Results
CRISPR knockout screens identify differential regulators of cisplatin sensitivity in WT and EXO1-deficient cells
EXO1 has been involved in multiple DNA repair mechanisms. DNA repair is a critical response to genotoxic chemotherapy such as cisplatin treatment. Previous studies have shown that loss of EXO1 renders cancer cells sensitive to cisplatin [48]. To further interrogate the biological roles of EXO1, we performed a series of genetic screens in WT (parental) and EXO1-depleted HeLa cells. We first created EXO1-knockout HeLa cells using the CRISPR method. Two different, independently obtained knockout clones (EXO1^KO#1^ and EXO1^KO#3^) (Supplementary Fig. S1) were subsequently used for the screens, for increased rigor. We then lentivirally infected the WT parental and the two EXO1-knockout cell lines with the Brunello genome-wide CRISPR knockout library [43], which targets 19 114 human genes with an average of four gRNAs for each gene, for a total of 76 441 unique gRNAs. After selection, taking care to maintain 250× fold library coverage (equivalent to 20 million cells) at all times, we divided the library-infected populations in two groups: one was grown under normal conditions, while the other was treated with 0.2 µM cisplatin for 11 days, splitting cells every 3–4 days with cisplatin added freshly every time (Fig. 1A). This low-dose, longer-term treatment was designed to mimic the use of cisplatin in the clinic.
Genome-wide CRISPR knockout screens for proliferation and cisplatin sensitization of HeLa WT and EXO1-knockout cells. (A) Overview of the CRISPR knockout screens to identify genes that are required for proliferation and cisplatin sensitivity of WT and EXO1-knockout HeLa cells. Created in BioRender. Moldovan, G. (2026); https://BioRender.com/c77sziu. (B) The cellular survival of WT and EXO1-knockout HeLa cells at each splitting time. Survival was calculated by dividing the number of live cells in the cisplatin-treatment population to the control (no treatment) population.
The cisplatin sensitivity of each cell line was assessed by dividing the number of live cells in the cisplatin-treatment population to the control (nontreatment) at every splitting time. In line with the previous studies cited above, both EXO1 knockout clones displayed a slight sensitivity to cisplatin (Fig. 1B and Supplementary Fig. S2). At the end of the cisplatin treatment, cells were collected, and genomic DNA was extracted. The gRNA region was amplified by PCR and identified by Illumina sequencing. Bioinformatic analyses using the MAGeCK algorithm [42] were used to analyze the screens.
We first compared the cisplatin treatment to nontreatment controls for each cell line. The MAGeCK algorithm was used to generate ranking lists of genes which were lost in cisplatin-treated cells compared to untreated cells (Supplementary Tables S1–S3). This represents genes which, when inactivated, result in reduced viability in cisplatin-treated cells compared to untreated cells. We then performed biological pathway analyses of the top hits with MAGeCK RRA score lower than 0.005. As expected from the known role of the DNA repair system in repairing cisplatin-induced DNA damage, DNA repair was the top enriched biological process suppressing cisplatin sensitivity in WT cells (Fig. 2A and Supplementary Fig. S3A). Gene Ontology pathway analyses of the top hits with MAGeCK score lower than 0.005 showed multiple FA (UBE2T, FANCA, FANCB, FANCF, FANCL), HR (RAD51B, RAD54L), and PCNA ubiquitination (RAD18, HLTF, HUWE1) factors as top hits (Fig. 2B), in line with the known roles of the FA/BRCA and TLS pathways in the bypass of cisplatin adducts during DNA replication.
Analyses of the cisplatin sensitivity CRISPR screens in WT and EXO1-knockout cells. (A) Functional annotation clustering of the top hits with MAGeCK score lower than 0.005 which cause cisplatin sensitivity in WT HeLa cells, using Gene Ontology and Uniprot terms. (B) Table showing the biological processes and corresponding genes from the Gene Ontology pathway analysis of the top hits with MAGeCK score lower than 0.005 which cause cisplatin sensitivity in WT HeLa cells. GO_BP terms with negative logP >1 are presented. (C, D) Functional annotation clustering of the top hits with MAGeCK score lower than 0.005 which cause cisplatin sensitivity in HeLa-EXO1KO#1 (C) and HeLa-EXO1KO#3 (D) cells using Gene Ontology and Uniprot terms. (E) Diagram showing the overlap of identical genes within the top cisplatin sensitivity hits in the two EXO1-knockout cell lines. The number of genes with MAGeCK score lower than 0.005 (left) and 0.015 (right) are shown. (F) The number of common genes within the top cisplatin sensitivity hits in the two EXO1-knockout cell lines compared to the random probability of identical hits. The top hits with MAGeCK score lower than 0.005 (left) and 0.015 (right) were included in the analysis. (G) Diagram showing the overlap of identical genes within the top cisplatin sensitivity hits in WT and EXO1KO#1 cells. The number of genes with MAGeCK score lower than 0.005 (left) and 0.015 (right) are shown. (H) The number of common genes within the top cisplatin sensitivity hits in WT and EXO1KO#1 cells compared to the random probability of identical hits. The top hits with MAGeCK score lower than 0.005 (left) and 0.015 (right) were included in the analysis. (I) Diagram showing the overlap of identical genes within the top cisplatin sensitivity hits in WT and EXO1KO#3 cells. The number of genes with MAGeCK score lower than 0.005 (left) and 0.015 (right) are shown. (J) The number of common genes within the top cisplatin sensitivity hits in WT and EXO1KO#3 cells compared to the random probability of identical hits. The top hits with MAGeCK score lower than 0.005 (left) and 0.015 (right) were included in the analysis.
In contrast to WT cells, DNA repair did not feature among the top biological processes suppressing cisplatin sensitivity in either of the EXO1-knockout cell lines. DNA repair was not found as an enriched biological process in the EXO1^KO#1^ cells and was ranked sixth out of nine processes in the EXO1^KO#3^ cells (Fig. 2C and D, and Supplementary Fig. S3B and C). This surprising result indicates that EXO1 is central to the DNA repair response upon cisplatin treatment. In the absence of EXO1, DNA repair as a biological process is not a meaningful component of cisplatin resistance any longer. This likely reflects the role of EXO1 in multiple DNA repair mechanisms, and suggests that in the absence of EXO1, the DNA repair system as a whole is compromised, at least in terms of cisplatin–DNA lesion repair.
We also analyzed the cisplatin sensitivity gene ranking lists in order to assess the robustness of our screens. Between the two EXO1 knockout lines, there were 11 genes in common within the top hits with MAGeCK score lower than 0.005 (329 genes for the EXO1^KO#1^ and 300 for EXO1^KO#3^) (Fig. 2E). This is higher than the random probability of common genes within two datasets of these sizes, which is 5.2 (Fig. 2F). When increasing the cutoff to a MAGeCK score lower than 0.015, there were 58 genes in common (Fig. 2E). This is higher than the random probability of common genes within two datasets of these sizes, which is 31.5 (Fig. 2F). In contrast, when comparing the cisplatin sensitivity gene list in WT cells to each of the knockout lines, the actual number of common genes was similar to the random probabilities (Fig. 2G–J). Overall, this analysis suggests that our screens were able to identify specific hits regulating the cisplatin sensitivity of WT and EXO1-deficient cells.
CRISPR knockout screens identify synthetic lethal interactions of EXO1
Synthetic lethal interactions have emerged as a new frontier in identifying novel targets for precision oncology, because of the large number of mutations observed in most tumors. Mutations in EXO1 have indeed been found in tumors. To identify EXO1 synthetic lethal interactions, we analyzed the control (nontreatment) arms of the WT and EXO1-knockout screens. We used the MAGeCK algorithm to generate ranking lists of genes which were lost in each of the EXO1-knockout cell lines compared to WT (Supplementary Tables S4 and S5). This represents genes which, when inactivated, reduce the viability of EXO1-knockout cells compared to WT cells.
When comparing the results of this synthetic lethality analysis, we found a high overlap between the top hits in the two EXO1-knockout cell lines. There were 28 genes in common within the top hits with MAGeCK score lower than 0.005 (315 genes for the EXO1^KO#1^ and 348 for EXO1^KO#3^) (Fig. 3A). This is much higher than the random probability of common genes within two datasets of these sizes, which is 5.7 (Fig. 3B). When increasing the cutoff to a MAGeCK score lower than 0.015, there were 130 genes in common (Fig. 3A). This is much higher than the random probability of common genes within two datasets of these sizes, which is 34.1 (Fig. 3B). Biological pathway analyses of these 130 genes revealed that DNA repair was the top process conferring synthetic lethality to EXO1-deficient cells (Fig. 3C). This further spotlights EXO1 as a crucial DNA repair factor, but also highlights the critical role of DNA repair in maintaining cellular homeostasis, even in the absence of exposure to DNA damaging agents.
Analyses of the EXO1 synthetic lethality CRISPR screens. (A) Diagram showing the overlap of identical genes within the top synthetic lethality hits in the two EXO1-knockout cell lines. The number of genes with MAGeCK score lower than 0.005 (left) and 0.015 (right) are shown. (B) The number of common genes within the top synthetic lethality hits in the two EXO1-knockout cell lines compared to the random probability of identical hits. The top hits with MAGeCK score lower than 0.005 (left) and 0.015 (right) were included in the analysis. (C) Functional annotation clustering of the common genes within the top synthetic lethality hits with MAGeCK score lower than 0.015 in the two EXO1-knockout cell lines, using Gene Ontology and Uniprot terms. (D) Diagram showing the overlap of identical genes within the top synthetic lethality hits in EXO1KO#1 cells (compared to WT) and the control comparison to EXO1KO#3. The number of genes with MAGeCK score lower than 0.005 (left) and 0.015 (right) are shown. (E) The number of common genes within the top synthetic lethality hits in EXO1KO#1 cells (compared to WT) and the control comparison to EXO1KO#3. The top hits with MAGeCK score lower than 0.005 (left) and 0.015 (right) were included in the analysis.
As a control analysis, we ranked the genes lost in one of the EXO1-knockout clones (EXO1^KO#1^) compared to the other clone (EXO1^KO#3^). We then identified the common genes within the top hits between this dataset and the ranking of genes which were lost in EXO1^KO#1^ compared to WT. The actual number of common genes (0 the top hits with MAGeCK score lower than 0.005 and 14 for the top hits with MAGeCK score lower than 0.015) was lower than the random probabilities (5.6 and 35.2, respectively) (Fig. 3D and E). Overall, these analyses suggest that our screens were able to identify specific synthetic lethality interactions of EXO1.
One of the top hits identified in this synthetic lethality analysis was CHAF1A, the large subunit of the histone chaperone CAF-1 [49]. CAF-1 has multiple roles in replication-associated histone deposition as well as in DNA repair. CHAF1A was a hit in both EXO1-knockout screens (ranked 234 in the EXO1^KO#1^ and 190 in EXO1^KO#3^) but not in the control analysis (ranked 7701) (Fig. 4A–G). To validate the screen results, we depleted CHAF1A using both siRNA and sgRNA in WT and EXO1-knockout cells. Clonogenic assays (without any DNA damaging treatment) indicated that CHAF1A depletion in both EXO1-knockout cell lines, with both siRNA and sgRNA, reduced their viability compared to the CHAF1A depletion in WT cells (Fig. 4H and I, and Supplementary Fig. S1B). These findings indicate that concomitant loss of EXO1 and CHAF1A reduced cellular survival under normal growth conditions.
Co-depletion of EXO1 and CHAF1A reduces cellular viability. (A, C, E). Volcano plots showing the results of genome-wide CRISPR knockout screens to identify EXO1 synthetic lethality interactions. Genes targeted by the library are presented based on their impact on the viability of EXO1KO#1 compared to WT cells (A), EXO1KO#3 compared to WT cells (C), and as control, EXO1KO#1 compared to EXO1KO#3 cells (E). Genes are plotted by the −log10 of their respective negative and positive P-values and associated log2 Fold Change values. The hit chosen for validation, namely CHAF1A, is indicated. (B, D, F) Scatterplots showing the results of genome-wide CRISPR knockout screens to identify EXO1 synthetic lethality interactions. Genes targeted by the library are plotted based on their impact on the viability of EXO1KO#1 compared to WT cells (B), EXO1KO#3 compared to WT cells (D), and as control, EXO1KO#1 compared to EXO1KO#3 cells (F). The hit chosen for validation, namely CHAF1A, is indicated. (G) Table showing the ranks in the synthetic lethality screens, and the biological roles of CHAF1A. (H, I) Clonogenic survival assays showing that siRNA (H) and sgRNA (I) depletion of CHAF1A reduces the viability of EXO1-knockout cells compared to WT HeLa cells. Clonogenic survival is presented normalized to WT control cells. The average of three independent experiments, with standard deviations indicated as error bars, is shown. Asterisks indicate statistical significance (t-test unpaired).
CAF-1 promotes the viability of EXO1-deficient cells by suppressing R-loop-associated DNA damage
Since both CAF-1 and EXO1 have roles in DNA replication, we tested if the genetic interaction observed may be caused by an increase in replication-associated DNA lesions. We treated cells with BrdU in order to label replicating cells, then performed alkaline comet assays. CHAF1A depletion in EXO1-knockout cells, but not in WT cells, resulted in increased BrdU olive tail moments (Fig. 5A). These results suggest that increased replication-associated DNA lesions may account for the synthetic interaction observed.
EXO1 and CHAF1A are recruited to R-loops and synergistically suppress R-loop accumulation. (A) BrdU alkaline comet assay showing that CHAF1A depletion causes increased accumulation of replication-associated single stranded DNA lesions in both EXO1-knockout lines compared to WT HeLa cells. At least 100 nuclei were quantified for each condition. The median values are marked on the graph and listed at the top. Asterisks indicate statistical significance (Mann–Whitney, two-tailed). Schematic representations of the assay conditions are shown at the top. (B, C) S9.6 PLA experiments showing increased EXO1 (B) and CHAF1A (C) recruitment to R-loops in HeLa cells. At least 100 cells were quantified for each condition. Bars indicate the mean values, error bars represent standard errors of the mean, and asterisks indicate statistical significance (t-test, two-tailed, unpaired). (D, E) S9.6-dsDNA PLA experiments showing increased R-loops in HeLa cells with concomitant inactivation of EXO1 and CHAF1A. RNAseH1 overexpression suppresses the PLA signal, indicating that it derives from R-loops. At least 100 cells were quantified for each condition. Bars indicate the mean values, error bars represent standard errors of the mean, and asterisks indicate statistical significance (t-test, two-tailed, unpaired).
We next sought to identify the nature of these lesions. DNA–RNA hybrids (R-loops) can interfere with DNA replication [50], and previous studies have shown that an increase in R-loops can be detected using the BrdU alkaline comet assay [51]. EXO1 was shown to play a role in suppressing R-loop accumulation, through the recruitment of the FANCM-BLM complex, which disassembles R-loops [38]. We thus investigated the recruitment of EXO1 to R-loops in normal and CHAF1A-depleted cells. We employed proximity ligation assays (PLA) using the S9.6 antibody recognizing R-loops. We found that EXO1 is recruited to R-loops, and this recruitment is greatly enhanced in CHAF1A-knockdown cells (Fig. 5B). We employed the same assay to investigate if CAF-1 is recruited to R-loops. Similar to the EXO1 findings, in S9.6-CHAF1A PLA assays we found that CAF-1 is recruited to R-loops, and this recruitment is greatly enhanced in EXO1-knockout cells (Fig. 5C). These findings indicate that CAF-1 and EXO1 are independently recruited to R-loops. The loss of one of them results in highly increased recruitment of the other, suggesting that they participate in separate, potentially synergistic pathways of R-loop processing.
We thus sought to evaluate if the synthetic interaction between EXO1 and CAF-1 may reflect R-loop suppression. To measure R-loops, we employed PLA using the S9.6 and an antibody recognizing dsDNA. In line with previous findings [38], EXO1-knockout cells show increased R-loops, as indicated by an increase in the S9.6-dsDNA PLA signal (Fig. 5D). Depletion of CHAF1A moderately increased R-loops in WT cells. In contrast, CHAF1A depletion in EXO1-knockout cells showed a large increase compared to mock-depleted EXO1-knockout cells. Confirming that the S9.6-dsDNA PLA signal observed derives from R-loops, overexpression of RNAseH1, which dissembles R-loops [52], abolished this PLA signal (Fig. 5E and Supplementary Fig. S1C). These results indicate that the concomitant loss of EXO1 and CAF-1 results in synergistic R-loop accumulation.
Since R-loops can be a source of DNA damage [50], we measured R-loops-associated DNA damage using PLA assays with the S9.6 antibody and an antibody recognizing γH2AX, a DNA damage marker. We observed that R-loop-associated DNA damage was increased in EXO1-knockout cells, and CHAF1A depletion in these cells further enhanced it (Fig. 6A). RNaseH1 overexpression reduced the S9.6-γH2AX PLA signal, confirming that it derives from R-loops. Overall, these findings indicate that, in the absence of EXO1, CAF-1 becomes critical for the suppression of R-loop-induced DNA damage.
Concomitant depletion of EXO1 and CHAF1A causes the accumulation of R-loop-associated DNA damage. (A) S9.6-γH2AX PLA experiments showing increased R-loop-associated DNA damage in HeLa cells with concomitant inactivation of EXO1 and CHAF1A. RNAseH1 overexpression suppresses the PLA signal, indicating that it derives from R-loops. At least 100 cells were quantified for each condition. Bars indicate the mean values, error bars represent standard errors of the mean, and asterisks indicate statistical significance (t-test, two-tailed, unpaired). The siCHAF1A/siControl mean ratios are also presented. (B–D) γH2AX immunofluorescence showing that CHAF1A depletion causes increased DNA damage in both EXO1-knockout lines compared to WT HeLa cells, which is suppressed by RNaseH1 overexpression. Quantifications (B, D) and representative micrographs with scale bars representing 10 µm (C) are shown. At least 100 cells were quantified for each condition. Bars indicate the mean values, error bars represent standard errors of the mean, and asterisks indicate statistical significance (t-test, two-tailed, unpaired). (E–G) Neutral comet assays showing that CHAF1A depletion causes increased DSB formation in both EXO1-knockout lines compared to WT HeLa cells, which is suppressed by RNaseH1 overexpression. Quantifications (E, G) and representative micrographs with scale bars representing 10 µm (F) are shown. At least 100 comets were quantified for each sample. The median values are marked on the graph, and asterisks indicate statistical significance (Mann–Whitney, two-tailed).
Since R-loops can block replication forks and cause fork collapse [50], we measured the formation of DNA DSBs, using both neutral comet assays and γH2AX foci formation. In the absence of any treatment with DNA damaging agents, CHAF1A depletion in EXO1-knockout cells caused a much higher increase in DNA damage compared to its depletion in WT cells (Fig. 6B–G). Importantly, using both neutral comet assays and γH2AX immunofluorescence assays, we found that RNAseH1 overexpression suppresses DSB accumulation in the doubly depleted cells (Fig. 6D and G), indicating that these DSBs are derived from R-loops. Overall, these findings suggest that R-loops accumulate in cells with concomitant loss of CHAF1A and EXO1, causing replication fork collapse, resulting in DSB formation. This potentially explains the synthetic interaction observed.
Finally, we investigated if the exonuclease catalytic activity of EXO1 is important for the R-loop suppression activity identified here. To this end, we complemented EXO1-knockout cells with either WT EXO1 or the catalytic inactive D173A EXO1 variant [53, 54] (Supplementary Fig. S1D). S9.6-γH2AX PLA assays showed that both WT and the D173A mutant were able to suppress the R-loop-associated DNA damage observed in EXO1-knockout cells (Fig. 6A). This is in line with the previously described catalytic-independent role of EXO1 in R-loop suppression [38]. However, knockdown of CHAF1A in cells complemented with the EXO1 D173A mutant resulted in an increase in the S9.6-γH2AX PLA signal, as evidenced by the siCHAF1A/siQ ratios, similar to its depletion in EXO1-knockout cells. In contrast, CHAF1A depletion in EXO1-WT cells did not. Similar results were obtained when measuring DSB accumulation using γH2AX immunofluorescence and neutral comet assays (Fig. 6D and G). Overall, these findings indicate that, unlike its previously described catalytic-independent role in R-loop suppression, EXO1 employs its catalytic activity for suppressing R-loops in CHAF1A-depleted cells.
Discussion
EXO1 is a multifaceted protein, with multiple activities in DNA homeostasis. Through its exonuclease activity, EXO1 is a component of multiple DNA repair processes, including HR and MMR. These functions promote genomic stability [20, 21]. On the other hand, unrestricted EXO1 activity, particularly in the BRCA-mutant background, leads to excessive degradation of stalled forks, potentially promoting genomic instability [16, 22–37]. Finally, noncatalytic roles of EXO1 have also been identified as important for DNA homeostasis [38]. These complex roles of EXO1 highlight the need for a better understanding of the genetic networks it is involved in, particularly in the context of cancer therapy.
Our CRISPR screens provide ranking lists of genes regulating the exposure to low-dose, long-term cisplatin treatment (mimicking the clinical deployment of cisplatin) in WT and EXO1-deficient cells. This allowed us to identify differential regulators of the cisplatin response in these two genetic backgrounds. A caveat of our experimental setup is the library coverage (250×), which is lower than the standard 500×. Another caveat is the use of WT Cas9, which induces DSBs and thus may result in confounding effects in DNA damage screens (although our use of HeLa cells, which are p53 pathway deficient, partially mitigates this concern). Nevertheless, by analyzing the genetic networks that these regulators belong to, we found that DNA repair as a biological process is critical for cellular resistance to cisplatin in WT cells, but was not a main component of cisplatin resistance in any of the EXO1-knockout cell lines. Surprisingly, there was limited overlap between the two knockout cell lines, which could potentially be explained by off-target effects of the CRISPR process used to generate the EXO1 knockout in the first place. Overall, our findings indicate that EXO1 is central to the DNA repair response to cisplatin-induced DNA damage. This epistatic interaction likely reflects the participation of EXO1 in multiple DNA repair mechanisms which handle DNA adducts.
Finally, our synthetic interaction screens revealed a comprehensive list of EXO1 genetic interactors. Of those, we validated that loss of CHAF1A, the major subunit of the histone chaperone CAF-1, reduces the viability of EXO1-deficient cells compared to WT cells. CAF-1 is involved in replication-associated chromatin assembly [55]. However, specific roles of CAF-1 in DNA repair have been described, including roles in DSB repair [56] and fork protection [33, 37]. Since both proteins play roles in the replication stress response, we investigated how the concomitant loss of EXO1 and CAF-1 impacts DNA replication. We found that both EXO1 and CAF-1 are independently recruited to R-loops, and loss of one of them greatly increases the R-loop recruitment of the other. Moreover, concomitant loss of EXO1 and CAF-1 results in R-loop accumulation, R-loop-associated DNA damage, and DSB formation. We moreover show that the role of EXO1 in R-loop suppression in CHAF1A-depleted cells requires its catalytic activity. We propose that R-loop accumulation, and subsequent replication fork collapse at these sites, causes the reduced proliferation observed in cells doubly depleted of EXO1 and CHAF1A, thus explaining the synthetic interaction.
How CAF-1 suppresses R-loops remains unclear. CAF-1 assembles nucleosomes during DNA replication and repair [55]. Interestingly, perhaps providing a connection to chromatin regulation, the R-loop suppression pathway in which EXO1 participates, involves acetylation of histone H4 a lysine 8 (K8) by PCAF [38]. EXO1 is recruited to this modification, and in turn, through a still unclear mechanism, facilitates the recruitment of FANCM and BLM, which disassembles the R-loops. Our data suggest that, in EXO1-deficient cells, other R-loop suppression pathways are able to partly mitigate R-loop accumulation. Loss of CAF-1 inhibits these other pathways, thus resulting in increased R-loop accumulation. It is possible that deficient chromatin assembly in CAF-1-deficient cells, coupled with increased H4K8 acetylation, creates an inhibitory milieu for R-loop-disassembling helicases such as senataxin [57]. Indeed, senataxin was shown to interact with chromatin remodelers including CHD4 [58], suggesting a less recognized involvement of chromatin organization in regulating R-loop resolution. In any case, our results are potentially relevant to precision oncology treatment of tumors, considering that both CAF-1 and EXO1 are found inactivated in clinical samples.
Supplementary Material
gkag226_Supplemental_Files
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Brown A, Kumar S, Tchounwou PB. Cisplatin-based chemotherapy of human cancers. J Cancer Sci Ther. 2019;11:97.32148661 PMC 7059781 · pubmed ↗
- 2Woods D, Turchi JJ. Chemotherapy induced DNA damage response: convergence of drugs and pathways. Cancer Biol Ther. 2013;14:379–89. 10.4161/cbt.23761.23380594 PMC 3672181 · doi ↗ · pubmed ↗
- 3Basu A, Krishnamurthy S. Cellular responses to cisplatin-induced DNA damage. J Nucleic Acids. 2010;2010:201367. 10.4061/2010/201367.20811617 PMC 2929606 · doi ↗ · pubmed ↗
- 4Swift LH, Golsteyn RM. Genotoxic anti-cancer agents and their relationship to DNA damage, mitosis, and checkpoint adaptation in proliferating cancer cells. Int J Mol Sci. 2014;15:3403–31. 10.3390/ijms 15033403.24573252 PMC 3975345 · doi ↗ · pubmed ↗
- 5Stornetta A, Zimmermann M, Cimino GD et al. DNA adducts from anticancer drugs as candidate predictive markers for precision medicine. Chem Res Toxicol. 2017;30:388–409. 10.1021/acs.chemrestox.6b 00380.27936622 PMC 5379252 · doi ↗ · pubmed ↗
- 6Zeman MK, Cimprich KA. Causes and consequences of replication stress. Nat Cell Biol. 2014;16:2–9. 10.1038/ncb 2897.24366029 PMC 4354890 · doi ↗ · pubmed ↗
- 7Berti M, Vindigni A. Replication stress: getting back on track. Nat Struct Mol Biol. 2016;23:103–9. 10.1038/nsmb.3163.26840898 PMC 5125612 · doi ↗ · pubmed ↗
- 8Barbari SR, Shcherbakova PV. Replicative DNA polymerase defects in human cancers: consequences, mechanisms, and implications for therapy. DNA Repair (Amst). 2017;56:16–25. 10.1016/j.dnarep.2017.06.003.28687338 PMC 5750057 · doi ↗ · pubmed ↗