RAD51C-XRCC3 complex regulates FANCM-mediated R-loop resolution to safeguard genome integrity

Abstract

Fanconi anemia (FA) is characterized by bone marrow failure, congenital abnormalities, and cancer predisposition. Mutations in RAD51 paralogs have been identified in FA-like disorders and cancers. Although the role of RAD51 paralogs is well established in homologous recombination (HR)–mediated DNA repair, little is known about their role during replication stress responses. Here, we report that the RAD51C-XRCC3 (CX3) complex of RAD51 paralogs participates in the FA pathway of R-loop tolerance mechanism. CX3 complex suppresses R-loops, transcription-replication collisions (TRCs), and associated genome instability under physiological and replication stress conditions. Mechanistically, the CX3 complex physically interacts with FANCM and facilitates its recruitment to the R-loop sites to promote its resolution. Notably, cells expressing the RAD51C R258H pathological mutant exhibit defective interaction with FANCM and display inefficient R-loop processing. The CX3 complex–mediated R-loop resolution is independent of its fork maintenance function. Collectively, we demonstrate a previously unidentified role of the CX3 complex in preventing R-loop–induced genome instability by regulating FANCM-mediated R-loop resolution.

RAD51 paralog CX3 complex distinctly participates in R-loop resolution and fork maintenance to safeguard the replicating genomes.

INTRODUCTION

The genome of a cell is constantly threatened by exogenous and endogenous sources of DNA damage (1). Unrepaired or aberrant repair of DNA lesions can lead to oncogenic mutations and inactivation of tumor suppressor genes, eventually leading to tumorigenesis (2–5). One of the major sources of genome instability arises from transcription-replication collisions (TRCs). Cells have mechanisms that tightly regulate transcription, replication initiation and progression, and replication fork stabilization to alleviate the risks of TRCs (3, 6–8). Defects in these mechanisms can lead to chromosome instability and cancer susceptibility genetic diseases (9, 10). Recent studies link TRCs to R-loop, which forms when a nascent RNA anneals with the template DNA strand, creating a three-stranded structure consisting of an RNA-DNA hybrid (11–13). RNA-DNA hybrid/R-loops have been implicated in diverse physiological processes, including transcriptional activation and repression, DNA replication, immunoglobulin class switch recombination, and homologous recombination (HR)–mediated repair of DNA double-strand breaks (DSBs) (13–17). However, unscheduled R-loop accumulation can impede the replication fork progression, leading to fork stalling and promoting TRCs and DNA breaks (12, 13).

Cells use various strategies to counteract the accumulation of R-loops and the associated genome instability. Splicing and RNA processing factors sequester nascent RNA and prevent R-loop formation (18); topoisomerases I and II restrict the annealing of RNA transcripts with DNA (19); and ribonuclease (RNase) H (11) or RNA-DNA helicase SETX (20), UAP56 (21), WRN (22), BLM (23), RECQL5 (23, 24), Aquarius (25), and several other DEAD/H-box helicases also resolve R-loops (25, 26). In addition, components of DNA damage response, chromatin remodelers, and replication fork stabilization factors have all been linked to R-loop resolution (11, 27–30). Among these, one of the emerging anti-R-loop mechanisms involves the participation of Fanconi anemia (FA) pathway proteins (31–35), which are critical for interstrand cross-link (ICL) repair and replication fork protection during replication stress (36). FA factors such as BRCA2/FANCD1, FANCD2, FANCA, FANCM, and BRCA1/FANCS have been shown to resolve R-loops and TRCs (37–39). However, the functional interplay between these FA proteins in R-loop/TRC resolution remains mechanistically undefined.

Mammalian RAD51 paralogs are a conserved family of five proteins (RAD51B, RAD51C, RAD51D, XRCC2, and XRCC3) essential for genome maintenance through HR-mediated DSB repair, ICL repair, replication fork protection and restart, telomere maintenance, and mitochondrial genome stability (40–43). Germline mutations in RAD51 paralogs cause breast and ovarian cancers, as well as FA-like disorder (40, 44, 45). RAD51 paralogs exist in two distinct complexes, RAD51B-RAD51C-RAD51D-XRCC2 (BCDX2) and RAD51C-XRCC3 (CX3) (41, 43, 46–51). While both these complexes are involved in the protection of stalled replication forks, only the CX3 complex participates in replication recovery (52–55). Moreover, a RAD51 paralog subcomplex (DX2) regulates replication fork progression during deoxynucleotide triphosphate (dNTP) pool alteration (56, 57). Somyajit et al. identified that RAD51 paralogs participate in the repair of replication-born lesions, especially during replication stress induced by the hard-to-replicate common fragile sites, which are also known to be R-loop accumulating sites in the genome (52, 58, 59). Nonetheless, whether RAD51 paralogs participate in resolving R-loops to maintain genome stability remains unknown.

In the present study, we uncovered a role of RAD51 paralog complex CX3 in suppressing R-loop accumulation. We find that depletion of the CX3 but not the BCDX2 complex proteins increases R-loops and R-loop–mediated DNA breaks under both physiological and replication stress conditions. Furthermore, we demonstrate that RAD51C participates in the FA pathway of R-loop resolution by recruitment of FANCM translocase to R-loop sites for its resolution. Notably, RAD51C R258H, identified in FA-like disorder, exhibits defective interaction with FANCM, affecting its localization to R-loop sites, resulting in impaired R-loop resolution. Moreover, the R-loop resolution function of the CX3 complex is separable from its role during fork protection and restart. Our study unveils a previously unidentified role of RAD51C in regulating FANCM-mediated R-loops and TRC resolution to maintain genome stability.

RESULTS

RAD51C and XRCC3, but not RAD51B, RAD51D, and XRCC2 knockdown leads to R-loop accumulation

Besides its role in ICL repair (36), the FA pathway also protects the genome against unscheduled R-loop accumulation (12, 13). As RAD51 paralogs are part of the FA pathway genes, we wondered whether RAD51 paralogs affect R-loop levels in the cells. We transiently depleted all five RAD51 paralogs and RAD51 in U2OS cells (fig. S1A). Analysis of R-loops by immunofluorescence using the S9.6 anti-DNA-RNA hybrid-specific monoclonal antibody revealed a significant increase in the S9.6 nuclear signal in RAD51C and XRCC3-depleted cells compared to control cells. Neither RAD51B, RAD51D, nor XRCC2 depletion showed elevated R-loop levels, suggesting a specific involvement of the CX3 complex in limiting R-loops (Fig. 1, A and B). Consistent with the earlier study (37), RAD51 depletion did not alter R-loop levels, suggesting that the CX3 complex prevents R-loops independently of the RAD51-mediated HR function. We overexpressed Flag-tagged RNase H1 (RNH1-Flag), an endoribonuclease that explicitly removes RNA hybridized to DNA (fig. S1, B and C). The increased S9.6 signals in RAD51C/XRCC3 (C/X3)–deficient cells were suppressed upon RNH1 overexpression, indicating that the signals are indeed specific to DNA-RNA hybrids (Fig. 1C and fig. S1D). S9.6 antibody gives bright staining at the nucleolus because of the presence of ribosomal DNA (rDNA) (37). C/X3-depleted cells showed increased R-loop accumulation even after negating the nucleolus-specific S9.6 signal, ruling out the possibility of a nucleolus-specific R-loop enrichment (fig. S1, E and F). A recent study reported that the S9.6 antibody can also bind to double-stranded RNA (dsRNA), causing artefacts during the detection of R-loops in certain immunofluorescence methods (60). To affirm that the detected signal in our staining protocol is specific to DNA:RNA hybrids, we treated control and the C/X3-depleted U2OS cells with recombinant RNase III, which cleaves dsRNAs, or RNase H, which cleaves DNA:RNA hybrids, and measured S9.6 nuclear intensities. Our results showed a significant increase in S9.6 nuclear intensities in C/X3-depleted cells even after treatment with RNase III. However, the cells treated with RNase H showed a much-reduced S9.6 nuclear signal (Fig. 1D). Furthermore, R-loop levels were evaluated by dot blot of genomic DNA with S9.6 antibody, and C/X3-depleted cells exhibited a significant increase in R-loop signals, which was entirely removed by in vitro treatment with RNase H (Fig. 1, E and F). Moreover, we also confirmed the CX3-specific R-loop phenotype by depleting RAD51C, XRCC3, or XRCC2 with two independent short hairpin RNAs (shRNAs) and ruled out the possibility of any off-target effect of shRNAs (fig. S1, G and H). Previous reports have shown that the catalytically dead RNH1 D210N mutant binds to DNA:RNA hybrids without removing them and can act as a sensitive tool to visualize R-loops in the cells (37, 61). To validate our observations, we overexpressed RNH1(D210N)-Flag in cells and scored Flag foci. Consistently, RAD51C and XRCC3, but not XRCC2, depleted cells showed higher RNH1(D210N)-Flag foci formation than control cells (Fig. 1G and fig. S1I). Next, we analyzed DNA:RNA hybrid accumulation at different previously characterized genomic loci prone to R-loop formation (37). Depletion of C/X3 increased the occupancy of RNH1(D210N)-Flag at the actively transcribed regions APOE and RPL13A, but not at a negative control site SNRPN, which is not susceptible to R-loop formation. Depletion of XRCC2 showed a similar profile to that of the control cells, confirming the involvement of the CX3 complex in R-loop removal (Fig. 1H).

Fig. 1. Depletion of RAD51C and XRCC3, but not RAD51B, RAD51D, and XRCC2, leads to R-loop accumulation in U2OS cells.

(A and B) Representative images (A) and quantification (B) of S9.6 staining in control, RAD51, and RAD51 paralog-depleted cells. Relative nuclear signal intensity data are plotted as box plot (n = 3, two-tailed unpaired t test with Welch’s correction on means of biological replicates, ≥253 cells were analyzed per condition). Scale bar, 5 μm. (C) Relative S9.6 nuclear signal intensity in control, C/X3-depleted cells with or without RNH1-Flag overexpression. Data are represented as a box plot (n = 3, two-tailed unpaired t test with Welch’s correction on means of biological replicates, ≥266 cells were analyzed per condition). (D) Relative S9.6 nuclear signal intensity in control, C/X3-depleted cells with or without RNase III/RNase H treatment. Data are represented as a box plot (n = 3, two-tailed, unpaired t test with Welch’s correction on means of biological replicates, ≥308 cells were analyzed per condition). (E) S9.6 dot blot analysis of control, C/X3-depleted genomic DNA with or without RNase H treatment. ssDNA levels served as loading controls. (F) Quantification of S9.6 dot blot signal in control, C/X3-depleted cells. Data are plotted as means ± SD (n = 3, two-tailed unpaired t test with Welch’s correction). (G) Chromatin-bound RNH1(D210N)-Flag foci in control, C/X3/X2-depleted cells. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥328 cells were analyzed per condition). (H) RNH1(D210N)-Flag ChIP-qPCR in control, C/X3/X2-depleted cells. Enrichment relative to No Ab control is plotted as means ± SD (n = 3, two-tailed unpaired t test). (I) Relative S9.6 nuclear intensity in EdU-positive and negative cells after C/X3 depletion. Data are represented as a box plot (n = 4, two-tailed unpaired t test with Welch’s correction on means of biological replicates, ≥132 cells were analyzed per condition). P values are indicated. ns, nonsignificant.

A recent study showed that R-loops accumulate during DNA replication in the S phase of the cell cycle without exogenous stress (62). Because RAD51 paralogs are enriched on chromatin during the S phase (52), we tested whether the accumulation of R-loops in CX3-deficient cells is specific to the S phase. We incorporated EdU to mark the actively replicating cells and investigated R-loop levels using immunofluorescence. As shown in a previous study (62), we also found a higher R-loop accumulation in replicating (EdU-positive) cells than in nonreplicating (EdU-negative) cells. C/X3-depleted cells showed a slight increase in R-loops in nonreplicating cells. However, R-loop accumulation was substantially higher in replicating C/X3-deficient cells (Fig. 1I and fig. S2A). Unregulated origin firing or licensing in cancer cell lines can cause elevated R-loops even in the non–EdU-positive cells (63). We studied R-loop levels in the immortalized noncancerous cell line hTERT RPE-1 to validate our observation from U2OS cells. Consistently, depletion of C/X3, but not X2, showed a significant increase in R-loop in hTERT RPE-1 cells (fig. S2, B and C). Similar to U2OS cells, replicating hTERT RPE-1 cells also showed significantly higher R-loops after depletion of C/X3 (fig. S2D). Collectively, these results suggest that among the two RAD51 paralog complexes, the CX3 complex is specifically required for R-loop suppression during active replication.

CX3 complex suppresses R-loop–dependent genome instability

Unscheduled R-loop accumulation in proliferating cells is a crucial source of endogenous genome instability (11). Loss of RAD51 paralogs, C/X3, leads to spontaneous DNA breaks and chromosomal abnormalities (59, 64). To understand whether accumulated R-loops underscore the genome instability, we analyzed the DNA damage status in C/X3-deficient cells. Analysis of γH2AX by immunofluorescence and comet tail length in alkaline comet assay revealed a significant increase in genome instability levels in C/X3-depleted cells compared to control cells. Both γH2AX foci and comet tail length were partially rescued upon expression of RNH1 in C/X3-deficient cells, implicating the involvement of R-loops in DNA break formation (fig. S3A and Fig. 2, A to C). The catalytic dead RNH1 D210N mutant binds and stabilizes the R-loops and has been reported to cause fork blockage, leading to genome instability in the absence of R-loop–preventing factors (37). As expected, expressing RNH1(D210N) mutant in cells depleted of C/X3, but not X2, showed elevated levels of γH2AX foci (Fig. 2D).

Fig. 2. CX3 deficiency leads to R-loop–induced genome instability.

(A) γH2AX foci in control, C/X3-depleted cells with or without RNH1-Flag overexpression. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥240 cells were analyzed per condition). (B and C) Representative images (B) and quantification (C) of alkaline comet assay tail moment in control, C/X3-depleted cells with or without RNH1-Flag overexpression. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥156 cells were analyzed per condition). Scale bar, 5 μm. (D) γH2AX foci in control, C/X3/X2-depleted cells with or without overexpression of catalytic dead RNH1(D210N)-Flag mutant. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥307 cells were analyzed per condition). (E) Quantification of γH2AX foci in EdU-positive and negative cells after depletion of C/X3. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥108 cells were analyzed per condition). (F and G) Representative images (F) and quantification (G) of RNAPIIS2P + PCNA PLA foci in control, C/X3/X2-depleted cells. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥316 cells were analyzed per condition). Scale bar, 5 μm. (H) Quantification of γH2AX foci in control, C/X3-depleted cells after treatment with transcription inhibitor DRB (100 μM, 4 hours). Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥263 cells were analyzed per condition). P values are indicated.

Furthermore, we analyzed the DNA breaks in replicating and nonreplicating cells after depletion of C/X3. Although we observed a moderate increase in γH2AX foci in nonreplicating cells, replicating C/X3-deficient cells showed a marked increase in γH2AX levels compared to control cells, implicating the role of the CX3 complex in preventing R-loop–mediated genome instability during S phase (Fig. 2E). TRC is one of the major causes of R-loop–induced genome instability in the cells (7, 63). Higher accumulation of S phase–specific DNA breaks in C/X3-deficient cells could be due to higher TRCs in these cells. To test this, we used the proximity ligation assay (PLA) to detect a collision between the elongating RNA polymerase II (RNAPIIS2P) and PCNA. C/X3 depletion significantly increased PLA foci compared to control cells, suggesting higher TRCs with C/X3 deficiency (Fig. 2, F and G, and fig. S3B). As expected, XRCC2-depleted cells showed a similar level of TRCs as control cells. Furthermore, to suppress TRCs, we treated cells with the transcription inhibitor 5,6-dichlorobenzimidazole 1-β-d-ribofuranoside (DRB) and scored S9.6 intensity and γH2AX in C/X3-deficient cells. As expected, the increased S9.6 signal and γH2AX foci observed in C/X3-depleted cells were significantly rescued after DRB treatment (fig. S3, C to E, and Fig. 2H). Together, our results indicate that CX3 deficiency causes DNA damage via R-loop–induced TRCs.

CX3 complex is critical for the resolution of replication stress–induced R-loops

Our earlier study with Chinese hamster cells devoid of RAD51C, XRCC2, and XRCC3 showed that RAD51C or XRCC3, but not XRCC2, cells were hypersensitive toward mild replication stress (52). The CX3-specific phenotype was explained by its role in HR-mediated repair mechanisms during replication fork recovery (52). However, recent studies also demonstrate that low doses of replication stress induced by hydroxyurea (HU) or aphidicolin (Aph) induce replication stalling, promoting TRCs and R-loop formation, especially at long genes in common fragile sites (63, 65). This prompted us to measure R-loops in C/X3-deficient cells in response to replication stress. We depleted C/X3/X2 and evaluated R-loop levels after treatment with a low dose of HU (1 mM for 4 hours). As reported earlier (63), we also observed a moderate but significant increase in R-loop accumulation after HU treatment in control shRNA-transfected cells. C/X3-deficient cells exhibited a further increase in R-loops accumulation (Fig. 3A and fig. S4A). Similar results were obtained with a low dose of Aph treatment, where C/X3-depleted cells showed significantly higher R-loops than control cells after Aph treatment, suggesting a role of the CX3 complex in preventing R-loop accumulation during replication stress (Fig. 3B). Depletion of XRCC2 showed comparable R-loop levels as those of control shRNA-transfected cells, both with HU and Aph treatments (Fig. 3, A and B, and fig. S4A). We studied DNA damage status in C/X3-deficient cells after HU-induced replication stress. As expected, C/X3-depleted cells showed significantly higher γH2AX staining compared to control cells after treatment with HU. Furthermore, overexpression of RNH1 could significantly rescue the observed γH2AX in C/X3-depleted cells (Fig. 3C). Moreover, C/X3-deficient cells showed elevated micronuclei formation after Aph-induced replication stress, which was also reduced with RNH1 overexpression (fig. S4B). Rescue of replication stress–induced genome instability by RNH1 overexpression in C/X3 deficiency prompted us to investigate the R-loop–induced hypersensitivity of these cells to replication stress. To examine this, we treated C/X3-depleted cells with HU (1 mM, 4 hours) and performed a cell survival assay with or without RNH1 overexpression. Suppressing R-loops by RNH1 overexpression partially rescued the sensitivity of the C/X3-depleted cells, suggesting that unresolved R-loops cause genome instability, leading to cell death in the absence of the CX3 complex (fig. S4C).

Fig. 3. CX3 complex protects cells from replication stress–induced R-loops.

(A) Quantification of S9.6 staining in control, C/X3/X2-depleted U2OS cells with or without HU (1 mM, 4 hours) treatment. Data are represented as a box plot (n = 3, two-tailed, unpaired t test with Welch’s correction on means of biological replicates, ≥242 cells were analyzed per condition). (B) Quantification of S9.6 staining in control, C/X3/X2-depleted U2OS cells with or without Aph (0.4 μM, 16 hours) treatment. Data are represented as a box plot (n = 3, two-tailed, unpaired t test with Welch’s correction on means of biological replicates, ≥211 cells were analyzed per condition). (C) Quantification of γH2AX intensity in the presence or absence of HU (1 mM, 4 hours) treatment in control, C/X3-depleted cells with or without overexpression of RNH1-Flag. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test, ≥241 cells were analyzed per condition). (D) Effect of RAD51C or XRCC3 depletion on fork progression/fork symmetry following 50 nM CPT treatment as indicated in the schematic. Top: Experimental schematic; middle: representative images; and bottom: quantification of sister fork ratio. Sister fork ratios were measured by dividing the longer sister fork by the shorter fork (IdU tracts). Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥40 sister fork pairs were analyzed per condition). P values are indicated.

Unresolved R-loops induce replication stress, and if the stalled fork is not repaired or restarted, it can lead to fork collapse and DSBs (66, 67). To understand the mechanism of R-loop–induced genome instability in C/X3-deficient cells, we investigated replication stalling events occurring during replication stress in C/X3-deficient cells. A short and mild pulse of camptothecin (CPT) treatment has recently been shown to cause R-loop–induced fork stalling (68). We treated cells with CPT (100 nM, 30 min) to promote R-loop formation and fork stalling. As expected, we observed significantly higher R-loop accumulation in U2OS cells treated with CPT (fig. S4D). Moreover, CPT treatment–induced FANCD2 foci formation, a replication stalling marker, was reduced after RNH1 overexpression (fig. S4, E and F), suggesting the R-loop–mediated fork stalling. Frequent fork stalling results in sister fork asymmetry during replication. To assess the frequency of R-loop–induced fork stalling, we analyzed fork asymmetry under CPT treatment in control and C/X3-deficient cells. We sequentially labeled cells with a pulse of CldU and CPT + IdU and measured the ratio of longer to shorter IdU tracts to assess asymmetry between sister forks. C/X3 depletion showed significantly higher RNH1-sensitive fork asymmetry than control cells (Fig. 3D), suggesting that C/X3 deficiency leads to higher R-loops and TRCs, ultimately affecting the fork progression during mild replication stress. Together, these data suggest the critical role of the CX3 complex in suppressing R-loop–mediated genome instability during replication stress.

CX3 complex participates in the FA pathway of R-loop resolution

XRCC3 undergoes ATM/ATR-mediated phosphorylation at S225, which is critical for HR-mediated DSB repair (56, 58). To understand the mechanism of C/X3-mediated R-loop processing, we investigated whether this phosphorylation plays a role in R-loop suppression. We expressed shRNA-resistant wild type (WT) and S225A mutant after depletion of XRCC3 and measured R-loop levels. XRCC3 S225A phosphorylation mutant showed similar complementation as WT XRCC3, suggesting that XRCC3 phosphorylation is dispensable for R-loop suppression (fig. S5, A to C).

Since both RAD51C (FANCO) and XRCC3 lack any nuclease or helicase domain (40, 48–50), we hypothesized that the CX3 complex suppresses R-loops by recruiting other anti-R-loop factors. To determine whether CX3 participates in the FA pathway of R-loop resolution, we performed an epistatic analysis with S9.6 staining after codepleting RAD51C/FANCO with FANCM, an FA pathway translocase that has been reported to unwind R-loop structures both in vivo and in vitro (31). Depletion of RAD51C or FANCM resulted in an expected increase in S9.6 nuclear signal; however, codepletion of both RAD51C and FANCM did not show any further increase in S9.6 intensities compared to single-depleted cells, suggesting that RAD51C participates in the same pathway as FANCM in processing R-loops (Fig. 4, A and B, and fig. S5D). Similar results were obtained when we analyzed R-loops after codepletion of RAD51C with FANCD2 (Fig. 4, C and D, and fig. S5E). Furthermore, codepletion of XRCC3 with FANCD2/FANCM showed a similar level of R-loops as single-depleted cells (Fig. 4E and fig. S5, F and G). To validate further, we scored for RNH1(D210N)-Flag foci after single and double depletion of RAD51C/FANCD2/FANCM. Codepletion of FANCD2/FANCM with RAD51C showed a similar enrichment of RNH1 D210N mutant foci compared to single-depleted cells (fig. S5H), suggesting that RAD51C and FANCM/FANCD2 participate in a common pathway of R-loop resolution. Next, we assessed DNA damage after codepletion of RAD51C with FANCD2 or FANCM. Single depletion of FANCD2/FANCM/RAD51C showed higher endogenous DNA damage (Fig. 4, F and G). However, consistent with the R-loop accumulation data, codepletion of RAD51C with FANCD2 or FANCM did not further increase DNA breaks compared to the single-depletion conditions (Fig. 4, F and G). Similarly, the TRC levels measured by RNAPIIS2P + PCNA PLA signals were similar in RAD51C-FANCD2/RAD51C-FANCM codepleted cells compared to RAD51C/FANCD2/FANCM alone-depleted cells (Fig. 4, H and I). Thus, our data suggest that the CX3 complex participates with FA pathway proteins to suppress R-loop accumulation and promote TRC resolution to safeguard genome integrity.

Fig. 4. CX3 complex is required for the FA pathway-mediated R-loop resolution.

(A and B) Representative images (A) and quantification (B) of S9.6 staining after transfection of cells with indicated shRNAs. Relative nuclear S9.6 intensity data are plotted as a box plot (n = 3, two-tailed, unpaired t test with Welch’s correction on means of biological replicates, ≥244 cells were analyzed per condition). Scale bar, 5 μm. (C and D) Representative images (C) and quantification (D) of S9.6 staining after transfection of cells with indicated shRNAs. Relative nuclear S9.6 intensity data are plotted as a box plot (n = 3, two-tailed, unpaired t test with Welch’s correction on means of biological replicates, ≥344 cells were analyzed per condition). Scale bar, 5 μm. (E) Quantification of S9.6 staining after transfection of cells with indicated shRNAs. Relative nuclear S9.6 intensity data are plotted as a box plot (n = 3, two-tailed, unpaired t test with Welch’s correction on means of biological replicates, ≥250 cells were analyzed per condition). (F and G) Representative immunofluorescence images (F) and quantification (G) of γH2AX foci after transfection with indicated shRNAs. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥264 cells were analyzed per condition). Scale bar, 5 μm. (H and I) Representative immunofluorescence images (H) and quantification (I) of RNAPIIS2P + PCNA PLA foci after transfection of cells with indicated shRNAs. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥302 cells were analyzed per condition). Scale bar, 5 μm. P values are indicated.

CX3 complex facilitates the recruitment of FANCM and FANCD2 to R-loop sites

In the ICL repair context, RAD51 paralogs act downstream of the FA core proteins and facilitate HR-mediated repair (36, 59, 69). To investigate the mechanism underlying CX3-mediated R-loop suppression, we asked whether the CX3 complex is physically present at the R-loop sites. Spontaneously generated R-loops or TRCs activate the FA pathway, which can be scored by FANCD2 foci in the cell population without exogenous stress. In line with the earlier studies (31, 70, 71), we observed FANCD2 and FANCM foci in control cells, which were significantly reduced by overexpression of RNH1 (fig. S6, A and B). Similarly, RAD51C foci were observed without exogenous stress, and these were reduced upon overexpression of RNH1, suggesting the presence of RAD51C at the R-loop sites (fig. S6C). To assess the localization of RAD51C/FA proteins to R-loop sites more directly, we performed PLA with S9.6 and FA proteins and RAD51C. All three proteins, FANCD2, FANCM, and RAD51C, showed PLA signals with S9.6, which were reduced with RNH1 overexpression (fig. S6, D to F). Furthermore, chromatin immunoprecipitation (ChIP) assay also showed enrichment of RAD51C and XRCC3, but not XRCC2, at the R-loop prone loci APOE and RPL13A in the genome (Fig. 5A). The enrichment of C/X3 at these R-loop sites were reduced with RNH1 overexpression (Fig. 5A). Consistently, RAD51C and XRCC3 exhibited significantly higher PLA foci with S9.6 than XRCC2 (Fig. 5B), suggesting that only the CX3 complex localizes to the R-loop sites. Moreover, RAD51C foci also colocalized with FANCM and FANCD2 foci in the cell (fig. S6, G and H). Together, these data indicate that the CX3 complex is physically present at the R-loop sites along with FANCD2 and FANCM.

Fig. 5. CX3 promotes the recruitment of FANCM to R-loop sites.

(A) ChIP-qPCR of RAD51C/XRCC3/XRCC2 with or without RNH1-Flag overexpression. Enrichment relative to No Ab control is plotted as means ± SD (n = 3, two-tailed unpaired t test). (B) Representative images (left) and quantification (right) of PLA foci of indicated proteins with S9.6. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥329 cells were analyzed per condition). (C) Co-IP of RAD51C with or without HU (1 mM, 4 hours) treatment. Endogenous protein levels were shown in the input. (D) Co-IP of XRCC3 with or without HU (1 mM, 4 hours) treatment. Endogenous protein levels were shown in the input. (E) Co-IP of FANCM with or without HU (1 mM, 4 hours) treatment. Endogenous protein levels were shown in the input. (F) The front and back views of the FANCM-RAD51C-XRCC3 complex. All monomers are shown as surface representation (FANCM: slate; RAD51C: light pink; and XRCC3: light orange). The binding interfaces of FANCM, RAD51C, and XRCC3 are encircled in black and are marked by pink, orange, and dark blue, respectively. (G) Zoom-in view of the two FANCM domains: N-terminal translocase (dark blue) and C-terminal ERCC4 region (smudge green), which exhibit interactions with different regions of RAD51C (pink) and XRCC3 (orange). The interacting residues are represented as sticks. (H) Representative images (left) and quantification (right) of S9.6 + FANCM PLA in control, C/X3/X2-depleted cells. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥322 cells were analyzed per condition). Scale bar, 5 μm. (I) FANCM ChIP-qPCR in control, C/X3/X2-depleted cells. Enrichment relative to No Ab control is plotted as means ± SD (n = 3, two-tailed unpaired t test). P values are indicated.

To gain mechanistic insights into the localization of C/X3 and FA proteins to R-loop sites, we examined whether C/X3 physically interacts with FANCM to promote R-loop resolution. Coimmunoprecipitation (co-IP) using RAD51C antibody showed an interaction of RAD51C with FANCM in control cells, and this interaction was enhanced by HU treatment (Fig. 5C). Similarly, co-IP using the XRCC3 antibody also identified an interaction with FANCM in both control and HU-treated cells (Fig. 5D). To confirm these interactions, we performed reverse co-IP using FANCM antibody and found an interaction with RAD51C and XRCC3 in both control and HU-treated cells (Fig. 5E). We could not detect XRCC2 in FANCM co-IP, suggesting that the interaction of FANCM with RAD51 paralogs is limited to the CX3 complex (Fig. 5E). To validate our IP data, we performed IP–mass spectrometry (MS) after RAD51C-IP and searched for FA proteins and other R-loop resolvases. IP-MS data also revealed an interaction of RAD51C with FANCM (fig. S6I). However, to rule out the possibility of DNA/RNA-mediated interactions between these proteins, we performed the co-IP in the presence of benzonase endonuclease, which degrades both DNA and RNA in the lysate. CX3-FANCM complex was intact in the presence of benzonase. Together, these data suggest that the CX3 complex is directly interacting with the FANCM translocase (fig. S6, J and K). To further validate the direct binding of FANCM to CX3 complex, we modeled the FANCM-RAD51C-XRCC3 complex using AlphaFold 3 (AF3) (Fig. 5F and fig. S7, A to C). The predicted structure indicated for a direct interaction of the CX3 heterodimer with FANCM. The FANCM N-terminal translocase and C-terminal ERCC4 domains interact with the CX3 complex (Fig. 5G). Between XRCC3 and RAD51C, the former showed an extended interaction with the FANCM, engaging its interactions with both the translocase and ERCC4 domains. The N-terminal (25–30) and C-terminal (250 to 260) residues of RAD51C are majorly involved in binding with the ERCC4 domain of FANCM. The binding interface energy of CX3-FANCM was observed to be −18.6 kcal/mol. The quality of the predicted structure was analyzed using pLDDT score (fig. S7D). Certain binding regions of the translocase and ERCC4 domains were observed to have very low confidence/pLDDT score (fig. S7E). Hence, to validate the correctness of the predicted structural fold of these regions, the AF3-predicted domains have been aligned with the reported crystal structures of their respective domains (fig. S7F) (72). Comparative analysis indicated that the predicted structures are similar to those of the reported domains (72), thus establishing the direct interaction between the FANCM and the CX3 complex.

Furthermore, to decipher whether CX3 is upstream or downstream of FA proteins, we measured FANCM and FANCD2 recruitment to R-loop sites via PLA with S9.6 in cells depleted for C/X3/X2. FANCD2 and FANCM PLA foci with S9.6 were significantly reduced in C/X3-depleted cells compared to control cells. As expected, depletion of XRCC2 did not alter the FANCM/FANCD2 PLA signal with S9.6 (Fig. 5H and fig. S8A). Consistently, ChIP analyses showed a reduction in FANCM enrichment at the R-loop sites upon depletion of C/X3 (Fig. 5I). However, depletion of either FANCD2 or FANCM did not alter the PLA signal between S9.6 and RAD51C, suggesting that RAD51C is upstream of FANCD2 and FANCM (fig. S8B). Together, these data reveal that the CX3 complex facilitates FANCM localization to R-loop sites via direct interaction to promote its resolution.

RAD51C R258H pathological mutant is defective for FANCM interaction and R-loop resolution

Germline mutation in RAD51C is known to cause FA-like syndrome (45). RAD51C homozygous mutation R258H was identified in a family with a characteristic FA-like phenotype (40, 45). However, the functional defect associated with this mutation leading to FA syndrome is still unclear. Because the CX3 complex participates in the FA pathway of R-loop resolution, we investigated the efficiency of R-loop processing by this FA-linked mutation. We transiently expressed shRNA-resistant WT and R258H RAD51C in RAD51C-depleted U2OS cells and scored R-loops (Fig. 6A). RAD51C R258H showed significantly higher R-loop accumulation than WT RAD51C, suggesting a functional defect associated with this mutation (Fig. 6B). Similarly, TRC levels were also significantly elevated in R258H-expressing cells compared to WT RAD51C-expressing cells (Fig. 6C). A defect in R-loop resolution by R258H RAD51C prompted us to examine its ability to localize to R-loop sites. The R258H RAD51C mutant was able to localize to the R-loop sites similar to that of WT RAD51C (Fig. 6D). Furthermore, to understand the inability of the R258H pathological mutant to resolve R-loops, we investigated its interaction with the FANCM translocase by performing co-IP with Flag-antibody after expressing WT and R258H RAD51C. R258H RAD51C failed to interact with FANCM (Fig. 6E). To gain further insights into the defective interaction of R258H RAD51C with FANCM, the structure of the FANCM-CX3 (R258H) complex was predicted using AF3 (fig. S9, A to D). The complex exhibited the loss of interface interactions between the residues 250 and 260 of RAD51C (R258H) with the ERCC4 domain of FANCM (fig. S9E), with a notable decrease in the binding interface energy to −12.3 kcal/mol as compared to its WT counterpart. Furthermore, R258H RAD51C exhibited defective recruitment of FANCM to R-loop sites compared to WT RAD51C-expressing cells (Fig. 6, F and G). These data suggest that the RAD51C R258H pathological mutant is defective in its interaction with FANCM and its recruitment to R-loop sites for its processing.

Fig. 6. RAD51C R258H pathological mutant exhibits impaired FANCM interaction and R-loop resolution.

(A) Representative immunoblot indicating levels of WT and R258H RAD51C-Flag after depletion of endogenous RAD51C in U2OS cells. MCM3 levels are shown as loading control. (B) Quantification of S9.6 staining after transfection of cells with the indicated DNA constructs, as shown in (A). Relative nuclear S9.6 intensity data are plotted as a box plot (n = 3, two-tailed, unpaired t test with Welch’s correction on means of biological replicates, ≥252 cells were analyzed per condition). (C) Quantification of RNAPIIS2P + PCNA PLA foci after transfection of cells with the indicated DNA constructs. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥303 cells were analyzed per condition). (D) Representative immunofluorescence images (top) and quantification (bottom) of S9.6 + Flag PLA foci in WT and R258H Flag-RAD51C–transfected cells. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥731 cells were analyzed per condition). Scale bar, 5 μm. (E) Co-IP of exogenously expressed Flag-tagged WT and R258H RAD51C, followed by immunoblotting with indicated antibodies. Endogenous protein levels were shown in the input. (F and G) Representative immunofluorescence images (F) and quantification (G) of S9.6 + FANCM PLA foci in cells transfected with the indicated DNA constructs. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, two-tailed unpaired t test on means of biological replicates, ≥251 cells were analyzed per condition). Scale bar, 5 μm. P values are indicated. res, shRNA-resistant.

Schwab et al. have reported that FANCM can directly unwind DNA:RNA hybrid structures through its translocase activity in both in vivo and in vitro systems (31). To support our findings further, we generated the FANCM translocase dead (K117R) mutant and scored for R-loops. As expected, expressing the K117R mutant failed to complement the R-loop phenotype to the level of the WT FANCM protein (fig. S10, A and B). Furthermore, we performed co-IP to assess whether the mutant is able to bind to RAD51C. We observed that the K117R mutant could still interact with RAD51C (fig. S10C) and be recruited to R-loop sites similar to the WT protein (fig. S10D). These data support our findings that RAD51C is upstream and recruits FANCM translocase to resolve R-loops in the genome.

R-loop resolution function of RAD51C is independent of its role in fork protection and restart

Upon encountering a lesion, the replication fork undergoes RPA-mediated protection of single-stranded DNA (ssDNA) followed by RAD51-mediated fork reversal (66, 73, 74). Fork protection factors like BRCA2, FA proteins, and RAD51 paralogs promote RAD51 assembly at the reversed nascent DNA, which prevents fork degradation against nucleases like MRE11 and DNA2 (66, 75–78). Since previous studies showed that the CX3 complex participates in fork protection and restart (45–47, 61, 71), we tested whether the R-loop resolution function of the CX3 complex is separable from its fork protection/restart. To this end, we transiently expressed shRNA-resistant adenosine 5′-triphosphate (ATP) binding–deficient K131A and ATP hydrolysis–deficient K131R RAD51C mutants after depleting endogenous RAD51C (fig. S10E) and assessed fork protection, restart, and R-loop resolution. While the RAD51C K131A mutant was defective for both fork protection and its restart, the K131R mutant exhibited only a restart defect (Fig. 7, A to C). These results are in agreement with our previous study (52). Both ATP-binding and hydrolysis mutants of RAD51C rescued R-loop accumulation to a level comparable to that of WT RAD51C, suggesting that ATP binding and hydrolysis activity are dispensable for R-loop resolution (Fig. 7, D and E). Furthermore, RAD51C K131A and K131R mutants interact with FANCM similar to WT RAD51C (fig. S10F). Collectively, our data suggest that the CX3 complex functions as an adaptor to recruit FANCM to R-loop sites and facilitates its resolution by FANCM-mediated translocase activity.

Fig. 7. ATP binding or hydrolysis activity of RAD51C is dispensable for its role in R-loop resolution.

(A) Experimental schematic and representative images, and (B) quantification of IdU/CldU ratio for fork protection assay in cells expressing the indicated mutants of RAD51C. Data are represented as a scatter plot. Black lines indicate mean values (n = 3, Mann-Whitney test, ≥203 fibers were analyzed per condition). (C) Effect of expressing WT, K131A, and K131R RAD51C mutants on replication restart following the treatment of U2OS cells with 2 mM HU for the indicated time. Top: Experimental schematic; and bottom: quantification of stalled forks. Data show percentage of stalled forks as means ± SD (n = 3, two-tailed unpaired t test). (D and E) Representative immunofluorescence images (D) and quantification (E) of S9.6 staining after transfection of cells with indicated shRNA and RAD51C constructs. Relative nuclear S9.6 intensity data are plotted as a box plot (n = 3, two-tailed, unpaired t test with Welch’s correction on means of biological replicates, ≥278 cells were analyzed per condition). Scale bar, 5 μm. P values are indicated. EV, empty vector. h, hours.

DISCUSSION

Although R-loops are formed naturally in the genome and play essential roles in various physiological processes, their persistence can cause genome instability (7, 11). Thus, their formation and removal must be temporally and dynamically regulated to alleviate the associated defect in cell growth and development. FA pathway proteins, such as BRCA1, BRCA2, FANCD2, and FANCM, have been linked to R-loop tolerance and TRC resolution (13). However, the molecular mechanism by which these proteins coordinate and participate with other anti-R-loop factors to preserve genome integrity is largely unclear. Here, we identify a previously unexplored role of RAD51 paralog CX3 complex in R-loop resolution, which is independent of its role in fork protection and restart functions. Mechanistic studies demonstrate that the CX3 complex serves as an adaptor to recruit FANCM to R-loop sites and promotes its resolution via FANCM-mediated translocase activity.

The RAD51 paralog complexes, BCDX2 and CX3, are essential for HR-mediated DSB repair (40, 79–82). However, previous studies, including ours, report that both complexes are required for replication fork protection, but only the CX3 complex promotes replication restart from stalled/collapsed forks (41, 52–54, 83), indicating a separation of function between these two complexes. Depletion of the CX3, but not the BCDX2 complex proteins, showed increased R-loop accumulation, suggesting a separation of function between these complexes. Aberrant R-loop accumulation impairs replication fork progression, promotes TRCs, and causes replication stress (7, 11, 12). On the other hand, replication stress and deregulated origin firing also promote the formation of R-loops and TRCs (63). In line with the R-loop accumulation, C/X3-deficient cells exhibited higher R-loop–induced genomic instability and TRCs. Previously, we reported that RAD51 paralogs localize to the S-phase chromatin without exogenous stress (52). We found a marked increase in γH2AX and R-loop levels in the S-phase cells after C/X3 depletion, which is consistent with the reports showing that unresolved R-loops in the S phase lead to TRCs and genome instability (62). The R-loop and γH2AX levels were exacerbated in C/X3-deficient cells in response to mild replication stress. Notably, our data show a notable rescue of genome instability and increased sensitivity of C/X3-deficient cells to spontaneous and replication stress conditions after RNH1 overexpression. These data suggest the crucial role of the CX3 complex in preventing R-loops and associated genome instability.

Unscheduled R-loop accumulation can act as an obstacle for the replication fork and cause fork stalling. These R-loops need to be actively cleared by R-loop–suppressing proteins for continuous fork progression. C/X3-deficient cells exhibited RNH1-sensitive fork asymmetry, suggesting a role for CX3 in the removal of R-loops to facilitate unrestrained fork progression. R-loop stalled replication forks undergo restart via a fork cleavage and religation cycle mediated by SLX4, MUS81-EME1, RAD52, and POLD3 (68). Moreover, a recent report showed that RAD51 is required to protect cells from TRCs and prevent RAD52-mediated BIR during early S phase (84). The CX3 complex also participates in the restart of stalled/collapsed replication forks (52–54, 58). However, further studies are required to understand whether the CX3 complex promotes the restart of replication from TRC/R-loop–induced fork stalling.

In addition to their role in ICL repair (36), recent studies have shown the participation of FA pathway proteins in R-loop and TRC resolution (29, 31, 32). However, the molecular mechanism by which FANCD2/FANCM participates in mitigating R-loop–induced toxicity remains obscure. The FANCI-FANCD2 (ID2) complex directly binds to R-loops, activating the FA pathway of R-loop resolution by promoting ID2 monoubiquitination (85). The translocase activity of FANCM is required for R-loop resolution (31). We find that CX3 depletion is epistatic to FANCD2 and FANCM, suggesting a previously unexplored role of the CX3 complex in the FA pathway of R-loop and TRC resolution. The CX3 complex localizes to the R-loop sites along with FANCD2 and FANCM. Notably, we demonstrate a previously unidentified interaction between the CX3 complex and FANCM (Fig. 8), which is further enhanced by replication stress. The interaction of RAD51 paralogs with FANCM is limited to the CX3 complex, suggesting a separation of function between RAD51 paralog complexes in FANCM-mediated R-loop resolution. In contrast to its downstream role in the FA pathway of ICL repair (40, 59), our data show that the CX3 complex has an upstream role in FANCD2/FANCM-mediated R-loop resolution by promoting their recruitment to R-loop sites (Fig. 8). The recruitment of CX3 complex to R-loop sites may be mediated by chromatin remodeling factors, which are known to participate in the R-loop resolution (27, 86). Whether the CX3 complex directly binds to R-loop sites or is mediated by chromatin remodeling or other factors requires further studies. FAAP24 exists as a stable complex with FANCM and facilitates localization of the FA core complex to ICL lesions to promote its repair (72, 87–89). Purified FANCM-FAAP24 complex has been reported to unwind R-loops (31). Our comparative analysis of CX3 and FAAP24 binding sites on FANCM using the AF3 predictions of FANCM-CX3 complex and FAAP24-FANCM (ERCC4) crystal structure revealed that a small overlap exists in the ERCC4 domain [amino acids (aa) 1928 to 1998] of FANCM for both the interacting partners, although both partners have their independent binding domains/regions on the FANCM surface apart from this overlapping region (fig. S11). This suggests that both CX3 and FAAP24 can be accommodated on the FANCM. However, whether FAAP24 directly participates along with the FANCM and CX3 complex in the FA pathway of R-loop resolution requires further studies. Recently, RAD18, a known interactor of RAD51C, has been shown to facilitate FANCD2 recruitment to R-loop–prone sites, promoting TRC tolerance (70, 90). In addition, the MRN complex also participates upstream of FA proteins in resolving R-loops independently of its nuclease activity (71). Nonetheless, the mechanisms by which the CX3 complex coordinates with multiple factors to preserve genome integrity against R-loops and TRCs require further studies.

Fig. 8. A model to explain the distinct roles of the CX3 complex in R-loop resolution and fork maintenance.

CX3 complex localizes to R-loops and recruits FANCM translocase by acting as an adaptor to regulate R-loop resolution. In the absence of the CX3 complex, FANCM recruitment to R-loops is perturbed, resulting in the accumulation of R-loop–induced genome instability. The role of the CX3 complex in R-loop resolution is independent of its fork protection/restart functions.

FA is a rare genetic disorder characterized by bone marrow failure, congenital abnormalities, chromosomal instability, and cancer susceptibility (36). Missense mutations in RAD51C have been identified to cause FA-like disorder and breast and ovarian cancers (40, 44, 45). However, the functional defects associated with these mutations leading to diseases are unclear. A homozygous missense mutation in RAD51C (R258H) was identified in a family with a characteristic FA-like phenotype (45). This mutant shows hypersensitivity to both ICL and replication stress–inducing agents and is partially defective in HR-mediated DSB repair (59). Our data show that R258H RAD51C is defective in FANCM interaction, resulting in impaired localization of FANCM to R-loop sites, leading to increased R-loops and TRCs. A recent study shows the importance of the FA pathway in coordinating transcription and replication for error-free genome duplication and the development of primordial germ cells (33). However, further studies are required to understand the defect associated with the R258H mutant in resolving R-loops and TRCs leading to FA syndrome.

Unresolved R-loops in the genome can cause fork stalling that requires fork protection and restart factors for efficient fork repair and recovery. Fork protection factors such as BRCA2, RAD51 paralogs, and FA proteins promote RAD51 assembly at the reversed forks to protect against nucleolytic degradation (41, 78, 91, 92). The CX3 complex also participates in the restart of stalled/collapsed replication forks, and this function is dependent on ATP binding and hydrolysis by RAD51C (52, 58). The ATP binding and hydrolysis–deficient mutants of RAD51C were competent to bind FANCM and promote FANCM localization to R-loop sites for its resolution. These data provide evidence for the distinct functions of RAD51C in R-loop resolution and fork repair/restart functions. Notably, the fact that RAD51C adenosine triphosphatase mutants were proficient in FANCM interaction and suppressing R-loop accumulation similar to WT RAD51C implies that the CX3 complex acts as an adaptor to recruit FANCM to the R-loop sites to promote its resolution (Fig. 8). Collectively, our findings highlight the importance of the RAD51 paralog CX3 complex in regulating FANCM-mediated R-loop resolution for the maintenance of genome integrity and provide evidence that this function may be dysregulated in the pathophysiology of FA and breast and ovarian cancers.

MATERIALS AND METHODS

Antibodies and reagents

Antibodies used in this study are as follows: anti-S9.6 (Sigma-Aldrich; catalog no. MABE1095; RRID: AB_2861387), anti-ssDNA (Sigma-Aldrich; catalog no. MAB3034), anti-FLAG (Cell Signaling Technology (CST); catalog no. 14793; RRID: AB_2572291), anti-FLAG (Sigma-Aldrich; catalog no. F7425; RRID: AB_439687), anti-H2AX (pS139) (BD Biosciences; catalog no. 560443; RRID: AB_1645592), anti-RAD51 [Santa Cruz (SC); catalog no. sc-8349; RRID: AB_2253533], anti-RAD51C (SC; catalog no. sc-56214; RRID: AB_2238197), anti-RAD51C (Abcam; catalog no. ab-72063; RRID: AB_2177279), anti-RAD51B (SC; catalog no. sc-377192), anti-RAD51D (SC; catalog no. sc-53432), anti-XRCC2 (SC; catalog no. sc-365854), anti-XRCC2 (Abcam; catalog no. AB180752), anti-XRCC3 (SC; catalog no. sc-271714), anti-XRCC3 (Abcam; catalog no. AB58467; RRID: AB_883592), anti-MCM3 (SC; catalog no. sc-365616; RRID: AB_10846721), anti-α-tubulin (SC; catalog no. sc-5286; RRID: AB_628411), anti-FANCD2 (Abcam; catalog no. AB108928; RRID: AB_10862535), anti-FANCM (Thermo Fisher Scientific; catalog no. PA5-76229; RRID: AB_2719956), anti-FANCM (Novus; catalog no. NBP2-50418; RRID: AB_2716711), anti-nucleolin (Cell Signaling Technology; catalog no. 14574; RRID: AB_2798519), anti-HA (Roche; catalog no. 10952100), mouse anti-rabbit immunoglobulin G (IgG)–horseradish peroxidase (HRP) (SC; catalog no. sc-2357; RRID: AB_628497), m-IgG binding protein–HRP (SC; catalog no. sc-516102; RRID: AB_2687626), anti-mouse tetramethyl rhodamine isothiocyanate (TRITC) (Sigma-Aldrich; catalog no. T5393; RRID: AB_261699), anti-rabbit fluorescein isothiocyanate (FITC) (Sigma-Aldrich; catalog no. F0382; RRID: AB_259384), anti-mouse IgG H&L (Alexa Fluor 488) (Abcam; catalog no. ab150125), donkey anti-rat Alexa Fluor 594 (Abcam; catalog no. ab150156; RRID: AB_2890252), rat anti-BrdU (Abcam; catalog no. ab6326; RRID: AB_305426), mouse anti-BrdU (BD Biosciences; catalog no. 347580; RRID: AB_400326), anti-PCNA (Cell Signaling Technology; catalog no. 2586; RRID: AB_2160343), and anti-RNAPIIS2P (Abcam; catalog no. ab5095; RRID: AB_304749).

Other commercial reagents used in this study are as follows: HU (Sigma-Aldrich; catalog no. H8627), Aph (Sigma-Aldrich; catalog no. A0781), CPT (Sigma-Aldrich; catalog no. C9911), 4′,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich; catalog no. D8417), 5-chloro-2-deoxyuridine (Sigma-Aldrich; catalog no. C6891), 5-iodo-2-deoxyuridine (Sigma-Aldrich; catalog no. I7125), protease inhibitor cocktail (Roche; catalog no.11836153001), thiazolyl blue tetrazolium bromide (Sigma-Aldrich; catalog no. M2128), agarose low gelling temperature (Sigma-Aldrich; catalog no. A9414), Protein G Sepharose beads (Cytiva; catalog no. GE17-0618-01), Protein A Sepharose beads (Cytiva; catalog no. GE17-0780-01), 5-ethynyl-2-deoxyuridine (Thermo Fisher Scientific; catalog no. A10044), benzonase (Sigma-Aldrich; catalog no. E1014), l-ascorbic acid (Sigma-Aldrich; catalog no. A92902), Mowiol 4-88 (Merck; catalog no. 81381), Immobilon Western Chemiluminescent HRP Substrate (Millipore; catalog no. WBKLS0500), Duolink In Situ PLA Probe Anti-Rabbit PLUS (Merck; catalog no. DUO92002), Duolink In Situ PLA Probe Anti-Mouse MINUS (Merck; catalog no. DUO92004), Duolink In Situ Detection Reagents Red (Merck; catalog no. DUO92008), and Click-iT EdU Alexa Fluor 647 Imaging Kit (Thermo Fisher Scientific; catalog no. C10340).

Cell culture

Human cell lines U2OS (American Type Culture Collection; HTB-96; RRID: CVCL_0042), HeLa Kyoto (Sachin Kotak laboratory; RRID: CVCL_1922), and hTERT RPE-1 (Sachin Kotak laboratory) were grown in Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin (Sigma-Aldrich), and 1% Glutamax (Gibco) under standard conditions (5% CO₂, humidified atmosphere) at 37°C.

Plasmids and transfections

Flag-tagged WT and D210N RNH1; Flag-tagged WT, K131A, K131R, and R258H RAD51C; and Flag-tagged WT and K117R FANCM constructs were generated using polymerase chain reaction (PCR)–based mutagenesis and cloned into the pcDNA3β vector. hXRCC3 constructs and all shRNA constructs used were previously reported (52, 56, 57). All plasmid transfections were performed using a Bio-Rad Gene Pulser Xcell (250 V and 950 μF), and experiments were performed between 36 and 48 hours after transfection.

Immunoblotting

Immunoblotting was performed as described previously (93). Cells were harvested and lysed in radioimmunoprecipitation assay buffer (without SDS) supplemented with cOmplete and PhosSTOP tablets (Roche). Protein concentrations were estimated by the standard Bradford assay. Proteins were resolved on an SDS–polyacrylamide gel electrophoresis gel and transferred onto polyvinylidene difluoride membranes (Millipore). The membranes were blocked using 5% dry milk (w/v) in TBST [50 mM tris-HCl (pH 8.0), 150 mM NaCl, and 0.1% Tween 20]. The membranes were then incubated with the primary antibody overnight at 4°C. The membranes were washed with TBST and incubated with respective HRP-conjugated secondary antibodies (1:10,000) for 1 hour at room temperature (RT). After TBST washes, membranes were developed with chemiluminescent HRP substrate (Millipore) and imaged using Chemidoc (Bio-Rad Chemidoc Imaging System). Primary antibodies used for immunoblotting were as follows: mouse anti-RAD51 (1:250, SC), mouse anti-RAD51B (1:250, SC), mouse anti-RAD51C (1:250, SC), mouse anti-RAD51D (1:100, SC), mouse anti-XRCC2 (1:250, SC), mouse anti-XRCC3 (1:250, SC), mouse anti-α-tubulin (1:2000, SC), mouse anti-MCM3 (1:1000, SC), mouse anti-Flag (1:2000, Sigma-Aldrich), rabbit anti-FANCM (1:1000, Abcam), and mouse anti-FANCD2 (1:500, SC).

Dot blot assay

Genomic DNA was isolated using the Thermo Fisher Scientific genomic DNA isolation kit, followed by quantification in a NanoDrop spectrophotometer. One hundred nanograms of genomic DNA was spotted on the Nylon N+ membrane (GE Lifesciences) and allowed to air dry. Then, the membrane was ultraviolet (UV)–cross-linked at 1200 μJ using UV Stratalinker. Following cross-linking, the membrane was blocked with 5% dry milk in TBST and further processed as described in the immunoblotting protocol with S9.6 antibody. Subsequently, the membrane was incubated in denaturing buffer (0.4 M NaOH and 0.6 M NaCl), followed by neutralizing buffer [1.5 M NaCl and 0.5 M tris (pH 7.4)] for 10 min each, and immunoblotted with anti-ssDNA antibody to detect total DNA. The primary antibodies against S9.6 (1:1000) (MABE1095) and ssDNA (1:10,000) (MABE3031) were purchased from Merck Millipore.

Coimmunoprecipitation assay

A total of 5 × 10⁶ cells per condition were harvested after indicated treatments and lysed in NTEN buffer [25 mM tris (pH 7.5), 150 mM NaCl, 1 mM EDTA, and 1% NP-40] containing cOmplete mini protease inhibitor and PhosSTOP phosphatase inhibitor cocktail (Roche), followed by a brief sonication. One milligram of protein from each sample was incubated either with no antibody or 2 to 5 μg of the indicated antibody overnight at 4°C. The following day, protein A/G beads were washed multiple times in the lysis buffer and added to the cell lysate + antibody mixture for 4 hours at 4°C under rotatory agitation. After incubation, the beads were washed three times in lysis buffer, and the protein was eluted by boiling the beads in 2× Laemmli buffer for 15 min. Immunoblotting was performed as described in the previous section. For experiments with benzonase treatment, cells were lysed in lysis buffer [25 mM tris (pH 7.5), 150 mM NaCl, and 1% NP-40] for 1 hour on ice. Salt concentration was lowered to less than 100 mM NaCl by adding the required amount of dilution buffer [25 mM tris (pH 7.5) and 1% NP-40] supplemented with protease inhibitors. Furthermore, the lysates were supplemented with 2 mM MgCl₂ and incubated with antibodies overnight in the presence or absence of benzonase (100 U/1 mg lysate). Washing, elution, and immunoblotting were performed as described above.

IP–mass spectrometry

Cells were harvested and lysed with Hepes-Triton buffer [35 mM Hepes KOH (pH 7.6), 200 mM NaCl, 1 mM EDTA (pH 8), 0.75% Triton X-100, and 8% glycerol, supplemented with protease inhibitor cocktail (Sigma-Aldrich)] on ice for 20 min. Clarified lysates (5 mg) were incubated with 6 μg of antibody overnight on a rotating wheel at 4°C. The following day, after incubating with 50 μl of protein A beads for 4 hours, the IP was washed once with lysis buffer and twice with IP wash buffer [100 mM tris-HCl (pH 8), 300 mM NaCl, 1 mM EDTA (pH 8), 1% Triton X-100, and 8% glycerol, supplemented with protease inhibitor cocktail (Sigma-Aldrich)]. Bound proteins were eluted by boiling beads in elution buffer [75 mM tris-HCl (pH 6.8), 2% SDS, 10% glycerol, 100 mM dithiothreitol (DTT), and 5% β-mercaptoethanol] for 30 min. Eluates were alkylated using DTT, reduced using iodoacetamide, digested with overnight incubation with trypsin (Merck, catalog no. 11047841001) at 37°C, and desalted. Digested samples were normalized to 0.2 μg/μl, and 10 μl of each sample was injected into the liquid chromatography (LC)–MS system. Peptides were analyzed in an LC-MS Orbitrap Fusion (Thermo Fisher Scientific). They were separated on a C-18 in-house column (15 cm in length, 1.5-μm particle size, internal diameter of 150 μm). Peptides were eluted with a gradient of 2 to 90% acetonitrile in 0.1% formic acid and analyzed in positive-ion mode of the nano-electrospray ionization source. The MS was operated in a data-dependent mode, automatically switching between MS and MS/MS acquisition. Automated MS data of peptides were acquired through an Orbitrap mass analyzer between 350 and 2000 mass/charge ratio above the 5000 count threshold. The 20 most intense ions were selected for MS/MS acquisition using an ion trap mass analyzer. The ion charge states were +2 to +8. The higher-energy C-trap dissociation (HCD) fragmentation energy was adjusted to 30%. Raw data were analyzed using Proteome Discoverer V2.5 (Thermo Fisher Scientific). The parent ions obtained from the MS/MS spectra were analyzed using the SEQUEST HT search engine, against the database of the proteins involved in DNA damage response that were retrieved from Reactome V93 (94) and manually curated references. Parameters that were set are as follows: The peptide tolerance was set to 15 parts per million in MS mode, and the fragment ion tolerance was set to 0.5 Da for HCD spectra (Orbitrap Fusion); two missed trypsin cleavages were allowed, and methionine oxidation was searched as variable modification, whereas cysteine carbamidomethylation was searched as a fixed modification. A t test was used to calculate P values of the quant ratios of protein, and the false discovery rate value was restricted to 1%. Statistical analysis was performed using MetaboAnalyst 6.0 (95) with the criteria of P < 0.05 and log 2 (fold change) > 2.

Cell survival assay

A total of 5000 cells per well were seeded in a 24-well plate for each condition and allowed to settle for 6 to 8 hours. Cells were treated with 1 mM HU for 4 hours. After treatment, cells were allowed to grow for 5 to 7 days. Later, cell survival was measured by MTT (0.3 mg/ml; Sigma-Aldrich) assay using a microplate reader (VersaMaxROM version 3.13). Percent cell survival was calculated as treated cells/untreated cells × 100.

Alkaline comet assay

Frosted glass slides (Bluestar) were coated with 1% agarose for at least 24 hours before the experiment. Cells were harvested and resuspended in phosphate-buffered saline (PBS). Thirty microliters (~50,000 cells) of cell suspension was mixed with 270 μl of 0.5% low–melting point agarose, and 100 μl of this mixture was spread onto precoated slides. Slides were incubated in chilled lysis buffer [2.5 M NaCl, 0.1 M EDTA, 10 mM tris-HCl (pH 8), 1% Triton X-100, and 10% dimethyl sulfoxide] at 4°C for 1 hour. After incubation, the slides were incubated in alkaline electrophoresis buffer (300 mM sodium hydroxide and 1 mM EDTA, pH > 13) for 20 min and transferred to an electrophoresis tank filled with chilled electrophoresis buffer. The electrophoresis was performed at 1 V/cm for 20 min at RT. Slides were then washed twice with PBS, fixed in ice-cold methanol for 5 min at RT, and washed again in PBS. Slides were washed with double-distilled H₂O and transferred to 70% ethanol for 15 min, followed by 100% ethanol for 30 min. Slides were then air-dried and stained with propidium iodide (2 μg/ml in Milli-Q). Images were acquired using an Apotome microscope (ZEISS Axio Observer). The comet tail moment was measured using CometScore software.

Immunofluorescence

Exponentially growing cells were seeded onto coverslips and allowed to settle for at least 12 hours. Indicated treatments were given. After treatment, cells were washed with PBS, preextracted with 0.5% Triton X-100 for 2 min on ice, and fixed in 4% formaldehyde for 10 min at RT. When Click-iT EdU staining was performed, cells were incubated with 10 mM EdU for 30 min before washing and preextraction, and EdU detection was performed according to the manufacturer’s recommendations (Thermo Fisher Scientific) before incubation with primary antibodies. After three PBS washes, coverslips were blocked in blocking buffer [0.5% bovine serum albumin (BSA) and 0.5% Triton X-100 in PBS] for 1 hour. The coverslips were incubated with the indicated primary antibodies for 2 hours at RT. After three washes with blocking buffer, the coverslips were incubated with respective FITC/TRITC-conjugated secondary antibodies for 1 hour at RT and then stained with DAPI (1 μg/ml; Sigma-Aldrich) for 5 min before mounting onto slides with Mowiol 4-88 (Sigma-Aldrich). Images were acquired using an Apotome microscope (ZEISS Axio Observer) and processed using ImageJ software. For S9.6 staining, cells were fixed with ice-cold methanol on ice for 15 min instead of 4% formaldehyde. Primary antibodies used for immunofluorescence were mouse anti-H2AX (pS139) (1:1000, BD Biosciences), mouse anti-Flag (1:1000, Sigma-Aldrich), mouse anti-S9.6 (1:200, Merck), rabbit anti-nucleolin (1:1000, CST), mouse anti-FANCD2 (1:500, SC), rabbit anti-FANCD2 (1:1000, Abcam), mouse anti-RAD51C (1:100, SC), rabbit anti-RAD51C (1:1000, Abcam), rabbit anti-FANCM (1:1000, Thermo Fisher Scientific), and mouse anti-FANCM (1:200, Novus).

Proximity ligation assay

PLA was performed using Duolink PLA Technology (Merck) as per the manufacturer’s guidelines. Samples were preextracted, fixed, permeabilized, and incubated with primary antibodies as described for immunofluorescence assays. Then, secondary antibody binding, ligation, and amplification reactions were performed according to the manufacturer’s guidelines. Duolink In Situ PLA Probe Anti-Rabbit PLUS, Duolink In Situ PLA Probe Anti-Mouse MINUS, and Duolink Detection Reagents Red (Merck) were used to perform the PLA reaction. Last, nuclei were stained with DAPI and mounted in Mowiol 4-88 (Sigma-Aldrich). Antibodies used for PLA reactions were anti-mouse PCNA (1:1000, CST), anti-rabbit RNAPIIS2P (1:1000, Abcam), and mouse anti-S9.6 (1:100, Merck). All other antibodies were used as described in the immunofluorescence assays.

DNA fiber assay

Cells were sequentially pulse-labeled with 25 μM CldU (Sigma-Aldrich) and 250 μM IdU (Sigma-Aldrich) for 30 min each or as indicated in the respective panel. After labeling, cells were incubated in ice-cold PBS on ice for 10 min, harvested, counted, and resuspended in 250 μl of PBS. Three microliters of the cell mixture was mixed with 7 μl of lysis buffer on glass slides (Thermo Fisher Scientific Superfrost) and allowed to stand for 7 min. Slides were inclined at a 45° angle to spread the suspension and fixed in methanol:acetic acid (3:1) solution at 4°C overnight. The following day, DNA was denatured by incubating in 2.5 M HCl for 1 hour and blocked with 2% BSA in 0.1% PBST solution (1× PBS and 0.1% Tween 20). Next, the slides were incubated with primary antibodies for 2.5 hours and secondary antibodies for 1 hour at RT. Coverslips were mounted on the slides with a Mowiol mounting medium and visualized using an Apotome microscope (ZEISS Axio Observer). Antibodies used for performing DNA fiber studies include rat anti-BrdU for CldU (1:500, Abcam), mouse anti-BrdU for IdU (1:250, BD Biosciences), rabbit anti-mouse IgG (Alexa Fluor 488) (1:500, Abcam), and donkey anti-rat IgG (Alexa Fluor 594) (1:500, Abcam).

Chromatin immunoprecipitation

ChIP was performed as reported earlier (64). Briefly, HeLa Kyoto cells were subjected to DNA protein cross-linking by cross-linking buffer [5 mM Hepes (pH 7.9), 0.1 mM EDTA, 10 mM NaCl, and 1.1% formaldehyde] for 15 min at RT in the dark. Excess formaldehyde was quenched by incubating the cells with 125 mM glycine for 5 min at RT. Cells were washed several times with ice-cold 1× PBS. Cytoplasmic fraction was removed by incubating cells with cell lysis buffer [5 mM PIPES (pH 8.0), 85 mM KCl, and 0.5% NP-40] for 10 min on ice. Later, cells were pelleted down and washed twice with 1× PBS containing 0.5% NP-40. The pellet was resuspended in lysis buffer [50 mM tris-HCl (pH 8.0), 10 mM EDTA, and 0.5% SDS] supplemented with protease inhibitor cocktail. The chromatin fraction was sonicated on ice (12 cycles of 30 s at low amplitude) to generate DNA fragments of length 200 to 800 base pairs. Lysates were then subjected to centrifugation at 12,500g for 2 min, and the resulting supernatant was precleared with Protein A Sepharose beads blocked with BSA (500 μg/ml) for 2 hours at 4°C. For each ChIP, 15 μg of precleared chromatin was incubated with 6 μg of RAD51C (AB 72063), XRCC2 (AB 180752), XRCC3 (AB 58467), Flag (CST 14793S), or FANCM (PA5 76229). An equivalent amount of chromatin was taken as a no-antibody control. Protein A beads (50 μl) were added to the chromatin-antibody complex and incubated for 2 hours at 4°C. The beads were washed for 5 min each with low salt buffer [0.1% SDS, 1% Triton X-100, 2 mM EDTA, 2 mM tris-HCl (pH 8.0), and 150 mM NaCl], high salt buffer [0.1% SDS, 1% Triton X-100, 2 mM EDTA, 2 mM tris-HCl (pH 8.0), and 500 mM NaCl], and LiCl wash buffer [250 mM LiCl, 1% NP-40, 1% sodium deoxycholate, 1 mM EDTA, and 10 mM tris-HCl (pH 8.0)] and twice with TE buffer [10 mM tris-HCl (pH 8.0) and 1 mM EDTA]. Last, the chromatin was eluted with elution buffer (1% SDS and 100 mM NaHCO₃). Elutes were de-crosslinked by adding NaCl to a final concentration of 300 mM and incubated overnight at 65°C. After incubation with RNase (0.1 mg/ml; Invitrogen, 12091-021) at 42°C for 1 hour, eluates were digested with proteinase K (0.1 mg/ml; Sigma-Aldrich, 70663-4) for 6 hours at 55°C. DNA was purified by phenol-chloroform-isoamyl alcohol extraction and ethanol precipitation. The isolated DNA was used as a template for quantitative PCR (qPCR) analysis on a Bio-Rad CFX Opus 96 Real-Time PCR system using iTaq Universal SYBR Green Supermix (Bio-Rad). Fold enrichment was calculated as follows: fold enrichment = 2^{−(CtIP − CtNo Ab)}, where CtIP and CtNo Ab mean threshold cycles of PCR done on DNA samples immunoprecipitated with a specific antibody and a no-antibody control, respectively. The sequences of primers used for qPCR analyses are as follows: APOE: 5′-CCGGTGAGAAGCGCAGTCGG-3′ (forward), 5′-CCCAAGCCCGACCCCGAGTA-3′ (reverse); RPL13A: 5′-GCTTCCAGCACAGGACAGGTAT-3′ (forward), 5′-CACCCACTACCCGAGTTCAAG-3′ (reverse); SNRPN: 5′-TGCCAGGAAGCCAAATGAGT-3′ (forward), 5′-TCCCTCTTGGCAACATCCA-3′ (reverse).

AlphaFold prediction and structural analysis

The full-length amino acid sequences for FANCM (UniProt ID: Q8IYD8), RAD51C (UniProt ID: O43502), and XRCC3 (UniProt ID: O43542) were obtained from the UniProt database. Structural models of complexes of FANCM with WT CX3 and R258H CX3 mutant were generated by AF3 (96). The interacting domains and interface binding energies of both FANCM-CX3 complexes were evaluated using PRODIGY and PDB-PISA, respectively (97, 98). The interactions were analyzed considering FANCM-CX3 models as a binary complex, which have been uploaded to the Dryad public repository (https://doi.org/10.5061/dryad.mw6m9069c). All five models were assessed for the interface binding energy (figs. S7B and S9A). Considering the relative orientation and similarity of the interface in the AF3-generated modeled protein complexes, Model 1 for both complexes was found suitable for analyzing the interacting domains and the respective interface binding energies. All the molecular interactions were analyzed and visualized with PyMOL 2.2.3 visualization software (99). The modeling files of complexes of FANCM with WT CX3 and R258H CX3 mutant have been deposited at https://modelarchive.org/ with accession numbers ma-xdl20 and ma-miymb, respectively (100).

Statistical analysis

Statistical parameters, including the number of biological replicates (n), SD, and statistical significance tests, are reported in the figure legends. All results were obtained from a minimum of three independent biological replicates. Single antibody control PLAs were realized only once per antibody. For immunofluorescence experiments, >80 cells were quantified for each experimental sample of each biological replicate. Statistical tests were performed on mean values of biological replicates. For box plots, boxes and whiskers indicate the 25th to the 75th and the 10th to the 90th percentiles, respectively, and median values are indicated. P values and the statistical test applied are described in the figure legends.

Supplementary Materials

This PDF file includes:

Figs. S1 to S11