Functional analysis of germline RAD51C missense variants highlight the role of RAD51C in replication fork protection

Abstract

Monoallelic or biallelic RAD51C germline mutations results in chromosome instability disorders such as Fanconi anemia and cancers. The bona fide function of RAD51C is to assist RAD51 nucleoprotein filament onto single-strand DNA to complete homologous recombination (HR) repair. In addition to HR repair, the role of RAD51C in DNA replication is emerging when replication forks are transiently or irreversibly stalled. We identified novel RAD51C variants of uncertain significance (VUS) from breast, ovarian, pancreatic and gastric cancer patients and functionally characterized the effect of these variants in replication fork protection and double-strand breaks (DSB’s) repair. In RAD51C-deficient Chinese hamster CL-V4B cells, expression of RAD51C F164S, A87E, L134S and E49K variants heightened sensitivity to mitomycin C (MMC), etoposide and PARP inhibition. Differently, expression of subset of RAD51C variants R24L, R24W and R212H displayed mild sensitivity to MMC, etoposide and PARP inhibition. Further functional characterization of a subset of variants revealed that Rad51C F164S, A87E, L134S and E49K variants displayed reduced RAD51 foci formation and increased overall nuclear single strand DNA levels in the presence of replication stress. Additionally, DNA fiber assay revealed that RAD51C F164S, A87E, L134S and E49K variants displayed defective replication fork protection upon prolonged fork stalling. Investigations using patient-derived lymphoblastoid cell line carrying heterozygous RAD51C L134S variant showed an impairment in RAD51 chromatin association and replication fork protection, suggestive of deleteriousness of this VUS variant. Overall, our findings provide more insights into molecular roles of RAD51C in replication fork integrity maintenance and in DSB repair.

RAD51C, replication fork, genome instability

Introduction

RAD51C is one among the essential homologous recombination (HR) genes that includes BRCA1, BRCA2, PALB2, XRCC3, RAD51D and BRIP1 whose germline mutation carriers are at high risk for breast and ovarian cancers (1). Germline pathogenic or likely pathogenic variants in known cancer-predisposing genes have implications for heritability. Nonetheless, the impact of variants of uncertain significance (VUS) on the etiology of disease and cancer predisposition has not been fully appreciated. In recent years, RAD51C novel VUS have been encountered in the clinics (2) and some variants are functionally characterized in vitro (3–5). RAD51C is a part of two protein complexes named BCDX2 (RAD51B, RAD51C, RAD51D and XRCC2) and CX3 (RAD51C and XRCC3) whose primary role is to facilitate HR by assisting RAD51 nucleofilament formation at double-strand breaks (DSBs) (6). Rad51C gene encodes a ~40 kDa protein, consisting of 376 amino acids. Rad51C contains a Holliday junction resolution domain (1–126 amino acids), a nucleotide binding region such as walker A motif (125–132 amino acids) and walker B motif (238–242 amino acids), a region for interaction with Rad51B, Rad51D and XRCC3 (79–136 amino acids) and the C-terminal nuclear localization signal (366–370 amino acids) (7,8) (Fig. 1A). Previous functional characterization of RAD51C germline variants revealed that RAD51C G125V and L138F are highly sensitive to mitomycin C (MMC) induced inter-strand crosslinks (ICL) in both DT-40 chicken cells and in Chinese hamster cells (3–5). Similarly, RAD51C D159N, G264S, T287A, R366Q and R258H hypomorphs displayed partial sensitivity to MMC in both chicken DT-40 and Chinese hamster cells (3,4). The sensitivity to DNA-damaging agents is attributed to the ability of these variants to repair DSBs through HR. In addition to human cell lines reports, identification and characterization of Rad51 paralog complex, from Caenorhabditis elegans, revealed that RFS-1/RIP-1 is essential for HR and Rad51 foci formation in vivo (9). Furthermore, several biochemical and biophysical studies demonstrated that RFS-1/RIP-1 complex structurally remodels Rad51 filaments to an open conformation and this remodeling activity facilitates strand exchange with the template DNA (9). Mechanistically, RFS-1/RIP-1 limits the Rad51-ssDNA turnover by binding to the 5′-end of individual Rad51-ssDNA filaments and mediate remodeling in a 5′–3′ direction (10).

Figure 1

Identification of Rad51C germline variants. (A) Schematic representation of the identified variants at the protein level, positioned on the basis of the known functional domains in Rad51C. (B) Human Rad51C protein has been modeled using Swiss-model server (https://swissmodel.expasy.org) on the basis of the archaeal RadA as template PDB ID: 2ZUB. The residues are indicated in yellow.

In addition to their classical HR-mediated DNA DSB’s repair, several studies uncovered single strand DNA gap prevention, replication fork protection and restart roles for HR and Fanconi anemia genes such as BRCA1, BRCA2, RAD51 and FANCD2 (11–15). However, the clear role of RAD51 paralogues in replication fork protection and stability is poorly understood. RAD51C, also a member of RAD51 paralog, has emerging roles in replication fork protection (5), and it promotes replication fork restart (16) in mammalian cell lines in the presence of hydroxyurea (HU)-induced fork stalling. Indeed, RAD51C G125V and L138F variants that lack functional RAD51C exhibited deficits in replication fork protection and restart (5). Hence, broader implementation of in vitro functional characterization including different chemotherapeutic drug sensitivity assay, RAD51 nucleation assay and DNA replication forks protection assay is essential for the pathogenicity assessment of the VUS variants.

In this study, we identified RAD51C VUS in patients diagnosed with breast, ovarian, pancreatic or gastric cancers and had familial history of cancers. Using multiple approaches such as drug sensitivity assay, RAD51 foci formation, replication fork protection and chromosome aberration assays, we report that a subset of RAD51C such as F164S, A87E, L134S and E49K are likely deleterious variants. The study has provided a weight of evidence that RAD51C functions in stalled replication fork protection beyond its key roles in DSB repair. Our results provide additional evidence suggesting the need for the functional characterization of missense RAD51C variants encountered in the clinic.

Results

Identification of RAD51C germline VUS

Demographic and clinical data of all 12 patients carrying RAD51C VUS mutation is summarized in Table 1. The list of the co-occurring mutations is summarized in Supplementary Table S1. Of all heterozygous RAD51C VUS carriers identified in our cohort, seven had breast cancer, three had ovarian cancer, one had pancreatic cancer and one had multiple cancers (gastric cancer, endometrial cancer, kidney cancer and squamous cell carcinoma). Among the seven breast cancer patients, two had a second breast cancer after their first diagnosis of breast cancer. The age at diagnosis of breast and ovarian cancer in the patients is between 30 and 49 years old. Overall, 9 of the 12 RAD51C carriers demonstrated family history of cancers. In summary, a total of 10 RAD51C missense variants was identified and are classified as VUS. Up to date, except that RAD51C c.635G > A (R212H) variant has been reported before yet functionally characterized (16–20), the rest RAD51C missense variants: c.70C > T (R24W), c.71G > T (R24L), c.145G > A (E49K), c.233C > T (T78I), c.253 T > G (C85G), c.260C > A (A87E), c.401 T > C (L134S), c.491 T > C (F164S) and c.589G > A (E197K) are novel. These missense variants are distributed around Walker A and Walker B motif on RAD51C genes (Fig. 1A). Rad51C 3D structure was homology modeled on the basis of archaeal RadA crystal structure and the variants discovered in this study are highlighted (Fig. 1B) The deleterious effects of RAD51C missense variants predicted by in silico tools are variable and inconsistent (Table 2); therefore, this has prompted us to functionally evaluate the significance of these germline missense variants identified in our patients.

Table 1

Clinicopathological characteristics of the RAD51C VUS variants

Patient no.	RAD51C VUS variants	Protein change	Gender	Race	Cancer types and age of cancer diagnosis	Family history of cancers
1	c.70C > T	R24W	Female	Chinese	Breast cancer (30 years old) [invasive mucinous carcinoma]	Breast cancer, kidney cancer, colorectal cancer
2	c.71G > T	R24L	Female	Chinese	Gastric cancer (49 years old) [Other poorly cohesive carcinoma], Endometrial cancer (37) [Endometroid], Kidney cancer (39), [Papillary urothelial carcinoma (TCC)], Squamous cell carcinoma (51)	Colorectal cancer, breast cancer, lung cancer, prostate cancer, skin cancer, gastric cancer, endometrial cancer, pancreatic cancer
3	c.145G > A	E49K	Female	Chinese	Ovarian cancer (37) [Serous carcinoma]	Breast cancer, colorectal cancer, kidney cancer
4	c.233C > T	T78I	Female	Chinese	Breast cancer (42)	Breast cancer
5	c.233C > T	T78I	Male	Chinese	Pancreatic cancer (55) [Adenocarcinoma]	Nil
6	c.253 T > G	C85G	Female	Chinese	Breast cancer (39) [Invasive ductal breast carcinoma], Second breast cancer (59)	Nil
7	c.260C > A	A87E	Female	Chinese	Breast cancer (37) [Invasive ductal breast carcinoma]	Nil
8	c.401 T > C	L134S	Female	Chinese	Breast cancer (49) [Invasive ductal breast carcinoma]	Breast cancer, brain cancer
9	c.491 T > C	F164S	Female	Chinese	Ovarian cancer (45) [Serous carcinoma]	Kidney cancer
10	c.589G > A	E197K	Female	Chinese	Ovarian cancer (48) [Transitional cell carcinoma]	Breast cancer, gastric cancer
11	c.635G > A	R212H	Female	Chinese	Breast cancer (34) [Invasive ductal breast carcinoma], Second breast cancer (42)	Gastric cancer
12	c.635G > A	R212H	Female	Chinese	Breast cancer (44) [Invasive lobular breast carcinoma], Thyroid cancer	Lung cancer, breast cancer

All the RAD51C variants identified in our study are listed in the table along with their clinicopathological characteristics.

Table 2

RAD51C missense variants and in silico predictions

No.	Nucleotide change	Protein change	Protein change	dbSNP	ClinVar	REVEL	SIFT	Mutation Taster	gnomAD	Reported in study
1	c.70C > T	p.Arg24Trp	R24W	rs878855180	VCV000241776	0.258	Damaging	Disease causing	Nil	Nil
2	c.71G > T	p.Arg24Leu	R24L	rs777554369	VCV000421982	0.242	Damaging	Polymorphism	Nil	Nil
3	c.145G > A	p.Glu49Lys	E49K	rs753709131	VCV000423572	0.216	Damaging	Disease causing	3.98E-06	Nil
4	c.233C > T	p.Thr78Ile	T78I	rs112832782	VCV000241767	0.063	Tolerated	Polymorphism	6.36E-05	Nil
5	c.253 T > G	p.Cys85Gly	C85G	rs1060502595	VCV000409853	0.156	Tolerated	Disease causing	3.98E-06	Nil
6	c.260C > A	p.Ala87Glu	A87E	rs1567786127	VCV000631280	0.562	Damaging	Disease causing	Nil	Nil
7	c.401 T > C	p.Leu134Ser	L134S	rs1233795152	VCV000538772	0.703	Damaging	Disease causing	4.05E-06	Nil
8	c.491 T > C	p.Phe164Ser	F164S	rs1060502589	VCV000409845	0.757	Damaging	Disease causing	Nil	Nil
9	c.589G > A	p.Glu197Lys	E197K	rs1555597094	VCV000660834	0.093	Tolerated	Disease causing	Nil	Nil
10	c.635G > A	p.Arg212His	R212H	rs200857129	VCV000187633	0.639	Damaging	Disease causing	8.36E-05	PMID: 29263802, 21 597 919, 25 338 684, 25 980 754, 32 566 746

For the identified Rad51C variants, in silico predictions were performed using REVEL, SIFT and Mutation Taster online tools to determine the pathogenicity of the variants.

RAD51C F164S, A87E, L134S and E49K variants are sensitive to MMC, etoposide and PARP inhibition

To evaluate the functional impacts of the RAD51C missense VUS variants, we took advantage of the established RAD51C-deficient (CL-V4B) Chinese hamster cell line model system together with their RAD51C proficient parental cell lines (V79B) (17), previously used to demonstrate the role of RAD51C in DSB repair (4,18–20). CL-V4B cells displayed remarkable sensitivity to MMC. For complementation assay, we stably overexpressed flag-tagged RAD51C wild-type, empty vector (EV) and various RAD51C mutants (R24L, R24W, E49K, T78I, C85G, A87E, L134S, F164S, E197K, R212H) in CL-V4B cells. The protein expression of various mutants in CL-V4B cells are shown in Figure 2A. In agreement to previous reports (4,5), our experiment showed that CL-V4B cells stably expressed EV are dramatically sensitive to MMC and etoposide compared with the parental V79B cells. The introduction of RAD51C wild-type plasmid was able to restore survival upon MMC and etoposide to a level comparable to the V79B cells (Fig. 2B and C). Compared with wildtype RAD51C (WT), CL-V4B cells expressing the RAD51C F164S, A87E, L134S and E49K variants were significantly more sensitive to MMC-induced ICLs and etoposide-induced DSBs (Fig. 2B and C). The rest of the RAD51C mutants exhibited very mild to no sensitivity to MMC and etoposide (Fig. 2B and C). Because error-free HR repair is required to resolve ICL-induced DNA breaks (21), we questioned whether poor RAD51 localization and foci formation to the sites of breaks might contribute to this heightened sensitivity. Consistently, RAD51 foci formation was severely impaired in F164S and A87E mutants (Fig. 2D). However, RAD51 foci formation was moderately but significantly affected in L134S and E49K mutants. (Fig. 2D). We next investigated whether RAD51C mutants that display reduced RAD51 foci formation can confer lethality to PARP inhibition because of poor HR function, as reported in BRCA1/2-deficient cells (22). As speculated, RAD51C F164S, A87E, E49K and L134S variants were significantly sensitivity to PARP inhibition in a decreased order of magnitude (Fig. 2E). Overall, these results suggest that these subset of RAD51C missense variants phenocopy Rad51C-deficient scenario evidenced by reduced RAD51 foci formation and displaying increased sensitivity different DSB’s-inducing agents such as to MMC, etoposide and PARP inhibition.

Figure 2

RAD51C F164S, A87E, L134S and E49K variants are sensitive to MMC, etoposide and PARP inhibition. (A) Overexpression of Flag-RAD51C mutants in CL-V4B cells (Rad51C^−/−). WT indicates Flag-tagged wildtype RAD51C. (B) MMC was treated for 5 days and percentage cellular viability was calculated for all the variants identified. (C) The variants expressing cells were treated with etoposide and recovered in fresh media for 4 days. Then, the percentage cell viability was calculated. (D) MMC-treated RAD51C variants and WT overexpressing cells were subjected RAD51 staining and cells with >5 foci were counted using confocal microscopy. (E) In total, 1 μM talozoparib was treated for 5 days and percentage cellular viability was calculated. Note: For B, C, D and E, unpaired t-test was performed in comparison with WT-RAD51C expressing cells. Each experiment was performed in triplicates.

A subset of RAD51C variants exhibit replication fork deprotection and chromosome instability in the presence of replication stress

Because F164S, A87E, L134S and E49K exhibited sensitivity to chemotherapeutics that induce DSBs accompanied with reduced RAD51 foci formation, we next focused our further functional analyses on these four potentially damaging variants. We first tested the effect of these variants in unperturbed DNA replication. Single molecule DNA fiber assay revealed that the expression of F164S, A87E, L134S and E49K variants did not affect replication fork speed in unperturbed S phase (Supplementary Fig. S1A). To address the effect of RAD51C variants in response to replication stress, we treated F164S, A87E, L134S and E49K expressing cells with HU that induces replication fork stalling, collapse and DSB’s induction upon prolonged treatment. Consistent to previous observations, Rad51C-deficient CL-V4B cells displayed dramatic sensitivity to HU treatment and expression of WT Rad51C rescued sensitivity in these cells (4,5). Interestingly, F164S and A87E displayed heightened sensitivity, whereas L134S and E49K showed moderate sensitivity to HU treatment (Fig. 3A). Replication fork stalling induced by HU remodels replication forks into a four-way junction called reversed forks whose controlled resection is required for prompt fork restart and stability (23). Hence, to this end, we next adopted DNA fiber nascent degradation assay as shown in Figure 3B to monitor the replication fork stability in Rad51C-deficient cells. In agreement with previous reports, Rad51C-deficient CL-V4B displayed reduced nascent DNA stability/protection in the presence of HU as measured by the degradation of nascent DNA tracks labeled with nucleotide analogs, IdU and CIdU (5) (Fig. 3B and C). Furthermore, expression of wild-type RAD51C or optional addition of MRE11 exonuclease inhibitor mirin both rescued nascent DNA degradation in CL-V4B cells (4) (Fig. 3C). This result further settles the role of RAD51C in protecting nascent DNA against MRE11 nuclease upon HU-induced replication fork remodeling in CL-V4B cell lines. Therefore, we next tested the replication fork stability of the four probably damaging RAD51C variants such as F164S, A87E, L134S and E49K in the presence of HU. Interestingly, Rad51C F164S, A87E, L134S and E49K variants expressing cells displayed compromised nascent DNA protection mirroring Rad51C-deficient state in CL-V4B cells (Fig. 3E). Unscheduled resection of the nascent DNA in the presence of replication stress leads to single-strand gaps accumulation, replication fork collapse and DSBs generation (12). Along these lines, we next monitored the overall nuclear ssDNA levels by incorporating BrdU into cells followed by HU treatment. RAD51C F164S, A87E, L134S and E49K mutants displayed increased overall nuclear ssDNA levels in the presence of replication stress induced by HU and these nuclear ssDNA accumulations were readily rescued by expression of wild-type RAD51C (Fig. 3E). RAD51 nucleoprotein filament formation is essential to prevent resection of nascent DNA by MRE11-mediated resection (13,24). Therefore, the ssDNA regions observed in the presence of HU in RAD51C mutants might be because of reduced RAD51 nucleoprotein filament formation, and this, in part, could have contributed to poor nascent DNA protection. In different mammalian cell lines, Rad51C loss has been shown to increase spontaneous sister chromatid gaps and chromosome breaks in the mitotic population (4,5). Hence, we next tested whether the unscheduled nascent DNA degradation observed in Rad51C variants in the presence of HU (Fig.3C and D), translates into chromosome abnormalities in metaphase spreads. Strikingly, in contrast to cells expressing WT Rad51C, expression of F164S, A87E, L134S and E49K Rad51C variants significantly increased chromosome breaks in the metaphase spreads (Fig.3F). Taken together, these data indicate that the identified Rad51C variants give rise to detectable impairment in replication fork stability, this further support the deleterious effects of these RAD51C missense variants.

Figure 3

A subset of RAD51C variants exhibit replication fork deprotection and chromosome instability in the presence of replication stress (A) Rad51C WT, EV, F164S, A87E, L134S and E49K variants expressing cells were treated with HU and the percentage cell viability was calculated. (B) Cartoon of the replication fork stability assay experimental setup and fiber interpretation is shown in the figure. Representative fiber image displaying fork degradation in CL-V4B-EV (Rad51C-deficient) cell lines is shown with respect to CL-V4B-WT (Rad51C-proficient). (C) CL-V4B-EV (Rad51C-deficient) is optionally treated with mirin and CL-V4B-WT (RAD51C-proficient) cells were labeled sequentially with nucleotide analog IdU and CldU for 30 min each followed by incubation with 4 mM HU for 5 h. At least 75 fibers were counted for each condition and Mann–Whitney test were performed. (D) RAD51C WT, EV, F164S, A87E, L134S and E49K variants expressing hamster cell lines were subjected to replication fork stability assay. At least 75 fibers were counted for each condition and Mann–Whitney test were performed. All the experiments were performed in triplicates. (E) RAD51C mutants expressing cells were treated with BrdU for 16 h followed by treatment with 2 mM HU for 4 h and stained for native BrdU signal omitting denaturing step and this represents the overall nuclear ssDNA intensity. At least 200 cells were counted for each condition and Mann–Whitney test were performed. The overall nuclear ssDNA signals were scored by measuring intensity of BrdU signal. (F) Chromosome breaks were detected by metaphase spread analysis. RAD51C WT, EV, F164S, A87E, L134S and E49K variants expressing cells were treated with 4 mM HU for 5 h. Left: Experimental plan and a representative metaphase spread with chromosome breaks indicated (Scalebar, 5 μm). Right: The graph bar depicts chromosome breaks per metaphase spreads. Unpaired t-test was performed in comparison with WT-RAD51C expressing cells. Each experiment was performed in triplicates. For each condition at least 50 metaphase spreads were analyzed.

Patient-derived cell line harboring RAD51C c.401 T > C (L134S) displayed compromised genome maintenance function

To further confirm the observed phenotype in patient samples, we next attempted to test the effect of these variants in genome stability maintenance. Because of limited biospecimen from patients, we were able to obtain only one EBV-immortalized B lymphoblastoid cell line (LCL) derived from a breast cancer patient harboring heterozygous RAD51C c.401 T > C (L134S) mutation. The patient presented with invasive ductal breast carcinoma at the age of 49 and has family history of breast and brain cancers (Fig. 3A). Because the RAD51C L134S variant expression in CL-V4B cells displayed moderate sensitivity to MMC, etoposide, PARP inhibition and exhibited poor fork protection function, we next questioned whether heterozygous RAD51C L134S LCL display these features of genome instability. At the basal level, the patient LCL showed mild reduction in endogenous RAD51C protein levels compared with healthy controls (Fig. 3B), whereas the total Rad51 levels and replication fork speed remained unaltered (Fig. 3C; Supplementary Fig. S1B). To access the HR function in LCLs, we induced DSB’s with etoposide and investigated the RAD51 chromatin association. Interestingly, treatment with etoposide increased the amount of chromatin bound phospho-RPA in both RAD51C L134S patient and healthy LCLs but only the patient cells showed impaired RAD51 chromatin association, suggesting the HR function is compromised (Fig. 3D). Like the observations from CL-V4B expressing Rad51C L134S variant (Fig.2C and3A), Rad51C L134S LCLs displayed mild sensitivity to etoposide and HU (Fig. 4D). Interestingly, replication fork stability assay revealed that RAD51C L134S carrier cells displayed reduced fork protection in comparison to healthy control (Fig.4F) that is consistent to results obtained from hamster cells from Figure 3D. Similarly, Rad51C L134S carrier treated with etoposide or HU displayed increased chromosome breaks in the metaphase spreads (Fig.4G), suggesting that these chromosome anomalies might be because of compromised HR and replication fork protection.

Figure 4

Patient-derived cell line harboring RAD51C c.401 T > C (L134S) displayed compromised genome maintenance function. (A) Pedigree of the affected family showed the proband presented with invasive ductal breast carcinoma at age of 49 and has family history of breast and brain cancers. (B) RAD51C total protein comparison between patient and two healthy controls level. (C) Rad51 total protein comparison between healthy and patient-derived LCLs. (D) Chromatin fractionation was performed after treating LCLs with 10 μM etoposide for 16 h and for RAD51, pRPA and MCM7. (E) Healthy and patient-derived Rad51C L134S LCLs were treated with 10 nM etoposide or 0.5 mM HU for 72 h and the percentage cellular viability was calculated. (F) Replication fork stability assay was performed as described previously in Figure 3B, at least 100 fibers were counted for each condition. The ratio of CIdU and IdU lengths were plotted and Mann–Whitney tests were performed. (G) Healthy and Rad51C L134S patient-derived cell lines were treated with 10 nM etoposide or 0.2 mM HU for 6 h and metaphase spreads were performed for chromosome breaks analyses. At least 50 metaphase spreads were analyzed for each condition.

Discussion

Monoallelic germline alteration in RAD51C confer susceptibility to breast and ovarian cancer was first reported by Meindl et al. (3) and the predominance of RAD51C missense variants in breast and ovarian cancer is further highlighted by Osorio et al. (1), together suggested that RAD51C missense variants that are usually clinically classified as VUS should be further assessed for pathogenicity.

Here, our study demonstrates roles for novel Rad51C variants in the control of DNA replication fork integrity. In our cohort, altogether, we identified ten RAD51C germline VUS variants from 12 patients that have not yet been functionally characterized. Importantly, our functional characterization identified four missense Rad51C variants, E49K, A87E, L134S and F164S that displayed features of genome instability. E49K is located at domain required for Holliday junction resolution (25); A87E is located at domain required for its interaction with other paralogs such as RAD51B, RAD51D and XRCC3 (25,26). Interestingly, L134S is located at a 4-amino acid distance from the walker–ATP motif. L138F in walker–ATP motif has been reported to compromise replication fork protection and restart activity (5). Finally, F164S variant is located at the center of the protein between walker A and B motifs.

Remarkably, A87E and F164S variants displayed strong defect in Rad51 foci formation (Fig. 2D), with heightened sensitivity to MMC, etoposide and PARP inhibition (Fig. 2B, C and E). Differently, Rad51C E49K and L134S displayed moderate defect in Rad51 foci formation (Fig. 2D), with modest sensitivity to MMC and etoposide (Fig. 2B and C). These results suggest that the heightened cellular sensitivity to chemotherapeutics is because of severe defect in HR in Rad51C A87E and F164S variants in presence of DSB’s-inducing agents.

Replication fork stalling in the presence of HU induces fork remodeling by SMARCAL1, ZRANB3, HLTF and Rad51 (12,27). The remodeled reversed forks are protected by BRCA2 that facilitate RAD51 nucleoprotein filament formation to counteract nucleases-mediated degradation of nascent DNA (11,13). In the process of fork remodeling and subsequent fork protection, lower amount of Rad51 is enough to promote replication fork reversal (23,28), whereas, higher levels of Rad51 is required for replication fork protection (29,30). This suggests that Rad51 levels at the stalled replication forks determine the outcome of the stalled DNA replication fork metabolism. In addition to BRCA2, Rad51 levels at the replication forks are regulated by Rad51 paralogs when challenged with replication stress-inducing agents such as CPT and HU (5,16). Intriguingly, it has been shown that loss of Rad51 paralogs such as Rad51C, XRCC2 and XRCC3 trigger MRE11-mediated nascent strand degradation in hamster cell lines (5). Furthermore, Rad51C K131A variant residing within the walker A motif and L138F variant identified in breast/ovarian cancers has also been shown to increase nascent DNA degradation (5).

Along these lines, our data identify novel Rad51C variants E49K, A87E, L134S and F164S, whose expression results in replication fork deprotection (Fig. 3D), as a result, there is an increase in the overall ssDNA accumulation upon HU-induced fork stalling (Fig. 3E). Although all the four variants displayed similar levels of replication fork deprotection, there were differences in the extent of cellular sensitivity and chromosome breaks between the variants under HU challenged conditions (Fig. 3A and F). A87E and F164S were extremely sensitive to HU mirroring Rad51C null background with increased chromosome breaks, whereas E49K and L134S displayed moderate sensitivity to HU with moderate chromosome breaks in the metaphase spreads (Fig. 3A and F). In A87E and F164S variant expressing cells, the unscheduled nascent DNA degradation could have been converted into DSB’s. Because of severe HR defects (Fig. 3D), the A87E and F164S expressing mutants could have triggered apoptosis with increased chromosome breaks (Fig. 3F) in the presence of HU.

In mammalian cells, it has been shown that the HR function of BRCA1/2 is essential for embryonic cell viability (31). But more recent studies suggest that in the absence of HR, preserving fork protection is sufficient to sustain cell viability in BRCA2-deficient mouse embryonic stem cells and tumor cells (32,33). In addition, another study showed that, preventing nascent strand degradation by suppressing fork reversal mechanism improved cellular fitness in BRCA1/2-deficient cancer cells (34). Further, different separation of function mutant studies in non-transformed human mammary cell lines revealed that HR-directed repair but not fork protection is critical for cellular survival (35). In this context, our data suggest that the variants that displayed strong HR defect showed heightened sensitivity for HU (e.g. A87E and F164S variants) in comparison with other two variants (e.g. E49K and L134S). Further, our data from patient-derived cell lines suggest that heterozygous missense Rad51C L134S mutation was sufficient to trigger nascent DNA degradation (Fig. 4E) but displayed only mild sensitivity to HU with slight increase in chromosome breaks. Taken together, our study hints that HR mechanism contributes to cellular fitness, whereas fork protection appears to contribute minimally for cell survival when challenged with HU in these cell lines. Future studies are required to understand the role of Rad51C pathogenic variants in replication fork protection, replication fork restart/HR and their relationship with cancer cell survival. Further biochemical reconstitution studies for these variants are essential to dissect their interaction ability with other RAD51 paralogues such as RAD51B, XRCC2, XRCC3 and RAD51D. Hence in vitro reconstitution of these variants with the DNA substrates will be highly informative in understanding the sequential role of Rad51 paralogs in replication fork remodeling and protection. The roles of RAD51C in cancer predisposition are not well understood, although several studies correlate RAD51C mutation with cancer susceptibility (36). Because PARP inhibitors are highly efficient in BRCA1/2-deficient cancers, narrowing down RAD51C variants that display HR and fork protection defect can help stratify patients beneficial for PARP inhibitors treatment. Therefore, evidence-based classification of RAD51C VUS variants is essentially needed for accurate molecular diagnosis and gene-informed genetic counseling.

Materials and Methods

Patient data and specimens

Patients consulted at the Cancer Genetics Service of National Cancer Centre Singapore with RAD51C germline variants identified through multi-gene panel testing were enrolled to the study with written informed consent. The genes included in the panel are listed in Supplementary Table S1 in detail. Testing was performed by commercial laboratories accredited by College of American Pathologists/Clinical Laboratory Improvement Amendments (CAP/CLIA). Patient demographics, personal medical information and family history were retrieved from electronic medical records. This study is approved by SingHealth Centralized Institution Review Board (CIRB 2011/826/B).

Cell cultures

Chinese hamster lung fibroblasts V79B (RAD51C-wild-type) and CL-V4B (RAD51C-deficient) (19,37) were cultured in Dulbecco's Modified Eagle Medium supplemented with 10% fetal bovine serum (FBS) (Life Technologies, Carlsbad, CA), 2 mM L-glutamine (Life Technologies) and 100 U/ml penicillin/streptomycin (Life Technologies). EBV-immortalized B LCLs were derived from a breast cancer patient harboring RAD51C c.401 T > C (L134S) and from healthy donors. LCLs were maintained in RPMI1640 supplemented with 10% FBS. All cells were maintained at 37°C in a 5% CO₂–95% air atmosphere.

Plasmid constructs and establishment of stable cell lines

The cDNA encoding full length human RAD51C was PCR-amplified and cloned in pcDNA3.1(+) expression vector using Gibson assembly method. RAD51C mutant plasmids were generated by introducing the mutation using site-directed mutagenesis. EV, wild-type and mutant plasmids were transfected into V79B or CL-V4B cells using Lipofectamine 3000(Life Technologies). Stable cell lines were created by selection with 800-1000ug/ml G418(Life Technologies) for 2 weeks.

Western blot

The expression of RAD51C protein was confirmed by western blot. Cells were lysed in radioimmunoprecipitation assay lysis buffer (Thermo Fisher Scientific, Waltham, MA) supplemented with Halt Protease and Phosphatase Inhibitor cocktail (Thermo Fisher Scientific) and lysates were clarified by centrifugation at 16000 g for 20 min at 4°C. In total, 20 μg of protein was separated by SDS gel and transferred to polyvinylidene fluoride membranes. Membranes were incubated in 5% non-fat milk in Tris-buffered saline consisting of 0.1% Tween20 for 1 h at room temperature. The membrane was incubated with primary antibody diluted in blocking buffer overnight at 4°C. After three washings in Tris-buffered saline consisting of 0.1% Tween20, the membrane was incubated with horseradish peroxidase secondary antibody (Dako, Santa Clara, CA) for 1 h at room temperature. The bands were visualized with a Western Lightning chemiluminescent kit (PerkinElmer) and ChemiDoc Imaging system (Bio-Rad Laboratories, Inc., Hercules, CA). Pan-actin served as internal loading control.

Cell viability assay

The parental V79B and CL-V4B stably expressed RAD51C mutants were treated with ICLs agent MMC, etoposide, PARP inhibitor and HU to assess cellular viability. Cells seeded onto 96-well plate at a density of5000 cells per well were treated with MMC (100 nM, 5 days), etoposide (10 nM for 6 h and recovered with fresh media for 4 days), talazoparib (1 μM, 5 Days) and HU (0.4 mM for 12 h and recovered with fresh media for 4 days). Cell viability was determined via the cell counting kit CCK-8 assay (Dojindo Laboratories, Kumamoto, Japan) at an optical density of 450 nm on the Victor spectrophotometer (PerkinElmer Life Sciences, Waltham, MA).

RAD51 foci formation and metaphase spreads

Cells were treated with 10 ng MMC/ml for 6 h and then observed for RAD51 foci formation by immunofluorescence staining. Cells containing five or more foci were considered positive, with 200 cells scored for each condition over triplicate experiments. For the chromosome breaks assay, CL-V4B cells (Rad51C^−/−) with the expression of WT and different Rad51C variants were treated with 4 mM HU for 5 h. The compound was washed of 3 times with 1× PBS and incubated with the fresh media for 12 h. Later the cells were incubated with colcemid (1 μg/ml) for 4 h and the metaphase spreads were performed as described previously (4,5).

Replication fork stability assay

We adopted single molecule DNA fiber assay to measure the replication fork stability upon HU-induced nucleotide depletion. We sequentially labelled with 50 μM IdU and 250 μM CIdU nucleotide analogs for 30 min each and treated with 4 mM HU for 5 h to stall replication forks optionally supplemented with 50 μM mirin. The labelled cells were harvested with ice–cold 1× phosphate-buffered saline (PBS) and mixed gently. In total, 4 μl of cellular suspension were mixed with 6 μl of lysis buffer (200 mM Tris pH 7.5, 50 mM EDTA and 0.5% SDS). This mixture was layered on the super frost slides, after 3-min incubation the slides were slowly tilted at 45 angle to allow cell suspension to flow to the end of the slide. Later the slides were dried and fixed with 3:1 methanol: acetic acid for overnight at 4°C. Dried slides were rehydrated and denatured with 2.5 M HCl for 60 min. The denatured slides were washed thrice with 1× PBS and incubated with blocking buffer (1% bovine serum albumin, 1× PBS and 0.1% Triton-X). IdU and CIdU were stained using rat-BrdU antibody (Abcam #ab6326) and mouse-BrdU antibody (Becton Dickinson, 347580) with their respective 488 and 594 conjugated fluorophores. At least 75 fibers were counted for each condition over triplicate experiments.

Acknowledgements

We are grateful to Professor. Ganesh Nagaraju, Indian Institute of Science for kindly providing us the V79B, Rad51C proficient and CL-V4B Rad51C-deficient Chinese Hamster Cell Lines. We thank Mr Gopishankar Thirumoorthy, CCBT for technical support.

Conflict of Interest statement. All authors have no conflict of interest to declare.

Funding

Singapore Ministry of Education (MOE) under its Academic Research Fund (AcRF) Tier 1 [2019-T1-001-018]; National Cancer Centre Research Fund Terry Fox Grant [NCCRF-YR2018-NOV-1]; National Medical Research Council (NMRC) Clinician Scientist Award [MOH-000654].

Data Availability Statement

All the data related to this article will be made available by Joanne Ngeow ([email protected]) and Arun Mouli Kolinjivadi ([email protected]) upon reasonable request.