The FANCM:p.Arg658* truncating variant is associated with risk of triple-negative breast cancer

Breast cancer is a common disease partially caused by genetic risk factors. Germline pathogenic variants in DNA repair genes BRCA1, BRCA2, PALB2, ATM, and CHEK2 are associated with breast cancer risk. FANCM, which encodes for a DNA translocase, has been proposed as a breast cancer predisposition gene, with greater effects for the ER-negative and triple-negative breast cancer (TNBC) subtypes. We tested the three recurrent protein-truncating variants FANCM:p.Arg658*, p.Gln1701*, and p.Arg1931* for association with breast cancer risk in 67,112 cases, 53,766 controls, and 26,662 carriers of pathogenic variants of BRCA1 or BRCA2. These three variants were also studied functionally by measuring survival and chromosome fragility in FANCM−/− patient-derived immortalized fibroblasts treated with diepoxybutane or olaparib. We observed that FANCM:p.Arg658* was associated with increased risk of ER-negative disease and TNBC (OR = 2.44, P = 0.034 and OR = 3.79; P = 0.009, respectively). In a country-restricted analysis, we confirmed the associations detected for FANCM:p.Arg658* and found that also FANCM:p.Arg1931* was associated with ER-negative breast cancer risk (OR = 1.96; P = 0.006). The functional results indicated that all three variants were deleterious affecting cell survival and chromosome stability with FANCM:p.Arg658* causing more severe phenotypes. In conclusion, we confirmed that the two rare FANCM deleterious variants p.Arg658* and p.Arg1931* are risk factors for ER-negative and TNBC subtypes. Overall our data suggest that the effect of truncating variants on breast cancer risk may depend on their position in the gene. Cell sensitivity to olaparib exposure, identifies a possible therapeutic option to treat FANCM-associated tumors.


INTRODUCTION
The genetic architecture of inherited breast cancer is complex and involves germline pathogenic variants in high and moderate-risk genes and polygenetic factors. The major high-penetrance breast cancer risk genes include BRCA1 and BRCA2, which are key factors in the DNA double-strand break repair through homologous recombination (HR) and in the inter-strand crosslink (ICL) repair as a part of the Fanconi Anemia (FA) pathway. 1,2 Recently, based on a prospective cohort of families carrying BRCA1 or BRCA2 pathogenic variants, the average cumulative risk by age 80 was estimated to be 72% and 69% for carriers of BRCA1 and BRCA2 pathogenic variants, respectively. 3 PALB2 has been previously considered a moderate-risk gene, but the latest estimate of about 44% lifetime risk associated with pathogenic variants may raise this gene to the high-risk group. 4 Pathogenic variants in moderate-penetrance genes ATM and CHEK2 are also associated with breast cancer, conferring a 20% average lifetime risk. 5,6 Recently, BARD1, RAD51D, BRIP1, and RAD51C have been proposed as risk factors for triple-negative breast cancer (TNBC) with BARD1 and RAD51D conferring high risk, and BRIP1 and RAD51C associated with moderate risk. 7 Thus, the risk associated with pathogenic variants in each gene may vary by breast tumor subtype.
Many of the BRCA/FA pathway genes when altered by biallelic mutations cause FA disease. The FANCM gene (FA complementation group M, OMIM #609644) encodes for a translocase, which is a member of the BRCA/FA molecular pathway but has been recently disqualified as a disease-causing factor for FA. 8,9 Some protein-truncating variants in the FANCM gene were described as moderate breast cancer risk factors with a greater risk of TNBC. In the Finnish population, FANCM:c.5101 C > T (p.Gln1701*, rs147021911) is relatively frequent and was reported to be associated with breast cancer with odds ratio (OR) of 1.86 with 95% confidence intervals (CIs) = 1.26-2.75. A larger effect was observed in familial cases (OR = 2.11; 95% CI = 1.43-3.32), for estrogen receptor-negative (ER-negative) breast cancer (OR = 2.37; 95% CI = 1.37-4.12) and for TNBC (OR = 3.56; 95% CI = 1.81-6.98). 10 We showed an increased risk (OR = 3.93; 95% CI = 1.28-12.11) of the FANCM:c.5791 C > T (rs144567652) truncating variant using familial cases and controls. In vitro analysis showed that this variant causes the skipping of the FANCM exon 22 and the creation of a downstream stop codon (p.Gly1906Alafs12*). 11 However, in the present study we refer to the FANCM:c.5791 C > T base change as to FANCM:p.Arg1931*, which is the conventional amino acid annotation (consistent with the stop codon creation according to genetic code). The FANCM:p.Arg1931* was also found to be associated with TNBC risk in the Finnish population (OR = 5.14; 95% CI = 1.65-16.0). 12 A burden analysis of truncating variants discovered by a re-sequencing analysis of the entire FANCM coding region in German cases and controls confirmed that FANCM pathogenic variants had a particularly high risk for TNBC (OR = 3.75; 95% CI = 1.0-12.85). 13 To study the effect of FANCM on breast cancer risk further, we tested three recurrent truncating variants FANCM:p.Arg658*, p. Gln1701*, and p.Arg1931*, within the OncoArray Consortium, a collaboration of consortia established to discover germline genetic variants predisposing to different human cancers (e.g., breast, colon, lung, ovary, endometrium and prostate cancers). 14 These three variants were tested for association with breast cancer risk in 67,112 breast cancer cases, 53,766 controls, and 26,662 carriers of pathogenic variants in BRCA1 or BRCA2. We also studied the functional effect of these three variants after their lentiviral transduction into a FANCM −/− patient-derived cell line in which we measured survival and chromosome fragility after exposure to diepoxybutane (DEB) or the poly (ADP-ribose) polymerase inhibitor (PARPi) olaparib.

Case-control analyses
We analyzed the association of three FANCM truncating variants, p.Arg658*, p.Gln1701*, and p.Arg1931*, with breast cancer risk for each variant separately and using a burden analysis. We tested 67,112 invasive breast cancer cases and 53,766 controls collected by the Breast Cancer Association Consortium (BCAC, http://bcac. ccge.medschl.cam.ac.uk/) and 26,662 carriers of BRCA1 or BRCA2 pathogenic variants collected by the Consortium of Investigators of Modifiers of BRCA1/2 (CIMBA, http://cimba.ccge.medschl.cam. ac.uk/), of whom 13,497 were affected with breast cancer and 13,165 were unaffected.
In the BCAC dataset we assessed the breast cancer risk associated with the FANCM variants in a primary overall analysis and in a restricted analysis including only countries in which the variant carrier frequencies were higher than the median of the frequencies. In these analyses we tested association with the variants in all available invasive breast cancer cases or in the ERpositive, ER-negative and TNBC subgroups (Table 1). In the overall analysis, no evidence of association was observed, either with the presence of any FANCM variant or with any of the three variants individually. However, FANCM:p.Arg658* showed a higher heterozygote frequency in ER-negative breast cancer cases (0.093%) than in controls (0.035%) with a greater than two-fold increased breast cancer risk (OR = 2.44, 95% CI = 1.12-5.34, P = 0.034). When only TNBC cases were considered, the association was stronger (OR = 3.79, 95% CI = 1.56-9.18, P = 0.009). No association with ERnegative breast cancer or TNBC was seen for p.Gln1701* or p. Arg1931* or for all mutations combined (Table 1). In the countryrestricted analyses, we confirmed the association found for p. Arg658* with risk of ER-negative disease and TNBC (OR = 2.31, 95% CI = 1.05-5.07, P = 0.047 and OR = 3.56, 95% CI = 1.46-8.69, P = 0.011, respectively). The restricted set also provided evidence for an association between p.Arg1931* and ER-negative subgroup (OR = 1.96, 95% CI = 1.24-3.10, P = 0.006), though not for TNBC. No significant association was observed for p.Gln1701* with either subgroups ( Table 1).

Functional studies
We tested the functional effect of FANCM:p.Arg658*, p.Gln1701*, and p.Arg1931* on DNA repair using genetic complementation assays (Fig. 1). These assays were based on the EGF280 cell line derived from immortalized fibroblasts from a patient who lacked the FANCM protein due to a homozygous c.1506_1507insTA (p. Ile503*, rs764743944) truncating variant. 8 Complemented FANCM −/− cells were tested for sensitivity to DEB and olaparib by measuring cell survival and chromosome fragility. The FANCM protein was not detectable in the EGF280 fibroblasts. The transduction of these cells with lentiviral vectors carrying wildtype (wt) FANCM cDNA and cDNAs harboring FANCM:p.Gln1701* and p.Arg1931* variants produced, as expected, different Cterminal truncated forms of FANCM. In the EGF280 cells transduced with FANCM:p.Arg658* no visible band was observed on western blot ( Fig. 1a and Supplementary Fig. 1). As we lack information on the epitope recognized by the antibody, we could not determine whether the p.Arg658*-derived truncated protein was unstable or if the epitope was lost due to the truncation. We therefore analyzed the mRNA expression of FANCM:p.Arg658* by reverse transcription and digestion of the PCR-amplified cDNAs. The c.1972C > T base substitution causing the p.Arg658* variant was expected to abolish a digestion site for the restriction enzyme TseI present in the wt sequence. TseI-digestion of wt and mutated cDNAs clearly indicated the presence of a mutated mRNA product in the EGF280 cells transduced with FANCM:p.Arg658* ( Fig. 1b and Supplementary Fig. 1).
In the DEB sensitivity-based assay (Fig. 1c), the EGF280 patientderived cell line showed a high-sensitivity phenotype, that was rescued by expression of the wt FANCM. EGF280 cells expressing FANCM:p.Arg658* failed to rescue DEB sensitivity and showed survival rates overlapping with those of the native EGF280 cells. In comparison, cells expressing FANCM:p.Gln1701* and p.Arg1931* variants showed an intermediate phenotype with survival rates significantly higher than those of EGF280 cells, though significantly lower than those of the cells expressing wt FANCM ( Fig. 1c and Supplementary Table 2). These results were confirmed in the chromosome fragility tests where the number of chromatidic breaks in cells harboring p.Gln1701* or p.Arg1931* variants was statistically lower than that of EGF280 cells or cells expressing the p.Arg658* and statistically higher than that of cells expressing wt FANCM (Fig. 1d). In the olaparib sensitivity-based assay, the survival rates of the cell lines transduced with the three FANCM truncating variants were not statistically different. Only at higher olaparib concentrations (>5000 nM) the survival rates of these cell lines were significantly lower than that of the wt FANCM cells and higher than that of the EGF280 cells ( Fig. 1e and Supplementary Table 3).

DISCUSSION
In this study we investigated the association of the three recurrent FANCM truncating variants p.Arg658*, p.Gln1701*, and p.Arg1931*, with breast cancer risk overall and by tumor subtype. While in non-Finnish Europeans these are the three most common FANCM truncating variants, their carrier frequency is low being 0.033, 0.21 and 0.21%, respectively (https://gnomad.broadinstitute.org/). 15 We conducted large case-control studies in 67,112 unselected breast cancer cases, 53,766 controls, and 26,662 carriers of BRCA1 or BRCA2 pathogenic variants. Furthermore, we performed functional analyses based on a patient-derived FANCM −/− cell line transduced with vectors carrying the three FANCM variants and tested for sensitivity to DEB or olaparib. Our genetic data suggest that FANCM:p.Arg658* is a risk factor for ER-negative and TNBC subtypes with statistically significant ORs of 2.44 and 3.79, respectively. These associations were confirmed when we restricted the analyses to countries with higher carrier frequencies.
In these restricted analyses we also found that the p.Arg1931* was associated with breast cancer risk in the ER-negative subtype with statistically significant OR = 1.96. (Table 1). These data, together with previously published genetic studies, [10][11][12][13] confirm that FANCM truncating variants are risk factors for breast cancer, with a stronger association for the ER-negative and TNBC subtypes. Our functional data, obtained in a background of a FANCM null cell line, support these findings showing that all three truncating variants were deleterious; hence, it is expected that, in the heterozygous state, any of these FANCM variants have partial activity. In the functional tests, we also observed that olaparib had a greater effect on survival of the cells harboring any of the FANCM:p.Arg658*, p.Gln1701*, or p.Arg1931* variants with respect to that on EGF280 cells complemented with wt FANCM (Fig. 1e). As this is consistent with previous results, 16 PARP1 inhibition might be a possible therapeutic approach to treat patients with breast tumors associated with germline FANCM pathogenic variants. On the contrary, the DEB sensitivity assays showed that FANCM:p.Arg658*, is associated with a stronger impairment of DNA repair activity, compared to p.Gln1701* and p.Arg1931*, possibly reflecting the position of protein truncation (Fig. 1c, d).
FANCM encodes for a key protein of the upstream FA/BRCA pathway mediating the assembly of the FA core complex. This protein is 2048 AA long, possesses in its N-terminal region an intrinsic ATP-dependent DNA translocase activity and, with its central region, recognizes the Bloom's complex, which is also involved in the DNA HR repair. By interacting with its C-terminal binding partner, the FA associated protein 24 (FAAP24), the FANCM protein brings to sites of ICL DNA lesions the FA and the Bloom's complexes initiating HR repair 17 (Fig. 2). We studied FANCM:p.Arg658*, p.Gln1701*, and p.Arg1931* in the same genetic FANCM −/− background and showed that, after exposure to DEB, the N-terminal FANCM:p.Arg658* had a statistically stronger effect on cell survival and chromosome stability (presumably due to less efficient DNA repair activity) than did p. Gln1701* and p.Arg1931*. This also suggests that in human living cells the FANCM:p.Arg658* variant might impair DNA repair more severely than p.Gln1701* and p.Arg1931*. We have shown that in vitro both the p.Gln1701*-and the p.Arg1931*-derived FANCM proteins are expressed and that the p.Arg658*-mRNA is transcribed ( Fig. 1a, b). An N-terminus fragment including the first 422 AA of FANCM was shown to be stable when expressed in human cell lines, 17 thus supporting the possibility that the FANCM:p. Arg658*-derived protein may also be expressed and stable. Hence, we hypothesize that the observed difference in survival and chromosome fragility of cells treated with DEB may be attributable to the diverse residual function of the different truncated forms of FANCM. In fact, the p.Gln1701*-and the p.Arg1931*-derived forms are expected to lose the interaction with FAAP24, but to retain the ability of binding other FANCM interacting proteins. Hence, our data suggest that the lack of interaction between FANCM and FAAP24 has a less severe impact on the DNA damage response than when protein truncation occurs upstream the FANCM domains AA 687-1104 and AA 1027-1362 mediating the interaction with the FA core complex and the Bloom's complex, respectively.
Previously published genetic and clinical data support our hypothesis of a position effect. FANCM pathogenic variants were shown to be associated with a moderate risk of developing highgrade serous epithelial ovarian cancer, but p.Arg1931* appeared to confer a lower risk. 18 Moreover, five female breast cancer These analyses were not possible in the country-restricted cases and controls as different countries were included for each variant. P-values were from Pearson chi-squared test Fig. 1 Functional studies of the FANCM:p.Arg658*, p.Gln1701* and p.Arg1931* truncating variants using the patient-derived FANCM −/− EGF280 cell line. a Western blot showing the FANCM expression in EGF280 cells complemented with lentiviral vectors harboring the three different variants. Bands corresponding to truncated FANCM protein were visible for EGF280 + p.Gln1701* and p.Arg1931*, and no bands were present for the EGF280 + p.Arg658*. b Study of the expression of the FANCM protein in EGF280 + p.Arg658*. The c.1972C > T base substitution, causing the p.Arg658* variant abrogates a digestion site for the restriction enzyme TseI that is present in the wild-type (wt) cDNA sequence. Total RNA was extracted from EGF280 + wtFANCM and from the EGF280 + p.Arg658* and subjected to reverse transcription. PCRamplified cDNA products were digested with TseI. Digested and undigested cDNAs were loaded. In the first two lanes are shown bands of 386 bp corresponding to uncut wt cDNA, and bands of 257 and 129 bp corresponding to cut wt cDNA. In next two lanes bands of 386 bp indicate that p.Arg658* cDNA was not cut due to the c.1972C > T base substitution abrogating the TseI site. In the two lanes after the molecular weight marker (M) undigested and digested products of the two previous PCR products were mixed 1:1 and loaded as a control. c Analysis of diepoxybutane (DEB) sensitivity on cell survival. The EGF280 cells expressing p.Arg658* are significantly more sensitive to DEB than the cells expressing p.Gln1701* or p.Arg1931* (P-values from Tukey's range test are reported in Supplementary Table 4). EGF280 and EGF280 + wtFANCM are used as controls (N = 3; error bars: standard deviation).  Table 5; N = 3; error bars: standard deviation). All blots derive from the same experiment and were processed in parallel probands carrying homozygous FANCM truncating variants were recently described. 9 Three of these, two homozygous for p. Gln1701*, and one for p.Arg1931*, developed breast cancer at age 52 years or later and their cells did not demonstrate chromosome fragility. The other two probands were homozygous for p.Arg658* and developed early-onset breast cancer (at age 29 and 32); in addition, one developed several cancers, and the other demonstrated chromosomal fragility. 9 Due to the rarity of the studied mutations in most populations, estimation of the risks is challenging. Preferably, the cases should be examined in comparison to geographically, ethnically and genetically matched controls. In the Finnish population, p. Gln1701* and p.Arg1931* are reported with carrier frequency of 1.62% and of 0.92%, respectively (https://gnomad.broadinstitute. org/). 15 Case-control studies based on the Finnish population showed a strong statistical evidence of association of p.Gln1701* with ER-negative disease, with OR of 2.37 (95% CI = 1.37-4.12, P = 0.0021), and with TNBC with ORs of 3.56 (95% CI = 1.81-6.98, P = 0.0002), 10 while p.Arg1931* was found associated with TNBC with an OR of 5.14 (95% CI = 1.65-16.0, P = 0.005). 12 However, as our 95% CI of risk estimates for TNBC included odds ratios of 2 for both the latter mutations, the published and our results are not mutually exclusive. Risk estimates associated with rare variants may depend on their frequency and the genetic background of the population studied. Hence, pooling the data from multiple outbred and admixed populations as it was done in the present study, may yield different risk estimates than those derived from geographically, ethnically and genetically matched controls, as in the Finnish studies. Indeed, it would have been interesting to test the FANCM variant position effect in the Finnish population, but unfortunately the p.Arg658* is very rare if not absent in this population (https://gnomad.broadinstitute.org/). 15 Recent attempts to identify novel, high-to moderate-risk breast cancer-predisposing genes have not been particularly fruitful. However, a few genes have emerged as potential risk factors for ER-negative disease and TNBC, with FANCM, BRIP1, and RAD51C being among those suggested to confer moderate risk of these subtypes. Other predisposing genes increasing the risk of ERnegative and TNBC may also exist. Hence, further gene discovery efforts should take into consideration that risk-associated variants may be associated with specific tumor subtypes and/or variation in risk may depend on the variant position. In addition, we provide evidence that lack of FANCM protein and truncating variants identified in breast cancer patients are associated with increased sensitivity to the PARPi olaparib suggesting a therapeutic opportunity to treat FANCM-associated breast tumors that warrants further investigation. The PARPi sensitivity test may also prove useful for preclinical investigation of further truncating or missense FANCM variants.
In summary, we have shown that FANCM:p.Arg658* is associated with risk of ER-negative breast cancer and TNBC. The outcomes of functional assays testing the DNA repair efficiency in complemented human cells support the hypothesis that breast cancer risk may be greater for N-terminal than C-terminal FANCM truncating variants. Further genetic studies and meta-analyses are warranted to derive more precise risk estimates for the different FANCM variants.

Study participants
The individuals included in this study were women of genetically confirmed European ancestry who were originally ascertained in 73 casecontrol studies from 19 countries participating in the BCAC or in 59 studies enrolling BRCA1 or BRCA2 pathogenic variants carrier from 30 countries participating in the CIMBA.

Ethics
All participating studies, listed in Supplementary Table 4 and Supplementary Table 5, were approved by their ethics review boards and followed national guidelines for informed consent. However, due to the retrospective nature of the majority of the studies, not all participant individuals have provided written informed consent to take part in the present analysis. The Milan Breast Cancer Study Group (MBCSG) was approved by ethics committee from Istituto Nazionale dei Tumori di Milano and Istituto Europeo di Oncologia, in Milan.
The BCAC studies contributed 67,112 invasive breast cancer cases and 53,766 controls. The majority of these studies were population-based, hospital-based or case-control studies nested within population-based cohorts (86%); few were family-clinic-based studies (14%; Supplementary Table 4). For each study subject, information on the disease status and the age at diagnosis or at interview were provided. Data on lifestyle risk factors were available for most subjects and clinical and pathological data were available for most cases. All these data were incorporated in the BCAC dataset (version 10). A total of 44,565 (66%) cases were ER-positive, 10,770 (16%) were ER-negative, and 4,805 (7%) were TNBC; 13,743 (20%) had a positive first-degree family history of breast cancer.
The CIMBA studies contributed 15,679 carriers of a pathogenic BRCA1 variant and 10,983 carriers of a pathogenic BRCA2 variant to this analysis (Supplementary Table 5). Nearly all (98%) of these carriers were ascertained through cancer genetic clinics; few carriers were recruited by populationbased sampling of cases or by community recruitment. In some instances, multiple members of the same family were included. For each pathogenic variant carrier, the information on the type of the BRCA1 or BRCA2 variant, disease status, and censoring variables (see below, Statistical analyses) were collected and included in the CIMBA database.

Genotyping
Genotyping of FANCM:p.Arg658*, p.Gln1701*, and p.Arg1931* truncating variants was conducted using a custom-designed Illumina genotyping array (the "OncoArray", Illumina, Inc. San Diego, CA, USA) at six independent laboratories. To ensure consistency of the genotype data, all laboratories used the same genotype-clustering file and genotyped the same set of reference-samples selected from the HapMap project. Samples with a call rate <95% and those with heterozygosity <5% or >40% were excluded. Further details of the genotype-calling and quality control have been described previously. 14 The cluster plots of the three FANCM truncating variants were curated manually to confirm the automatic calls ( Supplementary Fig. 2).

Statistical analyses
The BCAC data were analyzed to test the association between FANCM:p. Arg658*, p.Gln1701*, and p.Arg1931* and breast cancer risk. Logistic regression analyses were performed to estimate ORs with 95% CIs for variant carriers versus non-carriers, adjusting for country and the first ten principal components, as previously described. 19 P-values were calculated by applying the likelihood ratio test (LRT) comparing the model containing the variant carrier status as a covariate to a model without the variant carrier status. The primary analyses were performed including all invasive breast cancer cases and controls and subgrouping cases based on tumor hormonal status. We then performed a country-restricted analysis c.5791 C > T (p.Arg1931*) is also shown including the 50% of the countries with the higher variant carrier frequencies. Specifically, we included only countries in which the carrier frequencies in cases and controls combined were higher than the median of the carrier frequencies observed in all countries. Median frequencies were 0.007, 0.114 and 0.163 for p.Arg658*, p.Gln1701* and p.Arg1931* carriers, respectively.
The CIMBA data were analyzed to evaluate the association between each FANCM truncating variant and breast cancer risk in carriers of BRCA1 or BRCA2 pathogenic variant. A survival analyses framework was applied. Briefly, each variant carrier was followed from the age of 18 years until the first breast cancer diagnosis, or censored as unaffected at ovarian cancer diagnosis, bilateral prophylactic mastectomy, or age at last follow-up. The analyses were performed by modelling the retrospective likelihood of the observed genotypes conditional on the disease phenotype as detailed previously. 20 All analyses were stratified for country. The per-allele hazard ratio (HR), 95% CIs were estimated separately for each variant. A score test was used to derive P-values for the associations. The analyses of the BCAC data were performed using STATA version 15 (StataCorp LLC, College Station, Texas, USA). The analyses of the CIMBA data were carried out using custom-written code in Python and Fortran. All statistical tests were twosided and P-values <0.05 were considered statistically significant.

Western blot and mRNA expression studies
Cell lysis and western blot assays were performed as previously described. 8 The following primary antibodies were used: mouse monoclonal anti-FANCM antibody, clone CV5.1 diluted 1:100 (ref: MABC545, MERCK Millipore), mouse monoclonal anti-Vinculin diluted 1:3000 (ref: ab18058, abcam). Western blotting detection was achieved with Luminata TM Classico (Millipore) (Vinculin) and LuminataForte TM (Millipore) (FANCM). We used RT PCR to test the expression of the mutant FANCM:p.Arg658*. Total RNA was extracted (RNeasy Mini Kit Qiagen) from the wtFANCM and FANCM:p.Arg658* transduced EGF280 cell lines. Reverse transcription was performed using High-Capacity RNA-to-cDNA Kit (Thermofisher); a cDNA region corresponding to the FANCM sequence containing the amino acid (AA) position Arg658 was amplified by PCR using the forward: 5'-AGTAACAGGCAGGTCCTTCA-3´and reverse: 5'-TGATCTTGCCACAGTCTCCA-3' primers. The 386 bp PCR products were then digested with TseI restriction enzyme (New England Biolabs) for two hours at 65°C and analyzed by standard agarose gel electrophoresis.

Cell survival assay
The effect of the different FANCM variants on cell survival was measured with a Sulforhodamine B (SRB) assay. 21 One-thousand cells were seeded in 96-well plates and treated constantly with DEB or PARPi olaparib at the indicated concentrations until untreated cells reached confluency. Cell monolayers were fixed overnight at 4°C with 75 μl of 20% trichloroacetic acid (TCA). TCA was aspirated, and cells washed with tap water. Once dried, 50 μl of SRB was added to the wells and plates were incubated on a shaker at room temperature for 30 min. The excess of SRB dye was removed by washing repeatedly with 1% acetic acid, the plates were dried for 20 min, and the protein-bound dye was dissolved in 10 mM Tris for OD determination at 492 nm using a microplate reader (Tecan Sunrise™, Tecan Group Ltd. Männedorf, Switzerland). At least three independent experiments were performed for each cell line and in each experiment, 12 wells were measured per concentration point. These results were statistically analyzed using the Prism (GraphPad) software. Two-Way ANOVA test was used for single comparisons between different cell lines and statistical significance was assessed with the Tukey's range test. A Pvalue < 0.05 was considered statistically significant.

Chromosome fragility test
Chromosome fragility test was performed as previously described. 11 Twenty-five metaphases were scored for chromosome breakages using the Metafer Slide Scanning Platform from Metasystems. Results were graphed as distributions of metaphases presenting 0, 1, and >1 chromatid break. Statistical analysis was performed applying chi-squared test.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary

DATA AVAILABILITY
A subset of the genotype data analysed in this study is publicly available from the dbGaP repository and can be accessed at https://identifiers.org/dbgap:phs001265.v1. p1 (data generated as part of the BCAC studies) and at https://identifiers.org/dbgap: phs001321.v1.p1 (data generated as part of the CIMBA studies). The remaining genotype data analysed in this study (and generated as part of the BCAC and CIMBA studies listed in Supplementary Tables 4 and 5 of the related article, respectively) are not publicly available due to restraints imposed by the ethics committees of individual studies, but can be accessed from the corresponding author on reasonable request as described at https://doi.org/10.6084/m9.figshare.8982296. 22 Additional datasets generated during this study (and supporting Fig. 1 and Supplementary Tables 2 and 3 in the published article) are available on request as described above. The data generated and analyzed during this study are described in the following data record: https://doi.org/10.6084/m9.figshare.8982296. 22 F5-2012-305421 and P-SPHERE COFUND project), the Fanconi Anemia Research Fund Inc, and the "Fondo Europeo de Desarrollo Regional, una manera de hacer Europa" (FEDER). CIBERER is an initiative of the Instituto de Salud Carlos III, Spain. BCAC: we thank all the individuals who took part in these studies and all the researchers, clinicians, technicians and administrative staff who have enabled this work to be carried out.     Cancer" programgrant # CRN-87521 and the Ministry of Economic Development, Innovation and Export Tradegrant # PSR-SIIRI-701. The NBHS was supported by NIH grant R01CA100374. Biological sample preparation was conducted the Survey and Biospecimen Shared Resource, which is supported by P30 CA68485. The Northern California Breast Cancer Family Registry (NC-BCFR) and Ontario Familial Breast Cancer Registry (OFBCR) were supported by grant UM1 CA164920 from the National Cancer Institute (USA). The content of this manuscript does not necessarily reflect the views or policies of the National Cancer Institute or any of the collaborating centers in the Breast Cancer Family Registry (BCFR), nor does mention of trade names, Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.