RNF168-mediated localization of BARD1 recruits the BRCA1-PALB2 complex to DNA damage

DNA damage prompts a diverse range of alterations to the chromatin landscape. The RNF168 E3 ubiquitin ligase catalyzes the mono-ubiquitination of histone H2A at lysine (K)13/15 (mUb-H2A), forming a binding module for DNA repair proteins. BRCA1 promotes homologous recombination (HR), in part, through its interaction with PALB2, and the formation of a larger BRCA1-PALB2-BRCA2-RAD51 (BRCA1-P) complex. The mechanism by which BRCA1-P is recruited to chromatin surrounding DNA breaks is unclear. In this study, we reveal that an RNF168-governed signaling pathway is responsible for localizing the BRCA1-P complex to DNA damage. Using mice harboring a Brca1CC (coiled coil) mutation that blocks the Brca1-Palb2 interaction, we uncovered an epistatic relationship between Rnf168− and Brca1CC alleles, which disrupted development, and reduced the efficiency of Palb2-Rad51 localization. Mechanistically, we show that RNF168-generated mUb-H2A recruits BARD1 through a BRCT domain ubiquitin-dependent recruitment motif (BUDR). Subsequently, BARD1-BRCA1 accumulate PALB2-RAD51 at DNA breaks via the CC domain-mediated BRCA1-PALB2 interaction. Together, these findings establish a series of molecular interactions that connect the DNA damage signaling and HR repair machinery.


B
RCA1′s role in maintaining genome integrity requires numerous protein interactions. The BRCA1 really interesting new gene (RING), coiled-coil (CC), and BRCA1 C-terminal (BRCT) domains each perform specialized functions within homologous recombination (HR) DNA repair. Significantly, germline mutations found in patients with breast and ovarian cancer often cause disruption to these domains 1,2 . BRCA1 and BARD1 form a heterodimer with E3 ubiquitin ligase activity via their respective RING domains. BRCA1 and PALB2 each contain CC regions that directly interact with one another, permitting the formation of a larger BRCA1-PALB2-BRCA2-RAD51 (BRCA1-P) macro-complex that promotes the accumulation of RAD51 at sites of DNA damage, and consequently RAD51 filament formation [3][4][5] . While the BRCA1-PALB2 interaction is not necessary for the focal accumulation of BRCA1, it is required for efficient PALB2 and RAD51 foci formation, indicating that BRCA1 recruits PALB2-RAD51 to doublestranded DNA break (DSB) sites. The BRCA1 BRCT domainmediated interaction with ABRAXAS and RAP80 allows for the BRCA1 CC domain to simultaneously bind PALB2. Thus, it has been proposed that the BRCA1-ABRAXAS-RAP80 (BRCA1-A) complex is responsible for localizing PALB2 to DSBs 6 . However, several reports have demonstrated that the BRCA1-RAP80 association reduces cellular HR activity 7,8 . Moreover, RAP80 depletion was shown to have little effect on RAD51 foci formation 7 , making the role of RAP80-ABRAXAS in recruiting BRCA1-P ambiguous. Currently, the precise molecular mechanism responsible for the localization of the BRCA1-P complex has not been defined.
In this study, we aimed to further decipher the role of RNF168 in recruiting BRCA1-P to DSBs (Fig. 1a).
To determine whether mouse phenotypes were related to Brca1-P complex recruitment, we investigated the effects of Rnf168 KO and the Brca1 CC mutation on Brca1, Palb2, and Rad51 foci formation. As expected, Rnf168 KO resulted in loss of 53bp1 IRIF positive cells ( Supplementary Fig. 2a). Of note, Brca1 CC/CC cells produce a Brca1-CC protein that is stable and forms foci 15 . Strikingly, Brca1-WT and Brca1-CC IRIF were dramatically reduced in Rnf168 in KO cells ( Fig. 1f and Supplementary Fig. 2a, b). The number of Palb2 and Rad51 IRIF positive Brca1 +/+ and Brca1 +/CC cells also decreased in the absence of Rnf168. However, Brca1 CC/CC cells, which have defective Brca1-Palb2 interaction, had few detectable Palb2 and Rad51 foci; therefore, Rnf168 KO had little measurable impact ( Fig. 1f and Supplementary Fig. 2a, b). Overall, these results suggest a model whereby Rnf168 promotes Brca1 localization, and consequently, the Brca1-dependent recruitment of Palb2-Rad51 to DSBs.
RNF168-generated and K63-linked ubiquitin polymers are known to recruit RAP80 [11][12][13] , and thereby the BRCA1-A complex to DSB sites. To determine if RAP80 is the mediator of BRCA1-P complex recruitment, we directly compared Rap80 and Rnf168 KO MEFs for Brca1 and Rad51 IRIF ( Supplementary Fig. 2c). In contrast to Rnf168 −/− cells, there was a mild reduction in Brca1, and no effect on Rad51 foci positive Rap80 −/− cells (Supplementary Fig. 2c), suggesting that RNF168 recruits the BRCA1-P complex through another pathway.
We next measured the effects of RNF168 expression levels on BRCA1 foci formation. As expected, RNF168 siRNA reduced the number of foci positive cells as well as size of BRCA1-FL foci. Additionally, RNF168 siRNA resulted in a complete loss of detectable BRCA1-ΔBRCT foci; by contrast, BRCA1-ΔRING foci were unaffected (Fig. 2c). Thus, RNF168-induced BRCA1 focal accumulation requires the RING domain.
BARD1 ANK and BRCT domains promote BRCA1 foci. We aimed to decipher the mechanism by which the BRCA1 RING domain promotes foci formation. BRCA1 and BARD1 both contain N-terminal RING domains that directly interact and form an E3 ubiquitin ligase 21 . Because BRCA1-ΔRING does not interact with BARD1 nor engender ubiquitin ligase activity, we separately disrupted the E3 ligase activity versus the BRCA1-BARD1 interaction. The BRCA1-I26A mutation is commonly used to block ubiquitin ligase activity, but may retain residual activity. However, a triple I26A + L63A + K65A mutation (BRCA1-3A) fully abrogates E3 ligase activity, while retaining the ability to interact with BARD1 22 . The BRCA1-C61G protein retains the RING domain but prevents BRCA1-BARD1 heterodimerization and results in decreased protein stability 23 . However, we obtained cells with suitable ectopic protein expression ( Fig. 3a and Supplementary Fig. 3a). We found that BRCA1 foci formation was identical between BRCA1-FL and BRCA1-3A expressing cells. By contrast, BRCA1-C61G formed fewer foci that were smaller and similar in size to the BRCA1-ΔRING protein (Fig. 3b). Similar to BRCA1-ΔRING, RNF168 siRNA also had no effect on BRCA1-C61G foci ( Supplementary Fig. 3b). To further ensure that E3 ligase activity was absent, we generated a construct that physically linked BRCA1-ΔRING and BARD1-ΔRING peptides (BRCA1-ΔRING-L-BARD1-ΔRING), thereby creating a synthetic heterodimer where both proteins were devoid of RING domains ( Fig. 3a and Supplementary Fig. 3a). Here, BRCA1-ΔRING-L-BARD1-ΔRING IRIF was identical to BRCA1-FL (Fig. 3b), suggesting that the interaction with BARD1 is required for efficient BRCA1 focal accumulation.
Given the established role of RNF168 E3 ligase activity in generating mUb-H2A 10 , we reasoned that RNF168 promotes BRCA1-P recruitment to DSB sites via BARD1 BUDR binding to mUb-H2A. To further investigate this possibility, we assessed a series of ectopic add back RNF168 mutations using an MDA-MB-231 RNF168 −/− clone ( Supplementary Fig. 4c, d). Similar to results in MEFs, BRCA1, and RAD51 IRIF were significantly reduced in MDA-MB-231 RNF168 −/− cells ( Supplementary  Fig. 4e). The RNF168 ΔRING mutation results in loss of the RING domain and E3 ligase activity, whereas RNF168 R57D retains E3 ligase activity but specifically blocks the ability of RNF168 to mono-ubiquitinate H2A 10 . Additionally, RNF168 ΔPID lacks the C-terminally located PALB2-interacting domain (PID), which is required for RNF168 to directly interact with PALB2 and facilitate the latter's foci formation in the absence of BRCA1 27 Foci positive cells were determined to be those with ≥10 foci/nuclei. ***p < 0.001, **p < 0.01, *p < 0.05, ns not significant compared to FL (percentage positive cells: unpaired, two-tailed t-tests size; number and size per nucleus: nonparametric Mann-Whitney tests). See Source data for additional information. c Cells from (a) were subject to scrambled (Sc) and RNF168-targeting siRNA followed by IR and HA foci formation measured by immunofluorescence. Left; western blot showing RNF168 depletion. Above; representative cells (scale bar, 10 μm). Below; foci analyses as described in (b), n = 3 biological replicates. See Source data for additional information.
( Fig. 4d). The mRNA expression of ectopic constructs was within a similar range, but RING domain mutant protein levels were dramatically reduced compared to wild-type ( Supplementary  Fig. 4f), indicating that RING mutations are detrimental to protein stability. To bypass this issue, we used a CMV promoter for RING mutant constructs, which produced cell lines with comparable protein expression to UBC promoter-driven RNF168 WT and RNF168 ΔPID expressing cells (Fig. 4d). IRIF experiments revealed that RNF168 WT and RNF168 ΔPID formed robust foci and fully rescued 53BP1 IRIF. In contrast, RNF168 ΔRING and RNF168 R57D formed smaller foci and failed to rescue 53BP1 IRIF ( Supplementary Fig. 4g). Moreover, while RNF168 WT and RNF168 ΔPID rescued, RNF168 ΔRING and RNF168 R57D had no effect on BARD1 and BRCA1 ( Fig. 4e and Supplementary Fig. 4g), or RAD51 foci ( Fig. 4f and Supplementary Fig. 4g). These data indicate that the E3 ligase activity, and specifically the ability of RNF168 to generate mUb-H2A, is required for efficient BARD1-BRCA1-RAD51 DSB localization.
We postulated that in the absence of BRCA1, a direct association of only the BRCA1 RING domain with PALB2 may direct PALB2-RAD51 to DSBs and rescue HR. Thus, we generated a construct containing the BRCA1 RING domain linked to PALB2 cDNA (HA-RING-L-PALB2) and expressed in BRCA1 null MDA-MB-436 cells. RING-containing constructs  Fig. 5a). Because MDA-MB-436 cells are BRCA1 null, and end resection is necessary for efficient PALB2-RAD51 loading 28 , we also subject cells to GFP-or 53BP1-targeting sgRNA, which increased RPA32 IRIF (Fig. 4g and Supplementary Fig. 5a). HA-RING-L-PALB2 formed detectable foci in sgGFP cells, but HA-RING-L-PALB2 and RAD51 foci were dramatically increased ( Fig. 4g and Supplementary Fig. 5b), and PARPi resistance was observed in sg53BP1 cells (Fig. 4h). Moreover, we confirmed that HA-RING-L-PALB2 and RAD51 foci were entirely reliant on RNF168 activity ( Fig. 4i and Supplementary Fig. 5c). Thus, the BRCA1 RING domain is an RNF168-dependent DSB localization module.

RNF8-BRCA1-A provides a backup route of recruitment.
Despite numerous pathogenic BRCA1 mutations being located within the BRCT repeats, loss of the BRCT domain had relatively mild effects on BRCA1 protein localization (Fig. 2b). To gain additional insight into how BRCA1 BRCT domain mutations impact BRCA1 localization and HR, we compared several truncating mutations located at varying distances from the BRCT repeat region, which spans~aa's 1642-1856 (Fig. 5a). We confirmed previous observations 29 , with stop codons located close to (aa 1558*) or within (aa 1814*) the BRCT domain resulting in reduced protein expression despite similar mRNA levels (Fig. 5a), and consequently few foci forming cells ( Fig. 5b and Supplementary Fig. 5d). In contrast, the BRCA1-ΔBRCT protein (aa 1495*), which stops~147 aa prior to the first BRCT repeat, was stable and formed relatively robust foci (Fig. 5a, b). However, BRCA1-ΔBRCT expressing cells failed to promote CtIP, RPA32, and RAD51 IRIF, relative to full-length BRCA1 expressing cells ( Fig. 5b and Supplementary Fig. 5d). Hence, even when BRCA1 is stable and can localize to DSBs, in the absence of the BRCT domain, CtIP recruitment, and consequently end resection and HR, is inefficient.

Discussion
In the current study, we report an RNF168-governed pathway that is responsible for the recruitment of the BRCA1-P complex to DSBs. A BUDR was recently discovered within the BARD1 BRCT domain, which directly binds mUb-H2A 26 . In support of this, mutating residues within the BUDR, or specifically blocking the ability of RNF168 to generate mUb-H2A, reduced BRCA1-P complex recruitment to DSBs.
We found human and mouse cell lines to be highly dependent on RNF168 activity for efficient BRCA1, BARD1, and RAD51 foci formation. Loss of RNF168 activity has previously been shown to reduce BRCA1 foci formation 9,11,27 . However, others reported mild or little impact 16,31 , and we are unable to pinpoint reasons for these discrepancies, but speculate assessment of foci versus laser-stripes, or quantification methods, could contribute. RNF168 is capable of promoting BRCA1-A complex recruitment through the production of K63-linked ubiquitin polymers that form a binding module for RAP80 11,13 . Moreover, RNF8 generated K63-linked ubiquitin chains 6,32-34 , as well as SUMObinding [35][36][37] , can contribute to RAP80 foci. In agreement, RNF168 −/− cells had reduced, but detectable RAP80 foci. Interestingly, BRCA1-A complex disruption was highly detrimental for BRCA1-P activity in RNF168 −/− and BARD1 −/− cells, suggesting the RAP80-ABRAXAS axis confers a backup recruitment pathway (Fig. 5h). Importantly, individual disruption of the RING or BRCT recruitment pathways were insufficient to entirely abolish BRCA1 foci, shedding light on how BRCA1 mutant cancers utilize hypomorphic proteins, which often lack these domains, to contribute to residual HR activity and PARPi resistance 20,29,[38][39][40] .
We found that BRCA1 BRCT mutations often induce proteasomal degradation. Another study showed a point mutation within the BRCT domain did not affect protein expression but did reduce Brca1 foci formation 41 ; the reason for this discrepancy is unclear. Although the BRCA1-A complex provided inefficient BRCA1-P localization, it is likely sufficient to maintain residual HR and the relatively normal development of Rnf168 −/− mice. Brca1 +/CC reduces the number of intact Brca1-Palb2 complexes, and Brca1 +/CC , Rnf168 −/− KO mice had a greater reduction in Palb2 and Rad51 foci compared to either genotype alone, as well as similar developmental defects to Brca1 CC/CC , Rnf168 +/+ mice 15 . Moreover, while we were unable to detect Palb2 and Rad51 foci in Brca1 CC/CC cells, it is possible that residual Rad51 loading, below the foci detection threshold, account for the birth of Brca1 CC/CC live mice. In contrast, Brca1 CC/CC , Rnf168 −/− resulted in early embryonic lethality, potentially due to a complete abolishment of residual Rad51 loading ( Supplementary  Fig. 5g). Another possibility is that Rnf168 compensates for loss of Brca1 activity and directly loads Palb2 27 . However, we were unable to detect Palb2 or Rad51 foci in Rnf168 +/+ , Brca1 CC/CC cells, making the role of the Rnf168-Palb2 interface unclear in this setting.
Despite both Rnf168 and 53bp1 KO activating resection, we speculate that the effects on Brca1 mutant mouse viability differ due to the additional requirement of Rnf168 in recruiting Brca1, and subsequently Palb2-Rad51 to DSBs. Although we acknowledge that these phenotypes are not fully explained. HR and developmental defects associated with mice harboring Brca1 exon 2 mutations were previously shown to be rescued by Rnf168 KO due to loss of 53bp1 localization and DNA end resection activation 16 . Brca1 exon 2 mutations produce a Brca1 Ringless protein isoform, resulting in loss of Bard1 interaction 42 . Here, Rnf168 is dispensable for Rad51 foci and HR, possibly due to loss of the Brca1 Ring domain and an inability to influence Bard1mediated Brca1-P recruitment. Thus, the BRCA1 RING domain could determine phenotypes associated with modulation of RNF168 activity in cancer. BRCA1 185delAG and C61G mutations are commonly found in patients and produce proteins that fail to interact with BARD1 20,23,39,43 . Cancers with these mutations may reduce RNF168 activity as a means of promoting DNA end resection and inducing RAP80-mediated RINGless BRCA1-P loading, and consequently PARPi resistance. In contrast, hypomorphic proteins generated that retain the RING, may be highly dependent on RNF168 for residual BRCA1 protein activity 44 . Future studies will be required to investigate these possibilities. BARD1 and 53BP1 both bind the same mUb-H2A modification 26 , implicating protein occlusion as a DSB resection regulatory mechanism. However, the switch to resection is likely more complex, given BRCA1 exon 11 and BRCT regions are also required for efficient resection 15,16,45,46 . Indeed, the BRCA1 RING peptide was capable of stabilizing BARD1 and localizing to DSBs, but insufficient to activate resection (Fig. 4g). Overall, we establish a series of molecular interactions linking RNF168ubiquitin signaling to BRCA1-PALB2-associated HR. Intriguingly, BRCA1 molecules may simultaneously localize to DSBs, activate resection, and promote PALB2-RAD51 loading.

Methods
Mouse models. The Fox Chase Cancer Center (FCCC) Institutional Animal Care and Use Committee (IACUC) approved experiments involving mice and all experiments were conducted in compliance with ethical regulations. Mice were housed at the FCCC Laboratory Animal Facility in standard conditions with health and pathogen monitoring by facility staff. Generation of mice harboring the Brca1 CC allele was previously described 15 . Rnf168 − mice were generated with the help of the FCCC Transgenic Mouse facility using CRISPR/Cas9 technology. Guide RNA sequences targeting the Rnf168 Ring domain were selected based on high predicted cut efficiency and minimal predicted off-target sites (https://portals. broadinstitute.org/gpp/public/analysis-tools/sgrna-design). Several sgRNAs were tested and ultimately the sgRNA target sequence TCTGTCGTCGCCGGGTCTCT was used. Cas9 enzyme (IDT, 1081058) and 20 ng/μl sgRNA were resuspended in injection buffer (1 mM Tris-HCl pH 7.5, 0.1 mM EDTA) and injected into single cell embryos in a B6C3F1/J background (offspring of C57BL/6J females and C3H/ HeJ males, Jax Laboratory). Mice derived from injected embryos were subsequently crossed with C57BL/6 J mice, resulting in a mixed genetic background. Developmental abnormalities in mice such as tail kinks and hypopigmentation were documented after visual inspection. Kaplan-Meier survival was assessed upon mice reaching IACUC guidelines for euthanasia and organs collected, fixed in neutral buffered formalin, paraffin embedded, sections hematoxylin and eosin stained and evaluated by the FCCC Histopathology facility. T-cell lymphoma diagnoses were confirmed by immunohistochemical staining of sections with anti-CD3 antibody (Agilent, M725429-2) and DAB solution development.
Mouse genotyping, embryo analysis, MEF generation, and culture. DNA from mice were extracted with 240 μl 50 mM NaOH followed by a 10-min incubation at 95 degrees then neutralized with 60 μl 1 M Tris-HCl, pH 8. Genotyping for Brca1 CC status was accomplished by PCR amplification using primers (GGTGCACT CTCCTCCAACATC, CTCTTGCACCTGCCTCTCTGA). PCR products were subject to digestion with EcoNI and bands visualized using agarose gel electrophoresis. The Brca1 CC allele introduces an additional EcoNI cut site, as previously described 15 . Rnf168 − allele was detected using two PCR reactions with a common primer (GGGTCCTCACCGCGTAAGAAG) and one reaction that amplifies the wild-type gene (CGAAGAGACCCGGCGACGA) and another that amplifies the mutant allele (GTATGGTACCGAGTCCATAAGGGACA) and bands visualized using agarose gel electrophoresis. Pregnant mice were euthanized at 12.5 and 13.5 days post-coitum for assessment of Mendelian developmental ratios and generation of mouse embryonic fibroblast (MEF) cell lines, respectively. Embryos were separated and collected for genotyping. To generate MEF lines, individual embryos were washed with sterile Dulbecco's phosphate-buffered saline (DPBS). Embryos were aspirated 3 times through a 16G needle and incubated for 15 min in trypsin/EDTA at 37 degrees. Cells were pelleted by centrifugation for 5 min at 1000 × g, resuspended in 10 ml MEF media (Dulbecco's Modified Eagle's Medium with 4.5 g/L glucose, 15% fetal bovine serum, 1× GlutaMAX, 1× MEM nonessential amino acids, 1 mM sodium pyruvate, 1× penicillin/streptomycin). Prior to passage three, MEFs were immortalized with pBabe-SV40Tag by electroporation using the Amaxa Mouse/Rat Hepatocyte Nucleofector Kit (Lonza, VPL-1004). Immortalized MEF lines were genotyped, tested for mycoplasma and maintained in MEF media with 10% FBS.
Cell lines. HEK293T were obtained from Takara Bio, MDA-MB-436, MDA-MB-231, and MCF7 cell lines were obtained from ATCC, tested for mycoplasma (Lonza, LT07-705) and identities confirmed by short tandem repeat (STR) profiling using IDEXX analysis. Cells were grown at 37 degrees with 5% CO 2 in RPMI or DMEM media with 10% FBS and 1× penicillin/streptomycin. Fig. 4 RNF168-mediated H2A mono-ubiquitination recruits BARD1 via the BUDR. a MCF7 BARD1 −/− cells expressing GFP-BARD1-WT, GFP-BARD1-ΔBRCT, or GFP-BARD1-BUDR (R705A+D712A+Q715R) were assessed for BARD1 and BRCA1 foci size, number per nucleus, as well percentage foci positive cells post IR by immunofluorescence. See Supplementary Fig. 4b for representative images. Left; mean and S.E.M. percentage foci positive cells; middle, foci positive cells were assessed for the number of foci present in a single nucleus, red line: median value; right, foci positive cells were assessed for the mean size of foci present per nuclei, n = 3 biological replicates. Foci positive cells were determined to be those with ≥10 foci/nuclei. ***p < 0.001, **p < 0.01, *p < 0.05, ns not significant compared to WT (percentage positive cells: unpaired, two-tailed t-tests size; number and size per nucleus: nonparametric Mann-Whitney tests). See Source data for additional information. b Cells from (a) were assessed for RAD51 foci 8 h post IR by immunofluorescence. Mean and S.E M. percentage (≥5) foci positive cells, n = 3 biological replicates. **p < 0.01 compared to WT (unpaired, two-tailed ttests). See Supplementary Fig. 4b for representative images. c Cells from (a) were incubated with increasing concentrations of rucaparib for 2 weeks and colony formation assessed. Mean and S.E.M. colony formation as well as mean IC50 values, n = 3 biological replicates. d Cartoon of RNF168 protein domains and mutations analyzed. Below; expression of ectopic FLAG-RNF168 proteins by western blotting in MDA-MB-231 RNF168 −/− cells. e Cells from (d) were assessed for BARD1 and BRCA1 foci as described in (a) with statistical comparisons to cells expressing GFP. See Supplementary Fig. 4g for representative images. f Cells from (d) were assessed for RAD51 IRIF as described in (b) with statistical comparisons to cells expressing GFP. See Supplementary Fig. 4g for representative images. g MDA-MB-436 cells with sgRNA targeting GFP or 53BP1 were infected with virus expressing HA-RING, HA-PALB2, or HA-RING-PALB2 and assessed for HA (≥10), RPA32 (≥10), and RAD51 (≥5) IRIF positive cells. Mean and S.E M. percentage foci positive cells, n = 3 biological replicates. ***p < 0.001 compared to HA alone (unpaired, two-tailed t-tests). See Supplementary Fig. 5a, b for Western blots and representative images. h Cells from (g) were incubated with increasing concentrations of rucaparib for 2 weeks and colony formation assessed. Mean and S.E.M. colony formation, as well as mean IC50 values, are shown from n = 3 biological replicates. i MDA-MB-436 cells with sgRNA targeting 53BP1 expressing HA-RING-PALB2 were subject to scrambled (Sc) or RNF168 siRNA and HA and RAD51 IRIF measured. Mean and S.E M. percentage foci positive cells, n = 3 biological replicates. ***p < 0.001 and **p < 0.01 compared to Sc (unpaired, two-tailed t-tests). See Supplementary Fig. 5c for western blots and representative images.
CRISPR/Cas9 edited cell lines. MEF cell lines were transduced with lentivirus containing Cas9 and guide RNA in the lentiCRISPRv2 plasmid (Addgene, 52961) and selected using 2 μg/ml puromycin. Single guide RNA (sgRNA) target sequences were selected based on low predicted off-target score and predicted cutting efficiencies (https://portals.broadinstitute.org/gpp/public/analysis-tools/sgrna-design). Oligos (Supplementary Table 1) were phosphorylated, annealed, then cloned using BsmBI sites. The Rnf168-targeted guide utilized the sequence TCTGTCGTCGCCGGG TCTCT to generate MEF lines. Single cell clones were isolated where indicated and were considered a knockout (KO) if all alleles exhibited frameshifting mutations by Sanger sequencing using the primers described for mouse genotyping. KO was further confirmed by the absence of 53bp1 IRIF. Rap80 was targeted with ATAGTGA TATCCGATAGCGA sgRNA target sequence and KO confirmed by Sanger sequencing using the following primers; GGTAAAAGGATGCCACGAAGGA and CAAATAATATGACCACACGCGCT. MDA-MB-231 CRISPR/Cas9 edited cell lines were generated by first isolating clones with doxycycline-inducible expression of Cas9 (Addgene, 50661), then infected with lentivirus harboring the sgRNA and selected with 4 μg/ml blasticidin. Surviving cells were subjected to 5 days of 4 μg/ml doxycycline treatment for induction of Cas9 then subclones isolated and screened for KO. RNF168 was targeted with sequences TCGAAAAGGCGAGTTTATGC and ATCTGCATGGAAATCCTCG. Data is shown for a representative clone harboring sgRNA target sequence TCGAAAAGGCGAGTTTATGC. KO was confirmed by Sanger sequencing using primers; TGATACGCTTCTGGGCATAATA and ACCCCCTAACCTCTGAGAACTAT. MCF7 and MDA-MB-436 cells were transduced with lentivirus containing Cas9 and guide RNA in the lentiCRISPRv2 plasmid and selected using 2 μg/ml puromycin. The guide RNA targeting GCGACCATCC GGTTCCATGG was used to knockout BARD1 in MCF7 cells and confirmed by Sanger sequencing using the following primers; GTGCCCTGCGAGTCCCTAT and CAAAAACTACCGTTTCAGTTGGAT. The guide RNA targeting sequence CCTCAACACGGACTCGGACA was used to knockout ABRAXAS in MCF7 cells and confirmed by sequencing with primers; CAGCAGAAGCGAAGGAGGA and GAGGGCTAATGCTGGAGAAGA. The guide RNA targeting GCCATCCAG TCCTCAAGGAG was used to knockout 53BP1 in MDA-MB-436 cells. The expression of 53BP1 in pooled population was confirmed by Western blot using the nuclear extracts. . The GFP-BARD1 constructs were shuttled into a lentiviral destination vector containing a UBC promoter and a hygromycin resistance gene using the Gateway LR Clonase II Enzyme Mix (Thermo Fisher Scientific, 12538120). BRCA1-ΔRING (Δ1-127 aa, aka M128) and -ΔBRCT (Δ1496-1863 aa, aka 1495*) constructs were generated previously 20,29 . BRCA1-3A (I26A + L63A + K65A), -C61G, -ΔRING (Δ1-296 aa; M297), -ΔBRCT (Δ1559-1863 aa) and −ΔBRCT (Δ1815-1863 aa) constructs were generated using site-directed mutagenesis or are described previously 20,29 . BARD1 aa 127-777 was linked using (Gly4Ser)2 to BRCA1 aa 128-1863 to generate a RING deficient BARD1/BRCA1 conjugated protein. The RNF168 cDNA was a gift from Daniel Durocher and was cloned into the pENTR1A Gateway Entry vector with an N-terminal 3× FLAG tag. RNF168-ΔRING (aa 2-58) and -ΔPID (aa 524-571) domain deletions were generated by PCR and the R57D mutant by the QuickChange Lightning Site-Directed Mutagenesis Kit (Agilent, 210518) according to the manufacturer's instructions. RNF168 constructs were shuttled into identical lentiviral destination vectors containing either a UBC or CMV promoter and a hygromycin resistance gene using the Gateway LR Clonase II Enzyme Mix. Fulllength PALB2 cDNA was cloned into PENTR1A with an N-terminal HA tag, then shuttled into a lentiviral expression vector with a hygromycin resistance gene. BRCA1 RING domain (aa 2-104) was linked using (Gly4Ser)2 to the N-terminal of full-length PALB2 to generate the HA-RING-PALB2 construct. Lentivirus for each construct was produced in HEK293T cells (Takara Bio, 632180) using psPAX2 and VSV-G plasmids. Cell lines were infected with lentivirus using media with 10 μg/ml polybrene (Boston Bioproducts, BM-862W). Cells were then selected using the appropriate antibiotic (200 μg/ml hygromycin, 2 μg/ml puromycin, or 4 μg/ml blasticidin) or sorted for GFP expression. Construct expression was confirmed by Western blotting. PARPi colony assays. Cells were seeded in 6 well plates into media containing DMSO or 1-10000 nM of rucaparib. Cells were fixed with 4:1 methanol:acetic acid after 2 weeks, stained with crystal violet, and colonies counted. IC50 calculations were performed with GraphPad Prism software using rucaparib response curves resulting from 3 or more independent experiments. In PARPi resistance assays, cells were seeded at varying densities into DMSO or 50 nM rucaparib and data presented as rucaparib treated colonies normalized to the DMSO treated control colonies for wells where distinct colonies could be counted. Rucaparib was provided by Clovis Oncology.
Cell cycle analyses and RT-PCR. Exponentially growing cells were harvested and fixed with 50% ethanol at 4 degrees. Cells were washed with PBS and incubated in FxCycle PI/RNase staining solution (Thermo Fisher Scientific, F10797) for 30 min. Data were acquired using a BD FACScan flow cytometer and analyzed using FlowJo software. Total RNA was isolated from cell lines using the RNeasy Plus Mini Kit (Qiagen, 74134). The following primers are located within the 3× FLAG tag sequence and were used to quantify ectopic FLAG-RNF168 expression: TACAAAGACCATGACGGTGATTA and AGTCTCGCTGCCTGAGATA. Quantification of ectopic of BRCA1 FL and -ΔBRCT expression was performed using primers located on the HA tag and N terminal of the BRCA1 cDNA: CCTACGACGTGCCCGACTA and ATGGGACACTCTAAGATTTTCTGCA. Expression was normalized to 36B4 mRNA as previously described using q(quantitative) RT-PCR 20 .
Statistics and reproducibility. Statistical tests, significant p values, and number of replicates are indicated in the figure legends. Nonparametric Mann-Whitney and unpaired, two-tailed t-tests were performed using GraphPad Prism software. R was used for chi-square goodness of fit tests comparing expected and observed genotypes. Each experiment was performed with three biological replicates or as indicated in figure legends.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All relevant data are available from the authors upon reasonable request. The data generated in this study are provided in the Source data file. Guide RNA design was performed using https://portals.broadinstitute.org/gpp/public/analysis-tools/sgrnadesign. Source data are provided with this paper.