MDC1 and RNF8 function in a pathway that directs BRCA1-dependent localization of PALB2 required for homologous recombination

PALB2 is encoded by a breast cancer susceptibility (Erkko et al., 2007; Rahman et al., 2007) and Fanconi anemia (FA) (Reid et al., 2007; Xia et al., 2007) gene. We and others have demonstrated that PALB2 functionally links the products of the two major breast cancer susceptibility genes, BRCA1 and BRCA2, respectively, by direct interactions with its N- and C-termini (Sy et al., 2009a; Zhang et al., 2009a; Zhang et al., 2009b). Further, PALB2 mediates BRCA2-dependent oligomerization of the RAD51 recombinase required for strand invasion during homologous recombination (HR) and also directly interacts with RAD51 (Buisson et al., 2010; Dray et al., 2010).

BRCA1 is a large protein, including an N-terminal RING domain, a coiled-coil domain that interacts with PALB2 (amino acids 1391–1424), and two C-terminal BRCT repeats (amino acids 1646–1863) (Moynahan and Jasin, 2010). The two BRCT repeats are required for the localization of BRCA1 to sites of DNA damage (Scully et al., 1999) and mediate interactions with phosphoproteins, including CtIP, BRIP1/FANCJ, and Abraxas (Wang et al., 2007; Yu et al., 2003; Yu et al., 1998).

While BRCA1 is involved in HR (Moynahan et al., 2001), its exact function in this process and the role of the PALB2–BRCA1 interaction are not well understood. Among the possibilities, it has been suggested that the PALB2–BRCA2–RAD51 complex interacts with BRCA1 after it is independently recruited to sites of DNA damage (Sy et al., 2009a). Alternatively, the interaction with BRCA1 may localize PALB2 (Zhang et al., 2009a; Zhang et al., 2009b). A coiled-coil motif at the N-terminus of PALB2 (amino acids 9–44) mediates both homo-oligomerization of PALB2 and its hetero-oligomerization with BRCA1 (Sy et al., 2009a; Sy et al., 2009b; Zhang et al., 2009a; Zhang et al., 2009b). The specific role of these homo- and hetero-oligomers in mediating the various functions of PALB2 is unclear.

The MDC1 checkpoint protein is localized to sites of DNA damage by binding to γ-H2AX (Stewart et al., 2003; Stucki et al., 2005). RNF8, in turn, binds to MDC1, and mediates protein ubiquitination in response to DNA damage (Huen et al., 2007; Huen et al., 2010; Kolas et al., 2007; Mailand et al., 2007). RAP80 and Abraxas are also components of this signaling pathway. RAP80 is an ubiquitin-binding protein that forms a complex with Abraxas and BRCA1 (Kim et al., 2007b; Wang et al., 2007). The role of RAP80 and Abraxas in HR has remained controversial (Coleman and Greenberg, 2011; Hu et al., 2011; Wang et al., 2007), but they have a clear function in regulating the G2 DNA damage checkpoint (Kim et al., 2007a; Sobhian et al., 2007; Wang et al., 2007). As a measure of the complexity of DNA damage responses, other proteins, including CtIP and the MRE11–RAD50–NBS1 complex, are also involved in HR. CtIP and the MRE11–RAD50–NBS1 complex are important for end resection (Huen et al., 2010; Sartori et al., 2007).

Given controversy about the importance of the PALB2–BRCA1 interaction, and to better understand the regulation of PALB2 function, and of DNA repair by HR, we generated a PALB2 fusion protein containing the BRCT repeats of BRCA1. Subsequent mutation of the PALB2 coiled-coil domain yielded an additional fusion protein, BRCT–PALB2(L21P), which is incapable of binding to BRCA1 but which has the potential to bypass BRCA1 and localize to sites of DNA damage via BRCA1-derived BRCT repeats. The results have yielded important insights into the regulation and function of the products of the BRCA1 and PALB2 tumor suppressor genes in cellular responses to DNA damage. We clearly demonstrate that BRCA1 mediates HR, in part, by localizing PALB2. Further, we find that the BRCA1–PALB2 hetero-oligomer, and not the PALB2–PALB2 homo-oligomer, is essential for PALB2-dependent DNA repair. Additionally, our results suggest that BRCA1-dependent localization of PALB2 may link DNA repair by HR into a pathway with the MDC1–RNF8–RAP80–Abraxas signaling cascade. Thus, in this manner, PALB2-dependent HR may be responsive to signals generated at DSBs.

To resolve whether BRCA1 is required to localize PALB2 we fused the BRCT repeats of BRCA1 to PALB2, or to the PALB2(L21P) mutant that cannot bind to BRCA1 (Zhang et al., 2009a). The localization of BRCA1 to sites of DNA damage is mediated by this BRCT domain (Scully et al., 1999). Diagrams of the fusion proteins containing PALB2 or PALB2(L21P), and the BRCT repeats of BRCA1, are shown in Fig. 1A. We also generated a control fusion with PALB2 (BRCTΔC–PALB2), based upon the truncated and non-functional BRCT domain present in HCC1937 breast cancer cells (Scully et al., 1999).

HCC1937 cells, in which BRCA1 does not localize to foci normally (Scully et al., 1999), were stably transduced with various constructs containing PALB2 or with BRCA1 itself. BRCT–PALB2, BRCT–PALB2(L21P) and BRCTΔC–PALB2 were expressed at similar levels (Fig. 1B). Further, the fusion proteins were expressed at higher levels than endogenous PALB2. Transduction with BRCA1 by itself increased the levels of total BRCA1 detected on immunoblots (Fig. 1B), since the truncated, endogenous BRCA1 protein that is present in HCC1937 cells shows the same mobility.

Examples of foci assembled in each case following exposure to irradiation (IR), and detected with anti-PALB2 antibodies, are shown in Fig. 1C. Strikingly, in cells reconstituted with either the BRCT–PALB2 or BRCT–PALB2(L21P) fusion proteins or with BRCA1, many of the PALB2 foci showed colocalization with γ-H2AX, a marker for DSBs. Foci assembled by γ-H2AX in response to IR were detected in each of the cell lines, regardless of whether PALB2 foci were detected.

Quantification of the assembly of foci, detected with anti-PALB2 antibodies, in each of the reconstituted HCC1937 cells is shown in untreated populations or following exposure to IR (Fig. 1D). Both BRCT–PALB2 and BRCT–PALB2(L21P), and exogenously expressed BRCA1, supported the assembly of PALB2 foci but the fusion protein with truncated BRCT repeats (BRCTΔC–PALB2) did not. Since HCC1937 cells reconstituted with vector alone did not display PALB2 foci (Fig. 1D), the PALB2 foci observed here after transduction with BRCT–PALB2 or BRCT–PALB2(L21P) likely correspond to exogenously expressed protein rather than endogenous PALB2. Together, our results suggest that BRCA1-derived BRCT repeats can correctly address PALB2 to sites of DNA damage in the absence of intact BRCA1 that is itself localized.

Since PALB2 is required for the assembly of the RAD51 recombinase into foci (Xia et al., 2007; Zhang et al., 2009a), we also assayed the formation of RAD51 foci as a measure of the function of the various fusion proteins (Fig. 1E). The assembly of RAD51 foci paralleled the results for PALB2 foci, demonstrating that fusion with the BRCT repeats is sufficient for targeting a functional PALB2 to DNA damage foci. This is even true for the BRCT–PALB2(L21P) mutant fusion protein that cannot bind to BRCA1. It should be noted that while there is a minimal level of PALB2 foci in HCC1937 cells prior to transduction with the fusion proteins, RAD51 foci display an intermediate level. This may be due to BRCA1-independent mechanisms which upregulate the levels of RAD51 protein and foci, and which may thereby compensate for the deficiency in BRCA1 (Martin et al., 2007).

To further support the conclusion that fusion of PALB2 to the BRCT repeats of BRCA1 can mediate recruitment of functional PALB2 in the absence of BRCA1, we expressed the fusion proteins in U2OS cells depleted of endogenous BRCA1. Unlike HCC1937 cells, U2OS cells express full-length BRCA1. Consistent with results obtained in HCC1937 cells, both the wild-type (WT) and L21P mutant fusion proteins, and siRNA-resistant BRCA1, supported the assembly of foci by endogenous PALB2 or RAD51 (supplementary material Fig. S1). PALB2 alone or PALB2 fused to the truncated BRCT domain, however, did not support the assembly of PALB2 or RAD51 foci in U2OS cells depleted of full-length BRCA1. In fact, expression of PALB2 or BRCTΔC–PALB2 in cells depleted of BRCA1 was associated with lower levels of RAD51 foci than in BRCA1-depleted cells which contained the empty vector alone (supplementary material Fig. S1C). BRCTΔC–PALB2 did not similarly affect RAD51 foci in a different BRCA1-deficient background found in HCC1937 cells, however (Fig. 1C).

We next utilized EUFA1341 FA-patient derived fibroblasts, which contain truncated PALB2 (Y551X) (Xia et al., 2007), to address the question of fusion protein localization in cells with an inherited deficiency for PALB2 function. BRCA1 is intact in EUFA1341 cells. The different forms of PALB2 were expressed at similar levels (Fig. 2A). Examples from cells exposed to IR demonstrate that the BRCT–PALB2(L21P) fusion protein, detected with anti-PALB2 antibodies, assembled into foci that colocalized with γ-H2AX foci as well or better than normal PALB2 that was exogenously expressed in EUFA1341 cells (Fig. 2B). Since the anti-PALB2 antibodies which we utilized do not detect the PALB2–Y551X mutant in EUFA1341 cells (Zhang et al., 2009a), the foci which we observed resulted from the exogenously expressed proteins.

Quantification demonstrates that the BRCT–PALB2 and BRCT–PALB2(L21P) fusion proteins, along with exogenously expressed PALB2, assembled into foci, detected with anti-PALB2 antibodies, in untreated EUFA1341 cells (Fig. 2C). Further, assembly of foci was increased following exposure to IR (Fig. 2C). BRCT–PALB2 displayed increased assembly into foci, as compared to BRCT–PALB2(L21P) or PALB2, apparently because it can localize by two pathways: (1) via the BRCT repeats or (2) by binding to BRCA1. In contrast PALB2(L21P) did not assemble into foci (Fig. 2C). Similarly, BRCT–PALB2, BRCT–PALB2(L21P) and PALB2, but not PALB2(L21P), supported the assembly of RAD51 foci both in untreated populations and following exposure to IR (Fig. 2D). Together, these results suggest that the mutant fusion protein can bypass BRCA1, and that localization of PALB2 is both necessary and sufficient for the recruitment of RAD51 foci to sites of DNA damage.

To further test the specific requirement for the BRCT repeats of BRCA1 in localizing functional PALB2, as a control, we fused PALB2–L21P to the BRCT repeats of the BRCA1 partner, BARD1 (diagrammed in supplementary material Fig. S2). The BRCT repeats of BARD1 are similar to those in BRCA1 but are not exactly conserved. Foci formation by BRCT[BARD1]–PALB2(L21P) was greatly diminished as compared to BRCT[BRCA1]–PALB2(L21P) (Fig. 2E). We refer to BRCT[BRCA1]–PALB2(L21P) as BRCT–PALB2(L21P) elsewhere. While these results suggest that the BRCA1 BRCT repeats have a specific role in recruiting PALB2, there was a small degree of foci assembled by BRCT[BARD1]–PALB2(L21P). This may reflect the fact that BARD1 also forms DNA damage foci (Brodie and Henderson, 2010) and, thus, its BRCT domain may have some affinity for proteins at sites of DNA damage.

Given that PALB2 is required for resistance to MMC and for DSB-initiated HR (Xia et al., 2006), as a measure of the functional importance of BRCA1-dependent localization of PALB2, we assayed these processes in PALB2-deficient cells reconstituted with BRCT–PALB2(L21P). Again, PALB2(L21P) does not bind to BRCA1, so any function of the mutant fusion protein can be attributed to the BRCT repeats of BRCA1. BRCT–PALB2(L21P), as well as BRCT–PALB2 or PALB2 alone, restored resistance of EUFA1341 (PALB2-deficient cells) to MMC (Fig. 3A). In contrast, cells reconstituted with PALB2(L21P) displayed the same sensitivity to MMC as cells containing only the empty vector. Thus, localization of PALB2 to sites of DNA damage, in this case by fusion to the BRCT repeats of BRCA1, is linked with the cellular response to DNA interstrand crosslinking agents.

U2OS-DR cells containing the DR-GFP HR reporter, and various forms of exogenously expressed PALB2, were depleted of endogenous PALB2. Cells reconstituted with PALB2(L21P) or BRCTΔC–PALB2(L21P) were deficient for DSB-initiated HR, while BRCT–PALB2(L21P) restored recombination to the same levels as PALB2 itself (Fig. 3B). Taken together, results in Figs 1–f02,3 suggest that BRCA1-mediated localization of PALB2 is necessary for HR. Perhaps related to its increased assembly into nuclear foci, BRCT–PALB2 mediated higher levels of HR as compared to PALB2 or BRCT–PALB2(L21P) (Fig. 2C).

Neither BRCT–PALB2(L21P), nor BRCT–PALB2, was able to correct the HR deficiency of U2OS-DR cells depleted of BRCA1, however (Fig. 3C). Thus, while the interaction of PALB2 with BRCA1 is necessary for HR (Sy et al., 2009a; Zhang et al., 2009a; Zhang et al., 2009b), it is not by itself sufficient to mediate DNA repair by HR.

It was previously demonstrated that PALB2 can form a homo-oligomer. Deletion of the N-terminal 42 amino acids of PALB2, which contains most of its coiled-coil domain, abrogated the PALB2 oligomer and disrupted various PALB2-dependent functions (Sy et al., 2009b). Whether this mutant was also deficient for interactions with BRCA1 was not examined, however. Given that fusion to the BRCT repeats of BRCA1 can restore the function of a mutant of the PALB2 coiled-coil domain [BRCT–PALB2(L21P)], we sought to determine the effect of the L21P mutation on both PALB2–PALB2 and BRCA1–PALB2 interactions. For this purpose, we transiently expressed Flag-tagged WT or PALB2(L21P) in 293T cells along with either WT HA-PALB2 or MYC-tagged BRCA1 (Fig. 4). WT PALB2 interacted with PALB2 and BRCA1, as expected, but PALB2(L21P) was deficient for interactions with both proteins. Thus, BRCT–PALB2(L21P) restores PALB2 function through fusion to the BRCT repeats of BRCA1, without the capacity for PALB2–PALB2 or BRCA1–PALB2 interactions. Together, our results (Figs 1–Fig. 2,Fig. 3,4) suggest that the BRCA1–PALB2 interaction alone, and not the PALB2–PALB2 interaction, is sufficient to support the assembly of RAD51 foci, and for DNA repair by HR and for resistance to MMC.

The BRCT domain of BRCA1 was capable of accurately localizing PALB2(L21P) when fused to it (Fig. 1C; Fig. 2B). This suggested a potential role in localizing the mutant fusion protein [BRCT–PALB2(L21P)] for one of the three DNA damage response proteins that are known to bind to the BRCT domain of BRCA1: CtIP, FANCJ, and Abraxas (Wang et al., 2007; Yu et al., 2003; Yu et al., 1998). As a first test of this possibility, we immunoprecipitated epitope-tagged versions of PALB2 or BRCT–PALB2(L21P) that were stably expressed in U2OS cells (Fig. 5A). Indeed, CtIP, FANCJ and Abraxas, each co-immunoprecipitated with BRCT–PALB2(L21P). The specificity of these interactions for the BRCT repeats present in the fusion protein is demonstrated by the fact that none of these proteins showed strong co-immunoprecipitation with PALB2 alone or with the vector control. In contrast, BRCA2 co-immunoprecipitated both with PALB2 and BRCT–PALB2(L21P), which is consistent with the fact that its N-terminus binds to PALB2 directly (Oliver et al., 2009).

To determine whether the interaction of the BRCT repeats with CtIP, FANCJ, and/or Abraxas is involved in the recruitment of BRCT–PALB2(L21P) to nuclear foci, we depleted each protein from U2OS-DR cells (Fig. 5B). Foci formed by the fusion protein containing PALB2(L21P) were not influenced by depletion of CtIP or FANCJ (Fig. 5C). Depletion of Abraxas, however, significantly decreased assembly of the mutant fusion protein into foci following exposure to IR (Fig. 5C).

Importantly, depletion of Abraxas, but not depletion of CtIP or FANCJ, also significantly decreased the assembly of foci by endogenous PALB2 in U2OS cells that lacked the fusion protein construct (Fig. 5D). Effects of depletion of Abraxas on foci formation by the mutant fusion protein were more dramatic than the effects on endogenous PALB2, presumably because the BRCT repeats make the fusion protein more dependent upon Abraxas.

To further elucidate the upstream pathway that regulates recruitment of PALB2 to sites of DNA damage, we depleted NBS1, which is part of the MRE11–RAD50–NBS1 complex required for end resection (Sartori et al., 2007) or RAP80 (Fig. 6). As shown by representative images, depletion of RAP80 decreased the assembly of endogenous PALB2, but not γ-H2AX, into foci following exposure to IR (Fig. 6A). RAP80 is an ubiquitin-binding protein that forms a complex with Abraxas (Kim et al., 2007b; Wang et al., 2007). Consistent with results obtained for depletion of Abraxas in Fig. 5D, quantification demonstrates that depletion of RAP80, but not NBS1, decreased the assembly of foci by endogenous PALB2 in response to IR (Fig. 6B). As further support for the role of RAP80 in regulating the recruitment of PALB2 following exposure to IR, we expressed a siRNA-resistant form of RAP80 in U2OS cells (Fig. 6C). Assembly of PALB2 foci in cells depleted of endogenous RAP80 and expressing the siRNA-resistant form of the protein was recovered to the same levels as cells transfected with the control siRNA (Fig. 6D).

The conclusion that RAP80 regulates recruitment of PALB2 in response to DSBs is also supported by our observation that depletion of RAP80 decreased the percentage of cells that were positive for PALB2 foci at each time point tested, ranging from 1–16 h after exposure to IR (Fig. 7A). It should also be noted that RAP80 appears to regulate basal levels of PALB2 foci, since PALB2 foci were also decreased in untreated populations of U2OS cells depleted of RAP80 (Fig. 7A). Depletion of RAP80 decreased the percentage of cells with PALB2 (Fig. 7A) or BRCA1 (Fig. 7B) foci to similar degrees, consistent with the possibility that RAP80 regulates BRCA1-dependent recruitment of PALB2 to sites of DNA damage.

Since MDC1 and RNF8 are known to function upstream of RAP80, their role in the recruitment of PALB2 in response to DNA damage was also tested. MDC1 is a checkpoint mediator that binds to γ-H2AX at DSBs, while RNF8 is an E3 ubiquitin ligase that binds to MDC1 and ubiquitinates histones and other proteins (Huen et al., 2007; Kolas et al., 2007; Mailand et al., 2007). Representative images demonstrate that depletion of MDC1 or RNF8 compromised PALB2 recruitment to DSBs following exposure to IR, as determined by decreased colocalization with γ-H2AX foci (Fig. 6A). Further, quantification demonstrates that depletion of either MDC1 or RNF8 dramatically decreased the percentage of cells that displayed assembly of endogenous PALB2 into foci after exposure to IR (Fig. 6B). Given the roles of MDC1 and RNF8, and of RAP80, in ubiquitinating histone and binding to it, respectively, these results suggest that histone ubiquitination regulates PALB2 recruitment. In further support of this possibility, we find that PALB2 strongly co-localized with ubiquitin after exposure of cells to IR (Fig. 6E). It has been demonstrated previously that ubiquitin accumulates at DSBs after treatment with IR (Sobhian et al., 2007).

It should be noted that depletion of MDC1, RNF8 or RAP80 also strongly inhibited the assembly of foci by the BRCT–PALB2(L21P) fusion protein (supplementary material Fig. S3). Since deficient foci formation by PALB2(L21P) can be rescued through fusion to the BRCT repeats of BRCA1 (Fig. 1D; Fig. 2C; supplementary material Fig. S1B), this result suggests that MDC1, RNF8 and RAP80 regulate PALB2 through their actions on BRCA1. Because PALB2 recruits BRCA2 and RAD51 to sites of DNA damage, our results suggest that PALB2 may integrate the MDC1–RNF8–RAP80–Abraxas network, which is initiated after detection of DSBs, with HR.

Previous reports have demonstrated that BRCA1 and PALB2 interact through a coiled-coil domain on each protein and that this interaction is required for DNA repair by HR (Sy et al., 2009a; Zhang et al., 2009a; Zhang et al., 2009b). How this interaction mediates HR and resistance to MMC has remained controversial, however. There have also been discrepant reports as to whether BRCA1 recruits PALB2 to sites of DNA damage (Sy et al., 2009a; Zhang et al., 2009a; Zhang et al., 2009b). To address these issues, we have utilized a fusion protein that contains PALB2(L21P), which is incapable of binding to BRCA1, and the BRCT repeats that localize BRCA1. We find that this fusion protein is recruited to sites of DNA damage, both in BRCA1- and PALB2-deficient cells, and can therefore bypass the normal requirement for BRCA1 and its capacity to bind PALB2. The fusion protein localizes correctly to DSBs following exposure to IR, as determined by partial colocalization with γ-H2AX foci. Importantly, BRCT–PALB2(L21P) mediates the formation of RAD51 foci, both in BRCA1- and PALB2-deficient cells. As an important control, we find that fusion of PALB2(L21P) to incomplete BRCT repeats which are found in a breast cancer cell line and which are incapable of localizing BRCA1 (Scully et al., 1999), does not support the assembly of DNA damage foci by the fusion protein or by RAD51. On the basis of these results, we conclude that PALB2 is localized by its interaction with BRCA1.

Furthermore, we find that the BRCT–PALB2(L21P) mutant fusion protein supports HR and resistance to MMC in a PALB2-deficient background. Thus, our results suggest that correct, BRCA1-dependent localization of PALB2 and the machinery for HR are necessary for DNA repair by HR. In this context, it is noteworthy that we also find that the L21P mutant of PALB2 is defective for interactions either with other PALB2 molecules or with BRCA1. Together, these observations make the novel point that the BRCA1–PALB2 heterodimer, and not the PALB2–PALB2 homodimer, is required to mediate PALB2 functions such as assembly of RAD51 foci, HR, and resistance to MMC.

The BRCT–PALB2(L21P) mutant fusion protein has also been an important tool for understanding how recruitment of PALB2 is linked to upstream DNA damage signaling processes. We find that both BRCT–PALB2(L21P) and endogenous PALB2 are recruited by a network, including MDC1, RNF8, RAP80 and Abraxas, that generates and responds to ubiquitin signals at DSBs. This pathway is initiated following the generation and detection of DSBs. Thus, PALB2 is part of a pathway that may link DNA repair by the core machinery for HR to various signals generated at the DSB. The potential importance of this pathway is underscored by the fact that RAP80 (Akbari et al., 2009; Nikkilä et al., 2009), in addition to BRCA1, PALB2 and BRCA2, is linked to an inherited susceptibility to breast cancer.

Together, the results on foci formation by cells reconstituted with BRCT–PALB2(L21P) suggest that BRCA1 has an important role in localizing PALB2, either spontaneously or in response to DNA damage induced by IR. Further, combined with previous results (Xia et al., 2006; Zhang et al., 2009a), it appears that PALB2 recruits BRCA2 and RAD51 after it is localized by BRCA1. We suggest that recruitment of PALB2 and RAD51 requires two domains on BRCA1: (1) the BRCT domain that initially localizes it and (2) the coiled-coil domain that binds to PALB2. It should be noted that breast cancer-associated mutations of BRCA1 are observed in both domains of the protein (Gayther et al., 1996). Further, the fact that BRCA1 mediates PALB2 localization and function in HR may explain the shared association of mutations of BRCA1 or PALB2 with breast cancer.

While BRCA1-dependent localization of PALB2 is required for HR (Fig. 3B), the BRCT–PALB2(L21P) fusion protein does not support DSB-initiated HR in BRCA1-deficient cells (Fig. 3C). We therefore speculate that another domain of BRCA1, in addition to the BRCT repeats and the coiled-coil domain, is also required for HR. One possibility is the N-terminal interaction of BRCA1 with the MRE11–RAD50–NBS1 complex (Chen et al., 2008). E3 ligase activity at the N-terminus of BRCA1 is not required for HR, however (Reid et al., 2008; Shakya et al., 2011).

We propose that the ubiquitin-related signaling cascade initiated at DSBs by binding of MDC1 to γ-H2AX (Stewart et al., 2003; Stucki et al., 2005) is linked to HR by mediating BRCA1-dependent recruitment of PALB2. It appears that RNF8 is subsequently recruited by binding to MDC1 and generates ubiquitin-dependent binding sites for RAP80. In fact, RNF8 mediates histone ubiquitination at DSBs (Huen et al., 2007; Kolas et al., 2007; Mailand et al., 2007) and RAP80 binds to ubiquitinated histone (Mailand et al., 2007; Wu et al., 2009). It should be noted, in this context, that deficiency for MDC1 or RNF8 (Huang et al., 2009; Zhang et al., 2005), like deficiency for PALB2 (Xia et al., 2006), compromises DSB-initiated HR.

Importantly, Abraxas is recruited to DSBs via its interaction with RAP80 and binds to BRCA1 (Kim et al., 2007b; Wang et al., 2007). Thus, RAP80 and Abraxas potentially couple BRCA1-dependent recruitment of PALB2 to the upstream ubiquitin-dependent signaling cascade. Consistent with this possibility, and in agreement with others (Hu et al., 2011; Kim et al., 2007a; Sobhian et al., 2007; Wang et al., 2007), we find that BRCA1 foci were reduced by depletion of RAP80 (Fig. 7). Further, PALB2 foci were reduced to a similar degree as BRCA1 foci (Fig. 7). Additionally, depletion of Abraxas resulted in decreased assembly of PALB2 foci (Fig. 5D). Further, depletion of Abraxas also decreases the assembly of BRCA1 foci in response to IR (Kim et al., 2007b). Considering all of this together, we propose that a primary function of RAP80 and Abraxas is to promote the BRCA1-dependent recruitment of PALB2.

Depletion of RAP80 appears to have opposite effects on the recruitment of PALB2 (Fig. 6A) as compared to the recruitment of two proteins that bind to the BRCT repeats of BRCA1: CtIP and FANCJ (Hu et al., 2011). PALB2 foci were not tested in this previous study. We find that depletion of RAP80 in the same cell line and at the same time points they utilized resulted in decreased assembly of PALB2 foci after exposure to IR (Fig. 7). It should be noted that Abraxas, CtIP and FANCJ bind to the BRCT domain of BRCA1. Because PALB2 binds through a different domain of BRCA1, depletion of RAP80 may affect PALB2 differently. In accord with this possibility, Hu et al. (Hu et al., 2011) reported that depletion of RAP80 did not decrease levels of RAD51 foci (Hu et al., 2011). Like PALB2, RAD51 does not bind to the BRCT repeats of BRCA1. Interestingly, since BRCA1 is a component of both the CtIP–MRN (Chen et al., 2008) and PALB2–BRCA2–RAD51 (Sy et al., 2009a; Zhang et al., 2009a; Zhang et al., 2009b) complexes, respectively, RAP80-dependent recruitment of BRCA1 could coordinate the end resection and strand invasion activities that are together required for HR.

EUFA1341, HCC1937 and U2OS cell lines were cultured at 37°C in a 5% CO₂ environment and irradiated as described previously (Zhang et al., 2009a).

N-terminally tagged (HA-Flag) proteins were generated in pOZ as described previously and subcloned into pMMP, where appropriate (Zhang et al., 2009a). PALB2 was fused to the BRCT domain of BRCA1 [amino acids 1629–1863] (BRCT–PALB2) or a truncated BRCT domain [amino acids 1629–1794] (BRCTΔC–PALB2). BRCT–PALB2(L21P) was then generated by subsequent site-directed mutagenesis according to our published procedures (Zhang et al., 2010). As a control, we introduced the BRCT domain of BARD1 [amino acids 550–777] (supplementary material Fig. S2) in front of PALB2(L21P).

Retroviral transduction and selection with interleukin-2 beads (pOZ) or with puromycin (pMMP) were as described previously (Zhang et al., 2009a; Zhang et al., 2010). Unless otherwise noted PALB2 and its fusions, and BRCA1, were transduced using pOZ and pMMP retroviruses, respectively. No differences in foci formation or HR were noted in cells transduced into cells using the empty pOZ or pMMP vectors.

Cells were stably transduced with a retrovirus for siRNA-resistant BRCA1 [UCACAGUGUCCUUUAUGUA (Ganesan et al., 2002)] or siRNA-resistant RAP80 [GUAUUGACUCGGAGACAAA (Hu et al., 2011)] that were generated by site-directed mutagenesis. Cells were transiently depleted of endogenous PALB2, or BRCA1 or RAP80, using siRNAs against the 3′-UTR or the coding sequence, respectively, as described (Ganesan et al., 2002; Hu et al., 2011; Zhang et al., 2009a). For other depletion experiments, siRNAs (5′–3′) directed against FANCJ (GUACAGUACCCCACCUUAU) (Zhang et al., 2010), CtIP (GCUAAAACAGGAACGAAUC) (Yu and Chen, 2004), Abraxas (GUAAAAGGUGAAGCCAAGA) (Liu et al., 2007), RNF8 (GGACAAUUAUGGACAACAA) (Mailand et al., 2007), MDC1 (UCCAGUGAAUCCUUGAGGU) (Lou et al., 2003), and NBS1 (GGAGGAAGAUGUCAAUGUU) (Yoo et al., 2009) were utilized as previously described. All siRNAs were purchased from Dharmacon.

The following antibodies were utilized: mouse anti-HA (HA.11; Covance), mouse anti-Flag (M2,Sigma), rabbit anti-PALB2 (Zhang et al., 2009a), rabbit anti-BRCA1 (Millipore), goat anti-CtIP (Santa Cruz Biotechnology), rabbit anti-FANCJ (Zhang et al., 2010), rabbit anti-Abraxas (Bethyl Laboratory), rabbit anti-RAP80 (Bethyl Laboratory), mouse anti-γ-H2AX (Millipore), and mouse anti-ubiquitin conjugates [FK2; Millipore; (Huen et al., 2007)].

Cells were prepared and imaged as previously described (Zhang et al., 2009a). For each data point, three independent counts of at least 150 cells each were made. Cells that had three or more nuclear foci were scored as positive.

Labeled cells were observed with a Leica DMI6000 microscope, and images were collected with a Hamamatsu Camera using Openlab software (Improvision). Images were processed into figures using Photoshop (Adobe).

Immunoblotting was as previously described (Zhang et al., 2009a).

293T cells were transiently transfected with Flag-tagged PALB2 or PALB2(L21P), and HA-PALB2 or MYC-BRCA1, extracts prepared, and immunoprecipitations with anti-Flag antibodies performed as described previously (Zhang et al., 2009a).

Survival was calculated in triplicate, relative to the average corrected absorbance for each untreated cell line, as described (Zhang et al., 2009a).

U2OS-DR cells containing an integrated reporter for HR that stably expressed various constructs described above were depleted of endogenous PALB2 or BRCA1, transfected with pCBASce encoding the I-SCEI endonuclease, and assayed as described previously (Zhang et al., 2009a). The ratio of recombination with and without depletion of PALB2 or BRCA1, as appropriate, was calculated for each cell line and condition.

We are grateful to Maria Jasin and Koji Nakanishi (Memorial Sloan-Kettering Cancer Center) for U2OS-DR cells and pCBASce. We thank Niall Howlett (University of Rhode Island) and Satoshi Namekawa (CCHMC) for comments on the manuscript.