A PALB2-interacting domain in RNF168 couples homologous recombination to DNA break-induced chromatin ubiquitylation

DNA double-strand breaks (DSB) elicit a ubiquitylation cascade that controls DNA repair pathway choice. This cascade involves the ubiquitylation of histone H2A by the RNF168 ligase and the subsequent recruitment of RIF1, which suppresses homologous recombination (HR) in G1 cells. The RIF1-dependent suppression is relieved in S/G2 cells, allowing PALB2-driven HR to occur. With the inhibitory impact of RIF1 relieved, it remains unclear how RNF168-induced ubiquitylation influences HR. Here, we uncover that RNF168 links the HR machinery to H2A ubiquitylation in S/G2 cells. We show that PALB2 indirectly recognizes histone ubiquitylation by physically associating with ubiquitin-bound RNF168. This direct interaction is mediated by the newly identified PALB2-interacting domain (PID) in RNF168 and the WD40 domain in PALB2, and drives DNA repair by facilitating the assembly of PALB2-containing HR complexes at DSBs. Our findings demonstrate that RNF168 couples PALB2-dependent HR to H2A ubiquitylation to promote DNA repair and preserve genome integrity. DOI: http://dx.doi.org/10.7554/eLife.20922.001


Introduction
Chromosomal DNA double-strand breaks (DSB) that arise in S phase during DNA replication or early in G2 phase are removed by the homologous recombination (HR) machinery, which utilizes the intact genetic information from the sister chromatid as a template for repair. Following their detection, the ends of a DSB are resected to generate extended 3' single-stranded DNA (ssDNA) overhangs, which are bound by the ssDNA-binding protein RPA. One of the key players in HR is the partner and localizer of breast cancer susceptibility protein 2 (PALB2), which is mutated in a subset of breast cancer and Fanconi Anaemia patients (FANCN) (Xia et al., 2006;Reid et al., 2007;Antoniou et al., 2014). PALB2 forms oligomers to interact with the breast cancer susceptibility proteins 1 and 2 (BRCA1 and BRCA2) and the recombinase RAD51 (Sy et al., 2009;Zhang et al., 2009aZhang et al., , 2009bBuisson and Masson, 2012). The interaction between BRCA1 and PALB2 is tightly regulated in a cell-cycle dependent manner to ensure suppression of HR in G1  and activation of HR in S/G2 (Buisson et al., 2017). Once recruited to resected DSB ends by BRCA1, PALB2 facilitates the assembly of BRCA2 and RAD51 onto broken DNA ends, the latter of which catalyzes strand invasion and DNA transfer during HR (Sy et al., 2009;Zhang et al., 2009aZhang et al., , 2009b. This mechanism is responsible for the error-free repair of DSBs in S/G2 phase. The alternative pathway for DSB repair is non-homologous end-joining (NHEJ), which is dominant in G1 and late G2 phase (Karanam et al., 2012). The choice between HR and NHEJ during the cell cycle is regulated by a signaling pathway that involves the progressive modification of chromatin surrounding DSBs (Polo and Jackson, 2011). This pathway involves the concerted action of the E3 ubiquitin ligases RNF8 and RNF168, which promote the ubiquitin-dependent recruitment of the BRCA1-Abraxas-RAP80-MERIT40 (BRCA1-A) and 53BP1 complexes (Mailand et al., 2007;Escribano-Díaz et al., 2013;Stewart et al., 2009;Doil et al., 2009;Huen et al., 2007;Wang and Elledge, 2007;Kolas et al., 2007). It has been demonstrated that the RNF168-dependent recruitment of 53BP1, through its effectors RIF1 and MAD2L2, inhibits DNA end-resection and subsequently HR Boersma et al., 2015). However, the action of RNF168 and 53BP1 does not seem to be intrinsically inhibitory to HR. For instance, the recruitment of RIF1 and its inhibitory impact on HR are restricted to G1 cells, which favors repair by NHEJ in that cell cycle stage .
The accumulation of RIF1 is antagonized by BRCA1 in S/G2 phase (Chapman et al., 2013;Escribano-Díaz et al., 2013;Feng et al., 2013), alleviating the inhibitory impact on HR and allowing this repair pathway to reach maximal activation during these stages of the cell cycle (Karanam et al., 2012). With the inhibitory impact of RIF1 relieved, RNF168 is still capable of ubiquitylating chromatin in S/G2 phase (Doil et al., 2009), raising the question whether RNF168-induced ubiquitylation affects HR during these cell cycle stages. In this study, we reveal an unexpected role of RNF168 in physically coupling the HR machinery to H2A ubiquitylation in S/G2 cells through a newly identified PALB2-interacting domain (PID). These findings delineate a new RNF168-dependent mechanism that is paramount to PALB2 recruitment and efficient execution of HR in S/G2 phase.

RNF168 promotes PALB2 recruitment to chromatin
To explore potential links between RNF168 and the homologous recombination (HR) machinery, we used a chromatin-tethering approach, exploiting the strong interaction between the LacR protein and the LacO DNA sequence (Acs et al., 2011;Luijsterburg et al., 2012). Human U2OS cells with stably incorporated LacO arrays (Janicki et al., 2004) were transfected with LacR-tagged RNF168 ( Figure 1A), which triggered the accumulation of ubiquitin conjugates (FK2) and the ubiquitin-binding protein 53BP1 (Figure 1-figure supplement 1A,B) at the LacO array as previously reported (Acs et al., 2011;Luijsterburg et al., 2012;Panier et al., 2012). Surprisingly, we found that tethering of LacR-RNF168 also attracted PALB2 and RAD51 to the array, while LacR alone failed to do so ( Figure 1B-E).
The RNF8 pathway has previously been implicated in the recruitment of PALB2 by promoting the accrual of the BRCA1-Abraxas-RAP80-MERIT40 (BRCA1-A) complex to DSBs (Zhang et al., 2012). Conversely, it was demonstrated in several other studies that RAP80 and Abraxas inhibit PALB2dependent HR (Coleman and Greenberg, 2011;Hu et al., 2011;Kakarougkas et al., 2013). We recently reported that PALB2 and RAP80 reside in biochemically distinct and mutually exclusive BRCA1-containing complexes (Typas et al., 2015). In line with this observation, we found that RAP80 depletion did not affect PALB2 recruitment after RNF168 tethering ( Figure 1F,G). Importantly, immunofluorescence and western blot analysis confirmed the efficient knock-down of RAP80 under these conditions ( Figure 1-figure supplement 1C,D). Additionally, knock-down of BRCA1, which was confirmed by immunofluorescence analysis ( Figure 1H and Figure 1-figure supplement 1E), also did not change the RNF168-dependent assembly of PALB2 in this experimental system ( Figure 1H,I), suggesting that RNF168 can promote PALB2 accrual independently or downstream of BRCA1.
It should be noted that while RNF168-mediated recruitment of PALB2 in the tethering system does not require BRCA1 ( Figure 1H,I), its recruitment into ionizing radiation-induced foci (IRIF) that contain DSBs is fully dependent on BRCA1 (Figure 2A), which is in agreement with earlier reports (Zhang et al., 2009a(Zhang et al., , 2009b. We conclude that forced chromatin binding of RNF168 triggers PALB2 and RAD51 loading onto chromatin in a manner that does not require the RAP80-containing BRCA1-A complex.

RNF168 promotes PALB2 recruitment to DSBs and execution of HR
To examine the physiological relevance of our findings in the tethering system, we decided to examine the impact of RNF8 and RNF168 on HR. It should be noted that a number of studies have previously reported a role for RNF8 in regulating HR through different mechanisms (Huang et al., 2009;Lu et al., 2012;Zhang et al., 2012), while other studies have reported that loss of RNF8 or RNF168 did not affect HR (Meerang et al., 2011;Sy et al., 2011;Muñoz et al., 2012;Nakada et al., 2012). Thus, the function of RNF8 and particularly RNF168 in HR is currently unclear. To elucidate the role of these proteins in this repair process, we depleted cells of RNF8 or RNF168 and monitored the recruitment of HR factors into IRIF. Cells stably expressing the mAG-tagged cell cycle marker geminin were engineered to specifically monitor recruitment in S and G2 cells (Typas et al., 2015). As expected, mAG-geminin-expressing cells showed recruitment of endogenous PALB2 into foci after exposure to ionizing radiation (IR) (Figure 2A). Importantly, a quantitative analysis of IRIF formation revealed that knock-down of RNF8 or RNF168 largely abolished the accrual of PALB2 in mAG-geminin-positive cells (Figure 2A,D). The knock-down of either ubiquitin ligase was verified by immunoblotting using specific antibodies ( Figure 2C). Importantly, while BRCA1 depletion completely abolished PALB2 recruitment (Zhang et al., 2009a(Zhang et al., , 2009b, loss of RAP80 had no appreciable impact (Figure 2A,D and Figure 2-figure supplement 1A,B). These results suggest that RNF8/RNF168 contribute to the BRCA1-dependent loading of PALB2 at DSBs independently of RAP80, which is in accordance with our tethering results ( Figure 1F,G), and consistent with several studies showing that RAP80 does not contribute to, or may even antagonize HR (Coleman and Greenberg, 2011;Hu et al., 2011;Typas et al., 2015). Importantly, depletion of RNF8 or RNF168 also impaired efficient RAD51 assembly in cycling cells, following either IR-inflicted DNA damage ( Figure 2B,E) or UV-A laser micro-irradiation (Figure 2-figure supplement 1C,D). This is in line with the effect on PALB2 recruitment, which is required for RAD51 loading (Sy et al., 2009;Zhang et al., 2009aZhang et al., , 2009b. We confirmed these findings in an siRNA-independent manner by utilizing fibroblasts from an RNF168-deficient RIDDLE patient who suffers from radiosensitivity and immunodeficiency, and displays dysmorphic features (Stewart et al., 2007). Analysis of mAG-geminin-expressing (S/G2) RIDDLE cells revealed markedly reduced RAD51 recruitment after IR when compared to the same cells complemented with RNF168 (Stewart et al., 2009) ( Figure 2B,E). To corroborate these results, we depleted endogenous RNF168 and transfected cells with siRNA-resistant GFP-tagged RNF168. Importantly, the expression of RNF168 fully rescued the loss of PALB2 foci after IR in cyclin B1-positive (S/G2) cells, demonstrating that this phenotype is not due to an off-target effect of the siRNA ( Figure 2F,G). To further validate these findings, we assessed whether knock-down of RNF168 would render cells sensitive to PARP inhibition, which is a feature of HR-deficient cells (McCabe et al., 2006). RNF168-depleted cells were hypersensitive to treatment with PARP inhibitor (PARPi), although not to the extent observed after PALB2 knockdown cells ( Figure 2H), suggesting a role for RNF168 in HR. Interestingly, the additional knock-down of PALB2 further sensitized RNF168-depleted cells to the same levels as PALB2 single knock-down cells ( Figure 2H; Figure 2-figure supplement 1F), indicating that RNF168 is epistatic to PALB2 in HR. To establish whether RNF168 promotes HR, we turned to the well-established DR-GFP reporter for HR activity (Pierce et al., 1999). Control experiments revealed that knock-down of BRCA2 almost fully impaired the HR-mediated repair of a nuclease-induced DSB ( Figure 2I). Using this reporter, we found that depletion of either RNF8 or RNF168 caused a significant reduction in HR efficiency ( Figure 2I) without any impact on cell cycle distribution (Figure 2-figure supplement 1E). Finally, we generated RNF168 knock-out mouse embryonic stem (ES) cells using CRISPR-Cas9-mediated genome editing. The genetic ablation of RNF168 was confirmed in two independent clones by western blotting ( Figure 2J). Both RNF168 knock-out clones showed a significant reduction in HR efficiency measured by the DR-GFP reporter ( Figure 2K), confirming our data in human cells. Together our results suggest that RNF168 facilitates PALB2/RAD51 loading and efficient execution of HR.

BRCA1 recruitment to HR sites does not require RNF168
The recruitment of PALB2 during HR requires both RNF168 ( Figure 2) and BRCA1 (Sy et al., 2009;Zhang et al., 2009aZhang et al., , 2009b. This raises the question whether RNF168 promotes HR at the level of PALB2 itself, or whether RNF168 acts upstream of PALB2 at the level of BRCA1. To address this, we specifically monitored BRCA1 foci formation in geminin-positive S/G2 cells at late time points post IR (6 hr after 10 Gy), when functional HR complexes have been assembled. While control cells showed clear accumulation of BRCA1 and 53BP1 in S/G2 cells, knock-down of RNF168 still allowed BRCA1 recruitment in cells that failed to form 53BP1 foci ( Figure 3A-D). It has previously been established that proteins involved in DSB signaling show extensive spreading from DSB sites, while proteins involved in HR show a more limited spreading, which is likely confined to resected DNA (Bekker-Jensen et al., 2006;Hu et al., 2011;Typas et al., 2015). As a consequence of this difference in the spreading from DSBs, signaling proteins form significantly larger IR-induced foci when compared to HR factors (see Figure 3-figure supplement 1 for a graphical representation of this phenomenon). To confirm and extend these findings, we performed quantitative measurements on IR-induced foci size. Our analysis confirmed that signaling proteins, such as MDC1, RNF168, BRCA1 and RAP80 form foci in the order of 0.15 mm 2 in size, whereas HR proteins, such as RPA, PALB2 and RAD51 form considerably smaller foci in the range of 0.06 mm 2 ( Figure 3E,F). Interestingly, the knock-down of RNF168 shifted the average BRCA1 foci size from that of the typical large foci (0.16 mm 2 ) to that in the range observed for HR proteins (0.08 mm 2 ; Figure 3G). A similar phenotype has been observed in RAP80-depleted cells both by us and by other groups (Hu et al., 2011;Typas et al., 2015). Although RNF168-depleted cells do form IR-induced BRCA1 foci in S/G2, we noted that the absolute number of BRCA1 foci is decreased compared to control S/G2 cells ( Figure 3C). However, a direct comparison between BRCA1 foci numbers in these conditions is difficult, because RNF168-depleted cells only form small HR foci, while control cells show small HR foci in addition to large signaling foci ( Figure 3G). In summary, our findings suggest that while RNF168 loss affects the formation of large BRCA1 foci, which arise from RAP80-dependent BRCA1 recruitment during DSB signaling, it does not impair the formation of small BRCA1 foci in S/G2 cells that are typical for factors involved in HR. This conclusion also fits with the observation that the RNF168mediated recruitment of PALB2 in the tethering system does not require BRCA1 ( Figure 1H,I), and suggests that RNF168 promotes HR downstream of BRCA1, likely at the level of PALB2.

PALB2 interacts with RNF168 via its WD40 domain
Given that RNF168 could promote HR at the level of PALB2 (Figures 2 and 3), we sought to address if these two proteins interact with each other. To test this, we performed a pulldown assay from cell extracts that revealed an interaction between GFP-tagged RNF168 and endogenous PALB2 ( Figure 4A). To rule out an indirect interaction between these proteins, we purified GST-tagged PALB2 and His-tagged RNF168 and addressed whether these proteins interact in vitro. In accordance with our pulldown results, we found that recombinant PALB2 efficiently bound recombinant RNF168, demonstrating a direct protein-protein interaction between these proteins ( Figure 4B). We next set out to map the region in PALB2 that associates with RNF168. To this end, we tested a series of recombinant PALB2 deletion mutants for their interaction with RNF168 in in vitro binding assays ( Figure 4C). Regions encompassing the ChAM domain (T3) and the C-terminal WD40 domain (T5) of PALB2 most strongly associated with RNF168 ( Figure 4C,D). To verify whether these PALB2 domains are required for the association with RNF168 in living cells, we monitored the interaction between RNF168 and internal deletion mutants of PALB2 lacking the ChAM (T6) or WD40 domain (T7). Both PALB2 WT and PALB2 DChAM (T6) interacted equally strongly with RNF168, while PALB2 DWD40 (T7) showed a markedly reduced ability to interact ( Figure 4E). To reciprocally verify this result, we pulled down YFP-tagged PALB2 WT or PALB2 DWD40 (T7) and examined their association with RNF168. While RNF168 readily bound YFP-PALB2 WT , this interaction was nearly lost in the absence of the WD40 domain ( Figure 4F). Similarly, the interaction between PALB2 and RAD51 was lost when the WD40 domain was lacking, agreeing with earlier reports (Zhang et al., 2009a;Buisson et al., 2010). The C-terminal WD40 domain is also the region in PALB2 that interacts with BRCA2 (Oliver et al., 2009). Indeed, we confirmed in pulldown experiments that BRCA2 readily associates with YFP-PALB2 (Figure 4-figure supplement 1A). We therefore asked whether RNF168, perhaps through its interaction with PALB2, can also associate with BRCA2. In addition to PALB2, we indeed found BRCA2 to also associate with GFP-RNF168 in pulldown experiments ( Figure 4G and Figure 4-figure supplement 1A). To validate these findings, we performed in vitro binding assays using purified PALB2, BRCA2 and RNF168 proteins. In these experiments, we initially used a smaller chimeric BRCA2 protein of 1009 amino acids (piccolo BRCA2 or piBRCA2), which interacts with both PALB2 and RAD51 (Buisson et al., 2010). Indeed, we found that recombinant PALB2 readily interacted with either piBRCA2 or RNF168. When all three proteins were incubated together, we also detected an efficient association of PALB2 with both piBRCA2 and RNF168    Figure 4H). Similar results were obtained when using purified full-length BRCA2 protein of 3330 amino acids (Figure 4-figure supplement 1B). These findings show that PALB2 interacts with RNF168 through its WD40 domain in a manner that is non-mutually exclusive with BRCA2 association.

PALB2 recruitment to DNA damage requires its WD40 domain
We next addressed whether the WD40 domain of PALB2 is required for its recruitment to chromatin ( Figure 5A). This would be predicted considering that RNF168 facilitates the recruitment of PALB2 (Figure 2), and because PALB2 interacts with RNF168 through its WD40 domain (Figure 4). Using the chromatin-tethering system, we found that, similarly to endogenous PALB2, YFP-PALB2 WT was recruited to chromatin following tethering of RNF168 ( Figure 5B-D). In stark contrast, YFP-PALB2 DWD40 (T7) failed to be recruited following tethering of RNF168, suggesting that the accrual of PALB2 in this system is WD40 dependent ( Figure 5C,D). Subsequently, we assessed the recruitment of YFP-tagged PALB2 to sites of DNA damage. Importantly, we found that YFP-PALB2 DWD40 (T7) did not accumulate as readily as YFP-PALB2 WT at laser-induced DNA damage ( Figure 5E,F), as well as at nuclease-induced DSBs ( Figure 5G-I). To further establish the importance of the WD40 domain for PALB2 recruitment, we complemented PALB2-deficient FANCN cells with YFP-tagged PALB2 WT or PALB2 DWD40 (T7). While YFP-PALB2 WT efficiently localized to sites of IR-induced DNA damage in cycling cells, its DWD40 counterpart showed significantly reduced recruitment ( Figure 5J,K). Importantly, BRCA1 was clearly recruited to nuclease-induced DSB ( Figure 5H,I), as well as to sites of IRinduced DSBs ( Figure 5J,K) in cells that did not show PALB2 DWD40 (T7) recruitment, suggesting that the inability of PALB2 DWD40 to accumulate at DSBs is not related to a defect in BRCA1 recruitment. This is in line with the observation that the interaction between BRCA1 and PALB2 relies on the N-terminal coiled-coiled domain and not the C-terminal WD40 domain of PALB2 (Zhang et al., 2009a(Zhang et al., , 2009b. Together, our findings suggest that RNF168 recruits PALB2 to DSBs through its WD40 domain.

Ubiquitin-dependent recruitment of PALB2 to DNA damage
Having shown that RNF168 facilitates PALB2 recruitment in the tethering system ( Figure 1 and Figure 5C,D) and at IR-induced DSBs (Figure 2), and having established that these proteins interact through PALB2's WD40 domain (Figure 4), we reasoned that RNF168 may recruit PALB2 to DSBs through a direct protein-protein interaction. Interestingly, efficient recruitment of RNF168 itself to DSBs requires an auto-amplification loop that involves the binding of RNF168 to its own ubiquitylation product through its ubiquitin-dependent DSB-recruitment module 2 (UDM2) (Panier et al., 2012;Thorslund et al., 2015). We therefore first asked whether this auto-amplification loop contributes to PALB2 accrual. To this end, we tethered catalytically inactive RNF168 C16S to a LacO array, which resulted in substantially reduced ubiquitin conjugation compared to tethering RNF168 WT ( Figure 6A,B). Strikingly, catalytically dead RNF168 C16S failed to recruit PALB2 to the LacO array in most cells ( Figure 6A,B), with PALB2 assembly being exclusively observed in cells retaining residual ubiquitin conjugates at the array, suggesting this is due to leakiness of the RNF168 C16S mutant ( Figure 6B). These data suggest that PALB2 recruitment is strictly dependent on the catalytic activity of RNF168.
One explanation for these results is that PALB2 can be co-recruited to RNF168-mediated ubiquitin conjugates on DSB-neighboring chromatin through its direct interaction with RNF168. If this hypothesis is correct then PALB2 should be able to indirectly associate with ubiquitin and such an interaction should require its WD40 domain, which mediates its association with RNF168 ( Figure 4C-F). In agreement with this rationale, recombinant GST-ubiquitin efficiently pulled down both RNF168 and YFP-PALB2 WT from cell extracts ( Figure 6C-E). As a control we included ubiquitin with a mutation in one of the key residues of the hydrophobic patch (I44), which inhibits the interaction with many ubiquitin-binding domains . Indeed, both GST alone and GSTubiquitin I44A failed to pulldown either PALB2 or RNF168, showing that the hydrophobic patch on the surface of ubiquitin is required for the interaction with RNF168/PALB2. Strikingly, YFP-PALB2 DWD40 (T7) failed to interact with GST-ubiquitin, indicating that the PALB2-ubiquitin interaction requires an intact WD40 domain ( Figure 6C-E), and suggesting that RNF168 constitutes a bridge between ubiquitin and PALB2. To exclude that PALB2 itself is not capable of binding to ubiquitin,  we performed in vitro interaction experiments using recombinant ubiquitin and either full-length or fragments of recombinant PALB2. Indeed, the WD40 domain of PALB2 was unable to bind polyubiquitin chains of various linkages under our assay conditions (K48, K63 and K27; Figure 6F). Moreover, none of the five PALB2 fragments spanning the entire protein, nor the full-length protein itself, was able to interact with recombinant ubiquitin (Figure 6-figure supplement 1A,B). These findings suggest that the WD40-dependent association of PALB2 with ubiquitin is not mediated by a direct interaction, but likely involves its ubiquitin-binding partner RNF168.

PALB2 interacts with ubiquitin through RNF168
We next sought to address if PALB2 associates with ubiquitin conjugates through its ubiquitin-binding partner RNF168 (Doil et al., 2009;Stewart et al., 2009;Panier et al., 2012;Thorslund et al., 2015). To this end, we performed in vitro GST-binding assays using recombinant K63-linked polyubiquitin chains, as these are the chains preferentially bound by RNF168 (Pinato et al., 2011). As expected, we found that purified RNF168 could interact with K63-polyubiquitin chains, but this association was not stimulated in the presence of PALB2 ( Figure 6-figure supplement 1C). In line with our previous data, we found that purified PALB2 was unable to bind K63-polyubiquitin chains on its own, but it was proficient in doing so when recombinant RNF168 was included in the reaction ( Figure 7A). Thus, our in vitro reconstitution experiments reveal an RNF168-dependent association of PALB2 with poly-ubiquitin conjugates.

PALB2 recruitment requires H2A ubiquitylation at K13/K15
It was demonstrated that RNF168 specifically ubiquitylates the core histone H2A at N-terminally situated lysines 13 and 15 in response to DNA damage (Mattiroli et al., 2012;Fradet-Turcotte et al., 2013). To address if these histone marks contribute to PALB2 recruitment, we made use of a pointmutant of RNF168 (R57D), which is still able to ubiquitylate, but has no activity toward H2A in a nucleosomal context (Mattiroli et al., 2012). In contrast to RNF168 WT , tethering LacR-tagged RNF168 R57D failed to recruit 53BP1 and PALB2, suggesting that RNF168-mediated H2A ubiquitylation promotes PALB2 recruitment ( Figure 7B,C). To corroborate these findings, we used a replacement strategy in which cells depleted for endogenous RNF168 were transfected with siRNAresistant versions of GFP-tagged RNF168. Ectopic expression of GFP-RNF168 WT fully rescued both PALB2 ( Figure 7D,E) and 53BP1 (Figure 7-figure supplement 1A,B) recruitment into IRIF. Conversely, neither RNF168 C16S nor RNF168 R57D were proficient in restoring PALB2 or 53BP1 foci formation after IR ( Figure 7D,E and Figure 7-figure supplement 1A,B). Importantly, GFP-RNF168 R57D , like its wild-type counterpart, was still recruited into IRIF ( Figure 7D) or to nuclease-induced DSBs (Figure 7-figure supplement 1C), albeit not as efficiently as wild-type RNF168. To further test the contribution of RNF168-mediated ubiquitylation of H2A on K13/K15, we over-expressed FLAGtagged versions of either H2A WT or H2A K13/15R in U2OS cells harboring a stably integrated LacO array and an inducible version of FokI-mCherry-LacR. Tethering of FokI-mCherry-LacR and subsequent DSB formation at the array triggered the robust recruitment of YFP-PALB2 in FLAG-H2A WTexpressing cells ( Figure 7F). Conversely, expression of FLAG-H2A K13/15R caused a significant decrease in YFP-PALB2 recruitment even though the accumulation of FokI-mCherry-LacR was similar between cells expressing H2A WT or H2A K13/15R ( Figure 7F,G). These results indicate a strong dominant-negative effect of the FLAG-H2A K13/15R mutant on PALB2 recruitment. We also witnessed a considerable decrease in 53BP1 recruitment in H2A K13/15R -expressing cells, demonstrating the validity of this approach (Figure 7-figure supplement 2A,B). Together, these results show that the catalytic activity of RNF168 toward H2A on K13/15 is required to promote efficient PALB2 recruitment. We propose that the initial RING-dependent ubiquitylation of H2A by RNF168, and the subsequent MIU-dependent binding of additional RNF168 proteins to this ubiquitin mark creates a chromatin-

PALB2 recruitment requires its interaction with the C-terminus of RNF168
Having shown that PALB2 recruitment requires the RNF168-dependent ubiquitylation of H2A (Figure 7), and considering that RNF168 and PALB2 interact at the protein level (Figure 4), we next sought to solidify that the interaction between these proteins is a prerequisite for the efficient recruitment of PALB2 to sites of DNA damage. Already supporting this possibility was our finding that the WD40 domain in PALB2 is required for both its interaction with RNF168 (Figure 4), as well as its efficient recruitment to DSBs ( Figure 5). However, considering that the WD40 domain of PALB2 also serves as the interaction surface for a number of other proteins , we set out to map the region in RNF168 that is required for the interaction with PALB2. To this end, we generated a series of deletion mutants and tested their ability to interact with endogenous PALB2 in pulldown assays ( Figure 8A-D). Using this approach we found that GFP-tagged RNF168 WT , RNF168 D1-189 (T1) and RNF168 D191-382 (T2) associated with PALB2 with equal efficiency ( Figure 8A, B), suggesting that the N-terminus and central region of RNF168 are not required for this interaction. Conversely, RNF168 D384-571 (T3) showed an almost complete loss of PALB2 binding, indicating that the C-terminus of RNF168 mediates its interaction with PALB2 ( Figure 8A-D). Considering that RNF168 D384-571 (T3) lacks the UDM2 required for the recruitment of full-length RNF168, we generated additional C-terminal deletion mutants that do include this domain ( Figure 8A). Analysis of these mutants revealed that RNF168 D479-571 (T4) and RNF168 D525-571 (T5) also failed to efficiently bind PALB2 ( Figure 8A-D). We conclude that the C-terminal~50 amino acids of RNF168 comprise a new PALB2-interacting domain, which we have named PID (for PALB2-Interacting Domain). An alignment of the PID in human RNF168 revealed considerable conservation of this domain in mammalian species (Figure 8-figure supplement 1). Having identified which region in RNF168 mediates its association with PALB2, we finally examined the functional relevance of this region for PALB2 recruitment. We reconstituted RNF168-depleted cells with siRNA-resistant GFP-tagged wild-type RNF168 or C-terminal deletion mutants of RNF168 that lack the PID and are therefore unable to interact with PALB2 ( Figure 8A). For this analysis we only used RNF168 mutants RNF168 D479-571 (T4) and RNF168 D525-571 (T5) since these mutants, in contrast to RNF168 D384-571 (T3), contain the UDM2 and are therefore proficient in accumulating at DSBs. Expression of these GFP-RNF168 variants rescued the 53BP1 foci defect in RNF168-depleted cells (Figure 8-figure supplement 2A-C), which shows that these mutants are recruited to DSBs and are able to ubiquitylate H2A. Conversely, while expression of RNF168 WT fully restored PALB2 foci formation in geminin-positive S/G2 cells, the C-terminal deletion mutants RNF168 D479-571 (T4) and RNF168 D525-571 (T5) failed to fully rescue this PALB2 foci defect ( Figure 8E,F). These data further prove that the RNF168-mediated recruitment of PALB2 depends on a direct protein-protein interaction that is mediated by the C-terminal PID of RNF168. To further corroborate these findings, we assessed the relevance of the PID of RNF168 for PALB2driven HR in PARPi survival assays. To this end, we used the Flp-In/T-REx system to establish HeLa cells stably expressing inducible siRNA-resistant GFP-tagged RNF168 alleles that were either able (WT) or unable (T5) to interact with PALB2. HeLa cells stably expressing GFP-NLS were generated to serve as a control. To perform PARPi survival assays, we knocked-down endogenous RNF168 with siRNAs in these cells and induced ectopic expression of RNF168 alleles with doxycycline (DOX). The expression was shut down by a DOX wash-out 12-24 hr before cells were subjected to a 24 hr pulse of 1 mM PARPi ( Figure 8G), resulting in ectopic RNF168 expression levels in the near-physiological range during the treatment with PARPi ( Figure 8H; Figure 8-figure supplement 3). Using this approach, we found expression of GFP-RNF168 WT to almost fully rescue the PARPi sensitivity of RNF168-depleted cells, while expression of GFP-NLS or GFP-RNF168 D525-571 (T5) failed to rescue this defect ( Figure 8I). Notably, we observed this effect in two independent clones expressing RNF168 WT or RNF168 D525-571 (T5). These findings suggest that the interaction between PALB2 and the C-terminal PID of RNF168 contributes to efficient DSB repair by HR. In summary, our findings support a piggyback ride mechanism for PALB2 recruitment, which links the ability of RNF168 to amplify and bind ubiquitin conjugates at damaged chromatin to a direct protein-protein interaction between PALB2's WD40 domain and RNF168's C-terminal PID ( Figure 8J).

Discussion
In this study, we define a new ubiquitin-dependent mechanism that regulates efficient HR. We demonstrate that RNF168 drives the recruitment of PALB2 to DSBs in a manner depending on its ubiquitin ligase activity toward histone H2A on K13/K15. Moreover, we show that RNF168 recruits PALB2 to DSBs through a direct protein-protein interaction that is established by the newly identified C-terminal PID of RNF168 and the C-terminal WD40 domain of PALB2. These findings advance our mechanistic understanding of the RNF168-dependent response to DSBs by revealing a physical and functional link between RNF168 and the core HR protein PALB2.

RNF168 promotes DSB repair through HR
The repair of DSBs in S/G2 cells relies on HR, which involves the sequential assembly of a functional repair complex consisting of BRCA1, PALB2, BRCA2 and RAD51. The accumulation of PALB2 depends on BRCA1 and is subsequently required for the efficient recruitment of BRCA2 and RAD51 (Sy et al., 2009;Zhang et al., 2009aZhang et al., , 2009b. Our current study reveals an unexpected role of the ubiquitin ligase RNF168 in promoting HR in S/G2-phase cells. Firstly, we show that knock-down of RNF168 impairs the recruitment of PALB2 and RAD51 to DSBs in S/G2 phase. This phenotype is also observed in cells from RNF168-deficient RIDDLE patients. Moreover, re-expression of siRNA-resistant RNF168 in RNF168-depleted cells, and of RNF168 in RIDDLE cells fully restored the HR defects, showing that these are a specific consequence of impaired RNF168 protein function. Secondly, we demonstrate that knock-down of RNF168 in human cells or knock-out of RNF168 in mouse ES cells confers a defect in HR measured by the well-established DR-GFP reporter. Thirdly, we find that knock-down of RNF168 leads to pronounced PARP inhibitor sensitivity, which is an established hallmark of defective HR. Previous studies also lend support for a role of RNF168-induced H2A ubiquitylation in promoting HR-mediated DSB repair. For example, cells deficient of the USP51 deubiquitylase, or the TRIP12 and UBR5 ubiquitin ligases display elevated levels of RNF168-induced H2A ubiquitylation at DSBs, which is accompanied by an increase in HR efficiency (Gudjonsson et al., 2012;Wang et al., 2016). Our results provide a mechanistic basis for these reported findings by showing that RNF168 couples the HR machinery to H2A ubiquitylation in S/G2 cells. Although the role of RNF168, and to a lesser extent RNF8, in regulating HR has remained controversial largely due to conflicting results (Huang et al., 2009;Meerang et al., 2011;Sy et al., 2011;Lu et al., 2012;Muñoz et al., 2012;Nakada et al., 2012;Zhang et al., 2012), our study solidifies the importance of RNF168 in promoting efficient HR. Consistently, we find that RNF8, which facilitates the recruitment of RNF168 to damaged chromatin through ubiquitylation of histone H1 (Thorslund et al., 2015), also contributes to HR. However, we cannot exclude the possibility that RNF8 promotes HR by targeting substrates other than H1 for ubiquitylation (Huang et al., 2009;Lu et al., 2012).

Auto-amplification of RNF168 recruitment contributes to PALB2 accrual
The RNF168 protein contains at least three domains required for its function. Firstly, the UDM1 domain associates with RNF8-generated ubiquitin conjugates on histone H1, which is the initial step of ubiquitin-dependent DSB signaling (Panier et al., 2012;Thorslund et al., 2015). Secondly, once recruited to RNF8-generated ubiquitin conjugates through UDM1, RNF168 utilizes its catalytic RING domain to ubiquitylate histone H2A at K13 and K15 (Mattiroli et al., 2012). Thirdly, the UDM2 domain in RNF168 promotes its association with the ubiquitin conjugates on H2A. Once bound, RNF168 can further ubiquitylate neighboring H2A using its RING domain and bind to these moieties through the UDM2. This auto-amplification through the UDM2 is critical for efficient RNF168 recruitment to DSBs (Panier et al., 2012;Thorslund et al., 2015). Our findings suggest that the ability of RNF168 to promote its own recruitment is also an integral part of PALB2-dependent HR. We show that both the catalytically inactive mutant RNF168 C16S (Doil et al., 2009) and the RNF168 R57D mutant, which is not able to bind and ubiquitylate nucleosomal H2A (Mattiroli et al., 2012), fail to promote PALB2 recruitment. Both these mutants could neither support PALB2 recruitment in the LacR/LacO tethering system nor to DSBs in complementation experiments with RNF168-depleted cells. Furthermore, we find that over-expression of H2A K13/15R , which lacks the ubiquitylation sites targeted by RNF168, inhibited PALB2 recruitment to nuclease-induced DSBs. Together, these findings suggest that the ubiquitin-dependent accumulation of RNF168 on broken chromosomes is critical for PALB2 recruitment.

RNF168 recruits PALB2 through protein-protein interactions
The significance of RNF168 in PALB2 recruitment is further corroborated by in vitro studies with purified RNF168 and PALB2, which reveal a direct protein-protein interaction. Domain mapping uncovered that the C-terminally located WD40 domain of PALB2 is the key determinant for this interaction. In line with RNF168 supporting PALB2 recruitment through a direct interaction, we find that full-length PALB2 lacking its WD40 domain is recruited much less efficiently to laser-, nuclease-, and IR-induced DSBs. Interestingly, biochemical reconstitution experiments reveal that PALB2 indirectly associates with poly-ubiquitin chains through a WD40-dependent interaction with recombinant RNF168. Thus, we can fully reconstitute the interaction between PALB2, RNF168 and poly-ubiquitin, providing a molecular mechanism for the RNF168-dependent recruitment of PALB2. Domain mapping also revealed that PALB2 associates with a previously unrecognized region in the C-terminus of RNF168, which we termed the PALB2-Interacting Domain (PID). Intriguingly, RNF168 mutants lacking the PID proved to be separation-of-function mutants which fully support 53BP1 foci formation, but are defective in supporting PALB2 accumulation at DSBs. Similarly, we found that stable ectopic expression of RNF168 WT rescued the PARPi sensitivity of RNF168-depleted cells, while a mutant lacking the PID (RNF168 D525-571 (T5)) failed to do so. These findings illustrate the importance of the protein-protein interaction between PALB2 and RNF168 in HR, and unveil the presence of a previously unrecognized domain in the C-terminus of RNF168 that is critical for its role in HR. One interesting possibility is that RNF168 through this direct protein-protein interaction targets PALB2 for ubiquitylation. However, whether this is the case and whether this potential RNF168-dependent ubiquitylation of PALB2 has functional significance for HR remains to be established.
A new RNF168-dependent mechanism for PALB2 recruitment Based on our findings, we propose the following model for RNF168-dependent regulation of HR ( Figure 8J): RNF168 decorates histones and other proteins with ubiquitin following its binding to RNF8-modified chromatin in the ssDNA-containing compartment proximal to DSBs. Although it may be expected that resected DNA is devoid of histones, previous studies have shown that resection can actually occur in a nucleosomal context (Tsukuda et al., 2005;Costelloe et al., 2012). RNF168 auto-amplifies its accumulation in the ssDNA compartment and recruits PALB2 through a direct protein-protein interaction. Once recruited, PALB2, in turn, facilitates BRCA2 and RAD51 assembly and promotes efficient HR. We propose that these events contribute to a trimodular mechanism for PALB2 recruitment in which (1) BRCA1 promotes PALB2 recruitment in an RNF168-independent manner through a protein-protein interaction involving the coiled-coil domain situated at the N-terminus of PALB2 (Zhang et al., 2009a(Zhang et al., , 2009b, (2) the more centrally located chromatin-association motif (ChAM) of PALB2 stabilizes its binding to damaged chromatin in the ssDNA compartment, and (3) the C-terminal WD40 domain of PALB2 further facilitates its retention by associating with the C-terminal PID of RNF168. We propose that integrating these independent recruitment signals leads to the efficient and specific accrual of PALB2 to HR sites ( Figure 8J). Such a multifaceted recruitment mechanism is not uncommon in the DSB response. For instance, the assembly of 53BP1 at damaged chromatin requires the simultaneous presence of two distinct marks on histones H4 and H2A Tuzon et al., 2014). It is clear that the action of RNF168 in chromatin more distal to DSBs results in the ubiquitin-dependent recruitment of 53BP1 (Bunting et al., 2010) and the BRCA1-Abraxas-RAP80-MERIT40 complex (Coleman and Greenberg, 2011;Hu et al., 2011), both of which are inhibitory to DNA-end resection. It is important to note that the RNF168mediated modification of H2A at K13/15 and the subsequent association of 53BP1 to this mark are not intrinsically inhibitory to HR. The inhibition of end resection is mediated downstream of 53BP1 by the recruitment of RIF1, which is restricted to G1 cells. Thus, RNF168 is only inhibitory to HR in G1 (Chapman et al., 2013;Escribano-Díaz et al., 2013;Orthwein et al., 2015). Indeed, our results reveal that RNF168 promotes PALB2 accrual and subsequent RAD51 loading on resected DNA in S/ G2 cells to promote HR. These findings demonstrate the versatility of the RNF168-mediated ubiquitylation of H2A in regulating HR in a cell cycle-dependent manner.

Cell culture
All cells were grown in DMEM (Gibco) containing 10% FCS (Bodinco BV) at 37˚C in 5% CO 2 . Human U2OS and HEK293 cells were purchased from ATCC. RNF168-deficient immortalized human RIDDLE fibroblasts and their FLAG-RNF168-complemented counterparts were a kind gift of Grant Stewart. FANCN/PALB2-deficient immortalized EUFA1341 fibroblasts were a gift of Johan de Winter. U2OS 2-6-3 cells containing 200 copies of a LacO-containing cassette were a gift of Susan Janicki (Janicki et al., 2004). A modified version of these cells stably expressing an inducible version of ER-mCherry-LacR-FokI-DD (2-6-5 cells) were a gift of Roger Greenberg (Tang et al., 2013). U2OS and RIDDLE cells stably expressing mAG-or mCherry-geminin were generated by lentiviral infection. HeLa Flp-In/T-REx cells, which were generated using the Flp-In/T-REx system (Thermo Fisher Scientific), were a gift of Geert Kops and Stephen Taylor. These cells were used to stably express inducible version of GFP-NLS, GFP-RNF168 WT or GFP-RNF168 4525-571 (T5) by co-transfection of pCDNA5/FRT/TO-Puro plasmid encoding GFP or GFP-tagged RNF168 (WT or T5) proteins (5 mg), together with pOG44 plasmid encoding the Flp recombinase (0.5 mg). After selection on 1 ug/mL puromycin, single clones were isolated and expanded. Stable HeLa Flp-In/T-REx clones were incubated with 50 ng/mL doxycycline to induce expression of cDNAs. All cells were authenticated by STR profiling and tested negative in routinely performed mycoplasma tests.

Transfections
Plasmid transfections were performed using Lipofectamine 2000. Cells were transfected with plasmid DNA using Lipofectamine 2000 (Invitrogen) or JetPEI (Polyplus) according to the manufacturer's instructions. Cells were typically analyzed 24 hr after transfection. All siRNA transfections (see Supplementary file 1 for siRNA sequences) were performed with 40 nM siRNA duplexes using Lipofectamine RNAiMAX (Invitrogen). Cells were transfected once or twice with siRNAs at 0 and 24 hr and were typically analyzed 60 hr after the first transfection.

Multiphoton laser micro-irradiation
U2OS cells grown on 18 mm coverslips were placed in a Chamlide CMB magnetic chamber and the growth medium was replaced by CO 2 -independent Leibovitz's L15 medium supplemented with 10% FCS and penicillin-streptomycin (Invitrogen). Laser micro-irradiation was carried out on a Leica SP5 confocal microscope equipped with an environmental chamber set to 37˚C. DSB-containing tracks (1.5 mm width) were generated with a Mira modelocked titanium-sapphire (Ti:Sapphire) laser (l = 800 nm, pulse length = 200 fs, repetition rate = 76 MHz, output power = 80 mW) using a UV-transmitting 63 Â 1.4 NA oil immersion objective (HCX PL APO; Leica). Confocal images were recorded before and after laser irradiation at 5 or 10 s time intervals over a period of 2-3 min.

UV-A laser micro-irradiation
U2OS cells were grown on 18 mm coverslips and sensitized with 10 mM 5 0 -bromo-2-deoxyuridine (BrdU) for 24 hr as described (Luijsterburg et al., 2016). For micro-irradiation, the cells were placed in a Chamlide TC-A live-cell imaging chamber that was mounted on the stage of a Leica DM IRBE widefield microscope stand (Leica) integrated with a pulsed nitrogen laser (Micropoint Ablation Laser System; Andor). The pulsed nitrogen laser (16 Hz, 364 nm) was directly coupled to the epifluorescence path of the microscope and focused through a Leica 40x HCX PLAN APO 1.25-0.75 oilimmersion objective. The growth medium was replaced by CO 2 -independent Leibovitz's L15 medium supplemented with 10% FCS and penicillin-streptomycin and cells were kept at 37˚C. The laser output power was set to 72 to generate strictly localized sub-nuclear DNA damage. Following micro-irradiation, cells were incubated for the indicated time-points at 37˚C in Leibovitz's L15 and subsequently fixed with 4% formaldehyde before immunostaining. Typically, an average of 50 cells was micro-irradiated (two iterations per pixel) within 10-15 min using Andor IQ software (Andor).

Microscopy analysis
Images were acquired on a Zeiss AxioImager D2 widefield fluorescence microscope equipped with 40x, 63x and 100x PLAN APO (1.4 NA) oil-immersion objectives (Zeiss) and an HXP 120 metal-halide lamp used for excitation. Images were recorded using ZEN 2012 software. IRIF were evaluated in ImageJ, using a custom-built macro that enabled automatic and objective analysis of the foci as described previously (Typas et al., 2015). expression vector and the I-SceI expression vector pCBASce. 48 hr later the percentage of GFP-positive cells among mCherry-positive cells was determined by FACS on a BD LSRII flow cytometer (BD Bioscience) using FACSDiva software version 5.0.3.

PARP inhibitor survival
U2OS cells were transfected with siRNA on day 1 and 2. On day 4, the cells were trypsinized, seeded at low density and mock-treated or exposed to 0.01, 0.1 and 1 mM PARP inhibitor KU-0058948 (Astrazeneca). On day 11, the cells were washed with 0.9% NaCl and stained with methylene blue. Colonies of more than 20 cells were scored. Stable HeLa Flp-In/T-Rex cells with inducible versions of GFP-NLS, GFP-RNF168 WT or GFP-RNF168 4525-571 (T5) were transfected with siRNA on day 1 and 2 in the presence of 50 ng/mL doxycycline to induce expression of GFP-tagged proteins. To shut down ectopic expression of GFP-tagged proteins, doxycycline was removed by extensive washes with PBS between 12-24 hr prior to trypsinization, seeding at low density followed by mock-treatment or exposure to a 24 hr pulse of 1 mM PARPi on day 4. On day 11, the cells were washed with 0.9% NaCl and stained with methylene blue. Colonies of more than 20 cells were scored.

Cell cycle profiling
For cell cycle analysis cells were fixed in 70% ethanol, followed by DNA staining with 50 mg/ml propidium iodide in the presence of RNase A (0.1 mg/ml) (Sigma). Cell sorting was performed on an LSRII flow cytometer (BD Bioscience) using FACSDiva software (version 5.0.3; BD). Quantifications were performed using Flowing Software.

Protein purification
for 35 min at room temperature in GSTB buffer containing 150 mM KCl. The GST pulldowns were performed as described above. For GST pulldowns in the presence of piBRCA2 or full-length BRCA2, purified GST-PALB2-His or GST were incubated with RNF168-His (500 ng) for 10 min at 37˚C in GSTB buffer containing 150 mM KCl followed by addition of FLAG-piBRCA2-His or BRCA2-FLAG for 10 min at 37˚C. The GST pulldowns were performed as described above. For GST pulldowns in the presence of piBRCA2, the quantities of proteins used in the assay are the following: GST-PALB2-His, GST and FLAG-piBRCA2-His: 200 ng; RNF168-His: 500 ng. For GST pulldowns in the presence of BRCA2-FLAG, the quantities of proteins used are the following: GST-PALB2-His, GST and BRCA2-FLAG: 125 ng; RNF168-His: 250 ng.

Statistical analysis
Statistical significance was assessed by a two-tailed, unpaired t-test and is indicated as: ***p<0.001, **p<0.01, *p<0.05. ns, not significant. All error bars represent the S.E.M.