RNF168 and USP10 regulate topoisomerase IIα function via opposing effects on its ubiquitylation

Topoisomerase IIα (TOP2α) is essential for chromosomal condensation and segregation, as well as genomic integrity. Here we report that RNF168, an E3 ligase mutated in the human RIDDLE syndrome, interacts with TOP2α and mediates its ubiquitylation. RNF168 deficiency impairs decatenation activity of TOP2α and promotes mitotic abnormalities and defective chromosomal segregation. Our data also indicate that RNF168 deficiency, including in human breast cancer cell lines, confers resistance to the anti-cancer drug and TOP2 inhibitor etoposide. We also identify USP10 as a deubiquitylase that negatively regulates TOP2α ubiquitylation and restrains its chromatin association. These findings provide a mechanistic link between the RNF168/USP10 axis and TOP2α ubiquitylation and function, and suggest a role for RNF168 in the response to anti-cancer chemotherapeutics that target TOP2.

T OP2a is important for chromosome condensation and segregation owing to its role in DNA decatenation and the decatenation G2 checkpoint 1,2 . This enzyme forms a transient covalent linkage with DNA, known as the TOP2a-DNA cleavage complex, during S phase to resolve DNA supercoiling that occurs during DNA replication, recombination and transcription [1][2][3] . TOP2a functions by binding a DNA duplex and cleaving both strands to allow another intact double-stranded DNA to pass through the break. Following strand passage, TOP2a ligates the DNA double-strand break (DSB), and thus relieves DNA topological stress. Through its activities, and its role in the decatenation G2 checkpoint, TOP2a is critical for maintaining genomic integrity [4][5][6] .
Post-translational modifications of TOP2a, including ubiquitylation, SUMOylation and phosphorylation, have been implicated in the regulation of TOP2a localization, turnover and activity [1][2][3] . TOP2a ubiquitylation by BMI1/RING1A or APC/C-Cdh1 promotes its degradation 2,7,8 , while its ubiquitylation by BRCA1 stimulates its decatenation activity 9 , in addition to its proteasomal degradation 10 . Despite the importance of ubiquitylation on TOP2a function and steadystate levels, the mechanisms that mediate its deubiqyuitylation have not yet been identified.
The E3 ubiquitin ligase RNF168 is an important regulator of DSB repair 11 . This repair process is vital for maintaining genomic integrity; and its defects have been implicated in the pathogenesis of a number of human diseases such as cancer 12,13 . At the sites of DSBs, RNF168 ubiquitylates H2A type histones to promote the recruitment of downstream DNA damage response proteins, including BRCA1 (refs 11,14,15). Further underscoring the importance of this E3 ubiquitin ligase, biallelic mutations of this gene have been identified as the underlying cause of the human RIDDLE syndrome [16][17][18] . In keeping with this disease connection, Rnf168 knockout mice exhibit many features of the human disorder, including increased radiosensitivity, immunodeficiency and genomic instability, which, in combination with the loss of p53, promotes earlier onset of cancer 19 .
To gain further insight into RNF168 functions, we used a proteomics approach to identify its interacting partners. Here, we report that RNF168 interacts with TOP2a to mediate its polyubiquitylation, and we further show that RNF168 deficiency confers resistance to ICRF-193, a TOP2 catalytic inhibitor, and etoposide (VP-16), a TOP2 poison and cytotoxic anti-cancer drug. We also report that RNF168 deficiency impairs TOP2a-chromatin association and its decatenation activity. Finally, we show that the deubiquitylase USP10 interacts with RNF168 and TOP2a, and restrains ubiquitylation of TOP2a as well as its chromatin binding. Collectively, our data reveal a novel and critical RNF168-dependent mechanism for fine-tuning TOP2a's decatenation activity through regulation of its ubiquitylation status. Our data also uncover USP10 as an important deubiquitylase for TOP2a.

Results
RNF168 is a novel interacting partner for TOP2a. To further define the functions of the E3 ligase RNF168, we initially expressed Flag-tagged RNF168 in HEK293T cells and coupled Flag immunoprecipitation (IP) with SDS-polyacrylamide gel electrophoresis analysis and mass spectrometry to identify novel RNF168-binding proteins. We found a 170 kDa protein that was significantly enriched in the Flag-RNF168 immunoprecipitate, which we identified as TOP2a ( Fig. 1a and Supplementary Table 1).
To confirm the physical association of RNF168 with TOP2a, we first co-transfected HEK293T cells with HA-tagged RNF168 and Flag-tagged TOP2a. Anti hemagglutinin (HA) immunoblot (IB) analysis of Flag-RNF168 immunoprecipitates from these transfected cells confirmed that RNF168 was pulled down by the TOP2a IP (Fig. 1b). Next, we examined the interaction between endogenous RNF168 and TOP2a. Consistent with our data using tagged proteins, reciprocal IPs of endogenous RNF168 and TOP2a from HEK293T cells showed that they interacted with one another (Fig. 1c).
We also examined using immunofluorescence the localization of RNF168 and TOP2a in S phase cells. Mouse embryonic fibroblasts (MEFs) were stained with the S phase marker 5ethynyl-2 0 -deoxyuridine (EdU), anti-RNF168, anti-TOP2a and 4,6-diamidino-2-phenylindole (DAPI). We observed coexpression of RNF168 and TOP2a in S phase cells (EdU-positive). These studies also indicated the colocalization of RNF168 and TOP2a S phase foci (revised Fig. 1d and Supplementary Fig. 1). Collectively, these results validate the interaction between RNF168 and TOP2a.
Important role for RNF168 in DNA decatenation. TOP2amediated decatenation of newly replicated sister chromatids is required for chromosome segregation and the entry of cells into mitosis 1,2 . To further test whether RNF168 affects TOP2a function, we performed a standard in vitro kinetoplast DNA-based decatenation assay using total nuclear extracts from primary MEFs and thymocytes deficient for Rnf168, and compared their decatenation activities with their wild-type (WT) controls. Rnf168 À / À cells showed reduced decatenation activity compared with WT controls ( Fig. 2a; Supplementary  Fig. 2), thus supporting the importance of RNF168 in DNA decatenation.
TOP2a is also required for the decatenation G2 checkpoint 5,6 . This checkpoint is distinct from the DNA damage-induced G2 checkpoint and serves to delay entry of cells with insufficiently decatenated chromatids into mitosis. To further test the effects of RNF168 deficiency on TOP2a function, we carried out decatenation G2 checkpoint assays that rely on TOP2 inhibition by the bisdioxopiperazine ICRF-193. This catalytic inhibitor of TOP2 blocks decatenation of intertwined daughter chromatids in G2 phase, resulting in activation of the decatenation G2 checkpoint and delayed progression into mitosis 1,20,21 . While ICRF-193 does not generally produce DSBs, it has been reported to do so under certain conditions 22,23 . Therefore, we first tested whether ICRF-193 induces DSBs in WT and Rnf168 À / À MEFs under our treatment conditions. Using foci for Serine 139 phosphorylated H2a.x (g-H2a.x) as markers DSBs (ref. 24), we observed that ICRF-193 did not generate DSBs in these cells ( Supplementary Fig. 3a,b). Therefore, we used ICRF-193-based assays to measure the activation of decatenation G2 checkpoint under our experimental conditions. We first performed a mitotic inhibition assay to assess the activation of the decatenation G2 checkpoint 25 . Cells treated with ICRF-193 or dimethyl sulfoxide (DMSO) for 15 min were incubated in culture media for an additional 2 h and the frequency of mitotic cells (phosphohistone-H3 positive, denoted 'pHH3 þ ') was determined by flow cytometry. ICRF-193 treatment significantly reduced the frequency of mitotic cells in WT MEFs and HA-RNF168reconstituted RIDDLE cells compared with Rnf168 À / À MEFs and HA-vector-reconstituted RIDDLE cells, respectively ( Fig. 2b-d; Supplementary Fig. 4a). We next performed a mitotic entry assay 25 to further examine the effects of RNF168 deficiency on the decatenation G2 checkpoint. Cells were treated for different durations (2-6 h) with either the microtubule depolymerization inhibitor colcemid alone, to prevent mitotic exit, or in combination with ICRF-193 to also block decatenation and delay entry into mitosis. The rate of mitotic entry of cells measured by pHH3 staining and flow cytometry showed higher mitotic rates for Rnf168 À / À MEFs and RIDDLE cells compared with WT MEFs and RNF168-reconstituted RIDDLE cells, respectively ( Fig. 2e-h; Supplementary Fig. 4b,c). Taken together, RNF168 deficiency leads to defective DNA decatenation and impaired activation of the decatenation G2 checkpoint, thus supporting a model in which loss of RNF168 inhibits TOP2a activity.
RNF168 deficiency increases the resistance to TOP2 drugs. To directly determine the effect of RNF168 deficiency on TOP2a function, we examined the cytotoxic effects of different concentrations of ICRF-193 on Rnf168 À / À MEFs and their WT controls by performing clonogenic assays. Camptothecin, a TOP1 inhibitor 26 , was used as a control for comparison. Our data indicated that while Rnf168 À / À MEFs displayed a similar sensitivity to camptothecin as compared with WT controls ( Supplementary Fig. 5), they were more resistant to ICRF-193induced cytotoxicity (Fig. 3a,b).
Etoposide, a TOP2 poison widely used for cancer therapy 21,27 , was also used to examine the effect of RNF168 deficiency on its cytotoxic effects. Analysis of clonogenic assays indicated that, similar to ICRF-193 treatment, Rnf168 À / À MEFs were highly resistant to etoposide-induced cytotoxicity (Fig. 3c,d). We next examined a panel of human breast cancer cell lines for their RNF168 expression levels and their response to etoposide. IB analysis indicated lower expression levels of RNF168 in the breast cancer cell lines MDA-MB-415 and MCF7 compared with SKBR3, T47D and MDA-MB-231 ( Supplementary Fig. 6a). To determine whether RNF168 expression level correlates with the response to etoposide, we examined the cytotoxic effect of etoposide on breast cancer cell lines that have different levels of endogenous RNF168. SKBR3, T47D and MDA-MB-231 cancer cell lines, which highly express RNF168, were more sensitive to etoposide compared with the low RNF168-expressing MCF7 and MDA-MB-415 cancer cell lines ( Supplementary Fig. 6b,c). To further elucidate the effect of RNF168 expression on the response of breast tumour cells to etoposide, we examined the effect of RNF168 knockdown on etoposide-induced cytotoxicity of T47D and MDA-MB-231 cell lines. Clonogenic assays using different doses of etoposide indicated that knockdown of RNF168 in T47D and MDA-MB-231 promotes their resistance to etoposide (Fig. 3e,f).
Collectively these data indicate that RNF168 deficiency promotes resistance to cytotoxicity induced by TOP2 catalytic inhibitors and poisons, further suggesting the importance of RNF168 for TOP2a function.
RNF168 mediates polyubiquitylation of TOP2a. Ubiquitylation plays important roles in regulating the function and turnover of TOP2a (ref. 2). As RNF168 interacts with TOP2a, we tested whether RNF168 could mediate ubiquitylation of TOP2a. Intracellular ubiquitylation assays indicated that endogenous TOP2a immunoprecipitated from RNF168-reconstituted RIDDLE cells, but not from the RIDDLE cells deficient for RNF168, were highly reactive to antibodies specific for ubiquitin (Ub; Fig. 4a). Knockdown of TOP2a in RNF168-reconstituted RIDDLE cells confirmed that RNF168 indeed polyubiquitylates endogenous TOP2a (Fig. 4a). Some of the TOP2a immunoprecipates from RNF168-expressing cells that were reactive to anti-Ub were smaller in size compared with the 170 KDa full-length TOP2a, suggesting that these smaller proteins might be the result of TOP2a degradation or TOP2a co-immunoprecipitating proteins. We also performed in vivo ubiquitylation assays and examined the effect of RNF168 deficiency in human breast cancer cell lines on the level of TOP2a ubiquitylation. We observed that knockdown of RNF168 in the tumour cell lines T47D and MDA-MB-231 restrained ubiquitylation of their endogenous TOP2a (Fig. 4b). Consistent with these data, ubiquitylation levels of Top2a in Rnf168 À / À MEFs were greatly reduced compared with WT controls (Fig. 4c).
Given that the E3 ligases RNF8, RNF168 and BRCA1 promote DSB repair 11 , and that BRCA1 has been reported to ubiquitylate TOP2a (refs 9,10), we were interested to see the effects of the deficiency of these E3 ligases on TOP2a ubiquitylation. As expected, Brca1 À / À MEFs displayed similarly reduced Top2a ubiquitylation compared with Rnf168 À / À MEFs (Fig. 4c). In contrast, RNF8, which plays important roles in the recruitment of RNF168 and BRCA1 to DSB flanking sites 11,28 , failed to interact with TOP2a, nor mediate its ubiquitylation (Supplementary Fig. 7a,b). Ubiquitylation of Top2a was also not affected in MEFs deficient for either Rnf8 or 53bp1, an important regulator of the DSB response 29 (Supplementary Fig. 7c). Altogether, these data indicate that RNF168 and BRCA1, but not RNF8, are major E3 ligases that mediate TOP2a ubiquitylation.
To further examine RNF168-mediated ubiquitylation of TOP2a, we also performed intracellular ubiquitylation assays. We observed that TOP2a smears became stronger when both exogenous RNF168-WT and Ub were co-expressed with the full-length TOP2a in HEK293T cells (Fig. 4d, lane 4). However, ubiquitylation of TOP2a was reduced when TOP2a was co-expressed with the Ub ligase-dead mutant RNF168 (RNF168-C21S) compared with RNF168-WT (Fig. 4d, cf. lanes 4 and 5). These data verify the importance of the E3 ligase activity of RNF168 for the ubiquitylation of TOP2a.
Given that different Ub linkage types exist and can lead to distinct cellular outcomes 30 , and that K48 and K63-Ub-linkages are the predominant forms of polyubiquitylation in cells, we performed intracellular ubiquitylation assays to explore whether RNF168-ubiquitylated TOP2a contains these Ub-linkages. Using Ub in which all lysines (K) except K48-or K63 were mutated to arginines, we observed that RNF168 preferentially catalyses K63-linked ubiquitylation of TOP2a ( Supplementary Fig. 7d). Consistent with these data, immunoprecipitated TOP2a from RNF168-reconstituted RIDDLE cells showed increased reactivity to antibodies against K63-Ub but not K48-Ub (Fig. 4e). These data support that RNF168 polyubiquitylation of TOP2a is mediated predominantly by K63-linkages. Similar to ubiquitylation, SUMOylation also regulates TOP2a functions 2 . In Xenopus, TOP2a SUMOylation on K660 (K662 in Homo sapiens) inhibits its decatenation activity 2,31,32 . Substitution of K660 to arginine (R) in TOP2a of Xenopus eliminates its SUMOylation-dependent inhibition 32 . Given that Ub and SUMO modifications have been shown to compete for the same K residues on certain substrates 33 , we sought to examine whether substitution K662R of TOP2a affects its ubiquitylation by RNF168. In vitro ubiquitylation analysis indicated reduced level of RNF168 ubiquitylation of TOP2a-K662R compared with TOP2a-WT ( Supplementary Fig. 7e). These data suggest that K662 of TOP2a is a target for RNF168 ubiquitylation. However, our observation that RNF168 ubiquitylation of TOP2a is not fully abrogated in the presence of K662R substitution, suggests that additional TOP2a residues are also ubiquitylated by RNF168.

ICRF-193 (μM)
To further examine whether TOP2a is a direct substrate for RNF168, we performed in vitro ubiquitylation assays using recombinant TOP2a, recombinant RNF168, E1 and E2 (Fig. 4f). These in vitro ubiquitylation assays indicated the ability of recombinant RNF168 to ubiquitylate recombinant TOP2a, thus suggesting that RNF168 is a direct E3 ligase for TOP2a ubiquitylation.
RNF168 E3 ligase activity stimulates TOP2a function. Based on the interaction of RNF168 and TOP2a, and our observation that TOP2a ubiquitylation by RNF168 does not increase its turnover, we hypothesized that RNF168 may regulate TOP2a. Given that TOP2a inactivation results in abnormal chromosome segregation 1,2 , we first examined whether RNF168 deficiency may also affect chromosome segregation. Defects in chromosome segregation can lead to increased frequency of lagging chromosomes that promote the formation of micronuclei as well as formation of chromosome bridges 5,34 . Interestingly, in contrast to WT MEFs, but similar to cells knocked down for Top2a, Rnf168 À / À MEFs readily displayed increased frequency of abnormal interphase nuclei with chromosome bridges (Fig. 5a,b and Supplementary Fig. 8). Similar to Top2a-knocked down cells, Rnf168 À / À MEFs also displayed a significantly increased frequency of micronuclei (Fig. 5a,b and Supplementary  Fig. 8). Human cells from the RIDDLE patient 15-9BI also showed increased frequency of micronuclei compared with their controls complemented with RNF168-WT (Fig. 5c). TOP2a is also essential for the condensation of mitotic chromosomes; and cells knocked down for TOP2a or treated with its catalytic inhibitor ICRF-193 display impaired chromosome condensation and increased frequency of entangled    2,3,5,35 . Therefore, we further tested our hypothesis that RNF168 regulates TOP2a function by examining the effects of RNF168 deficiency on chromosome condensation. Metaphase spreads of WT and Rnf168 À / À MEFs as well as Top2a-knocked down MEFs were examined for their frequency of undercondensed and entangled chromosomes. As expected, the frequency of these aberrant chromosomes was increased in WT MEFs treated with ICRF-193 or following Top2a knockdown ( Fig. 5d and Supplementary Fig. 8). Under untreated conditions, and in contrast to WT cells, Rnf168 À / À MEFs already displayed elevated frequencies of entangled and undercondensed chromosomes (Fig. 5d,e). The requirement of RNF168 for normal segregation and condensation of mitotic chromosomes, and decreased ability of etoposide to generate DSBs in Rnf168 À / À MEFs compared with WT controls ( Supplementary Figs 9 and  10), support our hypothesized role for RNF168 in regulating TOP2a function.
Given that the Ub ligase-dead mutant RNF168-C21S was defective in mediating TOP2a ubiquitylation (Fig. 4d), we examined the importance of the E3 ligase function of RNF168 for TOP2a activity using Rnf168 À / À MEFs reconstituted with empty vector (mock), Rnf168-WT or the Ub ligase-dead Rnf168-C21S mutant 36 . We first examined the frequency of undercondensed and entangled chromosomes in metaphase spreads of these cells. While complementation of Rnf168 À / À MEFs with Rnf168-WT reduced their frequency of undercondensed and entangled chromosomes, their complementation with the Rnf168-C21S mutant failed to rescue these defects (Fig. 5d,e).
TOP2a function is required for the decatenation G2 checkpoint 5,6 ; therefore we examined the effect of RNF168 E3 ligase activity on this checkpoint using the mitotic entry assay 25 . Reconstitution of Rnf168 À / À MEFs with Rnf168-WT reduced their fraction of mitotic cells to a level comparable to WT MEFs ( Fig. 5f; Supplementary Fig. 11). This rescue was however entirely dependent on Rnf168's E3 ligase activity since Rnf168 À / À MEFs complemented with Rnf168-C21S mutant retained their defective decatenation G2 checkpoint ( Fig. 5f; Supplementary Fig. 11). This finding prompted us to perform clonogenic assays and examine the effect of the loss of Rnf168 Ub ligase function on the response to ICRF-193-induced cytotoxicity. While complementation of Rnf168 À / À MEFs with Rnf168-WT restored their sensitivity to ICRF-193 to a level similar to WT controls, complementation of these cells with Rnf168-C21S mutant failed to suppress their resistance to ICRF-193 cytotoxicity (Fig. 5g). The inability of the E3 ligase-dead Rnf168 to rescue the defects of Rnf168 À / À MEFs, including impaired segregation and condensation of chromosomes, the decatenation G2 checkpoint and the resistance to the TOP2 inhibitor ICRF-193; all support the importance of this E3 ligase for the stimulation of TOP2a activity.
We also performed the standard in vitro kinetoplast DNA-based decatenation assay and examined decatenation activity of recombinant TOP2a that was pre-ubiquitylated in vitro by recombinant RNF168 (Fig. 4f). We observed that in vitro ubiquitylation of recombinant TOP2a by RNF168 increases its decatenation activity ( Supplementary Fig. 12). Collectively, these results support the importance of the E3 ligase function of RNF168 in the stimulation of TOP2a activity.
RNF168 ubiquitylates TOP2a to activate its chromatin-binding. TOP2a binding to chromatin is a necessary step for its biological function 1,2 . We therefore examined whether impaired ubiquitylation of TOP2a due to the absence of RNF168 affects its chromatin binding. IB analysis of WT MEFs indicated enrichment of Top2a in the chromatin fraction compared with the nuclear fraction (Fig. 6a). However, Rnf168 À / À MEFs displayed decreased levels of Top2a in their chromatin fraction, whereas their nuclear fractions displayed higher levels of Top2a compared with WT controls (Fig. 6a).
To confirm these findings in human cells, we performed similar fractionation experiments of breast cancer cell lines T47D and MDA-MB-231 and their RNF168-knockdown counterparts. We observed that in contrast to controls, but similar to Rnf168 À / À cells, RNF168-knockdown T47D and MDA-MB-231 cells displayed enrichment of TOP2a nuclear fractions (Fig. 6b).
Our data showed that RNF168 ubiquitylates TOP2a, and defects of this ubiquitylation affect TOP2 levels in the nuclear and chromatin fractions. Therefore, we next examined the effects of RNF168 overexpression on the ubiquitylation level of TOP2a in nuclear and chromatin fractions. We observed that co-overexpression of RNF168 and Flag-TOP2a in HEK293T cells enriches ubiquitylated TOP2a in both fractions; although this enrichment was more prominent in the chromatin fraction (Fig. 6c).
TOP2a from both nuclear and chromatin fractions have been previously reported to possess decatenation activity 9,37 ; and therefore, we performed in vitro kinetoplast DNA-based decatenation assays and examined the effects of Rnf168 deficiency on decatenation activities in the chromatin and nuclear fractions. Despite the elevated level of Top2a in the nuclear fractions of Rnf168 À / À MEFs, this fraction exhibited reduced decatenation activity compared with the nuclear fraction of WT MEFs (Fig. 6d). Furthermore, while Rnf168 deficiency impaired Top2a-chromatin association (Fig. 6a,b), Rnf168 À / À MEFs also exibited decreased decatenation activity in the chromatin fraction (Fig. 6d). We propose that RNF168 ubiquitylates TOP2a in the nuclear and chromatin fractions, and promotes its decatenation activity and chromatin association. Taken together these data highlight the importance of RNF168 for the regulation TOP2a function.

USP10 deubiquitylates TOP2a and restrains its chromatin binding.
Although ubiquitylation is central to the regulation of TOP2a function 2 , deubiquitylases that reverse this ubiquitylation have not yet been identified. Given that our mass spectrometry analysis identified USP10 as a possible RNF168-interacting partner (Supplementary Table 1), we hypothesized that USP10 and RNF168 forms a complex with TOP2a to co-regulate its ubiquitylation level. We tested this hypothesis by coexpressing HA-USP10 with Flag-RNF168 or Flag-TOP2a in HEK293T cells. Indeed, IP/IB analysis confirmed interaction of USP10 with RNF168 (Fig. 7a). Furthermore, we also observed interaction of USP10 with TOP2a (Fig. 7b). Examination of synchronized MEFs, in response to serum starvation, aphidicolin or colcemid treatments, indicated no changes of USP10-TOP2a interaction ( Supplementary Fig. 13a). However, similar to RNF168, USP10 interaction with TOP2a was reduced in response to radiation ( Supplementary Fig. 13b).
Clonogenic assays were performed to examine the sensitivity to etoposide and ICRF-193 of Usp10 knockdown WT MEFs and their controls. These studies indicated no change of sensitivity to TOP2 drugs in Usp10-deficient MEFs (Supplementary Fig. 13c). We next examined the effects of USP10 on RNF168-mediated ubiquitylation of TOP2a. Intracellular ubiquitylation assays using HEK293T cells transfected with RNF168, Flag-TOP2a, HA-USP10 and Myc-Ub indicated that USP10 antagonizes RNF168 ubiquitylation of TOP2a (Fig. 7c, cf. lanes 3 and 4). Furthermore, we examined the effect of Usp10-deficiency on the levels of Top2a ubiquitylation. MEFs knocked down for Usp10 displayed increased Top2a ubiquitylation compared with WT controls (Fig. 7d). These data indicate that the deubiquitylase USP10 opposes RNF168-mediated ubiquitylation of TOP2a.
We also examined whether USP10 affects TOP2a's binding to chromatin. To this end, we transfected HEK293T cells with Flag-TOP2a alone or with various combinations of RNF168, HA-Ub, HA-USP10-WT or the catalytically inactive mutant USP10-C424A (ref. 38) and performed IB analysis on chromatinenriched fractions of these cells. The fraction of TOP2a bound to chromatin was increased in cells co-transfected with TOP2a and RNF168, and this fraction was further enriched by the additional coexpression of Ub (Fig. 7e, cf. lane 1 versus lanes 2 and 3). However, additional overexpression of USP10-WT in these cells significantly reduced the fraction of chromatin-bound TOP2a (Fig. 7e, cf. lane 3 versus lane 4). In contrast to USP10-WT, overexpression of USP10-C424A was inefficient in reducing the level of chromatin-bound TOP2a (Fig. 7e, cf.

lane 4 versus lane 5).
Collectively these data reveal that RNF168 and USP10 exert important, but opposing, effects on the regulation of TOP2a's ubiquitylation status and association with chromatin.

Discussion
TOP2a is essential for DNA decatenation and chromosome segregation, and is an important target for cancer therapy 1,2,39 . The physiological importance of TOP2a is highlighted by the early embryonic lethality associated with its deficiency 40 . In accordance with its important functions, a number of post-translational modifications have been reported to regulate TOP2a's turnover and activity [1][2][3] . For instance, TOP2a ubiquitylation by BMI1/RING1A or APC/C-Cdh1 stimulates its turnover 2,7,8 . TOP2a is also ubiquitylated by BRCA1, although this ubiquitylation was reported to stimulate both its activity and proteasomal degradation 9,10 . Given the biological importance of TOP2a, its inhibition in cancer therapy, and the role ubiquitylation plays in regulating its turnover and activity, it is of importance to identify other E3 ligases that may ubiquitylate TOP2a. Furthermore, there is also a need to identify mechanisms that regulate deubiquitylation of TOP2a.
In the present study, we report polyubiquitylation of TOP2a by RNF168, an E3 ligase known for its role in DSB signalling and being mutated in the human RIDDLE syndrome. In vitro ubiquitylation studies confirmed that TOP2a is a direct substrate for RNF168. We also provide data to support that RNF168 ubiquitylation of TOP2a is mediated mainly by K63-Ub-specific linkages that are rarely associated with proteolysis 30 . Consistent with this finding, we observed that RNF168 ubiquitylation of TOP2a does not promote its degradation, but it rather stimulates its function. Finally, we provide evidence that RNF168-mediated ubiquitylation of TOP2a stimulates its decatenation activity.
Previous studies reported decatenation activity of TOP2a from nuclear and chromatin fractions 9,37 . Consistent with a role of RNF168 in the regulation of both nuclear and chromatin-bound TOP2a, overexpression of this E3 ligase increased ubiquitylation levels of TOP2a in both fractions, although the chromatin-bound TOP2a showed more pronounced ubiquitylation.
Consistent with the importance of RNF168 ubiquitylation of TOP2a for stimulating its function, deficiency of RNF168 results in defective activation of the decatenation G2 checkpoint, elevated frequency of chromosome aberrations (undercondensed and/or entangled), and impaired chromosome segregation (bridges and micronuclei). The E3 ligase function of RNF168 is required for its regulation of TOP2a function. Our study indicate that in contrast to RNF168-WT, its E3 ligase inactive form fails to rescue defective TOP2a function in cells deficient for RNF168. Collectively these data reveal a novel and important role for RNF168 in mediating TOP2a ubiquitylation and the activation of its decatenation function. BRCA1 is a tumour suppressor with a major function in DSB signalling/repair 11,41 . Consistent with previous reports 9, 10 , we observed that BRCA1 deficiency impairs TOP2a ubiquitylation level. TOP2a in the nuclear fraction of BRCA1-deficient cells has been shown to have defective decatenation activity, suggesting that BRCA1 stimulates TOP2a decatenation activity 9 . Similar to BRCA1-deficient cells, Top2a from the nuclear fraction of Rnf168 À / À cells also displayed defective decatenation activity. This defect was readily detected despite increased level of Top2a in the nuclear fraction of Rnf168 deficient cells.
Furthermore, Rnf168 stimulates Top2a-chromatin binding and its loss reduces decatenation activity of Top2a in the chromatin fraction.
Deubiquitylases play major roles in several cellular processes and signalling pathways, and their defects have been associated with a number of diseases including cancer 42 . Deubiquitylases and E3 ligases exert opposing effects on the ubiquitylation levels of their substrates and, as such are critical for the regulation of ubiquitylation-dependent processes 43 . Despite the importance of TOP2a function and the role ubiquitylation plays in the regulation of its activity and turnover 1,2 , mechanisms that deubiquitylate TOP2a, and the deubiquitylases involved in this processes, remain elusive. USP10 is a deubiquitylase important for the deubiquitylation and regulation of turnover and function of a number of substrates including p53, H2A.Z, SIRT6, AMPK, TRAF6 and MSH2 (refs 38,42,44-48). Interestingly, in this study, we also identified USP10 as an interacting partner for both TOP2a and RNF168. Furthermore, we discovered that USP10 antagonizes RNF168-mediated ubiquitylation of TOP2a and decreases TOP2a binding to chromatin, suggesting a role for this deubiquitylase in opposing RNF168-mediated effects on TOP2a (Fig. 7f). TOP2 poisons (for example, etoposide) are chemotherapeutic agents extensively used to treat a broad spectrum of human cancers 21,27 . However, resistance to these drugs is common and significantly limits their potential therapeutic benefit. Several mechanisms have been ascribed to drive tumour resistance to TOP2 poisons, including down-regulation of TOP2a expression 27,49 . In this study, we observed that RNF168 deficiency promotes the resistance to cytotoxicity induced by the TOP2 poison etoposide and its catalytic inhibitor ICRF-193. Notably, knock down of RNF168 in human breast cancer cell lines T47D and MDA-MB-231 promoted their resistance to etoposide-induced cytotoxicity. These findings suggest that RNF168 defects (mutations or loss of expression) may contribute to the resistance of cancer patients to TOP2 poisons.
Similarities exist between BRCA1 (ref. 9) and RNF168 in mediating ubiquitylation of TOP2a and stimulation of its decatenation activity. However, in contrast to RNF168, BRCA1 deficiency increases cytotoxicity to TOP2a catalytic inhibitors and poisons 50,51 . The intriguing sensitivity of BRCA1 mutant cells and tumors suggests additional functions of BRCA1 compared with RNF168. Indeed, a recent study indicated the requirement of BRCA1, together with the Mre11-Rad50-Nbs1 complex and CtIP, for the removal of etoposide-induced TOP2 protein-DNA adducts and the subsequent resection of etoposidegenerated DSBs (ref. 52). Defects of these novel BRCA1 functions would sensitize cells to TOP2 catalytic inhibitors and poisons 53 . In addition, the critical roles of BRCA1 in the homologous recombination (HR) pathway of DSB repair and in the CtIP-dependent processing of etoposide-induced DSBs (ref. 52), support the importance of HR in the repair of etoposide-induced DSBs. Interestingly, in contrast to BRCA1, RNF168 deficiency only mildly affects HR-mediated DSB repair 36 . Therefore, the difference in the requirement for BRCA1 and RNF168 for HR-mediated repair could also contribute to the different response of BRCA1 and RNF168 deficient cells to TOP2 catalytic inhibitors and poisons.
In this study, we also report that in addition to RNF168, TOP2a also interacts with USP10. We show that USP10 antagonizes RNF168 ubiquitylation of TOP2a and restrains RNF168 stimulation of TOP2a binding to chromatin. USP10deficient cells displayed increased ubiquitylation level of TOP2a compared with WT controls, confirming that USP10 is a deubiquitylating enzyme for TOP2a. Although it is conceivable that deregulation of this deubiquitylase could also contribute to the resistance of tumors to TOP2 drugs, given the growing number of its substrates and their involvement in different cellular processes 38,42,[44][45][46][47][48] , this remains to be demonstrated. Further studies are also required to determine whether loss of RNF168 expression or its inactivation correlate with resistance to anti-cancer chemotherapeutics that target TOP2 in cancer patients.
In summary, our results reveal the importance of the E3 ligase RNF168 for mediating the TOP2a ubiquitylation and the stimulation of its decatenation activity. We also identified USP10 as an interacting partner for RNF168 and TOP2a, and provide evidence that USP10 deubiquitylates TOP2a and negatively regulates its chromatin binding (Fig. 7f). These novel functional interactions between RNF168, TOP2a and USP10 are likely to contribute to the maintenance of chromosomal integrity. In addition, our data suggest that these functional interactions between RNF168 and TOP2a may play important roles in the response and resistance of tumors to anti-cancer drugs that target TOP2.

Methods
Mass spectrometry analysis. Mock or Flag-RNF168 expressing HEK293 cells were used for affinity purification. Samples were resolved in a gel that was subsequently stained with silver staining for visualization. The entire protein containing gel chunk was split into two and processed for mass spectrometry as previously described 54 . Samples were resuspended in 0.1% (MS grade) formic acid (Sigma). Analytical columns (75 mm inner diameter) and pre-columns (100 mm) were prepared in-house from silica capillary tubing (InnovaQuartz, Phoenix, AZ), and packed with 5 mm 300 Å C18-coated silica particles (Michrom). Analytical columns were fitted with metal emitters (Thermo Proxeon) using zero dead volume connections. Peptides were subjected to nanoflow liquid chromatographyelectrospray ionization-tandem mass spectrometry (nLC-ESI-MS/MS) , using a 90 min reversed phase (10-40% acetonitrile and 0.1% formic acid) buffer gradient running at 250 nl min À 1 on a Proxeon EASY-nLC pump in-line with a hybrid linear quadrupole ion trap Orbitrap mass spectrometer (LTQ Orbitrap, Thermo Fisher Scientific). A parent ion scan was performed in the Orbitrap, using a resolving power of 60,000. Simultaneously, up to the four most intense peaks were selected for MS/MS (minimum ion count of 500 for activation) using standard collision-induced dissociation (CID) fragmentation. Fragment ions were detected in the LTQ. Dynamic exclusion was activated such that MS/MS of the same m/z (within a 0.1 Da window, exclusion list size 500) detected three times within 45 s were excluded from analysis for 60 s. Thermo RAW files were converted to mzXML (Proteowizard 55 ), and searched with the iProphet 56 pipeline in Prohits 57 and Comet 58 search algorithms, using a parent mass window of 20 p.p.m. and 0.4 Da fragment mass window, against a database containing human refseq v57 forward and reversed protein sequences. Up to two missed cleavages were allowed, and oxidation of methionine, deamidation of glutamine and asparagine were set as variable modifications. Polypeptides with a Trans-Proteomic Pipeline (TPP) probability of 1.00 (and identified with 4 or more unique peptides) are shown in Supplementary Table 1. Constructs and mutagenesis. Full-length mouse Rnf168 was amplified by PCR using complementary DNA from WT MEFs and subcloned in p3xFLAG-CMV10 (Sigma-Aldrich), pcDNA3-HA (Invitrogen) and pBabe-puro (Addgene; ID 1764). FLAG-TOP2a was kindly provided by Dr. Sullivan (H. Lee Moffitt Cancer Center and Research Institute). HA-RNF8 was kindly provided by Dr. Durocher (Samuel Lunenfeld Research Institute). Site-specific mutagenesis was performed using QuickChange kit (Stratagene).
Retroviral and lentiviral infections. Ecotropic retroviruses were generated by transient transfection of Phoenix cells 36 . For lentviruses, HEK293T cells were co-transfected by calcium phosphate precipitation with pLKO constructs containing specific ShRNA or ShRNA control together with psPAX-2 and pMDG1.vsvg. 24 h later, the media was replaced with fresh complete growth medium and cells were incubated at 37°C for an additional 24-48 h before harvesting the virus. Cells were infected with retrovirus or lentivirus supernatants containing polybrene (Sigma-Aldrich) and were subjected to puromycin selection (10 mg/ml).
Clonogenic assay. 3T3 MEFs (200 cells) and breast cancer cell lines (300 cells) were seeded on 6 cm dishes, treated with DMSO, ICRF-193, etoposide or camptothecin for 24 h and cultured for an additional 11 days. Colonies were stained with crystal violet and counted. beads were collected and washed five times with the lysis buffer and one time with detergent-free lysis buffer.
Assessment of nuclear abnormalities. Cells grown on glass coverslips were fixed with 2% paraformaldehyde for 5 min at room temperature, stained with DAPI for 5 min and then mounted with Mowiol mounting solution. More than 1,000 interphase nuclei were analysed using a Leica DM4000B fluorescence microscope.
Mitotic inhibition assay of decatenation G2 checkpoint. Asynchronous cells were treated for 15 min with DMSO or 4 mM ICRF-193 (Sigma-Aldrich), followed by incubation in drug-free medium for 2 h. Cells were then trypsinized into a single-cell suspension and fixed with 70% ethanol. Mitotic cells were quantified by flow cytometry using anti-pHH3 (pSer10) antibody (1:100; Cell signaling, 9708) and Propidium Iodide (Sigma-Aldrich).
Mitotic entry assay of decatenation G2 checkpoint. Cells were plated at a density of 10 6 cells per 10 cm dish (day 0) and fed with fresh medium on day 1. On day 2, cells were treated with 100 ng ml À 1 colcemid and either DMSO or 4 mM ICRF-193 for 2, 4 and 6 h. Cells were stained with anti-pHH3 antibody and PI and mitotic fraction was identified by flow cytometry.
EdU labelling and RNF168, TOP2a staining. For EdU labelling, we used Click-It Edu Alexa Fluor 647 Imaging Kit according to manufacturer's recommendations (Invitrogen). EdU was added to growth medium (10 mM) for 4 h and was subsequently detected with Click-It cocktail containing Alexa Fluor 647 azide as one of the components. After washing with 3% BSA in PBS, the sheep polyclonal against RNF168 (1:2,000; R&D systems cat# AF7217) and the rabbit polyclonal against TOP2a (Abcam) were detected using their respective secondary antibodies. Cells were examined using the confocal laser scanning microscope 700 (Carl Zeiss).
Comet assay. Neutral comet assay was performed as previously described 61 . Single-cell suspensions of MEFs were collected after treatment with 20 mM etoposide for 30 min. They were mixed with 0.7% low-melting agarose and spread on slides pre-coated with 1% normal melting agarose. Slides were then treated with lysis solution (2.5 mM NaCl, 100 mM EDTA, 10 mM Trizma base, 10% DMSO and 1% Triton-X100, pH 9.0) at 4°C for 72 h. Electrophoresis was performed in Tris-buffer with boric acid and EDTA for 30 min at 32 V. Slides were then air-dried and stained with 2 mg ml À 1 of ethidium bromide and scored using KOMET 5.0 software (Kinectic Imaging). A minimum of one hundred cells were scored at random for each treatment. Olive Tail Moment (% DNA in the tail/distance between centre of gravity of head and tail) and % Tail DNA were reported as a measure of DNA double-strand breaks.
x foci experiments, MEFs grown on glass coverslips were fixed with 2% paraformaldehyde for 5 min at room temperature and then incubated overnight at 4°C with mouse anti-g-H2a.x (Millipore). Labelling was detected using Alexa Fluor 555-labelled goat-anti-rabbit immunoglobulin secondary antibodies (Molecular Probes). Cells were stained with DAPI for 5 min and then mounted with Mowiol mount solution (Calbiochem). The slides were examined under a Leica DM4000B fluorescence microscope equipped with digital camera (Leica DFC490). Images were acquired under Â 100 magnification using Leica Image Manager software.
Statistical analysis. Data are presented as the mean ± s.e.m. The statistical significance of experimental data (P-values r0.05) was determined using ANOVA with or without Tukey-Kramer test.
Data availability. The authors declare that the data supporting the findings of this study are available from the corresponding authors upon request.