The de-ubiquitylating enzymes USP26 and USP37 regulate homologous recombination by counteracting RAP80

The faithful repair of DNA double-strand breaks (DSBs) is essential to safeguard genome stability. DSBs elicit a signalling cascade involving the E3 ubiquitin ligases RNF8/RNF168 and the ubiquitin-dependent assembly of the BRCA1-Abraxas-RAP80-MERIT40 complex. The association of BRCA1 with ubiquitin conjugates, which occurs in a RAP80-dependent manner, is known to be inhibitory to DSB repair by homologous recombination (HR). However, the precise regulation of this mechanism remains poorly understood. By performing genetic screens, we identified USP26 and USP37 as key deubiquitylating enzymes (DUBs) that limit the repressive impact of RNF8/ RNF168 on HR. Both DUBs are recruited to bona fide DSBs where they actively remove RNF168-induced ubiquitin conjugates. Depletion of USP26 or USP37 disrupts the execution of HR and this effect is alleviated by the simultaneous depletion of RAP80. In addition, we demonstrate that these DUBs prevent the ubiquitin-dependent sequestration of BRCA1 via the BRCA1-Abraxas-RAP80-MERIT40 complex, allowing BRCA1 to form a complex and cooperate with PALB2-BRCA2-RAD51 in HR. These findings reveal a novel ubiquitin-dependent mechanism that orchestrates the spatial assembly of distinct BRCA1-containing complexes for efficient repair of DSBs by HR. BRCA1 to form a complex with PALB2-BRCA2-RAD51 and execute HR. Thus, these enzymes promote HR by limiting the repressive impact of RAP80 on HR. These findings reveal a novel ubiquitin-dependent mechanism that orchestrates the spatial assembly and function of HR complexes at DSBs.


INTRODUCTION
DNA double-strand breaks (DSBs) pose a considerable threat to the stability of the human genome and their timely repair is essential to safeguard genome stability and counteract tumour development [1]. Cells activate robust signalling pathways in response to DSBs that coordinate cell cycle progression, changes in chromatin structure and DNA repair [2,3]. Eukaryotic cells primarily utilise homologous recombination (HR) or non-homologous end-joining (NHEJ) to remove DSBs from their genomes.
A key feature of the DNA damage response (DDR) is the rapid assembly of signalling and repair factors in the vicinity of DSBs, by progressively modifying histones and DNA repair enzymes [4,5]. An initial phosphorylation-dependent cascade of post-translational modifications in DSB-containing chromatin requires the ATM kinase and culminates into the association of MDC1 with phosphorylated histone H2A variant H2AX (γH2AX) [6]. The binding of the RNF8 E3 ubiquitin ligase to MDC1 subsequently initiates a ubiquitylationdependent cascade, involving the recruitment of the E3 ubiquitin ligase RNF168 in cooperation with the E2 ubiquitin-conjugating enzyme UBC13 [7,8]. The activity of these enzymes contributes to the ubiquitylation of K13/15 on histone H2A/H2AX [9,10], as well as the ubiquitin-dependent assembly of 53BP1 [11], RAD18 [12] and the BRCA1-Abraxas-RAP80-MERIT40 (or BRCA1-A) complex [13][14][15][16] onto DSB-neighbouring chromatin.
Although the principles underlying the RNF8 signalling pathway are by now well understood, we are only beginning to comprehend how this pathway is linked to the actual repair of DSBs through the major repair pathways NHEJ [27] and HR [28][29][30]. During HR, the ends of a DSB are resected to expose 3' single-stranded DNA (ssDNA) overhangs, which are rapidly coated with the ssDNA-binding protein RPA. Following resection, the PALB2 protein is recruited by BRCA1 and subsequently facilitates the assembly of BRCA2 [31,32]. This, in turn, promotes the exchange of RPA with RAD51, which drives the search for and pairing with a homologous sequence, as well as the exchange of homologous DNA during the final steps of HR [31][32][33]. BRCA1 is incorporated into distinct multi-protein complexes, including BRCA1-PALB2-BRCA2-RAD51 and BRCA1-Abraxas-RAP80-MERIT40 [34]. Strikingly, while the BRCA1-PALB2-BRCA2-RAD51 complex promotes HR, the BRCA1-Abraxas-RAP80-MERIT40 complex functionally antagonises this repair process by sequestering BRCA1 from HR sites by binding to RNF8/ RNF168-ubiquitylated chromatin [16,[35][36][37][38][39][40]. These findings suggest that distinct BRCA1-containing complexes can differentially affect HR in a manner dependent on DNA damage-induced ubiquitylation. Remarkably, little is known about the involvement of DUBs in regulating BRCA1-dependent HR.
Through genetic screens we identified the de-ubiquitylating enzymes USP26 and USP37 as key factors critical for DSB repair by HR. Mechanistically, we show that by removing RNF168-induced ubiquitin conjugates distal from DSBs, these enzymes prevent the ubiquitin-dependent sequestration of BRCA1 through the BRCA1-Abraxas-RAP80-MERIT40 complex, ultimately allowing BRCA1 to form a complex with PALB2-BRCA2-RAD51 and execute HR. Thus, these enzymes promote HR by limiting the repressive impact of RAP80 on HR. These findings reveal a novel ubiquitin-dependent mechanism that orchestrates the spatial assembly and function of HR complexes at DSBs.

A screen for DUBs reveals novel regulators of 53BP1 and RAD51
The BRCA1 protein is incorporated into distinct multi-protein complexes that are not all competent in promoting HR. While the BRCA1-PALB2-BRCA2-RAD51 complex promotes HR, the BRCA1-Abraxas-RAP80-MERIT40 complex functionally antagonises this repair process by sequestering BRCA1 from HR sites by binding to RNF8/RNF168-ubiquitylated chromatin [16,[35][36][37][38][39][40]. These findings suggest that distinct BRCA1-containing complexes can differentially affect HR in a manner dependent on RNF8/RNF168 damageinduced ubiquitylation. Although the responsible E3 ubiquitin ligases RNF8 and RNF168 have been characterised [7,8,[13][14][15], potential DUBs that play a role in this ubiquitin-dependent regulation of HR remain elusive. In order to identify such proteins, we performed an over-expression screen using a FLAG-tagged cDNA library of ~60 human DUBs (Supplemental Fig.   1A) in human U2OS cells (Fig. 1A). Specifically, we monitored if DUB overexpression simultaneously antagonises the ionising radiation (IR)-induced formation of 53BP1 foci, a read-out for RNF168-mediated ubiquitylation [11], as well as the IR-induced focal accumulation of RAD51, a measure of HR efficiency. Given that 53BP1 directly binds to RNF168-induced ubiquitin conjugates [9,11], we reasoned that DUBs modulating both these processes are likely to regulate RNF168-mediated HR events.

3
Chapter 3 fact that we identified various published DUBs demonstrates the validity of our screen. Importantly, among the DUBs that suppressed both 53BP1 and RAD51 IRIF formation, USP26 and USP37 emerged as novel candidates ( Fig. 1B and Supplemental Fig. 1D,E). We could not distinguish a common pattern in the impact of DUBs on 53BP1 or RAD51 foci formation within USP, UCH, MJD, JAMM, OTU or unclassified DUBs (Supplemental Fig. 1B-E), suggesting this is a unique property of the identified enzymes. Thus, via our screen, we identified USP26 and USP37 as potential novel regulators of 53BP1 and RAD51.    FokI nuclease to a genomic locus containing LacO repeats [43]. In line with the results obtained by laser micro-irradiation, both USP26 and USP37 accumulated at bona fide, FokI-induced DSBs marked by γH2AX ( Fig. 2A 3B) accumulation at the array, implying that they directly remove RNF168induced ubiquitylation. RNF168 targets H2A-type histones for ubiquitylation [9,10]. Interestingly, ectopic expression of USP37 moderately decreased RNF168-induced H2A ubiquitylation [9,46], while expression of USP26 nearly eliminated such ubiquitylation (Fig. 2E). Thus, our results suggest that USP26 and USP37 are able to bind to chromatin modified by RNF8/RNF168 3 Chapter 3 and reverse ubiquitylation induced by these E3 ligases at DSBs.

Loss of USP26 or USP37 impairs DSB repair
To address if the identified DUBs play a role in HR under physiological conditions, USP26 and USP37 were depleted using independent siRNAs.
Immunoblotting Flow cytometric analysis of DR-GFP cells confirmed that USP26 or USP37 depletion leads to a significant defect in HR (Fig. 4A). Notably, overexpression of mCherry-tagged RNF8 or RNF168 also strongly inhibited HR The effects on HR were not due to alterations in the cell cycle as cell cycle profiles were unchanged under these conditions (Supplemental Fig. 5A,B).
Depletion of USP26 or USP37 also rendered cells highly sensitive to a poly(ADP-ribose) polymerase (PARP) inhibitor, which is a hallmark of HRdeficient cells such as those lacking BRCA2 (Fig. 4C) [47].
Having shown that USP26 and USP37 regulate HR, we next sought to address if these enzymes affect the other major DSB repair pathway, nonhomologous end-joining (NHEJ). Using the flow cytometry-based EJ5-GFP reporter assay to monitor NHEJ efficiency [48], we found that USP26 or USP37 depletion substantially impaired this repair pathway (Fig. 4D). In line with a general defect in DSB repair, the combined knockdown of USP26 and USP37 led to a delay in the clearance of IR-induced γH2AX foci (Supplemental

Loss of USP26 or USP37 impairs HR by antagonising RAP80-dependent sequestration of BRCA1
To gain insight into the mechanism that disrupts HR under conditions of excessive ubiquitylation, we turned our attention to the BRCA1-Abraxas-RAP80-MERIT40 complex, which through RAP80 drives the ubiquitindependent recruitment of BRCA1 to RNF8/RNF168-modified chromatin [7,[13][14][15][16][35][36][37]. It was recently unveiled that RAP80-mediated recruitment of BRCA1 inhibits HR [38,39], by sequestering BRCA1 from HR sites, thus hampering the formation of a BRCA1-PALB2-BRCA2-RAD51 complex, which is essential for HR [38,40]. We reasoned that USP26 and USP37 may antagonise the RAP80-dependent sequestration of BRCA1 by removing RNF8/RNF168-mediated ubiquitylation and thereby promote HR. To address this, we established a quantitative, computer-assisted approach to measure BRCA1 foci size. In agreement with an earlier report [38], we found that BRCA1 foci were not reduced in number, but rather were considerably smaller in size following depletion of RAP80 (Fig. 5A). In contrast, depletion of either DUB had the opposite effect, leading to an increase in larger BRCA1 foci without affecting the total number of foci ( we depleted RAP80 and examined if this would restore HR proficiency in USP26 or USP37-depleted cells. Indeed, we found that defective IR-induced accrual of both PALB2 and RAD51 in USP26-or USP37-depleted cells could be fully rescued by additional depletion of RAP80 (Fig. 6A,B). Similarly, HR efficiency was completely restored upon co-depletion of either DUB and RAP80, as measured in the DR-GFP reporter assay (Fig. 6C). Cell-cycle profiles in these cells remained unchanged ruling out effects of cell cycle misregulation (Supplemental Fig. 5E). Together these results suggest that USP26 and USP37 promote the BRCA1-dependent loading of PALB2 and RAD51 by counteracting the repressive impact of RAP80-dependent BRCA1 sequestration during HR (Fig. 7).
USP26 and USP37 regulate HR by counteracting RAP80
In this study, we identify USP26 and USP37 as novel DUBs that reverse RNF168-mediated ubiquitylation ( Fig. 1-2, Supplemental Fig. 1-2), a process known to repress HR by sequestering the BRCA1-Abraxas-RAP80-MERIT40 complex through its ubiquitin-binding subunit RAP80 [38,40]. By removing RNF168-induced ubiquitin conjugates distal from DSBs, USP26 and USP37 prevent the RAP80-dependent assembly of this BRCA1-containing complex, allowing BRCA1 to function in the BRCA1-PALB2-BRCA2-RAD51 complex during HR (Fig. 7). These findings advance our conceptual understanding of the RNF168-dependent response to DSBs, by revealing pathways that differentially regulate the spatial assembly and function of HR complexes at DSBs.
We propose the following model for RNF168-dependent regulation of HR ( Fig. 7): RNF168-induced ubiquitin conjugates spread away from DSBs into more distal chromatin regions [20]. The BRCA1-Abraxas-RAP80-MERIT40 complex through RAP80 interactions associates with the RNF168-induced ubiquitin conjugates in these regions, thereby sequestering BRCA1 from the ssDNA compartment and inhibiting HR [38,40]. In line with this model, we demonstrate that supra-physiological levels of RNF168 triggered extensive ubiquitylation of H2A (Fig. 2E), concomitant with a substantial reduction in HR efficiency (Fig. 4B). We extend these findings by showing that this phenomenon is actively antagonised by USP26 and USP37. Loss of USP26/ USP37 function markedly impairs the assembly of PALB2, RAD51 and efficient HR (Fig. 3C, 4A and 6A-C). However, these defects can be rescued 3 Chapter 3 by the additional loss of RAP80 (Fig. 6A-C). Together, these data suggest that USP26/37 limit the magnitude of the BRCA1-Abraxas-RAP80-MERIT40 complex assembly in DSB-neighbouring chromatin, by actively removing RNF168-mediated H2A ubiquitylation ( Fig. 2E and Supplemental Fig. 3).
Indeed, depletion of either DUB resulted in an increase in the size of BRCA1 foci, indicative of more extensive spreading of the BRCA1-Abraxas-RAP80-MERIT40 complex from the DSB site, which could be rescued by additional loss of RAP80 (Fig. 5A,B and Supplemental Fig. 5D). This scenario explains how these DUBs limit the repressive impact of RAP80 on HR. An alternative, yet not mutually exclusive scenario, would be in line with recent findings showing that DSB-induced H2A/H2AX ubiquitylation needs to be reversed in the core of IRIF for DNA end-resection to occur [39]. Given that USP26 and USP37 are able to reverse RNF8/168-mediated ubiquitylation and promote HR, these enzymes would be ideal candidates to facilitate such events. 7. Molecular model for the role of USP26 and USP37 in HR. BRCA1 is sequestered from HR sites through RAP80, which is functionally antagonised by USP26-and USP37dependent de-ubiquitylation of RNF168-modified chromatin (see discussion for details). Loss of USP26 or USP37 leads to more extensive RNF168-dependent sequestration of BRCA1, thereby preventing BRCA1 to form a complex with PALB2-BRCA2-RAD51 in HR. Additionally, the more extensive spreading of RAP80 upon USP26 or USP37 depletion reduces DNA endresection, which also impairs HR.  harbouring a destabilisation domain (DD) were previously described [7,15,43,58]. The ViraPower system (Life Science) was used to produce lentivirus using mAG-or mCherry-geminin expression vectors [59].
U2OS cells stably expressing mAG-or mCherry-geminin were made by standard lentiviral transduction, followed by FACS sorting, in order to select homogeneously fluorescent cells.

Cell survival assay
VH10-SV40 or U2OS cells were transfected with siRNAs, trypsinised, seeded at low density and exposed to IR. 7 days later cells were washed with 0.9% NaCl and stained with methylene blue. Colonies of more than 10 cells were scored.

UV-A laser micro-irradiation
U2OS cells were grown on 18 mm coverslips and sensitised with 10 μM 5′-bromo-2-deoxyuridine (BrdU) for 24 hours as described [18,44] Leica). Confocal images were recorded before and after laser irradiation at 5 or 10 seconds time intervals over a period of 5 -10 minutes.

Microscopy analysis
Images of fixed cells 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. Fluorescent probes were detected using previously described filters [61]. Images were recorded using ZEN 2012 software and analysed using ImageJ. The average reflects the quantification of 50-150 cells from 3 independent experiments.

IRIF analysis
PALB2, RAD51, BRCA1 and RAP80 (except in Supplemental Fig. 2B) IRIF were analyzed in U2OS cells 6hr after 10Gy, RPA and CtIP IRIF were assayed 4hr after 10Gy, whereas conjugated ubiquitin (FK2), γΗ2ΑΧ, MDC1, FLAG-RNF8, RNF168, BRCA1 and RAP80 (at Supplemental Fig.  USP26 and USP37 regulate HR by counteracting RAP80 2B) and 53BP1 IRIF were examined 1hr after 2Gy, unless stated otherwise. IRIF were evaluated in ImageJ, using a custom-built macro that enabled automatic and objective analysis of the foci. Full details of this macro will be published elsewhere. In brief, cell nuclei were detected by thresholding the (median-filtered) DAPI signal, after which touching nuclei were separated by a watershed operation. The foci signal was background-subtracted using a Difference of Gaussians filter. For every nucleus, foci were identified as regions of adjacent pixels satisfying the following criteria: (i) the grey value exceeds the nuclear background signal by a set number of times (typically 2-4x) the median background standard deviation of all nuclei in the image, and is higher than a user-defined absolute minimum value; (ii) the area is larger than a defined area (typically 2 pixels). These parameters were optimised for every experiment by manually comparing the detected foci with the original signal.

Immunofluorescent labelling
Immunofluoresecent labelling was carried out as described previously [18,19]. Briefly, cells were grown on glass coverslips and treated as indicated in Samples were incubated with 0.1 μg/ml DAPI and mounted in Polymount.

Primary antibodies and secondary antibodies are listed in the Supplemental
Table.

Western blotting
Cell extracts were generated by boiling cell pellets in Laemmli buffer, separated by SDS-PAGE and transferred to PVDF membranes (Millipore).
Membranes were probed with the antibodies listed in Supplemental Table 1 3 Chapter 3 followed by protein detection using the Odyssey infrared imaging scanning system (LI-COR Biosciences).

HR and NHEJ assay
HEK293 and U2OS cells containing a stably integrated copy of either the DR-GFP of EJ5-GFP reporter were used to measure the repair of I-SceI-induced DSBs by HR or NHEJ, respectively [48,62].

RT-qPCR-based gene expression analysis
RNA isolation, reverse transcription (RT)-based cDNA synthesis and quantitative (q)PCR were carried out as previously described [61]. The primers used are listed in the Supplemental Table. USP26 and USP37 regulate HR by counteracting RAP80