TOPBP1 regulates RAD51 phosphorylation and chromatin loading and determines PARP inhibitor sensitivity

Synthetic lethality is a genetic concept whereby a combination or synthesis of mutations in multiple genes results in cell death, whereas inactivation of single genes does not affect cell viability. This concept has been exploited in cancer treatment with promising clinical results. Indeed, cancer patients with BRCA1 or BRCA2 gene mutations benefit from treatment with a poly(ADP-ribose) polymerase (PARP) inhibitor (PARPi; Lord et al., 2015). PARP1/2i olaparib has been recently approved for treatment of ovarian cancer patients with BRCA1/2 defects in Europe and the United States.

PARP1 plays an important role in DNA repair, especially in repair of DNA single-strand breaks via base excision repair. On DNA damage, PARP1 binds DNA via its N-terminal zinc finger motifs, accumulates at DNA damage sites, and regulates accumulation of DNA repair proteins by generation of PAR chains (Luo and Kraus, 2012). Because of negative charge of PAR polymers, autoPARylation of PARP1 itself eventually causes its dissociation from DNA. A recent model suggests that olaparib and other PARPis trap PARP1 at DNA and prevent its release (Murai et al., 2012), thereby creating obstacles for replication forks. The observation that stalled replication forks require functional homologous recombination (HR) for restart likely explains the synthetic lethality interaction between BRCA1/2 genes and PARPi. In addition to BRCA1 and BRCA2 genes (Bryant et al., 2005; Farmer et al., 2005), several other PARPi sensitivity-causing DNA damage response (DDR) defects, in several DDR kinases and repair proteins, have been reported (Lord et al., 2015).

There are two major pathways for DNA double-strand breaks (DSBs) repair: nonhomologous end joining and HR, which, unlike nonhomologous end joining, requires sister chromatid and therefore is restricted to S and G2 phases of the cell cycle. HR starts with 5′ to 3′ resection of DNA ends that generates single-stranded DNA (ssDNA) ends. The ssDNA is rapidly coated by replication protein A (RPA), which is then replaced by RAD51 (Jackson and Bartek, 2009). RAD51 filaments promote DNA strand invasion and ensue HR. Although a BRCA1–PALB2–BRCA2 complex promotes RAD51 loading on chromatin (Sy et al., 2009), regulation and additional factors involved in RAD51 chromatin loading are incompletely understood.

Topoisomerase IIb–binding protein 1 (TOPBP1) was initially identified as a factor interacting with C-terminal region of DNA topoisomerase IIβ (Yamane et al., 1997). TOPBP1 is a large, nine BRCT domain–containing protein with essential roles in cellular processes, including DNA repair, replication, and transcription (Sokka et al., 2010). TOPBP1 enhances ATR kinase activity (Kumagai et al., 2006) through interaction with ATR partner protein ATRIP (Mordes et al., 2008). Ectopic expression of the ATR-activation domain (AAD) of TOPBP1 is sufficient to activate ATR in the absence of DNA damage and leads to cell cycle arrest (Toledo et al., 2008). TOPBP1 does not possess any known enzymatic activity; it rather serves as a scaffold protein for numerous interacting proteins that bind to its BRCT domains. Although TOPBP1 contributes to DNA repair and was suggested to be involved in HR (Morishima et al., 2007), any mechanistic insights into TOPBP1’s functions in DNA repair are missing. Here we report on a mechanism through which TOPBP1 regulates HR and impacts PARPi sensitivity.

To identify factors that mediate sensitivity to PARPi olaparib, we performed a high-content RNAi screen in human osteosarcoma cell line U2OS (Frankum et al., 2015). Among other hits, we identified TOPBP1 as a candidate protein whose depletion enhanced the toxic effect of PARPi. These results suggested that loss of TOPBP1 could sensitize tumor cells to PARPi and that loss or inactivation of TOPBP1 could predict response to this class of agents. To validate the screen data we first used an independent siRNA to assess induction of micronuclei and DNA damage in TOPBP1-depleted cells exposed to olaparib for 3 d. TOPBP1 siRNA combined with olaparib caused micronuclei formation and increased the level of a DNA damage marker, histone H2A.X phosphorylated at serine 139 (γH2AX), to an extent equivalent to that achieved by a siRNA targeting BRCA2 (Fig. 1, A–C). Next, we tested olaparib sensitivity of U2OS and CAL51 (human breast cancer) cells using a clonogenic assay, exposing cells to olaparib for 2 wk. We observed in both cell lines synergistic reduction of colonies in olaparib-treated, TOPBP1-depleted cells at levels of suppression similar to that caused by siBRCA2 (Fig. 1 D and Fig. S1, A–C). Analysis of whole cell lysates (WCLs) from TOPBP1-depleted cells treated with olaparib showed synergistic induction of DNA DSB markers: activated CHK2 kinase (pCHK2 T68) and RPA32 phosphorylated at serines 4/8 and threonine 21, which were mostly absent in TOPBP1-depleted cells or control cells treated with olaparib (Fig. 1 E). As expected, olaparib treatment induced RPA32 phosphorylation at serine 33 and the activatory ATR-mediated phosphorylation of CHK1 at serine 317, the latter modification being impaired in the TOPBP1-depleted cells because TOPBP1 is required for ATR activation. These results indicated that TOPBP1 silencing causes olaparib hypersensitivity and raised a question of potential TOPBP1 aberrations in human tumor types most relevant to treatment with PARPi. Strikingly, although ovarian cancer is the only type of cancer so far approved for clinical treatment by olaparib, there has been no report on TOPBP1 protein expression in this tumor. Here we have performed an immunohistochemical analysis of 136 human ovarian carcinomas. Although most tumors showed normal levels of TOPBP1 comparable with a normal ovary, 8 and 10 carcinomas showed aberrant reduction and overabundance of TOPBP1 protein, respectively (Fig. S1 D). The rather modest frequencies of TOPBP1 aberrations in our cohort are reminiscent of studies of breast cancer (Going et al., 2007; Liu et al., 2009) and likely reflect the fact that TOPBP1 gene mutations are more common in some other types of cancer (https://www.intogen.org/search?gene=TOPBP1).

Because the known synthetic lethal interactions with PARPi commonly involve HR pathway genes (Lord et al., 2015) and we found elevated DSB markers in TOPBP1-depleted cells treated with olaparib (Fig. 1, C and E), we next examined whether olaparib sensitivity could reflect defective HR. As a surrogate of HR activity, we quantified formation of RAD51 ionizing radiation (IR)–induced foci (IRIF) in cyclin A–positive (S/G2-phase) cells. Indeed, TOPBP1-depleted U2OS cells showed a pronounced reduction in this RAD51 IRIF assay (Fig. 2, A–C). Abrogation of RAD51 IRIF was not caused by reduced RAD51 protein levels (Fig. 2 D) and was statistically significant at 2, 4, 10, and 15 h after IR (Fig. 2 B). Three TOPBP1 siRNAs, which efficiently reduced TOPBP1 protein levels, recapitulated RAD51 IRIF abrogation without affecting RAD51 abundance (Fig. S2, A and B). Depletion of TOPBP1 also impacted RAD51 foci formation on DSB induction by camptothecin (CPT; Fig. S2 C). There were no marked changes in cell cycle distribution on TOPBP1 depletion (Fig. S2 E), and a similar percentage of TOPBP1-depleted cells accumulated in mitosis after 4-h treatment with nocodazole (Fig. S2 F), indicating that cell cycle indirect effects were unlikely to account for our results. We then directly assessed HR activity in the TOPBP1-silenced cells using the traffic light reporter system (Certo et al., 2011). Consistent with Morishima et al. (2007), TOPBP1-depleted cells showed reduced HR activity (Fig. S2 D). To gain more mechanistic insight into TOPBP1’s role in RAD51 foci formation, we examined which TOPBP1 domains are required for formation of RAD51 IRIF. TOPBP1 contains nine BRCT domains (Rappas et al., 2011) and an ATR-activation domain (Kumagai et al., 2006). First, we generated three large RNAi-resistant GFP-tagged TOPBP1 deletion mutants (Fig. 2 E) in which N-terminal, central, or C-terminal parts were deleted and expressed those mutants to examine their ability to rescue the impaired RAD51 IRIF formation in cells depleted of endogenous TOPBP1. Whereas TOPBP1 full-length and TOPBP1 ΔN and Δcentral complemented RAD51 IRIF formation, TOPBP1 ΔC mutant was unable to rescue RAD51 IRIF formation (Fig. 2 F), suggesting that TOPBP1 domains within the C terminus are required for RAD51 IRIF. Second, we generated another set of RNAi-resistant GFP-tagged TOPBP1 mutants, including deletions of AAD, BRCT7, BRCT8, and BRCT domains 7/8. Among these, only the TOPBP1 ΔAAD could rescue RAD51 IRIF formation in TOPBP1-depleted cells (Fig. 2 F). Furthermore, expression of the BRCT domains 7/8–containing fragment alone did not rescue RAD51 IRIF in TOPBP1-depleted cells (Fig. S3 A). These rescue experiments indicated that both C-terminal BRCT domains 7/8 are essential, but not sufficient, for formation of RAD51 IRIF.

To further characterize the role of BRCT domains 7/8 in regulation of RAD51 IRIF formation, we exploited a recently identified small molecular compound, calcein AM (CalAM), that avidly binds to BRCT domains 7/8 (Chowdhury et al., 2014) and thereby alters their engagement in protein–protein interactions, including TOPBP1 oligomerization (Liu et al., 2013). Indeed, 2-h pretreatment of U2OS cells with CalAM resulted in reduced formation of RAD51 foci, and 16-h pretreatment completely inhibited RAD51 IRIF (Fig. 2 G). If the TOPBP1 function targeted by CalAM is required for RAD51 IRIF and therefore HR, exposure to CalAM should sensitize cells to PARPi. Indeed, although low doses of olaparib or CalAM used separately only slightly reduced numbers of colonies in a clonogenic assay, combined olaparib and CalAM showed a more pronounced effect (Fig. 2 H). Control experiments with coexpressed GFP- and mCherry-tagged wild-type TOPBP1 in 293T cells, and biochemical analyses of endogenous TOPBP1 in U2OS cells by nondenaturing gel electrophoresis and Western blotting, showed altered TOPBP1 protein patterns on treatment with CalAM, but without a pronounced alteration of TOPBP1’s oligomerization status (unpublished data). In addition, although treatment with CalAM reduced the known interaction between BRCT domains 7/8 of endogenous TOPBP1 and the DNA repair protein BACH1 (Fig. S3 D), BACH1 knockdown had only modest effect on RAD51 IRIF (Fig. S3, B and C). Overall, although these results support the essential role of BRCT domains 7/8 in TOPBP1’s ability to facilitate RAD51 IRIF formation, the mechanistic basis of such function remained unclear.

In search for the underlying mechanism, we assessed the potential role of TOPBP1 in HR, a process that can be separated into several distinct phases, including DNA end resection and chromatin loading of RPA and RAD51, steps that can be visualized by formation of microscopically detectable foci. First, we compared the extent of ssDNA by detection of BrdU incorporation under nondenaturing conditions (a surrogate marker for DNA end resection) and formation of RPA, BRCA1, and RAD51 foci in irradiated control and TOPBP1-depleted cells. Although neither DNA end resection nor RPA loading showed deviation from control values, formation of RAD51 foci was impaired by TOPBP1 knockdown (Fig. 3, A and B). Using biochemical cell fractionation, we compared accumulation of RPA70 and RAD51 proteins in chromatin-enriched fractions (CHEFs) on IR. TOPBP1 knockdown abolished RAD51 accumulation in CHEFs, whereas RPA70 accumulated normally (Fig. 3 C). These results suggested that TOPBP1 acts at the level RAD51 chromatin loading. BRCA2 physically interacts with RAD51 (Chen et al., 1998) and is essential for RAD51 loading on chromatin (Yuan et al., 1999; Reuter et al., 2014; Shahid et al., 2014); hence, we assessed BRCA2 recruitment to damage sites. Because we were unable to detect BRCA2 IRIF with available antibodies against BRCA2, we tested recruitment of endogenous BRCA2 to DNA damage sites induced by laser microirradiation (Bekker-Jensen et al., 2006). BRCA2 localized to laser stripes independently of TOPBP1 (Fig. 3 D), whereas RAD51 recruitment was reduced in TOPBP1-depleted cells (Fig. S3 E). These results indicated that TOPBP1 functions in HR at the level of RAD51 chromatin loading, apparently without affecting BRCA2 function.

RAD51 is phosphorylated by PLK1 kinase at serine 14 in response to DNA damage and subsequently by CK2 at threonine 13, thereby facilitating RAD51 recruitment to DNA damage sites (Yata et al., 2012). We found that CPT treatment increased RAD51 phosphorylation at serine 14, and TOPBP1 knockdown prevented this modification (Fig. 4 A). Notably, compared with efficient chromatin accumulation of wild-type TOPBP1, a nonphosphorylatable mutant RAD51 protein in which serine 14 was replaced by alanine showed impaired chromatin accumulation (Fig. 4 B). Consistently, pretreatment of cells with CK2 or PLK1 inhibitors reduced formation of RAD51 IRIF (Fig. 4 C). Efficiency of PLK1 and CK2 inhibitors was tested on phosphorylations of S1790 of LRRK1 (Hanafusa et al., 2015) and S380 of PTEN (Martins et al., 2014), respectively (Fig. S3 F). Because RAD51 phosphorylation at serine 14 and threonine 13 facilitates RAD51 recruitment to IR-induced damage sites (Yata et al., 2012), absence of RAD51 phosphorylation at serine 14 could explain the observed abrogation of RAD51 chromatin loading and RAD51 IRIF formation in TOPBP1-depleted cells.

Given that TOPBP1 provides a scaffold support for various protein–protein interactions, we considered that TOPBP1 might interact with PLK1. Indeed, reciprocal coprecipitation experiments with antibodies to PLK1 and TOPBP1 revealed interactions between endogenous PLK1 and TOPBP1 in extracts from U2OS cells (Fig. 5, A and B). Reminiscent of the scenario reported for the PLK1–BRCA2 interaction (Yata et al., 2014), the TOPBP1-PLK1 complex was more abundant in nocodazole-treated, mitotic cells with a high abundance of active PLK1. Consistent with its functional significance for HR, the complex between endogenous TOPBP1 and PLK1 was abolished on pretreatment of cells with the BRCT domains 7/8–targeting CalAM, used at the concentration that prevented RAD51 IRIF formation (Figs. 5 C and 2 G).

In this study, through an unbiased siRNA screen, we identified and validated TOPBP1 protein as a novel determinant of PARPi sensitivity. We conclude that TOPBP1 is required for RAD51 loading on chromatin and formation of RAD51 IRIF, whereas the earliest steps of HR, DNA end resection and RPA, and BRCA1 and BRCA2 recruitment are TOPBP1 independent. Furthermore, TOPBP1 C-terminal BRCT domains 7/8 are essential for this novel role in HR. We also show that TOPBP1 physically interacts with PLK1 kinase. Collectively, we propose a model for a plausible scaffold role of TOPBP1 in promoting PLK1-mediated phosphorylation of RAD51 at serine 14, a prerequisite for efficient formation of RAD51 IRIF and activity of HR (Fig. 5 D). Although this novel function is important for HR-mediated DSB repair, it does not exclude the possibility that additional aspects of TOPBP1 function contribute to HR or other DNA repair pathways, for example in response to replication stress. Apart from contributing to our understanding of genome integrity maintenance through TOPBP1, our results may also inspire assessment of TOPBP1 as a candidate biomarker for targeted treatment of tumor subsets with PARPis.

Human U2OS, CAL51, 293T, and HeLa cell lines were grown in DMEM with 10% FBS and penicillin/streptomycin (Sigma-Aldrich). To generate stable isogenic U2OS cell lines with inducible expression of FLAG-Rad51 variants, Flp-In T-REx U2OS cells were cotransfected with pOG44 and pcDNA5/FRT/TO/FLAG-Rad51 vectors (Yata et al., 2012) and stable clones were selected with hygromycin.

Olaparib was provided by Astra Zeneca. PLK1 inhibitor volasertib and CK2 inhibitor silmitaserib were purchased from Selleckchem. CalAM was purchased from Sigma-Aldrich.

All siRNA transfections were performed using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer’s instructions. All siRNA duplexes were purchased from Ambion: siCON (negative control #1, AM4635, 5′-AGUACUGCUUACGAUACGGTT-3′), siTOPBP1#1 (s21823, 5′-GGAUAUAUCUUUGCGGUUUTT-3′), siTOPBP1#2 (s21824, 5′-GCAGAACUGUUGCGGAUUATT-3′), siTOPBP1#3 (s21825, 5′-GCUCUGUAAUAGUCGACUATT-3′), siBRCA2 (s2085, 5′-GGAUUAUACAUAUUUCGCATT-3′), siBACH1#1 (s38386, 5′-GAAUAACCCAAGUCGCUAUTT-3′), and siBACH1#2 (s38385, 5′-GACUAUCUUUUUAGGCAAATT-3′).

Plasmid transfections were performed using FuGENE 6 (Roche) according to the manufacturer’s instructions. To generate siRNA-insensitive GFP-TOPBP1* mutants, silent mutations were introduced into the siTOPBP1#3 target sequence in the TOPBP1 coding region of the pEGFP-C1-hTOPBP1 plasmid using the QuikChange II Site-Directed Mutagenesis kit (Stratagene).

The construct pSNV2-GFPB-YTA-NLS-hTOPBP1-1258-1522 for GFP-BRCT7/8 expression was provided by T. Halazonetis (University of Geneva, Geneva, Switzerland; Cescutti et al., 2010).

The construct for expression of GFP-TOPBP1 ΔN (deleted aa 1–519) was generated by double digestion of pEGFP-C1-hTopBP1 with restriction enzymes ApaI and Bpu1102I followed by blunting by mung bean nuclease (New England Biolabs) and ligation of the blunted ends.

The construct for expression of GFP-TOPBP1 Δcentral (deleted aa 526–1,000) was generated by double digestion of pEGFP-C1-hTopBP1 with restriction enzymes Bpu1102I and Eco32I followed by ligation with annealed oligos RTP41 (5′-TGAGCCCTTGAATGATTCTACT-3′) and RTP42 (5′-AGTAGAATCATTCAAGGGC-3′).

The construct for expression of GFP-TOPBP1 ΔC (deleted aa 1,001–1,522) was generated by double digestion of pEGFP-C1-hTopBP1 with restriction enzymes Eco32I and BamHI followed by blunting by mung bean nuclease treatment and ligation of the blunted ends. NLS was added to this construct by site-directed mutagenesis.

Small internal deletion mutants of GFP-TOPBP1, including ΔAAD (deleted aa 993–1,196), ΔBRCT7 (1,259–1,351), ΔBRCT8 (1,389–1,486), and ΔBRCT7/8 (1,259–1,486), were generated by site-directed mutagenesis. All prepared constructs were verified by sequencing.

We used the following rabbit antibodies: TOPBP1 (ab2402; Abcam), RAD51 (sc8349; Santa Cruz; ab63801; Abcam), pRAD51 S14 (Yata et al., 2012), cyclin A (NCL-CYCLINA; Leica), RPA70 (ab79398; Abcam), pRPA32 T21 (ab61065; Abcam), pRPA32 S4/8 (A300-245A; Bethyl), pRPA32 S33 (NB100-544; Novus), pCHK1 S317 (2344; Cell Signaling), pCHK2 T68 (2661; Cell Signaling), H3 (ab1791; Abcam), pH3 S10 (06–570; Millipore), BACH1 (sc-28738; Santa Cruz), pPTEN S380 (9551; Cell Signaling), pLRRK1 S1790 (H. Hanafusa, Nagoya University, Nagoya, Japan; Hanafusa et al., 2015); mouse antibodies: γH2AX (05–636; Millipore), cyclin A (sc-751; Santa Cruz), BrdU (RPN20AB; AP Biotech), RPA32 (ab2175; Abcam), BRCA1 (sc-6954; Santa Cruz), BRCA2 (OP95; Millipore), CHK1 (sc-8408; Santa Cruz), Importin-β (ab2811; Abcam), Lamin A/C (sc-7292; Santa Cruz), α-tubulin (sc-8035; Santa Cruz), PLK1 (05–844; Millipore; 331700; Zymed); and goat antibody: γ-tubulin (sc-7396; Santa Cruz).

Cells were transfected with siRNAs; 24 h later, they were seeded to 6-cm-diameter dishes; and after the next 24 h, they were treated with olaparib or vehicle. Colonies were grown for 10 d, fixed in 70% ethanol, and stained with 1% crystal violet in ethanol. Colonies of >50 cells were counted, and the surviving fractions were calculated.

Cells cultured on glass coverslips were fixed in 4% formaldehyde for 15 min at RT, permeabilized for 5 min with 0.2% (vol/vol) Triton X-100 in PBS, washed in PBS, and incubated with primary antibodies for 60 min at RT. After the washing step, the coverslips were incubated with goat anti–rabbit or goat anti–mouse Alexa Fluor 488, 568, or 647 secondary antibodies (Invitrogen) for 60 min at RT, washed with PBS, and mounted using Vectashield mounting reagent with DAPI (Vector Laboratories). ssDNA was detected by BrdU antibody under nondenaturing conditions after 48-h incubation of cells in culture medium supplemented with 5 µM BrdU. For the RAD51 IRIF assay, sc8349 Ab (Santa Cruz) was used initially, and after the next batch failed completely, it was replaced by ab63801 (Abcam), accounting for different starting basal values of RAD51 in Figs. 2 B and 3 B versus Figs. 2 (F and G) and 4 C.

Microscope images were acquired using a confocal microscope (LSM510; Carl Zeiss) mounted on inverted microscope (Axiovert 100M; Carl Zeiss) equipped with a Plan-Apochromat 63×/1.4 oil immersion objective (Carl Zeiss) at RT. Image acquisition and analysis was performed using LSM ZEN software (Carl Zeiss). Dual-color confocal images were acquired using laser lines 488 and 543 nm for excitation of Alexa Fluor 488 and 568 dyes (Invitrogen), respectively. Automated multichannel image acquisition was performed using high-content screening station scan^R (Olympus) equipped with motorized microscope (IX81; Olympus), UPlanSApo 40×/0.95 air immersion objective (Olympus), and digital monochrome electron multiplying charge coupled device camera (C9100; Hamamatsu). Image acquisition and analysis was performed using scan^R acquisition and analysis software (Olympus).

Cells grown on glass coverslips were presensitized with 10 µM Hoechst 34580 (Life Technologies) for 5 min. For microirradiation, the cells were placed in a LabTek chamber and mounted on the stage of a custom-designed PALM MicroBeam with a 335-nm UV-A pulsed laser (Carl Zeiss).

The cohort (n = 136) of formalin-fixed, paraffin-embedded human ovarian carcinomas (Department of Pathology, University Hospital in Las Palmas) included the following histopathological subtypes of tumors: serous (n = 61), mucinous (n = 17), clear-cell (n = 22), endometrioid (n = 22), mixed type (n = 5), and undifferentiated (n = 9). For a sensitive immunohistochemistry procedure without nuclear counterstaining, the slides were deparaffinized, subject to antigen retrieval (15 min heating with the Target retrieval solution, pH 9, code S2367; Dako), and incubated with the primary rabbit antibody against human TOPBP1 (ab2402; diluted 1:4,000; Abcam) overnight, followed by the indirect streptavidin–biotin–peroxidase method using the Vectastain Elite kit (Vector Laboratories) and nickel sulfate-based chromogen enhancement detection. Normal rabbit serum served as a negative control. Staining patterns were evaluated by an experienced oncopathologist; TOPBP1 protein abundance was compared with sections of normal ovary and surrounding normal (stromal) cells within each tumor section (as an internal control) and was categorized as normal, aberrantly reduced, or overexpressed.

WCLs were prepared in Laemmli sample buffer (LSB; 50 mM Tris, pH 6.8, 100 mM DTT, 2% SDS, 0.1% bromophenol blue, and 10% glycerol), separated by SDS-PAGE, and transferred to nitrocellulose membranes (GE Healthcare). The membranes were blocked with 5% (wt/vol) dry milk in 0.1% (vol/vol) Tween-20 in PBS and probed with the primary antibodies, followed by HRP-labeled secondary antibodies (Vector Laboratories and Santa Cruz), and visualized using ECL detection reagents (GE Healthcare).

Cells were washed three times by PBS, and soluble proteins were removed by incubation of cells with 0.5% Triton X-100 in PBS for 5 min on ice. The remaining pellet was washed by 0.5% Triton X-100 in PBS and resuspended in 2× LSB.

To prepare lysates for immunoprecipitation, cells were washed three times in PBS and lysed in TNE buffer (150 mM NaCl, 50 mM Tris-HCl, pH 8.0, 1 mM EDTA, 0.5% NP-40) supplemented with cOmplete and PhosSTOP tablets (Roche). After 30 min incubation on ice, lysates were cleared by centrifugation. Where appropriate, antibodies were added to lysate and incubated for 16 h at 4°C. Lysates were then incubated with 25 µl of either Dynabeads Protein G (Novex) or Dynabeads M-280 Sheep anti Rabbit IgG (Novex) for 1 h at 4°C. Ig–antigen complexes were washed extensively before elution in 2× LSB before SDS-PAGE.

Cells were harvested by trypsinization, fixed in 70% ethanol, and resuspended in propidium iodide buffer Facsflow (BD). Samples were incubated for 30 min at 37°C before analysis. Cell cycle analysis was performed using flow cytometer FACSCalibur (BD). Mitotic entry was examined by staining with primary antibody for pH3 S10 followed by FITC-conjugated secondary antibody.

Fig. S1 shows sensitivity of TOPBP1-depleted CAL51 cells to olaparib and efficiency of siRNA-mediated knockdowns. Fig. S2 shows effects of multiple TOPBP1 siRNAs on RAD51 foci formation and documents effects of TOPBP1 knockdown on cell cycle. Fig. S3 shows RAD51 IRIF after complementation of TOPBP1-depleted cells with the BRCT domains 7/8 construct, effect of BACH1 knockdown on RAD51 IRIF, impact of CalAM on TOPBP1-BACH1 interaction, and impaired accumulation of RAD51 on microirradiated regions in TOPBP1-depleted cells.

We thank E. Dulina and V. Turcanova for technical assistance, Dr. Hanafusa for the pLRKK1 S1790 antibody, and Dr. Halazonetis for the pSNV2-GFPB-YTA-NLS-hTOPBP1-1258-1522 construct.

This study was supported by grants from the European Community’s Seventh Framework Programme (project DDResponse), the Kræftens Bekæmpelse, the Lundbeckfonden (R93-A8990), the Sundhed og Sygdom, Det Frie Forskningsråd (DFF-1331-00262), the Novo Nordisk UK Research Foundation (NNF14CC0001 to J. Lukas and NN13331 to J. Bartek), the Danmarks Grundforskningsfond (Center of Excellence: CARD), the Norway Grants (project PHOSCAN, no. 7F14061), the Vetenskapsrådet, the Kellner Familly Foundation, and Intra-European Fellowship (FP7-PEOPLE-2011-IEF, no. 303269 to K.M. Wolanin).

The authors declare no competing financial interests.