The ubiquitin-dependent ATPase p97 removes cytotoxic trapped PARP1 from chromatin

Poly (ADP-ribose) polymerase (PARP) inhibitors elicit antitumour activity in homologous recombination-defective cancers by trapping PARP1 in a chromatin-bound state. How cells process trapped PARP1 remains unclear. Using wild-type and a trapping-deficient PARP1 mutant combined with rapid immunoprecipitation mass spectrometry of endogenous proteins and Apex2 proximity labelling, we delineated mass spectrometry-based interactomes of trapped and non-trapped PARP1. These analyses identified an interaction between trapped PARP1 and the ubiquitin-regulated p97 ATPase/segregase. We found that following trapping, PARP1 is SUMOylated by PIAS4 and subsequently ubiquitylated by the SUMO-targeted E3 ubiquitin ligase RNF4, events that promote recruitment of p97 and removal of trapped PARP1 from chromatin. Small-molecule p97-complex inhibitors, including a metabolite of the clinically used drug disulfiram (CuET), prolonged PARP1 trapping and enhanced PARP inhibitor-induced cytotoxicity in homologous recombination-defective tumour cells and patient-derived tumour organoids. Together, these results suggest that p97 ATPase plays a key role in the processing of trapped PARP1 and the response of tumour cells to PARP inhibitors. Krastev et al. report that trapped PARP1 undergoes SUMOylation, followed by ubiquitylation, resulting in the recruitment of the p97 ATPase to remove trapped PARP1 from chromatin and prevent PARP inhibitor-induced cytotoxicity.

P oly (ADP-ribose)-polymerase (PARP) inhibitors (PARPi) selectively kill tumour cells with impaired homologous recombination and are approved for use in homologous recombination-defective breast, ovarian, pancreatic and prostate cancers 1 . PARP1 (also known as ARTD1), the key target of PARPi, is a ubiquitously expressed nuclear enzyme that uses NAD + to synthesise poly(ADP-ribose) (PAR) chains on substrate proteins (heteromodification) and itself (automodification). This catalytic activity (PARylation), which is enhanced by PARP1 binding to damaged DNA, initiates DNA repair by driving the recruitment/concentration of DNA-repair effectors and modulating chromatin structure. Once DNA repair is initiated, PARP1 is released from DNA via auto-PARylation. Most clinical PARPi bind the NAD + -binding site (catalytic domain) and inhibit catalytic activity, but also induce chromatin retention of PARP1 (PARP trapping); this latter characteristic is a key driver of PARPi-mediated cytotoxicity 2 . Consistent with this, deletion of PARP1 causes PARPi resistance, as do in-frame PARP1 insertion/deletion mutations that impair PARP1 trapping 3 . Moreover, the chemical modification of a PARPi with poor trapping properties into a derivative with enhanced trapping properties, but similar catalytic potency, enhances cytotoxicity 4 . Although it is known that specific PARP1 mutations alter PARP1 trapping 3 , as does modulation of the amount of residual PAR on PARP1 via PAR-glycosylase 5 , how trapped PARP1 is released from damaged DNA is poorly understood.
By generating a series of protein-protein interaction profiles of either trapped or non-trapped PARP1, we show that trapped PARP1 binds p97 ATPase (also known as valosin-containing protein, VCP). p97 ATPase is a hexameric unfoldase/segregase that unfolds and disassembles ubiquitylated substrates through its central pore 6,7 including Aurora B kinase, CMG helicases, the licensing factor CDT1 8 and the TOP1-cleavage complex [9][10][11] . We show that the PARP1-p97 interaction is mediated by sequential PIAS4-mediated SUMOylation and RNF4-mediated ubiquitylation of trapped PARP1. Ufd1-mediated p97 recruitment to trapped and modified PARP1 ultimately leads to the removal of PARP1 from chromatin. In addition, we show that p97 inhibition by a metabolite of the clinically used drug disulfiram, leads to prolonged PARP1 trapping and profound PARPi sensitivity, suggesting an approach to enhance PARPi-induced cytotoxicity. Collectively, our findings suggest that the PARP1-p97 axis is essential for the removal of trapped PARP1 and the cellular response to PARPi.

Identification of trapped PARP1-associated proteins.
To understand the nature of the trapped PARP1 complex, we used two orthogonal systems to generate mass spectrometry-based PARP1 protein-protein interactomes from cells with either trapped or non-trapped PARP1: (1) rapid immunoprecipitation mass spectrometry of endogenous proteins (RIME) 12

and (2) in vivo Apex2
The ubiquitin-dependent ATPase p97 removes cytotoxic trapped PARP1 from chromatin Articles NATURE CEll BIOlOgy peroxidase-mediated labelling of proximal proteins 13 . We used a PARPi-resistant PARP1-defective cell line 3 , CAL51 PARP1 -/-, into which we introduced either wild-type PARP1 (PARP1 WT ) or a trapping-deficient PARP1 del.p.119K120S transgene 3 (Fig. 1a) fused to enhanced green fluorescent protein (eGFP) for RIME or Apex2-eGFP for proximity labelling. We established single-cell clones that expressed the desired PARP1 fusion proteins (Extended Data Fig. 1a). Validating these transgenes, we found that expression of either PARP1 WT -eGFP or PARP1 WT -Apex2-GFP proteins re-established PARPi sensitivity in PARP1 -/cells (Fig. 1b,c), whereas expression of the PARP1 del.p.119K120S -eGFP did not. We also used a PAR-binding PBZ-mRuby2 probe and an ultraviolet light (UV) micro-irradiation assay to demonstrate that PARP1 WT -Apex2-eGFP localized to DNA-damage sites where it generated PAR. In the presence of PARPi, PARP1 was retained at the site of damage (Fig. 1d).
As PARP1 translocates to chromatin following DNA damage, we first used RIME-based immunoprecipitation 12,14 to identify proteins associated with trapped PARP1 (Fig. 1a). Cells expressing PARP1 WT -eGFP or PARP1 del.p.119K120S -eGFP were exposed to PARP1-trapping conditions (methyl methanesulfonate (MMS) + talazoparib), after which the protein interactions were stabilized by formaldehyde crosslinking. Talazoparib traps DNA-bound PARP1, whereas MMS was used to create DNA lesions. After trapping, the chromatin-bound proteins were isolated and the PARP1-associated complexes were immunoprecipitated using GFP-Trap beads. The proteins were then identified by mass spectrometry. We also included an analysis of the parental PARP1 -/cells lacking eGFP to identify proteins that bind non-specifically to GFP-Trap beads (Extended Data Fig. 1b). The non-specific bead-binding proteins were removed from the list of identified proteins (see Methods). As a result, we identified 50 PARP1-associated proteins in cells expressing wild-type PARP1 (either in the presence or absence of PARPi; Supplementary Table 1) and 144 PARP1-associated proteins in cells expressing PARP1 del.p.119K120S -eGFP (Supplementary Table 2). In both datasets, PARP1 was by far the most abundant protein identified by its mass spectrometry (MS) score (Fig. 1e,f). To prioritize proteins for further analysis, we used the MS score and the enrichment ratio of peptide spectrum matches (PSM) in the cells exposed to MMS + talazoparib compared with cells exposed to MMS alone. As expected, MMS + talazoparib increased the PARP1 PSM enrichment ratio in cells expressing PARP1 WT -eGFP but not in those expressing PARP1 del.p.119K120S -eGFP (2.5 versus 1.1, respectively). The trapping-defective PARP1 del.p.119K120S -eGFP mutant seemed to interact with the cohesion complex subunit PDS5A, regardless of the presence of PARPi (Fig. 1f), suggesting that some interaction between mutant PARP1 and chromatin did exist independently from trapping. Both mutant and wild-type PARP1 also seemed to interact with chromatin-associated proteins (for example, CHD4), but when compared with PARP1 del.p.119K120S -eGFP, PARP1 WT -eGFP showed a relative enrichment of the small ubiquitin modi-fier proteins SUMO1 (PSM ratio of five in PARP1 WT versus one in PARP1 del.p.119K120S ) and SUMO2 (PSM ratio of four in PARP1 WT versus unidentified in PARP1 del.p.119K120S ; Fig. 1e).
As an orthogonal MS approach, we employed Apex2-mediated proximity labelling. Western blotting confirmed PARP1 WT -Apex2-eGFP biotinylation in the presence of biotin-phenol (Extended Data Fig. 1c), which was further increased when PARP1 labelling was conducted under trapping conditions (Extended Data Fig. 1c). Biotinylated proteins were purified under stringent conditions and analysed by mass spectrometry. A caveat to our work was our inability to generate a trapping-defective PARP1 fused to Apex2-eGFP; this prevented us from using this control in the proximity labelling. We used cells expressing PARP1 WT -eGFP instead to filter out non-specific interactions with beads. As a result of this filtering, we identified a higher number of proteins-that is, 360-associated with PARP1 in our proximity labelling analysis than for RIME (either in the presence or absence of PARPi; Supplementary Table 3). A STRING network analysis, using a high-stringency cutoff (0.7) representing the trapped PARP1 interactome network (Extended Data Fig. 2a), was enriched in proteins associated with PARP1-mediated base-excision repair (PCNA, HMGB1, LIG3, PARP1 and POLE; P < 0.01; Extended Data Fig. 2a,b), giving us high confidence in the analysis. Gene-set ontology analysis also identified an enrichment in proteins involved in the spliceosome and ribosome biogenesis (Supplementary Table 4). We also identified a number of well-characterized PARylation targets (for example, PCNA, NCL, FUS and ILF3; refs. 15,16 ), strengthening the idea that we identified bona fide PARP1-proximal proteins. Notably, 'protein processing in endoplasmic reticulum' (P < 10 −3 ) and 'proteasome' (P < 0.01) seemed to be enriched in the gene-set ontology analysis, observations we focus on later in this manuscript. The MS and PSM scores showed a positive correlation and identified that PARP1, p97, UBA1 and TOP2A were among the most abundant proteins (Fig. 1g,h). Proteins that showed high enrichment ratios in PARP1-trapping conditions-for example, USP7-were generally identified with a low MS score pointing to a low abundance. As a trapping-deficient mutant was not available to perform a comparison similar to the RIME analysis, for the Apex2 analysis, we prioritized the high MS score over the high PSM ratio in our further considerations as it would represent higher stoichiometric interactions at DNA-damage sites. Among the most abundant labelled proteins were ubiquitin-like modifier-activating enzyme 1 (UBA1), which has been previously implicated in ubiquitylation events at the sites of DNA damage 17 , and the transitional endoplasmic reticulum ATPase p97, which acts as a central component of a ubiquitin-controlled process. The ATP-dependent unfoldase activity of p97 extracts proteins from chromatin before their proteasomal degradation or recycling [9][10][11][18][19][20] . Furthermore, p97, working with cofactors that often contain ubiquitin-binding domains, recognises client proteins via ubiquitylation events, mostly those involving ubiquitylation of the K48 and K6 residues 21,22 . In addition, p97 was identified in the PARP1 WT , but Fig. 1 | identification of trapped PARP1-interacting proteins. a, Schematic describing the identification of trapped PARP1 protein-protein interactomes via RIME or proximity labelling linked to mass spectrometry. The cells were exposed to either PARPi + MMS (to enable trapping) or MMS (no trapping) for 1 h, after which PARP1-interacting/proximal proteins were identified by mass spectrometry analysis. BP, biotin-phenol. b, Clonogenic assay illustrating the restoration of PARPi sensitivity in the complemented PARP1 -/-CAL51 cells as described in a. PARP1 protein expression in the different clones is shown in Extended Data Fig. 1a. c, Quantification of the colony formation assay shown in b; data are the mean of two biological replicates. d, PARP1 WT -Apex2-eGFP protein localizes to sites of DNA damage, generates PAR and can be trapped by PARPi. Cells expressing PARP1 WT -Apex2-eGFP were transfected with a PAR sensor, a PBZ PAR-binding domain fused to mRuby2 (left). PARP1 WT -Apex2-eGFP and PBZ-mRuby2 accumulate at the sites of UV micro-irradiation. Exposure to 100 nM talazoparib causes sustained accumulation of PARP1 WT -Apex2-eGFP (middle) but abolishes PAR production (right). Data represent two independent experiments with similar results. DMSO, dimethylsulfoxide. e,f, PARP1 interactions that are enriched under PARP1-trapping conditions (as defined by the PSM ratio ((talazoparib + MMS) ÷ MMS) and MS scores). Scatter plots are shown for PARP1 WT -eGFP (e) and PARP1 del.p.119K120S -eGFP (f) RIME. g, PARP1 interactions that are enriched under PARP1-trapping conditions for PARP1 WT -Apex2-eGFP proximity labelling. RIME and proximity labelling were performed in three independent experiments. h, A graph plotting the PSM against MS score for PARP1 WT -Apex2-eGFP proximity labelling interactions shows that p97 is among the most abundant proteins identified in the PARP1 WT -Apex2-eGFP proximity labelling.

NATURE CEll BIOlOgy
not in the PARP1 del.p.119K120S , RIME analysis, strengthening the idea that it may interact with trapped PARP1.

Trapped PARP1 is sequentially SUMOylated and ubiquitylated.
Our RIME analysis suggested that trapped PARP1 was associated with SUMO1 and SUMO2, whereas our proximity labelling analysis identified ubiquitylation and p97 to be associated with trapped PARP1. This raised the hypothesis that trapped PARP1 is modified by SUMOylation and ubiquitylation. This hypothesis was consistent with the observation that in cells cultured in MMS + PARPi, chromatin-associated PARP1 was present as multiple high-molecular-weight forms that could conceivably represent SUMOylated and/or ubiquitylated PARP1, forms that are absent in the nuclear-soluble fraction (Fig. 2a).
We first assessed whether ubiquitylated PARP1 was present in cells exposed to PARPi with different trapping properties. We used  4 . Cells were exposed to MMS + PARPi, after which the ubiquitylated pool of proteins was isolated from the chromatin fraction via ubiquitin-streptavidin-haemagglutinin (Ub-Strep-HA) isolation. In this fraction, high-molecular-weight isoforms of PARP1 were more prevalent in the talazoparib-or UKTT15-exposed cells compared with veliparib-exposed cells (Fig.  2b), suggesting that PARP1 ubiquitylation was enhanced by PARP1 trapping. We repeated the Ub-Strep-HA pulldown experiment in and PARP1 WT HEK293 cells were exposed to PARP1 trapping (MMS + talazoparib) and fractionated into nuclear-soluble and chromatin-bound fractions. High-molecular-weight (MW) forms of PARP1 are more prevalent in the chromatin fraction after trapping (lane 8). b, PARP1 trapping leads to PARP1 ubiquitylation. HEK293 cells were transfected with an Ub-Strep-HA-expressing construct and exposed to 0.01% MMS, 100 nM talazoparib, 10 μM veliparib or 10 μM UKTT15. Chromatin fractions were prepared in denaturing conditions, ubiquitylated proteins were immunoprecipitated and the presence of PARP1 was detected. c, As in b, HEK293 cells were transfected with an Ub-Strep-HA-expressing construct and exposed to combinations of MMS, talazoparib or 5 μM MLN-7243. The presence of high-molecular-weight ubiquitin forms of PARP1 were reduced by MLN-7243 exposure (lane 7 versus lane 5). The input controls for these experiments are shown in Extended Data Fig. 3a. b,c, EV, empty vector. d, Trapped PARP1 is SUMOylated. HEK293 PARP1 WT and PARP1 −/− cells were transfected with a HA-SUMO2-expressing construct and subsequently treated with 0.01% MMS or 100 nM talazoparib. The HA-SUMO2-modified proteins were purified from the chromatin fraction and PARP1 was detected via immunoblotting. High-exposure blots of PARP1 SUMOylation are shown in Extended Data Fig. 3b. e, SUMOylation and ubiquitylation inhibitors prevent trapped PARP1 modification. Similarly to c, the ubiquitylated pool of proteins was immunoprecipitated from the chromatin fraction of cells exposed to MLN-7243 (5 μM) or ML-792 (1 μM) and the presence of high-molecular-weight PARP1 isoforms was identified by immunoblotting. f, PARP1 is modified and interacts with RNF4 in a SUMO-dependent manner. PARP1 WT and PARP1 -/-HEK293 cells were exposed to trapping conditions in the presence of either 5 μM MLN-7243 or 1 μM ML-792 and PARP1 was immunoprecipitated under native conditions using PARP1-Trap beads. Western blotting for PARP1 revealed that modified PARP1 isoforms were abrogated by exposure to ML-792 but not MLN-7243. Abrogation of SUMOylation prevented the association between PARP1 and RNF4, whereas inhibition of ubiquitylation stabilized the interaction. Data represent two biological replicates. IP, immunoprecipitation; and WB, western blot.

NATURE CEll BIOlOgy
the presence of the E1 ubiquitin activating enzyme inhibitor MLN-7243 (also known as TAK243). Western blotting with an antibody to PARP1 revealed that trapping conditions led to the formation of high-molecular-weight PARP1 isoforms; these were almost completely abolished when the cells were exposed to MLN-7243, consistent with these high-molecular-weight isoforms representing ubiquitylated PARP1 ( Fig. 2c and Extended Data Fig. 3a). The poly-ubiquitylation of PARP1 was also observed in reciprocal denaturing immunoprecipitation experiments, where PARP1 was immunoprecipitated from HEK293 cells transfected with a FLAG-tagged PARP1 complementary DNA-expression construct (Extended Data Fig. 3b). We also identified poly-ubiquitin chains on PARP1 that were linked by K48 linkage (Extended Data Fig. 3c).
The presence of SUMO1 and SUMO2 in our trapped PARP1 interactome suggested that PARP1 may also be modified by SUMOylation in addition to ubiquitylated PARP1. We expressed HA epitope-tagged SUMO2 and isolated the SUMOylated pool of proteins under denaturing conditions from the chromatin fraction. Trapped PARP1 was clearly modified by SUMOylation (Fig. 2d). We also found that when cells were exposed to MMS alone (to induce DNA damage and activate PARP1) in the absence of PARPi, there was a depletion in the total pool of SUMO2 and a minimal level of PARP1 SUMOylation, as previously observed 23 ( Fig. 2d and Extended Data Fig. 3d). However, this did not occur to the same extent as seen under PARP1-trapping conditions. This suggested that PARP1 is SUMOylated when it becomes trapped. Interestingly, when cells cultured in PARP1-trapping conditions were incubated in the presence of a SUMOylation inhibitor (ML-792, which inhibits SUMO-activating enzyme), the levels of high-molecular-weight forms of ubiquitylated PARP1 decreased (Fig. 2e), suggesting PARP1 ubiquitylation following trapping could require previous PARP1 SUMOylation. Conversely, a ubiquitylation inhibitor had no effect on PARP1 SUMOylation (Fig. 2f), suggesting that the SUMOylation of trapped PARP1 is required for its ubiquitylation but that the ubiquitylation of trapped PARP1 is not a prerequisite for PARP1 SUMOylation.
Trapped PARP1 is sequentially modified by PIAS4 and RNF4. The pattern of SUMOylation and ubiquitylation of trapped PARP1 suggested the concert action of a SUMO E3 ligase and a SUMO-targeted ubiquitin ligase (STUbL). We assessed whether PIAS4 (a SUMO E3 ligase) and RNF4 (a STUbL) were responsible. PIAS4 has been previously implicated as a SUMO E3 ligase for PARP1 in its non-trapped state 24 and RNF4 has previously been implicated in modulating the transcriptional activity of PARP1 (ref. 25 ) as well as being involved in the repair of topoisomerase cleavage complexes, which also represent a 'trapped' nucleoprotein complex 26 . Chromatin co-immunoprecipitation of trapped PARP1 showed an increased interaction with RNF4 ( Fig. 2f), consistent with our hypothesis. This PARP1-RNF4 interaction was reduced following inhibition of SUMOylation and stabilized in cells exposed to a ubiquitylation inhibitor, indicative of a ligase-substrate interaction (Fig. 2f).
To delineate the relationship between SUMOylation and ubiquitylation of trapped PARP1 and a possible role for the SUMO E3 ligase PIAS4 in this process, we used PIAS4 -/-HCT116 and RNF4 -/-MCF7 cell lines 26 . Both cell lines were transfected with a FLAG-PARP1-expressing cDNA construct. After culturing the cells in MMS + PARPi, we immunoprecipitated PARP1 from the chromatin fraction. Western blotting with an antibody to SUMO2 and -3 revealed that PIAS4 is necessary for efficient SUMOylation and ubiquitylation of trapped PARP1 ( Fig. 3a and Extended Data Fig. 4a-c). Re-expression of wild-type PIAS4 in PIAS4 -/cells reversed these effects, but this was not achieved when we expressed a DNA binding-deficient form of PIAS4 (SAP domain deleted) or the catalytically inactive p.C342A PIAS4 mutant 27 (Fig. 3b,c and Extended Data Fig. 4d). Interestingly, although PARP1 ubiquitylation was decreased in RNF4 -/cells (confirming that RNF4 activity is responsible for this modification), PARP1 SUMOylation was increased ( Fig. 3d and Extended Data Fig. 4e-g). Re-expression of wild-type RNF4 in the RNF4 -/cells also reversed these effects, but this was not the case in cells expressing the SUMO-interacting motif-deleted 28 or catalytically inactive p.H156A mutant forms of RNF4 (Fig. 3e,f and Extended Data Fig. 4h). We also observed strong RNF4-dependent PARP1 ubiquitylation by overexpressing wild-type RNF4 in cells cultured in MMS + PARPi (Extended Data Fig. 5a), an effect that was not observed when we expressed a dominant-negative E2 binding-mutant form of RNF4 (p.M136S/ R177A). Using the RNF4 -/cells and dominant-negative mutants of RNF4, we found that RNF4 was responsible for 80-95% of the ubiquitylation of trapped PARP1. We also found that silencing of the RNF4 gene reduced ubiquitylation of trapped PARP1 (Extended Data Fig. 5b). Together, these data established RNF4 as a STUbL E3 ligase for trapped PARP1. Although PIAS4 and RNF4 accounted for the majority of SUMOylation and ubiquitylation of trapped PARP1, both PIAS4 -/and RNF4 -/cells exhibited some residual PARP1 SUMOylation and ubiquitylation, suggesting that other ligases might also contribute to the modification of trapped PARP1.
Finally, we tested the interdependency of PARP1 SUMOylation and ubiquitylation events using in vitro SUMOylation and ubiquitylation reactions. Incubation of recombinant PARP1 in the presence of a synthetic nicked DNA substrate, SUMO1 or SUMO2, SAE1 (SUMO E1), Ubc9 (SUMO E2) and an increasing concentration of PIAS4 drove a concentration-dependent SUMOylation of PARP1 ( Fig. 3g and Extended Data Fig. 5c). The addition of ubiquitin, UBE1 (Ub E1), Ubc5H (Ub E2) and an increasing concentration

NATURE CEll BIOlOgy
of RNF4 to this reaction led to efficient PARP1 ubiquitylation (Fig.  3h). In contrast, RNF4 displayed much lower PARP1 ubiquitylating activity in the absence of SUMOylation (Fig. 3h). Collectively, these data suggested a stepwise process, where following trapping, PARP1 is initially SUMOylated by PIAS4, followed by STUbL RNF4-driven ubiquitylation.  p97 interacts with modified trapped PARP1. Although the above experiments suggested a link between the trapping of PARP1 by PARPi and subsequent PARP1 SUMOylation and ubiquitylation, the functional significance of these post-translational modifications remained to be determined. Our mass spectrometry analysis suggested that under PARP1-trapping conditions there was an enhanced interaction between PARP1 and p97, an ATPase involved in the removal of ubiquitylated substrate proteins from chromatin. We therefore hypothesized that the SUMOylation and ubiquitylation of trapped PARP1 serve as a necessary prelude to the recruitment of p97 ATPase and the removal of trapped PARP1 from chromatin. We confirmed the interaction between p97 and PARP1 using both proximity ligation assays (PLAs; Fig. 4a) and co-immunoprecipitation experiments (Extended Data Fig. 6a). We then verified that the PARP1-p97 interaction was enhanced in a trapping-dependant manner in PARP1 wild-type cells but not in cells expressing a DNA binding-deficient PARP1 mutant (Fig. 4b,c). UKTT15, but not veliparib, also led to an increase in the PARP1-p97 interaction, consistent with PARP1 trapping being important for this interaction, as opposed to catalytic inhibition of PARP1 (Extended Data Fig. 6b).
CB-5083 is a small molecule that inhibits p97 ATPase activity and induces a p97 substrate-trapping effect 29 ; we found that CB-5083 caused an increase in PARP1-p97 interactions (Fig. 4d,e), suggesting that PARP1 could be a p97 substrate. Blocking p97 catalytic activity leads to the accumulation of ubiquitylated isoforms of its substrates 30,31 , which was also the case for trapped PARP1 (Fig. 4f). This was also observed using reciprocal immunoprecipitation of PARP1 under denaturing conditions ( Fig. 4f and Extended Data Fig.  3b,c). We reproduced the substrate-trapping effect of p97 inhibition by expressing a dominant-negative ATPase-deficient p97 mutant, p.E578Q 18,32,33 (Fig. 4g), consistent with trapped PARP1 being a p97 substrate. Furthermore, we demonstrated that the PARP1-p97 interaction was enhanced by expressing the p97 p.E578Q mutant in PARP1 -/cells reconstituted with wild-type PARP1 but not the trapping-defective PARP1 del.pK119S120 (Fig. 4h). This conclusion was further supported by co-localization immunofluorescence experiments, where p97 p.E578Q and trapped PARP1 foci 34 were found to substantially overlap (Extended Data Fig. 6d).
Ubiquitylation is a mediator of p97 interactions 6,7 . When cells were exposed to ubiquitylation (MLN-7243) or SUMOylation (ML-792) inhibitors (which decreased trapped PARP1 ubiquitylation; Fig. 2e), the PARP1-p97 interaction was reduced ( Fig. 4i and Extended Data Fig. 6e). p97 recognises and processes its ubiquitylated substrates using the NPL4-UFD1 complex, which mostly serves as a ubiquitin-binding receptor due to ubiquitin-binding domains in both NPL4 and UFD1 (refs. 35,36 ). When UFD1 was depleted, the interaction between trapped PARP1 and p97 was reduced (Fig. 4j). This was not the case when NPL4 was depleted, although, as expected, depletion of either subunit reduced the overall p97 recruitment to chromatin (Fig. 4j). Furthermore, only depletion of UFD1 led to a profound accumulation of trapped PARP1 (Fig. 4j). This suggested that the processing of trapped PARP1 is UFD1-but not NPL4-dependent. Although canonically, UFD1 is thought to function as an obligate heterodimer with NPL4, these observations seem consistent with previous work suggesting that NPL4 and UFD1 can recognise substrates independently of each other 7,8,10 . We also evaluated the effect of CuET, a metabolite of the approved alcohol-abuse drug disulfiram, which segregates p97 from chromatin into inactive agglomerates by disrupting NPL4 zinc finger motifs 37,38 and thus serves as a tool that inactivates the entire p97 pool. Because of its ability to inactivate the p97 pool by forming agglomerates, CuET has a distinct mechanism of action compared with CB-5083 and also NPL4-or UFD1-gene silencing. We found that the PARP1-p97 interaction was almost completely abrogated by CuET exposure (Fig. 4k). Together, these observations suggest that the p97 system and its ubiquitin-binding cofactor UFD1 (p97-UFD1) recognise and physically interact with trapped PARP1.
Trapped PARP1 is modulated by p97 activity. We used a 'trapchase' experimental approach to assess whether p97 removes trapped PARP1 from chromatin (Fig. 5a). Cells were exposed to MMS + PARPi to induce trapping (the 'trap') and then cultured in fresh media containing combinations of PARPi and p97-complex inhibitors (the 'chase'). The amount of trapped PARP1 was evaluated either by chromatin fractionation or measuring the proximity of PARP1 to phosphorylated H2AX (γH2AX PLA 39 ) at various time points during the chase. First, we followed the kinetics of trapped PARP1 in PIAS4 -/and RNF4 -/cells (Fig. 3a,d). Both PIAS4 -/and RNF4 -/cells showed slower resolution of chromatin-bound PARP1, especially at the later time points (Fig. 5b,c and Extended Data Fig. 7a,b), consistent with the idea that these SUMO/ubiquitin ligases promote the resolution of trapped PARP1 complex.
We then investigated the role of p97 activity with this assay by including talazoparib plus CB-5083 or CuET in the chase phase of the experiment and monitoring trapped PARP1 either via chromatin fractionation (Extended Data Fig. 7c) or PLA (Fig. 5d). After exposing cells to MMS + talazoparib, a significant amount of PARP1 was detected in the proximity of γH2AX (Fig. 5d), indicating that the 'trapping' part of the experiment was successful; after removing Fig. 4 | PARP1 interacts with p97 in a trapping-dependent manner. a, Images of a PLA for endogenous PARP1 and p97 in CAL51 cells. b, The PARP1-p97 interaction is increased following DNA damage. CAL51 cells expressing PARP1 WT -eGFP or PARP1 del.pK119S120 -eGFP were exposed to trapping conditions and PARP1-GFP was immunoprecipitated under native conditions. Data represent two biological replicates. c, PARP1-p97 PLA (anti-PARP1 and anti-p97) in CAL51 cells expressing either PARP1 WT -eGFP or PARP1 del.pK119S120 -eGFP. The geometric mean and 95% confidence interval (CI) are shown; n = 2,016 cells from three independent experiments. d, PARP1-p97 PLA in CAL51 cells under trapping. PLA with p97 antibody alone (top) or p97 + PARP1 antibody (bottom). a,d, Scale bars, 5 μm. Data represent three biological replicates. e, Number of PLA foci in d. The geometric mean and 95% CI are shown; n = 2,035 cells from three independent experiments. f, Inhibition of p97 increases the presence of ubiquitylated PARP1. HEK293 cells expressing Ub-Strep-HA were cultured in PARP1-trapping conditions in the presence or absence of 10 μM CB-5083 and the ubiquitylated proteins were immunoprecipitated under denaturing conditions (see Extended Data Fig. 6c for the input controls). Data represent three biological replicates. g, HEK293 cells expressing either wild-type p97-Myc or p97 p.E578Q-Myc were transfected with a FLAG-PARP1 construct, exposed to trapping conditions and PARP1 immunoprecipitated from the chromatin fraction. Data represent two biological replicates. h, CAL51 cells expressing PARP1 WT -eGFP or PARP1 del.p.119K120S -eGFP were transfected with p97 E578Q-Strep-Myc for 18 h, exposed to trapping conditions and then fractionated. The chromatin PARP1-eGFP immunoprecipitate was probed by antibody that detected both endogenous and ectopically expressed p97. Data represent two biological replicates. i, PARP1-p97 co-localization is reduced by ubiquitylation (5 μM MLN-7243) or SUMOylation (1 μM ML-792) inhibitors. Number of PARP1-p97 PLA (anti-PARP1 and anti-p97) foci. The geometric mean and 95% CI are shown; n = 1,316 cells from three independent experiments. c,e,i, ****P < 0.0001; NS, not significant; ordinary one-way analysis of variance (ANOVA). j, The p97 adaptor UFD1 mediates the interaction between p97 and trapped PARP1. Chromatin-bound co-immunoprecipitation. Data represent three biological replicates; siCon, control short interfering RNA (siRNA); siUFD1, siRNA to UFD1; and siNPL4, siRNA to NPL4. k, As per j, the PARP1-p97 interaction is disrupted by the p97 sequestration agent, CuET. Data represent two biological replicates. EV, empty vector; IP, immunoprecipitation; WB, western blot; and WT, wild type.

NATURE CEll BIOlOgy
the trapping agents, the amount of trapped PARP1 decreased (for example, the PARP1-γH2AX PLA signal disappeared). The PARP1-γH2AX PLA signal also diminished when cells were chased in the presence of one agent-that is, a PARPi or a p97 inhibitor. Conversely, when cells were chased in the presence of both inhibitors, PARPi (talazoparib) and p97 inhibitor (either CB-5083 or CuET), the amount of trapped PARP1 persisted (Fig. 5d,e). Consistent with the idea that RNF4 is an upstream factor involved in the processing of trapped PARP1, we also found that gene silencing of RNF4 led to the persistence of PARP1-γH2AX PLA foci (Fig. 5f). In addition, we assessed the effect on PARP1 trapping by the expression of a dominant-negative RNF4 p.M136S/R177A  Fig. 7d), a p97 p.E578Q mutant (Extended Data Fig. 7e-g) or UFD1 depletion (Extended Data Fig. 7h). All three interventions led to a higher level of trapped PARP1 in the chromatin fraction, confirming the importance of these proteins in the processing of trapped PARP1.
In homologous-recombination proficient cells, trapped PARP1 activates RAD51-mediated DNA repair, monitored by assessing nuclear RAD51 foci. We found that a 16 h exposure of cells to PARPi elicited both γH2AX and RAD51 foci, but γH2AX and RAD51 foci diminished after 3 h when PARPi was removed from the culture Articles NATURE CEll BIOlOgy media by washing, suggesting resolution of the DNA damage caused by trapped PARP1 (Fig. 5g-i). When we used p97 inhibitors (CB-5083 or CuET) in this chase period, the γH2AX and RAD51 foci persisted, indicating that the underlying trapped PARP1-related damage could not be resolved as efficiently. Incubating cells in the presence of p97 inhibitor alone did not induce γH2AX and RAD51 foci, suggesting that the persistence of γH2AX and RAD51 foci in experiments involving PARPi, followed by p97 inhibitor were indeed caused by PARPi. The effects on foci resolution were also not trivially explained by alterations in the cell cycle, as exposure of cells to p97 inhibitor for 3 h did not lead to significant changes in cell-cycle distribution (Extended Data Fig. 8a,b). We also noted that when we used fluorescence recovery after photobleaching (FRAP) to monitor the exchange of PARP1 WT -eGFP at a UV laser stripe in the presence of a PARPi 40 , the addition of a p97 inhibitor (CB-5083) led to a modestly slower FRAP (PARP1 WT -eGFP half-time of recovery (t 1/2 ) of 4.9 ± 1.3 s in the presence of talazoparib versus 7.8 ± 1.4 s in the presence of talazoparib + CB-5083; two-sided t-test P < 0.05; Extended Data Fig. 8c,d).
p97 inhibition potentiates PARPi cytotoxicity. Based on the prolonged PARP1-trapping effects described earlier, we hypothesized that p97 inhibition modulates the cytotoxic effects of PARPi. We assessed the effect of two p97 inhibitors (CB-5083 and CuET) on the cytotoxic effect of two trapping PARPi (talazoparib and olaparib) and observed a dose-dependent potentiation of the clonogenic effect of each PARPi by the presence of the p97 inhibitor ( Fig. 6a-c).
Bliss independence analysis confirmed that these drugs elicited supra-additive effects when used in combination (Extended Data Fig. 9a,b). This combinatorial effect was PARP1-trapping dependent as it was reversed in PARP1 -/cells ( Fig. 6a and Extended Data Fig. 9c,d), suggesting that it was also not due to other roles that p97 might play in DNA repair. Furthermore, p97 inhibitor (at concentrations used in the previous PARPi combinatorial experiments) did not enhance sensitivity to the alkylating agents MMS or temozolomide in either PARP1 WT or PARP1 -/cells (Fig. 6d,e). This implied that other roles p97 might play in alkylation DNA-damage repair are unlikely to explain its ability to evict trapped PARP1 from chromatin and, following from that, the ability of CB-5083/CuET to sensitize to PARPi. Because PARPi are approved for the treatment of cancers that have homologous-recombination defects and trapped PARP1 is the key cytotoxic event in homologous recombination-defective cells, we assessed the effect of combined exposure to CB-5083 + talazoparib in DLD1 cells with/without genetic ablation of BRCA2. When used alone, CB-5083 had a modest BRCA2 synthetic lethal effect (Extended Data Fig. 9e), but it had a far greater effect on BRCA2 -/-DLD1 cells than isogenic wild-type BRCA2 (BRCA2 WT ) cells when used in combination with talazoparib ( Fig. 6f and Extended Data Fig. 9f,g). In tumour organoids derived from mice with combined Brca1 and Tp53 loss-of-function mutations (WB1P) 41 we found that CB-5083 further sensitized tumour organoids to talazoparib (Fig. 6g and Extended Data Fig. 9h). We also assessed CB-5083 in combination with PARPi in a patient-derived organoid culture derived from a patient with triple-negative breast cancer harbouring a germline pathogenic BRCA1 p.R1203* mutation (BRCA1 c.3607C>T), which was homozygous in the organoid. CB-5083 led to a marked shift in talazoparib sensitivity (Fig. 6h and Extended Data Fig. 9i), suggesting that p97 inhibition has the potential to potentiate the effects of PARPi in human tumour cells.

Discussion
The effectiveness of PARPi in cancer treatment relies on their ability to trap PARP1 in the chromatin. Here we have delineated a biochemical cascade that processes trapped PARP1. Trapped PARP1 is sequentially SUMOylated by PIAS4 and then ubiquitylated by RNF4; these events recruit p97, whose ATPase activity removes PARP1 from chromatin (Fig. 6i). Importantly, interference with any of these processing steps leads to persistence of the trapped complex and enhanced PARPi sensitivity. Other factors might also influence this process, especially as other ubiquitin-processing enzymes are recruited to DNA damage and also PARP1 in a PAR-dependent manner (for example, the deubiquitylating enzyme ATXN3 (ref. 42 ) and the E3 ubiquitin ligase TRIP12 (ref. 43

)).
During our studies we considered whether the effects of p97 modulation on PARP1 trapping/PARPi sensitivity might not be solely due to an effect of p97 on trapped PARP1 but could also be due to p97 modulating other DNA-repair processes. However, we think this unlikely for the following reasons: (1) the p97 inhibitors, when employed as single agents, did not elicit biomarkers of DNA damage such as γH2AX or RAD51 foci (Fig. 5g-l), or alterations in cell-cycle dynamics (Extended Data Fig. 8a,b); and (2) p97 inhibitors enhanced PARPi sensitivity in a PARP1-dependent manner (Fig. 6a,b), the p97 inhibitor CB-5083 did not alter sensitivity to MMS or temozolomide (Fig. 6c,d).
Our observations may lead to further questions. First, although PIAS4 and RNF4 seem to act in a linear manner, it is still possible that the balance of SUMOylation and ubiquitylation is influenced by other E3 ligases. The effect of losing PIAS4 on trapped PARP1 resolution was indeed modest (Fig. 5b), suggesting other proteins might also be involved. Second, our data also suggest that UFD1 is required for the recruitment of p97 to trapped PARP1 (Fig. 4j). How UFD1 recruits p97 to trapped PARP1 remains to be established. We note that canonically, UFD1 is thought to function as an obligate heterodimer with NPL4; however, we observed NPL4 silencing did not alter PARP1 trapping or the PARP1-p97 interaction, whereas UFD1 depletion did (Fig. 4j). Although we are unable to entirely rule out a role for NPL4 in the processing of trapped PARP1, it is possible that our described function of p97, similar to the removal of CDT1 and other substrates from chromatin 7,8,10 , is dependent on UFD1 only. Third, in most systems the p97-dependent removal of Fig. 6 | inhibition of p97 potentiates the effect of PARPi. a, Inhibition of p97 potentiates the cytotoxicity of PARPi. CAL51 cells were exposed to PARPi (talazoparib (left) or olaparib (right)) in the presence of a p97 inhibitor (CB-5083 or CuET) for a period of 14 d. Images are shown for samples exposed to 100 nM CB-5083 and 8 nM CuET. b,c, Drug-response curves for CB-5083 (b) and CuET (c). See also Extended Data Fig. 9a,b. d,e, DNA alkylating agents that are used to induce PARP1 trapping do not enhance the cell-inhibitory effects of CB-5083. PARP1 WT and PARP1 -/-CAL51 cells were exposed to the alkylating agents MMS (d) or temozolomide (TMZ; e) in combination with either talazoparib (positive control) or CB-5083 for 7 d, after which the cell viability was measured. f, CB-5083 modulates the synthetic lethal effect of PARPi in BRCA2 -/cells. Survival curves from clonogenic survival assays in BRCA2 WT and BRCA2 -/-DLD1 cells treated with different doses of CB-5083 and talazoparib. Colony formation images and quantification are shown in Extended Data Fig. 9e. g, Inhibition of p97 sensitizes mouse cancer organoid cells to PARPi. WB1P breast cancer organoids with Brca1 and p53 loss-of-function mutations were cultured in the presence of the indicated drugs for 7 d. Bright-field images of organoids are shown in Extended Data Fig. 9f. h, Inhibition of p97 sensitizes a human BRCA1-mutant patient-derived breast cancer organoid to PARPi. KCL014BCPO organoids were cultured in the indicated drugs for 7 d. Bright-field images of the organoids are shown in Extended Data Fig. 9g. b-h, Data are the mean ± s.d. of three biological replicates. i, Model of the processing of trapped PARP1. PARP1 trapped by the presence of PARPi on DNA is processed in a stepwise manner. It is initially SUMOylated in a PIAS4-dependant manner and subsequently ubiquitylated in an RNF4-dependent manner. p97 is recruited to the ubiquitin chains and binds via UFD1 and the ATPase activity of p97 extracts the modified PARP1 from the chromatin. DMSO, dimethylsulfoxide.

NATURE CEll BIOlOgy
ubiquitylated proteins is coupled to proteasomal degradation, but this was not the case for PARP1. In addition, p97 is known to participate in substate recycling, as is the case for Aurora B 9 , yeast transcriptional repressor alpha 20 , Ub-LexA-VP16 (ref. 19 ) and MRE11 (ref. 44 ). This raises the possibility that PARP1 might also be a p97 substate that is recycled, not degraded.
Finally, the PARPi-generated DNA lesions seem to be processed in an analogous fashion to trapped TOP1-cleavage complexes 11 . Both PARPi and TOP1 inhibitors cause replication-fork stress and sensitivity in cells with homologous recombination defects and the sensitivity to both classes of agents is modulated by SLFN11 (ref. 45 46 and data shown here). Thus, it seems plausible that the sensing and processing machinery that activate the SUMOylation and ubiquitylation of trapped PARP1 as well as trapped TOP1 and TOP2 cleavage complexes are also shared and not necessarily private to the precise nature of the nucleoprotein complexes, and might be related with their ability to interfere with normal DNA metabolism.
In conclusion, our work elucidates an elegant and highly orchestrated molecular machinery of PIAS4, RNF4 and UFD1-p97 that recognises and removes trapped PARP1 from chromatin.

online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41556-021-00807-6.

Methods
Cells and cell culture. CAL51 (DSMZ, ACC 302), DLD1 (ATCC, CCL-221), DLD BRCA2 −/− (Horizon, HD 105-007), HeLa (ATCC, CCL-2, commonly misidentified cell line as set out by ICLAC; we did not authenticate but used directly from the ATCC) cells were maintained in DMEM medium supplemented with 10% fetal bovine serum (FBS) and 1×penicillin-streptomycin (Sigma-Aldrich). PARP1 −/− CAL51 cells were previously described 15 . They were transfected with a corresponding PARP1-expressing piggyBac construct in combination with hyPBase-expressing plasmid 47 . Single-cell clones were sorted by FACS 72 h after transfection and allowed to expand. These clones were characterized for the expression of the tagged protein by microscopy and western blotting. PARP1 -/-HEK293 cells were a gift from I. Ahel (University of Oxford) 48 . The HCT116 PIAS4 -/and MCF7 RNF4 -/cells were previously described 26 . The WB1P organoid line was previously described 49 . They were grown in a mix of 50% Matrigel (Corning) and 50% Advanced DMEM/F12 (Life Technologies) medium containing 10 mM HEPES (Sigma-Aldrich) pH 7.5, GlutaMAX (Invitrogen) and supplemented with 125 mM N-acetylcysteine (Sigma-Aldrich), B27 supplement and 50 ng ml −1 EGF (Life Technologies). The organoids were seeded at 10,000 cells per well of a 24-well plate and drugs were added at the indicated concentrations 24 h later. Cell viability was assessed using a 3D CellTiter-Glo assay (Promega). KCL014BCPO was derived (L.M.B. et al., manuscript in preparation), similarly to as described 50 . Briefly, human breast tumour samples were obtained from adult female patients after informed consent as part of a non-interventional clinical trial (BTBC study REC no. 13/LO/1248, IRAS ID 131133; principal investigator: A.N.J.T., study title: ' Analysis of functional immune cell stroma and malignant cell interactions in breast cancer in order to discover and develop diagnostics and therapies in breast cancer subtypes'). This study had local research ethics committee approval and was conducted adhering to the principles of the Declaration of Helsinki. Specimens were collected from surgery and transported immediately. A clinician histopathologist or pathology-trained technician identified and collected tumour material into basal culture medium. The tumour samples were coarsely minced with scalpels and then dissociated using a Gentle MACS dissociator (Miltenyi). The resulting cell suspension was mechanically disrupted, filtered and centrifuged. The resulting cell pellets were then plated into three-dimensional (3D) cultures at approximately 1-2 ×10 3 cells μl −1 in Ocello PDX medium and hydrogel. All cultures were maintained in humidified incubators at 37 °C and 5% CO 2 . All human cell-line identities were confirmed by short-tandem-repeat typing and verified to be free of mycoplasma infection using Lonza MycoAlert.

Plasmids, antibodies and reagents.
To generate PB-PARP1-eGFP, PARP1 cDNA was cloned in a previously described piggyBac vector 51 . To generate the PARP1-Apex2-eGFP construct, the Apex2 gene was amplified from Addgene vector 49386 and inserted between the PARP1 and eGFP coding sequences via InFusion (Clonetek, 648910). PBZ-mRuby2 is described in 15 . Ub-Strep-HA was a gift from V. D' Angiolella; HA-SUMO2 was a gift from E. Yeh (ref. 52 ), FLAG-PARP1 was a gift from I. Ahel, p97-GFP was a gift from H. Mayer.
The following other small molecules were used: olaparib (Selleckchem, S1060), veliparib (Selleckchem, S1004), UKTT15 from in-house synthesis as described in 4 , MMS (Sigma-Aldrich, 129925-5G), CB-5083 (Selleckchem, S8101), CuET from in-house synthesis as described in 37  Western blotting. Standard protocols for SDS-PAGE and immunoblotting were used 53 . Nitrocellulose (GE Healthcare) or PVDF (BioRad) membranes were used to transfer proteins from polyacrylamide gels, depending on the antibody. Cellular fractionation immunoprecipitation. Cells were washed twice with PBS and then resuspended in buffer A (10 mM HEPES, 10 mM KCI, 340 mM sucrose, 10% glycerol, 2 mM EDTA, protease and phosphatase inhibitors, N-ethylmaleimide (NEM)). Triton X-100 was added to a final concentration of 0.1% and left on ice for 2-5 min, depending on the cell line. The supernatant was harvested as the cytosolic fraction and the pellet (nuclei) was then washed twice with buffer A. Buffer B (3 mM EDTA, 0.2 mM EGTA, 5 mM HEPES pH 7.9, protease and phosphatase inhibitors, NEM) was then added to burst the nuclei, after which lysates were kept on ice for 10 min. The supernatant was then removed as the nuclear-soluble fraction. The remaining chromatin pellet was then washed with buffer B in 0.5% Triton X-100, followed by benzonase buffer without MgCl 2 (50 mM Tris-HCl pH 7.9, 100 mM NaCl). Benzonase digestion buffer (50 mM Tris-HCl pH 7.9, 10 mM MgCl 2 , 100 mM NaCl, protease and phosphatase inhibitors, NEM) supplemented with 125 U benzonase enzyme (Merk Millipore) was added to the pellet and rotated on a wheel at 4 °C for 1 h. The samples were then centrifuged at 20,000g for 15 min; chromatin input for the immunoprecipitation reaction was taken from the supernatant. The remaining supernatant was then incubated with the respective beads for 3 h at 4 °C with rotation. For native immunoprecipitations, ethidium bromide (1:200) was added to remove unwanted DNA-protein interactions. The beads were then washed three times with IP wash buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.5 mM EDTA, 0.05% Triton X-100) before elution with Laemmli buffer.
Whole-cell immunoprecipitation. Cells were lysed in IP lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.5% Triton X-100, protease and phosphatase inhibitors, NEM) and spun on a wheel at 4 °C for 10 min. The supernatant was removed, the pellet was washed once with benzonase buffer, and the supernatants were pooled together. Benzonase digestion buffer supplemented with 125 U benzonase enzyme (Merk Millipore) was added to the pellet and placed on a rotating wheel at 4 °C for 1 h. The samples were centrifuged at 20,000g and all supernatants were pooled. Input for the immunoprecipitation was then removed and the samples were incubated with the respective beads for 3 h at 4 °C with rotation. The beads were then washed three times with IP wash buffer before elution with Laemmlli buffer.
Denaturing immunoprecipitation. Cells were lysed according to either the cell fractionation immunoprecipitation or whole-cell immunoprecipitation protocol as described earlier. Before incubation with beads, SDS was added to the samples to a concentration of 1% and boiled at 95 °C for 5 min. The samples were then diluted in 1% Triton X-100 to achieve a dilution of 1:10 (SDS at 0.1%) along with beads and rotated on a wheel at 4 °C for 3 h. The beads were then washed three times with IP wash buffer before elution with Laemmlli buffer.
Cell viability and clonogenic survival assays. The viability of cells was measured using a CellTiter-Glo assay (Promega) after exposure to various concentrations of drugs for 6 d. The long-term effects of drug exposure were assessed by colony-formation assay after exposure of the cells to a drug-containing medium (refreshed weekly) for 12-14 d; the cells were stained at the end of the assay with sulforhodamine B. When plotting survival curves, the surviving fraction was calculated relative to the dimethylsulfoxide (solvent)-exposed cells.
The viability of the KCL014BCPO organoid line was measured using a 3D Cell Titre-Glo assay (Promega). Organoids were seeded in 24-well plates, with one 15 µl Matrigel droplet containing 3,000 cells per well. After 24 h, the organoids were treated with a drug-containing media (drug refreshed after 4 d) for 7 d before assessing their viability by measuring 3D Cell Titre-Glo luminescence.
Apex2-mediated proximity labelling. For each condition tested, 5-10 × 10 6 cells expressing PARP1-Apex2-eGFP were exposed to either 0.01% MMS or a combination of 0.01% MMS + 100 nM talazoparib for 1 h. In the last 30 min of the incubation, biotin-tyramide (Sigma-Aldrich, SML2135) was added to the media at a final concentration of 500 µM. To label proteins, H 2 O 2 (Sigma-Aldrich, H1009) was added for 60 s at a final concentration of 1 mM. The reaction was quenched by washing the cells three times with freshly prepared quench solution (PBS containing 10 mM sodium ascorbate, 10 mM sodium azide, 5 mM Trolox (Sigma-Aldrich, 238813)). Subsequently, the cells were scraped in quench solution and washed twice in 0.1% IGEPAL CA-630 quench solution. The remaining nuclei were lysed in nuclear RIPA buffer (50 mM Tris-HCl pH 7.5, 1 M NaCl, 1% IGEPAL CA-630, 0.1% sodium deoxycholate, 1 mM EDTA) for 10 min on ice. The lysates were diluted with RIPA buffer without NaCl, to obtain a final concentration of 200 mM, sonicated for 1 min and incubated with 250 U benzonase for 20 min at room temperature. The lysates were clarified by centrifugation at 13,000g for 15 min at 4 °C. The protein concentration was determined and 1 mg total protein was incubated with 30 μl streptavidin-magnetic beads (ThermoFisher, 88816) for 1 h at room temperature. The beads were washed stringently by sequential washes-twice Mass spectrometry and data analysis. After initial washes according to the purification method, the beads were further washed twice with 50 mM ammonium bicarbonate. The proteins on the beads were digested with 0.1 μg μl −1 sequencing-grade trypsin (Roche) at 37 °C overnight. The peptide solution was neutralized with 5% formic acid; acetonitrile was added (60% final concentration) and the solution was filtered through a Millipore Mutiscreen HTS plate (pre-washed with 60% acetonitrile). The peptide solution was lyophilized on a SpeedVac and the peptides were dissolved in 20 mM TCEP-0.5% formic acid solution. The liquid chromatography with tandem mass spectrometry analysis was conducted on the Orbitrap Fusion Tribrid mass spectrometer coupled with a U3000 RSLCnano UHPLC system (ThermoFisher). The peptides were first loaded on a PepMap C18 trap (100 μm inner diameter × 20 mm, 100 Å, 5 μm) at 10 μl min −1 with 0.1% formic acid in H 2 O; they were then separated on a PepMap C18 column (75 µm inner diameter × 500 mm, 100 Å, 2 μm) at 300 nl min −1 and a linear gradient of 4-32% acetonitrile in 0.1% formic acid in 90 min with the cycle at 120 min. Briefly, the Orbitrap full MS survey scan was m/z 375-1,500 with a resolution of 120,000 at m/z 200, with the automatic gain control set at 40,000 and maximum injection time at 50 ms. Multiply charged ions (z = 2-5) with an intensity above 8,000 (for Lumos) or 10,000 (for Fusion) counts were fragmented in a higher collision dissociation cell at 30% collision energy and the isolation window at 1.6 Th. The fragment ions were detected in ion-trap mode with the automatic gain control at 10,000 and a maximum injection time of 35 ms. The dynamic exclusion time was set at 40 s with ±10 ppm.
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE 54 partner repository with the dataset identifier PXD024337. Raw mass spectrometry data files were analysed using Proteome Discoverer 1.4 (Thermo). Database searches were carried out using Mascot (version 2.4) against the UniProt human reference database (January 2018; 21,123 sequences) with the following parameters: trypsin was set as digestion mode with a maximum of two missed cleavages allowed. Precursor mass tolerance was set to 10 ppm and the fragment mass tolerance was set to 0.5 Da. Acetylation at the amino terminus, oxidation of methionine, carbamidomethylation of cysteine, and deamidation of asparagine and glutamine were set as variable modifications. Peptide identifications were set at a false-discovery rate of 1% using Mascot Percolator. Protein identification required at least one peptide with a minimum score of 20. For the Apex2-based proximity labelling mass spectrometry, the following steps were taken. Proteins identified with a single peptide were removed from further analysis. The PARP1-eGFP mass spectrometry profile under trapping conditions (MMS + talazoparib) was used as a negative control. Proteins identified with >2 unique peptides in this sample were removed from further analyses. Peptide spectrum matches (PSM) was used as a proxy of protein abundance in the samples. A ratio was built between the PARP1-Apex2-eGFP MMS + talazoparib PSM and PARP1-Apex2-eGFP MMS PSM as an indicator for enrichment in the trapping conditions. Where PSM values were absent from the PARP1-Apex2-eGFP MMS PSM (that is, no detection in the sample) a value of one was added to calculate a meaningful ratio (data provided in Supplementary Table 1). The list of genes was then searched on the STRING database to build a network of the hits. A high-confidence threshold was set for mapping the network, using a minimum required interaction score of 0.7 for connecting nodes. Single unconnected nodes were excluded from the network plots. The gene list was searched in the Enrichr database to assess which Kyoto Encyclopedia of Genes and Genomes 2019 pathway annotations are enriched in the dataset. The list of annotations was filtered using −log(P) values of 1.3 (P = 0.05) or 2 (P = 0.01; data are provided in Supplementary Table 2). For RIME analysis, the proteins identified with single peptides were removed from further analyses. Proteins identified in PARP1 -/-CAL51 cells were considered as background and removed from further analyses when they were identified with more than two unique peptides. Subsequently, the mass spectrometry data obtained from PARP1 WT -eGFP or PARP1 del.p.119K120S -eGFP cells were considered separately. For each cell line, a ratio was built between the MMS + talazoparib PSM and the MMS PSM as an indicator for enrichment in the trapping conditions. Where PSM values were absent from the MMS PSM (that is, no detection in the sample), a value of one was added to calculate a meaningful ratio (data in Supplementary Tables 3 and 4 for PARP1 WT -eGFP and PARP1 del.p.119K120S -eGFP, respectively).
PLA. The PLA assays were carried out using a Duolink in situ red starter kit mouse/rabbit kit (Sigma-Aldrich) according to the manufacturer's protocol. The primary antibodies used were: mouse anti-PARP1 (Sigma-Aldrich, WH0000142M1), rabbit anti-PARP (Cell Signaling), mouse anti-p97 (Abcam, ab11433) and rabbit anti-phospho-H2AX (Cell Signaling). The antibodies were used at a 1:1,500 dilution. Images were acquired on a Marianas advanced spinning disk confocal microscope (3i) and analysed using a custom CellProfiler pipeline. Typically, several hundred nuclei were counted per condition from at least two independent biological repeats.
Micro-irradiation. Cells were cultured in glass-bottomed culture dishes (MaTek, P35G-0.170-14-C) in 10% FBS DMEM media and maintained at 37 °C and 5% CO 2 in an incubation chamber mounted on the microscope. Imaging was carried out on an Andor Revolution system, ×60 water objective with micropoint at 365 nm. For FRAP analysis, the cells were acquired one at a time; each cell was irradiated at a single spot with 1 µm diameter in the nucleus. After the signal of recruitment reached its maximum (typically 30-60 s after micro-irradiation), the recruitment spot was bleached with a 488 nm laser and imaging continued with one frame per interval of 2 s. For each experimental condition, 10-12 cells were acquired and the experiment was repeated independently on a different imaging day. From the raw intensities of the micro-irradiation site, the spot intensity immediately before bleaching was set to one and the intensity immediately after bleaching to zero. The recovery data were fitted with one site-specific binding model of nonlinear regression (GraphPad Prism software) and the extra-sum-of-squares F-test was used to calculate the t 1/2 .
Cell-cycle analysis. Cells were incubated in the presence of inhibitors for the corresponding amount of time. Ethylene-deoxyuridine (10 µM; ThermoFisher) was added to the media 1 h before fixation. Subsequently, the cells were trypsinized and fixed in ice-cold absolute ethanol. The cells were rehydrated via a PBS wash and permeabilized with 0.5% Triton X-100 in PBS for 15 min at room temperature with rotation. After a PBS wash, a click chemistry reaction cocktail was added to the cells (100 mM Tris-HCl pH 7.6, 4 mM CuSO 4 , 2.5 µM azide-Fluor 488 (Sigma), 100 mM sodium ascorbate (Sigma)) and incubated for 30 min at room temperature, protected from light. After a PBS wash, propidium iodide/RNase staining solution (Thermo) was added to the cells for 30 min. The cell-cycle profiles were acquired on a BD LSRII flow cytometer and analysed using the BD FACSDiva software.
Chromatin fractionation. The chromatin fractionation assay for PARP trapping was based on a previously published protocol 2 . For the trap-chase experiments, cells were cultured in six-well plates, exposed to 100 nM talazoparib and 0.01% MMS for 1 h and subsequently incubated in media containing the appropriate drugs (typically 100 nM talazoparib, 10 μM CB-5083 or 1 M CuET) for a chase period of 3 h. The cells were fractionated using a Subcellular protein fractionation kit for cultured cells (ThermoFisher, 78840) according to the manufacturer's recommendations.

Statistics and reproducibility.
No statistical method was used to pre-determine the sample size. No data were excluded from the analyses. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment. Denaturing and co-immunoprecipitations were performed twice, showing reproducibility, unless specified otherwise in the legends. For pre-extraction-based immunofluorescence microscopy, quantification and statistics were derived from n = 3 independent experiments. Immunofluorescence and PLA experiments were conducted in at least n = 3 independent biological repeats, and for each repeat, a few hundred cells were scored per condition. The data were pooled and analysed by ordinary one-way ANOVA (GraphPad Prism 9). Cellular growth inhibition assays were performed for at least n = 3 independent biological repeats and the statistical significance was derived using a two-way ANOVA (GraphPad Prism 9).
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository (dataset identifier PXD024337). All other data supporting the findings of this study are available from the corresponding authors on reasonable request. Source data are provided with this paper.
Last updated by author(s): Oct 11, 2021 Reporting Summary Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted

Software and code
Policy information about availability of computer code Data collection Western blots were imaged and processed using ImageLab (BioRad) v5.2.1 and Imagej 1.53c. PLA assays were imaged using Marianas Advanced Spinning disk confocal microscope 3i. Image acquisition for trapped PARP1 colocalisation immunofluorescence was achieved using Andor Dragonfly Fusion software. Image acquisition for trapped PARP1 accumulation after p97EQ expression was achieved using Nikon NIS-elements. FACs profiles were acquired and analyzed with BD FACSdiva software.

Data analysis
Clonogenic curves, PLA graphs, immunofluorescence graphs were produced using GraphPad Prism v7. Representative images for PLA and immunofluorescence were analysed using ImageJ 1.53c. Quantification of trapped PARP1 and PLA were analysed by CellProfiler 4.1.3. FACs profiles were analyzed with BD FACSdiva v9.0.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A list of figures that have associated raw data -A description of any restrictions on data availability The mass spectrometry proteomics data ( Figure 1) have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset April 2020 identifier PXD024337. Source data are provided with this study. All other data supporting the findings of this study are available from the corresponding author on reasonable request.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
For PLA, each experiment was replicated 3 times, with 200-250 total cells quantified for each condition, sample size was determined by accepted standard in the field for accurate assessment of interaction by PLA. For PARP1 trapping immunofluorescence, no sample size calculations were made. Instead sample size was dependent on the frequency of successful transfection and expression, three replicate experiments were carried out and in total for each condition 80-220 cells were quantified.
Data exclusions No relevant data was excluded from this study.

Replication
The experiments presented in this manuscript were typically performed in 2-3 biological replicates, with each attempt at replication successful. Number of times of experiments were replicated are indicated in the Methods section.
Randomization In each experiment, different cell samples started from similar conditions and treatments were randomly allocated.

Blinding
Data did not require blinding as no qualitative measurements were taken. Furthermore, quantification of foci from PLA, immunofluorescence and survival in clonogenic assays in different conditions were all achieved using automatic software based protocols in Cellprofiler, so blinding was not performed to remove bias.
Reporting for specific materials, systems and methods