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

Summary paragraph

Poly-(ADP-ribose) polymerase inhibitors (PARPi) elicit anti-tumour activity in homologous recombination-defective cancers by trapping PARP1 in a chromatin-bound state. How cells process trapped PARP1 remains unclear. Using wild-type or a trapping-deficient PARP1 mutant combined with rapid immunoprecipitation mass spectrometry of endogenous proteins (RIME) 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 upon trapping, PARP1 is SUMOylated by PIAS4 and subsequently ubiquitinated by the SUMO-targeted E3-ubiquitin ligase, RNF4, events that promoted 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 PARPi-induced cytotoxicity in homologous recombination-defective tumour cells and patient-derived tumour organoids. Taken together, these results suggest that p97 ATPase plays a key role in the processing of trapped PARP1 and the response of tumour cells to PARPi.

Introduction

PARP inhibitors (PARPi) selectively kill tumour cells with impaired homologous recombination (HR) and are approved for use in HR defective breast, ovarian, pancreatic or prostate cancers ¹ . The key target of PARPi, Poly(ADP-Ribose) Polymerase 1 (PARP1/ARTD1), is an 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 by 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 being a significant 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 modulating the amount of residual PAR on PARP1 via PAR-glycosylase (PARG) ⁵ , there is only a limited understanding of how trapped PARP1 is released from damaged DNA.

By generating a series of protein/protein interaction profiles of either trapped or non-trapped PARP1, we show that trapped PARP1 binds p97 ATPase (Valosin Containing Protein, VCP). p97 is a hexameric unfoldase/segregase which unfolds and disassembles ubiquitylated substrates through its central pore ^{6, 7} including Aurora B kinase, CMG helicases, the licensing factor CDT1 and the TOP1-cleavage complex ^8–10 . 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, using a metabolite of the clinically used drug disulfiram, leads to prolonged PARP1 trapping and profound PARPi sensitivity, suggesting an approach to enhancing PARPi-induced cytotoxicity. Collectively, our findings suggest that the p97-PARP1 axis is essential for removal of trapped PARP1 and the cellular response to PARPi.

Results

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: (i) Rapid Immunoprecipitation Mass spectrometry of Endogenous proteins (RIME ¹¹ ) and (ii) in vivo Apex2 peroxidase-mediated labelling of proximal proteins ¹² . We used a PARPi-resistant PARP1 defective cell line ³ , CAL51 PARP1 ^−/−, into which we introduced either PARP1^WT or a trapping-deficient PARP1^{del.p.119K120S} transgene ³ (Figure 1A), fused to eGFP for RIME or Apex2-eGFP for proximity labelling. We established single-cell clones that expressed the desired PARP1 fusion proteins (Extended Data 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 (Figure 1B, C), while expression of the PARP1^{del.p.119K120S}-eGFP did not. We also used a PAR-binding PBZ-mRuby2 probe and a UV microirradiation assay to demonstrate that PARP1^WT-Apex2-eGFP localised to DNA damage sites where it generated PAR. In the presence of PARPi, PARP1 was retained at the site of damage (Figure 1D).

Figure 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 to PARPi/MMS (to enable trapping) or MMS (no trapping) for 1 hour, after which PARP1 interacting/proximal proteins were identified by mass spectrometry analysis. B. Clonogenic assay illustrating the restoration of PARP inhibitor sensitivity in the complemented CAL51 PARP1 ^−/− cells as described in (A). PARP1 protein expression in the different clones is shown in Extended Data Figure 1A. C. Quantification of colony formation assay shown in (B); the mean of 2 biological replicas are shown. D. PARP1^WT-Apex2-eGFP protein localises to DNA damage, generates PAR and can be trapped by PARPi. PARP1^WT-Apex2-eGFP expressing cells were transfected with a PAR sensor, a PBZ PAR binding domain fused to mRuby2. PARP1^WT-Apex2-eGFP and PBZ-mRuby2 accumulate at the sites of UV microirradiation. Exposure to 100 nM talazoparib, causes sustained accumulation of the PARP1^WT-Apex2-eGFP, but abolishes PAR production. Data shown represents the 2 independent experiments with similar results. E and F. PARP1 interactions that are enriched under PARP1 trapping conditions (as defined by PSM ratio and MS scores). Scatter plots are shown for PARP1^WT-eGFP RIME (E) and PARP1^{del.p.119K120S}-eGFP RIME (F). G. PARP1 interactions that are enriched under PARP1 trapping conditions for PARP1^WT-Apex2-eGFP proximity labelling. RIME and proximity labelling have been performed in 3 independent experiments. H. A graph plotting the PSM against MS score for PARP1^WTApex2-eGFP proximity labelling interactions shows that p97 is among the most abundant proteins identified in the PARP1^WT-Apex2-eGFP proximity labelling.

As PARP1 translocates to chromatin upon DNA damage, we first used RIME-based immunoprecipitation ^{11, 13} , to identify proteins associated with trapped PARP1 (Figure 1A). PARP1^WT-eGFP and PARP1^{del.p.119K120S}-eGFP expressing cells were exposed to PARP1 trapping conditions (methyl methanesulfonate (MMS) + talazoparib) after which protein interactions were stabilised by formaldehyde crosslinking. MMS was used to create DNA lesions, whereas talazoparib traps DNA-bound PARP1. After trapping, chromatin-bound proteins were isolated and PARP1-associated complexes immunoprecipitated using GFP-Trap beads. Proteins were then identified by mass spectrometry. We also included an analysis of the parental PARP1 ^−/− cells lacking eGFP, in order to identify proteins that bind non-specifically to GFP-Trap beads (Extended Data 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 (Figure 1E, F). To prioritise proteins for further analysis, we used the MS score and the enrichment ratio of peptide spectrum matches (PSM) in the MMS + talazoparib exposed cells, compared to cells exposed to MMS alone. As expected, MMS + talazoparib increased the PARP1 PSM enrichment ratio in PARP1^WT-eGFP expressing cells but not in PARP1^{del.p.119K120S}-eGFP expressing cells (2.5 vs. 1.1 respectively). The trapping-defective PARP1^{del.p.119K120S}-eGFP mutant appeared to interact with the cohesion complex subunit PDS5A, regardless of the presence of PARP inhibitor (Figure 1F), suggesting that some interaction between mutant PARP1 and chromatin did exist independently from trapping. Both mutant and wild type PARP1 also appeared to interact with chromatin-associated proteins (e.g. CHD4), but when compared to PARP1^{del.p.119K120S}-eGFP, PARP1^WT-eGFP showed a relative enrichment of the Small Ubiquitin Modifier Proteins, SUMO1 (PSM ratio 5 in PARP1^WT vs. 1 in PARP1^{del.p.119K120S}) and SUMO2 (4 in PARP1^WT vs. not identified in PARP1^{del.p.119K120S}) (Figure 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 1C), which was further increased when PARP1 labelling was conducted under trapping conditions (Extended Data 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. Instead, we used PARP1^WT-eGFP-expressing cells to filter out non-specific interactions with beads. As a result of this filtering, we identified a higher number of proteins, 360, that 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 cut off (0.7) representing the trapped PARP1 interactome network (Extended Data 2A), was enriched in proteins associated with PARP1-mediated Base Excision Repair (BER), (PCNA, HMGB1, LIG3, PARP1 and POLE, p-value<0.01, Extended Data 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-characterised PARylation targets (e.g., PCNA, NCL, FUS, ILF3 ^{14, 15} ) strengthening the notion that we identified bona fide PARP1-proximal proteins. Of note, “protein processing in ER” (p-value<10^-3) and “proteasome” (p-value<0.01) appeared enriched in the gene set ontology analysis, observations we focus upon later in this manuscript. The MS score and PSM scores showed a positive correlation and identified that among the most abundant proteins were, among others, PARP1, p97/VCP, UBA1, TOP2A (Figure 1G, H). Proteins that showed high enrichment ratios in PARP1 trapping conditions e.g., USP7, were generally identified with a low MS score pointing to a low abundance. As trapping deficient mutant was not available to perform comparison similar to the RIME analysis, for the Apex2 analysis we prioritised 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 the ubiquitin-like modifier-activating enzyme 1 (UBA1), which has been previously implicated in ubiquitylation events at the sites of DNA damage ¹⁶ and the transitional endoplasmic reticulum ATPase, p97 (also known as valosin containing protein, VCP), which acts as a central component of a ubiquitin-controlled process. p97’s ATP-dependent unfoldase activity extracts proteins from chromatin prior to their proteasomal degradation or recycling ^{8–10, 17–19} . Furthermore, p97, working with cofactors that often contain ubiquitin binding domains (UBDs), recognises client proteins via ubiquitylation events, mostly those involving lysine-48 (K48) and lysine-6 (K6) ^{20, 21} ubiquitylation. p97 was also identified in the PARP1^WT, but not in the PARP1^{del.p.119K120S} RIME analysis, strengthening the notion 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, while our proximity labelling analysis identified ubiquitylation and p97 as being 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 which could conceivably represent SUMOylated and/or ubiquitylated PARP1, forms not present in the nuclear-soluble fraction (Figure 2A).

Figure 2. Trapped PARP1 is SUMOylated and ubiquitylated.

A. PARP1 trapping conditions elicit high MW forms of PARP1 in the chromatin fraction. PARP1-/- (KO) or PARP1 wild-type (WT) HEK293 cells were exposed to PARP1 trapping (MMS/Talazoparib), cells were fractionated into nuclear-soluble and chromatin-bound fractions. High 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 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 Ub-STREP-HA-expressing construct and exposed to combinations of MMS, talazoparib or 5 μM MLN-7243. The presence of high MW/Ub forms of PARP1 were reduced by MLN-7243 exposure (lane 7 vs lane 5). Input controls for these experiments are shown in Extended Data Figure 3A. D. Trapped PARP1 is SUMOylated. CAL51 PARP1^WT-eGFP-expressing cells were transfected with HA-SUMO2 expressing construct and subsequently they were treated with 0.01 % MMS or 100 nM talazoparib. HA-SUMO2-modified proteins were purified from the chromatin fraction and PARP1 was detected via immunoblotting. High exposure of blots of PARP1 SUMOylation are shown in Extended Data Figure 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 MLN-7243 (5 μM) or ML-792 (1 μM) exposed cells and the presence of high MW PARP1 isoforms identified by immunoblotting. F. PARP1 is modified and interacts with RNF4 in a SUMO-dependent manner. HEK293 WT or PARP1 ^−/− cells were exposed to trapping conditions either in the presence 5 μM MLN-7243 or 1 μM ML-792. Western blotting for PARP1 revealed that modified PARP1 isoforms were abrogated by ML-792, but not by MLN-7243 exposure. Abrogating SUMOylation prevented the association between PARP1 and RNF4, whereas inhibiting ubiquitinating stabilised the interaction. Data shown represent 2 biological replicas in A, B, D and F and 2 biological replicas in C and E.

We first assessed whether ubiquitylated PARP1 was present in cells exposed to PARPi with different trapping properties. We used either: (i) the potent PARP1 trapper, talazoparib; (ii) veliparib, a clinical PARPi that effectively inhibits PARP1 catalytic activity, but which has minimal trapping properties; or (iii) a recently described structural derivative of veliparib, UKTT15, that is able to elicit PARP1 trapping ⁴ . Cells were exposed to MMS + PARPi, after which the ubiquitylated pool of proteins was isolated from the chromatin fraction via HA-Strep-ubiquitin isolation. In this fraction, high molecular weight isoforms of PARP1 were more prevalent in talazoparib or UKTT15-exposed cells, when compared to veliparib-exposed cells (Figure 2B), suggesting that PARP1 ubiquitylation was enhanced by PARP1 trapping. We repeated the HA-Strep-ubiquitin pulldown experiment in the presence of the E1 ubiquitin activating enzyme inhibitor, MLN-7243 (TAK243). Western blotting with an anti-PARP1 antibody revealed that trapping conditions led to the formation of high molecular weight PARP1 isoforms; these were almost completely abolished when cells were exposed to MLN-7243, consistent with these high molecular weight isoforms representing ubiquitylated PARP1 (Figure 2C, Extended Data 3A). The poly-ubiquitylation of PARP1 was also observed in reciprocal denaturing IP experiments, where PARP1 was immunoprecipitated from HEK293 cells transfected with a FLAG-PARP1 cDNA-expression construct (Extended Data 3B). We also identified poly-ubiquitin chains on PARP1 that were linked by K48 linkage (Extended Data 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 (Figure 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 ²² (Figure 2D, Extended Data 3D). However, this was not to the same extent as seen under PARP1 trapping conditions. This suggested that PARP1 is SUMOylated when it becomes trapped. Interestingly, incubating cells grown in PARP1 trapping conditions in the presence of a SUMOylation inhibitor (ML-792, which inhibits SUMO-activating enzyme) decreased high MW forms of ubiquitylated PARP1 (Figure 2E), suggesting PARP1 ubiquitination upon trapping could require prior PARP1 SUMOylation. Conversely, a ubiquitylation inhibitor had no effect on PARP1 SUMOylation (Figure 2F), suggesting that the SUMOylation of trapped PARP1 is required for its ubiquitination, but that the ubiquitination of trapped PARP1 is not a pre-requisite 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 ²³ and RNF4, has previously been implicated in modulating PARP1’s transcriptional activity ²⁴ as well as being involved in repairing topoisomerase cleavage complexes, which also represent a “trapped” nucleoprotein complex ²⁵ . Chromatin co-immunoprecipitation of trapped PARP1 showed an increased interaction with RNF4 (Figure 2F) consistent with our hypothesis. This PARP1/RNF4 interaction was reduced upon inhibition of SUMOylation and stabilised in cells exposed to a ubiquitylation inhibitor, indicative of a ligase-substrate interaction (Figure 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 HCT116 PIAS4 ^−/− and MCF7 RNF4 ^−/− cell lines ²⁵ . 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 anti-SUMO2/3 antibody revealed that PIAS4 is necessary for efficient SUMOylation and ubiquitylation of trapped PARP1 (Figure 3A, Extended Data 4A, B and C). Re-expressing 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 ²⁶ (Figure 3B and C, Extended Data 4D). Interestingly, in RNF4 ^−/− cells, while PARP1 ubiquitylation was decreased (confirming that RNF4 activity is responsible for this modification), PARP1 SUMOylation was increased (Figure 3D, Extended Data 4E, F and G). Re-expressing wild-type RNF4 in RNF4 ^−/− cells also reversed these effects, but this was not the case in cells expressing SIM (SUMO-interacting motifs)-deleted ²⁷ or catalytically inactive p.H156A mutant forms of RNF4 (Figure 3E and F, Extended Data 4H). We also observed strong RNF4-dependent PARP1 ubiquitination by overexpressing wild-type RNF4 in cells cultured in MMS + PARPi (Extended Data 5A), an effect not seen 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 up to 80-95% of the ubiquitylation of trapped PARP1. We also found that RNF4 gene silencing reduced ubiquitylation of trapped PARP1 (Extended Data 5B). Taken 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.

Figure 3. Trapped PARP1 is modified in a PIAS4- and RNF4-dependent manner.

A. PARP1 is SUMOylated in a PIAS4-dependent manner in vivo. HCT116 wild-type or PIAS4 ^−/− cells were transfected with FLAG-PARP1 expressing plasmid, exposed to trapping, and chromatin-bound PARP1 investigated for SUMOylation and ubiquitylation. In PIAS4 ^−/− cells, SUMO1 (Extended Data Figure 4A), SUMO2 and ubiquitin were reduced (total ubiquitin in Extended Data Figure 4B and quantification in Extended Data Figure 4C). Data shown represent 2 independent experiments with similar results. B. HCT116 PIAS4 ^−/− cells were transfected with PIAS4-expressing plasmids (EV: empty vector, WT: wild type, SAP: SAP domain deleted, C342A catalytic dead) for 48 hours, followed by 30 min talazoparib (10 μM) treatment in the presence of 0.01% MMS and PARP1 immunoprecipitation. C. A quantification of the abundance of SUMO2/3 (top) and ubiquitin (bottom) modified PARP1 in (B). 2 biological replicates are shown. D. Similar to (A), trapped PARP1 was purified from MCF7 wildtype or RNF4 ^-/- cells. In RNF4 ^-/- cells, PARP1 amount ubiquitylation was reduced, whilst SUMO1- (Extended Data Figure 4E) and SUMO2-ylation was increased. Total ubiquitin input is shown in Extended Data Figure 4F and quantification of the blots is presented in Extended Data Figure 4G. Data shown represent 2 independent experiments with similar results. E. MCF7 RNF4 ^−/− cells were transfected with RNF4-expressing plasmids (EV: empty vector, WT: wild type, SIM: SUMO-interacting motif deleted, H156A catalytic dead) for 48 hours and processed as in (B). F. A quantification of SUMO2/3 (top) and ubiquitin (bottom) modified PARP1 in (E). 2 biological replicates are shown. G. PIAS4 mediates PARP1 SUMOylation in vitro. Recombinant PARP1 was incubated with nicked DNA, SUMO1 or SUMO2, SAE1/2 (SUMO E1), Ubc9 (SUMO E2) and an increasing concentration of PIAS4. PIAS4 led to a concentration-dependent increase of SUMOylation (Extended Data Figure 5C). Free SUMO2 is indicated by*. H. RNF4 mediates PARP1 ubiquitylation in a SUMO-dependent manner in vitro. Similar to (G), PARP1 SUMOylation reactions were supplemented with ubiquitin, Ube1 (E1), Ubc5H (E2) and an increasing concentration of RNF4. SUMOylated PARP1 was a better substrate for ubiquitylation. Free ubiquitin is indicated by*. Data shown represent 2 independent experiments with similar results.

Finally, we tested the interdependency of PARP1 SUMOylation and ubiquitylation events using in vitro SUMOylation and ubiquitylation reactions. Incubating 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 (Figure 3G, Extended Data 5C). Adding ubiquitin, UBE1 (Ub E1), Ubc5H (Ub E2) and an increasing concentration of RNF4 to this reaction led to efficient PARP1 ubiquitylation (Figure 3H). In contrast, RNF4 displayed much lower PARP1 ubiquitylating activity in the absence of SUMOylation (Figure 3H). Collectively, these data suggested a stepwise process, where upon 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 hypothesised 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 assessed the interaction between p97 and PARP1 using both Proximity Ligation Assays (PLA) (Figure 4A) and co-immunoprecipitation experiments (Extended Data 6A) which 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 (Figure 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 6B).

Figure 4. PARP1 interacts with p97 in a trapping-dependent manner.

A. Proximity Ligation Assay (PLA) of endogenous PARP1 and p97 in CAL51 cells; scale bar = 5 μm, Data shown represent 3 biological replicas. B. PARP1-p97 interaction is increased upon DNA damage. PARP1^WT-eGFP or PARP1^{del.pK119S120}-eGFP expressing CAL51 cells were exposed to trapping conditions and PARP1-GFP was immunoprecipitated under native conditions. Data shown represent 2 biological replicas. C. PARP1-p97 PLA in CAL51 cells expressing either PARP1^WT-eGFP or PARP1^{del.pK119S120}-eGFP; n=2016 cells from n=3 independent experiments; shown is the geometric mean ± 95CI, ordinary one-way ANOVA, **** - p < 0.0001. D. PARP1-p97 PLA in CAL51 cells under trapping. PLA with p97 antibody alone (top) or p97+PARP1 antibody (bottom); scale bar = 5 μm, Data shown represent 3 biological replicas. E. Quantification of PLA foci/nucleus in (D), n=2035 cells from three independent experiments; shown is the geometric mean ± 95% CI, ordinary one-way ANOVA, **** - p < 0.0001. F. p97 inhibition increases the presence of ubiquitylated PARP1. Ubiquitin-STREP-HA-expressing HEK293 cells were cultured in PARP1 trapping conditions in the presence or absence of 10 μM CB-5083 and ubiquitylated proteins were immunoprecipitated under denaturing conditions. Input controls in Extended Data Figure 6C, n=3 biological replicas G. HEK293 cells expressing either p97-Myc p.E578Q-Myc were transfected with FLAG-PARP1 construct, exposed to trapping conditions and PARP1 immunoprecipitated from the chromatin fraction; Data shown represent 2 biological replicas. H. CAL51 PARP1^WT-eGFP or PARP1^{del.p.119K120S}-eGFP expressing cells were transfected with p97-E578Q-Strep-MYC for 18 h, exposed trapping and then fractionated. Chromatin PARP1-eGFP IP was probed by antibody that detected both endogenous and ectopically-expressed p97; Data shown represent 2 biological replicas. I. PARP1/p97 co-localisation is reduced by ubiquitylation (5 μM MLN-7243) or SUMOylation (1 μM ML-792) inhibitors. PARP1-p97 PLA foci/nucleus from n=1316 cells, n=3 independent experiments; shown is the geometric mean ± 95CI, ordinary one-way ANOVA, **** - p < 0.0001. J. The p97 adapter UFD1 mediates the interaction between p97 and trapped PARP1 - chromatin-bound co-IP; Data shown represent 3 biological replicas. K. Similar to (J), the PARP1/p97 interaction is disrupted by the p97 sequestration agent, CuET; Data shown represent 2 biological replicas.

CB-5083 is a small molecule which inhibits p97 ATPase activity and induces a p97 substrate trapping effect ²⁸ ; we found that CB-5083 caused an increase in PARP1-p97 interaction (Figure 4D, E), suggesting that PARP1 could be a p97 substrate. Blocking p97 catalytic activity leads to the accumulation of ubiquitylated isoforms of its substrates ^29,30 , which was also the case for trapped PARP1 (Figure 4F). This was also observed using reciprocal immunoprecipitation of PARP1 under denaturing conditions (Figure 4F, Extended Data 3B and C). We reproduced the substrate trapping effect of p97 inhibition by expressing a dominant negative ATPase deficient p97 mutant, p.E578Q ^{17, 31, 32} (Figure 4G), consistent with trapped PARP1 being a p97 substrate. Furthermore, we demonstrated that the p97-PARP1 interaction was enhanced by expressing the p97 p.E578Q mutant in PARP1 ^−/− cells reconstituted with wild-type, but not trapping-defective PARP1^{del.pK119S120}, (Figure 4H). This conclusion was further supported by co-localisation immunofluorescence experiments where p97 p.E578Q and trapped PARP1 foci ³³ were found to substantially overlap (Extended Data 6D).

Ubiquitylation is a mediator of p97 interactions ^{6, 7} . Indeed, when cells were exposed to ubiquitylation (MLN-7243) or SUMOylation (ML-792) inhibitors (which decreased trapped PARP1 ubiquitylation, Figure 2E), the PARP1/p97 interaction was reduced (Figure 4I and Extended data 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 (UBDs) in both NPL4 and UFD1 ^34,35 . When UFD1 was depleted, the interaction between trapped PARP1 and p97 was reduced (Figure 4J). This was not the case when NPL4 was depleted, although, as expected, depletion of either subunit reduced overall p97 recruitment to chromatin (Figure 4J). Furthermore, only depletion of UFD1 led to a profound accumulation of trapped PARP1 (Figure 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 appear consistent with previous work suggesting that the NPL4 and UFD1 can recognise substrates independently of each other ^{7, 36,9} . 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 ZnF 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 to CB-5083 and also NPL4 or UFD1 gene silencing. We found that the PARP1-p97 interaction was almost completely abrogated by CuET exposure (Figure 4K). Taken together, these observations suggested that the p97 system and its ubiquitin binding cofactor UFD1 (p97-UFD1), recognises and physically interacts with trapped PARP1.

Trapped PARP1 is modulated by p97 activity

To assess whether p97 removes trapped PARP1 from chromatin, we used a “trap-chase” experimental approach (Figure 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”). At various time points during the chase, the amount of trapped PARP1 was evaluated either by chromatin fractionation or by measuring the proximity of PARP1 to phosphorylated H2AX (γH2AX proximity ligation assay, PLA ³⁹ ). Initially, we followed the kinetics of trapped PARP1 in PIAS4 ^−/− and RNF4 ^−/− cells (Figure 3A and 3D). Both PIAS4 ^−/− and RNF4 ^−/− cells showed slower resolution of chromatin bound PARP1, especially at the later time points (Figure 5B, C, Extended Data 7A, B), consistent with the notion that these SUMO/ubiquitin ligases promote the resolution of trapped PARP1 complex.

Figure 5. PARP trapping is modulated by the PIAS4-RNF4-P97/VCP axis.

A. Schematic of the trap-chase experiment. B. Trapped PARP1 is processed in a PIAS4-dependent manner. A trap-chase experiment in HCT116 wild-type or PIAS4 ^−/− cells. After PARP1 trapping, cells were chased in talazoparib-containing media. Samples were collected at indicated time point for chromatin fractionation and Western blotting. Data shown represent 2 biological replicas. C. Trapped PARP1 is processed in a RNF4-dependent manner. A trap-chase experiment in MCF7 wild-type or RNF4 ^−/− cells similar to (B); Data shown represent 2 biological replicas. D. Representative confocal microscopy images from a PARP1-γH2AX PLA trap-chase experiment, scale bar = 5 μm. E. PARP1-γH2AX PLA foci persist in cells chased in PARPi plus p97 inhibitors. Quantification of PARP1-γH2AX PLA foci from the trap-chase experiment in (D), n=5736 cells n=3 independent experiments; shown is the geometric mean ± 95CI, ordinary one-way ANOVA, **** - p < 0.0001. F. PARP1-γH2AX PLA foci persist in cells with RNF4 silencing. Quantification of PLA foci/nucleus in n=1235 cells from three independent experiments; shown is the geometric mean ± 95CI, ordinary one-way ANOVA, **** - p < 0.0001. G. PARPi-induced RAD51 and γH2AX foci persist in the presence of p97 inhibitors. Representative confocal microscopy images from a trap-chase experiment (trap: talazoparib overnight, chase: p97 inhibitor-containing media) are shown. The cells were stained for the presence of γH2AX and RAD51 foci. Representative images for each condition, scale bar = 5 μm. H and I. Quantification of γH2AX (H) and RAD51 foci (I), from experiment (G). Quantification in n=1750 cells from n=3 independent experiments; shown is the geometric mean ± 95 % CI, ordinary one-way ANOVA, **** - p < 0.0001.

Secondly, we investigated the role of p97 activity with this assay, by including talazoparib plus CB-5083 or CuET in the chase phase of the experiment, monitoring trapped PARP1 either via chromatin fractionation (Extended Data 7C) or PLA (Figure 5D). After exposing cells to MMS + talazoparib, a significant amount of PARP1 was detected in the proximity of γH2AX (Figure 5D), indicating the “trapping” part of the experiment was successful; after removing the trapping agents, the amount of trapped PARP1 decreased (e.g., the PARP1/γH2AX PLA signal disappeared). When cells were chased in the presence of single agent PARPi or p97 inhibitor, the PARP1/γH2AX PLA signal also diminished. Conversely, when cells were chased in the presence of both PARPi (talazoparib) and p97 inhibitor (either CB-5083 or CuET), the amount of trapped PARP1 persisted (Figure 5D, E). Consistent with the notion 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 (Figure 5F). We also assessed the effect on PARP1 trapping by the expression of a dominant negative RNF4 p.M136S/R177A mutant (Extended Data 7D), a p97 p.E578Q mutant (Extended Data 7E, F and G) or UFD1 depletion (Extended Data 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-hour exposure of cells to PARPi elicited both γH2AX and RAD51 foci but when PARPi was removed from culture media by washing, γH2AX and RAD51 foci diminished after three hours, suggesting resolution of the DNA damage caused by trapped PARP1 (Figure 5G-I). When we used p97 inhibitors (CB-5083 or CuET) in this “chase” period, 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 then p97 inhibitor were indeed caused by PARPi. The effects on foci resolution were also not trivially explained by alterations in the cell cycle as three hours exposure of cells to p97 inhibitor did not lead to significant changes in cell cycle distribution (Extended Data 8A, B). We also noted that when we used FRAP (fluorescence recovery after photobleaching) 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 t_1/2 in presence of talazoparib = 4.9 ± 1.3 s vs. t_1/2 = 7.8 ± 1.4 s in the presence of talazoparib + CB-5083, two-sided t-test p-value < 0.05, Extended Data 8C, D).

p97 inhibition potentiates PARP inhibitor cytotoxicity

Based on the prolonged PARP1 trapping effects described above, we hypothesised 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, olaparib) and observed a dose dependent potentiation of the clonogenic effect of each PARPi by the presence of p97 inhibitor (Figure 6A-C). Bliss independence analysis confirmed that these drugs elicited supra-additive effects when used in combination (Extended Data 9A, B). This combinatorial effect was PARP1 trapping dependent as it was reversed in PARP1 ^−/− cells (Figure 6A, Extended Data 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 prior PARPi combinatorial experiments, did not enhance sensitivity to the alkylating agents MMS or temozolomide in either PARP1^{wild type} or PARP1^−/− cells (Figure 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, CB-5083/CuET’s ability to sensitise to PARPi.

Figure 6. p97 inhibition potentiates the effect of PARP inhibitors.

A. p97 inhibition potentiates the cytotoxicity of PARP inhibitors. CAL51 cells were exposed to PARP inhibitor (talazoparib or olaparib) in the presence of p97 inhibitor (CB-5083 or CuET) for a period of 14 days. Images are shown for the 100 nM CB-5083 and 8 nM CuET exposed samples. Drug response curves are shown in (B), (C) and Extended Data Figure 9A, B. Shown are the mean ± SD, n=3 biological replicas. D. and E. DNA alkylating agents that are used to induce PARP1 trapping do not enhance the cell inhibitory effects of CB-5083. CAL51 WT or PARP1 ^-/- cells were exposed to alkylating agents MMS (D) or temozolomide (TMZ) (E) in combination with either talazoparib (the positive control) or CB-5083 for seven days, after which, cell viability was measured. Shown are the mean ± SD, n=3 biological replicas. F. CB-5083 modulates the synthetic lethal effect of PARPi in BRCA2 ^−/− cells. Survival curves from clonogenic survival assays in DLD1 BRCA2^wild-type and DLD1 BRCA2 ^−/− cells. Colony formation images and quantification are shown in Extended Data Figure 9E. Shown are the mean ± SD, n=3 biological replicas. G. p97 inhibition sensitises mouse cancer organoid cells to PARPi. Brca1/p53 mutant WB1P breast cancer organoids were grown in the presence of the indicated drugs for seven days. Shown are the mean ± SD, n=3 biological replicas. Brightfield images of organoids are shown in Extended Data Figure 9F. H. p97 inhibition sensitises a human BRCA1 mutant patient-derived breast cancer organoid to PARPi. KCL014BCPO organoids were grown in the indicated drugs for seven days. Shown are the mean ± SD, n=3 biological replicas. Brightfield images of organoids are shown in Extended Data Figure 9G. I. A 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.

Because PARPi are approved for the treatment of cancers that have HR defects, and because trapped PARP1 is the key cytotoxic event in HR-defective cells, we assessed the effect of combined exposure to CB-5083 + talazoparib in DLD1 cells with/without genetic ablation of BRCA2. CB-5083, when used alone, had a modest BRCA2 synthetic lethal effect (Extended Data 9E) but when used in combination with talazoparib, had a far greater effect on DLD1 BRCA2 ^−/− cells than in isogenic BRCA2 wild-type cells (Figure 6F and Extended Data 9F, G). In tumour organoids derived from mice with combined Brca1/Tp53 loss-of-function mutations (WB1P ⁴¹ ) we found that CB-5083 further sensitised tumour organoids to talazoparib (Figure 6G and Extended Data 9H). We also assessed CB-5083 in combination with PARPi in a human patient-derived tumour organoid culture (PDO) derived from a triple negative breast cancer patient harbouring a germline pathogenic BRCA1 p.R1203* mutation (BRCA1 c.3607C>T), which was homozygous in the organoid. CB-5083 led to marked shift in talazoparib sensitivity (Figure 6H and Extended Data 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 upon 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 (Figure 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 (e.g. the DUB, ATXN3 ⁴² and the E3 ubiquitin ligase TRIP12 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: (i) the p97 inhibitors, when employed as single agents, did not elicit biomarkers of DNA damage such as γH2AX or RAD51 foci (Figure 5 G-I), nor alterations in cell cycle dynamics (Extended Data 8A, B); (ii) p97 inhibitors enhanced PARPi sensitivity in a PARP1-dependent manner (Figure 6A,B), p97 inhibitor did not alter sensitivity to MMS or temozolomide (Figure 6C,D).

We also see a number of new questions that might now arise from our observations. Firstly, although PIAS4 and RNF4 appear to act in a linear manner, there remains the possibility that the balance of SUMOylation and ubiquitylation is influenced by other E3 ligases. Indeed, the effect of losing PIAS4 on trapped PARP1 resolution was modest (Figure 5B) suggesting other proteins might also be involved. Secondly, our data also suggest that UFD1 is required for the recruitment of p97 to trapped PARP1 (Figure 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 nor the PARP1-p97 interaction, whilst UFD1 depletion did (Figure 4J). Whilst 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, 36,9} , appears to be dependent on UFD1 only. Thirdly, in most systems the p97-dependent removal of ubiquitylated proteins is coupled to proteasomal degradation, but this was not the case for PARP1. p97 is also known to participate in substate recycling, as is the case for Aurora B ⁸ , yeast transcriptional repressor alpha ¹⁹ , Ub-LexA-VP16 ¹⁸ or MRE11 ⁴⁴ . This raises the possibility that PARP1 might also be a p97 substate that is recycled, not degraded.

Finally, the PARPi-generated DNA lesions appear to be processed in a analogous fashion to trapped TOP1-cleavage complexes ¹⁰ . 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 ⁴⁵ ; both are SUMOylated, ubiquitylated and modified by p97 (reviewed in ⁴⁶ and data shown here). Thus it seems plausible that the sensing and processing machinery that activate the SUMOylation and ubiquitylation of trapped PARP1 and 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.

Materials and 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 ATCC) cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FBS, 1x Penicillin-Streptomicyn (Sigma-Aldrich). CAL51 PARP1^-/- cells were previously described ¹⁴ . They were transfected with a corresponding PARP1-expressing piggyBac construct in combination with hyPBase-expressing plasmid ⁴⁷ . 72 hours after transfection, single-cell clones were sorted by FACS and allowed to expand. These clones were characterized for the expression of the tagged protein by microscopy and western blotting. HEK293 PARP1 ^−/− cells were a kind gift from Ivan Ahel, University of Oxford ⁴⁸ . The HCT116 PIAS4 ^−/− and MCF7 RNF4 ^−/− cells were previously described ²⁵ . WB1P organoid line was previously described ⁴⁹ . They were grown in a mix of 50% Matrigel (Corning) and 50% Advanced DMEM/F12 (Life Technologies) 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 EGF (Life Technologies). 24 hours prior to drug organoids were seeded at 10000 cells per well of a 24 well plate and drugs added at the indicated concentrations. Cell viability was assessed using 3D cell-titer glow (Promega). KCL014BCPO was derived (Badder 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. 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. Resulting cell pellets were then plated into 3D cultures at approximately 1 × 10³ to 2 × 10³ cells per μl in Ocello PDX medium (OcellO B.V) and hydrogel. All cultures were maintained in humidified incubators at 37°C, 5% CO₂. All human cell line identities were confirmed by STR typing and verified 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 PARP1-Apex2-eGFP construct, the Apex2 gene was amplified from Addgene vector 49386 and inserted inbetween PARP1 and eGFP coding sequences via InFusion (Clonetek, 648910). PBZ-mRuby2 is described in ¹⁴ . UB-STREP-HA was a kind gift from Vincenzo D’angiolella, HA-SUMO2 ⁵² , FLAG-PARP1 was a kind gift from Ivan Ahel, p97-GFP was a kind gift from Hemmo Mayer.

The wild-type PIAS4 expressing construct was obtained from Addgene (#15208) and RNF4 from Origene (RC207273). The corresponding SAP and SIM mutants were generated as described ²⁵ . The RNF4-M136S,R177A was a kind gift from Ronald Hay. Antibodies used were: GFP (Sigma-Aldrich, 11814460001, clones 7.1 and 13.1, dilution WB 1:5000 IF 1:500 PLA 1:1500); PARP (CST, 9532, 46D11, dilution 1:2000) for immunoblotting and PLA; p97 (Abcam, ab11433 [5], dilution WB 1:1000 PLA 1:2000) for immunoblotting and PLA; PAR (Trevigen, 4335-AMC-050, dilution WB 1:1000); HA (Roche, 11867423001, dilution WB 1:5000); FLAG (M2, Sigma-Aldrich, F1804, dilution WB 1:5000) for immunoprecipitation; FLAG (Sigma-Aldrich F7425, dilution WB 1:5000) for immunoblotting; Streptavidin-HRP (ThermoFisher, S911, WB dilution 1:1000); PARP1 (Sigma-Aldrich, WH0000142M1, dilution WB 1:1000 PLA 1:2000) for PLA; β-actin (Invitrogen, AM4302, dilution WB 1:5000); lamin-B1 (Thermo, PA5-19468, dilution WB 1:5000); vinculin (Abcam, ab18058. dilution WB 1:5000); phospho-H2AX (CST, 9718S, dilution 1:2000) for PLA; phospho-H2AX (Millipore, 05-636, dilution 1:1500) for foci immunostaining; RAD51 (Abcam, ab133534, dilution 1:1500) for foci immunostaining; Histone H3 (CST, 9715, dilution WB 1:5000); SUMO1 (CST, 4940, dilution WB 1:1000); SUMO2/3 (CST, 4971, dilution WB 1:1000); ubiquitin (Santa Cruz Biotechnology, sc-8017, dilution WB 1:1000); RNF4 (Novusbio, NBP2-13243, dilution WB 1:1000); UFD1L (Abcam, ab181080, dilution WB 1:1000); Anti-Rabbit IgG HRP (Rockland, 18-8816-31, dilution WB 1:5000). Talazoparib was supplied by Pfizer as part of the BCN Catalyst programme. Other small molecules were as follows: 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 ; MLN-7243 (Selleckchem, S8341); ML-792 (Medchemexpress, HY-108702). siRNAs were obtained from Dharmacon: RNF4 (L-006557-00-0005 and 3’UTR siRNA sequence 5’-GGGCAUGAAAGGUUGAGAAUU); UFD1L (L-017918-00-0005); NPL4 (L-020796-01-0005).

Western blotting

Standard protocols for sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS PAGE) and immunoblotting were used ⁵³ . Nitrocellulose membrane (GEHealthcare) or PVDF (BioRad) were used to transfer proteins from polyacrylamide gels depending on the antibody.

Cellular fractionation immunoprecipitation

Cells were washed two times with PBS 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 achieve a final concentration of 0.1% and left on ice for 2-5 min depending on cell line. The supernatant was harvested as the cytosolic fraction and the pellet (nuclei) was then washed two times 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 nuclei, after which lysates were kept on ice for 10 min. Supernatant was then removed as the nuclear soluble fraction. Remaining chromatin pellet was then washed with buffer B in 0.5% Triton X-100 followed by benzonase buffer without MgCl₂ (50 mM Tris-HCl pH 7.9, 100 mM NaCl). Benzonase digestion buffer (50 mM Tris-HCl pH7.9, 10 mM MgCl₂, 100 mM NaCl, protease and phosphatase inhibitors, NEM) supplemented with 125 U of Benzonase enzyme (Merk Millipore) and rotated on wheel at 4°C for 1 hour. Samples were then centrifuged at 20,000 g for 15 min, chromatin input for the immunoprecipitation reaction was then taken from supernatant. The remaining supernatant was then incubated with respective beads on a rotating wheel for 3 hours at 4°C. For native IPs, 1:200 ethidium bromide was added to remove unwanted DNA-protein interactions. Beads were then washed with IP wash buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.5 mM EDTA, 0.05% Triton X-100) three times before elution with lamelli 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 wheel at 4°C for 10 min. The supernatant was removed and pellet was washed with benzonase buffer once and the supernatants were pooled together. Benzonase buffer (50 mM Tris-HCl pH 7.9, 10 mM MgCl₂, 100 mM NaCl, protease and phosphatase inhibitors, NEM) supplemented with 125 U of Benzonase enzyme (Merk Millipore) was added to the pellet and left on wheel at 4°C for 1 hour. Samples were then centrifuged at 20,000 g and all supernatants were pooled. Input for the IP was then removed and samples were incubated with respective beads on wheel for 3 hours at 4°C. Beads were then washed with IP wash buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.5 mM EDTA, 0.05% Triton X-100) three times before elution with Laemmlli buffer.

Denaturing IP

Cells were lysed either according to fractionation immunoprecipitation or whole cell immunoprecipitation protocol as stated above. Before incubation with beads, SDS was added to samples to a concentration of 1% and boiled at 95°C for 5 min. 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 wheel at 4°C for 3 hours. Beads were then washed with IP wash buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.5 mM EDTA, 0.05% Triton X-100) three times before elution with Laemmlli buffer.

Cell viability and clonogenic survival assays

The viability of cells was measured after six days exposure to various concentrations of drugs using the Cell Titre-Glo assay (Promega). Long-term drug exposure effects were assessed by colony formation assay after 12–14 days exposure to a drug-containing medium (refreshed weekly) and cells stained at the end of the assay with sulforhodamine B. When plotting survival curves, the surviving fraction was calculated relative to DMSO (solvent)-exposed cells.

The viability of the KCL014BCPO organoid line was measured using the 3D Cell Titre-Glo assay (Promega). Organoids were seeded in 24-well plates, with one 15 μl Matrigel droplet containing 3000 cells per well. 24 hours after seeding, the organoids were treated with a drug-containing media (drug refreshed after 4 days) for 7 days, before assessing the viability by measuring 3D Cell Titre-Glo luminescence.

Apex2-mediated Proximity Labelling

For each condition tested, 5-10 x 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 hour. In the last 30 min of the incubation, 500 μM final concentration biotin-tyramide (Sigma-Aldrich, SML2135) was added to the media. To label proteins, 1 mM final concentration H₂O₂ (Sigma-Aldrich, H1009) was added for 60 s. The reaction was quenched by washing the cells tree 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 not containing NaCl to final 200 mM NaCl, sonicated for 1 min and incubated with 250 U benzonase for 20 min at room temperature. The lysates were clarified by centrifugation at 13000 g for 15 min at 4°C. 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 with RIPA lysis buffer, once with 1 M KCl, once with 0.1 M Na₂CO₃, once with 2 M urea, twice with RIPA lysis buffer and processed further for MS analysis.

Rapid immunoprecipitation mass spectrometry of tagged protein (RIME-based approach)

For each condition tested, 5-10 x 10 ⁶ cells were exposed to either 0.01% MMS or a combination of 0.01% MMS + 100 nM talazoparib for 1 hour. At the end of the incubation period, formaldehyde (ThermoFisher, 28908) was added to the media to 1% final concentration and incubated for 10 min at room temperature. The reaction was quenched by the addition of 125 mM glycine final concentration. The cells were collected and washed once in ice-cold PBS. The cells were resuspended in ice cold PBS containing 0.1% Triton X-100 and protease inhibitor cocktail (Merck, 4693116001). The nuclei were spun at 3000 g for 5 min at 4°C, and subsequently resuspended in PBS containing 1% IGEPAL CA-630 (Sigma-Aldrich) and protease inhibitors, and incubated on ice for 15 min. The remaining chromatin was spun at 13000 g for 5 min at 4°C and resuspended in PBS containing 0.1% IGEPAL CA-630 and protease inhibitors. The chromatin pellet was spun at 13000 g for 5 min at 4°C, resuspended in lysis buffer (20 mM HEPES pH 7.5, 150 mM NaCl, 0.5% sodium deoxycholate, 0.1% SDS, 10 mM MgCl₂) supplemented with 250 U benzonase (Sigma-Aldrich, E1014) and incubated for 30 min at room temperature with rotation to release the chromatin bound proteins. The supernatant was isolated after centrifugation (13000x g for 10 min at 4°C) and incubated with 25 μl GFP Trap (Chromotek, gtm-20) magnetic beads for 1 hour at 4°C with rotation. The beads were washed four times with the lysis buffer and processed further for MS analysis.

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 sequencing grade trypsin (Roche) overnight at 37°C. The peptide solution was neutralised with 5% formic acid, acetonitrile was added to 60% final concentration and the solution was filtered through a Millipore Mutiscreen HTS plate (pre washed with 60 % acetonitrile). The peptide solution was lyophilised on a SpeedVac and the peptides were dissolved in 20 mM TCEP-0.5% formic acid solution. The LC-MS/MS analysis was conducted on the Orbitrap Fusion Tribrid mass spectrometer coupled with U3000 RSLCnano UHPLC system (ThermoFisher). The peptides were first loaded on a PepMap C18 trap (100 μm i.d. x 20 mm, 100 Å, 5 μm) at 10 μl/min with 0.1% FA/H₂O, and then separated on a PepMap C18 column (75 μm i.d. x 500 mm, 100 Å, 2 μm) at 300 nl/min and a linear gradient of 4-32% ACN/0.1% FA in 90 min with the cycle at 120 min. Briefly, the Orbitrap full MS survey scan was m/z 375–1500 with the resolution 120,000 at m/z 200, with AGC (Automatic Gain Control) set at 40,000 and maximum injection time at 50 ms. Multiply charged ions (z = 2 – 5) with intensity above 8,000 counts (for Lumos) or 10,000 counts (for Fusion) were fragmented in HCD (higher collision dissociation) cell at 30% collision energy, and the isolation window at 1.6 Th. The fragment ions were detected in ion trap with AGC at 10,000 and 35 ms maximum injection time. 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 with Proteome Discoverer 1.4 (Thermo). Database searches were carried out using Mascot (version 2.4) against the Uniprot human reference database (January 2018, 21123 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 fragment mass tolerance set to 0.5 Da. Acetylation at the N terminus, oxidation of methionine, carbamidomethylation of cysteine, and deamidation of asparagine and glutamine were set as variable modifications. Peptide identifications were set at 1% FDR using Mascot Percolator. Protein identification required at least one peptide with a minimum score of 20. For the Apex2-based proximity labelling MS the following steps were taken. Proteins identified with a single peptide were removed from further analysis. PARP1-eGFP MS 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 analysis. Peptide Spectrum Matches (PSM) was used as a proxy of protein abundance in the samples. A ratio was built between 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 (i.e. no detection in the sample) a value of one was added in order to calculate a meaningful ratio (the data is 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 KEGG 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) (the data is provided in Supplementary Table 2). For RIME analysis the proteins identified with single peptides were removed from further analysis. Proteins identified in CAL51 PARP1 ^−/− cells were considered as background and removed from further analysis when they were identified with more than two unique peptides. Subsequently the MS 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 (i.e. no detection in the sample) a value of one was added in order to calculate a meaningful ratio (the data is provided in Supplementary Table 3 and 4 for PARP1^WT-eGFP or PARP1^{del.p.119K120S}-eGFP, respectively).

Proximity ligation assay (PLA)

Proximity ligation assays were carried out with Duolink® In Situ Red Starter Kit Mouse/Rabbit (Sigma-Aldrich) according to the manufacturer’s protocol. The primary antibodies used were mouse anti-PARP1 (WH0000142M1-100UG), rabbit anti-PARP (Cell Signalling), mouse anti-p97 (ab11433) and rabbit anti-phospho-H2AX (Cell Signalling). The antibodies were used in 1:1500 dilution. Images were acquired on Marianas advanced spinning disk confocal microscope (3i) and analysed with a custom CellProfiler pipeline. Typically, several hundred nuclei were counted per condition from at least two independent biological repeats.

Microirradiation

Cells were grown in glass-bottom culture dishes (MaTek,P35G-0.170-14-C) in 10% FBS DMEM media and maintained at 37 °C and 5% CO2 in an incubation chamber mounted on the microscope. Imaging was carried out on Andor Revolution system, Å~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 within the nucleus. After the signal of recruitment reached its maximum (typically 30 s to 60 s after microirradiation) the recruitment spot was bleached with a 488 nm laser and imaging continued with one frame per 2 s intervals. For each experimental condition ten to twelve cells were acquired and the experiment was repeated independently on a different imaging day. From the raw intensities of the microirradiation site, the spot intensity immediately prior to bleaching was set to 1 and immediately after bleaching to 0. The recovery data was fitted with one site-specific binding model of non-linear 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. One hour prior to fixation 10 μM ethylene-deoxyuridine (EdU, Thermofisher) was added to the media. Subsequently, the cells were trypsinized and fixed in ice-cold absolute ethanol. The cells were re-hydrated via 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 CuSO₄, 2.5 μM azide-fluor488 (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 with 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 grown in six-well plates, exposed to 100 nM talazoparib and 0.01% MMS for 1 h and subsequently incubated in a media containing the corresponding drugs (typically 100 nM talazoparib, 10 μM CB-5083, 1 μM CuET) for a chase period of 3 h. The cells were fractionated with the Subcellular Protein Fractionation kit for Cultured Cells (ThermoFisher #78840) according to the manufacturer’s recommendations.

Statistics and Reproducibility

No statistical method was used to predetermine 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-IPs have been performed 2 times showing reproducibility unless specified in the legends. For pre-extraction-based immunofluorescence microscopy, quantification and statistics were derived from n = 3 independent experiments. PLA and immunofluorescent 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 was pooled together and analysed by ordinary onew-way ANOVA (Graphpad Prism 9). Cellular growth inhibition assays were performed at least n=3 independent biological repeats and the statistical significance was derived with a 2way ANOVA (Graphpad Prism 9).

Extended Data

Extended Data Fig. 1. Proteomic profiling of PARP1 transgene-expressing CAL51 cells.

A. Western blot showing the expression of PARP1 transgenes, detected by an PARP1 antibody. Data shown represent 2 biological replicas. B. A Western blot analysis of the purified PARP1-associated proteins as described in the RIME experiment in Figure 1A. Data shown represent 3 biological replicas. C. Western blot analysis of the purified biotinylated proteins isolated in the PARP1WT-Apex2-eGFP proximity labelling experiment. Immunoblotting using Streptavadin-HRP is shown in the top panel, whilst anti-GFP immunoblotting is shown in the bottom panel. Endogenously biotinylated proteins are indicated as*. Data shown represent 3 biological replicas.

Extended Data Fig. 2. Bioinformatic analysis of trapped PARP1 proteomic data.

A. STRING network diagram of proteins identified by PARP1 proximity labelling under PARPi trapping conditions (as described in Methods). The graph shows connected nodes identified with a high stringency threshold of 0.7 (non-connected proteins are excluded from this visualisation). The colour coding corresponds to the following functional annotation: DNA damage repair-associated proteins (blue), base excision repair (green), ubiquitylation machinery (purple) and proteasome (magenta). Clusters, enriched for certain biological processes are indicated (e.g. “Ubiquitylation/proteosome”). B. Summary of gene set ontology analysis of the networks presented in (F). KEGG terms enriched at p-value <0.01 are shown.

Extended Data Fig. 3. Trapped PARP1 is SUMOylated and ubiquitinated.

A. Input controls for Figure 2C, showing the efficacy of MLN-7243 to inhibit ubiquitylation. Data shown represent 3 biological replicas. B. Reciprocal denaturing IP over PARP1-FLAG showed accumulation of trapped PARP1 ubiquitination in HEK293s. HEK293 cells were transfected with PARP1-FLAG-expressing construct for 24 hours then treated with 100 nM talazoparib/0.01% MMS and/or 10 μM CB-5083. Cells were lysed, chromatin was digested and then incubated with anti-FLAG beads. 4% of sample was harvested for input pre-incubation. Data shown represent 2 biological replicas. C. As in (C), but the immunoprecipitated proteins were analysed with an anti-K48 Ub chains recognising antibody. This experiment has been performed once. D. High exposure blot of PARP1 SUMOylation from Figure 2D, red arrows show SUMOylated PARP1 in MMS treated samples. Data shown represent 2 biological replicas.

Extended Data Fig. 4. Trapped PARP1 is modified in a PIAS4-dependent manner.

A. PARP1 SUMOylation by SUMO1 was detected as described in Figure 3A. Data shown represent 2 biological replicas. B. Western blotting for total ubiquitin input for Figure 3A. Data shown represent 2 biological replicas. C. A quantification of the SUMO2/3ylated and ubiquitylated PARP1 isoforms in the gels in Figure 3A. D. PARP1 SUMOylation by SUMO1 detected as described in Figure 3B. Data shown represent 2 biological replicas. E. PARP1 SUMOylation by SUMO1 was detected as described in Figure 3D. n=1 biological replicas. F. Western blotting for total ubiquitin input for Figure 3D. Data shown represent 2 biological replicas. G. A quantification of the SUMO2/3ylated and ubiquitylated PARP1 isoforms in the gels in Figure 3D. H. PARP1 SUMOylation by SUMO1 detected as described in Figure 3E. Data shown represent 2 biological replicas.

Extended Data Fig. 5. Trapped PARP1 is modified in a RNF4-dependent manner.

A. Overexpression of RNF4-WT increased PARP1 ubiquitination under trapping conditions. HEK293 cells were transfected with Ubiquitin-Strep-HA in combination with either FLAG-RNF4-WT or M136S/R177A mutant (E2 binding mutant, dominant negative) expressing constructs. After treatment with MMS + Talazoparib, the cells were fractionated and ubiquitylated proteins were purified from the chromatin-bound fraction via Streptactin beads. Data shown represent 2 biological replicas. B. RNF4 depletion prevents PARP1 ubiquitination under PARP trapping conditions. Denaturing IP of UB-HA-STREP-expressing HEK293 cells, similar to Figure 2B. Cells were depleted of RNF4 with either a 5’UTR sequence or Dharmarcon SMARTpool and were treated with 100 nM Talazoparib and 0.01% MMS. Pulldown was conducted with Streptactin beads. Data shown represent 2 biological replicas. C. In vitro SUMOylation assay as described in Figure 3G. The reactions were incubated in the presence of SUMO1, which was subsequently detected by anti-SUMO1 antibody. The asterisk indicates the free SUMO1. Data shown represent 2 biological replicas.

Extended Data Fig. 6. Trapped PARP1 interacts with p97.

A. Western blot analysis of Co-IP confirms PARP1-p97 interaction. CAL51 cells were transiently transfected with p97-WT-GFP-expressing construct. Subsequently, GFP was immunoprecipitated in native conditions and the presence of PARP1 investigated by Western blotting. Data shown represent 2 biological replicas. B. PARP1 interacts with p97 in a trapping-dependant manner. Cells were treated with 0.01% MMS in the presence of 100 nM talazoparib, 10 μM veliparib or 10 μM UKTT15. PARP1 associated proteins were immunoprecipitated and the presence of p97 was investigated by immunoblotting. Data shown represent 2 biological replicas. C. Western blots for denaturing IP experiment shown in Figure 4F. Data shown represent 3 biological replicas. D. p97-E578Q mutant colocalises with PARP1 under trapping conditions. CAL51 PARP1WT-eGFP and PARP1del.p.119K120S-eGFP cells were transfected with p97-WT-Strep-MYC or p97-E578Q-Strep-MYC constructs and then subsequently exposed to MMS + talazoparib to induce PARP1 trapping. Cell were then pre-extracted and fixed, and stained for trapped PARP1 and MYC (as described in ³³ ). The p97-E578Q-mutant colocalised with the trapped PARP1 signal in CAL51 PARP1WT-eGFP cells (yellow arrows) whereas PARP1del.p.119K120S-eGFP were unable to form trapped PARP1 foci. Scale bar = 5 μm. Data shown represent 2 biological replicas. E. Ubiquitin is required for the PARP1/p97 interaction in trapping conditions. Western blots of PARP1 co-immunoprecipitates from CAL51 PARP1^WT-eGFP-expressing cells. Trapping increases the PARP1/p97 interaction (lane 4), an effect reversed by MLN-7243 (5 μM). Data shown represent 3 biological replicas.

Extended Data Fig. 7. PARP1 trapping is modulated by PIAS4, RNF4 and p97.

A. Quantification of the chromatin bound PARP1 in Figure 5B; 2 biological replicas are displayed with individual points. B. Quantification of the chromatin bound PARP1 in Figure 5C; 2 biological replicas are displayed with individual points. C. As described in Figure 5A, trapping was induced in cells and subsequently chased as stated. Cells were then fractionated and the amount of chromatin-bound PARP1 was investigated by Western blotting. Data shown represent 3 biological replicas. D. CAL51 PARP1^WT-eGFP cells were transfected with FLAG-RNF4-WT or FLAG-RNF-M136S/R177A (E2 binding mutant, dominant negative) constructs. Treatment occurred 24 h after expression. Data shown represent 2 biological replicas. E. HeLa cells were transfected with either p97-Strep-MYC cDNA or a p97 E578Q mutant-Strep-MYC. Sixteen hours later, cells were exposed to MMS + talazoparib. Data shown represent 2 biological replicas. F. CAL51 cells expressing PARP1^WT-eGFP or PARP1^{del.p119K120S}-eGFP were transfected with p97^WT-Strep-MYC or p97^E578Q-Strep-MYC-expressing construct. After trapping and pre-extraction, cells were fixed and imaged. Scale bar = 5 μm. G. Graph of quantification of PARP1-eGFP foci of the experiment presented in (F). n=80-200 cells examined over 3 biologically independent experiments, mean ± SEM, p-values derived with a Kruskal-Wallis test. H. Western blots of trapped PARP1 from CAL51 PARP1-eGFP expressing cells transfected with a control siRNA (siLuc) or UFD1-targeting siRNA (siUFD1). 72 hours post transfection, cells were treated with MMS + talazoparib. Data shown represent 3 biological replicas.

Extended Data Fig. 8. Cell cycle and “PARP1 exchange” under p97 inhibition.

A. Cell cycle profiling for the experiment shown in Figure 5I. CAL51 cells were exposed to drugs as shown. One hour prior to fixation, 10 μM EdU was added to the media. EdU was stained by a click reaction with Alexa488-azide and DNA was stained by propidium iodide. B. A quantification of the G1, S and G2 populations from (A). C, D. CAL51 PARP1^WT-eGFP cells were subjected to UV-microirradiation, accumulation of PARP1^WT-eGFP at UV-laser induced DNA damage sites was monitored in the presence of DMSO (vehicle), talazoparib, CB-5083 or in combination. At the maximum time of PARP1^WT-eGFP recruitment (typically 1 min after microirradiation) the focus was bleached with a 488 nm laser and recovery of PARP1^WT-eGFP was monitored over time as described in ³⁹ . E. Image montages of the microirradiation site for (L). Scale bar = 2 μm. L. A quantification of the FRAP described in (K). The fluorescent signal was scaled according to the maximum PARP1^WT-eGFP immediately prior the photobleach to (equalling 1) and the signal immediately after photobleach (0), as in ³⁹ . The FRAP data was fitted with one site-specific binding model of non-linear regression and the extra sum of squares F test was used to calculate the t_1/2. The significance was determined with a two-sided t-test from two independent experiments, where 10 to 12 cells were quantified for each condition. * - p-value < 0.05.

Extended Data Fig. 9. PARP inhibitors synergise with p97 inhibition.

A. and B. Bliss synergy calculation, performed with the Combenefit software (CRUK Cambridge Institute), of the drug-response curves shown in Figure 6 B and C. C. and D. Drug response curves for the colony formation assays presented in Figure 6A. CAL51 WT or PARP1 ^−/− cells were treated with increasing concentrations of the PARP inhibitor Olaparib in the presence of either 100 nM CB-5083 (A) or 8 nM CuET (B). Surviving fractions were calculated based on the number of colonies after 14 days of exposure to the drugs. Shown are the mean ± SD of n=3 biological replicas. E. A quantification of the CB-5083 single agent effect on the surviving fraction of DLD1 and DLD1 BRCA2^−/− cells, respectively. Shown are the mean ± SD of n=3 biological replicas. F. Colony formation assays showing the synergistic effect between talazoparib and CB-5083 in DLD1 and DLD1 BRCA2^−/− cellular models. G. Area under the curve (AUC) analysis of the surviving fractions of DLD1 and DLD1 BRCA2^−/− cells in the presence of increasing concentrations CB-5083 combination as presented in Figure 6F. Shown are the mean ± SD of n=3 biological replicates; ordinary one-way ANOVA **** - p<0.0001. H. Brightfield images, showing the effect of talazoparib-CB-5083 combination on the GEMM WB1P organoid as described in Figure 6G. Scale bar = 200 μm. Data shown represent 2 biological replicas. I. Brightfield images showing the effect of talazoparib-CB-5083 combination on the KCL014BCPO organoid as described in Figure 6H. Scale bar = 200 μm. Data shown represent 3 biological replicas.

Data availability

Mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository (dataset identifier PXD024337). Source data are provided with this study. All other data supporting the findings of this study are available from the corresponding authors on reasonable request.