Single-Stranded DNA Gap Accumulation Is a Functional Biomarker for USP1 Inhibitor Sensitivity

Human tumors that are deficient in BRCA1 or BRCA2 are dependent on alternative DNA damage response pathways for maintaining genomic integrity (1). Homologous recombination (HR) deficiency renders these tumors susceptible to synthetic lethal targeting of those alternative pathways (2–4). Accordingly, the FDA has approved inhibitors of PARP for BRCA-deficient breast, ovarian, pancreatic, and prostate cancers (5–8). However, intrinsic and acquired PARP inhibitor resistance has become a major clinical problem. There is therefore an urgent need for strategies that can augment the benefit afforded by PARP inhibitors or that may overcome PARP inhibitor resistance in HR-deficient tumors.

In addition to their HR deficiency, BRCA1/2-deficient tumors accumulate single-stranded DNA (ssDNA) gaps behind advancing replication forks (9–12). These replicative gaps are genome-destabilizing structures, and their repair is required for the maintenance of genomic integrity. ssDNA gap accumulation is enhanced by treatment with PARP inhibitors or other DNA-damaging agents (9–11, 13, 14). For the maintenance of genomic stability, the ssDNA gaps left behind advancing replication forks must be properly repaired. In the absence of HR, gap filling requires postreplicative repair mechanisms, such as translesion DNA synthesis (TLS) repair (12, 13). Interestingly, PARP inhibitors further increase the number of ssDNA gaps (10, 14), contributing to their lethal effects in BRCA-deficient cells. Indeed, it is now recognized that exacerbation of ssDNA gaps is a key determinant of synthetic lethality between PARP inhibition and BRCA deficiency (10, 14). ssDNA gap induction provides a biomarker for response of patients to PARP inhibitors.

DNA polymerase theta (POLQ) is a microhomology-mediated end-joining (MMEJ) protein that is overexpressed in HR-deficient tumor cells. Genetic ablation of POLQ confers synthetic lethality in HR-deficient cancers (15, 16), and pharmacologic inhibition of POLQ, like PARP inhibition, kills HR-deficient tumor cells (17, 18) and promotes the accumulation of ssDNA gaps. POLQ inhibitors, such as novobiocin, block the ability of the polymerase to refill these gaps (19–21). Enhanced ssDNA gap accumulation may therefore be the mechanism of cytotoxicity of POLQ inhibitors in HR-deficient tumor cells.

The ubiquitin-specific protease 1, USP1, is a deubiquitinating (DUB) enzyme belonging to the cysteine protease family of DUB proteins (22, 23). The known substrates of USP1 include ubiquitinated-FANCD2 (Ub-FANCD2), ubiquitinated-FANCI (Ub-FANCI), and ubiquitinated–proliferating cell nuclear antigen (Ub-PCNA), all of which play a critical role in DNA repair (24–26). By cleaving ubiquitin from Ub-PCNA, USP1 controls TLS, a DNA damage tolerance pathway (24). During genotoxic stress or when replication forks encounter a lesion, PCNA is monoubiquitinated by an E3 ligase, RAD18 (27). Monoubiquitinated PCNA in turn recruits TLS polymerases, such as REV1, POLk, and POLi, via the monoubiquitin binding pockets of these proteins, which fill the postreplicative ssDNA gaps and maintain genome integrity (28–32). Efficient TLS polymerase filling of ssDNA gaps also requires USP1-dependent deubiquitination of Ub-PCNA. Loss of USP1 leads to increased levels of monoubiquitinated PCNA and dysfunctional TLS activity (33). Indeed, USP1 inhibition is synthetically lethal with BRCA1 deficiency (33). USP1 inhibitors are therefore now undergoing clinical development for HR-deficient tumors (NCT05240898, NCT06065059, and NCT05932862 listed in https://clinicaltrials.gov). USP1 inhibition drives the accumulation of Ub-PCNA, and this process may contribute, at least in part, to the cytotoxicity of these agents in HR-deficient tumor cells (33, 34). Therefore, like PARP inhibitors and POLQ inhibitors, USP1 inhibitors may cause accumulation of ssDNA gaps in HR-deficient tumor cells, suggesting that these drugs could be used in combinations.

In this study, we evaluated the molecular mechanism by which USP1 inhibitors kill HR-deficient tumor cells. We show that USP1 inhibition by a small molecule inhibitor caused an increase in ssDNA gaps resulting from enhanced ubiquitination of PCNA. Knockout of RAD18, known to monoubiquitinate PCNA, led to cellular resistance to USP1 inhibition and suppression of ssDNA gaps in BRCA1-deficient cancer cells. Furthermore, the sensitivity of patient-derived BRCA1-deficient ovarian tumor organoids to USP1 inhibition correlated with ssDNA gap accumulation, suggesting that this functional biomarker may be useful in ongoing clinical trials.

The following drugs were purchased from SelleckChem: olaparib (catalog #S1060), niraparib (catalog #S2741), and novobiocin (catalog #S2492). USP1 inhibitors I-138 (US patent # WO/2017/087837), TNG0240869 (TNG0869), and TNG0746132 (TNG6132) were obtained from Tango Therapeutics (Supplementary Fig. S5; ref. 34). Niraparib tosylate was purchased from MedChemExpress (catalog #HY10619B).

Culture conditions for human RPE1 cells, MDA-MB-436 cells, and UWB1.289 cells, as well as the generation of RPE1 P53 KO BRCA1 KO cells, were previously described (17, 33, 35).

Protein extraction from cell pellets, chromatin fractionation, and immunoblotting were performed as previously described (35, 36). Details of the Western blot protocol are provided in the Supplementary Materials and Methods section. Primary antibodies used were anti-USP1 (rabbit polyclonal; Bethyl, catalog #A301-699A) anti-RAD18 (mouse monoclonal; Cell Signaling catalog #9040), anti-PCNA (mouse monoclonal; Santa Cruz, catalog #sc56), antiubiquitinated PCNA (rabbit monoclonal; Cell Signaling catalog #13439), anti-vinculin (mouse monoclonal; Santa Cruz, catalog #SC25336), and anti-H3 (Abcam, catalog #ab1791).

Accelerated native isolation of proteins on nascent DNA (aniPOND) analysis was performed according to previously published protocols (37). Details of the protocol are provided in Supplementary Materials and Methods section.

Written informed consent was obtained from all patients involved in the study. The collection of patient data, ovarian cancer tissues, and pleural effusion fluids was performed at the Brigham and Women’s Hospital under Institutional Review Board-approved protocols in accordance with the Declaration of Helsinki and Belmont Report. Ascites fluid was obtained immediately at the time of paracentesis or thoracentesis and was centrifuged at 1,500 RPM for 5 minutes to create a cell pellet. The pellet was suspended with ammonium-chloride–potassium (ACK) Lysing Buffer (Thermo Fisher Scientific, catalog #A1049201) up to three times each incubated 2 minutes followed by washes with basal culture media Advanced DMEM/F12 with Glutamax (1%; Thermo Fisher Scientific, catalog #10565018) supplemented with 0.5% penicillin–streptomycin, and 20% FBS. The mixture was spun at 1,500 revolutions per minute (RPM) for 5 minutes for patient-derived organoid (PDO) generation. Solid tumors from ovarian cancer patients were transferred with transfer media and washed two times in sterile PBS (Thermo Fisher Scientific, catalog #10010023). Solid tissues then were transferred to a Petri dish and were cut into 3- to 5-mm³ pieces to be minced for enzymatic dissociation in collagenase (Sigma, catalog #C9407) containing media (final concentration of 2.5 mg/mL) supplemented with 5 µmol/L RHO/ROCK pathway inhibitor (Abmole Bioscience, catalog #Y27632) at 37°C shaking incubator for 30 minutes. The total homogenate was then strained in 100-μm cell strainers and mixed with basal media in a 1:2 ratio. The suspension was centrifuged at 1,500 RPM for 5 minutes. In case of visible red blood cells, the same process of ACK lysis was performed with an additional wash of basal culture media. Fifty percent of the cell pellets were mixed with growth factor–reduced Matrigel (Corning, catalog #CB40230C) at the ratio of 90% Matrigel/10% basal media with 20,000 cells in 20-µL droplets. Droplets were then solidified at room temperature, and the Petri dish was then flipped and incubated further at 37°C incubator for 4 minutes followed by the addition of 15-mL basal media to the dishes for organoid culturing. Generally, tissue digestion and cell plating were according to the protocols described previously (38).

The tumor cells from the corresponding patient-derived xenograft (PDX) model DF68 (35) were centrifuged at 1,500 RPM for 5 minutes to create a cell pellet. In case of visible red blood cells, ACK lysis protocol was performed with a wash of basal culture media and centrifugation at 1,500 RPM. The cell pellets were then mixed with growth factor–reduced Matrigel and organoids were grown using the procedures described for the PDOs.

Animal studies for PDX experiments were performed at Dana-Farber Cancer Institute with approval from the institute’s Animal Care and Use Committee in an Association for Assessment and Accreditation of Laboratory Animal Care–accredited facility. Eight-week-old female NSG (NOD.Cg-Prkdc^scid IL2rg^tm1Wjl/Sz) mice, purchased from Jackson Laboratory, were used for the animal studies. All mice were housed, treated, and handled in accordance with the guidelines set by the Animal Care and Use Committee of the Dana-Farber Cancer Institute.

A PARP inhibitor–resistant luciferized DF68 PDX model of high-grade serous ovarian cancer was used as described previously (39). Approximately 5 × 10⁶ luciferized tumor cells were injected intraperitoneally into NSG mice as described previously (35). Tumor burden was measured by bioluminescence imaging (BLI) using an IVIS Lumina III (PerkinElmer). Tumors were allowed to establish for 4 to 5 weeks and mice with increasing BLI were randomized into various treatment groups. The disseminated disease progression in the peritoneal cavity was measured serially once per week by BLI. Niraparib tosylate was formulated with 0.5% methylcellulose (4,000 cP) and 0.5% Tween 80 in water and administered orally at 45 mg/kg daily for 4 weeks. TNG6132 was formulated in 5% DMA + 30% PEG 400 + 20% Solutol HS15 + 45% water and administered orally at 200 mg/kg daily for 4 weeks.

For the DNA fiber assays, we used FiberPrep DNA extraction kit (catalog #IFU-M&S13-c), the FiberComb Molecular Combing System (catalog #MCS001) to comb DNA fibers, and FiberVision (Genomic Vision) scanner to scan the slides, following manufacture’s protocols. Briefly, cells were plated on six-well plates on the day before the experiment at 50% confluence. The cells were first incubated with 5-chloro-2'-deoxyuridine (CldU, 100 µmol/L) for 30 minutes, followed by 5-iodo-2′-deoxyuridine (IdU; 100 µmol/L) for 2 hours, both at 37°C. DMSO or the drug of interest was added concomitantly to the IdU labeling. Cells were washed with PBS 3× after each analog incubation. Cells were then permeabilized with CSK buffer [0.5% TritonX-100, 10 mmol/L 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 300 mmol/L sucrose, 100-mmol/L NaCl, and 3 mmol/L MgCl₂ in sterile water] for 10 minutes at room temperature followed by one wash with S1 nuclease buffer (NaCl 50 mmol/L, sodium acetate 300 mmol/L at pH 4.6, zinc acetate 10 mmol/L, and glycerol 5%) and incubation with S1 nuclease buffer ± S1 nuclease (20 U/mL) at 37°C for 30 minutes. Nuclei were then washed with PBS 1× and 1 mL of PBS + 0.1% BSA was added to each well. Nuclei were then harvested using a cell lifter, pelleted, and resuspended in 45 μL PBS. This solution was mixed with melted agarose and incubated at 4°C for 45 minutes to form agarose plugs that were incubated in proteinase K solution (buffer 3 + proteinase K) overnight at 50°C. The plugs were then washed with washing buffer (buffer 4) 4× and incubated overnight in the combing buffer (buffer 6) at 4°C. The samples were then combed into coverslips (catalog #COV002-RUO) and incubated with primary antibody rat anti-BrdU (Abcam #ab6326), mouse anti-BrdU (BD Biosciences, catalog #347580), and diluted in blockaid (Invitrogen, catalog #B10710) overnight at 37°C. The coverslips were incubated with secondary goat anti-rat Cy5 antibody (Abcam, catalog #ab6565) and goat anti-mouse Cy3 antibody (Abcam, catalog #97035) for 45 minutes at 37°C, followed by mouse anti-ssDNA antibody (catalog #DSHB autoanti-ssdna) for 1 hour and 15 minutes at 37°C, and goat antimouse BV480 (Jackson, catalog #115-685-166) for 45 minutes at 37°C). The coverslips were then dry mounted and scanned using a FiberVision scanner (Genomic Vision). The following final drug concentrations were used: olaparib 10 µmol/L, I-138 0.1 µmol/L, and novobiocin 200 µmol/L.

For the S1 nuclease assay in organoids, viable organoids in culture were processed for the partial removal of Matrigel. The partial removal was performed via 15-minute incubation of collected organoids on ice. Once the Matrigel was visibly melted, samples were centrifuged at 1,500 RPM for 5 minutes at 4°C. In case of lack of partial removal, samples were exposed to ice incubation for 10 more minutes, and the centrifuge step was repeated with additional removal of Matrigel by discarding the supernatant. Each sample was then incubated with CldU (200 µmol/L) for 1 hour followed by IdU (200 µmol/L) for 4 hours concomitantly to the drug treatment. Samples were washed 3× with PBS after the incubation with each analog. Before the last wash, samples were incubated for 10 minutes on ice to remove any residual Matrigel. The samples were then centrifuged at 1,500 RPM and resuspended in 45 μL of PBS. Plug preparation and overnight incubation were done according to the protocol for monolayer samples. On the second day, samples were washed with washing buffer (buffer 4) twice, followed by one wash with S1 buffer and one wash with S1 buffer ± S1 nuclease. The remaining steps are identical to the protocol for monolayer samples as described above.

For clonogenic colony formation assays, 500 to 8,000 cells per well were seeded in six-well plates and treated with graded concentrations of the desired drugs for 10 to 14 days. Colonies were washed with PBS, fixed with methanol for 10 minutes, and stained with crystal violet. Plates were imaged, and colonies were quantified with Oxford Optronix GelCount. For CellTiter-Glo assays, cells were plated in a 96-well plate (500–4,000 cells per well) on the first day and treated with graded concentrations of the desired drugs for 7 days. Cell viability was determined by CellTiter-Glo (Promega, catalog #G7571), and viability was calculated relative to DMSO-treated samples. GraphPad Prism was used to generate dose–response curves and calculate IC₅₀. Synergy between the drugs was calculated using Combenefit software (40).

Isolated patient cells and PDX cells were mixed with 10% Matrigel (10,000 cells per well) and plated to 96-well U-bottom ultralow attachment plates (Corning 7007), and then centrifuged at 2,500 RPM for 10 minutes at 4°C to produce identical organoids to be tested for drug sensitivity. Plates were then immediately transferred to a 37°C incubator and incubated for 48 hours prior to drug applications. CellTiter-Glo 3D Cell Viability Assay was performed after 7 days of treatment with selected drugs; 100 μL of CellTiter-Glo 3D Reagent (Promega, catalog #G9682) was added to 100 μL of medium containing organoids. The content was mixed for 10 minutes to induce organoid lysis and the plate was incubated for an additional 15 minutes to stabilize the luminescent signal, and the signal was recorded accordingly.

The targeted CRISPR screen in UWB1.289 cells was performed using a targeted CRISPR-Cas9 library as previously described (34). The library consists of single guide RNAs (sgRNA) targeting 1,145 DNA damage response genes. The library contains six sgRNAs per gene, and the sequences were sourced from Vienna Bio Center (https://www.vbc-score.org/, PMID: 32514112). Also included in the library were 600 nontargeting and 72 introcutting sgRNAs .

For data analysis, the raw NGS data were preprocessed to generate the raw sgRNA read counts as previously described (34). The raw sgRNA read counts were normalized with the nontargeting sgRNAs as log₂-transformed counts per million (log₂-CPM), using the calcNormFactors and cpm functions of the edgeR (41) package in R. The log₂-fold change (LFC) for each sgRNA between the USP1 inhibitor–treated experimental sample and the DMSO-treated control sample was computed by taking the difference between their log₂-CPM values. The hypergeometric distribution method (https://github.com/mhegde/volcano_plots) was performed on LFCs of the sgRNAs to derive the gene-level LFC estimates and P values. The volcano plots were generated using ggplot2 and ggrepel R packages.

Quantitative data were analyzed and graphed using GraphPad Prism 9 software. All data are represented as mean ± SEM unless indicated otherwise. Significance was tested using the Mann–Whitney test unless indicated otherwise. Assessment of synergy was calculated based on the Bliss model of synergy using Combenefit software (40, 42).

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Alan D. D’Andrea ([email protected]).

Additional methods are provided in the Supplementary Materials and Methods section.

USP1 is upregulated in BRCA1-deficient cells, and knockout or inhibition is synthetic lethal with BRCA1 deficiency (33). To dissect the mechanism of synthetic lethality between USP1 inhibition and BRCA1 deficiency, we employed a specific inhibitor of USP1, namely I-138, which inhibits USP1 at low nanomolar concentrations (34). Consistent with previous studies (33, 34), I-138 treatment caused an increase in Ub-PCNA levels in MDA-MB-436 BRCA1-deficient breast cancer cells (Supplementary Fig. S1A). The isolation of proteins on nascent DNA, via the accelerated native isolation of proteins on nascent DNA (aniPOND) method, showed that Ub-PCNA was enriched at the replication fork after I-138–mediated USP1 inhibition (Supplementary Fig. S1B). Aberrant localization of Ub-PCNA at the replication fork can interfere with replication and can increase ssDNA gaps (13). We therefore employed DNA fiber assays in BRCA1-deficient, PARP inhibitor-sensitive UWB1.289 ovarian cancer cells and MDA-MB-436 breast cancer cells to assess ssDNA gaps after USP1 inhibition. In these fiber assays, S1 nuclease treatment was used to digest ssDNA regions during replication (43). Consistent with previous reports (10, 11), PARP inhibition resulted in a shorter length of IdU-labeled DNA tracks after S1 treatment, demonstrating an accumulation of gaps in these cells (Fig. 1A and B; Supplementary Fig. S1C). Interestingly, USP1 inhibition by I-138, like PARP inhibition, also caused an accumulation of gaps (Fig. 1A and B; Supplementary Fig. S1C). Moreover, RPE cells with CRISPR-mediated knockout of USP1 (USP1-KO) also exhibited an increase in ssDNA gaps after BRCA1 knockdown (Supplementary Fig. S1D–F). Recent studies suggest that accumulation of ssDNA gaps due to unligated Okazaki fragments during replication causes PARP1 activation leading to an increase in S phase-specific PARylation in BRCA1-deficient cells (10, 44). Indeed, we observed that increased monoubiquitination of PCNA and ssDNA gap accumulation after USP1 inhibitor treatment also caused an increase in S phase-specific PARylation (Supplementary Fig. S1G). Collectively, these results indicate that pharmacologic inhibition of USP1 by I-138 increases Ub-PCNA at the replication fork and leads to accumulation of ssDNA gaps in BRCA1-deficient cells.

A previous study indicated that a USP1 inhibitor can cause an increase in the level of Ub-PCNA and a compensatory reduction in the level of free, unmodified PCNA. This low level of PCNA is a possible mechanism for the killing of BRCA1 deficient cells by the drug (34). To provide further insight into other killing mechanisms of USP1 inhibitors, we performed a targeted CRISPR screen using the TNG0869 USP1 inhibitor in BRCA1-deficient UWB1.289 cells. sgRNAs against RAD18 or against genes in the Fanconi anemia pathway, conferred resistance to USP1 inhibition (Fig. 1C; Supplementary Table S1). Interestingly, sgRNAs against genes in the MMEJ pathway, such as XRCC1 and APEX2, conferred sensitivity to USP1 inhibition (Fig. 1C). RAD18 is an E3 ligase that monoubiquitinates PCNA, and USP1 is the DUB enzyme that removes the ubiquitin (45). Indeed, RAD18 knockdown reduces synthetic lethality between BRCA1 and USP1 and results in USP1 inhibitor resistance (33, 34). We therefore further investigated the possible contribution of RAD18 to the accumulation of ssDNA gaps. Initially, we performed the aniPOND method, combined with mass spectrometry, to provide a quantitative measure of proteins bound to replication forks after USP1 inhibition in BRCA1-deficient breast cancer cells. Indeed, RAD18 was enriched at the replication fork after USP1 inhibition (Fig. S1H; Supplementary Table S2). Our results indicated an increased localization of RAD18, Ub-PCNA, and USP1 to the replication fork after USP1 inhibition (Fig. 1D). Consistently with this finding, USP1 inhibitor treatment of BRCA1-deficient cells caused trapping of RAD18 and Ub-PCNA in the chromatin fraction (Fig. 1E). As expected, USP1 autocleavage was inhibited after USP1 inhibition, and uncleaved USP1 was also enriched in the chromatin fraction (Fig. 1E). Altogether, uncleaved USP1 is trapped at the replication fork and may contribute to the mechanism of action of the USP1 inhibitor (46). To further confirm the enrichment of RAD18 at replication forks after USP1 inhibition, we next performed a proximity ligation–based assay and measured the association of proteins to the nascent DNA of replication forks. Again, the USP1 inhibitor enhanced the association of RAD18 with the replication fork (Fig. 1F).

We next reasoned that the high level of Ub-PCNA generated by RAD18 at the replication fork was the cause of ssDNA gap accumulation. As expected, RAD18 knockdown in BRCA1-deficient cells showed decreased levels of Ub-PCNA after USP1 inhibition (Fig. 1G). Consistent with the CRISPR screen results, RAD18 knockdown rescued the survival of BRCA1-deficient UWB1.289 cells and MDA-MB-436 cells in the presence of I-138 (Fig. 1H; Supplementary Fig. S1I and S1J). Strikingly, RAD18 knockdown also rescued ssDNA gap accumulation after USP1 inhibition in UWB1.289 cells (Fig. 1I). Altogether, the persistent RAD18-mediated ubiquitination of PCNA causes the accumulation of ssDNA gaps in USP1 inhibitor–treated cells and the synthetic lethality between BRCA1 deficiency and USP1 inhibition.

Therapeutic drug responses in fresh human tumor organoid models correlate, at least in part, with the corresponding patient responses in the clinic (38, 47, 48). To determine whether USP1 inhibitor-induced ssDNA gap accumulation also occurs in primary tumor cells, we generated ovarian cancer organoids from tumors of five patients with high-grade serous ovarian cancer (HGSOC). Two of the organoid models were generated from patients harboring germline BRCA1 pathogenic mutations (DF3888 and DF4440), and the corresponding tumor cells were BRCA1 deficient. The other three models were wildtype for BRCA1 (Supplementary Table S3). The clinical features of these five patients and the sampling procedures are also summarized in Supplementary Table S3.

The morphologic and cytologic features of the five organoid models matched the primary ovarian tumors from which they were derived. The organoids displayed extensive nuclear pleomorphism, prominent nucleoli, and dense chromatin and demonstrated positive staining for the Mullerian marker, PAX8, WT1, and mutant p53 protein recapitulating the tumors from which they were derived (Fig. 2A; Supplementary Fig. S2A and S2B; Supplementary Table S3). As expected, the two BRCA1-mutated organoid models, but not the BRCA1-WT organoid models, were sensitive to olaparib (Supplementary Fig. S2C). Absence of RAD51 foci, a functional marker of HR, confirmed the HR deficiency of the BRCA1-mutated organoids (Fig. 2B; Supplementary Fig. S2D and S2E). There were no differences in terms of USP1 protein expression between the BRCA1-mutant organoids and BRCA1-WT organoids (Supplementary Fig. S2F), Interestingly, the two BRCA1-mutated organoids had high levels of USP1 mRNA, further confirming that these two organoids with HR deficiency are unique and distinct from the BRCA1-WT organoids (Fig. 2C). We next tested a USP1 inhibitor, namely TNG6132, in organoids. TNG6132 and I-138 USP1 inhibitors had comparable potency in killing BRCA1-deficient cells (Supplementary Fig. S2G). TNG6132 is a highly selective inhibitor for DUB activity of USP1 and it increased Ub-PCNA in BRCA1-deficient cells (Supplementary Fig. S2H and S2I). Interestingly, the BRCA1-mutant organoids, but not the BRCA1-WT organoids, were sensitive to a USP1 inhibitor TNG6132 (Fig. 2D). The BRCA1-mutant organoids and the BRCA1-WT organoids grew equally without the drug, suggesting that difference in drug sensitivity of the organoids does not result from differences in their cell proliferation (Supplementary Fig. S2J). USP1 inhibition or PARP inhibition promoted ssDNA gaps in the BRCA1-mutated organoids but not in the BRCA1-WT organoids (Fig. 2E). Altogether, ssDNA gap accumulation induced by a USP1 inhibitor, as well as the high level of USP1 mRNA in the BRCA1 mutant PDOs, correlated with the drug sensitivity of BRCA1-mutated organoid models (Fig. 2F), thereby providing predictive biomarkers for patients whose tumors are most likely to respond to these agents.

Recent studies have shown that PARP inhibitors or POLQ inhibitors can induce ssDNA gaps in BRCA1-deficient tumor cells (10, 19, 20). Indeed, POLQ, a key component of the MMEJ pathway, functions by refilling the postreplicative gaps in BRCA-deficient cells (19). Interestingly, knockout of several genes in the MMEJ pathway can enhance the sensitivity of tumor cells to USP1 inhibition, based on CRISPR screens (Fig. 1C). We therefore reasoned that inhibitors of USP1 would synergize with inhibitors of proteins in the MMEJ pathway, including the proteins POLQ and PARP. As shown in the model in Fig. 3A, cells with BRCA1 mutations have ssDNA gaps, and the ssDNA gaps accumulate after PARPi treatment. We hypothesize that these PARP inhibitor–treated cells respond by upregulating POLQ and/or USP1 expression, thereby promoting the refill of the ssDNA gaps. Accordingly, inhibition of POLQ or USP1 would resensitize the cells to the PARP inhibitor. Consistent with this hypothesis, the USP1 inhibitor, I-138, enhanced the sensitivity of BRCA1-deficient cell lines to PARP inhibitor, resulting in a 40-fold reduction in the IC₅₀ for olaparib (Fig. 3B; Supplementary Fig. S3A). Combined USP1 and PARP inhibition also led to the accumulation of ssDNA gaps in BRCA1-deficient cells (Fig. 3C). Moreover, I-138 also enhanced the sensitivity of the BRCA1-deficient cells to a POLQ inhibitor, and this drug combination also led to the accumulation of ssDNA gaps (Fig. 3D and E; Supplementary Fig. S3B). Combination of I-138 with a PARP inhibitor or a POLQ inhibitor did not show synergy in BRCA1-proficient cells (Supplementary Fig. S3C and S3D).

We next determined whether the combination of a USP1 inhibitor and a PARP inhibitor or a POLQ inhibitor could enhance the drug sensitivity of the PDO models. Combined inhibition of USP1 and PARP showed no synergy in a BRCA1-mutated PDO model (Supplementary Fig. S3E). However, combined inhibition of PARP and USP1 resulted in significant synergistic killing of a BRCA1-WT PDO model (Supplementary Fig. S3F). The USP1 inhibitor alone or niraparib alone did not induce ssDNA gaps in this BRCA1-WT PDO model. However, combined treatment with both drugs resulted in accumulation of ssDNA gaps (Supplementary Fig. S3G). Similar results were obtained when USP1 and POLQ inhibition were combined. One of the BRCA1-mutated PDOs was sensitive to a POLQ inhibitor novobiocin (Supplementary Fig. S3H). The USP1 inhibitor further sensitized this BRCA1-mutant PDO model (Supplementary Fig. S3I and S3J). The combination of TNG6132 with novobiocin showed weak synergy in BRCA1-WT PDO models (Supplementary Fig. S3K and S3L).

Altogether, the pairwise combination of inhibitors of PARP, POLQ, or USP1 can enhance the drug sensitivity of BRCA1-deficient tumors and cause the accumulation of ssDNA gaps. Combined inhibition of these targets also generates ssDNA gaps in BRCA1 WT cells, suggesting that this strategy may overcome intrinsic PARP inhibitor resistance.

To determine whether the USP1 inhibitor can kill PARP inhibitor–resistant tumor cells in vivo, we conducted a study using a BRCA1-mutant, PARP inhibitor–resistant PDX model of HGSOC, DF68 (35). Tumor-bearing mice were treated with the PARP inhibitor niraparib, the USP1 inhibitor TNG6132 (suitable for in vivo studies), or the drug combination, and tumor growth was monitored by BLI (Fig. 4A and B). As expected, niraparib failed to kill this tumor in vivo. TNG6132-mediated USP1 inhibition alone also did not affect the growth of the tumor cells; however, the combination of niraparib and TNG6132 demonstrated substantial tumor growth inhibition (Fig. 4A and B). The drug combination was well tolerated, with minimal loss of body weight during the time of drug exposure (Supplementary Fig. S4A). Pharmacodynamic analyses of TNG6132 activity in tumors revealed an increase in Ub-PCNA as well as an increase in full-length (uncleaved) USP1, demonstrating target engagement by the USP1 inhibitor (Fig. 4C). USP1 inhibitor treatment also caused a decreased level of total PCNA in tumors, consistent with the previous study showing that depletion of free, unmodified PCNA is a mechanism of USP1 inhibitor–mediated cell death (34). Similar to the in vivo results, the combination of niraparib and TNG6132 resulted in synergistic cytotoxicity in PDXOs (Fig. 4D; Supplementary Fig. S4B and S4C). The monotherapies did not induce ssDNA gaps (Fig. 4E); however, combined treatment with both drugs resulted in accumulation of ssDNA gaps (Fig. 4E). Collectively, these results demonstrate that a USP1 inhibitor can overcome PARP inhibitor resistance in BRCA1-mutated tumors of ovarian cancer in vivo, with reversal of ssDNA gap suppression.

On the basis of the synthetic lethality (33), USP1 inhibitors are now under active clinical trials for BRCA1-deficient tumors. Despite these ongoing trials, little is known about the mechanism through which these agents specifically kill BRCA1-deficient tumors. Biomarkers predicting drug response are also unavailable. In this study, we show that USP1 can suppress ssDNA gaps in BRCA1-deficient tumor cells by deubiquitinating its key substrate, PCNA. USP1 inhibitors increase Ub-PCNA at the replication fork and increase ssDNA gaps in BRCA1 deficient cells. The accumulation of these ssDNA gaps correlates with drug sensitivity. Furthermore, RAD18, an E3 ligase for PCNA ubiquitination, has a key role in killing BRCA1-deficient cancer cells during USP1 inhibitor exposure. Previous studies have shown that knockdown of RAD18 rescues BRCA1-deficient cells from USP1 inhibitor–mediated toxicity (33, 34). We now confirm and extend these findings by showing that RAD18 is trapped at replication forks after USP1 inhibition, providing a likely impediment to TLS polymerase-mediated refill of ssDNA gaps. Knockdown of RAD18 rescues BRCA1 deficient cells from USP1 inhibitor–mediated cytotoxicity and ssDNA gap accumulation. Importantly, our results identify ssDNA gap accumulation as a functional biomarker and a key determinant for USP1 inhibitor cytotoxicity in BRCA1-deficient cancer cells.

SsDNA gap accumulation is emerging as a new mechanism of cellular toxicity caused by many DNA repair inhibitors and chemotherapy agents (9–11, 13, 14, 49). ssDNA gap accumulation correlates with chemotherapy response in BRCA-deficient tumors (9, 14). PARP inhibitors increase ssDNA gap accumulation in HR-deficient cells (10, 14), correlating with cytotoxicity; however, whether or not the presence of ssDNA gaps directly leads to lethality remains uncertain. Inhibition of the polymerase, POLQ, also kills BRCA1 or BRCA2-deficient tumor cells, including those that are PARP inhibitor–resistant, and also causes postreplicative ssDNA gap accumulation (19, 20).

Altogether, the three known proteins with synthetic lethal interactions with BRCA1 deficiency—namely, PARP1, POLQ, and USP1—all seem to function via a common mechanism. Each of these three proteins can suppress ssDNA gap accumulation in BRCA1-deficient tumor cells, and inhibitors of these proteins can increase ssDNA gaps and promote killing of these cells. Interestingly, the known inhibitors of these three targets are oral drugs and, in principle, could be used in combination for the treatment of HR-deficient tumors. To date, some combinations have been evaluated in preclinical models. For instance, the combination of a PARP inhibitor and a POLQ inhibitor demonstrated enhanced killing of BRCA1-deficient tumors, including PARP inhibitor-sensitive and PARP inhibitor–resistant (17). A previous study has shown that this combination promotes ssDNA gaps (19). Similarly, we now show that a USP1 inhibitor can also synergize with a PARP or POLQ inhibition and promote ssDNA gap generation and cell killing. These findings provide new opportunities for combination therapies that may be superior to monotherapies. Additional biomarkers may guide the use of one or more of these drugs in specific settings. For instance, a tumor with a high level of a USP1 mRNA may have a better response to a USP1 inhibitor, as opposed to a PARP inhibitor or a POLQ inhibitor.

Our study has other clinical implications. The fiber assay for the detection of ssDNA gaps in organoids allows the rapid identification of patients who may be included in clinical trials and are most likely to respond to these new oral agents, either as monotherapy or in combination. Future studies with a larger number of patient-derived organoids and samples from patients treated on USP1 inhibitor clinical trials will be required to validate the predictive power of this biomarker. This functional biomarker approach may also have limitations, including some technical difficulties in performing the fiber assays, the reproducibility of the assay, and interobserver variability of the data analysis. The original tumor microenvironment and the tumor heterogeneity may not be fully recapitulated in the organoids, and specific clones may emerge during culturing, thereby affecting drug sensitivity and ssDNA gap formation.

Finally, a recent report indicated that, in some cellular settings, a USP1 inhibitor can paradoxically cause an increase in ssDNA gap refill and a suppression of ssDNA gap formation (50). Specifically, in the setting of HR proficiency (i.e., BRCA1-proficient cells), cisplatin exposure, and PRIMPOL overexpression, a USP1 inhibitor can increase PCNA-Ub levels and enhance the level of TLS-refill of ssDNA gaps. Moreover, as new USP1 inhibitors become available, it will be important to determine their relative impact on ssDNA gap levels based on the specific tumor cell context.

J. Liu reports personal fees from AstraZeneca, Bristol Myers Squibb, Clovis Oncology, Daiichi Sankyo, Eisai, Genentech/Roche, GlaxoSmithKline, Regeneron Therapeutics, Revolution Medicine, Zentalis Pharmaceuticals, and Deciphera Pharmaceuticals outside the submitted work and institutional funding for clinical trials from 2X Oncology, Aravive, Arch Oncology, AstraZeneca, Bristol Myers Squibb, Clovis Oncology, GlaxoSmithKline, Impact Therapeutics, Regeneron, Seagen, Vigeo Therapeutics, and Zentalis Pharmaceuticals. G.I. Shapiro reports grants from Tango Therapeutics during the conduct of the study; grants and personal fees from Merck KGaA/EMD Serono; and grants from Bristol Myers Squibb, Merck & Co., Pfizer, Eli Lilly, Bicycle Therapeutics, Kymera Therapeutics, ImmunoMet, Concarlo Therapeutics, Janssen, and Xinthera outside the submitted work, as well as having a patent for Dosage regimen for sapacitabine and seliciclib, Cyclacel Therapeutics, and Compositions and methods for predicting response and resistance to CDK4/6 inhibition issued. L. Cornell has a patent for Dosage regimen for sapacitabine and seliciclib, Cyclacel Therapeutics, and Compositions and methods for predicting response and resistance to CDK4/6 inhibition issued. A.D. D’Andrea reports grants from Bristol Myers Squibb, EMD Serono, Moderna, and Tango Therapeutics; personal fees and other support from Impact Therapeutics, PrimeFour Therapeutics, and Covant Therapeutics; and personal fees from Servier Bio-Innovation LLC outside the submitted work. No disclosures were reported by the other authors.

A.A. da Costa: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. O. Somuncu: Validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. R. Ravindranathan: Validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. S. Mukkavalli: Validation, investigation, visualization, methodology, writing–review and editing. D.B. Martignetti: Validation, investigation, visualization, methodology, writing–review and editing. H. Nguyen: Validation, investigation, visualization, methodology, writing–review and editing. Y. Jiao: Validation, investigation, visualization, methodology, writing–review and editing. B.P. Lamarre: Validation, investigation, visualization, methodology, writing–review and editing. G. Sadatrezaei: Validation, investigation, visualization, methodology, writing–review and editing. L. Moreau: Validation, investigation, visualization, methodology, writing–review and editing. J. Liu: Methodology, writing–review and editing. D.R. Iyer: Validation, investigation, visualization, methodology, writing–review and editing. J.-B. Lazaro: Validation, investigation, visualization, methodology, writing–review and editing. G.I. Shapiro: Conceptualization, resources, supervision, methodology, writing–original draft, writing–review and editing. K. Parmar: Conceptualization, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. A.D. D’Andrea: Conceptualization, resources, supervision, methodology, writing–original draft, writing–review and editing.

We thank members of the A.D. D’Andrea laboratory for their helpful suggestions and comments. We thank Dr. David L. Kolin, MD, PhD, Department of Pathology at Brigham and Women’s Hospital, for providing the immunohistochemistry pictures of the patients’ tumors from which the organoids were generated. This work was supported by Sponsored Research Agreement funding from TANGO Therapeutics (A.D. D’Andrea, G.I. Shapiro), U.S. National Institutes of Health grants (R01HL052725 and P01HL048546), US Department of Defense (BM110181), Breast Cancer Research Foundation, Fanconi Anemia Research Fund, Ludwig Center at Harvard, and Smith Family Foundation (A.D. D’Andrea). This work was also supported by U.S. National Institutes of Health grants P50 CA168504 (A.D. D’Andrea, G.I. Shapiro) and P50 CA240243 (A.D. D’Andrea, G.I. Shapiro).