PARP Inhibitors: Extending Benefit Beyond BRCA-Mutant Cancers

The establishment of the relationship between the tumor-suppressive genes BRCA1 and BRCA2 (BRCA1/2) and hereditary breast and ovarian cancer syndrome (HBOC) revolutionized clinical cancer genetics, and led to increased research focused on germline variant testing, risk stratification, early detection, and cancer prevention for BRCA1/2 mutation carriers (1). BRCA1 and BRCA2 are key proteins in the DNA Damage Response (DDR), and over the past two decades, key advances in next-generation sequencing, as well as epigenetic and expression-level profiling technologies have rapidly expanded our understanding of the role of DDR pathway deficiencies and associated genomic instability in cancer initiation and evolution. Critically, they have also informed the field on how they can be rationally targeted for cancer therapy (2). Preclinical and clinical studies have revealed key gene networks that may directly or indirectly influence DDR, and which now include molecular aberrations beyond BRCA1/2 mutations, as well as tumor types outside of breast and ovarian cancers (3–6). Previously, the identification of a germline BRCA1/2 mutation generally only impacted cancer screening and prevention practices for that patient and their relatives; however, the discovery that small-molecule inhibitors of PARP selectively killed BRCA1/2-mutant cancer cells has now led to new therapeutic approaches in the clinic for patients with DDR gene aberrations (7–10). In this article, we review the PARP inhibitors currently in the clinic with respect to their known mechanism(s) of action, current single-agent applications, key similarities and differences, predictive biomarkers of response and resistance, and rational combinatorial strategies with other anticancer agents.

DNA damage activates a complex range of processes, including DDR signaling, DNA repair, cell-cycle regulation, and potentially also an immunologic response, all of which have been extensively reviewed previously (6, 7, 11–14). Briefly, DNA single-strand breaks resulting from processes that remove misincorporated rNTP incorporation, dNTP mismatches, or bases following oxidative damage are the most common; however, DNA double-strand breaks (DSB) are the most cytotoxic. The high-fidelity homologous recombination repair (HRR) pathway faithfully repairs DNA DSBs by using the replicated sister chromatid DNA when available, and the activities of key molecules including BRCA1/2 and RAD51 among others (15). If an undamaged template DNA is unavailable, then the more rapid but error-prone nonhomologous end joining (NHEJ) repair pathway is the primary method of DNA DSB repair in the cell, utilizing essentially a direct ligation approach (16). The PARP family of proteins plays a key role in a variety of cellular processes that includes DNA repair, chromatin modulation, and aspects of the replication stress response (7, 17). With regard to DNA repair, PARP1 and PARP2 play critical roles in DNA strand break repair through multiple DDR pathways, with HRR-deficient cells showing a greater reliance on PARP activity to maintain cell survival (7).

The primary activity of PARP1, and the closely related PARP2 protein, is the poly-ADP ribosylation (PARylation) of key components of chromatin and DDR as well as auto-PARylation (18). PARP1 activity both opens up chromatin and facilitates recruitment of downstream DNA repair factors to damaged sites (18). After completing this recruitment role, PARP auto-PARylation triggers the release of bound PARP from DNA to allow access for other DNA repair proteins to complete repair. Thus, the binding of PARP to damaged sites, its catalytic activity, and its eventual release from DNA are all necessary steps for a cancer cell to respond to DNA breaks introduced by certain chemotherapies, radiation, and various forms of endogenous damage (7). It has been shown that preclinical models with biallelic loss of PARP show viability under normal conditions, but are exquisitely sensitive to alkylating chemotherapy and DNA-damaging radiation.

Synthetic lethality describes a bimodal dependency whereby the loss-of-function of just one component in a cell or organism does not have a significant impact on viability, yet the combined loss of both components results in cell death due to the interdependent and/or compensatory nature of the two pathways (7). The finding that single-agent PARP inhibition selectively killed BRCA1 (8) or BRCA2 (9)-deficient cells was a pivotal discovery that ushered in new synthetic lethal therapeutic approaches in clinical oncology. This was soon followed by the demonstration that non-BRCA deficiencies in the HRR pathway also resulted in PARP inhibitor single-agent sensitivity (19). In this cardinal example of DDR-based synthetic lethality, it is the PARP inhibitor that can trap PARP protein onto the DNA at the single-strand break (SSB) as the initiating event. If the inhibitor stays bound within the PARP-active site and the PARP protein is trapped on the DNA long enough to be encountered by the replication machinery, this can result in a stalling of the replication fork, its collapse, and the generation of a DNA DSB. In cancer cells deficient for HRR, classical NHEJ is employed, generating error-prone repair which, after multiple rounds of DNA replication, can lead to unsustainable levels of genomic instability and cancer cell death. Conversely, normal cells with functional HRR are able to deal with the DSBs accurately and effectively. This example of PARP inhibitor synthetic lethality in HRR-deficient cancer cells heralded the promise of cancer cell–specific killing and importantly, an opportunity for a wide therapeutic window (2).

Currently, six small-molecule PARP inhibitors are available in the clinic—olaparib (the first PARP inhibitor to test the synthetic lethality concept in the clinic), rucaparib, niraparib, talazoparib, veliparib, and pamiparib (Fig. 1). Olaparib, rucaparib, and niraparib have all obtained FDA and/or EMA approval in ovarian cancer in different settings. Olaparib, and very recently talazoparib, are currently the only FDA-approved PARP inhibitors for metastatic BRCA1/2-mutant breast cancer (20, 21). Although pamiparib only recently entered phase I trial testing, a favorable safety profile and preliminary antitumor activity have resulted in the initiation of randomized phase III trials versus placebo for maintenance therapy in both platinum-sensitive gastric and ovarian cancers (22). Veliparib does not yet have an approved label, and its use is being investigated mostly in combination with chemotherapy or targeted agents. The likely reason for this is described below.

Preclinical and clinical data so far on the different PARP inhibitors have revealed many similarities, but also notable differences, resulting from the different chemical structures of each PARP inhibitor (Fig. 1; refs. 23, 24). In addition, sequential tumor and/or liquid biopsies in PARP inhibitor–treated preclinical models and patients are giving researchers a better understanding of the underlying mechanisms of sensitivity and resistance to this class of drugs, potentially paving the way forward for the development of functional biomarkers for rational patient selection.

The PARP inhibitors currently approved for clinical use and those still under trial development all share a similar capacity to outcompete NAD⁺ binding to the PARP catalytic domain and to inhibit the acute PARylation of downstream substrates and PARP itself, with anticatalytic activity seen in the nanomolar range for each drug (Fig. 1; refs. 24, 25). Clinical studies of PARP inhibitors have used the quantification of PAR chain formation in peripheral blood lymphocytes as a pharmacodynamic biomarker, demonstrating that they can cause the acute and complete inhibition of PARP enzymatic activity, even at subtherapeutic doses. Importantly, such abrogation of enzymatic activity was not found to correlate with clinical responses, and the reasons behind this have not at this time been satisfactorily explained.

Despite the mechanistic similarities in anticatalytic function, the six PARP inhibitors differ in their chemical structure, preclinical potency, and clinical doses used for patients (Fig. 1). With respect to chemical structure, veliparib is the smallest of the PARP inhibitors, whereas talazoparib is the largest in size and possesses a more rigid structure (25). These differences in size and rigidity are thought to be the basis for the off-rate of each PARP inhibitor, and therefore the distinct capacity of each drug to prevent the release of bound PARP1/2 from chromatin—a phenomenon known as “PARP trapping” (25). However, PARP inhibitors do not trap PARP to DNA simply by inhibiting the catalytic activity and auto-PARylation of PARP in the short term; otherwise the class of PARP inhibitors would not display such different PARP-trapping abilities and single-agent cytotoxicities. Rather, as described above, PARP trapping, in addition to the inhibition of catalytic activity, also requires a slow “off-rate,” and as long as the inhibitor is bound to the active site, NAD cannot be utilized for auto-PARylation and DNA dissociation. Talazoparib is able to bind chromatin and create these trapped PARP–DNA complexes to an approximately 100-fold greater degree than rucaparib, niraparib, or olaparib, whereas veliparib displays negligible PARP-trapping ability (23–26). Although preclinical data on pamiparib is more limited than the other PARP inhibitors, a study in cell line models showed that this drug is able to form PARP–DNA complexes at an IC₅₀ of 13 nmol/L. In addition, pamiparib displayed 10-fold greater potency and antitumor activity in a BRCA1-mutant breast cancer xenograft compared with olaparib (27, 28). However, while sufficient PARP trapping is required for single-agent activity in HRR-deficient cancers, it does not directly correlate with clinical efficacy. This is because the stronger PARP trappers often have to be used at lower doses in the clinic due to lower MTD achieved (25).

Novel insights are now emerging that may help to explain the basis of the balance between efficacy, specificity, and tolerability profiles in the clinic. For example, a recent study by Leo and colleagues comparing all aforementioned PARP inhibitors, apart from pamiparib, showed that the cytotoxic potential of each of the five PARP inhibitors differs between HRR-deficient (HRD) and HRR-proficient isogenic cell lines, whereby olaparib showed the most HRD-specific sensitivity, and talazoparib was the most agnostic to HRR status with regard to its cytotoxicity. Veliparib displayed the least efficacy in both the HRD and HRR-proficient lines (24).

Each of the PARP inhibitors have shown generally favorable safety profiles in clinical trials, with the most common side effects including fatigue, gastrointestinal (GI) symptoms, and bone marrow suppression. Anecdotally, PARP inhibitor–related toxicities, such as GI symptoms and fatigue, have interestingly been observed to improve over time while on therapy. Although the underlying mechanisms for this clinical phenomenon are still unclear, they have important implications for encouraging patients and physicians to continue PARP inhibitor therapy at their starting doses while optimizing supportive therapies. The frequency and grade of cytopenias do seem to differ between PARP inhibitors in clinical trials carried out to date, with talazoparib arguably showing the highest occurrence of cytopenias, particularly anemia and neutropenia (20), followed by niraparib (24, 29), with high rates of grade 3 or greater thrombocytopenia and neutropenia in particular, and then olaparib (21, 30, 31) and rucaparib (32) with similar lower rates of all cytopenias versus talazoparib and niraparib. Although a direct comparison between these trials is not possible given that they are in distinct patient populations, it is notable that the rate of occurrence of grade 3 or higher bone marrow suppression tends to mirror the respective PARP-trapping ability of these PARP inhibitors. Late-phase testing of pamiparib is currently underway, thus adverse event data are not mature, but early-phase data (33) did show that the most common grade 3 toxicities were anemia and neutropenia, in a similar vein to the other “PARP trappers.”

Interestingly, a number of unique secondary paharmacology–based toxicities specific to each PARP inhibitor have also been observed in the clinic (Fig. 1). For example, niraparib leads to hypertension in approximately 19% of patients (29), which is thought to be due to the off-target inhibition of dopamine transporters, among others (34). It is important for clinicians to be aware of such distinctive PARP inhibitor–specific toxicities, so that they can be managed effectively in the clinic if they arise, and also to guide the selection of the most appropriate PARP inhibitor to use to minimize any overlapping toxicities when designing combinations with other antitumor agents, which come with their own side effects.

Although there have not been any direct head-to-head comparisons of PARP inhibitors in clinical trials, the phase III monotherapy data of olaparib, niraparib, rucaparib, and talazoparib have been published, while phase III trial data involving veliparib in combination with chemotherapy have also been presented (Fig. 1). The clinical data so far have been encouraging for olaparib, rucaparib, niraparib, and talazoparib, mostly centered around their single-agent use in patients with ovarian or breast cancer harboring BRCA1/2 mutations. Importantly, statistically significant clinical benefit of maintenance olaparib, rucaparib, or niraparib extended to all patients with platinum-sensitive high-grade serous ovarian cancer, regardless of DDR gene mutation status or HRD as determined by specific genomic assays, likely due to the high incidence of HRR deficiency and continued platinum sensitivity of the cancers being treated in this patient population (29, 31, 35). The very recent SOLO1 data (36), where olaparib was given daily as a first-line maintenance therapy in advanced ovarian cancer harboring BRCA1/2 mutations, suggest that these agents are going to transform ovarian cancer patient outcomes. The 5-year survival rate for ovarian cancer is only 40%, with 70% of patients progressing within 3 years following their first line of platinum-based chemotherapy (36). However, in the SOLO1 trial, 60% of patients had not progressed after 3 years, in contrast to less than 30% on placebo, even though olaparib therapy in many cases had only been given for 2 years (36). These data far outweighed the expected benefits going from second-line maintenance to this earlier line of therapy. Moreover, there was no significant reduction in quality of life compared with the placebo arm, and 70% of patients remained on full doses of drug during the trial (36).

Veliparib given in combination with carboplatin/paclitaxel chemotherapy for BRCA1/2-mutant metastatic breast cancer in the BROCADE-2 trial showed some promise in phase II with an overall response rate (ORR) 77.8% in the veliparib arm and 61.3% in the placebo arm (37). However, there was no benefit in progression free survival (PFS) with the addition of veliparib in BROCADE-2, and phase III results are currently pending (BROCADE-3) (37). Phase III results available thus far including veliparib in combination with carboplatin/paclitaxel chemotherapy for 1) patients with treatment-naive metastatic smoking-related non-small-cell-lung cancer (NCT02106546) and 2) neoadjuvant veliparib in combination with carboplatin/paclitaxel for TNBC (NCT02032277) failed to meet their primary endpoints of improved overall survival (OS), and improved complete response (pCR), respectively (38, 39). A recent phase II study of veliparib in combination with temozolomide for recurrent small-cell lung cancer (SCLC) showed improved ORR for the veliparib arm versus placebo (39% vs. 14%); however, a median OS and progression-free survival (PFS) benefit were only seen in patients with overexpression of schlafen11 (SLFN11), an emerging biomarker of PARP inhibitor response that is discussed later in this review (40).

In SCLC xenograft studies, single-agent talazoparib displayed striking single-agent activity, similar to cisplatin (41). In another xenograft study, talazoparib was shown to be superior in its radiosensitizing capacity compared with veliparib, probably due to the enhanced PARP-trapping ability of talazoparib (42). In this study, talazoparib was shown to synergize with temozolomide independent of SLFN11 status in preclinical SCLC models (42). Thus, while phase II trial data for veliparib look promising in recurrent SCLC, the more potent PARP-trapping drugs may prove to be more adept therapies in this setting, although overlapping toxicities with chemotherapy, including bone marrow suppression, are likely to be dose limiting. There is currently a trial underway combining olaparib and temozolomide (NCT02446704), which has shown encouraging preliminary activity in phase I testing with an ORR of 46% in patients with SCLC (43). Novel strategies are being developed to mitigate toxicities observed with PARP inhibitor and chemotherapy combinations. For example, in preclinical studies, by conjugating talazoparib to a low-pH–sensing peptide, it was possible to selectively deliver the PARP inhibitor across the membrane of tumor cells and achieve target engagement in the tumor without free drug detected in the systemic circulation, thus having the potential to reduce chemotherapy combination bone marrow toxicity (44).

Although clinical benefit has been observed for olaparib in the aforementioned patient populations with breast or ovarian cancers, a phase III study of paclitaxel with or without olaparib for patients with metastatic gastric cancer failed to meet statistical significance in its primary endpoint of improved OS, even in those harboring a predicted biomarker of response—low ATM protein expression (45). This lack of efficacy is likely multifactorial, including a lack of standardization in the IHC assay to quantify loss of ATM protein expression, not enriching for patients with true HRD, the absence of an olaparib monotherapy arm as a comparator, as well as a suboptimal dose of olaparib and lack of synergy between PARP inhibition and paclitaxel, all of which may have been contributing factors. Further studies of PARP inhibitors and associated biomarkers are currently being pursued in gastric cancer.

PARP inhibitors are also under active investigation as part of treatment regimens for primary brain cancers such as glioblastoma (GBM), as well as metastatic brain disease. The ability of the drug to cross the blood–brain barrier (BBB) is different for each PARP inhibitor. A phase I study of olaparib in combination with temozolomide for patients with relapsed GBM showed a high incidence of grade 3 or higher toxicities (∼70% of 35 evaluable patients), predominantly myelosuppression (46). The study did show that olaparib was present in both core and margin GBM tissue specimens, despite olaparib not being able to penetrate the BBB at clinically relevant doses in preclinical xenograft models with an intact BBB. The 6-month PFS was 45% in 13 evaluable patients, providing supporting evidence to proceed to late-phase trial testing of olaparib in combination with radiotherapy ± temozolomide for newly diagnosed GBM (PARADIGM-2; ref. 47). In contrast to olaparib, niraparib was shown to effectively penetrate the BBB in preclinical models and have a significant antitumor effect in germline BRCA2-mutant intracranial xenografts at clinically relevant doses (48). In preclinical studies, talazoparib sensitized GBM cells to temozolomide in vitro and in flank GBM xenografts; however, there was no benefit for the addition of talazoparib in the intracranial xenograft models (49). In this study, talazoparib demonstrated poor penetration across the BBB because of multidrug-resistant 1 (MDR1) efflux activity at the barrier (49). However, in the EMBRACA trial, the PFS benefit of talazoparib extended to patients with central nervous system metastases in subgroup analysis (20). Xenograft studies of rucaparib showed a similar discordance in efficacy between in vitro and flank xenografts compared with intracranial xenografts, also due to drug efflux transporter activity leading to poor BBB penetration (50). Veliparib despite synergizing with temozolomide in temozolomide-resistant GBM lines in vitro, did not display efficacy at clinically relevant doses in vivo, even in a flank xenograft model (51).

Although it is unlikely that there will be a clinical trial comparing PARP inhibitors head-to-head in the near future, it is evident from preclinical data and clinical studies conducted so far that the success of a specific PARP inhibitor may ultimately be influenced by the specific biomarker-selected population being targeted. Although the synthetic lethal relationship between PARP inhibition and HRR deficiency resulting from BRCA1/2 deficiency is known, there are also cancer cell types that display other forms of PARP-dependency, including mechanisms that result in an HRD-like phenotype. For example, deleterious variants in non-BRCA HRR genes such as ATM, PALB2, and RAD51 lead to functional HRR deficiency and sensitivity to PARP inhibition. Outside of canonical HRR pathway genes, mutations in Krebs cycle genes such as IDH1 or FH lead to oncometabolite production that results in reduced expression of HRR gene activity (52, 53). In addition, mutations in key chromatin regulators, such as ARID1A (54) and BAP1 (55, 56), give way to an HRD phenotype via loss of sustained DDR signaling and ubiquitylation activity that is needed in DSB repair. Finally, there are biomarkers of PARP inhibitor sensitivity that are distinct from HRD altogether as further discussed below.

With the incorporation of PARP inhibitors in multiple ongoing trials, a wide range of companion biomarkers that relate to HRD, beyond BRCA1/2 mutations, are also under development, as previously discussed and extensively reviewed (6, 57).

Although the rationale for targeting BRCA1/2- and HRD-associated malignancies with PARP inhibitors is well described, several additional tumor types, including Ewing sarcoma, some aggressive variant prostate cancers, and SCLC, display increased sensitivity to PARP inhibitors in spite of a seemingly intact HRR pathway. The best characterized of these non-HRD examples is probably SCLC, where previous proteomic characterization has identified high expression of not only PARP itself, but also a number of additional DDR-related proteins including ATM, ATR, CHK1, and CHK2 (58). The unique genomic profile of SCLC includes the nearly ubiquitous genetic loss of TP53 and RB1, and thus the loss of the tumor-suppressive roles that these proteins play in response to DNA damage (59). In addition, SCLC is known to display frequent MYC amplification, thereby providing additional oncogenic stresses during the tumor cell cycle (59). The net effect of these genomic alterations is a rapidly dividing tumor under immense replication stress due to checkpoint loss and thus, a tumor that is heavily dependent on a robust DDR to maintain its survival (34). The loss of RB1 assists in enhancing DDR capacity due to the absence of the inhibitory effects of RB1 on the E2F1 transcription factor responsible for the expression of PARP1 and many other DDR-related genes (58). Relative to NSCLC, SCLC models showed increased sensitivity to olaparib in vitro in terms of cytotoxicity, but also reductions in PAR levels (58). Despite the initially robust DDR machinery in SCLC, PARP inhibitor treatment resulted in broad downregulation of multiple DDR proteins, likely due to the role of PARP1 as a coactivator of the aforementioned E2F1 (58). Thus, in SCLC, PARP inhibition is capable not only of abrogating the role of PARP itself but of reversing much of the DDR protein upregulation that maintains DNA integrity amidst high levels of replication stress in this cancer (Fig. 2).

It is now clear, however, that the sensitivity of SCLC and other high replication stress tumors to PARP inhibitors is not ubiquitous. For example, while a phase I trial of talazoparib in patients with advanced cancers reported robust responses in patients with BRCA1/2-mutant breast and ovarian cancers, objective responses were only observed in 2 of 23 patients with SCLC, and 2 of 10 patients with pancreatic cancer, and no objective responses in small populations of patients with Ewing sarcoma and metastatic castration-resistant prostate cancer (60). Significant efforts have been made to identify predictive biomarkers of response to PARP inhibitors and combinations in SCLC, with varying degrees of clinical success. In vitro, analyses in SCLC cell lines and cell line–derived xenografts identified expression of the DNA-PKcs protein, as well as a 5-gene expression score (GLS, UBEC2, HACL1, MSI2, LOC100129585), which predicted sensitivity to veliparib (61). In a separate preclinical study, a DNA repair score developed on the basis of the overexpression of 17 DDR proteins, rather than deficiency in DDR proteins, predicted sensitivity to talazoparib in SCLC in vitro models (41). Multiple studies have also identified baseline PI3K/mTOR pathway activation as a marker of PARP inhibitor resistance in SCLC and demonstrated that PARP inhibition itself leads to PI3K activation (41, 62). Another potential predictor of PARP inhibitor sensitivity in SCLC includes high expression of E-cadherin (63). However, none of these biomarkers have so far been supported by clinical data.

One of the most promising predictive biomarkers to emerge recently is the expression of SLFN11 (Fig. 2); this putative DNA/RNA helicase protein has been identified by different research groups as a predictor of sensitivity to multiple PARP inhibitors, as well as platinum chemotherapy, in both in vitro and patient-derived xenograft models (63–66). The predictive role of SLFN11 for talazoparib responses in vitro appears to extend beyond SCLC across the NCI-60 collection of cell lines, where SLFN11 expression is second only to BRCA inactivation in its correlation with drug sensitivity (67). Furthermore, SLFN11 is under transcriptional control of the EWS–FLI1 fusion protein that defines Ewing sarcoma and has been linked to PARP inhibitor sensitivity (68). SLFN11 helps orchestrate the DDR by binding to chromatin in response to replication stress through association with RPA (34), thus blocking replication fork progression. In the absence of SLFN11, this irreversible and lethal replication inhibition does not occur (67). SLFN11 expression is dynamic, owing to silencing by promoter methylation, which is responsible both for innate and acquired SLFN11-dependent resistance (69), although treatment with EZH2 and HDAC inhibitors may be capable of restoring SLFN11 expression, and therefore PARP inhibitor sensitivity (70, 71).

Efforts to analytically validate existing biomarkers and to identify new ones to predict PARP inhibitor response are ongoing, including novel strategies utilizing circulating tumor cells (CTC) and CTC-derived xenograft (CDX) models (72). In this study, patients enrolled on a phase I/II trial of temolozomide and olaparib for previously treated SCLC had blood collected at multiple time points prior to, during, and following trial therapy. These blood samples were subjected to CTC enrichment using the CTC-iChip microfluidic device and the resultant CTCs injected into mice to generate longitudinal CDX models. Preliminary analysis from this ongoing study revealed that basal protein PARylation distinguishes sensitive CDX models from both intermediate and resistant models, while PARylation, like SLFN11 expression, appears to decrease with treatment and the development of resistance (72). A similar CDX-based approach investigating the combination of olaparib with or without the WEE1 inhibitor adavosertib (AZD1775) identified multiple additional biomarker candidates, including inducers and markers of replication stress, such as MYC-family proteins, phospho-RPA, and cyclinE1 (73). Given the relative dearth of tissue available in SCLC and other tumors not commonly managed via surgical resection, approaches such as these may shape the next era of biomarker discovery for PARP inhibitors and beyond.

As highlighted above, there have been multiple positive late-phase trials and subsequent regulatory approvals for the use of PARP inhibitors in pretreated patients with metastatic BRCA1/2-mutant cancers. Neoadjuvant studies exploit the possibility of PARP inhibitors to potentially spare patients the detrimental toxicities and likely impact on quality of life of chemotherapy, while also obtaining vital information on the underlying tumor biology and responsiveness to PARP-targeted therapy. Recent phase II neoadjuvant trial data for the use of single-agent neoadjuvant talazoparib for patients with BRCA1/2 mutant, HER2 normal (predominantly triple negative) operable breast cancer were very encouraging, with 53% of women on study achieving a pathologic complete response, and over 60% ORR reported during preliminary analysis (74). Toxicities observed were similar to those reported in studies of talazoparib administered in the metastatic setting, including grade 4 thrombocytopenia in 1 of 18 patients, and dose reductions were required in 9 of 18 patients. Pretreatment and postoperative tissue profiling are expected to provide additional novel insights into how PARP inhibition alters genomic stability, DDR and oncogene pathway function, and the immune microenvironment in this unique patient population. A small phase II study of veliparib plus carboplatin added to a standard neoadjuvant backbone regimen of paclitaxel and adriamycin/cyclophosphamide for localized TNBC patients appeared to show improved pathologic complete response rates. However, the subsequent phase III BriGHTness trial failed to meet its primary endpoint, demonstrating that the addition of carboplatin alone to the standard chemotherapy regimen was sufficient to benefit patients, without the need for veliparib (38). The failure of veliparib in this setting may be multifactorial, but its value as a PARP inhibitor in the context of available, more potent “PARP trapping” agents remains questionable.

In the adjuvant setting, results are pending from the phase III OlympIA trial, which is assessing olaparib as adjuvant therapy for patients with BRCA1/2-mutant localized breast cancer who have completed neoadjuvant chemotherapy and definitive treatment.

Given the relationship between PARP inhibitors and germline BRCA1/2 mutation carriers, chemoprevention strategies have also been assessed in BRCA-mutant preclinical models (75). These studies showed that physiologic doses of veliparib or olaparib were able to delay mammary gland tumor onset in BRCA1-deficient mice, with olaparib showing an improved delay in tumor onset and also improved overall survival in the mice even when given intermittently rather than continuously. However, it is important to note that these mice had biallelic loss of BRCA1 and all still eventually developed tumors, thus implying that PARP inhibition delayed, but did not prevent cancer in this mammary tumor model. Moreover, the efficacy and risk-benefit of the use of PARP inhibitors in the case of monoallelic loss or haplo-insufficiency of BRCA1/2, as is the case in patients with germline mutations in these genes who have not yet developed a cancer, is not currently clear. In addition, given that: (i) not every patient with a HBOC-associated mutation will develop a cancer, (ii) clinical trials have shown that not all patients with BRCA1/2-mutant cancers respond to PARP inhibition, and (iii) long-term toxicities of PARP inhibitor use have not been fully elucidated (76, 77), such chemoprevention strategies still require thoughtful discussion and study. In particular, it is still unclear if the rare (<1%) incidence of acute myeloid leukemia and myelodysplasia observed in long-term ovarian cancer maintenance treatment can be attributed to PARP inhibition, or if it is due to prior platinum-based chemotherapy. Given the high risk of cancer in germline BRCA1/2 mutation carriers, chemoprevention with PARP inhibitors may warrant investigation, perhaps starting with an assessment of the potential to delay or prevent secondary cancers in such carriers.

In addition to the aforementioned strategy of combining PARP inhibitors with DNA-damaging chemotherapy and/or radiation, preclinical evidence has led to multiple biologically informed clinical trials combining PARP inhibitors with (i) other DDR inhibitors, such as those that target ATR, CHK1/2, or WEE1, (ii) agents that target oncogenes, and (iii) immune checkpoint therapy (Fig. 3).

Acquired or innate resistance to single-agent PARP inhibitors is frequently observed in both preclinical models and the clinic. Multiple potential mechanisms for resistance to PARP inhibitors have been described with the majority linked to routes by which HRR capability is restored (78). The mechanisms have been extensively previously reviewed (7, 78, 79), and include reversion to wild-type mutations in BRCA and other HRR genes, promoter demethylation of suppressed DDR genes, mitigation of replication stress, mutations in PARP itself, and/or drug efflux pumps, among others. This overarching mechanism of HRR restoration has been highlighted in multiple preclinical PDX studies of PARP inhibition in TNBC (15, 80–82), where functionality of HRR in virtually all cases was implied by the presence of RAD51 foci in untreated tumor samples, suggesting this could represent a useful clinical biomarker for PARP inhibitor response and/or resistance. Emerging data also suggest that these olaparib-resistant cancer models can be resensitized to olaparib when combined with a WEE1 inhibitor or an ATR inhibitor (67, 73, 83–85). Preclinical studies suggest that PARP inhibitor–resistant, BRCA-deficient cells have an increased reliance on ATR signaling for fork stabilization (83, 84), while synthetic lethal screens identified ATR as a target that was able to overcome PARP inhibitor resistance, leading to early-phase clinical trials combining ATR and PARP inhibitors (NCT02723864, NCT03462342, NCT03682289, NCT02576444; ref. 85). In addition to reversing PARP inhibitor resistance, WEE1 inhibitors may serve as a promising partner to PARP inhibitors, with the synergy dependent on the PARP-trapping ability of the PARP inhibitor (86). Encouragingly, in preclinical chemosensitive and chemorefractory SCLC models, the combination of olaparib and adavosertib provided superior efficacy versus the standard combination of cisplatin and etoposide (73).

The cytotoxic mechanism of action of these DDR inhibitor combinations is multifactorial, with contributions from an over-reliance on alternative DDR pathways in PARP-inhibitor–resistant cells, as well as catastrophic replication stress and nucleotide resource depletion when PARP inhibitors are combined with ATR inhibitors or WEE1 inhibitors. There are also other DDR inhibitors targeting multiple points along the cascade in clinical development, including ATM, CHK1/2, DNA-PK, and POLθ, which all have the potential for effective combinations with PARP inhibitors (78). Importantly, in designing these DDR–DDR inhibitor combinations, one must consider the potential for overlapping bone marrow toxicity. Further studies are thus warranted to establish the appropriate doses, schedule, and sequence of these therapies.

A major theme of PARP inhibitor resistance from preclinical studies is oncogene-driven expression of HRR genes and rescue of repair activity. For example, androgen receptor (AR) signaling has been shown to alter HRR and cell-cycle gene expression in castration-resistant prostate cancer (CRPC), with DDR genes upregulated at progression, and PARP1 required for maximal AR function (87, 88). The rational combination of AR blockade and PARP inhibition has thus been explored in preclinical models and clinical trials (89). Enzalutamide was shown to reduce BRCA1 expression in CRPC cell lines, inducing a “chemical BRCAness” phenotype, potentially resulting in the synergy observed with the combination of enzalutamide and olaparib in preclinical models. These findings were regardless of DDR gene mutation status, whereby lead-in therapy with enzalutamide reduced HRR capacity, increased apoptosis, and had antitumor effects in prostate cancer models (87, 89). The recently published phase II trial of abiraterone/prednisone with or without olaparib showed a clinical benefit for the addition of olaparib with improved radiographic PFS (13.8 vs. 8.2 months; ref. 90). Importantly, in an exploratory analysis, this benefit appeared to extend to patients who did not have DDR gene alterations, though less than half of the study population had the appropriate DNA sequencing performed and the study was not designed to detect differences in DDR gene mutation carriers (90). Unlike the aforementioned olaparib trial, a phase II multicenter trial of abiraterone/prednisone with or without veliparib for patients with pretreated CRPC did not show any benefit for veliparib over placebo (91). The selection of the specific PARP inhibitor to use in combination with targeted agents is clearly important and must be informed by preclinical mechanistic data.

The combination of PI3K/mTOR inhibition and PARP inhibition has also been suggested to result in efficacy in preclinical studies of various cancer types via multiple mechanisms centered on inducing HRD, including the suppression of DNA DSB repair protein SUV39H1, a histone methyltransferase, and suppression of HRR gene expression (92). An early-phase clinical trial of olaparib in combination with buparlisib, a pan-PI3K inhibitor, showed approximately 30% response rate for the combination (93). However, multiple dose reductions of buparlisib for treatment-related toxicities led to the discontinuation of this particular combination. Subsequently, olaparib was combined with a more specific PI3Kalpha inhibitor, alpelisib, and showed an ORR of 36% in patients with advanced ovarian cancer, who were largely platinum-resistant (94). Importantly, the benefit extended to patients with and without germline DDR gene mutations (94).

RAS-mutant cancers are notoriously difficult to target; however, preclinical evidence has shown that mitogen-activated protein kinase inhibitors reduce HRR gene expression and DNA damage checkpoint activation, increase FoxO3a expression, and synergize with PARP inhibition for antiproliferative activity in RAS-mutant cells (95). A phase I/II trial of selumetinib and olaparib for patients with RAS-altered cancers, as well as PARP-inhibitor–resistant ovarian cancers is underway (NCT03162627), among others in development. The combination of anti-VEGF small molecules with PARP inhibitors has also shown promise in patients with recurrent ovarian cancer. Phase II trial data of cedirinib in combination with olaparib showed a PFS benefit versus single-agent olaparib (16.5 vs. 8.2 months) in a platinum-sensitive, BRCA wild-type patient population. However, grade 3 or higher toxicities were frequent (75%; ref. 96). The mechanism of synergy with this combination is not as well understood, with the current thinking being that antiangiogenic therapy may create tumor hypoxia, which has been shown to alter DDR gene expression (97). Interestingly, patients with BRCA1/2 mutations derived no benefit from the addition of cedirinib to olaparib versus olaparib alone, suggesting that this combination may best be reserved for the patients with an intact HRR phenotype (96).

BET bromodomain (BRD4) protein promotes oncogene transcription, and BET inhibitors have been shown to suppress DDR genes, including TOPBP1, WEE1, and DNA DSB repair protein CtiP in some cell line models (98). The combination of BRD4 and PARP inhibition demonstrated enhanced activity in multiple tumor lineages, regardless of BRCA1/2, TP53, RAS, or BRAF mutation status both in vitro and in vivo (98). Consequently, BET inhibitors in combination with PARP inhibitors may have wide application in the clinical setting and are under active early-phase clinical trial investigation.

Preclinical studies are rapidly uncovering new targets that can induce HRD and/or overcome PARP inhibitor resistance with biologically informed clinical trials evolving from these novel mechanistic insights. The goal in the clinic is now to minimize overlapping toxicities by optimizing dose and schedule, for example, concurrent versus alternating versus sequential dosing, determine the order of drugs in the treatment sequence, and to limit the use of combinations to those patients who would not otherwise obtain the same benefit from PARP inhibitor monotherapy. Patient selection approaches incorporating tumor type and molecular profiles specific to each unique PARP inhibitor–based combination will also be essential to optimize efficacy, while minimizing toxicity of these rationale strategies. Although it is likely that combinations involving three or more agents are feasible, we will need to establish safety and efficacy in doublet combinations first before moving to assess strategies involving multiple drugs.

Patients with cancers harboring innate deficiencies in canonical DDR genes, including mismatch repair and HRR genes, have been shown to have increased CD8⁺ T-cell infiltration and improved ORR to immune checkpoint therapy (13). The intertwined relationship between immunotherapies and DDR pathways is extensive, and the synergy between DDR deficiency and immune activation against cancer cells is multifactorial (13, 99–101). Preclinical and clinical data indicate that neoantigen burden primarily drives this response in patients with hypermutated tumors, such as in the case of MMR and POLE deficiencies (102–104). For patients with HRR gene aberrations where mutational burdens are lower than in MMR (105), there is also likely to be a contribution from the generation of S-phase–specific DNA damage resulting from collapsed replication forks or underreplicated DNA, leading to the accumulation of cytosolic DNA, which in turn can activate the cGAS–STING innate immune pathway and type I IFN signaling (Fig. 3; ref. 106). Interestingly, PARP inhibition has also been shown in preclinical models to inactivate GSK3β and upregulate PD-L1 in a dose-dependent manner, suppressing T-cell activation and increasing cancer cell killing (107). These effects were observed in rucaparib-, olaparib-, and talazoparib-treated cancer models; and subsequent combination therapy with anti–PD-L1 blockade induced PARP inhibitor sensitization and led to antitumor activity to a greater degree than either drug alone (107).

A phase II trial of durvalumab in combination with olaparib for selected advanced solid cancers (MEDIOLA) showed that this combination is well tolerated with no significant overlapping toxicities (108–111). Clinical response measured as disease control rate at 12 weeks was 29%, 80%, and 81% in patients with advanced SCLC, germline BRCA1/2-mutant HER2-normal breast cancer, and germline BRCA1/2-mutant platinum-sensitive ovarian cancer, respectively (109–111). This same combination was also assessed in patients with CRPC, and reported a 12-month PFS of 51% in a population unselected for DDR variants (112). In molecularly driven tumor types where PARP inhibitor monotherapy activity is already observed, the hope is that such a combination will lead to deeper and more durable responses in a greater proportion of patients, including long-lasting complete responses.

Given the increasing importance of immunotherapy for the management of cancer patients with multiple tumor types, one of the most clinically important questions just beginning to be answered is to what extent PARP inhibitors may enhance response to immune checkpoint blockade or other immunotherapy approaches. Ongoing and future trials will determine the degree to which the therapeutic targeting of DNA repair by PARP inhibition or other DDR inhibitors may similarly enhance responses, and the underlying mechanisms through which this occurs. Finally, as with other targeted agents, optimal biomarkers to predict benefit and resistance from PARP inhibitors and specific combinations will be essential to enable the rational expansion of this class of drugs to the additional patient populations who may benefit.

Preliminary safety and efficacy data from the phase II trial of niraparib and pembrolizumab for patients with advanced platinum-resistant ovarian cancer (TOPACIO) demonstrated no additive toxicities, an ORR of 25%, and disease control rate (including stable disease) of 68% (113). A subgroup analysis of patients with BRCA1/2 mutations revealed that their disease control rate (73%) did not differ significantly from the general study population; however, the ORR in BRCA1/2-mutant patients was 45% (113). It is not yet clear which specific PARP inhibitor, or even other DDR inhibitors, may combine best with immune checkpoint blockade, and what the optimal predictive biomarker strategy is, that is, if there will be efficacy independent of HRD status. Nevertheless, the data reported from these early studies are certainly encouraging for this novel combination strategy.

PARP inhibitors are the first approved DDR-targeted medicines and have already transformed treatment paradigms for subgroups of patients with ovarian and breast cancers (20, 21, 29, 30, 32). DDR deficiencies are common in cancer and indeed have been postulated to be a necessary component of tumorigenesis, but they also represent an Achilles' heel that can now be targeted (2). It is likely that with the significant improvement in patient benefit observed in earlier therapeutic settings, along with the likelihood of long-term tolerability of PARP inhibitors, there is great potential for this drug class to become a foundation treatment for ovarian cancer, and for its impact to extend to multiple other cancers and far beyond BRCA1/2-mutant tumors. Preclinical mechanistic studies are now guiding hypothesis-testing, biomarker-driven clinical trials that look beyond BRCA1/2 mutations, toward a broader view of HRD phenotypes, and even beyond HRD itself to maximize the number of individuals who may benefit from PARP inhibition.

If we also take into account the potential to reverse acquired PARP inhibitor monotherapy resistance using PARP inhibitor–DDR inhibitor combinations (through a “PARP-after-PARP” inhibitor approach), while broadening the patient populations beyond what is obtainable through monotherapy via combinations with other targeted agents such as inhibitors of VEGFR, PI3K pathways, other oncogenic drivers such as AR and immunotherapy, as well as other DDR-targeted agents, it is clear that PARP inhibitor benefits have the potential to go well beyond the initial impressive advances experienced in recent years.

L.A. Byers reports receiving other commercial research support from and is a consultant/advisory board member for AbbVie and AstraZeneca. M.J. O'Connor is a shareholder in AstraZeneca. T.A. Yap reports receiving other commercial research support from AstraZeneca, Bayer, Pfizer, Tesaro, Jounce, Eli Lilly, Seattle Genetics, Kyowa, Constellation, and Vertex Pharmceuticals, speakers bureau honoraria from AstraZeneca, Merck, Pfizer, and Tesaro, and is a consultant/advisory board member for Aduro, Almac, AstraZeneca, Atrin, Bayer, Bristol-Myers Squibb, Calithera, Clovis, Cybrexa, EMD Serono, Ignyta, Janssen, Merck, Pfizer, Roche, Seattle Genetics, and Vertex Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

The University of Texas MD Anderson Cancer Center (Houston, TX) is supported by an NIH Cancer Center Support Grant (CA016672; to P.G. Pilié, C.M. Gay, L.A. Byers, and T.A. Yap).