PARP Inhibitors in Cancer Diagnosis and Therapy

Targeting DNA damage repair (DDR) signaling is a fast-expanding field for cancer therapy. DNA damage causes genomic instability in cells that require DDR for rescue. DDR signaling triggers activation of repair protein transcription, hence leading to overexpression of the related repair protein and activation of mechanisms for DNA repair (1). PARP is an essential and abundant DNA repair protein, which plays an important role in regulation of various DDR pathways, including base excision repair, homologous recombination (HR), classical and alternative nonhomologous end joining, nucleotide excision repair, microhomology-mediated end joining, maintenance of replication fork stability, and mismatch repair (2). In addition, PARP has been found overexpressed in various types of cancers, such as breast, ovarian, and oral cancers than their normal surrounding healthy tissues, making inhibition of PARP activity a promising strategy for cancer therapeutics by disrupting PARP functions, thereby impairing DDR pathways of cancer cells. PARP inhibition for the treatment of breast or ovarian cancers with BRAC1/2 mutations (deficiencies in the HR) is now a well-established approach. A PARP inhibitor (PARPi) is utilized to impair the process of single-strand break (SSB) repair, thereby converting SSB into double-strand breaks by stalling replication folk (1) or accelerating fork elongation (3), which requiring HR for repair during S phase (4). Therefore, if HR is defective for example, resulting from mutations in HR genes such as BRCA1 and 2, DSBs cannot be repaired, and the cell dies, a concept known as synthetic lethality. Over the past decade, four PARPi (olaparib, rucaparib, niraparib, talazoparib) have been FDA approved for clinical use as single agents, with veliparib expected to receive approval following promising phase III trial results (BROCADE3; NCT02163694). The ability of PARPi to impair DDR processes led to their employment in combination with chemotherapy and radiotherapy. Increasing attention is also being drawn to radiolabeled versions of PARPi that possibly allow direct visualization of PARP expression, quantification of PARPi biodistribution and tumor delivery, as well as target engagement, thereby potentially improve stratification of patients and enable the monitoring of treatment progress. In addition, radiolabeled PARPi-based radionuclide therapy agents are being proposed as therapies for PARP-expressing cancers. Here, we present an up-to-date understanding of PARP in DDR, the use of PARPis as mono- or combination therapies for cancer, and the potential of radiolabeled PARPi in clinical applications.

PARP enzymes form a 17-member group of mostly nuclear enzymes that share a conserved catalytic domain homologous to ADP-ribosyl-transferases diphtheria toxin (ARTD), which gives the PARP enzyme family its alternative name of ARTD-like enzymes (5). Upon DNA damage, members of the PARP family bind the DNA lesion, become catalytically active, and mediate ADP-ribosylation using NAD⁺ as substrate, leading to rapid recruitment of other DNA repair factors (6). PARP-1, PARP-2 and tankyrase (PARP-5A/5B) produce poly-ADP-ribose (PAR) modifications on target proteins, while other PARPs are responsible for generating mono-ADP-ribose and have potential cancer-related functions (7). The evidence to date suggests that only PARP-1, PARP-2, and PARP-3 are activated by DNA breaks and catalyze ADP-ribosylation after DNA damage, with PARP-1 responsible for 80%–90% of global PAR synthesis in response to DNA damaging agents (6). Apart from SSB recognition, recent studies showed that PARP-1 can also detect DSB rapidly, and plays an essential role in regulation of DNA end resection for DSB repair, with recruitment and activation of PARP-1 occurring within 0.1 μs after induction of DSBs (6).

PARP-1 is the most studied of the PARP enzyme family by far. Its structure comprises of three zinc finger domains for DNA binding, a BRCA-C-terminus domain (BRCT), a WGR (Trp-Gly-Arg) domain that contributes to, but is not required for, DNA interaction and catalytic activity regulation in response to DNA damage, and a catalytic domain consisting of a helical subdomain (HD) adjacent to a conserved ADP-ribosyl transferase fold (ART; ref. 8). Compared with PARP-1 (N-terminal region: >500 residues), PARP-2, and PARP-3 have more compact structures, with shorter N-terminal extensions (70 and 40 residues, respectively) from the WGR domain (Fig. 1A; ref. 9). PARP-1 requires the zinc finger domains for DNA break detection and binding (10), but PARP-2 and PARP-3 rely on their WGR domain for this function (2, 11).

Despite differences in DNA damage detection, PARP-1, PARP-2, and PARP-3 share allosteric regulatory activation mechanisms, where binding to damaged DNA leads to an increase of PARylation activity (12). Langelier and colleagues. demonstrated that the HD domain in PARP-1 can selectively and reversibly block or allow access of NAD⁺ or the NAD analog BAD upon DNA binding (13). The HD domain, in its native state, was found primarily in a folded conformation, sterically blocking the catalytic binding site. Upon DNA damage detection, the binding of NAD⁺ to the catalytic active site stabilizes the HD domain in its unfolded conformation and strengthens the interaction between the regulatory domains and DNA (Fig. 1B). This reversible blockage of the NAD⁺ binding site provides control of NAD⁺ access to the catalytic site, regulating the PARylation process.

Clinical PARPis compete for binding in the NAD⁺ binding pocket of the PARP enzymes, and are mostly selective for PARP-1 and PARP-2. However, different PARP inhibitors display markedly different levels of cytotoxicity, attributed to the ability of PARP inhibitors to prevent auto-PARylation of PARP, thereby preventing its dissociation from the broken DNA, thus stabilizing and trapping PARP on the DNA (14). This trapping leads to DSB formation by stalling and collapsing replication forks (15) or accelerating fork elongation (3). Hence, different degrees of trapping (talazoparib > niraparib ≥ olaparib = rucaparib > veliparib (16, 17)) lead to large differences in cellular toxicity. Although the trapping abilities of PARPi are dependent on catalytic inhibition and the resulting dissociation rates of PARPi (17, 18), reverse allostery effects also play a significant role. Molecular impacts of PARPi binding to PARP-1 corroborated by molecular dynamics simulation, hydrogen-exchange mass spectrometry and X-ray crystallography implied that different PARPi induce diverse degrees of structural deviations in the regulatory domains, thereby positively or negatively affecting the retention of PARP-1 trapped at the DNA damage site (19, 20). Three types of PARPi were discerned on the basis of their allosteric impacts on the PARP-1–DNA complex (19): type I, allosteric proretention on DNA, which induces HD conformational changes that reinforce contacts with the DNA break and result in slow release; type II, with no or mild allosteric effects; and type III, allosteric prorelease from DNA, which induce HD conformational changes opposite to type I inhibitors (Fig. 1C). These results have illuminated that the PARP-1 trapping induced by some PARPi, such as type II PARPi (talazoparib and olaparib) and type III PARPi (rucaparib and niraparib), does not rely on the allosteric effect, indicating that other factors may contribute to the trapping ability of these PARPi, and therefore further investigation is warranted.

The efficacy of PARPi monotherapy in HR deficient (HRD) cancers carrying BRCA1/2 mutations was first demonstrated in cancer cell lines, and later confirmed in patient-derived xenograft (PDX) models. Treatment with olaparib resulted in significant tumor graft regression (21, 22) in PDX tumors of BRCA1-mutant triple-negative breast cancer (TNBC). Other alterations, such as absence of PTEN or ATM, also cause sensitivity of TNBC (23) and small-cell lung cancer (SCLC; ref. 24) to talazoparib treatment in vivo, confirming the effectiveness of PARPi beyond germline BRCA1/2 mutation cancers.

A wealth of clinical data supports the use of PARPi for maintenance treatment in patients with platinum-sensitive high-grade serous ovarian cancer (HGSOC). In the randomized, placebo-controlled, double-blind phase III SOLO2 trial, patients with relapsed epithelial ovarian, fallopian tube, or primary peritoneal cancer carrying a BRCA1/2 mutation demonstrated a significantly longer median progression-free survival (PFS) under olaparib maintenance treatment than those receiving placebo (19.1 vs. 5.5 months; 95% confidence interval, 16.3–25.7 vs. 5.2–5.8; HR 0.3), without significant detrimental effects on health-related quality of life (HRQoL; refs. 25, 26).

The antitumor activity of olaparib was highlighted in the recent phase III SOLO3 (NCT02282020) and OlympiAD (27) trials, in germline BRCA-mutated ovarian cancer and HER2⁻ breast cancer, respectively. Similarly, the ABRAZO study (28), and the latest EMBRACA trial, where patients underwent talazoparib monotherapy, showed a significant PFS advantage and higher objective response rate (ORR) over standard chemotherapy (8.6 vs. 5.6 months, 62.6% vs. 27.2%, respectively) with greater HRQoL in favor of talazoparib (29). Several recently reported trials including POLO and PROfound also demonstrated the benefits of PARPi (olaparib or rucaparib) over placebo or androgen receptor (AR)-directed therapy, respectively, in patients with metastatic pancreatic ductal adenocarcinoma (PDAC) harboring germline BRCA mutation (30) and in patients with castration-resistant prostate cancer (mCRPC) with confirmed HRD phenotype (31). Ongoing efforts to expand the use of PARPi to a wider population of patients with cancer in different settings beyond BRCA mutation will undoubtedly expand the potential use of PARPi as monotherapy, and several studies are currently underway (NCT03402841, NCT03286842, NCT02000622).

Heterogeneous responses to single-agent PARPi are frequently observed in both preclinical models and clinical practice, even in BRCA-mutated tumors, associated with complex intrinsic or acquired resistance. The mechanisms of PARPi resistance are well described in several recent reviews, with the majority related to restoration of HR capability (32). Rational combination treatment regimens thus become the next strategy to improve patient outcome, for example, using PARPi with chemotherapy, radiation, targeted drugs, or immunotherapy. The choice of partner must be based on the fundamental insights into the molecular pathways that drive tumorigenesis and therapy response. Strategies to optimize the use of PARPi as chemopotentiators or radiosensitizers are being actively pursued to overcome or prevent the development of treatment resistance in patients.

One approach is the combination of genotoxic agents and PARPi, where chemotherapy-induced DNA damage can increase the cellular reliance on DNA repair and hence sensitize tumors to PARPi, such as in ref. 33. Combination of PARPi and ionizing radiation (IR) is another effective strategy to sensitize tumors to radiotherapy, without overt signs of toxicity at least in mice, warranting further its evaluation in ongoing clinical trials (34), despite an early recognition of the limitation of this combinatorial approach in cells (35). Given the relationship between PARP and HR repair (HRR) regulatory pathways, it is not surprising to see the potential benefits of combination treatments between PARPi and DDR signaling targeting or otherwise genotoxic therapies. The increased reliance of cancer cells on alternative or specific DRR pathways provides an opportunity to pharmacologically induce HR deficiency, which in combination with PARPi achieves synergistic response. Disruption of, for example, checkpoint kinase 1 (CHK1) activity with PARP inhibition leads to the accumulation of DSB and induces apoptosis following treatment of PDX mouse models using olaparib/MK8776 (36) or olaparib/prexasertib (37) combinations. Tretinoin, used to treat acute promyelocytic leukemia patients, also acts as an effective HR disrupter and is capable of sensitizing BRCA-proficient TNBC to olaparib (38). Other studies involved the use of cyclic-dependent kinase (CDK) inhibitors to disrupt DNA damage checkpoint signaling and prevent DNA repair in neuroendocrine prostate cancer (NEPC; ref. 39) and TNBC PDX models (40). In addition, inhibition of bromodomain-containing protein (BRD4) disrupts DDR signaling by reducing DNA repair proteins levels, thereby sensitizing cancer cells to olaparib in preclinical models (41, 42). Treatment of mCRPC PDXs with an AR inhibitor enzalutamide induces downregulation of AR-regulated HRR gene expression and sensitizes to PARPi (43). An overview is presented in Table 1.

Promising efficacy of PARPi with genotoxic agents was observed in preclinical models and successfully translated in various cancer types: The phase III VELIA study demonstrated the efficacy of concurrent administration of veliparib and carboplatin with paclitaxel (V/CP) followed by veliparib maintenance in advanced patients with HGSOC with and without germline BRCA mutation. A significant PFS advantage was observed in both BRCA-mutant (34.7 vs. 22 months, HR 0.44), HR defects cohorts (31.9 vs. 20.5 months), and the intention-to-treat population (23.5 vs. 17.3 months, HR 0.68) receiving veliparib versus placebo with chemotherapy (44). Similar efficacy was observed in patients with breast cancer in phase II and III studies (45, 46).

Other clinical studies showed improved patient outcomes for treatment regimens involving PARPi combined with targeted drugs, such as antiangiogenic and AR antagonists. The recent FDA approval of olaparib/bevacizumab combination therapy as a first-line maintenance treatment for patients with HGSOC regardless of their HRD phenotypes based on the PAOLA-1 trial (47) will undoubtedly encourage additional studies to investigate other combination pairs (such as ICON9; NCT03278717, NCT02502266). For example, the use of olaparib/cediranib showed appreciable benefits in patients with ovarian cancer (48). Other promising studies include the combination of abiraterone and olaparib in patients with mCRPC, irrespective of HRR status (49). Combination of immune checkpoint inhibitors with PARPi is under active investigation, with only a limited amount of clinical data available to date. The MEDIOLA trial showed initial promise in a subset of patients with ovarian cancer (platinum sensitive, germline BRCA-mutated) receiving this combination (50), but no improvement in ORR or PFS when a more heterogeneous patient population (without BRCA mutation) was considered (51). Notwithstanding, the preliminary outcomes supported the tolerability of this combination in patients and better patient selections in subsequent trials should help to create clinically meaningful responses (NCT02484404). An overview is presented in Table 2.

Molecular imaging using PARP imaging agents could serve as a powerful tool for functional imaging of PARP in vivo. Tagging a PARPi with a radionuclide allows PET (¹¹C, ¹⁸F) or SPECT (¹²³I) imaging to potential clinical benefits: interrogation of PARP expression levels, PARPi distribution, tumor uptake, and drug-target engagement in a noninvasive manner, which may allow tumor diagnosis and staging, selection of patient subgroups suitable for PARP inhibition, monitor the treatment response, or gauge the emergence of resistance. If, on the other hand, the same chemical scaffold is radiolabeled with a therapeutic radionuclide, emitting alpha (²¹¹At), beta (¹³¹I), or Auger (¹²³I) radiation, the resulting radionuclide therapy agent is capable of causing cell death by delivering IR to PARP-expressing tumor cells.

Over the past 15 years, numerous radiolabeled PARP agents have been developed (Fig. 2A) and undergone characterization in vitro and in animal models, with two compounds already evaluated in early-phase clinical trials (34, 52). [¹¹C]PJ34 was the first radiolabeled PARPi, intended for PET imaging of tissue necrosis with upregulation of PARP-1 expression levels in a rat models of diabetes (53). The radiolabeled imaging agents [¹⁸F]BO (54), [¹⁸F]PARPi (55), [^123/131I]PARPi (56, 57), [¹⁸F]PARPi-FL (58), and [⁶⁴Cu]PARPi-DOTA (59), all shared structural similarity to olaparib, and most displayed high specificity toward PARP. Also of note is the development of [¹⁸F]olaparib, an ¹⁸F-labeled isotopologue of olaparib, with an identical chemical structure as the parent molecule, hence with identical pharmacokinetic and pharmacodynamic profiles in vivo (60). Among the olaparib-derived radiotracers, [¹⁸F]PARPi-FL, the dual-modality PARP imaging agent (optical/PET), was used to image drug distribution and drug-target engagement of different PARPis in different cancer models (61, 62), demonstrating the potential application of radiolabeled-PARPi in monitoring of drug distribution and drug-target engagement. Rucaparib-inspired radiolabeled compounds [¹⁸F]FE-LS-75 (without publicly available preclinical characterization; ref. 63), [¹²⁵I]KX1 (64), [¹²⁵I]KX-02–019 (65), and [¹⁸F]FTT (66) also show excellent PARP selectivity.

Moving beyond preclinical models, both [¹⁸F]FTT (66, 67) and [¹⁸F]PARPi (55, 68) have been evaluated in different types of cancers in clinical practice. [¹⁸F]FTT is currently the subject of phase I trials in patients with ovarian cancer (Fig 2B; ref. 66), and also able to measure PARP-1 expression in different subtypes of patients with breast cancer (estrogen receptor positive, HER2⁺ and triple negative; ref. 67). [¹⁸F]PARPi is currently also in phase I trials but for head and neck cancer (Fig. 2C; ref. 55). Furthermore, recent clinical data have shown that the utility of [¹⁸F]PARPi can be potentially expanded to image brain cancer in patients for treatment monitoring (68). In addition to applying radiolabeled-PARPi in patients for cancer detection, a fluorescent-labeled PARPi, PARPi-FL, was also subjected to clinical studies for the detection of oral carcinoma, and potentially serve as the surgical-margin assessment of epithelial cancer of the upper intestinal tract (69). These translational studies established the encouraging potential of PARP imaging agents in cancer detection for clinical practice.

Although these agents showed promising tumor uptake, they display different affinity and selectivity for PARP-1 and other isoforms, with a range of IC₅₀ values (2.3–200 nmol/L) as a proxy for PARP affinity (Fig. 2A). Their selectivity toward other PARP family enzymes is not always fully characterized. Combined results indicate that small structural modifications may have significant impact on the target specificity of the compound. The Reiner group showed, for example, that altering the iodination position on the aromatic ring and the length of the linker between the iodinated-aromatic ring and the olaparib core resulted in very different IC₅₀ values for PARP-1 (57). Similarly, recently reported on the effects of structure modification on tumor uptake and target specificity of the olaparib-derived radiolabeled PARPi [¹⁸F]olaparib and [¹⁸F]AZD2461 (70).

The great potential of radiolabeled PARPi has now also expanded to the use of therapeutic radionuclides: ¹²³I (Auger electron therapy; ref. 71), ¹³¹I (high energy beta particles; ref. 72), and ²¹¹At (alpha radiation; ref. 73). These PARPi-based radionuclide therapy agents have shown some efficacy in brain cancer and neuroblastoma models with little toxicity to normal tissues, making use of the high ionizing potential and short ranges of these particle emitters for cancer treatment (Fig. 3A–C).

To date, PARPi-derived radiolabeled compounds are based on either olaparib-like structures or rucaparib-like structures. Given that small variations in structure can lead to large differences in PARP IC₅₀ values and trapping, it may not be assumed that all radiolabeled variants are similar in PARP trapping ability, or elicit the same allosteric effects under the same circumstances. Olaparib and rucaparib have a comparable ability for trapping PARP-1 on DNA via catalytic inhibition, yet their allosteric effects differ significantly, and may even be different for other PARP isoforms. These variances may result in diverse degrees of target engagement and PARP trapping, including PARPi trapping. This may in turn influence tumor versus normal tissue uptake, affecting imaging contrast and toxicity, of radionuclide imaging and therapy. Therefore, target specificity and PARP trapping ability of PARPi or their radiolabeled analog, and target expression should be taken into account in PARP-targeting approaches. For example, high accumulation of ¹⁸F-PARPi in healthy tissues such as lymph nodes, bone marrow, liver, and spleen may induce higher toxicity to these organs due to high physiologic expression of PARP-1 and PARP-2 (55). Careful consideration is needed in the interpretation of PARPi-derived imaging agents, which cannot necessarily predict the precise behavior of analogous, but nonidentical radionuclide therapy compounds. This can of course be overcome by using thera(g)nostic compounds, which allow imaging patient dosimetry calculation and therapy response, as demonstrated by the recent study using [¹²³I]PARPi in glioblastoma models (Fig. 3A; ref. 71).

The discovery and development of PARPi in BRCA-mutant cancers is arguably the most elegant example of synthetic lethality to target tumor suppression gene loss in clinical practice. PARPi therapy can be of considerable benefit to a wide variety of patients with cancer as a monotherapy and can synergize with appropriately selected drugs to improve treatment outcome. However, before the full potential of mono- and combination therapies may be achieved in clinical practice, it is imperative to recognize the limitations currently faced in the PARP inhibition in cancer therapy. Attention should be paid to identification and validation of predictive biomarkers to predict synthetic lethality, how distinct molecular mechanisms of action of different PARPi, off-target effects due to high physiologic PARP-expressing healthy organs, tumor heterogeneity, and genetic modifications in patients may lead to unexpected treatment responses. Genetic testing, for example, BRACAnalysis or myChoice test (Myriad Genetics) to select those patients who will benefit from PARPi treatment has been implemented in many clinical trials, and the use of diagnostic PARPi-based PET imaging may serve as a noninvasive method to monitor patient treatment response or study the degree of drug engagement in tumor (based on PARP-1 expression levels) at any time during the treatment cycle. Proof-of-concept studies using imaging agents to study PARP levels will add value to the interpretation of clinical data. The emergence of PARP-targeting radionuclide therapy represents another treatment strategy that warrants further investigation to explore its potential.

C.Y. Chan reports grants from Pancreatic Cancer UK and CRUK through the Oxford Institute for Radiation Oncology during the conduct of the study. K.V. Tan reports grants from University of Hong Kong during the conduct of the study. B. Cornelissen reports grants from Pancreatic Cancer UK, CRUK, and other from Theragnostics Ltd during the conduct of the study; in addition, B. Cornelissen has a patent licensed (application number 1804924.7).

This work was supported by Pancreatic Cancer U.K.; CRUK through the Oxford Institute for Radiation Oncology (for B. Cornelissen), and in part by Seed Fund for Basic Research of University of Hong Kong (for K.V. Tan and C.Y. Chan).

Because of space limitations, we regret that many primary and historical published works on the topic have not been cited in this review.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.