Laying a trap to kill cancer cells: PARP inhibitors and their mechanisms of action

See allHide authors and affiliations

Science Translational Medicine  26 Oct 2016:
Vol. 8, Issue 362, pp. 362ps17
DOI: 10.1126/scitranslmed.aaf9246


Poly(ADP-ribose) polymerase (PARP) inhibitors are the first DNA damage response targeted agents approved for cancer therapy. Here, we focus on their molecular mechanism of action by PARP “trapping” and what this means for both clinical monotherapy and combination with chemotherapeutic agents.


This year marks the 50-year anniversary of the discovery of the poly(ADP-ribose) polymer (PAR) by Chambon et al. (1). PARylation was the first described posttranslational modification associated with DNA damage and repair and chromatin remodeling (2). The discovery of PAR preceded by 1 year that of the first PAR polymerase, PARP1 (3). PARP1 belongs to a family of 17 enzymes with a common catalytic ADP-ribosyltransferase (ART) motif (46), which are referred to as ARTDs (diphtheria-toxin-like ADP-ribosyltransferases). Only four ARTDs generate PAR chains: PARP1 (ARTD1), PARP2 (ARTD2), and the two tankyrases, PARP5a (ARTD5) and PARP5b (ARTD6). Because the other ARTDs add only a single ADP-ribose, the two reactions are referred to as PARylation and MARylation for poly- versus mono-ADP ribosylation, respectively. Multiple reviews have been published over the past 2 years on PARP biology and inhibitors (513), but the present article focuses on PARP trapping and its relevance for the anticancer activity of PARP inhibitors, both as a monotherapy and in combination with DNA-damaging chemotherapy.


PARP1 and/or PARP2 are required to repair DNA single-strand breaks (SSBs). PARP1 also repairs DNA double-strand breaks (DSBs) and replication fork damage (Fig. 1) (1418). PARP1 and PARP2 have overlapping functions but may differ in their substrate preference (19). In mice, the single knockouts of PARP1 or PARP2 are viable, whereas the double knockout of PARP1 and PARP2 is lethal, and double knockout cells exhibit genomic instability (20). All clinical PARP inhibitors inhibit both PARP1 and PARP2, highlighting the difference between pharmacological inhibition and a deficiency of both proteins during embryonic development. PARP2 is also essential in hematopoietic stem/progenitor cell survival (21), whereas PARP3 acts in DSB repair by recruiting aprataxin-like factor (APLF) and scaffolding the DSB end-joining complexes (22). PARP3 does not appear to contribute to the antitumor activity of PARP inhibitors, because a compound (AZD2461) that inhibits PARP1 and PARP2 but not PARP3 is still effective at causing BRCA-mutated tumor regression in preclinical models (23).

Fig. 1. DNA repair by PARP1 and the effects of PARP inhibitors.

Upon the generation of an SSB, PARP1 binds to the break (A) and uses NAD+ (B) to generate PAR polymers on itself (auto-PARylation), as well as on histones and chromatin-associated proteins. This serves the purpose of relaxing chromatin and recruiting repair proteins. Cumulative auto-PARylation causes the dissociation of PARP1 from DNA (C), allowing access to other repair factors scaffolded by XRCC1 (D). PARylation is removed by PARG (E), a glycohydrolase, which allows PARP1 reactivation. PARP inhibitors block NAD+ binding and PARylation for as long as the inhibitor is bound to the NAD+ site (B), thereby preventing PARP dissociation from the SSB, resulting in both accumulation of unrepaired SSBs (F) and PARP trapping (G). Repairing the ensuing DSB and PARP trapping will require BRCA1, BRCA2, and other HRR factors, as well as ATM, Fanconi, and replication bypass pathways for cell survival (H). PARP1 is also involved in the repair of “collapsed forks” with DSEs (I), in the retraction and restart of stalled replication forks (J), and in the repair of DSBs (M). PARP inhibitors trap PARP at DSEs (K and L) and DSBs (N).


PARP1 is highly expressed and generates the majority of PAR polymers, which can be readily detected in the absence of exogenous DNA damage (24), although the process of lysing cells can artificially stimulate PARP1 (25). As one of the most ubiquitous posttranslational modifications, PARylation is recognized and processed by different macromolecular complexes related to DNA repair and epigenetic pathways, with specialized “writers” (the PARPs), “readers” that recognize specific modules within the PAR polymers, and “erasers” [poly(ADP-ribose) glycohydrolase (PARG); Fig. 1E] (6, 26). DNA-bound and activated PARP1 and PARP2 both add long PAR chains to a variety of nuclear target proteins, including histones and PARP itself (auto-PARylation) (Fig. 1B). Binding of PARP1 to nucleosomes, and PARylation, can be detected at promoters of actively transcribed genes and enhancers together with PARG, consistent with the normally transient nature of PARylation (5, 6). One proposed mechanism for recruiting PARP1 to active promoters is by the formation of DNA breaks by topoisomerase IIβ (27, 28). PARP1 also associates with heterochromatin to silence ribosomal DNA (rDNA) (29) and is a cofactor of the chromatin insulator CTCF (CCCTC-binding factor) (30, 31).

Tens of thousands of endogenous SSBs form transiently every day, and PARP1 and PARP2 are critical for their repair. In addition, SSBs are produced by the chemotherapy drugs temozolomide and camptothecins, which explains the synergy between temozolomide or camptothecins and PARP inhibitors, which was observed early on during the development of PARP inhibitors (9, 10). The broad DNA repair functions of PARP1 and PARylation are outlined in Fig. 1, including SSB repair (Fig. 1, A to D), replication damage repair (Fig. 1, I to L), and DSB repair (Fig. 1, M and N). In addition to classical DSBs, such as those produced by ionizing radiation, radiomimetic drugs, and endogenous recombination reactions (Fig. 1M), a DSB can occur when a replication fork encounters an SSB, thereby generating a single-ended DSB (DSE; Fig. 1F). Such breaks can also be actively induced to resolve blocked replication forks (“collapsed forks”) by the action of endonucleases (Fig. 1I) (32). PARP1 and PARylation are also implicated in replication fork retraction and restart, which enables the repair of the DNA template (Fig. 1J).


The clinical utility of PARP inhibitors as monotherapy is based on the concept of synthetic lethality (33), where neither PARP inhibition alone nor BRCA deficiency alone is lethal but the combination is. Initial data in vitro (34, 35) were followed by in vivo studies in BRCA knockout mice (36, 37) and then by clinical validation (38) with olaparib in patients with hereditary germline BRCA mutations. Initial thinking was that unrepaired SSBs resulting from PARP inhibition would be converted into the more genotoxic DSEs in replicating cells (Fig. 1F). Whereas these DSEs could be repaired effectively in normal cells with functional homologous recombination repair (HRR) (Fig. 1G), cancer cells with homologous recombination repair deficiencies (HRDs), such as BRCA loss-of-function mutations, would only have low-fidelity repair pathways to fall back on, resulting in an increase in genomic instability, ultimately leading to tumor-specific cell death (39).


The trapping of PARP on DNA was invoked to explain why the cytotoxicity of PARP inhibitors was markedly greater than that of not having PARP1 at all and why the cytotoxicity of PARP inhibitors was eliminated by removing PARP1 genetically (40, 41). Hence, the cytotoxicity of PARP inhibitors had to be mediated by the presence of PARP1 itself. PARP trapping was demonstrated by detecting PARP-DNA complexes in cells treated with PARP inhibitors (40, 41). Mechanistically, PARP trapping is not independent of catalytic inhibition of PARylation (Fig. 1, A and D) but rather is linked to it. PARP1 and PARP2 are effectively trapped and will not come off the DNA until the inhibitor dissociates from the active site, allowing the utilization of NAD+ (nicotinamide adenine dinucleotide) and auto-PARylation. Because PARP inhibitors with equal catalytic inhibition potency show markedly different PARP trapping ability, direct interactions of the drugs with the PARP NAD+ binding site have been invoked (40, 41). The molecular details underlying the tight binding (slow off-rate) of PARP inhibitors and the proposed allosteric effect of PARP inhibitors (40) are only partially understood because these drugs have not been cocrystallized with full-length PARP1 (or PARP2) bound to damaged DNA.


The single agent cytotoxicity of different PARP inhibitors does not correlate with their ability to inhibit PARylation (11, 40, 42). Hence, distinctions must be made between inhibitors based on their trapping potency, at least when considering single agent activity. Still, it is important to stress that all PARP inhibitors currently in the clinic are catalytic inhibitors, and where they vary is in their effectiveness in trapping PARP onto DNA.

Clinical PARP inhibitors can be ranked by their ability to trap PARP (from the most to the least potent): talazoparib >> niraparib > olaparib = rucaparib >> veliparib (40, 41). This ranking parallels the cytotoxic potency of the drugs as single agents, with veliparib being inactive at 100 μM in most cancer cell lines (40) and, at the other end, talazoparib being active at nanomolar concentrations (41). The chemical structures of PARP inhibitors (Fig. 2) reveal differences in overall size and flexibility. The smallest drug is veliparib (molecular mass = 244 versus 434 for olaparib), and the most rigid is talazoparib, with two racemic centers (Fig. 2) that are critical for inhibiting PARP (41, 43). Thus, the stereospecific and larger structure of talazoparib is likely to account for its potent trapping activity against PARP1 and PARP2. The most rigid and bulky drugs have the greatest propensity to exhibit low off-rate and allosteric effects.

Fig. 2. Structures of the clinical PARP inhibitors.

The nicotinamide moiety common to PARP inhibitors is shown in red. The blue arrows indicate the racemic centers for talazoparib, explaining the selectivity of the active enantiomer.



PARP trapping is markedly more deleterious than persistent SSBs in the absence of PARP (Fig. 1F) (40, 44). In addition, because trapped PARP1 and PARP2 are likely associated with other proteins (45, 46), large complexes are potentially more damaging and difficult to remove than PARP1 (or PARP2) alone.

PARP inhibitors have activity in tumors beyond those with BRCA loss of function, including deficiencies in repair proteins commonly defective in cancer cells such as ataxia telangiectasia mutated (ATM) (40, 47). It is therefore vital to establish how trapped PARP damages the genome and which pathways and proteins repair trapped PARP to aid in patient selection. To date, a broad range of factors affect the efficacy of PARP inhibitors, including NAD+ metabolism and core replication factors, DNA template switching proteins, and chromatin-remodeling proteins (40, 48, 49), implying that replication interference is a major cytotoxic mechanism for PARP inhibitors. One possibility is that ongoing replication forks collide with trapped PARP-DNA complexes and that, once collisions have occurred, stalled replication forks generate DSEs, commonly referred to as “replication fork collapse” (Fig. 1I) (7). This mechanism is reminiscent of the DNA damage generated by the trapping of topoisomerase I (TOP1) cleavage complexes (TOP1ccs) by TOP1 inhibitors (Fig. 3) (5052). In the context of PARP inhibitors, defective PARylation might also favor DSEs by blocking replication fork reversal (Fig. 1J) (18, 52). The availability of potent PARP trappers, such as talazoparib, and of assays to measure PARP trapping should allow the dissection of the lethal lesions induced by PARP trapping and of the associated repair pathways. If such pathways correspond to tumor suppressor genes whose function is lost in tumors, then additional synthetic lethal therapeutic approaches may be discovered (33).

Fig. 3. Synergistic DDR deficiencies for the combination of PARP inhibitors with TOP1 inhibitors.

TOP1ccs activate PARP1, which is recruited with TDP1 to remove the covalent attachment of TOP1 to the DNA 3′-end (A). An alternative pathway (B) uses endonucleases, which are commonly mutated in colon and lung cancers. Hence, cancers deficient for these endonucleases should be selectively responsive to the combination of PARP and TOP1 inhibitors.



A clinical consequence of PARP trapping is that the antitumor effect of these drugs requires PARP1: the more PARP1, the greater the opportunity for PARP trapping, and conversely, as described above, one of the resistance mechanisms is down-regulation of PARP1 (40, 48). Another consideration is that PARP inhibitors are substrates for drug efflux pumps such as MDR1/P-glycoprotein-1 (encoded by ABCB1). Both olaparib and rucaparib are MDR1 substrates (36, 5355), making drug efflux pump activation a potential mechanism of intrinsic and acquired drug resistance. Amplification and/or transcriptional hyperactivation of ABC drug transporters are therefore potentially testable determinants of resistance to clinical PARP inhibitors.

HRD is a strong determinant of response to PARP inhibitors, and reactivation of HRR has been observed in BRCA-deficient tumors that become refractory to PARP inhibitors. The clinical relevance of reactivating mutations of BRCA1 (56), up-regulation of the DNA end resection protein CTIP (57) to reactivate HRR, and inactivation of 53BP1 (58) to suppress nonhomologous end-joining (NHEJ), which competes with HRR, still needs to be established.


The two most synergistic combinations between PARP inhibitors and anticancer drugs are with the DNA alkylating agent temozolomide (9, 59) and with TOP1 inhibitors (10, 51, 60). Temozolomide acts by methylating guanines at positions O6 and N7 and adenines at position N3 (61). Although O6 methylation is the main mechanism of action for temozolomide in glioblastomas, which commonly lack MGMT (O6-guanine-methyl-transferase), N7 guanine methylation is the most prevalent lesion and is efficiently dealt with by PARP1-mediated repair (62). In the absence of PARP1, and in tumors with normal MGMT, temozolomide produces SSBs at N7-methylated guanines and N3-methylated adenines. Those SSBs recruit PARP1, which is susceptible to being trapped (see Fig. 1). The synergy between temozolomide and PARP inhibitors is based on PARP trapping in addition to PARP catalytic inhibition. As a result, talazoparib and olaparib are markedly more efficient at killing temozolomide-treated cells than veliparib or PARP1/2 genetic inactivation (59). Also, PARP trapping can be readily detected in cells treated with temozolomide and PARP inhibitors (59).

For the TOP1 inhibitors, including the widely used camptothecin derivatives, irinotecan and topotecan (10, 51, 60), synergy with PARP inhibitors is solely due to PARP catalytic inhibition and does not require PARP trapping. Indeed, PARP inhibitors enhance cell killing no further than the enhanced killing observed in PARP1/2 knockout cells (59), and PARP trapping is undetectable in cells treated with camptothecin in combination with olaparib (59). Three mechanisms account for the synergy: (i) inhibition of HRR and stimulation of NHEJ by PARP1 inactivation (63); (ii) inhibition of TOP1 PARylation, which otherwise promotes the reversal of the trapped TOP1cc (64, 65); and (iii) inhibition of tyrosyl-DNA-phosphodiesterase 1 (TDP1), the repair enzyme that excises the covalently bound TOP1 (Fig. 3) (66). The dominant role of PARP1 as a normal cofactor for TDP1 activity was recently confirmed by chemical screening for TDP1 inhibitors (42).


The cytotoxicity associated with the DNA trapping described here for the PARP inhibitors extends to the mechanism of action of at least three other important classes of anticancer drugs and targets: TOP1 (discussed above), type II topoisomerases (TOP2α and TOP2β), and DNA methyltransferase (DNMT). In all cases, the cytotoxic mechanism of the inhibitors is related not only to catalytic inhibition of the target enzymes but also to their trapping on DNA as DNA-protein complexes. The lethality of such complexes is born from the highly dynamic nature of chromatin during transcription, replication, genome segregation, and DNA repair. Stalling of large protein complexes on DNA interferes with such processes and thereby produces severe genomic lesions that tend to be less efficiently repaired in cancer cells with inactivating mutations in tumor suppressor genes.


The initial clinical development of PARP inhibitors was based on potentiating tumor cell kill by DNA-damaging agents; this included alkylating agents, topoisomerase inhibitors, and ionizing radiation. The first clinical trials focused on evaluating the tolerability of such combinations and ascertaining that the PARP inhibitor chemosensitization observed preclinically could be recapitulated in patients.

First PARP inhibitor phase 1 trial—Combination chemosensitization studies

The first PARP inhibitor evaluated in patients was rucaparib, administered to potentiate the antitumor activity of temozolomide (67). This trial used a single agent PARP inhibitor run-in period and demonstrated proof of mechanism with PARP blockade in peripheral blood lymphocytes. Although the initial phase 1 study suggested that full-dose temozolomide could be administered with low-dose rucaparib, a further phase 2 evaluation of this combination in patients suffering from metastatic melanoma revealed that myelosuppression was a major challenge to the administration of this combination (68). This has been recapitulated by other cytotoxic and PARP inhibitor combination trials; overall, these studies indicate that administering tolerable doses of these combinations requires truncated PARP inhibitor schedules. Radiation sensitization by PARP inhibitors may also face similar challenges and may require administering lower radiation doses or lower PARP inhibitor doses with truncated administration schedules.

Phase 1 single agent PARP inhibitor clinical trials

The study of continuous PARP inhibitor patient dosing followed the documentation that PARP inhibition was synthetic lethal with loss of BRCA1 or BRCA2 function. This led to the successful first evaluation of the antitumor activity of single agent olaparib (38, 69). Similar antitumor activity has been documented with other inhibitors that trap PARP, including niraparib (70), rucaparib (71), and talazoparib (72). The first-in-human olaparib trial demonstrated satisfactory pharmacokinetic-pharmacodynamic properties and proof of mechanism as well as proof of concept, showing inhibition of PAR formation and single agent durable antitumor activity in BRCA carrier patients suffering from breast, ovarian, orprostate cancer. These initial trials also documented that resistance to PARP inhibition in BRCA carrier ovarian cancer correlated with clinical measures of platinum sensitivity, with platinum-sensitive ovarian cancer being most sensitive to olaparib and platinum-refractory ovarian cancer being almost refractory; similar observations have since been made for other PARP inhibitors. Limited activity of PARP inhibitors in heavily pretreated patients may extend to tumor types other than ovarian cancer, including triple-negative breast cancer and Ewing’s sarcoma, and may at least in part be related to prior DNA-damaging agent exposure (73) and the development of resistance mechanisms that subsequently affect PARP inhibitor activity. Therefore, both platinum sensitivity status and prior lines of therapy have the potential to limit PARP inhibitor single agent activity and may require reevaluating molecular tumor profiles at relapse.

Relationship between PARP trapping and clinical maximum tolerated dose

The insights into PARP inhibitor mechanism of action and the trapping of PARP beg the question whether the greater potency of PARP trapping translates into greater clinical antitumor activity as a monotherapy. Although a direct comparison between different PARP inhibitors is difficult, emerging data suggest that there is no such simple correlation.

Although greater PARP1 trapping ability may be associated with greater preclinical potency, in the clinic, it is also associated with greater toxicity to normal tissue (11). For example, although talazoparib is about two to three orders of magnitude more effective at trapping PARP than niraparib, rucaparib, and olaparib; in the clinic, the recommended phase 2 dose (RP2D—effectively the maximum tolerated dose) of talazoparib is 300 to 1200 times lower than for these other PARP inhibitors. In all likelihood, this increased toxicity of talazoparib reflects the greater potency of PARP1 trapping in normal tissue.

When comparing the clinical activity of different PARP inhibitors, the challenge is comparing like with like. For example, recent data on rucaparib highlight a RECIST (response evaluation criteria in solid tumors) response rate in the phase 2 ARIEL trial of 69% partial or complete response (PR/CR) in BRCA-mutated ovarian cancer patients (74). This can be compared with 44% for talazoparib (72). Both data sets were from BRCA-mutated ovarian cancer patients. However, in the case of rucaparib, all patients only had relatively few lines of previous therapy, whereas for talazoparib, there was a mixture of platinum-resistant diseases as well as patients having had more lines of therapy, both affecting PARP inhibitor sensitivity. Talazoparib RECIST data based only on the platinum-sensitive tumors increased to a 50% PR/CR rate. Similarly, comparative data in platinum-sensitive tumors showing 50% PR/CR for niraparib (70) and 61.5% for olaparib (69) (where, again, patients had on average had more lines of therapy than those in the ARIEL 2 trial), suggest similar clinical activity when prior lines of therapy and tumors’ platinum sensitivity status are taken into account. Only data for veliparib, with a PR/CR rate of 35% in platinum-sensitive BRCA-mutated ovarian patients (75), suggest that the much weaker PARP trapping may affect PARP inhibitor monotherapy activity. This may become clearer once larger phase 3 clinical trials are completed. Overall, however, there appears to be little major difference in the clinical effectiveness of the majority of the PARP inhibitors to date (13), but more data are required.

Synthetic lethality beyond BRCA

Recent studies have now shown that PARP inhibition with olaparib also has clinically meaningful antitumor activity in metastatic prostate cancers, with this being almost exclusively reported in tumors with genomic aberrations of HRR genes, including BRCA2, ATM, BRCA1, and PALB2 (76). This work has resulted in olaparib being given breakthrough designation by the FDA for BRCA2 and ATM aberrant advanced prostate cancers. In contrast, most prostate cancer patients without a DNA repair defect did not respond to olaparib (74). Rucaparib, niraparib, talazoparib, and veliparib are all being explored as potential monotherapy beyond BRCA and/or beyond ovarian cancer, based on the potential to exploit genetic backgrounds where HRR-associated deficiencies mean that synthetic lethality can result from PARP trapping.


It has taken 50 years from the discovery of PARP1 to the approval of the first medicine based on the inhibition of this enzyme. PARP inhibition is now a standard of care for high-grade serous ovarian cancers with BRCA mutations. It is also likely that PARP inhibition will be clinically relevant in other diseases, including subpopulations of prostate and breast cancers. The importance of understanding the mechanism of action of PARP inhibitors, especially PARP trapping, and its contribution toward regulatory approval should not be underestimated. No doubt, there is still more to learn from the mechanisms of PARP inhibitors. Although PARP trapping may be central to the mechanism of action of these inhibitors, it is also clear that PARP plays multiple roles in DNA repair, including the repair of the DSBs, and that inhibiting PARP during this process may also be an important component of cancer cell killing (15, 77). Insights into drug mechanisms of action are not just something that is nice to have; they are fundamental for determining how a drug should be given to a patient in terms of dose and schedule, for patient selection, and for understanding mechanisms of resistance.


  1. Acknowledgments: We wish to thank J. Murai for suggestions and critical reading of this manuscript and for her invaluable contribution to the discovery of PARP trapping. Our studies are supported by the Center for Cancer Research and the Intramural Program of the National Cancer Institute (BC 006150). Funding: J.d.B. acknowledges support from the Institute of Cancer Research/Royal Marsden Drug Development Unit through an Experimental Cancer Medical Centre (ECMC) grant from Cancer Research U.K. and the Department of Health (C51/A7401) and NHS funding to the National Institute for Health Research Biomedical Research Centre to the Royal Marsden. Competing interests: The Institute of Cancer Research is a not-for-profit research organization that has a commercial interest in PARP inhibition in DNA repair–defective cancers. J.d.B. has served as an advisory board member for AstraZeneca, Medivation, Merck, Pfizer, and Tesoro and received honoraria from AstraZeneca. M.J.O. is a full-time employee of AstraZeneca and owns stock in the company.

Stay Connected to Science Translational Medicine

Navigate This Article