PARP inhibitors in metastatic prostate cancer: When, who, and how?

Types of DNA repair mechanisms

A plethora of endogenous/exogenous agents lead to DNA damage which is repaired, mainly by the following mechanisms: Nucleotide excision repair, Base excision repair (BER), Mismatch repair, homologous repair (HR), non-homologous end joining (NHEJ; Classical and alternate), and finally the Fanconi pathway.^[2] Impairment in these pathways may lead to genomic instability, a hallmark of cancer.^[3] Consequently, defects in these mechanisms lead to a multitude of inherited cancer syndromes.^[4] The tumor microenvironment may potentiate these DNA repair defects through local hypoxemia, hypoglycemia, epigenetic modification, and other mechanisms.^[2] In this complex conundrum of interactions, two pathways deserve special mention when dealing with prostatic adenocarcinoma: BER and homologous recombination repair (HRR).

BER corrects lesions that typically do not alter the DNA helix structure.^[5] These lesions (deamidation, oxidation, and methylation) are typically a manifestation of spontaneous decay of DNA, a process accelerated by reactive oxygen species generation in aerobic organisms.^[6] DNA glycosylases, enzymes that cleave bonds between the deoxyribose backbone and the mismatched DNA base, are the key effectors. Each specific lesion covered in the BER spectrum is recognized by a specific DNA glycosylase (although many have overlapping specificities) leading to the formation of an abasic site (AP), which is subsequently repaired through the sequential activity of an AP-endonuclease, exonuclease, and a DNA polymerase as well as ligase.^[7]

Double-strand breaks may be repaired by homologous recombination or non-homologous end-joining.^[8,9] Homologous recombination utilizes the complementary sequence on the sister chromatid, leading to high fidelity, but a dependence on the G2 and M phases of the cell cycle (when DNA is condensed into chromosomes). NHEJ shows lesser fidelity but is available throughout the cell cycle.^[10]

PARP: Functions in DNA damage repair

PARPs are a family of 17 proteins with varied functions including stress response, chromatin remodeling, DNA repair, and apoptosis. Its most well recognized member, PARP1, was initially identified as a part of BER with subsequent evidence hinting at the utility in the other aforementioned DNA repair mechanisms as well.^[11] It is now known that PARP1 is pivotal for single-strand break (SSB) repair and has a role in BER since an SSB is introduced as an intermediate step.^[12]

PARP1 detects DNA damage through zinc finger motifs^[13] and subsequently leads to the polymerization of ADP-ribose units from NAD+ molecules onto acidic residues.^[14] AutoPARylation leads to enhanced catalytic activity and causes PARylation of histones and chromatin associated proteins as well.^[15] This PARylated scaffold attracts DNA repair molecules such as x-ray repair cross-complementing protein 1 (XRCC1) to the site.^[16] Persistent PARP1 activation leads to depletion of nicotinamide adenine dinucleotide (NADH+) based energy sources intracellularly and may lead to cell death. PARP2 and PARP3 have similar functions with some redundancy, but both are essential for cell survival.^[15]

Into the precision medicine era

The advent of precision medicine in carcinoma prostate stems from the whole exome and transcriptome analysis of fresh metastatic site tissue reported by Robinson et al.^[23] Prior studies had focused on the specimen from primary prostate cancer and elucidated somatic mutations, copy number alterations, and chromoplexy involving, mainly the speckle - type POZ protein (SPOP), forkhead box A1 (FOXA1), and the TP53 gene along with abundant E26-transformed specific (ETS) translocations.^[24] While some of these alterations were prognostic (Copy number variations and evidence of tumor hypoxia could predict relapses), none were targetable.^[25,26] Mutational landscape of metastatic disease was less well known given the technical difficulties in sequencing decalcified biopsies (since bone is the most common metastatic site).^[27,28] The few studies focusing on this question did demonstrate mutations in the AR gene, as well as the androgen signaling pathway. However, these studies were based on a small number of autopsy specimens and pre-clinical models and utilized targeted sequencing, impacting both the internal and external validity of these findings.^[29,30] The lack of characterization of mutational landscape in metastatic settings combined with hardly any targetable mutations in primary settings made carcinoma prostate a rather unattractive choice for targeted drug discovery.

Targetable mutations, the holy grail of personalized medicine, are best evaluated by whole-exome/transcriptome sequencing done prospectively on specimens of clinical interest (therapy naïve vs. after failure). The study by Robertson et al. evaluated 150 such biopsies and reported that, nearly, all cases had the presence of a pre-established biological driver mutation. About 63% of patients had a mutation in the AR pathway, showing the continued dependence on AR signaling in castration-resistant diseases. Even after excluding all AR-pathway alterations, 65% of patients had a “clinically actionable mutation” involving the phosphoinositide 3-kinases (PI3K) pathway (49%), DNA repair pathways (19%), rapidly accelerated fibrosarcoma (RAF) kinases (3%), cyclin dependent kinase (CDK) inhibitors (7%), and the wingless-related integration site (WNT) pathway (5%). In addition, 8% of cases had a pathological germline variant as well.^[23]

The mutations detected in prostate cancer specimens have therapeutic implications hinting at the potential for personalized medicine in metastatic prostate cancer. A report of exceptional responders to platinum based chemotherapy reported deleterious HRR gene mutations in all three reported cases.^[31] Another study evaluating blanket Olaparib therapy revealed similar enrichment in responders.^[32] Thus, DNA repair pathway defects are targetable in carcinoma prostate, and studies further characterizing these lesions are detailed below.

DNA repair pathways defects are early drivers

To assess better the type, nature, and frequency of DNA repair pathway mutations, a study evaluated 504 tumor specimens from 451 patients through targeted sequencing.^[33] The frequency of mutations in the PI3K, RAF kinase family, and the WNT pathway was similar to prior reports.^[23] In addition, new insights regarding HRR machinery in carcinoma prostate were evident. First, around 22% of all cases had a somatic mutation in the gene involved in HRR, with mutations in the BRCA-2 gene being the most common (9%). Second, among patients undergoing germline testing, 19% had a germline pathogenic mutation in an HRR gene. Third, nearly 27% of patients had some HRR gene alteration when cotested with germline and somatic assays. Fourth, germline analysis alone accounted for only half of these patients. Fifth, although many gene mutations showed enrichment with progressive stages (AR, retinoblastoma 1 (RB1), phosphatase and tensin homolog (PTEN), and ataxia telangiectasia mutated (ATM)), the prevalence of mutations in TP53 and BRCA2 was relatively uniform.^[33] Taken together, these findings imply that nearly 1/3^rd of patients have targetable HRR mutations which are better detected by cotesting on somatic as well as germline assays. Furthermore, BRCA2 mutations seem to be early oncogenic drivers of disease.

Germline predisposition in carcinoma prostate

Findings of the aforementioned studies indicate a significant impact of germline alterations on the risk and natural history of the disease. The heterogenous pace of progression is well documented in carcinoma prostate.^[34] Furthermore, inherited susceptibility accounts for more than half of the relative risk of carcinoma prostate.^[35] Studies have shown that patients with HRR gene mutations have a higher propensity for distant spread and recurrence.^[36] Given these findings, germline HRR gene mutations may show enrichment in more aggressive diseases. This is supported by incremental rates of BRCA mutations in the normal population (<1%), patients with high risk localized carcinoma prostate (5.3%), and patients with metastatic disease at diagnosis (19%).^[33,37,38]

Along these lines, a study evaluated 692 samples of patients with metastatic carcinoma prostate at diagnosis (unselected by family history) and reported the following findings. First, 11.8% of all carcinoma prostate cases had a germline alteration involving an HRR gene (Most common being BRCA2: 44% of all). Second, the prevalence of carcinoma prostate in a first-degree relative did not vary with germline mutation status. However, the presence of an underlying germline mutation was associated with a higher Gleason score, as well as a higher risk of other cancers in first-degree relatives (Breast, ovarian, pancreatic, endometrial, gastrointestinal, and leukemia).^[39]

To summarize, advanced carcinoma prostate shows enrichment in targetable HRR gene mutations and cotesting through somatic and germline platforms are likely to maximize yield. This enrichment likely stems from a propensity to more aggressive disease rather than in response to prior lines of therapy. Finally, these mutations are actionable, providing us with a novel therapeutic option different from the conventional choices of taxanes and androgen axis-based therapies.

Olaparib

The TOPARP-A study was the initial phase two study which enrolled all metastatic CRPC patients regardless of the underlying mutational profile. While the overall efficacy was unimpressive, patients with a positive biomarker (i.e., HRD mutation) showed impressive ORR and PFS despite being used in a heavily pre-treated population.^[32] Given the likely futility of PARPi in non-HRD population, the subsequent TOPARP-B study compared 300 mg and 400 mg doses of Olaparib in phase two settings and documented no difference in efficacy (ORR, CR, PSA50, CTC reduction, and PFS). Notably, some non-BRCA HRD mutations showed slightly better responses at higher dates, but, further, data are needed before this is applied clinically.^[41] Olaparib approval in HRD metastatic CRPC mainly stems from the PROfound study. Evaluating Olaparib versus physician’s choice in phase three settings, this study documented superior ORR, CR, PFS, and OS with targeted agents in patients progressing on at least one prior AR-directed therapy (taxanes agnostic design) in patients with mutations in BRCA1, BRCA2, or ATM.^[42] While these may suggest Olaparib as the preferred option in these settings, the following merit consideration. First, there is no basis for the inclusion of ATM mutated patients in the primary efficacy cohort (covered below). Second, options in the physician’s choice arm were limited to abiraterone or enzalutamide, leading to a weak comparator arm given the sequential use of AR-directed therapy. A subsequent analysis of the trial reported improvement in PFS as well as OS in the primary efficacy arm.^[43] Based on these findings, Olaparib therapy is recommended in patients with mutations in BRCA1, BRCA2, or ATM genes after the failure of enzalutamide or abiraterone therapy.

Rucaparib

The TRITON2 study evaluated rucaparib monotherapy at a dose of 600 mg BD in 28-day cycles in patients with metastatic CRPC and prior exposure to at least one AR directed, and a taxane. Patients needed to have somatic or germline mutations in any of the following HRD mutations: BRCA1, BRCA2, ATM, BARD1, BRIP1, CDK12, CHEK2, FANCA, NBN, PALB2, RAD51, RAD51B, RAD51C, RAD51D, or RAD54L. The BRCA cohort had an ORR of 33% (mostly partial responses). PSA response was reported in 63% of patients. Responses did not vary by germline versus somatic mutation status. Notably, BRCA1 mutants responded less well compared to BRCA2 mutants, and no benefit was documented in patients with hepatic metastasis, perhaps hinting at more aggressive disease biology.^[44] Based on these findings, rucaparib is approved for patients with CRPC with germline/somatic mutations in BRCA 1 or 2 after the failure of one AR-directed and taxane based therapy. A subsequent phase three study TRITON3 is underway and builds on the lessons learned from the PROfound study in the form of inclusion of a taxane naïve subset with BRCA 1 or BRCA2 mutations only. Furthermore, docetaxel therapy was included in the standard of care arm, coming closer to real-world settings.^[45]

Other PARP inhibitors tested in these settings

The TALAPRO-1 study reported similar efficacy with Talazoparib in patients with HRD (except CDK12 mutations) and prior chemotherapy, as well as AR-directed therapy.^[46] The GALLAHAD study reported similar findings in these patients with niraparib therapy, except that all patients had biallelic/germline HRD mutations.^[47]

Efficacy in non-BRCA HR gene mutations

While the efficacy in BRCA 1 and BRCA 2 mutants remains undisputed, the candidacy of other HRD genes remains controversial, especially ATM gene mutations. The TRITON2 study reported an ORR of 10.5%, 0%, 11%, and 28.6% in patients with ATM, CDK12, CHEK2, and other HRD genes, respectively. PSA responses and PFS outcomes mirrored the same, with no predictive capabilities of germline versus comatic status.^[44] Similar poor responses were documented in the GALLAHAD and TALAPRO-1 trials as well.^[46,47] However, since the overall cohort in the PROfound study reported a PFS benefit, Olaparib is approved in patients with PALB2, BRIP1, BARD1, RAD51, RAD54, ATM, CDK12, and CHEK2/1 mutations as well. However, subgroup analysis shows a lack of benefit with Olaparib therapy in patients with ATM, CDK12, and CHEK2 mutations.^[42] Notably, the TALAPRO-1 study excluded patients with CDK12 mutations altogether given the lack of benefit in prior studies.^[46]

Further, evidence for preferential therapy in BRCA mutants only comes from a meta-analysis by Wu et al. reporting a significant impact of PARPi monotherapy on ORR and PFS in HRD compared to non-HRD. Within the HRD subgroup, BRCA mutated patients had higher ORR and PFS compared to non-BRCA HRD mutated which was more than non-HRD patients.^[48] Another meta-analysis by Arsalan et al. suggests that the benefit may be limited to BRCA 1/1 and PALB2 mutations.^[49]

PARP inhibitors: Preferred agent selection

Given the presence of phase three evidence, Olaparib is the most commonly prescribed agent in patients with AR-directed therapy failure. Furthermore, it may be given in aforementioned non-BRCA HRD patients as well. However, lack of benefit on sub-group analysis and a weak comparator arm need to be kept in mind.^[43] Furthermore, given the taxane agnostic nature of the PROfound study, rucaparib may be preferred in patients failing on docetaxel as well although Olaparib may also be used given the efficacy outcomes in the TOPARP-A and the TOPARP-B study.^[32,41]

Basis behind combination therapy

Seminal conceptualization of AR-inhibition enhancing DNA-damage mediated killing stems from trials reporting enhanced efficacy of radiotherapy in patients receiving androgen deprivation therapy (ADT),^[50,51] which reflects in biopsy specimen as well.^[52] In vitro models also suggest enhanced killing with radiotherapy in presence of ADT.^[53] Among multiple postulated mechanisms, AR-blockade-mediated impairment of DNA repair mechanisms has gained the most traction.^[54] In this regard, a study reported an increase in the expression of DNA repair proteins in response to AR signaling.^[55] PARP1 specifically modulates the transcriptional activity or AR and ETS activity and consequently can lead to ligand independent AR-signaling.^[56] In addition, AR-signaling in CRPC is known to potentiate the expression of HRR genes and its inhibition may induce a state of “BRCAness.”^[57] In vitro studies have corroborated this through synthetic lethality assays as well.^[58]

Current evidence behind combination therapy

Among the first studies evaluating PARP inhibitor, in addition to AR-directed therapy was the NCI-9012 study which evaluated the addition of veliparib to abiraterone therapy regardless of DRD status. Regardless of the treatment given, patients with an underlying HRD mutation fared better. However, veliparib addition did not lead to improved outcomes in the entire cohort, or in the ETS translocation subset. The impact of veliparib addition specifically in the HRD subset remains unknown.^[59] Although a negative study, the NCI-9012 study informs us of the potential efficacy of AR-directed therapy in HRD subsets, the futility of universal PARP inhibitor prescription, and futility of ETS translocations as a potential biomarker. Given these findings, the BRCAAway study is currently evaluating HRD CRPC patients in terms of optimal combinations and sequencing.^[60] A similar phase two study evaluating Olaparib addition to abiraterone therapy after docetaxel failure reported a higher PFS in all patients regardless of the HRR gene status. The findings of this study prompted the PROpel trial (discussed below).

Current phase three evidence for upfront combination therapy in CRPC is limited to the recently presented PROpel, and the MAGNITUDE study.^[61,62] The PROpel study randomized patients with no prior therapy for CRPC (Docetaxel for CSPC allowed) to receive abiraterone with a placebo or Olaparib regardless of their HRD status. Less than 15% of patients had visceral disease and around a quarter of patients had an HRD mutation. The risk of progression was reduced by 34% in the entire cohort regardless of the HRD status, and without an increase with Grade 3 or higher ADRs or discontinuations. Notably, rates of thromboembolic events were higher (7.3% vs. 3.3%).^[61] These findings seem to support the universal utility of Olaparib addition regardless of mutational status.

The MAGNITUDE trial, evaluating niraparib in almost equivalent settings (therapy naïve CRPC randomized to abiraterone with placebo vs. niraparib), gives us a clearer picture. This comes from a prudent difference in its schema, namely, stratification of patients with HRR status at baseline (ATM, BRCA1, BRCA2, BRIP1, CDK12, CHEK2, FANCA, HDAC2, and PALB2). Nearly, 40% of patients had BRCA2 mutations, and around a quarter of all patients had visceral metastasis at baseline. Similar to the PROpel study, nearly, 20% had already received docetaxel for prior CSPC. While PFS was prolonged in the HRD cohort (HR: 0.73; 95% CI: 0.56–0.96), it was more pronounced in the BRCA 1 or 2 mutated cohort (HR: 0.53; 95% CI: 0.36–0.79), and nonexistent in the HRR cohort (at the pre-specified early futility analysis itself). In addition, niraparib was associated with more anemia, and thrombocytopenia leading to dose reductions and discontinuations. While these findings argue against combination therapy for all patients, doses of niraparib used here (200 mg) were lesser than in the initial monotherapy trials (300 mg).^[62] Evidence from initial phase two studies on Olaparib seems to hint at a dose-effect curve in CRPC and whether these findings would remain antithetical to the ones from PROpel study that had higher doses of niraparib been used remains an unanswered question.^[32,41]

A similar dichotomy of findings is expected in the current ongoing combination therapy trials with the CASPAR trial administering rucaparib regardless of HRR status (without stratification), while the TALAPRO-2 study is evaluating Talazoparib in all patients but is stratified by HRR status. In addition, both studies are using enzalutamide as the AR-therapy backbone.^[63-66] Of note, a recent pharmacovigilance meta-analysis has found an association between PARP inhibitors and subsequent risk of myelodysplastic syndromes/ acute myeloid leukemia (Absolute risk: 0.73%).^[67] While hardly any cases have been reported in the studies done in CRPC [Table 2], a longer follow-up is prudent. This is of paramount importance since most of the treatment-emergent myeloid neoplasms reported in ovarian cancer patients harbor a monosomal karyotype with TP53 mutations, a subset of AML with particularly dismal outcomes.^[68,69] Whether baseline NGS-based testing for CHIP will allow for better, patient selection is currently unknown and the characteristics of myeloid neoplasms associated with prior PARPi use are beyond the scope of this review.

Mechanisms of PARP inhibitor resistance: A primer

Four major mechanisms which may contribute to clinical resistance to PARPi are altered drug availability, alterations in PARylation enzymes, PARP independent restoration of HR, and restoration of replication fork stability.^[85] Tracing the steps involved in HRR allows us to elucidate the various alterations which may lead to resistance.

HRR: Basic steps

In response to a double stranded DNA break (DSB), cells activate HRR and NHEJ mechanisms simultaneously and the phase of the cell cycle decides the dominant mechanism.^[86] The following sequence simplifies the usual steps for HRR, the details of which are beyond the scope of this review:

1. Depending on the phase of the cell cycle the cell is in, appropriate CDK activates the effector proteins (CtIP and ATM by CDK18).^[87-89]

2. ATM and ATR kinases phosphorylate the variant H2A histone leading to its expansion and trapping of TP53 binding protein 1 (53BP1) (promotes NHEJ, negates HRR), as well as BRCA1 (part of HRR machinery)^[90]

3. Resection of the ends of the DSB lesion (done by MRN, CtIP, and various nucleases) is the first step in committing toward HRR^[91]

4. These resected ends and then coated by hyperphosphorylated replication protein A^[92]

5. The PALB2 protein allows interaction between BRCA1 and BRCA2, responsible for the unloading of the RAD51 protein on the lesion^[93]

6. The RAD51 protein is responsible for finding and invasion of the homologous sequence (formation of the D-loop structure). It also protects the single stranded DNA formed in the interim.^[94]

The NHEJ machinery, mainly the 53BP1 and the RIF1 protein, inhibits BRCA1.^[95] The 53BP1, specifically, also protects the DSB ends from resection through nucleosome barrier formation,^[96] as well as recruitment of the SHIELDIN complex.^[97] Of note, the ATM is needed for 53BP1 and RIF1 activation, implying that ATM plays a role in both HRR and NHEJ.^[98]

Potential mechanisms of resistance (In vitro)

Increased RAD51 activity, either through downregulation of inhibitors such as EMI1/DDB2 or through upregulation of the TOPβ1 activator, allows HRR to bypass BRCA1 or 2 functions and leads to resistance.^[99,100] Notably, bromodomain protein 4 overactivity leads to a similar rise in RAD51 activity which may be negated by using bromodomain inhibitors.^[101,102]

Another common mechanism of PARP inhibitor resistance observed in in vitro studies is reversion mutations.^[2]

A separate method of the resistance stems from replication fork stabilization downregulation of EZH2.^[103] PTIP/CHD4/ MELL 3- or 4-mediated reduction of MRE11 activity,^[104] FANCD2-mediated MRE11 suppression,^[21,105] SMARCAL1 deletion,^[106] or RADX deletion^[107] all lead to PARPi resistance in in vitro models through replication fork stabilization. While epigenetic mechanisms of PARPi resistance exist, they are yet to be targeted clinically.^[2]

Testing for possible mechanisms

Detection of three possible alterations may guide us regarding the possible mechanism of resistance to PARP inhibitors: Mutations, expression patterns, and functional assays.^[108] The most common mutations are reversion mutations in BRCA 1 or 2.^[2] These mutations may cause resistance to both PARPi, as well as platinum agents.^[109,110] Compared to other BRCA deficient malignancies, reversion mutations seem, especially common in CRPC (40% in one study).^[111] Liquid biopsy may detect.^[112] Mutations in 53BP1/ REV7 are found in up to 20% of PARPi resistant cases.^[113,114] Some PARP mutations may confer resistance through reduced trapping (R591C mutations).^[115] Assessment of expression of the PTIP/EZH2 proteins as well as the PARG moieties may also define mechanisms of resistance.^[104,116] In this regard, quantitative assessment of RAD51 foci on the genome may allow a functional assessment of HRR activity in a cell.^[114]

Overcoming PARP inhibitor resistance

Therapeutic interventions reversing the likely resistance mechanism are being actively investigated. In addition, certain resistance mechanisms may be targetable, a phenomenon known as collateral sensitivity.^[117] Among many others, radiotherapy is an accessible and proven modality for this purpose. PARP inhibitors sensitize cancer cells to radiation through impairment of DNA repair.^[118] Along these lines, studies are evaluating a combination of PARPi with radiotherapy in clinical settings and some have reported encouraging outcomes.^[119] Conversely, radiotherapy can overcome PARPi resistance resulting from a multitude of mechanisms. Concerning reversion mutations leading to the resumption of HRR, HRR requires nuclear localization of BRCA1 protein, while radiotherapy causes cytoplasmic extrusion of the same.^[120-122] Functional p53 is required for this phenomenon.^[118] On the other hand, tumors dependent on 53BP1 over-expression for PARPi resistance are exquisitely sensitive to radiotherapy. This probably stems from impairment in NHEJ. While PARPi is lethal in the G2/M phase of the cell cycle when HRR predominates, radiotherapy is cell cycle non-specific and NHEJ is the only mechanism of DSB repair in the G1 and early S-phase.^[123] Although at least six clinical trials are evaluating the utility of radiotherapy in overcoming PARPi resistance, none are enrolling patients with CRPC (since radiotherapy is not the standard of care).^[2] However, a phase one study (LuPARP) is evaluating Olaparib in addition to 177Lu-PSMA therapy.^[124]

An additionally acquired vulnerability in PARPi resistant cancers stems from a reliance on ATM and ATR kinases. The PALB2/BRCA2 proteins, indispensable to the restoration of HRR, are dependent on ATR-kinase for activation. The RAD51 protein is another ATR-kinase-dependent protein mediating resistance through fork stabilization.^[125] A similar relationship exists between TP53 binding protein 1 (TP53BP1)/DNA polymerase zeta processivity subunit 7 (REV7) proteins and the ATM-kinase.^[126] Consequently, studies are evaluating a combination of PARPi with pharmacological inhibitors of ATM or ATR-kinases in various malignancies,^[2] including CRPC.^[127]

PARPi and immune checkpoint inhibitors also demonstrate in vitro synergy, mainly through PARPi-mediated upregulation of PD-L1 expression (through GSK3β).^[128] STING-mediated inflammation in the tumor microenvironment is another potential mechanism of this synergy.^[129] While a multitude of trials is evaluating this combination, none are focused on CRPC.^[2] Another strategy is combining oncolytic herpes simplex viruses (oHSV) therapy with PARPi. Specifically, the MG18L oHSV knocks out RAD51 enhancing efficacy and overcoming resistance.^[130] Clinical studies evaluating this combination are eagerly awaited. Finally, pharmacological inhibition of CDKs, leading to preferential utilization of HRR, is another option with strong pre-clinical data.^[131]