PARP inhibitors in prostate cancer: time to narrow patient selection?

1. Introduction

Poly ADP-ribose polymerase (PARP) enzymes are critical to base excision repair of single-stranded DNA breaks. PARP inhibitors (PARPi) have been shown to induce synthetic lethality in some patients with metastatic cancers that harbor germline or somatic mutations in homologous recombination (HR) DNA repair genes. For example, BRCA1 and BRCA2 have been long studied in breast and ovarian cancer, which resulted in the U.S. Food and Drug Administration (FDA) approval of PARPi for widespread use in these biomarker-selected cancers. More recently, the FDA also approved two PARP inhibitors (olaparib and rucaparib) for metastatic prostate cancer patients with germline or somatic pathogenic mutations in one of 14 HR genes (olaparib) or BRCA1/2 specifically (rucaparib).

HR proteins have various functions in repairing double-stranded DNA breaks. BRCA1 and BRCA2, in concert with PALB2 and RAD51, directly mediate and execute the repair of double-strand DNA damage. Other HR family members, such as ATM/ATR and CHEK2, may act as sensors of double-strand DNA damage, facilitating subsequent recruitment and activation of BRCA1/2 and other effector proteins. Given these divergent roles, all of which may broadly be considered as DNA repair functions, genetic aberrations in specific HR DNA repair genes may confer variable sensitivities to PARP inhibition and should be dissected at the individual-gene level.

2. Commentary

Germline and somatic mutations in HR DNA repair genes have been observed in approximately 10–15% and 20–25% of patients with metastatic prostate cancer, respectively [1,2]. Given the prevalence of HR deficient (HRD+) prostate cancer, multiple clinical trials have studied the efficacy of PARPi in patients with metastatic castration-resistant prostate cancer (mCRPC). Now that these studies have matured, it is becoming increasingly clear that not all HR DNA repair genes (when inactivated) predict for favorable response to PARP-directed therapy.

The TOPARP-A study was a Phase II clinical trial investing olaparib in heavily-pretreated patients with mCRPC[3]. Although not a requisite for enrollment, this study was enriched for patients that harbored an HR DNA repair gene mutation. HRD+ patients in that study, in comparison to the HR-proficient patients, had an improved composite response rate (88% vs. 6%), progression-free survival (9.8 vs. 2.7 months) and overall survival (13.8 vs. 7.5 months). The majority of HRD+ patients that responded to olaparib in that trial had BRCA1/2 or PALB2 alterations. Interestingly, 4 of 5 patients with an inactivating ATM mutation also derived benefit as measured by the composite endpoint, suggesting that DNA damage sensing members of the HR DNA repair family may also be important in predicting PARPi sensitivity. However, subsequent retrospective data have suggested that mCRPC patients with ATM mutations had inferior outcomes to PARPi compared to those harboring BRCA1/2 mutations[4]. More recent prospective studies have also suggested limited benefit to PARPi in non-BRCA1/2 mutated patients, specifically ATM, CHEK2, and CDK12. A follow-up Phase II study (TOPARP-B) was conducted using olaparib in HRD+ mCRPC patients in the post-chemotherapy setting[5]. Twenty-eight of thirty-seven patients (75.7%) with a pathogenic BRCA1/2 or PALB2 mutation achieved a PSA decline or objective response, compared to only 2 of 39 patients (5.1%) with an underlying ATM or CDK12 mutation. Although circulating tumor cell (CTC) conversion rates were favorable for some patients with ATM, CDK12 and other non-BRCA1/2 mutations, this metric may not be a reliable indicator of clinical benefit. PROfound was a Phase III trial for HRD+ mCRPC patients (who had received at least one novel AR-directed therapy; prior chemotherapy was allowed but not mandated), who were randomized to olaparib or the alternative AR-targeted therapy. In the cohort of patients with a BRCA1, BRCA2 or ATM mutation (the primary study population), radiographic progression-free survival (rPFS) was significantly longer following PARPi vs. AR-targeted therapy (7.4 vs. 3.6 months, HR 0.34, P < 0.001)[6]. In a prespecified secondary analysis that considered all patients with one or more mutations in a larger panel of 15 HR genes, the rPFS benefit of olaparib remained significant, leading the FDA to approve olaparib on 19 May 2020 for mCRPC patients with a broad panel of HR gene mutations. However, the benefit of PARP inhibition in PROfound was largely found in the BRCA2-mutated population (rPFS olaparib vs. control: 10.8 vs. 3.5 months). BRCA1 mutation did not appear to portend benefit to olaparib therapy (rPFS: 2.1 vs. 1.8 months), nor was a benefit apparent in ATM-mutated patients (rPFS: 5.4 vs. 4.7 months). This finding merits further subgroup analyses of prior PARP inhibitor studies in prostate cancer examining the differential responses between BRCA1 vs. BRCA2 mutations. Interestingly, in the PROfound study, RAD51B (rPFS: 10.9 vs. 1.8 months) and RAD54L (rPFS: 7.2 vs. 2.4 months) mutations also appeared to favor olaparib, suggesting a key role for RAD51/RAD54 proteins in HR DNA damage repair.

A recent publication now sheds additional light on the response to PARPi treatment in non-BRCA1/2-mutated patients. TRITON-2 is a single-arm open-label Phase II trial of rucaparib in post-chemotherapy mCRPC patients that harbor a DNA damage repair gene mutation. Preliminary results had shown that treatment with rucaparib induced a PSA and objective response rate of 53.6% and 47.5%, respectively, in patients with BRCA1/2 mutations[7]. Final survival data remain unpublished, but these findings further support the use of PARPi in BRCA1/2-mutated mCRPC patients. In fact, this non-randomized single-arm study of rucaparib in the BRCA1/2-mutated prostate cancer population led the FDA to approve rucaparib on 15 May 2020 for the treatment of mCRPC patients with pathogenic BRCA1 or BRCA2 alterations, specifically. The non-BRCA1/2 mutated patient cohort [8] is the focus of this editorial. In that cohort, Abida et al. observed low PSA response rates in prostate cancers with ATM (2/49 = 4.1%), CDK12 (1/15 = 6.7%) or CHEK2 (2/12 = 16.7%) mutations. However, anecdotal PSA responses were observed in patients with PALB2 (2/2 = 100%), BRIP1 (1/2 = 50%) and RAD51B mutations (1/1 = 100%). Therefore, the totality of the clinical data thus far using olaparib and rucaparib both show high clinical activity in BRCA1/2-mutated patients with limited activity in the setting of ATM, CHEK2 or CDK12 mutations.

Other PARPi, including niraparib and talazoparib, have also been explored in patients with advanced mCRPC. The GALAHAD study is a Phase II clinical trial using niraparib in HRD+ patients with mCRPC after progression on an AR-targeted therapy and a chemotherapy. Clinical response rates were notably lower in non-BRCA1/2 vs. BRCA1/2 mutated patients[9]. PSA and objective response rates following niraparib were 57% vs. 6%, and 38% vs. 11%, respectively, favoring those patients with BRCA1/2 mutations. Similar findings were observed in the TALAPRO-1 study, a Phase II trial using talazoparib in post-chemotherapy patients with mCRPC and an HR DNA repair gene mutation[10]. BRCA1/2-mutated patients had a 42% objective response rate and a 61% PSA response rate with talazoparib treatment. Only 12% and 6% of ATM-mutated patients achieved an objectiveor PSA response, respectively, to talazoparib. In our attempt to summarize the overall data, response rates to different PARPi across studies and stratified by individual HR DNA repair gene mutations are collated in Table 1.

Table 1.

PSA response rates (PSA RR) and objective response rates (ORR) stratified by type of homologous recombination DNA repair gene mutations in various phase II clinical trials using PARP inhibitors in patients with metastatic castration resistant prostate cancer.

	BRCA1/2		Non-BRCA1/2		ATM		PALB2		CHEK2		CDK12
	PSA RR	ORR	PSA RR	ORR	PSA RR	ORR	PSA RR	ORR	PSA RR	ORR	PSA RR	ORR
TOPARP-B (olaparib)	77%	52%	11%	6%	5%	8%	67%	33%	NR	NR	0%	0%
TRITON-2 (rucaparib)	54%	48%	NR	NR	4%	11%	100%*	50%*	17%	11%	7%	0%
GALAHAD (niraparib)	57%	38%	6%	11%	NR	NR	NR	NR	NR	NR	NR	NR
TALAPRO-1 (talazoparib)	61%	42%	50%	33%	7%	7%	NR	NR	NR	NR	NR	NR

PSA RR = PSA response rate, ORR = Objective response rate, NR = not reported

^*N = 2 patients.

Based on these collective studies, several overarching trends have emerged. First, PARPi appear to be highly effective in patients with BRCA1/2-mutated mCRPC (particularly BRCA2). Several different PARPi all demonstrated high clinical response rates, but whether these agents can lead to an overall survival benefit remains to be determined. It is also possible that the true benefit of PARPi in non-BRCA1/2 patients might be uncovered only when we have overall survival data from the randomized trials. Second, less common HR DNA repair genes such as PALB2, FANCA, RAD51B and BRIP1 may also confer PARPi sensitivity, but more patients are needed to confirm this potential benefit. Definitive conclusions are difficult to draw at this time, due to the small number of patients studied with each of these gene defects. Third, clinical benefit of PARPi in ATM-mutated mCRPC appears to be minimal, and alternative strategies should be sought for such patients. Although CTC conversion has been observed following PARPi in some ATM-mutated cases, the general lack of PSA or objective responses are concerning for future drug development in these patients. However, the PROfound study combined ATM together with BRCA1/2 in their primary study cohort, which did show an rPFS benefit following olaparib treatment compared to AR-targeted therapy in that entire cohort, justifying the potential use of olaparib in certain patients with ATM alterations. Fourth, CDK12 and CHEK2 mutations do not generally appear to confer significant clinical benefit to PARPi in mCRPC patients. This is important because CDK12 and CHEK2 mutations are quite common, accounting for about 6% and 2% of mCRPC cases, respectively. Interestingly, CDK12 mutations are virtually always somatic alterations (and are often biallelic), while CHEK2 mutations are usually germline alterations. However, to our knowledge, biallelic inactivation of the second somatic allele in tumor cells has not been reported in patients with germline CHEK2 mutations (and double-somatic CHEK2 inactivation is also extremely rare), perhaps reflecting one reason that such patients may not respond favorable to PARPi therapy. Finally, further study is needed to explore resistance mechanisms to PARPi, including reversion mutations that restore the open reading frame and rescue protein function, thus abolishing synthetic lethality. In particular, can reversion mutations be delayed if PARPi are administered in combination with other agents (i.e. platinum chemotherapy)? Should PARPi be given intermittently or earlier in the treatment course to prolong efficacy? Many such questions remains unanswered.

The case of CDK12 alterations deserves further consideration[12]. While this gene was previously thought to be one of the HR DNA repair genes (and might function as such in some tumors including ovarian cancer)[13], it appears to have a distinct role in maintaining genomic stability in prostate cancer[14]. In mCRPC, biallelic CDK12 inactivation leads to a unique pattern of genome-wide chromosomal instability characterized by widespread focal tandem genomic duplications that lead to gene fusions and consequently fusion-induced neoantigens [15,16]. While these CDK12-altered mCRPC patients do not generally respond to PARPi as outlined above, a subset may have favorable responses to PD-1 inhibitors such as pembrolizumab or nivolumab. The percentage of such patients that derive clinical benefit from a PD-1 inhibitor is somewhere in the 20–40% range [17,18], which is on par with the response rate to immune checkpoint inhibitors in prostate cancer patients with mismatch repair (MMR) deficiency [19,20]. Therefore, both MMR mutations and CDK12 mutations seem to be more predictive of benefit to PD-1 blockade rather than PARP inhibition[21]. Dedicated studies targeting the CDK12-mutated and MMR-deficient mCRPC populations are needed before the efficacy of immunotherapy in these patient subsets is confirmed.

3. Conclusion

In mid-May 2020, the FDA approved two PARP inhibitors for mCRPC: olaparib for HR gene-mutated patients (14-gene panel: BRCA1, BRCA2, ATM, BARD1, BRIP1, CDK12, CHEK1, CHEK2, FANCL, PALB2, RAD51B, RAD51C, RAD51D, or RAD54L), and rucaparib for BRCA1/2-mutated patients specifically. This heralds a new era in the precision treatment of advanced prostate cancer. It is becoming clear, however, that not all DNA repair deficiencies can be considered in one broad bucket[22]. Each repair gene (and protein) has a different function, and it is also possible that different mutations within the same gene (e.g. BRCA2) may also result in different sensitivities to PARPi therapies. To this end, there are clear examples even of patients with biallelic BRCA2 mutations who do not benefit from PARP inhibitors, while there are other examples of patients who do derive benefit from PARP inhibitors without any apparent DNA repair gene mutations (either because they were not present, or they were not detected, or the gene was inactivated by some other means e.g. methylation). Therefore, perhaps a better way of defining PARPi sensitivity would be to focus on functional assays of HR repair activity (or deficiency). These assays have been challenging to develop and validate thus far. Alternatively, RNA-based assays of genomic scarring or DNA-based genomic signatures associated with HR deficiency might prove useful, but such work has not been performed in prostate cancer. Although the presence or absence of a detectable genetic mutation in one or more individual HR genes might be the easiest way of capturing benefit of PARPi presently, it is naïve to think that this simplistic method will actually reflect HR status in a given tumor at a particular moment in time. The challenge that lies before us, therefore, is to develop and implement assays interrogating broad HR repair function (or deficiency) in order to better select our patients for treatment with PARP inhibition. At the same time, we should also focus on developing other non-PARPi–based synthetic lethality strategies for patients with non-BRCA1/2 mutations (particularly ATM, CDK12 and CHEK2 alterations) by targeting other protein components of the DNA repair machinery in these subsets of patients.