BRCA1 and BRCA2 loss-of-function mutations are the most well-recognized predictive biomarkers of response to PARPi (6, 17). BRCA1/2 genes are characterized by different functional domains (2). BRCA2 has three main functional domains, namely the RAD51-binding domain (RAD51-BD), a C-terminal DNA-binding domain (DBD), and the BRC or TR2 domains that interact with RAD51 filaments. All are involved in HR pathways (2). Moreover, BRCA1 functional domains include a highly conserved N-terminal Really Interesting New Gene (RING), a DBD, and a C-terminal domain of BRCA1 (BRCT). While they are all involved in DNA repair, BRCT is also implicated in the cell cycle through G2/M and S-phase checkpoint. Consequently, some studies reported distinct outcomes in BRCA1 or BRCA2 loss of function mutation carriers treated with PARPi. A post-hoc analysis of study 19 and ARIEL2 trials showed better outcomes in BRCA2 mutation carriers, with olaparib and rucaparib, respectively (30, 31). Of note, patients with BRCA1 promoter hypermethylation derived benefit from rucaparib, but no one was observed among long-term responders, and two experienced poor outcomes despite olaparib therapy in study 19 (30). The location and type of mutation also confer distinct sensitivity to PARPi. In vitro studies showed that tumors with mutation in exon 11 of BRCA1 are less sensitive to PARPi than those with mutation outside the exon 11 (32). Moreover, deletion in DBD of BRCA2 increases olaparib and cisplatin sensitivity of engineered cell lines while deletion within the C-terminal domain retains partial HR activity and confers less sensitivity to DNA-damaging agents (33). Recently, ancillary analysis from the PAOLA-1 phase III clinical trial (29), assessing olaparib and bevacizumab first-line maintenance therapy, was the first large effort to assess the impact of the location of BRCA1/2 mutations on PARPi sensitivity in patients (6). Labidi-Galy and colleagues analyzed the location of BRCA1/2 mutation in 233 out of 806 randomized patients. Interestingly, they found that patients with BRCA1/2 mutation involving exon 11 derived greater benefit from the addition of olaparib compared to patients with mutation outside the exon 11 (HR = 0.2 [95% CI = 0.11–0.36] and HR = 0.41 [95% CI, 0.22–0.75], respectively) (29), as well as those with mutations in DBD of BRCA1 (HR = 0.08 [95% CI = 0.02–0.28]). These results are, however, surprising since previous studies reported worse outcomes and sensitivity to PARPi and platinum in cell lines with BRCA1 mutation inside versus outside exon 11 due to the presence of a hypomorphic BRCA1 protein (32). Nevertheless, the median PFS (mPFS) of 16 months in these patients in the PAOLA-1 trial suggests lower platinum sensitivity (29) but still deserves further analyses, specifically regarding cross-resistance between platinum and PARPi and BRCA1 protein functionality. In BRCA2 mutation carriers, whereas patients with mutation in RAD51-BD had a significantly longer PFS with the addition of olaparib (HR = 0.31 [95% CI = 0.11–0.92]), those with mutation in DBD had an excellent outcome in both arms (24 months PFS of 90% and 100% with and without olaparib, respectively), consistent with in vitro studies showing a substantial PARPi and platinum sensitivity in DBD BRCA2-mutated cell lines (33) and very rare reversion mutations in that domain (25). Of note, subgroup analysis from the PAOLA-1 and SOLO1 trials showed that BRCA1 mutation carriers derived greater benefit from PARPi (HR = 0.40 [95% CI = 0.29–0.56] and HR = 0.20 [95% CI = 0.10–0.38] for PFS in BRCA2 and BRCA1 carriers, respectively, in SOLO1 and HR = 0.5 [95% CI = 0.34–0.73] and HR = 0.2 [95% CI = 0.11–0.39]), respectively, for PFS in PAOLA-1) (29, 34). Altogether, these data explain, in part, different outcomes in patients with BRCA1 or BRCA2 mutations treated with PARPi, with some patients experiencing a very long response and a potential cure, while others have quick progression despite a BRCA1/2 mutation or HRD profile.

A few BRCA1 mutations lead to a hypomorphic BRCA1 protein that retains HR activity and promotes PARPi resistance. For instance, a frameshift mutation in exon 11 leads to the BRCA1-Δ11q splice variant transcript and a BRCA1 protein with HR activity measured by RAD51 γ-irradiation-induced foci formation in vitro. Even though BRCA1-Δ11q HR activity is inferior to BRCA full-length, this drives partial PARPi resistance (32). Importantly, this variant has been found in postprogression tumor samples from patients previously treated with rucaparib in the ARIEL2 trial (35). Similarly, alterations in the RING domain of BRCA1, like BRCA1185delAG or BRCA1C61G, also translate into a hypomorphic RING-less BRCA1 protein with a partial DNA damage response that decreases sensitivity to PARPi and platinum therapy. Mice with a hypomorphic RING-less BRCA1 protein become rapidly resistant to PARPi and platinum without acquiring reversion mutations (36, 37). Altogether, these data showed that the BRCA1 RING domain and exon 11 are dispensable, to some extent, for HR activity. In contrast, hypomorphic BRCA1 proteins that lack domains in the C-terminal region might need the alteration of additional pathways to acquire PARPi resistance. For instance, a stop codon in the coiled-coil domain of BRCA1 results in a hypomorphic protein that acts downstream of end resection and brings PARPi resistance only in a 53BP1 gene knockout mouse model (38). In addition, the BRCT-less BRCA1 chimeric protein is usually destroyed and therefore needs to be stabilized by HSP-90 to escape proteasome degradation, interact with the PALB2-BRCA2-RAD51 complex, and drive PARPi and platinum resistance (39). Even if less described, some hypomorphic BRCA2 proteins have also been reported. Although no relationship between PARPi and platinum sensitivity was reported, BRCA2 C-terminal DNA-binding domain deletion but conserved BRC repeat motifs allowing complex with RAD51 and HR activity has been described in vitro (40). Another in vitro study also showed that duplication of the BRCA2 mutant allele lacking the DBD along with the overexpression of truncated protein led to PARPi resistant cell lines. However PARPi resistance requires a Disruptor Of Telomeric silencing 1-Like (DOT1L) to interact with BRCA2 truncated protein, suggesting that hypomorphic BRCA2 is not sufficient by itself to promote PARPi resistance and needs an additional mechanism of HR recovery (41).

It is now well-admitted that, beyond BRCA1/2 mutation and HRD status, patients responding to platinum-based chemotherapy are the ones who benefit most from PARPi (42). Furthermore, platinum resistance confers a poor benefit of subsequent PARP inhibition. Recent data also showed a poorer survival and overall response rate (ORR) with platinum rechallenge after PARPi maintenance therapy, regardless of platinum-free interval (PFI) and BRCA1 or BRCA2 mutations (43–46). Altogether, these data suggest that platinum and PARPi resistance mechanisms partially overlap. BRCA1/2 reversion mutations after platinum-based chemotherapy have been reported to occur in 20% to 40% of platinum-resistant or refractory diseases (26, 47, 48), while in less than 5% in platinum-sensitive settings (27, 28, 48) and jeopardize primary PARPi efficacy. Of note, the occurrence of reversion mutations in platinum-sensitive diseases also highlights the limit of the platinum-free interval as the only definition of platinum sensitivity. Indeed, post-hoc analysis of OlyimpiAD (8), SOLO3 (49), and LIGHT (50) trials assessing olaparib in BRCA1/2 mutant breast cancers and recurrent platinum-sensitive ovarian cancer patients showed that 4% (N = 4/114) and 3% (N = 4/130) of patients with breast and ovarian cancers, respectively, already harbored BRCA1/2 reversion mutations before PARPi, despite a platinum-sensitive disease as defined with PFI (28). Moreover, ancillary analysis from the ARIEL2 study assessing rucaparib in platinum-resistant and platinum-sensitive patients reported enrichment of reversion mutations in platinum-resistant or refractory diseases (13% and 18%, respectively) compared to platinum-sensitive diseases (2%) (31). Of note, for three patients, including one with platinum-refractory disease and two patients with platinum-resistant diseases, five, four, and eight reversion mutations, respectively, were detected using ctDNA. Patients with BRCA1/2 reversion mutations before PARPi had a poorer PFS with rucaparib compared to patients without reversion mutations (mPFS 9 months and 1.8 months, respectively, HR = 0.12 [95% CI = 0.05–0.26]; p < 0.0001). Interestingly, among patients with platinum-resistant or refractory disease, those with a reversion mutation had a poorer prognosis (mPFS 7.3 months and 1.7 months, respectively, HR = 0.16 [95% CI = 0.07–0.42]; p < 0.0001, in patients without and with BRCA1/2 reversion mutation) (27), suggesting that we should refine the prognostic groups based on molecular characteristics. Similarly, Norquist et al. reported a reversion mutation in 46% of platinum-resistant HGSC patients versus only 5% in platinum-sensitive tumors (48). Among patients with platinum-resistant ovarian cancers and mutation reversion, two out of three were resistant to subsequent PARPi, whereas, surprisingly, one patient still derived benefit from PARPi and experienced a partial response despite BRCA1 reversion mutation to wild type and a platinum-resistant disease. Moreover, three patients with platinum-resistant disease did not have reversion mutations and experienced a partial or complete response with PARPi (51). Thus, although reversion mutations usually confer crossresistance between platinum and PARPi, some might specifically drive platinum resistance while retaining PARPi sensitivity, and some mechanisms of platinum resistance might be independent of reversion mutations and consequently not overlap with PARPi resistance. Reversion mutation types and underlying mechanisms might also help to predict PARPi resistance in patients with reversion mutations occurring under platinum-based chemotherapy. As more extensively described below, types of reversion mutation usually overlap with those that appear under PARPi or platinum-based chemotherapy, and microhomology end joining (MMEJ) is involved in both situations (26, 47). These strengthen the idea that most of the reversion mutations might confer PARPi and platinum crossresistance. However, larger deletions are enriched in reversion mutations affecting the BRCA2 gene after PARPi compared to platinum therapy (26), suggesting that some mechanisms could also be drug-specific.

The most reported mechanism of HR restoration under PARPi is the reversion mutation of the BRCA1/2 gene, which restores the expression of BRCA1/2 functional protein and HR pathways. BRCA1/2 reversion mutations encompass second-site deletion or insertion and the in-frame deletion of the original mutation, both restoring the open reading frame, and the true reversion mutation, namely “mutation reversion to wild type”, restoring the wild-type gene sequence (25). BRCA1/2 reversion mutations have first been studied using in vitro cancer cell lines, demonstrating the restoration of open reading frame and HR activity that confer platinum and PARPi crossresistance (52, 53). Later, reversion mutations were studied on PDX models treated with DNA-damaging agents. Under cisplatin or olaparib treatments, Ter Brugge and colleagues described a BRCA1/2 original mutation deletion that restored the open reading frame and, thus, the production of BRCA1 functional protein. They also described the loss of methylation of the BRCA1 promotor as well as the fusion of BRCA1 with a heterologous promotor that allows for transcription of hypermethylated BRCA1 and confers PARPi and platinum resistance in PDX model with BRCA1 promotor hypermethylation (54). Next, using samples from PARPi-resistant patients, the reversion mutation under PARP inhibitors was transferred into the clinic. Interestingly, several teams report different clinical cases involving breast (55, 56), ovarian (35, 57), pancreatic (58), and prostate cancers (59), highlighting that PARPi resistance mechanisms would be shared between cancers of the BRCA1/2 spectrum. BRCA1/2 reversion frequency under PARPi accounts for 10% to 40%, depending on the studies (26–28, 47, 60, 61), and can occur on BRCA1/2 somatic or germline original mutants and in platinum-sensitive or resistant settings. In ARIEL2 post-hoc analysis (31), reversion mutations that appear under rucaparib were detected on cfDNA samples for eight out of 78 patients (10%), regardless of platinum sensitivity, and were detected at a median of 3.4 months before radiological progression. Interestingly, among the three patients who harbored multiple reversion mutations acquired before PARP inhibition, likely secondary to previous platinum-based chemotherapy, the mutation allele frequency (MAF) of the different reversion mutations changed under PARPi pressure, suggesting that some reversion mutations may preferentially drive PARPi resistance while others might retain, to some extent, PARPi sensitivity. More recently, Pettitt et al. and Tobalina et al. reviewed all published clinical cases reporting BRCA1/2 reversion mutations in cancers of the BRCA1/2 spectrum, namely ovarian, breast, prostate, and pancreatic cancers (25, 26). Pettitt et al. also included relevant studies from cell lines and PDX models (25). In both studies, most patients were treated for ovarian cancer, which was expected regarding the extensive use of platinum and PARPi in this tumor type. They reveal that most of the reversion mutations are unique, reported as “single-patient mutations”, except for the stronger propensity of true-reversion mutation to wild-type recurrent reversion mutations (25). Reversion mutations occur more frequently in BRCA2 than in BRCA1 genes. Within each gene, they also identified regions more prone to be impacted by reversion mutations, namely “hotspot” regions, and those called desert regions with very few reversion events ( Figure 1 ). Within BRCA2, reversion mutations were frequently observed in N-terminal or RAD51BD domains, namely hotspot, while they were relatively rare in the highly conserved C-terminal region, encompassing the DBD domain, namely “a desert” (25, 26) ( Figure 1 ). Of note, reversion mutations occurring in this region were a true reversion to wild-type or missense reversion mutations, whereas deletions were more frequently observed in the hotspot region. In BRCA1, BRCT and RING domains are found to be the hotspots (26). The type of original loss-of-function alteration also impacts the likelihood of being reversed. Reversion of original frameshift deletion or insertion pathogenic mutations is more frequent that reversion of original missense or splice site mutations, which are relatively rare (25–27). Secondary mutations themselves are mainly deletions, with a larger size observed in the BRCA2 gene. Larger deletions are especially seen in exon 11 of both BRCA1 and BRCA2, highlighting less conserved and dispensable sequences with regard to BRCA function, as previously described (32). However, more substitution or reversion to wild type occurs in BRCA1, suggesting that the mechanism of reversion mutation in the BRCA1 gene might arise from a larger range of DNA repair mechanisms. Thus, the type of secondary mutation depends on the location and type of primary mutation (25, 26). While the mechanism of reversion mutation is not fully understood, the identification of DNA sequence microhomology at the end of the site of reversion mutation highlights that MMEJ may be responsible for, at least a part of reversion mutations. The detection of microhomology sequences is more frequently reported in BRCA2 than BRCA1, although surround reversion mutations occur in more than half of the cases and are longer in BRCA2 (25, 26). Notably, whereas reversion mutations seem to be more frequent in the BRCA2 gene, BRCA2 mutant carriers do not have a worse outcome, and the opposite has even been reported. As previously described, reversion mutations are often large deletions that could produce hypomorphic proteins. As mentioned above, while BRCA1 hypomorphic protein has been reported as PARPi primary resistance mechanism (32), hypomorphic BRCA2 protein might need additional alteration to drive HR restoration (40, 41). Moreover, longer sequences of microhomology and secondary deletion or insertion, more frequent in BRCA2, introduce novel amino-acid sequences and may constitute neoantigens and drive immunogenic antitumor activity (25). However, we should be aware that these data are only hypothesis-generating and should be further explored. Beyond reversion mutations, heterozygosity also plays a role in PARPi efficacy. Lheureux et al. reported upregulation of BRCA1 and BRCA2 along with gene copy number gain as a mechanism of resistance in two patients who experienced a very long response under PARPi (57). Of note, no one had germline BRCA1/2 loss of function at baseline. Similarly, loss of homozygosity or complete loss of BRCA1 promotor hypermethylation in PDX models of BRCA1-methylated HGSC is also associated with PARPi resistance, demonstrating that hypermethylation of all BRCA1 copies are required to predict response to PARPi (54, 62). Thus, deciphering BRCA1/2 mutation reversion might help tailor the patient’s follow-up and treatment strategy. Patients with BRCA2 mutations in DBD have an excellent prognosis and a low risk of reversion mutation occurrence, while patients with BRCA2 mutations in the N-terminal domain or in BRCT or RING domain of BRCA1 should be closely monitored. Moreover, beyond BRCA1/2 mutations, copy number and gene expression should be taken into consideration to estimate the probability of both response and resistance.

The use of gene panels searching for mutations in HR-related genes other than BRCA1/2 failed to predict response to PARPi in ovarian cancers (63), limiting such use in clinical practice. However, loss of function of specific HR-related genes, such as RAD51C and RAD51D, has still been associated with PARPi response (31, 64), and reversion mutations have also been reported in this setting. Post-hoc analysis from ARIEL2 part I study assessing rucaparib in platinum-sensitive HGSC patients showed that secondary mutations in RAD51C and RAD51D that restore the open reading frame allow the production of functional protein and resistance to PARPi, further validated in vitro (35). More specifically, one patient harbored four distinct RAD51C reversion mutations in a postprogression sample, all responsible for multidrug resistance in culture cell lines, including platinum (cisplatin and carboplatin) and PARPi (rucaparib, olaparib, veliparib, talazoparib, and niraparib). Interestingly, different adjacent core biopsies collected to generate PDX models revealed that one core predominantly contained one specific reversion mutation (c.574_577delinsGGCG mutation) and underlined substantial intratumor heterogeneity and the emergence of resistant clones under PARPi selective pressure. In another patient, a RAD51D reversion mutation was also reported, but only within a splenic lesion that was progressing on rucaparib, while it was not found in liver metastasis still responding to rucaparib. In addition, loss of RAD51C promoter methylation has also been reported to confer PARPi resistance (65), with loss of methylation in a single copy being sufficient to cause PARPi resistance (66), as was previously shown in BRCA1-hypermethylated ovarian cancers. In the EVOLVE study, authors also described a reversion mutation of RAD51B as well as an overexpression of RAD51C in a second patient after olaparib failure (61). RAD51 amplification has also been reported in PDX models (67). Beyond RAD51, the PALB2 reversion mutation has also been described in vitro (68) and then reported in patients treated with olaparib for prostate cancer (69, 70). Therefore, reversion mutation that drives PARPi resistance is not restricted to BRCA1/2 genes and should be considered among biomarkers of response and resistance to PARPi.

Liquid biopsies and ctDNA analysis have been increasingly studied in the past few years (71). In ovarian cancer, several studies described reversion mutations detected on ctDNA with high sensitivity and specificity and a strong correlation with mutations found within tumors (27, 60). In a post-hoc analysis of the ARIEL2 trial, cell-free DNA (cfDNA) was able to detect ctDNA in 81% and 96% of pre- and posttreatment samples, respectively (27). Similarly, Weigelt and colleagues also detected ctDNA in 95% of blood samples from patients with ovarian cancers (47), and another team reported a sensitivity and specificity of 60% and 90%, respectively, to detect reversion mutations in ctDNA (60). In addition to the advantage of being minimally invasive, one of the major interests of liquid biopsy is to detect intratumor heterogeneity, which allows for a more global tumor assessment. Indeed, distinct reversion mutations are found within tumor subclones that arise with disease evolution and PARPi selective pressure. For instance, Barber et al. reported a single BRCA2 reversion mutation occurring in a metastatic lymph node that was the only site of progression. No reversion mutation was found in the primary tumor or peritoneum metastasis (56). Another particularly notable example is from Patch et al., who reported 12 distinct BRCA2 reversion mutations using multiple metastatic lesion collections from an autopsy patient. Some reversion mutations were shared between several metastatic sites, while others were site-specific and therefore could be used to describe cancer phylogeny (12). Therefore, in clinical practice, ctDNA might help to detect a larger number of reversion mutations stemming from different tumor subclones that would not have been detected with a single tumor solid biopsy. Lin et al., in an ancillary analysis from the ARIEL2 trial, reported additional reversion mutations detected on ctDNA compared to NGS of solid tumor tissue (27). Furthermore, MAF variation of the reversion mutations acquired on platinum-based chemotherapy under PARPi highlighted subclone expansion under PARPi pressure. Multiple reversion mutations were also more frequently found in ctDNA than in paired tumor samples in the prospective analysis from Weigelt et al. on 24 HGSC patients (47). Importantly, a patient with reversion mutations of BRCA1 found in ctDNA was detected neither in primary tumors nor in peritoneum metastasis samples. Of note, Christie and colleagues reported the reversion mutations from paired ctDNA and ascite samples of 30 retrospectively selected HGSC patients. Unlike previously described, more reversion mutations were detected in ascite samples than in ctDNA, which is consistent with the highly conserved somatic mutation described in ascitic fluid (72). Another opportunity that ctDNA offers is the detection of reversion mutations occurring during the course of the disease in a “real-time” manner. Jacob and colleagues experimented with this concept in a patient treated for an advanced HGSC. By collecting the ctDNA at several time points, they described several reversion mutations that emerged during the course of the disease along with successive treatments (73). Moreover, some studies showed that ctDNA may detect reversion mutations before radiologic or clinical progression (27, 47, 60), although the benefit of an earlier modification of treatment strategy is still unanswered. Thus, even more promising would be the use of ctDNA as a surrogate biomarker of response to PARPi. In a recent prospective phase I clinical trial assessing camonsertib, an ATR inhibitor, in 120 patients with solid tumors and alteration in DNA damage repair genes, the investigators collected ctDNA at baseline and at each cycle of treatment to perform targeted sequencing. They defined molecular response as a 50% decline in the mean variant allele frequency (MAF) of the original somatic variant. A total of 54% (N = 7/13) of evaluable HGSC patients and 53% (N = 15/28) of patients with BRCA1/2 mutations had a molecular response. Although not statistically significant, the molecular response was more frequently observed in patients with clinical benefit versus not (66% vs. 25% and 69% vs. 40% in ovarian cancer and BRCA1/2 mutant subgroups, respectively) (74). However, ctDNA analysis does not come without drawbacks, and technical specificity might miss some alterations. For instance, in the ARIEL 2 trial, a large deletion of BRCA1 that restored the open reading frame in a platinum-resistant patient was detected in the tumor sample but not in ctDNA (27). Thus, the short cfDNA fragments and short paired-end read sequencing usually used for ctDNA sequencing should be overcome in the future to detect specific alterations, such as large deletions, by the implementation of ctDNA sequencing new cutting-edge technologies (75, 76). Thus, ctDNA should be used in the near future to assess the molecular response to PARPi as well as to detect the earlier occurrence of the BRCA1/2 reversion mutation. Of note, the frequency of ctDNA assessment should be tailored to the risk of progression under PARPi considering, among others, the original loss of function mutation of BRCA1/2, with a higher risk of reversion mutation occurrence and earlier recurrence if original mutation is a frameshift deletion or insertion or is within the BRCT or RING domain of BRCA1 and in RAD51-BD of BRCA2 ( Figure 2 ).

For a long time, surgery has been the cornerstone of advanced ovarian cancer treatment. Importantly, complete macroscopic resection is the most important prognostic factor in ovarian cancer. Residual disease might be associated with the presence of tumor subclones and intratumor heterogeneity and thus increase resistance to subsequent therapy. Of note, this statement remains true with the emergence of PARPi. Indeed, in the PAOLA-1 trial, BRCA1/2-mutated low-risk patients, i.e, patients with FIGO stage III disease who underwent macroscopic complete primary surgery, had a greater benefit of bevacizumab and olaparib maintenance therapy as compared to BRCA1/2-mutated higher-risk patients, i.e., patients with stage IV disease or with residual disease after upfront surgery or who received neoadjuvant chemotherapy (HR = 0.11 [95% CI = 0.03–0.31] and HR = 0.37 [95% CI = 0.23–0.59], respectively) (82). Moreover, in the case of oligoprogression under PARPi, surgery, as well as other local therapies and the continuation of PARPi, may have a role in overcoming PARPi resistance. Recently, a retrospective analysis from Gauduchon et al. showed that PARPi prolongation after local therapy for oligometastatic progression offers 11.5 months of PFS (83). Palluzi et al. also demonstrated a median prolongation of the treatment-free interval without platinum of 6 months and 10 months after surgery or stereotactic body radiotherapy, respectively, and PARPi continuation for patients with ovarian cancers experiencing oligoprogression under PARPi (84). Thus, surgery plays a major role in disease management.

PARP modulates the expression of genes involved in angiogenesis, with a particular impact on HIF-1a, which plays a significant role in tumor progression by orchestrating a comprehensive response to hypoxia (85). On the other hand, induction of hypoxia by antiangiogenic agents demonstrated a decrease in HR pathways’ efficiency through the downregulation of key genes such as BRCA1/2 and RAD51 (86). Thus, several preclinical data suggest the potential synergistic effect of PARPi in combination with antiangiogenic agents. In the first-line setting, the PAOLA-1 phase III clinical trial demonstrated a noteworthy improvement in PFS and OS of olaparib and bevacizumab first-line combination maintenance therapy versus bevacizumab alone in BRCA1/2-mutated patients and the HRD population (6, 87). Although the absence of an olaparib maintenance arm does not allow direct and strong conclusions, several data points suggest the potential benefit of the addition of bevacizumab to prolonge olaparib efficacy and therefore overcome PARPi resistance. First, a population-adjusted indirect treatment comparison was conducted that pooled patients from SOLO-1, assessing first-line olaparib maintenance therapy, and PAOLA-1 clinical trials. Results showed a numerical improvement in PFS in favor of bevacizumab and the olaparib arm as compared to olaparib alone in the BRCA1/2-mutated population (HR = 0.71 [95% CI = 0.45–1.09]) (88). Furthermore, while olaparib monotherapy as first-line maintenance seemed to have greater efficacy in BRCA2- compared to BRCA1-mutated patients in the SOLO1 trial (HR = 0.20 [95% CI = 0.10–0.38] and HR = 0.40 [95% CI = 0.29–0.56], respectively) (34), similar efficacy was observed in the PAOLA-1 trial with bevacizumab and olaparib combination (HR = 0.22 [95% CI = 0.09–0.54] and HR = 0.26 [95% CI = 0.16–0.41]) (29), highlighting the likely role of the combination in less sensitive BRCA1-mutated tumors. Moreover, in the high-risk (i.e., patients with residual disease after primary debulking surgery, neoadjuvant chemotherapy, or FIGO stage IV disease) BRCA1/2-mutated population from the PAOLA-1 trial, only 15% of patients experienced a progression within the first 15 months of olaparib and bevacizumab maintenance therapy (82). However, in the PRIMA trial, including high-risk patients comparable to the high-risk population from the PAOLA-1 trial, 34% of patients had progressed within the first 15 months of niraparib maintenance monotherapy (18). The same comparison can be done between BRCA1/2-mutated patients from the PAOLA-1 trial, including high- and lower-risk patients, and the BRCA1/2-mutated population from the ATHENA-mono trial, which assessed rucaparib in monotherapy as first-line maintenance. In the PAOLA-1 trial, a disease progression was observed for 12% of the patients within the first 15 months of combination maintenance therapy (87), while 23% experienced a disease progression with rucaparib monotherapy at the same time in the ATHENA-mono trial (17). However, these comparisons should be considered carefully, and the hypothesis must be confirmed with a prospective randomized trial. To this end, the ongoing NIRVANA and AGO-OVAR 28/ENGOT-ov57 trials assessing niraparib and bevacizumab first-line maintenance therapy versus niraparib alone address this question (89, 90). Also, the NIRVANA-R phase II single-arm study assessing the potential of combining niraparib and bevacizumab as maintenance therapy in patients with recurrent platinum-sensitive ovarian cancer who already received PARPi therapy (NCT 05183984) will provide more information about the role of bevacizumab in reversing PARPi resistance. In recurrent platinum-sensitive ovarian cancers, Mirza et al. reported a benefit in PFS with niraparib and bevacizumab compared to niraparib alone in intention-to-treat populations as well as in HRD and HRP (homologous recombination proficient). Patients (HR = 0.35 [95% CI = 0.21–0.57], HR = 0.38 [95% CI = 0.20–0.72], and HR = 0.40 [95% CI = 0.19–0.85], respectively) (91). In addition to bevacizumab, other antiangiogenics have been assessed in a recurrence setting, although less convincing. While the NRG-GY004 phase III trial assessing olaparib versus olaparib and cediranib versus chemotherapy in platinum-sensitive relapsed ovarian cancer patients did not show any difference between treatment arms in the whole population, subgroup analyses suggested the efficacy of the combination therapy in BRCA1/2-mutated patients. Indeed, the median PFS was 10.5 months, 18.0 months, and 12.7 months with chemotherapy, olaparib and cediranib, and olaparib alone, respectively, in patients with BRCA1/2 mutant tumors (92). However, the EVOLVE study assessing olaparib and cediranib combination in HGSC patients after progression on PARP inhibitors was disappointing, with only 9% of ORR (three out of 34 patients). Of note, patients with BRCA1/2 reversion mutations correlated with poor outcomes and did not benefit from the experimental combination (61), highlighting the restricted role of that combination in recurrence settings and suggesting the importance of using antiangiogenic with PARPi in earlier stages with the likely potential to delay PARPi resistance. Importantly, more translational research works are needed to understand how it might impact, if so, the occurrence of BRCA1/2 reversion mutations.

PIK3/AKT/mTOR pathway inhibition has also been associated with a decrease in the expression of BRCA1/2 genes and impaired HR pathway activity, leading to a BRCAness profile (93, 94). Two phase I clinical trials investigating the efficacy of PI3KCA inhibitors in combination with olaparib showed promising results with an ORR of more than 30% (78). However, of the 10 partial responses, six occurred in BRCA1/2 wild-type patients, while only three occurred in BRCA1/2 mutation carriers, leaving a doubt about efficacy in BRCA1/2 mutation carriers. Currently, a phase III randomized study, EPIK-O (NCT04729387), is undergoing assessment to evaluate the combination of olaparib and alpelisib versus standard chemotherapy in platinum-resistant ovarian cancer patients (95). Beyond antiangiogenic agents, targeting other protumorogenic pathways such as RAS/RAF/Mek pathways (96, 97) or targeting genes involved in epigenetics such as bromodomain-containing 4 (BET) (98) might also enhance PARPi efficacy. An ongoing phase I/II (NCT03162627) trial assessing olaparib and selumetinib, a MEK inhibitor, also includes an expansion cohort of PARPi-resistant ovarian cancers.

Inhibition of the ataxia-telangiectasia and Rad-3 (ATR) kinases has garnered significant attention in the context of HGSC. This interest arises from the central role that ATR plays in the DNA damage response as well as in response to replication stress (99). ATR, along with its downstream effectors, checkpoint kinase 1 (CHK1) and WEE1-like protein kinase (WEE1), instigate cell cycle arrest and collaborate to rectify halted replication forks (100–102). Of note, some activity of ATR or WEE1 inhibitors in combination with PARPi in patients who are resistant to PARPi has been shown (80, 81, 103). More importantly, the underlying mechanism is mainly related to the restoration of replication fork stability. Nevertheless, the synergistic effect of ATR and PARPi has been observed across a wide range of PARPi and platinum-resistant models, including PDX models, that harbor different genetic alterations responsible for PARPi resistance, including the BRCA1/2 reversion mutation (104).

MMEJ serves as a compensatory mechanism for repairing DSBs in the absence of HR. In addition, we previously showed that MMEJ pathways might drive BRCA1/2 reversion mutations, which makes it a preferential target to overcome PARPi resistance and the occurrence of reversion mutations. Therefore, the inhibition of POLθ, an essential enzyme involved in MMEJ repair pathways, has gained specific interest (105, 106). The antibiotic novobiocin (NVB) is a specific inhibitor that binds to the ATPase activity of POLθ, leading to the inhibition of MMEJ repair. Both in vitro and in vivo evidence has demonstrated that NVB-mediated POLθ inhibition leads to synthetic lethality in HR-deficient tumor cells (107, 108). Recently, ART558, a small POLθ inhibitor, also induced DNA damage and synthetic lethality in BRCA1/2 mutant tumor cells (109). However, the antitumor activity has been demonstrated in PARPi-resistant tumors driven by the defect of 53BP1/Shieldin complex, and efficacy in tumors that acquired BRCA1/2 reversion mutation or the ability to delay reversion mutations still needs to be assessed. Also, the first oral-specific POLθ inhibitor, ART4215, is currently undergoing evaluation in a phase I/II trial, either alone or in combination with a PARPi (NCT04991480).

Therapeutic strategies combining PARPi and immunotherapy have raised interest in BRCA1/2-mutated ovarian cancers. Indeed, BRCA1/2-mutated tumors exhibit a higher mutational load, an increased amount of neoantigens, tumor-infiltrating lymphocytes, and PD-L1 expression (110, 111). Single-cell analyses have recently revealed a higher immune infiltration by CD8+PD-L1+ T cells and a heightened co-occurrence of T cells, B cells, and antigen-presenting cells in BRCA1/2-mutated tumors, advocating for a more intricate collaboration of the immune system components compared to their wild-type counterparts (112). Interestingly, a recent pan-cancer research work also demonstrated an increase in T cells, natural killers, macrophages, and dendritic cells in BRCA2-mutated tumors, whereas BRCA1 mutant tumors exhibited a higher presence of myeloid suppressive cells. Importantly, an enhanced OS with immune checkpoint inhibitors was noted in BRCA2 mutation carriers compared to those with BRCA1 mutations (113). In the prospective phase Ib/II TOPACIO/KEYNOTE-162 trial, the combination of niraparib with pembrolizumab was evaluated in patients with recurrent platinum-resistant ovarian cancers. The ORR was 18%, and the DCR reached 65%. However, ORR did not differ between patients harboring BRCA1/2 mutations or not, and among eight patients who experienced long-term response, five were treated for a BRCA1/2 wild-type tumor (114). The MEDIOLA phase II clinical trial investigated the combination of olaparib and durvalumab in BRCA1/2-mutated platinum-sensitive recurrent ovarian cancer patients. The ORR was 72%, and the median PFS was 11.1 months (115). However, the ARIEL2 trial assessing rucaparib in monotherapy in platinum-sensitive recurrent ovarian cancer patients showed a median PFS of 12.8 months and an ORR of 80% in the BRCA1/2-mutated population (116) This suggests that the addition of durvalumab might have only a limited impact, if any, on top of PARPi therapy. Of note, the JAVELIN PARP 100 phase III trial (NCT03642132), aiming to assess avelumab in combination with chemotherapy followed by maintenance avelumab and talazoparib in first-line ovarian cancer patients, closed for futility after interim analysis. The DUO-O phase III randomized trial (NCT03737643) assessing durvalumab in association with chemotherapy and bevacizumab and continuing with bevacizumab and olaparib as maintenance therapy in a first-line setting is ongoing to further answer these questions.