Summary: Reversion mutations associated with PARP inhibitor resistance have been identified in tumors with RAD51C, RAD51D, and PALB2 as well as BRCA1 and BRCA2 mutations. Multiple different reversion mutations can occur in a single patient, and they can be detected by analysis of circulating cell-free DNA. Cancer Discov; 7(9); 937–9. ©2017 AACR.
See related article by Kondrashova et al., p. 984.
See related article by Quigley et al., p. 999.
See related article by Goodall et al., p. 1006.
The approval of multiple PARP inhibitors (olaparib, rucaparib, and niraparib) since 2014 highlights the utility of synthetic lethality as a therapeutic approach. Preliminary observations in germline BRCA1 and BRCA2 (gBRCA1/2) mutation carriers demonstrated that PARP inhibitors in combination with a defect in homologous recombination DNA repair could result in tumor regression. These initial findings were followed by randomized phase III trials of PARP inhibitors in gBRCA1/2 breast cancer and ovarian cancer, establishing significant clinical benefit (1–4), with more limited data demonstrating efficacy in gBRCA1/2 prostate and pancreatic cancers (5, 6). PARP inhibitors appear to be active not only in gBRCA1/2 mutation carriers, but in ovarian cancers in general, particularly in those with somatic BRCA1/2 mutations or homologous recombination deficiency (HRD; refs. 2, 3). Studies in metastatic castration-resistant prostate cancer revealed responses to olaparib in tumors with mutations in genes in the homologous recombination (HR) pathway beyond BRCA1/2, such as ATM and PALB2 (5). Multiple studies are ongoing to define the scope of PARP inhibitor–sensitive tumors.
As exciting as this work has been and although some patients are “long-term responders,” most cancers become resistant to therapy. In gBRCA1/2 breast cancer, median progression-free survival with olaparib treatment is 7.0 months, and even in patients whose tumors responded, the median duration of response is 6.4 months (4). Multiple mechanisms of primary and acquired resistance to both platinum-based chemotherapy and PARP inhibitors have been proposed, including inactivation of 53BP1 or REV7, loss of PARP1 expression, hypomorphic BRCA mutations with partial function, and acquisition of reversion mutations. The relevance and prevalence of each of these mechanisms for resistance in the clinical setting have not been established. However, reversion mutations have been observed in human studies of platinum and PARP resistance.
What is meant by a “reversion mutation”? True reversion refers to restoration of a wild-type sequence. Phenotypic reversion leads to restoration of the open reading frame and thus normal, or at least “normal-enough,” function. Reversion mutations may occur through multiple proposed mechanisms, all of which lead to the expression of protein with some functional capacity. Mutational mechanisms may include large in-frame deletions removing the region of the primary frameshift mutation; deletions or insertions which when found in conjunction with the original mutation restore the reading frame; or splice-site mutations that may, for example, result in a hypomorphic protein without exon 11. In BRCA1/2 mutation– associated tumors with reversion mutations, our current understanding is that enough HR repair function has been restored that treatment with cisplatin no longer is effective and/or PARP inhibitors can no longer work via synthetic lethality. Although reversion mutations were first reported in 2008, studies to date have largely been related to platinum resistance in ovarian cancer, with limited reports of tumors from patients treated with PARP inhibitor therapy.
In this context, three studies in this issue of Cancer Discovery examine reversion mutation in clinical trials of PARP inhibitors in ovarian and prostate cancers and substantially advance the field.
In the first, Kondrashova and colleagues (7) examined 12 patients with recurrent platinum-sensitive ovarian cancer receiving rucaparib on the Ariel 2 trial, and with pretreatment and postprogression biopsies available. Of the 12 tumors, six had truncating mutations in HR genes: four in BRCA1, one in RAD51C, and one in RAD51D. In five of the six patients, one or more reversion mutations that restored the open reading frame were identified in the postprogression tumor biopsies. In the tumor with a RAD51C mutation, four distinct reversion mutations were found. The authors conducted detailed analyses of the RAD51C reversion mutations and present compelling evidence for restoration of HR function.
Goodall and colleagues (8) investigated reversion mutations in the TOPARP-A trial of olaparib in metastatic castration-resistant prostate cancer. Targeted and whole-exome sequencing of circulating cell-free tumor DNA (cfDNA) was performed on serially collected specimens. Decreases in cfDNA concentration were independently associated with outcome. Six subjects had tumors with somatic mutations in HR genes (ATM, BRCA2, and PALB2); all were also detectable in baseline cfDNA. Analyses were performed on postprogression specimens of 10 of 16 patients with an initial response. At time of disease progression, analysis of cfDNA revealed multiple subclonal aberrations reverting germline and somatic DNA repair mutations in BRCA2 (n = 3) and PALB2 (n = 1).
Finally, Quigley and colleagues (9) examined two patients with metastatic prostate cancer with germline mutations in BRCA2 who received PARP inhibitors (one talazoparib and one olaparib). In the first case, two reversion BRCA2 mutations (with deletions of 177 and 66 nucleotides) were identified in a postprogression tumor biopsy, each of which eliminated the baseline pathogenic mutation and restored the open reading frame. Analysis of cfDNA at the same time point detected both of these reversion mutations, as well as five additional alleles with deletions eliminating the pathogenic mutation and restoring the open reading frame. In the second case, cfDNA at the time of tumor progression on olaparib revealed 105 distinct reads with somatic alterations upstream of the germline stop-gain mutation. There were 34 distinct alleles resulting from secondary indels in exons 9 and 10, immediately preceding the germline mutation, all of which were predicted to restore the open reading frame.
What do these studies tell us? First, reversion mutations are not restricted to BRCA1 and BRCA2. Reversion mutations have now been identified in tumors associated with RAD51C, RAD51D, and PALB2 mutations (although each in a single case). Germline mutations in RAD51C and RAD51D are rare and associated with an increased risk of ovarian cancer, whereas germline PALB2 mutations are associated with an increased risk of breast and pancreatic cancers. These studies, in addition to elucidating resistance mechanisms, also provide evidence that the initial mutations in these genes (germline or somatic) are causally associated with response to therapy. In these three cases, the pathogenic mutations in RAD51C, RAD51D, and PALB2 were synthetically lethal with PARP inhibitors leading to a response, and reversion mutations were seen at the time of PARP resistance.
A second, quite striking finding is that many of the patients developed multiple reversion mutations when exposed to PARP inhibitors. These reports complement and extend the findings of Patch and colleagues (10) wherein multiple reversion mutations were seen in two cases of platinum-resistant ovarian cancer. Studies with detailed pre- and post-PARP inhibitor progression analyses (particularly requiring multiple tumor biopsies) are difficult to do from a logistical standpoint and thus the total number of patients for whom such data are available remains very limited. However, current evidence suggests that in the clinical setting in tumors associated with HR mutations, reversion mutations may be the major driver of platinum and PARP resistance. Why might this mechanism of resistance develop and potentially occur multiple times within a single patient? Goodall and colleagues examined the nucleotide sequences flanking the original frameshift mutations in BRCA2 and PALB2 and detected microhomology regions. This finding raises the possibility that defective HR and subsequent reliance of alternative error-prone DNA repair mechanisms may make this phenomenon more likely to occur.
Third, reversion mutations can be detected readily by cfDNA. Indeed, more mutations were found by this approach than by tumor sequencing. Serial tumor biopsies are challenging in the clinic, and even when postprogression biopsies are obtained, it is rare that more than one site is biopsied. Tumor heterogeneity in metastastic cancer is well described, and cfDNA may be advantageous for identifying reversion mutations stemming from multiple sites of tumors.
The importance of undertaking correlative science in clinical trials cannot be overstated. The careful work in these three studies has led to an improved understanding regarding the casual role of RAD51C, RAD51D, and PALB2 in their respective tumors as well as provided additional insight into PARP resistance. Multiple questions remain. What strategies can we use to prevent reversion mutations from developing? If it is confirmed that reversion mutations are generally acquired following exposure to therapy (rather than existing in very rare subclones in the primary tumor), it is possible that administering PARP inhibitors earlier in the course of a patient's therapy, rather than after relapse, will be beneficial. How soon before clinical progression do reversion mutations become detectable in the cfDNA? Will alteration of therapy at that time improve patient outcome compared with waiting for clinical progression? Clinical trials will be needed to demonstrate that early detection of reversion mutations, particularly in patients without clinical progression who are feeling well on therapy, adds value. Preclinical and clinical studies are needed to develop therapeutic approaches when reversion mutations are identified, such as combination and sequencing approaches. Switching from a PARP inhibitor to an alternate treatment (such as standard chemotherapy) at detection of a reversion mutation and back to a PARP inhibitor if and when the revertant clone decreases is one approach. Combination studies of PARP inhibitors with ATR inhibitors, chemotherapy, immune checkpoint inhibitors, and vascular endothelial growth factor inhibitors (such as cediranib) are all under way. A comprehensive approach is necessary to achieve the goal of more successful therapies.
Disclosure of Potential Conflicts of Interest
S.M. Domchek has received honoraria from the speakers bureau of AstraZeneca.