Cells carrying mutated BRCA1 or BRCA2 genes are defective in DNA repair by homologous recombination and, as a consequence, are highly sensitive to inhibitors of poly (ADP-ribose) polymerase (PARP). This provides the basis for a novel “synthetic lethal” approach to cancer therapy. We have recently shown that this sensitivity can be reversed, and resistance to PARP inhibition can be acquired by deletion of a mutation in BRCA2. Furthermore, a similar mechanism seems to be associated with carboplatin resistance in some BRCA2 mutation carriers with ovarian cancer. [Cancer Res 2008;68(24):10021–3]
- PARP inhibitor
Carriers of heterozygous germ-line mutations in the BRCA1 or BRCA2 genes are strongly predisposed to cancer of the breast, ovary, and other organs ( 1). BRCA1 and BRCA2 proteins are critically important for the repair of double strand breaks (DSB) by homologous recombination (HR) and loss of the wild-type BRCA1 or BRCA2 allele in tumors likely fosters cancer progression by promoting genome instability and mutation ( 2). We have developed therapeutic strategies based on exploiting this specific DNA repair defect. First, following up on observations that BRCA1- or BRCA2-deficient cells in culture were very sensitive to DNA cross-linking agents, we have established a clinical trial to test the hypothesis that tumors arising in BRCA1 and BRCA2 mutation carriers show enhanced sensitivity to carboplatin ( 3). Our second approach has used the principle of synthetic lethality. Synthetic lethality is defined as the situation when mutation in either of two genes individually has no effect but combining the mutations leads to death ( 4). This effect was first described over 60 years ago in studies on the fruit fly ( 5). The implication of synthetic lethality for therapeutic strategies is that targeting one such genes in a cancer where the other is defective should be selectively lethal to the tumor cells but not toxic to the normal cells. In principle, this should lead to a large therapeutic window ( 4).
We and others showed that BRCA1 or BRCA2 deficiency, as well as deficiency in other HR components, profoundly sensitizes cells to killing by potent inhibitors of the enzyme poly(ADP-ribose) polymerase (PARP), such as KU0058948 ( 6– 8).
We hypothesized that inhibition of PARP would lead to the persistence of DNA lesions that would normally be repaired by BRCA2-mediated HR. Therefore, PARP inhibitors (PARPi) could be selectively lethal to cells lacking functional BRCA1 or BRCA2 but be associated with minimal toxicity to normal cells. This synthetic lethal interaction between PARP and BRCA1 or BRCA2 suggested a potential new mechanism-based approach for the treatment of patients with BRCA1 or BRCA2 mutation–associated cancers. In these patients, tumor cells lack normal BRCA1 or BRCA2 function, but normal tissues retain a single normal copy of the relevant gene potentially providing a large “therapeutic window” and reduced side effects. A number of clinical trials have been developed to test the efficacy of this approach. The potent PARPi KU-0059436/AZD2281 (AstraZeneca) is currently being tested in a phase I trial. Preliminary observations in BRCA mutation carriers with ovarian cancer suggest low toxicities with some promising indications of response measured radiologically and using tumor markers ( 9). Phase II trials of KU-0059436/AZD2281 as a single agent for the treatment of breast and ovarian cancers in BRCA1 or BRCA2 mutation carriers are now also under way, and a number of other PARPi are in clinical development ( 10). Although these studies are still in their infancy, it seemed likely that some tumors may be de novo resistant to PARP inhibition or that they may acquire resistance during treatment. To model this, we derived PARPi-resistant clones from the human BRCA2 deficient cell line, CAPAN1 ( 11).
CAPAN1 cells were originally derived from a pancreatic epithelial cancer occurring in an individual carrying the protein truncating frame-shifting c.6174delT BRCA2 mutation. These cells have lost the wild-type copy of the BRCA2 gene but express a truncated protein of 2,002 amino acids lacking 1,416 amino acids compared with the wild-type protein. This truncated protein is missing important functional domains, and as a result, CAPAN1 cells are defective in the ability to form DNA damage–induced RAD51 foci and HR and are extremely sensitive to treatment with potent PARPi ( 12). To generate cells resistant to PARP inhibition, we exposed CAPAN1 cells to the potent, drug-like, PARPi KU0058948 ( 6) and produced multiple independent PARP inhibitor–resistant (PIR) clones ( 11). These were all >1,000-fold resistant to PARPi compared with the parental cell line and had acquired the ability to form damage-induced RAD51 nuclear foci and limit genomic instability, both hallmarks of a restored HR pathway. We, therefore, examined the structure of the BRCA2 gene and protein expressed in the PIR clones. This revealed the unexpected finding that each of these clones carried different BRCA2 alleles harboring deletions of up to 58 kb, all of which result in elimination of the c.6174delT mutation. Each deletion had different consequences on the encoded protein, but all restored the open reading frame (ORF) so that the COOH terminus of the protein was produced in frame with the NH2 terminus ( 11). One allele produced a protein with a small, 153 amino acid deletion, but other alleles encoded proteins with major (>1,400 amino acid) deletions encompassing 3 RAD51 binding BRC repeats and either all or most of the proposed DNA Binding Domain (DBD). Notably the restored COOH terminus of BRCA2 has recently been shown to contain an important RAD51 binding domain (reviewed in ref. 13).
It was formally possible, although unlikely, that the restored BRCA2 ORF and cellular PARP resistance were coincidental; the novel alleles had arisen under conditions where DNA repair was ineffective and after exposure to PARPi likely causing multiple cellular mutations which could have been responsible. Therefore, we performed a series of experiments to link the novel alleles to PARPi resistance ( 11). These included resensitising PIR clones to PARPi with BRCA2 siRNA and re-expressing the variant BRCA2 proteins in PARPi-naïve CAPAN1 cells and conferring resistance. Finally, we were able to show that these truncated proteins were fully competent, at least in the assay we used, in facilitating HR. Taken together, these observations implicate the novel BRCA2 proteins present in PIR clones in resistance to PARPi and, importantly, suggest that this resistance occurs by restoration of DNA repair by HR. Therefore, we consider these gain-of-function mutants of BRCA2 to be, at least in part, reversion alleles ( 11).
The use of potent PARPi for the treatment of BRCA-associated cancer is still in the early stages of assessment ( 9), and it is not yet possible to assess mechanisms of resistance in patient material. However, ovarian cancer patients who carry BRCA1 or BRCA2 mutations have long been treated with agents such as cisplatin and carboplatin, which as discussed above are thought to exert their BRCA-selective effects by a mechanism similar to that of PARPi. Several reports suggest that ovarian cancer patients who carry BRCA1 or BRCA2 mutations gain more benefit from carboplatin compared with patients with no family history of the disease, but resistance does eventually occur ( 14). Numerous mechanisms of platinum drug resistance have been proposed ( 15, 16). To investigate the mechanism in BRCA2 mutation carriers, we determined the sequence of the BRCA2 gene in tumor material from two patients bearing the BRCA2 c.6174delT mutation, whose ovarian carcinomas progressed on carboplatin treatment. In both cases, deletions were present downstream of the c.6174delT mutation, which restored the BRCA2 ORF. Taken together with our observations in PIR cells, this links these reversion alleles with clinical resistance to carboplatin. Similar observations were made by Taniguchi's group ( 17).
Over the past few years, resistance to targeted therapy has been well-documented. The prime example of this is resistance to imatinib (gleevec) in chronic myelogenous leukemia, a disease caused by expression of the BCR-ABL fusion protein ( 18). Here, resistance occurs via mutation of the drug-binding pocket of the ABL protein kinase so that imatinib can no longer bind and, as a consequence, kinase activity cannot be inhibited. The mutations we observe are different from this paradigm in a number of ways. First, the reversion alleles we described result in the restoration of a previously lost function rather than by blocking drug binding. Second, they constitute deletions, in some cases of very large size, rather than missense mutations. Finally, mirroring our synthetic lethal strategy, the resistance inducing mutation is in the synthetic lethal partner (BRCA2) rather than in the drug target itself (PARP).
Our observation that variants of BRCA2 lacking a significant fraction of the protein including putative functionally important domains are competent in mediating PARP resistance, forming RAD51 foci, and mediating HR in the assay we performed, is surprising. Although these domains are clearly not absolutely required for these functions, their strong evolutionary conservation suggests they are important and may be involved in processes such as the fine tuning of DNA repair or in meiosis. Induction of resistance to PARPi in the context of other HR deficiencies might be a means of identifying similar functionally dispensible modules in proteins.
It seems probable that tumors arise in carriers of mutations in BRCA1 and BRCA2, at least in part, due to an inability to maintain genome stability caused a by a defect in a specific DNA repair pathway, HR. The synthetic lethal strategy that we have used exploits this defect, which creates an enormous sensitivity to PARPi. Ironically, resistance to PARPi may be fostered by the same DNA repair defect. In the absence of functional BRCA2, pathways such as single-strand annealing and nonhomologous end-joining compensate and repair DSBs by aligning short regions of homology flanking the DSB, deleting the intervening ( 19, 20). The nucleotide sequences surrounding many of the BRCA2 deletions in the PIR clones and in one of the platinum-resistant ovarian tumors revealed short regions of sequence identity ( 11). Therefore, it seems possible that these deletions arose as a consequence DNA repair dysfunction caused by the BRCA2 mutation. This new mechanism of drug resistance seems to operate under Darwin's principle of natural selection—very rare stochastic genetic events allowing cells to escape from the selection that chemotherapy imposes ( Fig. 1 ). However, the added twist here is that the stochastically occurring reversion mutations may be favored by the underlying DNA repair defect. Therefore, it will be important to study the nature and frequency of the reversion events and the stage of the disease in which they occur. Next Generation Sequencing technologies may allow this issue to be addressed.
We believe that the observations we and others have reported ( 11, 17) may have important clinical ramifications. A key question is, can other mutations in BRCA2 or BRCA1 be reverted in this way and cause resistance to PARPi and to platinum salts? If this is the case, it seems possible that individual mutations will not all revert at the same frequency. For example, large deletions may not be prone to this mechanism of resistance. Therefore, mutation site or nature may be predictive of therapeutic response/time to relapse. Another issue is when should these potentially potent therapeutic approaches be used, given that mechanism based resistance can occur? If as discussed above, these reversion events are indeed stochastic, it means that resistance is statistically much more likely to occur in advanced disease where the challenge is to kill a large tumor mass compared with the adjuvant setting where many fewer cells are present, which can provide the raw material for resistance to develop. Finally, it is intriguing to speculate whether similar mechanisms might underlie resistance to therapy in some sporadic cancers.
Disclosure of Potential Conflicts of Interest
The author is a named coinventor on a number of patents held by AstraZeneca relating to the use of PARP inhibitors. He may benefit from this financially through the Institute of Cancer Research's “Rewards to Inventors” scheme.
I thank Breakthrough Breast Cancer and Cancer Research UK for supporting the research in my laboratory over many years.
- Received June 16, 2008.
- Revision received August 12, 2008.
- Accepted August 31, 2008.
- ©2008 American Association for Cancer Research.