Abstract
The maintenance of a pristine genome, free from errors, is necessary to prevent cellular transformation and degeneration. When errors in DNA are detected, DNA damage repair (DDR) genes and their regulators are activated to effect repair. When these DDR pathways are themselves mutated or aberrantly downregulated, cancer and neurodegenerative disorders can ensue. Multiple lines of evidence now indicate, however, that defects in key regulators of DNA repair pathways are highly enriched in human metastasis specimens and hence may be a key step in the acquisition of metastasis and the ability of localized disease to disseminate. Some of the key regulators of checkpoints in the DNA damage response are the TP53 protein and the PARP enzyme family. Targeting of these pathways, especially through PARP inhibition, is now being exploited therapeutically to effect significant clinical responses in subsets of individuals, particularly in patients with ovarian cancer or prostate cancer, including cancers with a marked metastatic burden. Targeting DNA repair–deficient tumors with drugs that take advantage of the fundamental differences between normal repair–proficient cells and repair-deficient tumors offers new avenues for treating advanced disease in the future. Clin Cancer Res; 22(13); 3132–7. ©2016 AACR.
Background
The DNA damage response pathway
One of the hallmarks of the human cancer genome is the prevalence of apparently random aberrations in the normal order of genetic information on the chromosomes. This genomic instability is directly linked to the acquired loss of function in any one of six DNA damage repair (DDR) pathways (Fig. 1), including those of mismatch repair (1, 2), homologous recombination repair (3, 4), nonhomologous end joining (5–7), translesion DNA synthesis (8), base excision repair (9), or nucleotide excision repair pathways (10–12).
DNA damage repair pathways and regulators. The main nuclear DNA damage repair (DDR) pathways are depicted, along with regulators of their function. Target sites for therapeutic intervention via PARP, the ATM/TP53, and immune checkpoint inhibitors are also shown. Modifiers of DNA damage repair pathways, in particular the DNA damage checkpoint control TP53/ATM/ATR pathway, which can modulate DDR function, are also depicted. BER, base excision repair; HMR, homologous recombination repair; MMR, mismatch repair; NER, nucleotide excision repair pathways; NHEJ, nonhomologous end joining; TLS, translesion DNA synthesis.
DNA damage repair pathways and regulators. The main nuclear DNA damage repair (DDR) pathways are depicted, along with regulators of their function. Target sites for therapeutic intervention via PARP, the ATM/TP53, and immune checkpoint inhibitors are also shown. Modifiers of DNA damage repair pathways, in particular the DNA damage checkpoint control TP53/ATM/ATR pathway, which can modulate DDR function, are also depicted. BER, base excision repair; HMR, homologous recombination repair; MMR, mismatch repair; NER, nucleotide excision repair pathways; NHEJ, nonhomologous end joining; TLS, translesion DNA synthesis.
The molecular machinery governing these processes is both detailed and complex and beyond the scope of this review. We refer the reader wishing a more thorough exposition of the biochemistry of DNA repair to a number of excellent recent reviews (13–16). Instead, we focus on those DNA repair mechanisms and key proteins that have been specifically linked to advanced disease and metastasis, such as DNA damage checkpoint control and TP53, as well as other repair proteins, such as PARP, that are exploitable therapeutically. We then explore the potential for clinically targeting these pathways, as well as their effect on cellular function (in particular the generation of neoantigens) to impede disease progression and improve patient survival.
DNA damage checkpoints and TP53
Once DNA damage has occurred, it is vital that the cell does not proceed through the S-phase of the cell cycle and permanently “fix” damage in daughter cells. To prevent this from happening, once DNA damage has been detected, cell-cycle checkpoints are activated, orchestrated by two master regulators, ATM and ATR, kinases that phosphorylate key checkpoint effector proteins leading to cell-cycle arrest. Loss of function or imbalance in either ATM or ATR, as well as in some key mediators such as BRCA1, has been linked to aggressive and metastatic tumors in preclinical models of breast cancer and in clinical prostate cancer (17–19), underlying the importance of cell-cycle checkpoint pathways in the spread of tumor cells. One important downstream target of ATM/ATR is TP53, the key effector of apoptosis and senescence following DNA damage. Loss of TP53 function renders cells unable to induce apoptosis or senescence programs in response to DNA damage and therefore primes these cells for transformation. It is not surprising then that mutation in TP53 is the most commonly occurring mutation in human cancers with more than 50% of all tumors having aberrations in the TP53 gene. However, it has become increasingly apparent that distinct defects in TP53 produce differing cellular effects. For instance, the metastatic potential of tumors is associated with missense mutations in the DNA-binding domain, which are predicted to result in “gain of function,” as opposed to the more frequently observed loss-of-function mutations that lead to a hypofunction or absence of TP53 in cells. For instance, although TP53-null mice do form tumors, they rarely metastasize or display an invasive phenotype (20, 21). Conversely, mice expressing missense mutations in the “hotspot” DNA-binding domain of the TP53 protein display a markedly higher incidence of metastatic carcinomas and osteosarcomas (22, 23). The presence of TP53 mutations confers a poor prognosis for patients with breast cancer, a surrogate for metastasis formation (24). Recent work also suggests that the late acquisition of missense TP53 mutations in subclonal populations of tumor cells is also the driver of metastatic expansion in clinical prostate cancer (25).
Regulating DNA repair pathways: The PARP family
The PARP family of posttranslational modifying enzymes regulates protein function and cofactor binding by catalyzing the covalent attachment of one or more ADP–ribose units to client substrates (26). Three family members (PARP 1–3) play a critical role in DNA repair. Following single- and double-strand breaks (DSB), PARP proteins bind to damaged DNA and promote ADP-ribosylation of both themselves and a number of other chromatin proteins, thereby activating either base excision repair or homologous recombination repair (27). In addition, poly ADP-ribosylated PARP blocks access to free DNA ends by proteins involved in the error-prone nonhomologous end-joining repair of DSBs. The PARP proteins are also involved in the repair of single-strand breaks; however, if PARP function is impeded, then persistent single-strand breaks can lead to the formation of replication cycle–induced DSBs (14). Such inhibition in the presence of a preexisting DNA repair defect can result in catastrophic genomic instability and cellular demise and is discussed further below.
DNA repair pathway aberrations in human metastases
Although a number of individual studies have reported the prevalence of DNA repair pathway aberrations in metastatic samples from individual datasets (25, 28–30), to date, no systematic analysis across large combined metastatic datasets has been reported. To address this issue, we screened DNA copy number and mutation data (31) across 6 different human metastasis datasets comprising in total 317 metastatic samples across 3 different tumor types, namely, melanoma (32), colorectal cancer (33), and prostate cancer (25, 28–30). We used a comprehensive list of 180 DNA repair pathway genes across 18 different categories (34), which are either directly involved or act as modifiers of DNA repair protein function. In a very high proportion of the metastatic specimens, at least one alteration in a DNA repair gene was identified (Fig. 2). Across the tumor types, 70% of prostate cancer metastases exhibited a deletion, amplification, or point mutation in a DNA repair pathway gene, and this proportion was even higher for the melanoma (75%) and colorectal (82%) metastases (Fig. 2). There were direct parallels in the spectrum of DNA repair pathway genes undergoing point mutation in prostate cancer with those seen in the colorectal cancer and melanoma datasets. Across the entirety of the metastatic datasets, mutations in the TP53 gene dominated, with 65% of the metastases harboring point defects in this gene. The next most frequently mutated target gene was the BRCA2 gene, at 5.3%. Prostate cancer metastases also exhibited a high frequency of DNA rearrangements, with 70% exhibiting genomic instability at a DNA repair pathway locus. Strikingly, 76% of all point mutations in the prostate metastases were in the TP53 gene, with the next highest, the ATM gene, a distant second at 3.5% mutation frequency. These results confirm that there is a very high rate of aberrations in DNA repair genes in metastases across three common human cancer types and that a marked enrichment in metastases over localized primary cancers. These findings strongly suggest that clinical strategies that target mutant TP53, destabilize PARP repair function, or can exploit the “tumor neoantigens,” which are formed as a consequence of a high somatic mutation rate, may impede the distant dissemination of cancer cells.
DNA repair aberrations in human metastases. A, classes of aberrations in 180 DNA repair genes in human metastases across three tumor types. B, classes of aberrations observed in the TP53 gene in the same cohorts. Prostate cancer cohort comprises 223 metastases across four studies. Melanoma cohort comprises 25 metastases. Colorectal cancer comprises 69 metastases. Amplification refers to the percentage of cancer patients with an amplification of one or more DNA repair genes. Deletion and mutation refer to homozygous deletions and point mutations. Multiple alterations indicate the proportion of patients presenting with two or more of the previous classes.
DNA repair aberrations in human metastases. A, classes of aberrations in 180 DNA repair genes in human metastases across three tumor types. B, classes of aberrations observed in the TP53 gene in the same cohorts. Prostate cancer cohort comprises 223 metastases across four studies. Melanoma cohort comprises 25 metastases. Colorectal cancer comprises 69 metastases. Amplification refers to the percentage of cancer patients with an amplification of one or more DNA repair genes. Deletion and mutation refer to homozygous deletions and point mutations. Multiple alterations indicate the proportion of patients presenting with two or more of the previous classes.
Clinical–Translational Advances
DNA repair and PARP inhibition
A burgeoning set of data is implicating the acquisition of metastatic potential with the onset of DNA damage repair deficits in tumors. If this premise holds true, the possibility exists of targeting these tumor-specific repair deficits much earlier in the disease process to impede the spread of cancer cells to distant sites before frank metastatic lesions become clinically detectable. Current strategies however, are focusing on three cancer-associated defects of the DNA repair pathway in patients with frank metastatic lesions. One of these strategies showing considerable clinical promise is targeting the key DNA repair protein, PARP. In systems that harbor a preexisting DNA repair defect, typically in mutations in the BRCA1 or BRCA2 genes, DSBs accumulate, and administration of PARP inhibitor drugs can precipitate fatal genomic instability, a feature termed “synthetic lethality” (35, 36).
Certainly, clinical evidence to date supports the concept of synthetic lethality. For instance, in a phase II study of the PARP inhibitor olaparib in high-grade serous ovarian cancer, patients with either germline or somatic BRCA1/2 mutations experienced significantly greater progression-free survival (PFS) than those with wild-type BRCA1/2, with extended benefit observed in a large subset (37). Interestingly, clinical responses, including prolonged PFS, were also observed in the wild-type group, raising the possibility that these women had other defects in DNA repair that were uncharacterized at the time of treatment. In this regard, a recent study has shown that stratification of patients on the basis of their underlying DNA repair status can also have an impact on predicting their response to traditional chemotherapy (38). In a large cohort of patients with primary ovarian cancer, those having either germline or somatic loss-of-function mutations in any one of 13 genes in the homologous recombination pathway had higher rates of response to platinum chemotherapy as well as improved overall survival (OS; ref. 38). Across the cohort of 367 patients, 31% had a detectable germline or somatic mutation in one of the 13 homologous recombination DNA repair genes. As expected, the majority of these were in BRCA1 or BRCA2 (74%); however, 26% occurred in the other 11 genes, and importantly, these patients also had higher rates of response to platinum and improved OS (38). This study highlights the clinical utility of performing targeted capture and deep sequencing of DNA repair gene panels for selecting patients for traditional chemotherapy as well as for PARP inhibitor trials. PARP inhibitors have also brought responses in patients with BRCA1/2-mutated breast cancer (39), although less consistently than ovarian cancer, as well as in subsets of patients with prostate and pancreatic carcinomas (40). Evaluation is ongoing in a number of tumor types, including open phase III trials in ovarian cancer (NCT02446600, NCT02502266, NCT02470585) and in breast cancer (NCT02163694, NCT01945775) and also in patients with solid tumor metastases (NCT01366144; ref. 27). In a recent landmark trial, patients with advanced prostate cancer, molecularly stratified on the basis of defined DNA repair gene defects and treated with the PARP inhibitor olaparib, had high response rates (41). Not all patients with BRCA1/2 mutations respond favorably, due to secondary mutations that restore function or compensation via other pathways. In addition, it is clear that some patients without BRCA1/2 mutations do derive clinical benefit, and given that the full repertoire of proteins involved in DNA repair may not be fully characterized, assays that target broad panels of DNA repair genes and their regulators will likely have the most clinical utility in stratifying patients for therapy (38). The relative frequency of multiple aberrations in human metastases, especially those from prostate and ovarian cancer, suggests that there may be further scope to use synthetic lethality approaches to treat metastases other than those having just BRCA1/2 or mismatch repair defects. Drugs that inhibit ATM are capable of inducing synthetic lethality in vitro in the presence of mutations in other DNA damage repair genes, including TP53, BRCA1, RAD50, XRCC1, MRE11A, and FANCD2 (42). In addition, ATM-deficient cancers have also been shown to be susceptible to inhibition of ATR (42). Non–small cell lung cancers deficient in both ATM and TP53 showed high sensitivity to ATR inhibition in a preclinical setting (42). This targeting strategy is in the early trial stage in the clinical setting, with two ATR inhibitors in open phase I clinical trials for patients with advanced solid tumors, AZD6738 (NCT02223923) and VX-970 (NCT02157792).
Clinical targeting of mutant TP53
The function of TP53 is tightly regulated by targeting the protein for degradation via the E3 ubiquitin ligase, MDM2. In normal, unstressed cells, this ensures low levels of TP53 are maintained. A variety of cellular stresses lead to stabilization of TP53 and induction of either senescence or apoptosis, which act as safeguards against transformation. Tumors circumvent these safeguards by inactivating or redirecting the function of TP53. In addition to mutation and deletion, TP53 is inactivated by amplification of the negative regulator MDM2 in 7% of all human cancers (43). Development of clinical agents targeting TP53 has mainly focused on inhibition of MDM2 function, restoration of wild-type TP53 activity to mutants, and disruption of the p53-mutant activity (44).
Up to seven different inhibitors of MDM2 are currently in early phase I clinical trials (45). Encouragingly, in a number of cases, these compounds have been shown to induce TP53 expression in tumors, induction of apoptosis, and stabilization of disease.
Compounds that restore wild-type function to mutant TP53, such as PRIMA-1met, are thought to bind and stabilize the mutant protein in the wild-type conformation (46). In a phase I clinical trial, this agent showed some activity in a patient with acute myeloid leukemia with a TP53 mutation (A355V; ref. 46).
Tumors can become addicted to mutant TP53 oncogenic gain-of-function activities. To circumvent the decreased protein stability exhibited by some of these mutants, stabilizers like the HSP90/HDAC6 chaperones are often upregulated in tumors. Mice lines with the TP53 mutations R248Q and R172H that were treated with the HSP90 inhibitor ganetespib were protected against the development of T-cell lymphoma normally seen in these strains (47). This was associated with degradation of TP53 and induction of tumor apoptosis (47).
Neoantigens and immune checkpoint inhibitors
Recent insights into the mechanisms of action of immune checkpoint inhibitors clearly show that high mutational loads, a common result of defects in DNA repair mechanisms, particularly those of mismatch repair involving the MSH2/6 and MLH1/2 and PMS2 genes, render tumors intrinsically vulnerable to activation of the adaptive immune response (Fig. 1; refs. 48–50). Transcription of exons bearing nonsynonymous mutations may give rise to novel sequences of amino acids, which, when fragmented and presented by the MHC class I or II histocompatibility complex to the T-cell receptor, are recognized as “non-self,” leading to T-cell activation. These neoepitopes, termed “neoantigens,” are frequently derived from proteins that are not necessary for tumor progression (passenger mutations) but are nonetheless abundantly expressed (51, 52). Tumors with a higher mutational load are therefore more likely to generate neoantigens, and from first principles more likely to derive benefit from immune checkpoint inhibitors, such as those that target CTLA-4 [e.g., ipilimumab (53)], PD-1 [e.g., pembrolizumab (54)], or the ligand PD-L1 [(e.g., MPDL3280A (48)] and which are now approved for clinical use in a number of tumor types.
The tumors that show the greatest clinical benefit, including melanoma, non–small cell lung cancer, and urothelial carcinoma of the bladder, have a higher rate of somatic mutations compared with that of other common tumor types (55). In addition, a recent phase II study in treatment-refractory colorectal cancer patients with either germline (Lynch syndrome) or somatically acquired mismatch repair defects demonstrated a significantly better tumor response rate to the anti–PD-1 inhibitor pembrolizumab than matching patients with repair-proficient tumors (56). As expected, patients with mismatch repair–deficient tumors had a significantly higher rate of somatic mutations, which was associated with a longer PFS. An open phase III trial will now test the clinical efficacy of this approach (NCT02563002). Taken together, these findings suggest that metastases associated with acquired defects in DNA repair mechanisms may be more sensitive to immune checkpoint inhibitors than paired primary tumors.
Although in general, clinical response correlates with mutational burden, not all patients with a high somatic mutation benefit from treatment, indicating that genomic alterations alone are not sufficient to predict tumor response. Recent investigations have identified that although mutation rate increases the probability of neoantigen formation, the amplitude of immune response is also influenced by a number of other factors, including peptide abundance, processing efficiency, MHC affinity, as well as orientation and position of the neoepitope within the presented peptide fragment (57). Experiments in which immunogenic peptides presented on MHC I have been positively identified by mass spectrometry suggest that neoantigens may be patient rather than tumor specific, so predictive tests may need to be individualized (57). A number of algorithms have been developed to predict neoantigen formation based on both exome and transcript sequencing, although their utility remains to be demonstrated in prospective studies (50, 57, 58).
Conclusions and Future Directions
Fundamental advances in our basic understanding of DNA repair processes have now permitted the rational targeting of these pathways in cancer and have already led to measurable clinical responses in patients with defined perturbations in these pathways.
As our understanding of how the complex DNA repair machinery detects and coordinates the repair of genetic aberrations advances, new opportunities for perturbing this process to impede cancer dissemination and frank metastasis formation will likely arise. It may eventually be possible to tailor therapeutic combinations based on the underlying DNA repair defects to block distant tumor spread, rather than attempt to eradicate all tumor cells with cytotoxic approaches. This raises the exciting possibility of one day transforming advanced disease into a chronic condition rather than a lethal one.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: N.M. Corcoran, R. Stuchbery, C.M. Hovens
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.M. Corcoran, C.M. Hovens
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.M. Corcoran, R. Stuchbery, C.M. Hovens
Writing, review, and/or revision of the manuscript: N.M. Corcoran, M.J. Clarkson, R. Stuchbery, C.M. Hovens