ATM Dysfunction in Pancreatic Adenocarcinoma and Associated Therapeutic Implications

This article is featured in Highlights of This Issue, p. 1897

Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal solid malignancies with fewer than 10% of patients surviving 5 years. Disease incidence is increasing, and PDAC is projected to be the second most common cause of cancer-related death by 2030 (1, 2). The high mortality is due to the majority of patients presenting with locally advanced or metastatic disease at the time of diagnosis. Unfortunately, despite recent improvements in outcomes with newer chemotherapy regimens, the median survival remains less than 1 year (3, 4). Comprehensive genetic analysis is being pursued to identify mutational pathways for potential treatment options, and a recent study utilizing genomic sequencing identified therapeutically relevant (highly actionable) alterations in 27% of PDAC samples (5). These findings were consistent with several other publications, all of which have demonstrated findings of actionable targets in 17% to 48% of PDAC samples (6–11). One commonality to these large-scale next-generation sequencing efforts is the identification of mutational defects in the genes that regulate the DNA damage response and repair (DDR) system, found in 17% to 25% of PDACs.

During the cell cycle, there is a replication of over 6 billion base pairs of DNA. Such genomic replication is subject to numerous insults and replication stressors, which rely on essential response and repair mechanisms to ensure DNA's integrity. Furthermore, the chemotherapies used to treat cancers, particularly pancreatic cancer, result in specific types of DNA damage. For example, alkylating agents, such as platinums, and topoisomerase inhibitors, such as irinotecan, cause double-strand DNA breaks (DSB), whereas antimetabolites such as 5-fluorouracil and gemcitabine cause singe base-pair damage that can lead to single-strand DNA breaks (12). Deficiencies in DDR mechanisms have revealed targets for therapy, and research has led to FDA-approved treatments targeting cancers that harbor such deficiencies. For example, BRCA1 and BRCA2 play an integral role in the maintenance of genomic integrity, and germline mutations in either gene lead to increased risks for breast, ovarian, pancreatic, and prostate cancers (13). However, the presence of a BRCA1/2 mutation also predicts for an improved response and improved overall survival with platinum-based chemotherapy in both triple-negative breast cancers and PDAC (14, 15). Exploiting this DNA-repair defect not only improves sensitivity to chemotherapy, but also allows targetable therapy through the inhibition of the poly [adenosine diphosphate (ADP)–ribose] polymerase (PARP), leading to the accumulation of single-strand breaks which compromise DNA double-strand integrity at the replication fork. PARP inhibitors increase progression-free survival in advanced BRCA1/2-mutated ovarian and breast cancer (16–18), and are now FDA approved for these diseases. Additionally, responses to PARP inhibitors are also frequently seen in BRCA-mutated castrate-resistant prostate cancer with, for example, a response rate of 88% in a 50-patient trial; and in BRCA-mutated pancreatic cancer, with responses of 16% to 22% in small trials of 19 and 23 patients, respectively (16, 19–21).

ATM also plays a critical role in DDR. The ataxia telangiectasia mutated (ATM) gene, located on chromosome 11q 22–23, was first identified in 1995 during the evaluation of the ataxia telangiectasia syndrome. Germline mutations of ATM result in a well-characterized syndrome, as well as an increased predisposition for breast, pancreatic, and prostate cancers (22–28). Relevant here, mutations in the ATM gene, whether germline or somatic, are found in up to ∼6% of PDACs (further details below), and thus may represent a more prevalent DDR mutation than BRCA1/2 (7). In this review, we detail the function of ATM, review the current data on ATM deficiency in PDACs, and examine the therapeutic implications of ATM alterations.

The ATM gene consists of 66 exons that encode a PI3K-related serine/threonine protein kinase that plays a central role in the response to, and ultimately the repair of DNA DSB. Structurally, this large protein (350 kDa) contains serine or threonine residues susceptible for phosphorylation, followed by a glutamine amino acid located near its hydrophobic target region. Similar sites for posttranslational modifications (PTM) are found in the ataxia telangiectasia and RAD 3-related (ATR) kinase and in the DNA protein kinase (DNA-PK) proteins (29). ATM has important functions in the cell, including the maintenance of (i) telomere length (30, 31) and (ii) the mitotic spindle structure during mitosis (32). However, this review solely focuses on the central role of ATM in the process of DDR, including as it relates to targeted therapies in cancer. As depicted in Fig. 1, in order to repair damaged DNA, the MRE11-RAD50-NBS1 (MRN) complex acts as the primary sensor for DSBs and creates a physical bridge between the two broken ends (33). ATM can then interact directly with NBS1 (part of the MRN complex) through the direct binding of the C-terminus of NBS1 to several of the HEAT repeats that reside in ATM (34). It is believed that several PTMs are required for subsequent ATM activation. For instance, ATM has been shown to be activated through acetylation of K3016 by TIP60, a histone acetyltransferase that binds to ATM through recognition of the C-terminal FATC domain (35). ATM also requires autophosphorylation at S1981, which allows the kinase domain to dissociate from the FAT domain, enabling, in turn, the kinase to become active (36). These modifications allow ATM to transition from an inactive homodimer into an active monomer in response to DNA damage (36). This mechanism has been supported in the literature, but also questioned by others, demonstrating the need for further work in the field to clearly identify the role of S1981 and other ATM autophosphorylation events (29, 36, 37). Once activated, ATM phosphorylates multiple substrates, protein kinases, and sensor proteins in order to carry out DSB repair and also regulate normal cell-cycle processes, such as apoptosis and checkpoint activation (36, 38, 39).

ATM plays a role in the signaling required to initiate DNA repair, and thus, ATM defects can lead to genomic instability and malignancy. Hereditary and sporadic ATM mutations span the functional domains of the entire ATM gene (Fig. 2). These mutations occur mostly in the C-terminal end, which interacts with the PI3 kinase domain. This domain is involved with acetylation and activation of ATM (40). DDR is impaired when the ATM protein is dysfunctional, and loss of this DDR mechanism designed for DSBs can possibly lead, over time, to the accumulation of mutations, which can, in theory initiate the process of tumorigenesis. For instance, germline point mutations in ATM result in increased risks of breast cancers, specifically those associated with the S49C and S707P mutations. Melanoma, prostate, and oropharyngeal cancers are specifically associated with the S49C mutation. Although thyroid or endocrine cancers are generally associated with the S707P mutation (41, 42). Next-generation sequencing has also revealed somatic ATM mutations in many tumor types, including PDAC (Fig. 2; ref. 40).

Multiple studies have reaffirmed the importance of TP53, KRAS, CDKN2A, and SMAD4 mutations in PDAC (7, 11, 43, 44). Inherited risks of PDAC are also well established to be in large part due to germline mutations in BRCA1/BRCA2 and CDKN2A, identified in 7.4% of familial pancreatic cancers (N = 727; ref. 45), as well as individuals affected by Lynch syndrome (mutations in the genes: MLH1, MSH2, PMS2, and MSH6; ref. 46). Additionally, a study of familial PDAC patients showed that deleterious ATM mutations were significantly higher than the control group, suggesting ATM's role in malignancy (25). This knowledge has led to germline and somatic mutations in ATM as being identified and added to the catalog of predisposing gene mutations.

A genomic characterization of PDAC revealed 37% of the 71 samples carried alterations in DNA-repair genes (6). This significant discovery has also been confirmed in a recent large PDAC profiling study that identified targetable alterations in 50% of 640 PDAC patients, of which 8.4% expressed BRCA1/2 or ATM mutations (5). A review of the International Cancer Genome Consortium, a large database of sporadic PDACs, in 2015 identified ATM mutations in 9% to 18% of PDACs, with an average of 12% of the 591 samples (47). An additional large population study that identified the prevalence of homologous recombination–related gene mutations in 15.4% of PDACs (48). This study also revealed that ATM, ATRX, and CHEK2 mutations are present in 1.3% of 50,000 tumor samples. These mutations were most prevalently identified in PDAC (48). Further genomic studies support the high prevalence of ATM, CHEK2, and ATR mutations in PDAC (8). Our literature search, capturing 5,234 pancreatic cancer patients overall, shows that the total prevalence of ATM mutations (germline or somatic) in PDAC is 6.4% (range, 1%–34%; refs. 5, 6, 8, 9, 11, 25, 47–57). Importantly, in one study that showed that nearly 10% of PDAC patients carried a germline ATM mutation, in 44% of these patients a somatic second hit was identified (58). Although this loss-of-heterozygosity (LOH) seems to occur frequently in tumors arising from patients with germline ATM mutations (e.g., as seen in breast and pancreatic cancers; refs. 59, 60), the necessity of LOH to confer therapeutic sensitivity (i.e., to platinums and DDR inhibitors, as discussed below) is uncertain.

It is now well established that tumor cells with underlying defects in DDR are exquisitely sensitive to DNA-damaging agents, notably platinums, and more recently to PARP inhibitors in a phenomenon known as synthetic lethality (61). Similarly, inactivating mutations in ATM can also set up synthetic lethality in the presence of DNA-damaging agents (62). Historically, ataxia telangiectasia patients were found to be profoundly radiosensitive at the chromosomal level, suggesting the link between ATM and DNA repair (63). Preclinical data have demonstrated that knockdown of ATM results in radiosensitization (64, 65). Additional studies in the laboratory have demonstrated that ATM alterations in malignant cells can sensitize cells to platinum drugs (44, 66), and outside of platinum therapy, preclinical studies have also demonstrated ATM suppression in p53-deficient mouse fibroblasts sensitizes them to doxorubicin (67). Previous work has also demonstrated that ATM deficiency in p53-deficient cell lines causes a modest increase in 5-FU sensitivity (68). Lastly, one study took advantage of ATM-mutated PDAC cells in a mouse model and showed that treatment with the PARP inhibitor olaparib or the ATR inhibitor VE-822 led to dramatic accumulation of DSBs and reduced tumor cell viability in vitro and in vivo (62). The authors noted the compensation of alternate signaling routes to bypass ATM deficiency, including ATR in the replicative stress response. Thus, ATR inhibition was efficient in promoting intolerable mitotic damage, an effect that was enhanced when combined with gemcitabine (62).

The clinical experience in PDAC patients with confirmed ATM, ATR, or CHEK2 mutations is very limited, and focused on the efficacy of oxaliplatin-based chemotherapy. One case series (N = 71) utilizing real-time whole-exome sequencing demonstrated that a majority of patients with such mutations experienced a partial response or stable disease with oxaliplatin-based chemotherapy (6). In this case series, 80% of those with ATM, ATR, or CHEK2 mutations were treated with an oxaliplatin-based chemotherapy and 62.5% demonstrated partial response or stable disease on first follow-up scans (6). Another small study (N = 13) demonstrated a 37.5% response rate to oxaliplatin-based chemotherapy regimens in patients with DDR-mutated tumors (69). There was also a significantly longer progression-free survival compared with those patients whose tumors were DDR-nonmutated (20.8 months vs. 1.7 months, respectively P = 0.049; ref. 69). Specifically, 4 of 30 patients had known pathogenic ATM mutations, with at least one patient experiencing a prolonged partial response of nearly 40 months on 5-FU, irinotecan, and oxaliplatin (FOLFIRINOX; ref. 69). There are currently multiple ongoing trials targeting ATM-deficient tumors, including and especially with PARP inhibitors (Table 1).

Because ATM, ATR, and CHK1 are all important for resolving DNA damage (Fig. 1), utilizing an underlying DDR defect and inducing synthetic lethality by inhibiting an additional kinase is an innovative way to induce cancer cell death. The use of small-molecule inhibitors of ATM, ATR, and CHK1 is a promising avenue of cancer treatment due to the malignant cells' rapid and unregulated cell division. There are currently phase I and II clinical trials utilizing ATM or ATR inhibitors as monotherapy as well as in combination with chemotherapy (70, 71).

The first compound described to inhibit ATM was wortmannin (72); however, there are now a host of newer, more potent compounds that inhibit ATM. One of the newer generation of ATM inhibitors published in 2004 was KU55933. This compound was shown to inhibit downstream ATM phosphorylation after radiation, and it also enhanced responses to the topoisomerase inhibitors etoposide, camptothecins, and doxorubicin (73). A similar sensitization to topoisomerase inhibitors was later demonstrated with the ATM inhibitor AZ31, which was shown to increase the efficacy of irinotecan in resistant tumors in PDX models (74). KU60019 is another compound that was introduced in 2009 as an improved analogue of KU55933 (75), and early work demonstrated that KU60019 is a potent radiosensitizer (75). This molecule is currently being studied in in clear-cell renal cell carcinoma in combination with another known sensitizing agent, CX4945, which is an inhibitor of the protein kinase CK2 (76). The rationale for this combination comes from a compound screen where CK2 and ATM inhibitors were found to be highly synergistic in renal cancer. Interestingly, when CK2 inhibitors were tested in isogenic ATM-proficient and -deficient mouse cell lines, there was little difference in downstream effectors of DNA repair, such as AKT1 and BID, although overall viability in the ATM proficient and deficient cell lines treated with CK2 inhibitors was not assessed (77). This brings to light an important point when comparing the efficacy of interventions performed in combination with ATM inhibitors, as compared with these same interventions performed in patients with ATM-deficient tumors. The results may potentially be divergent as tumors with constitutive deficiency in ATM may have adapted to chronic loss of ATM function as opposed to acute loss as induced by ATM inhibitors. Conversely, other interventions may be both synergistic with ATM inhibitors and more potent in ATM-deficient patients.

Beyond the preclinical explorations of ATM inhibitors, currently two ATM inhibitors in clinical trials are being investigated in combination with other therapies (Table 1). AZD0156, an oral ATM inhibitor, is currently in clinical trials in combination with olaparib or FOLFIRI (78). These combinations are rational because, as previously mentioned, ATM inhibitors have been shown to sensitize cells to PARP inhibitors, and also to both 5-FU and irinotecan (61, 62, 68). AZD1390, another oral ATM inhibitor that penetrates the blood–brain barrier, is currently being tested in combination with radiation, given that radiation has been demonstrated to be more effective in ATM-deficient cancers (65, 79, 80). Importantly, in considering the potential adverse events for this trial, knockout of ATM in healthy tissue as compared with cancerous tissue was shown to induce less radiation sensitivity (80). This work demonstrated that increased sensitivity to radiation through ATM inhibition was primarily seen in cells that were rapidly replicating. As ATM inhibitors are further explored in the clinic, it will, of course, be important to monitor the side effects of ATM inhibitors particularly in combination with other therapies.

Increased sensitivity of ATM-deficient tumors to PARP inhibitors has previously been shown (64, 75, 81), and in a clinical trial of 124 patients, it has been demonstrated that dosing with olaparib and paclitaxel was more effective at increasing overall survival in patients with less ATM activity (HR, 0.35; 80% CI, 0.22–0.56; P = 0.002; median OS, not reached vs. 8.2 months; ref. 82). Unfortunately, the subsequent phase III trial with 525 patients did not enrich for patients with ATM-deficient tumors and was a negative study (83). Nevertheless, there are multiple ongoing trials of PARP inhibitor–based therapy targeting patients, at least in part, with ATM-deficient tumors (Table 1).

ATR is a phosphoinositide 3-kinase-related protein kinase that primarily responds to and repairs single-strand DNA breaks. It also shares functional sequences with ATM and DNA-PK, which respond to DSBs (29, 76). Upstream protein phosphorylation by ATM and autophosphorylation at the T1989 site stimulates ATR activity as well as TopBP1, which contains an ATR-activation domain to stimulate the kinase's activity (29, 84). The ATR kinase responds to a wide array of cell stressors, maintains DNA's integrity during replication, and is essential for proliferating cell survival. In the rapidly dividing cancer cell, there exists a high degree of replicative stress, creating an environment in which, as preclinical research has shown, suppression of ATR activity further increases replication stress leading to cell death (84). Furthermore, although normal dividing cells utilize ATM-dependent pathways for assistance in DNA repair, cancer cells, which are often deficient in ATM/p53 signaling, may rely solely on the ATR pathway for survival (67, 85, 86). This was demonstrated in genetically engineered mouse models of cancer, in which 90% genetic reduction of ATR expression suppressed the development of fibrosarcomas and acute myeloid leukemias with minimal side effects in normal tissues. This work affirmed the tumor selectivity of ATR inhibition (84). Moreover, inhibition of ATR selectively sensitizes tumor cells, but not normal cells, to radiation and chemotherapy (87).

Thus, small-molecule inhibitors of ATR may be particularly potent in PDACs with somatic mutations in ATM given that the lack of ATM's function may lead to increased dependence on ATR, and ATR inhibition could thus significantly promote cancer cell death. The ATR inhibitor VE-821 sensitizes cancer cells but not normal cells to chemotherapy (87), and these effects were synergistic in ATM-deficient cells (87). Another ATR inhibitor, AZD6738, causes accumulation of DNA damage, S phase arrest, and apoptosis in ATM dysfunctional gastric cells while not affecting those with functional ATM (88). Similar preclinical studies also suggest synthetic lethality between ATR inhibition with VE-822 and ATM deficiency in PDAC as well as lung adenocarcinoma cell lines, reaffirming the actionable molecular dependencies on ATR (62, 81). VE-822 has also been shown to potentially synergize with cisplatin in ATM-deficient esophageal squamous cells (89). This effect of potentiating the cytotoxicity of cisplatin and gemcitabine is also seen with AZD6738 in ATM-deficient non–small cell lung cancer cells (90). Several ongoing trials of ATR inhibitor-based therapies are listed in Table 1.

Downstream to and activated by ATR is the checkpoint kinase 1 (CHK1) pathway. CHK1 promotes proteasomal degradation of CDC25A in response to genome stress (29). The combined activity of ATR, CHK1, and CDC25A results in cell-cycle arrest and stabilization of replication stress at DNA forks. The inhibition of this complex leads to a decreased rate of fork progression, massive fork collapse in S phase cells, and ultimately cell death (82). Preclinical studies utilizing the CHK1 inhibitors MK8776 and LY2603618, with gemcitabine-based chemoradiation, showed synergistic effects to induce apoptosis of PDAC cells (91, 92). The combination of gemcitabine, a CHK1 inhibitor, PF-477736, and Lutetium-177–labeled anti-EGFR antibody leads to extensive DNA damage, apoptosis, and tumor degeneration in patient-derived xenografts (93). Additional preclinical studies with a tumor stem cell marker Doublecortin-like kinase 1 (Dclk1) inhibitor, LRRK2-IN-1 (LRRK), showed decreased expression of phosphorylated Chk1 (94). This same study demonstrated the combination of gemcitabine with LRRK significantly reduced cell survival compared with treatment with gemcitabine alone (94). Thus, CHK1 and Dclk1 are both potential targets in ATM-deficient malignancies as they also play a large role in single-strand break DNA repair.

CHK1 activation is primarily dependent on ATR at stalled replication forks and single-strand DNA, whereas CHK2 is activated mainly by ATM induced by DNA DSBs. One preclinical study examined the antitumor effects of a CHK2 inhibitor, NSC109555, in combination with gemcitabine. This combination increased apoptosis in pancreatic cells (95). Clinical trials of CHK1/2 inhibitors are also listed in Table 1.

There are currently no FDA-approved targeted therapies for patients with pancreatic cancer. However, genomic profiling of pancreatic adenocarcinomas is revealing therapeutically relevant alterations, and 17% to 25% of pancreatic cancers harbor mutations in the DDR pathway. Therapy targeted toward inhibiting the DNA damage response, including with PARP inhibitors, is proving to be highly effective particularly in DDR-deficient cancers, as has been established in BRCA1/2 mutated cancers. However, DDR is a highly complicated process, involving several overlapping pathways. It is reasonable to hypothesize that, depending upon the specific DDR mutation, there may be different optimal therapies to be utilized. PARP inhibitors are showing early promise in PDACs that harbor BRCA1/2 or PALB2 mutations, but consistently the most common DDR gene mutated in PDAC is actually ATM. It will be critical in the coming years to explore what DDR-targeted therapies might work best in ATM-deficient tumors. As with any therapeutic breakthrough, the future exploration of the complexity of the DDR pathway also justifies the need for a better understanding of compensatory and resistance mechanisms that may arise in the setting of ATM/ATR/CHK1–targeted therapies.

ATM deficiency may provide sensitivity for other elements of conventional therapies for PDAC, including radiation (96) and oxaliplatin. However, emerging targeted strategies, including immunotherapeutic combination approaches (97), will likely provide even better matches for ATM-deficient tumors. For example, mechanistically, it seems reasonable to consider that PARP inhibitors may be effective in treating ATM-deficient tumors. But potentially more promising might be the combination of a PARP inhibitor with an ATR inhibitor in ATM-deficient PDAC—essentially exploiting a new node of synthetic lethality in the DDR pathway. Similarly, there is a mechanistic reason to explore the role of CHK1 inhibition in ATM-deficient tumors. In both cases, understanding the need for inducing DNA damage with DNA-damaging chemotherapy will be critical as well. There are several ongoing clinical trials as discussed above, but clinical trials in PDAC in which there is such a high unmet need, and where ATM deficiency is common, would be ideal.

Additionally, recent genetic studies have revealed that specific ATM genotypes correlate to susceptibility to different diseases including cancer, which may provide valuable clinical information with regard to early detection, the subtyping of, and the treatment of PDACs (98). These genetic studies may complement and/or be evaluated in published genetically engineered mouse models (47, 62, 99) that have identified ATM's various roles (i.e., EMT, genetic instability, and metastases) in the progression model of PDAC. Moving forward, the research community should evaluate novel agents and combination therapies discussed above in these isogenic, in vivo models with the ultimate aim of classifying each ATM pathogenic genotype observed in patients with an optimally tailored, matched targeted therapeutic strategy.

No potential conflicts of interest were disclosed.

J.R. Brody and M.J. Pishvaian are supported by an NIH–NCI R01 CA212600 grant, and J.R. Brody is also supported by an NIH–NCI P30CA056036 core grant given to K. Knudsen. Research is supported by the 2015 Pancreatic Cancer Action Network–AACR Research Acceleration Network Grant, grant number 15-90-25-BROD, to J.R. Brody and M.J. Pishvaian.