Abstract
Background
Radiation therapy is one of the most commonly used cancer therapeutics but genetic determinants of clinical benefit are poorly characterized. Pathogenic germline variants in ATM are known to cause ataxia-telangiectasia, a rare hereditary syndrome notable for marked radiosensitivity. In contrast, somatic inactivation of ATM is a common event in a wide variety of cancers, but its clinical actionability remains obscure.
Methods
We analyzed 20 107 consecutively treated advanced cancer patients who underwent targeted genomic sequencing as part of an institutional genomic profiling initiative and identified 1085 harboring a somatic or germline ATM mutation, including 357 who received radiotherapy (RT). Outcomes of irradiated tumors harboring ATM loss-of-function (LoF) mutations were compared with those harboring variants of unknown significance. All statistical tests were 2-sided.
Results
Among 357 pan-cancer patients who received 727 courses of RT, genetic inactivation of ATM was associated with improved radiotherapeutic efficacy. The 2-year cumulative incidence of irradiated tumor progression was 13.2% vs 27.5% for tumors harboring an ATM LoF vs variant of unknown significance allele, respectively (hazard ratio [HR] = 0.51, 95% confidence interval [CI] = 0.34 to 0.77, P = .001). The greatest clinical benefit was seen in tumors harboring biallelic ATM inactivation (HR = 0.19, 95% CI = 0.06 to 0.60, P = .005), with statistically significant benefit also observed in tumors with monoallelic ATM inactivation (HR = 0.57, 95% CI = 0.35 to 0.92, P = .02). Notably, ATM LoF was highly predictive of outcome in TP53 wild-type tumors but not among TP53-mutant tumors.
Conclusions
We demonstrate that somatic ATM inactivation is associated with markedly improved tumor control following RT. The identification of a radio-sensitive tumor phenotype across multiple cancer types offers potential clinical opportunities for genomically guided RT.
Several large cancersequencing efforts have recently uncovered frequent and recurrent somatic or germline alterations in DNA-damage response (DDR) pathways (1,2). Some of these alterations are therapeutically targetable, as exemplified by pan-cancer approval of anti-PD1 therapy for mismatch repair defective cancers and poly (ADP-ribose) polymerase (PARP) inhibitors for BRCA1/2 mutant breast and ovarian cancers. Whether other DDR pathway alterations are clinically actionable remains unclear. The ataxia-telangiectasia mutated (ATM) gene encodes an apical serine-threonine kinase essential in the detection and signaling to repair DNA double-strand breaks (3) and frequently undergoes somatic mutation in multiple cancer types, including lung (4), prostate (5), and pancreatic cancer (6). Ionizing radiation is known to potently induce double-strand breaks, with in vitro cell line studies demonstrating increased radio-sensitivity after genetic or pharmacologic inhibition of ATM (7,8). Although we and others have previously provided anecdotal evidence that ATM may be therapeutically targetable with select DDR therapies, adequate phenotypic evaluation from human data is poor (9,10). Here, we sought to examine the landscape of ATM alterations in a large multi-histology cancer cohort and rigorously evaluate whether ATM alterations are associated with response to one of the most commonly used cancer therapies, radiation therapy.
Methods
Patients and Clinical Annotation
We analyzed a cohort of 20 107 patients treated at Memorial Sloan Kettering Cancer Center who underwent clinical genomic profiling, as part of a prospective genomic profiling initiative, using Integrated Mutation Profiling of Actionable Cancer Target (11,12). All patients provided written informed consent, including a specific consent for analysis of germline variant data in a subset of patients (n = 4701). After receiving institutional review board approval, we identified all patients harboring an ATM gene alteration who received radiotherapy (RT) at our institution. Among 1085 patients harboring an ATM mutation, 357 received RT to a total of 727 target lesions and were included in the present analysis. Irradiated tumor control was defined as the absence of radiographic or pathologically confirmed progression and/or relapse within the RT planning target volume (Supplementary Methods, available online).
Sequencing, Variant Calling, and Analysis
The Integrated Mutation Profiling of Actionable Cancer Target assay was performed as previously described (Supplementary Methods, available online) (11,13). Loss-of-function (LoF) mutations in ATM were defined a priori as mutations resulting in a frameshift or start-stop codon change (including nonsense), deleterious splice site mutations, or any missense mutation known to cause ataxia-telangiectasia in the germline setting (Supplementary Table 1, available online) (14–28), because such variants would be expected to have a deleterious effect on protein function. Allele-specific copy number analysis was performed using FACETS, as previously described (29), to assess ATM loss of heterozygosity as well as to identify tumors harboring homozygous deletion of ATM.
Statistical Analysis
Competing risks analyses, including both the nonparametric Gray’s test and the semiparametric Fine-Gray regression model, were used to assess the cumulative incidence of irradiated tumor progression as well as covariates’ effect, with death treated as a competing event. In analyses that included multiple treatment courses per patient, cluster effects were taken into account in the competing risks regression model. Interaction tests were conducted to examine potential differences of effects between patient subsets, including tumor histology or cancer subtype. Patient characteristics and their differences between groups were evaluated using the Wilcoxon rank sum and χ2 tests. P values and 95% confidence intervals (CI) were 2-sided, and a P value less than .05 was considered statistically significant. The statistical software R 3.6.0 with packages cmprsk and crrSC was used.
Results
Patient and Tumor Characteristics
We analyzed a cohort of 20 107 patients who underwent targeted clinical sequencing as part of a prospective genomic profiling initiative. Among all patients in our cohort, we identified 1085 (5.4%) whose genotyped tumor harbored a nonsynonymous somatic or germline mutation in ATM, including 443 (2.2%) harboring a LoF mutation. To evaluate the relative prevalence of somatic vs germline alterations, we next analyzed the subset of patients who consented to both germline and somatic genotyping (n = 4701) and identified 326 (6.9%) patients with a nonsynonymous ATM mutation, including 175 (3.7%) with a LoF mutation. Somatic and germline ATM LoF mutations were observed in 2.5% and 1.2%, respectively (Supplementary Figure 1, A, available online), similar to rates observed in The Cancer Genome Atlas pan-cancer cohort (Supplementary Figure 1, B, available online). Notably, ATM LoF events were observed across a wide variety of solid tumors (Figure 1, A).
Figure 1.
Genetic inactivation of ATM is common and is associated with clinical benefit from radiation therapy across cancer histologies. A) Distribution of loss-of-function (LoF) ATM mutation frequency by cancer type. Black bars represent the number of cases of each cancer type in the overall cohort, red bars represent ATM mutational frequency within each cancer type. B) Cumulative incidence of irradiated tumor progression by ATM genotype for all histologies. ATM LoF associated with decreased incidence of tumor progression (P = .001; Fine-Gray competing risk regression with clustering). C) Forest plot of irradiated tumor progression for ATM LoF tumors (vs variant of unknown significance [VUS]), stratified by histology or cancer type. All statistical tests were 2-sided. CI = confidence interval; RT = radiotherapy.
Association of ATM Genotype With Clinical Outcomes Following RT
Our primary objective was to investigate the association between ATM genotype and irradiated tumor progression among patients who received RT during their disease course. We hypothesized that tumors harboring ATM LoF mutations would have improved response to RT compared with tumors harboring an ATM variant of unknown significance (VUS), given that the majority of VUS are likely passenger events not expected to affect protein function. Among the 1085 patients harboring a nonsynonymous ATM mutation in our overall cohort, we identified 357 patients who received RT to a total of 727 sites. All patients in our study received RT based on a clinical indication that was unrelated to tumor genotype. Given the focus of our institutional sequencing initiative on guiding treatment decisions for advanced and metastatic cancer patients, RT in our cohort was most commonly (75.0%) for palliation (Table 1). All germline ATM LoF mutations were found in heterozygous carriers; there were no patients with biallelic germline ATM inactivation.
Table 1.
Patient characteristicsa
| Variable | Overall n = 357 | ATM VUS n = 180 | ATM LoF n = 177 | P a |
|---|---|---|---|---|
| Mean age at RT (range), y | 62 (9-97) | 63 (9-94) | 61 (31-97) | .11 |
| Female sex, No. (%) | 192 (53.8) | 89 (49.4) | 103 (58.2) | .09 |
| Metastatic disease, No. (%) | 187 (52.4) | 97 (53.9) | 90 (50.8) | .70 |
| Cancer type, No. (%) | .72 | |||
| Lung | 110 (30.8) | 55 (30.6) | 55 (31.1) | |
| Endometrial | 42 (11.8) | 19 (10.6) | 23 (13.0) | |
| Breast | 35 (9.8) | 18 (10.0) | 17 (9.6) | |
| Prostate | 30 (8.4) | 11 (6.1) | 19 (10.7) | |
| Colorectal | 20 (5.6) | 11 (6.1) | 9 (5.1) | |
| Thyroid | 12 (3.4) | 6 (2.8) | 6 (3.4) | |
| Other | 108 (30.3) | 60 (33.3) | 48 (27.1) |
P values calculated by Wilcoxon rank sum test for continuous variables and χ2 test for categorical variables. All tests were 2-sided. ATM = ataxia-telangiectasia mutated gene; LoF = loss of function; RT = radiotherapy; VUS = variant of unknown significance.
Given the nonrandomized nature of our study, we performed a detailed accounting of patient factors that may have been imbalanced between groups. Baseline patient, tumor, and treatment characteristics were generally well balanced (Tables 1 and 2) except for the presence of gross disease at RT (84.2% vs 77.9%, P = .04). There were no statistically significant differences between groups regarding cancer subtype, stage, tumor size, or radiation dose. Bone metastasis, central nervous system radiosurgery, and female sex were numerically more common in the ATM LoF group, but these differences did not reach statistical significance (P = .26, P = .06, and P = .09, respectively).
Table 2.
Tumor and treatment characteristics
| Variable | No. | Overall n = 727 | ATM VUS n = 361 | ATM LoF n = 366 | P a |
|---|---|---|---|---|---|
| Mean radiation doseb (range) | 727 | 51 (12-151) | 52 (12-115) | 51 (14-151) | .50 |
| Mean non-CNS tumor size (range) | 147 | 4.42 (0.8-14.3) | 4.30 (0.8-13.4) | 4.60 (1.0-14.3) | .73 |
| Mean CNS tumor size (range) | 104 | 1.18 (0.20-6.7) | 1.32 (0.20-6.7) | 1.03 (0.30-3.9) | .70 |
| Presence of gross tumor at RT | 727 | 589 (81.0) | 304 (84.2) | 285 (77.9) | .04 |
| RT type for localized disease, No. (%) | 182 | .95 | |||
| Definitive RT | 70 (38.6) | 34 (38.6) | 36 (38.3) | ||
| Adjuvant RT | 101 (54.5) | 48 (54.5) | 53 (56.4) | ||
| Neoadjuvant RT | 11 (6.8) | 6 (6.8) | 5 (5.3) | ||
| RT type for metastatic disease, No. (%) | 545 | .01 | |||
| Primary site | 48 (8.8) | 24 (8.8) | 24 (8.8) | ||
| Bone metastasis | 257 (47.2) | 122 (44.7) | 135 (49.6) | ||
| Lymph node metastasis | 24 (4.4) | 16 (5.9) | 8 (2.9) | ||
| Visceral metastasis | 43 (7.9) | 30 (11.0) | 13 (4.8) | ||
| CNS: whole-brain RT | 38 (7.0) | 23 (8.4) | 15 (5.5) | ||
| CNS: radiosurgery | 135 (24.8) | 58 (21.2) | 77 (28.3) |
P values calculated by Wilcoxon rank sum test for continuous variables and χ2 test for categorical variables. All tests were 2-sided. ATM = ataxia-telangiectasia mutated gene; CNS = central nervous system; LoF = loss of function; RT = radiotherapy; VUS = variant of unknown significance.
bBiologically effective dose in Gy (see Methods).
Median follow-up among the entire cohort and among surviving patients was 16.9 and 22.1 months from RT, respectively. The 2-year cumulative incidence of irradiated tumor progression was 13.2% vs 27.5% for ATM LoF vs VUS tumors, respectively (P = .001; Figure 1, B). Irradiated tumors were half as likely to progress when harboring an ATM LoF allele compared to an ATM VUS allele (hazard ratio [HR] = 0.50, 95% CI = 0.33 to 0.73, P = .001; Table 2). Given limited sample sizes for each cancer subtype, point estimates and confidence intervals for the hazard of local progression varied among cancer subtypes; however, similar trends were observed for the most common cancer subtypes in our cohort (Figure 1, C). Consistent with this, we tested for and did not find a statistically significant interaction between ATM genotype and cancer subtype in multivariable regression analysis.
Clinical and genetic factors associated with irradiated tumor control on univariate analysis included ATM genotype and absence of gross tumor (adjuvant treatment for microscopic disease; Table 3). ATM LoF remained independently associated with superior irradiated tumor control in our multivariable model (HR = 0.51, 95% CI = 0.34 to 0.77, P = .001; Table 3). Likewise, ATM LoF remained statistically significant in subset analysis restricted to patients with gross disease (P < .001; Supplementary Figure 2, A, available online). No statistically significant difference in tumor control was observed for germline vs somatic LoF mutation (P = .74; Supplementary Figure 2, B, available online). ATM genotype was independently predictive of superior irradiated tumor control among both men (HR = 0.56, 95% CI = 0.34 to 0.91, P = .02) and women (HR = 0.39, 95% CI = 0.24 to 0.63, P < .001).
Table 3.
Analysis of factors associated with irradiated tumor controla
| Factor | Univariate analysisb |
Multivariable analysisb |
||
|---|---|---|---|---|
| Hazard ratio (95% CI) | P | Hazard ratio (95% CI) | P | |
| ATM genotype | ||||
| LoF (VUS = reference) | 0.50 (0.33 to 0.74) | .001 | 0.51 (0.34 to 0.77) | .001 |
| TP53 genotype | ||||
| Mutant (wild type = referent) | 1.24 (0.84 to 1.87) | .26 | 1.32 (0.87 to 2.02) | .20 |
| Radiation dose | ||||
| BED (continuous) | 0.99 (0.98 to 1.00) | .23 | 0.99 (0.98 to 1.01) | .24 |
| Age | ||||
| At time of RT (continuous) | 0.99 (0.97 to 1.00) | .14 | 0.99 (0.97 to 1.00) | .12 |
| Gross tumor at RT | ||||
| Present (absent = referent) | 1.93 (1.18 to 3.16) | .009 | 2.39 (1.33 to 4.29) | .004 |
| Extent of disease at RT | ||||
| Metastatic (localized = referent) | 1.08 (0.73 to 1.60) | .70 | 0.84 (0.48 to 1.47) | .55 |
| Cancer type | ||||
| Lung | 1.00 (referent) | .09 | 1.00 | |
| Endometrial | 1.89 (0.98 to 3.67) | 2.19 (0.98 to 4.91) | .06 | |
| Breast | 1.19 (0.57 to 2.48) | 1.08 (0.53 to 2.18) | .83 | |
| Prostate | 1.98 (1.02 to 3.85) | 2.09 (1.07 to 4.08) | .03 | |
| Colorectal | 1.80 (0.68 to 4.77) | 1.62 (0.70 to 3.76) | .26 | |
| Thyroid | 2.87 (1.09 to 7.56) | 3.08 (1.36 to 6.96) | .007 | |
| Other | 2.01 (1.2 to 3.36) | 2.07 (1.20 to 3.59) | .009 | |
N = 727 observations from N = 357 unique patients. BED = biologically effective dose; CI = confidence interval; LoF = loss of function mutation; RT = radiotherapy; VUS = variant of unknown significance.
Hazard ratio and P values calculated using a competing risks regression model with data clustered on patient identifier.
To evaluate whether inclusion of multiple RT courses per patient may have influenced our observations, we performed a confirmatory analysis limited to the first radiation course for each patient, which yielded highly concordant results (Supplementary Table 2, available online). Furthermore, to ensure the association of ATM genotype and irradiated tumor control was due to the superior response of LoF-harboring tumors and not suboptimal response of VUS-harboring tumors, we compared tumors with ATM LoF with a cohort of ATM wild-type tumors harboring mutations in FAT1, a gene similar in size to ATM (139 vs 146 kb) with no known role in mediating DNA repair. This analysis demonstrated superior tumor control for ATM LoF tumors compared with ATM-wt/FAT1-mut tumors (P = .005; Supplementary Figure 2, C, available online) and similar rates of tumor control in both control cohorts (ATM-wt/FAT1-mut and ATM-VUS; P = .30).
ATM Phenotypes in TP53-Wild Type vs TP53-Mutant Tumors
To assess for potential confounding genomic alterations, we compared the genomic landscape of ATM LoF and ATM VUS tumors. We observed no statistically significant differences in genes affected by copy number alteration or mutation (Supplementary Figure 3, A, available online), with the exception of TP53, which we have previously described (30). Tumors harboring ATM LoF mutations were statistically significantly less likely to carry TP53 mutations compared with ATM VUS tumors (44% vs 29%, q < .0001), suggesting functional epistasis. Given this observed tendency toward mutual exclusivity, we next examined whether TP53 mutation influenced the effect of ATM inactivation on radiation response. We found ATM inactivation to be statistically significantly associated with improved irradiated tumor control in TP53 wild-type (HR = 0.42, 95% CI = 0.26 to 0.69, P = .001; Figure 2, A) but not TP53 mutant tumors (HR = 0.70, 95% CI = 0.37 to 1.30, P = .26; Figure 2, B). A statistical test for interaction between ATM and TP53 genotype regarding irradiated tumor control was limited by the relative paucity of TP53-mutant tumors harboring an ATM LoF allele but demonstrated a trend towards statistical significance (P = .09 in multivariable model; Supplementary Table 3, available online).
Figure 2.
Clinical outcomes stratified by TP53 genotype and loss of ATM heterozygosity. A) Cumulative incidence of irradiated tumor progression stratified by ATM genotype among TP53 wild-type tumors. B) Cumulative incidence of irradiated tumor progression stratified by ATM genotype among TP53 mutant tumors. ATM loss-of-function (LoF) was associated with decreased tumor progression for TP53 wild-type tumors (P < .001) but not TP53 tumors (P = .26; Fine-Gray competing risk regression with clustering). C) Cumulative incidence of irradiated tumor progression stratified by ATM allelic status. D) Kaplan-Meier overall survival analysis from time of radiation stratified by ATM allelic status (P = .72). All statistical tests were 2-sided. VUS = variant of unknown significance.
Monoallelic vs Biallelic ATM Inactivation
Given prior preclinical work demonstrating marked radio-sensitivity in cells with homozygous ATM inactivation and an intermediate phenotype for heterozygous ATM inactivation (31), we next sought to examine whether tumors with biallelic ATM inactivation were more likely to be controlled with RT than those with monoallelic inactivation. Biallelic ATM inactivation was defined as a germline or somatic LoF mutation with loss of the wild-type allele or compound heterozygosity with 2 different LoF alleles. Among 306 (85.7%) patients with available allele-specific copy number data, biallelic ATM inactivation was observed in 34 of 177 (19.2%) tumors harboring an LoF mutation and in 29 of 180 (16.1%) tumors harboring a VUS (Supplementary Tables 4 and 5, available online).
Consistent with our hypothesis, tumors harboring a biallelic ATM LoF mutation displayed the lowest rate of irradiated tumor progression, whereas those with monoallelic ATM inactivation displayed an intermediate phenotype. The hazard ratio for irradiated tumor progression in multivariable analysis for biallelic LoF, monoallelic LoF, and biallelic VUS (compared with monoallelic VUS) was 0.19 (95% CI = 0.06 to 0.60, P = .005), 0.57 (95% CI = 0.35 to 0.92, P = .02), and 0.92 (95% CI = 0.48 to 1.76, P = .81), respectively (Supplementary Table 6, available online). The 2-year cumulative incidence of irradiated tumor progression was 3.4% and 16.9% for biallelic and monoallelic ATM LoF tumors and 19.8% and 32.1% for biallelic and monoallelic ATM VUS tumors, respectively (P < .001; Figure 2, C). Review of individual clinical histories revealed anecdotal evidence of exceptional responses to RT, including a woman with bulky leptomeningeal disease from breast cancer harboring a biallelic ATM nonsense mutation (L186*) who had a complete and durable response to whole-brain RT (Supplementary Figure 4, available online).
Because biallelic inactivation can result from homozygous deletion, we additionally identified all patients in our denominator whose irradiated tumor harbored homozygous deletion of ATM and identified an additional 5 patients (receiving 8 courses of RT) whose tumors harbored such an alteration (see Methods). Consistent with observed outcomes among patients with biallelic LoF mutation or LOH, we observed no progression events (0 of 8) among this subset of irradiated tumors.
Functional Evaluation of ATM Missense Alleles via Analysis of Clinical Outcomes
Given the association between ATM genotype and outcome following RT, we next performed an exploratory analysis to determine whether analysis of patient-level data could inform the functional classification of ATM VUS alterations. The human ATM protein is organized into 3 main structural units, including an N-terminal “spiral” domain, a middle “pincer” domain, thought to function in regulating the accessibility of the kinase domain to substrates, and a conserved C-terminal kinase domain containing the catalytic core (32). Despite exclusion of known pathogenic alleles, we were able to observe statistically significant differences in the radiation phenotype dependent on missense mutation location (Supplementary Figure 4, B and C available online), highlighting the possibility that a small subset of VUS may affect protein function. Tumors harboring a VUS in the pincer (HR = 0.38, 95% CI = 0.16 to 0.89, P = .03) or kinase domains (HR = 0.57, 95% CI = 0.29 to 1.11, P = .09) were more likely to be controlled following RT than tumors harboring a VUS in the N-terminal spiral domain, although only the former reached statistical significance.
Systemic Therapy and Unirradiated Tumor Control
To account for potential confounding caused by differential responses to systemic therapy, we examined patient exposures to pre- and post-RT systemic therapies. Specifically, we examined pre-RT systemic therapy in patients with metastatic disease as defined by number of lines of prior treatment and post-RT systemic therapy as defined by the first line of systemic therapy completed following RT. The number of lines of systemic therapy before RT in patients with metastatic disease was well-balanced between patients harboring ATM LoF and VUS alleles, respectively. The median number of lines of systemic therapy before RT was 1 in both cohorts, and the number of lines of systemic therapy before RT was not associated with irradiated tumor control (HR = 1.02, 95% CI = 0.91 to 1.14, P = .71). We additionally found no difference in the type of systemic therapy administered after RT for patients harboring LoF vs VUS alleles (P = .60; Supplementary Table 9, available online).
To confirm these observations, we additionally examined whether ATM inactivation was associated with superior tumor control outside of the irradiated field. Because RT was most commonly delivered with palliative intent in patients with additional metastatic disease outside of the irradiated field, we hypothesized that genetic ATM inactivation would not be associated with improved overall survival or disease progression outside of the irradiated field. A comparison between patients with tumors harboring ATM LoF alleles with those harboring a VUS revealed no statistically significant difference in overall survival (P = .72; Figure 2, D) or the rate of disease progression outside of the irradiated field (HR = 0.82, 95% CI = 0.61 to 1.09, P = .18; Supplementary Figure 5, available online), suggesting that ATM inactivation is predictive of improved outcomes for irradiated lesions but not otherwise prognostic of superior outcome.
Non-ATM DDR Genes and Clinical Outcomes Following RT
We next queried our database of patients for alterations in DDR signaling genes (ATR, CHEK1, CHEK2) as well as genes specifically required for the core homologous recombination pathway (BRCA1, BRCA2, PALB2, RAD51B/C/D) and Fanconi anemia/interstrand cross-link repair pathway (FANCC, FANCA). To limit confounding, we limited our analysis to patients in the control cohort (ATM VUS or FAT1 VUS/LoF) and identified 190 patients whose irradiated tumor harbored either an LoF mutation (n = 72) or VUS/B mutation (n = 118) in 1 of the aforementioned DDR genes (121 irradiated LoF tumors, 226 irradiated VUS/B tumors). In contrast to our results for ATM, we did not observe any statistically significant difference in irradiated tumor control among tumors harboring LoF mutations (vs VUS/B) for any of the DDR genes we analyzed (Supplementary Table 7, available online).
Discussion
Therapeutic radiation is commonly used in cancer treatment, but genetic determinants of clinical benefit are poorly characterized. In this study, we found genetic ATM inactivation to be strongly associated with clinical benefit from RT across a wide variety of solid tumor types. Although the greatest clinical benefit was observed in tumors harboring biallelic ATM inactivation, monoallelic ATM inactivation was statistically significantly associated with improved irradiated tumor control. Collectively, these data are consistent with a genetic basis for benefit from RT and, to our knowledge, are the first to identify a pan-cancer genomic alteration associated with clinical benefit from RT. Several prior studies have examined associations between genotype and clinical radiation response but either focused on gene expression signatures (33) or were restricted to small, single histology cohorts (34,35).
Given that radio-sensitivity is a hallmark of ataxia-telangiectasia caused by bi-allelic germline ATM inactivation, our findings are not entirely surprising. Yet, the clinical value of heterozygous ATM inactivation has remained poorly understood. Although analyses of heterozygous Atm+/− murine models have previously suggested premature aging and sensitivity to ionizing radiation, whether monoallelic/heterozygous ATM loss can potentially serve as a biomarker for radiation therapy has remained elusive. In this context, the data presented here provide novel evidence for the clinical actionability of both heterozygous and homozygous ATM inactivation.
A notable finding from our study was the observation that ATM inactivation is predictive of improved irradiated tumor control in TP53 wild-type but not TP53 mutant tumors. The role of p53 in the radiation response is known to be highly dependent on cellular context and lineage, leading to sometimes paradoxical effects in determining susceptibility to radiation. For example, Trp53−/− mice have been shown to be resistant to hematopoietic failure after radiation exposure but prone to death induced by gastrointestinal radiation injury (36). Nevertheless, the mechanism by which TP53 mutation attenuates radio-sensitivity in ATM-deficient tumors remains unclear and requires further investigation.
We additionally examined whether LoF alterations in non-ATM DNA repair genes were associated with irradiated tumor control among patients in our database. In contrast to our results for ATM, we did not observe any statistically significant difference in irradiated tumor control for any of the DDR genes we analyzed. Given the preliminary, hypothesis-generating nature of this analysis, we cannot rule out the possibility that LoF alterations in these genes affect clinical responses to RT, but our observations are nevertheless consistent with preclinical observations that inactivation of core HR or FA pathway genes produce a much more limited radio-sensitivity than observed with inactivation of ATM.
Our study has limitations. Given the broad focus of our institutional sequencing initiative, our patient cohort was heterogeneous regarding histology, disease stage, timing of radiation in the disease course, and other clinical factors. Additionally, our analysis assumed concordance in ATM genotype for cases where genotyping was performed using tissue from a nonirradiated site. It is likely that this assumption reduced the signal to noise ratio in the LoF phenotypes we observed, but further analysis will be needed to confirm this.
Current approaches to decision making in radiation oncology largely rely on traditional clinicopathologic factors such as histology, tumor size, and lymph node involvement. The identification of a radio-sensitive tumor phenotype among patients with somatic ATM mutations offers potential applications to the personalization of radiation treatment decisions given that tumor genotypes can be readily ascertained using clinically available assays and minimal tissue or DNA. A potential first step to incorporating genomic data in radiation treatment decisions is individualization of radiation dose based on expected intrinsic radio-sensitivity. For example, a reduced dose of palliative RT could be evaluated in metastatic tumors harboring somatic ATM mutations, with the goal of reducing radiation-related toxicity in patients whose tumors are likely to be controlled with a lower dose of ionizing radiation. In addition to modulation of radiation dose, the observation that biallelic ATM inactivation is associated with exceptional radio-sensitivity may allow for the use of low-dose RT in clinical scenarios where it is not currently employed (37,38).
In summary, examining a cohort of consecutively genotyped advanced cancer patients, we find genetic ATM inactivation to be strongly associated with clinical benefit from RT across multiple cancer types. Prospective studies are clearly needed to both validate genetic biomarkers of radiation response and test personalized approaches to delivery of RT.
Funding
This work was supported by National Institutes of Health K12 CA184746 (to JS and PBR), R35 CA232097 (to TAC), and RO1 CA205426 (to TAC and NAR); the National Cancer Institute at the National Institutes of Health Cancer Center Support Grant (P30 CA008748); the PaineWebber Chair (TAC); and the Breast Cancer Research Foundation (JR-F).
Notes
Role of the funder: The funders had no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; and the decision to submit the manuscript for publication.
Disclosures: P is a co-founder of Gritstone Oncology and holds equity. T.A.C. holds equity in An2H. T.A.C. acknowledges grant funding from Bristol-Myers Squibb, AstraZeneca, Illumina, Pfizer, An2H, and Eisai. T.A.C. has served as an advisor for Bristol-Myers, MedImmune, Squibb, Illumina, Eisai, AstraZeneca, and An2H. T.A.C. holds ownership of intellectual property on using tumor mutation burden to predict immunotherapy response, with pending patent, which has been licensed to PGDx. A.R. acknowledges grant funding from Varian Medical Systems, Boehringer Inhelheim, Pfizer, and AstraZeneca. A.R. has received honorarium from AstraZeneca, Merck, Research to Practice, Cybrexa, and MoreHealth, and non-financial support from Philips/Elekta. C.A.B. acknowledges funding from Elekta, Amgen and Merck for investigator initiated clinical trials. P.B.R has received honorarium from Corning and has served as a consultant for AstraZeneca and EMD Serono. J.S.R-F. reports personal/consultancy fees from VolitionRx, Page.AI, Goldman Sachs, Grail, Ventana Medical Systems, Invicro, Roche Diagnostics, and Genentech outside the scope of the submitted work. N.R. acknowledges research funding support from Pfizer, Bristol Myers Squibb, and AstraZeneca, and has served as an advisor to Mirati.
Data availability statement
De-identified datasets analyzed in the current study are available from the corresponding author on reasonable request.
Role of the authors: KLP: Data curation; Formal analysis; Investigation; Writing—original draft; Writing—review and editing. DLC: Data curation; Investigation; Writing—original draft; Writing—review and editing. YCL: Data curation; Investigation. MH: Data curation; Formal analysis; Writing—original draft; Writing—review and editing. ZZ: Formal analysis; Methodology; Supervision; Writing—original draft; Writing—review and editing. XS: Data curation. IP: Data curation; BM: Data curation. JM: Data curation. RMS: Data curation; Investigation; Writing—original draft. IXP: Data curation; Formal analysis; Software; AJK: Investigation; Writing—original draft. LZB: Investigation; Writing—original draft. LGTM: Investigation; Writing—original draft. CAB: Investigation; Writing—original draft. AR: Investigation; Writing—original draft. KMA: Investigation; Writing—original draft. PBR: Investigation; Writing—original draft. CHC: Investigation; Writing—original draft. JY: Investigation; Writing—original draft. MJZ: Investigation; Writing—original draft. HIS: Investigation; Writing—original draft. JLB: Investigation; Writing—original draft. DLM: Investigation; Supervision; Writing—original draft. BW: Investigation; Supervision; Writing—original draft. JSR-F: Investigation; Supervision; Writing—original draft; Writing—review and editing. NYL: Investigation; Supervision; Writing—original draft. SNP: Investigation; Supervision; Writing—original draft; Writing—review and editing. TAC: Conceptualization; Investigation; Supervision; Writing—original draft; Writing—review & editing. NR: Conceptualization; Data curation; Formal analysis; Investigation; Writing—original draft; Writing—review and editing.
Supplementary Material
References
- 1. Mutter RW, Riaz N, Ng CK, et al. Bi-allelic alterations in DNA repair genes underpin homologous recombination DNA repair defects in breast cancer. J Pathol. 2017;242(2):165–177. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Mandelker D, Zhang L, Kemel Y, et al. Mutation detection in patients with advanced cancer by universal sequencing of cancer-related genes in tumor and normal DNA vs guideline-based germline testing. JAMA. 2017;318(9):825–835. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Shiloh Y, Ziv Y.. The ATM protein kinase: regulating the cellular response to genotoxic stress, and more. Nat Rev Mol Cell Biol. 2013;14(4):197–210. [PubMed] [Google Scholar]
- 4.Cancer Genome Atlas Research Network. Comprehensive molecular profiling of lung adenocarcinoma. Nature. 2014;511(7511):543–550. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Cancer Genome Atlas Research Network. The molecular taxonomy of primary prostate cancer. Cell. 2015;163(4):1011–1025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Cancer Genome Atlas Research Network. Integrated genomic characterization of pancreatic ductal adenocarcinoma. Cancer Cell. 2017;32(2):185–203.e113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Taylor AM, Harnden DG, Arlett CF, et al. Ataxia telangiectasia: a human mutation with abnormal radiation sensitivity. Nature. 1975;258(5534):427–429. [DOI] [PubMed] [Google Scholar]
- 8. Rainey MD, Charlton ME, Stanton RV, Kastan MB.. Transient inhibition of ATM kinase is sufficient to enhance cellular sensitivity to ionizing radiation. Cancer Res. 2008;68(18):7466–7474. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Choi M, Kipps T, Kurzrock R.. ATM mutations in cancer: therapeutic implications. Mol Cancer Ther. 2016;15(8):1781–1791. [DOI] [PubMed] [Google Scholar]
- 10. Golmard L, Castera L, Krieger S, et al. Contribution of germline deleterious variants in the RAD51 paralogs to breast and ovarian cancers. Eur J Hum Genet. 2017;25(12):1345–1353. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Cheng DT, Mitchell TN, Zehir A, et al. Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT): a hybridization capture-based next-generation sequencing clinical assay for solid tumor molecular oncology. J Mol Diagn. 2015;17(3):251–264. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Zehir A, Benayed R, Shah RH, et al. Mutational landscape of metastatic cancer revealed from prospective clinical sequencing of 10,000 patients. Nat Med. 2017;23(6):703–713. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Cheng DT, Prasad M, Chekaluk Y, et al. Comprehensive detection of germline variants by MSK-IMPACT, a clinical diagnostic platform for solid tumor molecular oncology and concurrent cancer predisposition testing. BMC Med Genomics. 2017;10(1):33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Camacho E, Hernández L, Hernández S, et al. ATM gene inactivation in mantle cell lymphoma mainly occurs by truncating mutations and missense mutations involving the phosphatidylinositol-3 kinase domain and is associated with increasing numbers of chromosomal imbalances. Blood. 2002;99(1):238–244. [DOI] [PubMed] [Google Scholar]
- 15. Grønbæk K, Worm J, Ralfkiaer E, Ahrenkiel V, Hokland P, Guldberg P.. ATM mutations are associated with inactivation of the ARF-TP53 tumor suppressor pathway in diffuse large B-cell lymphoma. Blood. 2002;100(4):1430–1437. [DOI] [PubMed] [Google Scholar]
- 16. Fang NY, Greiner TC, Weisenburger DD, et al. Oligonucleotide microarrays demonstrate the highest frequency of ATM mutations in the mantle cell subtype of lymphoma. Proc Natl Acad Sci U S A. 2003;100(9):5372–5377. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Sandoval N, Platzer M, Rosenthal A, et al. Characterization of ATM gene mutations in 66 ataxia telangiectasia families. Hum Mol Genet. 1999;8(1):69–79. [DOI] [PubMed] [Google Scholar]
- 18. Demuth I, Dutrannoy V, Marques W Jr, et al. New mutations in the ATM gene and clinical data of 25 AT patients. Neurogenetics. 2011;12(4):273–282. [DOI] [PubMed] [Google Scholar]
- 19. Anheim M, Fleury M, Monga B, et al. Epidemiological, clinical, paraclinical and molecular study of a cohort of 102 patients affected with autosomal recessive progressive cerebellar ataxia from Alsace, Eastern France: implications for clinical management. Neurogenetics. 2010;11(1):1–12. [DOI] [PubMed] [Google Scholar]
- 20. Stankovic T, Kidd AM, Sutcliffe A, et al. ATM mutations and phenotypes in ataxia-telangiectasia families in the British Isles: expression of mutant ATM and the risk of leukemia, lymphoma, and breast cancer. Am J Hum Genet. 1998;62(2):334–345. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Becker-Catania SG, Chen G, Hwang MJ, et al. Ataxia-telangiectasia: phenotype/genotype studies of ATM protein expression, mutations, and radiosensitivity. Mol Genet Metab. 2000;70(2):122–133. [DOI] [PubMed] [Google Scholar]
- 22. Broeks A, de Klein A, Floore AN, et al. ATM germline mutations in classical ataxia-telangiectasia patients in the Dutch population. Hum Mutat. 1998;12(5):330–337. [DOI] [PubMed] [Google Scholar]
- 23. Barone G, Groom A, Reiman A, Srinivasan V, Byrd PJ, Taylor AM.. Modeling ATM mutant proteins from missense changes confirms retained kinase activity. Hum Mutat. 2009;30(8):1222–1230. [DOI] [PubMed] [Google Scholar]
- 24. Jacquemin V, Rieunier G, Jacob S, et al. Underexpression and abnormal localization of ATM products in ataxia telangiectasia patients bearing ATM missense mutations. Eur J Hum Genet. 2012;20(3):305–312. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Cavaciuti E, Lauge A, Janin N, et al. Cancer risk according to type and location of ATM mutation in ataxia-telangiectasia families. Genes Chromosom Cancer. 2005;42(1):1–9. [DOI] [PubMed] [Google Scholar]
- 26. Allinen M, Launonen V, Laake K, et al. ATM mutations in Finnish breast cancer patients. J Med Genet. 2002;39(3):192–196. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Renault AL, Mebirouk N, Fuhrmann L, et al. Morphology and genomic hallmarks of breast tumours developed by ATM deleterious variant carriers. Breast Cancer Res. 2018;20(1):28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Telatar M, Teraoka S, Wang Z, et al. Ataxia-telangiectasia: identification and detection of founder-effect mutations in the ATM gene in ethnic populations. Am J Hum Genet. 1998;62(1):86–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Shen R, Seshan VE.. FACETS: allele-specific copy number and clonal heterogeneity analysis tool for high-throughput DNA sequencing. Nucleic Acids Res. 2016;44(16):e131–e131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Weigelt B, Bi R, Kumar R, et al. ; kConFab Investigators. The landscape of somatic genetic alterations in breast cancers from ATM germline mutation carriers. J Natl Cancer Inst. 2018;110(9):1030–1034. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Lavin MF, Scott S, Gueven N, Kozlov S, Peng C, Chen P.. Functional consequences of sequence alterations in the ATM gene. DNA Repair (Amst). 2004;3(8-9):1197–1205. [DOI] [PubMed] [Google Scholar]
- 32. Wang X, Chu H, Lv M, et al. Structure of the intact ATM/Tel1 kinase. Nat Commun. 2016;7(1):11655. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Scott JG, Berglund A, Schell MJ, et al. A genome-based model for adjusting radiotherapy dose (GARD): a retrospective, cohort-based study. Lancet Oncol. 2017;18(2):202–211. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Jeong Y, Hoang NT, Lovejoy A, et al. Role of KEAP1/NRF2 and TP53 mutations in lung squamous cell carcinoma development and radiation resistance. Cancer Discov. 2017;7(1):86–101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Wang M, Han J, Marcar L, et al. Radiation resistance in KRAS-mutated lung cancer is enabled by stem-like properties mediated by an osteopontin-EGFR pathway. Cancer Res. 2017;77(8):2018–2028. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Lee CL, Blum JM, Kirsch DG.. Role of p53 in regulating tissue response to radiation by mechanisms independent of apoptosis. Transl Cancer Res. 2013;2(5):412–421. [PMC free article] [PubMed] [Google Scholar]
- 37. Rahal R, Frick M, Romero R, et al. Pharmacological and genomic profiling identifies NF-kappaB-targeted treatment strategies for mantle cell lymphoma. Nat Med. 2014;20(1):87–92. [DOI] [PubMed] [Google Scholar]
- 38. Meissner B, Kridel R, Lim RS, et al. The E3 ubiquitin ligase UBR5 is recurrently mutated in mantle cell lymphoma. Blood. 2013;121(16):3161–3164. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
De-identified datasets analyzed in the current study are available from the corresponding author on reasonable request.
Role of the authors: KLP: Data curation; Formal analysis; Investigation; Writing—original draft; Writing—review and editing. DLC: Data curation; Investigation; Writing—original draft; Writing—review and editing. YCL: Data curation; Investigation. MH: Data curation; Formal analysis; Writing—original draft; Writing—review and editing. ZZ: Formal analysis; Methodology; Supervision; Writing—original draft; Writing—review and editing. XS: Data curation. IP: Data curation; BM: Data curation. JM: Data curation. RMS: Data curation; Investigation; Writing—original draft. IXP: Data curation; Formal analysis; Software; AJK: Investigation; Writing—original draft. LZB: Investigation; Writing—original draft. LGTM: Investigation; Writing—original draft. CAB: Investigation; Writing—original draft. AR: Investigation; Writing—original draft. KMA: Investigation; Writing—original draft. PBR: Investigation; Writing—original draft. CHC: Investigation; Writing—original draft. JY: Investigation; Writing—original draft. MJZ: Investigation; Writing—original draft. HIS: Investigation; Writing—original draft. JLB: Investigation; Writing—original draft. DLM: Investigation; Supervision; Writing—original draft. BW: Investigation; Supervision; Writing—original draft. JSR-F: Investigation; Supervision; Writing—original draft; Writing—review and editing. NYL: Investigation; Supervision; Writing—original draft. SNP: Investigation; Supervision; Writing—original draft; Writing—review and editing. TAC: Conceptualization; Investigation; Supervision; Writing—original draft; Writing—review & editing. NR: Conceptualization; Data curation; Formal analysis; Investigation; Writing—original draft; Writing—review and editing.