Germline Pathogenic Variants Impact Clinicopathology of Advanced Lung Cancer

Lung cancer is the leading cause of cancer mortality worldwide, with a five-year survival rate of 15% (1). Small cell lung cancer (SCLC) accounts for 15% to 17% of cases, and the various histologies that comprise non–small cell lung cancer (NSCLC, including lung adenocarcinoma and squamous cell carcinoma) account for 85% (2). SCLC and NSCLC differ in risk factors, somatic mutational landscapes, invasiveness, and response to treatment. Although tobacco smoking is a primary cause for both categories of lung cancer (3, 4), only about 15% of lifelong smokers develop the disease, indicating there is significant individual variation in susceptibility. The smoking prevalence is much lower among adenocarcinoma patients compared with squamous and SCLC patients. Additionally, an estimated 10% to 25% of lung cancer cases occur in never smokers (5), highlighting the potential role of genetic predisposition in lung cancer etiology. Lung cancer is significantly more common among those with a positive family history (odds ratio = 1.57–2.7), and this is particularly true for young-onset lung cancers (6). Genetic background or ancestry may also influence the somatic evolution of lung cancer (7). Rare familial forms of lung cancer have been associated with germline pathogenic or likely pathogenic variants (germline PV) of EGFR, TP53, RB1, CDC147, BAP1, HER2, PARK2, and YAP1 (8–10). Several lung cancer susceptibility loci with small effect sizes were identified using genome-wide association studies (GWAS; refs. 11–13). Recent studies reported that 2.5% to 4.5% of lung cancer patients carried germline PV that have been linked to cancer risk in Mendelian syndromes (14, 15). Using whole-exome sequencing, we and others have recently identified rare, deleterious variants in ATM, BRCA2, and Fanconi anemia genes that associate with lung cancer risk (16–18). The increased use of clinical sequencing for both tumor profiling and germline testing has incidentally detected clinically actionable germline PV in lung cancer patients (19, 20). Although genetic risk models have been applied to inherited breast, ovarian, pancreatic, and prostate cancer (20–23), the clinicopathologic significance of germline PV in lung cancer has not been elucidated.

Here, we report the prevalence and clinical implications of germline PV in 5,118 patients with lung cancer who underwent paired tumor–matched normal tissue sample clinical sequencing using the MSK-IMPACT assay (24). We performed integrated germline–somatic mutation data analyses to identify germline PV in lung cancer patients and correlated to individual and familial outcomes data to assess clinical and biological significance. Our study defined a subset of lung cancer patients at hereditary risk for a range of tumor types, including cases whose kindreds show clustering of TP53, ATM, or BRCA1/2-associated tumor types, as well as rarer subsets with lung tumor predispositions due to inherited mutations of EGFR, MEN1, or BAP1.

In total, we analyzed 5,118 patients with lung cancer at Memorial Sloan Kettering Cancer Center (MSKCC) who underwent tumor and normal DNA sequencing as part of their clinical care utilizing MSK-IMPACT (Integrated Mutation Profiling of Actionable Cancer Targets; ref. 24) to interrogate 468 cancer-related genes from January 2014 until August 2019. All patients provided written informed consent, and the study was approved by the MSKCC Institutional Review Board (NCT01775072) and conducted in accordance with recognized ethical guidelines. Baseline clinical data including gender, histologic subtypes, age at diagnosis (binned into five-year intervals), self-reported ancestry/ethnicity, smoking status (current/prior smoker vs. never-smoker), and stage (metastatic or not) for all patients were obtained from institutional electronic medical records (EMR). A selected 152 patients who had either a family history of any cancer, personal history of multiple primary cancers or early age at diagnosis (≤55 years old) provided consent to return of germline results for up to 88 cancer predisposition genes tested on the targeted gene panel including all cancer-susceptibility genes identified in American College of Medical Genetics and Genomics (ACMG) guidelines (25). The detailed clinical data and family history of any cancer were available for high-risk patients’ analysis. Patients with | $ \ge $ |2 primary cancers were identified through EMR database query, manually reviewed, and classified as multiple primaries based on IARC rules (26). Patients who received the platinum-based chemotherapy drugs cisplatin, carboplatin, or both, with or without other chemotherapies were identified through automated EMR database query.

Tumor and blood lymphocyte DNA from patients were sequenced using the MSK-IMPACT platform interrogating up to 468 cancer-associated genes. The analytical methods including germline and somatic variant calling were performed as described in Srinivasan et al. (27) and Jonsson et al. (28). The analytical approach is represented in the flow diagram (Supplementary Fig. S1). All patients were anonymized prior to germline variant pathogenicity assessment and linking of germline, somatic, and clinical data for the integrated analyses. The pathogenicity assessment for germline variants with <1% population allele frequency in the genome aggregate (gnomAD r2.1.1; ref. 29) database was performed according to ACMG criteria (25). The germline variants annotated as pathogenic/likely pathogenic in ClinVAR (January 2020; refs. 25, 30) were manually curated, and only high confidence (ClinVar gold star ≥ 2) pathogenic variants were reported here. Here, we identified and reported germline pathogenic variants in the full cohort including patients from European, Ashkenazi Jewish (AJ), Asian, African, and others/unknown ancestries. We excluded germline variants likely derived from clonal hematopoiesis or from tumor-derived circulating cell-free DNA as previously described (28). We further classified germline PV according to the current disease modeling into high-penetrance [relative risk (RR) of disease > 4]; moderate-penetrance (RR 2–4), low-penetrance (RR<2), uncertain penetrance, or associated with an autosomal recessive condition as previously described (20). The common truncating variant BRCA2 p.Lys3326Ter and missense variant ATM p. Leu2307Phe reported in lung cancer GWAS as a risk allele (12, 13) and CHEK2* p.Ile157Thr reported as protective allele (12) were excluded in the current analysis.

The high-risk patients were defined as a lung cancer patient who had either a strong family history of any cancers, or personal history of multiple primary tumors, or early age at diagnosis (≤55 years). We compared the prevalence of germline pathogenic variants aggregated by genes with high/moderate or low penetrance, genes with uncertain significance or recessive between high-risk patients and unselected patients and performed Fisher exact test to report the odds ratio and P value.

Tumor sequence data were analyzed as previously described to detect somatic mutations, small insertions and deletions (indels), DNA copy-number alterations, and select translocations (31). Loss of heterozygosity (LOH) in the tumor at the locus of the detected germline PV was evaluated using the FACETS algorithm (32). Modeling of zygosity and filtering of these calls was performed as previously described (27, 28). The zygosity for each germline variant in tumor was determined to be heterozygous, loss of the wild-type (WT) allele or loss of the mutant allele. Patients harboring germline PV were considered to have biallelic inactivation if either loss of WT allele was observed or a second somatic truncating mutation was detected in their tumors. Since our method could not phase a second somatic truncating mutation with a germline variant, we considered the zygosity of a germline variant as indeterminate if two or more detectable somatic pathogenic events (e.g., patients with germline PV who had both loss of WT allele and a somatic truncating mutation in tumors were excluded). We estimated the rate of biallelic inactivation as the number of germline mutation carriers who had biallelic inactivation compared with total number of evaluable germline mutation carriers (after excluding indeterminate cases). We next assessed the enrichment of biallelic inactivation (either loss of WT allele or a second somatic truncating mutation) in patients carrying germline PV compared with patients without any germline PV (germline PV absent) and carrying nonpathogenic variants such as benign/likely benign variants that did not have any functional consequence [benign/likely benign variants annotated in ClinVAR database (January 2020); refs. 25, 30]. The distribution of biallelic inactivation in germline benign/likely benign variants was determined in the corresponding tumor in a manner identical to that of pathogenic variants as described above. We annotated somatic mutations as oncogenic or likely oncogenic using OncoKB, a precision oncology knowledge base (28). We performed all analyses separately for histologic subtypes (lung adenocarcinoma or lung squamous cell carcinoma) since there are significant differences in mutational landscapes between them. We compared frequency of key somatic driver mutations, oncogenic fusions, and copy-number alterations in germline PV carriers and compared with patients without any germline mutations.

The binned clinical data were anonymized before linking to genomic data. Clinical characteristics of patients with germline PV aggregated by gene were compared with those without any germline PV using Fisher exact test. To assess the associations between germline PV carrier status and clinical characteristics or somatic driver mutations, logistic regression was used for binary phenotypes. All patients (including AJ patients) were analyzed for association tests, and self-reported ancestry was used as covariate in the multivariate model. Multiple linear regression was performed for quantitative traits to adjust for clinical variables including sex, self-reported ancestry (European, Asian, African, or others), smoking status (never vs. current/former smoker), age at diagnosis, and tumor stage (metastatic yes/no). All analyses were performed separately for patients with lung adenocarcinoma, lung squamous cell carcinoma, and SCLC. In total, 15 genes with ≥5 patients carrying germline PV were tested for association analysis, and using Bonferroni correction for multiple testing, the results with P < 0.003 were reported as statistically significant.

Overall survival (OS) data were available and analyzed for 1,876 patients with lung adenocarcinoma. OS was measured from the date of diagnosis to the date of death and censored at the date of last follow-up for survivors. Survival times were binned for the purpose of anonymization such that each time interval had more than 15 patients. To assess the association between OS and germline PV status, univariate Cox regression was performed with likelihood ratio tests. Because survival times were binned to preserve anonymization of data, point estimates for survival times were chosen randomly within the time interval bins. 100 replicates were performed to assure stability of estimates (Supplementary Fig. S2). The proportional hazards assumption was tested with Schoenfeld residuals. Kaplan–Meier curves were utilized to visualize survival differences using the R package “survival” in R software version 3.4.2. Multivariable Cox regression was performed to assess the independent association of germline PV grouped by specific genes with OS using age at diagnosis (binned as 5-year interval time), sex, smoking status (current/former smoker vs. never-smoker), stage (metastatic or not), self-reported ancestry (European, Asian, African, or others) and platinum-based chemotherapy (platinum-based chemotherapy/non-platinum/none). The OS analysis was not possible for the other subtypes because of insufficient sample size.

To estimate the carrier frequency of germline PV aggregated at gene level in the general population, we used the genome aggregate (gnomAD r2.1.1) database (29) as described in recent publications (33–35). We annotated germline variants in the gnomAD database as PV using the same variant interpretation pipeline as applied to our patient cohort. The carrier frequency of germline PV in patients with lung adenocarcinoma or with squamous cell carcinoma from European ancestry was computed separately and compared with the estimated carrier frequencies computed using non-Finnish European individuals from gnomAD. We excluded patients with self-reported AJ ancestry and carriers of AJ and European founder mutations for this comparison (burden analysis).

Raw genomic data for this study were generated at Molecular Diagnostics Service in the Department of Pathology, the Integrated Genomics Operation Core, and Marie-Josée and Henry R. Kravis Center for Molecular Oncology at MSKCC (ClinicalTrials.gov Identifier: NCT01775072). Derived summary data supporting the findings of this study are available from the corresponding author upon request.

We evaluated a cohort of 5,118 patients with advanced lung cancer consenting to a research study utilizing tumor–normal clinical sequencing targeted to a panel of 468 cancer-associated genes (MSK-IMPACT assay; ref. 24). Germline–somatic variant and clinical data were anonymized for this study. Clinical characteristics of the patients are summarized in Table 1. Most patients (n = 3,886; 75.9%) had a diagnosis of lung adenocarcinoma (LUAD); 517 (10.1%) cases had lung squamous cell carcinoma (LUSQ) and 275 (5.4%) cases had SCLC (Table 1). Smoking data were available for 91% of the patients. Most SCLC and LUSQ patients were current or former smokers (ever-smokers) in our cohort. Among those diagnosed with LUAD, 2,611 (67.5%) patients were classified as ever-smokers. A selected 152 patients who had either a family history of any cancer, personal history of multiple primary tumors, or early age at diagnosis (≤55 years) were consented to receive both tumor and germline sequencing results and identified as high-risk patients (Supplementary Table S1). Of these, 79% had a family history of any cancer. The median age at diagnosis for these 152 patients was 64 years compared with 70 years for the unselected cohort (P < 0.002).

In our cohort of 5,118 patients with lung cancer, 584 harbored one or more germline PV in 49 genes, including 462 patients diagnosed with LUAD, 57 patients with squamous cell carcinoma, 29 patients with SCLC and 34 patients who had other histologic subtypes (Supplementary Table S2). Of all germline PV carriers, 222 (4.3%) harbored germline PV in high/moderate-penetrance genes, 152 harbored uncertain penetrance PV and 210 patients carried low penetrance or recessive PV; 36 patients had two or more germline PV (Fig. 1A; Supplementary Table S3). Thirty-five patients harbored germline PV in genes previously associated with risk for lung cancer including PARK2 p.Arg275Trp (n = 21), TP53 (n = 7), BAP1 (n = 3), EGFR p.Thr790Met (n = 2), and MEN1 (n = 2; Fig. 1B). In addition to lung cancer risk genes, germline PV were most frequently observed in DNA damage repair (DDR) pathway genes including high/moderate-penetrance genes BRCA2 (n = 54), BRCA1 (n = 34), CHEK2 (n = 30) and ATM (n = 26; Fig. 1B; Supplementary Table S2). Almost all patients carrying ATM germline PV were diagnosed with LUAD (n = 24/25). The frequency of germline PV in BRCA2 was significantly higher in lung squamous cell carcinoma (1.9%) compared with other histologies after adjusting for age at diagnosis, gender, smoking status, stage, and ethnicity (P = 0.03).

Among 152 high-risk lung cancer patients who had either family history of any cancer, personal history of multiple primary tumors, or early age of diagnosis (≤55 years old), 26 patients harbored germline PV; of these 25 patients and their families were referred for screening and counseling. A significantly higher proportion of germline PV in high/moderate-penetrance genes were detected in patients clinically identified as high-risk compared with patients in the unselected cohort (10.5% vs. 4.1%; odds ratio = 3.0; 95% CI, 1.7–5.2, P = 1.7e−04; Supplementary Table S4).

The biological impact of germline PV on lung tumors was assessed using Knudson's two-hit hypothesis for biallelic inactivation (36). Biallelic inactivation of germline PV was considered if either the WT allele was lost in the tumor (LOH) or a second somatic truncating mutation was observed. We excluded germline PV carriers with two or more detectable somatic pathogenic events in the same gene (e.g., presence of both somatic truncating mutation and loss of WT allele) to eliminate the bias that biallelic inactivation may be resulted by somatic mutations. Among patients with LUAD, biallelic inactivation was observed in 75% of patients harboring germline PV in high-penetrance lung cancer predisposing tumor suppressors (TP53, BAP1, and MEN1 in Fig. 2A). Biallelic inactivation in four DDR genes (ATM, BRCA2, CHEK2, and MRE11A) was observed in greater than 40% of carriers, whereas only 14% of BRCA1 germline PV carriers showed biallelic inactivation (Fig. 2A) in patients with LUAD. ATM and CHEK2 had a significantly higher rate of biallelic inactivation in LUAD patients harboring germline PV compared with patients harboring benign/likely benign variants (56.5% vs. 26.2% in ATM and 51.8% vs. 20.4% in CHEK2; P <0.005 statistically significant after correcting for multiple testing; Fig. 2B). In the stratified analysis using smoking history of LUAD patients, we observed that among never smokers, BRCA2 showed a marginally higher rate of biallelic inactivation in patients carrying germline PV compared with patients with benign/likely benign variants at the same locus (47.3% vs. 25.4%, P = 0.06; Supplementary Table S5). The biallelic inactivation in lung squamous cell carcinoma patients carrying germline PV in TP53, RAD51D, BRIP1, and BRCA2 was observed in 60% of the patients (Fig. 2C).

Tumor mutation burden and the frequencies of somatic driver mutations or somatic copy-number alterations were not significantly different between patients with and without germline PV after correcting for multiple testing.

Twenty-eight LUAD patients who carried germline PV in lung cancer risk genes (TP53, BAP1, MEN1, EGFR, and PARK2) had significantly younger age at diagnosis compared with patients without any germline PV, after adjusting for gender, smoking status, ethnicity, and tumor stage (P = 0.0008; Fig. 3A). Excess of germline PV in BRCA1 and BRCA2 were observed in LUAD patients who were never smokers compared with current/former smokers, based on a multivariate model (adjusted odds ratio = 1.3; 95% CI, 1.1–1.6; P < 0.001; Fig. 3B). The germline PV status was not associated with other clinical features in patients with LUAD or other histologies. We observed that the prevalence of germline PV in patients from AJ ancestry was significantly higher than that observed in patients from European, Asian, or African ancestry, reflective of the presence of founder mutations in patients from AJ ancestry (Supplementary Tables S6 and S7).

An exploratory OS analysis was performed for 1,876 advanced-stage LUAD patients using carrier status for germline PV. The mutational status of high/moderate-penetrance genes or DDR pathway genes (combined) was not associated with OS. However, gene-centric analysis showed that the carrier status for PV in BRCA2 was associated with improved OS (HR = 0.29; 95% CI, = 0.07–0.99; P = 0.03; Figure 4A). In multivariable Cox regression analysis stratified by smoking history, which did not satisfy the proportional hazards assumption (Supplementary Table S8), and adjusted by age at diagnosis, sex, tumor stage (metastatic yes/no), self-reported ancestry/ethnicity, and receipt of platinum-based chemotherapy (cisplatin or carboplatin treated/other chemotherapy/none), the BRCA2 germline PV carriers were marginally associated with OS (adjusted HR = 0.26; 95% CI = 0.07–1.0; P = 0.06; Fig. 4B); additional studies are needed. The germline mutational status of other genes including ATM, BRCA1, or CHEK2 was not associated with OS (Supplementary Fig. S3).

We observed that the carrier frequencies of germline PV in genes ATM and BRCA2 in patients with LUAD from European ancestry were statistically higher than expected frequencies observed in individuals from the non-Finnish European population in the gnomAD database (odds ratio = 2.3; 95% CI = 1.3–3.6 for ATM and odds ratio = 2.2; 95% CI = 1.3–3.4 for BRCA2; P = 0.002; Supplementary Table S9). We also observed a higher than expected frequency of germline PV in BRCA2 in lung squamous cell carcinoma patients compared with individuals from the non-Finnish European population in the gnomAD population database (refs. 33, 37; odds ratio = 5.8; 95% CI = 2.3–12.2, P = 0.0003) (Supplementary Table S9).

Here, we report a large integrated analysis of germline and somatic mutations in 5,118 patients with advanced lung cancer, identifying germline PV in high- or moderate-penetrance genes in 4.3% of patients. A significantly increased prevalence of germline PV in high/moderate-penetrance genes was observed in patients with either a family of any cancers, personal history of multiple primary tumors, or early age of diagnosis (≤55 years age). In our cohort, patients harboring germline PV in lung cancer predisposing genes (TP53, EGFR, BAP1, and MEN1) had earlier age at diagnosis compared with patients without germline PV; 75% of these with germline PV in the tumor suppressor gene (TP53, BAP1, and MEN1) demonstrated biallelic inactivation in their tumors. Germline PV in BAP1, a mesothelioma susceptibility gene (38), was identified in three LUAD patients, with two of them demonstrating LOH in their tumors, suggesting a role of BAP1 in a subset of LUAD.

The most frequently mutated genes were associated with the DDR pathway, including ATM, BRCA1, BRCA2, and CHEK2, that were typically associated with breast, ovarian, pancreatic, prostate, and other cancer susceptibility (39–42). Our data support previous findings that ATM and BRCA2 constitute lung cancer susceptibility genes (13, 16, 18, 43). The observation that among LUAD cases, 56.5% of ATM germline PV carriers, 51.8% of CHEK2 germline PV carriers, and 45% of BRCA2 germline PV carriers demonstrated biallelic inactivation in their tumors supports their biological roles in lung cancer etiology. We observed higher than expected frequencies of germline PV in ATM and BRCA2 in patients with LUAD, and BRCA2 in patients with lung squamous cell carcinoma. BRCA1 and BRCA2 heterozygotes with LUAD were more likely to be never smokers. LUAD patients with germline BRCA2 mutations had an improved OS presumably due to improved outcome following platinum therapies; however, additional studies are needed due to limited sample size to clarify the interactions between germline PV and treatment response in lung cancer patients. Recently, a report of 87 patients with SCLC demonstrated that the subset with inherited mutations in DDR genes including BRCA2 had more favorable responses to platinum-based chemotherapy (43). Our findings also suggested that germline PV in BRCA1 and BRCA2 in lung cancer patients might have distinct consequences on lung tumors. Although 193 lung cancer patients harbored germline PV in high/moderate-penetrance genes in the DDR pathway, it also remains unclear if such patients will benefit from trials of targeted therapies, as observed for patients with DDR deficiencies and breast, ovarian, pancreatic, prostate, and other cancer types (28).

The observed prevalence of all germline PV in the current NSCLC patient cohort (n = 584; 11.4%) is somewhat higher than that reported in the larger pan-cancer study led by TCGA (8%; ref. 44), and similar prevalence in SCLC was reported in a recent study (43). Given that the TCGA study was primarily based on early-stage resected cancers, the difference may reflect an ascertainment bias given the high proportion of lung cancer patients with advanced metastatic disease referred for tumor profiling in our cohort; we have previously shown an enrichment in germline PV in metastatic compared with nonmetastatic tumors of various types (20). A limitation of the current study was the incomplete annotation of family history typically assessed at time of genetic counseling, as the vast majority of lung cancer patients reported here, in contrast to patients with other cancer types seen at our center (20), did not receive germline findings or follow-up genetic counseling. Sequencing analysis was also limited to 468 cancer-associated genes; however, this included 88 known ACMG cancer predisposition genes. We utilized publicly available population data (gnomAD r2.1.1) to compute the estimated carrier frequencies of germline PV by gene as previously described in recent publications (33–35); however, we acknowledge the limitation of such methods due to technical differences in sequencing platforms/bioinformatic pipelines and the presence of population stratification as confounders. We reported survival analysis for only 1,876 patients with advanced LUAD due to limited data availability. Additional studies are warranted to address the effects of germline PV status on response to therapy and clinical outcome.

In conclusion, our study systemically characterized the germline PV in cancer-predisposing genes in patients with lung cancer, emphasizing the role of genetic counseling for unaffected family members who may benefit from proactive surveillance. The study expands the known lung cancer susceptibility genes to include ATM and BRCA2, in addition to TP53, EGFR, and MEN1, and provides supporting evidence for BAP1, a gene previously associated with mesothelioma. Additional research is warranted to further establish the criteria for genetic counseling in lung cancer patients to improve precision oncology practices. Screening with low-dose CT may be incorporated into the care of the high-risk patients with lung cancer and their unaffected relatives carrying germline mutations independent of smoking history. Our study also demonstrates the biological importance of germline mutations in DDR genes in patients with advance lung cancer. Such patients may benefit from personalized therapeutic modalities as several DDR inhibitors have been developed and are either approved for other cancer types (e.g., PARP inhibitors) or are in clinical trials. Additional studies will provide a better understanding of response to such therapies in lung cancer cohorts defined by germline PV status. The identification of germline PV in highly penetrant, clinically actionable genes in 4.3% of patients with lung cancer, particularly those with early age at diagnosis (≤55 years), multiple primary tumors, or a family history of cancer, is important given the high global incidence of lung cancer and the potential benefits of cascade testing for relatives (45) to guide surveillance for early cancer detection.

S. Mukherjee reports grants from Robert and Kate Niehaus Center for Inherited Cancer Genomics, Andrew Sabin Family Foundation, and the Sharon Levine Corzine Research Fund, and NCI Core grants P30 CA008748 and 1R01CA227534-01A1 during the conduct of the study. M.D. Hellmann reports grants and personal fees from BMS, personal fees from Achilles, Adagene, Adicet, Arcus, AstraZeneca, Blueprint, DaVolterra, Eli Lilly, Genentech/Roche, Genzyme/Sanofi, Janssen, Immunai, Instil Bio, Mana Therapeutics, Merck, Mirati, Natera, Pact Pharma, Shattuck Labs, and Regeneron during the conduct of the study; in addition, M.D. Hellmann has a patent filed by Memorial Sloan Kettering related to the use of tumor mutational burden to predict response to immunotherapy (PCT/US2015/062208) pending and licensed by PGDx pending and licensed to PGDx; and as well as equity options from Factorial, Immunai, Shattuck Labs, Arcus, and Avail Bio. Subsequent to the completion of this work, Hellmann began as an employee (and equity holder) at AstraZeneca. H. Rizvi reports subsequent to the completion of this work, H.R. began as an employee at AstraZeneca. M.G. Zauderer reports grants from NCI during the conduct of the study; grants and personal fees from Takeda, GSK, personal fees from Novocure, Aldeyra, Ikena, grants from Epizyme, Polaris, Sellas Life Sciences, BMS, Curis, and Atara outside the submitted work; and Chair, Board of Directors, Mesothelioma Applied Research Foundation, uncompensated. S. Topka reports grants from NIH/NCI, Sharon Levine Corzine Research Fund, Andrew Sabin Family Foundation, and Cycle for Survival during the conduct of the study; in addition, S. Topka has a patent for use of Illudin class of alkylating agents in patients harboring mutations in the ERCC3 gene (PCT/US2018/022588) issued. P. Srinivasan reports current employment at Natera Inc. M. Esai Selvan reports grants from LUNGevity Foundation and R33 during the conduct of the study. M.I. Carlo reports personal fees from Suo and Onclive outside the submitted work. K.A. Cadoo reports personal fees from Astra Zeneca, MJH Life Sciences, MSD Ireland, other support from MSD Ireland, GSK Ireland, Immunogen, nonfinancial support from Pfizer, nonfinancial support from Roche Ireland, and personal fees from Nextcure outside the submitted work. J.G. Hamilton reports grants from NCI during the conduct of the study. Y.L. Liu reports grants from REPARE Therapeutics, GSK, and AstraZeneca outside the submitted work. S.M. Lipkin has received stock from AnaNeo Therapeutics, which has no relationship to this study. Z.H. Gümüş reports grants from LUNGevity Foundation and NIH R33 CA263705-01 during the conduct of the study. D.B. Solit reports personal fees from Pfizer, Loxo/Lilly Oncology, Vividion Therapeutics, FORE Therapeutics, Scorpion Therapeutics, and BridgeBio outside the submitted work. D.R. Jones reports other support from AstraZeneca and Merck outside the submitted work. M.G. Kris reports personal fees from Janssen, AstraZeneca, Pfizer, Daiichi-Sanyko, Novartis, Sanofi, and other support from Genetech outside the submitted work. J. Vijai reports a patent for diagnosis and treatment of ERCC3-mutant cancer pending. Z.K. Stadler reports other support from Alcon, Adverum, Gyroscope Therapeutics, RegenexBio, and Neurogene outside the submitted work. C.I. Amos reports grants from NCI and Cancer Prevention Research Institute of Texas during the conduct of the study. B.S. Taylor reports grants from Genentech, Inc., personal fees from Boehringer Ingelheim, and Loxo Oncology at Lilly outside the submitted work. M.F. Berger reports personal fees from Eli Lilly and PetDx outside the submitted work. C.M. Rudin reports personal fees from AbbVie, Amgen, Astra Zeneca, Daiichi Sankyo, Epizyme, Genentech/Roche, Ipsen, Jazz, Kowa, Lilly, Merck, Syros, Bridge Medicines, Earli, and Harpoon Therapeutics outside the submitted work. No disclosures were reported by the other authors.

S. Mukherjee: Conceptualization, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. C. Bandlamudi: Resources, data curation, software, formal analysis, methodology, writing–review and editing. M.D. Hellmann: Conceptualization, resources, data curation, funding acquisition, project administration, writing–review and editing. Y. Kemel: Data curation, formal analysis, project administration, writing–review and editing. E. Drill: Software, formal analysis, visualization, methodology, writing–review and editing. H. Rizvi: Data curation, validation, project administration. K. Tkachuk: Resources, data curation, project administration. A. Khurram: Data curation, project administration. M.F. Walsh: Validation, investigation, methodology, writing–review and editing. M.G. Zauderer: Resources, data curation, funding acquisition, writing–review and editing. D. Mandelker: Resources, data curation, software, validation, investigation, project administration. S. Topka: Data curation, validation, writing–review and editing. A. Zehir: Resources, data curation, software, methodology, project administration. P. Srinivasan: Resources, data curation, software, formal analysis, validation, methodology, writing–review and editing. M. Esai Selvan: Validation, writing–review and editing. M.I. Carlo: Resources, validation, writing–review and editing. K.A. Cadoo: Data curation, writing–review and editing. A. Latham: Resources, data curation, writing–review and editing. J.G. Hamilton: Writing–review and editing. Y.L. Liu: Resources, data curation, investigation, methodology, writing–review and editing. S.M. Lipkin: Resources, data curation, writing–review and editing. S. Belhadj: Data curation, writing–review and editing. G.L. Bond: Resources, data curation, writing–review and editing. Z.H. Gümüş: Resources, validation, writing–review and editing. R.J. Klein: Resources, validation, writing–review and editing. M. Ladanyi: Resources, data curation, software, supervision, funding acquisition, writing–review and editing. D.B. Solit: Resources, data curation, supervision, funding acquisition, writing–review and editing. M.E. Robson: Resources, data curation, funding acquisition, writing–review and editing. D.R. Jones: Resources, funding acquisition, project administration, writing–review and editing. M.G. Kris: Resources, funding acquisition, writing–review and editing. J. Vijai: Resources, writing–review and editing. Z.K. Stadler: Resources, data curation, supervision, funding acquisition, writing–review and editing. C.I. Amos: Validation, methodology, writing–review and editing. B.S. Taylor: Resources, software, investigation, methodology, writing–review and editing. M.F. Berger: Resources, supervision, funding acquisition, investigation, methodology, project administration, writing–review and editing. C.M. Rudin: Resources, data curation, supervision, funding acquisition, writing–review and editing. K. Offit: Conceptualization, resources, data curation, supervision, funding acquisition, investigation, methodology, writing–original draft, project administration, writing–review and editing.

K. Offit was supported by the Robert and Kate Niehaus Center for Inherited Cancer Genomics, the Andrew Sabin Family Foundation, and the Sharon Levine Corzine Research Fund; MSKCC is supported by the NCI Core grant P30 CA008748. C.I. Amos is a research scholar of the Cancer Prevention and Research Institute of Texas and supported by RR170048 and U19CA203654; M.F. Berger received the funding NIH/NCI 1R01CA227534–01A; D.R. Jones received the NIH/NCI awards 1R01 CA217169 and 1R01 CA234617. D.B. Solit received the grant Cycle for Survival and Marie-Josée and Henry R. Kravis Center for Molecular Oncology; and Z.H. Gümüş received award from the LUNGevity Foundation Fund. We would like to thank all individuals who participated in this study. We gratefully acknowledge the members of the Molecular Diagnostics Service in the Department of Pathology and the use of the Integrated Genomics Operation Core and Clinical Genetics Service at Memorial Sloan Kettering Cancer Center.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.