Somatic Genomics and Clinical Features of Lung Adenocarcinoma: A Retrospective Study

Abstract

Background: Lung adenocarcinoma (LUAD) is the most common histologic subtype of lung cancer and has a high risk of distant metastasis at every disease stage. We aimed to characterize the genomic landscape of LUAD and identify mutation signatures associated with tumor progression.

Methods and findings: We performed an integrative genomic analysis, incorporating whole exome sequencing (WES), determination of DNA copy number and DNA methylation, and transcriptome sequencing for 101 LUAD samples from the Environment And Genetics in Lung cancer Etiology (EAGLE) study. We detected driver genes by testing whether the nonsynonymous mutation rate was significantly higher than the background mutation rate and replicated our findings in public datasets with 724 samples. We performed subclonality analysis for mutations based on mutant allele data and copy number alteration data. We also tested the association between mutation signatures and clinical outcomes, including distant metastasis, survival, and tumor grade. We identified and replicated two novel candidate driver genes, POU class 4 homeobox 2 (POU4F2) (mutated in 9 [8.9%] samples) and ZKSCAN1 (mutated in 6 [5.9%] samples), and characterized their major deleterious mutations. ZKSCAN1 was part of a mutually exclusive gene set that included the RTK/RAS/RAF pathway genes BRAF, EGFR, KRAS, MET, and NF1, indicating an important driver role for this gene. Moreover, we observed strong associations between methylation in specific genomic regions and somatic mutation patterns. In the tumor evolution analysis, four driver genes had a significantly lower fraction of subclonal mutations (FSM), including TP53 (p = 0.007), KEAP1 (p = 0.012), STK11 (p = 0.0076), and EGFR (p = 0.0078), suggesting a tumor initiation role for these genes. Subclonal mutations were significantly enriched in APOBEC-related signatures (p < 2.5×10-50). The total number of somatic mutations (p = 0.0039) and the fraction of transitions (p = 5.5×10-4) were associated with increased risk of distant metastasis. Our study's limitations include a small number of LUAD patients for subgroup analyses and a single-sample design for investigation of subclonality.

Conclusions: These data provide a genomic characterization of LUAD pathogenesis and progression. The distinct clonal and subclonal mutation signatures suggest possible diverse carcinogenesis pathways for endogenous and exogenous exposures, and may serve as a foundation for more effective treatments for this lethal disease. LUAD's high heterogeneity emphasizes the need to further study this tumor type and to associate genomic findings with clinical outcomes.

Conflict of interest statement

The authors have declared that no competing interests exist.

Figures

Fig 1. Somatic mutations of lung adenocarcinoma in EAGLE data.

(A) Distribution of point somatic mutations across nine mutation types. (B) The top panel shows the number of nonsilent mutations detected by whole-exome analysis for 101 EAGLE samples. Tumor samples were arranged from left to right by the number of nonsilent mutations. The middle panel shows the mutations for previously reported significantly mutated genes based on the TCGA data, reported in the TumorPortal website. The next panel shows the mutations for the three new driver genes. The bottom panels show smoking status. The right panel shows the frequency of nonsilent mutations in EAGLE data for each driver gene. Each column represents one patient.

Fig 2. Somatic mutations in three LUAD candidate driver genes (POU4F2, ZKSCAN1, and ASEF) in EAGLE, TCGA and Broad Institute studies.

The protein sequences from these three genes are schematically described using grey bars along with their respective structural and functional domains in color-coded blocks. Each mallet represents an independent nonsilent mutation with potential functional relevance in the three studies (the complete list of mutations is reported in S1 Table). Numbers below each sequence representation mark the total length of the transcript, the domain ranges, and the locations of mutations.

Fig 3. The associations between DNA methylation and somatic mutation signatures based on EAGLE and TCGA data.

(A) The number of CpG probes significantly associated with the TNSM and the fractions of various types of point mutations (p < 1.5×10−7, based on Bonferroni correction). (B) CpG probe cg00042837 was strongly associated with TNSM, the fractions of C→A mutations, C→T mutations, and transversions. Each point represents one sample. The blue line was generated by “lowess,” a nonparametric statistical procedure for nonlinear regression. (C) The enrichment fold change of CpG probes mapping to different categories in the association with somatic point mutation types. “CGI” represents CpG island regions; “NonCGI” includes shore and shelf regions. (D) The enrichment fold change of CpG probes mapping to different gene regions in the association with point somatic mutation types. (E) and (F) show The proportion of identified CpG probes showing positive associations with different somatic point mutation types.

Fig 4. Clonal and subclonal point mutations in EAGLE data. Mutations in amplification regions were not included in the analysis.

(A) The number of clonal and subclonal mutations in 37 driver genes for lung adenocarcinoma. (B) Fraction of clonal and subclonal mutations in each of the nine point mutation types. (C) The fraction of APOBEC-mediated mutations significantly differed in clonal and subclonal mutations. (D) Estimated fraction of subclonal mutations for each sample. (E) Estimated fractions of subclonal mutations for patients at different tumor stages.

Fig 5. Mutual exclusivity of driver genes detected in 825 patients combining TCGA, Broad Institute, and EAGLE WES of lung adenocarcinoma.

(A) A MEGS with six genes covering 60.3% of patients. Samples without nonsynonymous mutations in these six genes are not shown. Samples labelled as blue carry a nonsynonymous mutation in the gene region, while samples labelled as gray do not carry a synonymous mutation in the gene region. (B) A MEGS with four genes covering 33.3% of patients. Samples without nonsynonymous mutations in these four genes are not shown.

Fig 6. Association between genomic features and clinical outcomes.

(A) The mutational status of TP53 and KRAS and the time of developing distant metastasis. p-values were two-sided. Red: mutated; blue: not mutated. (B) The association between the fraction of nine point mutation types and overall transversions and the time of developing distant metastasis after initial diagnosis. Relative risks and their 95% confidence intervals were estimated based on a Cox regression model adjusted for age, sex, and disease stage. p-values were two-sided. (C) Cancer-free survival was not associated with the mutational status of TP53 or KRAS. p-values were two-sided. Red: mutated; blue: not mutated.