Elucidating the genomic architecture of Asian EGFR-mutant lung adenocarcinoma through multi-region exome sequencing

Abstract

EGFR-mutant lung adenocarcinomas (LUAD) display diverse clinical trajectories and are characterized by rapid but short-lived responses to EGFR tyrosine kinase inhibitors (TKIs). Through sequencing of 79 spatially distinct regions from 16 early stage tumors, we show that despite low mutation burdens, EGFR-mutant Asian LUADs unexpectedly exhibit a complex genomic landscape with frequent and early whole-genome doubling, aneuploidy, and high clonal diversity. Multiple truncal alterations, including TP53 mutations and loss of CDKN2A and RB1, converge on cell cycle dysregulation, with late sector-specific high-amplitude amplifications and deletions that potentially beget drug resistant clones. We highlight the association between genomic architecture and clinical phenotypes, such as co-occurring truncal drivers and primary TKI resistance. Through comparative analysis with published smoking-related LUAD, we postulate that the high intra-tumor heterogeneity observed in Asian EGFR-mutant LUAD may be contributed by an early dominant driver, genomic instability, and low background mutation rates.

Fig. 1

Landscape of clonal and subclonal mutations in Asian EGFR-mutant tumors. a Phylogenetic trees generated for the 16 Asian EGFR-mutant LUADs. Trunks, branches and tips are depicted in blue, green, and red, respectively, while non-silent mutations carrying LUAD specific drivers are in red and other cancer drivers are in blue (Methods section). All patients are never-smokers except A103 (marked *) who was a light ex-smoker. Truncal mutation burden followed by total mutation burden for all sectors is indicated below the trees. b Bar plot representing truncal and non-truncal mutation burden per sector. c Oncoprint heatmap for mutations in LUAD drivers depicting the presence (see color legend) or absence (gray box) and type of non-silent mutation. d Proportions of the three mutation signatures identified for each sector. Signature numbers are according to the COSMIC nomenclature. e Pie charts representing contribution of the three mutation signatures in early (trunk) and late (branch/private) mutations

Fig. 2

Genomic instability and variegated copy number landscape of EGFR-mutant tumors. a Bar plot representing the fraction of genome altered by copy number alterations relative to ploidy of the sector, which is termed as the genomic instability index (GII). b Bar showing genome doubling status. Blue indicates significant evidence for genome doubling (Supplementary Data 3, Methods section) and gray indicates no genome doubling. c Heatmap depicting gains in known driver or recurrently amplified cytobands in LUAD (Methods section). Light red represents gain in one copy beyond the ploidy while dark red represents gain in ≥2 copies beyond the ploidy. d Heatmap depicting losses in known driver or recurrently amplified cytobands in LUAD (Methods section). Light blue represents loss of one copy relative to the ploidy while dark blue represents loss of ≥2 copies beyond the ploidy. Samples for all panels are as depicted in d. TP53 wild-type samples are depicted in gray

Fig. 3

EGFR is a dominant driver with few co-drivers. Dot plots comparing mutation burden on a trunk and b branches between the Asian EGFR-mutant and smoker dominated Caucasian cohorts^, . c Dot plot comparing number of branch/private mutated drivers (extended driver list) between the EGFR-mutant and smoker dominated Caucasian cohorts^, . Welch’s t-test was used to compare the two groups. d Dot plot showing that EGFR-mutant LUADs have significantly fewer truncal drivers (extended driver list; Welch’s t-test) compared to smoker Caucasian LUADs. Three random sectors were picked 20 times iteratively and averages of the iterations are represented as circles in a–d. Horizontal line indicates the median for that cohort. e “Driver dominance score” measuring driver self-sufficiency for each of 78 LUAD driver genes calculated across published 412 tumors^, is plotted against the fraction of patients carrying the mutated driver

Fig. 4

TP53 mutations, genomic instability, high-driver burden lead to poor outcome. a Lower panel is a heatmap representing number of copies for selected genes involved in EGFR TKI resistance or associated with prognosis. Upper panel represents features of a tumor which are associated with patient outcome like TP53 mutation status, genomic instability index, presence of whole-genome doubling, above and below median number of drivers (LUAD specific or extended driver list) and the relapse status. All these features tend to coincide in many tumors. b Total mutation burdens and c driver burdens (extended driver list) are compared between TP53 mutant (mt) and wild-type (wt) tumors. Three random sectors were picked iteratively (n = 20) and averages across iterations are represented in b and c. The first p-value is taking all 16 patients into consideration and the second p-value is after eliminating the outlier A102 in the analysis. P-values are calculated using Welch’s t-test. d Survival plots using TCGA LUAD EGFR-mutant cases (those with non-silent mutations in tyrosine kinase domain, n = 26) after stratifying above or below median number of LUAD drivers (median = 3). P-value from χ²-test is indicated

Fig. 5

Schematic of evolutionary trajectories in smokers and EGFR-mutant non-smoker tumors. a In smokers, accumulation of many truncal mutations and drivers before branched evolution. Top: driver mutations are represented as colored rectangles. The dashed line indicates the point of diversification on the tree or the last common ancestral clone (carrying the last truncal driver mutation). Middle: shaded background represents clonal expansion (y-axis) as mutations accumulate (x-axis) in individual clones (colors). Bottom: schematic representation of phylogenetic mutation tree. b The non-smoker scenario where a dominant driver like EGFR is hit early leading to a big fitness advantage, fewer clonal sweeps and early diversification. Panel structure as in a. In both scenarios, evolutionary trajectory is influenced by clonal dynamics, which in turn is related to competitive fitness of the individual cell populations. Example of such factors include but are not limited to mutations rates, driver nature, cytokine milieu, immune cell infiltration, and metabolic conditions