Tumor evolutionary trajectories during the acquisition of invasiveness in early stage lung adenocarcinoma

Lung adenocarcinoma (LUAD) is the most common subtype of lung cancer and a leading cause of cancer death. Despite increased detection of malignant pulmonary nodules (MPNs) via CT screening, early-stage LUAD remains prognostically heterogeneous. Invasive components such as minimally invasive adenocarcinoma (MIA) and invasive adenocarcinoma (IAC) are important prognostic discriminators, yet the initiation, early progression, and evolutionary patterns of invasive components are poorly understood. Although a linear progression from atypical adenomatous hyperplasia (AAH) to adenocarcinoma in situ (AIS) and then IAC has been proposed, the evolutionary trajectory from pre-invasive to invasive LUAD within the same lesion has not been fully elucidated. EGFR and KRAS are frequent drivers with differing prognostic implications in early-stage disease. This study aims to delineate driver molecular events, characterize genetic heterogeneity between paired pre-invasive and invasive components within MPNs, define early evolutionary trajectories, and assess their prognostic relevance, with particular focus on the role of truncal EGFR mutations and the tumor microenvironment.

Prior studies reported significant genetic differences among AAH, AIS, and LUAD, and hypothesized a linear progression from pre-invasive to invasive states; however, paired pre-invasive and invasive components within a single MPN had not been jointly analyzed. Early-stage EGFR-mutated NSCLC generally shows better outcomes than KRAS-mutated cases, and invasive component status is a strong prognostic marker. Genome-wide somatic analyses have advanced understanding of LUAD evolution, but the influence of dominant driver genes (e.g., EGFR, KRAS, STK11) on the transition from pre-invasive to invasive states, and their interaction with the immune microenvironment, remained unclear.

Study design and cohorts: 53 T1-stage LUAD cases with malignant pulmonary nodules (MPNs) underwent micro-dissection to separate paired pre-invasive and adjacent invasive components. In total, 113 MPN components (including 61 of 69 MPNs concordant for microdissection), 8 whole MPNs, 5 metastatic lymph nodes (MLNs), and 29 cDNA/cfDNA specimens were sequenced. Additional cohorts included 496 T1-stage patients from the Boston Lung Cancer Study (BLCS) for prognosis analyses and TCGA LUAD data for immune infiltration and mutation abundance analyses. Sequencing: Two phases of targeted sequencing were performed. Phase 1 used a 1021-gene panel on 6 whole MPNs and 35 micro-dissected components; Phase 2 used a 425-gene panel on 2 whole MPNs and 78 micro-dissected components. Somatic mutations were identified across components. Variant calling pipeline included quality control with Trim Galore, alignment to GRCh37 with BWA, PCR deduplication (Picard), local realignment and variant calling with GATK/Mutect2, with manual review in IGV. Copy-number alterations were inferred using Control-FREEC, with amplification defined as CN ≥ 6 and gene loss as CN ≤ 0. Phylogenetic analysis: Somatic alterations were assigned to trunks (shared) or branches (private) between paired pre-invasive and invasive components using Bayesian inference with Treemix-based approaches to construct phylogenetic trees and quantify intratumor heterogeneity (ITH). Driver events in trunk and branches were annotated by known oncogenic pathways. Immune infiltration and IHC: TIMER-based deconvolution of TCGA LUAD RNA-seq data estimated tumor-infiltrating lymphocytes (B cells, CD4+ T cells, CD8+ T cells, macrophages, dendritic cells). Immunohistochemistry on serial sections evaluated CD20+ B cells and CD3+ T cells in paired components from EGFR-mutated cases. Circulating cfDNA: Somatic mutations were assessed in 12 pre-operation and 11 post-operation cfDNA samples. Statistics: ITH comparisons used Kruskal–Wallis or Wilcoxon tests; mutual exclusivity by Fisher’s exact test; survival by Kaplan–Meier and log-rank tests; dN/dS ratios computed across trunks and branches.

- Mutation burden: 1–34 somatic mutations per MPN component (median 8) in Phase 1; 1–15 (median 4) in Phase 2. EGFR was the most frequently mutated gene; TP53, MED12, and ERBB2 were notable drivers. EGFR L858R was the most recurrent variant and was more frequent in females. No significant differences in driver frequencies were observed between pre-invasive and invasive components; invasive components had a higher proportion of C>G transversions. - Evolutionary trajectories: Three modes were identified among paired components. EM1: no shared driver mutations between pre-invasive and invasive components (indicative of distinct drivers). EM2: presence of truncal drivers with private branch alterations; subdivided into EM2A (branching evolution) and EM2B (linear-like evolution). EM3: strong divergence between pre-invasive and invasive component mutations suggesting inter-clonal evolutionary shifts. In 62% (31/50) of MPNs, evolution was branched (EM1 and EM2A), while EM2B was consistent with linear progression. - Intratumor heterogeneity: EM1 showed the highest ITH; EM2B the lowest. In EM2B, invasive components exhibited significantly higher ITH than adjacent pre-invasive components, consistent with clonal expansion during early progression. - Truncal drivers and pathways: Truncal mutations were enriched for known drivers, notably EGFR, TP53, KRAS, and STK11, with RTK–RAS pathway genes contributing most to trunks. EGFR, CDK4, and TP53 were frequent in invasive branch mutations; TP53 was prominent in pre-invasive branches. In EM2, the proportions of truncal vs branching mutations differed significantly for KRAS, STK11 (P=0.035 each), and EGFR (P=6.96×10^-8). - Mutual exclusivity and prognosis: Truncal EGFR mutations were mutually exclusive with truncal KRAS and STK11 in the JSCH cohort and validated in BLCS and TCGA. In BLCS T1-stage patients, EGFR-mutated cases had better survival than KRAS/STK11-mutated cases; TCGA showed a consistent trend. - Selective pressure on EGFR trunks: In MPNs with truncal EGFR mutations, variant allele frequency of EGFR significantly decreased from pre-invasive to invasive components, indicating negative selection during acquisition of invasiveness. Overall truncal mutation abundance decreased more in EGFR-mutant MPNs, and dN/dS ratios were reduced, supporting strong selective pressure. - Immune microenvironment: TCGA analyses showed higher B-cell infiltration in EGFR-mutated T1-stage tumors compared with KRAS/STK11 groups. IHC confirmed higher CD20+ B-cell density in invasive versus pre-invasive components in EGFR-mutated cases, with no significant difference in CD3+ T cells. These findings implicate B-cell–mediated selective pressure on EGFR-mutant clones. - cfDNA: No significant differences in mutation scores were observed between pre- and post-operation cfDNA, and EGFR mutations were not detected in cfDNA in this T1-stage cohort, potentially reflecting microenvironmental selection and low shedding. - Additional observations: Dual EGFR hotspot variants were identified in some EM1 MPNs, segregating between pre-invasive and invasive components. Approximately 80% (36/45) of EM2 tumors harbored EGFR exon 19–21 variants, highlighting EGFR’s dominant role in initiation.

This study demonstrates that early LUAD frequently follows branched evolutionary trajectories from pre-invasive to invasive states within the same lesion, although a subset displays a linear-like path. The identification of distinct modes (EM1, EM2A, EM2B, EM3) highlights the heterogeneity of early invasive progression. Truncal RTK–RAS pathway alterations, especially EGFR, commonly initiate tumorigenesis; yet, EGFR-mutant truncal clones undergo reduced abundance upon invasion, coinciding with increased B-cell infiltration, suggesting microenvironment-driven negative selection. These dynamics help explain favorable prognoses in EGFR-mutated T1-stage cases relative to KRAS/STK11-mutated tumors and the low detectability of EGFR mutations in cfDNA at early stages. Mapping phylogenetic relationships between matched pre-invasive and invasive components provides insight into clonal selection, intratumor heterogeneity, and the interplay between driver genotype and immune contexture during the acquisition of invasiveness. The findings underscore the need to consider evolutionary mode and immune milieu when designing early intervention and surveillance strategies.

By micro-dissecting and sequencing paired pre-invasive and invasive components of early-stage LUAD, this study defines three principal evolutionary trajectories and shows that most lesions evolve via branched evolution. Truncal EGFR, KRAS, TP53, and STK11 are key drivers, with RTK–RAS alterations dominating trunks. EGFR-mutant truncal clones decrease in abundance during invasion, likely due to B-cell–mediated selective pressure, correlating with better prognosis compared to KRAS/STK11-mutant cases. These results clarify how genetic heterogeneity and immune microenvironment shape early invasive progression and provide a framework for risk stratification and early therapeutic intervention. Future research should dissect the mechanistic basis of B-cell–tumor interactions, validate evolutionary modes in larger, multi-omic cohorts, and test tailored adjuvant or neoadjuvant strategies informed by evolutionary trajectory and immune context.

In some MPNs, no private mutations were detected between components, indicating limited sensitivity to capture certain key alterations with the targeted approaches used. The reliance on targeted panels (varying between phases) may miss driver events outside panel coverage. Sample sizes for certain analyses (e.g., IHC, cfDNA) were modest. Technical and algorithmic thresholds in phylogenetic inference may influence trunk/branch assignments.