Formalin-fixed paraffin-embedded (FFPE) blocks of lung cancer cases obtained by the biobank with appropriate waiver of consent approvals from the institutional ethics committee (IECs) of Apollo hospitals were used in this study. The use of the biobanked FFPE samples and associated diagnostic data for NGS analysis was further approved by the IEC of the biobank, constituted as per the Indian Council of Medical Research (ICMR) 2017 and DHR guidelines (Protocol SBS-IEC-2020-05 titled “Utilization of retrospective FFPE tissue blocks for research and development at Thermo Fisher Scientific” and Protocol SBS-IEC-2022-01 titled “Analysis of sequencing data of biobanked cancer tissue samples for translational research and publication”). FFPE samples and data were coded by the biobank to protect patient confidentiality and privacy as per ICMR guidelines.

A total of 103 lung cancer resection cases drawn from the years 2004–2021 were profiled. Demographic data such as age and gender of the patient, and diagnostic data for the surgical samples were retrieved from the biobank’s curated medical records. A haematoxylin and eosin (H&E) stain of each block was used to re-confirm their quality and histological diagnosis by a clinical pathologist in order to remove any risk of inaccuracies stemming from the use of archived blocks and records.

Sections of 20-microns were used from FFPE blocks and sequenced by ThermoFisher Scientific (TFS) using their proprietary Oncomine™ Dx Express Test (ODxET™) on their Ion Torrent™ Genexus™ Integrated Sequencer. TFS performed the raw data processing, variant calling, and annotation, and provided a comprehensive list of annotated variants based on their gene class (gain-of-function GOF, or loss-of-function LOF) and variant class annotations (Supplementary data).

The list of genetic variants generated by TFS was further filtered to remove synonymous mutations that lacked variant identifiers such as COSMIC or arbitrarily assigned identifiers. Three LUAD cases (n = 3) were excluded from the study due to “poor” DNA or RNA QC scores. The specifics for calculating composite DNA and RNA QC scores are outlined in Supplementary methods.

MutationMapper and OncoPrinter from the cBioPortal Cancer Genomics (Cerami et al., 2012; Gao et al., 2013) were used for data visualization purposes. Additional annotations and drug-alteration matching were performed using OncoKB (Chakravarty et al., 2017). All statistical analyses were performed using the GraphPad Prism 6 software (GraphPad software Inc., La Jolla, CA, United States).

In this study, 53 cases of primary non-metastatic LUAD were selected from the 103 lung cancer cases that were profiled using a 46 gene hotspot Oncomine™ Dx express test. These 53 cases included 2 cases that had NACT prior to surgery and 2 that had locally relapsed within 6 months. Among the 53 cases, 29 (54.7%) were classified as adenocarcinoma with features not otherwise specified, while the remaining cases were categorized into various subtypes according to the 2021 WHO classification of lung tumors (Table 1).

The patient demographic data including gender, age, and age distribution, are summarized in Table 2. Age frequency distribution revealed that most patients were aged 51–70, with only 4 cases being below the age of 50. The age distribution between male and female patients did not differ significantly (Mann-Whitney test p = 0.138, data not shown).

Of the 53 LUAD cases analysed, 50 (94.3%) were found to carry at least one pathogenic or likely pathogenic variant using a variant allele frequency (VAF) cut-off of 2% (Supplementary data). These driver variants were detected in 31 genes out of the 46 cancer-relevant genes from the ODxET panel (See Supplementary methods). A total of 252 genetic alterations of different types were found in the hotspot regions of the 31 genes (Figure 1A). Some samples had multiple alterations in the same gene. The frequency distribution of distinct gene alterations with at least one mutation detected in them varied across the sample set, ranging from samples with no variants detected (n = 3, 5.7%), samples with only a single driver variant (n = 5, 9.4%), to samples with 10 genes alterations (n = 2, 3.8%) that included multiple alterations in oncogenes such as in EGFR, PIK3CA, ERBB2, FGFR3 (Figure 1B).

The landscape of commonly mutated genes along with other clinical parameters was visualized using OncoPrinter analysis in the cBioPortal database (Figure 2). The most frequently altered genes comprised TP53 (52.8%), and EGFR (50.9%), followed by RET (26.4%), PIK3CA (24.5%), ERBB2 (22.6%), ALK (20.8%), PTEN (17.0%), STK11, KEAP1 and IDH2 (15.1%), FGFR4 (13.2%), MET, FGFR3, KRAS, and FGFR2 (11.3%).

Most TP53 mutations were LOF missense mutations (n = 35), primarily occurring in exons 5 to 8 within the DNA binding domain (Figure 3A). EGFR mutations were detected in the cytoplasmic region from exons 18 to 21, located in the protein tyrosine kinase domains or the C-terminal phosphorylation domains. The most common EGFR mutation was Exon19 in-frame deletion (n = 9) (Figure 3B). RET variants were mostly GOF missense mutations that were predominantly located in the kinase domain, within the cytoplasmic region from exons 13 to 16. The most frequently observed mutation was RET V804M (Figure 3C). RET was the sole driver mutation in three cases: one case with RET C618Y mutation located in the cysteine-rich domain, another with RET V804M, and the third with a novel RET-KIAA1468 fusion.

PIK3CA mutations were mostly observed in the catalytic subunit of PI3-kinase, including R38C in the p85 binding domain (n = 6), R108H in the adaptor-binding domain ABD (n = 3), V344M residue between the RAS binding domain and the C2 domain (n = 3), and H1047R in the kinase domain of the protein (n = 2) (Figure 3D). PIK3CA H1047R was the single driver mutation in one patient sample. Amplification in ERBB2 (HER2 or ErbB2) was not observed in these 53 NSCLCs; instead activating oncogenic mutations were identified such as exon20 insertion p.Y772_A775dup (n = 3), V842I (n = 1), and G660D (n = 1). The most common ERBB2 mutation observed was R896 C/H located in the kinase domain of the protein (n = 8) (Figure 3E).

ALK oncogenic mutations were primarily located in exons 23 to 26 of the gene, within the intracellular tyrosine kinase domain, with the most frequent ones being R1192P, L1198P, G1202R, D1203N, S1206Y and R1275Q (Figure 3F). Rearrangements of ALK were detected in 2 cases. PTEN LOF missense mutations were predominantly located in the phosphatase domain, with the truncating PTEN p.R130* being the most frequent mutation observed (n = 5) (Figure 3G). STK11 mutations were identified at 3′-splicesite, 5′-splicesite, and tyrosine kinase domains, that included p.A205T and p.D194N/Y (Figure 3H). Other gene alterations with greater than 10% frequency included IDH2 R140W/Q (n = 8) present in the catalytic domain of the protein (Figure 3I), MET exon 14 in-frame deletions (n = 2) (Figure 3J), KRAS exon 2 mutations (n = 6) located in the P-loop of the catalytic G-domain of the protein, HRAS exon 2 mutations (n = 4), NRAS G12V, Q61L (n = 2) (Figures 3K–M).

There were two cases of invasive adenocarcinoma treated with neoadjuvant chemotherapy (NACT) included in this study, each with distinct mutation profiles (SB00046727 and SB00046728). The first case exhibited a rare fusion mutation ALK-PRKAR1A, along with FGFR4 V550M, FGFR2 A648T, and EGFR V769M SNV mutations. The second case showed mutations in IDH2 R140W and EGFR exon19 deletion. Furthermore, our study included two cases of locally relapsed NSCLCs. One case, classified as Bronchioloalveolar carcinoma with local recurrence (SB00036833), exhibited a mutation spectrum including EGFR exon19 deletion, EGFR A750P, TP53 R249S, FGFR4 V550M, and FGFR2 S252L mutations. The second case, characterized as acinar adenocarcinoma with relapse within 6 months (SB00036841), presented mutations in TP53 E286V, TP53 R282W, and EGFR exon19 deletion.

To check the correlation between the distribution of gene alterations with clinical features, an enrichment analysis was performed. No significant correlation was observed with age, tumor stage or subtype diagnosis of LUAD with mutations across any of the 46 genes.

Enrichment analysis of gene alterations by gender however, revealed that certain genes such as ALK and IDH2 were more frequently enriched in females, while TP53 and PTEN were more frequently enriched in males. TP53 mutation frequency was 64.3% (18/28) in males as compared to 40% in females (10/25), with a p-value of 0.102. The mutation frequency of IDH2 was found to be 7.1% (2/28) in males as compared to 24% in females (6/25), with a p-value of 0.129. Although not significant, these indicate a tendency towards gender-wise enrichment. These results were supported by scatter and volcano plots shown in Figure 4.

Potentially actionable somatic alterations were analysed using a higher, 5% VAF mutation sensitivity, that is more often used for clinical reporting. As defined by OncoKB Therapeutic Levels of Evidence V2 framework, these alterations were identified in 42 (79%) cases with at least one potentially actionable alteration eligible for adjuvant targeted therapy (Figure 5A). Of these 42 cases, 64.3% of patients (n = 27) had a Level 1 alteration, 2.4% (n = 1) had Level 2, 21.4% (n = 9) Level 3B, and 11.9% (n = 5) Level 4 as their highest level of actionability. More than one potentially actionable alteration was observed in 39.6% of patients (n = 21) that could help in selection of multiple lines of therapy. Biomarkers predictive of resistance or futility to FDA-approved therapies in NSCLC (R1) were also identified in 3.8% of cases (n = 2), and biomarkers with compelling preclinical evidence of resistance to a drug (R2) in 7.5% of cases (n = 4).

EGFR Exon19 in-frame deletions and L858R were the most frequent predictive biomarkers (n = 11, 20.8%), indicative of response to various EGFR TKIs (Figure 5B). Actionable ERBB2 mutations (Exon20 insertion, G660D, V842I) were detected in 5.7% of cases (n = 3); these are predictive of response to Trastuzumab Deruxtecan, the first FDA-approved drug for HER2-mutant NSCLC. EGFR T790M detected in 1 case has been associated with acquired resistance to first- or second line- EGFR TKI therapies but predictive of a favourable response to Osimertinib. Oncogenic fusions in ALK, RET, NTRK family, MET exon14 alterations, all of which predict significant clinical benefit from targeted inhibitors of these kinases, were identified in 11.3% of cases (n = 6). Less common ALK oncogenic mutations (R1192P, L1198P, D1203N, S1206Y and R1275Q) were identified in 5.7% of cases (n = 3), associated with resistance to first and second-generation ALK inhibitors, and sensitivity to Lorlatinib. These resistance mutations, such as EGFR T790M and the less common ALK mutations, which are more frequently associated with disease progression on tyrosine kinase inhibitor (TKI) therapy, were not identified in the two cases of primary NSCLCs treated with NACT and the two cases of relapsed NSCLCs in our cohort. Previous reports have also indicated that these mutations have been identified in TKI-naive patients with NSCLC (Li et al., 2018); or as primary resistance mutations, leading to the lack of benefit from TKI treatment (Lin et al., 2017).

Other less common EGFR mutations observed, G719X, L861Q, S768I (n = 2) included as Level 1 are predictive of response to Afatinib. KRAS G12C (n = 2) is prognostic of poor survival and associated with responsiveness to Sotorasib and Adagrasib.

Among other genes not included yet in the NCCN guidelines for molecular testing in NSCLC, clinically informative alterations were detected in 9 cases (17.0%). These alterations included PIK3CA oncogenic mutations, namely, R38C, R108H, V344M, E542A, H1047R (n = 7); IDH2 R140W/Q (n = 3); RET oncogenic mutations such as RET C618Y, V804M, R886W (n = 3); HRAS G12S, G13S (n = 2); NRAS G12V, Q61L (n = 2); and PDGFRA Y849C (n = 1) (Figure 5B). Consistent with other studies, most oncogenic or likely oncogenic PIK3CA alterations were identified in tumors with a co-occurring higher OncoKB-level alteration (Jordan et al., 2017).