An in-depth exploration of four heterogeneity structure-based EGFR mutation subgroups in Chinese non-small cell lung cancer

Abstract

Background

EGFR mutations have been classified into four functional subgroups (Classical-like, P-loop and αC-helical compressing (PACC), T790M-like, and exon 20 loop insertions) based on their influence on EGFR protein structure, as well as response to various types of EGFR-tyrosine kinase inhibitors (TKIs). However, the differences in molecular phenotypes and clinical outcomes for patients carrying these different forms of EGFR mutations are not fully understood. Here we sought to investigate the distribution of different EGFR structural types in Chinese NSCLC patients and the biological characteristics of each subgroup.

Methods

2992 EGFR mutant NSCLC patients with available next-generation sequencing result were collected for mutation analysis. 118 patients with targeted RNA sequencing data were further analyzed to compare transcriptome differences across mutation subgroups.

Results

Across the entire cohort, 80.82% of patients were Classical-like, 5.92% were PACC, 10.76% were T790M-like, and 2.51% were Ex20ins. TP53 was the most common co-occurring mutation across the four subgroups, occurring in 60% of the T790M-like subgroup. Interestingly, the Ex20ins group exhibited a notable proportion of genomic alterations related to DNA repair processes. Additionally, both the Ex20ins and T790M subgroups demonstrated higher tumor mutational burden (TMB) scores. Furthermore, for the first time, we observed transcriptomic heterogeneity within these four subgroups. Classical-like group displayed enrichment of immune-related pathways, including PD1 signaling, CD28 family, and TCR signaling. Notably, the L858R group showed significant enrichment in immune activation signatures, including effector memory CD8 T cells, natural killer cells, and MHC I/II. This suggests a potentially robust immune response in that group. In contrast, the T790M-like subgroups showed lower anti-tumor immune signatures but were marked by a significant enrichment in tumor proliferation signatures.

Conclusion

In this study, we have employed a structure-based classification approach for EGFR mutants to comprehensively characterize the mutational landscape and heterogeneous biological traits at the transcriptional and functional levels in Chinese patients with NSCLC.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12890-025-03774-y.

Introduction

Targeted therapy has revolutionized treatment for a subset of patients with Non-Small Cell Lung Cancer (NSCLC), and driver gene testing is now standard practice for patients newly diagnosed with NSCLC [1–4]. Classical EGFR mutations (L858R and exon 19 deletions (Ex19del) show marked improvements in clinical outcomes when treated with first-, second-, or third-generation inhibitors [5, 6]. Howerer, NSCLC containing EGFR mutations is a genetically heterogeneous disease, with more than 200 distinct mutations having been identified so far [1, 7, 8]. The response of these rare mutations to existing EGFR-TKIs varies widely and is not well understood, since most of the pivotal studies with EGFR-TKIs in the first line, have focused on patients with classic EGFR mutations [5, 9, 10]. Recently, Jacqulyne et al. found that EGFR mutations can be separated into four distinct subgroups on the basis of sensitivity and structural changes, as well as predicting patient outcomes more accurately than traditional exon-based classifications following treatment with EGFR inhibitors [11].

Understanding the molecular differences between different EGFR mutation subgroups is particularly important for evaluating the suitability of NSCLC patients to receive immunotherapy. Clinical analysis has revealed that EGFR L858R mutant tumors exhibit higher tumor mutational burden (TMB) levels and increased infiltration of CD8 + PD-1 + T cells compared to EGFR exon 19 deletions [12, 13]. Retrospective studies have also suggested that patients with uncommon EGFR mutations, such as G719X (c. 2155G > X, p. Gly719X) and exon 20 insertions (Ex20ins), may derive more clinical benefits from immune checkpoint inhibitors (ICIs) than those with common mutations [14, 15]. However, much remains unknown regarding the suitability of each mutation subgroup for ICI treatment.

Given the high frequency of EGFR mutations among the East-Asian population [16], it is clinically crucial to investigate the distribution and biological characteristics of structure-based EGFR classification in Chinese NSCLC patients.

Methods

Patient cohort

In total, we retrospective analyzed 2992 patients with NSCLC at Shanghai Chest Hospital who underwent tumor DNA and/or RNA sequencing as part of their clinical care from November 2011 and November 2021. All patients provided written informed consent, and this study was approved by the institutional review board of the Shanghai Chest Hospital (Ethics ID:IS22059). We judged patients eligible for the study if they carried an EGFR non-synonymous mutation that could be confirmed through histological or in blood sample sequencing. EGFR mutations can be classified into four subtypes (Classical-like, P-loop and αC-helical compressing (PACC), T790M-like, and exon 20 loop insertions) based on unique structures as Robichaux et al. previously reported (Supplementary Table 1) [11]. In brief, the specific identity of EGFR mutations was determined from formalin-fixed paraffin-embedded tumors or blood samples by CLIA/CAP-certified methods. And samples subjected to comprehensive genomic profiling by targeted DNA/RNA Next-Generation Sequencing (NGS) panels (Supplementary Table 2). The mutational landscape and transcription characteristics of different EGFR mutations among patients were analyzed. The flow chart of this study is shown in Fig. 1.

Fig. 1.

Flowchart of the study design. Study design and analytical workflow for EGFR mutation profiling in NSCLC. This retrospective cohort included 2,992 non-small cell lung cancer patients with EGFR mutations. Tumors were stratified into structure-based EGFR mutation subtypes: Classical-like, PACC, T790M-like, and exon 20 loop insertions

Genomics data analysis

In addition to EGFR mutations, information on concurrent genetic alterations were available in 544 patients with EGFR mutations who were profiled by targeted 559 gene DNA sequencing using AmoyDx® Master Panel (Amoy Diagnostics Co., Xiamen, China). In brief, variant detection was conducted through GATK's Mutect2 pipeline with subsequent refinement using FilterMutectCalls, producing preliminary variant calls. These variants underwent functional annotation through ANOVA-based analysis prior to rigorous filtering. The selection criteria for clinically relevant variants consisted of four tiers: 1) Threshold requirements of ≥ 5 supporting reads and ≥ 5% allele frequency; 2) Exclusion of population polymorphisms (frequency > 2% in 1000 Genomes, ExAC, or gnomAD); 3) Restriction to protein-coding sequences; 4) Prioritization of variants with oncogenic potential as classified in OncoKB. Only variants satisfying all these parameters were retained for downstream biological interpretation and statistical evaluation. The analytical methods were performed as described in Dong et al. [17]. Genomic aberrations identified in these patients and Tumor mutational burden (TMB) were described. The TMB was defined as the total number of somatic mutations per mega base (Mb) of the genome examined.

Differentially Expressed Genes (DEGs) and functional pathways analysis

1813 gene expression was analyzed in 118 patients using using AmoyDx® Master Panel (Amoy Diagnostics Co., Xiamen). The analytical methods were performed as described in Dong et al. Differential gene expression of log2(TPM + 1) was determined by using limma R package (fold change > 1.2, p value < 0.05) among the four subtypes. Gene Set Enrichment Analysis (GSEA) enrichment (BH < 0.05) analysis through R package clusterProfiler (v 4.2.2) were performed to search for biological functions and pathways using Reactome gene sets from MSigDB.

Analysis of immune landscape

Tumor microenvironment analysis was performed using gene expression data. Estimation of cell populations counter based on the mean marker gene expression that is specifically expressed in the cell type. Microenvironment analysis was performed by GSVA analysis from GSVA R package (v 1.42.0) using 28 and 29 signatures [18, 19].

Statistical analysis

Statistical comparisons were evaluated by using the Wilcoxon rank sum test and Kruskal–Wallis tests. Two-sided P values < 0.05 were considered statistically significant. All analyses were conducted using R version 4.1.2 and its associated packages.

Results

Distribution for structure-based EGFR mutation subtype among NSCLC patients

2992 patients with EGFR-mutant NSCLC were stratified by EGFR mutation subtype as described in the methods above. Among these patients, the distribution of four subtypes were as follows; 80.82% of patients were Classical-like (including classical 19del and L858R), PACC (5.92%), T790M (c.2369C > T, p.Thr790Met)-like (10.76%), and Ex20ins (2.51%) (Fig. 2A). The EGFR mutations were manually curated as classical or atypical EGFR mutations (Classical EGFR mutations are L858R and/or Ex19del with or without T790M, T790M. Atypical EGFR mutations are defined as non-classical). 84.96% had classical EGFR mutations; 15.04% had atypical EGFR mutations (Fig. 2B). In the atypical EGFR mutations, the primary EGFR mutation subtypes across the four subgroups were outlined below: L861Q (c.2582 T > A, p.Leu861Gln) (45.39%) and L858R compound mutation (42.76%) in Classical-like subgroup, G719X + S768I (c.2303G > T, p.Ser768Ile) compound mutation (28.25%) and E709X + G719X compound mutation (12.43%) in PACC subgroup, 19del + T790M + C797S (c.2389 T > A, p.Cys797Ser) compound mutation (50%) and L858R + T790M + C797S (26.09%) in T790M-like subgroup, and D770_N771insSVD (c.2303_2311dup, p.Asp770_Asn771insSerValAsp) (30.67%) and A767_V769dup (c.2300_2308dup, p.Ala767_Val769dup) (25.33%) in Ex20ins subgroup (Fig. 2C-F).

Fig. 2.

Distribution for structure-based EGFR mutation subtype among NSCLC patients. A Distribution EGFR mutation subtype among NSCLC patients. B Distribution of atypical EGFR mutations in different subtype. C Distribution of EGFR mutations subtype in Classical-like group. D Distribution of EGFR mutations subtype in PACC group. E Distribution of EGFR mutations subtype in T790M-like group. F Distribution of EGFR mutations subtype in Ex20ins group

Genetic alterations in patients with different structure-based EGFR mutation subtype

To further analyze the likelihood of other driver mutations being present in EGFR-mutant patients, 544 patients were used for further analysis. Within these samples, TP53 was the most common co-occurring mutation in each of the four distinct subgroups, accounting from an average of ~ 40% up to 60% of T790M-like subgroups. Other mutations were overrepresented in select groups; for instance, APC mutations were more common in the T790M-like group, while not being found in the Ex20ins group (Fig. 3A). Interestingly, a high proportion of genomic alterations involving DNA repair were observed in Ex20ins group, including in MSH2 and MSH6. Overall, the Ex20ins group and T790M group displayed higher TMB scores in comparison to the Classical-like and PACC groups (Fig. 3A-B).

Fig. 3.

Co-occurring genomic alterations and tumor mutational burden (TMB). A Mutational landscape of frequently altered genes (top 30) in EGFR-driven tumors, stratified by EGFR subgroups: Classical-Like, PACC, T790M-like, and Ex20 Loop Insertions. Mutation types are color-coded (frameshift, non-frameshift, nonsense, nonsynonymous substitutions, splicing). B Comparative TMB analysis among EGFR subgroups

Functional enrichment analysis in different structure-based EGFR mutant groups

Beyond differences in co-occurring mutations, the four subtypes also showed significant difference in gene expression profile, as observed through DEG analysis (top DEGs shown in heatmap) in Fig. 4A. Functional analysis and signaling pathway enrichment of these candidate DEGs were conducted using Reactome. The Classic-like subtype was characterized by an enrichment of immune-related pathways, including those involved in PD-1 signaling, the CD28 family, and TCR signaling. This suggests that these patients may be associated with immune activation. In contrast, the other subtypes were enriched for other types of biological functions. For the PACC subgroup tumors demonstrated activation of the GPCR signaling cascade towards cellular events. Cell cycle pathways including E2F targets, and cell cycle checkpoints (G2M) were significantly upregulated in the T790M-like group. The Ex20ins subgroup has been identified with a distinctive enrichment of oncogenic processes, which encompass the activation of PTK6 and MYC, the modulation of gene expression through epigenetic mechanisms, and the Neddylation pathway (Fig. 4A-B).

Fig. 4.

Transcriptomic Characteristics of structure-based EGFR mutation subtype. A Heatmap for DGEs (top 10) across samples in Classical-Like, PACC, T790M-like, and Ex20 Loop Insertions. B Gene Set Enrichment Analysis (GSEA) enrichment were performed to search for biological functions and pathways using Reactome gene sets from MSigDB. Heatmap for immune-related signatures across EGFR mutation subtype with. C 28 Immune cell signatures (Bindea G et al.,2013), and D 29 Immune cell signatures (Keenan TE et al.,2020)

Tumor immune microenvironment in the structure-based EGFR mutations subtype

Subsequently, the immune characteristics of tumors with different subtypes were investigated. Consistent with the analysis above, T790M-like subgroups displayed lower anti-tumor immune signature, such as T cell infiltration (Fig. 4C-D). On the other hand, significant enrichment of tumor proliferation signatures was observed in T790M-like group (Fig. 4C-D). It is interesting to note that the heterogeneous immune microenvironment has been observed of in the Classical-like groups too. Significant enrichment of effector memory CD8 T cell, Natural killer cell and MHC I/II signatures were displayed in L858R group. In addition, Th2 and Neutrophil signature were significantly enriched in L858R-compound subgroup (Fig. 5A-B).

Fig. 5.

Immune-related signatures that were altered among Classical-like group. The fraction of TME cells in four gene clusters estimated by A 28 Immune cell signatures (Bindea G et al.,2013), and B 29 Immune cell signatures (Keenan TE et al.,2020). Note. * < 0.05, * < 0.01

Discussion

Exon 19del (about 45% of mutations) and exon 21 p.L858R (about 40% of mutations) lie in the tyrosine kinase domain of EGFR, and can be targeted by TKI [1]. It is crucial to emphasize that approximately 10% of patients exhibit a rare EGFR mutation, resulting in diverse responses to EGFR-TKI therap [20]. As the understanding of the genetic and clinical complexities of this condition continues to expand, there is a pressing need to evaluate treatment responses among these uncommon mutation subtypes to inform clinical management decisions. Based on the findings of Robichaux et al. [11], it has been identified that four distinct subgroups of EGFR mutations can be categorized depend on their sensitivity and structural characteristics. This classification method offer a more precise categorization compared to exon-based groups, when considering patient responses to treatment with TKI.

In our dataset, for patients with NSCLC, 84.96% had classical EGFR mutations, 15.04% had atypical EGFR mutations occurred primarily in exons 18 and 20 insertions. Further categorization of these atypical EGFR mutations revealed that 33.78% were Classical-like, 39.33% were PACC, 10.22% were T790M-like, and 16.67% were Ex20ins (Fig. 2B). Prevalent hotspots for atypical mutations were the P-loop (L718–V726) and the C-terminal loop of the αC-helix (A767–G779). The diversity of these mutations is much higher than previously appreciated [1, 21], and highlights the necessity of comprehensive test for patients with NSCLC. These results indicate that there exists a complex landscape of distinct EGFR mutations in NSCLC patients. Integrating this structure–function-based EGFR classification into clinical practice has the potential to enhance precision medicine for patients with EGFR mutant NSCLC.

In general, EGFR-mutant NSCLCs represent a big challenge for the immunotherapy treatments. However, certain EGFR mutant tumors have shown promising responses to these treatments. EGFR L858R tumors had a similar response rate and OS outcomes to an EGFR wild-type lung cancer population, while EGFR 19del cases did substantially worse [22]. Moreover, individuals with rare EGFR mutations (like G719X and exon 20 insertion) may experience greater clinical advantages from ICIs compared to those with more common mutations [23]. However, current clinical and preclinical evidence for each mutation type is insufficient, with the presence of co-occurring mutations being a significant concern. For example, EGFR exon 20 insertions exhibit a wide range of variations and are frequently accompanied by co-occurring mutations [23]. In this study, we find here that a high proportion of genomic alterations involving DNA repair were observed in Ex20ins group, and consequently displayed higher TMB score, indicating a potential to benefit from ICIs.

Previous reports have indicated that decreased programmed cell death-ligand 1 (PD-L1) expression and tumor mutation burden (TMB) may be the root cause of poor response to immune checkpoint inhibitors in EGFR mutation NSCLC cases [24]. However, the comprehensive understanding of the tumor microenvironment (TME) characteristics in different EGFR mutation subgroups in NSCLC is still lacking. Our analysis revealed that the Classical-like subgroup exhibited a significant enrichment of pathways related to immune response. Interestingly, significant heterogeneity was also observed within the Classical-like subgroup. In particular, when compared to other EGFR mutant subtypes, the L858R and L858R-compound group showed a significant enrichment of immune activation signatures. In contrast, T790M-like subgroups displayed lower anti-tumor immune signatures, while significant enrichment of tumor proliferation signatures. Similarly observed in previous analyses of T790M positive/negative NSCLC tumors [25, 26], patients lacking the T790M mutation demonstrated elevated PD-L1 levels, increased CD8 + TIL infiltration, and decreased FOXP3 + TIL infiltration in comparison to T790M-positive patients. Taken together, these results show that different subtypes of EGFR mutations are indeed correlated with differential biological activity within tumors. Further studies are required to clarify the specific immune features and response patterns to immunotherapeutic approaches.

This study has several limitations. Firstly, this was a retrospective study with selected patients, which could have inevitably involved some selection bias. Second, not all patients underwent the same molecular testing. Thirdly, due to the lack of patient’s information, it is not possible to perform analysis of the biological characteristics among different subgroups based on distinct clinicopathologic features. Furthermore,

due to the lack of data on the treatment and prognosis of these patients, it was unclear whether targeted therapy could benefit them. Thus, further large-scale and prospective investigations are warranted to solve some of these limitations.

In conclusion, our study depicted the landscape of structural types of EGFR mutations in Chinese NSCLC patients. We applied the structure-based of EGFR-mutant classification approach to characterize the mutational landscape and heterogenous biology trait at transcriptional and functional level.

Authors’ contributions

SL and CZ conceived and designed the experiments. YY, FY and TH collected the data. FY, ZH, JW and CZ analyzed the data. YY and FY analyzed the data and wrote the paper. All authors read and approved the final manuscript.

Funding

This research was funded by National Natural Science Foundation of China (82030045, 82241227); National Multi-disciplinary Treatment Project for Major Diseases (2020NMDTP); Collaborative Innovation Center for Clinical and Translational Science by Ministry of Education & Shanghai (TM202112, CCTS202204, CCTS202304).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This study was approved by the institutional review board of the Shanghai Chest Hospital, Shanghai Jiao Tong University School of Medicine (Ethics ID:IS22059).

Consent for publication

All patients provided written informed consent before molecular diagnosis, participation in this study was covered by this protocol.

Competing interests

The authors declare no competing interests.