co-mutations are associated with elevated PD-L1 expression and high tumor mutational burden in non-small cell lung cancer: insights from comprehensive genomic profiling.

Introduction
Lung cancer remains the leading cause of global cancer mortality, accounting for approximately 1.8 million deaths annually (1). The World Health Organization (WHO) Classification of Thoracic Tumors has underscored the critical role of molecular diagnostics in contemporary lung cancer management (2). Over the past decade, next-generation sequencing (NGS) has become the cornerstone of molecular profiling, enabling massively parallel sequencing and comprehensive detection of somatic mutations and genomic alterations (3,4). Current clinical guidelines recommend tyrosine kinase inhibitors (TKIs) as first-line therapy for patients with tumors harboring EGFR, ALK, ROS1, and BRAF-positive non-small cell lung cancer (NSCLC). The therapeutic scope of EGFR TKIs has been expanded to include patients with resectable early-stage and inoperable locally advanced NSCLC, marking a crucial milestone in the precision management of EGFR-mutant NSCLC (5). Patients with resected ALK-positive NSCLC of stage IB, II, or IIIA, adjuvant alectinib significantly improved disease-free survival as compared with platinum-based chemotherapy (6). There is an expanding list of approved targeted therapies, including for KRAS G12C mutations, MET exon 14 alterations, and NTRK and RET rearrangements (7). Notably, non-actionable mutations, including TP53, PIK3CA, and SMAD4, play important roles in NSCLC. TP53 mutations are associated with accelerated tumor progression and chemoresistance (8,9); PIK3CA dysregulation promotes cellular growth and proliferation processes (10); and SMAD4 loss facilitates histological transformation to small cell carcinoma (11). Therefore, systematic characterization of both targetable and non-targetable genomic profiles through NGS is essential for guiding NSCLC therapy. Growing evidence indicates that co-mutation patterns involving actionable and non-actionable genes carry important implications for tumor progression and therapeutic decision-making. For example, EGFR co-mutations have been linked to an increased risk of recurrence in invasive lung adenocarcinoma with micropapillary components (12). TP53 co-alterations have been reported to adversely affect survival in patients with EGFR-mutated advanced NSCLC receiving standard first-line therapy (13). In addition, concurrent KRAS and TP53 mutations predict the benefit of immune checkpoint blockade in lung adenocarcinoma (14). However, there is currently a lack of data regarding the distribution of actionable mutations and non-actionable co-mutations in NSCLC.
Programmed cell death ligand 1 (PD-L1) expression and tumor mutational burden (TMB) have garnered increasing recognition as predictive biomarkers for response to PD-L1 blockade therapy. Clinical evidence indicates that PD-L1 expression ≥1% correlates with higher rates of major pathological response (MPR) and pathological complete response (pCR). Similarly, elevated TMB associates with improved MPR and pCR outcomes in early-stage patients (15). Advanced NSCLC Patients frequently exhibit elevated PD-L1 expression. Immune checkpoint blockade through direct inhibition of the programmed cell death 1 (PD-1)/PD-L1 pathway currently serves as the cornerstone of first-line therapy for advanced NSCLC in the absence of actionable tumor genomic alterations (16). NGS enables precise quantification of TMB, with higher levels further enhancing the likelihood of clinical benefit from immune checkpoint inhibitors (17). Therefore, the selection of clinical treatment strategies for NSCLC should take into account the combined influence of co-mutation patterns, PD-L1 expression, and TMB. To our knowledge, the systematic assessment of immunological biomarkers in patients with these co-mutation profiles remains relatively underexplored.
On one hand, our study enrolled a large Chinese cohort of NSCLC cases to explore the associations between driver gene alterations and clinicopathological characteristics using a pan-solid tumor panel of 654 genes; on the other hand, it characterized actionable and non-actionable co-mutation patterns in NSCLC and analyzed their relationships with PD-L1 expression and TMB levels. We present this article in accordance with the STROBE reporting checklist (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-1030/rc).

Methods

Patient cohort
This is a retrospective study on NSCLC patients who attended Peking University People’s Hospital from 2019 to 2023, diagnosed with NSCLC by histopathology. The inclusion criteria were: (I) patients who underwent 654-gene panel testing by NGS; (II) a cytological or pathological diagnosis confirming lung cancer; and (III) no history of radiotherapy or chemotherapy prior to surgery. A total of 473 patients met the inclusion criteria. The exclusion criteria were: (I) small cell lung cancer patients (n=23); (II) patients lacking clinical pathological information (n=20). Demographic and clinical characteristics were collected from patients. Demographic and clinical characteristics were collected from all patients. Basic clinical data, including age, sex, and smoking history, were recorded. Patients included in this study had undergone NGS-based gene testing, which may introduce a degree of selection bias. The final study population consisted of 431 individuals, all of whom underwent PD-L1 and TMB testing, with complete datasets available. A flowchart is presented in Figure 1. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Peking University People’s Hospital (No. 2022PHA024-001), and individual consent for this retrospective analysis was waived.

The determination of actionable and non-actionable mutations
Based on the results of previous studies and European Society for Medical Oncology (ESMO) guidelines, we finally determined that the actionable mutations involved in the study include EGFR-activating mutations, KRAS G12C, BRAF V600E, MET amplification, MET exon 14 skipping mutations, and ALK-EML4 fusions (18), and non-actionable mutations comprised TP53 mutations, PIK3CA mutations, CTNNB1 mutations, SMAD4 mutations, and non-G12C KRAS mutations (19). NSCLC patients with co-mutations were classified into three categories: patients with no less than two actionable alterations, patients with no less than two non-actionable alterations, as well as patients with no less than one actionable and one non-actionable mutation.

NGS-based assay
First of all, DNA and RNA were extracted from formalin-fixed paraffin-embedded samples or fresh tissue samples using MagMAX™ FFPE DNA/RNA MLtra Kitextraction kit (Thermo fisher, Waltham, USA), and genomic DNA from fresh peripheral blood samples was extracted using the MagPure Blood DNA LQ Kit (Thermo fisher), in accordance with the manufacturer’s instructions. The nucleic acid quantification was performed using nanoDrop 2000C Spectrophotometer (Thermo Fisher) and Qubit 4.0 (Waltham, MA, USA). Secondly, libraries were constructed if the DNA amount from tumor tissue was ≥30 ng with ≥20% tumor cells and plasma samples ≥25 ng. Enriched libraries were amplified and analyzed on the Illumina CN500 platform (Illumina, Inc., San Diego, CA, USA) following the manufacturer’s instructions. The sequencing data were then subjected to quality control again by the following criteria: fraction of target covered at least 0.2× average depth ≥80%, Q30 ≥80%, Q20 ≥85%, mapped reads ≥85%, fraction of effective bases on target ≥40%, coverage ≥97%, and average effective depth ≥1,000× in tissue and 15,000× in plasma. Finally, single-nucleotide variants (SNVs) and small insertions/deletions (indels <50 bp) were identified from the cleaned sequencing data using GATK MuTect2 (v1.1.4) with default settings, followed by removal of variants located within ENCODE blacklisted regions. High-confidence somatic variants were retained through a series of quality filters, excluding SNVs with <20× tumor depth or <4× alternate allele depth in tumor samples, <10× depth in matched normal samples, or those with alternate allele reads exceeding 1% in the normal control. Variant annotation was performed using ANNOVAR, incorporating Human Genome Variation Society (HGVS) nomenclature, population frequency databases [1000 Genomes, The Single Nucleotide Polymorphism Database (dbSNP), Exome Aggregation Consortium (ExAC)], functional prediction tools [Polymorphism Phenotyping version 2 (PolyPhen-2), Sorting Intolerant From Tolerant (SIFT)], and phenotype or disease-related databases [Online Mendelian Inheritance in Man (OMIM), Catalogue Of Somatic Mutations In Cancer (COSMIC), ClinVar]. Variants annotated as genomicSuperDups were removed. Additional filtering excluded SNVs with variant allele frequency (VAF) <0.2 or a maximum population frequency (PopFreqMax) >0.05. The final variant set included SNVs located in curated cancer hotspots with VAF ≥1% and all other SNVs with VAF ≥3% (20).

TMB calculation and PD-L1 expression
TMB is defined as the total number of somatic, coding, non-synonymous mutations per megabase (mut/Mb) within the examined genomic region. In this study, a 654-gene panel was utilized, which encompasses a sequencing region size of 1.2 Mb. The formula for calculating TMB is: the total number of somatic non-synonymous mutations divided by the size of the sequenced exonic region (in Mb). The PD-L1 mRNA expression level was quantified by calculating the transcripts per million or fragments per kilobase of transcript per million mapped reads values from the normalized RNA-seq data. For validation, PD-L1 protein expression was assessed by immunohistochemistry using the 22C3 antibody clone. The tumor proportion score (TPS) was calculated based on standard criteria, defined as the percentage of viable tumor cells showing partial or complete membrane staining at any intensity. PD-L1 TPS ≥1% was considered positive expression, while PD-L1 TPS <1% was considered negative.

Statistical analysis
Mutations in cancer-related driver genes were also analyzed. Fisher’s exact test was used to analyze the distribution difference across gene mutation groups. The Kruskal-Wallis test was employed to assess the association between PD-L1 expression or TMB levels and patterns of co-occurring gene mutations. Additionally, intergroup comparisons were performed. The statistical software package SPSS 22.0 was used for statistical analysis. The oncoplot mutations were plotted by using the “MAfTools” R package. All tests were bilateral, with P<0.05 indicating a significant statistical difference.

Results

Clinical characteristics of 431 NSCLC patients
The research included 431 NSCLC patients, with clinicopathological characteristics detailed in Table 1. The cohort comprised 52.9%female patients (228/431), 19.0% individuals (82/431) reporting smoking history, and a mean age of 59 years (range: 27–99 years). Histological classification revealed lung adenocarcinoma as the predominant subtype (394/431, 91.4%), followed by squamous cell carcinoma (29/431, 6.7%). Disease staging demonstrated early-stage predominance, with 76.1% patients (328/431) categorized as stage I/II and 18.6% patients (80/431) as stage III/IV.

The profiling of somatic variants and the most frequently mutated genes in patients with NSCLC
Genomic profiling revealed 2,669 somatic variants across the cohort, comprising 209 pathogenic, 641 likely pathogenic, and 1,819 variants of uncertain significance. Missense variants predominated (1,891/2,669, 70.9%; Figure 2A), with a C>T/G>A substitution bias evident (Figure 2B). Mutational events spanned 509 genes, where EGFR (n=290), TP53 (n=184), MUC16 (n=67), RBM10 (n=64), KRAS (n=43), PIK3CA (n=34), NAV3 (n=24), ATM (n=21), ARID1A (n=20), and KMT2D (n=20) constituted the top 10 most frequently altered gene (Figure 2C). The detailed mutation information in the top 10 genes was shown in available online: https://cdn.amegroups.cn/static/public/10.21037tlcr-2025-1030-1.pdf. Functional domain analysis identified recurrent hotspot mutations including EGFR L858R (tyrosine kinase domain), TP53 R273C (DNA-binding domain), and KRAS G12C/G12D (Ras family domain), with spatial distributions detailed in Figure 2D.
The top 20 most frequently altered genes across 431 patients were illustrated in Figure 3. The predominant mutated genes included EGFR (59.1%, 255/431), TP53 (39.7%, 171/431), RBM10 (14.6%, 63/431), MUC16 (12.7%, 55/431), and KRAS (10.0%, 43/431), with subsequent alterations in PIK3CA (7.4%), ATM (4.9%), NAV3 (4.9%), KMT2C/D (4.4% each), ARID1A (4.2%), and PTPRD (4.2%).

The significant difference among various clinical indicators and molecular features
Significant differences in clinicopathological characteristics were observed across patients with mutations in the five most frequently altered genes (Table 1). EGFR mutations were more frequently detected in females (P<0.001), whereas MUC16 (P=0.004) and KRAS (P<0.001) mutations were more common in males. Higher MUC16 (P=0.001) and KRAS (P=0.003) mutation frequencies were noted in elderly patients compared with younger individuals. TP53 (P=0.01) and KRAS (P<0.001) mutations occurred more often in smokers, while EGFR mutations were more frequently found in non-smokers (P<0.001). Histologically, EGFR (P<0.001) and RBM10 (P=0.02) mutations appeared predominantly in LUAD, whereas TP53 mutations were more commonly observed in lung squamous carcinoma (LUSC) (P<0.001). Lymphatic metastasis showed distinct distribution patterns: TP53, MUC16, and KRAS mutations occurred more frequently in metastasis-positive cases, whereas EGFR and RBM10 mutations were more common in metastasis-negative tumors (all P<0.05). Patients with advanced disease stages exhibited higher frequencies of TP53 mutations (P<0.001), while EGFR (P<0.001) and RBM10 mutations (P=0.04) were more frequently observed in early-stage tumors.

TP53-EGFR and TP53-KRAS co-mutations emerged as a highly frequent pattern in patients with actionable and non-actionable mutations
Our analysis revealed that 84.7% (365/431) of NSCLC patients harbored detectable driver mutations. Within this subgroup, 32.9% (120/365) exhibited concurrent actionable and non-actionable mutations (Figure 4A). Among patients with actionable mutations, EGFR L858R and EGFR exon 19 deletions (19del) represented the predominant subtypes, accounting for 50% and 37.2% of cases, respectively (Figure 4B). In the non-actionable mutation group, TP53 mutations (66%), PIK3CA mutations (12.4%), and non-G12C KRAS mutations (12.4%) constituted the most frequent alterations (Figure 4C). Co-mutation analysis demonstrated that patients with ≥1 actionable mutation alongside non-actionable alterations (83.3%) comprised the majority of those with multiple mutations (Figure 4D). The most prevalent combination involved EGFR actionable mutations co-occurring with TP53 mutations (70%, 84/120), followed by EGFR actionable mutations co-occurring with PIK3CA co-mutations (17.5%, 21/120). Among patients with ≥2 non-actionable alterations, KRAS G12C co-mutations with TP53 (13.0%, 3/23) emerged as the most frequent pattern (Figure 4E).

TP53 co-mutations associated with elevated PD-L1 expression and TMB
To explore potential clinical implications, we examined the relationships between these immunotherapy biomarkers and the most common co-mutation patterns identified in our cohort: TP53-EGFR and TP53-KRAS co-mutations. A total of 169 patients harbored TP53 mutations, 256 patients had EGFR mutations, and 43 patients carried KRAS mutations. Among them, 58.0% (98/169) of patients had TP53-EGFR co-mutations, and 27.9% (12/43) had TP53-KRAS co-mutations. TPS ≥1% was defined as PD-L1 positivity. Among patients with EGFR mutations but without TP53 mutations, 10.1% (16/158) were TPS-positive; among patients with TP53 mutations but without EGFR mutations, 52.1% (37/71) were TPS-positive; and among patients with TP53-EGFR co-mutations, 23.5% (23/98) were TPS-positive. Among patients with KRAS mutations but without TP53 mutations, 22.6% (7/31) were TPS-positive; among patients with TP53 mutations but without KRAS mutations, 33.1% (52/157) were TPS-positive; and among patients with TP53-KRAS co-mutations, 66.7% (8/12) were TPS-positive. Figure 5A showed that significant differences in the level of positive TPS were present across mutation subgroups (P=0.001). Patients with TP53-only mutations [44%±38%, 95% confidence interval (CI): 32–57%] had significantly higher TPS compared to EGFR-only mutations (7%±7%, 95% CI: 4–11%, P=0.001). Notably, patients with TP53-EGFR co-mutations (19%±23%, 95% CI: 9–28%) had higher TPS levels than EGFR-only mutations (P=0.03). A similar pattern was seen for TMB, with significant differences across these subgroups (P<0.001) (Figure 5B). Patients with TP53-only mutations had significantly higher TMB (8.90±5.81, 95% CI: 7.57–10.27) compared to those with EGFR-only mutations (2.82±2.87, 95% CI: 2.73–3.28, P<0.001) or TP53-EGFR co-mutations (3.82±3.09, 95% CI: 3.20–4.44, P=0.03). TMB levels also varied significantly between EGFR-only mutations and TP53-EGFR co-mutations (P<0.001). Figure 5C displayed that further analysis of TP53-KRAS co-mutations revealed the differences of the level of positive TPS (P=0.007). Patients with TP53-KRAS co-mutations (67%±36%, 95% CI: 36–97%) showed significantly higher TPS than TP53-only mutations (31%±31%, 95% CI: 23–40%, P=0.02) or KRAS-only mutations (15%±15%, 95% CI: 1–29%, P=0.01). However, there was no significant difference between KRAS-only mutations and TP53-only co-mutations (P=0.69). Similarly, the difference was seen in TMB analysis (P=0.02) (Figure 5D). Patients with TP53-KRAS co-mutations (9.04±5.59, 95% CI: 5.48–12.59) had significantly higher TMB compared to those with KRAS-only mutations (4.79±5.81, 95% CI: 2.66–6.93, P=0.01). However, no statistically significant difference in TMB was observed between the TP53-KRAS co-mutation group and the TP53-only mutation group (5.72±4.99, 95% CI: 4.94–6.51; P=0.13). There were no different TMB levels between TP53-only mutations and TP53-KRAS co-mutations (P=0.19).

Discussion
Our study delineated the mutational landscape of 431 Chinese patients with NSCLC using a targeted NGS panel covering 654 genes, identifying a total of 2,669 somatic variants. We specifically focused on evaluating the associations between TP53-EGFR and TP53-KRAS co-mutations and key immunologic biomarkers, including PD-L1 expression and TMB. To our knowledge, this is among the first studies to systematically assess immunological biomarkers in patients harboring these co-mutation patterns.
In our study, EGFR was the gene with the highest frequency of somatic mutations detected. A study about the mutation profile of 32 driver genes in a population from Taiwan showed that the highest mutation frequency was found in EGFR, followed by TP53 (21), which was the same as our findings. Another study showed that genes with the most frequent mutations were KRAS and EGFR in the Portuguese population (22). The reasons for this slight discrepancy may be racial and regional. In addition, a genomic profiling study conducted in smoking patients reported that EGFR and KRAS were the most common driver mutations, whereas TP53 represented the most frequent concomitant mutation in advanced NSCLC with a smoking history (23). In our study, however, a potential selection bias resulted in a relatively small proportion of smokers, accounting for only approximately 19% of the cohort, which may partly explain the discrepancy between the findings.
The prevalence rate of EGFR mutations was 59.1%, aligning with reported EGFR mutation rates in lung adenocarcinoma (50–60%) (24), consistent with our cohort composition, where 91.4% of patients had adenocarcinoma histology. Two mutations, deletions in exon 19 and the single amino acid substitution L858R in exon 21, often referred to as “classical” EGFR mutations, together account for about 85% of observed EGFR mutations in NSCLC (25). This oncogenic driver induces constitutive EGFR activation through structural conformation changes, leading to persistent downstream signaling (e.g., PI3K/AKT, MAPK pathways) that promote tumor progression (26). EGFR mutations are more frequently observed in female patients, non-smokers, individuals with early-stage disease, and those without lymph node metastasis, which is consistent with the findings of a recent meta-analysis. This analysis further demonstrated that, compared to other histological subtypes, adenocarcinoma is significantly associated with a higher prevalence of EGFR mutations (27). These results reinforce the critical need for early EGFR testing in NSCLC patients, and such testing enables timely TKI initiation.
TP53 mutations emerged as the second most prevalent genomic alteration (39.7%), trailing only EGFR mutations. Most of the TP53 variants were clustered exclusively within exons 4–9 on chromosome 17, regions encoding the critical DNA-binding domain essential for tumor suppressor function. NSCLC with TP53 alterations carries a worse prognosis and may be relatively more resistant to chemotherapy (28). TP53 R273C hotspot mutation may attenuate the effectiveness of osimertinib in a previous study (29). In our study, TP53 R273C mutation was identified in 5 patients, who may require closer monitoring and careful assessment of treatment response due to the potential impact of this mutation on drug sensitivity. A separate study suggests that a substantially higher frequency of TP53 mutations was seen in squamous cell carcinoma compared to adenocarcinoma, and the number of mutations increased with tobacco consumption (30), supporting the results in our study.
KRAS represents the most frequently mutated oncogene in advanced NSCLC in Western populations, with the G12C variant constituting the predominant mutation subtype (31). The recent FDA approval of KRAS G12C-specific inhibitors, sotorasib and adagrasib, effectively transforms KRAS from an “undruggable” target to an actionable oncogenic driver (32). Beyond G12C, our study identified additional prevalent KRAS variants, including G12D and G12V. Specifically, KRAS G12D promotes an immunosuppressive tumor microenvironment through higher PD-L1 expression and T-cell exclusion, thereby conferring resistance to immune checkpoint inhibitors (33). In contrast, KRAS G12V preferentially activates the MEK/ERK signaling axis over the PI3K/AKT pathway (32), potentially explaining its association with EGFR TKIs resistance in NSCLC.
Early reports from the Lung Cancer Mutation Consortium suggested that dual driver mutations in NSCLC were rare. However, with the advent of newer NGS platforms, the prevalence of dual TP53-EGFR mutations ranges from 30% to 60% of TP53 mutations in EGFR-mutated tumors (8,34), with 58.0% observed in our study. The incidence of EGFR and KRAS co-mutations is relatively low, around 2% to 5% (35), while they occurred independently in our study. In our study, 12 patients exhibited TP53-KRAS co-mutations and 98 patients harbored TP53-EGFR co-mutations. Previous studies have suggested that TP53 co-mutations may be associated with poorer progression-free survival (36,37). On one hand, the cellular tumor antigen p53, encoded by TP53 gene, plays a central role in responding to DNA damage and determines the outcome of the DNA damage checkpoint response and dysfunctional p53 results in cells that, despite a damaged genome, continue to proliferate, thus fueling malignant transformation (38). On the other hand, concurrent TP53 mutations facilitate the evolution of resistance to TKIs in EGFR-mutant lung adenocarcinoma (8,39). In addition, EGFR-PIK3CA co-mutations have been identified as another genetic mechanism of TKI resistance (40), and a relatively high frequency of EGFR-PIK3CA co-mutations was also observed in our cohort. Notably, patients harboring TP53 co-mutations have been reported to derive longer progression-free survival from immune checkpoint blockade therapy (14,41,42).
A previous study demonstrated that high PD-L1 expression was significantly associated with EGFR co-mutation with TP53, while EGFR mutation alone was not associated with high PD-L1 expression (43). TMB has emerged as an important genomic feature with both biological and clinical implications in lung cancer. Patients with TP53-EGFR or TP53-KRAS co-mutations in NSCLC have been reported to exhibit higher TMB levels (8,14). In our cohort, TMB exhibited heterogeneous distribution across different mutation subgroups, with TP53 co-mutations showing markedly higher TMB. PD-L1 expression and TMB level are predictive biomarkers for response to immunotherapy in NSCLC. One study specifically reported that high TMB and elevated PD-L1 expression are predictive of benefit from immune checkpoint blockade treatment in oncogene-driven NSCLCs (44). Therefore, integrating TP53 co-mutation patterns with PD-L1 expression and TMB levels may provide a valuable framework for refining patient selection and expanding the population that benefits from immune checkpoint inhibitor therapy.
While our findings demonstrate clinical relevance, study limitations, including sample size constraints and potential selection bias, warrant validation through large-scale prospective trials to establish robust therapeutic algorithms for NSCLC patients with co-occurring driver mutations.

Conclusions
In conclusion, systematic characterization of somatic mutation profiles and co-mutation patterns in NSCLC provides critical insights for elucidating the interplay between concurrent genetic alterations and immunotherapy biomarkers. The identification of specific co-mutation signatures, particularly TP53-EGFR and TP53-KRAS co-alterations associated with elevated PD-L1 expression and increased TMB, offers compelling biological rationale for immunotherapy selection in these molecularly defined subgroups.

Supplementary
The article’s supplementary files as