Screening air pollutants-related genes to construct a prognostic risk model for lung adenocarcinoma and analyzing its immune microenvironment

Abstract

Background

As the most common histological type of lung cancer, lung adenocarcinoma (LUAD) remains a major global health concern. The acceleration of worldwide industrialization has led to deteriorating air quality, which is recognized as a contributing factor in the development and advancement of numerous malignancies. This study aims to investigate the value of air pollutants-related genes (APRGs) as potential biomarkers for LUAD.

Methods

Drawing on data from The Cancer Genome Atlas-LUAD and the GSE42127 cohort, this study identified key prognostic genes for LUAD through an integrated approach combining differential expression analysis with univariate/multivariate Cox regression and least absolute shrinkage and selection operator regression. Patients were stratified into high- and low-risk groups based on the median risk score derived from these prognostic genes. Subsequently, the patterns of immune cell infiltration were evaluated between the two groups. Drug sensitivity analysis was also performed to predict patient responses to conventional chemotherapy drugs. Furthermore, consensus clustering of LUAD samples was conducted to identify molecular subtypes with distinct biological characteristics.

Results

Prognostic modeling based on eight air pollutants-related genes (APRGs) revealed that LUAD patients in the low-risk group experienced significantly superior overall survival. This survival benefit was accompanied by a notably enriched tumor microenvironment, characterized by elevated infiltration of B cells and resting memory CD4⁺ T cells. Furthermore, patients in the low-risk group may demonstrate greater sensitivity to crizotinib while exhibiting reduced responsiveness to gefitinib. Two robust molecular subtypes of LUAD were identified through consensus clustering.

Conclusions

By constructing prognostic models centered on APRGs, this investigation systematically elucidated the immune microenvironment and molecular underpinnings of LUAD, contributing fresh perspectives on disease mechanisms and potential treatment avenues.

Highlight box.

Key findings

• This study developed a robust predictive model that integrates eight genes associated with air pollutants.

What is known and what is new?

• Multiple air pollutants have been demonstrated to possess clear pulmonary pathogenicity and even carcinogenicity.

• This study developed a prognostic prediction model for lung adenocarcinoma that incorporates air pollutant characteristics and identified two subtypes with significantly distinct molecular regulatory and immune profiles.

What is the implication, and what should change now?

• Eight genes associated with air pollutants show potential as biomarkers for predicting the prognosis of lung adenocarcinoma. Future studies should validate these findings using prospective cohort samples and conduct further in vitro and in vivo experiments to explore them in greater depth.

Introduction

Lung cancer is a major driver of cancer-related deaths worldwide, with its pathological subtypes primarily including small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC) (1). Lung adenocarcinoma (LUAD), a predominant subtype of NSCLC, poses an increasingly significant disease burden and major public health challenge (2). Surgical intervention offers definitive benefits for patients with early-stage LUAD; however, a notable subset remains susceptible to postsurgical recurrence or distant metastasis, representing a major clinical challenge (3,4). With the advancement of precision medicine and the emergence and widespread application of targeted/immunotherapy, survival in LUAD patients has been significantly extended (5,6). However, clinical practice still faces severe challenges. First, the limited number of exploitable drug targets leaves some patients without effective treatment options; second, patients initially responsive to treatment commonly develop acquired resistance, leading to disease progression (7). Consequently, elucidating LUAD pathogenesis, pinpointing actionable therapeutic targets, and developing safer, more effective targeted therapies have emerged as pivotal research priorities.

With the rapid advancement of urbanization and industrialization, air pollution has become a significant environmental challenge facing many countries (8). Air pollution is detrimental to human health and is associated with numerous diseases, such as cardiovascular diseases (9), respiratory diseases (10,11), and metabolic disorders (12). The lungs, connected to the external atmosphere via the upper respiratory tract, are particularly vulnerable to environmental pollutants (10). Multiple air pollutants have been demonstrated to possess clear pulmonary pathogenicity and even carcinogenicity. For instance, inhalable particulate matter (PM) can penetrate deep into the lungs, enter the alveoli, and subsequently enter the bloodstream, and has been classified as a carcinogen (13). For example, long-term exposure to PM2.5 can induce pulmonary immunosuppression (14). The compound auramine O released by incense burning can accumulate within the nuclei of lung cancer cells, significantly enhancing their migration and invasive capabilities (15). Furthermore, exposure to elevated levels of PM10 and volatile organic compounds is significantly associated with an increased risk of LUAD (16-18). Mechanistically, airborne carcinogens generate reactive oxygen species that directly damage DNA and organelles, while simultaneously hijacking signaling networks that regulate inflammation, cell cycle, and gene expression. The repeated or sustained activation of these pathways leads to the accumulation of mutations and epigenetic reprogramming, thereby fostering a microenvironment conducive to malignant transformation (19). In summary, air pollutants play a significant role in lung cancer development. However, their prognostic value in post-diagnosis assessment for LUAD patients—particularly the molecular mechanisms centered on air pollutants-related genes (APRGs) and their impact on survival outcomes—remains incompletely elucidated.

To address these knowledge gaps, this study leveraged bioinformatic (including differential expression analysis, univariate/multivariate Cox regression analysis, and least absolute shrinkage and selection operator (LASSO) regression analysis) approaches both to identify APRGs as prognostic markers in LUAD and to establish molecular subtypes via consensus clustering. Our systematic investigation underscores the multifaceted roles of APRGs in LUAD development and progression, offering mechanistic insights into the intersection of environmental pollution and tumor biology, while providing a foundation for improved prognostic stratification and personalized therapeutic approaches. We present this article in accordance with the TRIPOD reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0078/rc).

Methods

Data sources and processing

Transcriptomic data for The Cancer Genome Atlas-LUAD (TCGA-LUAD; 59 normal and 526 tumor samples) were acquired from the UCSC Xena database (https://xena.ucsc.edu/) to form the training set. The validation set comprised microarray data from the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/) under the accession number GSE42127 (tumor: 176). Additionally, 257 APRGs were acquired from the study by Pu et al. (20). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Preliminary screening of LUAD prognostic genes

Differential expression analysis was performed between normal and tumor samples in the LUAD training set utilizing the limma package, employing thresholds of |logFC| >0.585 and an adjusted P value <0.05 to identify differentially expressed genes (DEGs). The intersection between DEGs and APRGs was calculated to obtain differentially expressed APRGs (DEAPRGs). Further, univariate Cox regression analysis was performed on DEAPRGs employing the survival package to identify candidate prognostic genes for LUAD.

Development of a prognostic model for LUAD

To minimize redundancy and refine the prognostic gene set, LASSO regression with cross-validated lambda selection was performed using the glmnet package to filter highly correlated DEGs. The selected genes were subsequently analyzed via multivariate Cox regression in the survival package to establish a prognostic model. The risk score was calculated according to the following formula:

$$Riskscore={\displaystyle \sum \left(Geneexpressio{n}_i\times coefficien{t}_i\right)}$$ [1]

Model validation was conducted in both training and validation cohorts using Kaplan-Meier (K-M) survival analysis, receiver operating characteristic (ROC) curves, and risk score distribution patterns according to survival outcomes.

Gene enrichment analysis

To explore functional differences between risk groups, we employed GSEA software (v4.3.2) for gene set enrichment analysis (GSEA). Additionally, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed on DEGs meeting the threshold of |logFC| >1 and false discovery rate (FDR) <0.05 using the clusterProfiler package.

Construction and evaluation of nomogram

Following the identification of significant predictors through univariate and multivariate Cox analyses—including clinical features, risk score, and prognostic genes—a nomogram was constructed to quantify individualized risk for LUAD patients. The clinical utility of this predictive tool was rigorously assessed using calibration curves for 1-, 3-, and 5-year survival predictions.

Characterization of immune infiltration and assessment of immunotherapy response potential

To elucidate the immune contexture associated with the prognostic model, we applied CIBERSORT (CIBERSORT package) and single sample gene set enrichment analysis (ssGSEA, GSVA package) algorithms to quantify immune cell infiltration. Pearson correlation analysis was used to examine the relationship between gene expression and both cellular and functional immune signatures. The translational potential was further assessed by evaluating differential responses to immune checkpoint inhibition: based on the IMvigor210 cohort (21), patients were stratified into responders (R) and non-responders (NR) to anti-PD-L1 therapy for comparison across risk groups. Furthermore, immunophenotype scores (IPS) from the TCIA database were analyzed to predict the likelihood of response to CTLA-4 and PD-1 blockade in each risk group.

Tumor mutational burden (TMB) and drug sensitivity analysis

Genomic alterations in high- and low-risk groups were characterized by analyzing mutation frequencies in the LUAD training set; waterfall plots constructed with the maftools package displayed the 20 genes exhibiting the highest mutation rates in each group. Pharmacological profiling was conducted by querying the CellMiner database to identify compounds whose activity correlated with prognostic gene expression. Chemotherapeutic sensitivity was further quantified by calculating half maximal inhibitory concentration (IC₅₀) values for standard agents via the pRRophetic package.

Identification of LUAD molecular subtypes

Molecular subtypes of LUAD were identified through consensus clustering implemented via the ConsensusClusterPlus package. Clustering stability was assessed by analyzing the cumulative distribution function (CDF) and the corresponding delta area curve, ensuring robust stratification of the training set samples.

Statistical analysis

R software (v4.4.0) was applied to data analysis and visualization. Wilcoxon test was utilized for inter-group comparison. Correlation analyses were performed using Pearson and Spearman coefficients, with P<0.05 considered statistically significant.

Results

Identification of candidate prognostic genes in LUAD

From the LUAD training set, 4,064 DEGs were identified using thresholds of |logFC| >0.585 and adjusted P<0.05 (Figure 1A). Intersection with APRGs produced 82 DEAPRGs (Figure 1B), of which 30 were significantly associated with prognosis in univariate Cox regression analysis (P<0.05, Figure 1C). These 30 candidate genes exhibited strong correlations (Figure 1D) and were significantly differentially expressed between normal and tumor tissues (Figure 1E).

Figure 1.

Identification of candidate prognostic genes in LUAD. (A) Volcano plot displaying DEGs. (B) UpSet plot displaying the intersection between DEGs and APRGs. (C) Univariate Cox regression analysis of DEAPRGs (P<0.05). (D) Correlation analysis heatmap of DEAPRGs. (E) Comparative analysis of DEAPRG expression in normal versus tumor groups. *, P<0.05; **, P<0.01; ***, P<0.001. APRG, air pollutant-related gene; DEAPRG, differentially expressed air pollutant-related gene; DEG, differentially expressed gene; LUAD, lung adenocarcinoma.

Construction and validation of a prognostic model for LUAD

Multivariate refinement of the prognostic signature began with LASSO regression, which reduced candidate genes to 15 (Figure 2A,2B). Subsequent multivariate Cox analysis identified eight genes for inclusion in the final risk model (Figure 2C), all of which demonstrated significant differential expression between risk groups (Figure S1A) and independent prognostic capacity (Figure S1B). Comprehensive validation confirmed model robustness: time-dependent ROC analysis yielded strong predictive performance in both training and validation cohorts (Figure 2D); survival analysis revealed significantly superior outcomes for low-risk patients (Figure 2E); and risk score distribution plots visually reinforced the survival disparity (Figure 2F). Expression patterns revealed upregulation of TYMS, EPHB2, and GPR35 in tumors, with concurrent downregulation of the remaining five genes (Figure 2G).

Figure 2.

Construction and validation of a LUAD prognostic model. (A,B) LASSO regression analysis identified 15 candidate genes through coefficient profiles and cross-validation for optimal λ. (C) Multivariate Cox regression further refined the signature to eight independent prognostic genes. (D-F) Model validation in training (upper panels) and validation (lower panels) cohorts. (D) ROC curves demonstrated predictive accuracy, (E) Kaplan-Meier survival analysis revealed significantly longer survival in the low-risk group, and (F) risk score distribution plots confirmed outcome disparities between risk groups. (G) Differential expression of the eight prognostic genes in tumor versus normal tissues. *, P<0.05; **, P<0.01; ***, P<0.001. AUC, area under the curve; LASSO, least absolute shrinkage and selection operator; LUAD, lung adenocarcinoma; ROC, receiver operating characteristic.

Gene enrichment analysis

GSEA revealed significant enrichment in pathways such as Amplified MYC to p27 cell cycle G1/S, DNA Replication Termination, and DNA Replication Licensing within the high-risk group (Figure 3A); while the low-risk group showed significant enrichment in variant mutation inactivated atp1a1 to angiotensin aldosterone signaling pathway, and reference bcr plcg calcineurin signaling pathway (Figure 3B). GO enrichment analysis revealed that DEGs were primarily involved in biological processes related to cell division, including sister chromatid segregation and mitotic sister chromatid segregation (Figure 3C). Additionally, KEGG pathway analysis showed significant enrichment in cell cycle, cellular senescence, and oocyte meiosis pathways (Figure 3D).

Figure 3.

Functional enrichment profiling. GSEA revealing differentially activated pathways in the (A) high-risk and (B) low-risk groups. (C) GO and (D) KEGG enrichment analysis of DEGs across different risk groups. DEGs, differentially expressed genes; FC, fold change; GO, Gene Ontology; GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes.

A nomogram for LUAD prognosis

We observed a strong association between the risk score and multiple clinical parameters, with significantly elevated scores observed in specific subgroups: males relative to females, advanced local invasion (T3 + T4 versus T1 + T2), late-stage disease (Stage IV versus I–III), presence of nodal involvement (N1−3 versus N0), and distant metastasis (M1 versus M0) (Figure 4A). Univariate Cox regression confirmed that risk score, pathological stage, and T, N, and M classifications each had significant prognostic value (Figure 4B). In the multivariate analysis, risk score, T stage, and N stage emerged as robust independent predictors of patient outcomes (Figure 4C). To translate these findings into a clinically applicable tool, we developed a nomogram integrating the risk score with established clinicopathological variables to predict survival in LUAD patients (Figure 4D). The model’s predictive accuracy was substantiated by calibration plots, which showed excellent concordance between predicted and actual survival probabilities at 1, 3, and 5 years, underscoring its potential for guiding clinical decision-making (Figure 4E).

Figure 4.

A nomogram for LUAD prognosis. (A) Comparison of risk scores for various clinical characteristics. (B) Univariate and (C) multivariate Cox regression analysis. (D) The nomogram illustrated the combined contribution of clinical variables and the risk score. (E) Calibration curve of nomogram. LUAD, lung adenocarcinoma.

Immune infiltration analysis and prediction of immunotherapy response

The immune microenvironment landscape was compared between risk groups through the following analyses. Based on ssGSEA algorithm results, most immune cells (e.g., B cells) and immune functions (e.g., APC co-stimulation) exhibited significantly higher infiltration levels in the low-risk group compared to the high-risk group (Figure 5A). CIBERSORT analysis showed elevated M0 and M1 macrophages in the high-risk group, while monocytes were enriched in the low-risk group (Figure 5B). ESTIMATE algorithm analysis confirmed these findings, with the low-risk group displaying higher Stromal, Immune, and ESTIMATE scores alongside reduced tumor purity (Figure 5C). IPS analysis predicted that low-risk patients would respond better to anti-CTLA-4 therapy but poorly to anti-PD-1 therapy compared to high-risk patients (Figure 5D), consistent with the observed upregulation of most immune checkpoint molecules in the low-risk group (Figure 5E). Clinically, low-risk patients exhibited superior overall survival (Figure 5F) and significantly higher response rates to anti-PD-L1 therapy (Figure 5G), with responders showing markedly lower risk scores than non-responders (Figure 5H). Correlation analysis further identified positive associations between macrophage M1 infiltration and TYMS/TLR8 expression, while CFTR expression was negatively correlated with M1 infiltration (Figure 5I).

Figure 5.

Characterization of immune infiltration and assessment of immunotherapy response potential. (A) Immune infiltration analysis based on ssGSEA, (B) CIBERSORT, and (C) ESTIMATE algorithms. (D) IPS violin plot. (E) Comparative analysis of immune checkpoint gene expression between risk groups. (F) Overall survival of high- and low-risk patients in the IMvigor210 cohort. (G) Anti-PD-L1 therapy response rates stratified by risk group. (H) Risk score distribution in NR and R groups. (I) Correlations between immune cell infiltration and prognostic gene expression levels. ns, P≥0.05; *, P<0.05; **, P<0.01; ***, P<0.001. IPS, immunophenotype score; NR, no response; PD-L1, programmed death-ligand 1; R, response; ssGSEA, single sample gene set enrichment analysis.

Profiling of TMB and drug sensitivity

Assessment of somatic mutation landscapes using TMB analysis demonstrated that high-risk patients harbored a higher frequency of TP53 mutations (28%) than their low-risk counterparts (17%) (Figure 6A,6B). Moreover, the high-risk group exhibited significantly elevated global TMB levels relative to the low-risk group (Figure 6C). Drug sensitivity profiling indicated that low-risk status correlated with increased sensitivity to several agents, most notably crizotinib, erlotinib, and dabrafenib compared to low-risk patients, while gefitinib showed the opposite trend (Figure 6D). Furthermore, correlation analysis indicated a positive association between EPHB2 expression levels and drug sensitivity to afatinib and erlotinib (Figure 6E; Table S1).

Figure 6.

Profiling of TMB and drug sensitivity. (A,B) Distinct mutational profiles of (A) high-risk and (B) low-risk groups, showing the top 20 mutated genes. (C) Significantly elevated TMB in the high-risk group compared to the low-risk group. (D) Differential chemosensitivity between risk groups. (E) Relationship between prognostic gene expression and multidrug sensitivity, revealing potential therapeutic targets. IC₅₀, half maximal inhibitory concentration; TMB, tumor mutational burden.

Molecular subtyping of LUAD

Consensus clustering using prognostic gene expression profiles identified two molecular subtypes of LUAD (group 1 and group 2; Figure 7A). The optimal number of clusters was determined through examination of the CDF and delta area curve (Figure 7B). Principal component analysis confirmed distinct transcriptomic profiles between the two subtypes (Figure 7C). K-M analysis revealed significantly superior overall survival for group 1 relative to group 2 (Figure 7D). Differential expression analysis (Figure 7E) between subtypes followed by GO enrichment analysis revealed significant enrichment in biological processes including microtubule binding, water transmembrane transporter activity, and water channel activity. KEGG analysis indicated that DEGs were mainly implicated in the bile secretion and cell cycle signaling pathways (Figure 7F). Furthermore, gene expression analysis revealed that TLR8, NOS1, CYP2F1, CFTR, and ID2 were upregulated in group 2, while TYMS and EPHB2 were downregulated (Figure 7G). Immune characterization of the two molecular subtypes revealed distinct microenvironmental profiles. Group 2 displayed higher ESTIMATE, Stromal, and Immune scores with lower tumor purity (Figure 7H), indicating a more immunologically active microenvironment. Consistently, multiple algorithms demonstrated enhanced B cell-related infiltration in group 2: CIBERSORT showed elevated memory B cells (Figure 7I), ssGSEA confirmed increased overall B cell abundance (Figure 7J), and MCP-counter validated higher B lineage cell infiltration (Figure 7K).

Figure 7.

Molecular subtyping of LUAD. (A) Consensus clustering matrix (k=2) showing sample stratification. (B) Cluster stability assessed by the CDF and delta area curve. (C) Principal component analysis plots of the two groups. (D) Kaplan-Meier survival analysis across groups. (E) DEGs between groups. (F) Functional enrichment (GO and KEGG) analysis of the DEGs. (G) Comparison of prognostic gene expression levels across groups. (H) Immune infiltration analysis using ESTIMATE, (I) CIBERSORT, (J) ssGSEA, and (K) MCP-counter algorithms. ns, P≥0.05; *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001. BP, biological process; CC, cellular component; CDF, cumulative distribution function; DEGs, differentially expressed genes; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; LUAD, lung adenocarcinoma; MCP, microenvironment cell populations; MF, molecular function; ssGSEA, single sample gene set enrichment analysis.

Discussion

The acceleration of global industrialization is inextricably linked to the intensification of air pollution. Long-term exposure to air pollutants, including PM2.5, NO₂, SO₂, and O₃, has been established as a significant contributor to the incidence and mortality of multiple cancer types (22,23). However, its relationship with LUAD remains poorly understood, particularly the role of air pollutant-associated genes in LUAD progression. To address this, this study integrates multiple bioinformatics approaches to systematically investigate the molecular mechanisms and immune characteristics of key APRGs in LUAD, evaluating their potential value as novel biomarkers.

This study integrated multiple bioinformatics approaches to identify eight key APRGs as prognostic markers for LUAD. These genes exhibited distinct differential expression patterns in LUAD tumor tissues: TYMS, EPHB2, and GPR35 were upregulated, while TLR8, NOS1, CYP2F1, CFTR, and ID2 were downregulated. Among highly expressed genes, thymidylate synthetase (TYMS)—an essential folate-dependent enzyme—exhibits elevated mRNA and protein levels correlated with poor prognostic outcomes in multiple hematologic malignancies and solid tumors (24,25). Knockdown of TYMS reduces LUAD cell proliferation capacity (26). Low EPHB2 expression predicts improved patient survival and reduced mortality (27), and its knockdown also impairs cell proliferation and migration (28). GPR35 serves as a potent stimulator of tumor growth; its knockout significantly inhibits tumor growth and immune infiltration (29). Among the downregulated genes, CFTR is closely associated with environmental exposures. For instance, nicotine exposure reduces CFTR expression and increases cell migration in lung cancer cells (30-32); both tobacco smoke extract and diesel exhaust particles cause CFTR dysfunction (33). CYP2F1 is primarily expressed in the lungs and is responsible for metabolizing various substances with specific respiratory toxicity. It is considered a key initiating factor in lung injury caused by environmental toxins (34,35). As a member of the Toll-like receptor family, TLR8 exhibits reduced expression in LUAD. Given that PM2.5 has been reported to inhibit TLR9 activation by interfering with endocytosis (36,37), we pose the question: Does PM2.5, within the pathological microenvironment of LUAD, interfere with TLR8’s innate immune function through similar or alternative mechanisms? Could this create a synergistic effect with the low expression of TLR8, collectively weakening the body’s immune defense against tumors or concurrent infections? Further investigation is warranted. Additionally, ID2 overexpression may suppress LUAD cell migration, invasion, and proliferation (38,39). NOS1 also plays a similar role in esophageal cancer (40). In summary, these genes are not only associated with LUAD prognosis but also interact with environmental exposures and the tumor microenvironment. Notably, CFTR, CYP2F1, and TLR8 are linked to air pollutants, offering new insights into the molecular mechanisms by which environmental factors drive LUAD progression.

As central mediators within the tumor microenvironment, immune cells critically orchestrate tumor progression and influence therapeutic outcomes (41). Comparative analysis of immune cell infiltration demonstrated that the low-risk group exhibited enrichment of B cells and CD4⁺ memory resting T cells, while the high-risk group showed elevated levels of CD4⁺ memory activated T cells. This finding aligns with existing research: B cells exert antitumor functions by enhancing T cell immunity, promoting interferon-γ production, and aiding natural killer cell antitumor activity (42-44). Moreover, high B cell infiltration correlates with longer recurrence-free survival in LUAD patients (45). A notable imbalance in CD4⁺ memory T cell subsets was observed in LUAD, with tumor tissues showing lower abundance of resting memory cells and higher abundance of activated memory cells compared to normal controls (46). Exposure to air pollutants is an established factor driving the alteration of immune cell composition in the tumor microenvironment (47-49). Studies indicate particulate matter exposure alters immune cell infiltration patterns, including reduced B cells and increased IL-4-producing CD4⁺ T cells (50). Furthermore, chronic low-level particulate matter exposure sufficiently exacerbates acute lung injury and alters pulmonary CD4⁺ T cells, with this effect potentially persisting after exposure cessation (51). These results point to a potential mechanism whereby air pollutants could promote LUAD progression: the regulation of specific immune cell subsets within the tumor microenvironment. However, whether pollutants directly promote LUAD progression through this mechanism requires further experimental and clinical validation. Additionally, this study identified differences in treatment response to PD-1 and PD-L1 inhibitors among patients in the low-risk group, suggesting that the two may hold distinct clinical significance in predicting immunotherapy efficacy. This phenomenon may relate to their respective mechanisms of action: PD-1 inhibitors primarily bind to the PD-1 receptor on T-cell surfaces, blocking its interaction with ligands to restore the antitumor function of effector T cells; whereas PD-L1 inhibitors primarily target PD-L1 molecules on tumor cell surfaces, interfering with their binding to T-cell PD-1 and thereby releasing immune suppression (52,53). Relevant clinical studies also support the clinical significance of this mechanism difference. A meta-analysis incorporating 32 real-world studies demonstrated that anti-PD-1/PD-L1 immunotherapy significantly improves patient survival in first-line NSCLC treatment, with median progression-free survival and overall survival of 3.35 months and 9.98 months, respectively (54). These findings suggest that patient individuality and tumor immune microenvironment characteristics should be considered during immunotherapy strategy selection to achieve more precise treatment decisions.

The TMB analysis results of this study indicate that tumor protein p53 (TP53) mutations are the most common in LUAD, with a mutation frequency of 28% in the high-risk group, significantly higher than that in the low-risk group (17%). This suggests that TP53 may play a key role in LUAD disease progression. While TP53 induces alveolar type 1 cell differentiation in LUAD (55), its mutation disrupts homeostasis, thereby driving malignant progression and contributing to adverse clinical outcomes (56). Additionally, lung cancer patients who never smoked but resided in areas with severe air pollution were more likely to carry TP53 mutations and exhibit shorter telomeres (57). This suggests environmental exposure factors may synergistically promote LUAD progression with TP53 mutations, though the specific mechanisms require further investigation. Notably, epidermal growth factor receptor (EGFR) kinase domain activating mutations represent one of the most common and targetable oncogenic driver mutations in LUAD (58,59). According to our TMB analysis results, the EGFR mutation frequency was 8% in the low-risk group and 4.6% in the high-risk group. This discrepancy may be influenced by multiple factors, including geographic background and cohort composition. Specifically, EGFR mutations occur significantly more frequently in Asian populations than in non-Asian populations, whereas the data in this study were derived from the TCGA database. Additionally, EGFR mutations are more common among non-smokers, suggesting smoking history as a potential confounding factor (60,61). Interestingly, recent studies indicate a strong association between air pollution and the development of EGFR-mutant lung cancer. For instance, in environments with high air pollution exposure, the cumulative incidence of EGFR-driven lung cancer cases within three years was significantly higher than in low-exposure populations (62). Mechanistic studies further reveal that prolonged exposure to PM2.5 promotes sustained activation of the EGFR signaling pathway in LUAD cells harboring EGFR-sensitive mutations (such as L858R and T790M), thereby enhancing cellular proliferation and accelerating tumor progression (62,63). These findings underscore the public health significance of improving air quality for lung cancer prevention. Therapeutically, TP53 mutations reduce sensitivity to crizotinib in ALK-rearranged NSCLC patients and correlate with poor prognosis (64). Consistent with this, our study found that patients in the low-risk group demonstrated superior response to crizotinib and better prognosis. Additionally, we observed a slightly higher KRAS proto-oncogene, GTPase (KRAS) mutation rate (14%) in the low-risk group compared to the high-risk group (13%), with reduced sensitivity to gefitinib in this subgroup. This may result from the PI3K/AKT pathway promoting gefitinib resistance in KRAS-mutated LUAD through deacetylase-dependent mechanisms (65). In summary, TP53 and KRAS mutations exhibit differential distribution across distinct risk groups in LUAD and may influence treatment response and disease progression through distinct molecular pathways. Collectively, these observations yield critical clues for advancing personalized medicine in LUAD.

While this study offers a systematic characterization of APRGs in LUAD, several limitations should be acknowledged. The clinical applicability of the identified molecular subtypes requires validation in prospective clinical trials, and our reliance on public databases may limit sample diversity. Furthermore, the mechanistic interplay among air pollutants, prognostic genes, immune cells (B cells and CD4⁺ T cells), and targeted therapies (crizotinib and gefitinib) remains to be fully elucidated. Future work should prioritize multicenter validation studies, functional experiments in cellular and animal models.

Conclusions

This study systematically revealed the close association between APRGs and the LUAD immune microenvironment, key driver mutations (such as TP53 and KRAS), and targeted therapy response by constructing a prognostic model for these genes. This study offers novel molecular insights and suggests potential intervention strategies, contributing to both a deeper understanding of environmental drivers in LUAD progression and the advancement of personalized treatment.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.