Skip to main content
NCBI home page
European Journal of Medical Research logoLink to European Journal of Medical Research
. 2025 Nov 27;30:1301. doi: 10.1186/s40001-025-03544-w

Molecular subtyping and a seven-gene immune signature reveal heterogeneity in tumor microenvironment and prognosis of lung adenocarcinoma

Hongzhi Li 1,✉, Xian Gao 1, Chengde Chen 1, Zhongfeng Cui 2, Xiaojiu Cao 1, Jing Su 1, Guangming Li 3,✉
PMCID: PMC12752188  PMID: 41310864

Abstract

Background

Lung adenocarcinoma (LUAD) is a leading cause of cancer deaths. Given that traditional pathologic features to diagnose LUAD do not fully reflect the biological differences in patients, the search for novel biomarkers is necessary.

Methods

In this study, we obtained immune-related genes (IRGs) from ImmPort and performed cluster analysis on The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) to mine LUAD subtypes with different immune characteristics. Quantitative analysis of IRGs was performed by single-sample gene set enrichment analysis (ssGSEA). Based on the univariate cox and LASSO regression methods, we screened the characteristic genes that significantly affected LUAD and built the model based on the RiskScore coefficients. The relative expressions of characteristic genes in LUAD were determined using qRT-PCR. Transwell and wound healing assays were utilized to verify the practical regulation of these genes on the migration and invasion levels of LUAD. Correlations were established between RiskScore and LUAD drug sensitivity by oncoPredict.

Results

We acquired three LUAD subtypes and demonstrated heterogeneous IRGs scores and clinical features. The molecular subtypes were differentially enriched in bile acid metabolism, fatty acid metabolism, and ECM–receptor interaction. This study identified seven genes (MS4A1, EXO1, CPS1, ZNF750, S100P, NT5E, KCNN4) as a signature affecting prognosis, from the differentially expressed genes (DEGs) among the molecular subtypes, and constructed a RiskScore for the prognosis of LUAD. Cellular experiments verified that 6 of 7 characteristic genes were expression dysregulation in LUAD cell line. Silencing of EXO1 significantly suppressed the migration and invasion of LUAD cell lines. RiskScore and immune checkpoints such as CD276, TNFSF4, and TNFSF9 showed a positive correlation.

Conclusions

This study identified three LUAD subtypes with distinct immune characteristics and constructed a seven-gene prognostic model. This model correlates with immune checkpoint and chemotherapy sensitivity, providing new targets and strategies for clinical diagnosis and treatment.

Keywords: Lung adenocarcinoma, Molecular subtype, RiskScore, Immunity, Immune checkpoints, Biomarkers

Introduction

Based on the American Cancer Society's data, lung cancer was the major cause of cancer-correlated mortality worldwide, with an incidence rate of 11.7% in 2020, and a mortality rate of 18%, ranking it first in cancer fatalities [1, 2]. Lung cancer is categorized into small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC), with the latter accounting for ~ 85% of all cases. Among NSCLCs, lung adenocarcinoma (LUAD) is a particularly common histological subtype [3]. For some time, surgical resection, chemotherapy, molecular therapies and their derivatives have been used to varying degrees in oncology, including novel approaches like targeting tumor prognostic biomarkers to alleviate cancer progression. However, no breakthrough has been achieved in alleviating cancers (including LUAD), primarily due to the lack of more effective targeted molecules [4, 5]. In addition, the lack of obvious early clinical symptoms leads to most patients being diagnosed at advanced stages, while the complex immune environment and diverse immune subtypes of advanced LUAD complicate treatment [6]. Thus, LUAD requires more targeted therapeutic strategies to be developed, and finding therapeutic targets that work is a high priority.

Numerous studies have indicated that tumor microenvironment heterogeneity influenced the degree of cancer malignancy, including the heterogeneity of immunocyte, stromal cell type, inflammatory and extracellular factors [7]. Among them, tumor-infiltrating immune cells are considered valuable for the diagnosis and prognostic prediction of the tumor [8]. A study summarizing the correlation between T cells and clinical outcomes in cancer found that cytotoxic T cells combined with Th1 cells, were positively correlated with favorable clinical outcomes in several cancers, comprising lung cancer [9]. Beyond that report, studies have shown that immunogens can influence the prognosis of cancer patients by regulating the TME. For example, genes, such as CCR6, ITK, CCR4, DOK2, and AMPD1, which influenced the tumor immune microenvironment and pathological staging, may serve as potential targets for tumor immunotherapy [10]. NPM1, a prognostic biomarker of LUAD, participates in the regulation of immune infiltration of LUAD through the activation of the B-cell receptor, and affects metabolic pathways, such as glycolysis, to regulate the LUAD progression [11, 12]. It has also been shown that the BTK, a gene that play a role as immune-regulating divisor in lung tissue, has been shown to be significantly related to elevated levels of B cells and CD8 + T cells, indicating its potential as a gene for lung cancer prognosis [13]. These findings demonstrate the relationship between cancer progression, immune-related genes, and the TME, offering potential targets for enhancing the treatment outcomes for patients suffering from malignant tumors.

In this study, based on consensus clustering of LUAD samples from The Cancer Genome Atlas (TCGA), we characterized the heterogeneity of LUAD samples in terms of immune patterns. Then, genes that significantly affected the prognosis of LUAD were mined based on the differentially expressed genes among immune subgroups and used to develop a prognostic assessment model. The correlation between model characterized genes and LUAD immune checkpoint expression and drug sensitivity was profoundly revealed in this study. The molecular markers and prognostic models screened in this study are expected to provide guidance for comprehensively understanding the molecular mechanisms of LUAD progression and improving clinical treatment strategies.

Materials and methods

Data acquisition

In this study, LUAD sample data were obtained from the Cancer Genome Atlas (TCGA) data set. Specifically, FPKM data from the LUAD data set were gained access to from the UCSC xena (https://xena.ucsc.edu/). The samples that did not contain clinical follow-up data were removed, and a total of 501 LUAD samples were obtained after removing those with a de-escalation time of survival of ≤ 30 days, and 59 control samples. A sum of 226 cancer tissues were obtained after downloading RNA-Seq data and clinical data for GSE31210 from the gene expression omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/), deleting normal tissue and converting probes to symbol based on the annotation file.

Cancer immunomodulation-related genes (IRGs) were obtained from the ImmPort base (https://www.immport.org/home). These data were downloaded in June 2024, and its corresponding data release version is Data Release DR54.

LUAD molecular subtypes based on IRGs

In this study, univariate Cox analysis was used to identify IRGs significantly associated with LUAD prognosis (p < 0.05). Subsequently, consensus clustering was performed using the R package ConsensusClusterPlus to identify molecular subtypes in LUAD samples [14]. The clustering was conducted with the following parameters: a maximum evaluated cluster count (maxK) of 10, bootstrap repetitions (reps) of 500, a subsampling ratio of 80% for each iteration (pItem = 0.8), and no feature resampling (pFeature = 1). Through this process, different molecular subtypes were mined and their prognoses were analyzed.

Single sample gene set enrichment analysis (ssGSEA)

ssGSEA is conducted through the GESA extension. ssGSEA calculates an enrichment score, which reflects the enrichment of a gene set within each sample in a certain data set [15]. This study assesses the associated enrichment in each sample by comparing the gene expression data of that sample to specific IRGs. The ssGSEA enrichment score indicates the degree to which a specific gene set is consistently up- or down-regulated in a given sample [16].

Modeling of prognostic risk assessment

In this study, variant expressed genes among the above LUAD immunological subtypes were obtained using the limma software [17], and gene screening was carried out by Cox univariate operation (p < 0.05). To further compress the number of differentially expressed genes, lasso operation method was used to mined differentially expressed molecules between subtypes that proved significantly inter-related to LUAD prognosis. We then determined the risk score for each tissue using the formula: RiskScore = Σβi × Expi, where i and β refer to the gene expression and the Cox operation associated coefficient for each gene, respectively. The RiskScore was subjected to zscore processing, and tissue samples were consisted of RiskScore high- and low-score groups ground to threshold “0”, and KM curves were displayed for validity of the patient's clinical diagnosis prediction under Kaplan–Meier method.

Correlation analysis of risk scores and drug sensitivity

In this study, R package oncoPredict was used to predict drugs' IC50 value for TCGA–LUAD data set samples [18]. Furthermore, we applied spearman correlation so as to mark the bond of chemotherapeutic sensitivity and RiskScore, setting p < 0.05 as well as |cor| > 0.3 as significant correlation.

Cell culture and transfection

The LUAD cell line A549 (BNCC337696, BeiNa Culture Collection, Xinyang, China) and the lung epithelial cell line BEAS-2B (BNCC359274, BeiNa Culture Collection, China) were purchased for this study for subsequent experiments. Dulbecco's Modified Eagle Medium (DMEM) containing 10% fetal bovine serum (FBS, 26140-095, Thermo Fisher Scientific, Waltham, Massachusetts, United States of America) and 1% antibiotics (15070-063, Thermo Fisher Scientific, United States of America) were used for culturing the cells. Cells were stored at 37℃ and 5% CO2. Transfection of EXO1 siRNA (sequence: 5’-CCUCUUUGCCUGAGAAUAATT-3’) with Lipofectamine 2000 (11668027, Invitrogen, Carlsbad, California, United States of America) and the negative control with scramble target sequence were ordered and applied for transfection.

Quantitative real-time fluorescence PCR (qRT-PCR)

RNA was reverse-transcribed to cDNA using the Qiagen One-Step RT-PCR kit (210212, Qiagen GmbH, Hilden, Germany) and subjected to qRT-PCR. An ABI 7500 system (4351106, Thermo Fisher Scientific, Waltham, Massachusetts, United States of America) using SYBR Green PCR Mix (D7260, Beyotime, Shanghai, China) by 7500 system (4351106, Thermo Fisher Scientific, United States of America) was used to conduct qRT-PCR [19]. All relative mRNA expression levels were normalized to GAPDH and calculated with the 2−ΔΔct method. The primer sequences of GAPDH were 5’-GTCTCCTCTGACTTCAACAGCG-3’ (Forward) and 5’-ACCACCCTGTTGCTGTAGCCAA-3’ (Reverse); of MS4A1 were 5’-CATTCTGTCGGCGATGCTGATC-3’ (Forward) and 5’-TCTCCAGCTGACAGCAGAACCA-3’ (Reverse); of EXO1 were 5’-TCGGATCTCCTAGCTTTTGGCTG-3’ (Forward) and 5’-AGCTGTCTGCACATTCCTAGCC-3’ (Reverse); of CPS1 were 5’-CATGGAACATCCAGCCGAATTGG (Forward) and 5’-GATGGCACATCCTCAGAGCCTT-3’ (Reverse); of ZNF750 were 5’-GGCGTAGAGATGCACCTGAT-3’ (Forward) and 5’-CCCAAGTTAAGTGCCTCTGC-03’ (Reverse); of S100P were 5’-CTCAAGGTGCTGATGGAGAAGG-3’ (Forward) and 5’-GAACTCACTGAAGTCCACCTGG-3’ (Reverse); of NT5E were 5’-CGCTCAGAAAGTTCGAGGTGTG-3’ (Forward) and 5’-CGCAGGCACTTCTTTGGAAGGT-3’ (Reverse); of KCNN4 were 5’-CACGCTGAGATGTTGTGGTTCC-3’ (Forward) and 5’-CTCCTTGGCATGGAAGACCACA-3’ (Reverse).

Cell counting kit-8 (CCK-8) assay

The assessment of cell proliferation was conducted using a 96-well plate, with 1000 cells introduced into each compartment following cell counting. After intervals of 0, 24, 48, and 72 h, 10 μL of CCK-8 solution (Dojindo Molecular Technique, Inc.,  Kumamoto, Japan) was administered to every well and allowed to incubate for 1 h. The absorbance (optical density) at 450 nm was measured for each well using an enzyme labeling apparatus (GEN10S-BASIC, Tecan, Männedorf, Switzerland).

Wound healing test

The cell line was seeded into a 6-well plate and grew until they covered the entire bottom. A vertical scratch was created with a 200 μL pipette tip. Cells were washed twice with phosphate buffered saline and photographed with an inverted microscope at 0 and 48 h after scratching. The wound healing rate was calculated as [(0 h width–48 h width)/0 h width] × 100%. The experiment was conducted three times.

Transwell analysis

The cells were supplemented into the upper chamber of a transwell instrument (pore: 8 μm, 3422, Corning, Inc., Corning, New York, United States of America) containing serum-free medium. Subsequently, PRMI-1640 medium (11875093, Thermo Fisher Scientific, United States of America) added with 10% FBS (26140-095, Thermo Fisher Scientific, United States of America) was added to the lower chamber. After incubation at 37 ℃ for 1 day, cells that did not migrate to the lower chamber were removed from the upper chamber. Migrated cells were fixed by 4% paraformaldehyde (P1110, Solarbio Lifesciences, Beijing, China) and dyed by crystal violet (G1063, Solarbio Lifesciences, China) for 30 min. The cells were observed under a microscope and counted.

Statistical analysis

The unpaired t test and two-way ANOVA were utilized to compare differences in continuous variables between the two groups or more than groups. All calculations were performed using R language (version 4.3.1) and GraphPad prism (version 8.0.2). A p value of less than 0.05 was considered statistically significant.

Results

Molecular typing of LUAD based on IRGs

Regarding to our research, we calculated the enrichment scores of IRGs in TCGA sample sets of LUAD by ssGSEA, and it was observed that the IRGs scores of LUAD were significantly more less than those of healthy tissues (Fig. 1A). Next, we applied a consensus clustering method to classify patients in reliance on the IRG expression pattern, and determined that k = 3 had more stable clustering results, and finally we obtained three molecular subtypes (Fig. 1B). The prognostic characteristics of these molecular subtypes above were remarkably different, and in overall point of view, C1 and C3 had relatively good prognosis, while C2 subtype had relatively poor prognosis (Fig. 1C). In addition, we investigated the clinicopathologic characteristics across various subtypes in TCGA–LUAD and discovered that the C2 subtype exhibited a higher clinical grade in comparison with the other LUAD molecular subtypes (Fig. 1D).

Fig. 1.

Fig. 1

Open in a new tab

Molecular typing of LUAD based on IRGs. A Boxplot of ssGSEA based on IRGs in TCGA–LUAD. ****p < 0.0001. B PCA analysis of TCGA–LUAD subtypes based on IRGs. C KM curves of overall survival differences of the three LUAD molecular subtypes. D Clinicopathological features of the three molecular subtypes of LUAD in TCGA–LUAD

Characterization of pathways between LUAD molecular subtypes

Enrichment analysis of differentially activated pathways among molecular subtypes showed that pathways related to glycolysis, angiogenesis, epithelial mesenchymal transition were significantly activated in the C2 subtype, whereas pathways related to fatty acid metabolism, bile acid metabolism, and other pathways were significantly activated in the C1 and C3 subtypes (Fig. 2A). Our subsequent calculation of the presence of differentially expressed genes (DEGs) between C1 vs. other, C2 vs. other, and C3 vs. other subtypes revealed 239 genes differentially expressed between molecular subtypes (Fig. 2B). Enrichment analysis of these genes showed significant enrichment in Focal adhesion and ECM–receptor interaction pathways (Fig. 2C). Meanwhile, we revealed the distribution location of these genes on chromosomes (Fig. 2D).

Fig. 2.

Fig. 2

Open in a new tab

Differential expression analysis between LUAD molecular isoforms and their enrichment pathways. A GSEA analysis of TCGA–LUAD. B Results of differential expression analysis between molecular isoforms in TCGA–LUAD. C Differentially expressed genes between molecular isoforms were subjected to KEGG functional enrichment analysis. D Positions of the differentially expressed genes in different molecular subtypes on the chromosome and log2FC values in the corresponding clusters

Development and inspect of a prognostic assessment equation for LUAD

To construct biomarkers for assessing the prognosis of LUAD, this study further performed Univariate COX regression on DEGs and further compressed these genes remarkably corresponding to prognosis by lasso operation, and finally screened seven genes as signatures affecting the prognosis of LUAD, which were MS4A1, EXO1, CPS1, ZNF750, S100P, NT5E, and KCNN4 (Fig. 3A–C). A risk assessment model for LUAD was then established using the formula: RiskScore = -0.128*MS4A1 + 0.215*EXO1 + 0.054*CPS1—0.104*ZNF750 + 0.064*S100P + 0.166*NT5E + 0.101*KCNN4. Next, we categorized the LUAD samples into high and low RiskScore groups. Survival analysis results indicated that in TCGA–LUAD, the prognosis for LUAD samples in the high RiskScore group was significantly poorer than the low RiskScore group. Furthermore, the model demonstrated satisfactory accuracy in predicting 1–5-year survival outcomes, with all area under the curve (AUC) values exceeding 0.6 (Fig. 3D). To ensure the reliability of the IRG-based clinical prognostic model predictions, we validated it using the GEO validation set GSE31210. Consistent results in the TCGA–LUAD data set were obtained, confirming the robustness of the model (Fig. 3E).

Fig. 3.

Fig. 3

Open in a new tab

Construction of LUAD prognostic model and validation. A Trajectory of each independent variable with lambda. B Confidence intervals under lambda. C Multivariate forest plot of prognostic model genes. D From left to right are: the RiskScore, time of survival and status of survival in TCGA–LUAD data set, ROC curve evaluating the performance of the RiskScore model, and KM survival curves of TCGA–LUAD clinical prognostic model. E From left to right are: RiskScore, time of survival vs. status of survival, ROC curve evaluating the performance of the RiskScore model, and KM survival curves distribution in the GSE31210 data set

Regulatory role of key genes related to malignant phenotype of LUAD cell lines

This study further investigated the regulatory roles of the key genes in LUAD cell lines. The molecular assays demonstrated that 6 key genes were noticeably expression dysregulation in LUAD cell lines, among which EXO1 was not only up-regulated at a significant expression level, but also closely linked with the prognostic outcomes of LUAD in the existing studies, and thus became the focus of the subsequent study (Fig. 4A). Based on the results of the CCK-8 assay, we found that silencing EXO1 led to a decrease in the proliferative capacity of A549 cells (Fig. 4B). Subsequently, we also observed that silencing EXO1 significantly reduced the migration and invasion levels of A549 cells (Fig. 4C, D). The above results supported the regulatory roles of EXO1 in the malignant phenotype of LUAD and confirmed the potential regulatory role of these key genes on LUAD progression.

Fig. 4.

Fig. 4

Open in a new tab

Regulation of the malignant phenotypes in LUAD cell line by prognostic genes. A Relative expressions of MS4A1, EXO1, CPS1, ZNF750, S100P, NT5E and KCNN4 in LUAD cell line were measured by qRT-PCR. B CCK-8 assay evaluated the effect of EXO1 gene silencing on the proliferation capacity of A549 cells. C Wound healing assays evaluated the effect of EXO1 gene silencing on the migration capacity of A549 cells. D Transwell assay was used to evaluate the effect of EXO1 gene silencing on the invasive capacity of A549 cells. All procedures were performed in triplicate (n = 3). ns p > 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

RiskScore-based LUAD nomogram construction

Univariate and multivariate Cox regression analysis on the TCGA–LUAD data set showed that RiskScore was the most significant prognostic factor independent of other factors, such as Stage, pathologic_T, pathologic_N, pathologic_M, age, and gender (Fig. 5A, B). To evaluate the survival and risk assessment for TCGA–LUAD, we developed a nomogramintegrating RiskScore with other clinicopathologic characteristics. RiskScore showed the most significant impact on survival prediction (Fig. 5C). The calibration curve demonstrated the high accuracy of RiskScore in forecasting patient survival over 1–5 years. In addition, the decision curve analysis supported these findings, confirming the strong predictive performance of the nomogram (Fig. 5D).

Fig. 5.

Fig. 5

Open in a new tab

Development of nomogram and validation. A Results of univariate Cox regression analysis of LUAD-related clinical indicators; B Results of multivariate Cox regression analysis of LUAD-related clinical indicators; C RiskScore combining clinical characteristics to create a nomogram; D calibration and decision curves for the nomogram

The regulation of the LUAD immune microenvironment and chemotherapeutic drug sensitivity by signatures

The correlation between immune microenvironment and RiskScore, this study employed MCP-counter and TIMER to evaluate the infiltration of different stromal and immune cells. We observed significant differences in cell infiltration between the high- and low-risk groups, including dendritic cells (DC), T cells, myeloid dendritic cells, CD8 + T cells, CD4 + T cells, B lineage cells, and fibroblasts, among others (Fig. 6A, C). The infiltration of these cells was notably negatively correlated with RiskScore, indicating that immune cell activity may be markedly suppressed during LUAD progression (Fig. 6B, D). In addition, our analysis of immunotherapy responses across different molecular subtypes revealed that compared to the low-risk group, the high-risk group had markedly higher TIDE score, suggesting less immunotherapy benefit to high-risk patients (Fig. 6E). In addition, Riskscore was closely positively linked to the genes related to immune checkpoint (CD276, TNFSF4, and TNFSF9) (Fig. 6F). In addition to the in-depth analysis of the immune microenvironment, this study further explored the correlation between LUAD drug sensitivity and prognostic model, and we identified a significant correlation between 22 drugs and RiskScore (FDR < 0.05 and |cor| > 0.3) (Fig. 6G). The above results revealed a regulatory correlation between the immune microenvironment and RiskScore, suggesting that model signature genes may influence LUAD progression by mediating cancer immune escape. In addition, the drugs tapped in this study are expected to alleviate LUAD progression through immunosuppressive effects.

Fig. 6.

Fig. 6

Open in a new tab

Correlation analysis of RiskScore with LUAD immune microenvironment and chemotherapy resistance. A Differences in TCGA–LUAD immune cell MCP-Counter scores between low- and high-risk groups. B Association of immune infiltration levels obtained by RiskScore and MCP-Counter analysis. C Differences in TCGA–LUAD immune cell TIMER scores between high- and low-risk groups. D Correlation of immune infiltration levels obtained by RiskScore and TIMER analysis to get correlation of immune infiltration level. E Results of TIDE prediction in TCGA–LUAD. F Correlation of immune checkpoints and RiskScore. G Correlation analysis of RiskScore with drugs' IC50 in TCGA–LUAD

Pathway characterization and genomic landscape differences between low- and high-RiskScore LUAD groups

KEGG enrichment analysis on the differential genes in the high- and low-risk groups of TCGA–LUAD data set showed that the high-risk LUAD patient group, pathways closely related to cell proliferation showed significant activation, especially the DNA replication and cell cycle pathways were significantly activated (Fig. 7A). Further TMB analysis between the two risk groups revealed a remarkably higher TMB in the high RiskScore group than the low RiskScore group (Fig. 7B). The mutation profiles of the high- and low-risk groups showed that the mutation frequencies were not consistent between the groups as a whole, specifically, the high-risk group showed notably higher frequency of mutations in comparison with the low-risk group (Fig. 7C).

Fig. 7.

Fig. 7

Open in a new tab

Pathway characterization and genomic landscape of prognostic model. A GSEA analysis between TCGA–LUAD high- and low-risk subgroups. B Differences in the number of mutations between TCGA–LUAD high- and low-risk groups. C Gene expression status of high-frequency mutations top 20 in TCGA–LUAD high- and low-risk groups

Discussion

LUAD has a high degree of malignancy, poor biological behavior, and poor prognosis, and it is so needed to dig effective biomarkers for prognostic determination. Relevant reports found that molecules related to immune regulation in cancer have a very important role to play in understanding the prognostic mechanisms of patients with tumors lung cancer, for example, FOXM1, a gene related to the heterogeneity of tumor infiltration microenvironment, predicting the poor prognostic outcomes of small cell carcinoma of the lung [20]. In terms of regulatory mechanisms, FOXM1 expression in lung tissues influences immune escape from lung cancer by mediating macrophage polarization and activating the tumor immune pathway p38/MAPK [21]. Another study also screened lung cancer patients for prognostic markers through genes associated with the cancer immune microenvironment, in which genes such as IRF1 and STAP1 were shown to be associated with human leukocyte antigen presentation and B-cell activation, and to affect the progression of many types of cancers, including lung cancer, by mediating the patient's immune response [22]. This also provided a reference for this study, which used LUAD immune-related genes as a starting point, identified LUAD immune subtypes by consensus clustering, and explained the different features in immunogen between subtypes, and combined with multiple bioinformatics tools or methods to mine and build a LUAD prognostic module. This study provides novel molecular biomarkers and models for the immunophenotyping and prognostic assessment of LUAD, and offers a direction for exploring its immunotherapy strategies.

In this study, seven LUAD and immunomodulation-related genes were mined to develop a prognosis model, in which MS4A1, ZNF750 were prognostic protective factors, while EXO1, CPS1, S100P, NT5E, KCNN4 were prognostic risk factors. MS4A1 exists as a prognostic predictive marker for lung adenocarcinoma in the existing studies, but its dysregulated expression is usually mediated by lymphocytes that in turn regulate the progression of MS4A1 [23]. ZNF750, a zinc finger protein, has been demonstrated in existing studies to be usually significantly negatively correlated with cellular immunogenicity in squamous cell carcinoma of the lung, which can target and inhibit the expression of the tendon protein TNC to regulate the malignant behaviors of tumor cells, laying the groundwork for the improvement of the efficacy of immunotherapy in lung cancer [24]. EXO1 proves a potential marker for cancer progression that affects the abundance of immune cells and their types in LUAD, specifically, upregulated expression of this gene in LUAD was strongly linked to reduced dendritic cells, B cells, and CD4 + T cells [25]. CSP1 is commonly upregulated in LUAD samples, and may be considered as a LUAD diagnostic and prognostic biomarker, and studies have demonstrated that the gene upregulation is highly correlated with high infiltration of cancer-associated fibroblasts and expressions of immune checkpoint molecules, for instance, PD-L1 [26, 27]. S100P interacts with integrin α7, which increases the migratory and invasive phenotypes of lung cancer cells, whereas knockdown of S100P inhibits cancer cell migration and invasion, and results in EMT in highly invasive cancer cells reversal [28]. NT5E interacts with microRNAs in cancer tissues, and the combined effects of the two mediate malignant progression of lung cancer by promoting non-small cell lung cancer cell growth, proliferation, and migratory activity and inhibiting apoptosis levels [29]. KCNN4, on the other hand, regulates lung adenocarcinoma through the P13K/AKT and MEK/ERK signaling pathway progression, and proliferation and metastasis of lung adenocarcinoma cells were also significantly inhibited using inhibitors targeting AKT and ERK [30]. The AKT and ERK pathways have regulatory activity on the cancer immune microenvironment, which mediates immune checkpoint-associated responses to promote cancer progression [31]. In summary, the molecules tapped in this study that are responding to the prognosis of LUAD are usually related to the malignant phenotype of tumor cell lines or the regulation of immune factors, and are expected to serve as potential targets for the treatment of LUAD.

This study found significant differential level of the immune checkpoint molecules CD276, TNFSF4, and TNFSF9 between high- and low-risk groups. CD276 (B7–H3) is one of the molecules of the B7 superfamily, and proteins encoded by this gene belong to the immunoglobulin family, which prove certain roles in mediating of the T-cell-mediated immunity response. CD276 has been suggested to act as a stimulatory protein for T cells, but many studies have shown that CD276 acts as a suppressor of T cells and promotes tumor invasion and proliferation, and that CD276 is associated with adhesion between cells and between cells and extracellular matrix [32]. Some certain reports have illustrated that this gene seems a target for immunotherapy and can influence immune escape by mediating squamous cell carcinoma heterogeneity [33]. As a member of the tumor necrosis factor superfamily, TNFSF4 shows upward trend in lung fibroblasts in conditions of stress, and reports illustrated that serum and tissue levels of TNFSF4 are higher in patients with LUAD who have received chemotherapy, promoting chemotherapy resistance in cancer cells by inhibiting their apoptosis [34]. In addition, TNFSF4 significantly increased the reactivity of NF-κB/BCL-XL pathway in LUAD cell line, which influences the anti-tumor immune response process through the regulation of macrophage and T-cell activity [35]. TNFSF9, on the other hand, induces M2 polarization phenotype of macrophages through Src/FAK/p-Akt/IL-1β pathway, which in turn affects cancer cell metastatic phenotype, and thus this gene is often present as a potential target for cancer immunotherapy in existing studies [36]. In addition, it has also been shown that the up-regulated trend of this molecule proves associated with certain infiltration level of CD8 + T cells, which can be used as a reference to uncover clear infiltration phenotypes of miscellaneous kinds of immune cells in cancer immune microenvironment [37]. The role of these immune checkpoint-related genes in mediating immunosuppression in LUAD tumors has been confirmed, and provides a basis for applying the potential biomarkers revealed in our task to clinical immunotherapy research.

However, this study has some limitations. First, this study relies entirely on retrospective data from public databases, which may introduce batch effects and selection bias, and lacks independent validation from prospective clinical cohorts. In the future, we will collect multi-center, prospective clinical sample cohorts and conduct independent validation using techniques such as RNA sequencing to enhance the model's generalizability and clinical applicability. Second, the cell-based experiments focused solely on the migration and invasion functions of the EXO1 gene, failing to comprehensively validate the specific roles of other selected key genes in the malignant phenotype of LUAD. Furthermore, there is a lack of in vivo animal experimental evidence. Future studies will utilize mouse xenograft models to systematically evaluate the roles of all characteristic genes in tumor growth, metastasis, and the remodeling of the immune microenvironment in vivo, thereby elucidating their complete biological functions. Finally, this study identified associations between RiskScore and immune cell infiltration as well as checkpoint expression. However, we were unable to experimentally elucidate how these characteristic genes directly regulate immune cell function and immune escape within the tumor microenvironment. To address this, we will subsequently employ in vitro co-culture systems combined with flow cytometry and other techniques to investigate the direct effects of the selected key genes on immune cell function. This will enable us to decipher the specific molecular pathways through which they mediate immune suppression.

Conclusion

This study defines three immune-derived molecular subtypes of LUAD and innovatively constructs a potent seven-gene prognostic signature. This signature serves as an independent prognostic indicator and demonstrates strong clinical relevance by correlating with immune checkpoint expression and chemotherapy sensitivity. Functional validation confirms that the key gene, EXO1, promotes tumor migration and invasion. Our findings provide a valuable framework for risk stratification and pave the way for developing tailored immunotherapeutic and chemotherapeutic strategies for LUAD patients.

Acknowledgements

None

Abbreviations

LUAD

Lung adenocarcinoma

IRGs

Immune-related gene sets

UCSC-Xena

University Of Cingifornia Sisha Cruz-Xena

TCGA

The cancer genome atlas

DEGs

Differential expression genes

LASSO

Least absolute shrinkage and selection operator

ROC

Receiver operating characteristic

AUC

Area under the curve

ssGSEA

Single-sample gene set enrichment analysis

Author contributions

All authors contributed to this present work: [HZL] and [GML] designed the study. [XG] and [CDC] collected data. [XJC] analyzed data. [JS] drafted the manuscript. [ZFC] reviewed and revised the manuscript. The manuscript has been approved by all authors for publication.

Funding

This study was supported by Special Project on Traditional Chinese Medicine Research in Henan Province (2024ZY2171) and the Science and Technology Research Program of Henan Province (252102310236) with Hongzhi Li.

Availability of data and material

The datasets generated and/or analyzed during the current study are available in the [GSE31210] repository, [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE31210].

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Contributor Information

Hongzhi Li, Email: lihongzhi9908@163.com.

Guangming Li, Email: lgm177@sina.com.

References

  • 1.Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71(3):209–49. [DOI] [PubMed] [Google Scholar]
  • 2.Liu B, Yang C, Liu J, Peng M, Mao J, Tang S, et al. Targeting of lysosomes as a therapeutic target in cancer. Curr Mol Pharmacol. 2025. 10.2174/0118761429354659250320051057. [DOI] [PubMed] [Google Scholar]
  • 3.Herbst RS, Morgensztern D, Boshoff C. The biology and management of non-small cell lung cancer. Nature. 2018;553(7689):446–54. [DOI] [PubMed] [Google Scholar]
  • 4.Chen D, Wang R, Yu C, Cao F, Zhang X, Yan F, et al. FOX-A1 contributes to acquisition of chemoresistance in human lung adenocarcinoma via transactivation of SOX5. EBioMedicine. 2019;44:150–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Song Z, Yu J, Wang M, Shen W, Wang C, Lu T, et al. CHDTEPDB: transcriptome expression profile database and interactive analysis platform for congenital heart disease. Congenit Heart Dis. 2023. 10.32604/chd.2024.048081. [Google Scholar]
  • 6.Song Y, Yan S, Fan W, Zhang M, Liu W, Lu H, et al. Identification and validation of the immune subtypes of lung adenocarcinoma: implications for immunotherapy. Front Cell Dev Biol. 2020;8:550. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Hanahan D, Coussens LM. Accessories to the crime: functions of cells recruited to the tumor microenvironment. Cancer Cell. 2012;21(3):309–22. [DOI] [PubMed] [Google Scholar]
  • 8.Kang HJ, Oh JH, Chun SM, Kim D, Ryu YM, Hwang HS, et al. Immunogenomic landscape of hepatocellular carcinoma with immune cell stroma and EBV-positive tumor-infiltrating lymphocytes. J Hepatol. 2019;71(1):91–103. 10.1016/j.jhep.2019.03.018. [DOI] [PubMed] [Google Scholar]
  • 9.Fridman WH, Pages F, Sautes-Fridman C, Galon J. The immune contexture in human tumours: impact on clinical outcome. Nat Rev Cancer. 2012;12(4):298–306. [DOI] [PubMed] [Google Scholar]
  • 10.Xu ZY, Zhao M, Chen W, Li K, Qin F, Xiang WW, et al. Analysis of prognostic genes in the tumor microenvironment of lung adenocarcinoma. PeerJ. 2020;8:e9530. 10.7717/peerj.9530. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Liu XS, Zhou LM, Yuan LL, Gao Y, Kui XY, Liu XY, et al. NPM1 is a prognostic biomarker involved in immune infiltration of lung adenocarcinoma and associated with m6A modification and glycolysis. Front Immunol. 2021;12:724741. 10.3389/fimmu.2021.724741. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Lin Q, Ma Y, Chen P. Identification of potential biomarkers in follicular thyroid carcinoma: bioinformatics and immunohistochemical analyses. Oncologie. 2024;26(2):311–22. [Google Scholar]
  • 13.Bi KW, Wei XG, Qin XX, Li B. BTK has potential to be a prognostic factor for lung adenocarcinoma and an indicator for tumor microenvironment remodeling: a study based on TCGA data mining. Front Oncol. 2020;10:424. 10.3389/fonc.2020.00424. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Wilkerson MD, Hayes DN. Consensusclusterplus: a class discovery tool with confidence assessments and item tracking. Bioinformatics. 2010;26(12):1572–3. 10.1093/bioinformatics/btq170. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Barbie DA, Tamayo P, Boehm JS, Kim SY, Moody SE, Dunn IF, et al. Systematic RNA interference reveals that oncogenic KRAS-driven cancers require TBK1. Nature. 2009;462(7269):108–12. 10.1038/nature08460. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Jiang T, Zheng J, Li N, Li X, He J, Zhou J, Sun B, Chi Q. Dissecting the mechanisms of intestinal immune homeostasis by analyzing T-cell immune response in Crohn's disease and colorectal cancer. Current Gene Therapy. 2024;24(5):422–40. [DOI] [PubMed]
  • 17.Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. Limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43(7):e47. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Jiang P, Gu S, Pan D, Fu J, Sahu A, Hu X, et al. Signatures of T cell dysfunction and exclusion predict cancer immunotherapy response. Nat Med. 2018;24(10):1550–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Zhang L, Yang H, Liu J, Wang K, Cai X, Xiao W, Wang L, Wang M, Zhang C, Zhang J. Metabolomics-based approach to analyze the therapeutic targets and metabolites of a synovitis ointment for knee osteoarthritis. Curr Pharmaceutical Anal. 2023;19(3):222–34.
  • 20.Liang SK, Hsu CC, Song HL, Huang YC, Kuo CW, Yao X, et al. Correction: FOXM1 is required for small cell lung cancer tumorigenesis and associated with poor clinical prognosis. Oncogene. 2021;40(49):6705. [DOI] [PubMed] [Google Scholar]
  • 21.Goda C, Balli D, Black M, Milewski D, Le T, Ustiyan V, et al. Loss of FOXM1 in macrophages promotes pulmonary fibrosis by activating p38 MAPK signaling pathway. PLoS Genet. 2020;16(4):e1008692. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Wu J, Li L, Zhang H, Zhao Y, Zhang H, Wu S, et al. A risk model developed based on tumor microenvironment predicts overall survival and associates with tumor immunity of patients with lung adenocarcinoma. Oncogene. 2021;40(26):4413–24. [DOI] [PubMed] [Google Scholar]
  • 23.Wright CM, Savarimuthu Francis SM, Tan ME, Martins MU, Winterford C, Davidson MR, et al. MS4A1 dysregulation in asbestos-related lung squamous cell carcinoma is due to CD20 stromal lymphocyte expression. PLoS ONE. 2012;7(4):e34943. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Xia L, Lin H, Cao H, Lian J. Tenascin C as a novel zinc finger protein 750 target regulating the immunogenicity via DNA damage in lung squamous cell carcinoma. BMC Cancer. 2024;24(1):561. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Zhou CS, Feng MT, Chen X, Gao Y, Chen L, Li LD, et al. Exonuclease 1 (EXO1) is a potential prognostic biomarker and correlates with immune infiltrates in lung adenocarcinoma. Onco Targets Ther. 2021;14:1033–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Luo Y, Zhang S, Xie H, Su Q, He S, Lei Z. Prognosis and immunotherapy significances of a cancer-associated fibroblasts-related gene signature in lung adenocarcinoma. Cell Mol Biol (Noisy-le-grand). 2023;69(14):51–61. [DOI] [PubMed] [Google Scholar]
  • 27.Wu G, Zhao Z, Yan Y, Zhou Y, Wei J, Chen X, et al. CPS1 expression and its prognostic significance in lung adenocarcinoma. Ann Transl Med. 2020;8(6):341. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Hsu YL, Hung JY, Liang YY, Lin YS, Tsai MJ, Chou SH, et al. S100P interacts with integrin alpha7 and increases cancer cell migration and invasion in lung cancer. Oncotarget. 2015;6(30):29585–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Dong L, Zheng J, Gao Y, Zhou X, Song W, Huang J. The circular RNA NT5E promotes non-small cell lung cancer cell growth via sponging microRNA-134. Aging (Albany NY). 2020;12(4):3936–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Xu P, Mo X, Xia R, Jiang L, Zhang C, Xu H, et al. KCNN4 promotes the progression of lung adenocarcinoma by activating the AKT and ERK signaling pathways. Cancer Biomark. 2021;31(2):187–201. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Zhang Z, Richmond A, Yan C. Immunomodulatory properties of PI3K/AKT/mTOR and MAPK/MEK/ERK inhibition augment response to immune checkpoint blockade in melanoma and triple-negative breast cancer. Int J Mol Sci. 2022. 10.3390/ijms23137353. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Zhou WT, Jin WL. B7-H3/CD276: an emerging cancer immunotherapy. Front Immunol. 2021;12:701006. 10.3389/fimmu.2021.701006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Wang C, Li Y, Jia L, Kim JK, Li J, Deng P, et al. CD276 expression enables squamous cell carcinoma stem cells to evade immune surveillance. Cell Stem Cell. 2021;28(9):1597-1613 e1597. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Li Y, Chen Y, Miao L, Wang Y, Yu M, Yan X, et al. Stress-induced upregulation of TNFSF4 in cancer-associated fibroblast facilitates chemoresistance of lung adenocarcinoma through inhibiting apoptosis of tumor cells. Cancer Lett. 2021;497:212–20. [DOI] [PubMed] [Google Scholar]
  • 35.Sun SC. The non-canonical NF-kappaB pathway in immunity and inflammation. Nat Rev Immunol. 2017;17(9):545–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Wu J, Wang Y, Yang Y, Liu F, Jiang Z, Jiang Z. TNFSF9 promotes metastasis of pancreatic cancer by regulating M2 polarization of macrophages through Src/FAK/p-Akt/IL-1beta signaling. Int Immunopharmacol. 2022;102:108429. [DOI] [PubMed] [Google Scholar]
  • 37.Wu J, Wang Y, Jiang Z. TNFSF9 is a prognostic biomarker and correlated with immune infiltrates in pancreatic cancer. J Gastrointest Cancer. 2021;52(1):150–9. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The datasets generated and/or analyzed during the current study are available in the [GSE31210] repository, [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE31210].

Articles from European Journal of Medical Research are provided here courtesy of BMC

RESOURCES