A T-cell exhaustion-related signature for predicting response to immunotherapy among patients with lung adenocarcinoma

Introduction

Lung cancer is broadly categorized into two primary histological subtypes: non-small cell lung cancer (NSCLC) and small-cell lung cancer (SCLC). In 2022, lung cancer was the most frequently diagnosed malignancy and continues to be the predominant driver of cancer-associated mortality, accounting for approximately 18.7% of all cancer-related deaths (1). Although advances in therapy have markedly extended both the progression-free survival (PFS) and overall survival (OS) of patients with lung cancer, this disease remains a leading cause of global morbidity and mortality. Critically, most cases of advanced NSCLC ultimately develop resistance to treatment regimens—particularly immunotherapy and targeted agents—leading to disease progression (2). Recent research indicates that reversing T-cell exhaustion may amplify the efficacy of tumor immunotherapy. For instance, Zhou et al. reported that N-acetylcysteine impedes the differentiation of T cells into a terminally exhausted state, thereby potentiating the response to PD-1 blockade in colorectal cancer model (3). Similarly, Cheng et al. identified ETV7 as a promoter of T-cell exhaustion that accelerates tumorigenesis (4). Furthermore, Zhu et al. demonstrated that KLF2 plays a key role in preserving T-cell effector capacity while suppressing exhaustion, supporting it as a novel target for immunotherapeutic intervention (5). Consequently, mitigating T-cell exhaustion to restore functional effector activity is being recognized as a priority for optimizing outcomes among patients treated with immunotherapy.

In 1998, a state of T-cell suppression was first identified in a mouse model of chronic lymphocytic choriomeningitis virus infection (6), and this immunological process is now known as T-cell exhaustion. When under chronic infection and antitumor immunity, T cells are exposed to persistent antigenic stimulation which leads to the progressive loss of effector functions, the expression of multiple inhibitory receptors, dysregulated metabolism, poor memory recall response, and homeostatic proliferation (7). A number of mechanisms can cause T-cell exhaustion, and several studies have suggested that the transcription factor network exerts significant regulatory functions in this regard. For example, Martinez et al. demonstrated that NFAT controls CD8⁺ T-cell exhaustion by binding directly to regulatory regions of many exhaustion-associated genes, including the Pdcd1 (PD-1) and Havcr2 (TIM3) promoters and cis-regulatory regions (8). Another study found that T-cell exhaustion during chronic infection is driven by a high abundance of T-cell receptor-induced transcription factors IRF4, BATF, and NFATc1, which promote the expression of PD-1 and mediate impaired cellular metabolism (9). Moreover, it was reported that NFAT can induce TOX, TOX2, and NR4A transcription factor family proteins, all of which upregulate the expression of inhibitory receptors, with a positive feedback regulation existing between NR4A and TOX (10). Other research demonstrated that BATF can directly bind to regulatory regions near Tbx21 and Klf2, modulating their enhancer accessibility to facilitate the development of CD8⁺ T cells during chronic viral infection (11). One study found that other transcription factors, such as T-bet, can directly compete with eomesodermin at gene promoters, such as Pdcd1, to promote T-cell homing and activation during PD-1 blockade (12). These findings suggest that the regulation exerted by transcription factors can be applied clinically and clarify, to some extent, the nature of the transcription factor network; however, further research is necessary to refine the use of these transcription factors and guide the next generation of treatment.

Few studies have investigated transcription factors previously linked to T-cell exhaustion in LUAD, particularly their associations with the immune microenvironment and treatment response. We therefore evaluated 10 transcription factors reported in the literature (8-10,12-16)—IRF4, BATF, NFATC1, TCF7, NR4A1, NR4A2, NR4A3, TBX21, EOMES, and TOX—and examined their prognostic and immune-microenvironment associations in bulk LUAD transcriptomic datasets. We further developed a prognostic score and evaluated its association with computational estimates of immunotherapy response. Because these analyses are associative, they do not establish a direct exhausted-T-cell phenotype or causal function for the selected transcription factors in LUAD. We present this article in accordance with the TRIPOD reporting checklist (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2026-1101/rc).

Methods

Collection of data on patients with LUAD

Transcription factors previously linked to T-cell exhaustion were selected from the relevant literature. The Cancer Genome Atlas LUAD (TCGA-LUAD) dataset (https://portal.gdc.cancer.gov) served as the training cohort; transcriptomic data reported as fragments per kilobase of transcript per million mapped reads, clinical data, somatic mutation data, and proteomic data were used. Six datasets with follow-up information (GSE72094, GSE31210, GSE68465, GSE26939, GSE42127, and GSE37745) were obtained from the Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) and analyzed as independent validation cohorts rather than pooled into a single expression matrix. Cohort-specific gene identifiers were mapped to official gene symbols, and overlapping genes were retained for cross-cohort evaluation. GSE43458, GSE32863, GSE75037, and GSE40791, which contained adjacent non-tumor and LUAD tissues, were used to evaluate expression of the signature genes. For GEO datasets, the data were downloaded as Series Matrix Files. These datasets were generated based on the Expression profiling by array platform. Unnormalized datasets were normalized using log₂(X+1) transformation, while already normalized datasets were used without further processing.

Identification of T-cell exhaustion-related transcription factors

First, we used the R package RCircos to display the genomic locations of transcription factors previously linked to T-cell exhaustion. A correlation network was then generated using the training-set expression data. Univariate Cox regression was used to evaluate associations between transcription-factor expression and prognosis, and forest plots were generated using the reshape2 (v1.4.4), survival (v3.2-13), and survminer (v0.4.9) R packages. These analyses assess statistical association and do not demonstrate cell-specific function or causality.

T-cell subpopulation and exhaustion analysis

T cells were extracted from the integrated scRNA-seq dataset and subjected to secondary clustering using Seurat. After normalization, identification of highly variable genes, scaling, and PCA, the first 20 principal components were used for UMAP visualization and unsupervised clustering. T-cell subsets were annotated based on differentially expressed genes and canonical markers. Exhausted T cells (Tex) were subsequently extracted based on the expression of established exhaustion markers. The expression and co-expression patterns of exhaustion-related transcription factors (IRF4, NFATC1, EOMES, and TOX) and canonical exhaustion markers (PDCD1, CTLA-4, LAG3, and HAVCR2) were evaluated, and their associations were assessed using Spearman correlation analysis.

Consensus clustering analysis

By using the ConsensusClusterPlus (v1.54.0) and scatterplot3d (v0.3-41) R packages, we performed a consensus clustering analysis of the training group based on four transcription factor genes that regulate T-cell exhaustion and that were associated with prognosis. From this, we determined the optimal number of types and drew principal component analysis (PCA) maps. Next, a heatmap with clinical features was drawn via the pheatmap R package. Kaplan-Meier (KM) curves were used to analyze the differences in OS, PFS, and disease-specific survival (DSS) of patients with high and low T-cell exhaustion via the survival and survminer R packages.

Analysis of the TME

First, we obtained the immune score, stromal score, ESTIMATE score, and tumor purity score of the patients with LUAD in the training set using the ESTIMATE algorithm. The ESTIMATE score was calculated as the sum of the immune score and stromal score. To analyze the degree of immune cell infiltration in LUAD tumor tissues, seven algorithms from the Tumor Immune Estimation Resource (TIMER) database were used: TIMER, CIBERSORT, CIBERSORT-ABS, QUANTISEQ, MCPCONT, XCELL, and EPIC (http://timer.cistrome.org) (17-22). The estimate (v1.0.13), limma (v3.46.0), ggpubr (v0.4.0), reshape2, pheatmap (v1.0.12), scales (v1.4.0), ggplot2 (v3.3.5), ggtext (v0.1.2), and tidyverse (v1.3.1) R packages were used for these analyses.

Identification of differentially expressed genes (DEGs) between the different risk groups of T-cell exhaustion and functional enrichment analysis

Using |log fold change (FC)| >1 and false-discovery rate (FDR) <0.05 as filter conditions, we identified DEGs between different clusters of T-cell exhaustion and visualized them in a volcano plot using the limma R package. Next, the clusterProfiler (v4.0.5), org.Hs.eg.db (v3.13.0), enrichplot (v1.10.2), and ggplot2 R packages were used to conduct Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses, with P<0.05 and FDR <0.05 being the selection criteria.

Construction of a prognostic risk score associated with T-cell exhaustion

To identify DEGs reproducibly observed across cohorts, we intersected cohort-specific gene lists without merging the six GEO expression matrices and visualized the overlap with an UpSet plot, yielding 19 common DEGs. Univariate Cox regression was then used to evaluate their associations with prognosis. Subsequently, LASSO Cox regression with 10-fold cross-validation (nfolds =10, family = “cox”, standardize = TRUE) was used to establish the T-cell exhaustion-related prognostic score (TEXRGPS), with the optimal λ selected as lambda.min (λ=0.0108773), which was calculated as follows:

TEXRGPS=∑i=1ncoef(genei)∗expr(genei)[1]

We divided patients into high- and low-TEXRGPS groups based on the median TEXRGPS obtained. Risk curves, survival status plots, and risk heatmaps were drawn to visualize the distribution of the TEXRGPS, the relationship between the TEXRGPS and survival status, and the expression of T-cell exhaustion-related genes, respectively.

The KM curve was used for survival analysis, and the receiver operating characteristic (ROC) curve was used to verify the ability of TEXRGPS to predict patient survival. In addition, the correlation between the TEXRGPS and clinical information was analyzed via bar charts and boxplots. Following this, KM survival analysis was performed to assess the association between the expression levels of the six signature genes and patient survival. To ensure the suitability of the model, we performed a layered analysis. Finally, univariate and multivariate Cox regression analyses were conducted to determine whether the TEXRGPS was an independent prognostic factor. The above-mentioned analyses were completed via the UpSetR, survival, survminer, glmnet, ggpubr, ggExtra, reshape2, pheatmap, timeROC, limma, and ggplot2 R packages.

Validation of the TEXRGPS

To demonstrate the stability of the TEXRGPS, six data sets were selected for validation. Applicability was demonstrated by the KM curve, ROC curve, clinical correlation, and independent prognostic analyses. Next, the six genes that constituted the signature were validated in four data sets. We used the UALCAN tool (http://ualcan.path.uab.edu/analysis.html) to analyze the protein expression levels in normal tissues and LUAD specimens obtained from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) database. Finally, the Human Protein Atlas (HPA) database (https://www.proteinatlas.org/) was used to further verify the protein expression of the six genes in normal and LUAD tissues. Analyses of the validation set was conducted with the same R package as those used for the training set analyses.

Single-cell analysis

The GSE131907 single-cell dataset was analyzed via the Tumor Immune Single Cell Hub (TISCH; http://tisch.comp-genomics.org) database. The single-cell dataset included 208,506 single-cell transcriptomic profiles from 58 LUAD specimens. Importantly, the dataset included tumor, stromal, and immune cell information.

Correlation of the TEXRGPS with response to immunotherapy

The expression levels of multiple immune checkpoint genes in the high- and low-TEXRGPS groups were visualized via boxplots. Subsequently, the expression level of the PD-L1 (CD274) protein was determined in the two groups, and the immunophenotype score (IPS) was obtained from The Cancer Immune Atlas (TCIA; https://tcia.at/) database to predict the response to immunotherapy. In addition, the submap algorithm was used to further verify whether there was a difference in response to immunotherapy between the two groups. Finally, the results were validated in an independent immunotherapy cohort (IMvigor210). The above-mentioned analyses were completed via the limma, ggplot2, ggpubr, reshape2, pheatmap, survival, and survminer R packages.

Drug sensitivity analysis

To analyze the response of patients in the high- and low-TEXRGPS groups to specific antitumor agents, we accessed the Genomics of Drug Sensitivity in Cancer (GDSC) database. The half-maximal inhibitory concentration (IC₅₀) in each patient was calculated to predict sensitivity to certain agents. The pRRophetic (v0.5), ggplot2, limma, and ggpubr R packages were used to implement these analyses and draw related boxplots.

Gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA)

To investigate the different biological functions of genes in the high- and low-TEXRGPS groups, we conducted GSEA and GSVA. GSEA was performed with 1,000 permutations (nPerm =1,000) to ensure robust statistical inference. Moreover, heatmaps were used in the GSVA to visualize the differential pathways of the two groups. Analyses were conducted via the clusterProfiler, org.Hs.eg.db, enrichplot, and limma R packages.

Construction of the nomogram

By using the survival, survminer, timeROC (v0.2), regplot (v0.1), and rms (v6.2-0), we developed a nomogram based on patients’ clinical characteristics and the TEXRGPS to predict patients’ 3-, 4-, and 5-year OS probabilities. ROC and calibration curves were plotted for further validation.

Clinical specimens and reverse-transcription quantitative polymerase chain reaction (RT-qPCR)

Patient selection and study design

This was a single-center, retrospective study conducted at The Second Affiliated Hospital of Harbin Medical University. Twenty paired LUAD and adjacent non-tumor tissue specimens were collected from patients who underwent surgical resection between December 2022 and March 2023. Patients were included if they had histologically confirmed primary LUAD, underwent curative-intent surgical resection, had available paired tumor and adjacent non-tumor tissue specimens with adequate tumor cellularity (>70%) for RNA extraction, were treatment-naïve with no prior immunotherapy, chemotherapy, targeted therapy, or radiotherapy before surgery, and had complete clinical data. Patients were excluded if they had received neoadjuvant therapy, had a history of other malignancies within the past 5 years, had incomplete clinical data, had poor specimen quality (tumor cellularity <70%), or had recurrent or metastatic disease. Pathological staging was performed according to the American Joint Committee on Cancer (AJCC) 8th edition criteria. Histological subtypes were classified according to the 2021 World Health Organization (WHO) classification of lung tumors by two independent pathologists. The cohort comprised 20 patients (16 females, 4 males) with a median age of 59 years (range, 43-69 years). All patients were of Han Chinese ethnicity. Nineteen patients were non-smokers and 1 patient was a smoker. Pathological stages included stage IA1 (n=1), IA2 (n=11), IA3 (n=5), and IB (n=3). Histological subtypes included acinar (n=13), papillary (n=2), lepidic (n=2), solid (n=2), and micropapillary (n=1). A summary of patient characteristics is provided in Table S1. Detailed clinicopathological characteristics of these 20 patients are provided in Table S2. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of The Second Affiliated Hospital of Harbin Medical University (approval No. KY2023-139), and written informed consent was obtained from all participants.

RNA Extraction and RT-qPCR

Specimens were preserved in liquid nitrogen. Total RNA was extracted using TRIzol reagent (cat. no. SM129-02; Sevenbio, Beijing, China). Complementary DNA was synthesized using a reverse-transcription kit (cat. no. 1119ES60; Yeasen Biotechnology, Shanghai, China), and RT-qPCR was performed using SYBR Green Master Mix (cat. no. 11184ES03; Yeasen Biotechnology). GAPDH was used as the candidate reference gene, and relative target-gene expression was calculated using the 2^−ΔΔCt method. Primer sequences are provided in Table S3. The RT-qPCR raw Ct values are provided in Table S4.

Statistical analysis

All data in this study were analyzed with R software version 4.1.2. The nonparametric Wilcoxon test was used to determine the differences between groups, while the paired nonparametric Wilcoxon test was used to compare normal and tumor samples from the same patient. Spearman analysis was applied to evaluate correlation coefficients, and the chi-squared test was used to analyze categorical data. Survival data were compared via KM analysis and the log-rank test, while the ROC curve was used to evaluate diagnostic performance. The calibration curve was used to assess the agreement between the predicted probabilities and the observed outcome. Univariate and multivariate Cox regression analyses were used to analyze hazard ratios and identify whether the TEXRGPS was an independent risk factor. P<0.05 was considered to indicate a statistically significant difference.

Results

The overall workflow and analytical pipeline of this study are illustrated in Figure 1.

Figure 1 Flowchart. *, P<0.05; **, P<0.01; ***, P<0.001. Group A: TEXCluster A (high-TEXRGPS); Group B: TEXCluster B (low-TEXRGPS). Fourth row left panel (HPA database, CD79A): available at https://www.proteinatlas.org/ENSG00000105369-CD79A/cancer/lung+cancer#img; staining method: IHC staining with DAB chromogen and hematoxylin counterstaining; scale bar: 200 μm. The HPA image is used under the HPA licence (images under CC BY 4.0; database content under CC BY-SA 3.0). AUC, area under the curve; DC, dendritic cell; DEGs, differentially expressed genes; GEO, Gene Expression Omnibus; HPA, Human Protein Atlas; IHC, immunohistochemistry; LASSO, least absolute shrinkage and selection operator; LUAD, lung adenocarcinoma; PD-1, programmed cell death protein 1; ROC, receiver operating characteristic; RT-qPCR, reverse transcription quantitative polymerase chain reaction; TEX, T-cell exhaustion; TME, tumor microenvironment.

Transcription factor genes related to T-cell exhaustion and construction of clusters

Ten transcription factors previously linked to T-cell exhaustion were identified from published studies, and their genomic locations and expression correlations were examined (Figure 2A,2B). In univariate Cox analyses, IRF4, NFATC1, EOMES, and TOX were associated with overall survival and were used for consensus clustering (Figure 2C). Based on these four prognosis-associated genes, patients in TCGA-LUAD were stratified into two transcriptomic clusters (Figure 2D). These clusters represent expression-based groups and should not be interpreted as direct measurements of T-cell functional exhaustion. PCA and heatmap analyses showed differences between the clusters (Figure 2E,2F), and Kaplan-Meier analyses evaluated OS, PFS, and DSS (Figure 2G-2I). Five-year OS and PFS differed significantly between clusters, whereas DSS did not.

Figure 2 T-cell exhaustion-related transcription factors in the TCGA cohort. (A) Gene location of transcription factors that regulate T-cell exhaustion. (B) Gene correlation network; red indicates a positive correlation and blue a negative correlation. (C) Univariate Cox regression analysis and forest map. (D) Consensus cluster analysis of four prognostic transcription factors regulating T-cell exhaustion. (E) Principal component analysis. (F) Heatmap of clinical features. (G-I) KM curve comparing OS, PFS, and DSS between patients with high and low TEX. Group A: TEXCluster A (high-TEXRGPS); Group B: TEXCluster B (low-TEXRGPS). DSS, disease-specific survival; KM, Kaplan-Meier; PFS, progression-free survival; TCGA, The Cancer Genome Atlas; TEX, T-cell exhaustion.

Immune cell infiltration of the TME in LUAD

The ESTIMATE algorithm was used to calculate the immune score, stromal score, ESTIMATE score, and tumor purity score of the different risk groups (Figure 3A) from which a heatmap was drawn of the distribution of immune cells in LUAD (Figure 3B). A number of immune cells showed significant differences in infiltration across the risk groups. Most importantly, we found that many T-cell types, such as naive CD8⁺ T cells, central memory CD8⁺ T cells, and effector memory CD8⁺ T cell had a more significant degree of infiltration in the B cluster according to the results obtained from the XCELL algorithm.

Figure 3 Analysis of the TME in the TCGA cohort. (A) Immune score, stromal score, ESTIMATE score, and tumor purity score of patients with LUAD in the TCGA cohort. (B) Heatmap showing the immune cell infiltration in LUAD tissues. ***, P<0.001. Group A: TEXCluster A (high-TEXRGPS); Group B: TEXCluster B (low-TEXRGPS). LUAD, lung adenocarcinoma; TCGA, The Cancer Genome Atlas; TEX, T-cell exhaustion; TME, tumor microenvironment.

Identification of DEGs in TEXClusters

The DEGs between the two clusters were screened (|logFC| >1 and FDR <0.05) and are presented in Figure 4A. GO and KEGG pathway analyses indicated that these DEGs were related to multiple immune processes, including immunoglobulin production, production of molecular mediator of the immune response, antigen receptor-mediated signaling pathway, cytokine-cytokine receptor interaction, hematopoietic cell lineage, and viral interaction with cytokine and cytokine receptor (Figure 4B,4C).

Figure 4 Identification of the key DEGs and construction of the TEXRGPS. (A) Volcano plot of DEGs between the A and B clusters of T-cell exhaustion. (B,C) KEGG and GO analysis of DEGs. (D) The UpSet plot showing the 19 genes overlapping between the TCGA and the six test sets. (E) Univariate Cox regression analysis and forest plot showing the 13 genes of these 19 genes associated with prognosis. (F) LASSO coefficient profiles of 13 DEGs. (G) The tuning parameters were cross-validated in the LASSO model. BP, biological process; CC, cellular component; DEGs, differentially expressed genes; FC, fold change; GO, Gene Ontology; IgA, immunoglobulin A; IgG, immunoglobulin G; KEGG, Kyoto Encyclopedia of Genes and Genomes; LASSO, least absolute shrinkage and selection operator; MF, molecular function; TCGA, The Cancer Genome Atlas.

Construction of the risk score based on T-cell exhaustion

We selected 19 DEGs that overlapped in the six data sets (Figure 4D), 13 of which were found to correlate with survival according to univariate Cox regression analysis (Figure 4E). LASSO regression ultimately identified six eligible genes (Figure 4F,4G), including CCL19, CD79A, CHRDL1, MS4A1, S100P, and SFTPC that were incorporated into the TEXRGPS.

Based on the median TEXRGPS in the training cohort, patients with LUAD from the training (TCGA) and validation (GEO) cohorts were divided into low- and high-TEXRGPS groups, respectively. In the training cohort, the distribution of TEXRGPS was determined via the risk curve, and survival status was compared between the different groups (Figure 5A). We observed that S100P was highly expressed in the high-TEXRGPS group, while the other five genes were highly expressed in the low-TEXRGPS group; further analysis was applied to determine the association of the TEXRGPS with the OS, PFS, and DSS among patients with LUAD. It was found that the high-TEXRGPS group had a significantly worse prognosis (Figure 5B-5D). The area under the ROC curve (AUC) of the TEXRGPS in predicting 3-, 4-, and 5-year survival was 0.649, 0.642, and 0.650, respectively (Figure 5E). We additionally analyzed the correlation between clinical characteristics and the TEXRGPS. We found that patients in the high-TEXRGPS group had a higher proportion of males and patients with stage Ⅲ–Ⅳ disease, stage T3–4, disease, and stage N1–3 disease, while there was no difference in age or M stage (Figure 5F). The TEXRGPS was significantly higher among males and patients with advanced disease, lymph node metastasis, and distant metastasis (Figure 5G). Moreover, a higher expression of S100P was associated with lower survival probability, but this was not the case for the other five genes (Figure 5H).

Figure 5 The verification of the TEXRGPS in the TCGA. (A) Distribution of the TEXRGPS in the TCGA cohort. Survival status in the high- and low-TEXRGPS groups in the TCGA cohort. The heatmap of DEG distribution between the high- and low-TEXRGPS groups of the TCGA cohort. (B-E) The correlation between the TEXRGPS and survival status in the TCGA cohort in terms of OS, PFS, and DSS; AUC for 3-, 4-, and 5-year survival. (F,G) Bar chart and boxplots for the association between the TEXRGPS and clinical features in the TCGA cohort. (H) KM survival analysis of the six DEGs in the TEXRGPS. *, P<0.05; ***, P<0.001; ****, P<0.0001; ns, not significant. AUC, area under the curve; DEG, differentially expressed gene; DSS, disease-specific survival; KM, Kaplan-Meier; M, metastasis; N, node; OS, overall survival; PFS, progression-free survival; T, tumor; TCGA, The Cancer Genome Atlas.

A stratified analysis was conducted to ensure the robustness of the TEXRGPS (Figure 6A). Subsequent univariate and multivariate regression analyses demonstrated that the TEXRGPS and stage were independent prognostic factors for survival (Figure 6B,6C).

Figure 6 Validation of the TEXRGPS in the TCGA cohort. (A) Stratification analysis of the TEXRGPS for different clinical features, including age, TNM stage, and gender. (B,C) Univariate and multivariate Cox regression analyses of the TEXRGPS and clinical features. M, metastasis; N, node; T, tumor; TCGA, The Cancer Genome Atlas; TNM, tumor-node-metastasis.

Prognostic validation of the TEXRGPS in LUAD

To ensure the stability of TEXRGPS, we assessed it in the validation cohorts, and the results showed good consistency with the previous analyses (Figure 7, Figures S1,S2). We then examined the expression levels of the six genes that constitute the TEXRGPS. For CCL19, CD79A, and CHRDL1, the TCGA dataset revealed that CCL19 and CD79A were significantly upregulated, while CHRDL1 was significantly downregulated in tumor tissues compared to normal tissues (Figure 8A). These mRNA-level differences were further validated by RT-qPCR in 20 pairs of clinical samples from our hospital (Figure 8B). At the protein level, CPTAC database analysis showed that CD79A was upregulated and CHRDL1 downregulated in LUAD primary tumors compared to normal tissues, while CCL19 showed a borderline difference (P=0.06) (Figure 8C). Immunohistochemistry data from the HPA database corroborated these proteomic findings (Figure 8D). As for MS4A1, S100P, and SFTPC, the TCGA dataset demonstrated that MS4A1 and S100P were significantly upregulated, whereas SFTPC was downregulated in tumor tissues (Figure 9A). These results were also confirmed by RT-qPCR in our clinical samples (Figure 9B); however, the expression pattern of MS4A1 was not consistent in the GSE32863 and GSE75037 datasets (Figure S3). Protein-level analysis from CPTAC revealed that MS4A1 and S100P were upregulated and SFTPC was downregulated in primary tumors (Figure 9C). These findings were further supported by HPA immunohistochemistry data (Figure 9D). Collectively, these results supported differential expression of the six genes in LUAD. Some genes did not reach statistical significance in every validation cohort, which may reflect limited sample size, platform differences, and biological heterogeneity. To determine the differences in the infiltration of immune cells in the TME between the low- and high-TEXRGPS groups, 208,506 single-cell LUAD transcriptomic profiles from 58 specimens in the GSE131907 dataset were clustered and defined into 14 cell types, including B lymphocytes, CD4⁺ conventional T (CD4Tconv) cells, CD8⁺ T cells, exhausted CD8⁺ T (CD8Tex) cells, dendritic cells (DCs), endothelial cells, epithelial cells, fibroblasts, mast cells, monocytes/macrophages, oligodendrocytes, and plasma cells (Figure 10A). We observed that SFTPC had a significantly higher expression in several cell types, including CD8⁺ T cells and CD8Tex cells; CD79A and MS4A1 had a significantly higher expression in B lymphocytes and plasma cells; and S100P and CHRDL1 had a significantly higher expression in epithelial cells and fibroblasts, respectively (Figure 10B,10C). Results from immunohistochemistry and analysis of the TCGA further indicated that CD79A and IRF4 had a significantly higher expression in LUAD tissues, and there was a significant correlation between them (Figure S4A-S4C). Furthermore, the correlation between the TEXRGPS and immune cell distribution was determined (Figure 11).

Figure 7 Validation of the TEXRGPS in the GSE72094 dataset. (A) Distribution of the TEXRGPS and survival status. Heatmap of the six DEGs. (B) KM curve for the different TEXRGPS groups. (C) AUC for 3-, 4-, and 5-year survival. (D,E) Bar chart and boxplots of analysis for the association between the TEXRGPS and clinical features, including age, gender, and stage. (F,G) Univariate and multivariate Cox regression analyses of the TEXRGPS and clinical features. AUC, area under the curve; DEGs, differentially expressed genes; KM, Kaplan-Meier.

Figure 8 Validation of CCL19, CD79A, and CHRDL1 expression in lung adenocarcinoma. (A) Box plots showing the expression differences of CCL19, CD79A, and CHRDL1 between tumor and normal tissues in the TCGA dataset. (B) Comparison of the expression of the three characteristic genes between 20 pairs of tumor and normal tissues. (C) UALCAN analysis of the protein expression levels (Z-value) of these genes in normal and LUAD samples from the CPTAC database. (D) Representative images showing the protein expression of CCL19, CD79A, and CHRDL1 in normal and LUAD tissues from the HPA database: CCL19 (https://www.proteinatlas.org/ENSG00000172724-CCL19/cancer/lung+cancer#img), CD79A (https://www.proteinatlas.org/ENSG00000105369-CD79A/cancer/lung+cancer#img), and CHRDL1 (https://www.proteinatlas.org/ENSG00000101938-CHRDL1/cancer/lung+cancer#img). Staining method: IHC staining with DAB chromogen and hematoxylin counterstaining. Scale bars: 200 μm for CCL19, CD79A and CHRDL1. The HPA images are used under the HPA licence (images under CC BY 4.0; database content under CC BY-SA 3.0). *, P<0.05; **, P<0.01; ***, P<0.001. CPTAC, Clinical Proteomic Tumor Analysis Consortium; HPA, Human Protein Atlas; IHC, immunohistochemistry; LUAD, lung adenocarcinoma; TCGA, The Cancer Genome Atlas.

Figure 9 Validation of MS4A1, S100P, and SFTPC expression in lung adenocarcinoma. (A) Box plots showing the expression differences of MS4A1, S100P, and SFTPC between tumor and normal tissues in the TCGA dataset. (B) Comparison of the expression of the three characteristic genes between 20 pairs of tumor and normal tissues. (C) UALCAN analysis of the protein expression levels (Z-value) of these genes in normal and LUAD samples from the CPTAC database. (D) Representative images showing the protein expression of MS4A1, S100P, and SFTPC in normal and LUAD tissues from the HPA database: MS4A1 (https://www.proteinatlas.org/ENSG00000156738-MS4A1/cancer/lung+cancer#img), S100P (https://www.proteinatlas.org/ENSG00000163993-S100P/cancer/lung+cancer#img), and SFTPC (https://www.proteinatlas.org/ENSG00000168484-SFTPC/cancer/lung+cancer#img). Staining method: IHC staining with DAB chromogen and hematoxylin counterstaining. Scale bars: 200 μm for MS4A1, S100P, and SFTPC. The HPA images are used under the HPA licence (images under CC BY 4.0; database content under CC BY-SA 3.0). ***, P<0.001; ns, not significant. CPTAC, Clinical Proteomic Tumor Analysis Consortium; HPA, Human Protein Atlas; IHC, immunohistochemistry; LUAD, lung adenocarcinoma; TCGA, The Cancer Genome Atlas.

Figure 10 Single-cell analysis of the GSE131907 dataset from the TISCH database. (A) Distribution of immune cells and stromal cells in tumor tissues. (B,C) The expression of six signature genes in different cells. DC, dendritic cell; TISCH, Tumor Immune Single Cell Hub.

Figure 11 Immune microenvironment analysis. (A) Difference in immune score, stromal score, ESTIMATE score, and tumor purity between the high- and low-TEXRGPS groups. (B) Correlations between cell infiltration and TEXRGPS determined by seven different algorithms. ***, P<0.001.

Association of the TEXRGPS with response to antitumor therapy

First, we evaluated several immune-checkpoint genes and found that TIGIT, LAG3, HAVCR2, PDCD1, and CTLA-4 were more highly expressed in the low-TEXRGPS group than in the high-TEXRGPS group (Figure 12A). Proteomic data similarly showed higher PD-L1 expression in the low-TEXRGPS group (Figure 12B). The low-TEXRGPS group also had a higher immunophenotype score, suggesting greater predicted sensitivity to immune-checkpoint inhibition (Figure 12C). Subclass mapping indicated a stronger predicted response to anti-PD-1 therapy in the low-TEXRGPS group (Bonferroni-corrected P<0.02; Figure 12D). The association between TEXRGPS and immunotherapy outcome was further examined in IMvigor210, an independent urothelial-cancer immunotherapy cohort; because this cohort is not LUAD, it provides supportive cross-tumor rather than disease-specific validation (Figure 12E). Drug sensitivity was estimated using the GDSC database (Figure S5).

Figure 12 Immunotherapy-related analysis of the TEXRGPS in patients with LUAD. (A) Boxplots showing the expression levels of several immune checkpoint genes in the high- and low-TEXRGPS groups. (B) The expression level of PD-L1 (CD274) protein in the high- and low-TEXRGPS groups. (C) IPS was obtained from the TCIA in the high- and low-TEXRGPS groups. (D) Prediction of the immunotherapy effect in the high- and low-TEXRGPS groups according to the submap algorithm. (E) IMvigor210, an independent immunotherapy cohort, was used to validate the ability of the TEXRGPS to predict response to immunotherapy. *, P<0.05; **, P<0.01; ***, P<0.001. IPS, immunophenotype score; LUAD, lung adenocarcinoma; PD-L1, programmed death-ligand 1; TCIA, The Cancer Immune Atlas.

Second, we performed GSEA and GSVA to evaluate the biological pathways in the low- and high-TEXRGPS groups and visualized differential pathway enrichment in a heatmap. We found that pathways related to immune cell function, such as IL-2/STAT5 signaling and IL-6/JAKSTAT5 signaling pathways, were enriched in the low-TEXRGPS group, and pathways correlated with inhibitory factors in the TME, such as glycolysis, fatty acid metabolism, and cell proliferation, were enriched in the high-TEXRGPS group (Figure 13A-13C).

Figure 13 GSEA, GSVA, and the nomogram. (A,B) GSEA of the low- and high-TEXRGPS groups. (C) GSVA of the low- and high-TEXRGPS groups. (D) Performance of the nomogram composed of the TEXRGPS and clinical factors in predicting the 3-, 4-, and 5-year OS probabilities of patients with LUAD. (E,F) The ROC curve and the calibration curve of the nomogram. ***, P<0.001. AUC, area under the curve; GSEA, gene set enrichment analysis GSVA, gene set variation analysis LUAD, lung adenocarcinoma; OS, overall survival; ROC, receiver operating characteristic; TCGA, The Cancer Genome Atlas.

Development of a nomogram for predicting the survival of patients with LUAD

A nomogram consisting of the TEXRGPS and clinicopathological parameters was developed to predict the 3-, 4- and 5-year OS rates. (Figure 13D). The AUCs of the nomogram in predicting 3-, 4- and 5-year OS were 0.724, 0.706, and 0.724, respectively (Figure 13E). The calibration curve indicated good agreement between the predicted and observed survival probabilities, indicating that the nomogram was well calibrated (Figure 13F).

Discussion

Among malignancies, LUAD is one of the most common and is associated with high global mortality. Consequently, research on effective treatments, especially anti-PD-1/PD-L1 and anti-CTLA-4 therapy, is considered a particularly high priority. In contrast to conventional therapies such as surgical treatment, radiotherapy, targeted therapy, and chemotherapy, immunotherapy exerts its effect by remodeling the immunosuppressive TME, but more capable agents are needed in this regard. Although a considerable volume of research has identified an array of important regulatory factors in the TME and provided the theoretical basis for next-generation immunotherapy therapy, how transcription factors related to T-cell exhaustion affect the TME and their influence on the response to immunotherapy remain unclear. T-cell exhaustion-related immunity is considered to be a key factor that can modulate antitumor treatment; for example, it has been found that chimeric antigen receptor T-cell exhaustion is a major limitation to this type of therapy (23). Therefore, characterizing T-cell exhaustion status may help predict the survival of patients with LUAD and identify populations amenable to cancer immunotherapy. In our study, we sought to clarify if transcription factors related to T-cell exhaustion influence immune infiltration features, prognosis, and immunotherapy sensitivity in patients with LUAD.

We developed a LUAD prognostic risk score (TEXRGPS) using TCGA-LUAD and evaluated it in six independent GEO cohorts. TEXRGPS was associated with prognosis and computational estimates of immunotherapy sensitivity. To explore potential clinical utility, we integrated TEXRGPS with age, sex, and tumor-node-metastasis stage in a nomogram. These retrospective findings support further validation but do not establish clinical utility or treatment-selection benefit.

The GSEA and GSVA results showed that nutrient metabolism and cell cycle pathways, which are recognized targets for immunotherapy, were enriched in the high-TEXRGPS group. The drug sensitivity analysis we performed may provide insight into developing targeted drugs. By contrast, pathways associated with inflammatory response were enriched in the low-TEXRGPS group, and this we had a higher degree of immune cell infiltration, particularly for T cells and B cells. We further confirmed the ability of the TEXRGPS to predict response to immunotherapy via an analysis of gene expression, proteomics, and IPS, with the TEXRGPS demonstrating satisfactory performance.

The TEXRGPS includes six genes, namely CCL19, CD79A, CHRDL1, MS4A1, S100P, and SFTPC. CCL19, CD79A, MS4A1, and S100P were more highly expressed in LUAD tissues, whereas CHRDL1 and SFTPC were less expressed. These genes, except for S100P, were all highly expressed in the low-TEXRGPS group and correlated with a positive prognosis. S100P belongs to the S100 family and is recognized for its role in promoting tumor cell proliferation, survival, angiogenesis, and metastasis. Its biological functions are mediated through several signalling pathways and regulatory molecules, including calcium-dependent signalling, the receptor for advanced glycation end products-dependent pathway, ezrin, calcyclin-binding protein/Siah1-interacting protein and cathepsin D (24), with its biological effect being particularly evident in LUAD (25,26). Moreover, the expression of S100P may be closely related to tumor size and tumor staging (27). As to how S100P promotes the progression of lung cancer, previous studies suggest that it may be through increasing its interaction with integrin α7 or interferon beta (IFN-β) (28,29). Other studies have indicated that S100P is an important link in multiple pathways related to lung cancer invasion such as the Keap1-Nrf2 axis, the SIX3-Wnt/β-catenin signaling axis, the PI3K/AKT pathway, and the YAP pathway (30-33). Analyses of the TCGA and GEO suggest that S100P is an effective prognostic gene for LUAD, correlating with immune cell infiltration, and may also predictive response to immunotherapy (34,35).

CCL19 exerts tumor-suppressive functions through the induction of antitumor immune responses and inhibition of angiogenesis but may also promote tumor development by inducing inflammation, cell growth, and metastasis. CCL19 dysregulation has been reported to be a tumor biomarker for diagnosis and prognosis in several cancers, including colorectal, breast, pancreatic, and lung cancer (36). For example, Itakura et al. demonstrated that chemokine receptor 7 (CCR7) and its ligand CCL19 are prognostic factors and that CCR7 is strongly related to Crk-like adaptor protein (CrkL) and c-ABL kinase messenger RNA expression, which play a role in malignant cell transformation in LUAD (37). In one study on the immunosuppressive TME, CCL19 was found to function as a molecular rheostat, balancing immune activation and suppression within the TME (38). However, a study using a murine model of spontaneous bronchoalveolar cell carcinoma revealed that CCL19 can induce CD4⁺ and CD8⁺ T-cell infiltration into the tumor, increasing the secretion of antitumor cytokines and decreasing the expression of immunosuppressive cytokine TGF-β (39).

SFTPC has also been extensively investigated. Although it remains unclear whether it exerts a similar inhibitory effect on lung cancer progression to that of other surfactant proteins, the study by Li et al. suggested that the miR-629-3p-meditated downregulation of SFTPC may promote the LUAD progression and that low SFTPC expression correlated with the poor OS of patients with LUAD (40). Another study showed that SFTPC was positively correlated with the infiltration of six immune cell types (CD8⁺ T cells, B cells, CD4⁺ T cells, macrophages, neutrophils, and DC) both in LUAD and lung squamous cell carcinoma; moreover, its low expression was associated with shorter OS in patients with lung cancer (41). CHRDL1 is a specific antagonist of bone morphogenetic protein (BMP), and BMP signaling participates in many cancer processes, including cell proliferation, migration, and invasion. A previous analysis of the TCGA and GEO databases suggested SFTPC to be a signature of endoplasmic reticulum stress-related genes and potentially correlated with the expression level of immunomodulatory molecules including CD28, CD27, and IDO2, among others (42).

CHRDL1 is an inhibitor of BMPs, and, according to Deng et al., is related to processes governing the homeostasis of cell proliferation, including cell cycle checkpoint, protein translation, and G2M phase checkpoint; moreover, it may suppress the cancer invasion through innate immunity and improve the effectiveness of immunotherapy by regulating immune infiltration in LUAD (43). Although its specific mechanism of action in LUAD remains unclear, database analyses indicated that CHRDL1 is an important prognostic factor for patients with LUAD (43-45).

CD79A and MS4A1 are the specific markers of B cells, and the different expression of these two genes in our study implies a distinct degree of infiltration of B cells between the low- and high-TEXRGPS groups of LUAD. Furthermore, it suggests that B cells may play an important role in the emergence and development of tumors and may be used to predict antitumor efficacy. The favorable clinical outcomes provided by tumor-infiltrating B cells for patients with LUAD was demonstrated in a study that analyzed RNA-sequencing data from the TCGA. It was found that the expression levels of B-cell and plasma cell signature genes were associated with improved OS in patients with LUAD (46).

Research has established that B cells are important components of the TME, with approximately two-thirds of lung tumor-infiltrating immune cells being composed of T and B cells (47-49). B cells have different subtypes that exerts different immunological effects on lung cancer. For example, Wu et al. found that the abundance of tumor-infiltrating B cells is positively associated with delayed cancer progression, improved therapeutic response, and extended survival in patients lung cancer (50). Another study reported that B-cell infiltration is positively correlated with PD-L1 expression and that PD-L1 acts as a negative regulator to influence B-cell infiltration-mediated tumor immunity in LUAD, which suggests that B-cell infiltration may be involved in the effect that anti-PD-L1 therapy exerts (51). In contrast, regulatory B cells have been shown to inhibit T cell-mediated immunity through inhibitory cytokines such as IL-10, IL-35, and transforming growth factor beta (TGF-β) (52,53). Moreover, certain type of B cells, by activating several STAT3 downstream proangiogenic genes, may promote tumor progression through angiogenesis (54). However, the exact role of B cells in LUAD and their effect on immunotherapy have not been sufficiently clarified.

After further analysis of the TCGA database, we found a significant correlation between CD79A and IRF4. In addition, through further immunohistochemical analysis, we found that the expression of CD79A and IRF4 in tumor group was significantly higher than that in the normal group. Other studies have shown that CD79A is highly correlated with the prognosis of patients with lung cancer and is an effective predictor of response to immunotherapy (55-58). Transcription factor IRF4, an important regulator of a variety of immune cells, has been demonstrated to promote B-cell differentiation and maturation (59). Moreover, by enhancing the suppressive effect of regulatory T cells, it can promote the exhaustion of T cells and thus tumor development (60). Other research found that a subset of DCs with low IRF4 expression can restore immunotherapy resistance (61). Other studies have shown that IRF4-associated signaling pathways are important predictors of immunotherapy efficacy and immune-related adverse events (62,63). Although our immunohistochemical analysis did not indicate a significant relationship between CD79A and IRF4, the literature and our analysis suggest that they may be nonetheless connected.

The TME, a key link in T-cell exhaustion, not only plays an important role in antitumor immunity but is also being increasingly regarded as marker of immunotherapy efficacy, although it is still not used in clinical practice. In our study, the low-TEXRGPS subgroup had higher immune and stromal scores but a higher ESTIMATE score and survival rates. In addition, there was a significant difference in the degree of immune cell infiltration between the high- and low-TEXRGPS group. Patients in the low-TEXRGPS subgroup showed a relatively high infiltration of immune cells, especially different subtypes of T cells and B cells. The low-TEXRGPS had highly similar results to those of the high-TEXRGPS subgroup with T-cell exhaustion. Moreover, the low-TEXRGPS group not only expressed a variety of inhibitory receptors but had an elevated expression of a variety of T-cell exhaustion-related DEGs. The differences between these groups were closely related to the changes in the composition of the microenvironment caused by the dynamic progression of the tumor (64,65). Finally, we found that T-cell exhaustion was associated with poor response to treatment.

Thus, considering these differences, we deemed that it may be productive to assess the effect of ICI treatment in the two groups of patients with LUAD and subsequently analyzed the relationship between the TEXRGPS and both PD-1/PD-L1 and CTLA-4 expression via proteomics and the IPS. We found that the proteomics and IPS scores of PD-1/CTLA-4 or PD-L1 were higher in the low-TEXRGPS group than in the high-TEXRGPS group, suggesting that immunotherapy may be more effective in the low-TEXRGPS group. This is consistent with the results of subsequent subclass mapping analysis and validation in an independent cohort. Compared to the high-TEXRGPS, the low-TEXRGPS group had a higher degree of immune cell infiltration and a greater response to anti-PD-1/PD-L1 and anti-CTLA-4 therapy, suggesting that the TEXRGPS is associated with the inhibition of antitumor immunity in LUAD. Therefore, the TEXRGPS may serve as tool for effectively evaluating the prognosis of patients with LUAD receiving immunotherapy.

Several limitations should be acknowledged. First, standardized markers and criteria for T-cell exhaustion remain incomplete, and the bulk-transcriptomic and deconvolution analyses used here cannot directly establish a dysfunctional T-cell phenotype or a causal role for NFATC1, EOMES, TOX, or the TEXRGPS genes. The available single-cell analysis described cell-type expression patterns but did not provide functional validation. Second, sequential feature filtering and model fitting may introduce selection instability and TCGA-specific overfitting. Although the signature was evaluated in six independent GEO cohorts, these retrospective public datasets may share platform-, processing-, and selection-related biases. Third, IMvigor210 is a urothelial-cancer rather than LUAD immunotherapy cohort, so its findings represent cross-tumor supportive evidence. Fourth, while GAPDH was selected as the reference gene for RT-qPCR normalization based on its widespread use in the literature and its demonstrated stability in our preliminary experiments, we acknowledge that GAPDH is a glycolytic enzyme and may be subject to expression alterations in cancers with metabolic reprogramming. Formal validation of GAPDH expression stability across LUAD tissues—potentially using multiple reference genes or algorithmic approaches—would further strengthen the normalization strategy. Finally, the clinical RT-qPCR series was small and conducted at a single center. Prospective multicenter validation using prespecified coefficients and thresholds, clinically defined treatment endpoints, and single-cell, protein-level, and functional experiments is required before clinical implementation.

Conclusions

Our study examined associations between transcription factors previously linked to T-cell exhaustion, the tumor microenvironment, and clinical outcomes in LUAD. We constructed the six-gene TEXRGPS and evaluated it in six independent GEO cohorts—GSE72094, GSE31210, GSE68465, GSE26939, GSE42127, and GSE37745. TEXRGPS was associated with overall survival and with computational estimates of immune infiltration and immunotherapy response. These retrospective results support further investigation but do not establish a direct functional relationship with exhausted T cells or justify clinical treatment selection. Prospective multicenter and mechanistic validation is required.

Acknowledgments

The authors wish to thank all the patients and staff who participated in this study.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2026-1101/rc

Data Sharing Statement: Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2026-1101/dss

Peer Review File: Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2026-1101/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2026-1101/coif). The authors have no conflicts of interest to declare.

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. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of The Second Affiliated Hospital of Harbin Medical University (approval No. KY2023-139), and written informed consent was obtained from all participants.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. Wang M, Herbst RS, Boshoff C. Toward personalized treatment approaches for non-small-cell lung cancer. Nat Med 2021;27:1345-56. [Crossref] [PubMed]
3. Zhou W, Qu M, Yue Y, et al. Acetylcysteine synergizes PD-1 blockers against colorectal cancer progression by promoting TCF1+PD1+CD8+ T cell differentiation. Cell Commun Signal 2024;22:503. [Crossref] [PubMed]
4. Cheng J, Xiao Y, Peng T, et al. ETV7 limits the antiviral and antitumor efficacy of CD8+ T cells by diverting their fate toward exhaustion. Nat Cancer 2025;6:338-56. [Crossref] [PubMed]
5. Zhu Z, Lou G, Teng XL, et al. FOXP1 and KLF2 reciprocally regulate checkpoints of stem-like to effector transition in CAR T cells. Nat Immunol 2024;25:117-28. [Crossref] [PubMed]
6. Zajac AJ, Blattman JN, Murali-Krishna K, et al. Viral immune evasion due to persistence of activated T cells without effector function. J Exp Med 1998;188:2205-13. [Crossref] [PubMed]
7. Kurachi M. CD8+ T cell exhaustion. Semin Immunopathol 2019;41:327-37. [Crossref] [PubMed]
8. Martinez GJ, Pereira RM, Äijö T, et al. The transcription factor NFAT promotes exhaustion of activated CD8⁺ T cells. Immunity 2015;42:265-78. [Crossref] [PubMed]
9. Man K, Gabriel SS, Liao Y, et al. Transcription Factor IRF4 Promotes CD8+ T Cell Exhaustion and Limits the Development of Memory-like T Cells during Chronic Infection. Immunity 2017;47:1129-1141.e5. [Crossref] [PubMed]
10. Seo H, Chen J, González-Avalos E, et al. TOX and TOX2 transcription factors cooperate with NR4A transcription factors to impose CD8+ T cell exhaustion. Proc Natl Acad Sci U S A 2019;116:12410-5. [Crossref] [PubMed]
11. Chen Y, Zander RA, Wu X, et al. BATF regulates progenitor to cytolytic effector CD8+ T cell transition during chronic viral infection. Nat Immunol 2021;22:996-1007. [Crossref] [PubMed]
12. McLane LM, Ngiow SF, Chen Z, et al. Role of nuclear localization in the regulation and function of T-bet and Eomes in exhausted CD8 T cells. Cell Rep 2021;35:109120. [Crossref] [PubMed]
13. Quigley M, Pereyra F, Nilsson B, et al. Transcriptional analysis of HIV-specific CD8+ T cells shows that PD-1 inhibits T cell function by upregulating BATF. Nat Med 2010;16:1147-51. [Crossref] [PubMed]
14. Alfei F, Kanev K, Hofmann M, et al. TOX reinforces the phenotype and longevity of exhausted T cells in chronic viral infection. Nature 2019;571:265-9. [Crossref] [PubMed]
15. Jung S, Baek JH. The Potential of T Cell Factor 1 in Sustaining CD8+ T Lymphocyte-Directed Anti-Tumor Immunity. Cancers (Basel) 2021;13:515. [Crossref] [PubMed]
16. Seo H, González-Avalos E, Zhang W, et al. BATF and IRF4 cooperate to counter exhaustion in tumor-infiltrating CAR T cells. Nat Immunol 2021;22:983-95. [Crossref] [PubMed]
17. Newman AM, Liu CL, Green MR, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods 2015;12:453-7. [Crossref] [PubMed]
18. Becht E, Giraldo NA, Lacroix L, et al. Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression. Genome Biol 2016;17:218. [Crossref] [PubMed]
19. Li B, Severson E, Pignon JC, et al. Comprehensive analyses of tumor immunity: implications for cancer immunotherapy. Genome Biol 2016;17:174. [Crossref] [PubMed]
20. Aran D, Hu Z, Butte AJ. xCell: digitally portraying the tissue cellular heterogeneity landscape. Genome Biol 2017;18:220. [Crossref] [PubMed]
21. Racle J, de Jonge K, Baumgaertner P, et al. Simultaneous enumeration of cancer and immune cell types from bulk tumor gene expression data. Elife 2017;6:e26476. [Crossref] [PubMed]
22. Finotello F, Mayer C, Plattner C, et al. Molecular and pharmacological modulators of the tumor immune contexture revealed by deconvolution of RNA-seq data. Genome Med 2019;11:34. [Crossref] [PubMed]
23. Zhang X, Zhang C, Lu S, et al. Miltefosine reinvigorates exhausted T cells by targeting their bioenergetic state. Cell Rep Med 2024;5:101869. [Crossref] [PubMed]
24. Wang X, Zhao D, Zhao E, et al. Multiple roles of S100P in pan carcinoma: Biological functions and mechanisms (Review). Oncol Rep 2025;53:62. [Crossref] [PubMed]
25. Rehbein G, Simm A, Hofmann HS, et al. Molecular regulation of S100P in human lung adenocarcinomas. Int J Mol Med 2008;22:69-77. [PubMed]
26. Watanabe T, Miura T, Degawa Y, et al. Comparison of lung cancer cell lines representing four histopathological subtypes with gene expression profiling using quantitative real-time PCR. Cancer Cell Int 2010;10:2. [Crossref] [PubMed]
27. Li J, Li W, Qiao R, et al. Peripheral blood-based S100P hypomethylation as a biomarker for the early-stage non-small cell lung cancer. Epigenomics 2025;17:1423-33. [Crossref] [PubMed]
28. Hsu YL, Hung JY, Liang YY, et al. S100P interacts with integrin α7 and increases cancer cell migration and invasion in lung cancer. Oncotarget 2015;6:29585-98. [Crossref] [PubMed]
29. Kazakov AS, Mayorov SA, Deryusheva EI, et al. Highly specific interaction of monomeric S100P protein with interferon beta. Int J Biol Macromol 2020;143:633-9. [Crossref] [PubMed]
30. Chien MH, Lee WJ, Hsieh FK, et al. Keap1-Nrf2 Interaction Suppresses Cell Motility in Lung Adenocarcinomas by Targeting the S100P Protein. Clin Cancer Res 2015;21:4719-32. [Crossref] [PubMed]
31. De Marco C, Laudanna C, Rinaldo N, et al. Specific gene expression signatures induced by the multiple oncogenic alterations that occur within the PTEN/PI3K/AKT pathway in lung cancer. PLoS One 2017;12:e0178865. [Crossref] [PubMed]
32. Tan BS, Yang MC, Singh S, et al. LncRNA NORAD is repressed by the YAP pathway and suppresses lung and breast cancer metastasis by sequestering S100P. Oncogene 2019;38:5612-26. [Crossref] [PubMed]
33. Liu S, Tian Y, Zheng Y, et al. TRIM27 acts as an oncogene and regulates cell proliferation and metastasis in non-small cell lung cancer through SIX3-β-catenin signaling. Aging (Albany NY) 2020;12:25564-80. [Crossref] [PubMed]
34. Li Q, Xie D, Yao L, et al. Combining autophagy and immune characterizations to predict prognosis and therapeutic response in lung adenocarcinoma. Front Immunol 2022;13:944378. [Crossref] [PubMed]
35. Zhou E, Wu F, Guo M, et al. Identification of a novel gene signature of lung adenocarcinoma based on epidermal growth factor receptor-tyrosine kinase inhibitor resistance. Front Oncol 2022;12:1008283. [Crossref] [PubMed]
36. Gowhari Shabgah A, Al-Obaidi ZMJ, Sulaiman Rahman H, et al. Does CCL19 act as a double-edged sword in cancer development?. Clin Exp Immunol 2022;207:164-75. [Crossref] [PubMed]
37. Itakura M, Terashima Y, Shingyoji M, et al. High CC chemokine receptor 7 expression improves postoperative prognosis of lung adenocarcinoma patients. Br J Cancer 2013;109:1100-8. [Crossref] [PubMed]
38. Lee H, Lim SH, Hwang WC, et al. PLD1 and PLD2 promote an immunosuppressive tumor microenvironment via CCL19-dependent macrophage polarization and PD-L1 induction. Exp Mol Med 2026;58:1870-84. [Crossref] [PubMed]
39. Hillinger S, Yang SC, Batra RK, et al. CCL19 reduces tumour burden in a model of advanced lung cancer. Br J Cancer 2006;94:1029-34. [Crossref] [PubMed]
40. Li B, Meng YQ, Li Z, et al. MiR-629-3p-induced downregulation of SFTPC promotes cell proliferation and predicts poor survival in lung adenocarcinoma. Artif Cells Nanomed Biotechnol 2019;47:3286-96. [Crossref] [PubMed]
41. Ren L, Wen X, Liu M, et al. Comprehensive Analysis of the Molecular Characteristics and Prognosis value of AT II-associated Genes in Non-small Cell Lung Cancer. Comput Math Methods Med 2022;2022:3106688. [Crossref] [PubMed]
42. Yang C, Wei Y, Li W, et al. Prognostic Risk Signature and Comprehensive Analyses of Endoplasmic Reticulum Stress-Related Genes in Lung Adenocarcinoma. J Immunol Res 2022;2022:6567916. [Crossref] [PubMed]
43. Deng B, Chen X, Xu L, et al. Chordin-like 1 is a novel prognostic biomarker and correlative with immune cell infiltration in lung adenocarcinoma. Aging (Albany NY) 2022;14:389-409. [Crossref] [PubMed]
44. Deng H, Hang Q, Shen D, et al. Low expression of CHRDL1 and SPARCL1 predicts poor prognosis of lung adenocarcinoma based on comprehensive analysis and immunohistochemical validation. Cancer Cell Int 2021;21:259. [Crossref] [PubMed]
45. Wu Y, Yang L, Zhang L, et al. Identification of a Four-Gene Signature Associated with the Prognosis Prediction of Lung Adenocarcinoma Based on Integrated Bioinformatics Analysis. Genes (Basel) 2022;13:238. [Crossref] [PubMed]
46. Iglesia MD, Parker JS, Hoadley KA, et al. Genomic Analysis of Immune Cell Infiltrates Across 11 Tumor Types. J Natl Cancer Inst 2016;108:djw144. [Crossref] [PubMed]
47. Stankovic B, Bjørhovde HAK, Skarshaug R, et al. Immune Cell Composition in Human Non-small Cell Lung Cancer. Front Immunol 2018;9:3101. [Crossref] [PubMed]
48. Wang SS, Liu W, Ly D, et al. Tumor-infiltrating B cells: their role and application in anti-tumor immunity in lung cancer. Cell Mol Immunol 2019;16:6-18. [Crossref] [PubMed]
49. Lu G, Liu H, Wang H, et al. Potentially functional variants of INPP5D and EXOSC3 in immunity B cell-related genes are associated with non-small cell lung cancer survival. Front Immunol 2024;15:1440454. [Crossref] [PubMed]
50. Wu H, Chen C, Gu L, et al. B cell deficiency promotes the initiation and progression of lung cancer. Front Oncol 2022;12:1006477. [Crossref] [PubMed]
51. Ho KH, Chang CJ, Huang TW, et al. Gene landscape and correlation between B-cell infiltration and programmed death ligand 1 expression in lung adenocarcinoma patients from The Cancer Genome Atlas data set. PLoS One 2018;13:e0208459. [Crossref] [PubMed]
52. Schwartz M, Zhang Y, Rosenblatt JD. B cell regulation of the anti-tumor response and role in carcinogenesis. J Immunother Cancer 2016;4:40. [Crossref] [PubMed]
53. Sarvaria A, Madrigal JA, Saudemont A. B cell regulation in cancer and anti-tumor immunity. Cell Mol Immunol 2017;14:662-74. [Crossref] [PubMed]
54. Yang C, Lee H, Pal S, et al. B cells promote tumor progression via STAT3 regulated-angiogenesis. PLoS One 2013;8:e64159. [Crossref] [PubMed]
55. Guo Z, Qi X, Li Z, et al. Development and validation of an immune-related gene signature for prognosis in Lung adenocarcinoma. IET Syst Biol 2023;17:27-38. [Crossref] [PubMed]
56. Zhang J, Chen A, Xue Z, et al. Identification of immune-associated prognostic biomarkers in lung adenocarcinoma on the basis of gene co-expression network. Immunopharmacol Immunotoxicol 2023;45:334-46. [Crossref] [PubMed]
57. Xu L, Zhao Y, Song S, et al. Construction and validation of a lung adenocarcinoma prognostic model based on neutrophil extracellular traps and oxidative stress-related genes. Eur J Med Res 2025;30:1275. [Crossref] [PubMed]
58. Sambanthamoorthy S, Ren Y, Galvez TK, et al. The Presence of CD11c+ B Cells with Potent Effector Memory Phenotype in Lung Adenocarcinoma Correlates with Overall Patient Survival. Cancer Immunol Res 2026;14:811-26. [Crossref] [PubMed]
59. Lu R. Interferon regulatory factor 4 and 8 in B-cell development. Trends Immunol 2008;29:487-92. [Crossref] [PubMed]
60. Alvisi G, Brummelman J, Puccio S, et al. IRF4 instructs effector Treg differentiation and immune suppression in human cancer. J Clin Invest 2020;130:3137-50. [Crossref] [PubMed]
61. Brassard J, Gill ME, Bernatchez E, et al. Countering the advert effects of lung cancer on the anticancer potential of dendritic cell populations reinstates sensitivity to anti-PD-1 therapy. PLoS One 2021;16:e0260636. [Crossref] [PubMed]
62. Gao J, Ao YQ, Zhang LX, et al. Exosomal circZNF451 restrains anti-PD1 treatment in lung adenocarcinoma via polarizing macrophages by complexing with TRIM56 and FXR1. J Exp Clin Cancer Res 2022;41:295. [Crossref] [PubMed]
63. Zhang L, Shi Y, Han X. Immunogenomic correlates of immune-related adverse events for anti-programmed cell death 1 therapy. Front Immunol 2022;13:1032221. [Crossref] [PubMed]
64. Kim N, Kim HK, Lee K, et al. Single-cell RNA sequencing demonstrates the molecular and cellular reprogramming of metastatic lung adenocarcinoma. Nat Commun 2020;11:2285. [Crossref] [PubMed]
65. Maynard A, McCoach CE, Rotow JK, et al. Therapy-Induced Evolution of Human Lung Cancer Revealed by Single-Cell RNA Sequencing. Cell 2020;182:1232-1251.e22. [Crossref] [PubMed]