This study integrated multi-omics data with experimental validation. Publicly available datasets used included: (1) The Cancer Genome Atlas (TCGA)-LUSC Cohort (https://cancergenome.nih.gov): RNA sequencing (RNA-seq) data, clinical survival records, and proteomic profiles of patients with LUSC were retrieved from the TCGA database using the TCGAbiolinks package (v2.28.4) in R 4.3.1 (27). Data were collected on August 21, 2024, using the GDCquery function with the following filters: project = “TCGA-LUSC”, data.category = “Transcriptome Profiling”, data.type = “Gene Expression Quantification”, and workflow.type = “STAR - Counts”. Genes with zero counts in over 50% of samples were excluded during quality control. Raw counts were normalized to transcripts per million (TPM) using trimmed mean of M-values (TMM) scaling via the edgeR::calcNormFactors function, followed by log2 transformation. Batch effects from technical confounders were corrected with the ComBat algorithm. The final processed matrix was validated for robustness in downstream analyses. A total of 496 samples with complete clinical data and transcriptomic data, along with 322 samples with proteomic profiles, were included. (2) GDSC Database (https://www.cancerrxgene.org) (28): TGM2 expression levels in LUSC cell lines were screened through this database. The NCI-H520 (low TGM2 expression) and SK-MES-1 (high TGM2 expression) cell lines were selected for functional experiments.(3) UALCAN Database (https://ualcan.path.uab.edu) (29): This platform was used to identify signaling pathways potentially influenced by variations in TGM2 protein expression levels. (4) TIMER 2.0 Database (http://timer.cistrome.org/) (30): This tool enabled analysis of to analyze immune cell infiltration within the samples.

Using the immune phenotype scoring model previously developed, expression profiles of the 17 genes and patient survival data were integrated. Random forest modeling was performed with the randomForestSRC package (version 3.2.3), using parameters ntree = 1000 and nodesize = 15, to assess the significance of these genes in predicting overall survival (OS) and disease-free survival (DFS). TGM2, identified as the gene with the highest prognostic significance, was selected for further analysis. Survival analysis was conducted with the survival package (version 3.5-7), and survival curves were generated with the survminer package (version 0.4.9). The optimal cutoff value for OS (TPM = 210.89) was determined using the surv_cutpoint function, based on survival-specific maximally selected rank statistics, which accounts for to respect the time-to-event nature of OS data, classifying patients into high-risk (n = 51) and low-risk (n = 444) groups. Multivariate Cox proportional hazards regression analysis confirmed the independent prognostic value of TGM2, and the results were visualized using the forestploter package (version 1.1.2).

Cell Lines and Culture Conditions: Human LUSC cell lines NCI-H520 and SK-MES-1 (Shanghai QuiCell Biotechnology Co., Ltd.) were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin at 37°C under 5% CO₂.

TGM2 Overexpression and Knockdown: Lentiviral vectors were transfected into NCI-H520 cells to establish stable TGM2-overexpressing (TGM2-OE) cell lines. Empty vector and untreated cells served as negative controls (NC) and blank controls (Ctrl), respectively. SK-MES-1 cells were transfected with three specific siRNAs (siRNA-1, siRNA-2, and siRNA-3), with NC and Ctrl used as controls. Transfection efficiency was validated by quantitative Reverse Transcription Polymerase Chain Reaction (qRT-PCR) under the following conditions: 95°C for 3 min (initial denaturation); 40 cycles of 95°C for 15 sec, 60°C for 30 sec, 72°C for 30 sec, followed by melt curve analysis (95°C for 15 sec, 60°C for 1 min, 95°C for 15 sec) to verify amplification specificity, and Western blot analysis. Primer and siRNA sequences are listed in Supplementary Table S1 . Primary antibodies were diluted as follows: anti-TGM2 (HUABIO, ET1706-35; 1:1000), anti-p38 MAPK (HUABIO, HA722036; 1:1000), anti-Phospho-p38 MAPK (T180/Y182) (HUABIO, ER1903-01; 1:1000), anti-NR3C1 (Cell Signaling Technology, #12041; 1:1000), anti-GAPDH (Proteintech, 60004-1-Ig; 1:5000), anti-β-tubulin (CST, #2146; 1:1000). HRP-conjugated secondary antibodies (Cell Signaling Technology, #7074 for rabbit, #7076 for mouse) were used at a 1:10,000 dilution.

Cell Proliferation Assay: Cell proliferation was assessed using a CCK-8 kit (Shanghai Sangon Biotech Co., Ltd.) at 0, 24, 48, 72, and 96 hours post-transfection by measuring absorbance at 450 nm. Proliferation curves were generated accordingly.

Migration and Invasion Assays: (1) Wound Healing Assay: Cells were seeded in 6-well plates to form a monolayer. A scratch was made using a sterile pipette tip, and wound closure was observed after 24 hours. (2) Transwell Assay: For migration assays, 2×10⁴ cells/well in serum-free medium were seeded into the upper chamber of uncoated 24-well Transwell plates. For invasion assays, 2×10⁴ cells/well in serum-free medium were seeded into Matrigel-coated chambers. After incubation, cells that migrated through the membrane were fixed, stained, and counted.

Colony Formation Assay: Cells (1,000 per dish) were cultured in 6 cm dishes for 10 days, followed by crystal violet (Shanghai Beyotime Biotechnology Co., Ltd.) staining was performed after fixation. Colonies were counted manually.

Apoptosis Detection: Apoptosis rates were assessed using an Annexin V-APC/7-AAD apoptosis detection kit (Wuhan Elabscience Biotechnology Co., Ltd.) by flow cytometry.

Reverse-Phase Protein Array (RPPA) Analysis: RPPA data from TCGA-LUSC samples were downloaded to identify differentially expressed proteins (DEPs) between TGM2 high-risk and low-risk groups. Correlation analysis was performed to identify proteins significantly associated with TGM2 expression.

Pathway Screening: The UALCAN database was used to identify key cancer-related signaling pathways potentially influenced by TGM2 expression.

Pathway Validation: Western blotting was performed to detect key proteins and their phosphorylation levels identified in RPPA and pathway screening analyses, thereby validating the regulatory role of TGM2 in these pathways.

Differentially Expressed Genes (DEGs) Screening: DEGs between TGM2 high-risk and low-risk groups were identified using the limma package (version 3.56.2) (31), DEGs between TGM2 high-risk and low-risk groups were identified with thresholds set at |Fold Change (FC)| ≥2 and false discovery rate (FDR) <0.05. Results were visualized using the ggVolcano package (version 0.0.2).

Gene Set Enrichment Analysis (GSEA): The clusterProfiler package (version 4.8.3) (32) was used to analyze Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment of DEGs via the gseKEGG function, which directly accesses the KEGG pathway database (https://www.kegg.jp) through its application programming interface (API) at the time of access (9^th September 2024). Enrichment plots were generated using the GseaVis (version 0.0.9) and enrichplot (version 1.20.3) packages.

Immune Infiltration Analysis: Transcriptomic data from TCGA-LUSC were uploaded to TIMER2.0 with species set to “human” and cancer type to “LUSC”. Immune cell infiltration scores were obtained and correlated with TGM2 TPM values. Statistically significant results (P < 0.05) were selected for presentation.

Potential transcriptional regulators of TGM2 were predicted by integrating data from two databases: (1) Cistrome DB Toolkit (http://cistrome.org/db/#/): Species = “Human hg38”, gene symbol = “TGM2”, data type = “Transcription factor/chromatin regulator”, half-decay distance to TSS = 10 kb. (2) hTFtarget (https://guolab.wchscu.cn/hTFtarget/#!/): Direct search for ‘TGM2’ in the Quick Search module. Overlapping TFs between both databases and DEGs were prioritized for further analysis. A heatmap was generated to display TFs based on TGM2 risk groups. The top-ranked TF, NR3C1, was selected for additional investigation: (1) Expression Correlation: The GEPIA database (http://gepia.cancer-pku.cn) (33) was used to analyze the correlation between NR3C1 and TGM2 expression in LUSC. (2) Group Differences and Survival Analysis: Differences in NR3C1 expression differences between TGM2 risk groups were assessed. Survival analysis and survival curve generation were performed based on NR3C1 expression and TCGA-LUSC patient survival data. (3) Binding Sites Prediction: The JASPAR database was used to predict NR3C1 binding motifs within the TGM2 gene sequence.

To validate the direct transcriptional regulation of TGM2 by NR3C1, a chromatin immunoprecipitation (ChIP) assay was performed using the SimpleChIP^® Enzymatic Chromatin IP Kit (Agarose Beads) (#9002, CST, USA). Briefly, cross-linked chromatin was sonicated to generate appropriately sized fragments, followed by immunoprecipitation with an NR3C1 antibody, with IgG serving as a negative control. The enriched DNA fragments were quantitatively analyzed by qRT-PCR. Primer sequences used for ChIP-PCR are listed in Supplementary Table S1 .

All in vitro phenotypic assays were conducted with at least three independent biological replicates per experimental condition. Data analysis and visualization were performed using R 4.3.1 and GraphPad Prism 9.0.0 (San Diego, California, USA). Band intensities were quantified using ImageJ software (v1.53t, NIH), normalized to GAPDH. Unless otherwise stated, default parameters were applied for the software packages mentioned, and data visualization was performed using the ggplot2 package (version 3.3.5). Continuous variables are presented as mean ± standard deviation (SD). Comparisons between two groups were made using the Student’s t-test or Mann-Whitney test, depending on the assumptions of homogeneity of variance and normal distribution. For comparisons among multiple groups, ordinary one-way Analysis of Variance (ANOVA) with multiple comparisons was used, except for the CCK-8 assay, which employed two-way ANOVA with multiple comparisons. Correlation analysis between two groups was conducted using the Spearman’s test, and survival analysis was performed using the Log-rank test. A P-value <0.05 was considered statistically significant, with the following notation: * (P < 0.05), ** (P < 0.01), *** (P < 0.001), and **** (P < 0.0001).

The flowchart of this study is shown in Figure 1 . The previously developed immune phenotype scoring model (13) was reconstructed and analyzed using the random forest method. Among the 17 modeling genes, TGM2 was identified as the most important predictor for both OS ( Figure 2A ) and DFS ( Figure 2B ). Univariate survival analysis revealed that the high TGM2 expression group had significantly shorter OS (P = 0.00018, Figure 2C ) and DFS (P = 0.00019, Figure 2D ) compared to the low-expression group. Multivariate Cox regression analysis confirmed that TGM2 served as an independent prognostic factor for both OS ( Figure 2E ) and DFS ( Figure 2F ) in patients with LUSC. These results suggest TGM2 as a promising candidate for further investigation.

In functional studies of LUSC cells, five LUSC cell lines (NCI-H520, HARA, NCI-H2170, NCI-H226, SK-MES-1) were selected from the GDSC database, covering a wide range of endogenous TGM2 expression levels (high/medium/low), and based on commercial availability with authenticated short tandem repeat (STR) profiles. Among these, the NCI-H520 cell line exhibited the lowest TGM2 expression level, while SK-MES-1 showed the highest TGM2 expression ( Figure 3A ). Experimental validation confirmed a statistically significant difference in TGM2 expression between these two cell lines (P < 0.0001, Figure 3B ). Consequently, SK-MES-1 was chosen for TGM2 knockdown experiments to assess loss-of-function effects, while NCI-H520 used for TGM2 overexpression studies to examine gain-of-function outcomes.

Lentiviral transfection successfully generated stable TGM2-OE NCI-H520 cells, exhibiting a more than 30,000-fold increase in TGM2 expression compared to the NC group (P < 0.0001, Figure 3C ). Functional assays showed that TGM2 overexpression significantly enhanced malignant phenotypes in NCI-H520 cells: (1) Proliferation: The CCK-8 assay revealed significantly higher proliferation rates in the TGM2-OE group at 72 hours compared to controls (P = 0.0021, Figure 3D ). (2) Migration and Invasion: The wound healing assay showed a 1.3443-fold increase in migration capacity (P = 0.0013, Figure 3E ). Transwell assays revealed a 3.0107-fold increase in migrated cells ( Figure 3F ) and a 2.3543-fold increase in invaded cells ( Figure 3G ). (3) Colony Formation: The colony formation assay showed a 1.6625-fold increase in efficiency in the TGM2-OE group (P = 0.0050, Figure 3H ). (4) Apoptosis: Flow cytometry analysis showed a 0.5333 reduction in the apoptosis rate in the TGM2-OE group (P < 0.0001, Figure 3I ).

Notably, TGM2 knockdown in SK-MES-1 cells ( Supplementary Figure S1 ) produced phenotypic effects opposite to those observed in the overexpression experiments. Three siRNAs were validated for mRNA knockdown efficacy ( Supplementary Figure S1A ), and siRNA-1, selected for its balance of efficacy and specificity, was used in all functional assays, with protein-level silencing confirmed in Figure 4E . TGM2 knockdown in SK-MES-1 cells significantly suppressed proliferation ( Supplementary Figure S1B ), migration ( Supplementary Figures S1C, D ), invasion ( Supplementary Figure S1E ), and colony formation ( Supplementary Figure S1F ), while promoting apoptosis ( Supplementary Figure S1G ), confirming TGM2’s oncogenic role. These findings are consistent with the survival analysis results presented in Figure 2 . Collectively, this study demonstrates that TGM2 promotes LUSC progression by enhancing proliferation, migration, invasion, and colony formation, while suppressing apoptosis.

To investigate the regulatory mechanisms of TGM2 on its downstream pathways, patients were stratified into high- and low-risk groups based on the optimal cutoff value of TGM2 expression level (TPM = 210.89), which was significantly associated with OS ( Figure 2C ), followed by proteomic analysis. RPPA analysis identified 42 DEPs between the two groups ( Figure 4A ), of which 35 exhibited significant correlations with TGM2 expression ( Figure 4B ). Notably, the protein levels of p38 MAPK and PD-L1 were significantly upregulated in the TGM2 high-risk group, while CTLA-4 was notably downregulated. Although several proteins (e.g., ACVRL1, DUSP4, PREX1) showed increased expression in the high-risk group, PD-L1 and p38 MAPK were prioritized for mechanistic validation due to their established roles in tumor-immune interactions (34, 35). This targeted approach enabled robust experimental confirmation of the TGM2-p38 MAPK regulatory axis in LUSC pathogenesis, while the complete DEPs profile serves as a resource for future investigations into additional candidates. To further explore the potential mechanisms of TGM2 in LUSC, the relationship between TGM2 expression and signaling pathways was analyzed using the UALCAN database. Statistically significant differences were observed in receptor tyrosine kinases (RTKs)-related pathways between LUSC and normal tissues ( Figure 4C ). Experimental validation confirmed that TGM2-OE NCI-H520 cells exhibited significantly higher levels of phosphorylated p38 MAPK proteins (P < 0.0001 vs. NC, Figure 4D ), whereas TGM2-silenced SK-MES-1 cells showed marked reductions (P<0.0001 vs. NC, Figure 4E ). These results suggest that TGM2 deficiency activates the p38 MAPK signaling pathway, thereby promoting tumor progression.

Samples were stratified based on TGM2 expression levels and DEGs analysis identified a total of 8,562 DEGs. Specifically, compared to the TGM2 high-risk group, 3,714 genes were significantly downregulated, while 4,848 genes were upregulated in the low-risk group ( Figure 5A ). These DEGs, along with their log₂FC values, were subjected to GSEA, which revealed significant enrichment in 42 potential functional pathways based on adjusted P-values. The top-ranked pathways are visually highlighted im Figure 5B . Notably, pathways such as antigen processing and presentation (35), natural killer (NK) cell-mediated cytotoxicity (36), and Th1/Th2 cell differentiation (37) were of particular interest. To further investigate these pathways, a gene network was constructed ( Figure 5C ) and the GSEA results for these three key pathways were detailed ( Figure 5D ). Notably, all these pathways were closely associated with the p38 MAPK signaling pathway (34–37). This led to the hypothesis that TGM2 promotes LUSC progression by activating the p38 MAPK signaling pathway, while potentially exerting immunosuppressive effects via modulation of this pathway.

The regulatory role of TGM2 in immune cell populations, particularly CD4⁺ T cells and NK cells, was further explored by analyzing correlations between TGM2 TPM values and immune cell infiltration scores. TGM2 expression was significantly negatively correlated with Th1 cell scores (R = -0.186, P < 0.0001; Figure 6A ), Th2 cell scores (R = -0.193, P < 0.0001; Figure 6B ), and CD4⁺ central memory T cell scores (R = -0.108, P = 0.0153; Figure 6C ). In contrast, TGM2 expression showed a positive correlation with resting CD4⁺ memory T cell scores (R = 0.307, P < 0.0001; Figure 6D ). Additionally, TGM2 expression negatively correlated with activated myeloid dendritic cell scores (R = -0.107, P = 0.0161; Figure 6E ) and NK cell activation scores (R = -0.116, P = 0.0092; Figure 6F ). Conversely, TGM2 exhibited positive correlations with immunosuppressive cell populations, including M2 macrophages ( Figures 6G-I ), cancer-associated fibroblasts (R = 0.365, P < 0.0001; Figure 6J ), hematopoietic stem cells (R = 0.469, P < 0.0001; Figure 6K ), and endothelial cells (R = 0.628, P < 0.0001; Figure 6L ). These results suggest that TGM2 plays a critical role in immune regulation, potentially influencing antigen presentation, CD4⁺ T cell and NK cell activity, and contributing to the formation of an immunosuppressive microenvironment.

To further validate the immunomodulatory role of TGM2, infiltration scores between TGM2 high- and low-risk groups were compared. TGM2-high tumors showed significant reductions in Th1 cells (P = 0.0048) and Th2 cells (P = 0.002), with concurrent increases in M2 macrophages (QUANTISEQ/XCELL, P < 0.0001), cancer-associated fibroblasts (P < 0.0001), hematopoietic stem cells (P < 0.0001), and endothelial cells (P < 0.0001) ( Supplementary Figure S2 ). These findings align with our correlation analyses ( Figure 6 ) and support the role of TGM2 in driving immunosuppression. Although activated NK cell infiltration did not show a significant difference (P = 0.75), the suppression of NK-mediated cytotoxicity pathways identified by GSEA ( Figures 5B–D ) suggests that TGM2 may impair NK cell function rather than abundance, potentially through p38 MAPK-mediated immunomodulation.

To explore the direct regulatory mechanisms of the TGM2 gene, its upstream regulators, particularly TFs, were identified. The Cistrome DB and hTFtarget databases were employed to predict potential TFs regulating TGM2, resulting in 23 key candidates identified through intersection analysis with DEGs ( Figure 7A ). The expression profiles of these TFs in LUSC are shown in Figure 7B . Based on Cistrome DB prediction scores, NR3C1, the top-ranked TF, was prioritized for further investigation. Analysis using the GEPIA database revealed a significant positive correlation between TGM2 and NR3C1 expression in both normal and tumor tissues (R = 0.53, P = 7.6×10⁻⁶⁰; Figure 7C ). Furthermore, NR3C1 expression significantly differed between TGM2 high- and low-risk groups (P < 0.0001; Figure 7D ). Although survival analysis did not reach statistical significance, there was a trend suggesting poorer prognosis for patients with LUSC exhibiting high NR3C1 expression (P = 0.075; Figure 7E ). To investigate potential direct regulatory interactions, NR3C1 binding sites in the TGM2 promoter region were predicted using the JASPAR database. Figure 7F illustrates the nucleotide sequence features of NR3C1 binding regions, and Figure 7G lists the top ten predicted binding sites and their sequences. These findings laid the groundwork for further exploration of TGM2’s regulatory mechanisms.

To assess the regulatory role of NR3C1, two distinct siRNAs were used for knockdown experiments ( Supplementary Figure S3 ). siRNA1, selected for its efficacy, was used in subsequent functional studies, demonstrating that NR3C1 knockdown led to a concomitant decrease in TGM2 mRNA levels ( Figures 8A, B ). Conversely, NR3C1 overexpression resulted in the opposite effect, increasing TGM2 mRNA expression ( Figures 8C, D ), suggesting that NR3C1 promotes TGM2 transcription. Western blot analysis further confirmed that NR3C1 knockdown reduced TGM2 protein levels, while NR3C1 overexpression enhanced them ( Figures 8E, F ), indicating that NR3C1 upregulates TGM2 protein expression likely by enhancing its transcription. To directly test whether NR3C1 promotes TGM2 transcription, ChIP assays were performed targeting four predicted NR3C1 binding sequences (selected based on a prediction score threshold of >0.65, as shown in Figure 7G ) within the TGM2 promoter region. The locations of the four validated sequences are shown in Figure 8G . The results identified Site 1 (921-935) as the most probable NR3C1 binding site on the TGM2 promoter ( Figure 8H ), suggesting that NR3C1 likely binds to Site 1 (921-935) to stimulate TGM2 transcription, thereby upregulating TGM2 protein expression. In summary, NR3C1 was identified as a potential upstream transcriptional regulator of TGM2, and their expression correlation and direct regulatory relationship were preliminarily revealed.

These findings deepen our understanding of the role of TGM2 in LUSC pathogenesis, provide directions for future research and offer valuable insights for developing potential therapeutic strategies targeting TGM2 and its regulatory network. The schematic diagram of the scientific hypothesis in this study is shown in Figure 9 , created using Figdraw (https://www.figdraw.com).