Identification of prognostic immune-related genes and evaluation of chemotherapy and immunotherapy responses in pancreatic cancer

Introduction

Pancreatic cancer is generally considered as a solid tumor with a poor prognosis (5-year survival rate below 9%). In recent years, the incidence and mortality of pancreatic cancer have been increasing worldwide (1). Pancreatic cancer was expected to have 60,430 new cases and cause 48,220 fatalities in the United States in 2021 (2). Pancreatic ductal adenocarcinoma (PDAC) is the most common pathological type of pancreatic cancer (more than 90%) (3). PDAC is usually insidious and diagnosed in advanced stages, resulting in lost opportunities for surgical treatment. Therefore, FOLFIRINOX and gemcitabine plus nab-paclitaxel are considered to be the main treatment strategies for patients with advanced disease (4). However, in PDAC, extracellular matrix accounts for 90% of the total tumor mass (5). As a physical barrier, the dense matrix can prevent the penetration of conventional chemotherapy drugs, and infiltrated immune cells in the tumor microenvironment might lead to poor chemotherapy efficacy (6). Fortunately, with the deepened understanding of the key role of the immune system in the occurrence and development of pancreatic cancer, increasing research indicates that immunotherapy has become a promising therapy for PDAC (7-9). However, recent research found that the efficacy of immunotherapy is affected by many factors such as tumor microenvironment and tumor mutational burden (TMB), hence only a part of PDAC patients could benefit from immune checkpoint inhibitors [programmed death-ligand 1 (PD-L1) and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4)], chimeric antigen receptor (CAR) T cells, immunomodulators, and vaccines (10,11). Therefore, it is urgent to find effective biomarkers to evaluate the immunotherapy and prognosis of PDAC, and provide novel ideas for the development of tumor-targeted adjuvant drugs.

Herein, we identify the immune gene, integrin subunit alpha 3 (ITGA3), highly associated with prognosis in multiple datasets. Receiver operating characteristic (ROC) curve and concordance index (C-index) indicated that ITGA3 had excellent ability to predict prognosis of PDAC. Furthermore, we analyzed ITGA3 expression associated tumor infiltrating immune cells (TIICs) by Cell-Type Identification by Estimating Relative Subsets of RNA Transcripts (CIBERSORT) based on The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) cohorts. Finally, we evaluated the reactivity of ITGA3 expression to chemotherapy and immunotherapy in PDAC patients. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0299/rc).

Methods

Data sources

GEO

The gene expression profiles of PDAC patients with survival data were searched in the GEO database (https://www.ncbi.nlm.nih.gov/geo/). Finally, five datasets, GSE28735 (45 normal samples and 45 tumor samples), GSE57495 (63 tumor samples), GSE62452 (61 normal samples and 69 tumor samples), GSE78229 (50 tumor samples) and GSE85916 (80 tumor samples) were included in the study. These five datasets all adopted homo sapiens annotation files for gene annotation of probes. The “SVA” R package was used to merge and standardize the five datasets (12). The normalize result was validated using principal component analysis (PCA). Eventually, 307 tumor specimens and 106 normal specimens were obtained for subsequent analysis. Data from the GEO databases used in this study are freely available, and details of those data are shown in the appendix available at https://cdn.amegroups.cn/static/public/tcr-2026-1-0299-1.xlsx.

TCGA

The transcriptome expression profiles, somatic mutation information and clinical information of PDAC were obtained from the TCGA (https://cancergenome.nih.gov/). The transcriptome expression data [transcripts per million (TPM)] were downloaded from TCGA, including 179 tumor specimens and 4 normal specimens. TMB was acquired from the from the calculation of tumor-specific mutated genes. The cases without follow-up data and normal samples were excluded, and 178 cases were finally included for subsequent analysis in this study. Data from the TCGA databases used in this study are freely available, and details of those data can be found in the available at https://cdn.amegroups.cn/static/public/tcr-2026-1-0299-2.xlsx.

Identification of differentially expressed immune-related genes

The “limma” R package was used to analyze the expression data of five standardized datasets from the GEO database, as well as selected differentially expressed genes (DEGs) between groups with adjusted P<0.05 and |log₂fold change (FC)| >1 (13). The immune-related genes were acquired from the InnateDB (https://www.innatedb.com/) and ImmPort (https://www.immport.org/shared/home) databases. After the duplication was removed, 2,660 immune-related genes were retained (appendix available at https://cdn.amegroups.cn/static/public/tcr-2026-1-0299-3.xlsx). The abnormal expression of immune-related genes in PDAC was revealed by integrating DEGs with immune-related genes.

Protein-protein interaction (PPI) analysis

The PPI network of differentially expressed immune-related genes were established using the STRING database (http://string-db.org/), and then calculated the degree of nodes. Following this, the central nodes were identified and visualized using “igraph” R package (14).

Functional annotation analysis

The “clusterProfiler” R package was used for functional enrichment analysis, including biological processes (BPs), cell components (CCs), molecular functions (MFs), and Kyoto Encyclopedia of Genes and Genomes (KEGG). Screening for enrichment information, P<0.05 was considered statistically significant.

Weighted gene co-expression network analysis

The immune-related gene coexpression networks were constructed through the “WGCNA” R package (15). The soft-thresholding power (β) was determined based on the scale-free topology criterion using the “sft$powerEstimate” function. When β=7, the scale-free topology fit index (R²) approached 0.85 while maintaining acceptable mean connectivity; therefore, β=7 was selected as the optimal parameter. A signed network was constructed to preserve the directionality of gene co-expression relationships. The adjacency matrix was then transformed into a topological overlap matrix (TOM) to measure network connectivity. The minimum module size was set to 10. To merge similar modules, a cut height of 0.25 (mergeCutHeight =0.25) was applied, corresponding to merging modules with eigengene correlation greater than 0.75. Finally, key modules were identified based on the Pearson correlation between module eigengenes (MEs) and clinical traits.

Identification of hub genes highly associated with prognosis

First, the hub gene was identified by intersection of the differentially expressed immune-related genes and the key module genes using the Wayne diagram tool (http://bioinformatics.psb.ugent.be/webtools/Venn/). Next, univariate and multivariate Cox regression analysis was used to identify genes associated with prognosis using the GEO dataset. In order to visualize the results, we divided the patients into high and low expression groups respectively according to calculated survival cut-off value by ROC curve. Subsequently, survival curves were plotted using the Kaplan-Meier method. In addition, the prognostic value of key genes was reconfirmed in TCGA database, and prognostic ability was evaluated by ROC and C-index.

Immune infiltration analysis

The “CIBERSORT” R package was utilized to illustrate the immune cell patterns of PDAC samples from TCGA and GEO databases, respectively (16). With 1,000 aligned default signature matrices, the relative proportions of 22 infiltrating immune cells in tumor samples and CIBERSORT P values were obtained using Monte Carlo sampling. Only samples with a CIBERSORT value of P<0.05 were included in the subsequent analysis.

Evaluation of chemotherapy and immunotherapy responses

We used the “pRRophetic” R package (version 0.5) to calculate the half-maximal inhibitory concentration (IC₅₀) of chemotherapeutic agents and evaluate the chemotherapy responses of hub gene (17). To explore the relationship between the hub gene and immunotherapy responses, the correlation between key gene expression levels and TMB was assessed in the TCGA cohort. We further investigated the association between key genes and immune checkpoints (PD-L1 and CTLA-4). TMB was defined as the total number of somatic, non-synonymous mutations per megabase (Mb) based on mutation data from the TCGA cohort. Synonymous mutations were excluded to focus on coding region alterations. The expression levels of PD-L1 and CTLA-4 were obtained from RNA sequencing (RNA-seq) data in TCGA, representing mRNA expression levels. For subsequent analyses, patients were stratified into high and low expression groups according to the optimal cutoff values determined by ROC curve analysis. The prediction model was trained on pharmacogenomic data from the Genomics of Drug Sensitivity in Cancer (GDSC) database. Ten-fold cross-validation was applied to ensure model stability. Default parameters were used unless otherwise specified.

Gene set enrichment analysis (GSEA)

GESA was carried out to estimate whether a priori specified collection of genes revealed substantially different expression between high and low hub gene expression groups, so as to confirm the impact of hub gene expression on pathway-level modifications of pancreatic cancer patients, and adjusted P<0.05 indicated statistical significance. The analyses were conducted using the MSigDB Hallmark gene set collection. The ranking metric was based on the log₂FC of gene expression between high and low ITGA3 groups, and gene sets with FDR q value <0.25 were considered significantly enriched, following standard GSEA guidelines.

Patients and design

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Medical Ethics Committee of Zhongda Hospital Affiliated to Southeast University (No. 2024ZDSYLL378-P01), and written informed consent was obtained from all participants or their legal guardians. The study included patients who were treated at the Surgical Oncology Department of Zhongda Hospital Affiliated to Southeast University between 2022.8 and 2024.6. The participants were patients with pathologically confirmed primary PDAC and were excluded from association with other oncologic diseases.

For patients who completed enrollment, we collected baseline data for all cases. Specifically, this included gender, age, tumor stage and grade, and suspiciousness of regional lymph nodes. Tissue samples were obtained by tumor resection or core needle biopsy during surgery, then fixed with fixative and paraffin-embedded.

Immunohistochemistry (IHC) and immunofluorescence (IF) staining

The paraffin-embedded blocks were cut into slides for staining. The tissue slides were subjected to heat-induced antigen retrieval using citrate buffer (pH 6.0). Endogenous peroxidase activity was blocked with 30% H₂O₂, followed by blocking with 3% bovine serum albumin (BSA) for 30 minutes. For IHC/IF, the following primary antibodies were applied at 4 ℃ overnight: anti-ITGA3 rabbit polyclonal (Abcam, catalog ab131055; dilution 1:200); anti-CD206/MRC1 rabbit monoclonal (Cell Signaling Technology, clone E2L9N #91992; dilution 1:200); anti-PD-L1 rabbit monoclonal (Abcam; clone 28-8; catalog ab205921; dilution 1:100–1:200); anti-CD8 mouse monoclonal (e.g., clone C8/144B; Novus Biologicals; prediluted or ~1:100); and anti-CD19 mouse monoclonal (Cell Signaling Technology; catalog #90176; dilution ~1:100). Following primary incubation, for IHC, species‑appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies were added and incubated at room temperature for 30 minutes and visualized with 3,3'-diaminobenzidine (DAB); for IF, fluorescence-conjugated secondary antibodies (e.g., Alexa Fluor 488/594) were applied. Positive signals were analyzed and quantified using ImageJ across representative fields to assess relative expression levels for each marker.

Statistical analysis

Continuous variables were expressed as median [interquartile range (IQR)], and comparison was made by Wilcoxon test. Differences in overall survival (OS) between groups were compared by log-rank test. To identify the independent risk factors of OS, only variables with P<0.05 in the univariate Cox regression analysis were included into the multivariate Cox regression analysis. Differences were considered statistically significant when P<0.05. Statistical analysis in this study was carried out using R software (version 4.1.0).

Results

Identifying DEGs and functional annotation

The scattered graphs based on PCA of normalized sequencing after batch effect removal demonstrated that cross-platform normalization had successfully removed the batch effect (Figure 1A,1B). Then, 307 tumor samples and 106 normal samples were obtained by merging five datasets from the GEO database (Figure 1C). Subsequently, we explored DEGs between groups. With adjusted P<0.05 and |log₂FC| >1, 310 DEGs were obtained (Figure 1D). After screening of 2,660 immune-related genes, 60 genes were finally obtained, among which 36 genes were significantly up-regulated and 24 genes were significantly down-regulated in tumor samples compared with normal samples (Figure 1E). Then, we analyzed the Gene Ontology (GO) and KEGG pathways to determine the function of the 60 DEGs. Two hundred and fifteen GO terms and 3 KEGG enriched pathways were identified (appendix available at https://cdn.amegroups.cn/static/public/tcr-2026-1-0299-4.xlsx). Figure 1F,1G present the first 12 GO terms and all 3 enrich pathways, respectively. KEGG enriched pathways indicated that these genes were involved in complement and coagulation cascades, cytokine-cytokine receptor interaction, and viral protein interaction with cytokine and cytokine receptor pathway. Besides, GO enrichment analysis demonstrated that these genes were enriched in cytolysis, cytokine-mediated signaling pathway, myeloid leukocyte migration, receptor internalization, type I interferon signaling pathway, etc. Finally, we used STRING database to establish the PPI network of those genes, and the network of the genes with a threshold weight >0.4 was showed in Figure 1H.

Figure 1 Identifying DEGs and further analysis. PCA of the datasets for gene expression. The scattered graphs represent the samples based on the top two principal components of the gene expression consequences without (A) and with (B) the batch effect removed. (C) Five data from the GEO database were merged. (D) Volcano plot of DEGs between tumor samples and normal samples. (E) Volcano plot of differentially expressed immune-related genes between tumor samples and normal samples. (F,G) GO and KEGG enriched analyses of differentially expressed immune-related genes. (H) The network of the genes in differentially expressed immune-related genes (weight of edge >0.4). DEGs, differentially expressed genes; FC, fold change; FDR, false discovery rate; GEO, Gene Expression Omnibus; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; N, normal samples; PCA, principal component analysis; T, tumor samples.

WGCNA

To identify the immune-related hub genes, WGCNA was carried out on the 2660 immune-related genes downloaded from ImmPort and InnateDB. At first, the “Hclust” function was utilized for sample clustering analysis (307 tumor samples and 106 normal samples from the GEO database), no outliers were observed (Figure S1). Then, “sft$powerEstimate” was used to calculate the soft threshold β (Figure 2A). After TOM network construction and with height as 0.25 (Figure 2B), 11 immune-related gene modules were detected (Figure 2C). Finally, the magenta module was identified as having the highest significant positivity of correlation between MEs and features (Cor =0.64, P<0.001) (Figure 2D). The magenta module contained 247 genes. Module membership (MM) and gene significance (GS) of the magenta module also showed significant relationship (Cor =0.66, P<0.001) (Figure 2E).

Figure 2 Gene modules identified by WGCNA. (A) Network topology analysis for various soft-thresholding powers. (B) Clustering of MEs. The red line indicates cut height (0.25). (C) Dendrogram of immune-related genes clustered based on a dissimilarity measure (1 − TOM). (D) Heatmap of correlation between the MEs and trait. (E) Scatter plot for correlation between gene MM in the magenta module and GS. GS, gene significance; ME, module eigengene; MM, module membership; TOM, topological overlap matrix; WGCNA, weighted gene co-expression network analysis.

Identification of hub genes highly associated with prognosis

To identify important genes in PDAC, the intersection of the differentially expressed immune-related genes and the magenta module genes was initially selected, and 30 genes were eventually obtained (Figure 3A). Then, univariate Cox regression analysis indicated that 11 genes were significantly correlated with prognosis in the five merged data from the GEO database [ITGA3: hazard ratio (HR) =1.38, P<0.001; CCL20: HR =1.12, P=0.02; SLPI: HR =1.18, P=0.003; S100A16: HR =1.28, P=0.007; PLAU: HR =1.30, P<0.001; DKK1: HR =1.18, P=0.001; GREM1: HR =1.19, P=0.01; IL1RAP: HR =1.24, P=0.005; MET: HR =1.35, P<0.001; MMP7: HR =1.15, P=0.003; C7: HR =0.87, P<0.001] (Figure 3B). Kaplan-Meier survival curves showed that patients with high ITGA3 expression had a worse prognosis than patients with low ITGA3 expression (Figure 3C). By including these 11 genes into multivariate Cox regression analysis, only ITGA3 was identified to be an independent risk factor for OS (ITGA3: HR =1.30, P=0.04) (Figure 3D). Furthermore, through the Gene Expression Profiling Interactive Analysis 2 (GEPIA2) (http://gepia.cancer-pku.cn/index.html), we demonstrated that ITGA3 had a significant relationship with OS and disease-free survival (DFS) (OS: HR =1.8, P=0.007; DFS: HR =2.4, P<0.001) (Figure 3E,3F). To further verify the accuracy of ITGA3 in predicting survival, we found that the ROC curves in the 178 TCGA pancreatic cancer cohort had the 1-year survival area under the curve (AUC) value of 0.717, the 3-year survival AUC value of 0.772, and the 5-year survival AUC value of 0.743 (Figure 3G). Intriguingly, comparison of the AUC value and C-index revealed that ITGA3 had a higher predictive efficacy compared to other variables (age, gender, grade, and stage) (Figure 3H,3I).

Figure 3 Identification of hub genes highly associated with prognosis. (A) Venn plot of differentially expressed immune-related genes and the magenta module genes. (B) Univariate Cox regression analysis of genes with prognostic impact. (C) Kaplan-Meier analysis showing the association of ITGA3 expression with OS in 178 pancreatic cancer TCGA cohort. (D) Multivariate Cox regression analysis of genes with prognostic impact. (E,F) Kaplan-Meier analysis showing the association of ITGA3 expression with OS and DFS in pancreatic cancer patients through the GEPIA2 database. (G) ROC curves of ITGA3 expression in the prediction of OS at 1-, 3-, and 5-year. (H,I) ROCs and C-index of ITGA3 expression and other clinical prognostic factors for OS. AUC, area under the curve; C-index, concordance index; CI, confidence interval; DEGs, differentially expressed genes; DFS, disease-free survival; GEPIA2, Gene Expression Profiling Interactive Analysis 2; HR, hazard ratio; ITGA3, integrin subunit alpha 3; OS, overall survival; ROC, receiver operating characteristic; TCGA, The Cancer Genome Atlas; WGCNA, weighted gene co-expression network analysis.

Patterns of TIICs related to ITGA3 expression

First, we explored whether ITGA3 expression was related to the 22 TIICs in GEO cohort. After filtering with CIBERSORT P<0.05, 282 of 307 pancreatic cancer samples were retained for subsequent analysis, while 25 samples were excluded due to low deconvolution confidence and the patterns of TIICs was analyzed in Figure 4A. Then, we explored the difference of TIICs with the high ITGA3 expression group and the low ITGA3 group. Obviously, the high ITGA3 expression group had a lower B cell naive and CD8⁺ T cells compared with the low ITGA3 expression group (P=0.002; P=0.001). On the contrary, M0 macrophages and M2 macrophages were found to be significantly higher (P<0.001; P=0.02) (Figure 4B). Whereafter, we verified the above results in the TCGA database, and the same pipeline was used to elucidate the patterns of TIICs. Ultimately, 98 of 178 pancreatic cancer samples were retained for subsequent analysis, while 80 samples were excluded based on the CIBERSORT P<0.05 threshold. Figure 4C displayed the landscape of TIICs. Similarly, the infiltrations of B cells naive and CD8⁺ T cells were statistically lower in high expression group (P=0.05; P<0.01), whereas infiltrations of M0 macrophages was higher in high expression group (P=0.007) (Figure 4D).

Figure 4 Immune infiltration pattern in PDAC patients. (A) Landscape of immune infiltration in the PDAC GEO cohort. (B) Analysis of TIICs related to ITGA3 expression in the PDAC GEO cohort. (C) Landscape of immune infiltration in the PDAC TCGA cohort. (D) Analysis of TIICs related to ITGA3 expression in the PDAC TCGA cohort. GEO, Gene Expression Omnibus; ITGA3, integrin subunit alpha 3; NK, natural killer; PDAC, pancreatic ductal adenocarcinoma; TCGA, The Cancer Genome Atlas; TIICs, tumor-infiltrating immune cells.

Response of ITGA3 expression to chemotherapy and immunotherapy

We had identified ITGA3 as an important immune gene that affects the prognosis of PDAC. However, the response of ITGA3 expression to chemotherapy and immunotherapy remained unclear, which is unfavorable for the application of ITGA3 in clinical treatment. Therefore, we first predicted the efficacy of common chemotherapy drugs for PDAC in the high and low ITGA3 expression groups by pRRophetic algorithm. As shown in Figure 5. Surprisingly, five drugs (5-fluorouracil, cisplatin, gemcitabine, paclitaxel, and mitomycin C) had lower IC₅₀ in the high ITGA3 expression group (all P<0.05; Figure 5F,5H-5K), suggesting that the high ITGA3 expression patients were more sensitive to these five drugs. On the contrary, camptothecin had a lower IC₅₀ in the low ITGA3 expression group, suggesting that the low ITGA3 expression patients were more sensitive to camptothecin (all P<0.05) (Figure 5G).

Figure 5 Response of ITGA3 expression in PDAC to chemotherapy and immunotherapy. (A,B) Kaplan-Meier analysis showed that patients with high TMB and high ITGA3 levels had worse OS. (C-E) The correlation of ITGA3 expression with TMB and immune checkpoint marker (PD-L1 and CTLA-4) expression. (F-K) Sensitivity analysis of six common chemotherapeutic drugs (5-fluorouracil, camptothecin, cisplatin, gemcitabine, paclitaxel, and mitomycin C) in patients at high and low ITGA3 expression. *, P<0.05; **, P<0.01; ***, P<0.001. CTLA-4, cytotoxic T-lymphocyte-associated protein 4; H, high; IC₅₀, half-maximal inhibitory concentration; ITGA3, integrin subunit alpha 3; L, low; OS, overall survival; PD-L1, programmed death-ligand 1; PDAC, pancreatic ductal adenocarcinoma; TMB, tumor mutational burden.

To explore the relationship between ITGA3 expression and immunotherapy responses, we compared the correlation of ITGA3 expression with TMB and immune checkpoint marker (PD-L1 and CTLA-4) levels in the TCGA cohort. TMB was calculated by tumor-specific mutant genes from the TCGA database (appendix available at https://cdn.amegroups.cn/static/public/tcr-2026-1-0299-5.xlsx). First, the patients were divided into two groups using cut-off value calculated by ROC curve, and the TMB high patients had significantly worse OS compared to TMB low patients (P=0.008) (Figure 5A). Moreover, ITGA3 could further reduce the OS rate of patients (P<0.001) (Figure 5B). Figure 5C displayed that patients with high ITGA3 expression had higher TMB and PD-L1 than those with low ITGA3 expression, indicating that patients with high ITGA3 expression might respond better and had better outcome when receiving immune checkpoint inhibitors (all P<0.05) (Figure 5C,5E). However, there was no significant difference in CTLA-4 expression between groups (P=0.97) (Figure 5D).

GSEA of different ITGA3 expression subgroups

GSEA was carried out in the TCGA cohort between high and low ITGA3 expression patients to further confirm the role of ITGA3 in PDAC. The gene sets of the high ITGA3 group were enriched in tumor proliferation, metabolism, and metastasis-related pathways [E2F targets, epithelial-mesenchymal transition (EMT), G2M checkpoint, glycolysis, hypoxia, interferon-alpha response, etc.] (Figure 6A). While the gene sets of the low ITGA3 samples were enriched in bile acid metabolism, KRAS signaling and pancreas beta cells (Figure 6B).

Figure 6 GSEA of different ITGA3 expression subgroups. (A) Gene sets enriched in the ITGA3 high subgroup. (B) Gene sets enriched in the ITGA3 low subgroup. GSEA, gene set enrichment analysis; ITGA3, integrin subunit alpha 3.

The expression of ITGA3 exhibits a correlation with PDAC progression in tumor tissues of PDAC patients

Next, we sought to validate the correlation between ITGA3 and PDAC progression in tumor tissue samples from PDAC patients. We collected clinical data from eligible PDAC patients and performed immunohistochemical staining in pathological tissues obtained by surgical resection or biopsy.

Patient characteristics

A total of 132 patients were enrolled in the study between 2022.8 and 2024.6. The results showed that there was a higher percentage of males and patients aged 65 years; the tumor stage was approximately equally divided between those in T1–T2 and those in T3–T4, with similar occurrences in tumor grade G1–G2 and G3; in addition, the majority of cases did not present with obvious suspicious local lymph nodes. Specific data characteristics are shown in Table 1.

Immunohistochemical staining reveals the association of ITGA3 with tumor immune response

First, we applied immunohistochemical staining to detect the expression of ITGA3 and classified 132 PDAC patients into the low ITGA3-expressing group (n=63) and the high ITGA3-expressing group (n=69) based on the intensity of expression (Table 1, Figure 7A). We then performed IF staining of tumor section tissues to determine the correlation between the expression of ITGA3-positive signals and immune-related molecules. The results showed that the fluorescence intensity of the proteins CD206 and PD-L1 was also evident in fixed regions where ITGA3-positive signals were evident, although they were not always present in the same smaller unit of cells. In contrast, ITGA3 expression tended to show an opposite trend to CD8 and CD19 (Figure 7B).

Figure 7 ITGA3 is associated with immune effects in tumors. (A) Immunohistochemical staining for ITGA3 of tissues from PDAC patients. (B) IF staining against CD206, PD-L1, CD8, and CD19 antibodies on the tissue slides from PDAC patients. Scale bar =50 µm. DAPI, 4’,6-diamidino-2-phenylindole; IF, immunofluorescence; ITGA3, integrin subunit alpha 3; PD-L1, programmed death-ligand 1; PDAC, pancreatic ductal adenocarcinoma.

Discussion

In recent years, with the rapid development of molecular biology and high-throughput RNA-seq, the contribution of the tumor microenvironment to cancer development and metastasis has been increasingly recognized (18-20). There is a complex dynamic crosstalk between tumor and non-cancer cells (infiltrating immune cells and adjacent stromal cells). Increasing studies indicate that the prognosis of cancer patients is significantly influenced by the tumor microenvironment (21,22).

Compared with other types of cancers, PDAC has a dense extracellular matrix and a highly immunosuppressive tumor microenvironment, suggesting that immunotherapy might be a promising therapy for PDAC (23). The TMB, PD-L1, and CTLA-4 are considered biomarkers to predict the efficacy of immunotherapy, but not sensitive for all patients (24). Some patients can still benefit from immunotherapy, even if they carry low TMB or immune checkpoint levels. Therefore, it is important to explore the impact of tumor immune microenvironment on the prognosis of PDAC and to identify patients who could benefit from immunotherapy. This can help to understand the pathogenesis of pancreatic cancer and to guide clinical anti-tumor therapy.

Firstly, five gene expression profiles about PDAC were searched and downloaded from the GEO database, with all of these datasets having survival data. Then, these datasets were merged and standardized using “SVA” R package, and 307 tumor samples and 106 normal samples were obtained. To acquire the immune-related genes, we downloaded them from InnateDB and ImmPort databases. Combining DEGs and WGCNA, we obtained 30 immune genes that were significantly related to the occurrence and development of PDAC. Further, using Cox regression analysis to identify genes associated with prognosis, we found that ITGA3 was an independent risk factor for OS (HR =1.30, P=0.04). Surprisingly, the significant correlation between ITGA3 expression and the prognosis of PDAC was also observed in the TCGA cohort (OS: HR =1.8, P=0.007; DFS: HR =2.4, P<0.001). Besides, the AUC and C-index of ITGA3 were significantly higher than other prognostic indicators (age, gender, grade, stage), suggesting that ITGA3 is a marker highly correlated with the prognosis of PDAC. Subsequently, we investigated the expression of ITGA3 in association with 22 TIICs in PDAC. We found that the high ITGA3 expression group had a lower amount of naive B cells and CD8⁺ T cells, while macrophages were found to be significantly more in GEO cohort and TCGA cohort, respectively (all P<0.05), suggesting that patients with high ITGA3 expression were more prone to immune escape, which might contribute to the poorer prognosis.

Although we identified ITGA3 as an important gene that affects immune escape and prognosis in PDAC, whether ITGA3 could be utilized as a biomarker for predicting the efficacy of immunotherapy remains unclear. Therefore, we compared the correlation between ITGA3, TMB, and immune checkpoint (PD-L1 and CTLA-4) levels in the TCGA cohort, respectively. We found that patients with high ITGA3 expression had higher TMB and PD-L1 levels than those with low ITGA3 expression, suggesting the value of ITGA3 expression in predicting immunotherapy response. Although tumors with high ITGA3 expression exhibit elevated TMB and PD-L1 levels, features generally associated with improved responsiveness to immune checkpoint inhibitors, the concurrent enrichment of macrophages and reduction of CD8⁺ T-cell infiltration suggest the presence of a highly immunosuppressive and stromal-rich microenvironment that physically and functionally limits T-cell penetration. Similarly, the increased sensitivity to chemotherapeutic agents predicted by the pRRophetic algorithm likely reflects tumor-intrinsic susceptibility, whereas the poor clinical outcomes may be driven by extrinsic factors such as dense extracellular matrix, impaired drug delivery, and aggressive tumor biology, as supported by the enrichment of proliferation- and metastasis-related pathways in the high ITGA3 group. From a clinical perspective, these findings suggest that ITGA3 may serve as a dual biomarker reflecting both high-risk disease and potential therapeutic vulnerability. Patients with high ITGA3 expression may benefit from combination treatment strategies, such as immune checkpoint blockade combined with stromal modulation or macrophage-targeting therapies, rather than monotherapy alone. Therefore, integrating ITGA3 expression with other microenvironmental and clinical indicators may provide a more precise framework for therapeutic decision-making in PDAC. In addition, we predicted the efficacy of common chemotherapy drugs for pancreatic cancer in different ITGA3 expression group by pRRophetic algorithm. The results indicated that PDAC with high ITGA3 expression were more sensitive to 5-fluorouracil, cisplatin, gemcitabine, paclitaxel, and mitomycin C. On the contrary, PDAC with low ITGA3 expression were more sensitive to camptothecin (all P<0.05). Notably, an apparent inconsistency was observed between the increased predicted sensitivity to chemotherapeutic agents and the poorer prognosis in the high ITGA3 expression group. This discrepancy may be attributed to the limitations of in silico drug sensitivity prediction models. Specifically, the pRRophetic algorithm is based on cancer cell line data and primarily reflects tumor-intrinsic drug responses, without accounting for the complex tumor microenvironment in vivo. In PDAC, the dense desmoplastic stroma and impaired vascularization can significantly hinder drug delivery and reduce therapeutic efficacy. Moreover, our GSEA results demonstrated that the high ITGA3 expression group was enriched in proliferation- and metastasis-related pathways, such as EMT and cell cycle progression, indicating a more aggressive tumor phenotype that may override the intrinsic chemosensitivity and lead to worse clinical outcomes.

ITGA3 belongs to the integrin family. ITGA3 binds to the β1 subunit and acts as a receptor for fibronectin, laminin, and collagen. It was reported that ITGA3 could promote the remodeling of the ECM, contributing to cancer metastasis (25). Besides, Chen et al. showed that ITGA3 promotes cancer development and is associated with tumor proliferation, migration and invasion, possibly through activation of PI3K-Akt signaling (26). Jiao et al. found that ITGA3 expression was correlated with histological grade, stage and prognosis, which was consistent with our study (27). Furthermore, previous studies have shown that ITGA3 could mediate M2 polarization and promote the stiffness of the ECM (28,29). Our present study demonstrates that ITGA3 could recruit macrophages, reduce the infiltration of B cells and CD8⁺ T cells, and cause tumors to evade immune surveillance. ITGA3 might be a potential prognostic marker for immunotherapy, while further studies are needed to verify it.

Finally, we applied immunohistochemical to the validation of ITGA3. CD206 is a characteristic macrophage marker that is often associated with immunosuppression-related microenvironmental features. Meanwhile, PD-L1 is considered an important mechanism for tumor cells to overcome immune surveillance. In the present study, we observed that the expression of ITGA3 showed consistency with the expression levels of CD206 and PD-L1, suggesting that ITGA3 may be part of the tumor-promoting mechanism that contributes to the formation of an immunosuppressive environment, which leads to the immune escape of tumor cells. On the contrary, we observed a negative correlation between the expression trend of ITGA3 and the expression of CD8 and CD19. CD8 T cells are the main immune cells exerting effects in the tumor microenvironment, and their activity directly affects the effectiveness of the tumor immune response, while CD19 is a marker of B cells, reflecting the degree of activity of humoral immunity. High expression of ITGA3 may inhibit the infiltration or function of CD8 T cells and B cells, further suggesting its specific role in regulating the tumor microenvironment. Combined with these observations, we speculate that ITGA3 may promote immune escape and tumor growth by inhibiting the infiltration of effector immune cells.

However, there are still some limitations of this study that need to be addressed. First, although we identified ITGA3 as a potential prognostic marker and explored its role in the tumor immune microenvironment, our conclusions were primarily based on bioinformatics analyses of publicly available datasets. Therefore, the findings may be limited by the quality and heterogeneity of the original datasets. Second, although we validated ITGA3 expression through immunohistochemical analysis, further experimental validation in a larger cohort of clinical samples is needed to strengthen our findings and confirm the role of ITGA3 in immune escape and its association with immunotherapy efficacy. Third, this study focused on the correlation between ITGA3 and immune cell infiltration but did not fully elucidate the underlying molecular mechanisms by which ITGA3 regulates immune responses in the tumor microenvironment. Future studies should aim to clarify these mechanisms, preferably using in vivo and in vitro models. In addition, the prediction of chemotherapeutic sensitivity in this study was performed using the pRRophetic algorithm, which is trained on pharmacogenomic data derived from cancer cell lines (e.g., the GDSC database). Therefore, these results primarily reflect tumor-intrinsic drug responses and may not fully capture the complexity of in vivo tumor behavior, particularly in PDAC, where stromal barriers and tumor microenvironment play critical roles in drug delivery and therapeutic efficacy. Consequently, these findings should be interpreted with caution, and further validation in well-designed prospective clinical studies is required before clinical application. Finally, the predictive value of ITGA3 for immunotherapy response requires prospective clinical validation to determine its true potential as a treatment-guiding biomarker for PDAC patients.

Conclusions

In conclusion, our study indicates that ITGA3 is a potentially valuable immune-related biomarker that can distinguish immunological and molecular features of PDAC and predict its prognosis.

Acknowledgments

All authors thank the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) databases for providing publicly available data. We are grateful to all patients and investigators who contributed to the independent clinical cohort. We also acknowledge the support of our institutions and colleagues for their valuable discussions and technical assistance.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0299/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0299/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0299/prf

Funding: This project was supported by the General Surgery Innovation Team of Jiangsu Province (No. 2017ZXKJXW09).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0299/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. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Medical Ethics Committee of Zhongda Hospital Affiliated to Southeast University (No. 2024ZDSYLL378-P01), and written informed consent was obtained from all participants or their legal guardians.

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. GBD 2017 Pancreatic Cancer Collaborators. The global, regional, and national burden of pancreatic cancer and its attributable risk factors in 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet Gastroenterol Hepatol 2019;4:934-47. [Crossref] [PubMed]
2. Siegel RL, Miller KD, Fuchs HE, et al. Cancer Statistics, 2021. CA Cancer J Clin 2021;71:7-33. [Crossref] [PubMed]
3. Hidalgo M, Von Hoff DD. Translational therapeutic opportunities in ductal adenocarcinoma of the pancreas. Clin Cancer Res 2012;18:4249-56. [Crossref] [PubMed]
4. Mizrahi JD, Surana R, Valle JW, et al. Pancreatic cancer. Lancet 2020;395:2008-20. [Crossref] [PubMed]
5. Kong X, Li L, Li Z, et al. Targeted destruction of the orchestration of the pancreatic stroma and tumor cells in pancreatic cancer cases: molecular basis for therapeutic implications. Cytokine Growth Factor Rev 2012;23:343-56. [Crossref] [PubMed]
6. Gu J, Saiyin H, Fu D, et al. Stroma - A Double-Edged Sword in Pancreatic Cancer: A Lesson From Targeting Stroma in Pancreatic Cancer With Hedgehog Signaling Inhibitors. Pancreas 2018;47:382-9. [Crossref] [PubMed]
7. Banerjee K, Kumar S, Ross KA, et al. Emerging trends in the immunotherapy of pancreatic cancer. Cancer Lett 2018;417:35-46. [Crossref] [PubMed]
8. DeSelm CJ, Tano ZE, Varghese AM, et al. CAR T-cell therapy for pancreatic cancer. J Surg Oncol 2017;116:63-74. [Crossref] [PubMed]
9. Akce M, Zaidi MY, Waller EK, et al. The Potential of CAR T Cell Therapy in Pancreatic Cancer. Front Immunol 2018;9:2166. [Crossref] [PubMed]
10. Wang W, Xu Z, Wang N, et al. Prognostic value of eight immune gene signatures in pancreatic cancer patients. BMC Med Genomics 2021;14:42. [Crossref] [PubMed]
11. Wu J, Cai J. Dilemma and Challenge of Immunotherapy for Pancreatic Cancer. Dig Dis Sci 2021;66:359-68. [Crossref] [PubMed]
12. Leek JT, Johnson WE, Parker HS, et al. The sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics 2012;28:882-3. [Crossref] [PubMed]
13. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015;43:e47. [Crossref] [PubMed]
14. Csardi G, Nepusz T. The igraph software. Complex Syst 2006;1695:1-9.
15. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 2008;9:559. [Crossref] [PubMed]
16. 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]
17. Geeleher P, Cox NJ, Huang RS. Clinical drug response can be predicted using baseline gene expression levels and in vitro drug sensitivity in cell lines. Genome Biol 2014;15:R47. [Crossref] [PubMed]
18. Garner H, de Visser KE. Immune crosstalk in cancer progression and metastatic spread: a complex conversation. Nat Rev Immunol 2020;20:483-97. [Crossref] [PubMed]
19. Joyce JA, Fearon DT. T cell exclusion, immune privilege, and the tumor microenvironment. Science 2015;348:74-80. [Crossref] [PubMed]
20. Tang HD, Wang Y, Xie P, et al. The Crosstalk Between Immune Infiltration, Circulating Tumor Cells, and Metastasis in Pancreatic Cancer: Identification of HMGB3 From a Multiple Omics Analysis. Front Genet 2022;13:892177. [Crossref] [PubMed]
21. Budhu A, Forgues M, Ye QH, et al. Prediction of venous metastases, recurrence, and prognosis in hepatocellular carcinoma based on a unique immune response signature of the liver microenvironment. Cancer Cell 2006;10:99-111. [Crossref] [PubMed]
22. Chew V, Chen J, Lee D, et al. Chemokine-driven lymphocyte infiltration: an early intratumoural event determining long-term survival in resectable hepatocellular carcinoma. Gut 2012;61:427-38. [Crossref] [PubMed]
23. Jiang J, Zhou H, Ni C, et al. Immunotherapy in pancreatic cancer: New hope or mission impossible? Cancer Lett 2019;445:57-64. [Crossref] [PubMed]
24. Carbone DP, Reck M, Paz-Ares L, et al. First-Line Nivolumab in Stage IV or Recurrent Non-Small-Cell Lung Cancer. N Engl J Med 2017;376:2415-26. [Crossref] [PubMed]
25. Kawataki T, Yamane T, Naganuma H, et al. Laminin isoforms and their integrin receptors in glioma cell migration and invasiveness: Evidence for a role of alpha5-laminin(s) and alpha3beta1 integrin. Exp Cell Res 2007;313:3819-31. [Crossref] [PubMed]
26. Chen WL, Wang XK, Wu W. Identification of ITGA3 as an Oncogene in Human Tongue Cancer via Integrated Bioinformatics Analysis. Curr Med Sci 2018;38:714-20. [Crossref] [PubMed]
27. Jiao Y, Li Y, Liu S, et al. ITGA3 serves as a diagnostic and prognostic biomarker for pancreatic cancer. Onco Targets Ther 2019;12:4141-52. [Crossref] [PubMed]
28. Li Y, Li F, Bai X, et al. ITGA3 Is Associated With Immune Cell Infiltration and Serves as a Favorable Prognostic Biomarker for Breast Cancer. Front Oncol 2021;11:658547. [Crossref] [PubMed]
29. Cavaco ACM, Rezaei M, Caliandro MF, et al. The Interaction between Laminin-332 and α3β1 Integrin Determines Differentiation and Maintenance of CAFs, and Supports Invasion of Pancreatic Duct Adenocarcinoma Cells. Cancers (Basel) 2018;11:14. [Crossref] [PubMed]