Introduction

Pancreatic adenocarcinoma (PDAC) is anticipated to become the second leading cause of cancer-related mortality by 20301, as the management of PDAC continues to present significant clinical challenges. Surgical resection continues to be the cornerstone of curative treatment for resectable PDAC. However, a significant proportion of patients are diagnosed with unresectable disease. For these patients, chemotherapy serves as the primary treatment option, typically including agents such as gemcitabine, albumin-bound paclitaxel, and fluorouracil. Emerging targeted therapies have exhibited some benefit for patients with BRCA1/BRCA2 and KRASmutations2. Nevertheless, despite these advancements, the improvement in the five-year survival rate has been modest, increasing from 2 to 11%, while the median survival remains between 10 and 12 months2. Consequently, the identification of novel therapeutic targets and the development of innovative treatment strategies are anticipated to enhance the prognosis of PDAC.

Immunotherapy has revolutionized the outcomes of many cancers, however, advancements in the management of PDAC have been limited. PDAC exhibits a marked resistance to FDA-approved immunotherapies, primarily due to its dense and immunosuppressive tumor microenvironment (TME)3,4. This TME is characterized by abundant cancer-associated fibroblasts (CAFs), tumor-associated macrophages (TAMs), regulatory T cells, and myeloid-derived suppressor cells, coupled with a notable deficiency of CD8+T cells2,4. Recent researches indicate that alterations in the TME probably influence cellular responses to immunotherapy and the prognosis of PDAC2. Consequently, contemporary strategies are focused on targeting immunosuppressive cells, recruiting immune effector cells, employing immune checkpoint inhibitors, and developing therapeutic vaccines5,6. With the advancement of precision medicine, there is an increasing demand for immune biomarkers to optimize therapeutic decision-making and improve prognostic predictions for patients with PDAC.

Traditionally, tumor prognosis has been evaluated using the TNM (tumor-node-metastasis) and pathological staging system. However, these system does not adequately incorporate additional characteristic variables, such as the immune biomarkers, and fails to fully meet the contemporary clinical requirements7. In contrast, the nomogram presents a more comprehensive prognostic model in a graphical format8. Specifically, following rigorous discrimination and validation processes, the nomogram assigns points to the variables, aggregates these scores to assess individual risk, and predicts clinical outcomes. The nomogram is extensively employed across different cancer types9,10,11and represents a significant advancement in clinical prognostication. Nevertheless, the research on immune-related nomograms in PDAC remains limited12.

This study aims to investigate the role of immune-related genes and immune microenvironments in predicting the prognosis of patients with PDAC, and to identify potential targets for immunotherapy (Fig. 1). Our findings suggest that immune-related genes can effectively distinguish between different levels of immune cell infiltration, providing a more accurate evaluation of prognostic risks than the conventional staging system. This approach offers valuable insights for therapeutic decision-making.

Fig. 1
figure 1

Overview of the general workflow.

Full size image

Materials and methods

Data collection and identification of immune-related genes

For the training cohort, RNA sequencing data from 178 PDAC samples and 4 normal samples, along with their corresponding clinical phenotypes, were downloaded from the Cancer Genome Atlas (TCGA) database (https://www.cancer.gov/tcga/). Six PDAC samples were excluded from the subsequent analysis due to deficiencies in the data. For the validation cohort, RNA sequencing data and clinical phenotypes of 317 PDAC patients were downloaded from the International Cancer Genome Consortium (ICGC) database (https://dcc.icgc.org/releases/current/Projects/PACA-CA/), with 204 patients excluded for the similar reasons. The characteristics of patients with PDAC in the TCGA and ICGC cohorts are summarized in Tables 1 and 2, respectively.

Table 1 Characteristics of patient with PDAC using the TCGA cohort.
Full size table
Table 2 Characteristics of patient with PDAC using the ICGC cohort.
Full size table

Construction and evaluation of an immune-related gene signature

A comprehensive list of 1,793 immune-related genes was downloaded from the ImmPort database (https://www.immport.org/shared/home). For the subsequent analysis, 320 genes were excluded due to the lack of RNA sequencing data in the TCGA cohort. The remaining 1,473 genes, along with their survival outcomes, were analyzed using univariate Cox regression via the “coxph” function from the “survival” R package. Hazard ratios (HRs) and P values were computed. A total of 227 genes were selected based on a significance threshold of P value < 0.05. To further refine feature selection and reduce the risk of overfitting, least absolute shrinkage and selection operator (LASSO) regression was conducted using the “cv.glmnet” function from the “glmnet” R package. This process reduced the gene candidates to 38 based on the optimal log lambda. Subsequently, multivariate Cox regression was employed to establish the immune-related gene signature, again using the “survival” R package. The risk score was calculated as the sum of the products of each gene and its corresponding coefficients within the signature. Based on the median risk score, patients were categorized into high- and low-risk groups.

To evaluate the efficacy of the immune-related gene signature, a heatmap depicting the signature alongside the distribution of survival times was created using the “pheatmap” R package. The Kaplan-Meier (KM) survival curve was constructed using the “survminer” R package. To determine the accuracy of the signature, the receiver operating characteristic (ROC) curve and the area under the curve (AUC) values were computed using the “survivalROC” R package. The decision curve analysis (DCA) was conducted using the “rmda” R package. The concordance index (C-index) was calculated employing the “survcomp” R package.

For the validation cohort, the relevant gene expression data and corresponding survival outcomes were obtained from the ICGC cohort. The heatmap, survival time distribution, K-M curve, and C-index were generated using a consistent methodology.

Construction and evaluation of a prognostic nomogram

Univariate and multivariate Cox regression analyses were performed using the “survival” R package. A total of twelve clinical phenotypes from the TCGA database were evaluated, which included the risk score derived from the immune-related gene signature, age, gender, tumor grade, pathological stage, family history of cancer, postoperative treatment, primary therapy outcome, history of chronic pancreatitis, diabetes, tobacco use, and alcohol consumption. Three clinical phenotypes were selected based on their clinical and statistical significance, with a defined threshold of P value < 0.05: the risk score, tumor stage, and postoperative treatment. The prognostic nomogram was constructed using the “survival” and “rms” R packages.

To evaluate the discrimination of the nomogram, ROC curves and AUC values were generated using the “survivalROC” R package. The calibration of the nomogram was evaluated using the calibration curves and C-index, which were computed using the “rms” and “survcomp” R packages, respectively.

For the validation cohort, similar clinical phenotypes were collected from the ICGC database. The ROC curve, AUC values, calibration curves, and C-index were generated using the same methodology.

CIBERSORT analysis

The cell-type identification by estimating relative subsets of RNA transcripts, known as CIBERSORT analysis, was performed using the “CIBERSORT” R package and RNA sequencing data obtained from the TCGA database13. The cell types were classified based on the leukocyte signature matrix (LM22), which consists of 547 genes and is capable of distinguishing 22 distinct human hematopoietic cell types. A heatmap depicting the compositions of immune cell infiltration was created using the “pheatmap” R package.

Based on the prognostic nomogram, patients were categorized into high-risk and low-risk groups. A box plot for the comparison of these groups was generated using the “ggpubr” and “ggcorrplot” R packages.

ESTIMATE and xCell analysis

The Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data (ESTIMATE) analysis was performed using the “estimate” R package and RNA sequencing data obtained from the TCGA database, with validation provided by the xCell algorithm14,15. Tumor purity, ESTIMATE scores, stromal scores, immune scores, and microenvironment scores were calculated using the gene expression profiles of stromal and immune cells. The interrelationships among these scores were analyzed and visualized using the “ggpubr” and “ggplot2” R packages.

RNA extraction and quantitative reverse transcription polymerase chain reaction (qPCR)

Written or signed informed consent was obtained from all patients for the use of paired normal and tumor tissues. RNA was extracted using the precipitation method with TriQuick Reagent (R1100, Solarbio) and grinding beads. Reverse transcription was conducted using ReverTra Ace qPCR RT Master Mix (FSQ-201, TOYOBO). qPCR was performed using Hieff® qPCR SYBR Green Master Mix (11201ES08, YEASEN) on a LightCycler® 480 Instrument II (05015243001, Roche). The comparative CT method was employed to quantify the relative RNA transcript levels. The primers were as follows: IL6R-human-forward: GGTTGTGGAATCTTGCAGCC; IL6R-human-reverse: TGATGCTGGAGGTCCTTGAC; YM1-mouse-forward: TTTGGACCTGCCCCGTTC; YM1-mouse-reverse: CCTTGGAATGTCTTTCTCCACA; MRC1-mouse-forward: TTCAGCTATTGGACGCGAGG; MRC1-mouse-reverse: GAATCTGACACCCAGCGGAA; IL10-mouse-forward: GGTTGCCAAGCCTTATCGGA; IL10-mouse-reverse: CACCTTGGTCTTGGAGCTTATT.

Isolation and polarization of murine peritoneal macrophages

Murine peritoneal macrophages were isolated as previously described16. To induce M2-like polarization, macrophages were treated with recombinant IL-4 (20 µg/ml, PeproTech, 214-14-20UG) and anti-IL6R (10 µg/ml, BioXcell, BE0047) or IgG (10 µg/ml, BioXcell, BE0090) for 24 h. Additionally, macrophages were treated with tumor supernatants from KPC cells along with anti-IL6R or IgG for 48 h.

Statistical analysis

Statistical analyses were performed using R version 4.3.2 software. An unpaired Student’s t-test was employed to compare two groups, with P values < 0.05 deemed considered statistical significance.

Results

Identification of an immune-related gene signature of PDAC

The training cohort contained 172 PDAC samples and 4 normal samples sourced from the TCGA cohort, all of which included comprehensive clinical information. A total of 1,473 genes were identified by intersecting of immune-related genes from the ImmPort database with genes from the TCGA cohort. Univariate Cox regression analysis, which utilized survival times and outcomes as dependent variables, identified 227 genes with a P value < 0.05. Following this, LASSO regression analysis was then performed to determine the optimal log lambda, thereby narrowing the gene list to 38 candidate genes (Fig. 2A, B). Ultimately, multivariate Cox regression analysis was finally employed to identify the 18 most prognostic genes, which together form the immune-related gene signature (Fig. 2C). The risk score formula was constructed based on this signature and is expressed as follows: risk score = 0.83 × ACKR3 + (−0.30) × BMP7 + (−2.07) × BPIFB4 + (−5.34) × CGA + 0.21 × CXCL9 + 0.37 × ERAP2 + (−0.82) × FYN + 0.28 × GKN1 + (−0.52) × IGHV7-81 + 0.44 × IL18 + (−0.46) × IL6R + (−1.77) × KLRC2 + 1.01 × PF4 V1 + (−1.01) × SEMA6 C + 0.54 × TAP2 + (−0.42) × TNFRSF4 + 1.83 × TRAJ52+(−2.67) × TSHB.

Fig. 2
figure 2

Construction of an immune-related gene signature using the TCGA cohort. (A) Profiles of LASSO regression coefficients and (B) the optimal log lambda for immune-related genes identified through univariate Cox regression analysis. (C) Forest plot depicting the immune-related gene signature analyzed via multivariate Cox regression.

Full size image

To evaluate the reliability of the signature, patients were categorized into high-risk and low-risk groups based on the median risk score. The resulting heatmap of prognostic gene expressions and the distribution of survival times are presented (Fig. 3A-B). The KM survival curve demonstrated that patients in the low-risk group exhibited higher survival probabilities compared to those in the high-risk group, thereby underscoring the significant discriminative power of the signature (Fig. 3C). The area under the curve (AUC) values of the receiver operating characteristic (ROC) curves were recorded as 0.742, 0.827, and 0.779 for overall survival at 1, 3, and 5 years, respectively (Fig. 3D). Besides, the signature achieved a C-index of 0.76, reflecting its predictive accuracy. In comparison, the AUC values for the traditional pathological tumor staging system were 0.470, 0.649, and 0.671 for overall survival at 1, 3, and 5 years, respectively, with a C-index of 0.56 (Fig. 3E). DCA further demonstrated that the signature provided a higher net benefit than the staging system, reinforcing its clinical relevance (Fig. 3F).

Furthermore, we analyzed the differentially expressed genes between high-risk and low- risk patient groups using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses (Fig. 3G). The GO analysis demonstrated significant differences in the processes related to the killing of cells from other organisms and the defense response to bacteria infections. The KEGG analysis revealed the involvement of pancreatic secretion. These findings suggested potential correlations among risk stratification, immune responses, and pancreatic function.

Fig. 3
figure 3

Evaluation of the immune-related gene signature. (A) Heatmap illustrating the clustering of the immune-related gene signature into high-risk and low-risk groups. (B) Distribution of risk scores and survival durations of patients with PDAC. (C) Kaplan-Meier survival curve comparing high-risk and low-risk patients. (D) ROC curves evaluating the overall survival at 1, 3, and 5 years based on the gene signature. (E) ROC curves evaluating the overall survival at 1, 3, and 5 years based on the tumor stage. (F) Decision curve analysis comparing the predictive value of the gene signature and stage. (G) GO and KEGG analyses of the genes included in the signature.

Full size image

To further validate the effectiveness of this immune-related gene signature in PDAC stratification, we employed 113 PDAC samples from the ICGC cohort as the validation cohort. The resulting heatmap of the gene expressions and the distribution of survival times are presented (Fig. 4A-B). As anticipated, the KM survival curve demonstrated that the low-risk group had higher survival probabilities than the high-risk group (Fig. 4C). The C-index of the ICGC cohort was calculated to be 0.79, thereby confirming that the prediction accuracy is consistent with that of the training cohort. These findings substantiate the robustness of the gene signature within the validation cohort as well.

Fig. 4
figure 4

Construction and evaluation of the signature and nomogram using the ICGC cohort. (A) Heatmap illustrating the clustering of the immune-related gene signature into high-risk and low-risk groups. (B) Distribution of risk scores and survival durations of patients with PDAC. (C) Kaplan-Meier survival curves for patients categorized into high- and low-risk groups. (D) Nomogram designed to predict the 1-, 3-, and 5-year survival probabilities of patients. (E-F) Calibration curves for 1- and 3-year survival rates. (G) ROC curves assessing 1- and 3-year overall survival outcomes.

Full size image

Development of a prognostic nomogram of PDAC

To enhance the accuracy of prognosis predictions for patients with PDAC, we performed both univariate and multivariate Cox regression analyses to identify clinically significant or statistically valuable parameters. These parameters included the immune-related gene signature, tumor stage, and postoperative treatment. Given the relatively low 5-year survival rate for PDAC and the constraints imposed by a limited sample size, our nomogram was designed to focus on 1- and 3-year survival outcomes. Consequently, we developed a prognostic nomogram to estimate the survival probabilities at these years (Fig. 5A). our findings indicated that higher risk scores were associated with an increase in points on the nomogram. Moreover, patients who did not receive postoperative treatment exhibited a correlation with elevated points. As the total points increased, a corresponding decrease in survival probability was observed.

To evaluate the accuracy of the nomogram, calibration curves were constructed. The curves for 1-year and 3- year survival displayed a robust correlation between the nomogram-predicted and the actual observed survival rates (Fig. 5B-C). The AUC values of the ROC curves were 0.742 and 0.827 for overall survival at 1 year and 3 years, respectively (Fig. 5D), while the C-index was calculated to be 0.76.

Fig. 5
figure 5

Construction and evaluation of a prognostic nomogram using the TCGA cohort. (A) Nomogram designed to predict the 1-, 3-, and 5-year survival rates of patients. (B-C) Calibration curves for 1- and 3-year survival. (D) ROC curve for the evaluation of 1- and 3-year overall survival.

Full size image

We also applied the consistent methodology to the ICGC cohort, which served as the validation cohort, yielding similarly significant parameters and facilitating the construction of a comparable nomogram (Fig. 4D). The calibration curves demonstrated a good fit (Fig. 4E-F). The AUC values of the ROC curves were 0.738 and 0.750 for overall survival at 1 year and 3 years, respectively (Fig. 4G). The C-index for this cohort was calculated to be 0.90. These findings reinforce the accuracy and consistency of the nomogram from multiple perspectives.

Comparison of immune microenvironment between high-risk and low-risk PDAC patients

In addition to identifying the immune-related gene signature and developing the prognostic nomogram, we also investigated immune cell infiltration in PDAC. The relative proportions of 22 immune cell types were estimated in 172 PDAC samples using the CIBERSORT algorithm (Fig. 6A)13. Consistent with findings from other studies17, a significant presence of immune cells was observed in the tumor samples, particularly macrophages and T cells.

Subsequently, we categorized patients into high- and low-risk groups based on the prognostic nomogram and analyzed their immune cell compositions (Fig. 6B). In alignment with previous researches, macrophages were identified as the major immune cell type18,19,20. Notably, the high-risk group exhibited higher proportions of M0 and M2-like macrophages compared to the low-risk group. This observation may elucidate the anti-inflammatory environment associated with the high-risk group, which could potentially promote tumor progression. Additionally, resting memory CD4+T cells presented the third most abundant cell type, with the low-risk group presenting a greater quantity, which is critical for immune surveillance21,22. These findings suggest that the immune microenvironment of the high-risk group is relatively immunosuppressed.

To further evaluate the stromal and immune cell infiltration, we employed the ESTIMATE and xCell algorithms14,15. The ESTIMATE algorithm revealed no significant differences in stromal scores between the two groups (Fig. 6C). Additionally, the high-risk group exhibited a lower immune score and higher tumor purity. The xCell algorithm also revealed a negative correlation between the risk score and immune score, consistent with the clinical outcomes predicted by the nomogram (Fig. 6D).

Fig. 6
figure 6

Characterization of the PDAC immune microenvironment using a prognostic nomogram. (A) Heatmap illustrating the composition of 22 immune cell types as estimated using CIBERSORT. (B) Comparative analysis of immune cell compositions between high-risk and low-risk groups. (C) Correlations between the risk score and the stromal score, immune score, ESTIMATE score, and tumor purity, respectively. (D) Correlations between the risk score and the immune score, microenvironment score, as assessed using the xCell algorithm. *P < 0.05, **P < 0.01, ***P < 0.001.

Full size image

IL6R as a protective factor in PDAC

We further investigated the implications of genes within the immune-related gene signature. By evaluating differential expression between normal and tumor tissues, and their correlation with survival outcomes (Supplementary Fig. 1), we identified potential genes using the following thresholds: P < 0.03, | coefficient |> 0.03, and log2 (TPM + 1) > 1. Through this analysis, IL6R, ERAP2, FYN, and TAP2 were selected as candidates. A literature review was conducted to assess their functional relevance. ERAP2 and TAP2 are primarily localized in the endoplasmic reticulum, while FYN functions as a tyrosine-protein kinase; these genes are not as directly associated with immune responses as IL6R, which encodes a subunit of the interleukin 6 (IL-6) receptor23,24,25. Additionally, IL-6 inhibitors have already received approval for non-oncology indications, which would facilitate translational research and clinical applications26. Based on these considerations, we focused on IL6R for subsequent analyses. The expression of IL6R was notably increased in tumor samples from the TCGA cohort and GTEx database compared to those in normal samples (Fig. 7A). For experimental validation, we collected local paired normal and tumor tissues from patients with PDAC and performed qPCR analysis (Table 3). The results corroborated that IL6R expression was indeed upregulated in tumor tissues (Fig. 7B). Moreover, high IL6R expression was associated with improved survival outcomes, suggesting its potential as a prognostic marker in PDAC (Fig. 7C-D).

IL6R also knowns as membrane glycoprotein (Gp80), interacts with IL-6 to induce the homodimerization of Gp130, which leads to the formation of the IL-6/IL-6R/Gp130 complex. This complex subsequently triggers downstream signaling cascades, including the RAS-MAPK, JAK1-STAT3, and PI3 K-AKT pathways26. IL-6, which is produced during early stages of inflammation, has the capacity to promote effector T cell differentiation27. We observed elevated IL6Rexpression in tumor samples, suggesting enhanced sensitivity to IL-6/IL-6R signaling. Besides tumor cells, TME contains numerous CAFs, TAMs, and myeloid-derived suppressor cells4. Single-cell transcriptomic analyses of PDAC samples confirmed substantial IL6R expression in mono/macrophages (Fig. 7E), identifying these cells as primary source of IL-6R. The immune-related gene signature demonstrated that low IL6R expression correlates with reduced sensitivity to IL-6/IL-6R signaling and is associated with a high-risk score. Additionally, patients with high-risk PDAC exhibited a significantly greater proportion of M2-like macrophages (Fig. 6B). Based on theses observations, we hypothesized that attenuated IL-6/IL-6R signaling in macrophages facilitate the M2-like differentiation. To test this hypothesis, we polarized murine peritoneal macrophages into M2-like macrophages through IL-4 stimulation while concurrently administering an IL-6R blocker. The results revealed that blocking of IL-6 binding to IL-6R promotes M2-like polarization (Fig. 7F). This finding was further validated using tumor supernatant to culture murine peritoneal macrophages, which yielded consistent results (Fig. 7G). These findings collectively suggest that diminished IL-6/IL-6R signaling in macrophages contributes to an immunosuppressive TME and may represent as a promising target for immunotherapeutic interventions.

Fig. 7
figure 7

Functional significance of the key gene IL6R. (A) Differential expression of IL6R in normal and tumor samples from patients with PDAC, using data from the TCGA cohort and the GTEx database. (B) Validation of IL6R expression in normal and tumor samples from local PDAC patients using qPCR. Kaplan-Meier survival curves illustrating the survival outcomes associated with low and high IL6R expression levels in (C) the TCGA cohort and the GTEx database, and (D) the ICGC cohort. (E) Heatmap representing the expression of IL6R across major cell lineages, using public databases. (F) qPCR analysis of M2-like polarization markers (YM1, MRC1, and IL10) in murine peritoneal macrophages following treatment with IL-4 and an IL-6R blocker. (G) qPCR analysis of M2-like polarization markers (YM1, MRC1, and IL10) in murine peritoneal macrophages following treatment with tumor supernatant and an IL-6R blocker. *P<0.05, **P<0.01, ***P<0.001.

Full size image
Table 3 Characteristics of patient with PDAC using the local cohort.
Full size table

Discussion

Due to the immunosuppressive TME, PDAC is characterized as an immunologically “cold” tumor, which demonstrates limited responsiveness to immune checkpoint inhibitors28. The administration of durvalumab, an anti-programmed death-ligand1 (PD-L1) agent, either as monotherapy or in conjunction with tremelimumab, an anti-cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) agent, has resulted in an objective response rate of only 0–3.1%29. To address this challenge, various immunomodulatory strategies have been developed to reprogram immunosuppressive cells, including TAMs and CAFs. TAMs are typically categorized into two phenotypes: classically activated M1-like macrophages, which secrete inflammatory cytokines that promote anti-tumor immune responses, and alternatively activated M2-like macrophages, which support tissue repair and tumor progression. The predominant subtype of TAMs in PDAC is M2-like, with extensive researches conducted in this field30. For instance, colony-stimulating factor-1 (CSF-1) and its receptor (CSF-1R) are crucial for the maintenance and differentiation of TAMs. Blocking CSF-1R has been shown to reprogram TAMs towards the M1-like phenotype and enhance T cell activation, as evidenced by a phase I trial involving patients with advanced PDAC; however, this intervention did not result in improved progression-free survival among patients receiving standard chemotherapy3,31. In this study, we examined a comprehensive range of immune-related genes from the ImmPort database, with the objective of identifying more promising immunotherapeutic targets for the treatment of PDAC.

To achieve this, we developed an immune-related gene signature utilizing RNA sequencing data, along with patient survival times and outcomes. We performed univariate Cox regression, Lasso regression, and multivariate Cox regression analyses to refine our selection from over a thousand immune-related genes to eighteen key genes that were incorporated into the signature. The signature demonstrated competitive advantages over existing research in at least three significant aspects32,33. Firstly, from a methodological perspective, we adapted the gene selection process to accommodate the unique TME of PDAC. Specifically, acknowledging the role of the dense extracellular matrix produced by CAFs in hindering immune cell infiltration, we opted not to restrict the number of immune-related genes to a small subset34. We also did not limit the candidate genes to those with differential expression, recognizing that certain immune-related gene expression levels might not serve as adequate internal reference genes. These methodological adjustments enabled a more comprehensive understanding of immune characteristics, ensuring that potentially valuable insights were not missed. Secondly, we identified several previously neglected genes, such as IL6R, which are highly clinically relevant and show promise as immunotherapeutic targets. Our study provides important clues and directions for further research to investigate the potential of the remaining genes. Thirdly, our signature outperformed the conventional pathological tumor staging system in predicting patient prognosis. It achieved higher AUC values, C-index, and net benefit compared to the staging system. Additionally, the KM survival curve confirmed the high accuracy of the signature in both training and validation cohorts. These results underscore the clinical significance and utility of the signature.

Furthermore, drawing inspiration from the staging system, we integrated the gene signature with clinical phenotypes to develop a prognostic nomogram. Univariate and multivariate Cox regression analyses revealed that the signature, tumor stage, and postoperative treatment are significant predictors of survival probability. The high-risk predictions associated with the immune-related signature corresponded with low survival probabilities, thereby emphasizing the effectiveness of the nomogram. Calibration curves, AUC values, and C-index further confirmed the consistency of the nomogram across both training and validation cohorts. Beyond prediction accuracy, the nomogram introduces several innovations. While limited research has incorporated postoperative treatment into nomograms, our study demonstrates its correlation with decreased survival probabilities. This may due to the fact that patients requiring postoperative interventions often present with more advanced tumors or from the detrimental effects of radiotherapy and chemotherapy on patients’ immune systems. Additionally, multiple immune infiltration analyses affirmed the value of integrating the immune-related signature into the nomogram. Given that PDAC is characterized by limited immune cell infiltration, this represents a significant advancement. In future studies, we intend to collect additional samples and comprehensive clinical phenotypes from patients with PDAC to refine the immune-related signature and prognostic nomogram. This endeavor will facilitate a deeper understanding of clinical outcomes in PDAC patients and aid in identifying novel immunotherapeutic targets.

CIBERSORT, ESTIMATE and xCell algorithms demonstrated that patients with low survival probabilities, as predicted by the nomogram, exhibited a higher prevalence of immune-suppressed cells within the TME, including M2-like TAMs. In light of these findings, we conducted a comprehensive review of literature and database concerning eighteen candidate genes associated with the immune-related signature. For each gene, we evaluated its expression variations between normal and tumor samples, analyzed the survival outcomes based on high and low expression levels, and investigated the immunological functions and clinical significances. Our primary focus was on IL6R, which encodes a subunit of the IL-6 receptor.

IL-6 features with pleiotropic activity: IL-6 could induce synthesis of acute inflammation in hepatocytes, and stimulate antibody production and effector T cell differentiation; continual dysregulation of IL-6 could lead to the onset of chronic inflammation cancers27. Our cohort data and experimental findings revealed that elevated IL6Rexpression in tumor samples was correlated with favorable outcomes, indicating enhanced sensitivity of the pro-inflammatory function of IL-6. Additionally, much of the research has centered on the IL-6R/STAT3 signaling in tumor cells activated by TAMs35,36. Notably, in PDAC samples, we observed an increased response of IL-6R in TAMs rather than in tumor cells. Integrated with risk score and immune infiltration analysis, experimental findings evidenced that attenuated IL-6/IL-6R signaling in TAMs promoted the M2-like polarization, serving as a novel immunotherapeutic target for PDAC. Regarding immunotherapy applications, monoclonal antibodies targeting IL-6 and IL-6R have received approval for the treatment of rheumatoid arthritis, although their efficacy in tumor therapy is still under investigation in clinical trials. A phase I trial involving patients with epithelial ovarian cancer demonstrated that anti-IL-6R monoclonal antibodies were feasible and safe for immune enhancement37. Blockade of IL-6 signaling has been shown to mitigate paclitaxel-induced neuropathy38. However, despite high expectations, clinical trials exploring IL-6 and IL-6R blockades in PDAC remain scare, underscoring the necessary for further research in this field.

In this study, we developed and validated an immune-related gene signature along with a corresponding prognostic nomogram to predict clinical outcomes of patients diagnosed with PDAC. The proposed model demonstrated superior predictive performance compared to the traditional staging system. Through analysis of immune cell infiltration between high-risk and low- risk groups, we identified IL6R as a promising target for immunotherapy. Elevated IL6R expression was correlated with improved survival outcomes, whereas attenuated IL-6/IL-6R signaling in macrophages contributed to the establishment of an immunosuppressive TME. Our findings not only provide a novel approach for prognostic prediction and therapeutic decision-making in PDAC, but also highlighting IL6R as a new target for immunotherapeutic intervention.