A novel immune checkpoint score system for prognostic evaluation in pancreatic adenocarcinoma

Abstract

Background: Pancreatic adenocarcinoma (PAAD) remains a lethal malignancy making the detection of novel prognostic biomarkers urgent. Limited studies have investigated the predictive capability of immune checkpoints in PAAD.

Method: Gene expression data and correlative clinical information of PAAD cohort were obtained from public databases, including TCGA, ICGC, GTEX and GEO databases. Risk factors were screened and used to establish a risk score model through LASSO and Cox regression analyses. The prognostic ability of the risk score model was demonstrated. The association between risk score with immune cells infiltration, immune checkpoint genes expression, immunogenic cell death, somatic mutations and signaling pathways enrichment were analysed. scRNA-seq data were collected to confirmed the immune checkpoints expression in PAAD samples. The prognosis prediction ability of OX40/TNFRSF4 was identified. The mRNA and protein expression of OX40 in our clinical specimens were examined by RT-PCR and IHC method and its prognosis ability was verified.

Results: First of all, the difference of immune microenvironment between pancreatic cancer and adjacent tissues was shown. A risk score system based on three immune checkpoints (OX40, TNFSF14 and KIR3DL1) was established. The risk score model was an independent prognostic factor and performed well regarding overall survival (OS) predictions among PAAD patients. A nomogram was established to facilitate the risk model application in clinical prognosis. Immune cells including naive B cells, CD8⁺ T cells and Tregs were negatively correlated with the risk score. The risk score was associated with expression of immune checkpoint genes, immunogenic cell death related genes and somatic mutations. Glycolysis processes, IL-2-STAT5, IL-6-STAT3, and mTORC1 signaling pathways were enriched in the high-risk score group. Furthermore, scRNA-seq data confirmed that TNFRSF4, TNFSF14 and KIR3DL1 were expressed on immune cells in PAAD samples. We then identified OX40 as an independent prognosis-related gene, and a higher OX40 expression was associated with increased survival rate and immune environment change. In 84 PAAD clinical specimens collected from our center, we confirmed that higher OX40 mRNA expression levels were related to a good prognosis. The protein expression of OX40 on tumor-infiltrating immune cells (TIICs), endothelial cells and tumor cells was verified in PAAD tissues by immunohistochemistry (IHC) method.

Conclusions: Overall, our findings strongly suggested that the three-immune checkpoints score system might be useful in the prognosis and design of personalized treatments for PAAD patients. Finally, we identified OX40 as an independent potential biomarker for PAAD prognosis prediction.

Keywords: Immune checkpoints; Immune risk score model; OX40; Pancreatic adenocarcinoma; Prognostic evaluation.

Fig. 1

Expression of tumor-infiltrating immune cells and immune checkpoint genes between PAAD and normal cases. A Differences in the abundance of tumor-infiltrating immune cells between PAAD and normal tissues calculated by CIBERSORT in TCGA and GTEX database. B Differences in the abundance of tumor-infiltrating immune cells between PAAD and normal tissues calculated by CIBERSORT in GSE62452 dataset. C Differences in the expression of immune checkpoint genes between PAAD and normal tissues. *p < 0.05, **p < 0.01, ***p < 0.001

Fig. 2

Three immune checkpoint genes were screened to construct the immune risk scoring model. A The LASSO regression analysis identified three immune checkpoint genes with the best prognostic value to be incorporated into the prognostic signature. B Distribution of immune risk scores, survival status, and heatmap of gene expression patterns (TNFRSF4, TNFSF14, and KIR3DL1) in the high-risk and low-risk groups in the training cohort. C ROC curve in the training cohort. D Kaplan–Meier survival curves of overall survival in the training cohort

Fig. 3

Validation of the risk scoring model in testing and total cohorts. A Risk score distribution, survival time, and expression patterns of the three genes in risk groups in the testing cohort. B Kaplan–Meier survival curves between different risk groups in the testing cohort. C Distribution of risk score, survival time, and heatmap of the expression of the three genes in the total cohort. D Kaplan–Meier survival curves in the total cohort

Fig. 4

Cox’s model of related factors in PAAD patients. A Univariate Cox analysis for eight OS-related clinicopathological parameters. B Multivariate Cox analysis for eight OS-related clinicopathological parameters. C A prognostic nomogram predicting the OS of PAAD patients. OS: overall survival

Fig. 5

Relationship between immune checkpoint-related risk scores and immune cells infiltration in PAAD. Radar graph of the relative abundance of immune cells in different risk groups in PAAD samples

Fig. 6

Relationship between risk scores with immune checkpoint and immunogenic cell death-related gene expression in PAAD. A Differentially expressed immune checkpoint-related genes in different risk groups. B Differentially expressed immunogenic cell death-related genes in different risk groups

Fig. 7

GSEA enrichment plots and mutation profile between different risk groups. A The GSEA was based on the risk score median value in PAAD patients. B The gene mutation frequency is shown on the left panel, while the mutational prevalence of synonymous/nonsynonymous mutations is presented on the upper panel. The gene mutation landscape is displayed in the middle. Different risk groups are depicted on the bottom

Fig. 8

Single-cell RNA-Seq data of PAAD. A Single-cell RNA-Seq data of two PAAD datasets in TISCH database. B t-SNE of normalized single-cell RNA-seq data for three immune checkpoints in CancerSEA database

Fig. 9

Survival and immune infiltration analysis of prognostic-related genes in PAAD. A Kaplan–Meier plot of overall survival between differentially expressed TNFRSF4, KIR3DL1, and TNFRSF14 groups. B Correlation analysis between the expression level of TNFRSF4 and immune infiltration

Fig. 10

Survival analysis of OX40 in clinical PAAD specimens. Kaplan–Meier plot between different OX40 expressions in clinical PAAD specimens

Fig. 11

IHC staining of OX40 in clinical PAAD specimens. Human PAAD tissue slides were stained with anti-OX40 antibody. Magnifcation (10 × and 40x) images of high and low expression of OX40 were showed. Scale bar = 100 μm (black line at the bottom left). A OX40 expression on endothelial cells. B OX40 expression on immune cells. C OX40 expression on tumor cells