Despite the great progress made in treating pancreatic cancer over the last few decades, the prognosis has not been effectively improved (Neoptolemos et al., 2018). Genetic concepts and tools are increasingly being applied to clinical practice, especially in precision medicine (Lomberk et al., 2019). However, the biomarkers related to the prognosis of pancreatic cancer are still limited.

Recently, more and more evidence has revealed the significant effect of the reprogrammed metabolism on pancreatic cancer in terms of carcinogenesis, progression, treatment, and prognosis (Qin et al., 2020). The so-called metabolic reprogramming refers to the significant changes in metabolic patterns during cell carcinogenesis, which involves glycolysis, tricarboxylic acid cycle, oxidative phosphorylation, as well as metabolism of amino acid, fatty acid, and nucleic acid (Ward and Thompson, 2012; Huo et al., 2021a,b). During proliferation, tumor cells rely on metabolic reprogramming to provide nutrition, energy, and biosynthetic activity (Pavlova and Thompson, 2016; Neoptolemos et al., 2018). Pancreatic cancer is a malignant tumor with metabolic heterogeneity. Changes in glucose, lipid metabolism as well as amino acid in pancreatic tumors, from cells to microenvironment, and even at the systemic level, can significantly impact tumor progression (Daemen et al., 2015; Qin et al., 2020). Even for the same patients with pancreatic cancer, the metabolic gene expression of the primary focus and the metastatic focus were relatively different (Chaika et al., 2012; Qin et al., 2020). Although the metabolic targeted therapy for pancreatic cancer is not mature at present (Biancur and Kimmelman, 2018), successive clinical trials have shown that metabolic treatment of pancreatic cancer may improve the prognosis of patients (Zachar et al., 2011; Raez et al., 2013; Alistar et al., 2017). Hence, more metabolic biomarkers related to pancreatic cancer prognosis need to be identified. Considering that the effective clinical treatment of pancreatic cancer is still limited, it is urgent to explore new treatment strategies.

The microenvironment around pancreatic cancer cells is composed of immune cells, stellate cells/fibroblasts, and extracellular matrix (ECM). The rapid proliferation of tumor cells leads to a lack of nutrients in the microenvironment, increased release of lactic acid and other metabolites, and metabolic remodeling such as hypoxia and oxidative stress imbalance. Pancreatic cancer cells rely on metabolic reprogramming to adapt to the lack of energy and nutrition in the tumor microenvironment, abnormal oxidative stress, and so on (Bapat et al., 2011). Therefore, it is necessary to deeply understand the impact of metabolic reprogramming on the occurrence and development of pancreatic cancer, so as to provide new ideas for the targeted intervention of metabolic characteristics for the treatment of pancreatic cancer.

In this study, we identified metabolic genes with different expressions between pancreatic cancer and normal tissues through the TCGA and GTEx databases and explored their prognostic value. The prognostic model, composed of 15 metabolic genes, can accurately predict the survival rate of pancreatic cancer and is an independent predictor related to prognosis. In addition, we integrate the GEO database and ICGC database to verify the model and build a survival predictive nomogram.

Pancreatic cancer is a highly malignant digestive tract tumor. Because of its concealed early symptoms, rapid disease progression, low resection rate, and low effective rate of chemotherapy, patients have a very poor prognosis (Ilic and Ilic, 2016). With the accumulation of high-throughput sequencing data, more and more biomarkers have been developed for diagnosing and treating pancreatic cancer. These prognostic signatures involve m6A methylation, autophagy, immunity, and many other aspects (Zheng et al., 2018; Wu et al., 2019; Zhou et al., 2019; Tian et al., 2020; Yue et al., 2020). In recent years, more and more evidence shows that reprogramming metabolism could greatly affect pancreatic cancer in terms of the occurrence, the development, as well as the treatment (Qin et al., 2020). However, the prognostic signatures related to metabolic reprogramming in pancreatic cancer are far from fully cleared.

Patients (572) with complete prognostic information were included in this study. First, we compared 178 pancreatic cancer tissues with 171 normal pancreatic tissues in TCGA and GTEx databases, and identified 379 DEMRGs. Then univariate Cox regression analysis together with the Lasso regression assisted in constructing a novel prognostic model. The unified risk score formula together with the cutoff value were considered to divide patients into a group with a high risk and a group with a low risk. The ROC curve showed the prognostic model with high accuracy in predicting OS, DSS, and PFS of patients. There were 15 genes included into our signature. Among them, ABCA5 is a member of the ATP binding cassette (ABC) transporters, which play a variety of roles in cancer biology and drug resistance. Low expression of ABCA5 is associated with poor prognosis of serous ovarian cancer (Hedditch et al., 2014). Irene Aksoy and others (Aksoy et al., 2017) combined sequencing technology with IPSC technology to identify that GTDC1 is related to neurodevelopmental disorders. Ema et al. (2015) found that SLC25A27 was amplified in advanced gastric cancer with lymph node metastasis. Sulfate endonuclease SULF2 regulates heparan sulfate protein polysaccharide 6-O-sulfation. Alhasan reported that the increase in SULF2 in PDAC is related to advanced tumor stage, vascular invasion, short interval between imaging progression, and short OS (Alhasan et al., 2016). GPD2 is a component of glycerol phosphate shuttle, which can promote the oxidation of glucose, thus, promoting the production of acetyl-CoA. Langston found that GPD2 is involved in the regulation of macrophage inflammation (Langston et al., 2019). MTHFD1 is an enzyme that provides tetrahydrofolic acid-carbon derivatives. Yu found that the high expression of MTHFD1 in hepatocellular carcinoma is associated with a lower survival rate and higher recurrence rate (Yu et al., 2019). Type II inositol polyphosphate 4-phosphatase (INPP4B) is a member of the PI3K/Akt signaling pathway. Zhai found that the overexpression of INPP4B in pancreatic cancer could lead to poor OS and DFS (Zhai et al., 2019). Glycosylation can remarkably affect tumor invasion and immune escape. Zhang found that the high expression of GALNT10 in high-grade ovarian serous cancer (HGSC) is related to immunosuppressive microenvironment, thus promoting tumor progression (Zhang et al., 2020). No reports focus on studying the effect of the remaining genes on cancer.

The group with a high risk presented a worse prognosis compared with the group with a low risk. The external validation results of the ICGC and the GEO cohorts further confirm the effectiveness of this prognostic model. GSEA revealed the oncological characteristics with significant enrichment in the group with a high risk, and pancreatic cancer is one of them, while the group with a low risk was associated with multiple metabolic pathways, indicating that the imbalance of tumor metabolic microenvironment may affect the progression of pancreatic cancer. The tumor microenvironment is a hot topic in the field of tumor research in recent years. Multiple studies have shown that metabolic reprogramming can have a significant impact on the tumor microenvironment (Lyssiotis and Kimmelman, 2017; Reina-Campos et al., 2017). Immune cells are an important component of the tumor microenvironment, which has been proved to be valuable in predicting the prognosis of tumors (Gentles et al., 2015). YIno found that tumor-infiltrating CD8T cells can be used to independently predict the prognosis of pancreatic cancer, and the high infiltration of CD8T cells is associated with longer survival (Ino et al., 2013). In this study, we also found that the proportion of CD8T cell infiltration in the group with a low risk was higher than the group with a high risk, further confirming the prognostic value owned by tumor-infiltrating CD8T cells in pancreatic cancer. Programmed cell death ligand 1 is one protein encoded by the CD274 gene. When it binds to PD1, it transmits a negative regulatory signal to T-cells, induces T-cells to enter a resting state, reduces the proliferation of CD8T cells in lymph nodes, making them unable to recognize cancer cells, reduces T-cell proliferation or apoptosis, effectively relieves the immune response of the body, and promotes further proliferation of cancer cells (Chen and Han, 2015; Naidoo et al., 2015). This study found that the risk score was positively related to the expression level of CD274 (PDL1), so the group with a high risk exhibited a poor prognosis possibly caused by the mechanism of immune escape. Besides, we can also predict the degree of tumor differentiation, clinicopathological stage, and lymph node metastasis according to the risk score, which has important reference value for clinical decision making. As revealed by the univariate and multivariate Cox regression analyses, individual neoplasm status, lymph node metastasis, as well as risk score were independent predictors of prognosis. We combined three indicators to construct one nomogram for the prediction of 1-, 3-, and 5-year OS of pancreatic cancer. The nomogram further enriches the prognosis evaluation system of pancreatic cancer, and the predictive ability of the risk score is further improved. The nomogram has a better prediction effect than a single predictor.

The study integrated as well as analyzed high-throughput sequencing data from multiple databases, and a personalized nomogram for survival prediction was gradually created. However, due to the lack of corresponding clinical data in the validation queue, we only performed internal validation on nomogram. Metabolic genes in the model may be potential targets for diagnosis or treatment of pancreatic cancer, and their detailed mechanisms need to be explored with the help of in vivo and in vitro verification experiments. This study is only a retrospective study, and further prospective results are needed to support each other.

The study focused on constructing a signature and a nomogram associated with metabolic reprogramming for predicting the prognosis of pancreatic cancer, which may help to further improve the treatment strategy of pancreatic cancer.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

JH and LW designed this study. JH analyzed the data in this study, interpreted the findings, and drafted the manuscript. LW and YZ carried out data management and revised the manuscript. All authors reviewed the final version of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.