The prognosis of pancreatic cancer was extremely poor. The overall 5-year survival rate was only 9%, and merely 37% in stage I or IIA postoperative PC patients 2. There were many prognostic factors for PC patients, and it was difficult to predict the prognosis accurately by relying on solo prognostic indicator. Some prognostic prediction systems were constructed with clinical parameters to improve the predictive capability, classical ones such as Botsis score, Heidelberg score, MSKCC score, etc. 3,16,17. However, their performances were still not entirely unsatisfactory. More accurate indicators and prediction systems should be used as the reference for individualized clinical decision-making in the future. As the mechanism of TME was gradually revealed, it was a reasonable and feasible way to explain clinical characteristics and improve clinical practice guide by using the information of TME. ESTIMATE score system was developed in 2013. The principle was to use the gene expression data of immune and stromal cells to infer the contents of them in tumor samples 15. Currently, RNA sequencing data of 25 cancer species from TCGA had been analyzed by ESTIMATE. In this study, a total of 210 candidate genes related to TME and prognosis were identified by using ESTIMATE and survival analysis. These genes were significantly enriched in the construction of extracellular matrix and Cytokine-cytokine receptor interaction by GO and KEGG enrichment analysis. Finally, seven genes were screened for the construction of the prognostic risk score system which were rare reported. Now, these genes would be briefly described below. The mechanism of FAM57B and LY6D in malignant tumors has not been reported. HTRA3, which expresses a pro-apoptotic protease, promotes drug-induced cytotoxic effects and was thought to have the function of anti-tumor 18. It showed inhibition of tumor proliferation and metastasis in lung cancer cells and endometrial cancer cells, and high expression of HTRA3 was associated with longer disease-free survival in non-small cell lung cancer patients 19–21. SPRR1B is overexpressed in oral tumor stem cells and has been found to induce tumor proliferation by activating MAPK signaling, but its specific functional mechanism in PC has not been investigated 22. CXCL10, also known as IP-10, high expression in the TME of PC could influence lymphocytes recruitment and was correlate with poor survival 23. GABRP, significantly higher expression in PC tissue than normal tissues, could promote tumor growth and metastasis. Higher expression of GABRP was also associated with poor prognosis of PC. In mechanism, GABRP interacts with KCNN4 to induce calcium influx, activate the nuclear factor κB signaling, and ultimately promote macrophage infiltration by inducing the expression of CXCL5 and CCL20 24. Overexpression of FAM83A activated TGF-β signaling pathway and the Wnt/β-catenin signaling. In vitro and vivo, FAM83A expression was associated with the characteristics of PC tumor stem cell and the generation of chemotherapy resistance 25. The role of these novel genes in tumorigenesis and tumor development should be further investigated.
The predictive capabilities of the prognostic risk score systems were different by construction using different principles and angles. In the past decade, with the rapid development of bioinformatics, tumor molecular biology and tumor immunology, a large number of molecular markers have been discovered and were considered to have potential diagnostic and therapeutic value. The number of the prognostic models constructed by tumor molecular markers were gradually increasing, and most of them showed acceptable performance 26–30. However, there was still no perfect prognostic model due to the heterogeneous tumors and unknown pathogenesis. Current research on the pathogenesis of malignant tumors, a large number of TME related molecular markers have been revealed had the relationship with tumor proliferation, immune escape, drug resistance, metastasis and so on. They also play as mediators for tumor - environment interaction. Many researchers have transferred their focuses from tumor cells to the surrounding environment of tumor cells. PC, as an inflammatory tumor, has changes in the TME during its development of inflammatory - cold process 14,31,32. In clinical research on TME and PC, phase Ib study of PEGylated human recombinant hyaluronidase (PEGPH20) in combination with gemcitabine in advanced PC patients has shown promising efficacy 33. Phase I study of CCR2 inhibition in combination with FOLFIRINOX in borderline resectable and locally advanced PC patients also has yielded a good clinical outcome 34. It is very important to study TME related prognostic genes of PC, which can further reveal the tumor development and progression. Establishment of a prognostic system based on TME related prognostic genes can more accurately and conveniently evaluate the prognosis of PC patients. Pu et al. analyzed the TME related prognostic genes through the similar way, and identified 53 genes related to immune score and 17 genes related to stromal score, which were related to recurrence free survival. Unfortunately, prognostic model had not further constructed 35. To our knowledge, this prognostic prediction system was the first model constructed by ESTIMATE for PC. According to this prognostic risk score system, Harrell's C-index was 0.73, and AUC value of 1-year, 2-year and 3-year OS period was 0.67, 0.76 and 0.86 in the internal validation set. Harrell's C-index was 0.71, and AUC value of 1-year, 2-year and 3-year OS period was 0.81, 0.72 and 0.78 in the external validation set. By reviewing other reported prognostic prediction systems of PC, this prognostic system contains fewer variables and has a good predictive capacity 26–29. Furthermore, it was found this prognostic risk score system was significantly associated with the expression of some ICGs, suggesting that it may have predictive implications for immunotherapy response and efficacy.
Since the data were based on TCGA and ICGC, the limited sample size and the quality of data set were the biggest bottleneck of this study. In addition, the functions and mechanisms of the novel genes should be further explored by basic experiments. The effectiveness of this prognostic prediction system needs to be confirmed by multi-center and prospective research.