Serum Metabolomic Profiling Reveals Biomarkers for Early Detection and Prognosis of Esophageal Squamous Cell Carcinoma

Pan Pan Wang,Xin Song,Xue Ke Zhao,Meng Xia Wei,She Gan Gao,Fu You Zhou,Xue Na Han,Rui Hua Xu,Ran Wang,Zong Min Fan,Jing Li Ren,Xue Min Li,Xian Zeng Wang,Miao Miao Yang,Jing Feng Hu,Kan Zhong,Ling Ling Lei,Liu Yu Li,Yao Chen,Ya Jie Chen,Jia Jia Ji,Yuan Ze Yang,Jia Li,Li Dong Wang

DOI: https://doi.org/10.3389/fonc.2022.790933

IF: 4.7

2022-01-28

Frontiers in Oncology

Abstract:Esophageal squamous cell carcinoma (ESCC) is one of the most common aggressive malignancies worldwide, particularly in northern China. The absence of specific early symptoms and biomarkers leads to late-stage diagnosis, while early diagnosis and risk stratification are crucial for improving overall prognosis. We performed UPLC-MS/MS on 450 ESCC patients and 588 controls consisting of a discovery group and two validation groups to identify biomarkers for early detection and prognosis. Bioinformatics and clinical statistical methods were used for profiling metabolites and evaluating potential biomarkers. A total of 105 differential metabolites were identified as reliable biomarker candidates for ESCC with the same tendency in three cohorts, mainly including amino acids and fatty acyls. A predictive model of 15 metabolites [all-trans-13,14-dihydroretinol, (±)-myristylcarnitine, (2S,3S)-3-methylphenylalanine, 3-(pyrazol-1-yl)-L-alanine, carnitine C10:1, carnitine C10:1 isomer1, carnitine C14-OH, carnitine C16:2-OH, carnitine C9:1, formononetin, hyodeoxycholic acid, indole-3-carboxylic acid, PysoPE 20:3, PysoPE 20:3(2n isomer1), and resolvin E1] was developed by logistic regression after LASSO and random forest analysis. This model held high predictive accuracies on distinguishing ESCC from controls in the discovery and validation groups (accuracies > 89%). In addition, the levels of four downregulated metabolites [hyodeoxycholic acid, (2S,3S)-3-methylphenylalanine, carnitine C9:1, and indole-3-carboxylic acid] were significantly higher in early cancer than advanced cancer. Furthermore, three independent prognostic markers were identified by multivariate Cox regression analyses with and without clinical indicators: a high level of MG(20:4)isomer and low levels of 9,12-octadecadienoic acid and L-isoleucine correlated with an unfavorable prognosis; the risk score based on these three metabolites was able to stratify patients into low or high risk. Moreover, pathway analysis indicated that retinol metabolism and linoleic acid metabolism were prominent perturbed pathways in ESCC. In conclusion, metabolic profiling revealed that perturbed amino acids and lipid metabolism were crucial metabolic signatures of ESCC. Both panels of diagnostic and prognostic markers showed excellent predictive performances. Targeting retinol and linoleic acid metabolism pathways may be new promising mechanism-based therapeutic approaches. Thus, this study would provide novel insights for the early detection and risk stratification for the clinical management of ESCC and potentially improve the outcomes of ESCC.

oncology