Causal relationship between key genes and metabolic dysfunction‐associated fatty liver disease risk mediated by immune cells: A Mendelian randomization and mediation analysis

• Published in: Diabetes, obesity & metabolism Vol. 26; no. 12; pp. 5590 - 5599
• Main Authors: Feng, Gong, He, Na, Gao, Jing, Li, Xiao‐Cheng, Zhang, Fen‐Na, Liu, Cheng‐Cheng, Targher, Giovanni, Byrne, Christopher D., Mi, Man, Zheng, Ming‐Hua, Ye, Feng
• Format: Journal Article
• Language: English
• Published: Oxford, UK Blackwell Publishing Ltd 01.12.2024; Wiley Subscription Services, Inc
• Subjects: Biomarkers; Biomarkers - metabolism; Biopsy; CD11c antigen; Cell culture; diagnostic markers; DNA microarrays; Fatty liver; Female; Hepatocytes; Humans; Insulin-Like Growth Factor Binding Protein 2 - genetics; Insulin-like growth factor-binding protein 2; L-selectin; Liver diseases; Male; Mediation Analysis; Mendelian randomization; Mendelian Randomization Analysis; metabolic dysfunction–associated fatty liver disease; Metabolism; Middle Aged; Monocytes; Non-alcoholic Fatty Liver Disease - genetics; Non-alcoholic Fatty Liver Disease - metabolism; Quantitative Trait Loci; Risk Factors; ROC Curve; Transcriptome; Transcriptomes; diagnostic markers; metabolic dysfunction–associated fatty liver disease; mediation analysis; Mendelian randomization
• ISSN: 1462-8902; 1463-1326; 1463-1326
• DOI: 10.1111/dom.15925

Aim Non‐invasive diagnostics for metabolic dysfunction–associated fatty liver disease (MAFLD) remain challenging. We aimed to identify novel key genes as non‐invasive biomarkers for MAFLD, elucidate causal relationships between biomarkers and MAFLD and determine the role of immune cells as potential mediators. Materials and Methods Utilizing published transcriptome data of patients with biopsy‐proven MAFLD, we applied linear models for microarray data, least absolute shrinkage and selector operation (LASSO) regressions and receiver operating characteristic (ROC) curve analyses to identify and validate biomarkers for MAFLD. Using the expression quantitative trait loci database and a cohort of 778 614 Europeans, we used Mendelian randomization to analyse the causal relationships between key biomarkers and MAFLD. Additionally, mediation analysis was performed to examine the involvement of 731 immunophenotypes in these relationships. Results We identified 31 differentially expressed genes, and LASSO regression showed three hub genes, IGFBP2, PEG10, and P4HA1, with area under the receiver operating characteristic (AUROC) curve of 0.807, 0.772 and 0.791, respectively, for identifying MAFLD. The model of these three genes had an AUROC of 0.959 and 0.800 in the development and validation data sets, respectively. This model was also validated using serum‐based enzyme‐linked immunosorbent assay data from MAFLD patients and control subjects (AUROC: 0.819, 95% confidence interval: 0.736–0.902). PEG10 was associated with an increased MAFLD risk (odds ratio = 1.106, p = 0.032) via inverse variance–weighted analysis, and about 30% of this risk was mediated by the percentage of CD11c + CD62L– monocytes. Conclusions The MAFLD panels have good diagnostic accuracy, and the causal link between PEG10 and MAFLD was mediated by the percentage of CD11c + CD62L– monocytes.