ABSTRACT
BACKGROUND: Metabolic associated fatty liver disease (MAFLD) is a novel concept proposed in 2020, which is more practical for identifying patients with fatty liver disease with high risk of disease progression. Fatty liver is a driver for extrahepatic complications, particularly cardiovascular diseases (CVD). Although the risk of CVD in MAFLD could be predicted by carotid ultrasound test, a very early stage prediction method before the formation of pathological damage is still lacking. METHODS: Stool microbiomes and plasma metabolites were compared across 196 well-characterized participants encompassing normal controls, simple MAFLD patients, MAFLD patients with carotid artery pathological changes, and MAFLD patients with diagnosed coronary artery disease (CAD). 16S rDNA sequencing data and untargeted metabolomic profiles were interrogatively analyzed using differential abundance analysis and random forest (RF) machine learning algorithm to identify discriminatory gut microbiomes and metabolomic. RESULTS: Characteristic microbial changes in MAFLD patients with CVD risk were represented by the increase of Clostridia and Firmicutes-to-Bacteroidetes ratios. Faecalibacterium was negatively correlated with mean-intima-media thickness (IMT), TC, and TG. Megamonas, Bacteroides, Parabacteroides, and Escherichia were positively correlated with the exacerbation of pathological indexes. MAFLD patients with CVD risk were characterized by the decrease of lithocholic acid taurine conjugate, and the increase of ethylvanillin propylene glycol acetal, both of which had close relationship with Ruminococcus and Gemmiger. Biotin l-sulfoxide had positive correlation with mean-IMT, TG, and weight. The general auxin pesticide beta-naphthoxyacetic acid and the food additive glucosyl steviol were both positively correlated with the increase of mean-IMT. The model combining the metabolite signatures with 9 clinical parameters accurately distinguished MAFLD with CVD risk in the proband and validation cohort. It was found that citral was the most important discriminative metabolite marker, which was validated by both in vitro and in vivo experiments. CONCLUSIONS: Simple MAFLD patients and MAFLD patients with CVD risk had divergent gut microbes and plasma metabolites. The predictive model based on metabolites and 9 clinical parameters could effectively discriminate MAFLD patients with CVD risk at a very early stage.