Nonalcoholic steatohepatitis (NASH) is a condition which can progress to cirrhosis and hepatocellular carcinoma. Markers for NASH diagnosis are still lacking. We performed a comprehensive lipidomic analysis on human liver biopsies including normal liver, nonalcoholic fatty liver and NASH. Random forests-based machine learning approach allowed characterizing a signature of 32 lipids discriminating NASH with 100% sensitivity and specificity. Furthermore, we validated this signature in an independent group of NASH patients. Then, metabolism dysregulations were investigated in both patients and murine models. Alterations of elongase and desaturase activities were observed along the fatty acid synthesis pathway. The decreased activity of the desaturase FADS1 appeared as a bottleneck, leading upstream to an accumulation of fatty acids and downstream to a deficiency of long-chain fatty acids resulting to impaired phospholipid synthesis. In NASH, mass spectrometry imaging on tissue section revealed the spreading into the hepatic parenchyma of selectively accumulated fatty acids. Such lipids constituted a highly toxic mixture to human hepatocytes. In conclusion, this study characterized a specific and sensitive lipid signature of NASH and positioned FADS1 as a significant player in accumulating toxic lipids during NASH progression.

Mice with hepatocytes that express hREG3A, which travels to the intestinal lumen, are less sensitive to colitis than control mice. We found hREG3A to alter the colonic microbiota by decreasing levels of ROS. Fecal microbiota from REG3A-TG mice protect non-TG mice from induction of colitis. These findings indicate a role for reduction of oxidative stress in preserving the gut microbiota and its ability to prevent inflammation.