Hepatic and serum lipid signatures specific to nonalcoholic steatohepatitis in murine models

Abstract

Nonalcoholic fatty liver (NAFL) is a precursor of nonalcoholic steatohepatitis (NASH), a condition that may progress to cirrhosis and hepatocellular carcinoma. Markers for diagnosis of NASH are still lacking. We have investigated lipid markers using mouse models that developed NAFL when fed with high fat diet (HFD) or NASH when fed using methionine choline deficient diet (MCDD). We have performed a comprehensive lipidomic analysis on liver tissues as well as on sera from mice fed HFD (n = 5), MCDD (n = 5) or normal diet as controls (n = 10). Machine learning approach based on prediction analysis of microarrays followed by random forests allowed identifying 21 lipids out of 149 in the liver and 14 lipids out of 155 in the serum discriminating mice fed MCDD from HFD or controls. In conclusion, the global approach implemented allowed characterizing lipid signatures specific to NASH in both liver and serum from animal models. This opens new avenue for investigating early and non-invasive lipid markers for diagnosis of NASH in human.

Figure 1. Characteristics of mouse models fed a high-fat diet and a methionine-choline deficient diet.

(a) Body weights of mice fed a high-fat diet 60% (HFD, ■), a methionine choline deficient diet (MCDD, ▲) and their respective control groups fed a normal diet (ND, •). Mice on MCDD were fed during 5 weeks (black bare). (b) Blood glucose follow-up during the 5 weeks when mice were fed a MCDD (▲) compared to HFD (■) and their respective control groups fed a ND (•). (c) Hematoxylin and eosin staining of livers from mice fed a HFD, a MCDD and their respective controls (ND). Left panels are 20x magnification and black square area of 40x magnification on the right panels. Scale bares: 100 μm. Black arrows show inflammatory cells and black dash square focus on hepatocytes presenting ballooning and Malory’s body (magnification of black dash square focus on upper-right panel) with presence of hepatic lipid droplets characteristic of NASH. (microvesicular steatosis; white arrows). (d) Total hepatic triglycerides (TG) in mice fed a HFD and a MCDD compared to their respective control (ND). (e) Hepatic mRNA gene expression levels involved in the inflammatory process and partially controlled by lipids. Data are means ± SEM. *p < 0.05, by unpaired t-test compared to mice fed a ND and +p < 0.05, by unpaired t-test compared to mice fed a HFD, after ANOVA-test. ND n = 10, HFD n = 5, MCDD n = 5. Il: interleukin; Tgf: transforming grouth factor; Tlr: toll-like receptor; Tnf: tumor necrosis factor.

Figure 2. Specific hepatic lipids discriminating mice fed a MCDD.

(a) Identification of 38 lipids among 149 lipids based on the calculating threshold 2.301 to have the minimum misclassification rate between the 3 groups of mice using the algorithm based on prediction analysis of microarray. (b) 32 lipids identified among 38 lipids discriminating the three groups of mice based on random forests analysis. (c) Principal component analysis (PCA) using 32 lipids identified in (A). Dots represent each mouse, lines are the ellipses centred to the mean (coloured squares) representing 95% interval confidence, and p the probability associated with the F- test of the analysis of variance along the axes of the first and the second dimensions (α = 0.05). (d) Boxplots of the 21 lipids from PCA discriminating mice fed a MCDD. ^$p < 0.05 compared to mice fed ND; ‡p < 0.05 compared to mice fed HFD based on ANOVA-test follow by unpaired t-test. 12-HETE: 12-hydroxy-5Z,8Z,10E,14Z-eicosatetraenoic acid; 18-HEPE: 18-hydroxy-5Z,8Z,11Z,14Z,16E-eicosapentaenoic acid; CE: cholesteryl ester; DG: diacylglycerol; HFD: high-fat diet (red); FAME: fatty acyl methyl ester; MCDD: methionine-choline deficient diet (green); ND: normal diet (black). PC: phosphatidylcholine; PE: phosphatidylethanolamine; PI: phosphatidylinositol; TG: triglycerides. Most representative phospholipids: PC(C32:0) = PC(16:0/16:0); PC(C32:1) = PC(16:0/16:1); PC(C34:1) = PC(16:0/18:1); PC(C34:2) = PC(16:0/18:2); PC(C34:3) = PC(16:1/18:1); PC(C36:4) = PC(18:2/18:12); PC(C38:6) = PC(18:2/20:4); PE(C38:6) = PE(18:1/20:5); PE(C40:6) = PE(18:0/22:6); PI(C36:2) = PI(18:1/18:1); PI(C36:3) = PI(18:1/18:2).

Figure 3. Specific serum lipids discriminating mice fed a MCDD.

(a) Identification of 37 lipids among 155 lipids based of the calculating threshold 1.934 to have the minimum misclassification rate between the 3 groups of mice using prediction analysis of microarray. (b) 33 lipids identified among 37 lipids discriminating the three groups of mice based on random forest analysis. (c) Principal component analysis (PCA) using 33 lipids identified in (A). Dots represent each mouse, lines are the ellipses centred to the mean (coloured squares) representing 95% interval confidence, and p the probability associated with the F- test of the analysis of variance along the axes of the first and the second dimensions (α = 0.05). (d) Boxplots of the 14 lipids from PCA discriminating mice fed MCDD. ^$p < 0.05 compared to mice fed ND; ^‡p < 0.05 compared to mice fed HFD based on ANOVA test follow by Student t-test. CE: cholesteryl ester; HFD: high-fat diet (red); FAME: fatty acyl methyl ester; FFA: free fatty acid; MCDD: methionine-choline deficient diet (green); ND: normal diet (black); PC: phosphatidylcholine; PI: phosphatidylinositol; SM: sphingomyelin. Most representative phospholipids: PC(C38:6) = PC(18:0/20:6); PC(C36:2) = PC(18:1/18:1); SM(C36:1) = SM(18:1/18:0); SM(C40:1): SM(18:1/22:0); SM(C40:2): SM(18:1/22:1).