Abstract
Non-alcoholic fatty liver disease (NAFLD) is now the most common liver disease worldwide. Given that NAFLD can progress from steatosis to non-alcoholic steatohepatitis (NASH), fibrosis and potentially hepatocellular carcinoma, early diagnosis and accurate disease staging are primary clinical concerns. Hypothesizing that a subset of liver proteins will exhibit differential expression in NAFLD and that these proteins may represent candidate disease biomarkers; the aims of this project were to use proteomics to identify differentially expressed proteins both in an in vitro and an in vivo model of NASH. Preliminary studies developed and characterised both models used here; experiments utilized a relative quantitative proteomic approach with isobaric tags for relative and absolute quantitation labelling combined with nano-liquid chromatography and tandem mass spectrometry. Initial proteomic experiments in vitro with palmitate treated human hepatocytes (HuH7 cell line), identified 298 proteins with high stringency (>2 peptides and false discovery rate < 0.05). Application of a novel unsupervised machine learning approach, (quantitation state representation; QSR) developed specifically for these experiments, resulted in the identification of twelve differentially expressed proteins. Remarkably, the three proteins (peroxiredoxin, glyoxalase 1 and non-specific lipid transfer protein) that were selected for follow up in immunocytochemistry and immunoblotting experiments, were all confirmed as modulated in response to palmitate-induced lipid loading, validating the QSR strategy. In vivo, ApoE(-/-) animals clearly developed the histological features of NASH following 12 weeks of a high fat/high cholesterol diet (HFD) and subsequently, this time point was used for liver proteome analysis. Across biological replicates, an average of 249 cytosolic and 343 membrane proteins were identified with high stringency (0.05 FDR; >2 peptides), and quantified by Scaffold Q+S software. Of these, 83 cytosolic and 80 membrane proteins were found to be differentially expressed with statistical significance by both randomized permutation and ANOVA tests. Liver fatty acid binding protein and fatty acid synthase, both previously found down-regulated in NASH patients, were among those proteins identified as down-regulated (-1.2; P=4.3E-15 and -0.9; P=7.9E-11) in the HFD-fed ApoE-/- animals. Network analysis demonstrated major disruptions to fatty acid, amino acid and carbohydrate metabolic pathways. Strikingly, glyoxalase 1 and non-specific lipid transfer protein were identified as differentially expressed in both the in vitro and in vivo models and so were independently examined in clinical biopsy samples from paediatric patients with NAFLD. Non-specific lipid transfer protein and glyoxalase 1 were both confirmed as down-regulated in the biopsies examined. In conclusion, these data support the hypothesis that a subset of hepatic proteins will exhibit differential expression in NAFLD. The identification of non-specific lipid transfer protein and glyoxalase 1 as dysregulated in the liver of paediatric NASH patients illustrates that proteomics offers the potential to identify protein biomarkers of potential clinical relevance to NAFLD.