Non-alcoholic fatty liver disease (NAFLD) refers to a spectrum of diseases ranging from simple steatosis (isolated hepatic fat accumulation) to steatohepatitis (NASH; hepatic fat accumulation with lipotoxicity, hepatic cell damage and inflammation), eventually progressing to fibrosis, cirrhosis and potentially hepatocellular carcinoma. NAFLD is a common disease highly associated with the Metabolic Syndrome. Hence, it has been rising in prevalence in parallel with the increasing incidence of diabetes and obesity. The transition between the different stages of the disease could be a discriminant for a benign prognosis or a higher mortality risk due to cardiovascular or chronic liver disease. The diagnosis of NAFLD can be based on imaging studies (e.g., ultrasound technology or MRI spectroscopy). However, the specific stage of the disease, the presence of hepatocellular damage, inflammation, and amount of fibrosis in the liver can be detected only by biopsies. Such invasive methods cannot be applied outside specialist practice, present a significant risk of complications and are subject to sampling error; thus, they constitute an imperfect gold standard. To date, there is a lack of 1) tractable non-invasive biomarkers that could aid the diagnosis, risk stratification and monitoring of patients; 2) approved therapies; 3) reliable pre-clinical models of the disease to test drug and biomarker effectiveness/accuracy. A better understanding of NAFLD/NASH pathophysiology in humans will help achieve identifying new targets, improving the management of NAFLD patients. During my PhD, I have integrated and analysed transcriptomics data from rodent models and human NAFLD patients of different stages of the disease to identify specific mechanisms that explain the progression between the different stages of the disease. I have also compared these data to preclinical disease models to rank them against human pathophysiology and point to the most desirable ones. The thesis is organised into six chapters. In addition to the “Introduction” and the “Final Discussion”, the four “Results” chapters address the following topics:

1. Molecular characterisation of rodent models to determine their suitability for preclinical studies using human data as reference. Using an unbiased approach that allows ranking the murine models based on metabolic phenotyping, histology, and transcriptomics (compared to human data), I have ranked multiple preclinical models of NAFLD based on their “proximity to human disease”. My results suggest that rodents fed with diets enriched in refined carbohydrates, saturated lipids, and cholesterol (Western Diets), with/without sugar water (American Lifestyle approach) are the closest models to human NASH and, therefore, the most representative of human pathophysiology. Additionally, some genetic models of obesity (ob/ob, MC4r) augment liver damage induced by these diets, making them valuable tools to achieve more advanced disease stages and/or faster models.
2. Identification of the molecular landscape of NAFLD progression. All previous work describing NAFLD progression has been based on the division of the datasets in artificially defined, discrete disease stages. Here, I implemented a pseudotemporal ordering method that successfully captures the disease trajectory in a continuum. Based on the pathway enrichment and upstream regulator analysis results, I show that my analysis matches the results of the discrete approach supervised by histology, providing additional granularity and better defining the transition among disease stages. An expression-module-based analysis also defined relevant processes representing the molecular signature of the disease during its progression. Using Random Forest models, I identified a list of genes predictive of the disease progression; intriguingly, this list of hits features multiple drivers of NASH well characterised mechanistically plus novel targets, worth future investigation. I expect this detailed molecular profile of NAFLD progression to help understand the mechanisms underpinning NASH progression and improve the stratification of patients and the ranking of pre-clinical models.
3. Comparison of NAFLD progression for patients with and without Type 2 diabetes. Using transcriptomics data of NAFLD patients with and without diabetes, the aim is to answer whether there are differences between the two populations in NAFLD progression. Utilising the pseudotemporal ordering approach mentioned above, I identified specific T2DM-associated biological processes that due to the complex nature of the disease and the limitations of relatively small human datasets, could not be identified with standard bioinformatics approaches.
4. Contribution of MBOAT7 and INSIG1 in NAFLD progression in mice and humans. This chapter provides a proof of concept of how transcriptomic studies can be reverse-translated into mice to study the contribution of the targets I identified in mechanisms of NASH progression. Here I show that the partial genetic ablation of MBOAT7 alters the hepatic transcriptome and the dependence of these effects on the dietary challenge. Additionally, I show that an INSIG1 ablation induces an increased lipid remodelling and cholesterol biosynthesis, acting as a protective mechanism that prevents NASH progression. Overall, my work paves the way for a better understanding of the NAFLD disease progression and defines new approaches to study NASH with system biology and translational approaches, exploiting novel methods that had never been used in NAFLD research.