Exploratory Data Analysis of Cell and Mitochondrial High-Fat, High-Sugar Toxicity on Human HepG2 Cells

• Main Authors: Ricardo Amorim, Inês C. M. Simões, Caroline Veloso, Adriana Carvalho, Rui F. Simões, Francisco B. Pereira, Theresa Thiel, Andrea Normann, Catarina Morais, Amália S. Jurado, Mariusz R. Wieckowski, José Teixeira, Paulo J. Oliveira
• Format: Article
• Language: English
• Published: MDPI AG 2021-05-01
• Series: Nutrients
• Subjects: non-alcoholic fatty liver disease (NAFLD); in vitro cell model; Hepg2 cells; lipid accumulation; mitochondria dys(function); exploratory data analysis
• Online Access: https://www.mdpi.com/2072-6643/13/5/1723

Non-alcoholic steatohepatitis (NASH), one of the deleterious stages of non-alcoholic fatty liver disease, remains a significant cause of liver-related morbidity and mortality worldwide. In the current work, we used an exploratory data analysis to investigate time-dependent cellular and mitochondrial effects of different supra-physiological fatty acids (FA) overload strategies, in the presence or absence of fructose (F), on human hepatoma-derived HepG2 cells. We measured intracellular neutral lipid content and reactive oxygen species (ROS) levels, mitochondrial respiration and morphology, and caspases activity and cell death. FA-treatments induced a time-dependent increase in neutral lipid content, which was paralleled by an increase in ROS. Fructose, by itself, did not increase intracellular lipid content nor aggravated the effects of palmitic acid (PA) or free fatty acids mixture (FFA), although it led to an up-expression of hepatic fructokinase. Instead, F decreased mitochondrial phospholipid content, as well as OXPHOS subunits levels. Increased lipid accumulation and ROS in FA-treatments preceded mitochondrial dysfunction, comprising altered mitochondrial membrane potential (ΔΨm) and morphology, and decreased oxygen consumption rates, especially with PA. Consequently, supra-physiological PA alone or combined with F prompted the activation of caspase pathways leading to a time-dependent decrease in cell viability. Exploratory data analysis methods support this conclusion by clearly identifying the effects of FA treatments. In fact, unsupervised learning algorithms created homogeneous and cohesive clusters, with a clear separation between PA and FFA treated samples to identify a minimal subset of critical mitochondrial markers in order to attain a feasible model to predict cell death in NAFLD or for high throughput screening of possible therapeutic agents, with particular focus in measuring mitochondrial function.