Mechanistic analyses of hepatocyte plasticity

Abstract

Adult hepatocytes are highly plastic cells which have a molecular toolbox for efficient regeneration. Yet, end stage liver diseases are becoming an increasing clinical burden. Currently, most knowledge on the underlying mechanisms of regenerative processes in the liver as well as pre-clinical prediction of drug induced liver injury stems from models with limited translational capacity. To narrow the translational gap in these fields, we used a recently developed 3D culture system for adult primary hepatocytes which faithfully recapitulates in vivo phenotypical features.

Experiments using 3D spheroids revealed that dedifferentiation of hepatocytes which entails loss of hepatocyte nuclear factor (HNF) activity and NFκB induction is a crucial step that “primes” human hepatocytes to re-enter cell cycle. We also found that activation of Wnt/β-catenin constitutes the major mitogenic signal which actively drives “primed” hepatocytes to cell cycle progression. Furthermore, this 3D culture system represents a scalable tool for screening novel molecular probes that modulate hepatocyte regeneration.

In addition, we characterized major changes at the whole-transcriptome level that occur upon isolation of primary hepatocytes from the in vivo niche, and lead to cell dedifferentiation. These changes are transient in 3D culture, as they last only during the few first days, while in 2D monolayers they persist over culture time. Importantly, the transient reprogramming in 3D-aggregating human hepatocytes highly resembled dedifferentiation phenomena of regenerating livers after partial hepatectomy.

Finally, by comparing the exposure outcomes of new anti-cancer drugs in 3D spheroids vs 2D cultured hepatocytes, we identified atypical CYP gene induction only in the latter. This occurred because of inhibition of the inherent dedifferentiation in the 2D method and indicates that this strategy, which is a current standard for pre-clinical assessment of drug safety issues, is of suboptimal predictive accuracy.