DISTRIBUTION OF LIPID VESICLE SIZE AS POTENTIAL MARKER OF NAS SEVERITY: ASSESSMENT WITH MORPHOQUANT-NASH AND MRI
Sabrina Doblas1, Cindy Serdjebi2, Felicia Julea1, Karine Bertotti2, Valérie Paradis3,4, Bernard E. Van Beers1,5, Philippe Garteiser1
1 LBI, CRI – U1149 Inserm, Université de Paris, 75018 Paris
2 Biocellvia, 13001 Marseille
3 INDiD, CRI – U1149 Inserm, Université de Paris, 75018 Paris
4 Pathology department, Beaujon University Hospital, AP-HP, 92110 Clichy
5 Radiology department, Beaujon University Hospital, AP-HP, 92110 Clichy
Objectives: To identify quantitative histopathological biomarkers and non-invasive MRI surrogates which inform on the NAFLD activity score (NAS) in mice.
Methods: After approval by the ethical committee, mice were fed a choline-deficient diet to induce NASH (n = 30), a high-fat diet to induce simple steatosis (n = 30) or a standard diet in the control group (n = 14). Liver magnetic resonance imaging (MRI) was performed at 4, 10 and 16 weeks after diet introduction to measure liver fat fraction, transverse relaxation rate R2* [de Haan MRM 2011] and visco-elastic parameters including stiffness, elasticity, viscosity and damping ratio [Yin Eur Radiol 2019]. At each timepoint, livers were collected and automated quantitative histopathological analyses were performed on whole liver picrosirius-red stained sections using MorphoQuant-NASH to yield the macrosteatosis surface area (%), the shape of the size distribution of lipid vesicles (Zipf’s parameter α, independent from the amount of steatosis), F4/80-labelled area, and the number of hepatic crown-like structures (hCLS). NAS [Kleiner Hepatology 2005] was blindly graded by an experienced pathologist.
Results: Steatosis and α were visually different between NAS groups (Fig. 1). After multiple regression analysis including the steatosis surface area, α, the F4/80 labelling and hCLS, α was found to be a strong predictor of NAS (p < 0.0001, rpartial = -0,76) while F4/80 labelling was a moderate predictor (p = 0.01, rpartial = 0.3). Multiple regression analysis was then conducted to identify non-invasive parameters that would inform on these histopathological predictors of NAS. Among all MRI parameters, the best one reflecting α was R2* (p < 0.0001, rpartial = -0.65, Fig. 2) while the best parameter reflecting F4/80 labelling was the stiffness (p < 0.0001, rpartial = 0.62). Grading of NAS severity using α yielded AUROC of 0.973 (NAS ≥ 3), 0.972 (NAS ≥ 4) and 0.986 (NAS ≥ 5), while its non-invasive imaging surrogate R2* yielded AUROC of 0.976, 0.969 and 0.984, respectively (Fig. 3).
Conclusion: NAS severity is characterized by altered size distribution of MorphoQuant-NASH-derived lipid vesicles which can be assessed non-invasively with MRI measurements of R2*.
