Validity of biopsy-based drug effects in a diet-induced obese mouse model of biopsy-confirmed NASH

The prevalence of nonalcoholic fatty liver disease (NAFLD) is increasing worldwide alongside the increased incidences of diabetes and obesity [1, 2]. NAFLD ranges from benign nonalcoholic fatty liver (NAFL) with simple steatosis to the necroinflammatory state non-alcoholic steatohepatitis (NASH) and cirrhosis [3‐5], which is estimated as the leading cause of end stage liver disease within a few years [6‐8]. To date there is no licensed treatment for NASH, however, numerous clinical trials are ongoing [9]. Most advanced are obeticholic acid (a farnesoid X nuclear receptor (FXR) agonist), elafibranor (a dual peroxisome proliferator-activated receptor (PPAR)-α/δ agonist), selonsertib (an apoptosis signal-regulating kinase 1 (ASK1) inhibitor), and liraglutide (a long-acting glucagon-like peptide-1 (GLP-1) analogue) [9].

In addition to the difficult task of developing therapeutics for NASH, clinical diagnosis and follow-up data are hampered by the unmet need for reliable non-invasive diagnostic and prognostic tools [10, 11]. NASH development is unpredictable and vary in both disease severity and progression rates [12]. Non-invasive imaging procedures, including ultrasonography, magnetic resonance imaging (MRI) and magnetic resonance elastography (MRE) have shown potential in diagnosing NAFL and can be repetitively performed during the disease monitoring period [11]. However, their utility is inadequate due to a lack of sensitivity to differentiate between intermediate levels of fibrosis severity, their limited availability and associated costs [11]. Highly sensitive and predictive blood chemistry tests for circulating surrogate biomarkers of liver injury have still not reached FDA approval [11]. Accordingly, invasive and risky paired liver biopsies still remain the gold standard for staging and grading of NASH, and for monitoring drug efficacy in clinical trials [13‐17].

To aid the development of pharmaceutical therapeutics, animal models reflecting the clinical NASH phenotype are of uppermost importance. Several models have been developed and are generally categorized into diet-induced, chemically-induced, or genetic models (knockout or transgenic) [18]. Different obesogenic Western-type diets have proven to promote a NASH phenotype in mice, though the disease severity is often mild [19]. However, when kept ≥26 weeks on a diet high in fat, fructose, and cholesterol (the Amylin liver NASH diet; AMLN [20, 21]), C57BL/6J mice have been shown to develop the hepatic pathological hallmarks of NASH, including steatosis, lobular inflammation, and ballooning degeneration, as well as mild to moderate fibrosis [21‐26]. These hallmarks are further accentuated in leptin-deficient Lep^ob/Lep^ob mice [20, 25, 27, 28]. The pharmacological efficacy on metabolic and hepatic endpoints have already been extensively characterized in these models [20, 26, 29].

Elafibranor, INT-767 and liraglutide have previously been shown to induce diverse pharmacodynamic effects on liver histopathology [20, 26, 29‐34]. The three compounds represent three completely different drug classes with three different mechanism of action [29, 35‐37] and are also known to affect total liver mass. While findings based on small tissue biopsies are encouraging, no studies have previously used gold standard stereological sampling to evaluate the homogeneity of liver morphometry across liver lobes nor to evaluate the validity of liver biopsy assessments to reflect pharmacologically induced changes on the whole mouse liver. This study aims to evaluate if biopsy-based quantitative image analysis efficiently reflects whole liver remodelling following drug treatment by comparison with stereology-based quantitative digital image analysis of the whole liver.

The present study aimed to verify the validity of a liver biopsy, representing less than 1 % of the total liver, to reflect whole liver disease remodeling following pharmaceutical treatment in male Lep^ob/Lep^ob-NASH mice. By comparing morphometric analyses on biopsies with stereologically sampled sections across the whole liver, we demonstrate that the biopsy is overall representative of the whole liver status and is applicable for preclinical evaluation of pharmacological intervention studies. Notably, however, we also demonstrate that pharmacologically induced effects on liver weight should be carefully considered when comparing NASH related endpoints in preclinical studies.

Whereas liver lipid content showed little variation within and between lobes, intra-lobe variability was more evident for both relative gal-3 and col1a1 levels. This difference is not surprising and emphasizes the need to take biopsies in the same part of the lobe when comparing tissue dynamics between different animals, or to use unbiased stereological sampling principles covering the whole liver. The differences in col1a1 levels is mainly related to the fraction of Glisson’s capsule in the tissue section. The Glisson’s capsule [40], a collagenous layer covering the liver, increases in thickness during progression of fibrosis [41, 42]. Accordingly, levels of col1a1 was markedly higher at apical parts compared to slaps containing a higher ratio of central parts of the lobe. The same reason may apply for the variability of gal-3, as, macrophage-derived gal-3 is known to be linked to myofibroblasts and hence fibrosis [43, 44].

In addition to the in-depth assessment of lobe variability and the validation of liver biopsy assessments in mice, we characterized the effects of liraglutide, elafibranor, and INT-767 in Lep^ob/Lep^ob-NASH mice. Liraglutide, a human GLP-1 analogue, is already FDA approved for the treatment of obesity (Saxenda®) and type 2 diabetes (Victoza®) [45, 46], and is in addition to it's well-described incretin effects [47], also reported to improve liver enzymes, oxidative stress, and steatosis [20, 26, 30, 31, 37, 48]. In contrast, elafibranor, a high-affinity agonist for PPAR-α/δ, exerts its effect on NASH amelioration mainly by increasing clearance of fatty acids, as well as inhibition of pathways involved in inflammation and fibrosis [26, 33, 34]. Finally, INT-767, a dual FXR and transmembrane G-protein-coupled receptor 5 (TGR5) agonist, dose dependently reduce cholesterol and liver triglyceride levels, reduce steatosis, inflammation, and fibrosis stage [29, 32]. In human liver and plasma samples both FXR and TGR5 levels correlates with NAFLD disease severity [49‐51]. All three compounds exerted marked effects on relative liver lipid content, whereas only INT-767 and elafibranor affected relative gal-3 levels. Only stereology-based assessment of INT-767 efficacy revealed improvements on relative col1a1 levels. However, when incorporating compound specific effects on liver size, both INT-767 and liraglutide significantly improved liver fibrosis, as well as total liver lipid and inflammation. In contrast, elafibranor did not reduce total col1a1, as also reported previously in both C57bl/6 and Lep^ob/Lep^ob mice [26]. Thus, the presented data highlight the importance of looking at whole organ dynamics, instead of reporting relative values. Since liraglutide and INT-767 significantly reduce liver weight, mainly by reducing lipid content, relative values of col1a1 and gal-3 content would tend to show no regulation or even upregulation if not affected directly by the compound. Conversely, the peroxisome proliferating mechanism of elafibranor, which may lead to hepatomegaly in rodent models of NASH [26], would indirectly lead to biased reduced relative values of all other liver components if not addressed directly.

It should be noted that the comparison was based on image analyses and not a histopathological assessment of NAFLD activity scores and fibrosis stage, as reported previously [26]. Image analysis allows for an objective analysis of the liver histomorphology, whereas scoring and staging by a trained pathologist is more subjective. Image analysis of relative hepatic lipid levels is based on the actual amount of lipids in a histologic section (i.e. area or volume fractions) [52], whereas steatosis scores are graded based on the percentage of hepatocytes having lipid droplets, irrespective of the size of the lipid droplets [38]. Similarly, staging of fibrosis is based on the localizations of fibrotic bands, and not the area or thickness of fibrotic bands which is estimated in image analysis [25, 26, 53]. Lastly, scoring of lobular inflammation depends on the number of inflammatory foci (clusters of inflammatory cells) in 200X field of view [38], and not the total content of inflammatory cells (here assessed by gal-3 IHC). Thus, image analyses of NASH components are not necessarily directly correlated to histopathological scoring and staging. This inherent variability is an appropriate feature of clinical studies where only a small fraction of the organ can be sampled. However, at the preclinical stage, when one is trying to differentiate compounds within or across modalities it is readily possible to gain a more accurate assessment of the true disease state of the total organ.

Finally, it should be stated that the analyses presented here were based on a quantitative assessment of gal-3 and col1a1 immunohistochemistry. These “pan-markers” of inflammation and fibrosis are used extensively in preclinical and clinical research but may of course not represent all inflammatory of extracellular matrix remodelling during NASH development. Accordingly, the validity of biopsy-based drug efficacy presented here should be considered cautiously for other markers.

In conclusion, we report that a liver biopsy can be considered representative for the remodeling occurring in the entire liver of Lep^ob/Lep^ob-NASH mice following pharmaceutical treatment, though changes are slightly different using an in-depth stereological assessment of the whole organ, as evidenced here for liraglutide and INT-767 for col1a1 assessments. In a recent study, repeated liver biopsies were extracted from the left lateral lobe, medial right lobe and medial left lobe in mice over a period of three months [54]. Although these data showed that repeated liver biopsies from different lobes were feasible, our data suggests that continuous biopsy-based measurements should be performed in the same lobe. Finally, we highlight the importance of introducing effects on total liver remodeling when assessing liver histomorphometry, as clearance of steatosis or hepatomegaly would bias relative values.