Activation of thyroid hormone receptor‐β improved disease activity and metabolism independent of body weight in a mouse model of non‐alcoholic steatohepatitis and fibrosis

Activation of hepatic thyroid hormone receptor β (THR‐β) is associated with systemic lipid lowering, increased bile acid synthesis, and fat oxidation. In patients with non‐alcoholic steatohepatitis (NASH), treatment with THR‐β agonists decreased hepatic steatosis and circulating lipids, and induced resolution of NASH. We chose resmetirom (MGL‐3196), a liver‐directed, selective THR‐β agonist, as a prototype to investigate the effects of THR‐β activation in mice with diet‐induced obesity (DIO) and biopsy‐confirmed advanced NASH with fibrosis.


| INTRODUCTION
Non-alcoholic fatty liver disease (NAFLD) describes a spectrum of liver abnormalities ranging from fatty liver or simple steatosis to non-alcoholic steatohepatitis (NASH) without or with hepatic fibrosis.
NAFLD is a common condition, with a global prevalence of approximately 25%, strongly linked to obesity, diabetes, and systemic dyslipidaemia (Younossi et al., 2016). NASH is characterized by marked hepatic steatosis, lobular inflammation and hepatocyte ballooning (Diehl & Day, 2017). The prevalence of NASH is increasing (Estes, Anstee, et al., 2018;, and NASH with advanced fibrosis is associated with a strong increase in liver-related and overall mortality (Dulai et al., 2017;Ekstedt et al., 2015). No pharmacological therapy is approved for the treatment of NASH, though some experimental drugs are currently in later stages of clinical development (Garber, 2019). First-line therapy is lifestyle intervention with a focus on weight loss (Chalasani et al., 2018). Thus, new therapies for advanced NASH with fibrosis are urgently needed.
The thyroid hormone receptor β (THR-β) is the most abundant THR isoform in the liver but is also expressed in other tissues. Its expression in the liver is reduced in patients with NASH (Krause et al., 2018). Activation of hepatic THR-β is associated with systemic lipid lowering, increased bile acid synthesis and fat oxidation (Sinha et al., 2019). Liver-directed agonists with high selectivity for THR-β over THR-α have been generated (Erion et al., 2007;Kelly et al., 2014) and have shown metabolic benefits in preclinical models of diabetes and obesity, including reduced hepatic steatosis and inflammation (Sinha et al., 2019). However, there is little information on their activity in obese animals with manifest NASH and fibrosis. Recently, resmetirom (MGL-3196) was evaluated in a 36-week clinical phase 2 study in patients with biopsy-confirmed NASH, was well tolerated, and led to a significant reduction in hepatic fat fraction, circulating LDL cholesterol and triglycerides, and a higher rate of NASH resolution, compared to placebo (Harrison et al., 2019).
Here, we report the evaluation of resmetirom as a prototype of a liver-directed, selective THR-β agonist in a diet-induced obese (DIO) and biopsy-confirmed mouse model of advanced NASH with fibrosis. This is, to our knowledge, the first study describing the effects of resmetirom in a preclinical model of NASH, fully replicating recent clinical observations. Thirty-six male C57BL/6J mice (5 weeks old), obtained from JanVier (JanVier Labs, France), were included in the study. Before treatment with resmetirom, animals had ad libitum access for 34 weeks to a regular rodent diet (n = 12; Altromin 1324) or a diet high in fat (40%), of these 18% trans-fat, 40% carbohydrates (20% fructose), and 2% cholesterol (n = 24; ) previously described as the Amylin Liver NASH diet (AMLN diet) (Clapper et al., 2013) and tap water. Mice were group housed at five animals per cage (UNO Plastic cages Type 3 for group housing, 427 × 267 mm, floor area 840 cm 2 ) under a 12/12-h dark-light cycle (lights off at 3 pm).

| Preparation of model and study design
The effects of THR-β activation were investigated in mice with dietinduced obesity and biopsy-confirmed NASH and fibrosis (DIO-NASH model), as described earlier by Kristiansen et al. (2016). The study design is illustrated in Figure 1a. A baseline liver biopsy was conducted 3 weeks before the intervention for histological assessment of individual fibrosis (stage ≥1) and steatosis (score ≥2), as described (Kristiansen et al., 2016). After the baseline biopsy,

What is already known
• Resmetirom reduces hepatic steatosis in patients with non-alcoholic steatohepatitis (NASH).

What this study adds
• Resmetirom improves liver histopathology in dietinduced obese mice with NASH without causing weight loss.
• Whole genome expression analysis indicates activation of canonical THR-β signalling pathways and down-regulation of fibrogenesis.

What is the clinical significance
• Long-term resmetirom treatment may lead to resolution of hepatic fibrosis. animals were single housed (UNO Plastic cages Type 1 L Cage for single caging, 300 × 120 mm, floor area 360 cm 2 ) and remained so until the end of the study. A week before the intervention, the animals were randomized into groups of equal size and stratified according to liver Col1a1 quantification into three groups: Group 1, lean-chow control (n = 12); Group 2, DIO-NASH + vehicle (n = 12); and Group 3, DIO-NASH + resmetirom (n = 12). Group sizes were selected based on previous studies with other drugs in the same model (Tølbøl et al., 2018). Group 1 contained all animals that had been on the regular rodent chow diet, whereas mice on the AMLN diet were randomized to Groups 2 and 3. Resmetirom was administered once daily, in the morning, by oral gavage at a dose of 3 mgÁkg −1 . The vehicle used for compound formulation and control injections was 0.6% methyl cellulose with 0.5% Tween 80.
The resmetirom concentration in the dosing solution was 0.6 mgÁml −1 ; injection volume was 5 mlÁkg −1 body weight. The intervention lasted for a period of 8 weeks. The DIO-NASH mice in the vehicle control and resmetirom groups remained on the NASHinducing AMLN diet throughout the intervention period. At the end of the intervention, animals were killed in the fed state by heart puncture under isoflurane anaesthesia, and liver tissue and plasma were collected. One animal from the vehicle control group showed abnormal morphology in the portal tract on study termination and was therefore excluded from further analyses.
The dose of 3 mgÁkg −1 was selected based on the dose-response to resmetirom (MGL-3196) on cholesterol lowering in DIO mice (Kelly et al., 2014). For mice with a body weight of 38 g (Figure 1b), this would correspond to a human equivalent dose of about 19 mg for an 80-kg person (Nair & Jacob, 2016) which is about four times lower than the dose of 80 mg used in a recent phase 2 study in NASH patients (Harrison et al., 2019). Thus, it can be assumed that resmetirom was not overdosed in our study, though the relationship between exposure profile and target engagement is not known and may be different across species.
F I G U R E 1 (a) Study design. (b) Body weight development over the intervention period. (c) Food intake over the course of the study. N = 12 per group, N = 11 for vehicle control group. *P < .05, significantly different from DIO-NASH vehicle control Body weight determination, body composition analysis, blood sampling, plasma biochemistry, endotoxin determination and liver tissue biochemistry were performed in a blinded fashion as previously described (Kristiansen et al., 2016).

| Histological analyses
For histology, baseline liver biopsy and terminal samples were collected from the left lateral lobe (about 50-100 mg at baseline and 200 mg at the end) and fixed overnight in 4% paraformaldehyde.
Sections were stained with haematoxylin and eosin and Sirius Red to assess hepatic steatosis and fibrosis, respectively, followed by analysis with Visiomorph software (Visiopharm, Denmark). Collagen Antibodies were diluted in 5% pig serum; solutions were used only once. A pathologist blinded to the study performed the histological assessment and scoring. NAFLD activity score (NAS) combining steatosis, lobular inflammation, and hepatocyte ballooning scores and fibrosis stage were quantified applying the criteria proposed by Kleiner et al. (2005).

| RNA sequencing (gene expression analysis)
Liver tissue was harvested from the left lateral lobe, stabilized over-

| Data and statistical analyses
For all experiments, statistical analysis was only undertaken if group or sample size was at least n = 5, with group size referring to the number of independent values and not technical replicates. All available data were included in the analysis without removal of potential outliers. The data and statistical analysis comply with the recommendations of the British Journal of Pharmacology on experimental design and analysis in pharmacology (Curtis et al., 2018). Values of body weight, body composition analysis, blood sampling, plasma biochemistry, endotoxin determination and liver tissue biochemistry were analysed by one-way ANOVA with Dunnett's test. Data from histology and immunohistochemical analyses were assessed for differences in the number of animals showing improvement or worsening, using a one-sided Fisher's exact test with Bonferroni correction with the DIO-NASH vehicle group as comparator. Results from the gene expression analysis were corrected for variable multiplicity and falsediscovery rate (FDR)-adjusted P values calculated, using the Benjamini-Hochberg correction (Benjamini & Hochberg, 1995). Differentially expressed gene data were further interrogated on pathways and further causal relationships through the use of Ingenuity Pathway Analysis (RRID:SCR_008653, Ingenuity Qiagen) as described in Krämer et al. (2014). For all statistical analyses, P values below 0.05 were considered to indicate significant differences between groups. 3 | RESULTS Figure 1b shows the change in body weight over the treatment period of 8 weeks. Resmetirom had no effect on weight or food intake ( Figure 1c) and was well tolerated throughout the treatment period with no obvious adverse effects or signs of toxicity. Body weight remained stable throughout the treatment period. The transient weight loss observed at day 39 in all three groups resulted from the 4-h fasting period before taking blood for glucose determination.

| Materials
Activation of THR-β had a profound effect on plasma transaminase activities, leading to a reduction by about 50% for alanine aminotransferase (ALT) and 26% for aspartate aminotransferase (AST), compared with the DIO-NASH group treated with vehicle controls (Figure 2a,b). Plasma total cholesterol levels that were strongly elevated in DIO-NASH mice, compared with lean controls, were nearly normalized upon treatment with resmetirom ( Figure 2c). There was a trend towards lower plasma triglycerides that was, however, not statistically significant (Figure 2d). Four-hour fasting blood glucose levels determined on treatment day 39 were significantly lower in resmetirom-treated animals ( Figure 2e) whereas resmetirom had no influence on plasma insulin concentration (Figure 2f).
Liver hypertrophy observed in DIO-NASH mice was markedly reduced upon treatment with the THR-β agonist (Figure 3a), accompanied by a strong reduction in liver triglycerides, total cholesterol, and hepatic total lipid content (Figure 3b-d). Liver col1a1, α-SMA, and gal-3 contents were lower after resmetirom treatment (Figure 4). The difference from vehicle-treated DIO-NASH mice remained statistically significant upon normalization to liver weight for α-SMA, but not for col1a1 (not shown).
Resmetirom treatment was associated with changes in liver morphology (Figure 5a,b) and a significant improvement (posttreatment vs. pretreatment biopsy) in the histological NAS, relative was significantly reversed by treatment with resmetirom ( Figure S3b and Table S1). Furthermore, DIO-NASH vehicle-treated mice demonstrated a differential down-regulation (>1.5-fold) of 1,750 genes in comparison with lean control mice, with 79 of these transcripts being significantly reversed by resmetirom treatment.

| DISCUSSION
Here, we report that the liver-selective THR-β agonist resmetirom, administered at a dose that did not influence body weight or food intake, had a profound effect on hepatomegaly, hepatic steatosis, liver injury, circulating cholesterol, and the histological NAS in a DIO and biopsy-confirmed mouse model of advanced NASH with fibrosis. Our study was performed as an intervention (therapeutic) study in DIO-NASH mice with established metabolic disease also comprising obesity, hypercholesterolaemia, and insulin resistance, reflecting clinical characteristics of patients with NASH.
Our results are in very close agreement with a recently reported Phase II trial of resmetirom in patients with NASH in which, after 36 weeks of treatment, a 40% reduction in hepatic fat content measured by MRI was observed that was accompanied by a significant improvement in circulating liver transaminases, total and LDL F I G U R E 4 Representative immunohistochemistry images and quantitative analysis of (a) hepatic collagen 1α1, (b) liver α-SMA, and (c) liver galectin-3. The length of the scale bar is 85 μm. N = 12 per group, N = 11 for vehicle control group. Data shown are individual values with means ± SEM (shown as box). *P < .05, significantly different from DIO-NASH vehicle control cholesterol. Furthermore, resmetirom responders that had lower hepatic fat fractions at 36 weeks also exhibited significantly reduced NASs (Harrison et al., 2019).
Several THR-β agonists have been investigated in animal models of metabolic disease with fatty liver (see Sinha et al., 2019).
Sobetirome (GC-1) and eprotirome (KB2115), for example, reduced hepatic steatosis and liver triglycerides in ob/ob mice (Martagon et al., 2015). Notably, at the selected doses, treatment also resulted in significant body weight loss, and ob/ob mice do not develop manifest NASH or hepatic fibrosis under the investigated conditions. Similarly, both compounds, administered over a period of 10 days, led to strong reductions in liver triglyceride content in Sprague-Dawley rats on a high-fat diet (Vatner et al., 2013) that, however, did not develop NASH. Sobetirome also caused a marked decrease in hepatic and circulating triglycerides and serum ALT in rats fed a high-fat, methionine-and choline-deficient diet (Perra et al., 2008). However, treatment was accompanied by weight loss, and the presence of NASH was not investigated. MB07344/VK2809, a THR-β agonist prodrug selectively taken up by the liver and converted into the active form, reduced hepatic triglycerides and serum lipids in C57BL/6 mice on a high-fat diet and in Zucker diabetic fatty rats (Cable et al., 2009;Erion et al., 2007), in the absence of NASH and fibrosis. Thus, there is evidence from several models of metabolic disease for an improvement in liver fat and circulating lipids. Our results add to this body of evidence by showing that activation of THR-β can reduce hepatic steatosis and liver injury in a DIO model combining metabolic syndrome with advanced NASH and fibrosis.
Many preclinical models of NAFLD have been described that vary in their translatability to the human situation (Hansen et al., 2017).
Our data indicate that with respect to THR-β agonists, data obtained in biopsy-confirmed DIO-NASH mice have a good human translatability. So far, it is not established to which extent resmetirom is able to also hold or to reverse fibrosis in late-stage NAFLD. In the recent Phase II trial of resmetirom in patients with NASH and predominantly mild fibrosis, no significant improvement was observed (Harrison et al., 2019). It is unclear whether this results from lack of efficacy or whether the treatment duration of 36 weeks was too short to elicit an observable improvement in hepatic fibrosis. Similar to the population investigated in this trial, our model showed mild fibrosis (approximately 60% F2, 40% F3; see Figure 5d), and no significant improvement in fibrosis score was seen with resmetirom. However, reduced expression of markers of fibrogenesis in resmetirom, compared with vehicle-treated DIO-NASH animals indicates an anti-fibrotic effect of THR-β activation that may translate into a more robust reduction in F I G U R E 5 (a) Representative images of liver morphology at termination of the study. H&E staining (20×, scale bar = 100 μm), (b) Sirius red staining (20×, scale bar = 100 μm). (c) Change in NAFLD activity score (pretreatment compared to post-treatment) for individual animals in the different treatment groups: Data shown are from lean control mice, DIO-NASH vehicle controls (Vehicle PO) and DIO-NASH mice treated with resmetirom (Resmetirom). (d) Change in fibrosis stage; data as in (c ). N = 12 per group, N = 11 for vehicle control group. *P < .05, significantly different from DIO-NASH vehicle group; one-sided Fisher's exact test with Bonferroni correction F I G U R E 6 Hepatic gene expression assessed by RNA sequencing with subsequent Ingenuity Pathway Analysis: (a) Principal component analysis over all 53,431 measured transcripts displaying the first two main components, PC1 and PC2. (b) Visualization of upstream factor analysis performed with Ingenuity, which predicts THR-β as one important factor activated in the resmetirom-treated group. (c-e) Expression of selected genes in normalized counts (FKPM) and shown as box-whisker blots with median (line) and mean (+) and confidence intervals of 5% to 95%. *P < .05, significantly different from DIO-NASH vehicle group (false-discovery rate adjusted for multiplicity by Benjamini-Hochberg correction) fibrosis upon longer duration of treatment. Thus, the 8-week treatment schedule in our study, chosen on the basis of previous studies in the same model with different types of intervention (Kannt et al., 2019(Kannt et al., , 2020Tølbøl et al., 2018) may have been too short for a significant improvement in fibrosis score. Furthermore, as there was no worsening of fibrosis staging in the vehicle control group over the course of the intervention period (Figure 5b), prevention of fibrosis worsening by resmetirom could not be assessed. It is also possible that longer pre-feeding with the AMLN diet might further aggravate advanced fibrosis in DIO-NASH mice and could be a useful model for preclinical evaluation of the compound with respect to anti-fibrotic efficacy.
A drug for which an improvement in fibrosis has recently been observed is obeticholic acid (OCA): In a recently reported interim analysis of a Phase III trial, improvement of fibrosis by at least one stage, was seen in a significantly larger percentage of individuals treated with OCA, compared with placebo after 18 months (Younossi et al., 2019) which is about twice as long as the resmetirom Phase II trial (Harrison et al., 2019). In the DIO-NASH mouse model, OCA treatment over 8 weeks led, similarly to resmetirom in this study, to a reduction in expression of genes linked to fibrogenesis and a non-significant trend to an improvement in fibrosis score (Tølbøl et al., 2018). Thus, it is tempting to suggest that a longer treatment with OCA may also have translated into a more pronounced improvement in fibrosis in DIO-NASH mice or, likewise, longer treatment with resmetirom could lead to fibrosis improvement in patients.
Of note, although our whole genome expression analysis indicated activation of the canonical THR-β pathway, the data did not explain the strong reduction of hepatic lipids by resmetirom, as there was no clear down-regulation of fatty acid synthesis genes or stimulation of genes associated with fat oxidation. Thus, the anti-steatotic effect of the compound in this model is not due to transcriptional effects. It is possible that activation of THR-β redirects metabolic fluxes by post-transcriptional mechanisms.
An important limitation of our study is the restriction to a single dose level. The dose of 3 mgÁkg −1 was selected based on a previous study in DIO mice over 23 days where this dose led to $60% reduction in circulating cholesterol levels (Kelly et al., 2014). Whereas the chosen dose of resmetirom showed a marked improvement in hepatic and systemic metabolic health, in the absence of obvious adverse effects, additional studies investigating a range of doses will be required to, for example, establish the maximal tolerated or minimally or maximally effective doses of resmetirom for the treatment of NASH.

ACKNOWLEDGEMENT
Claire Kammermeier is gratefully acknowledged for excellent technical assistance.

AUTHOR CONTRIBUTIONS
A.K. and A.N.M. designed the study. A.K., P.W., A.N.M., and S.S.V. performed experiments and analysed data. M.F. and D.S. analysed data. A.K. wrote the paper. All authors edited and approved the manuscript.

RIGOUR
This Declaration acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research as stated in the BJP guidelines for Design & Analysis, Immunoblotting and Immunochemistry, and Animal Experimentation and as recommended by funding agencies, publishers, and other organizations engaged with supporting research.

CONFLICT OF INTERESTS
A.K., P.W., and D.S. are or were employees of Sanofi. A.N.M., S.S.V., and M.F. are employees of Gubra.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request. Some data may not be made available because of privacy or ethical restrictions.