How effective are nonalcoholic fatty liver disease models for drug discovery?

ABSTRACT

The spectrum of nonalcoholic fatty liver disease (NAFLD) ranges from simple steatosis to nonalcoholic steatohepatitis (NASH) with hepatic fibrosis up to liver cirrhosis and hepatocellular carcinoma, displaying a global health problem with no effective therapy yet. Multiple preclinical models reflecting different aspects of the disease helped to identify a variety of different targets over the last years. However, some recent clinical trials have revealed a lack of translatability, emphasizing the need for more effective preclinical research. In this editorial, we discuss different NAFLD mouse models as well as emerging ex vivo and in vitro models that have been used in drug discovery and dissect the translational challenges that have to be considered in drug development.

KEYWORDS:

• Nonalcoholic fatty liver disease (NAFLD)
• nonalcoholic steatohepatitis (NASH)
• drug discovery
• preclinical models
• liver fibrosis

1. Introduction

The worldwide prevalence of nonalcoholic fatty liver disease (NAFLD) and nonalcoholic steatohepatitis (NASH), the severe inflammatory stage of NAFLD, has dramatically increased over the last decade and has become the main cause of chronic liver diseases, ultimately leading to advanced fibrosis up to hepatocellular carcinoma (HCC) [1]. Reliable and adequate preclinical models are therefore crucial to identify new therapeutic targets and develop novel, urgently needed drugs. Although several molecular mechanisms underlying the pathogenesis of NAFLD are well characterized and multiple potential drug targets have been identified over the last years, no specific pharmacotherapy has been approved yet. Recent findings have identified hepatic inflammation as well as toxic accumulation of lipids as the central drivers for the progression of the disease. As a consequence, most treatment interactions focus on the improvement of metabolic pathways to reduce fat accumulation (e.g. PPAR agonists), modulations of immune related mechanisms (e.g. inhibition of chemokine receptors CCR2/5) or the development of fibrosis (e.g. inhibition of Galectin-3). Tremendous interest in drug development has led to a number of different compounds with various mechanisms of action that are currently evaluated in ongoing clinical trials, with promising trials involving, for example, obeticholic acid (OCA), resmetirom, aramchol and cenicriviroc (CVC). However, meeting clinical end points has been challenging, and findings from animal models could not fully be reproduced in patient trials, which is likely due to the very complex, multifactorial character of the disease that is difficult to mimic in preclinical models. Further, metabolomic and gene expression studies from patients with NAFLD suggest different distinct disease subtypes that might even hint toward different pathological pathways [2].

Ideal preclinical models should show a high resemblance to human NAFLD to efficiently evaluate drug responses. An effective NAFLD animal model should preferably match the following criteria (1) development of obesity, (2) insulin resistance, (3) dyslipidemia and (4) hepatic inflammation. From a drug efficacy testing point of view, development of hepatic fibrosis would also be desirable. However, at the moment no single animal model can sufficiently mimic all pathological aspects of human NAFLD.

Preclinical NAFLD studies are mostly based on rodents, mainly mouse models, as those are easy to handle, relatively low in costs, easy to manipulate and have a short lifespan. Various different approaches, including dietary, genetic and chemically- or mechanically induced animal models that mirror different aspects of the disease have been proposed in the last years (Table 1). For drug development it is crucial to select a suitable model that best reflects the pathological features that is targeted by a new drug [3]. Commonly used preclinical NAFLD models have been reviewed in greater detail elsewhere [4].

Table 1. Characteristics of commonly used mouse models of NAFLD used in preclinical studies

			NAFLD/NASH phenotype
Intervention	Model	Obesity	Insulin resistance	Hypertriglyceridemia	Hepatic steatosis	Hepatic inflammation	Hepatocyte ballooning	Fibrosis	HCC	Preclinical studies with compounds in clinical trials
dietary	WD	+	+	+	+	+	±	(+)	-	Liraglutide, Lanifibranor
	HFD	+	+	+	+	+	+	-	-	Emricasan, Selonsertib, Liraglutide
	CDA-HFD	(+)	-	-	+	+	+	+	+	CVC, Lanifibranor
	MCD	-	-	-	+	+	+	+	-	CVC
genetic	ob/ob	+	+	+	+	-	-	-	-	OCA
	db/db	+	+	+	+	-	-	-	-	Elafibranor
	foz/foz	+	+	+	+	+	-	-	-	Elafibranor
chemical	STZ	-	-	-	-	+	-	-	-	CVC
	CCl4	-	-	-	-	+	+	+	+	CVC, Elafibranor, Simtuzumab
	TAA	-	-	-	-	+	+	+	+	CVC
mechanical	BDL	-	-	-	-	+	-	+	-	OCA, CVC
combined	ALIOS/AMNL	+	+	+	+	+	+	±	-	Tropifexor
	STAM	+	-	+	+	+	-	+	+	CVC, Tropifexor

ALIOS, American Lifestyle-Induced Obesity Syndrome; AMLN, Amylin liver NASH model; BDL, Bile duct ligation; CCl₄, Carbontetrachloride; CDA-HFD, choline-deficient, L-amino acid-defined, high-fat diet; CVC, Cenicriviroc; HCC, hepatocellular carcinoma; HFD, High fat diet; MCD, methionine deficient diet; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis; OCA, Obeticholic acid; STAM, Stelic model NASH (STZ + HFD); STZ, Streptozotocin; TAA, Thioacetamide; WD, Western diet

2. Reflecting NAFLD in animal models

As NASH naturally arises from overnutrition, most in vivo models are based on diet-induced obesity that include feeding mice with high amounts of fat, cholesterol and/or fructose. Common dietary interventions include the Western diet (WD) model or the high fat diet model (HFD). As those dietary models usually show low levels of fibrogenesis, nutrient deficient models, lacking methionine (methionine choline deficient diet, MCD) or choline (choline deficient L-amino-acid defined diet, CDAA) have been used alone or in combination to overcome this problem. Recently, the reduction of steatosis, inflammation and fibrosis by the pan-PPAR agonist lanifibranor (currently under investigation in phase 2b) has been proven using the CDA-HFD and WD model [5]. While selonsertib, an ASK1 inhibitor, was shown to improve hepatic steatosis as well as fibrosis in MCD and CDA-HFD fed mice [6], no anti-fibrotic effect was observed in patients [7].

Further common models comprise genetically modified mouse models as the ob/ob, db/db or foz/foz mice, resulting in hepatic lipid accumulation. Additional stimuli are necessary in those models to sufficiently induce features of NAFLD. For instance, beneficial effects of combination therapy with OCA and elafibranor were reported in a trans-fat containing amylin liver NASH (AMLN) diet-induced ob/ob mouse model [8]. Additionally, commonly used monogenic knock-out mouse models from atherosclerosis research (e.g. LDL, ApoE, SREBP) can provide useful information on mechanistic pathways.

Chemical models as the carbon tetrachloride (CCl₄) mouse model can be useful to investigate mechanisms of liver fibrosis up to HCC, but the initiating mechanisms of NAFLD development (i.e. steatosis and lipotoxicity) are lacking in this model. Further, the STAM model combines a low-dose injection of streptozotocin to promote type 1 diabetes with a high fat diet to induce NAFLD. The anti-inflammatory and anti-fibrotic effect of CVC could, among others, be observed in this model [9].

While diet-induced and genetic models mostly induce a metabolic phenotype, chemical and mechanical (e.g. bile duct ligation, BDL) models exhibit more fibrotic features. To reflect the whole spectrum of NAFLD it is therefore beneficial to use combined approaches. An overview of commonly used NAFLD mouse models and related preclinical studies is shown in Table 1.

3. Human in vitro and ex vivo models for NAFLD

To circumvent translational problems of animal models related to species differences, human in vitro and ex vivo models have been evolved over the last years that aim to better address pathological processes and drug responses in men.

As an ex vivo tool, human precision cut liver slices (hPCLS) have for example been shown to respond to OCA treatment by upregulation of FXR target genes and thereby demonstrated their suitability for drug testing [10]. One major advantage of hPCLS lies in keeping the architecture of the liver, but interventions are limited to a short time frame due to stability of the slices in culture. Moreover, steatohepatitis could be induced by fatty acid treatment of human liver organoids created from pluripotent stem cells [11].

Emerging organotypic liver systems try to recreate the complex three-dimensional (3D) architecture of the liver including various cell types under dynamic conditions. Newer studies attempt to use these to model influences of NAFLD in vitro for example by adding stimuli as free fatty acids to the media and thereby pave the way for in vitro drug screening and testing for drug safety [12]. Gori et al. developed a microfluidic perfused liver model of NAFLD with HepG2 cells that were stimulated with free fatty acids [13]. Another attempt by Kostrewski et al. could prove the anti-steatotic effect of pioglitazone in a microfluidic NAFLD model using primary human hepatocytes [14].

However, important pathological aspects of the disease as the recruitment of circulating immune cells or the impact of extrahepatic signals e.g. from the gut are difficult to reflect in those models, and drug effects on other organs (e.g. adipose tissue) cannot be evaluated.

4. Translating preclinical findings to human NAFLD

Some recently published clinical trials failed to transfer the promising research findings from rodents to human disease, which is likely due to principle differences in disease initiating mechanisms between species and also a high heterogeneity in human NAFLD patients [2]. A comparison of gene expression profiles revealed large transcriptomic differences between mice and men during NAFLD progression [15] that might cause differences is drug metabolism. Multiple other factors such as age, sex, hormonal status or microbiota have also been shown to drive disease heterogeneity in humans and thereby cause lack of drug efficacy, indicating that effective treatment may require more individual strategies. It has been suggested to rename NAFLD to metabolic (dysfunction) associated fatty liver disease (‘MAFLD’) in order to better capture this heterogeneity [16]. Moreover, it was shown that NASH development in mice significantly depends on the strain, suggesting similar divergent mechanisms in humans [17]. It is further important to note that preclinical experiments are also often designed to minimize external influences as gender, genetic background, diet or age that might limit translatability. Additionally, only a few models include comorbidities as atherosclerosis, insulin resistance or obesity that further shape the complex etiology of human NAFLD. Further, insufficient evidence in preclinical data before moving on to clinical trials may be the cause of decreased drug efficiency in patients.

5. Conclusions

Treatment efficiency heavily depends on the choice of appropriate preclinical models that help to mimic certain aspects of the disease and thereby evaluate drug efficacy and safety. Although translating findings obtained from animal models to humans is challenging, they still represent the best possibility to investigate liver disease mechanisms and the development of novel drugs in an in vivo setting. Compared to currently used ex and in vitro systems they are able to reflect the complex interactions between multiple organs which is especially relevant regarding crosstalk between liver and gut or adipose tissue. Additionally, newly emerging 3D human in vitro cell culture systems allow to specifically study hepatic cell functions and create new in vitro NAFLD models that can be used for drug testing. Although, further progress in this area is necessary they might provide helpful information and complement insights gained from animal models.

6. Expert opinion

Due to the complex and multifactorial processes underlying NAFLD progression it will be challenging to find one preclinical model reflecting the entire pathology of the disease. A lot of progress has been made in the development of preclinical models over the last years that helped to gain valuable information on NAFLD pathology. Animal models are crucial to understand disease mechanisms and drug metabolism regarding organ crosstalk and other systemic effects. However, their efficiency in drug development is sometimes limited by translational challenges, which motivated an evolution of novel human in vitro and ex vivo liver models. A lot more work has to be invested in the development of those systems to have reliable models that can be used for drug discovery.

To increase the chances for efficacy of newly identified drugs in patients, it is beneficial to use least two individual NAFLD models to evaluate drug efficacy or a combination of different approaches targeting metabolic, inflammatory as well as fibrotic pathways. Moreover, the power of omics technologies can provide a broad unbiased overview about disease mechanisms and can help to dissect drug related effects in depth and link them to human disease.

As there is no clear evidence that drug efficacy can be completely translated from preclinical models toward human disease it is important to be aware of the above-mentioned limitations while interpreting research findings.

Abbreviations

3D    Three-dimensional

AMLN  Amylin liver NASH

BDL   Bile duct ligation

CCl₄   Carbon tetrachloride

CDAA   Choline deficient L-amino-acid defined diet

CVC   Cenicriviroc

HCC   Hepatocellular carcinoma

HFD   High fat diet

hPCLS   Human precision cut liver slices

MAFLD  metabolic (dysfunction) associated fatty liver disease

MCD   Methionine choline deficient diet

NAFLD  Non-alcoholic fatty liver disease

NASH   Non-alcoholic steatohepatitis

OCA    Obeticholic acid

PPAR   Peroxisome proliferator-activated receptors

STAM   Stelic model NASH (STZ + HFD)

WD    Western diet

Declaration of interest

F Tacke has received research funding from Allergan, Bristol-Myers Squibb, Galapagos and Iventiva. F Tacke is also supported by the German Research Foundation (DFG; Ta434/5-1, CRC1382 and SFB/TRR57). The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This work is supported by the German Research Foundation (DFG; Ta434/5-1, CRC1382 and SFB/TRR57).