Animal Models of Alcoholic Liver Disease

Abstract

The risk of alcohol-induced liver disease (ALD) increases dose- and time-dependently with consumption of alcohol. The progression of the disease is well characterized; however, although the progression of alcohol-induced liver injury is well characterized, there is no universally-accepted therapy available to halt or reverse this process in humans. With better understanding of the mechanism(s) and risk factors that mediate the initiation and progression of this disease, rational targeted therapy can be developed to treat or prevent it in the clinics. Several models for experimental ALD exist, including non-human primates, micropigs and rodents. However, most researchers employ rodent models of ALD. Furthermore, the advent of genetically modified strains of rodents (e.g. ‘knockout’ mice) has increased the specificity of the hypotheses that can be directly tested. Based on these models systems, several plausible hypotheses to explain the mechanism(s) by which alcohol leads to liver damage have been proposed, including consequences of alcohol metabolism, oxidative/nitrosative stress, altered inflammatory responses, and increased sensitivity to cytotoxic stimuli. These studies have also identified candidate genes for polymorphism studies to explain potential increased genetic risk in some individuals. However, despite significant advances in our understanding of the mechanisms by which ALD develops based on studies with these models, this work has yet to translate to a viable therapy for ALD in the clinics. This talk will also discuss potential reasons for these limitations to date and suggest future prospects to improve the translational utility of modeling ALD.

Introduction

It is well known that alcoholic liver disease (ALD) results from the dose- and time-dependent consumption of alcohol [1]. Progression of ALD is well characterized and is actually a spectrum of liver diseases which ranges initially from simple steatosis, to inflammation and necrosis (steatohepatitis), to fibrosis and cirrhosis. Although the progression of ALD is well characterized, there is no universally accepted therapy to prevent or treat this disease in humans. Instead, clinical treatment focuses predominantly on the decompensation caused by the disease and/or transplantation [2]. Gaps in our knowledge of how the disease progresses (and regresses) must be filled before rational targeted therapy can be developed. These gaps are due in part to the lack of a standardized animal model of experimental ALD that develops liver pathology that is completely analogous to the clinical situation. A major goal of current experimental research is indeed to develop effective and accurate animal models of ALD.

Animal Models of Ethanol Delivery

Numerous species have been used to study liver pathology caused by alcohol, including rodents, minipigs, and non-human primates [3,4,5]. Of these species, the pathology in non-human primates (e.g. baboons) is arguably the most similar to that in humans [6]. For example, ad libitum alcohol exposure in baboons leads to the progression of all stages of liver damage associated with ALD in humans (see above). However, the relevance of non-human primate models is outweighed by the prodigious costs of maintaining them, which limits their utility to the field as a whole. Cost issues are a similar concern with the micropig models of ALD. It is then not surprising that the majority of alcohol research performed in animal models involves rodents (i.e. rats and mice). Further, as described below, the advent of genetically-altered rodent strains (e.g. ‘knockout’ mice) has advanced the field rapidly in recent years. Research using these rodent models will therefore be the major focus of this review.

Animal models of alcohol exposure include many routes of delivery, such as inhalation, intravenous and intraperitoneal [7,8,9]. Although these routes can be useful to investigate other pathologies associated with alcoholism (e.g. addiction), they have little use in alcohol-induced liver damage. The major route of ethanol delivery in humans is clearly oral. Although intraperitoneal administration is ‘closer’ to oral administration, there are significant concerns about the effect of injecting large volumes of concentrated ethanol into the peritoneal cavity. The best route of ethanol administration is therefore oral and the remainder of this chapter will focus on acute and chronic models of oral/enteral exposure.

Acute Animal Models of Alcohol Exposure

Acute administration over the course of hours or a few days is the simplest model of alcohol exposure. Acute models of alcohol exposure can clearly be used to model damage due to an acute binge drinking episode. However, it also appears that mechanisms of acute and chronic alcohol-induced liver injury overlap to some extent [10,11,12]. For example, Enomoto et al. [13] showed that compounds previously shown to protect against chronic alcohol-induced liver damage [14,15,16] also protected rat liver against acute ethanol-induced steatosis (table 1). Therefore, acute exposure can in principle be used to mimic very early changes caused by ethanol [17,18]. However, the major effects of ethanol on liver are generally biochemical (e.g. alterations in lipid/glucose metabolism) in acute models [19]. Pathologic changes caused by acute ethanol are generally minor under these conditions (e.g. lipid accumulation) and do not model more severe changes caused by alcohol.

Table 1.

Mechanistic overlap between acute and chronic liver damage caused by alcohol

	Protected against hepatic changes caused by
	acute ethanol	chronic ethanol
GdCl3	yes [13]	yes [14]
Antibiotics	yes [13]	yes [15]
Nimodipine	yes [13]	yes [16]
TNFR1–/– mice	yes [61]	yes [29]
PAI-1–/– mice	yes [61]	yes [61]

Summary of some studies in which pharmacologic or genetic (i.e. knockout mice) interventions protected against both acute (steatosis) and chronic (steatohepatitis) experimental alcohol-induced liver injury in rats or mice.

Although acute ethanol causes limited pathological changes in the liver, it is useful for studying the altered responses of liver to other xenobiotics. Specifically, acute ethanol at doses that are relatively non-toxic enhance the toxicity of other compounds. For example, acute ethanol exposure is well known to enhance liver pathology induced by xenobiotics, such as bacterial cell wall products (e.g. lipopolysaccharide (LPS)). One such model employs acute ethanol (4–6 g/kg) as the first ‘hit’, followed by a subsequent exposure to LPS ∼24 h later [20]. Under these conditions, liver damage caused by endotoxin is greatly exacerbated by ethanol administration. Interestingly, such a priming effect of ethanol to LPS-induced liver damage is not observed if the LPS is injected sooner after alcohol exposure (e.g. 2 h), indicating that the effect of ethanol is not likely ethanol (or acetaldehyde) per se, but indirect effects of the drug on the response, most likely mediated by altered gene expression [20]. In addition to liver pathology, acute ethanol exposure has been utilized to study the effects of in vivo ethanol exposure on subsequent responses to stimuli in culture. The apparent ‘priming’ effect of ethanol on inflammatory cells observed in vivo can also be mimicked in cell culture. For example, Kupffer cells exposed to acute ethanol in vivo are more easily stimulated and produce more inflammatory cytokines than naive cells [20]. Additionally, hepatocytes are sensitized to cytotoxic cell killing by ethanol administration [21].

Chronic Models of Alcohol Exposure: Ad libitum Feeding

A major advantage of ad libitum ethanol feeding of ethanol is that it is not technically demanding, generally requiring only daily changes of the ethanol-containing liquid. As mentioned earlier, oral alcohol administration by drinking is clearly the most relevant to clinical ALD. Therefore, it is not surprising that initial animal models of ethanol exposure attempted to mimic the human condition and utilized ad libitum access to ethanol-containing water [e.g. [22]]. While this model was quite simple, there are major concerns with such an administration protocol, especially in rodents. Specifically, rodents have a strong aversion to the taste or smell of ethanol and therefore do not consume large enough quantities of ethanol to produce significant liver pathology. Even ‘alcohol-preferring’ rodent strains consume relatively low amounts of ethanol [23,24]. Indeed, in the ethanol drinking water model, Best et al. [22] observed that there was ‘no more evidence of a specific toxic effect of pure ethyl alcohol upon liver cells than there is for one due to sugar’. Based on these observations, it was proposed at that time that ALD was due to accompanying nutritional deficiencies and not alcohol per se [25].

In an attempt to overcome the above-mentioned aversion of ethanol, alternative techniques of ad libitum ethanol delivery were developed, such as ‘forced choice’ models of alcohol exposure, in which ethanol is incorporated into a liquid diet that serves as the only source of calories. Of these liquid diet models, the most studied is that first described by Lieber et al. [26]. In that model, rats are given a liquid diet formula containing ethanol that typically makes up 36% of the total caloric intake [25]. On this regimen, a rodent's natural aversion to alcohol is partially overcome thus producing blood alcohol levels higher than observed if ethanol is only included in the drinking water. Indeed, daily ethanol intake for rats on ad libitum diet is sufficient to produce liver injury indicative of early ALD in humans, especially fat accumulation (steatosis) [25,26]. It was use of this model that solidified the hypothesis that alcohol is indeed a direct hepatotoxin [25]. An interesting point therefore is that while the early assumption that ethanol alone is not toxic is incorrect, the assumption that nutrition plays a key role in ALD is still supported experimentally. Specifically, ethanol alone is not sufficient to cause significant liver injury even in this model. Instead, liver pathology is highly dependent upon dietary composition, especially the amount and type of fat in the diet [27].

While a distinct improvement over alcohol exposure in drinking water, the Lieber-DeCarli and related models of ad libitum feeding of ethanol-containing liquid diet still have limitations. For example, despite partially overcoming their dislike of ethanol-containing diet, rodents on ethanol diet still consume less than animals on carbohydrate control diet. To address this issue, animals that are fed control diet are restricted calorically to match the alcohol-consuming group [25]. Therefore, while this model employs more physiologically relevant alcohol exposure, there is an inherent concern in comparing ‘pair-fed’ rodents that typically consume their daily caloric allowance in a short time period (e.g. 3–6 h) with ethanol-fed rodents that consume the same amount of calories over the course of a day. Indeed weight gain in both liquid diet groups is suboptimal compared to chow-fed animals. However, the major disadvantage of this model with regard to experimental ALD is that the liver pathology obtained is limited predominantly to steatosis, with some necroinflammatory changes. More severe steatohepatitis and advanced liver damage observed in humans (fibrosis and cirrhosis) is generally not observed in rodents on ad libitum liquid diet protocols.

Chronic Models of Alcohol Exposure: Enteral Feeding

One mechanism of bypassing aversion to ethanol in experimental animals is by eliminating ad libitum control of ethanol exposure; specifically, ethanol can be enterally delivered directly to the stomach via a surgically-implanted intragastric tube. Almost 20 years ago, Tsukamoto et al. [3] developed an enteral ethanol-feeding model of ALD that was designed primarily to increase alcohol exposure beyond that of ad libitum liquid diets. Similar to the Lieber-DeCarli protocol, ethanol is incorporated into a liquid diet, but it is infused at a constant rate over the course of the day instead of given ad libitum. Animals fed an enteral ethanol diet generally have average blood/urine alcohol concentrations that are significantly higher than levels observed during ad libitum feeding [28]. Interestingly, despite the fact that ethanol is being infused at a constant rate, daily alcohol levels fluctuate between 0 and 500 mg/dl [29,30]. Although not completely understood, it is likely that this phenomenon is mediated by fluctuations in hormones from the hypothalamic-pituitary-thyroid axis [31].

The major advantage of the enteral feeding model is that the pathology obtained in this model, in as little as 4 weeks of enteral diet, mimics early ALD in humans, which includes micro- and macrovesicular steatosis, apoptosis, mixed inflammatory cell infiltrate, and focal necrosis [32]. Furthermore, fibrosis and cirrhosis can also be produced in rats using this model by including carbonyl iron in the diet or prolonging the time course of enteral feeding [33,34,35]. The enhanced pathology caused by higher ethanol dosing is not limited to only liver. For example, Kono et al. [36] demonstrated that enteral ethanol causes pancreatic damage in rats characteristic of early stages of chronic alcoholic pancreatitis within 1 month of enteral feeding. Fibrotic changes associated with later stages of alcohol-induced pancreatic damage also occurred as early as after 8 weeks enteral feeding in this model [36]. Such pathology is generally not observed in animals fed ad libitum alcohol diet. It is likely that higher amounts of alcohol are necessary to achieve pancreatic damage in the rat in the absence of coadministration of other inducers of pancreatitis (e.g. cholecystokinin [37]).

Another advantage of the enteral model is the high level of control by the researcher on diet delivery. This effect avoids caloric concerns associated with ad libitum feeding, as discussed above. Indeed, animals gain weight at or near the same rate as animals on isocaloric control diet and both groups gain weight at similar rates as chow-fed animals. Furthermore, orally available drugs can be delivered at constant, controlled rates to the test animals. Furthermore, the modification of this model to mice has allowed the use of genetically-altered (e.g. knockout) mice in the studies of experimental ALD.

Although the pathology caused by enteral alcohol feeding is more severe and therefore more relevant to the situation in humans than ad libitum feeding (table 2), there are limitations associated with use of this model. First, the enteral model is a technically demanding method of ethanol delivery. Although the surgery is not complex, it does require a dedicated rodent surgery facility to perform this model. Furthermore, due to the higher levels of alcohol achieved in this model, the relative amount of husbandry required during enteral feeding compared to ad libitum protocols is greater. These issues have limited the number of laboratories that can successfully perform this protocol.

Table 2.

Animal models of alcoholic liver injury

Model	Comments	Steatosis	Inflammation and necrosis	Fibrosis and cirrhosis
Ad libitum
Drinking water	Not relevant due to the lack of significant pathology	–/–	–/–	–/–
Liquid diet	Reliable method to model early fatty changes caused by alcohol exposure. Easy to employ and well characterized. Pathology does not progress much past steatosis.	++	+	–/+
Enteral feeding
Liquid diet	Causes enhanced liver pathology due to higher levels of ethanol. Relevance of high doses of ethanol to humans a concern.	+++	++	–/+
Alternative models
Choline deficiency	Dietary model that mimics all stages of human ALD, including hepatocellular carcinoma. Relevance of lipotrope deficiency to humans is questionable.	+++	+++	+++
Other hepatotoxins (e.g. CCl4, TAA, DMNS)	Can cause severe fibrosis or cirrhosis relatively rapidly. Severity of liver damage is usually higher than that of ALD.	–/+	+++	+++

General animal models utilized in the study of alcoholic liver disease.

There are also concerns about the level and pattern of alcohol observed in this model. Specifically, the relevance of the high concentrations of ethanol achieved under these conditions to human alcohol consumption has been challenged. It should be emphasized that the concentrations of ethanol achieved in rodents is the result of careful increases in alcohol delivery over time; if such levels were achieved in a naive rodent, lethality would likely occur. These data indicate that tolerance to higher ethanol levels is achieved in the rat during chronic ethanol exposure. A similar effect also occurs in humans [38,39]. A striking example of this point was demonstrated by Urso et al. [40] in which blood alcohol levels in emergency room patients was measured. In that study, blood alcohol concentrations (BAC) ranged from 0 to 540 mg/dl (mean 290 ± 9); importantly, all of these patients were determined ‘non-intoxicated’ by behavioral assessment [40]. Therefore, the concentrations of alcohol observed in the enteral feeding model may be similar to values achievable in humans that chronically abuse the drug.

Another concern has been raised about the relevance of the cyclic pattern of ethanol concentrations observed in this model (see above). It is unclear as to whether this phenomenon in humans has ever been directly tested. If cycling occurs in humans analogous to rodents, BACs would have to be observed over the course of a number of days during continued alcohol exposure. In one study, Goldman [41] monitored the BAC in 4 alcoholics during ad libitum alcohol exposure for a number of days. Despite free access to alcohol, the BACs in each individual varied from as low as ∼20 mg/dl to as high as 380 mg/dl during the course of the study [41]. Although not directly an example of ‘cycling’ in humans consuming alcohol per se, these data indicate that alcohol concentrations in an individual can fluctuate quite dramatically despite free access to alcohol. Although the above-mentioned studies do not necessarily invalidate the concerns raised about the enteral feeding model, they do serve as a possible counterpoint to the argument that the enteral feeding model is not relevant to human alcohol exposure.

Alternative Models of Experimental Alcoholic Liver Disease

Pathological changes in the liver of human abusers of alcohol range from mild (steatosis) to moderate (steatohepatitis and early fibrosis) to advanced (late fibrosis and cirrhosis) dependent upon both the daily dose and time of exposure [42]. For example, daily intake of alcohol for 10 years or more with doses in excess of 40–80 g/day for men and 20–40 g/day for women is generally sufficient to induce ALD [42]. While mild and moderate stages of ALD are in general reversible processes, advanced liver injury is thought to be less reversible, at least with current clinical intervention. Despite progressive improvements in animal models of ALD, achieving significant fibrosis and cirrhosis in rodents with alcohol feeding has been elusive with only few studies demonstrating such effects [33,34,35]. Since these pathologic changes are the major clinical concern in ALD in humans, many animal models have been developed to specifically induce fibrosis and cirrhosis in order to identify critical cellular pathways involved in such liver pathology.

Several hepatotoxins have been used in animal models of fibrosis and cirrhosis, such as carbon tetrachloride (CCl₄), dimethylnitrosamine (DMNS), and thioacetamide (TAA) (table 2) [43,44]. Carbon tetrachloride is given to rodents typically through either subcutaneous injection or gavage of 1–2 ml/kg of 50% CCl₄ twice weekly [45]. In this model, significant hepatocyte necrosis, inflammatory infiltration, proliferation of hepatic stellate cells, and connective tissue deposition are prominent features observed in the liver of rats after 6–9 weeks of CCl₄ treatment [45]. DMNS and TAA are hepatic carcinogens that effectively induce severe liver fibrosis and cirrhosis in rodents in as little as 1–2 months of treatment [46,47]. In rats, DMNS is injected intraperitoneally three times a week, whereas TAA is placed in drinking water at a dose of 300 mg/l [43]. DMNS typically causes central-portal bridging fibrosis and TAA causes both portal-portal and portal-central fibrotic lesions. Although these hepatotoxins have proven useful for studying the mechanisms behind fibrotic lesion formation in the liver (e.g. stellate cell activation and transformation into myofibroblast as well as transforming growth factor-β₁ activation) [48,49,50,51,52], the relevance of such models to fibrosis and cirrhosis in alcoholics is questionable. Specifically, none of these experimental models reproduce exactly human liver fibrosis by etiology [44].

Another reliable approach for induction of fibrosis and cirrhosis in rats is the feeding of a diet deficient in choline [53,54]. Choline deficiency causes significant steatosis presumably via decreases in lipotropes as early as 1–2 weeks following initiation of diet (table 2). Long-term feeding (≥3 months) of choline-deficient diet causes hepatitis, fibrosis and cirrhosis and after prolonged feeding, hepatocellular carcinoma [55]. The advantages of this model are technical ease of feeding diet and induction of advanced liver pathology associated with chronic ALD. While the relevance to such dietary manipulations to the human population is low, the similarity of the course of pathology in this model to human ALD is compelling. The finding that alcoholics may have a functional impairment of the transsulfuration pathway via inhibition of methionine adenosyltransferase, which leads to a similar lipotrope deficiency as the choline-deficient diet [56], has renewed interest in this model. Furthermore, such models may also serve for the study of chronic non-alcoholic steatohepatitis (i.e. NASH), which shares an almost identical pathologic spectrum to ALD [57].

Summary and Conclusions

Major advances have been made in our understanding of the mechanisms for the development and progression of ALD. In experimental animals, one can block fatty liver and initial liver injury by blocking TNF/proinflammatory cytokine production, or inhibiting oxidative stress. Moreover, many of these pathways are highly interactive. For example, there are major interactions between oxidative stress and proinflammatory cytokine production [58]. Unfortunately, these seemingly straightforward in vitro and animal observations related to mechanisms for ALD have not translated into effective therapy for human ALD.

There are likely multiple reasons for this apparent lack of translation of therapeutic efficacy into humans. In most experimental models (both in vitro and in vivo), the experimental design focuses on preventing the development of liver disease rather than treating already developed liver injury. In the clinical situation, patients present with alcoholic hepatitis or cirrhosis, and treatment instead of prevention is required. Few studies to date [26,59] have focused on the recovery/regression from established experimental ALD.

Next, clinical trials have shown that small biologic molecules, such as anti-TNF, appear not to be effective in acute alcoholic hepatitis. In animal models, such therapy is highly effective for prevention [29], but in humans it does not appear to be effective for the treatment of established disease. A ‘baseline’ low concentration of TNF appears to be required for liver regeneration [60], while excess TNF can be hepatotoxic. Thus, inhibiting all TNF activity with biologic therapy may not be an appropriate strategy for treatment of alcoholic hepatitis. This concept may be true for several acute phase proteins that may promote regeneration/restitution, as well as mediate liver damage. Rather, drugs such as pentoxifylline that downregulate TNF production may be a more effective therapeutic approach.

In summary, we have markedly expanded our understanding of all phases of ALD. This knowledge now needs to be translated into the development of effective therapy, and this will require close interactions between basic scientists, clinicians, and industry.

Disclosure Statement

The author has no conflicts of interest to disclose.