It has been 10 years since we initiated a pilot study of troglitazone in human nonalcoholic steatohepatitis (NASH).1 Troglitazone was the first clinically available member of the thiazolidinedione (TZD) class of peroxisome proliferator–activated receptor-γ (PPAR-γ) ligands which now includes pioglitazone and rosiglitazone. Our rationale to study troglitazone was based on its beneficial effects in type 2 diabetes and on the potential effects of a TZD on mitochondria, which appeared to be involved in NASH pathophysiology. As a harbinger of the greater degree of complexity to come, our study was cut short by reports of troglitazone-induced fulminant liver failure (leading to the drug's withdrawal) although we observed histological improvement in follow-up biopsies albeit with puzzling changes in mitochondrial morphology. Since then, a substantial body of literature has emerged on the related subjects of hepatic steatosis, insulin resistance, energy homeostasis, and mediators of cellular injury, all of which are perhaps best united under the concepts of the insulin resistant (thrifty) genome and systemic lipotoxicity.2, 3
In the present issue of HEPATOLOGY, two contrasting articles raise new questions regarding the effects of TZDs in fatty liver disease. In the study by Garcia-Ruiz et al., rosiglitazone increased hepatic triglyceride levels 10-fold in ob/ob mice and impaired the mitochondrial respiratory chain function with increased markers of oxidative stress.4 Hepatic steatosis was associated with increased body weight but it was not described in lean controls. In contrast to the ob/ob mouse, prior studies of TZDs in human NASH have usually demonstrated decreased hepatic fat and improved markers of cell injury on follow-up biopsies. In the current issue of HEPATOLOGY, Lutchman et al. have extended this work by showing that the initial improvement in human NASH at the end of defined treatment periods is not sustained after drug discontinuation.5 What should we make of the apparent divergent response in mice compared to humans and the limited positive response evident in the latter?
TZDs are synthetic ligands for the nuclear receptor superfamily member PPAR-γ which participates extensively in regulation of fat metabolism and energy homeostasis. After ligand binding, the PPAR-ligand complex forms a heterodimer with the retinoid X receptor, which binds to the response element of specific target genes (transactivation pathway) to stimulate transcription. The tissue distribution of the receptor in large part governs the overall systemic receptor-mediated response to the agent. In humans, PPAR-γ and particularly the PPAR-γ2 isoform, is most abundant in adipose tissue where TZDs stimulate adipocyte differentiation. To a lesser extent, the receptor is present in the colon, kidney, liver, and small intestine. Importantly, it is not well known to what extent this distribution might change in disease states such as the metabolic syndrome. However, it is known that increased adipose expression of PPAR-γ2 is evident in obese versus lean subjects and that there is a correlation between the ratio of PPAR-γ isoforms and body mass index.6 This suggests that the systemic PPAR-γ–mediated response to TZDs may vary based on acquired host factors and resulting changes in receptor distribution. As confirmed by Garcia-Ruiz et al., hepatic PPAR-γ receptor expression is 7-fold to 9-fold greater in ob/ob mice compared to wild-type animals, whereas it is relatively sparse in normal human liver. Thus, variation in hepatic PPAR-γ expression offers a probable explanation for the contrasting hepatic response to TZDs in humans versus the ob/ob mouse. It is less likely to be related to the leptin-deficient state of ob/ob mice because other rodent models without leptin deficiency also experience an increase in hepatic steatosis with TZD treatment.7 Furthermore, TZDs decrease hepatic steatosis in human lipodystrophy in which there is also leptin deficiency.
The interspecies variation in TZD response leads to a question touched on in the article by Garcia-Ruiz et al.: should we anticipate variable hepatic expression of PPAR-γ receptor in some human populations and might this underlie some forms of TZD toxicity? Polymorphisms of PPAR-γ2 have been reported and influence plasma lipoprotein phenotypes.8 In addition, increased expression of adipocyte-like genes in experimental hepatic regeneration raises concern for acquired changes in receptor expression related to hepatic injury.9 Repopulation of a diseased liver with mesenchymal stem cells could theoretically result in different receptor expression compared to native hepatocytes, although at present, replacement of senescent hepatocytes in NASH appears to be mediated by endogenous oval cells.10 Thus, although it remains to be established, the consistent hepatic response to TZDs in human studies of NASH suggests that hepatic PPAR-γ expression does not vary widely in humans. On the other hand, there is variation in total body fat response to TZDs which could reflect variation in PPAR-γ expression peripherally (discussed below).
In general, TZDs cause a shift of fat from visceral to peripheral fat depots. This process has been observed in Sprague-Dawley rats fed a high-sucrose and high-fat diet, where it results from relatively greater TZD-stimulated energy expenditure in visceral fat in association with increased mitochondrial biogenesis.11 In humans, similar results have been shown both in vitro in studies of white adipocyte cell cultures and in vivo in studies of subcutaneous adipose tissue obtained after 12 weeks of treatment with pioglitazone.12 These studies have also demonstrated stimulation of adipocyte mitochondrial biogenesis, increased UCP (uncoupling protein) expression and increased brown fat–like morphology in adipocyte cell lines.13 The findings offer a plausible explanation for the effects of TZD therapy on hepatocyte mitochondria in human NASH, where crystalline inclusions are increased with TZD therapy (in parallel to improved light microscopic histologic analysis).1, 14 Similarly, Garcia-Ruiz et al. noted morphological changes in the hepatocyte mitochondria with TZD therapy, although crystalline inclusions, which appear to be unique to humans, were not observed.
PPAR-γ receptor binding is likely the main pathway by which the TZDs act, but other non–receptor mediated TZD effects have been observed experimentally, including direct effects on mitochondrial respiratory chain function in cell cultures and activation of adenosine monophosphate–activated protein kinase (AMPK).15 The latter pathway may be particularly relevant because TZD therapy increases adiponectin, which activates AMPK in humans. It is also especially interesting that TZD-induced steatosis in the ob/ob mouse reported by Garcia-Ruiz et al. was predominantly microsteatotic, similar to that reported with other diabetic mouse strains. Small-droplet fat accumulation is increasingly recognized in patients with NASH although its significance is controversial regarding its relative stability versus its possible role in hepatocyte ballooning.16 In this regard, the term “adipogenic transformation” has been applied to some forms of experimentally induced fatty liver, but the validity of this term in the TZD-treated mouse model has been questioned because of the lack of induction of perilipin—an adipocyte fat droplet–associated protein—with PPAR-γ2 ligands.17 However, there again may be divergence between human fatty liver disease and rodent models of fatty liver. In collaboration with Dr. Yoshihiro Ikura and colleagues at Osaka City University in Japan, we recently observed that perilipin and adipophylin (another fat droplet–associated lipase) are present in the rim of small fat droplets in human NASH, and the pattern of their expression appears to be related to cellular enlargement (ballooning) and the presence of oxidized phosphatidylcholine.18 Furthermore, because hepatic small fat droplets are intimately related to lipidation of the endoplasmic reticulum and secretion of very low density lipoprotein versus intracellular fat storage, disturbances in small droplet trafficking provides a link between simple fat storage (nonalcoholic fatty liver disease) and NASH with oxidative stress and endoplasmic reticulum injury. Thus, the accumulation of small droplet fat in the article by Garcia-Ruiz et al. may be especially relevant to mechanisms of cell injury in NASH and to TZD-induced changes in lipoproteins, both of which warrant further investigation.
Weight gain due to peripheral fat accumulation is a common but inconsistent problem in TZD therapy of human NASH. For example, weight gain was described in 67% of patients in the study by Neuschwander-Tetri et al. (rosiglitazone)19 and in 72% of subjects in the study by Promrat et al. (pioglitazone).20 However, weight gain was not observed in a Finnish report from Tiikkainen et al. (rosiglitazone).21 Could this represent genetic variation in peripheral PPAR-γ isoforms as has been described in the Finnish population?22 Or are weight changes during and after therapy equally dependent on diet and physical activity? The U.S.-based study by Lutchman et al. clearly documented the frequent lack of sustained benefit after stopping TZD therapy in human NASH. However, long-term follow-up of the subjects in our original study of troglitazone revealed that the only patients with durable histological responses more than 6 years after stopping treatment were those who had achieved and maintained weight loss through dietary changes and increased physical activity.23
Clearly, the presence of histologically determined NASH predicts both liver-related and cardiovascular deaths.24 However, the promise of TZD therapy in ameliorating steatohepatitis is tempered by its complexity. Adding to the uncertainty, a recent meta-analysis reported an increased probability of cardiac events in diabetic patients taking rosiglitazone.25 Although this study suffers from data quality issues, its results should provoke caution because response to these agents depends on a number of incompletely defined factors related to both lifestyle and genetic parameters. The former are intimately related to diet and exercise whereas the latter are likely the product of evolution and natural selection inherent to the survival benefit of the “thrifty genome” during times of resource deprivation. Both appear to influence the distribution of PPAR-γ receptors which are central to the maintenance of energy homeostasis. From this perspective, interspecies (and perhaps intraspecies) variation in response to PPAR-γ ligands should be expected.