We thank Dr. Tahan and colleagues for their interest in our article describing the effects of rosiglitazone (RGZ) on ob/ob mice with nonalcoholic steatohepatitis (NASH).1 In these animals, we found that RGZ increases liver triglyceride concentration, serum levels of aminotransferases, liver steatosis, particularly microvesicular steatosis, oxidative stress, and impairs the activity of complex I of the mitochondrial respiratory chain (MRC). In contrast to these effects, Tahan et al. found in a rat model of NASH that RGZ attenuates liver inflammation and decreases proinflammatory cytokines, but increases markers of oxidative stress and does not change liver steatosis.2, 3 We do not have a definitive explanation for these partially divergent responses to RGZ treatment in both studies. However, there are a number of differences in the methodology that might justify, at least in part, these conflicting results. Hence, animals (rats versus mice), NASH model (diet versus leptin deficiency), and doses (4 mg/kg body weight versus 1 mg/kg body weight) were different in both studies. Difference in the animal species used in these experiments is of particular interest because variation in drug metabolism among species and even individuals may account for dissimilar response to this drug. In this respect, as Drs. Caldwell and Argo point out in their editorial,4 polymorphisms or interspecies and intraspecies variations in the expression and tissue distribution of peroxisome proliferator-activated receptor γ might offer an explanation for the partially divergent hepatic response to RGZ in humans, rats, and mice.
The major divergences between our results and those found by Tahan et al. concern the anti-inflammatory effect of RGZ they described in their abstracts. Although in our article, we have not included the effect of RGZ on the liver concentration of proinflammatory cytokines, we found that this drug decreased the levels of tumor necrosis factor-α in liver tissue from 2.13 ± 0.19 pg/μg protein in ob/ob mice to 0.38 ± 0.02 pg/μg protein in RGZ-treated ob/ob mice (lean mice, 0.12 ± 0.004 pg/μg protein). Hence, in our animal model of NASH, RGZ treatment also seems to have some anti-inflammatory effects. Nevertheless, serum aminotransferase levels increased during treatment, and electron microscopy demonstrated unviable degenerated hepatocytes. Thus, these apparent hepatotoxic effects of RGZ on ob/ob mice might be attributed to its worsening effect on the oxidative phosphorylation, in particular, to its inhibitory effect on the complex I of the MRC. This effect has also been described by others5–7 in cell cultures and liver homogenates. Furthermore, using proteomic technology, including in-gel activity assay, we confirmed that RGZ treatment markedly decreases the activity of complex I in normal mice and that the 2-dimensional SDS-PAGE system demonstrated that this complex was incompletely assembled (unpublished observations). Failure of this complex to transport electrons through the MRC results not only in a decreased oxidative phosphorylation, but also in the generation of excessive amount of superoxide anion, and consequently in oxidative stress.8 In both rat and mouse models, there is evidence showing that RGZ increases oxidative stress (increased lipid peroxidation, tyrosine nitrated proteins, decreased mitochondrial glutathione concentrations). Therefore, this stress might contribute to hepatocyte degeneration and apoptotic cell death. The apoptotic effect of glitazones has been demonstrated by others in a variety of cells, and it seems to be mediated by the generation of reactive oxygen species.9–11 Obviously, these cytotoxic effects do not depend on the inflammatory activity of the liver lesion.