Peroxisome proliferator-activated receptor-γ: too much of a good thing causes harm

Abstract

The nuclear receptor peroxisome proliferator-activated receptor-γ (PPARγ) helps to translate 'what you eat' into 'what you are' because it allows dietary fatty acids (PPARγ ligands) to modulate gene transcription. Treatments for diabetes include PPARγ activators, as they sensitize the body to insulin. Our understanding of PPARγ function has recently been enhanced by a flurry of human and mouse genetic studies, and the characterization of new PPARγ ligands. This insight has led us to propose that modulating PPARγ activity, rather than activating it, might be the most effective strategy for treating metabolic disorders, as this will improve glucose homeostasis while preventing adipogenesis.

Introduction

The nuclear receptor peroxisome proliferator-activated receptor-γ (PPARγ) is a transcription factor that is crucial for whole-body energy homeostasis and adipogenesis (reviewed in Auwerx, 1999; Rosen & Spiegelman, 2001). The actions of PPARγ are mediated by two protein isoforms: the widely expressed PPARγ1 and the adipose tissue-restricted PPARγ2. The activity of PPARγ is governed by the binding of small lipophilic ligands—mainly fatty acids—that are derived from nutrition or metabolism, and the activation of PPARγ leads to adipocyte differentiation and fatty-acid storage. The modern westernized lifestyle—characterized by a high caloric intake and a lack of physical exercise—exposes people to prolonged chronic levels of fatty-acid-like PPARγ ligands, which, through a feed-forward pathway, often results in obesity (Fig 1) (Burdge et al, 2003; Friedman, 2003; Sanders, 2003). Obesity is more prevalent in affluent societies, along with associated metabolic diseases such as hyperlipidaemia, insulin resistance, type 2 diabetes and cardiovascular diseases, which constitute a heavy social and economic burden. It therefore seems ironic that synthetic PPARγ ligands, such as thiazolidinediones (TZDs), are used to treat diabetes because they sensitize the body to insulin; their down side is that they also promote fat accretion, and the long-term consequences of this are unknown. Moreover, the mechanism by which TZDs act and the reasons why they are effective are still not understood. It is therefore essential that these pathways be clearly defined to pave the way for better treatments for these metabolic diseases. Now that recent data have shed more light on the roles of PPARγ, we propose that insulin sensitization without the accompanying increase in fat deposition might be possible through the controlled regulation of PPARγ activity.

Figure 1.

The balancing act of lipids and tissues in energy homeostasis. Interactions between the main metabolic tissues that store (adipose tissue) and oxidize (skeletal muscle and liver) free fatty acids (FFAs). These interactions are mediated by FFAs and adipokines. Adipokines are subdivided into insulin sensitizing (for example, adiponectin) and insulin resistance (tumour necrosis factor-α) adipokines. The thickness of the arrows reflects the strength of the effects. Three conditions are compared: normal, lipoatrophy and obesity. Both lipoatrophy and obesity are characterized by insulin resistance (IR), which is, however, mediated by different signalling factors.

PPARγ: a thrifty transcription factor

Four lines of evidence have established PPARγ as the centrepiece of a feed-forward pathway that favours energy storage by adipocytes and coordinates a 'thrifty response' that once had substantial evolutionary benefits.

The first line of evidence comprises molecular and cellular studies. Early observations showed that the expression of PPARγ in cells is sufficient to induce adipocyte differentiation (Tontonoz et al, 1994). Moreover, targeted mutagenesis of the PPARγ gene in embryonic stem (ES) cells and knocking down the endogenous PPARγ2 in cell lines confirmed the role of PPARγ in adipocyte differentiation (Rosen et al, 1999; Ren et al, 2002). Consistent with this, PPARγ has been shown to increase the expression of genes that promote fatty-acid storage, whereas it represses genes that induce lipolysis and the release of free fatty acids (FFAs) in adipocytes.

Second are human genetic studies. The role of PPARγ in adipogenesis has been underscored by studies that link the PPARG locus on chromosome 3p25–p24 with obesity in Pima Indians (Norman et al, 1997). The partial 'loss of function' of the PPARγ2specific Pro12Ala mutation is associated with a decrease in body mass index (BMI), and improvement in insulin sensitivity and lipid profile (Deeb et al, 1998; Altshuler et al, 2000) (Fig 2). This association between the hypomorphic Ala substitution and insulin sensitivity disappears when the data are corrected for BMI, which indicates a primary effect on body-fat mass accretion (Deeb et al, 1998). By contrast, the rare Pro115Gln substitution renders PPARγ constitutively active and carriers of this mutation are obese but remain insulin sensitive (Hu et al, 1996; Ristow et al, 1998). The dominant-negative/loss-of-function mutations Phe388Leu, Pro467Leu and Arg425Cys have been associated with partial lipodystrophy, resulting in a loss of fat from some areas of the body (Agarwal & Garg, 2002; Combs et al, 2002; Hegele et al, 2002; Savage et al, 2003), and with severe insulin resistance, diabetes and hypertension (Barroso et al, 1999). Overall, PPARγ activity in humans corresponds directly to adipose mass and not necessarily to insulin sensitivity (Fig 2), which indicates that only part of the effect of PPARγ on glucose homeostasis is dependent on white adipose tissue (WAT).

Figure 2.

Genetic and pharmacological evidence supporting the relationship of peroxisome-proliferator-activated receptor-γ (PPARγ) with adipogenesis and insulin sensitivity. Human mutations, mouse models and pharmacological studies show that the level of PPARγ activity directly corresponds to adipose mass (adipogenesis). By contrast, insulin sensitivity can be achieved by both the inhibition and activation of PPARγ, as illustrated by human genetic and pharmacological studies.

Third are mouse genetic studies. Pparγ^−/− mice that are generated by conventional gene targeting die in utero owing to placental and cardiac defects. Although one such Pparγ^−/− animal was rescued by tetraploid aggregation, it died within days as a result of severe lipodystrophy (Barak et al, 1999). This example, together with the characterization of mice that are chimeric for Pparγ^−/− ES cells (Rosen et al, 1999), show the importance of PPARγ in adipose-tissue development in vivo. Interestingly, heterozygous Pparγ^+/− mice are resistant to the obesity and insulin resistance that is induced by a high-fat diet (Kubota et al, 1999; Miles et al, 2000). To overcome embryonic lethality, mice with tissuespecific deletions of Pparγ have been generated to help determine its tissue-specific activities (summarized in Table 1). In contrast to adipose tissue, muscle and liver (discussed below), the deletion of Pparγ in the pancreas did not result in a metabolic phenotype, but it underscored the antiproliferative role of Pparγ (Rosen et al, 2003).

Table 1.

Comparison of PPARγ tissue-specific knockout mouse models

Adipose		Muscle	Liver	Pancreas
Adiposity	Congenital lipodystrophy1 Progressive lipodystrophy2	Normal3 ↑adipose mass4	Progressive obesity5 Normal6	Normal
Plasma profile	↑glucose, ↑insulin, ↑TG, ↑FFA1 Normal glucose, insulin, ↑TG, ↑FFA2	Normal glucose, insulin3 Normal glucose, ↑ insulin, ↑TG, ↑FFA4	↑glucose, ↑TG, ↑insulin5 Normal glucose6	Normal
Adipokine production	↓leptin, ↓adiponectin1,2	Normal leptin3 ↑leptin, ↓adiponectin4	↑leptin, ↓adiponectin5	N/D
Liver steatosis	Normal liver TG1 Steatosis2	Normal liver TG3 Steatosis4	Normal liver TG5,6	Normal
Insulin resistance	IR1,2	IR3,4	IR5 No IR6	Normal
Background (age in months)	Mixed, SV129 & C57BL/6J(5)1 or C57BL/6J (14)2	Mixed (12)3 or C57BL/6J (14)4	Mixed (40)5 or mixed (10)6	Mixed (13)
References	1Koutnikova et al, 2003 2He et al, 2003	3Norris et al, 2003 4Hevener et al, 2003	5Gavrilova et al, 2003 6Matsusue et al, 2003	Rosen et al, 2003

The different mouse models have various age- and diet-dependent responses; therefore, for simplicity, the descriptions are for the adult phenotype in the postprandial state on a chow diet. FFA, free fatty acid; IR, insulin resistance; N/D, not determined; TG, triglycerides.

The specific reduction of PPARγ in the adipose tissue showed the essential role of PPARγ in adipogenesis and also highlighted its role in maintaining the integrity and function of the mature adipocyte (He et al, 2003; Koutnikova et al, 2003). Moreover, the essential role of adipose tissue in whole-body metabolism was exemplified by the significant mortality rate (>40%) of the WATspecific hypomorphic Pparγ knockdown mice (PPARγ^hyp/hyp), which were severely lipo-dystrophic (Koutnikova et al, 2003). When an aP2-driven Cre recombinase transgene was used to excise PPARγ from the mature adipocytes, a more moderate reduction of adipose mass was observed, which was accompanied by hyperlipidaemia and liver steatosis (He et al, 2003). This was in contrast to the surviving adult PPARγ^hyp/hyp mice, which did not have liver steatosis or dyslipidaemia (Koutnikova et al, 2003). Liver steatosis and dyslipidaemia were predominantly prevented in the PPARγ^hyp/hyp mice by the efficient oxidation of excess lipids in the muscle by Pparα and Pparβ/δ. Intriguingly, both adipose Pparγ-deficient models had relatively normal glucose tolerance (He et al, 2003; Koutnikova et al, 2003).

An abnormal lipid balance was a constant feature of all tissue-specific knockout models. As expected, when Pparγ was absent in the adipose tissue, FFAs in the postprandial state were elevated (He et al, 2003; Koutnikova et al, 2003). Lipid balance was also significantly altered when either muscle or liver Pparγ was obliterated. Deletion of Pparγ in the livers of two mouse models with significant steatosis (leptin-deficient ob/ob or lipodystrophic A-ZIP/F-1 mice) reduced the triglyceride content of the liver, although it elevated serum FFAs and lipoproteins and induced insulin resistance, which illustrates the role of Pparγ in liver lipogenesis (Gavrilova et al, 2003; Matsusue et al, 2003). Postprandial FFAs and serum lipids were also increased in mice lacking Pparγ specifically in the muscle. These mice had increased adiposity and developed insulin resistance (Hevener et al, 2003; Norris et al, 2003). It therefore seems that muscle Pparγ directly or indirectly induces FFA clearance by adipose tissue and FFA combustion in muscle. Pparγ expression varies significantly between these three metabolic tissues, with Pparγ expressed predominantly in the adipose tissue, moderately in the liver and only at low levels in the muscle. It is therefore fascinating that, irrespective of the level of Pparγ expression in each of these tissues, its absence in any of them has an effect on whole-body lipid homeostasis and insulin sensitivity. Each tissuespecific Pparγ mouse model shows the in vivo repartitioning of lipids between metabolic tissues, which illustrates the cross-talk between the liver, adipose and muscle to maintain energy homeostasis. This is, in part, owing to the adaptation of Pparγ in the non-targeted tissues and by the other Ppar isoforms, Pparα and Pparβ/δ, which enhance fatty-acid oxidation to minimize hyperlipidaemia and the resulting insulin resistance.

The complete absence of adipose tissue in Pparγ^hyp/hyp mice established that adipose tissue is crucial for the insulin-sensitizing effects of TZDs, as TZD treatment of these mice ameliorated the glucose intolerance but not the insulin resistance (Koutnikova et al, 2003). By contrast, mice with liver or muscle deletions of Pparγ responded to TZD treatment by reducing serum lipid levels, and by normalizing plasma glucose and insulin levels (Gavrilova et al, 2003; Hevener et al, 2003; Matsusue et al, 2003; Norris et al, 2003), which indicates that the muscle and liver Pparγ are not essential for this response. Despite this, the characterization of mice with a musclespecific deletion of Pparγ showed that muscle Pparγ is important for insulin-stimulated glucose disposal (Hevener et al, 2003; Norris et al, 2003). Studies with Pparγ obliterated in the liver showed that, in the presence of WAT, the impact of Pparγ on glucose homeostasis was minimal. However, in the absence of WAT, the ability of TZDs to lower triglycerides and glucose was dependent on liver Pparγ (Gavrilova et al, 2003). The fact that in this 'fatless' model the lowering of FFAs in response to TZDs was not dependent on liver Pparγ further points to a role for muscle Pparγ in FFA combustion. Pparγ in each of these metabolic tissues therefore contributes, in part, to maintaining whole-body insulin sensitivity. Taken together, these results indicate that WAT, rather than liver and muscle, is predominantly required for the insulinsensitizing effects of Pparγ (Gavrilova et al, 2003; He et al, 2003; Koutnikova et al, 2003; Matsusue et al, 2003; Norris et al, 2003). Therefore, Pparγ is not only the master regulator of adipogenesis in vivo, but also a driving force of glucose and lipid homeostasis. The knowledge obtained from the tissuespecific deletions of Pparγ will be further refined by the generation of mouse models that carry specific Pparγ mutations, such as the recently described knock-in of Ala at position 112 (S112A). This mouse, in which Pparγ is constitutively active (similar to the human Pro115Gln mutation), retains insulin sensitivity during diet-induced obesity (Rangwala et al, 2003).

The fourth and final line of evidence comprises pharmacological studies. PPARγ binds multiple ligands that can modulate its activity and induce a range of activities from full inhibition (antagonist) to activation (agonist) (Fig 2). The widely differing effects of various PPARγ ligands are attributed to differential cofactor recruitment (Rocchi et al, 2001). Moreover, selective PPAR modulators (SPPARMs) are compounds that activate or inactivate the PPAR isoforms in a tissuespecific manner. Full PPARγ agonists, such as the TZDs, improve insulin sensitivity, glucose tolerance and the lipidemic profile in type 2 diabetic patients. TZDs increase WAT mass, redistribute WAT from visceral to subcutaneous deposits and induce the appearance of small, newly differentiated adipocytes (reviewed in Picard & Auwerx, 2002). The observation that PPARγ agonists increase fat mass and improve glucose control supports the theory that WAT mediates some of their effects on glucose homeostasis. Besides FFAs, WAT secretes several adipokines, such as TNFα, leptin, resistin and adiponectin, which affect insulin signalling in other tissues (reviewed in Aldhahi & Hamdy, 2003). Whereas enhanced PPARγ activity is invariably associated with increased fat mass, suboptimal PPARγ activation or PPARγ antagonism is neutral or even reverses weight gain. This principle is illustrated by the binding of the partial agonist FMOC-L-Leu to PPARγ, which induces the differential recruitment of co-regulators to PPARγ such that glucose levels are still lowered but there is no weight gain (Rocchi et al, 2001). Other weak or partial PPARγ agonists, such as NC-2100 (Fukui et al, 2000) and MCC-555 (Reginato et al, 1998), also have little effect on adipocyte differentiation, but have potent antidiabetic activities. Partial inhibition of either PPARγ or its heterodimerization partner the retinoid X receptor (RXR) by antagonists (Mukherjee et al, 2000; Yamauchi et al, 2001; Rieusset et al, 2002) also improves insulin sensitivity, which is consistent with human and mouse genetic studies. Taken together with the mouse models of ablated PPARγ expression in metabolic tissues and human mutational analysis, these pharmacological studies show that PPARγ activity corresponds directly to adiposity in a linear fashion (Fig 2). However, unlike the relationship between PPARγ and fat mass, PPARγ activity is not linearly related to insulin sensitivity, as both inactivation and activation of PPARγ can enhance insulin sensitivity (Fig 2). This nonlinear relationship indicates that insulin sensitivity is an integrated effect, which is achieved predominantly by modulating PPARγ actions in the adipose tissue with effects on adipokine secretion and lipid storage, in addition to other tissuespecific PPARγ responses, as shown by tissue-specific PPARγ-obliterated models.

PPARγ: from inflammation to atherosclerosis

In addition to these thrifty activities in WAT, other PPARγ activities might also be beneficial, such as its role in modulating lipid homeostasis in arterial macrophages. Macrophage uptake of atherogenic lipoproteins in the arterial wall results in cholesteryl ester deposition and foam-cell formation, which are hallmarks of early and late atherosclerosis. PPARγ targets both lipoprotein uptake and cholesterol efflux, which are two competing processes that are involved in macrophage lipid homeostasis. PPARγ activation induces expression of the scavenger receptor CD36, thereby promoting oxidized low-density lipoprotein (oxLDL) uptake and foam-cell formation. In addition to the acquisition of cholesterol, however, macrophage uptake of oxLDL provides the cell with naturally occurring PPARγ ligands, thereby promoting further PPARγ activation and CD36 upregulation. Such a feed-forward cycle predicts that PPARγ is predominantly pro-atherogenic (Nagy et al, 1998; Tontonoz et al, 1998). PPARγ activation, however, might also be anti-atherogenic by reducing the expression of another scavenger receptor, scavenger receptor type AI/II (Moore et al, 2001), and by decreasing inflammatory cytokine production by macrophages (Jiang et al, 1998; Marx et al, 1998; Ricote et al, 1998). In addition to its effect on lipoprotein uptake, PPARγ activation promotes the removal of cholesterol from macrophages through enhancing the cholesterol efflux mediated by ATP-binding cassette transporter A1 (ABCA1). This stimulates high-density lipoprotein (HDL) formation and reverse cholesterol transport. The expression of ABCA1 is tightly regulated by cellular cholesterol content through the oxysterol-dependent activation of another nuclear receptor, the liver X receptor (LXR). PPARγ has been shown to activate ABCA1 expression indirectly through enhanced transcription of LXR (Chawla et al, 2001; Claudel et al, 2001).

These data emphasize the potential involvement of PPARγ in the pathogenesis of atherosclerosis, which was underscored in humans by the association of the PPARγ Pro12Ala polymorphism with protection from coronary heart disease (Ridker et al, 2003). The efficacy of PPARγ agonists in reducing atherosclerosis in mice (Chen et al, 2001; Claudel et al, 2001; Collins et al, 2001; Li et al, 2000) and exerting vasculoprotective effects in humans (Takagi et al, 2000; Koshiyama et al, 2001; Sekiya et al, 2001; de Dios et al, 2003) can be attributed to its effects on both inflammation and cholesterol efflux. Further studies are required to define how much these effects are due to the beneficial systemic metabolic effects of PPARγ versus its vascular and immune effects, which themselves might be indirect and mediated through LXR. It is also interesting to speculate that the increased frequency of atherosclerosis could be a PPARγ-driven maladapted macrophage response that occurs when the inherent beneficial effects (stimulation of the innate immune response and cholesterol efflux) are overwhelmed by the pro-atherogenic effects (increased oxLDL uptake). Such conditions are probably created by the chronic overload of lipids that is associated with an affluent westernized lifestyle.

PPARγ, caloric restriction, fat and ageing

From an evolutionary perspective, a 'thrifty response' clearly favours survival. However, caloric restriction, which refers to a dietary regimen that is low in calories without undernutrition, is also known to extend lifespan in species ranging from yeast to non-human primates (Weindruch, 1988; Sohal & Weindruch, 1996; Lane et al, 2001) (Fig 3). The beneficial effects of caloric restriction have been associated with alterations in metabolism, particularly the insulin/insulin-like growth factor 1 (IGF-1) pathways, and a decreased fat mass (Weindruch, 1988). These pathways converge on the FOXO forkhead transcription factors, which are activated when insulin/IGF-1 signalling decreases, and this translates ultimately into increased stress resistance, which is the hallmark of caloric restriction (Guarente, 2000; Nemoto & Finkel, 2002; Tran et al, 2002). The most plausible hypothesis to explain the anti-ageing effects of caloric restriction states that the reduced flow of carbon through the glycolytic pathways slows down the conversion of NAD⁺ to NADH, which is recognized by neuroendocrine sensors that ultimately reduce the production of growth hormone by the pituitary (and, in turn, IGF-1) and of insulin by the pancreas (Guarente, 2000). It has been proposed that the sirtuins, which are a family of NAD⁺-regulated protein deacetylases, have a role in these signalling events (Koubova & Guarente, 2003). The importance of insulin/IGF-1 signalling in longevity is supported by the extended lifespan of humans, mice, Caenorhabditis elegans and Drosophila that carry mutations in the insulin/growth hormone/IGF-1/FOXO signalling pathways (Bluher et al, 2003; and reviewed in Koubova & Guarente, 2003; Tatar et al, 2003). In contrast to caloric restriction, mice with increased fat mass have a shorter lifespan (Harrison et al, 1984; Smith et al, 1991). It is, at present, unclear how PPARγ interfaces with this signalling pathway, but the fact that PPARγ coordinates adipogenesis and glucose homeostasis leads us to speculate that it might also affect longevity (Fig 3).

Figure 3.

The effect of caloric restriction on energy metabolism translates into longevity. Caloric restriction decreases energy levels, which leads to the activation of a signalling cascade to enhance longevity. Decreased glucose intake reduces the flow of carbon through the glycolytic pathway and slows down the conversion of ADP to ATP, which eventually alters the NAD⁺:NADH ratio. SIRT1, which is a NAD⁺-regulated chromatin deacetylase, prolongs lifespan in response to caloric restriction in lower organisms. Also, signalling by insulin/insulin-like growth factor 1 (IGF-1) is attenuated under these conditions, which allows the FOXO forkhead transcription factors to increase stress resistance (pro-ageing). As extremes in fat mass are inversely related to lifespan, this indicates that PPARγ could also affect longevity, although the mechanisms are not yet established.

Too much of a good thing

Collectively, these data indicate that moderate levels of PPARγ activation coordinate an evolutionarily beneficial and adaptive response (Auwerx, 1999). Throughout human evolution, longevity has been improved because efficient energy conservation and storage has allowed survival through periods of food shortages, whereas an enhanced innate immune response combated infections, uncontrolled cell proliferation and cancers. Our present affluent lifestyle, which exposes us to excessive levels of natural PPARγ activators, throws this tightly regulated system out of balance. Being born with a 'silver spoon in your mouth' (that is, enjoying an affluent and often sedentary lifestyle) now turns this once favourable energy conservation response into a detrimental one, which contributes to the pathogenesis of lifestyle-associated diseases, such as obesity, type 2 diabetes and atherosclerosis. This also indicates that modulating (or inhibiting) PPARγ activity, rather than activating it, might be the preferred therapeutic strategy to treat metabolic disorders, in order to improve glucose homeostasis yet prevent adipogenesis.