Peroxisome proliferator-activated receptor gamma (PPARγ) is a transcription factor that was originally cloned as a regulator that binds to the promoter of the FABP4/aP2 gene . FABP4/aP2 is a gene that codes for a fatty acid-binding protein highly expressed in adipocytes. PPARγ controls both pre-adipocyte differentiation and lipid storage (reviewed in Rosen & Spiegelman ) and is recognized as a key regulator of adipocyte function. A cohort of later studies have also revealed that PPARγ is expressed and active in a variety of other cell types, including (but not limited to) immune cells, skeletal muscle cells, gastrointestinal epithelial cells, osteoblasts and osteoclasts . It is an important transcription factor that plays roles in an ever increasing list of regulatory mechanisms in metabolism, inflammation, endothelial function, cancer, atherosclerosis and bone morphogenesis. Additionally, PPARγ has gained instant medical relevance through the fortuitous discovery that members of the thiazolidinedione (TZD) class of drugs (comprising troglitazone, rosiglitazone, pioglitazone and ciglitazone) can be used to improve insulin sensitivity in type 2 diabetic patients and are high-affinity ligands for PPARγ. PPARγ can also be used as a model transcription factor to study the paradigms of transcriptional regulation. For these reasons PPARγ has always been in the spotlight of intense research. It is surprising therefore that some of the most basic questions of PPARγ biology are still unresolved. A review that attempts to summarize the current status of PPARγ research and to highlight the open questions will lead to an inherently diverse account. In this review we will focus on a selected few fields in which new discoveries have provided important insights into the mode of action of PPARγ in inflammation, metabolism and cancer research.
PPARs: fatty acid sensors among nuclear receptors
A nuclear protein, PPARα, was originally identified as a target of xenobiotic compounds that triggered proliferation of peroxisomes in rodent liver . It was later realized that there are three isoforms of PPARs, which include PPARα, PPARδ (also known as PPARβ) and PPARγ. It has also become evident that PPARs are key regulators of metabolism. PPARs belong to a large group (48 members in humans) of related transcription factors referred to as the nuclear receptor superfamily. Nuclear receptors can bind small lyophilic ligands that induce their transcriptional activities. For certain nuclear receptors, including PPARs, the identity of their endogenous ligands remains largely elusive. These members of the superfamily are called orphan receptors. Once it becomes apparent that a nuclear receptor can bind endogenous ligands they become ‘adopted’. Upon ligand binding, the ‘adopted’ nuclear receptors can activate or repress the transcription of a broad spectrum of genes (for a review on nuclear receptors, see Mangelsdorf et al. ).
PPARs have a prototypic nuclear receptor domain structure. The highly variable N terminal domain is a weak transcriptional activator (AF-1), whose activity is independent of the presence of ligands. The DNA-binding domain contains two zinc finger domains that can bind to specific DNA sequences. This domain is linked to the ligand-binding domain through a flexible hinge region. The ligand-binding domain has a conserved structure and is responsible for the strong ligand-dependent transcriptional activity (AF-2). The DNA-binding domain and ligand-binding domain regions together ensure that nuclear receptors act as part of a multiprotein complex. These two domains serve as a dimerization interface and can also bind co-activators or co-repressors.
PPARγ: a transcriptional activator at first sight
Due to its importance in the treatment of type 2 diabetes, PPARγ is the most intensively studied PPAR isoform. There are two distinct isoforms of PPARγ, termed PPARγ1 and PPARγ2. Both isoforms are transcribed from the same gene, but PPARγ2 differs by an extra N terminal motif (28 amino acids in human or 30 in mouse) . PPARγ2, which has a stronger transcriptional activity, is expressed at a high level almost exclusively in the adipose tissue, while PPARγ1 shows a more ubiquitous and lower level of expression. PPARγ, like other nuclear receptors, binds lipophylic ligands and regulates transcription upon ligand binding. Since the identification of endogenous ligands can reveal the function of a receptor, the nature of the ligands is one of the most fundamental information on any nuclear receptor. As a result of significant efforts to identify PPARγ ligands several candidates were found, including unsaturated fatty acids, arachidonic acid metabolites 15-deoxy-Δ12,14-prostaglandin J2 (15d-PGJ2) [4,7] and 15-HETE , and fatty acid-derived components of oxidized low-density lipoproteins (oxLDL), 9- and 13-HODE . All of these endogenous ligands can activate PPARγin vitro. However, due to their low affinity to PPARγ and to their low or undetectable level within cells, definitive evidence for these candidates being bona fide in vivo ligands has been lacking. Recent findings (Itoh et al. in press) raise the possibilities that PPARγ covalently binds a subsets of fatty acids and that PPARγ can bind two ligand molecules at the same time. This latter could explain why the ligand binding pocket of PPARγ is unusually large compared to that of related nuclear receptors. The proposed simultaneous ligand binding would also support the hypothesis that PPARγ is not a specific target for one particular fatty acid but it is a sensor molecule that samples the intracellular mixture of various fatty acid molecules.
Upon ligand binding PPARγ binds to canonical peroxisome proliferator response elements (PPREs) that are usually located upstream of acutely regulated PPARγ target genes. PPREs consist of a direct repeat of the nucleotide sequence TGA(A/C)CT separated by a single nucleotide to form the so-called DR1 promoter element. PPARγ belongs to the group of nuclear receptors (along with, for example, liver X receptor and retinoic acid receptor) that bind to their respective response elements when they form obligate heterodimers with their obligate partner, a promiscuous nuclear receptor, retinoid X receptor (RXR). PPARγ/RXR heterodimers recruit a protein complex that consists of co-activators or co-repressors to direct transcriptional activation or repression, respectively (Fig. 1). So far, the binding of the PPARγ/RXR heterodimer to PPREs has been shown to result in the stimulation of transcription. Our current understanding is that in the absence of ligands, PPARγ/RXR heterodimers bind to PPREs in a conformation that favours the binding of co-repressor molecules leading to the inhibition of transcriptional activity. Upon ligand binding, however, the heterodimers undergo conformational changes that promote co-activator recruitment and positive regulation of gene expression. Although the general statement above that PPARγ activation usually directly stimulates transcription is true, it must be emphasized that PPARγ activation by ligand binding can also lead to the repression of gene expression (Fig. 1). However, the ligand-dependent gene repression by PPARγ is thought to occur through indirect regulatory mechanisms. This form of gene repression is particularly relevant because PPARγ is thought to exert its anti-inflammatory action mainly via such indirect regulatory mechanisms. The full spectrum of the indirect mechanisms responsible for ligand-dependent gene repression is not known, but at least some of the negative regulatory effects of PPARγ are mediated through protein–protein interactions resulting in a phenomenon termed trans-repression. There are several models that aim at to explain how trans-repression occurs [10,11]. According to the best characterized model, there is an inhibitory protein complex consisting of co-repressors on the promoter of inflammatory genes (e.g. inducible nitric oxide synthase 2 or iNOS2) and this inhibitory complex represses transcription in the absence of inflammatory signals. The inhibitory complex needs to be removed by ubiquitin-dependent proteosomal degradation to ensure gene expression upon inflammatory stimuli. Liganded PPARγ is modified by sumoylation and binds to and stabilizes this inhibitory complex by blocking the proteosomal degradation of the co-repressor complex and the recruitment of the co-activators. In this way, liganded PPARγ can maintain the repression of inflammatory genes even in the presence of inflammatory stimuli. A more indirect mechanism for trans-repression might be squelching, in which case ligand-activated PPARs sequester some of the co-activator molecules (such as CREB-binding molecule) that are present in a limited amount in cells and would be needed for the full activity of the other transcription factors. PPARγ ligands (particularly 15d-PGJ2) were also shown to interfere with nuclear factor-kappa B (NF-κB) signalling either by inhibiting NF-κB DNA binding or by activating IκB kinase, a negative regulator of the NF-κB pathway [12,13]. However, at least a subset of this inhibitory effect on NF-κB is independent of PPARγ. Although the above mechanisms can explain a subset of the anti-inflammatory effects of PPARγ, it cannot be excluded that positive transcriptional regulation of inhibitory factors rather than these trans-repression mechanisms play important roles in ligand-induced repression. Furthermore, the in vivo relevance and the contribution of these proposed mechanisms to inhibition of gene expression remains to be explored and further established.
PPARγ at the interface of metabolism and inflammation
The notion that PPARγ lies at the interface of metabolism and immune biology is derived from several sources. The interrogation of the transcriptional activities of PPARγ has led us and others to the realization that PPARγ is proficient in the direct regulation of a module of genes mostly responsible for lipid transport and metabolism. The surprising finding is that PPARγ is also active in inflammatory cells and exerts its biological function in immune cells primarily through the regulation of principally a very similar module of genes . The idea that metabolism and immunity are inherently linked is also suggested by studies showing that metabolic syndrome, a group of related symptoms including obesity, high blood pressure, dyslipidaemia and type 2 diabetes, has distinct inflammatory components . Patients with metabolic syndrome exhibit low grade chronic inflammation that contributes to their metabolic disease. Conversely, inhibition of some of the regulators of this chronic inflammatory state relieves metabolic syndrome . Considering that PPARγ is expressed in key metabolic tissues and in inflammatory cells alike, it is a prime candidate for being one of the key regulators that links metabolism to immune regulation.
Models for studying PPARγ activation
There are several types of experimental approaches from which we have learned a great deal about PPARγ. The most basic approach is to interrogate transcriptional events by treating cells in vitro with either one of the potential endogenous ligands or with high-affinity synthetic ligands, some of which are used in therapy of type 2 diabetes. The synthetic ligands of PPARγ are also able to activate other nuclear receptors (most often the closely related PPARδ) and possibly other proteins that are not related to PPARs. As a result, the cellular phenotypes resulting from PPARγ ligand treatments should be considered as a combined effect of PPARγ dependent and independent transcriptional activities. The other difficulty in the interpretation of these experiments derives from the fact that the synthetic ligands (such as the TZD drug rosiglitazone that is marketed as Avandia) can bind to and activate PPARγ with a much higher affinity than the potential endogenous ligands. This raises the important question if some of the transcriptional events initiated by binding of a synthetic ligand to PPARγ are, in fact, not part of the physiological repertoire of endogenous PPARγ action, but rather they represent a superphysiological response. It is also a point for consideration that it is difficult, if not impossible, to culture some of the cell types in which PPARγ should be studied. In summary, this line of research can lead to results that need to be corroborated in in vivo experiments. For this reason in vivo approaches are being pursued to relieve us from the limitation of the in vitro systems.
There are two main strategies for deciphering the in vivo function of PPARγ. Unfortunately, both approaches have their own weaknesses. One strategy, that is analogous to the in vitro approach, is to treat wild-type animals with synthetic agonists to activate PPARγ. This is a simple solution that can give results relatively fast but suffers from the following pitfalls. As it has already been mentioned, the PPARγ dependence of the ligand action needs to be proven in these experiments. It is also difficult to determine which cell types contribute to the resulting phenotype since all the cell types that express PPARs are exposed to the ligand.
The use of genetically modified mouse strains in which one or both alleles of PPARγ are inactivated has also been exploited. The major obstacle in using mouse genetics to dissect PPARγ function is that PPARγ−/– animals are not viable . For this reason the first reports used mice that were heterozygous for PPARγ. Later developments in conditional knock-out strategies made it possible to create mouse strains in which homozygous deletion of PPARγ was achieved in select cell types [18–21].
In these Cre/lox conditional knock-out models, two mouse strains are needed to generate a homozygous gene deletion. One mouse strain carries a pair of short recognition sequences (termed loxP sites) flanking a critical region of the gene that is to be deleted (e.g. PPARγ). This modification results in a so-called floxed allele. The other mouse strain expresses the gene for the Cre endonuclease, which can recognize the loxP sites and delete the genomic region that lies between the two loxP sites. Neither of these parent strains shows deletion of the targeted gene (e.g. PPARγ) because each of them carries only one of the prerequisites of a successful deletion, namely, the presence of floxed alleles or the expression of Cre. However, the offsprings that derive from the crossing of the two parental strains carry both the floxed alleles and the gene for the Cre recombinase, and therefore will delete the floxed PPARγ alleles in cells in which the Cre protein is expressed.
The Cre expressing mouse strains used in the experiments to delete PPARγ exhibited cell-specific Cre expression. This ensured that the offsprings that are derived from the crossings of the floxed PPARγ and the Cre expressor strains were viable and only a specified subset of cells harboured a homozygous deletion of PPARγ. There are several advantages of using such conditional mutagenesis in PPARγ research. The most obvious one is that although the whole body knockout of PPARγ is embryonic lethal, the cell type-specific (e.g. adipocyte or macrophage specific) PPARγ−/– animals are viable and breed easily. Additionally, any phenotype observed in these models is strictly derived from a deficiency in PPARγ activity. In addition, cell type-restricted PPARγ deletion can help us reveal which cells are relevant in different subsets of PPARγ activities. This can be illustrated in the treatment of insulin resistance by TZD therapy in current medical practice. Although TZDs have been in use to improve insulin resistance in patients with type 2 diabetes for years, the question of which tissues are the targets of TZDs was left unanswered. Adipose tissue, skeletal muscle, liver and pancreatic β-cells were among the candidates. As it will be discussed later, a recent flurry of publications raised the intriguing and surprising possibility that macrophage PPARγ played an important role to ensure proper insulin sensitivity [22,23].
However, a cautionary note must be added to the use of cell type-specific Cre mouse strains. Most of the Cre lines that have been used in the PPARγ field were originally designed to exhibit strict cell type-specific Cre expression. However, the majority of these Cre lines express Cre in a broader population of cells than they were originally thought to do. Good examples for this are the LysozymeM-Cre line used for macrophage-specific PPARγ deletion that also caused deletion in neutrophils and possibly in some dendritic cells (DC), and the Tie2 line that was designed to be an endothelial-specific Cre strain but showed partial deletion in almost all haematopoietic lineages. This represents a trade-off in the conditional mutagenesis strategy that cannot be mitigated until a new generation of truly cell-specific Cre mouse strains is developed.
Finally, an important source of information that can help dissect PPARγ function is derived from medical practice. There are millions of people worldwide currently being treated with TZDs for insulin resistance. Due to the chronic nature of their disease, these patients have been taking PPARγ ligands for years. Statistical analysis of the incidence of diseases that were linked to PPARγ activity in animal models can reveal yet unidentified long-term effects of PPARγ agonists and can corroborate or refute the results derived from animal models. The importance of this kind of statistical analyses cannot be overstated. We have several well characterized PPARγ ligands at our disposal that show tolerable side-effects and have already undergone rigorous clinical trials. It is plausible that some of the diseases that show involvement of PPARγ activity (in colitis, allergic encephalomyelitis, colon cancer, etc.) could be ameliorated by a PPARγ agonist treatment [24,25].
Although all of the above-mentioned model systems have their own advantages, none of them can be considered to be absolutely superior. Clinical data cannot be properly interpreted without a solid knowledge of the underlying molecular mechanisms including transcriptional activities. On the other hand, the results of in vitro transcriptional studies must be challenged in in vivo models. It is also important to realize that PPARγ links the regulation of lipid metabolism to immune regulation. Both metabolism and immune biology can be studied in rodent models, but due to our 65 million years of divergent evolution, caution must be exercised when interpreting human metabolism and immune biology based on rodent experiments.
PPARγ as a metabolic regulator
The best characterized example of PPARγ activity is its role in adipose tissue. The finding that ectopic expression of PPARγ triggered adipocyte differentiation in NIH-3T3 cells, a non-adipogenic mouse fibroblastic cell line, demonstrated that PPARγ was sufficient to induce adipocyte differentiation . Additionally, no transcription factor has been identified that could promote adipogenesis in the absence of PPARγ. It was demonstrated that PPARγ was also required for the maintenance of the differentiated state of adipocytes. Inhibiting PPARγ activity, either in 3T3-L1 adipocytes or in the adipose tissue of mice, was shown to lead to dedifferentiation or adipocyte death . In line of the in vitro studies, the importance of PPARγ in adipogenesis was also shown in a chimeric mouse model. In this model, chimeric mice derived from wild-type and PPARγ−/– ES cells displayed lack of contribution of PPARγ−/– cells to the white adipose tissue of these animals .
Several studies aimed at identifying those genes that are under the control of PPARγ in adipocytes . The changes in gene expression in adipocytes in the presence of PPARγ ligands are difficult to interpret. The main reason for this is that beside PPARγ, a handful of other key regulatory transcription factors (such as C/EBPs or CCAAT/enhancer binding proteins, and KLFs or Krüppel-like factors) are switched on during adipocyte differentiation (for a review, see Rosen & MacDougald ). Despite this hurdle, several direct PPARγ target genes have been identified, including phosphoenolpyruvate carboxykinase, uncoupling protein 1 in brown adipocytes and lipoprotein lipase. Importantly, functional PPARγ response elements were identified in the regulatory regions of these genes [7,31,32].
Adipose tissue is a key regulator of energy balance (for a review on adipose tissue and the role of adypocyte dysfunction in insulin resistance, see Guilherme et al. ). White adipose tissue is specialized for lipid storage, while brown adipose tissue plays an important role in thermogenesis. White adipose tissue is not a passive lipid depot, but can be regarded as a dynamic buffer system that controls the level of triglycerides and non-esterified fatty acids in the circulation. Adipose tissue also has an endocrine function, because it secretes a range of metabolically active hormones (such as adiponectin and leptin)  and other bioactive substances (adipocytokines) such as interleukin-6 (IL-6), IL-18, tumour necrosis factor-alpha (TNFα), monocyte chemoattractant protein (MCP-1) and plasminogen activator (PAI-1) [35–38]. While leptin and adiponectin seem to be exclusively produced by adipocytes, it is possible that a significant portion of IL-6, TNFα and other secreted inflammatory factors within adipose tissue derive from cells other than adipocytes, such as macrophages that infiltrate adipose tissue.
Outside of the scope of fat tissue development and lipid metabolism, the most relevant metabolic consequence of PPARγ activation is insulin sensitization. While the adipocyte-specific PPARγ activity cannot account for the whole amplitude of the insulin sensitization effect of PPARγ ligands, it is still inherently linked to insulin sensitization. Because PPARγ is a positive regulator of adipogenesis, it is possible that PPARγ improves insulin sensitivity through its adipocyte-specific activities by expanding adipose tissue. The suggestion was put forward that PPARγ instructs adipose tissue to retain the bulk of the dietary lipids. This sequesters fatty acids that, if their levels are increased, could impair the insulin sensitivity of other insulin target tissues, including skeletal muscle and liver.
PPARγ is also expressed in other insulin target tissues besides adipose tissue and it is not clear in which tissues and to what extent does PPARγ activation in these cells play a role in the improvement of insulin sensitivity. It is possible that PPARγ in skeletal muscle contributes to the overall improvement of insulin sensitivity. PPARγ is expressed at a low level in skeletal muscle, but due to the large muscle mass, the net result of skeletal muscle PPARγ activation might have profound effects. Two similar mouse models to study skeletal muscle-specific PPARγ deficiencies gave inconsistent results showing either a modest or severe insulin resistance [39,40].
The metabolic consequences of PPARγ deficiency in humans were highlighted in a series of studies in which patients harbouring dominant mutations of PPARγ were studied. In this study heterozygous PPARγ mutations lead to a dominant negative inhibition of PPARγ functions . These mutations caused partial (and stereotyped) lipodystrophy, hepatic steatosis, insulin resistance and marked dyslipidaemia.
PPARγ is a key regulator of adipogenesis, regulates genes responsible for lipid metabolism and its ligand-mediated activation leads to improvement in insulin sensitivity; therefore, it seems to be a metabolic regulator at first sight. However, an increasing body of experimental evidence suggests that PPARγ also has an anti-inflammatory effect. An emerging picture within obesity research shows that metabolic syndrome has a chronic, subacute inflammatory component. The findings, that metabolic syndrome has an inflammatory component, that PPARγ activation has an anti-inflammatory effect, and that PPARγ is expressed in both adipocytes and macrophages within adipose tissue, raise the intriguing possibility that the beneficial effect of PPARγ activity on insulin sensitivity might partially derive from its anti-inflammatory action. It must be added that obesity is associated with macrophage accumulation in adipose tissue and macrophage dysfunction is thought to contribute to metabolic syndrome . The inflammatory aspect of the metabolic syndrome and the role of the dysfunctional (sometimes referred to as ‘bad’) adipocytes in the aetiology of the disease is a new but rapidly evolving field (for a review, see Guilherme et al. ). Here we would only like to emphasize the results of a group of studies on the role of TNFα in adipocyte function. According to these findings TNFα can affect the level of PPARγ at multiple layers of regulation and can interfere with the proposed function of PPARγ as a regulator of triglyceride synthesis and storage . As it will be discussed later, a mutual negative regulation between PPARγ and TNFα is also present in macrophages.
While the importance of PPARγ in adipocyte biology is uncontested, several lines of in vitro experiments show that PPARγ is functional in macrophages and DCs where it has anti-inflammatory functions as well. The observations, that immune cells pretreated with PPARγ ligands show abated inflammatory responses [44,45], and pretreatment with inflammatory agents (like TNFα in the case of adipose tissue) represses PPARγ activity, show us a mutual negative regulation. In vivo observations that certain inflammatory conditions (e.g. inflammatory bowel disease  and allergic encephalomyelitis [46,47], a mouse model for sclerosis multiplex) can be ameliorated by treatment with PPARγ ligands also point to the same direction. Here we will provide an overview of the transcriptional activity and functions of PPARγ in two closely related cell types, macrophages and DCs, which are instrumental in inflammation.
The importance of characterizing PPARγ activity in macrophages or DCs is twofold. As mentioned before, the analysis of the transcriptional activity of PPARγ in adipocytes is difficult because PPARγ activity is superimposed on an extensive gene expression profile change that occurs during adipocyte differentiation in which there is an ongoing PPARγ activity due to the presence of endogenous ligand(s). It is therefore difficult to dissect the transcriptional events that exclusively depend on PPARγ activation. Studying PPARγ activity in macrophages and DCs (especially in the monocyte-derived macrophage/DC differentiation model), however, provides a paradigm that is easier to interpret. A practical reason for this is that, compared with adipocytes, it is much easier to isolate myeloid primary cells (e.g. monocytes, or different subclasses of macrophages/DCs) for in vitro analysis. A more fundamental reason is that monocytes do not express or show PPARγ activity, and the rapid gene expression change that occurs during the early events of differentiation in the presence of PPARγ ligands is not confounded with the consequences of a pre-existing PPARγ activation. There is also a practical reason why studying macrophage/DC PPARγ activation is important. Inflammation is a strictly regulated process and dysregulation of inflammation contributes to a wide variety of diseases. The analysis of PPARγ mediated events in macrophages/DCs has direct medical relevance, because PPARγ activity has an anti-inflammatory effect.
PPARγ activity in macrophages
Macrophages are antigen-presenting cells of the innate immune system that are located throughout the body and monitor local environment by phagocytosis. It is well documented that PPARγ plays an important role in macrophage immunobiology. Pretreatment of macrophages with PPARγ ligands inhibits transcriptional activation of inflammatory genes upon inflammatory stimulus (e.g. lipopolysaccharide) [48,49]. The underlying mechanisms are not fully charted but it has been suggested that PPARγ activation exerts its anti-inflammatory function by trans-repressing the NF-κB and MAPK pathways that are central regulators of inflammatory genes. It must be added that PPARγ does not inhibit all subsets of NF-κB target genes and this rules out that a general inhibition of NF-κB signalling is responsible for the full spectrum of the anti-inflammatory effects of PPARγ. To confound the picture, it was also found that natural and synthetic ligands of PPARγ (15d-PGJ2 and TZDs) at high concentration could down-regulate a subset of inflammatory genes (TNFα and IL-6) even in PPARγ deficient macrophages . Therefore, it was suggested that at least a fraction of the immune repression is mediated independently of PPARγ, most probably through PPARβ/δ. Regardless of the possible anti-inflammatory contribution of PPARβ/δ to TZD treatments, PPARγ is an important regulator of inflammation in macrophages . It has even been suggested that PPARγ-positive macrophages are not simply repressed or inefficient inflammatory macrophages. Instead, PPARγ is proposed to be a key regulator of a distinct cell type, termed alternatively activated macrophages. The exact functions of the alternatively activated macrophages have not fully been characterized yet. They are generally produced in Th2 type immune responses upon macrophage exposure to IL-4 or IL-13. Due to the fact that Th2 cytokines positively regulate the alternative activation of macrophages, these macrophages are thought to regulate immune response to allergic, parasitic or extracellular pathogen challenges. Additionally, alternatively activated macrophages are also implicated in wound healing/repair, atherosclerosis, angiogenesis and tumour immunity (reviewed in Gordon ).
Another function of PPARγ in macrophages is the regulation of lipoprotein uptake. Lipoprotein uptake ensures that cells take up sufficient amount of cholesterol from the environment. While cholesterol is an essential lipid and required in every cell, excess cholesterol can lead to pathological conditions, including atherosclerosis. Lipoprotein uptake must be therefore tightly regulated. Several mechanisms exist for lipoprotein uptake. Lipoproteins can enter cells via the low density lipoprotein (LDL) receptor in most cells . Additionally, there is a distinct pathway for lipoprotein uptake in macrophages and this pathway appears to be regulated by PPARγ. This alternative pathway is based on the fact that LDLs can be modified and taken up in their oxidized form . Upon ligand binding PPARγ positively regulates two scavenger receptors, CD36 and SR-A, that are efficient receptors for oxLDLs [54,55]. Oxidized lipids inside the oxLDL particle such as 9-HODE and 13-HODE can activate PPARγ. If PPARγ up-regulates CD36 and SR-A in the presence of oxidized lipids, then this could lead to a positive feedback loop in which an ever-increasing intracellular oxidized lipid influx instructs PPARγ to further increase the expression of oxLDL receptors. The fact that TZD treatment can increase macrophage CD36 expression, but does not lead to the differentiation of abnormal, lipid-laden macrophages (termed foam cells) that are early hallmarks of atherosclerosis suggests that PPARγ can also activate a pathway that reduces intracellular cholesterol level despite the continuing oxLDL uptake. Such a pathway has been studied in our laboratory. We have shown that PPARγ activation with a concurrent retinoid pathway activation leads to the up-regulation of CYP27, a P450 enzyme. This enzyme can convert cholesterol to 27-hydroxycholesterol, a potential endogenous ligand of liver X receptor. Liver X receptor is another member of the nuclear receptor superfamily that regulates cholesterol efflux by inducing ABC transporters. In short, PPARγ signaling can boost lipoprotein uptake and at the same time can also enhance cholesterol efflux via liver X receptor dependent and independent pathways .
A thorough analysis of PPARγ ligand-treated mouse peritoneal macrophages and bone marrow-derived macrophages has found that rosiglitazone was able to induce only a handful of genes which played a role in lipid transport and metabolism, including CD36 and ABCG1 . The finding that macrophage PPARγ activation leads to a very limited change in gene expression in peritoneal macrophages raised the question how one can reconcile the limited change in gene expression with the results of the in vitro and in vivo models in which PPARγ ligand treatment or macrophage-specific PPARγ deletion brought about inflammatory and metabolic phenotypes. From the lessons our laboratory learned from studying PPARγ activation during the monocyte–DC transition several possible answers can be deduced. As it will be discussed, PPARγ is induced in a narrow developmental window in monocyte-derived DCs and it exerts its effects through several successive waves of changes in gene expression that last for several days. It is possible that mouse peritoneal macrophages are not an ideal model cells to study PPARγ action. It can be hypothesized that, similarly to DCs, macrophages respond to PPARγ activation only in a narrow developmental window, outside of which they are relatively refractory to PPARγ ligands. Macrophages also exhibit a remarkable heterogeneity and it is possible that PPARγ can robustly regulate gene expression in some but not all types of macrophages. Isolation of different subsets of macrophages in sufficient quantity and in a reproducible purity is technically demanding if not impossible for some cell types. Hence, a thorough characterization of macrophage subsets has been lacking. It is therefore yet to be seen if other macrophage subtypes can exhibit more robust PPARγ-mediated gene expression changes.
Our inadequate understanding of macrophage specific PPARγ activity is further illustrated by recent reports in which macrophage specific PPARγ deletion lead to insulin resistance and glucose intolerance in mice [22,23]. Although the results of the two original reports show inconsistencies regarding the site of the protective macrophage PPARγ effect, these reports illustrate that non-cell-autonomous effects of macrophage PPARγ activity might play a so far neglected role in the regulation of metabolism. These reports elegantly prove the central notion that PPAR links metabolism to inflammation and raise the question: which are the relevant PPARγ target cells that contribute to full body insulin sensitivity.
PPARγ in dendritic cells
Dendritic cells are professional antigen presenting cells of lymphoid or myeloid origin with a unique capacity to prime naive T cells. DCs are Janus-faced cells that decide about immune response. On one hand, mature DCs presenting both foreign antigens and costimulatory molecules can elicit a robust T-cell response. On the other hand, DCs can also inhibit adaptive immunity by triggering T-cell anergy or Treg expansion which negatively regulate immune response. Immature DCs constantly sample the local antigen pool through receptor mediated endocytosis, phagocytosis and macropinocytosis. Upon activation by inflammatory agents their antigen uptake decreases, they up-regulate costimulatory molecules and migrate to lymph nodes where they can activate T cells bearing cognate receptors against the antigen–major histocompability complex II.
Dendritic cells that express high level of PPARγ can be generated in vitro. DCs treated with PPARγ ligands exhibit dynamic changes in gene expression and therefore they can serve as a general model system to study the mechanisms of PPARγ action. Additionally, they can be used as a model system to investigate how transcriptional activities of PPARγ can lead to the modulation of immune functions.
Human DCs can be generated in vitro from CD34+ haematopoietic stem cells, from CD14+ peripheral monocytes and from blood peripheral DC precursors. In each model system, PPARγ is induced during the in vitro differentiation process. We have characterized the transcriptome of monocyte-derived DCs using gene expression profiling [14,41,56]. In this model system, monocytes are isolated from peripheral blood and then cultured in the presence of granulocyte-macrophage colony-stimulating factor (GMCSF) and IL-4. The monocytes which initially lack detectable PPARγ expression differentiate into immature DCs by day 5. The addition of the PPARγ ligand rosiglitazone modulates the gene expression events and the resulting phenotype of the DCs. There are several general features of the gene expression changes detected during the monocyte-immature DC transition upon the addition of rosiglitazone that are echoed in similar in vitro gene expression studies carried out in a variety of other cell types. The first such feature is that there is only a limited gene expression change that can be detected very early (6 h) upon differentiation, but successive wave(s) of secondary transcriptional activities will ensue leading to a massive change in the gene expression profile impacting the level of more than 1000 transcripts by day 5. The second general feature is that the majority of the genes that are up-regulated very early are responsible for lipid transport or lipid metabolism. It seems therefore that PPARγ does in DCs what it does best in fat cells, and that is regulation of lipid homeostasis. A third feature, which might be specific to this in vitro system, is that the expression of PPARγ is induced immediately at the start of differentiation but is down-regulated soon afterwards, meaning that PPARγ is expressed at a high level only in a narrow timeframe in a certain developmental window.
The question inevitably arises asking how the observed gene expression changes can modulate the function of a cell, whose primary function is immune regulation. As it has already been mentioned, genes that play a role in lipid homeostasis are over-represented among the immediate early genes that are acutely up-regulated upon rosiglitazone treatment in differentiating DCs. Therefore, PPARγ does not solely link metabolism and immune biology on the phenotypical level, but its connecting role is mirrored by the function of the genes it regulates. There is a much smaller set of transcripts (80 vs. 12) that are down-regulated at early time points of DC differentiation. This fits into the general assumption that PPARγ primarily acts as a transcriptional activator. Interestingly, however, this trend seems to be reversed by day 5 leading to a balanced change in gene expression profile showing, among others, regulation of genes of immunoregulatory functions. The question that needs to be resolved is that how a primarily positive regulation of early gene expression of lipid homeostasis genes is translated into a balanced expression profile change with a broad negative regulatory component.
The next question that needs to be addressed is what is the phenotypic consequence of PPARγ activation in DCs. PPARγ activation by synthetic ligands has been reported to affect several facets of DC functions in in vitro experiments. Immature DCs excel at antigen uptake by endocytosis. We have observed that rosiglitazone treatment enhanced this endocytic capacity . We also found that PPARγ directly induced the expression of ABCG2, a multidrug transporter molecule , that confers protection from xenobiotics. Murine splenic DCs and human monocyte-derived DCs also expressed less inducible IL-12 subunit and other Th1-specific inflammatory cytokines upon lipopolysaccharide stimulation if they were pretreated with rosiglitazone . These experiments suggest that PPARγ is an important regulator in immature DCs and maturation of DCs in the presence of PPARγ ligands might promote Th2 over Th1 T-cell responses.
From the interrogation of gene expression profile of PPARγ instructed DCs, we have identified another facet of DC function in which PPARγ played a decisive role. DCs cannot only present peptide derived antigens that are recognized by cognate T-cell receptors, but they are also able to present lipid antigens to invariant natural killer T cells. This is possible because PPARγ can regulate the expression of a group of cell surface molecules that are responsible for processing and presenting lipid antigens  (Fig. 2). During monocyte-derived DC differentiation, the cell surface level of the group I lipid presenting molecules (including CD1a, CD1b, CD1c and CD1e), that are expressed at a very low level in monocytes, increases sharply. In stark contrast, the closely related but distinct group II lipid presenting cell surface protein, termed CD1d, is expressed highly in monocytes and falls sharply during differentiation. If the DCs are differentiated in the presence of the PPARγ ligand rosiglitazone, however, the expression of the group I and II lipid presenting molecules is fundamentally changed. The cell surface expression of CD1a and other group I molecules is now kept repressed, while the robust monocyte-specific CD1d expression is sustained throughout the differentiation process. We have found that the high level expression of CD1d is sustained, at least in part, through the PPARγ-dependent activation of retinoic acid synthesis. The high level expression of CD1d in cells that express PPARγ ensures effective lipid antigen presentation to invariant natural killer T cells bearing semi-invariant T-cell receptors. The invariant natural killer T cells become activated and begin to proliferate when lipids are presented to them on DC CD1d molecules. Thus, the activation of PPARγ in DCs leads to the expansion of a T-cell subset that possesses robust IFNγ and IL4 producing capabilities and has the capability to mature DCs and licence them to prime antigen-specific T- and B-cell responses.
In vivo model systems in which PPARγ is inactivated strictly in DCs could corroborate the results of the in vitro or the in vivo experiments that were mainly carried out by ligand treatments. They would tell us what is the functional consequence of the PPARγ-modulated peptide or lipid presentation. This has been hindered by the unavailability of a suitable DC-specific Cre mouse strain. A recent publication of the CD11c-Cre mouse , however, opened the opportunity to study PPARγ-deficient DCs and further studies of the function of PPARγ in DCs can be anticipated.
PPARγ in intestinal epithelial cells: a case study of combined experimental approaches
PPARγ function has been implicated in the homeostasis of the gastrointestinal biology and in colon cancer. There are several reasons why PPARγ is a good candidate for being an important regulator of epithelial function in the gastrointestinal tract. On one hand, PPARγ1 is expressed in the epithelium from duodenum through rectum with the highest expression level found in the proximal colon . It is also highly expressed in crypt/villus junctions in the small intestine, at the site of terminal differentiation of epithelial cells. The gastrointestinal tract is also a place where the lumen epithelium is constantly exposed to a lipid-rich environment that contains potential activators of PPARγ. Hence, both prerequisites of a strong PPARγ action, high level of PPARγ expression and endogenous ligands, are present in the intestine. It is very important to note that gut epithelium is a major interface that is site of an extensive array of metabolic and immunological processes. Therefore, gut epithelial PPARγ could be an ideal candidate for a regulatory factor that connects metabolism and immunobiology.
The fact that PPARγ is induced in villus epithelial cells  that are about to enter their terminal differentiation stage can suggest two overlapping models. It might be necessary for successful completion of some of the steps required for differentiation of epithelial cells (e.g. block in proliferation or migration to villus tips). Alternatively, PPARγ might be required for physiological functions that are fulfilled by differentiated epithelial cells.
The role of PPARγ in colon cancer has been a subject of debate. PPARγ was found to either inhibit tumour growth or promote carcinogenesis depending on the cancer model used in the experiments. The history of this field provides an excellent example for showing how the improvements in mouse genetic technologies enabled the field to investigate the role of PPARγ in colon cancer in more and more refined mouse models.
One group of studies has shown that PPARγ ligands (TZDs) could inhibit sporadic colon carcinogenesis in a model system in which small rodents were treated with azoxymethane or dextran sulfate and the formation of aberrant crypt foci was monitored . This inhibition in tumorigenesis was not only seen in rats but in several strains of mice (namely, BALB/c and C57BL/6 J) . Additionally, several distinct synthetic PPARγ ligands could achieve dose-dependent inhibition of aberrant crypt foci in this model. These observations therefore indicate that TZDs are tumour suppressors in the azoxymethane model of sporadic colon cancer.
Another cohort of in vivo experiments, however, suggested that synthetic PPARγ ligands acted as tumour promoters [63,64]. In these experiments a mouse strain with a predisposition to colon cancer due to a point mutation in the adenomatosis polyposis coli gene (Apc+/Min) was treated with synthetic PPARγ ligands and was found to exhibit increased tumour incidence in the colon, but not in the small intestine. PPARγ constantly monitors lipid environment and modulate transcriptional activities as dictated by its locally available lyophilic ligands; therefore, it was hypothesized that high-fat diet increases the risk of colon via the activation of PPARγ.
Yet another twist was introduced in the story when heterozygous PPARγ+/– knockouts were used to investigate tumour incidence in the chemically induced azoxymethane model either in the absence or in the presence an additional APC mutation (APC+/1638N). While the heterozygous knockout of PPARγ did increase tumour incidence in the azoxymethane model in the absence of the APC mutation, it did not affect tumour incidence in the APC+/1638N model [65,66]. To explain this discrepancy, the hypothesis was put forward that PPARγ acted as a tumour suppressor in the azoxymethane model because it down-regulated β-catenin, an important activator of proto-oncogenes in gastrointestinal cancers. This down-regulation was partially relieved in the heterozygous PPARγ knockouts leading to increased tumour incidence. Because APC is a key regulator of β-catenin, the APC–β-catenin pathway was severely affected in the APC+/1638N mouse strain regardless of the status of PPARγ and, therefore, heterozygous PPARγ deletion could not affect tumour incidence in the APC+/1638N mouse strain. Paradoxically, it has also been reported that pioglitazone suppressed tumour growth in APC+/1309 mice .
Finally, further evolution in transgenic technologies lead to another model in which biallelic PPARγ deletion was achieved in a colon epithelium-specific manner. The villin-Cre/PPARγfl/flAPC+/Min mouse strain was predisposed to colon cancer due to its APC+/Min status and also had PPARγ-deficient colonic epithelial cells. This strain exhibited increased tumour incidence and size, corroborating some of the earlier studies .
Despite current efforts to delineate the downstream effect of PPARγ activation in epithelial cells of either healthy or malignantly transformed gastrointestinal system, no clear answers have been obtained. This is partly due to the fact that there are no good model systems in which PPARγ action could easily be studied in non-malignant epithelial cells. Recent attempts for gene expression profiling of PPARγ activation has focused either on non-malignant rat intestinal cells , on colorectal cancer cell lines  or on colonic epithelial cells isolated from mice . The results from these experiments point to the same direction showing PPARγ ligand-induced expression profile changes in genes responsible for metabolism, cell mobility, proliferation and cell signalling.
If we focus our attention to the role of PPARγ in malignant transformation, then it seems plausible that PPARγ activity leads to dysregulation of certain cellular functions that are considered to play key roles in tumorigenesis, such as cell proliferation, apoptosis, terminal differentiation and cell renewal capacity. Additionally, non-cell-autonomous processes can greatly affect the fate of malignant cells. Tumours need an ample supply of nutrients to grow and angiogenesis that leads to proper vascularization of the growing tumour mass can ensure that these nutrients are delivered. In addition, a successful immune response to malignant cells could eradicate tumours. The following question arises: which of the above functions are affected by PPARγ activity? This is a difficult question to answer, partly because as it is illustrated in the examples above, PPARγ is often implicated but not always conclusively proven to play a critical role in tumorigenesis. In vitro studies carried out in cell lines of different origin (including pancreatic cancers, breast tumours and non-small-cell lung cancers), suggest that PPARγ activity might lead to cell cycle arrest, trigger apoptosis or induce differentiation [53,69–73]. It is also possible that PPARγ activity interferes with non-cell-autonomous functions that are essential for tumour growth. Particularly interesting is the fact that PPARγ activation in endothelial cells leads to inhibition of proliferation, induces apoptosis and inhibits VEGF-induced cell migration. These and other observations suggest that PPARγ activity might have an anti-angiogenic effect . Finally, it is also possible that PPARγ activity might interfere with tumour growth by regulating immune response to tumour cells.