Endogenous Ligands for Nuclear Receptors: Digging Deeper*

Nuclear receptors (NRs) are hormone-sensing transcription factors that translate dietary or endocrine signals into changes in gene expression. Therefore, the adoption of orphan NRs through the identification of their endogenous ligands is a key element for our understanding of their biology. In this minireview, we give an update on recent progress in regard to endogenous ligands for a cluster of NRs with high sequence homology, namely peroxisome proliferator-activated receptors α and γ, Rev-erbα, and related receptors. This knowledge about the nature and physiology of these ligands may create new opportunities for therapeutic drug development.


General Considerations of Endogenous Ligands for Nuclear Receptors
The superfamily of nuclear receptors (NRs) 3 controls processes as diverse as development, inflammation, toxicology, reproduction, and metabolism (1). There are 48 members encoded in the human genome (49 in mouse) (1). The term endogenous ligand in regard to NRs describes a naturally occurring small molecule that elicits a conformational change in the NR upon binding (2). This conformational change alters the cellular location of the NRs and/or their interaction with cofactors, which ultimately translates into changes in gene expression and explains why these transcription factors are referred to as "ligand-activated" (3). Until the identification of such endogenous ligands, the receptor is called an orphan NR (1).
The known endogenous ligands for NRs consist of a wide range of chemical structures, such as bile acids, phospholipids, steroid hormones, thyroid hormone, retinoids, and vitamin D (4 -10). Thus, although many of these are derived from cholesterol, a definition of the term endogenous ligand based on chemical structure is impossible. Some endogenous ligands, such as estrogens, were used as radiolabeled reagents to find their corresponding NR by identifying their binding partners (1). Other ligands were found integrated in the NR ligand-binding domains (LBDs) by solving their crystal structure (11)(12)(13) or by immunoprecipitating the NR (14).
Because there are NRs that have been described to bind a variety of different physiological ligands, which is especially the case for NRs with a spacious LBD (15), it can be hard to determine which one is the "real" endogenous ligand (14,16,17). Future studies may present evidence beyond the proof of in vitro binding to consolidate these ongoing debates. As an example, the cellular levels of a previously identified ligand (18) were shown to be far lower than required to induce NR activation (19). Therefore, the required features for such ligands usually comprise nuclear availability, high binding potency (in the nanomolar range), ability to induce a conformational change in the protein structure, and a NR-related physiological function as a hormone (1).
Considering the abundant expression of several NRs, it has also been suggested that one and the same NR may have distinct endogenous ligands in distinct tissues or cell types (20). This is of specific interest because a therapeutic intervention could selectively target the availability of one ligand without interfering with the desired effects of another. These interventions could range from controlling the biosynthesis of endogenous ligands to the modulation of processes involved in their inactivation. Therefore, detailed knowledge will be required to design therapeutics that specifically modify the availability of endogenous ligands. In this minireview, we focus on peroxisome proliferator-activated receptors ␣ (PPAR␣; NR1C1) and ␥ (PPAR␥; NR1C3), Rev-erb␣ (NR1D1), and retinoic acid receptor-related orphan receptor ␣ (ROR␣; NR1F1), a cluster of NRs with high sequence homology for which much has recently been learned about endogenous ligands.

PPAR␣
PPAR␣ is highly expressed in liver and controls the expression of genes involved in fatty acid oxidation, ketogenesis, lipid transport, gluconeogenesis, glycogen metabolism, and inflammation (21,22). Mice with a disruption of the PPAR␣ gene develop hepatic steatosis due to impaired fatty acid oxidation and fasting hypoglycemia (22). PPAR␣ is therefore considered the key player in the hepatic fasting response. Activation of PPAR␣ by therapeutically used fibrates leads to a decrease in serum triglycerides and an increase in HDL cholesterol (23).
PPAR␣ and PPAR subtypes ␤ and ␥ bind to DNA as heterodimers with retinoid X receptor ␣ (RXR␣) (24). The binding motifs for all PPARs consist of a direct repeat of the core consensus half-site ATTGCA spaced by one nucleotide. These motifs, located in the promoter and enhancer regions of target genes, are referred to as PPAR-response elements (25)(26)(27).
Soon after its discovery, several studies described the activation of PPAR␣ by enzymatically derived derivatives of arachidonic acid, such as the eicosanoids leukotriene B 4 and (8S)-hydroxyeicosatetraenoic acids, carbaprostacyclin, and unsaturated fatty acids (20, 28 -32). Most of these ligands were identified using in vitro approaches. Because the actual source of these lipids is mainly dietary uptake, the concept of PPAR␣ as a physiological lipid sensor developed soon after (28,29).
In addition, oxidized phospholipids, which are components of oxidized LDLs, were identified to activate PPAR␣ in endothelial cells (5). This activation was dependent on the activity of phospholipase A 2 , which suggests that these phospholipids may be precursors for endogenously generated activators (5). Similarly, the enzyme that locally hydrolyzes triglyceride-rich lipoproteins and releases fatty acids, lipoprotein lipase, was shown to generate PPAR␣-activating species in these cells (33). Addition of albumin abrogated the PPAR␣ activation, which implicates protein binding of the lipoprotein lipasegenerated fatty acids as an important factor (34). Another phospholipid-related species, oleoylethanolamide (OEA), was shown to regulate feeding and body weight through PPAR␣ activation. OEA can bind and induce PPAR␣ in the nanomolar range (EC 50 ϭ 120 nM), increases lipolysis and fatty acid oxidation, and is found at PPAR␣-activating concentrations in the small intestine (35). Interestingly, also adipose tissue has been shown to generate OEA (36). Thus, OEA could link fasting-induced lipolysis to hepatic PPAR␣ activation. Further studies are required to fully understand OEA in the context of the hepatic fasting response.
In an elegant recent study, Chakravarthy et al. (14) re-expressed FLAG-tagged PPAR␣ in the livers of PPAR␣ Ϫ/Ϫ mice in the presence or absence of hepatic FAS and gently affinitycaptured the NR without disrupting the binding of potential endogenous ligands. Tandem mass spectrometry identified 1-palmitoyl-2-oleoly-sn-glycero-3-phosphocholine (16:0/18:1 GPC) as the only PPAR␣-bound lipid that was dependent on the presence of FAS, one of the putative ligand-synthesizing enzymes (37). 16:0/18:1 GPC was competitively displaced by synthetic PPAR␣ ligands, induced hepatic gene expression similar to synthetic ligands, and enhanced the interaction of the NR with cofactor peptides (14), thus fulfilling most of the criteria for a specific ligand. These findings were corroborated by studies showing that an application of 16:0/18:1 GPC to the portal vein system alleviated hepatic steatosis. Interestingly, the authors found that peripheral lipids, entering the liver via the liver artery, were unable to activate hepatic PPAR␣, suggesting that the origin of 16:0/18:1 GPC may be important (14). More studies are necessary to fully understand which role 16:0/18:1 GPC plays in the activation of hepatic PPAR␣ during fasting. Part of this is the determination of the mechanism that allows this hydrophilic/charged ligand to access the nucleus for binding PPAR␣. These findings may allow us to evaluate whether the endogenous synthesis/metabolism of 16:0/18:1 GPC could present a potential drug target for lipid disorders. Potential components of ligand formation for PPAR␣ are summarized in Fig. 1A.

PPAR␥
PPAR␥ is highly expressed in adipose tissue and is the master regulator of adipocyte differentiation (41). Lower expression of PPAR␥ can be found in other cell types, such as macrophages, where it regulates inflammation and glucose and lipid metabolism (42,43) through cell type-specific DNA binding (44). Thiazolidinediones (TZDs), which are used clinically as insulin-sensitizing drugs, are potent synthetic PPAR␥ activators that induce adipogenesis (45). Thus, knowledge about the bona fide endogenous PPAR␥ ligand could create new opportunities in the treatment of insulin resistance and obesity.
As for PPAR␣, oxidized LDL has been identified early on to contain putative PPAR␥ ligands. Examples are the dehydration product of prostaglandin D 2 , 15-deoxy-⌬ 12,14 -prostaglandin J 2 , and derivatives of certain mono-and polyunsaturated fatty acids (18, 31, 46 -48). Most of these ligands bind PPAR␥ with low affinity and may not be present at cellular concentrations necessary to activate this NR (19). By contrast, nitroalkene derivatives of linoleic acid, such as nitrolinoleic acid, were shown to bind PPAR␥ as potently as TZDs and were found at activating concentrations in human red cells and plasma (49,50). Although the chemistry of nitrolinoleic acid has been largely understood (51), the physiological context of these nitro derivatives in regard to PPAR␥ signaling is still unclear.
Early studies proposed the SREBF (sterol regulatory element-binding transcription factor) to link PPAR␥ ligand production to fatty acid synthesis (52). Although the SREBF-dependent synthesized compound was secreted into cell culture media (52), its identity has not been determined. Other approaches investigated the stage of adipocyte differentiation, during which the production of an endogenous ligand is most likely. Using a reporter cell system, Tzameli et al. (53) demonstrated that a PPAR␥ LBD-activating species was produced downstream of C/EBP␤ (CCAAT/enhancer-binding protein ␤), concomitant with the induction of PPAR␥ expression in 3T3-L1 cells. Follow-up studies implicated the enzyme xanthine oxidoreductase (XOR) in this transitional rise in PPAR␥ activity, and mice deficient in XOR were found to have less adipose tissue (54). However, the enzymatic product of XOR driving adipogenesis is still unknown. Depletion of the enzyme retinol saturase (RetSat), a direct transcriptional target of PPAR␥ and induced early during differentiation, inhibited adipocyte conversion (55). Providing the putative product 13,14-dihydroretinol did not rescue adipocyte differentiation, which suggests that this enzyme may have an additional role than the sole generation of 13,14-dihydroretinol (55). RetSat overexpression enhanced adipocyte conversion and PPAR␥ activity; thus, RetSat enzymatic activity may be required for the production of an endogenous ligand. Mice with a RetSat deletion exhibited a surprising increase in adiposity and PPAR␥ target gene expression (56). This shows the complexity of RetSat function in vivo, which needs further investigation.
eLOX3 (epidermis-type lipoxygenase 3) has been described to play a role in early adipocyte differentiation by the generation of the arachidonic acid derivatives hepoxilins (16). Interestingly, Hallenborg et al. (16) could show that eLOX3 cooperates with XOR in the activation of a Gal DNA-binding domain/PPAR␥ LBD reporter, which suggests that both enzymes converge in the same pathway. This pathway could work via reactive oxygen species (ROS) as intermediates, which are generated by XOR and further metabolized by eLOX3 (16,54). Indeed, neutralizing ROS by the antioxidant N-acetylcysteine inhibits adipocyte differentiation (16). Moreover, an RNAi-based genetic screen for oxidative stress resistance discovered that silencing RetSat increased the survival of oxidant-exposed fibroblasts (57). This could implicate RetSat in ROS metabolism. However, the resulting hypothesis, that the synthesis of an endogenous PPAR␥ ligand involves radical lipid modifications, needs further investigation.
By using a reverse strategy, searching for natural metabolites that contain the indole acetate structure of the synthetic PPAR␥ ligand indomethacin, a recent study identified cellular 5-hydroxytryptamine (5-HT; serotonin) metabolites as PPAR␥ ligands (58). Waku et al. (58) showed that 5-methoxyindole acetate and 5-hydroxyindole acetate were able to activate PPAR␥ in adipocytes and macrophages. Indeed, 5-HT receptor-independent effects of 5-HT on glucose and lipid metabolism had been reported previously (59). 5-Methoxyindole acetate and 5-hydroxyindole acetate bound the PPAR␥ LBD in a distinct way because the PPAR␥ antagonist T0070907, which blocks binding of synthetic TZD drugs, had little effect on the activation of PPAR␥ by these 5-HT metabolites. Supplementing THP-1 macrophages with 5-HT induced PPAR␥ target gene expression, and the pharmacologic inhibition of 5-HT-metabolizing enzymes, such as the monoamine oxidase, prevented this induction. Because the highest peripheral levels of 5-HT are reached in the gastrointestinal tract, the authors argued that these metabolites interconnect the previously shown positive effects of PPAR␥ activation on experimental colitis (60) and that macrophages transport these compounds to PPAR␥-expressing target tissues (58). Additional studies will reveal more PPAR␥-dependent physiology of the 5-HT metabolites and how they relate to other endogenous ligands.
Tsukahara et al. (61) identified the phospholipid cyclic phosphatidic acid (CPA) as a potent (binding constant of ϳ125 nM) PPAR␥ antagonist, unlike the PPAR␥-activating property of other endogenous ligands. This discovery challenges the paradigm of endogenous ligands as positive regulators and is also surprising in regard to its structure because other lysophospholipids act as PPAR␥ agonists (62,63). However, the authors demonstrated that CPA is generated endogenously by phospholipase D 2 , induces association of PPAR␥ with NR corepressor (NCoR) 2 due to its cyclic phosphate moiety, and represses PPAR␥-dependent adipocyte differentiation and lipid accumulation in macrophages (61). This finding raises the intriguing question as to which, if any, of the above-described endogenous activators of PPAR␥ are physiologically antagonized by CPA. It also implicates C 18 carboxylic acid-containing phospholipids once more as ligands for PPARs. An overview of the described pathways leading to endogenous ligand generation for PPAR␥ is shown in Fig. 1B.

Rev-erb␣
The NR Rev-erb␣ is most abundantly expressed in metabolic tissues and is a negative feedback regulator of the circadian clock (64 -66). Mice deficient in Rev-erb␣ display changes in their circadian activity (65). Rev-erb␣ constitutively represses transcription because it lacks the C-terminal helix that is present in other NR called H12, a domain responsible for coactivator binding (67). Rev-erb␣ binds as a monomer to target promoters and recruits repression complexes containing the NCoR and HDAC3 (histone deacetylase 3) (67,68).
Rev-Erb␣ was originally cloned as an orphan NR (69,70). In 2005, the Rev-erb␣ homolog E75 from Drosophila was shown to contain protoporphyrin IX coordinated to iron (which is commonly referred to as heme) in its LBD (71). Soon after, two groups independently found that heme also binds human Rev-erb␣ and Rev-erb␤ (72,73). Heme was shown to bind reversibly at a 1:1 stoichiometry, enhance the thermal stability of the protein, and reversibly induce recruitment of the NCoR and HDAC3 by Rev-erb␣, thus fulfilling all criteria for an NR ligand (Fig. 2, left). Furthermore, heme depletion or addition of the synthetic analog hemin controlled the expression of the Rev-erb␣ target gene BMAL1 (72,73). Heme also mediated the sensitivity of the transcriptional activity of Rev-erb␤ to carbon monoxide and nitric oxide (74); whether this is true for Rev-erb␣ is not clear at this point.
Although the identification of heme as a Rev-erb ligand came as a surprise, heme itself had been already implicated in circadian regulation. It was shown that heme, as a cofactor for the gas responsive-transcription factor NPAS2 (neuronal PAS domain-containing protein 2) (75), controls the expression of mammalian Period genes mPer1 and mPer2 (76). Also ALAS1 (aminolevulinate synthase 1), the rate-limiting enzyme in circadian heme biosynthesis, was shown to be under the control of NPAS2 (76). Moreover, cellular heme levels oscillated in a circadian manner in NIH3T3 cells (77). It should be pointed out that the interaction of heme with Rev-erb␣ as a saturable reversible ligand extends the role of heme beyond its function as a gas sensor and identifies an exciting link between the mammalian clock and the control of metabolism by this NR.
Functionally, besides the described roles in circadian regulation, heme-dependent repression of glucose-6-phosphatase by Rev-erb␣ implicated the newly identified ligand in gluconeogenesis (73). In addition, it was shown that heme controls its own biosynthesis also through Rev-erb␣. Binding of heme to Rev-erb␣ repressed the expression of PGC-1␣, a potent inducer of heme biosynthesis (78), which decreased cellular heme concentrations (Fig. 2, left) (79). This negative feedback is in contrast to the heme/NPAS2-mediated induction of ALAS1, and further studies are needed to consolidate these two mechanisms. Hemin has also been shown to be a positive regulator of adipocyte differentiation (80), which could be mediated by binding Rev-erb␣, a known regulator of adipogenesis (81,82). Other Rev-erb␣ functions, such as lipoprotein expression and bile acid metabolism (83), have not been linked to heme regulation so far.

ROR␣
ROR␣ was cloned in the early 1990s as the first member of the ROR subfamily and named because of its sequence similarities to the retinoic acid receptor and RXR (84,85). An excellent review on ROR␣ and ROR subtypes ␤ and ␥ was published recently (86).
ROR␣ is expressed rather ubiquitously, including cerebellar Purkinje cells, liver, thymus, skeletal muscle, skin, lung, adipose tissue, and kidney (86). An intragenic mutation of ROR␣ found in staggerer mice revealed the involvement of ROR␣ in cerebellar development and glucose and lipid metabolism (87,88). ROR␣ and ROR␥ were also shown to regulate the maturation of a specific lineage of helper T cells called Th17 cells (89,90).
ROR␣ binds DNA as a monomer on ROR-response elements, which consist of the core consensus half-site and an adjacent 5Ј A/T-rich sequence (91). This specific binding site can also be bound by Rev-erb␣ or Rev-erb␤ (67), which suggests that it could compete with ROR␣ for occupying these elements (Fig. 2). Interestingly, ROR␣ has been shown to be a constitutive activator of gene transcription, which is in contrast to Rev-erb as a repressor. This dynamic model of activation and repression by ROR and Rev-Erb has been demonstrated for BMAL1 expression (66,92). Consistent with the effects of Rev-erb␣ deficiency, staggerer mice also exhibit an abnormal circadian activity pattern (93).
The first potential endogenous ligand for ROR␣ was identified by a crystallographic study showing that ROR␣, when expressed in baculovirus-infected Sf9 insect cells, integrated cholesterol in its LBD (94). Follow-up studies extended the group of potential ligands to several position 7-, 24-, and 25substituted cholesterol derivatives and cholesterol sulfate (95,96). All these sterols were shown to be high-affinity ligands and modulate the interaction of ROR␣ with coactivators (97).
The model of ROR␣ as a cholesterol sensor is intriguing because this NR had been shown to control several genes involved in lipid metabolism (86). Furthermore, it suggested that ROR␣ may not be a constitutively active receptor but rather rendered active by binding of ubiquitously occurring cellular cholesterol and its derivatives. However, ROR␣ expressed in Escherichia coli, without any endogenous sterols present, still exhibited constitutive activity in binding coactivator peptides, suggesting an intrinsic activating property of this NR (86).
It should be pointed out that only a subset of these ligands was able to regulate ROR␣ target gene expression in a receptor-dependent manner (98). This observation indicates that although several different cholesterol-related ligands bind ROR␣ in vivo, only a few are able to modify the receptor's intrinsic activity. Those that bind under physiological conditions without affecting the transcriptional activity of ROR␣ are referred to as silent ligands, a concept evolving from ligands binding several NRs (86,99,100). Silent ligands, such as cholesterol or cholesterol sulfate in the case of ROR␣, may compete with endogenous agonists or inverse agonists, thus adding another level of regulation in the biology of NRs (86). Because the role of cholesterol derivatives in the functional aspects of ROR␣ is still vague, more studies are needed to fully understand their potential as endogenous ligands.

Conclusions
Some of the most effective therapeutic agents available today are derived from endogenous ligands of NRs, with the great example of anti-inflammatory corticosteroids (1). As we have described, many potential candidates have been identified as endogenous ligands for PPAR, Rev-erb, and related receptors. Further research will be required to put them in perspective with regard to their physiological importance and to each other.
A recurring observation is that the cellular metabolism of these ligands is regulated by the corresponding NR, establishing feedback and feed-forward loops to balance NR activation. This may provide clues for the search of physiologically relevant ligands regarding other orphan NRs. Another observation is that the integration of newly identified ligands into the previously known NR physiology is complex, even more with the identification of endogenous antagonists and silent ligands. Cell and animal models in which several different NR ligands are present and can be manipulated will have to be used to consolidate these individual studies.
A way to "dig deeper" for endogenous ligands of the remaining orphan NR should make use of more sophisticated techniques, such as the combination of NR immunoprecipitation from in vivo material and mass spectrometry or crystallographic studies. Once the identification has led to ligands that are generally agreed upon, new therapeutic opportunities that interfere with how NRs work will become evident. We therefore believe that NR research still harbors a great potential for future drug development targeting metabolic diseases and their circadian components.