Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Introduction

In living organisms, cells need to adapt their intracellular activities to environmental conditions in order to control development, homeostasis and metabolism. One family of proteins playing a key role in the cellular response to internal and external stimuli is the nuclear hormone receptor superfamily, which includes Retinoid X Receptors (RXRs). RXRs occupy a central position within the nuclear receptor (NR) superfamily, serving as obligatory partners for numerous other NRs. This property sets RXRs at the crossroads of multiple signaling pathways where NRs play a coordinating role. The functionality of NRs is facilitated by two conserved domains in the proteins: the central DNA-Binding Domain (DBD) and the C-terminal Ligand-Binding Domain (LBD). The DBD is required for NR binding to specific DNA sequences and is folded into two small structural motifs called “zinc fingers”. In the absence of a ligand, NRs physically interact with transcriptional corepressors, such as Nuclear receptor Co-Repressors 1 and Silencing Mediator of Retinoid acid and Thyroid hormone receptor (NCOR1 and SMRT) [26, 52], that trigger the repression of gene expression when they are bound to DNA. Upon ligand binding, the receptors undergo a conformational change that can influence their intracellular localization and/or their affinity for coregulator proteins, thus increasing the recruitment of transcriptional coactivators and the release of corepressors. These events promote interaction of NRs with the transcription initiation complex which leads to the transcription of target genes. In this way, NRs are able to modulate gene expression in response to a variety of stimuli.

History: RXRs—Central Players Within the NR Superfamily

First Evidence for a New Receptor for Retinoids

The cloning of the first NR, the glucocorticoid hormone receptor, dates back to almost three decades ago [46, 101, 136]. Thereafter, many other hormone-dependent transcription factors sharing similar structure and properties have progressively enriched this superfamily of proteins. Humans express 48 NR genes, some of them generating more than one isoform (Table 5.1). RXRs were first described in 1990 by Mangelsdorf et al. [93] as NRs able to respond to vitamin A metabolites with a peculiar ligand specificity for 9-cis retinoic acid (9cis-RA).

Table 5.1 The human 48 nuclear receptors, categorized according to their ligand binding properties
Full size table

Early work in this field showed that NRs, such as the estrogen receptor (ER), bind to target DNA in the form of homodimers. New evidence generated in the early 1990s suggested there might also be heterodimer interactions between different NRs [44]. In particular, it became evident that nuclear extracts from different cell types contained proteins that were necessary for NRs, including Retinoic Acid Receptors (RARs) [45], Thyroid hormone Receptors (TRs) [18, 103], and Vitamin D Receptors (VDRs) [86], to bind DNA with high affinity. After the identification and first characterization of RXRs, several groups independently demonstrated that RXRs were indeed the common missing factors that were forming heterodimeric complexes with RARs, TRs and VDRs [17, 70, 82, 95, 133, 144, 147].

The ability of RXRs to form heterodimers was determined using Electrophoretic Mobility Shift Assay (EMSA). This technique evaluates the ability of protein complexes to bind to specific DNA sequences in an in vitro context (Fig. 5.1). Continued reports showed that RXRs are also able to form homodimers [54, 147] as well as to heterodimerize with Chicken Ovalbumin Upstream Promoter Transcription Factors (COUP-TFs) [6, 69], Peroxisome Proliferator-Activated Receptors (PPARs) [5, 40, 64, 71], Farnesoid X Receptors (FXRs) [37], Liver X Receptors (LXRs) [138], Nerve Growth Factor-induced protein I-B (NGFI-B) and transcriptional inducible NUclear Receptor Related (Nurr1) [111]. Thus, RXRs appear to play a unique role in integrating the action of several NRs by the forming heterodimers with multiple partners. Other work has shown that RXRs can also form homodimers [54, 147].

Fig. 5.1
figure 1

Traditional techniques employed to evaluate RXR functionality. a In transactivation assays, the LBD of RXR or other NR is fused with the DBD of Gal4, a yeast-derived transcription factor. This construct is transfected in cells together with a vector expressing a reporter gene under the control of a series of Gal4 responsive elements (UAS). In the presence of the NR ligand, the transcriptional activity of Gal4 is enhanced and the expression reporter gene is increased. b In electrophoretic mobility shift assays (EMSA), a radiolabeled DNA probe is incubated with the NR of interest and loaded on a gel. If the NR binds to the DNA probe, the migration of the DNA is delayed and we observe the shift (RXR-bound probe). The formation of a heterodimer complex is also revealed (heterodimer-bound probe)

Full size image

Classification of Nuclear Receptors

To unify the nomenclature of all NRs, a classification according to their phylogeny was adopted in 1999 (Nuclear Receptors Nomenclature [29]). However other criteria, more related to their functionality, are often used to categorize NRs [4]. For instance, when NRs are ordered based on ligand binding properties, they fall into three groups (Table 5.1). The first class is composed of the “classic” hormone receptors, including ERs, Androgen Receptor (AR), Mineralocorticoid Receptor (MR), Glucocorticoid Receptor (GR), TRs, VDR and RARs. The receptors belonging to this group can bind DNA in form of homodimers, like the ER and the GR, or in form of heterodimers with RXR, like RARs, VDR and TRs. The peculiarity of these NRs is their ability to bind a narrow range of small molecules with very high affinity, thus mediating a corresponding endocrine function.

By contrast, receptors belonging to the so-called metabolic sensors bind, with relatively low affinity, a large number of molecules, which are often intermediate or final metabolites of different metabolic pathways. Within this group are PPARs, LXRs and FXRs, which are activated by a broad spectrum of fatty acids, cholesterol and bile acid derivatives, respectively. Due to the nature of their activators and due to their target genes, these receptors regulate energy metabolism at multiple levels, but also cell proliferation, cell differentiation and cell survival (reviewed in [20, 30, 35]). Other members of this group are the pregnane X receptor (PXR) and the constitutive androstane receptor (CAR), which both induce cellular pathways important for detoxification [58, 140]. Most of the metabolic sensors bind DNA in the form of heterodimers with RXRs.

A third group is composed of “orphan” receptors that so far have no identified ligand. A number of studies have proposed putative ligands for some of these receptors [19, 68, 78, 92], that, thereafter, are now referred as “adopted” NRs. In contrast, for some others, such as Nurr1, the lack of a ligand binding pocket large enough to accommodate a ligand suggests that such receptors may have no ligand at all [135]. The orphan NRs are the most varied in terms of DNA-binding properties: indeed, they can act in the form of homodimers, such as the Hepatocyte Nuclear Factor 4 (HNF4), heterodimers with RXR, like COUP-TFs, or monomers, such as NGFI-B. Furthermore, the atypical receptor, Small Heterodimer Partner (SHP), lacks a DBD and cannot bind DNA.

The central role of RXRs as heterodimerization partners for NRs belonging to all three classes sets them apart from the above-described functional classification scheme. Complete occupation of the ligand-binding pocket by high-affinity ligands is a typical feature of classic NRs, a property exhibited by RXR, which has a small ligand binding cavity (between 400 and 500 Å) [51] that tightly accommodates 9-cis RA. Nonetheless, it is not possible to strictly categorize the RXRs as classic NRs because a physiological role for 9-cis RA has yet to be confirmed in vivo, and other natural ligands have much lower affinities for the receptor (see “Relevance: Endogenous and Synthetic RXR Ligands”).

RXR Isotypes

In both humans and mice, there are three RXR isotypes, RXRα (NR2B1), RXRβ (NR2B2) and RXRγ (NR2B3), that are encoded by separate genes sharing strong sequence homology [82, 91]. Each gene can further give rise to different isoforms by means of alternative splicing and/or alternative promoter utilization. In mice, three RXRα isoforms (RXRα1, 2 and 3), differing in the N-terminal domain, have been functionally characterized [16], and a fourth possible isoform has been predicted. Four RXRβ and two RXRγ isoforms have been identified [80, 89].

RXR isotypes and isoforms have a specific tissue distribution with partially redundant functions. In mice, RXRα1 is most abundant in the liver, but is also highly expressed in kidney, spleen, placenta and epidermis. RXRα2 and 3 are found in adult testis [16]. RXRβ is ubiquitous, with higher levels present in the central nervous system. RXRγ1 is expressed in skeletal muscle, olfactory bulb and pituitary gland, while RXRγ2 appears in both cardiac and skeletal muscle [10]. In humans, the tissue expression profile is more ubiquitous for RXRα and RXRβ, while RXRγ is predominantly in pineal gland (http://biogps.org [139]).

Development of the Field: RXR Behavior as Partners of Different Functional Complexes

How RXR-Containing Heterodimers Bind DNA

Many factors contribute to determining the recognition of the response element by the NR dimer in a specific target promoter, including the properties of the DNA consensus motif, its polarity, the NR ligand, and the DNA itself. The ligand may alter the DNA binding properties of a given NR heterodimer [146], and reciprocally, the DNA may itself play a role in the allosteric regulation of NR activity [54, 114, 120, 146].

NRs bind to specific DNA sequences, called consensus motifs. The DNA sequence recognized by NR dimers consists of two copies of a derivative of the consensus hexamer AGGTCA. Orientation and spacing of these two motifs are important determinants of their specificity toward each receptor dimer. Most of the consensus sequences for homo- or heterodimers containing RXRs are direct repeats of the core motif, with one to five nucleotides spacing (DR1 to DR5). In particular, RXR-homodimers and PPAR/RXR heterodimers preferentially bind to DR1, RAR/RXR complexes can bind to DR2 or DR5, whereas VDR/RXR and TR/RXR favor binding to DR3 and DR4 respectively. However, the rule underlying the recognition of a response element by NR heterodimers is difficult to decipher, due to some heterogeneity. For instance, RAR/RXR was also observed to associate with DR1 motifs, in some specific promoter contexts [119]. TR/RXR [34] and FXR/RXR [75] may also bind to inverted repeats (IR).

Another crucial feature of the heterodimer interaction with DNA is its polarity (see Chap. 3 and [77]). DR motifs are asymmetrical and can be read correctly only from one direction [90, 104]. This property provides a further level of specificity as gene regulatory responses may differ depending on whether RXRs occupy the 5′ upstream or 3′ downstream half-site [71, 73, 74, 94]. For instance, when complexed with RARs, RXRs always occupy the 3′ half-site in a DR1 but the 5′ half-site in a DR5. This difference in position influences the ability of the NR complex to be activated by the ligand. In contrast, PPARs always binds to the 5′ extended half-site of the DR1, while RXRs occupy the 3′ half-site [53]. This was confirmed by the crystal structure of the heterodimer PPARγ/RXRα complexed with its DNA response element [25].

Many homo/heterodimers compete for the same DR elements (for example, PPAR/RXR, HNF4/HNF4, and RXR/RXR bind to DR1; TR/RXR and LXR/RXR bind to DR4). This may have a strong functional impact depending on the protein abundance of RXRs and/or their partners and the heterodimer associating with DNA. As a consequence, the resulting expression of target genes may be significantly different. Alternately and not exclusively, the importance of promoter context may be a key determinant.

Structural Basis of Hetero- and Homo-Dimer Formation

As described, RXR is a peculiar member of the NR family since it can form homodimers or heterodimers with many different NRs. The first crystal structure of a RXR-dimer to be obtained was the RXR-LBD homodimer [12]. The analysis of its structure showed that both DBD and LBD are involved in dimerization. However, the LBD is more crucial due to a larger and stronger dimerization interface that is composed of amino acids arranged in a hydrophobic cluster and surrounded by polar residues. For example, the short region of RXRα between amino acids 387 and 429, called the I-box, has been shown to be required for heterodimerization [112]. A similar I-box, with a number of highly conserved amino acids, is also present in all RXR heterodimerization partners. Other crystal structures of RXR/partner LBD have been resolved, including RAR/RXR [13, 113], CAR/RXR [126, 142], LXR/RXR [59, 127] and PPAR/RXR [39, 47, 141]. The overall structure of these different heterodimers is highly similar, but within the molecules, there is variability between the different heterodimer interfaces, suggesting that some partners have a higher affinity for RXR than others.

More recently, thanks to tremendous progress in development of the techniques used to analyze protein structure and protein-protein interactions [i.e. Cryo Electron Microscopy (EM), nuclear magnetic resonance (NMR), small angle X-ray and neutron scattering (SAXS and SANS) and FRET (fluorescence resonance energy transfer)], the structures of intact heterodimers bound to DNA in their native environment have been reported (see Chaps. 2 and 3 and reviewed in [15]).

In 2008, the crystal structure of the full-length PPARγ /RXRα complex bound to ligands, coactivator peptides, and DNA was resolved, fully highlighting the domain–domain interactions involving these intact NRs. The heterodimer formed by PPARγ and RXRα is asymmetric, allowing the PPARγ-LBD to lie at the center of the complex and to make direct contacts with multiple domains in both proteins. Three distinct heterodimerization interfaces are formed between RXRα and PPARγ and, interestingly, some of them are DNA-dependent [25].

A different conformation was observed by Rochel et al. [117], who noticed that, in solution, the structure of several heterodimer complexes bound to DNA (VDR/RXR bound to a DR3, RAR/RXR to a DR5, RAR/RXR to a DR1 and PPAR/RXR to a DR1) is characterized by an asymmetric open conformation with separate DBDs and LBDs, connected by extended hinges. Subsequent analyses of the full VDR/RXR visualized the hinge domains of both receptors that stabilize the complex in a precise conformation [109]. Besides these physical interactions, an extensive communication throughout the heterodimer, influenced by the recognized sequence of DNA, has been documented by Hydrogen/Deuterium Exchange (HDX) experiments [146].

Based on these results, the currently accepted model for heterodimerization is that, first, RXR and the partner NR form a complex through their dimerization interfaces, and second, some remodeling of the formed heterodimer occurs once the complex recognizes the response element.

Functional Behavior of RXR-Containing Complexes

In the last twenty years, a great effort has been made to understand the behavior of the different RXR-containing complexes, namely homotetramers, homodimers and heterodimers, and the specific role of RXR proteins within the heterodimer. Unexpectedly considering the behavior of most members of the nuclear hormone receptor family, RXRs are able to self-associate in solution and form transcriptionally inactive homotetramers [38, 65, 66]. Upon ligand binding, these tetramers dissociate, allowing the formation of homo- and heterodimers, exhibiting transcriptional activity [28]. The function of RXR homotetramers is not completely understood. A first hypothesis regarding the role of these complexes was that they could act as structural elements able to fold the DNA, in order to approach distant RXR response elements [143]. Subsequently, it was proposed that homotetramers could function as a way for RXR molecules to be stored and then be mobilized for formation of heterodimers, depending on the ligand availability for the partner [36, 131]. No further evidence has been reported and we are left with only hypotheses.

Besides aggregating as homotetramers, RXR can homodimerize. 9-cis RA induces RXR homodimer formation in vitro, suggesting the existence of a retinoid response pathway distinct from that activated by the heterodimer RAR/RXR [147]. In cells and in animals treated with 9-cis RA, the actual presence of RXR homodimers on specific DNA response elements was subsequently demonstrated by in vivo chromatin immunoprecipitation (ChIP) studies that are extremely powerful in indicating whether a protein is recruited to specific DNA binding sites in a given cell [54] (the principles of the technique are presented in Fig. 5.2). Additionally, analysis of fluorescence fluctuation brightness in living cells expressing RXRs labeled with molecules that emit fluorescence showed that homodimerization is triggered by 9-cis RA binding to RXR [27] (Fig. 5.3). This event induces the recruitment of coactivators, such as Steroid Receptor Coactivator 1 (SRC-1) and Transcriptional Intermediary Factor 2 (TIF2), thus stabilizing the association of the complex to DR1 elements on the DNA. However and as mentioned, all these experimental studies were performed in the presence of 9-cis RA. Thus, the physiological role of RXR homodimers, which seems to highly depend on the presence of RXR ligand, will remain an open question until the existence and nature of endogenous ligands for RXR is revealed [54].

Fig. 5.2
figure 2

Chromatin Immunoprecipitation (ChIP). In ChIP assays, NR proteins are cross-linked with formaldehyde to regions of DNA that have recognition sites (red) for the proteins (NR). In the following step, DNA is fragmented in short pieces (300–1,000 bases). Fragments of DNA that are associated with the protein of interest are then captured using an antibody (blue inverted Y) that specifically recognizes the NR. In the last step, cross-linking is reversed in the captured DNA-NR fragments: the DNA is detached from the NR and the captured DNA fragments are identified by amplifing in a PCR or by DNA sequencing

Full size image
Fig. 5.3
figure 3

Fluorescence resonance energy transfer (FRET). FRET assays evaluate the interaction between two proteins (i.e. heterodimer partners). The two NR are linked to two different fluorophores (F1 and F2). If the two partners do not interact, the excitation of the F1 fluorophore at the appropriate wavelength (λ1) results in energy emission (λ2). If the two partners interact, the interaction brings the fluorophores in tight proximity, and, in this case, excitation of the F1 fluorophore results in an energy transfer to the F2 fluorophore, which gets excited and emits energy at wavelength λ3

Full size image

First attempts to understand the functional consequences of heterodimerization were based mostly on transient transfection experiments in which RXR proteins and their partners were overexpressed and activated by their respective ligands (Fig. 5.1a). These studies highlighted two categories of functionally distinct RXR partners, so-called permissive and non-permissive partners. The activity of a non-permissive NR partner is enhanced by its own agonist, but is unaffected by the RXR-agonist. Non-permissive partners include hormone receptors with a high affinity for their ligands, such as TRs, RARs and VDR. In contrast, permissive RXR heterodimers can be activated by either an RXR and/or the NR partner agonist. Permissive partners are typically metabolic sensors like PPARs, LXRs, FXRs, PXR and CAR, that sense mainly small lipid molecules with low affinity. One attractive hypothesis to explain the appearance during evolution of these two classes of receptors is that non-permissive RXR partners diverged from permissive receptors as they acquired the ability to recognize high-affinity hormonal ligands. In parallel, the amino acid sequence of permissive partners evolved to allow them to communicate trough allosteric mechanisms with RXR. Thus, the evolutionary process allowed NRs to differentially couple metabolic and endocrine regulation [123].

The molecular mechanisms underlying the allosteric coupling of partners in permissive heterodimers are still poorly understood. A crucial element underlying the functional synergy of activation could reside in the recruitment of coactivators. Several reports indicate that, upon ligand binding, a single coactivator molecule binds to receptor heterodimers [84]. Each heterodimer partner can be autonomously anchored to one of the LXXLL signature motifs that are present in the coactivator [50, 137]. In the presence of both ligands, synergy would originate from the cooperative binding of one coactivator molecule with the two heterodimer partners, through the simultaneous establishment of two NR-coactivator interfaces [41]. For permissive heterodimers, such as LXR/RXR, the existence of allosteric interactions, which can influence the activity of the protein complex in response to the binding of the RXR ligand, was demonstrated by recent structural data [124]. The RXR region mediating these interactions can be different depending on the partner. As an example, the RXR-AF2 region is necessary for RXR-mediated activation of PPARβ, but not of PPARγ [21]. Such allosteric interactions seem to be excluded within the complex RAR/RXR, in agreement with the non-permissiveness of this heterodimer [113]. In contrast, and quite unexpectedly, the non-permissive partner TR is needed to specifically activate the prolactin promoter in response to 9-cis RA in cells [24]. Taken together, these observations suggest that the mechanisms underlying the molecular crosstalk between the partners are strongly specific for each heterodimer and, for a given heterodimer, these mechanisms may differ depending on the evaluated target gene.

Therefore, to deeply understand RXR biology, it will be fundamentally important to couple the innovative techniques used to analyze protein structure and interactions with new genome-wide approaches as this will allow comprehensive evaluation of RXR binding and its effects on the expression of all target genes. Noteworthy, the physiological relevance of the distinction between non-permissive and permissive partners may have to be revisited once the nature of the natural ligand for RXR in vivo is elucidated.

Regulation of RXR Activity Through Post-Translational Modifications

The dynamic assembly/disassembly of retinoid receptors to promoters of target genes is regulated by proteasome-mediated degradation. Both partners within DNA-bound RXRα/RAR heterodimers can be degraded in response to retinoids acting either on RXR or on RAR (9-cis RA or derivatives and all-trans-retinoic acid) [11, 42, 43, 72, 130]. The impact of the ubiquitin-proteosome machinery on heterodimer activity appears to be isoform specific. It has been shown in different cell lines and adipose tissue that RXRα is degraded by this process, but RXRβ is not [79].

Proteasome-mediated degradation of RXRα is linked to its phosphorylation at specific serine residues by mitogen-activated kinases (MAPK) and/or protein kinase C (PKC) that can interfere with its function [1, 43, 97]. These post-translational modifications seem to be important for the potential involvement of RXRs in cell proliferation and in cancer development. RXR in tumors can be degraded by cathepsin L and calpain producing a fragment of 44KDa which goes to the cytosol and activates a cancer cell survival pathway [129, 148]. Thus, RXRα may also have extra-nuclear action and mechanisms governing its cellular localization could control its transcriptional activity, further complicating RXR biology.

Current State of the Field: Addressing RXR Functions on a Global Level

Targeted Gene Disruption: RXR-Null Mice

The generation of mice carrying RXR gene deletions or mutations has been a valuable tool to investigate the impact that RXR has on the biological functions of other NRs and the specific role, if any, of each RXR isotype. As expected, phenotypic characterization of the mutant mice demonstrated that RXR is involved in a plethora of developmental and physiological pathways. The analysis of RXRβ mutants revealed a reduction of spermatid formation in the mice [63]. Specific ablation of RXRγ impaired cholinergic responses in nigrostriatal pathways [118] and caused important metabolic effects, including an increased metabolic rate with an accompanying resistance to gain fat mass in response to high-fat feeding [49].

Whereas the RXRβ and RXRγ null mice had only mild developmental defects, ubiquitous inactivation of RXRα resulted in embryonic lethality at mid-gestation due to hypoplastic development of the ventricular myocardium [62, 125]. To bypass this lethality, several conditional, tissue-specific, RXRα knockout mice have been generated. Conditional hepatic ablation of RXRα results in alteration of fatty acid oxidation and hepatocyte lifespan suggesting that the absence of RXRα in the liver cannot be compensated for by RXRβ or RXRγ. This supports the hypothesis that each isotype may have specific functions [56, 134]. Tissue-specific deletion of RXRα in adipose tissue resulted in resistance to diet-induced obesity in these mice, due to impaired adipocyte differentiation, likely reflecting an altered PPARγ function [55].

More recently, characterization of mice lacking RXRα in myeloid cells has demonstrated an important role of RXRα in the innate immune response to inflammatory stimuli. Interestingly, this function seems to be mediated by RXR homodimers, evidence that supports the existence of an RXR signaling pathway in vivo [108].

In most cases, the phenotypes observed in RXR mutant mice can be related to alterations in pathways regulated by other NRs (reviewed in [31]). It remains difficult to unambiguously define the functions of individual heterodimers in regulating signaling pathways that influence the expression of specific genes via direct and/or indirect regulatory loops. Thus, much is left to be learned about the functional roles of RXRs in vivo. Besides a thorough investigation using tissue-specific knockout models, the systematic analysis of transcriptional programs regulated by agonists for RXRs or their partner is essential to bring new insight into the complex field of RXR biology.

Global Analyses of RXR-Dependent Transcriptional Programs

With the rapid development of quantitative PCR and microarray technologies, the evaluation of how the expression of large gene sets is affected by different experimental conditions has become routine. These approaches are suitable in order to identify potential target genes of transcription factors, including NRs. The application of such techniques has elucidated tissue- and gene-specific differences in the permissiveness of RXR-heterodimers, by using synthetic ligands.

One study performed in rats showed that, in adipose tissue, the effect of the RXR-specific agonist LG268 was not the same as that of the PPARγ-specific ligand rosiglitazone, suggesting that in this tissue, the PPARγ/RXR heterodimer is not permissive [2]. Differently, in mouse liver, the rexinoid bexarotene induced the expression of LXR-target genes that participate in lipogenesis, but did not affect those involved in cholesterol homeostasis. The latter findings support the concept that the permissivity of LXR/RXR heterodimers also depends on the target gene [76].

A systematic analysis of the different transcriptional programs induced by permissive and non-permissive heterodimers was performed in monocyte-derived dendritic cells, as a case study [128]. To comprehensively address all aspects of RXR signaling, the global gene expression profile was evaluated in cells treated with RXR ligands (LG628 and 9-cis RA), ligands for the non-permissive partners RAR and VDR (AM580 and 1,25-vitamin D, respectively), or ligands for the permissive partners PPARγ, PPARβ/δ and LXRs (rosiglitazone, GW1516 and GW3965, respectively). The results showed that RXR ligands regulated only a few of the target genes of the non-permissive partners, while most of the LXR and PPAR responsive genes were also induced by RXR-agonists, albeit to a lesser degree than was observed with LXR and PPAR agonists. While confirming the different functional behaviors of permissive and non-permissive heterodimers, this report also emphasized the fact that RXR-mediated activation on permissive heterodimers may have specific outcomes. For example, the combination of RXR and PPARγ agonists had a differential effect depending on the PPARγ targets: either synergism (most of the PPARγ/RXR responsive genes) or negative interaction, resulting in a reduced induction compared to that triggered by the PPARγ ligand alone (for example, the Fatty acid binding protein 4 (FABP4) gene), were observed [128]. Similar observations were made for other permissive heterodimers, such as FXR/RXR [61], strongly indicating that the heterodimer activation state is strongly dependent not only on the tissue, but also on the gene.

The molecular mechanisms underlying this differential tissue- and gene-dependent regulation are far from being understood. Presently, the most likely hypothesis is that cofactor availability in different cell types could be the determinant in such activities. The observation that the coactivator recruited on PPARγ/RXR heterodimers is different if induced by RXR or PPAR agonists [84] supports this hypothesis. This suggests that the amount of each cofactor in given cells and the nature of the available ligand influence RXR effects on target gene activation and expression.

Taken together, these observations underscore the existence of very intricate RXR signaling mechanisms. Further, this highlights the importance of performing analyses in different tissues or cell-types to fully decipher RXR behavior, especially in view of potential therapeutic uses of RXR-specific activators.

RXR Cistrome

In recent years, key technological advancements have allowed chromatin immune-precipitation to be coupled, initially with tiled oligonucleotide microarray (ChIP-chip) and later on with ultra high-throughput sequencing (ChIP-seq), providing a valuable tool to identify the NR cistrome, namely all DNA binding sites (referred to as cis as these sites are located on the DNA strand) recognized by a particular DNA-binding factor or complex (referred to as trans-acting factors), analyzed at the whole genome level. The first NR whose DNA binding sites were characterized genome-wide was the ER [22]. Thereafter, dozens of studies reported the cistrome of NRs in different cell lines or tissues. Currently, the landscape of RXR binding sites has been elucidated in combination with PPARγ during adipogenesis [106], with VDR in osteoblasts [99], with the oncofusion protein PML-RARα in an acute promyelitic leukemia cell line [96], with LXRβ in human keratinocytes [121] and with PPARα and LXR in the liver [9]. Most of these studies were primarily aimed at gaining a better understanding of the biology of the partner NR, rather than RXR, but nonetheless, they did highlight several features that are relevant to RXR, both as a NR and as a common heterodimerization partner.

One surprising finding that came out of analyzing the RXR-heterodimer cistrome was that the number of DNA binding sites was one, or even two, orders of magnitude greater than the number of effective target genes (Fig. 5.4). This difference can be partially explained by an underestimation of the ligand-regulated gene number due to confounding effects, such as RNA abundance and stability, which can influence the sensitivity of microarrays. However, the disparity is pronounced and raises the question of the functional significance of RXR, and more generally, NR, binding to these numerous binding sites that are apparently not related to transcriptional regulation. Unexpectedly, the analysis of NR binding sites localization showed that most of them (>60 %) are located in distal intergenic regions or within introns, whereas only a few sites are found in the proximal promoter region. This observation correlates well with the fact that only 16 % of the chromatin regions that are open and more accessible to transcription factors are located in promoter regions, as demonstrated by DNAse I hypersensitivity assays coupled to DNA sequencing [14]. Accordingly, the identified binding sites for RXR and LXR in the liver, in the absence of specific ligands, are located within open chromatin regions, whereas treatment with ligands enables LXR and RXR to bind also to less accessible sites [9].

Fig. 5.4
figure 4

The number of NR binding sites to DNA is one or two orders of magnitude higher than the number of their target genes. Upon ligand treatment, ChIP on chip or ChIP-seq experiments were performed to identify genome-wide DNA binding sites of RXR with VDR in murine osteoblasts, with LXR in mouse liver, and with PPARγ in murine adipocytes. Compared to a set of thousands of putative target genes that were proposed for each heterodimer based on several criteria, including proximity of the RXR/partner binding site to the transcription start site (TSS), the global gene expression analysis indicated that only hundreds of genes (effective target genes) were modulated by the treatment

Full size image

As expected, RXR binding sites are more numerous than those for a particular partner. Intriguingly, RXR binding to sites recognized by one heterodimer is not strictly dependent on the presence of the partner. For instance, many binding sites that are co-occupied by RXR and PPARγ in mature adipocytes, are already bound by RXR in preadipocytes, possibly in the form of a homodimer or as a heterodimer with PPARβ [106] or another unidentified partner. Similarly, VDR binding sites pre-marked by the presence of RXR have been identified in osteoblasts [99].

The constant recruitment of RXR on a number of DNA binding sites that are shared with other NRs was observed also in mouse liver [9]. The characterization of RXR binding sites in LXR double knockout mice livers reveals that, despite LXR absence, RXR is still binding to the majority of the shared LXR/RXR binding sites identified in the wild type mice. Interestingly, when a motif search was performed on sequences under the LXR/RXR peaks, it turned out that they were not only enriched in DR4 motifs, which are the canonical binding sites for LXR/RXR and TR/RXR heterodimers, but they were also enriched in other DRs, Inverted Repeats (IR1) and Everted Repeats (ER2), which suggests that different homo/heterodimers may bind to these regions. Indeed, a sub-group of these sites can be bound also by PPARα, HNF4α, FXR, and Reverbα, suggesting the existence of NR “hot spot” sequences, to which multiple NRs bind depending on the precise signaling acting within the cell at a certain moment. Notably, while LXR binds to sequences containing DR1 elements in vivo, it fails to bind the same sequences in vitro, indicating the chromatin context plays a major role in determining the recognition of a certain response element. The functional outcome of this extensive crosstalk appears to also be determined by the context of the single genomic binding sites. The observation that a very low percentage of RXR binding sites are overlapping in different tissues (i.e. only 12 % of RXR binding sites are common between adipocytes and liver) further supports the importance of the chromatin context in determining NR recruitment on specific DNA regions. However, the molecular mechanisms underlying the promiscuity and the regulation of the activity of individual receptors at particular sites are not yet fully understood.

The global maps of RXR binding sites obtained to date have been valuable tools for identifying novel transcription factor binding sites. With respect to RXR biology, the comparison of different global genome-wide binding profiles has clarified genomic regions occupied by different partners. Future work will reveal new levels of regulation by individual heterodimers that are strictly dependent on the chromatin context.

Relevance: Endogenous and Synthetic RXR Ligands

Given the widespread relevance of the superfamily of NRs to multiple aspects of human physiology and etiopathogenesis and the role of RXR as partner for many of these receptors, a detailed characterization of RXR ligands and their effects has major implications for the understanding of retinoid biology, as well as for the development of new potential drug treatments.

The physiological ligand for RXR is still unknown. A study performed in mouse epidermal keratinocytes demonstrated the concomitant presence of a transcriptional repressor complex composed of RXRα heterodimerized with unliganded RARγ and an active heterodimer PPARβ(δ)/RXRα. The induced expression of the target gene 3-Hydroxy-3-Methylglutaryl-CoA Synthase 2 (hmgcs2) indicates that the RXR agonist must have been present in these cells [21]. These results imply that the ligand for RXR must be different from the ligand for RAR. This excludes 9-cis RA from the possible candidates since this molecule also activates RAR.

More recently, other vitamin A metabolites acting as ligand for RXRs have been detected in vivo, such as dihydroretinoids [102], all-trans-retinaldehyde [149] and b-apo-14′-carotenal [150]. A group of flexible unsaturated fatty acids, either endogenously produced or derived from the diet, can act as RXR ligand. Two monocyclic terpenoid compounds, methoprenic acid and phytanic acid have been shown to be highly selective RXR ligands, albeit at much higher concentrations than 9-cis-RA [48, 67, 83]. In addition, the structural analysis of RXR-RAR LBD led to the identification of an oleic acid molecule into the RXR ligand binding pocket [13]. Docohexanoic acid (DHA) can also be accommodated in the LBD of RXRs and activate the receptors, similarly to other n-3 PUFA, such as docosatetraenoic acid (C22: 4 all-cis-Δ7,10,13,16) or arachidonic acid (C20: 4 all-cis-Δ5,8,11,14) [33, 132]. These findings suggest that RXRs, besides regulating metabolic homeostasis in association with their heterodimerization partners, can also directly act as intracellular sensors.

Besides natural ligands, many synthetic compounds target RXRs (reviewed in [23, 110]. Several environmental pollutants, such as the organotin tributyltin chloride (TBT), activate RXRs and other NRs, thus interfering with the physiological activity of these receptors [60].

An interest in RXR ligands as potential pharmacological treatments arose some time ago from the observation of their strong apoptotic effect [98, 105]. Accordingly, an RXR selective ligand, bexarotene (LGD1069, Targretin®) [8], has been successfully used in the therapy of refractory or persistent cutaneous T-cell lymphoma. Preclinical studies have shown that rexinoids are also promising drugs for acute promyelocytic and myeloid leukemias [3, 145], and clinical trials have demonstrated their efficacy in the treatment of non-small-cell lung cancer [7, 115], as well as in the prevention of both lung and mammary tumors [85, 87, 88].

Given their pro-apoptotic effects and their multi-partnerships with NRs that act as metabolic regulators, rexinoids are clearly molecules with potential to exert broad impact on metabolic regulation. LG1069 administration to patients results in hypertriglyceridemia, hypercholesterolemia and hypothyroidism [32, 116]. The characterization of rexinoid effects in vivo showed that they mimic PPARγ ligands such as thiazolidinediones (TZDs), improving insulin sensitivity. However, while TZD acts primarily at the level of adipose tissue, RXR ligands exert their activity mainly at hepatic and muscular levels [2, 122], suggesting that PPARα, PPARβ(δ) and LXRs are likely the partners accounting for rexinoid action.

The multiple metabolic effects of rexinoids highlight their potential as drugs for complex disorders, such as the metabolic syndrome. The striking gain in knowledge on how ligand binding influences RXR structure and the resulting ability of the complex to interact with cofactor proteins has made it possible to start designing rexinoids with heterodimer selective activities, such as selective NR modulators (SNuRM). Several reports describe rexinoids that are able to activate different combinations of heterodimers including HX630 (selective for PPARγ/RXR), PA024 (acting on both PPARγ/RXR and LXR/RXR) [107], and LG101506 (selectively activates PPARγ/RXR and not LXR/RXR or RAR/RXR) [81]. The distinct behavior is reflected in the different patterns of gene expression induced by HX630 and PA024 when used in combination with an RAR agonist [57]. Rats treated with LG101506, which activates only PPARγ/RXR, have improved insulin sensitivity but unchanged triglyceride levels, likely due to a lack of LXR/RXR activation [100].

Future Directions

RXRs can form heterodimers with a number of NRs involved in the regulation of multiple cellular functions. This peculiar ability represents a formidable example of how activity of one single receptor can broadly modulate a very complex regulatory network. In recent years, striking technological advancements have made it possible to improve our understanding of RXR biology.

Significant steps forward have been made in our understanding of how changes in RXR three-dimensional structure are induced by ligands and in how the DNA consensus sequence affects the ability of the receptor to interact with cofactor proteins. In parallel, the possibility to globally address the transcriptional effects of RXR activation, together with the genome-wide identification of DNA binding sites of RXRs and their partners, have highlighted an extremely intricate regulatory network in which multiple factors, such as cofactor availability, competition among possible partners, and ligand presence, play a role. All of this has led to an appreciation of the increasing levels of complexity that underly RXR biology.

There is a need for more in depth studies of the individual signaling pathways that are specifically activated by each RXR /partner NR heterodimer and/or single RXR isoforms. New approaches and new concepts, particularly based on the development of bioinformatics and modeling, are also needed to push the field forward and foster a way to grasp and control the inherent levels of complexity.

Such global analyses are essential in order to understand to what extent the unique properties of RXR signaling network can be exploited as a new therapeutic target for the treatment of diseases and metabolic disorders. Studies have paved the way for the design of pharmaceutical rexinoids with selective activities, and clinical trials have indicated their potential efficacy. But at the same time, the difficulty to keep in vivo responses to such promiscuous receptors under control speaks to a pressing need for additional research aimed at producing a deeper understanding of their direct and indirect actions.