Visualizing the Architectures and Interactions of Nuclear Receptors

Abstract

Nuclear receptors (NRs) are master regulators of broad genetic programs in metazoans. These programs are regulated in part by the small-molecule ligands that bind NRs and modulate their interactions with transcriptional coregulatory factors. X-ray crystallography is now delivering more complete pictures of how the multidomain architectures of NR homo- and heterodimers are physically arranged on their DNA elements and how ligands and coactivator peptides act through these complexes. Complementary studies are also pointing to a variety of novel mechanisms by which NRs access their DNA-response elements within chromatin. Here, we review the new structural advances together with proteomic discoveries that shape our understanding of how NRs form a variety of functional interactions with collaborating factors in chromatin.

The nuclear receptor (NR) transcription factor family consists of evolutionarily related members that can be regulated by binding to naturally occurring or synthetic small molecules as well as to DNA. A total of 48 receptors exist in humans, and these proteins control a broad array of genetic programs responsible for embryonic development, reproduction, immune function and metabolic homeostasis (1). By binding directly to fat-soluble hormones, vitamins, dietary lipids, heme and xenobiotic compounds, NRs can regulate gene expression programs in a variety of cell types (2–4). The binding of ligand or DNA produces conformational changes that are sensed across surfaces of NR polypeptides. These changes impart significant consequences in NR intermolecular interactions that ultimately culminate transactivation or transrepression of target genes (5–7). With their abilities to control gene programs in direct response to small-molecule ligands, many NRs have been actively explored as therapeutic drug targets.

At the molecular structure level, NR-mediated transcriptional function involves small-molecule binding at the ligand-binding domain (LBD), and the separate interactions of the DNA-binding domain (DBD) with response elements. Abundant structural studies on these isolated domains have provided fundamental insights about functional consequences of ligands and coregulator peptides binding to the LBD as well as DNA-response element binding to the DBD (4, 8). A more integrated picture of how all of the domains in NRs can be architecturally arranged in their active form on DNA has been generated only in a few cases (9–12). Advances in structural understanding of full-length NRs are providing new glimpses and inferences about domain-domain coupling and allosteric regulations. These are likely initial snapshots of a far more complex picture of interacting components that are required for NRs functions in the genome.

At the same time, our knowledge about NR-interacting partners in the epigenomic landscape remains incomplete, because they associate with functionally diverse regions in coding and noncoding DNA. NR DNA accessibility is highly regulated in the chromatin environment and varies between cells or during developmental stages (13, 14). Chromatin architecture, the packaging of DNA in nucleosomes, and composition orchestrate far fewer than predicted motifs to become accessible. For example, the mouse genome contains over 2 million predicted binding motifs for the glucocorticoid receptor (GR), but GR only targets 3000–8000 sites depending on cell type (15). Variability in NR ligands profoundly influences their genomic targeting (16). Access to chromatin hinges on NR cross talk with other transcription factors (14, 17, 18). In addition, a variety of posttranslational modifications (eg, Ref. 19) and numerous long noncoding RNA (lncRNA) molecules (20) regulate the interactions of NRs in chromatin.

Genome-wide studies are indicating that ligand binding leads to de novo NR recruitment to promoters and enhancers of target genes, but NRs bind to far more other sites for which the significance of targeting remains elusive. A majority of target sites have poor conservation of NR consensus elements. Furthermore, the binding interactions of NRs in the absence and presence of ligand have been found to be highly dynamic with little DNA footprints (21, 22). NRs also coordinate with diverse ATP-dependent chromatin remodeling enzymes to reach accessible DNA (14). Remarkably, NR cross talk with other transcription factors, some of which are called pioneer factors (eg, Forkhead box protein A1, [FoxA1]), influences DNA targeting in a reciprocal manner (21–23). Here, we discuss molecular observations that suggest how NR subdomains cooperate to selectively integrate ligand stimulation with DNA binding and coregulator interaction.

The Diversity of NR Functions and Ligands

Table 1 lists the mammalian NRs and their broad array of ligands. A number of NRs are considered orphan receptors as information about their ligands is still missing, or continues to evolve and requires further exploration (4, 5, 24). NRs fall into 7 (0–6) subfamilies according to sequence similarity; subfamily 0 corresponds to NRs that lack a DBD. About half of NRs form homodimers in their functional associations, whereas the other half function as heterodimers and integrate 2 distinct ligand signaling mechanisms. Retinoic X receptor (RXR) is a special member of the NR family that acts as a common heterodimeric partner to retinoic acid receptor (RAR), the thyroid hormone receptor (TR), vitamin D receptor (VDR), liver X receptor (LXR), peroxisome-proliferator-activated receptors (PPARs), farnesoid X receptor (FXR), pregnane X receptor (PXR), and constitutive androstane receptor (CAR) among others (25–30).

Table 1.

Mammalian Nuclear Receptors Are Listed According to Nomenclature, Ligands, and Functional States

Subfamily	Receptor	Symbol	Ligand Type	Functional Association
1	TRα	NR1A1	Thyroid hormone	Heterodimer/monomer
	TRβ	NR1A2
	RARα	NR1B1	All-trans-Retinoic acid	Heterodimer
	RARβ	NR1B2
	RARγ	NR1B3
	PPARα	NR1C1	Polyunsaturated fatty acids, prostaglandins	Heterodimer
	PPARβ/δ	NR1C2
	PPARγ	NR1C3
	Rev-Erbα	NR1D1	Heme	Monomer/homodimer
	Rev-Erbβ	NR1D2
	RORα	NR1F1	Oxysterols, cholesterol intermediate products, all-trans-retinoic acid	Monomer
	RORβ	NR1F2
	RORγ	NR1F3
	LXRβ	NR1H2	Oxysterols	Heterodimer
	LXRα	NR1H3
	FXR	NR1H4	Bile acids	Heterodimer
	VDR	NR1I1	1,25-dihydroxy vitamin D3	Heterodimer
	PXR	NR1I2	Xenobiotics	Heterodimer
	CAR	NR1I3	Androstenol	Heterodimer
2	HNF-4α	NR2A1	Fatty acids	Homodimer
	HNF-4γ	NR2A2
	RXRα	NR2B1	9-cis-retinoic acid	Heterodimer
	RXRβ	NR2B2
	RXRγ	NR2B3
	TR2	NR2C1	Unknown	Homodimer/heterodimer
	TR4	NR2C2
	TLX	NR2E1	Unknown	Monomer/homodimer
	PNR	NR2E3
	COUP-TF I	NR2F1	Unknown	Homodimer/heterodimer
	COUP-TF II	NR2F2
	EAR2	NR2F6
3	ERα	NR3A1	17β-estradiol	Homodimer
	ERβ	NR3A2
	ERRα	NR3B1	Unknown	Monomer/homodimer
	ERRβ	NR3B2
	ERRγ	NR3B3
	GR	NR3C1	Glucocorticoid	Homodimer
	MR	NR3C2	Aldosterone	Homodimer
	PR	NR3C3	Progesterone	Homodimer
	AR	NR3C4	Testosterone	Homodimer
4	NGFIB	NR4A1	Unknown	Monomer/homodimer/heterodimer
	NURR1	NR4A2
	NOR1	NR4A3
5	SF-1	NR5A1	Phosphatidylinositols	Monomer
	LRH-1	NR5A2	Phosphatidylinositols	Monomer
6	GCNF	NR6A1	Unknown	Homodimer
0	DAX-1	NR0B1	Unknown	Heterodimer
	SHP	NR0B2	Unknown	Heterodimer

Subfamily division is based on sequence similarity in DBD and LBD regions. A symbol corresponds to each NR followed by a 3-character code indicating subfamily number, group letter, and the specific gene number. AR, androgen receptor; DAX-1, Dosage-sensitive sex reversal-adrenal hypoplasia congenita critical region on the X chromosome, gene 1; LRH-1, liver receptor homolog 1; GCNF, germ cell nuclear factor; NGFIB, NR growth factor inducible-B; NURR1, Nur-related protein 1; NOR1, neuron-derived orphan receptor 1; PR, progesterone receptor; MR, mineralocorticoid receptor; PNR, photoreceptor-specific NR; SF-1, steroidogenic factor-1; SHP, small heterodimeric partner; TR2/4, testicular orphan receptor 2/4.

A variety of cellular functions are attributed to NRs, and some examples are mentioned here. GR, androgen receptor, estrogen receptor (ER), progesterone receptor, and mineralocorticoid receptor are essential for mammalian developmental and reproductive programs that respond to steroid molecules (31). RAR and RXR recognize ligands derived from the lipid-soluble vitamin A and are essential for differentiation and growth (27, 32). LXRs and FXRs regulate pathways in cholesterol and bile acid production and their transport (33–35). PPARs regulate carbohydrate and lipid metabolism in a number of metabolic organs (5). The Rev-Erbs regulate a broad program of lipid metabolism and energy use, also linking these programs to circadian timing (36–38). The estrogen-related receptors (ERRs) are critical for the control of energy use and mitochondrial bioenergetics program (39). Subfamily 4 members (NGFIB/NURR1/NOR1) also regulate energy metabolism, and contribute to development of dopaminergic neurons in the brain (40). Members of ROR subfamily regulate lymphoid organogenesis and the differentiation of thymocytes into proinflammatory T helper 17 cells (41–43). The xenobiotic sensors include PXR (44, 45) and CAR (46–48), which help defend against a broad variety of small-molecule environmental toxins.

The Basic Components of NRs, Response Elements, and Coregulators

NRs have similar modular architectures with distinct embedded domains in their polypeptides (8). Their N termini contain divergent segments that can harbor an activation function (AF)1. A highly conserved DBD allows their binding to hexameric DNA half-sites present as direct repeats (DRs) or inverted repeats (IRs) with distinct number of spacing bases (DR1 corresponds to DRs with a single nucleotide in between). A structurally conserved LBD has a pocket for small molecules and typically contains an AF2. DBDs and LBDs are linked through a nonconserved hinge region that in some receptors also harbors a nuclear localization signal. Hinge regions can also provide additional DNA-binding specificity (8). In some receptors, AF1 and AF2 physically and functionally cooperate in recruiting transcriptional coregulators (12, 49, 50).

For a subset of subfamily 3 receptors, such as GR, ligand binding appears to take place in the cytoplasm. This step triggers the release of receptor from chaperone proteins and the subsequent translocation of NR to the nucleus (51, 52). By contrast, ER, ERR, and other NRs do not appear to interact with chaperones in the cytoplasm and are readily detected within the nucleus even without their ligands. For all NRs, stimulation with ligand produces a number of critical changes in their functional interactions while inside the nucleus.

The NR endogenous ligands have the shared property of being small molecules (typically <400 Da) with characteristically high hydrophobicity. These 2 properties provide them an ease of movement in crossing between tissues and cells. Their properties also favor their binding inside the hydrophobic pockets of receptor LBDs (4, 53). NR LBDs fold into a 3 layer antiparallel α-helical sandwich that consists of 12 helices (H1–H12) and 2 short β-strands (Figure 1A). The ligand-binding pocket is above H5 and is lined by residues from H3, H5–H7, H11–H12, and β-strands. Ligand binding completes the otherwise empty hydrophobic holes inside the LBD pockets of receptors and leads to repositioning of H12. Agonists and antagonists stabilize different orientations of the H12. Interestingly, crystal structures of LBDs from orphan receptors without ligand resemble the agonist bound state with regard to the orientation of H12 (4).

Figure 1.

RXR LBD (shown in brown) forms unique heterodimeric complexes with PPAR and TR LBDs (shown in gray). A, The crystal structure of the intact complex of PPAR-RXR heterodimer on DNA is used to highlight architecture of the active conformation of RXR LBD bound to agonist (blue) and coactivator peptide (yellow). Helices and strands that form the ligand-binding pocket are labeled. AF2 surface is where the coactivator peptide associates. B, The crystal structure of the complex of TR-RXR LBD heterodimer in the presence of T₃ shows the architecture of a silenced RXR LBD where ligand binding is allosterically blocked by alterations in the dimer interface and distinct conformational changes at the ligand-binding pocket and AF2 surface.

The AF2 in LBDs corresponds to a surface formed near H12, H3, and H4, where coregulator peptides selectively bind as short α-helices and control transactivation (Figure 1A) (8, 54–56). Coactivator peptides have conserved leucines (LXXLL motif where X is any amino acid) that insert in AF2 grooves, but they also have polar residues that form stabilizing charge clamp interactions. The corepressor peptides also insert hydrophobic side chains near AF2, and these peptides fall into 2 categories, conventional with CoRNR boxes (LXXXIXXXI/L motif), and nonconventional (ALXXLXXY) found for tailless homolog (TLX) and related other orphan receptors (4, 57). These coactivator and corepressor peptides compete for overlapping surfaces near AF2.

Solution Studies Point to a Highly Inducible Conformation in RXR LBD

The crystal structures of some LBDs in the absence of ligand show a stable conformation similar to that of the ligand bound state, suggesting ligand binding creates subtle changes in the architecture. However, solution analyses of structures by NMR spectroscopy and hydrogen-deuterium exchange mass spectrometry have indicated otherwise (58). Importantly, NMR studies have identified a molten globule conformation for RXR LBD in the absence of agonist. The apo-RXRα displays missing or broad line widths for resonances corresponding to residues within the ligand binding pocket and AF2 surface, whereas the holo-RXRα bound to 9-cis-retinoic acid has sharp resonances corresponding to a well-defined conformation (59).

Because RXR agonist can stimulate transactivation of the PPAR-RXR, but it has no effect on TR-RXR, the NMR structure of RXR LBD was examined to compare conformational changes in the context of these LBD heterodimers to understand mechanisms of signal transduction (60). Examination of the RXRα-PPARγ LBD heterodimer showed the apo-RXR adopted a molten globule conformation, which became stable upon addition of RXR agonist (9-cis-retinoic acid). In the holo-RXR, the dimer interface (H7, H9, and H10/H11) became altered differentially whether agonist or antagonist of PPARγ were added. This study identified distinct routes for signal transduction through the RXRα-PPARγ LBD dimer interface, including a central role for RXR H5 to integrate signals from the dimer interface, ligand, and peptide-binding sites. A different scenario was found for TRβ-RXRα LBD heterodimer that supported RXR “silencing”. Upon addition of TR to the holo-RXR, the stable conformation of RXR converted to a molten globule state. Remarkably, addition of TR agonist T₃ led to stabilization of the RXR structure with major alterations. Crystallography was used to capture the detailed structure of this stable conformation, which showed RXR H5 rotated into a new position and the ligand binding pocket and AF2 surface were both disrupted (Figure 1B). Observations of allosteric signaling pathways in both these heterodimers agreed with predictions from the sequence-based method of statistical coupling analysis that had identified a network of energetically coupled residues in RXR LBD (60, 61).

Signal Transduction Through the DBD

As with LBDs, the DBDs of NRs have also been shown to interact with their partner as independent modules. NRs interact with DNA-response elements by anchoring DBDs to hexanucleotide sequences in their response elements (8). Sequence conservation is highest in the DBD portions of NRs, and accordingly, the hexameric binding sites appear to conform to 2 consensus types (5′-AGGTCA-3′ and 5′-AGAACA-3′). The DBDs fold into 2 zinc-binding modules that have a helix for DNA-recognition, which forms base-specific and phosphate-specific contacts. With another helix that lies perpendicular to the DNA-recognition helix, DNA recognition is stabilized. As homo- or heterodimers, NRs discriminate their binding sites by recognizing the orientations and spacing of 2 DNA half-sites. Studies on GR DBD homodimer binding to DNA with different sequences have identified an allosteric pathway that suggests GR dimer partners read DNA shape to direct sequence-specific gene activity (62, 63).

The DBD region of NRs may also function in non-DNA partner recognition to mediate transactivation (via coactivator recruitment) or transrepression (via lncRNA binding) function. For example, posttranslational modification of ERα DBD by the histone lysine methyltransferase G9a is shown to recruit the MOF coactivator complex to physically associate with ERα in an agonist-dependent manner (19). Furthermore, some receptor DBDs can alternatively recognize lncRNA that mimics their response element as found in Gas5 lncRNA. Structural and binding studies on GR DBD identified a response element mimic in Gas5 forms a stem loop to effectively associate with the DNA-binding surface and prohibit DNA interactions of GR (64).

High-Resolution Structures of Full-Length NRs on DNA Show Unique Quaternary Organizations

Three crystal structures have emerged that show the complex quaternary architectures for the PPARγ-RXRα heterodimer, the hepatocyte nuclear factor HNF-4α homodimer, and the RXRα-LXRβ heterodimer on their DNA-response elements (9–11). As shown in Figure 2, these pictures show how LBDs and DBDs can become physically coupled in a variety of different ways within each complex (65). Critical functional components were visualized and their positions clearly defined in each crystal structure, including their binding modes to small-molecule ligands, coactivator peptides and DNA-response elements. In the case of the PPAR-RXR/DNA structure, several independently conducted solution-based studies have confirmed the compacted nature of its quaternary structure on DNA (9, 65–67). In addition, molecular dynamics studies on this crystal structure have identified the trajectories of dynamics and identified the interdependent motions and preferential allosteric pathways (68).

Figure 2.

Three crystal structures showing the interactions of multidomain NR homo- and heterodimers on their idealized DNA-response elements. A, The PPAR-RXR heterodimer on DNA direct-repeats with single-base spacer. B, The HNF-4α homodimer as bound to the same DNA as in A. C, The LXR-RXR heterodimer on DRs with 4-base spacer. Coactivator LXXLL motifs are seen in all 3 structures on both subunits at their LBDs (yellow).

These intact structures are the first clear snapshots of the complex pictures of NRs in their physiologically functional states. As more NR/DNA structures are visualized, we will learn more about their complex quaternary states and interaction modes. For example in the case of RXR-TR heterodimer, biochemical studies suggest that the DNA causes a displacement of the TR DBD away from its LBD (69). The DNA-induced displacement of the TR DBD away from the LBD makes available a surface on the LBD required for coregulator binding (69). This suggests a very different way in which DNA acts as allosteric effector for TR-RXR complexes than currently known from the 3 crystal structures shown in Figure 2 (70).

The most intriguing lesson learned from the analysis of the currently available multidomain NR/DNA complexes is that their quaternary architectures and patterns of domain-domain interactions are unique in each case and do not adhere to a commonly shared mold (65). The architectural connections in such structures were found to be strikingly different even when the PPAR-RXR structure was compared with the HNF-4α structure, with both being positioned on the same DNA element (Figure 2, A and B) (9, 10). These different quaternary structures are due to the different length and sequence of the NR hinge regions. Also, the nonconserved surface residues on the outside of their LBDs establish different DBD-LBD interfaces in each case. As for the divergent amino terminal segments and the AF1 functions that may be contained there, these receptor regions could not be visualized within any of these complexes. The use of full-length coactivator proteins in future studies should better reveal how the AF1 segments may be cooperating with the LBDs (12).

Although the 3 crystal structures show differences in quaternary architectures (domain-domain interaction patterns), they still share several key aspects in their basic molecular organizations. All 3 NR structures have ligands that are clearly inside their 2 LBD pockets, precisely as known from previous structural studies using isolated LBDs of RXR, PPAR, LXR, and HNF-4α (53). Therefore, access to ligands is not blocked by the DNA or by the domain-domain interfaces in these complexes. All 3 structures also contain bound LXXLL peptides (Figure 2) that are positioned precisely as expected from previous LBD-only studies (53). The AF2 surfaces are positioned furthest from the DNA-binding surface. This distal positioning of the DNA away from the LBD allows the larger coregulator proteins to attach to their binding sites without being blocked or entangled by DNA. In each case, the coactivator binding stoichiometry is clear and as predicted. Each of these NR structures shows its 2 subunits binding independently and equivalently to separate coactivator peptides (Figure 2).

The complex and unique domain-domain interactions in these NRs also suggest that physical pathways are well in place to allow for allosteric signal communication across the coupled surfaces of these receptor complexes (65). Developing general concepts of allosteric signal transmission has been a challenge to the field. In the case of HNF-4α homodimer, a convergence zone was identified that coupled surfaces from both LBDs, the DBD of the upstream subunit and the hinge region of the downstream subunit (Figure 2B). Some of the disease-linked mutations (maturity onset diabetes of the young, MODY1) as well as posttranslational modifications occur in this zone and were correlated with a weakening of DNA interaction (10). Although positioned on the LBDs of HNF-4 α, the small perturbations caused by posttranslational modifications and disease mutations can be efficiently transmitted allosterically so as to weaken distal DBD-DNA interactions.

Aside from the 3 high-resolution (2.8–3.2 Å) crystal structures of NR/DNA complexes, 2 other structures have been reported at substantially lower resolution (10–12 Å). These other studies focused on the VDR-RXR/DR3 heterodimer and the Drosophila EcR-USP/IR1 heterodimer (71, 72). Both were determined by cryo-electron microscopy (cryo-EM), and owing to the limiting resolutions, only the broad molecular envelopes were visualized. Both complexes appear to lack physical interaction between the DBD and LBD portions of subunits. In the case of RXR-VDR, the absence of such interactions within the cryo-EM model has proved inconsistent with the conformation described for intact RXR-VDR/DR3 from hydrogen-deuterium exchange mass spectrometry. Using the latter technique, this heterodimer was examined for interacting with DNA in the presence and absence of agonists of VDR and RXR. Long-range allosteric interactions between the various domains and AF2 surfaces were identified (73). These studies further showed that response elements with variable nucleotide composition such as the natural sequence for a VDR target gene (consisting of a single half-site) and the consensus DR3 element, differentially affect the conformational dynamics of the RXRα-VDR heterodimer and stability of their AF2 surfaces.

Proteomic Analyses of NR Complexes Identify A Multitude of Coregulators/Cofactors

As mentioned in the introduction, a variety of interactions can mediate the functional associations of NRs with response elements, suggesting elaborate protein-protein interactions for NRs. The use of proteomic mass spectrometry can help bridge the gap in the identification of collaborating factors that interact with NRs. One method relies on tandem affinity purification (TAP), whereby a tag is cloned onto a gene of interest to generate a fused protein (74, 75). The protein TAP-tag fusion is expressed in a cell line and immunoprecipitation is used to isolate the proteins that associate with that NR. The precipitated complex is separated into its individual components by gel electrophoresis and examined by mass spectrometry to identify each constituent. A search for ER-interacting proteins using TAP-tagged approaches was conducted in HeLa cells, revealing a set of novel ER-associated proteins, including the nucleosome remodeling deacetylase complex that may interact in a cell cycle-dependent fashion (76).

A technical improvement in that approach uses formaldehyde cross-linking to maintain transient protein-protein interaction, and permitted improved sensitivity in detecting NR-associated proteins (77). This approach has now revealed over a hundred different proteins that can associate with ER. The list of enriched proteins includes GREB1, NRIP1, NCOA3, FoxA1, GATA3, CBP, AIB1, RXRα, TLE1, and AP2γ (77). Moreover, the high sensitivity of this method can also allow for detection of NR interactions in clinical tumor samples derived from patients (75).

Given DNA binding of NRs influences the association with diverse partners, a better approach has used biotin attached estrogen-response elements in pull-down assays coupled to mass spectrometry for identifying components of the ER interaction network (78). The results show that ERα as bound to ligand 17β-estradiol forms a complex with a limited number of coregulators including all 3 steroid receptor coactivator p160 family of proteins (SRC-1, SRC-2, and SRC-3), the histone acetyltransferases CREB binding protein (CBP/CREBBP) and adenovirus early region 1A binding protein p300 (p300/EP300) and mediator subunit (78). The study also found RIP-140 and C-terminal binding protein, both of which were thought to be corepressors, to be present in the active complex (Figure 3A). A significantly smaller list of 17 endogenous coregulators was observed from HeLa and MCF-7 cells with a subset of these interactions being biochemically stable (78). The estrogen ligand establishes the stable, poised complex that contains coactivators, corepressor, and the enzyme DNA-PK. But when DNA-PK finds its substrate ATP, the protein complex takes on a more dynamic state, wherein some coregulators leave and others join in (Figure 3A). ATP addition causes ER and SRC-3 both to become phosphorylated.

Figure 3.

The complex interactions of ER are revealed through proteomic and structural studies. A, The components of the ERα homodimer on DNA. B, The cryo-EM detected components and stoichiometry of proteins bound to the ERα homodimer on DNA.

More recently, promoter enrichment-quantitative mass spectrometry has been used to define the composition of a NR regulatory complex formed in macrophages at the promoter of the Abca1 gene, a membrane-associated lipid transporter (79). The attenuation of Abca1 expression has implications for the development of atherosclerosis as it impedes cholesterol efflux in macrophages. This study included the use of a biotinylated DNA to enrich for Abca1 promoter-associated proteins at a 321-bp regulatory sequence that flanks the LXR-response element. An important aspect of this study was that it assessed not only the effect of LXR ligand stimulation, but also the subset of the response strictly dependent on LXR binding (79). A total of 19 proteins were detected upon stimulation with a synthetic LXR ligand, including LXRβ, RXRβ, and a few other transcription factors. Only 7 among 19 proteins bound in an LXR-dependent manner, and NCOA5 and H/ACA ribonucleoprotein complex subunit 2 protein (NHP2) were identified as 2 new LXR coregulators. Using reporter assays, it was confirmed that NCOA5 expression repressed Abca1 transcriptional response, and pull-down studies established interaction between NCOA5 and LXRβ. Thus, stimulation with an LXR ligand impedes cholesterol efflux in macrophages by recruiting NCOA5 to promoter of Abca1 to attenuate LXRβ-RXRβ-dependent gene expression (79).

Higher Order Structural Visualizations of NR Mega Dalton Complexes

With global gene analyses and proteomic studies now uncovering ever more complex arrays of NR-interacting proteins, structural biologists must establish more sophisticated ways in which to visualize elements within these larger complexes. Major challenges come from the AF1 regions of NRs, which can be disordered and thus prevent crystallization. But including larger coregulator complexes in future studies may stabilize the AF1 segments and make it possible to study these complexes effectively. Also, substantial portions of both the p160 coactivators and the p300 protein are also disordered on their own, but these parts may establish stable conformations when arranged in complexes with NRs (12).

The proteomic studies with ER are now paving the way for more elaborate structural studies on larger NR complexes. O’Malley and coworkers have successfully visualized the full-length ER and full-length coactivator complexes using cryo-EM (Figure 3B) (12). They took advantage of their findings that showed ER formed a stable complex with p160 coactivators on DNA. They then assembled the full-length ER homodimeric protein with its DNA-response element and both SRC-3 and p300 proteins. This large complex was assembled using individually purified components to visualize the entire 720-kDa complex at a resolution of approximately 25 Å. This resolution limit does not allow for the details of the complex to be established but does allow for the positioning of the components. The cryo-EM study on ER was confirmed by many validation steps to ensure the positions of the components were clearly interpreted. These validation studies included the use of specific monoclonal antibodies to unambiguously identify the p300 protein and the location of the ER AF1 in this structure.

Their map provides us a solid understanding of the spatial organization of the ER/DNA/SRC-3/p300 complex (Figure 3B) (12). Two SRC-3 proteins become associated with the ER homodimer, but only one p300 molecule engages both SRC proteins. Therefore, 2 coactivator molecules cooperate for binding to ER homodimer and the p300 protein. The AF1 regions of each ER are near the LBDs, suggesting this ER segment cooperates with the LBD AF2 surface in securing SRC-3 protein binding. The study further reveals that 4 regions of p300 are involved in interacting with the SRC-3 proteins, and the p300 protein does not interact directly with DNA as it sits over SRC-3 proteins. The authors noted that other p160 coactivator family members (SRC-1/2) could be effectively accommodated in such complexes by substituting in the binding site observed for SRC-3 (12).

Concluding Remarks

Crystal structures reported now for 3 multidomain NR complexes suggest that the DBD and LBD segments can be physically and functionally interconnected, making possible the transmission of signals between distal domains in an allosteric fashion. Therefore, DNA-binding, ligand binding, and coregulator binding are interdependent events that use the participation of different parts of NR polypeptides. However, these are still early days in our understanding of full-length NR complexes. Despite overall domain conservation, NRs vary significantly in their sequences allowing them to associate with diverse ligands, peptides and DNA. Therefore, it is likely that many other variations in quaternary states and allosteric pathways will emerge as structural biology methods are further applied to NRs. We now know that NRs collaborate with a far greater number of protein partners than their well-known coactivators and corepressors. Their elaborate interactions are critical for defining ligand-specific and cell-specific responses. The new lessons obtained from the proteomic and cryo-EM studies on ERα pave toward efforts at observing the larger, more functionally relevant complexes of NRs that can establish the basis for their interactions and allosteric signaling networks.

Abbreviations

AF

activation function

CAR

constitutive androstane receptor

cryo-EM

cryo-electron microscopy

DBD

DNA-binding domain

DR

direct repeat

ER

estrogen receptor

ERR

estrogen-related receptor

FXR

farnesoid X receptor

GR

glucocorticoid receptor

HNF

hepatocyte nuclear factor

LBD

ligand-binding domain

lncRNA

long noncoding RNA

LXR

liver X receptor

NCOA

nuclear receptor coactivator

NMR

nuclear magnetic resonance

NR

nuclear receptor

p300

adenovirus early region 1A binding protein p300

PPAR

peroxisome-proliferator-activated receptor

PXR

pregnane X receptor

RAR

retinoic acid receptor

RXR

retinoic X receptor

SRC

steroid receptor coactivator

TAP

tandem affinity purification

TLX

tailless homolog

TR

thyroid hormone receptor

VDR

vitamin D receptor.