antagonist- and inverse agonist-driven interactions of the ......sence of ligand (fig. 1a, lane 3),...

Antagonist- and Inverse Agonist-Driven Interactionsof the Vitamin D Receptor and the ConstitutiveAndrostane Receptor with Corepressor Protein

Harri Lempiainen, Ferdinand Molnar, Manuel Macias Gonzalez, Mikael Perakyla, andCarsten Carlberg

Departments of Biochemistry (H.L., F.M., M.M.G., C.C.) and Chemistry (M.P.), University of Kuopio,FIN-70211 Kuopio, Finland

Ligand-dependent signal transduction by nuclearreceptors (NRs) includes dynamic exchanges ofcoactivator (CoA) and corepressor (CoR) proteins.Here we focused on the structural determinants ofthe antagonist- and inverse agonist-enhanced in-teraction of the endocrine NR vitamin D receptor(VDR) and the adopted orphan NR constitutive an-drostane receptor (CAR) from two species with theCoR NR corepressor. We found that the pure VDRantagonist ZK168281 and the human CAR inverseagonist clotrimazole are both effective inhibitors ofthe CoA interaction of their respective receptors,whereas ZK168281 resembled more the mouseCAR inverse agonist androstanol in its ability torecruit CoR proteins. Molecular dynamics simula-tions resulted in comparable models for the CoR

receptor interaction domain peptide bound toVDR/antagonist or CAR/inverse agonist com-plexes. A salt bridge between the CoR and a con-served lysine in helix 4 of the NR is central to thisinteraction, but also helix 12 was stabilized by di-rect contacts with residues of the CoR. Fixation ofhelix 12 in the antagonistic/inverse agonistic con-formation prevents an energetically unfavorablefree floatation of the C terminus. The comparablemolecular mechanisms that explain the similarfunctional profile of antagonist and inverse ago-nists are likely to be extended from VDR and CARto other members of the NR superfamily and maylead to the design of even more effective ligands.(Molecular Endocrinology 19: 2258–2272, 2005)

NUCLEAR RECEPTORS (NRs) form the largestfamily of metazoan transcription factors and reg-

ulate the expression of target genes that affect pro-cesses as diverse as reproduction, development, andmetabolism (1). Dietary lipids, such as cholesterol andfatty acids, and their metabolites, such as steroids andoxysterols, act as ligands to many of the 48 humanNRs and those of other metazoan species (2). SeveralNRs are pharmaceutical drug targets for the treatmentof diverse diseases, such as type 2 diabetes, athero-sclerosis, osteoporosis, and cancer. To aid interven-

tion in these diseases, multiple synthetic NR agonists,antagonists and inverse agonists have been synthe-sized and characterized (3). The physiology of NRsand their natural bioactive lipid and synthetic analogligands shows a broad variety, but their actions can besummarized to a couple of common gene regulatorymechanisms.

Basis of the common actions of NRs is their con-served structure. Members of the NR superfamily areidentified by the presence of a highly conserved DNAbinding domain and a structurally conserved ligandbinding domain (LBD) (4). The LBD of most NRs is acharacteristic three-layer antiparallel �-helical sand-wich formed by 11–13 �-helices. In the lower half ofthe domain, there is no central helical layer but a largenonpolar pocket, to which the various lipophilic li-gands bind. One side of this pocket is sealed by theC-terminal helix of the receptor, often called helix 12.This helix serves as a molecular switch by allowing theLBD in its agonistic conformation to interact with co-activator (CoA) proteins, such as steroid receptor co-activator-1, transcription intermediary factor 2 (TIF2),and receptor-associated coactivator 3 (5). In the ab-sence of an agonistic ligand, NRs interact with core-pressor (CoR) proteins, such as nuclear receptor core-pressor (NCoR), silencing mediator of retinoic acid andthyroid hormone receptor, and Alien (6). CoA and CoRproteins both contain multiple, short receptor interac-tion domains (RIDs), composed of the sequence

First Published Online May 19, 2005Abbreviations: ANF Atrial natriuretic factor; CAR constitu-

tive androstane receptor; CITCO 6-(4-chlorophenyl)imi-dazo[2,1-b] [1,3]thiazole-5-carbaldehyde O-3,4-dichloroben-zyl) oxime; CoA, coactivator; CoR corepressor; CYP P450cytochrome mono-oxygenase; DR direct repeat; GST gluta-thione-S-transferase; LBD ligand binding domain; MD mo-lecular dynamics; NCoR, nuclear receptor corepressor; NRnuclear receptor; 1�,25(OH)2D3; 1�,25-dihydroxyvitamin D3;PBREM phenobarbital-responsive enhancer module; PDB,Protein Data Bank; PPAR peroxisome proliferator-activatedreceptor; RA retinoic acid; RE response element; RID recep-tor interaction domain; RMSD root mean square deviation;RXR retinoid X receptor; TCPOBOP 1,4-bis[2-(3,5-dichloro-pyridyloxy)] benzene; TIF2, transcription intermediary factor2; VDR 1�,25(OH)2D3 receptor.

Molecular Endocrinology is published monthly by TheEndocrine Society (http://www.endo-society.org), theforemost professional society serving the endocrinecommunity.

0888-8809/05/$15.00/0 Molecular Endocrinology 19(9):2258–2272Printed in U.S.A. Copyright © 2005 by The Endocrine Society

doi: 10.1210/me.2004-0534

2258

on August 25, 2005 mend.endojournals.orgDownloaded from

http://mend.endojournals.org

LXXLL in case of CoAs (7) and LXXXIXXX[I/L] in case ofCoRs (8). Both type of coregulators interact withlargely overlapping surfaces on the LBD suggestingthat their binding is mutually exclusive (9). The mouse-trap model (10) proposes that helix 12 acts as a lid tothe ligand-binding pocket of the LBD, which has to beclosed to allow NR interaction with CoAs and open,when the receptor contacts CoRs. This implies thatwhereas helix 12 takes only one defined position in theagonist-bound receptor, multiple positions of the helixare possible in antagonist-bound or apo-receptor.

The NR for the seco-steroid 1�,25-dihydroxyvitaminD3 (1�,25(OH)2D3), the vitamin D3 receptor (VDR), isone of the 11 classic endocrine members of the NRsuperfamily that bind their respective ligands with highaffinity [dissociation constant (Kd) value of 1 nM orlower] (2). 1�,25(OH)2D3 is a key player in calciumhomeostasis and bone mineralization (11) and also hasantiproliferative and prodifferentional effects on vari-ous cell types (12). Adopted orphan NRs form anothersubclass within the NR superfamily. These NRs bind avariety of structurally diverse compounds with a rela-tively low affinity (Kd in the order of 1 �M) (13). Con-stitutive androstane receptor (CAR) is an interestingadopted orphan NR because it has an exceptionallyhigh constitutive activity (14) and therefore is function-ally opposite to the low basal activity of endocrineNRs, such as VDR. CAR plays a key role in the re-sponse to chemical stress and regulates an overlap-ping set of genes, some of which encode proteins,such as P450 cytochrome monooxygenases (CYPs)that are involved in the detoxification of potentiallyharmful xenobiotics and endobiotics (15). Primary NRtarget genes are defined through the presence of par-ticular binding sites, referred to as response elements(REs), in their promoter regions (16, 17). Peroxisomeproliferator-activated receptors (PPARs), CAR, VDR,and several other members of the NR superfamily formheterodimers with the retinoid X receptor (RXR) on REsthat are composed of a direct repeat (DR) of hexamericbinding sites (18). Multiple CAR RE clusters are com-monly called phenobarbital-responsive enhancermodules (PBREMs). The mouse CYP2B10 (ortholog tohuman CYP2B6) gene contains two DR4-type REswith an additional binding site for the transcriptionfactor NF-1 (19).

Most natural and synthetic NR ligands are agonists,such 1�,25(OH)2D3 for the VDR, the imidazothiazole de-rivative 6-(4-chlorophenyl)imidazo[2,1-b] [1, 3]thiazole-5-carbaldehyde O-3,4-dichlorobenzyl)oxime (CITCO) forhuman CAR (20) and the hepato-mitogen 1,4-bis[2-(3,5-dichloropyridyloxy)]benzene (TCPOBOP) (21) formouse CAR. The antimycotic clotrimazole (13) and thetestosterone metabolite androstanol (14) deactivatehuman and mouse CAR, respectively, and are there-fore considered as inverse agonists (22). The two-sidechain 1�,25(OH)2D3 analog Gemini can act, at highCoR levels, as inverse agonist of the VDR (23, 24). Incontrast, the 25-carboxylic ester 1�,25(OH)2D3 analogZK168281 (23) is a pure VDR antagonist (25, 26). It is

thought that inverse agonists and antagonists stabilizeNR-LBDs in different conformations (27), whereas wedemonstrate in this study that ZK168281 and clotrim-azole are both effective inhibitors of the CoA interac-tion of their respective receptors. ZK168281 resem-bled more androstanol in potent CoR recruitment.Molecular dynamics (MD) simulations resulted in com-parable models for the LBDs of VDR and both CARscomplexed with a CoR-RID peptide and ZK168281,clotrimazole, and androstanol, respectively. Weshowed that a salt bridge between the CoR and aconserved lysine in helix 4 of the NR-LBDs are impor-tant for the stability of the antagonist-/inverse agonist-stabilized NR-LBD-CoR complex. Moreover, the sta-bilization of helix 12 by direct contacts with residues ofthe CoR is common to these models. This demon-strates that NR antagonists and inverse agonists re-semble each other in their functional profile and mo-lecular mechanisms.

RESULTS

Individual Ligand-Triggered Interaction Profilesof VDR, CARs, and RXR with Coregulatorsin Solution

The ligand-triggered physical interaction between theNRs human VDR, human and mouse CAR with theCoA TIF2, and the CoR NCoR was assessed by glu-tathione-S-transferase (GST)-pull-down assays (Fig.1). For this purpose, in vitro-translated, [35S]-labeledwild-type NR proteins were incubated in the presenceof their respective agonistic or antagonistic/inverseagonistic ligands (for their structures see Fig. 1D) withbacterially produced GST alone, GST-TIF2 or GST-NCoR fusion protein immobilized on Sepharosebeads. The GST-TIF2 protein has all three LXXLL RIDsof the CoA, whereas the GST-NCoR protein containsonly the second of the two RIDs of the CoR. We testedalso the an alternative GST-NCoR fusion protein con-taining the first RID but did not obtain efficient inter-action with the three NRs neither in GST-pull-down norin supershift assays (data not shown). GST proteinalone showed only weak residual association with thethree NRs (lane 2 in Fig. 1, A–C), which was consid-ered as unspecific background binding. VDR showedreasonable association with TIF2 already in the ab-sence of ligand (Fig. 1A, lane 3), which could be in-creased significantly by the agonist 1�,25(OH)2D3

(lane 4) and decreased by the pure antagonistZK168281 (lane 5). In addition, VDR displayed ligand-independent binding of NCoR (lane 6), which was de-creased by 1�,25(OH)2D3 (lane 7) and increased byZK168281 (lane 8).

In the GST-pull-down assays, i.e. in solution, humanand mouse CAR show a ligand-independent interac-tion with TIF2 comparable to that of VDR (comparelane 3 in Fig. 1, A–C). The species-specific agonistsCITCO and TCPOBOP could slightly increase this as-

Lempiainen et al. • CoR Interactions of VDR and CAR Mol Endocrinol, September 2005, 19(9):2258–2272 2259



sociation between the two CAR orthologs and CoAprotein (Fig. 1, B and C, lane 4), whereas the humanand mouse inverse agonists clotrimazole and andro-stanol, respectively, both significantly reduced thisprotein-protein interaction (Fig. 1, B and C, lane 5).Human and mouse CAR showed a less prominentbasal interaction with NCoR than VDR (compare lane6 in Fig. 1, A–C). The mouse CAR-CoR complex was

reduced by the application of the agonist TCPOBOP(Fig. 1C, lane 7), whereas the agonist CITCO increasedthe interaction between human CAR and NCoR (Fig.1B, lane 7). The inverse agonist androstanol behavedas expected and increased the interaction betweenmouse CAR and NCoR (Fig. 1C, lane 8), but its humancounterpart clotrimazole (13) reduced the interactionbetween human CAR and NCoR (lane 8). Taken to-

Fig. 1. Individual Ligand-Enhanced Interaction Profiles of VDR and CAR with CoAs and CoRs in SolutionGST-pull-down assays were performed with bacterially expressed wild-type GST-TIF2 or GST-NCoR and full-length in

vitro-translated, [35S]-labeled human VDR (A), human CAR (hCAR, B), and mouse CAR (mCAR, C) in the absence and presenceof their respective ligands (1 �M for VDR ligands and 10 �M for CAR ligands). GST alone (�) was used as control. After precipitationand washing, the samples were electrophoresed through 15% sodium dodecyl sulfate-polyacrylamide gels and the percentageof precipitated NRs in respect to input was quantified using a Fuji FLA3000 reader. Representative gels are shown. Columnsrepresent the mean values of at least three experiments, and the bars indicate SD. Statistical analysis was performed using atwo-tailed, paired Student’s t test, and P values were calculated in reference to respective solvent controls (*, P � 0.05).Two-dimensional structures of VDR and CAR ligands are shown (D).

2260 Mol Endocrinol, September 2005, 19(9):2258–2272 Lempiainen et al. • CoR Interactions of VDR and CAR



gether, both VDR and the CAR orthologs display indi-vidual ligand-enhanced interaction profiles with CoAand CoR proteins in solution, which may be related tothe divergent structure of the ligands.

Ligand-Dependent Interaction Profiles of VDR-RXR and CAR-RXR Heterodimers withCoregulators on DNA

To test whether the ligand-enhanced interactions be-tween VDR, both CARs, TIF2, and NCoR were alsovalid for DNA-bound heterodimeric complexes withRXR supershift assays (Fig. 2) were performed with thesame panel of ligands as in the GST-pull-down assays(Fig. 1). GST alone did not induce any supershift (lane1 in Fig. 2, A–C). Both the VDR ligands induced the

known increase in VDR-RXR heterodimer complex for-mation on DNA (28) (Fig. 2A, lanes 2 and 3) and aminority of human CAR molecules displayed DNAbinding as a monomer as described previously (29)(Fig. 2B, lanes 1, 3, 6, and 9). In the absence of ligand,DNA-complexed VDR did not show any associationwith TIF2 or NCoR (Fig. 2A, lanes 4 and 8), which is incontrast to the interaction profile of VDR in solution(Fig. 1A). The high affinity of apo-VDR in solution forCoAs could have technical reasons being related to asignificant molar excess of bacterially produced CoAfusion protein. However, more important for the un-derstanding of vitamin D signaling is the demonstra-tion that DNA-bound VDR is able to attract a reason-able amount of CoA proteins after a conformational

Fig. 2. Ligand-Dependent Interaction Profiles of VDR-RXR and CAR-RXR Heterodimers with CoAs and CoRs on DNACombined gel shift and supershift experiments were performed with equal amounts of in vitro-translated wild-type human VDR

(A), human CAR (B), or mouse CAR (C), RXR protein and [32P]-labeled DR3- (A) or DR4-type RE (B and C). VDR-RXR and CAR-RXRheterodimers were preincubated with solvent, 100 nM 1�,25(OH)2D3 or 1 �M ZK168281 (A), 10 �M CITCO or clotrimazole (B), and10 �M TCPOBOP or androstanol (C) as indicated. Equal amounts of bacterially expressed wild-type GST (�), GST-TIF2, orGST-NCoR were then added. Protein-DNA complexes were resolved from free probe through 8% nondenaturing polyacrylamidegels. The relative amounts of supershifted complexes were quantified using a Fuji FLA3000 reader. Representative gels areshown. NS, Nonspecific complexes. Columns represent the mean of at least three experiments, and bars indicate SD. A two-tailed,paired Student’s t test was performed and P values were calculated in reference to the respective solvent control (*, P � 0.05;**, P � 0.01; ***, P � 0.001).




change induced by 1�,25(OH)2D3 (lane 5). The antag-onist ZK168281 did not induce any interaction of VDRwith CoA protein (lane 7) and a combination of agonistand antagonist resulted only in a very faint complex ofVDR-RXR heterodimers with TIF2 (lane 6). Applicationof 1�,25(OH)2D3 alone did only induce residual inter-action of VDR with NCoR (lane 9), whereas binding ofZK168281 to the VDR resulted in a strong associationwith NCoR (lane 10). The combined application ofagonist and antagonist resulted in the interaction ofsignificant NCoR amounts with DNA-bound VDR-RXRheterodimers (lane 11).

In contrast to VDR, DNA-complexed human andmouse CAR showed significant interaction with TIF2even in the absence of ligand (Fig. 2, B and C, lane 4).Interestingly, both receptors interacted also withNCoR in a ligand-independent fashion (lane 7). Asexpected from an agonist, CITCO increased the inter-action of human CAR with CoA, but surprisingly, it alsoinduced a significant interaction of the receptor withNCoR (Fig. 2B, lanes 5 and 8). Clotrimazole bluntedthe constitutive interaction of human CAR with TIF2and also reduced the interaction of the receptor withNCoR (lanes 6 and 9). Similar to CITCO in case ofhuman CAR, TCPOBOP increased the interaction ofmouse CAR with CoA protein (Fig. 2C, lane 5), but incontrast to CITCO it decreased the interaction of itsreceptor with CoR protein (lane 8). Finally, androstanolbehaved as expected from an inverse agonist in that itdecreased the interaction of mouse CAR with TIF2(lane 6) and increased the contact with NCoR (lane 9).In summary, as a DNA-bound heterodimer with RXRVDR shows no constitutive interaction with CoA pro-

tein and no significant interaction with CoR protein inthe presence of agonist. However, in all other aspectsthe supershift assays confirm the ligand-enhanced in-teraction profile of VDR and the two CARs with co-regulator proteins as observed in the GST-pull-downassays (Fig. 1B). This is an indirect proof that RXR hasonly minor contributions to the ligand profile of VDRand the two CARs. Moreover, it means that the VDRantagonist ZK168281 and the human CAR inverse ag-onist clotrimazole are both effective inhibitors of theCoA interaction of their respective receptors, whereasconcerning potent CoR recruitment ZK168281 re-sembles more the mouse CAR inverse agonistandrostanol.

Coregulator-Triggered Ligand Responsiveness ofVDR and CAR in MCF-7 Cells

To compare the functional consequences of agonistand antagonist/inverse agonist application to VDR andthe two CARs, we performed reporter gene assays inthe transiently transfected model cell line MCF-7 (Fig.3). The transactivation potential of VDR was assessedon four copies of the DR3-type RE of the rat atrialnatriuretic factor (ANF) gene (30), whereas the twoCARs were tested on the PBREM (which contains twoDR4-type REs) of the mouse CYP2B10 gene (31); bothREs were fused individually with the thymidine kinasepromoter driving the luciferase reporter gene. At en-dogenous coregulator levels, the very low basal levelof VDR on the rat ANF DR3-type RE (Fig. 3A, lane 1)was induced nearly 50-fold by 10 nM 1�,25(OH)2D3

(lane 2), whereas 100-times higher concentrations of

Fig. 3. CoA- and CoR-Triggered Ligand Responsiveness of VDR and CAR in MCF-7 CellsReporter gene assays were performed with extracts from MCF-7 cells that were transiently transfected with a luciferase

reporter construct containing four copies of the DR3-type RE of the rat ANF gene (A) or the PBREM of the mouse CYP2B10 gene(B and C). Wild-type human VDR (A), human CAR (B), and mouse CAR (C) expression vectors as well as plasmids coding forfull-length TIF2 or NCoR were also cotransfected as indicated. Cells were treated for 16 h with solvent, 10 nM 1�,25(OH)2D3 or1 �M ZK168281 (alone and in combination, A), 1 �M CITCO or 1 �M clotrimazole (B), and 1 �M TCPOBOP or 1 �M androstanol(C) as indicated. Relative luciferase activities were measured in reference to agonist-induced values of cells not overexpressingcoregulator protein. Fold inductions in relation to solvent are shown on the tip of the columns, and the relative basal activities atthe different coregulator levels are indicated below the respective solvent columns. Columns represent the mean of at least threeexperiments, and bars indicate SD.




ZK168281 (1 �M) resulted only in less than a 4-foldinduction (lane 4). This very low residual agonisticactivity of high concentration of ZK168281 were ob-served already previously (26). The combined applica-tion of agonist and antagonist led to a 16.7-fold induc-tion (lane 3). The overexpression of TIF2 resulted in asignificant increase of the basal level (3.8-fold, lane 5)and subsequently a less prominent increase of ago-nist-stimulated values, so that only an approximately15-fold induction was observed (lane 6). CoA overex-pression increased the response to the antagonist(4.9-fold, lane 8), which is known from a previousreport (32). However, the combined application of ag-onist and antagonist did not provide significantlyhigher induction (6.5-fold, lane 7) than agonist aloneand clearly less than in case of endogenous CoA con-centrations. The overexpression of NCoR reduced thebasal activity by 40% (lane 9). The effects of agonistand antagonist alone or in combination were alsoblunted (34.3-, 3.5-, and 2-fold, respectively, lanes10–12).

Human CAR showed a high basal activity on thePBREM (Fig. 3B, lane 1), which could only be induced2.5-fold by CITCO (lane 2) and was reduced by 40%with clotrimazole. TIF2 overexpression had only minoreffects on the basal activity and ligand response ofhuman CAR (lanes 4–6). In contrast, NCoR overex-pression clearly blunted the responsiveness of thereceptor. In this respect, the basal activity was signif-icantly reduced to 30% of the basal activity (lane 7),the induction by CITCO was only 1.6-fold (lane 8) andin the presence of clotrimazole still 90% of basal ac-tivity level was observed (lane 9). The response patternof mouse CAR (Fig. 3C) was similar to that of humanCAR, but the mouse-specific ligands had more prom-inent effects than that of human CAR.

Mouse CAR also showed high basal activity on thePBREM (lane 1), which was induced 3-fold by TC-BOBOP (lane 2) and reduced by 70% with androstanol(lane 3). CoA protein overexpression slightly increasedthe basal activity (1.4-fold, lane 4), reduced the re-sponse to the agonist (2.6-fold induction, lane 5) andto the antagonist (still 70% of basal activity, lane 6).Also with mouse CAR the overexpression of CoR pro-tein showed more prominent effects than the overex-pression of CoA proteins. The basal activity was re-duced significantly by 50% (lane 7), the response toTCPOBOP lowered to a 2.2-fold induction (lane 8) andandrostanol application reached 40% of the basal ac-tivity (lane 9). Taken together, these data indicate thatVDR mediates low basal activity and high agonist in-ducibility, whereas the two CARs show high basalactivity and only moderate inducibility. The VDR an-tagonist showed the known residual agonist activity(32) and could reduce at equimolar concentrationseffectively agonist-induced gene activity, whereas theinverse agonists of the two CARs reduced the highbasal activity. CoA protein overexpression signifi-cantly increased the basal activity and antagonist re-sponse of VDR and reduced its agonist inducibility but

had only minor effects on the two CARs. In contrast,the basal activity and the ligand responsiveness of allthree NRs were significantly reduced by CoR proteinoverexpression.

Modeling of CoR Interactions of VDR and CAR

Because CoRs seem to have a significant effect on theligand response and basal activity of both endocrineand adopted orphan NRs, we next investigated in thestructural determinants of the interactions of the LBDof VDR and the two CARs with CoR in the presence ofantagonist or inverse agonist. The complexes of hu-man VDR-LBD with 1�,25(OH)2D3 and CoA peptide(Fig. 4A, top), human CAR-LBD with CITCO and CoApeptide (Fig. 4B, top) and mouse CAR-LBD withTCPOBOP and CoA peptide (Fig. 4C, top) were mod-eled on the basis of the crystal structure and ourprevious MD simulations of the human VDR-LBD (26,33) and on the recently solved x-ray structures ofhuman and mouse CAR (34–36). In parallel, we dockedthe inverse agonists clotrimazole and androstanol tothe structures of human and mouse CAR, respectively,whereas for the ZK168281-bound VDR-LBD, we al-ready had a structure from a previous MD simulationstudy (26). To each of the three LBDs, we docked apeptide representing the amino acids 2275–2291 ofthe second RID of NCoR and performed MD simula-tions. The resulting structures (Fig. 4) represent theaverage of the last 50 psec of the MD simulations. Thedetailed views on these structures indicate interac-tions of helices 3, 4, and 12 of the NR-LBDs with theCoA- and CoR-RID peptide. The most remarkable andconsistent observation of the three CoR-NR modelstructures was that helix 12 seems to be not flexiblebut takes a stabilized position. This is also visible in theProtein Data Bank (PDB) file of the cocrystal of PPAR�with CoR-RID peptide (1KKQ), although this was notdiscussed in the respective publication (37).

In the VDR-LBD-ZK168281 complex, V418 of helix12 interacts with I2280 of NCoR and S427 of the Cterminus makes a backbone contact to F2289 ofNCoR (Fig. 4A). In a similar way, helix 12 of CAR isstabilized by an interaction between the positivelycharged K2283 of the CoR-RID peptide and L343 andL353 of human and mouse CAR, respectively (Fig. 5, Band C). In addition, K264 of helix 4 in VDR and thehomologous residues K195 and K205 in human andmouse CAR, respectively, form a salt bridge withE2278 of NCoR. Moreover, K246 of helix 3 in VDR andits homologous residues K177 and K187 in human andmouse CAR, respectively, contact the CoR-RID pep-tide at the backbone of L2285. Additionally, K246 alsointeracts with the backbone of the CoR residue A2284.The antagonist ZK168281 contacts H397 in helix 11,but the side chains of the homologous amino acidsY326 and Y336 in human and mouse CAR, respec-tively, have shifted their orientation (compare Fig. 4, Band C, top and center) and do not contact the two




inverse agonists clotrimazole and androstanol. Thestructures of agonist- and antagonist-bound CARdemonstrate that all four ligands are relatively smalland do neither contact helix 12 nor the coregulatorRID, i.e. they are not directly involved in the attractionof CoA or CoR protein. In contrast, the long side chainof ZK168281 is directly contacting NCoR at positionL2277 (Fig. 4A, center). In summary, the models ofVDR and both CARs with antagonist or inverse agonistand CoR-RID peptide suggest that both type of li-gands stabilize helix 12 in a position that allows sev-eral salt bridge and hydrogen bond-based interactionsbetween residues of helices 3, 4, and 12 of the recep-tors and the backbone and side chains of their partneramino acids in the NCoR RID. However, the antagonistis contacting both helix 11 and the CoR-RID, which isnot observed with the smaller inverse agonists.

Critical Amino Acids of VDR Ligand-EnhancedInteraction with CoRs

To challenge the above-described models, for each ofthe three receptors a series of point mutants of com-parable residues was created and tested by supershiftassays. The assays were performed with GST-NCoRin the presence of agonist and antagonist/inverse ag-onist (Fig. 5). GST alone and GST-TIF2 without ligandwere used as controls. As already shown in Fig. 2A, inthe absence of ligand VDR-RXR heterodimers failed tointeract with either TIF2 or NCoR (Fig. 5A, lanes 2 and3), whereas 1�,25(OH)2D3 and ZK168281 inducedfaint and strong interactions with NCoR, respectively(lanes 4 and 5). The mutants K246A, K264A, H397A,V418A, S427A, and N424* (resulting in a truncatedhelix 12) did not induce any ligand-independent inter-

Fig. 4. MD Simulations of Coregulator Interactions of VDR and CAROn the basis of the crystal structure and our previous MD simulations of the human VDR-LBD (26, 33) and recently solved x-ray

structures of human and mouse CAR (34–36) supported by information of the rat VDR-CoA peptide complex (55) and the humanPPAR�-CoR peptide complex (37), MD simulations were performed with the human VDR-LBD in complex with 1�,25(OH)2D3 andCoA peptide (A, top) or with ZK168281 and CoR peptide (A, center), with the human CAR-LBD in complex with CITCO and CoApeptide (B, top) or clotrimazole and CoR peptide (B, center) and mouse CAR-LBD in complex with TCPOBP and CoA peptide (C,top) or androstanol and CoR peptide (C, center). Only helices 3, 4, 11, and 12 (blue) and the side chains of the most importantamino acids are shown. The CoA-RID is shown in green and the CoR-RID peptide in red. Dashed lines indicate interactions witha distance below 3.4 Å. The interactions of the CoR-RID peptide are schematically depicted below each structure. Dashed lineswith horizontal bars at their end symbolize backbone interaction. The core NR interaction motif is underlayed in dark red; redindicates negatively charged amino acids and blue positively charged residues.




action of the receptor with CoA or CoR protein (lanes7, 8, 12, 13, 17, 18, 22, 23, 27, 28, 32, and 33), butreduced the antagonist-induced interaction withNCoR by approximately 25% in case of K246A,V418A, S427A, and N424* (lanes 10, 25, 30, and 35),50% in case of H397A (lane 20) and completely (by100%) in case of K264A (lane 15). The weak,1�,25(OH)2D3-induced interaction of VDR with NCoRwas abolished in case of K246A, K264A, H397A,V418A, and S427A (lanes 9, 14, 19, 24, and 29) andsignificantly increased in case of N424* (lane 34).Taken together, for the antagonist-triggered interac-tion of VDR with NCoR K264 is of central importanceand also the indirect effect of H397 is critical, whereasthe stabilization of the helix 12-NCoR contact via theresidues V418 and S427 and the helix 3-NCoR inter-action via K246 seem to have minor impact.

Critical Amino Acids of Human CAR for Ligand-Enhanced Interaction with CoR

The ligand-independent coregulator interaction of hu-man CAR (Fig. 5B) and mouse CAR (Fig. 5C) made thepattern of the homologous mutations more complex.The level of interaction of human CAR with TIF2 in theabsence of ligand was not significantly affected by themutant C347A (Fig. 5B, lane 27) and was reduced byapproximately 35% with Y326A and the extension ofhelix 12 by three amino acids (lanes 17 and 32), by50% with K177A and L343A (lanes 7 and 22) and by90% with K195A (lane 12). The ligand-independentinteraction of human CAR with TIF2 was approxi-mately doubled compared with NCoR (compare lanes2 and 3), but the latter was slightly increased withK177A (lane 8), not affected with L343A (lane 23) andreduced by approximately 40% with C347A (lane 28)and by more than 70% with K195A, Y326A and theextension of helix 12 (lanes 13, 18, and 33). Interest-ingly, the CITCO-induced interaction of human CARwith CoR protein was increased by 10–20% withK177A and L343A (lanes 9 and 24) and reduced byapproximately 25% with C347A and the extension ofhelix 12 (lanes 29 and 34) and by more than 80% withK195A and Y326A (lanes 14 and 19). Although CITCOinduced the interaction of human CAR with NCoR by150%, clotrimazole application reduced it by 35%(compare lanes 4 and 5). This ratio between the CITCOand clotrimazole effect remained approximately thesame with K177A, L343A, and C347A (compare lanes9 with 10, 24 with 25, and 29 with 30), whereas, com-bined with a low basal level, it became nearly equalwith K195A and Y326A (compare lanes 14 with 15 and19 with 20). The only exception was the extension of

Fig. 5. Critical Amino Acids of VDR and CAR for Ligand-Triggered Interaction with CoRs

Combined gel shift and supershift experiments were per-formed with equal amounts of in vitro-translated wild-type ormutant human VDR (A), human CAR (B), or mouse CAR (C),RXR protein, and [32P]-labeled DR3- (A) or DR4-type RE (Band C). VDR-RXR and CAR-RXR heterodimers were preincu-bated with solvent, 100 nM 1�,25(OH)2D3, or 1 �M ZK168281(A); 10 �M CITCO or clotrimazole (B); and 10 �M TCPOBOP orandrostanol (C) as indicated. Equal amounts of bacteriallyexpressed wild-type GST (�), GST-TIF2 or GST-NCoR werethen added. Protein-DNA complexes were resolved from freeprobe through 8% nondenaturing polyacrylamide gels. Therelative amounts of supershifted complexes were quantifiedusing a Fuji FLA3000 reader. Representative gels are shown.

NS, Nonspecific complexes. Columns represent the mean ofat least three experiments, and bars indicate standard devi-ations. Two-tailed, paired Student’s t test was performed,and P values were calculated in reference to the interaction ofwild-type receptor (*, P � 0.05; **, P � 0.01; ***, P � 0.001).




helix 12, which doubled the ratio of the CITCO- andthe clotrimazole-mediated interaction of human CARwith NCoR from approximately 4–8 (compare lanes 34and 35). In summary, residues K195 and Y326 and ashort helix 12 seem to be critical for both the direct andindirect human CAR and NCoR complex stabilizationand the effect of the antagonist clotrimazole. More-over, K195 and Y326, but not the length of helix 12, areimportant for CoR recruitment via CITCO.

Critical Amino Acids of Mouse CAR for Ligand-Enhanced Interaction with CoR

In the absence of ligand, the interaction of mouse CARwith TIF2 was not significantly affected by the exten-sion of helix 12 by three amino acids (Fig. 5C, lane 32).However, it was reduced by 40–60% with K187A,K205A, and Y336A (lanes 7, 12, and 17) and abolishedentirely with L353A and C357A (lanes 22 and 27). K187is part of the charge-clamp and its respective mutanthas already been tested in our own (38, 39) and othergroups (40). The strength of the effects of this mutantis inversely correlated with the concentration of thebacterially expressed CoA protein, which was rela-tively high in this study. The ligand-independent inter-action of mouse CAR with TIF2 was 30% higher thanwith NCoR (compare lane 2 and 3). The latter was notaffected by L353A and reduced by 10–30% withK187A and C357A (lanes 8 and 28), by approximately50% with K205A and the extension of helix 12 (lanes13 and 33) and by 80% with Y336A (lane 18). Theagonist TCPOBOP reduced the basal interaction ofmouse CAR with NCoR by 35% (compare lanes 3 and4), but this effect was decreased to approximately25% with K187A (compare lanes 8 with 9) and bluntedwith K205A, Y336A, L353A, C357A, and the extensionof helix 12 (compare lanes 13 with 14, 18 with 19, 23with 24, 28 with 29, and 33 with 34). The antagonistandrostanol induced the interaction of mouse CARwith NCoR by approximately 80% (compare lanes 3and 5). This effect was not affected with the mutationK187A (compare lanes 8 and 10), but abolished withthe five other mutants (compare lanes 13 with 15, 18with 20, 23 with 25, 28 with 30, and 33 with 35). Takentogether, the lysine in helix 4 and the tyrosine in helix11, K205 and Y336, are the most important residuesfor the ligand-enhanced interaction of mouse CARwith CoR protein. This is similar to the findings withhuman CAR.

Functional Analysis of Critical Amino Acids ofVDR and Human and Mouse CAR

To analyze the impact of critical amino acids on theagonist and antagonist/inverse agonist responsive-ness of VDR and the two CARs, we performed reportergene assays in transiently transfected MCF-7 cells(Table 1) using the same experimental conditions as inFig. 3. Concerning modulation of basal activities andagonist inducibilities, we obtained essentially the

same results as in our previous studies on the criticalrole of helix 12 on the constitutive activity and CoArecruitment of VDR and human and mouse CAR (38,39). More interesting is the observation that ZK168281showed no agonistic potential with the VDR mutantsK246A, K264A, H397A, and V418A but lost most of itsantagonistic potential (from 66% with wild-type VDRdown to 12–25%). In contrast, the mutant S427Ashowed a profile similar to wild-type VDR. Compara-bly, with the homologous human CAR mutants K177A,K195A, Y326A, and L343A and their mouse orthologsK187A, K205A, Y336A, and L353A clotrimazole andandrostanol lost most (K177A/K187A) or even all oftheir inverse agonistic potential. In comparison, theinverse agonists were still functional with C347A andC357A. In conclusion, these data confirm the centralrole of the lysine in helix 4 (K264, K195, and K205).However, most of the tested residues have also animpact on the basal activity and agonist inducibility ofthe receptors, allowing no trivial functional distinctionof the effects of antagonists and inverse agonists.

Critical Amino Acids of the CoR NRInteraction Motif

The mutational analysis of all three NRs (Fig. 5) andtheir functional test (Table 1) suggested that the con-served lysine in helix 4 (K264, K195, and K205, re-spectively) is the most important residue in the directcontact of the NR with CoR-RID. Therefore, we mu-tated in each of the three receptors the positivelycharged lysine into a negatively charged glutamate. Inparallel, we performed similar charge inversion mu-tagenesis of the CoR-RID, where we mutated E2278into a lysine and K2283 into a glutamate. Differentcombinations of these receptor and NCoR mutantswere assessed in supershift assays (Fig. 6). TheZK168281-induced interaction of VDR-RXR het-erodimers with NCoR was reduced by 70% when wild-type CoR protein was replaced with its E2278K mutant(Fig. 6A, compare lanes 2 and 3). Although the K2283Emutant resulted in some 40% reduction of CoR inter-action (lane 4), this NCoR residue appears not to con-tact VDR directly (see Fig. 4A). As expected, theK264E mutant of VDR completely abolished the CoRcontact (lane 6). However, in combination with theNCoR mutant E2278K, 70% of the interaction level ofboth wild-type proteins was restored (lane 7). Compa-rable observations were made with human and mouseCAR. The E2278K mutant of NCoR reduced the inter-action with human CAR by 50% and that with mouseCAR even by 70% (Figs. 6, B and C, lane 3). TheK195E and K205E mutants of human and mouse CAR,respectively, blunted the interaction of the receptorwith the CoR protein (lane 6) and the combination ofmutated receptor with mutated NCoR restored 90% ofthe interaction between CoR and human CAR. In caseof mouse CAR, even a level of 160% was reached(lane 7). A test of the NCoR mutant K2283E with thetwo CARs showed that it also reduced the receptor-




CoR interaction for both receptors by 55–70% (lane 4),but the combination with mutated receptor could notrestore the interaction (data not shown). In summary,the salt bridge between E2278 of NCoR and the lysineof helix 4 of VDR and the two CARs seems to be themajor direct fixation point between receptor and CoR;however, K2283 is also critical for the stability of helix12 and the effective interaction of the three NRs withNCoR.

DISCUSSION

Antagonists and Inverse Agonists Have SimilarFunctional Profiles

In this study, we demonstrate that despite their diver-gent structure, the VDR antagonist ZK168281, and theCAR inverse agonists clotrimazole and androstanolhave comparable effects on the coregulator interac-tion of their receptors. Therefore, antagonists and in-verse agonists may be taken into the same ligandclass that can be clearly distinguished from the classof agonists, even if the example of CITCO suggestssome overlap between both ligand families exist. Suchdual effects are also known for other CAR ligands,such as estrogens (41). According to present knowl-

edge, in very most cases the large ligand-bindingpockets of adopted orphan NRs can accommodateonly one ligand molecule at the time. This suggeststhat agonists and antagonists/inverse agonists com-pete for binding to the pocket. Therefore, effectiveagonist/antagonist or agonist/inverse agonist ligandcouples should have comparable affinities for the NR-LBD. Mutational analyses coupled with molecularmodeling of human and mouse CAR (42, 43) as well asthe crystal structures of both receptors (34–36) haveindicated that the ligand-binding pocket of theadopted orphan NR has a volume of approximately600 Å3 and is formed by some 30 amino acid residuesof in helices 5–7 and 10. The ligand-binding pocket ofVDR is as large as that of CAR (33) but is bound byZK168281 and 1�,25(OH)2D3 with the same high af-finity [Kd � 0.1 nM (23)]. The stricter structural require-ment of VDR ligands is related to their contact withH397 in helix 11 (Fig. 4A). In addition, the extendedside chain of ZK168281 both pushes helix 12 from itsagonistic position (26) and directly contacts and sta-bilizes the CoR-RID. In contrast, clotrimazole and an-drostanol do not contact the H397 homologs Y326and Y336 in human and mouse CAR and also do notinteract with the CoR-RID (Fig. 4, B and C). Theseproperties make ZK168281 a more effective regulatorof the coregulator exchange than clotrimazole andandrostanol.

Table 1. Functional Test of Critical Residues of VDR, Human and Mouse CAR MCF-7 Cells

Receptor Mutant Relative BasalActivity

Fold Induction byAgonist Treatment

Fold Induction byAgonist/Antagonist

Cotreatment

Fold Induction byAntagonist or

Inverse AgonistTreatment

VDR wt 100 48.9 17.4 3.8K246A 81 5.2 3.3 1.2K264A 69 31.6 25.5 1.1H397A 98 10.5 8 1.5V418A 103 7.2 5.8 1.0S427A 106 61.2 18.3 3.2

hCAR wt 100 2.6 0.5K177A 24 1.1 0.8K195A 46 4.8 1.5Y326A 43 2.4 1.1L343A 31 0.9 1.2C347A 40 3.8 0.5

mCAR wt 100 3.1 0.3K187A 19 1.3 0.7K205A 26 6.7 1.4Y336A 51 4.0 1.2L353A 11 1.9 1.3C357A 10 5.9 0.4

Reporter gene assays were performed with extracts from MCF-7 cells that were transiently transfected with a luciferase reporterconstruct containing four copies of the DR3-type RE of the rat ANF gene (for VDR) or the PBREM of the mouse CYP2B10 gene(for human and mouse CAR) and indicated expression vectors for wild type and mutant VDR, human and mouse CAR. Cells weretreated for 16 h with solvent, 10 nM 1�,25(OH)2D3 or 1 �M ZK168281 (alone and in combination), 1 �M CITCO or 1 �M clotrimazoleand 1 �M TCPOBOP or 1 �M androstanol. Luciferase activities were measured and fold inductions were calculated to in relationto respective solvent controls. Relative basal activity of the mutants was calculated in relation to the activity of wild type in theabsence of ligand. The values represent the mean of at least three experiments and standard deviations were below 15%.




The CoA/CoR Ratio May Be the Main Parameterfor Changing the Activity State of a NR

Common in the molecular mechanism of antagonistsand inverse agonists is their ability to promote theinteraction of their receptor with CoR proteins and toinhibit contact with CoA proteins. Classical endocrineNRs, such as VDR, show very low basal activity in theabsence of ligand, whereas adopted orphan NRs,such as CAR and PPARs (Molnar, F., and C. Carlberg,unpublished results), display significant amount ofconstitutive activity. Therefore, antagonists act after apreceding activation of their target receptor by anagonist, whereas inverse agonists antagonize the li-gand-independent high basal activity of their receptor.The ligands that were used in this study demonstratethat there are transitions between these two extremestates and that the potential of the ligands to perform

each of the two functions is molecule specific. Forexample, although clotrimazole prevents CoA contactof human CAR, it only weakly recruits CoR binding. Incontrast, CITCO supports both CoA and CoR interac-tion of human CAR. In particular, the latter exampleraises the question, whether the view that NR ligandsactively recruit coregulators of one variety or another isvalid for all compounds. Alternatively, it may be thatthe relatively large ligand-binding pocket of adoptedorphan receptors allows the binding of a small ligand,such as CITCO, without inducing significant changesin its conformation. Human and mouse CAR are ableto bind in their apo-state both CoA and CoR proteins.This suggests that the ratio between CoA and CoRproteins may be the main parameter for changing theconformation and activity state of a NR, such as CAR,and that the effect of the ligand may be of secondaryimportance.

Fig. 6. Critical Amino Acids of the CoR NR Interaction MotifSupershift experiments were performed with equal amounts of in vitro-translated wild-type or mutant human VDR (A), human

CAR (B), or mouse CAR (C), RXR protein, and [32P]-labeled DR3- (A) or DR4-type RE (B and C). VDR-RXR and CAR-RXRheterodimers were preincubated with solvent, 1 �M ZK168281 (A), 10 �M clotrimazole (B) and 10 �M androstanol (C). Equalamounts of bacterially expressed wild-type GST (�), wild type of mutant GST-NCoR, were then added. Protein-DNA complexeswere resolved from free probe through 8% nondenaturing polyacrylamide gels. The relative amounts of supershifted complexeswere quantified using a Fuji FLA3000 reader. Representative gels are shown. NS, Nonspecific complexes. Columns represent themean of at least three experiments, and bars indicate SD. Two-tailed, paired Student’s t test was performed, and P values werecalculated in reference to the interaction of wild-type receptor with wild-type NCoR (*, P � 0.05; **, P � 0.01).




Helix 12 Takes a Stabilized Position duringCoR Interaction

According to the mouse-trap model the helix 12 of aNR has a central role in determining the agonist-trig-gered interaction of NR-LBDs with CoA proteins (10),and this has been proven extensively in many studies.For apo-NRs, which should be able to interact withCoR proteins, the model suggests free movement ofhelix 12 as observed in the RXR-LBD crystal structure(44). Amino acids that were identified in this study asbeing important for CoR interaction, such as the con-served lysines in helices 3 and 4, have already beendescribed to be critical for the CoA contacts of therespective receptors (38). Although we did not directlyaddress the hydrophobic residues on the surface ofthe NR-LBDs and the CoR-RID, our investigation ofcharged and polar amino acids suggests that bothCoA and CoR proteins contact the same surface re-gion on NR-LBDs. However, the larger RIDs of CoRscompared with CoAs (9) make a move of helix 12necessary for a coregulator exchange. In fact, in vitroa complete truncation of helix 12 is favorable for CoRinteraction (Ref. 40 and data not shown).

In living cells, NRs have to deal with a transientlydispensable helix 12, although CoR binding. There-fore, our model of a stabilized position of helix 12 incase of CoR contacts (Fig. 4) represents energeticallya more favorable state than that of a free-floating helixas suggested in the mouse-trap model (10). Our model(Fig. 4) suggests that the contact between the CoR-RID and helix 12 stabilizes the position of the latterhelix. This means that, in contrast to CoA interaction,the stabilized position of helix 12 does not supportCoR interaction, but that the CoR helps the receptor toget its flexible helix under control. This would explainwhy the mutagenesis of the contact points betweenhelix 12 and the CoR-RID has less consequences thanthe mutagenesis of the salt bridge between the con-served lysine in helix 4 and E2278 in the CoR-RID. Forthe CoR-NR-LBD interaction, the latter salt bridgeseems to have a similar impact than the charge clampbetween the conserved lysine in helix 3 and the glu-tamate in helix 12 has for CoA interaction. However, incontrast to the glutamate in helix 12, which changes itsdistance in relation to its partner lysine with everyligand-triggered move of helix 12, the lysine in helix 4has a stabilized, ligand-independent position. This ex-plains why in the absence of ligand the LBDs of en-docrine NRs favor CoR interaction. Adopted orphanNRs, which display constitutively activity seem to bean exception because their mechanisms of stabilizinghelix 12 in the absence of ligand (38) prevent access ofCoR-RID to their interface on the surface of the LBD.

Conclusion

The ligand-triggered dynamic exchange of CoA andCoR proteins binding to NRs is the molecular basis ofthe action of agonists, inverse agonists, and antago-

nists. The structural determinants of the antagonist-and inverse agonist-triggered interaction VDR and hu-man and mouse CAR with the second RID of NCoR ledto the main conclusion of this study that antagonists ofendocrine NRs and inverse agonists of adopted NRshave a comparable functional profile. A second, im-portant finding of this study is the stabilization of helix12 in all three receptors by direct contacts with resi-dues of the CoR. However, in contrast to the CoAinteraction, which is dependent of a fixed position ofhelix 12, the helix is not needed for CoR interaction. Infact, helix 12 has to move from its position in theagonistic LBD conformation to a perpendicular posi-tion, where it does not disturb the contact betweenLBD and CoR. Therefore, fixation of helix 12 in theantagonistic/inverse agonistic conformation seems tobe only energetically favorable but of no specific func-tion. The comparable molecular mechanisms that ex-plain the comparable functional profile of antagonistand inverse agonists are likely to be extended fromVDR and CAR to other members of the NR superfamilyand may lead to the design of even more effectiveligands.

MATERIALS AND METHODS

Compounds

1�,25(OH)2D3 was kindly provided by Dr. L. Binderup (LeoPharma, Ballerup, Denmark), ZK168281 was a gift from Dr. A.Steinmeyer (Schering, Berlin, Germany) and TCPOBOP wassynthesized and purified according to Honkakoski et al. (45).CITCO was obtained from Biomol (Copenhagen, Denmark),androstanol from Steraloids (Newport, RI) and clotrimazolefrom Sigma-Aldrich (St. Louis, MO). The two-dimensionalstructures of the ligands are shown in Fig. 1D. The VDRligands were dissolved in propan-2-ol, whereas the othercompounds were dissolved in dimethylsulfoxide. Further di-lutions were made in either dimethylsulfoxide (for in vitroexperiments) or ethanol (for cell culture experiments).

DNA Constructs

Protein Expression Vectors. Full-length cDNAs for humanVDR (46), human CAR (47), human RXR� (48) and humanTIF2 (49) were subcloned into the T7/SV40 promoter-drivenpSG5 expression vector (Stratagene, La Jolla, CA). The full-length cDNAs for mouse CAR (50) and mouse NCoR (51)were subcloned into the T7/CMV promoter-driven pCMX ex-pression vector. The amino acid substitution mutants of hu-man VDR, human CAR, and mouse CAR were generatedusing the QuikChange point mutagenesis kit (Stratagene) andconfirmed by sequencing. The truncation of helix 12 of VDRby four amino acids (N424*) was created by mutating triplet424 into a stop codon. The extensions of helix 12 in humanand mouse CAR by three amino acids were generated by adouble mutant that converted the original stop codon into acoding triplet and the third downstream triplet into a stopcodon. The same constructs were used for both T7 RNApolymerase-driven in vitro transcription/translation of the re-spective cDNAs and for viral promoter-driven overexpressionof the respective proteins in mammalian cells.GST Fusion Protein Constructs. Critical domains of humanTIF2 (spanning from amino acids 646–926 including three




RIDs) (49) and mouse NCoR (spanning from amino acids2218–2453 including the second RID) (51) were subclonedinto the GST fusion vector pGEX (Amersham Pharmacia,Uppsala, Sweden).Reporter Gene Constructs. The luciferase reporter gene,fused with the thymidine kinase minimal promoter, was drivenby four copies of the DR3-type 1�,25(OH)2D3 response ele-ment of the rat ANF gene promoter (30) or one copy of thePBREM of the mouse CYP2B10 gene promoter (containingtwo DR4-type REs) (19).

In Vitro Translation and Bacterial Overexpressionof Proteins

In vitro-translated wild-type or mutated human VDR and hu-man and mouse CAR proteins were generated by coupled invitro transcription/translation using rabbit reticulocyte lysateas recommended by the supplier (Promega, Madison, WI).Protein batches were quantified by test translations in thepresence of [35S]-methionine. The specific concentration ofthe receptor proteins was adjusted to approximately 4 ng/�lafter taking the individual number of methionine residues perprotein into account. Bacterial overexpression of GST-TIF2,wild-type, and mutant GST-NCoR or GST alone was ob-tained from the Escherichia coli BL21(DE3)pLysS strain(Stratagene) containing the respective expression plasmids.GST-TIF2 and GST protein expression were stimulated with0.25 mM isopropyl-�-D-thio-galactopyranoside for 3 h at 37 Cand GST-NCoR expression was induced with 1.25 mM iso-propyl-�-D-thio-galactopyranoside for 5 h at 25 C. The fusionproteins were purified and immobilized by glutathione-Sepharose 4B beads (Amersham Pharmacia) according tothe manufacturer’s protocol. For gel shift experiments, thefusion proteins were eluted by glutathione.

GST-Pull-Down Assays

GST-pull-down assays were performed with 50 �l of a 50%Sepharose bead slurry of GST, GST-TIF2, or GST-NCoR(preblocked with 1 �g/�l BSA) and 20 ng in vitro-translated,[35S]-labeled NRs in the presence or absence of their respec-tive ligands. Proteins were incubated in immunoprecipitationbuffer [20 mM HEPES (pH 7.9), 200 mM KCl, 1 mM EDTA, 4 mM

MgCl2, 1 mM dithiothretiol, 0.1% Nonidet P-40 and 10%glycerol] for 20 min at 30 C. In vitro-translated proteins thatwere not bound to GST-fusion proteins were washed awaywith immunoprecipitation buffer. GST-fusion protein bound,[35S]-labeled NRs were resolved by electrophoresis through15% sodium dodecyl sulfate-polyacrylamide gels and quan-tified on a FLA3000 reader (Fuji, Tokyo, Japan) using ImageGauge software (Fuji).

Gel Shift and Supershift Assays

Gel shift assays were performed with equal amounts (�10 ng)of the appropriate in vitro-translated protein. The proteinswere incubated for 15 min in a total volume of 20 �l bindingbuffer [10 mM HEPES (pH 7.9), 150 mM KCl, 1 mM dithiothre-tiol, 0.2 �g/�l poly(deoxyinosine-deoxycytosine) and 5%glycerol]. For supershift experiments, 0.4–3 �g of bacteriallyexpressed wild-type or mutant GST fusion proteins (or GSTalone as negative control) were added to the reaction mix-ture. Approximately 1 ng of [32P]-labeled double-strandedoligonucleotides (50,000 cpm) corresponding to one copy ofthe DR3- or DR4-type RE (core sequences are indicated inFigs. 2, 5, and 6) was then added and incubation was con-tinued for 20 min at room temperature. Protein-DNA com-plexes were resolved by electrophoresis through 8% nonde-naturing polyacrylamide gels in 0.5� TBE [45 mM Tris (pH8.3), 45 mM boric acid, 1 mM EDTA] and quantified on aFLA3000 reader using Image Gauge software.

Transfection and Luciferase Reporter Gene Assays

MCF-7 human breast cancer cells were seeded into six-wellplates (105 cells/ml) and grown overnight in phenol red-freeDMEM supplemented with 5% charcoal-stripped fetal bovineserum. Plasmid DNA containing liposomes were formed byincubating a reporter plasmid and expression vectors forwild-type or mutated human VDR, human CAR, mouse CAR,human TIF2, or mouse NCoR (each 1 �g as indicated) with 10�g N-[1-(2,3-dioleoyloxy)propyl]-N,N,N-trimethylammoniummethylsulfate (DOTAP, Roth, Karlsruhe, Germany) for 15 minat room temperature in a total volume of 100 �l. After dilutionwith 900 �l phenol red-free DMEM, the liposomes wereadded to the cells. Phenol red-free DMEM supplementedwith 500 �l 15% charcoal-stripped fetal bovine serum wasadded 4 h after transfection. At this time, NR ligands orcontrol solvents were also added. The cells were lysed 16 hafter onset of stimulation using the reporter gene lysis buffer(Roche Diagnostics, Mannheim, Germany) and the constantlight signal luciferase reporter gene assay was performed asrecommended by the supplier (Canberra-Packard, Gro-ningen, The Netherlands). The luciferase activities were nor-malized with respect to protein concentration and inductionfactors were calculated as the ratio of luciferase activity ofligand-stimulated cells to that of solvent controls.

Structural Modeling and MD Simulations

The initial coordinates of VDR were obtained from the crystalstructure of the human VDR-LBD-1�,25(OH)2D3 complex[PDB code 1DB1 (33)]. The missing amino acid residues (no.118, 119, 375–377, and 424–427) were built using the Quan-ta98 molecular modeling package (Molecular SimulationsInc., San Diego, CA). The four residues missing from the Cterminus (424–427) were built in an �-helical conformation(� � � 57 °, � � � 47 °). The coordinates of mouse CAR weretaken from its recently solved crystal structure bound to theinverse agonist androstenol [PDB code 1XNX (36)]. Residues344–349 of helix 12, which were missing from the structure,were modeled using the targeted MD method (52). In thismethod, an additional term is added to the energy function ofthe system based on the mass-weighted root mean squaredeviation (RMSD) of a set of atoms in the current structurecompared with a reference structure. The additional energyterm acts as a positional restraint, which forces the currentstructure to move toward a reference structure during a tar-geted MD simulation. Here the crystal structure of mouseCAR in the agonistic conformation [PDB code 1XLS (35)], inwhich the residues 344–349 are present, was used as astarting structure and the mouse CAR-androstenol structurewas the target. In practice, during a 95-psec MD simulation at340 K, the coordinates of residues 326–343 and 350–357 ofthe starting structure were forced using a force constant of 5kcal (mol�1Å�2) to move toward the coordinates of the target.The six residues missing from the mouse CAR-androstenolx-ray structure were allowed to move freely. During the tar-geted MD simulation, the RMSD was linearly decreased from7.4 Å, which is the RMSD of residues 326–343 and 350–357of the starting structures, to 0.0 Å. The slightly increasedtemperature (340 K) was used to speed up the conforma-tional changes taken place in the targeted MD simulation. Theconformation of residues 344–349 obtained from the targetedMD was used to complete the mouse CAR-androstenol x-raystructure. The initial coordinates for residues 103–315 ofhuman CAR were taken from the human CAR/RXR� het-erodimer structure [PDB code 1XVP (34)], whereas the resi-dues 316–348 were built using the modeled mouse CAR.

The helices 12 of VDR and mouse and human CAR wererepositioned on the basis of the crystal structure of the hu-man PPAR�-CoR-RID peptide-GW6471 complex [PDB code1KKQ (37)]. The CoR-RID peptide was build and docked tothe LBDs of human VDR, human CAR, and mouse CAR using




the coordinates of the CoR-RID peptide of the PPAR� struc-ture. Finally, ligands were placed to the ligand-binding sitesof LBD-CoR-RID peptide complexes. ZK168281 was dockedto VDR on the basis of earlier MD simulation results (26) andandrostanol on the basis of the mouse CAR-androstenolx-ray structure (36). Clotrimazole was docked to the ligand-binding site of human CAR, respectively, using the GOLDprotein-ligand docking program (53). For the energy minimi-zations and MD simulations, human VDR, human CAR, andmouse CAR complexes were solvated by 11379, 12256, and11178 TIP3P water molecules in a periodic box of 63 � 69 �91 Å, 62 � 75 � 89 Å and 62 � 76 � 83 Å, respectively. Thewater molecules of the complexes were first energy-mini-mized for 1000 steps, heated to 300 K in 5 psec and equili-brated by 10 psec at a constant temperature of 300 K andpressure of 101,300 Pa. After that, the simulation systemswere minimized for 1000 steps, the temperature of the sys-tems was increased to 300 K in 5 psec and equilibrated for100 psec while keeping the protein backbone atoms (N, C�,C) restrained by an atom-based harmonic potential of 1 kcalmol�1Å�2. The purpose of these simulation steps was toremove atom-atom clashes and let the protein side chainspack efficiently. After that, the restraints were removed and150 psec MD simulations were carried out. In the simulations,the electrostatics were treated using the particle-mesh Ewaldmethod. A time step of 1.5 fsec was used, and bonds involv-ing hydrogen atoms were constrained to their equilibriumlengths using the SHAKE algorithm. The simulations weredone using the AMBER8.0 simulation package (University ofCalifornia, San Francisco, CA) and the parm99 parameter setof AMBER. The parameters of the ligands were generatedwith the Antechamber suite of AMBER8.0 in conjunction withthe general amber force field. The atomic point charges of theligands were calculated with the two-stage RESP (54) fit atthe HF/6–31G* level using ligand geometries optimized withthe semiempirical PM3 method using the Gaussian03 pro-gram (Gaussian Inc., Pittsburgh, PA).

Acknowledgments

We would like to thank Drs. S. Kliewer (SouthwesternMedical School, Dallas, TX) for CAR expression vectors, L.Binderup (LEO Pharma, Ballerup, Denmark) for1�,25(OH)2D3, A. Steinmeyer (Schering AG, Berlin, Germany)for ZK168281, P. Honkakoski (University of Kuopio) forTCPOBOP and discussions, and T. W. Dunlop for criticalreading of the manuscript.

Received December 24, 2004. Accepted May 3, 2005.Address all correspondence and requests for reprints to:

Professor Carsten Carlberg, Department of Biochemistry,University of Kuopio, P.O. Box 1627, FIN-70211 Kuopio,Finland. E-mail: [email protected].

This work was supported by the Academy of Finland(Grants 50319, 50331, and 203926).

REFERENCES

1. McKenna NJ, O’Malley BW 2002 Combinatorial controlof gene expression by nuclear receptors and coregula-tors. Cell 108:465–474

2. Chawla A, Repa JJ, Evans RM, Mangelsdorf DJ 2001Nuclear receptors and lipid physiology: opening theX-files. Science 294:1866–1870

3. Willson TM, Moore JT 2002 Minireview: genomics versusorphan nuclear receptors-a half-time report. Mol Endo-crinol 16:1135–1144

4. Mangelsdorf DJ, Thummel C, Beato M, Herrlich P,Schutz G, Umesono K, Blumberg B, Kastner P, Mark M,Chambon P, Evans RM 1995 The nuclear receptorsuperfamily: the second decade. Cell 83:835–839

5. Leo C, Chen JD 2000 The SRC family of nuclear receptorcoactivators. Gene 245:1–11

6. Polly P, Herdick M, Moehren U, Baniahmad A, Heinzel T,Carlberg C 2000 VDR-Alien: a novel, DNA-selective vita-min D3 receptor-corepressor partnership. FASEB J 14:1455–1463

7. Heery DM, Kalkhoven E, Hoare S, Parker MG 1997 Asignature motif in transcriptional co-activators mediatesbinding to nuclear receptors. Nature 387:733–736

8. Hu X, Lazar MA 1999 The CoRNR motif controls therecruitment of corepressors by nuclear hormone recep-tors. Nature 402:93–96

9. Li Y, Lambert MH, Xu HE 2003 Activation of nuclearreceptors: a perspective from structural genomics.Structure 11:741–746

10. Moras D, Gronemeyer H 1998 The nuclear receptor li-gand-binding domain: structure and function. Curr OpinCell Biol 10:384–391

11. Suda T, Ueno Y, Fujii K, Shinki T 2003 Vitamin D andbone. J Cell Biochem 88:259–266

12. Mørk Hansen C, Binderup L, Hamberg KJ, Carlberg C2001 Vitamin D and cancer: effects of 1,25(OH)2D3 andits analogs on growth control and tumorigenesis. FrontBiosci 6:D820–D848

13. Moore LB, Parks DJ, Jones SA, Bledsoe RK, Consler TG,Stimmel JB, Goodwin B, Liddle C, Blanchard SG, WillsonTM, Collins JL, Kliewer SA 2000 Orphan nuclear recep-tors constitutive androstane receptor and pregnane Xreceptor share xenobiotic and steroid ligands. J BiolChem 275:15122–15127

14. Forman BM, Tzameli I, Choi HS, Chen J, Simha D, SeolW, Evans RM, Moore DD 1998 Androstane metabolitesbind to and deactivate the nuclear receptor CAR-�. Na-ture 395:612–615

15. Honkakoski P, Negishi M 2000 Regulation of cytochromeP450 (CYP) genes by nuclear receptors. Biochem J 347:321–337

16. Carlberg C 1995 Mechanisms of nuclear signalling byvitamin D3. Interplay with retinoid and thyroid hormonesignalling. Eur J Biochem 231:517–527

17. Glass CK 1994 Differential recognition of target genes bynuclear receptor monomers, dimers, and heterodimers.Endocr Rev 15:391–407

18. Sueyoshi T, Kawamoto T, Zelko I, Honkakoski P, NegishiM 1999 The repressed nuclear receptor CAR responds tophenobarbital in activating the human CYP2B6 gene.J Biol Chem 274:6043–6046

19. Honkakoski P, Zelko I, Sueyoshi T, Negishi M 1998 Thenuclear orphan receptor CAR-retinoid X receptor het-erodimer activates the phenobarbital-responsive en-hancer module of the CYP2B gene. Mol Cell Biol 18:5652–5658

20. Maglich JM, Parks DJ, Moore LB, Collins JL, Goodwin B,Billin AN, Stoltz CA, Kliewer SA, Lambert MH, WillsonTM, Moore JT 2003 Identification of a novel human con-stitutive androstane receptor (CAR) agonist and its use inthe identification of CAR target genes. J Biol Chem 278:17277–17283

21. Honkakoski P, Moore R, Washburn KA, Negishi M 1998Activation by diverse xenochemicals of the 51-base pairphenobarbital-responsive enhancer module in theCYP2B10 gene. Mol Pharmacol 53:597–601

22. Tzameli I, Moore DD 2001 Role reversal: new insightsfrom new ligands for the xenobiotic receptor CAR.Trends Endocrinol Metab 12:7–10

23. Bury Y, Steinmeyer A, Carlberg C 2000 Structure activityrelationship of carboxylic ester antagonists of the vitaminD3 receptor. Mol Pharmacol 58:1067–1074




24. Macias-Gonzalez M, Samenfeld P, Perakyla M, CarlbergC 2003 Corepressor excess shifts the two-side chainvitamin D analog Gemini from an agonist to an inverseagonist of the vitamin D receptor. Mol Endocrinol 17:2028–2038

25. Toell A, Gonzalez MM, Ruf D, Steinmeyer A, Ishizuka S,Carlberg C 2001 Different molecular mechanisms of vi-tamin D3 receptor antagonists. Mol Pharmacol 59:1478–1485

26. Vaisanen S, Perakyla M, Karkkainen JI, Steinmeyer A,Carlberg C 2002 Critical role of helix 12 of the vitamin D3receptor for the partial agonism of carboxylic ester an-tagonists. J Mol Biol 315:229–238

27. Carlberg C 2003 Molecular basis of the selective activityof vitamin D analogues. J Cell Biochem 88:274–281

28. Quack M, Carlberg C 2000 Ligand-triggered stabilizationof vitamin D receptor/retinoid X receptor heterodimerconformations on DR4-type response elements. J MolBiol 296:743–756

29. Frank C, Gonzalez MM, Oinonen C, Dunlop TW, CarlbergC 2003 Characterization of DNA complexes formed bythe nuclear receptor constitutive androstane receptor.J Biol Chem 278:43299–43310

30. Kahlen JP, Carlberg C 1996 Functional characterizationof a 1,25-dihydroxyvitamin D3 receptor binding sitefound in the rat atrial natriuretic factor promoter. Bio-chem Biophys Res Commun 218:882–886

31. Honkakoski P, Negishi M 1997 Characterization of aphenobarbital-responsive enhancer module in mouseP450 Cyp2b10 gene. J Biol Chem 272:14943–14949

32. Herdick M, Steinmeyer A, Carlberg C 2000 Carboxylicester antagonists of 1�,25-dihydroxyvitamin D3 showcell-specific actions. Chem Biol 7:885–894

33. Rochel N, Wurtz JM, Mitschler A, Klaholz B, Moras D2000 Crystal structure of the nuclear receptor for vitaminD bound to its natural ligand. Mol Cell 5:173–179

34. Xu RX, Lambert MH, Wisely BB, Warren EN, Weinert EE,Waitt GM, Williams JD, Collins JL, Moore LB, Willson TM,Moore JT 2004 A structural basis for constitutive activityin the human CAR/RXR� heterodimer. Mol Cell 16:919–928

35. Suino K, Peng L, Reynolds R, Li Y, Cha JY, Repa JJ,Kliewer SA, Xu HE 2004 The nuclear xenobiotic receptorCAR: structural determinants of constitutive activationand heterodimerization. Mol Cell 16:893–905

36. Shan L, Vincent J, Brunzelle JS, Dussault I, Lin M, Ian-culescu I, Sherman MA, Forman BM, Fernandez EJ 2004Structure of the murine constitutive androstane receptorcomplexed to androstenol: a molecular basis for inverseagonism. Mol Cell 16:907–917

37. Xu HE, Stanley TB, Montana VG, Lambert MH, ShearerBG, Cobb JE, McKee DD, Galardi CM, Plunket KD, NolteRT, Parks DJ, Moore JT, Kliewer SA, Willson TM, Stim-mel JB 2002 Structural basis for antagonist-mediatedrecruitment of nuclear co-repressors by PPAR�. Nature415:813–817

38. Frank C, Molnar F, Matilainen M, Lempiainen H, CarlbergC 2004 Agonist-dependent and agonist-independenttransactivations of the human constitutive androstanereceptor are modulated by specific amino acid pairs.J Biol Chem 279:33558–33566

39. Andersin T, Vaisanen S, Carlberg C 2003 The critical roleof carboxy-terminal amino acids in ligand-dependentand -independent transactivation of the constitutive an-drostane receptor. Mol Endocrinol 17:234–246

40. Dussault I, Lin M, Hollister K, Fan M, Termini J, ShermanMA, Forman BM 2002 A structural model of the consti-tutive androstane receptor defines novel interactions thatmediate ligand-independent activity. Mol Cell Biol 22:5270–5280

41. Makinen J, Reinisalo M, Niemi K, Viitala P, Jyrkkarinne J,Chung H, Pelkonen O, Honkakoski P 2003 Dual action ofoestrogens on the mouse constitutive androstane recep-tor. Biochem J 376:465–472

42. Jyrkkarinne J, Makinen J, Gynther J, Savolainen H, PosoA, Honkakoski P 2003 Molecular determinants of steroidinhibition for the mouse constitutive androstane recep-tor. J Med Chem 46:4687–4695

43. Jyrkkarinne J, Windshugel B, Makinen J, Ylisirnio M,Perakyla M, Poso A, Sippl W, Honkakoski P 2005 Aminoacids important for ligand specificity of the human con-stitutive androstane receptor. J Biol Chem 280:5960–5971

44. Bourguet W, Ruff M, Chambon P, Gronemeyer H, MorasD 1995 Crystal structure of the ligand binding domain ofthe human nuclear receptor RXR-�. Nature 375:377–382

45. Honkakoski P, Moore R, Gynther J, Negishi M 1996Characterization of phenobarbital-inducible mouseCyp2b10 gene transcription in primary hepatocytes.J Biol Chem 271:9746–9753

46. Carlberg C, Bendik I, Wyss A, Meier E, Sturzenbecker LJ,Grippo JF, Hunziker W 1993 Two nuclear signalling path-ways for vitamin D. Nature 361:657–660

47. Baes M, Gulick T, Choi H-S, Martinoli MG, Simha D,Moore DD 1994 A new orphan member of the nuclearhormone receptor superfamily that interacts with a sub-set of retinoic acid response elements. Mol Cell Biol14:1544–1552

48. Mangelsdorf DJ, Ong ES, Dyck JA, Evans RM 1990Nuclear receptor that identifies a novel retinoic acid re-sponse pathway. Nature 345:224–229

49. Voegel JJ, Heine MJS, Zechel C, Chambon P, Gronem-eyer H 1996 TIF2, a 160 kDa transcriptional mediator forthe ligand-dependent activation function AF-2 of nuclearreceptors. EMBO J 15:3667–3675

50. Choi HS, Chung M, Tzameli I, Simha D, Lee YK, Seol W,Moore DD 1997 Differential transactivation by two iso-forms of the orphan nuclear hormone receptor CAR.J Biol Chem 272:23565–23571

51. Horlein AJ, Naar AM, Heinzel T, Torchia J, Gloss B,Kurokawa R, Ryan A, Kamei Y, Soderstrom M, Glass CK,Rosenfeld MG 1995 Ligand-independent repression bythe thyroid hormone receptor mediated by a nuclearreceptor co-repressor. Nature 377:397–404

52. Schlitter J, Engels M, Kruger P, Jacoby E, Wollmer A1993 Targeted molecular dynamics simulation of confor-mational change—application to the t-r transition in in-sulin. Mol Sim 10:291–308

53. Verdonk ML, Cole JC, Hartshorn MJ, Murray CW, TaylorRD 2003 Improved protein-ligand docking using GOLD.Proteins 52:609–623

54. Bayly CI, Cieplak P, Cornell WD, Kollman PA 1993 Awell-behaved electrostatic potential based method usingcharge restraints for deriving atomic charges: the RESPmodel. J Phys Chem 97:10269–10280

55. Vanhooke JL, Benning MM, Bauer CB, Pike JW, DeLucaHF 2004 Molecular structure of the rat vitamin D receptorligand binding domain complexed with 2-carbon-substi-tuted vitamin D3 hormone analogues and a LXXLL-con-taining coactivator peptide. Biochemistry 43:4101–4110

Molecular Endocrinology is published monthly by The Endocrine Society (http://www.endo-society.org), the foremostprofessional society serving the endocrine community.




antagonist- and inverse agonist-driven interactions of the ......sence of ligand (fig. 1a, lane 3),...

Documents