Identification and functional separation of retinoic acid receptor neutral antagonists and inverse agonists.

Inverse agonists are ligands that are capable of repressing basal receptor activity in the absence of an agonist. We have designed a series of C-1-substituted acetylenic retinoids that exhibit potent antagonism of retinoic acid receptor (RAR)-mediated transactivation. Comparison of these related retinoid antagonists for their ability to repress basal RAR transcriptional activity demonstrates that the identity of the C-1 substituent differentiates these ligands into two groups: RAR inverse agonists and neutral antagonists. We show that treatment of cultured human keratinocytes with a RAR inverse agonist, but not a RAR neutral antagonist, leads to the repression of the serum-induced differentiation marker MRP-8. While RAR-selective agonists also repress expression of MRP-8, cotreatment with a RAR inverse agonist and a RAR agonist results in a mutual repression of their individual inhibitory activities, indicating the distinct modes of action of these two disparate retinoids in modulating MRP-8 expression. Our data indicate that RARs, like beta2-adrenoreceptors, are sensitive to inverse agonists and that this new class of retinoids will provide insight into the molecular mechanisms of RAR function in skin and other responsive tissues.

Inverse agonists are ligands that are capable of repressing basal receptor activity in the absence of an agonist. We have designed a series of C-1-substituted acetylenic retinoids that exhibit potent antagonism of retinoic acid receptor (RAR)-mediated transactivation. Comparison of these related retinoid antagonists for their ability to repress basal RAR transcriptional activity demonstrates that the identity of the C-1 substituent differentiates these ligands into two groups: RAR inverse agonists and neutral antagonists. We show that treatment of cultured human keratinocytes with a RAR inverse agonist, but not a RAR neutral antagonist, leads to the repression of the serum-induced differentiation marker MRP-8. While RAR-selective agonists also repress expression of MRP-8, cotreatment with a RAR inverse agonist and a RAR agonist results in a mutual repression of their individual inhibitory activities, indicating the distinct modes of action of these two disparate retinoids in modulating MRP-8 expression. Our data indicate that RARs, like ␤ 2 -adrenoreceptors, are sensitive to inverse agonists and that this new class of retinoids will provide insight into the molecular mechanisms of RAR function in skin and other responsive tissues.
The concept of an inverse agonist, or negative antagonist, has grown out of studies of G-protein-coupled receptors where certain antagonists have been demonstrated to inhibit the activity of unliganded receptors (1)(2)(3). Recent evidence from transgenic mice with myocardial overexpression of the ␤ 2 -adrenoreceptor (4) has provided further support for a two-state model of ligand-regulated G-protein-coupled receptors. In this model, an equilibrium is postulated to exist between inactive receptors and spontaneously active receptors, which are capable of G-protein coupling in the absence of ligand. This model has provided the conceptual framework for the existence of agonists that shift the equilibrium toward active receptors, inverse agonists that shift the equilibrium toward inactive receptors, and neutral antagonists that themselves do not effect the receptor equilibrium but are capable of competitive antagonism of both agonists and inverse agonists.
In the nuclear hormone-steroid receptor superfamily, antagonists for the retinoic acid, progesterone/glucocorticoid, and estrogen receptors have been described (5)(6)(7)(8). The anti-proges-tin effects of RU-486 and the anti-estrogen effects of 4-hydroxytamoxifen have been demonstrated to have clinical utility in termination of pregnancy (9) and treatment of hormone-dependent breast cancer (10), respectively. Inverse agonists have not been previously reported for the nuclear receptor family. We have recently described a C-1-substituted acetylenic retinoid, AGN 193109, which exhibits potent antagonism of alltrans-retinoic acid (ATRA) 1 -mediated transactivation of RAR␣, RAR␤, and RAR␥ (6) as well as blockade of retinoid-mediated topical irritation in animal models (11). Using a series of AGN 193109 analogs, which differ in their C-1 substituent, we demonstrate that these RAR antagonists can be differentiated on the basis of their ability to inhibit, or trans-repress, RAR basal transcriptional activity in the absence of exogenously added retinoid agonist. We show that a RAR inverse agonist, but not a neutral antagonist, is capable of regulating gene expression of a differentiation marker in cultured human keratinocytes in a manner that is distinct from that of a classical retinoid agonist. As such, RARs, like G-protein-coupled receptors, are sensitive to inverse agonists, and such activity may be amenable to the design of retinoids with unique therapeutic potential.

EXPERIMENTAL PROCEDURES
Transfections and DNA Constructs-For ATRA antagonism studies, 4 ϫ 10 5 CV-1 cells (12-well Costar plate) were transiently transfected via calcium phosphate precipitation (12) with 0.7 g of the reporter plasmid MTV-4(R5G)-Luc containing four copies of the DR-5 RARE R5G (5Ј-AGGGTTCACCGAAAGAACAGT-3Ј) inserted into the HindIII site of the plasmid ⌬MTV-Luc (13), 0.1 g of the ␤-galactosidase expression plasmid pCH110 (Pharmacia Biotech Inc.), 0.01 g of the plasmid pRS-hRXR␣ (14), and 0.05 g of either pRS-RAR␣-P-GR (15), pcDNA3-RAR␤-P-GR, or pcDNA3-RAR␥-P-GR. RAR␤-P-GR and RAR␥-P-GR vectors were constructed via polymerase chain reaction-mediated mutagenesis of the P-box followed by insertion of the altered cDNAs into the mammalian expression vector pcDNA3 (Invitrogen). Receptors expressed from these plasmids contain DNA binding domains in which the P-box of the retinoid receptor (EGCKG) has been altered to that of the glucocorticoid receptor (GSCKV), allowing for RXR/RAR recognition of the R5G RARE. Eighteen hours after introduction of the DNA precipitants, cells were rinsed with phosphate-buffered saline and fed with Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 10% activated charcoal-extracted fetal bovine serum (Gemini Bio-Products). Cells were treated for 18 h with 10 nM ATRA in conjunction with the compounds indicated in the figure. After rinsing with phosphatebuffered saline, cells were lysed and luciferase activity was measured as described previously (16). Luciferase values represent the mean Ϯ S.E. of triplicate determinations normalized to ␤-galactosidase activity.
For analysis of ER-RAR chimeric receptor transactivation, 4 ϫ 10 5 CV-1 cells (per well of a 12-well Costar plate) were transiently transfected via calcium phosphate precipitation with 0.5 g of pERE-tk-Luc (containing the estrogen-regulated element of the Xenopus vitellogenin * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. § To whom correspondence should be addressed: Mail Code RD-3D, 2525 Dupont Dr., Irvine, CA 92715-9534. Tel.: 714-246-4895; Fax: 714-246-6207; E-mail: ElliottK@mail011.usirvine.allergan.sprint.com. A2 gene (17) inserted into the plasmid tk-luciferase (18)), 0.1 g of pCH110, and 0.05 g of the SV40-based vector pECE (19) expressing chimeric ER-RAR receptors consisting of the estrogen receptor A/B and DNA binding domains fused to the DEF domain of RAR␣, -␤, or -␥ (20). Treatment of transfected cells and determination of luciferase activity were performed as described for ATRA antagonism studies (above).
For analysis of ligand regulation of RAR␥-VP-16, CV-1 cells were transfected as outlined above with 0.5 g of pERE-tk-Luc, 0.1 g of pCH110, 0.1 g of ER-RXR␣ expression vector, and 0.2 g of the chimeric expression vector RAR␥-VP-16. ER-RXR␣ contains the hormone binding domain (amino acids 181-458) of RXR␣ (14) fused downstream from the estrogen receptor A/B and DNA binding domains (20). RAR␥-VP-16 has been previously described (21) and contains the activation domain of herpes simplex virus VP-16 fused to the N terminus of full-length RAR␥ cDNA. The strong constitutive activity of RAR␥-VP-16 is directed to pERE-tk-Luc by heterodimerization of the RAR␥ moiety to ER-RXR␣ bound to the ERE.
Ligand Binding Assays-Compounds were analyzed, as described previously (22), for their ability to competitively inhibit specific binding of [ 3 H]ATRA or [ 3 H]9-cis-retinoic acid to baculoviral expressed RARs and RXRs, respectively. Dissociation constants (K d ) were calculated via application of the Cheng-Prussof algorithm (23). Values represent the average of three independent assays performed in duplicate Ϯ the standard error of the mean.
Analysis of Cultured Human Keratinocytes-Human foreskin keratinocytes (passage 3), prepared as described previously (24), were maintained in keratinocyte growth medium (Clonetics) supplemented with 0.15 mM Ca 2ϩ and hydrocortisone (1 M). At 50 -70% confluence, the medium was changed to keratinocyte growth medium without hydrocortisone. After 24 h, cells were treated with 10% serum (charcoal extracted) and retinoids every day for 5 days. Control cultures were treated with 10% serum and vehicle alone (0.02% dimethyl sulfoxide). Cell lysates were prepared in 300 l of cell lysis buffer (10 mM Tris, pH 8.0, 1 mM EDTA, 0.5 mg/ml phenylmethylsulfonyl fluoride, 0.5 g/ml leupeptin, and 2 g/ml aprotinin) and centrifuged at 10,000 ϫ g for 10 min. 30 g of lysate was subjected to SDS-polyacrylamide gel electrophoresis (15%) and analyzed for MRP-8 protein by Western blot using a monoclonal anti-MRP-8 antibody, CF-145 (25).

C-1-substituted Acetylenic Retinoids Are RAR Antagonists-
The structurally related acetylenic retinoids in Table I were tested for their ability to competitively bind the three RAR subtypes. All of the compounds exhibit high affinity binding to the RARs and no measurable binding to the RXRs. With the exception of AGN 192870 and 193840, these ligands do not transactivate the RARs as measured in CV-1 cells transiently transfected with either chimeric ER-RARs or full-length RARs using appropriate reporter constructs (data not shown). AGN 192870 and 193840 do exhibit partial agonist (approximately 30% compared with ATRA) activity at RAR␤ only (data not shown). As shown in Fig. 1, the ligands of Table I are effective RAR antagonists, inhibiting ATRA-mediated transactivation with the expected RAR␤ partial antagonist behavior of AGN 192870 and 193840. Determined IC 50 values for these RAR antagonists (see Table I) are in agreement with their binding affinities (K d ) for the RARs.
Repression of Basal RAR Transcriptional Activity-By analogy with inverse agonists for the ␤ 2 -adrenoreceptor, a RAR inverse agonist should inhibit RAR basal transactivation. We analyzed the ability of AGN 193109, the highest affinity RAR ligand of Table I, to repress the basal activity of chimeric ER-RAR receptors containing the DEF domain of RAR fused to the estrogen receptor A/B and DNA binding domains (see Fig.  2a). These chimeric receptors provide a higher unstimulated signal relative to full-length wild type RARs. Comparison of luciferase activity from cells transfected with the reporter plasmid ERE-tk-Luc alone with that of cells cotransfected with both reporter and receptor constructs indicated a stimulation of basal luciferase activity for ER-RAR␤ and ER-RAR␥ only (Fig.  2b, inset). AGN 193109 repressed in a dose-dependent manner the basal activity of ER-RAR␤ and ER-RAR␥ (Fig. 2b) but had a negligible effect upon the basal activity of ER-RAR␣ cotransfectants. This refractoriness of ER-RAR␣ to AGN 193109 treatment may be due to the relatively weak basal transactivation activity of this receptor in the absence of agonist.
We next tested the ability of AGN 193109 to repress the basal activity of a RAR␥-VP-16 chimeric receptor where the constitutively active transactivation domain of herpes simplex virus (HSV) VP-16 is fused to the N terminus of RAR␥ (21). CV-1 cells were transfected either with ERE-tk-Luc and ER-RXR␣ or with ERE-tk-Luc, ER-RXR␣, and RAR␥-VP-16 (see Fig. 2a). As expected, luciferase activity in cells transfected with only ERE-tk-Luc and ER-RXR␣ is not regulated by AGN 193109 (Fig. 2c). Similarly, cells cotransfected with only EREtk-Luc and RAR␥-VP-16 do not exhibit regulation of luciferase activity by AGN 193109 or RAR agonists (data not shown). As has been previously demonstrated (21), addition of RAR␥-VP-16 to the transfection mixture results in a significant increase in the basal level of luciferase activity over that of ER-RXR␣ alone, consistent with heterodimerization of RAR␥-VP-16 and ER-RXR␣ receptors at the ERE. While ATRA treatment resulted in a mild induction of activity, AGN 193109 treatment gave a strong, dose-dependent reduction of RAR␥-VP-16 basal activity (Fig. 2c)  acetylenic retinoids Dissociation constants (K d ) were determined using baculovirus-expressed RARs as previously described (22). Values represent the average of three independent assays performed in duplicate Ϯ S.E. Sf21 cells infected with RXR expression constructs exhibited no detectable binding of these RAR ligands. IC 50 values (below K d values and in parentheses) represent the mean Ϯ S.E. of at least three independent assays, essentially as shown in Fig. 1, performed using antagonism of a 10 Ϫ8 M dose of ATRA at each of the RARs. AGN 192870 and 193840 exhibit RAR␤ partial agonism (pa); therefore no IC 50 value is provided. assay would be shared by the four other RAR antagonists in Table I. Treatment of CV-1 cells transfected as described in Fig.  2c revealed two distinct categories of compounds (Fig. 2d). In contrast to that shown for AGN 193109, AGN 192870 and 193840 failed to repress RAR␥-VP-16 activity even though they are potent and effective RAR␥ antagonists (Fig. 1). Similar to AGN 193109, 193385 and 193389 exhibit trans-repression of RAR␥-VP-16, consistent with the concept that these RAR ligands are capable of active repression, or inverse agonism, in the absence of added RAR agonist. Thus the identity of the C-1 substitution is capable of differentiating neutral antagonists from inverse agonists. The two classes of compounds are characterized by distinct structural features in that the inverse agonists (AGN 193109, 193385, and 193389)  RAR Inverse Agonist Function in Cultured Keratinocytes-Retinoids inhibit the expression of differentiation markers in cultured normal human keratinocytes (27,28) induced by confluence, elevated calcium, or serum treatment. We have recently shown that RAR-specific agonists inhibit the expression of the differentiation-specific gene MRP-8 (calgranulin A) in cultured keratinocytes. 2 While expression of MRP-8 is not detectable in normal human skin, it is highly expressed in psori-atic epidermis and in cultured human keratinocytes differentiated with serum (25,29,30). Surprisingly, treatment of serum-differentiated human keratinocytes with the RAR inverse agonist AGN 193109 also mediates a dose-dependent repression of MRP-8 protein levels (Fig. 3a). In contrast to both RAR agonists and inverse agonists, the neutral antagonist AGN 193840 does not repress MRP-8 expression (Fig. 3a). Consistent with their competitive activities upon RAR␥-VP-16 trans-repression (above), AGN 193840 cotreatment provides a dose-responsive antagonism of MRP-8 repression by both the RAR agonist TTNPB and the RAR inverse agonist AGN 193109 (data not shown). Interestingly, simultaneous administration of TTNPB together with AGN 193109, retinoids which as single agent treatments provide repression of MRP-8, results in a mutual antagonism of MRP-8 repression (Fig. 3b). Thus, a retinoid inverse agonist and a retinoid agonist are both capable of mediating repression of MRP-8 through apparently distinct mechanisms that are mutually exclusive in this system.
A growing body of evidence supports a general model of transcriptional activation of nuclear receptors involving ligandmediated control of receptor interactions with both positive and negative associated factors (reviewed in Ref. 31). One possible mechanism underlying the inverse agonism of AGN 193109 may involve its ability to increase the interaction between the RAR and recently identified corepressor molecules (32,33). Such modulation of corepressor-RAR interaction may provide a means to regulate gene expression directly at the RAR as well as through cross-talk with other nuclear receptors that share a common corepressor (34). Alternatively, a RAR inverse agonist may confer upon the RAR an inhibitory interaction with components of the core transcriptional machinery. AGN 193109 does not repress 12-O-tetradecanoylphorbol-13-acetate-stimulated AP-1 activity in HeLa cells. 3 Further, this inverse agonist has no effect upon the ability of RAR/RXR heterodimers to bind DNA in vitro. 4 Taken together, we propose that RAR inverse agonists function via shifting the equilibrium between RAR and RAR ϩ corepressor toward the latter.
Given the level of relatedness among the nuclear hormone receptor family members and the apparent conservation of mechanistic paradigms associated with their regulation by 2 S. Nagpal, manuscript in preparation. 3  The weak activation by ATRA is consistent with conversion to 9-cis-retinoic acid and consequent activation of ER-RXR␣ homodimers (26). CV-1 cells cotransfected with ERE-tk-Luc, pCH110, and ER-RXRa and RAR␥-VP-16 were treated with ATRA (E) or AGN 193109 (Ç). d, separation of RAR inverse agonists from neutral antagonists. CV-1 cells cotransfected with ERE-tk-Luc, pCH110, and ER-RXR␣ and RAR␥-VP-16 were treated with the compounds described in Table I. their respective ligands, it seems reasonable that inverse agonists for other family members may be identified. Our observations suggest that it is possible, with appropriate structural modifications, to design retinoid ligands with agonist, neutral antagonist, or inverse agonist properties, each with distinct biological properties.