Transactivation Properties of Retinoic Acid and Retinoid X Receptors in Mammalian Cells and Yeast CORRELATION WITH HORMONE BINDING AND EFFECTS OF METABOLISM*

The binding affinities of Beis-retinoic acid (Seis-RA) and all-trans-retinoic acid (t-RA) for retinoic acid receptors (RAR) a, p, and y and for retinoid X receptors (RXR) a, p, and y were determined using the recombinant re- ceptor proteins and were compared with each hor-mone’s ability to activate transcription through the re- ceptors in mammalian and yeast cell systems. S-cie-RA bound to both the RXRS (& = 1.4-2.4 nm) and the RARS (& = 0.2-0.8 nm). The ability of Beis-RA to bind to the RARa and RXRs correlated with its ability to produce similar transactivation profiles with these receptors in mammalian and yeast cell assays. t-RA bound to the RARs (& a 0.2-0.4 n ~ ) and activated transcription through the RARa in mammalian and yeast cells. In contrast, while t-RA did not bind to the RXRs, it did activate the RXRs, albeit less potently than Seis-RA, in mamma- lian cells. In yeast, however, the RXRs activated transcription only in the presence

receptors (RXRs). The RARs are encoded by three different genes, a, p, and y, each subtype expressing several isoforms that differ in their amino termini due to alternative mRNA splicing and different promoters (Giguere et al., 1987;Petkovich et al., 1987;Brand et al., 1988;Zelent et al., 1989;Ishikawa et al., 1990). The RXRs are also encoded by three genes, a, p, and y, and, at present, isoforms of these subtypes have not been described Leid et al., 1992b;Mangelsdorf et al., 1992). The RARs and RXRs are classified into these two subfamilies on the basis of their ( a ) primary structural differences, ( b ) binding affinity characteristics to synthetic and naturally occurring retinoids, and (c) differential regulation of target genes (Mangelsdorf et al., 1994). The complexity of retinoid receptor regulation of gene expression is further complicated by recent findings that RXRs can form heterodimers with the RARs, vitamin D receptor (VDR), and thyroid hormone receptor (TR) subtypes (Yu et al., 1991;Kliewer et al., 1992a;Zhang et al., 1992a), as well as various orphan receptors (Kliewer et al., 1992b. Naturally occurring and synthetic ligands have been described that have distinctive binding properties and transactivation effects on the various RAR and RXRs subtypes, thereby allowing differential modulation of retinoid receptor-induced gene expression. For example, t-RA binds directly to the RARs with high afiinity, thereby activating the RARs to modulate gene expression. However, t-RA does not bind to the RXRs, and only at high concentrations does it transactivate via the R X R s . This latter observation led to the hypothesis that t-RA was metabolized in cells to a retinoid compound capable of binding to, and thereby activating, the R X R s . Consistent with this hypothesis, we (Heyman et al., 1992) as well as others (Levin et al., 1992) recently reported the discovery of a n isomer of t-RA, 9-cis-retinoic acid (94s-RA) that binds recombinant human RXRa and stimulates transcription of =-dependent promoters in cell-based assays. The discovery that 9-ci.s-RA binds directly to RXRa has led us to further investigate its functional properties and to determine its ability to bind and transactivate through the other described RXRs, RXRp and RXRy.
Moreover, since 9-ci.s-RA is an activator of RARa (Heyman et al., 1992), we also examined its ability to bind and modulate the transcriptional properties of the RAR subtypes. Finally, we have compared the receptor-retinoid binding affinities with the ability of 9-cis-RA and t-RA to regulate gene expression in both mammalian cells and yeast. The results demonstrate that repeat with spacing of 1 base between repeats; DR3, HRE containing a direct repeat with spacing of 3 bases; DR5, HRE containing a direct from the RARP promoter; HAP, hydroxylapatite. repeat with spacing of 5 bases; PRE, retinoid response element derived 9-cis" is a high affinity ligand for members of both the RAR and RXR subfamilies, whereas t-RA only binds the M s . The transcriptional assays in yeast demonstrate that there is a direct correlation between 9-cis-RA binding to RXR and transcriptional activation by 9-cis" through the RXRS, implying that the binding of g-cis-RA results in an RXR homodimer complex that is transcriptionally active on retinoid responsive promoters. Both 9-cis-RA and t-RA stimulate transactivation in yeast through R A R y alone or via an RARylRXRy combination, with the highest level of transcription resulting from the heterodimeric complex when both receptors are occupied with ligand.
Antibodies-Rabbit antiserum against an hFtXRa peptide was kindly provided by J. Dyck (Heyman et al., 1992). We prepared subtype-specific rabbit antisera against hRXRcl, mRXRp, mRXRy, hRARa, hRARp, and h R A R y peptides, as well as a mouse monoclonal antibody against an peptide. A mouse monoclonal antibody against mouse RARp peptide 429448 was kindly provided by W. Vedeckis (Ali et al., 1992).
CV-1 Co-transactiuation Assays-Assays were performed exactly as described in Heyman et al. (1992). Reporter-luciferase constructs used were the palindromic thyroid hormone response element (TRE-pal) for RAR co-transfections and the CRBPII response element for RXR co-
Polymerase chain reaction technology was used to create restriction sites for the insertion of hRXRa (Mangelsdorfet al., 1990;BspHI, KpnI) and mRXRp  AflIII, KpnI) cDNAs into yeast expression plasmid YEpE2, driven by the copper-inducible yeast promoter CUP1 (McDonnell et al., 1991). This plasmid directs synthesis of a ubiquitin-receptor fusion protein which is subsequently cleaved by yeast ubiquitinase to release receptor. Both were subcloned into anNcoI site immediately 3' of the reading frame of ubiquitin and a downstream KpnI site. Site-directed mutagenesis was utilized to generate an EcoRI site 5' to the ATG of mRXRr (Mangelsdorfet al., 1992) in-frame with the ubiquitin gene. This construct was ligated to a shuttle vector containing the carboxyl-terminal 6 codons of ubiquitin. The AflII-KpnI fragment was ligated into YEpE46 (Sone et al., 1990). Oligonucleotides encoding the CRBPII sequence , the RARp response element @RE) sequence Hoffmann et al., 1990;Sucov et al., 1990), or a VDRE (Ozono et al., 1991) were designed to include XhoI or Sal1 restriction site overhangs. These sequences were annealed and subcloned into the XhoI restriction site of the yeast reporter plasmid YRpC2 (Pham et al., 1991).
Protein Extract Preparation-Sf21 cells were grown in suspension culture and infected at a density of 1.2 x lo6 celldml with recombinant virus at a multiplicity of infection of 2. Cells were harvested and washed 48 h postinfection by centrifugation at 1000 x g for 10 min at 4 "C. Cell pellets were resuspended in 2 volumes of lysis buffer (10 m Tris, pH 7.5, 5 m~ dithiothreitol, 2 m EDTA, 1 m~ phenylmethylsulfonyl fluoride, 1 pg/ml aprotinin, and 1 pg/ml leupeptin) at 4 "C and incubated on ice for 15 min. Lysates were produced by using a Dounce homogenizer and a B pestle followed by addition of 2 M KC1 to a final concentration of 0.4 M and centrifugation at 100,000 x g for 60 min at 4 "C.
Transformed yeast strains were grown to an Am of 1.0 and induced with 100 1.1~ CuS04 for 6 1 6 h at 30 "C. Cells were harvested, washed, and broken in 0.4 M KCl, 10 m~ Tris, pH 7.5, 2 m~ EDTA, 0.5 m~ phenylmethylsulfonyl fluoride, 1 pg/ml aprotinin, and 5 m~ dithiothreitol, by vortexing with acid-washed glass beads at 4 "C or via a Bead Beater (BioSpec Products, Bartlesville, OK). Lysates were obtained by centrifugation at 100,000 x g for 45 min at 4 "C to yield the soluble extract.
Retinoic Acid Binding Assay-For saturation binding analyses, whole cell high-salt extracts (5-50 pg total protein) were added to 12 x 75 borosilicate glass tubes in buffer containing 0.12 M KCl, 8 m~ Tris phosphate, pH 7.4, 8% glycerol, 4 m~ dithiothreitol, and 0.5% CHAPS detergent (Boehringer Mannheim). [3HlRetinoid was added at final concentrations of 0.3 to 10 IIM under dim light. Nonspecific binding was measured at each point in the presence of a 200-fold excess of unlabeled retinoid. The cell extracthetinoid mixture was incubated for 1618 h at 4 "C shielded from the light. Specific ligand binding to receptor was determined by a hydroxylapatite ( H A P ) separation method (Williams and Gorski, 1974) with modifications. HAP slurry (50% in 0.1 M KCl, 10 m Tris phosphate, pH 7.4; 100 pl) was added to each tube and vortexed every 10 min for 40 min at 4 "C. The suspension was centrifuged for 5 min at 1500 x g, and the pellet was washed three times with cold buffer containing 0.1 M KCl, 10 m~ Tris phosphate, pH 7.4,0.5% CHAPS. The HAP pellet was resuspended in 0.5 ml of wash buffer, then in 10 ml of Ecoscint (National Diagnostics, Manville, NJ), and tritium countdmin were measured on a Beckman LSC 6000 scintillation counter and quench-corrected to determine disintegrationdmin. Zmmunoblot Analysis-Receptor or null extracts (5-50 pg of total protein) were added to 2 x SDS-PAGE buffer (Laemmli, 1970). subjected to SDSPAGE, and transferred to nitrocellulose membranes (Bio-Rad).
Membranes were blocked in 5% Carnation instant nonfat dry milk in Tris-buffered saline (TBS, Bio-Rad). Primary antibodies were incubated with membranes overnight at room temperature in 1% milk in TBS, 0.1% Tween 20. Goat anti-mouse or anti-rabbit secondary antibodies conjugated to alkaline phosphatase (Bio-Rad) were incubated with membranes for 2 h at room temperature in 1% milldTBS, 0.1% Tween 20. Washes and development of color were via the Bio-Rad protocol.
Yeast p-Galactosidase Assays-Yeast co-transforinatiodtransactivation assays were performed as described previously (McDonnell et al., 1989) with the following modifications. Growing yeast cultures were diluted to an Asm of 0.02-0.05; 0-10 1.1~ CuS04 and various concentrations of 94s-RA or t-RA were added to separate flasks, and the cells (100 pl) were plated into 96-well culture dishes and incubated in the dark at 30 "C for 16 h. Am was determined, the cells were lysed, substrate was added, and A415 was read after 10-30 min. Normalized P-galactosidase values were determined from triplicate samples as follows: (A41dAgw) x 1000/min developed.
Metabolism Studies-t-RA or 94s-RA (final concentration, 1 1.1~) was added to separate 250-ml cultures of yeast strain BJ5409 at Am = 0.1 or to CV-1 on ten 150-mm plates at density of 4 x lo6 celldplate.
Cultures were protected from light and incubated with hormones for 16 or 40 h at 30 "C with shaking (yeast) or at 37 "C (CV-1). The cultures were harvested by cell scrapers or centrifugation, washed, resuspended in an equal volume of phosphate-buffered saline, and extracted three times with ethyl acetate (Thaller and Eichele, 1987). Media (15 ml) from the incubated cells and media incubated without cells were also extracted in the above manner with ethyl acetate. The extracted material was dried using a Rotovap system prior to resuspension in 0.1 ml of methanol. Sixty pl of ethyl acetate-extracted material in methanol was injected onto a C18 reversed phase column (TSK, 250 x 4.6 mm, 5-pm particle size) at 1.2 mVmin using an isocratic mobile phase consisting of 56% acetonitrile, 16% methanol, and 0.56% acetic acid in HzO, and absorbance was monitored at 350 nm. All reagents were HPLC quality.
3H-Labeled retinoid standards were co-chromatographed to confirm the identity of either t-RA or 9-cis-RA.   RXR subfamilies, transcriptional activation studies were performed in a mammalian cell-based assay system. CV-1 cells co-transfected with a retinoid receptor expression vector and a corresponding luciferase reporter plasmid display a ligand-dependent increase in reporter activity. The transactivation potencies or EC50 values (concentrations which give 50% of the efficacy at M retinoid) for 9-cis-RA and t-RA via the retinoid receptors are summarized in Table I. t-FU activates the RARs over a wide range of potencies, from 10 m to 350 m, for RARy and m a , respectively, and is considerably less potent via the RXRs (920-1490 m). In contrast, 9-cis-RAis a potent activator of both retinoid receptor classes and exhibits a narrower range of sensitivity (45-250 m). Interestingly, in this assay, 9-cis-RA is a more potent activator of the FUR subtypes than of the RXRS.

Potencies of t-RA and 9-cis-RA via the retinoic acid receDtors and retinoid X receotors in CV-1 cells t-RA
Receptor Expression and Immunoblot Analysis-We have previously employed recombinantly expressed RXRa to determine its binding affinities for the retinoic acid isomers (Heyman et al., 1992). In order to characterize the ligand binding properties of the other retinoid receptors for 9-cis-RA and t-RA, the RARS and RXRs were produced in Sf21 cells via a baculovirus expression system (O'Reilly et al., 19921, and Western blot technology was utilized to confirm the production of intact receptors. In control extracts (made from wild-type virus-infected Sf21 cells without specific receptor DNA), there is no detectable immunoreactivity with any of the receptor antibodies ( Fig. 1, A,  lanes 2, 4, and 6, and C, lanes 1, 4, and 6). Extracts prepared from insect cells (Fig. L4, lanes 1,3, and 5) with the respective RXR cDNAs show immunoreactive species (hRXRa = 55 kDa, mFtXFtp = 50 kDa, mRXRy = 53 kDa) which correspond to their approximate predicted molecular weights based upon the open reading frames of their cDNA sequences. The h W s expressed in Sf21 cells and probed with RAR subtype-specific antibodies (Fig. IC, lanes 2, 3, and 5) also show immunoreactive bands that correlate with the expected receptor sizes (hRARa = 54 kDa, hRARp = 54 kDa, hRARy = 50 ma). These protein extracts yield about 0.05-0.2% of their total protein as intact, nonproteolyzed retinoid receptor. In some extracts there is also degraded receptor as evidenced by immunoreactive species of lower molecular weight that are not present in the null extracts.  (Table 11). We conclude that 9-cis-RA binds to the RxRs with high affinity in a saturable and specific manner. The Kd value obtained here for RXRa is %fold lower than our previously reported value of 11.7 m, measured under different conditions (Heyman et al., 1992), or for other reported values of 9.5-18.3 m (Levin et al., 1992;Allenby et al., 1993). In the present study, a modification (inclusion of the detergent CHAPS) has led to an improvement in the binding assay by significantly reducing the amount of nonspecific binding of 94s-RA. These low nanomolar values are more in accord with the higher affinities traditionally observed with the intracellular hormone receptors for their cognate ligands.

Receptor-Hormone Binding Properties and Comparison to
In addition to binding to the RXRs, 9 4 s -R A also binds to all three members of the RAR subfamily with high affinity (  (Table I). However, since g-cis-RA is able to bind to both R A R s and RXRS, the endogenous complement of these recep- Nonspecific binding activity (H) was subtracted from total binding activity (0) to generate specific binding activity (0). Scatchard analysis (0) was performed on specific binding data (duplicates at each point) to yield the indicated Kd values for each receptor. A representative experiment for each receptor-ligand combination is shown. tors may complicate the interpretation of the CV-1 data. Whether g-cis-RA is also more potent through the RARs versus the RXRs in other cell types has yet to be determined.
RARa and RARP display Kds of 0.37 m for t-RA, while R A R y has a value of 0.22 m for t-RA ( Fig. 2C and Table 11). These values are in agreement with those determined for t-RA with bacterially expressed RARa (Yang et al., 1991), and with COS-1-expressed RARs (Allenby et al., 19931, and are at the low end of a range of values that have been reported in several other studies (Nervi et al., 1989;Crettaz et al., 1990;Delescluse et al., 1991;Apfel et al., 1991;Keidel et al., 1992). The RXRs do not display specific binding to t-RA. Therefore, while RARa and RARP bind to 9-cis-RA and t-RA with almost equal affinities (0.20-0.37 m), the binding affinity of R A R y to 9-cis-RA (0.78 m) is somewhat lower than its binding to t-RA (0.22 I " ) . Importantly, while t-RA binds equally well to all three RARs (Table 11), a 35-fold difference in potencies is observed among the different RARs in the CV-1 cell co-transactivation assay (Table I). Hence, there is a greater difference in potencies of t-RA via the transfected RARs in this cell-based assay than is observed in the in vitro binding affinities of the RARs for t-RA.

'inding and Dansactivation
In addition, although R A R y exhibited a statistically significant weaker binding affinity for 9-cis-RA than did the other RARS (Table 111, 9-cis-RA was not less potent with R A R y (EC50 = 45 m) than with RARa (EC50 = 191 m) or RARP (EC50 = 51 m) in the CV-1 assay (Table I). These data indicate that the binding affinities of 9-cis-RA for the various RAR subtypes do not parallel their transactivation potential values in the CV-1 cell assay (Table 11). The presence of other factors in CV-1 cells, such as endogenous retinoid receptor subtypes and/or receptor subtype-specific transcription factors, may contribute to the various levels of activation that are observed with 9-cis-RA via the RARs.
Retinoid Receptor Expression and Dansactivation in Yeast -Mammalian cells contain endogenous receptors and factors that promote receptor heterodimer formation which is known to modulate the transcriptional properties of the retinoid receptors. Therefore, we examined the biochemical and transactivational properties of the RXRs in yeast, a system that lacks intracellular receptors (McDonnell et al., 1989(McDonnell et al., , 1991 and is apparently devoid of factors known to facilitate receptor heterodimer formation. The RXRs were expressed in yeast strains BJ5409, BJ3505, and BJ2168, and their level of expression and ligand binding dissociation constants for 94s-RA were determined. The R X R s expressed in yeast are intact proteins, as analyzed by Western blot analysis (Fig. 1B). The receptor expression levels are 0.145% of the total soluble yeast protein, depending on yeast strain. The yeast-expressed RXRs have DNA binding properties indistinguishable from those expressed in Sf21 or mammalian cells (data not shown). The yeast-expressed RXRs a, P, and y bind 9-cis-RA with mean Kd values of 1.44 m, 1.86 m, and 1.46 m, respectively, which are similar to those obtained from Sf21 RXRs (Table 11). Quantitation of receptor-hormone binding indicates that 3040% of the total soluble yeast receptor is ligand binding-competent. A similar percentage of receptor expressed in Sf21 cells is capable of binding hormone.
To examine the transcriptional properties of the RXRS in yeast, the yeast strain BJ5409 was co-transformed with an RXR subtype along with a P-galactosidase reporter construct containing an RXR response element derived from the cellular retinol binding protein I1 (CRBPII) promoter  linked to the yeast cycl promoter (McDonnell et al., 1989). The addition of 9-cis-RA to these cells results in a concentration-dependent increase in P-galactosidase activity for each RXR subtype, with a maximal response of 5-10-fold occurring at M and an ECS0 of approximately 50-150 m for each subtype (Fig. 3A). These EC50 values obtained in yeast are similar to those obtained for 9-cis-RA acting through the RXRs in CV-1 cells (Table I). 9-cis-RA does not elicit an activation response when the reporter plasmid was transformed into yeast in the absence of a receptor construct (Fig. 3, A X ) , indicating that endogenous RXRs are not present in yeast and that this response is receptor-dependent. The RXRs also activate gene expression in yeast in response to 9-cis-RA via two copies of the PRE (PRE(2)) ( Fig. 3C) and via one copy of the PRE (pRE(1)) (Fig. 4A), a retinoid response element derived from the promoter of the RARP gene Hoffmann et al., 1990;Sucov et al., 1990) with efficacies and potencies similar to that seen with the CRBPII promoter (Fig. 3A). Importantly, the addition of 1 p~ 9-cis-RA (or t-RA) to BJ3505 yeast cultures does not increase the RXR (or RAR) protein concentration in the resultant extracts, as determined by Western blot analysis (data not shown) as has been observed for VDR in yeast after the addition of 1,25-dihydro&tamin D3 (Sone et al., 1990). Thus, the observed increase in P-galactosidase activity following the addition of 9-cis-RA to the yeast in the in vivo assays is due to a ligand-dependent transcriptional  activation of existing receptor and not to an increased amount of constitutively active receptor protein, as is the case for VDR in yeast. While 9-cis-RA activated the RXRS to stimulate transcription through response elements from both the CRBPII (DR1) and RARP (DR5) promoters, it did not transactivate via the RXRs on a promoter containing a VDRE (DR3) sequence from the osteocalcin gene promoter (Ozono et al., 1991;Fig. 30). These data suggest that in yeast all three RXRs are capable of specifically regulating gene expression through retinoid-responsive promoters in the presence of 9-cis-RA. In contrast, t-RA does not stimulate the CRBPII response element (Fig. 3B ) via the FXRs in this yeast transactivation assay. This result is contrary to the transcriptional activation of the CRB-PII-containing promoter by the RXRs with t-RA in CV-1 or Schneider cells Heyman et al., 1992; Table I). Additionally, no stimulation is observed with t-RA in yeast transformed with reporter plasmid alone (Fig. 3B), as expected, since endogenous RARs are not present in yeast. The RARs were also expressed in yeast strains as intact full-length receptors as analyzed via immunoblots (data not shown). We examined the ability of RARy to transactivate the PRE(1)-containing reporter in response to both 9-cis-RA and t-RA. R A R y gives a 2-%fold induction of the PRE-reporter P-galactosidase construct in response to 9-cis-RA (Fig. 4 . 4 ) or t-RA ( Fig. 4B) with ECS0 values of approximately 100 m. 9-cis-RA displays a similar potency via R X R y on the PRE(1) (Fig.  4A), while t-RA does not show substantial induction through R X R y (Fig. 4B). RARy co-transformed with R X R y and the PRE(l)-containing reporter gives a higher basal level of transcription than with either receptor alone (Fig. 4, A and B 1. The addition of t-RA increases the level of transcription about 2-fold above this basal level (Fig. 4B), and 9-cis-RA elicits a 4-5-fold enhancement of transcription (Fig. 4.4). That 9-cis-RA results in higher levels of transcription implies that optimal levels of transcription occur when both receptors are ligand-occupied. In fact, a combination of t-RA and a potent RXR-selective compound gives a response characterized by even greater -fold induction and potency via the heterodimeric pair than does 94s-RA (our data not shown).
Metabolism Studies-Since t-RA does not directly bind to the RXRs, one possible explanation for t-RA activation of the RXRresponsive pathways in mammalian cells could be potential metabolism of t-RA to g-cis-RA, which then binds to and activates the RXRs. Indeed, when mammalian or Schneider cells were incubated with t-RA for 24 h, 9-cis-RA was detected within the cell (Heyman et al., 1992;Levin et al., 1992). This conversion could be enzymatic, and the inability of t-RA to activate the CRBPII-containing reporter via the R X R s in yeast might indicate that yeast do not isomerize t-RA to 9-ci.s-RA. To test this hypothesis, we incubated CV-1 cells, yeast, or media without cells, with t-RA or 9-cis-RA under conditions identical with those in the transcriptional assays (see "Experimental Procedures"). The composition of retinoic acid isomers present in the organic-extracted material from each source (cell pellet or media) was determined via reversed-phase HPLC monitored at 350 nm for the presence of retinoids (Thaller and Eichele, 1990). Hormone-treated yeast media or CV-1 media without cells shows 5-9% thermal-and/or photoinduced conversion of either added retinoid (Table I11 and Fig. 5, C and F). Only 3 4 % of t-RA is converted to 9-cis-RA in CV-1 cells aRer 40 h, and approximately 16-20% of the t-RA is converted to other metabolites including 13-cis-RA in both CV-1 cell extracts and in CV-1 media without cells (Table 111). Therefore, under the experimental conditions employed here, the conversion of t-RA to other retinoids in CV-1 cells can be accounted for by processes that are not dependent on cellular components. Yeast also show a similar conversion (3-7%) of t-RA to 94s-RA when t-RA is incubated either in yeast media alone or in yeast cells (Table I11 A RAR and RXR Hormone Binding and Pansactivation been detected in target tissues (Heyman et al., 1992). These data imply that sufficient endogenous ligand is present in certain tissues to modulate the activity of the RXRs.
We have demonstrated previously in CV-1 co-transfection assays that 9-cis-RA activates not only RXRa, but also RARa (Heyman et al., 1992). Unexpectedly, the present experiments indicate that 94s-RA not only binds to the RARs, but does so with greater affinity than to the RXRs. These data are consistent with the observation that 9-cis-is more potent in stimulating transcription via RARP and RARy than through the RXRs in CV-1 cells. Thus, 9-cis-RA appears to be a "bifunctional" ligand in that it binds and functions as a potent transcriptional activator of both retinoid subfamilies. 9-cis" and t-RA both bind equally well to the F U R S and induce the transactivation potential of the F U R S , implying that both naturally occurring retinoid isomers function to regulate "mediated pathways. The case of a n intracellular receptor binding with equal affinities to both a hormone and its naturally occurring metabolite is unprecedented in the steroid receptor field. Whether the dual high affinity interactions of the RAR subtypes with 9-cis" and t-RA are physiologically relevant will require further investigation into their circulating levels, stability, and metabolism. These data imply that the metabolism and/or interconversion between t-RA and 9-cis-RA in different target tissues may help dictate which gene pathways are regulated. Examination of the transcriptional properties of the retinoid receptors in a mammalian cell-based assay system is complicated by the fact that most, if not all, mammalian (and insect) cells contain endogenous retinoid andor other intracellular receptors. Mammalian and insect cell extracts have been shown to biochemically complement many of the receptors in receptorspecific DNA binding assays (Burnside et al., 1990;Sone et al., 1991;Yang et al., 1991). The RXRs are known to heterodimerize with at least six other known receptors (VDR, two subtypes of TR, and the three RARs; Yu et al., 1991;h i d et al., 1992b;Zhang et al., 1992a;Kliewer et al., 1992a) as well as with orphan receptors (Kliewer et al., 1992b. Moreover, the Drosophila homologue of RXR, ultraspiracle, can also pair with several of these intracellular receptors (Oro et al., 1992). For these reasons, it is difficult to directly correlate in. vitro ligand binding affinities with their ex vivo transactivation properties in mammalian or insect systems. In an attempt to avoid this complicating issue, we examined the transactivation properties of the RXRs in yeast, an organism that does not appear to exhibit retinoid receptor activity or activity of any other intracellular receptors (McDonnell et al., 1989(McDonnell et al., , 1991 and does not contain factors that complement intracellular receptors in HRE-DNA binding assays. Importantly, the activity of glucocorticoid receptor, estrogen receptor, and other receptors has been reconstituted in yeast, indicating that fundamental coupling to the transcriptional machinery is conserved in yeast (Schena and Yamamoto, 1988;Metzger et al., 1988).
We show here that 94s-RA stimulates transactivation of the CRBPII response element (DR1)-containing promoter in an --dependent manner in yeast with potencies similar to those seen in CV-1 cells. In contrast, t-RA did not stimulate transactivation of this promoter via the R X R s in this yeast system. Thus, in yeast there is a direct correlation between ligand-receptor binding and ligand-induced transcriptional activation; a correlation that is not apparent in mammalian cells. Interestingly, 9-cis-RA also induces transcriptional activation of a PRE(DR5)-containing promoter (but not a VDRE(DR3)) in the presence of RXR. This indicates that R X R s require a specific nucleotide spacing between the half-sites of the direct repeat, implying that they function as homodimers. It is clear from our data that the R X R s function alone in yeast in response to g-cis-RA, most likely by forming homodimers on promoter elements. These data corroborate and extend the study by Zhang et al. (1992b) that indicates that 9-cis-RA can elicit RXR homodimer formation as assayed by electromobility shift assays, and that the RXRs activate the PRE in insect Schneider cells (a cell type also devoid of RARs), in the presence of 9-cis-RA (our data not shown).
We also show evidence that RAR-y transactivates via the PRE in yeast, in response to 94s-RA or t-RA. The heterodimeric RARmxR combination gives an enhanced basal level activity which is further induced when both receptors are occupied by ligand. These functional data from the retinoid receptors, together with the information that yeast do not contain intracellular receptors, indicate the usefulness of yeast as a system of choice for receptor expression and for use as a null receptor background to study the mechanism of receptor transactivation. It will be interesting to test other retinoid compounds in yeast co-transformed with the other RAR subtypes and RAR/ RXR combinations via various promoters to further examine the mechanism of retinoid action. The fact that t-RA does not activate the RXRs in yeast as it does in CV-1 or Schneider cells (Mangelsdorfet al., 1990;Heyman et al., 1992) does not seem to be due to substantial differences in the metabolism of t-RA to 9 4 s in mammalian cells versus yeast. In fact there is the same, small percentage of conversion of t-RA to 9-cis-RA in all of the mammalian or insect cells that we have studied, as well as in yeast, or from media without cells. Thus, under the experimental conditions employed here, while 9-cis-RA is present in these cell or media sources at low levels, it does not appear to be generated from t-RA by a cell-dependent enzymatic process. Although the ratio of 9-cis-RA to t-RA is similar in yeast and CV-1 cells, 94s-RA may be present in higher concentrations in CV-1 cells, which would explain why the addition of t-RA stimulates transcription via the RXRs in CV-1 cells, but not in yeast. The ability of t-RA to transactivate via the co-transfected RXRs in CV-1 cells may also be modulated by endogenous RARs (absent in yeast) which may bind t-RA (or converted 9-cis-RA) and heterodimerize with R X R s to stimulate these promoters. Additionally, other factors present in CV-1 cells (but absent in yeast) may help modulate the response of RXRs to t-RA. Also, yeast may not be as efficient in the uptake of t-RA versus 9-cis" or as compared with CV-1 or Schneider S2 cells and may even excrete retinoids from the cells as has been shown with estrogen in yeast (Gilbert et al., 1993).
While no cell-dependent conversion of t-RA to 9-cis-RA is evident in CV-1 or yeast cells (or in HepG2 or Schneider S2 cells, our data not shown), mammalian cells do exhibit celldependent conversion of 94s-RA to t-RA. These data may indicate that t-RA is more stable than 94s-RA or that an enzymatic pathway that converts 9-cis-RA to t-RA may exist in the cultured cells that we have studied, while the reverse reaction is less favored. While enzymatic conversion of t-RA to 9-cis-RA is not evident in any of the cell lines described here, it is possible that this activity is lost in cultured cells. The 25-hydroxyvitamin D3-l-a-hydroxylase enzyme that converts 25hydroxyvitamin D3 to its active metabolite, 1,25-dihydroxyvitamin D3 is located in the kidney (Fraser and Kodicek, 1973), but not in most cultured cells, including a variety of kidney cell lines. It will be necessary to examine endogenous retinoid levels in tissues and serum, as well as metabolism in tissues, to gain a better understanding of the normal circulating levels and metabolism of retinoid isomers. The complexity of the mechanism of action of retinoids is quite apparent. There are at least six subtypes and several isoforms of the nuclear retinoid receptors and the possibility for various combinations thereof in different cell types. The presence and availability of 94s-RA and t-RA will determine which receptors are activated, whether homodimers or heterodimers are formed, and ultimately which genes' expression levels are regulated and to what extent. Moreover, different cells may exhibit unique or favored metabolic schemes for interconverting retinoids, depending on the enzyme(s) that they produce. Different cell types may also vary in their pools of transcriptional accessory proteins which may display different specificities for the various retinoid receptor subtypes. In addition, the cellular complement of CRBPs and CRABPs and their affinities for various retinoid metabolites will also determine availability and effective concentration of the ligands. For example, t-RA binds to recombinant CRABPI with an apparent Kd of 6 m, while 9-cis-RA does not display saturable binding to CRABPI (data not shown). Allenby et al. (1993) have also shown recently that while t-RA binds to CRABPI and -11, 94s-RA does not.
CRABP overexpression in cells that are normally responsive to t-RA renders them less sensitive to t-RA treatment (Boylan and Gudas, 1992), presumably because the ligand is less available for nuclear receptor uptake. Likewise, 94s-RA may be more accessible to the nuclear retinoid receptors if it does not bind to a cellular binding protein. Additionally, retinoid movement through serum to other cells via serum retinoid binding proteins (RBPs) and/or serum albumin may play a role in the function of g-cis-RA. The fundamental question of whether 9-cis-RA can be transported in serum or whether it is restricted to metabolizing cells must be answered. These and other questions regarding retinoid metabolism and mechanism of action await further study.