(Received for publication, November 15, 1996, and in revised form, December 17, 1996)
From the Laboratory of Tumor Biology, Department of
Pathology, Uppsala University, S-751 85 Uppsala, Sweden and the
§ Ludwig Institute for Cancer Research, Box 240, S-171 77 Stockholm, Sweden
The receptor for 9-cis-retinoic acid, retinoid X receptor (RXR), forms heterodimers with several nuclear receptors, including the receptor for all-trans-retinoic acid, RAR. Previous studies have shown that retinoic acid receptor can be activated in RAR/RXR heterodimers, whereas RXR is believed to be a silent co-factor. In this report we show that efficient growth arrest and differentiation of the human monocytic cell line U-937 require activation of both RAR and RXR. Also, we demonstrate that the allosteric inhibition of RXR is not obligatory and that RXR can be activated in the RAR/RXR heterodimer in the presence of RAR ligands. Remarkably, RXR inhibition by RAR can also be relieved by an RAR antagonist. Moreover, the dose response of RXR agonists differ between RXR homodimers and RAR/RXR heterodimers, indicating that these complexes are pharmacologically distinct. Finally, the AF2 activation domain of both subunits contribute to activation even if only one of the receptors is associated with ligand. Our data emphasize the importance of signaling through both subunits of a heterodimer in the physiological response to retinoids and show that the activity of RXR is dependent on both the identity and the ligand binding state of its partner.
Vitamin A metabolites such as all-trans-(atRA)1 and 9-cis-retinoic acid (9cRA) play critical roles during embryonic development and adult physiology (1). At the cellular level retinoids influence processes such as growth and differentiation by specific effects on the regulation of gene expression. Both atRA and 9cRA bind to specific types of nuclear hormone receptors that belong to a large family of conserved proteins, including receptors for steroid hormones, thyroid hormone, and vitamin D3 (2). In addition, a large number of related proteins have been identified which lack known ligands and are therefore refered to as orphan receptors (3).
Nuclear hormone receptors act as ligand inducible transcription factors
and bind to specific DNA binding sites, termed hormone response
elements, in promoters of regulated genes. The receptors contain
distinct functional domains, including a well conserved DNA binding
domain and a somewhat less conserved C-terminal ligand binding domain
(LBD). Recent studies have revealed that ligand binding mediates a
structural transition, which involves repositioning of a -helix in
the core of the C-terminal activation domain (AF2) (4-9). In activated
receptors the AF2 domain binds to accessory proteins named
co-activators which in turn are believed to mediate contacts with the
basal transcriptional machinery and thereby trigger transcriptional
activation (10-13). In several cases receptors have been shown to
function as repressors of transcription, an activity that is believed
to be mediated by a distinct type of accessory proteins termed
co-repressors (14-17).
Two different types of retinoid receptors have been identified. RAR is activated by both atRA and 9cRA, whereas RXR is activated only by 9cRA (1). RXR has been shown to regulate gene expression in response to 9cRA both as homodimers and in heterodimeric complexes with orphan receptors such as LXR, NGFI-B, and FXR (18). In addition to its function as a 9cRA-inducible receptor, RXR plays a central role as a heterodimerizing co-factor in gene regulatory events mediated by other members of the nuclear receptor super family, i.e. RAR, the thyroid hormone receptor (TR) and the vitamin D3 receptor (VDR) (3). In these heterodimers RXR has been shown to be allosterically blocked and function as a silent co-receptor (19-21).
In contrast, previous studies have demonstrated that both RAR and RXR contributes to the activation of certain retinoic acid-responsive elements (RAREs) (6). Also, several cell lines which respond to retinoids require activation of both RAR and RXR pathways for efficient induction of cellular processes, e.g. growth arrest of cervical carcinoma cells (22), differentiation of embryonic carcinoma cells (23), and granulocytic differentiation of HL-60 cells (24). These results either suggest that two distinct pathways synergize in mediating the biological response of retinoids or, alternatively, in contrast to the concept of RXR as a silent heterodimerization partner of RAR, point to the possibility that both subunits of the RAR/RXR heterodimer can be activated by ligand.
In this study we have used synthetic RAR- and RXR-selective ligands to demonstrate that the monoblastic cell line U-937 requires activation of both RAR and RXR for efficient induction of the monocytic differentiation program. To analyze the molecular mechanism behind this cooperativity between RAR and RXR, we utilized the GAL4 system, which enabled us to study the retinoid response through RAR/RXR heterodimers independent of endogenous receptor complexes that bind to RA response elements. Although RXR can be shown to be in a repressed state in complex with RAR, we show that this inhibition of RXR is relieved by RAR ligand binding. Even binding of an RAR antagonist allows activation of the RXR moiety in the RAR/RXR heterodimer. Moreover, we demonstrate that the RXR AF2 domain is required for efficient activation of ligands binding to RAR, TR, VDR, and PPAR also in the absence of RXR ligand. In conclusion, our data demonstrate that RXR is subject to both positive and negative allosteric regulation by its heterodimeric partner that results in repression, partial activation, or full activation of RXR depending on the availability of ligands for either or both subunits.
U-937-1 cells (25)
were maintained in RPMI 1640 medium supplemented with 10% fetal calf
serum (Life Technologies, Inc., Uxbridge, United Kingdom) and
antibiotics (100 units/ml of penicillin and 50 µg/ml of
streptomycin). Cells at a concentration of 0.2 × 106/ml were exposed to different retinoid compounds for 4 days. Retinoids used were atRA (Sigma), 9cRA (Hoffmann
LaRoche, Basel, Switzerland), and TTNPB and SR11237 (19).
Differentiation antigen expression was analyzed by flow immunocytometry
as described (26). Primary antibodies used were LeuM5 (CD11c; Becton
Dickinson, Mountain View, CA), GoH3 (CD49f; Immunotech, Marseilles,
France), and YTH71.3 (CD66a; kindly provided by Prof. H. Waldman,
Department of Pathology, University of Cambridge, Cambridge, United
Kingdom). Fluorescein isothiocyanate-labeled goat-anti-mouse
F(ab)2 fragments (Central Laboratory of the Netherlands
Red Cross Blood Transfusion Service, Amsterdam, Netherlands) were used
as a secondary reagent for CD11c-labeled cells and fluorescein
isothiocyanate-labeled rabbit-anti-rat F(ab
)2 fragments
for CD66a- and CD49f-labeled cells. Cell cycle analysis was performed
as described (27). The data were processed with the MacCycle software
(Phoenix Flow Systems, San Diego, CA) to obtain DNA distribution
histograms as shown in Fig. 1C. The cellular morphology was
examined by light microscopy after cytospin centrifugation and staining
with May-Grünewald-Giemsa solution.
Plasmids
The luciferase reporters used in transfection
experiments contain three copies of the hRAR2 gene promoter RARE
(
REx3-tk-luc) or four copies of the GAL4 binding site
(MH100×4-tk-luc) cloned upstream of the herpes simplex virus thymidine
kinase gene minimal promoter (28). Receptors and derivatives of
receptors were expressed from pCMX expression vectors containing the
cytomegalovirus promoter and enhancer sequences (29). pCMX-GAL4-RAR and
pCMX-GAL4-RXR contain the sequences encoding the ligand binding domains
of hRAR
(from Glu-185) and hRXR
(from Glu-203), respectively,
cloned in frame after the sequence encoding the DNA binding domain of yeast GAL4 amino acids 1-147) (21, 28). pCMX-GAL4-RARmAF2 was cloned
by a two-step polymerase chain reaction strategy and contains the
sequence encoding the complete ligand binding domain (from Glu-185) of
the hRAR
mutated in the core of the C-terminal AF2 changing Glu-411
and Met-412 into alanines. pCMX-GAL4-RXRmAF2 was also mutated by
polymerase chain reaction and contains the sequence encoding the
complete ligand binding domain (from Glu-203) of the hRXR
mutated in
the C-terminal AF2 changing Met-454 and Leu-455 into alanines.
pCMX-RXRmAF2 contains the identical two-amino acid substitution in the
context of the complete coding sequence of hRXR
. A similar mutation
in RXR has been described previously (11). CMX-VP16-RXR contains the
complete coding sequence of the hRXR
cloned in frame after a
sequence encoding the 78 amino acids from the transactivation domain
encoded in pVP16C1 (Novagen, Madison, WI) (28). The entire coding
sequences of hRAR
, hRXR
, hTR
, VDR, and mPPAR
were all
cloned in pCMX and have been described previously (29, 30).
Human chorion carcinoma JEG-3 cells
were maintained in Dulbecco's modified Eagle's medium supplemented
with 10% fetal calf serum. Transfections were performed in duplicate
in 24-well plates by the calcium phosphate method as described
previously (28). Each well was transfected with 100 ng of reporter
plasmid, 100 ng of receptor expression vector, and 200 ng of
pCMX-gal reference plasmid containing the
-galactosidase gene.
Additions to each well were adjusted to contain a constant amount of
DNA and of pCMX expression vector. After 6-12 h, the cells were washed
once with phosphate-buffered saline, and fresh medium was added
containing 10% charcoal-stripped fetal calf serum and the indicated
ligands. The cells were harvested after 36-h incubation and lysed, and extracts were assayed for luciferase and
-galactosidase activity in
a microplate luminometer/photometer reader (Lucy-1, Anthos, Salzburg,
Austria). Variance between duplicate data points were less than 20% in
each experiment. Twenty million U-937 cells were harvested, washed, and
suspended in a cuvette in 0.5 ml of RPMI 1640 containing 20 µg of
REx3-tk-luc plasmid and 20 µg of RSV-
gal vector. The cells were
transfected by electroporation at 0.30 kV/960 microfarads (Gene Pulser,
Bio-Rad), left on ice for 10 min, and then transferred to culture
medium. Multiple transfections of cells were pooled and allowed to
recover in the incubator for 1-2 h. Then the cells were split and
stimulated over night in triplicate with retinoids as indicated. Cell
extracts were prepared by three cycles of freezing and thawing in 0.1 M potassium phosphate (pH 7.4) and assayed for luciferase
and
-galactosidase activity (luminescent
-galactosidase detection
kit, Clontech, Palo Alto, CA) in a luminometer (Lumat LB9501, Berthold,
Bad Wildbad, Germany). Variance between data points within each
experiment were less than 25%. All luciferase activities were
normalized to
-galactosidase activity.
The human U-937 cell line is arrested at a monoblastic stage of hematopoietic development and differentiates along the monocytic lineage upon atRA treatment (31). During the differentiation process the cells up-regulate surface antigens associated with the terminally differentiated phenotype and acquire a monocyte-like morphology. High concentrations of atRA (in the micromolar range) in cell cultures results in isomerization to 9cRA, which can activate both types of retinoid receptors. Therefore, to analyze the individual contribution of RAR and RXR activation for the differentiation process, we stimulated U-937 cells with TTNPB and SR11237, synthetic retinoids that specifically bind and activate RAR and RXR, respectively (19, 32). After 4 days of 9cRA stimulation over 90% of the U-937 cells express the surface antigens CD11c, CD49f, and CD66a, which are characteristic markers of the differentiated phenotype (Fig. 1A) (26). In contrast, TTNPB and SR11237 are by themselves poor inducers of antigen expression. However, when administered together, TTNPB and SR11237 induce differentiation antigens as efficiently as 9cRA (Fig. 1A).
RA-differentiated U-937 cells develop a characteristic morphology with lobulated nuclei as shown in Fig. 1B. Cells stimulated for 8 days with either TTNPB or SR11237 retained the blast-like phenotype of nondifferentiated cells. However, co-stimulation with TTNPB and SR11237 resulted in the typical morphology of mature atRA-differentiated cells (TT+SR; Fig. 1B). Similarly, TTNPB and SR11237 synergize in the induction of growth arrest in the G0/G1 phase of the cell cycle, consistent with the tight coupling between cessation of proliferation and terminal differentiation in these cells (Fig. 1C). Flow cytometric analysis of propidium iodide-stained cell nuclei prepared from cells treated with either TTNPB or SR11237 showed a slightly larger fraction of the cells in the G1/G0 peak, compared with unstimulated exponentially growing cells, indicating some inhibition of proliferation. However, a complete cell cycle arrest in the G0/G1 phase, similar to that observed for atRA, required co-stimulation with TTNPB and SR11237. Thus, the activation of both RAR and RXR is critical for the efficient induction of the hallmarks of RA induced differentiation, i.e. growth arrest in G1/G0, up-regulation of differentiation antigens, and development of the distinct morphology of mature monocytes.
We next wished to examine if the effect of combined RAR and RXR
stimulation on U-937 cell differentiation could result from synergism
in transactivation of an RARE. Transient transfections of U-937 cells
were performed with a reporter construct containing the RARE from the
RAR2 promoter (
RE) (33, 34). 9cRA is a strong inducer
of
RE reporter activity through the endogenous receptors of U-937
cells (35) (Fig. 2). TTNPB stimulation resulted in a
moderate activation while SR11237 only induced a minor response. Co-stimulation with TTNPB and SR11237 potently activated the reporter. This effect was more than additive, indicating synergism between RAR
and RXR activation. Since the
RE is recognized by RAR/RXR heterodimers in U-937 cells (35), the results support the possibility that both subunits in such complexes are activated in the presence of
TTNPB and SR11237.
Both Subunits in RAR/RXR Heterodimers Can Be Activated when Receptors Are Simultaneously Exposed to RAR and RXR Ligands
To
investigate the possibility that both RAR and RXR subunits can be
activated in RAR/RXR heterodimers, we wished to assay RXR activation
independently of RAR. Therefore we utilized GAL4 receptor hybrids and a
luciferase reporter containing GAL4 binding sites in its promoter (Fig.
3A). A two-amino acid substitution was
introduced in the RAR activation domain (AF2) that would eliminate activation through the RAR subunit without affecting its ability to
bind ligand. This assumption relies on previous mutations introduced in
RAR in which activation, but not ligand binding, is abolished (6). The
GAL4 hybrid receptor containing the ligand binding domain of such an
RAR derivative (GAL4-RARmAF2; Fig. 3A) can
heterodimerize with a VP16-fused RXR as efficiently as GAL4-RAR as
demonstrated in the two-hybrid transfection experiment performed in
human JEG-3 cells (Fig. 3B). VP16-fused RXR alone did not
activate the reporter (data not shown). Importantly, the data in Fig.
3C demonstrate that heterodimers formed between the
GAL4-RARmAF2 and RXR can be activated by SR11237 in the presence but
not in the absence of the RAR agonist TTNPB. Similarly, a low dose of
atRA also allows such activation through the RXR subunit of the
GAL4-RARmAF2/RXR heterodimer. Notably, activation by RXR is weak when
heterodimerizing with GAL4-RARmAF2, indicating that the RAR AF2
mutation affects the efficiency by which RXR can be activated (see
below). The data corroborate previous observations showing that RXR is
allosterically blocked in complex with RAR, but clearly demonstrate
that inhibition can be relieved by binding of an RAR agonist to the RAR
subunit of the heterodimer.
An RAR Antagonist Can Relieve the Allosteric Block of RXR
We
next wished to test if also an RAR antagonist could lead to a
conformational shift that would allow RXR to be activated in an RAR/RXR
heterodimer. Therefore, transfected JEG-3 cells were treated with
SR11237 and an RAR antagonist, Ro41-5253 (36), which binds to RAR but
does not lead to its activation (Fig. 4A). Interestingly, heterodimers formed between GAL4-RAR and RXR respond weakly to SR11237 in the presence, but not in the absence, of Ro41-5253, demonstrating that RXR inhibition is partly relieved under
these conditions (Fig. 4B).
RXR/RXR Homodimers and RAR/RXR Heterodimers Are Pharmacologically Distinct in Their Response to Synthetic RXR Ligands
Our results
together with previously published data clearly show that RAR can
allosterically influence RXR. Therefore we analyzed the pharmacological
properties of the RAR-RXR complex in comparison with RXR/RXR homodimers
in response to different RXR activators. In these experiments
dose-response curves using different RXR activators were determined for
activation by GAL-RXR or GAL-RARmAF2/RXR heterodimers (Fig.
5A). Cells expressing either GAL-RXR or the combination of GAL-RARmAF2 and RXR were treated with increasing amounts
of the different RXR activators. Moreover, GAL-RARmAF2- and
RXR-transfected cells were simultaneously incubated with a constant
amount of TTNPB, since a ligand for RAR is a prerequisite for RXR
activation as demonstrated above. Activation by three compounds,
SR11237, LG153, and atRA, showed similar dose-response curves when
comparing GAL4-RXR (homodimer) and GAL4-RARmAF2/RXR (heterodimer)
activation (Fig. 5A). Similar results were seen with RXR
activators such as 9cRA and methoprene acid (37) (data not shown).
Interestingly, as demonstrated in Fig. 5B, one of the tested
ligands, SR11234 (19), was a weak activator of the GAL4-RARmAF2/RXR
heterodimer, whereas it potently activated GAL4-RXR. These results
demonstrate that the two complexes are pharmacologically distinct and
also suggest that RXR homodimers and RAR/RXR heterodimers respond
differently to RXR ligands.
The AF2 Domain of Both Subunits Contributes to Activation of RXR Heterodimeric Complexes
Notably, RXR is only weakly activated in
complex with RAR in the experiments displayed in Figs. 3C
and 4B. In these experiments the AF2 of RAR is inactive,
either because the RAR AF2 is mutated (Fig. 3) or an RAR antagonist is
used (Fig. 4). Thus, the data indicate that efficient activation is
dependent on the functional integrity of both AF2 domains of a
heterodimeric complex. To test if the activation of RAR, in the context
of an RAR/RXR heterodimer, also depended on two AF2 domains, a
two-amino acid substitution was introduced in RXR AF2 (Fig.
6A). This RXR derivative (GAL4-RXRmAF2) is
inactive when transfected cells are treated with SR11237 but heterodimerizes efficiently with RAR, as indicated by the two-hybrid interaction between GAL4-RXRmAF2 and a VP16-RAR derivative (data not
shown). Indeed, a heterodimer formed between GAL4-RXRmAF2 and RAR is
only weakly active in the presence of TTNPB, demonstrating that the AF2
function of RXR is required, although TTNPB only binds to the RAR
subunit of the complex. In addition, other heterodimeric partners,
including TR, VDR, and PPAR, also require an intact RXR AF2 domain,
since activation is diminished when these receptors are co-expressed
with GAL4-RXRmAF2 and treated with triiodothyronine, vitamin
D3, and the specific PPAR
ligand BRL49653 (38),
respectively (Fig. 6B). Consistent with this result, a
truncated derivative of RXR (RXR443), which lacks an intact AF2 domain,
acts as a dominant negative inhibitor of RAR activation from a reporter
containing the
RE (Fig. 6C). Together, these results
demonstrate that an intact RXR AF2 is required for efficient ligand
activation of several RXR partners.
RXR participates in several signaling pathways by forming heterodimers with receptors such as RAR, TR, and VDR. In such heterodimers; however, RXR has been suggested to function as a silent partner, which promotes high affinity DNA binding of the ligand binding subunit to its specific hormone response elements (18). To support this notion we show in this report that RXR ligands have little effect on U-937 cells. Still, some reports favor the idea that both subunits in an RXR heterodimer can be activated by ligand, as supported by cooperativity between RAR and RXR ligands in RARE activation and induction of RA-induced genes (6, 23). Consistent with this concept our results demonstrate that both subunits in the RAR/RXR heterodimer can indeed be activated when both RAR and RXR ligands are present simultaneously. This also appears to be a prerequisite for the biological response, i.e. differentiation of U-937 cells.
The combined use of the GAL4 system and the RAR AF2 mutant demonstrated how RAR and RXR ligands affected RXR activation in a context where the luciferase reporter gene was not affected by ligand dependent trans-activation of RAR. Also, this experimental approach ensured that transcriptional influence from various endogenous receptor complexes was avoided. The results show that RXR is responsive to selective RXR ligands in the presence, but not in the absence, of RAR ligands. Previously published experiments demonstrated that RXR is allosterically inhibited in complex with RAR, which could be mechanistically explained by RXRs inability to bind ligand in RAR/RXR heterodimers (20, 21). Our experiments confirm this view but, in addition, suggest that RAR ligand induces a conformational change which not only affects the ligand binding subunit, RAR, but is also translated by RXR to relieve inhibition of ligand binding. Previous biochemical experiments support this view and have suggested that both subunits of RXR/RAR heterodimers can indeed associate with ligand (21, 24).
Surprisingly, even an RAR antagonist combined with RXR ligands allow
RXR activation. It was recently shown that a combination of an RAR
antagonist and an RXR agonist can induce differentiation and apoptosis
of promyelocytic NB4 cells, which express the PML/RAR fusion
oncoprotein (39). Our experiments demonstrate that cooperation between
RAR antagonists and RXR agonists in these cells may be explained by
allosteric receptor interactions within RAR/RXR heterodimer complexes.
Using several synthetic RXR ligands we could show that the RAR/RXR heterodimer is pharmacologically distinct from RXR/RXR homodimers. The experiments demonstrated that the synthetic RXR ligand SR11234, as opposed to other tested RXR ligands, is a weak activator of RAR/RXR heterodimers. Since a low level of activation could be observed (Fig. 5; data not shown) we conclude that SR11234 can associate with RAR/RXR heterodimers and that the deficiency resides in an inability to promote efficient activation rather than ligand binding to the RAR/RXR heterodimer. SR11234 and SR11237 are structurally similar benzoic acid derivatives with ketal and thioketal substituent groups, respectively (19). Thus, we speculate that the ketal as opposed to the thioketal group may interfere with the efficiency by which co-activators and/or co-repressors assemble with RAR/RXR heterodimers but not RXR/RXR homodimers. Such differences in factor assembly have been demonstrated recently comparing the activities of novel synthetic RXR ligands with opposite effects on RAR/RXR heterodimers (40).
RA has been used in treatment of a number of different cancers, such as promyelocytic leukemia, head and neck carcinoma, and skin cancer (41). Despite these interesting clinical developments, remarkably little is known as to how RA acts to suppress cancer cell growth. Previously published data have established that certain RAR antagonists, termed "dissociated retinoids," are functional in their ability to inhibit AP1 activity, although they fail to trigger transcriptional activation by RAR (42, 43). These retinoids specifically inhibited the growth of lung and breast cancer cells, whereas growth suppression of other cell types required "classical" activation of retinoid receptors by atRA (42). In addition, several naturally occurring retinoids exist, some of which are potent activators of RAR (44-46). It is plausible that also naturally occurring compounds, similar to SR11234, will eventually turn out to selectively affect distinct RXR complexes. Therefore highly selective cellular responses could be achieved with different combinations of synthetic and natural retinoids. Clinically, it can be anticipated that combinations of pharmacologically distinct retinoids could increase both the specificity and the potency of retinoid action in cancer therapy.
The emerging knowledge on co-activators/repressors and the
determination of the crystal structure of nuclear receptor ligand binding domains may provide clues as to the mechanism behind the versatile allosteric control of RXR reported in this paper. Although the structure for one and the same receptor has yet to be shown both in
a ligand-bound and unbound state, several important conclusions can be
made on the nature of the conformational change occurring in liganded
receptors (7-9). Comparison of the apo-LBD of RXR with the
holo-LBDs of RAR
and TR
suggests that a ligand-induced conformational shift creates a more compact LBD fold (47). Furthermore, the protruding AF2 core
-helix, encompassing helix 12 in the described nuclear receptor LBD structure, folds back against the LBD,
where it will be in close contact with the bound ligand. Thus, ligand
induces both local and global conformational changes that easily could
be envisioned to affect a heterodimerization partner. Also, several
studies have demonstrated the importance of the AF2 core for
association with co-activators in the presence of ligand (10-13).
Unexpectedly, our results demonstrated that the integrity of the AF2
domain of both subunits in RXR heterodimers are important for efficient
activation, even when only one subunit of the heterodimer is activated
by ligand. Experiments with an RXR AF2 mutant emphasized the importance
of two AF2 domains in heterodimers such as RAR/RXR, TR/RXR, VDR/RXR,
and PPAR/RXR, in which ligand-induced transcription depended on the AF2
domain of the nonliganded partner RXR. Conceivably, ligand binding to the heterodimerization partner could promote a conformational shift,
which also affects the RXR AF2 and induces a conformation that
resembles the "activated" state. Alternatively, co-activators may
associate weakly with RXR AF2 domains also in the absence of RXR
ligands. In a heterodimer with only one liganded receptor, synergistic
co-activator binding may result from a combined weak and strong
association with the nonliganded and liganded receptor partners,
respectively.
In conclusion, ligand binding to RAR has two important influences on RAR/RXR heterodimers. First, RAR ligand binding results in a partial transcriptional activation that is dependent on the RXR AF2. Second, an allosteric change is induced that allows RXR to bind ligand. Furthermore, in the presence of both ligands, heterodimers mediate synergistic transcriptional activation. Our data help to explain why synergistic physiological responses to both RAR and RXR are observed in many cell types such as U-937 cells. Also, a common and consistent feature of most, if not all, mammalian cells is the combined expression of RAR and RXR, suggesting a critical and central role for RAR/RXR heterodimers in responses to retinoids. The results emphasize the importance of observing each receptor in a heterodimer as an integrated component in a complex in which the receptors are regulated by mutual interactions involving the transactivation and ligand binding domains of both receptors.
We thank Michael Klaus (Hoffmann-La Roche, Basel, Switzerland), Louise Foley (Hoffmann-La Roche, Nutley, NJ) Rich Heyman (Ligand Pharmaceuticals), Steve Kliewer (GlaxoWellcome), and U. Fischer (Hoffmann-La Roche, Basel) for the gifts of nuclear receptor ligands and Ralf Pettersson for critical comments on the manuscript.