(Received for publication, December 18, 1996, and in revised form, March 5, 1997)
From the A subset of nuclear receptors, including those
for thyroid hormone (TR), retinoic acid, vitamin D3,
and eicosanoids, can form heterodimers with the retinoid X receptor
(RXR) on DNA regulatory elements in the absence of their cognate
ligands. In a mammalian two-hybrid assay, we have found that
recruitment of a VP16-RXR chimera by a Gal4-TR The thyroid hormone receptor (TR)1 is
a member of the nuclear receptor superfamily of ligand-inducible
transcription factors. A subset of receptors, including those for
estrogen and glucocorticoid, bind to regulatory sequences in target
gene promoters as homodimers, with cooperative DNA binding and
dimerization mediated by regions within their DNA- and ligand-binding
domains (1-7). The thyroid hormone receptor is also capable of binding
to thyroid hormone response elements (TREs) as a monomer or homodimer,
but the latter complex is known to dissociate in the presence of
thyroid hormone (T3) (8-11). More recently, it has been
shown that TR as well as receptors for vitamin D3, retinoic
acid (RAR), and eicosanoids form heterodimeric complexes with an
auxiliary factor, the retinoid X receptor (RXR) (12-22). The TR·RXR
heterodimer binds TREs with higher affinity than TR alone and remains
stable in the presence of T3 (8-11), suggesting that this
complex mediates ligand-dependent transcriptional
activation.
The isolated DNA-binding domains (DBDs) of TR and RXR are capable of
both cooperative binding to and discrimination of direct repeat TREs
(23). The dimerization interface between DBDs was mapped initially by
mutational analyses (24) and then elucidated from the crystal structure
of TR·RXR DBDs bound to a DR4 TRE (25). A separate dimerization
interface in the carboxyl-terminal domains of nuclear receptors was
first delineated as a series of nine conserved hydrophobic heptad
repeats (26) predicted to form Several studies have shown a solution interaction between RXR and other
nuclear receptors in the absence of DNA or ligand. Thus, RAR·RXR or
TR·RXR complexes can be coimmunoprecipitated in the absence of ligand
(18-20), and a constitutive interaction between the LBDs of RXR and
RAR (31) or TR (10, 32) has been demonstrated in mammalian cells and
yeast (33). However, the influence of ligand on dimerization between
RXR and TR is unclear. In this study, we have investigated the role of
T3 in modulating the interaction between the LBDs of TR The wild-type
pSG-TR and pSG-RXR mammalian expression vectors, respectively, contain
the LBDs of human TR JEG-3 human
choriocarcinoma cells were cultured in Opti-MEM supplemented with 2%
(v/v) fetal bovine serum and 1% penicillin/streptomycin/Fungizone and
transferred to Opti-MEM with 1% penicillin/streptomycin/Fungizone prior to transfection by a 4-h exposure to calcium phosphate as described previously (10). Following a 36-h incubation, cells were
lysed, and luciferase activity was measured and normalized using
SFY526 yeast cells were
cultured and cotransformed with the appropriate expression vectors and
assayed following the protocol detailed by the manufacturer
(CLONTECH). GST-RXR fusion protein
was expressed in Escherichia coli and purified as described
previously (32). [35S]Methionine-labeled Gal4-TR To examine the
interaction between the ligand-binding domains of TR and RXR, we used a
mammalian two-hybrid assay in which chimeric proteins containing the
LBDs of human TR
Since transcriptional activation by nuclear receptors is
generally dependent on cognate ligand, we next examined the influence of T3 in the above paradigm. As noted previously, there was
a significant constitutive interaction between Gal4-TR(wt) and
VP16-tagged full-length RXR or the RXR LBD (Fig. 2,
bars 2 and 3). The addition of thyroid hormone
enhanced reporter gene activity by 25-50-fold. One interpretation of
this finding was that the additional activation represented increased
recruitment of VP16-RXR to the promoter via its interaction with
Gal4-TR. However, T3 also significantly augmented the
activity of Gal4-TR alone (Fig. 2, bar 1), presumably by
stimulating the known ligand-dependent transcriptional
activation function (AF-2) contained within the TR LBD. This raised the
possibility that the ligand-dependent increase in reporter
activation was due to the AF-2 activity of TR superimposed on
activation associated with the VP16 moiety.
To differentiate between these possibilities, we used a
mutant Gal4-TR fusion containing a single amino acid substitution (L454A) in the carboxyl-terminal amphipathic Although full-length TR
Unliganded TR is known to mediate repression of basal
gene transcription, and this silencing function has been localized to its LBD (35, 36). Recently, two putative corepressor proteins (N-CoR
and SMRT) have been identified that bind TR·RXR heterodimers in the
absence of ligand, but that dissociate following ligand binding (37,
38). This raised the possibility that ligand-induced release of
corepressor(s) by promoter-bound Gal4-TR·VP16-RXR complexes could
account for the T3-dependent increased reporter
activity observed in two-hybrid assays. To discount this possibility,
we studied another mutant Gal4-TR fusion, AHTm, which contains a triple
substitution (A223G/H224G/T227A)2 that has
been shown to abolish the interaction of the TR·RXR heterodimer with
N-CoR (38). When tested independently (Fig. 4, bar
2), hormone-dependent reporter gene activation by
Gal4-TR(AHTm) was preserved, but significantly reduced (15%) compared
with Gal4-TR(wt) (bar 1), indicating that this mutant
receptor retains ligand binding and activation functions. The
Gal4-TR(AHTm) mutant unexpectedly showed negligible constitutive
interaction with coexpressed VP16-RXR compared with the wild-type
receptor fusion (Fig. 4, bars 3 and 4). Following
T3 treatment, there was a significant enhancement of
reporter activity with Gal4-TR(wt), as seen previously, but in
addition, Gal4-TR(AHTm) mediated a similar response, suggesting that
corepressor release did not account for this effect.
To determine whether T3 could enhance
dimerization between the TR and RXR LBDs, we utilized an in
vitro protein-protein interaction assay in which GST-RXR
Finally, we examined the influence of
9-cis-RA, the ligand for RXR, on the interaction of Gal4-TR
and VP16-RXR in the mammalian cell two-hybrid system. Either the Gal4
DBD or Gal4-TR(L454A) was coexpressed with VP16-tagged full-length RXR
or the RXR LBD. The presence of 1 µM 9-cis-RA,
in the absence of T3, attenuated the constitutive
interaction of Gal4-TR(L454A) with VP16-RXR, particularly with the
full-length VP16-RXR fusion (Fig. 6, bar 9 versus
bar 11). The enhanced interaction of Gal4-TR(L454A) with VP16-RXR
following T3 was also attenuated by 9-cis-RA
(Fig. 6, bar 6 versus bars 8 and 10 versus bar
12).
Heterodimerization between TR and RXR is a central mechanism for
T3-induced regulation of many target gene promoters. While a number of studies in vitro have suggested that the
formation of TR·RXR complexes on DNA is not significantly influenced
by ligand, the role of hormone in mediating solution interactions between these receptors has not been elucidated. Here we have shown a
constitutive heterodimeric interaction between the LBDs of TR and RXR
in the absence of ligand, confirming earlier observations (20, 32, 40),
while in addition demonstrating a lack of homodimeric interaction for
either LBD. This latter finding suggests that the two receptor LBDs
possess complementary dimerization interfaces and, furthermore, that
their DBDs and the response element to which they bind also play a
major role in determining the nature of the dimeric complex since
TR·TR and RXR·RXR homodimers do form readily on DNA (24, 41, 42).
We have made the additional observation that the addition of
T3 markedly increased reporter gene activity up to 50-fold
in cells containing coexpressed Gal4-TR and VP16-RXR fusion proteins,
suggesting that T3 enhances the recruitment of VP16-RXR to
promoter-associated Gal4-TR. However, alternative explanations were
also examined.
The possibility that the enhanced activity represented
T3-mediated induction of the ligand-dependent
transactivation function (AF-2) present in the TR LBD was discounted
since this effect was also observed with the mutant thyroid hormone
receptor fusion protein Gal4-TR(L454A), which lacks AF-2 activity, as
well as in yeast, where the AF-2 function of the TR LBD is not
manifest. The suggestion that the apparent enhanced recruitment of
VP16-RXR was due to T3-mediated release of corepressor(s)
was eliminated since Gal4-TR(AHTm), which contains mutations shown to
abrogate the interaction with the putative transcriptional corepressor N-CoR (38), also showed T3-dependent
recruitment of VP16-RXR. Furthermore, while previously reported
T3-induced dissociation of TR homodimers may facilitate
interaction of the Gal4-TR monomer with RXR (9, 11), the data in Fig. 1
show no evidence of TR LBD homodimer formation. We considered the
possibility that the apparent enhanced recruitment of VP16-RXR was not
due to augmentation of a direct interaction between TR and RXR LBDs,
but the result of ligand-dependent recruitment of
additional cofactors, capable of binding to either receptor partner and
promoting their interaction (43). Using a two-hybrid assay system in
mammalian cells, Kraus et al. (44) have shown a
ligand-dependent interaction between the amino- and
carboxyl-terminal regions of the estrogen receptor that is markedly
attenuated by a mutation in the AF-2 region of the carboxyl-terminal
domain. One mechanism proposed by the authors involved
ligand-dependent recruitment of an accessory factor to the
AF-2 region, thereby facilitating the interaction between the two
receptor regions. However, our data do not favor such a mechanism for
two reasons. First, we have recently shown that the L454A mutation in
the AF-2 region of TR abrogates its ligand-dependent interaction with at least two putative intermediary proteins (SRC-1 and
RIP140) (45), yet Gal4-TR(L454A) retained the ability to recruit
VP16-RXR in response to ligand; second, we have also shown that
T3 can enhance the association of Gal4-TR and GST-RXR by up
to 10-fold (Fig. 5, A and B) in vitro,
a context in which facilitation by auxiliary factors is unlikely.
Furthermore, the in vitro data also rule out the possibility
that the augmented reporter gene activity in vivo was
mediated by a ligand-dependent alteration in Gal4-TR
expression or stability leading to increased VP16-RXR recruitment.
There is precedent for our observations from previous studies with
other members of the nuclear receptor family.
Ligand-dependent homodimerization between fusion proteins
containing the activation or DNA-binding domains of Gal4 linked to the
full-length estrogen receptor has been demonstrated using a two-hybrid
assay system in yeast (46). The solution interaction of the vitamin
D3 receptor with GST-RXR, in the absence of DNA, has been
examined using surface plasmon resonance (47). The addition of the
ligand 1,25-dihydroxyvitamin D3 enhanced the affinity of
the vitamin D3 receptor for RXR by 7-fold as a result of an
increase in the association rate constant, and we hypothesize that the
T3-mediated enhancement of the TR-RXR interaction might
involve a similar mechanism. Another example in which ligand promotes
dimerization is provided by the ecdysone receptor-ultraspiracle
interaction, where muristerone has been shown to promote formation of
ecdysone receptor-ultraspiracle complexes as well as formation of
ecdysone receptor-ultraspiracle or ecdysone receptor-RXR heterodimers
on DNA (48). Last, a genomic footprinting study has shown that
occupancy of the retinoid response element in the RAR Studies using circular dichroism spectroscopic analysis have provided
evidence for a ligand-induced conformational change in TR, while gel
shift experiments have shown that the mobility of DNA-bound complexes
is increased with T3 (9, 11, 50). Crystallographic studies
of the LBD of the Although RXR enhances DNA binding by TR and subsequent
T3-induced receptor signaling, the role of
9-cis-RA in TR·RXR dimerization is unclear. The RXR moiety
of DNA-bound receptor heterodimers typically binds to the 5 We have shown that the solution interaction between the ligand-binding
domains of TR and RXR is markedly enhanced by T3. Although TR can clearly form unliganded complexes with RXR on cognate response elements to mediate silencing of basal transcription in many contexts (40), the ability to form TR·RXR heterodimers in solution may confer
certain advantages. First, since RXR is a factor common to the
signaling pathways of a number of other nuclear receptors, it might not
always be expedient for RXR to be sequestered into DNA-bound unliganded
receptor complexes, except when in a repressing role. Accordingly,
ligand-dependent dimerization in solution would facilitate
channeling of RXR to the desired signaling pathway and may also confer
a negative regulatory influence on other receptor signaling pathways
requiring this factor. We therefore propose that ligand-enhanced
heterodimerization prior to DNA binding may constitute an additional
early level of control in nuclear receptor-mediated gene
expression.
Department of Medicine,
ligand-binding domain
fusion is enhanced up to 50-fold by thyroid hormone (T3).
This was also observed with a mutant fusion, Gal4-TR(L454A), lacking
ligand-inducible activation function (AF-2) and unable to interact with
putative coactivators, suggesting that the AF-2 activity of TR or
intermediary cofactors is not involved in this effect. The wild-type
and mutant Gal4-TR fusions also exhibited hormone-dependent
recruitment of RXR in yeast. Hormone-dependent recruitment
of RXR was also evident with another Gal4-TR mutant, AHTm, which does
not interact with the nuclear receptor corepressor N-CoR, suggesting
that ligand-enhanced dimerization is not a result of
T3-induced corepressor release. Finally, we have shown that
the interaction between RXR and TR is augmented by T3
in vitro, arguing against altered expression of
either partner in vivo mediating this effect. We propose
that ligand-dependent heterodimerization of TR and RXR in
solution may provide a further level of control in nuclear receptor
signaling.
-helices with the potential to
mediate a coiled-coil protein interaction. A 40-amino acid region
encompassing the ninth heptad has been shown to be critical for
heterodimerization (27). This region is almost coincident with helical
segments (helices 9 and 10) that constitute the dimer interface in the
crystal structure of the ligand-binding domain (LBD) of hRXR
(28).
In addition, mutation of hydrophobic residues at either end of this
heptad in TR or RAR abolishes the interaction with RXR (10, 29,
30).
and RXR
.
Expression Vectors and Reporter Constructs
(amino acids 174-461) and RXR
(amino acids
198-462) coupled to the Gal4 DBD (amino acids 1-147) in pSG424.
Mutant expression vectors include pSG-TR(L454A), containing a L454A
substitution (32), and pSG-TR(AHTm), with the triple substitution
A223G/H224G/T227A in pSG-TR. pVP-RXR-(1-462) and pVP-RXR(DE) consist
of either full-length or the LBD (amino acids 198-462) of hRXR
,
respectively, coupled to the activation domain of VP16 in the
expression vector pAASV. The reporter plasmid UAStkLUC contains Gal4
recognition sequences and the thymidine kinase promoter in
pA3LUC. The internal control plasmid Bos-
gal contains
the promoter of human elongation factor 1
linked to
-galactosidase (32). The yeast expression vectors pMA-TR(wt) and
pMA-TR(L454A) contain the LBDs of either wild-type human TR
or the
L454A mutant described above coupled to the Gal4 DBD in pMA424,
respectively, and pMA210 contains full-length Gal4 cDNA. pGAD-RXR
consists of residues 198-462 of hRXR
coupled to the activation
domain of Gal4 in pGAD10. The bacterial expression vector pGEX-RXR
contains full-length hRXR
linked to glutathione S-transferase (GST) in pGEX2T.
-galactosidase activity.
-Galactosidase Assays
-Galactosidase activity in liquid
culture assays was corrected for cell density. Cells were grown in the
absence or presence of 1 µM triiodothyroacetic acid
(TRIAC).
was
synthesized by coupled in vitro transcription and
translation. Aliquots of GST-RXR immobilized on glutathione-Sepharose
were preincubated with 4 mg ml
1 bovine serum albumin in P
buffer (50 mM Tris-HCl, 0.1 M KCl, 0.14 M NaCl, 0.5% Nonidet P-40, 10% glycerol, and 2 mM dithiothreitol, pH 8.0) at room temperature prior to
10-fold dilution in P buffer and equilibration at 37 °C. Aliquots of
in vitro translated Gal4-TR were diluted in P buffer in the
absence or presence of 5 µM T3 and
equilibrated at 30 °C for 1 h to allow saturation with ligand. Equal volumes of the GST-RXR and Gal4-TR solutions were then mixed and
incubated at 37 °C for the specified duration, followed by washing
four times with 20 mM Tris-HCl, 0.1 M NaCl, 1 mM EDTA, and 0.5% Nonidet P-40, pH 8.0, and analysis by
SDS-polyacrylamide gel electrophoresis. Gels were stained with
Coomassie Blue to check for equal loading of fusion protein and then
exposed to autoradiography.
Gal4-TR and VP16-RXR Chimeras Interact in Vivo
or hRXR
or full-length hRXR
were fused to
either the Gal4 DBD or the VP16 activation domain and coexpressed in
JEG-3 cells (Fig. 1). Transfection of either receptor
fusion with the Gal4 DBD or VP16 alone resulted in negligible
transcriptional activity. Similarly, coexpression of Gal4-TR and
VP16-TR or Gal4-RXR and VP16-RXR did not enhance promoter activity.
However, cotransfection of TR and RXR chimeras resulted in a
significant increase in reporter gene activity, suggesting that a
constitutive interaction between the LBDs of these receptors does occur
in vivo.
Fig. 1.
Constitutive heterodimeric, but not
homodimeric, interaction of TR and RXR LBDs in vivo.
JEG-3 cells in six-well plates were transfected with 0.5 µg of each
of the specified Gal4- and VP16-receptor fusion vectors, along with 5 µg of UAStkLUC reporter and 0.3 µg of Bos-gal internal control
plasmid. -Fold induction of luciferase activity is relative to Gal4-TR + VP16. Data are the means ± S.E. of three separate
experiments.
[View Larger Version of this Image (25K GIF file)]
Fig. 2.
T3 enhances the interaction
between TR and RXR in mammalian cells. JEG-3 cells in 24-well
plates were cotransfected with 50 ng of pSG-TR(wt) (Gal4-TR(wt)),
pSG-TR(L454A) (Gal4-TR(L454A)), pAASV (VP16), pVP-RXR
(VP16-RXR-(1-462)), and pVP-RXR(DE) (VP16-RXR(DE)) as specified, along
with 500 ng of UAStkLUC reporter and 100 ng of Bos-gal internal
control. The cells were then incubated in the absence or presence of
0.1 µM T3. -Fold induction of luciferase activity is relative to Gal4-TR(wt) in the absence of ligand. Data are
the means ± S.E. of at least three individual experiments. Note
the discontinuous scale of the vertical axis.
[View Larger Version of this Image (23K GIF file)]
-helix that abolishes the AF-2 activity of TR without altering its ligand binding function (32). The mutant Gal4-TR(L454A) chimera also exhibited constitutive interaction with VP16-RXR (Fig. 2, bars 5 and 6),
and again, reporter activity was markedly enhanced with the addition of
T3. However, the mutant Gal4-TR(L454A) fusion exhibited no
ligand-dependent transcriptional activation when tested
alone (Fig. 2, bar 4), suggesting that the
T3-dependent increase in reporter activity was
not due to AF-2 activity, but rather to an increased recruitment of
VP16-RXR.
has been shown to exhibit both
constitutive and ligand-dependent activation functions with
TRE-containing reporter genes in yeast (34), a chimera consisting of
the TR
LBD coupled to the DBD of LexA was transcriptionally inactive in Saccharomyces cerevisiae (33). This lack of AF-2 activity in yeast prompted us to examine the influence of ligand on the interaction of TR and RXR LBDs to substantiate our findings in mammalian cells. Wild-type or mutant (L454A) Gal4-TR fusions were cotransformed into yeast with either the activation domain (AD) of Gal4
alone or a chimera containing the LBD of RXR fused to the activation
domain (AD-RXR). Recombinant yeast cells were grown either in the
absence or presence of TRIAC prior to detection of
-galactosidase
expression (Fig. 3A). Treatment with TRIAC did not induce
-galactosidase activity with coexpressed Gal4-TR(wt) and AD, but ligand markedly enhanced
-galactosidase expression with
a Gal4-TR(wt) and AD-RXR combination. To exclude the possibility that
the coexpressed RXR was simply facilitating the inherent AF-2 activity
of Gal4-TR, we tested the Gal4-TR(L454A) mutant and found a comparable
TRIAC-dependent enhancement of
-galactosidase activity
with AD-RXR. To quantify the magnitude of the
ligand-dependent enhancement, yeast cells containing either
wild-type or mutant (L454A) Gal4-TR fusions, together with either AD or
AD-RXR, were grown in the absence or presence of TRIAC, and cell
extracts were assayed for
-galactosidase activity (Fig.
3B). The results confirmed that neither the wild-type nor
the mutant (L454A) Gal4-TR fusions exhibit AF-2 activity following
TRIAC treatment. Both fusions showed a comparable constitutive
association with coexpressed AD-RXR that was enhanced 20-25-fold by
the addition of TRIAC.
Fig. 3.
Interaction between TR and RXR in yeast is
enhanced by ligand. A, in a yeast two-hybrid assay, cells
were transformed with pMA210 + pGAD10 (GAL4), pMA424 + pGAD-RXR (Gal + AD-RXR), pMA-TR(wt) + pGAD10
(Gal-TR(wt) + AD), pMA-TR(wt) + pGAD-RXR (Gal-TR(wt) + AD-RXR), or pMA-TR(L454A) + pGAD-RXR (Gal-TR(L454A) + AD-RXR) and grown on plates in the absence or presence of 1 µM TRIAC. Colonies were transferred to filters and
assayed for -galactosidase reporter expression using
5-bromo-4-chloro-3-indolyl
-D-galactopyranoside as
substrate. B, yeast cells were transformed with pMA-TR(wt) + pGAD10 (WT), pMA-TR(L454A) + pGAD10 (L454A),
pMA-TR(wt) + pGAD-RXR (WT + AD-RXR), or pMA-TR(L454A) + pGAD-RXR (L454A + AD-RXR) and then grown in liquid culture
in the absence (open bars) and presence (closed
bars) of 1 µM TRIAC. Data are the means ± S.E.
of at least four individual experiments. Note the discontinuous
vertical axis.
[View Larger Version of this Image (34K GIF file)]
Fig. 4.
T3-induced TR-RXR interaction is
not due to corepressor release. JEG-3 cells were transfected as
described in the legend to Fig. 2, but with the inclusion of
pSG-TR(AHTm) (Gal-TR(AHTm)). Cells were grown in the absence
or presence of 0.1 µM T3. Data are the
means ± S.E. of three individual experiments. Note the discontinuous vertical axis.
[View Larger Version of this Image (16K GIF file)]
was
incubated with the [35S]methionine-labeled Gal4-TR(wt)
chimera in the absence or presence of T3. The results
indicated a weak constitutive association between these two proteins in
the absence of hormone, which did not change upon extended coincubation
(Fig. 5, A and B). However, in the presence of 1 µM T3, the TR-RXR interaction
was enhanced by up to 10-fold, reaching a maximum after ~40 min of
coincubation (Fig. 5 (A and B) and data not
shown).
Fig. 5.
T3 enhances the interaction
between TR and RXR in vitro. Bacterially expressed GST
or GST-RXR fusion protein was incubated with 35S-labeled
Gal4-TR for up to 60 min in the absence or presence of 2.5 µM T3. Gal-TR Input represents
10% of the total Gal4-TR used in each incubation. The Gal4-TR doublet
is due to initiation of translation from an internal methionine.
A, autoradiograph of a representative experiment;
B, quantitation of radiolabeled Gal4-TR binding in
A. ,
T3;
, +T3.
[View Larger Version of this Image (35K GIF file)]
Fig. 6.
9-cis-RA attenuates the TR-RXR
interaction in vivo. JEG-3 cells were transfected as
described in the legend to Fig. 2. Cells were then incubated in the
absence or presence of 0.1 µM T3 or 0.1 µM 9-cis-RA as indicated. -Fold induction is
relative to the activity of Gal4-TR(L454A) + VP16 in the absence of
ligand. Data are the means ±S.E. of at least three independent
experiments. Note the discontinuous vertical axis.
[View Larger Version of this Image (30K GIF file)]
2 gene occurs
only after treatment of cells with retinoic acid (49). Although it was
suggested that this might represent an enhanced affinity of the
liganded RAR·RXR heterodimer for DNA, the effect may also be due to a
ligand-dependent increase in solution heterodimer
formation.
isoform of TR, which is highly homologous to
TR
(51), also suggest an influence of ligand binding on
dimerization. The ninth heptad repeat of TR is at the amino-terminal
end of helix 11, on the surface of which are many residues implicated
in dimerization. However, residues at the carboxyl-terminal end of
helix 11 border the hormone-binding cavity and form critical contacts
with ligand. This involvement of helix 11 in both functions is in
keeping with the colocalization of ligand binding and dimerization to a
similar region in the estrogen receptor (3). Evidence in favor of this
relationship is provided by the observation that the effect of a ninth
heptad mutation (L421R in human TR
, L365R in chick TR
, and M377R
in human RAR
), which abolishes the interaction of a number of
receptors with RXR can be reversed by the addition of cognate ligand,
suggesting a role for hormone in stabilizing dimerization (10, 30). The Gal4-TR(AHTm) triple mutant used in this study also failed to interact
with RXR in the absence of ligand, suggesting that these residues,
which lie within helix 1, either constitute part of the dimerization
interface or allosterically influence the integrity of dimerization
regions. However, with T3, the interaction of Gal4-TR(AHTm)
with VP16-RXR was fully restored, emphasizing the central role of
ligand binding in stabilizing the integrity of the LBD and thereby
promoting the formation of a dimerization interface.
-half site
of a TRE (23, 52, 53), but is unable to bind 9-cis-RA due to
the negative allosteric influence of its dimer partner (54, 55). We
found that 9-cis-RA inhibited both the constitutive and
T3-induced interaction of Gal4-TR with VP16-RXR in the
two-hybrid assay. 9-cis-RA-mediated inhibition of the
constitutive TR-RXR interaction suggests that this effect is not due to
disruption of T3 occupancy of TR by an inhibitory
allosteric influence of liganded RXR. An alternative possibility is
that 9-cis-RA induces dimerization of the VP16-RXR fusion to
itself or endogenous RXR or orphan receptors, titrating it away from
Gal4-TR. Indeed, studies in vitro have shown that 9-cis-RA enhanced homodimerization by RXR on DNA (56), while with coexpressed RXR
and TR
in CV-1 cells, the addition of
9-cis-RA or RXR-specific ligands augmented formation of RXR
homodimers, thereby inhibiting T3-dependent
activation via TR (57).
*
This work was supported by the Wellcome Trust, the Medical
Research Council (United Kingdom), and the Imperial Cancer Research Fund.The costs of publication of this
article were defrayed in part by the
payment of page charges. The article
must therefore be hereby marked
"advertisement" in
accordance with 18 U.S.C. Section
1734 solely to indicate this fact.
§
Equal first authors.
Wellcome Senior Clinical Research Fellow. To whom
correspondence should be addressed: Dept. of Medicine, University of
Cambridge, Level 5, Addenbrooke's Hospital, Hills Rd., Cambridge CB2
2QQ, UK. Tel.: 11-44-1223-336842; Fax: 11-44-1223-336846; E-mail:
kkc1{at}mole.bio.cam.ac.uk.
1
The abbreviations used are: TR, thyroid hormone
receptor; TRE, thyroid hormone response element; T3,
thyroid hormone; RAR, retinoic acid receptor; RXR, retinoid X receptor;
hRXR, human retinoid X receptor
; DBD, DNA-binding domain; LBD,
ligand-binding domain; GST, glutathione S-transferase;
TRIAC, triiodothyroacetic acid; AD, activation domain;
9-cis-RA, 9-cis- retinoic acid.
2
The original codon nomenclature has been altered
to correspond to the sequence containing 461 residues (39).
©1997 by The American Society for Biochemistry and Molecular Biology, Inc.