(Received for publication, December 16, 1996, and in revised form, April 1, 1997)
From the Department of Biochemistry, College of
Medicine, The University of Arizona, Tucson, Arizona 85724 and the
§ Laboratory of Molecular Growth Regulation, NICHD, National
Institutes of Health, Bethesda, Maryland 20892
To investigate a potential
ligand-dependent transcriptional activation domain (AF-2)
in the C-terminal region of the human vitamin D receptor (hVDR), two
conserved residues, Leu-417 and Glu-420, were replaced with alanines by
site-directed mutagenesis (L417A and E420A). Transcriptional activation
in response to 1,25-dihydroxyvitamin D3
(1,25-(OH)2D3) was virtually eliminated when
either point mutant was transfected into several mammalian cell lines.
Furthermore, both mutants exhibited a dominant negative phenotype when
expressed in COS-7 cells. Scatchard analysis at 4 °C and a
ligand-dependent DNA binding assay at 25 °C revealed
essentially normal 1,25-(OH)2D3 binding for the
mutant hVDRs, which were also equivalent to native receptor in
associating with the rat osteocalcin vitamin D responsive element as a
presumed heterodimer with retinoid X receptor. Glutathione S-transferase-human transcription factor IIB (TFIIB) fusion
protein linked to Sepharose equally coprecipitated the wild-type hVDR and the AF-2 mutants. These data implicate amino acids Leu-417 and
Glu-420, residing in a putative -helical region at the extreme C
terminus of hVDR, as critical in the mechanism of
1,25-(OH)2D3-stimulated transcription, likely
mediating an interaction with a coactivator(s) or a component of the
basal transcriptional machinery distinct from TFIIB.
The actions of the 1,25-dihydroxyvitamin D3
(1,25-(OH)2D3)1
hormonal ligand are mediated by the vitamin D receptor (VDR). This nuclear protein belongs to a large superfamily, which includes receptors for steroids, retinoids, and thyroid hormone (1). As with
other members of this superfamily, the VDR possesses an N-terminal DNA
binding domain (DBD) that contains two zinc finger DNA binding motifs
(2). The C-terminal portion of the VDR contains a ligand binding domain
(LBD) that associates with the 1,25-(OH)2D3 hormone (3). This binding initiates a presumed conformational change in
VDR that enhances its interaction with any one of several isoforms of
retinoid X receptor (RXR) to form a heterodimer, the apparent active
species in recognizing and binding with high affinity to vitamin
D-responsive elements (VDREs) located in the upstream promoter region
of genes regulated by 1,25-(OH)2D3 (4, 5). In
addition to binding 1,25-(OH)2D3, the
C-terminal LBD of VDR contains residues necessary for
heterodimerization with RXRs (6, 7). When the activated, ligand-bound
VDR:RXR heterodimer is associated with the VDRE, it stimulates the
transcription of downstream target genes (8). Several VDREs have been
characterized to date, with the prototypical sequence consisting of an
imperfect direct repeat of six nucleotide bases, GGGTGA, separated by a 3-base pair spacer (9-13). VDR has been shown to contact the base pairs in the 3 half-site of the VDRE, while RXR interacts with the 5
half-site (14).
The precise mechanism of transcriptional regulation by the activated
VDR:RXR heterodimer is not well understood. Functional analyses of
members of the nuclear receptor superfamily, including truncation and
point mutagenesis studies, have demonstrated the presence of at least
two major domains involved in receptor-mediated transcriptional
stimulation. The N-terminal regions of several nuclear receptors
contain a constitutive activation domain referred to as AF-1 (also
designated 1) (15-17) that can be linked to heterologous DNA
binding domains to create functional transcription factors (18-20).
Moreover, fusion of the large C-terminal domain of various receptors to
the GAL4 or glucocorticoid receptor (GR) DBD produces a chimeric
protein capable of activating transcription in response to the cognate
ligand (18, 21, 22). A subdomain of this hormone-dependent,
C-terminal activation function is known as AF-2 (also termed
4) (21,
23-25). Unlike the N-terminal AF-1 transcription domain, the
C-terminal AF-2 activation unit exhibits a higher degree of homology
among the nuclear receptor superfamily, suggesting a common or related
mechanism may be involved in ligand-mediated gene regulation. In one
study (26), the AF-2 region of the thyroid hormone receptor (TR) was
reported to interact with the basal transcription factor IIB (TFIIB),
implying that association of nuclear receptors with the basal
transcriptional machinery may serve as one means of achieving
transcriptional control of hormone-stimulated genes.
As part of a preliminary study designed to probe the regions of VDR involved in heterodimerization, 1,25-(OH)2D3 binding, and transcriptional activation, our laboratory reported previously that truncation of the C-terminal 25 amino acids in hVDR generated a transcriptionally inactive receptor that retained heterodimerization capacity and also partial ligand binding ability (6), suggesting the presence of an AF-2 domain in the extreme C-terminal region of hVDR. In the present study, we have refined the mapping of this domain by identifying specific residues essential for competent, hormone-dependent transcriptional activation that are resolved from other biological activities of the receptor.
The hVDR expression vector, pSG5hVDR (27), was employed in synthesizing point mutants by in vitro site-directed mutagenesis (28). Two residues in the extreme C terminus of hVDR, Leu-417 and Glu-420, were separately altered to alanines (mutants designated as L417A and E420A, respectively). These mutations were confirmed by dideoxy sequencing.
Transfection of Cultured Cells and Transcriptional Activation AssayCOS-7 monkey kidney epithelial cells (700,000 cells/60-mm
plate) were transfected with 0.1 µg of wild-type or mutant hVDR expression plasmid and 10 µg of a reporter plasmid
((CT4)4-TKGH) containing four copies of the rat osteocalcin
VDRE (13) inserted upstream of the viral thymidine kinase
promoter-growth hormone reporter gene (Nichols Institute, San Juan
Capistrano, CA) by the calcium phosphate-DNA coprecipitation method as
described previously (29). The pTZ18U plasmid was used as carrier DNA, and each transfection contained a constant amount of total DNA (20 µg). The transfected cells were washed, then grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with
10% fetal bovine serum, 100 units/ml penicillin, 100 µg/ml streptomycin and various concentrations of
1,25-(OH)2D3 in ethanol vehicle. After 24 h of incubation at 37 °C, the level of growth hormone secreted into
the culture medium was assessed by radioimmunoassay using a commercial
kit (Nichols Institute). Transfections and treatments of HeLa cells and
a transformed primary human embryonal kidney cell line (HEK293) were
carried out similarly except these cells were cultured in minimum
essential medium supplemented with 10% fetal bovine serum and
antibiotics. Cellular lysates prepared from transfected cells were
analyzed for VDR expression by immunoblotting using the 9A7
monoclonal anti-VDR antibody as described previously (30).
Heterodimeric DNA binding activity
was assessed by the gel mobility shift assay essentially as described
elsewhere (7). In ligand-independent studies, cellular extracts from
COS-7 cells transfected with either pSG5 (control) alone, pSG5
containing wild-type cDNA, or pSG5 containing mutant hVDR cDNAs
were incubated with 32P-labeled rat osteocalcin VDRE
(5-AGCTGCACTGGGTGAATGAGGACATTACA-3
; half-sites comprising an imperfect direct repeat are underlined). In
experiments employing the 1,25-(OH)2D3 ligand
(31), COS-7 cells were transfected with 50 ng of hVDR expression
plasmids, and the extracts from these cells were initially incubated
with 10
8 M
1,25-(OH)2D3 for 25 min at 22 °C followed by
incubation with the rat osteocalcin VDRE for 25 min at 22 °C.
COS-7 cells transfected with 50 ng of wild-type or mutant hVDR expression plasmids were lysed in KETZD-0.3 buffer (0.3 M KCl, 10 mM Tris-HCl, pH 7.4, 1 mM EDTA, 0.3 mM ZnCl2, 5 mM dithiothreitol) containing 0.5% Triton X-100 and supplemented with protease inhibitors (2 µg/ml aprotinin, 0.5 µg/ml leupeptin, 50 µg/ml trypsin inhibitor). Lysates were tested for 1,25-(OH)2D3 binding activity as described previously (31).
GST Fusion Protein Binding AssayHuman transcription factor
IIB (hTFIIB)-GST fusion protein was expressed from pGEX-2T-hTFIIB (26),
and GST alone was expressed from pGEX-4T, both in Escherichia
coli strain DH5. The overexpressed proteins were coupled to
glutathione-Sepharose (1 µg of protein/µl of resin) according to
the protocol of the manufacturer (Pharmacia Biotech Inc., Uppsala,
Sweden) and stored as a 50% slurry in KETZD-0.3 containing 30%
glycerol at
20 °C. COS-7 cells transfected with either 20 µg of
control, wild-type, or mutant hVDR expression plasmids were resuspended
in 550 µl of KETZD-0.2 (0.2 M KCl, 10 mM
Tris-HCl, pH 7.4, 1 mM EDTA, 0.3 mM
ZnCl2, 5 mM dithiothreitol, 1.0 mg/ml bovine
serum albumin, 0.2% Tween 20, 2 µg/ml aprotinin, 0.5 µg/ml
leupeptin, 50 µg/ml trypsin inhibitor) and sonicated on ice. The
sonicates were clarified by centrifugation for 15 min at 16,000 × g at 4 °C, and 500 µl were incubated with either 20 µl (50% slurry) of hTFIIB-GST-Sepharose or GST-Sepharose alone for
1 h at 4 °C on a rocker tray. The Sepharose beads were then washed four times with 1 ml each of KETZD-0.2, resuspended in 40 µl
of 2 × final sample buffer (2% SDS, 5%
-mercaptoethanol, 125 mM Tris-HCl, pH 6.8, 20% glycerol) and boiled for 3 min.
These samples, along with 25 µl of original lysate (5% of input)
were analyzed for VDR content by immunoblotting (30).
The
C-terminal LBD of the nuclear receptor superfamily possesses several
regions of high homology (32). One such subdomain, located at the
extreme C terminus of the LBD, is depicted in Fig. 1.
Examination of this region of hVDR reveals Leu-417, which is conserved
in the VDR subfamily of nuclear receptors (VDR, TR, and retinoic acid
receptor (RAR)), as well as in the estrogen receptor (ER), and Glu-420,
which is positionally conserved throughout the superfamily. These two
amino acids probably reside in an -helical segment of the VDR based
on the recently elucidated three-dimensional crystal structures of the
homologous regions in rat TR
(33), human RAR
(34), and human
RXR
(35) LBD. Thus, these residues were individually altered to
alanine in hVDR to preserve the potential
-helical character in this
region and to probe their potential functional significance.
Binding of 1,25-(OH)2D3 by VDR Mutants
Considering that Leu-417 and Glu-420 are located in the
LBD, and the residues corresponding to Val-418 and Phe-422 in VDR (Fig.
1) are ligand contact sites in either RAR (34) or TR (33), we tested
the 1,25-(OH)2D3 ligand binding activity of the
above generated mutant hVDRs in an equilibrium binding assay performed at 4 °C. Extracts from COS-7 cells transfected with expression plasmids encoding either wild-type or a mutant hVDR were incubated with
increasing concentrations of 1,25-(OH)2D3
overnight at 4 °C. The data from saturation binding curves (Fig.
2, A and B) were then transformed
and plotted by the method of Scatchard, yielding an estimate of the
dissociation constant (Kd) for the wild-type and
mutant receptors. Results from two independent experiments (Fig.
2C) revealed that the apparent Kd for the
1,25-(OH)2D3 ligand is not significantly
altered by replacement of Leu-417 with alanine, at least under the
conditions of this standard binding assay. Similarly, only a negligible
increase in the apparent Kd occurs when Glu-420 is
altered to alanine (Figs. 2, B and C). These data
extend previous results (6) that revealed truncation of the last 25 amino acids in hVDR does not abolish ligand binding but does cause an
approximate 10-fold elevation in the Kd. The extreme
C terminus of hVDR, therefore, is not absolutely required for
relatively high affinity interaction with
1,25-(OH)2D3, suggesting that this region of
the receptor, and more specifically the conserved residues Leu-417 and
Glu-420, performs a different function.
Transcriptional Activation Capacity of Mutant hVDRs in Transfected Cells
The transcriptional activity of wild-type and point mutated
hVDRs was evaluated in three transfected cell lines. Fig.
3A illustrates a
1,25-(OH)2D3-stimulated,
dose-dependent increase in transcription of a rat
osteocalcin VDRE-linked reporter gene in transfected COS-7 cells. The
maximal hormone-mediated increase in transcription with the wild-type
receptor in this cell line is approximately 40-fold (at or above
107 M 1,25-(OH)2D3).
Under parallel transfection conditions, the L417A and E420A hVDRs do
not exhibit detectable transactivation even at the highest dose
(10
6 M) of
1,25-(OH)2D3 tested (Fig. 3A). The
fact that this pharmacological level of
1,25-(OH)2D3 is ineffective in enhancing
transcription (Fig. 3A) is consistent with the conclusion
that the mutant VDRs are not partially impaired in ligand binding (Fig.
2). Expression of these mutant hVDRs was similar to that of the
wild-type receptor (data not shown; see also Figs.
4C, 5B, and 6B),
eliminating differential expression as a trivial explanation for these
results. To evaluate the possibility that the abrogation of
transcriptional activity in these mutant hVDRs is a cell-specific
phenomenon, a similar set of experiments was carried out in a
transformed primary human embryonal kidney cell line (HEK293; Fig.
3B) and in HeLa cells (Fig. 3C). The maximal fold
increase in transcriptional activity by wild-type hVDR was 5-fold (Fig.
3B) and 17-fold (Fig. 3C) in HEK293 and HeLa
cells, respectively. The dose-response profile of wild-type hVDR in
these additional cells lines was somewhat similar to that observed in
COS-7 cells, although 10
8 M
1,25-(OH)2D3 elicited a near-maximal
transcription effect in HEK293 and HeLa cells, and 10
9
M ligand approximated the ED50 in HeLa cells.
Importantly, neither the L417A nor the E420A mutant hVDR mediates
significant ligand-dependent transactivation in these other
two cell lines upon exposure to 10
9 M or even
10
8 M
1,25-(OH)2D3.
Thus, taken together, the data in Fig. 3 identify residues 417 and 420 in the transactivation function of VDR. A minor but significant repression of basal transcription is detected in the case of E420A hVDR, especially in HeLa cells (Fig. 3C), which is relieved by addition of increasing concentrations of 1,25-(OH)2D3. This observation is consistent with the findings that unoccupied nuclear receptors associate with a transcriptional corepressor that is released by binding of the cognate ligand (36) and that corepressor interaction may be enhanced by mutation of the receptor AF-2 domain (37). These results support a model (38) of mutually exclusive binding of a corepressor and coactivator to overlapping regions in the C termini of nuclear receptors, a domain that includes Leu-417 and Glu-420.
Transcriptionally Defective VDRs Possess Normal Heterodimeric DNA Binding ActivityWe next evaluated the ability of the
transcriptionally defective mutant hVDRs to interact with a
prototypical rat osteocalcin VDRE as a heterodimeric complex with
endogenous RXR in transfected COS-7 cells. Extracts from cells
transfected with the wild-type hVDR expression plasmid form an intense
shifted complex when incubated with the VDRE probe in a gel mobility
shift assay (Fig. 4A, lane 2). This retarded band is
referred to as VDR:RAF:VDRE because the exact identity of the
receptor auxiliary factor (RAF) as
one of the RXRs is not defined for COS-7 cells. This complex is absent when extracts from cells transfected with the expression vector lacking
the VDR insert are utilized (lane 1) or when wild-type VDR-containing extracts are preincubated with anti-VDR monoclonal antibody 9A7 (lane 3). When extracts from cells
transfected with either the L417A or E420A mutant are incorporated into
this assay (lanes 4 and 5), a shifted complex of
equal intensity and migration position compared with the wild-type
receptor is observed. Thus, there is no evidence to this point that
either mutation in hVDR affects its heterodimerization activity, with
the caveats that the VDR-containing extracts used in the experiments
pictured in Fig. 4A were obtained from COS-7 cells
expressing large amounts of each receptor protein and, additionally,
these cells were not treated with the
1,25-(OH)2D3 hormone. Thus, a similar analysis was performed with extracts from COS-7 cells expressing limiting amounts of VDR, which more closely approximate levels found in 1,25-(OH)2D3 target tissues, such as bone,
intestine, or kidney (39). Extracts from these cells were treated with
10
8 M 1,25-(OH)2D3 or
ethanol vehicle and used in a hormone-dependent gel shift
assay (31). Under these near physiologic conditions, wild-type and both
hVDR mutants formed a VDR-containing shifted complex that was enhanced
by the addition of the 1,25-(OH)2D3 ligand
(Fig. 4B, compare lanes 4 with 3,
7 with 6, and 9 with 8).
Significantly, the level of augmentation by
1,25-(OH)2D3 (2-fold) based on densitometric
scanning of the images shown in Fig. 4B (data not shown) was
similar for the wild-type receptor and mutant hVDRs. Also, the level of
expression of each receptor was essentially equivalent and unmodified
by sterol ligand as monitored by immunoblotting (Fig. 4C).
The apparent decrease in VDRE binding exhibited by the L417A mutant
(Fig. 4B, lanes 6 and 7) is actually the result of less total protein loaded into these lanes as demonstrated by the
decreased intensity of the nonspecific bands migrating below the
hVDR-containing complex. The nonspecific binding of the two lower
complexes was deduced by their appearance in non-VDR-transfected cells
(Fig. 4B, lanes 1 and 2) as well as their lack of
elimination by the VDR antibody (Fig. 4B, lane 5). The level
of L417A-containing complex formation approaches that of the wild-type
band when corrected for this loading difference (as assessed by
densitometric scanning), and repeat experiments (not shown) indicate
that similar levels of hVDR complexes are formed on the VDRE by all
three hVDRs tested (see also Fig. 4A). Taken together, the
results illustrated in Fig. 4 strongly suggest that the alteration of
Leu-417 or Glu-420 to alanine does not attenuate heterodimeric DNA
binding by the VDR protein, and such a phenomenon cannot therefore
account for the paucity of transactivation by these mutants.
Additionally, the essentially equivalent enhancement in VDRE binding
elicited by the 1,25-(OH)2D3 hormone is
consistent with the similar Kd values exhibited by
the wild-type and mutant receptors (Fig. 2C).
Because the above described DNA binding studies were
carried out at nonphysiologic temperatures, in vitro, with
cell extracts, it is not possible to conclude rigorously that the
transcriptionally inactive hVDR mutants actually bind to the VDRE in
intact cells (40, 41). Therefore, we determined whether the two mutant hVDRs under study could act in a dominant negative fashion in transactivation assays. Cotransfection experiments in COS-7 cells were
designed utilizing the rat osteocalcin VDRE reporter vector and 0.1 µg of wild-type hVDR expression plasmid (similar to the experiment
shown in Fig. 3). Under these conditions, the
1,25-(OH)2D3 hormone results in about a 15-fold
increase in transcription of the GH reporter gene (Fig.
5A, WT). Additional cotransfection of a
5-fold excess (0.5 µg) of L417A or E420A expression plasmid reduced
ligand-dependent transcription by approximately 50%
(L417A) and 66% (E420A), respectively. These results are analogous to those obtained previously in which a mutant hVDR possessing an altered
T-box region was shown to be an effective dominant negative at ratios
of mutant to wild-type hVDR expression plasmid of five and greater
(42). In rescue experiments (Fig. 5A), the level of
transfected mutant expression vector was maintained at 0.5 µg, while
the amount of wild-type hVDR expression plasmid was increased to 0.35 µg (designated ++ in Fig. 5A). The corresponding increase
in wild-type hVDR expression restores
1,25-(OH)2D3-dependent transcriptional activation to fold effects approaching (WT/E420A, 12X), or even slightly exceeding (WT/L417A, 19X), the
fold effects observed with transfection of only the wild-type control
(WT, 15X). As expected, immunoblotting of extracts from
these transfected cells revealed a close correlation between the total
amount of VDR receptor plasmid transfected (denoted schematically by
the total number of + symbols in each lane, Fig. 5B) and the
level of receptor expression. Because both L417A and E420A mutant hVDRs retain wild-type levels of hormone and heterodimeric DNA binding, in vitro (Figs. 2 and 4), and act in a dominant negative
fashion in transfected cells (Fig. 5), we conclude that these AF-2
mutants exhibit
1,25-(OH)2D3-dependent binding to
the VDRE, in vivo, but as transcriptionally
inactive heterodimers. As demonstrated in Fig.
5A, the expression of additional wild-type VDR effectively competes with the mutant receptor for RXR association, and subsequent binding of a normal heterodimer to the response element leads to
enhanced activation. In contrast, overexpression of RXR does not
reverse the dominant negative effect (data not shown), presumably because the mutated VDR in the heterocomplex contains an inactive AF-2
domain.
Recently, it has been reported that the VDR interacts
physically and functionally with the general transcription factor IIB in a 1,25-(OH)2D3 ligand-independent fashion
(43, 44). In one of these reports, it was suggested that the extreme
C-terminal portion of VDR may represent a contact domain for TFIIB
(44), an observation also reported for TFIIB interaction with TR (26). Consequently, we probed the potential interaction of L417A and E420A
with TFIIB, in vitro, to determine if a lack of TFIIB
association could explain the transcriptionally defective and dominant
negative phenotypes of these mutant receptors. Extracts from COS-7
cells transfected with expression plasmids for each receptor protein were incubated in the absence of the
1,25-(OH)2D3 ligand (Fig. 6A) with hTFIIB-GST fusion protein linked to
glutathione-Sepharose beads (GST-TFIIB-S) or GST-Sepharose
(GST-S) as a control (Fig. 6A). After extensive
washing, the presence of VDR protein coprecipitated by the GST-TFIIB-S
or GST-S complex was detected by immunoblotting using an anti-VDR
monoclonal antibody. The results illustrate that wild-type VDR (Fig.
6A, lane 4), as well as both mutant hVDRs (Fig. 6A,
lanes 6 and 8), interact efficiently and specifically with TFIIB; similar results were obtained in the presence of
108 M 1,25-(OH)2D3
(data not shown). The levels of VDR protein were essentially equivalent
as verified by Western blot analysis (Fig. 6B) using 5% of
the total extracts from the experiment illustrated in Fig.
6A. These data argue against a role for Leu-417 and/or Glu-420 in contacting TFIIB.
Previously, we reported that the C-terminal 25 amino acids of hVDR appeared to be vital for hormone-dependent transactivation (6). It is now demonstrated that two well conserved amino acids within this region of hVDR (Fig. 1), Leu-417 and Glu-420, are essential for 1,25-(OH)2D3 ligand-stimulated transcription (Fig. 3) and likely define the AF-2 function in the VDR. Because mutation of these residues does not affect hormone (Fig. 2) or DNA binding (Fig. 4) to any significant degree, these amino acids are thought to participate directly in the mechanism of transcriptional activation following the recruitment of ligand-bound VDR:RXR heterodimers to the VDRE. Such a mechanism might involve the interaction of VDR with basal or TATA-binding protein-associated factors, thus facilitating the formation of an active transcriptional complex. VDR also physically interacts with TFIIB (43), and inferences from studies utilizing TR and VDR (26, 44) suggested that the C terminus of these receptors might be involved in TFIIB interaction. However, the fact that the wild-type receptor and both mutant VDRs interact similarly with TFIIB, in vitro (Fig. 6), argues against any interaction between these residues and TFIIB. Rather, the present results suggest that Leu-417 and Glu-420 represent contact sites for the interaction of VDR with another coactivator protein(s) required for ligand-dependent transcriptional stimulation. Further studies of wild-type and AF-2 mutant VDRs employing methodologies such as the yeast two-hybrid system will be required to identify proteins which physically and functionally associate with this region of VDR.
Other members of the nuclear receptor superfamily also possess a
transactivation domain in the C-terminal region. The chicken TR has
been shown to contain a ligand-dependent activation
function in the last 35 amino acids of the receptor (45), and within this domain several acidic and hydrophobic residues were found to be
critical for transcriptional activity. Interestingly, this region of
the protein is absent in the transcriptionally defective v-erbA protein, the viral homolog of cTR
. Similar results
have been reported for the human TR
1, where mutations in the two
residues corresponding to Leu-417 and Glu-420 generated receptors with wild-type hormone and DNA binding activity but abrogated
transactivation (46, 47), a phenotype shared by the present hVDR
mutants. In addition, point mutagenesis of the mouse RAR
1 and
truncation analysis of mouse RXR
have also revealed the importance
of this domain in gene activation in these and other receptors (25, 48), an observation consistent with the relatively high level of
conservation of this region among the nuclear receptor superfamily (Fig. 1). In fact, it appears that the homologous amino acids identified in the present study for VDR may be essential for
transactivation even in the more distantly related ER and GR (23, 49),
although in one study (23), mutation of the residue in mouse ER (E546A) corresponding to Glu-420 in hVDR only modestly affected transcriptional activation. In contrast, an E546A mutant that also lacked the ER AF-1
domain could not stimulate transcription. These results suggest that
the previously described synergism between AF-1 and AF-2 (19) might
depend on the presence of Glu-546 in ER. If the VDR, with its
relatively small N terminus, lacks an AF-1 region, this could explain
the almost complete abolishment of transcriptional activity when
Glu-420 is mutated in hVDR, while the ER E546A mutant still possesses
significant activity. The importance of the AF-2 region in the nuclear
receptor superfamily is further supported by the observation that ER
and RXR truncation mutants (41, 49) and a deletion mutant of RAR
(48), all of which are missing amino acids homologous to Leu-417 and
Glu-420 in hVDR, act as effective dominant negative receptors similar
to the hVDR mutants described herein (Fig. 5). The concept of a
dominant negative phenotype for AF-2 mutants (34), including those in
the VDR, is in agreement with wild-type hormone and dimeric DNA binding properties exhibited by these altered receptors, since this would allow
the mutants to form inactive dimers that compete with the wild-type
protein for binding to the cognate responsive elements, thereby
suppressing gene stimulation.
The recently described crystal structures for the hormone binding
domains of rTR (33), hRAR
(34), and hRXR
(35) indicate that
all three receptors share a large degree of
-helical content, with
many of the helices being generally conserved within the overall
structure of the protein. In particular, the terminal
-helix in all
three molecules (Fig. 1,
-helix 12) encompasses the
conserved region that contains Leu-417 and Glu-420. It has been
postulated that this
-helical segment projects away from the core of
the LBD and that hormone binding induces a conformational change in the
receptor causing the repositioning of helix 12 so that it essentially
covers the opening of the ligand binding pocket. This hypothesis,
termed the "mouse trap" model (34), further predicts that Glu-414
in hRAR
(homologous to Glu-420 in hVDR) forms a crucial salt bridge
with Lys-264 (homologous to Lys-264 in hVDR) which is thought to
"lock" helix 12 in a position over the ligand binding pocket
following interaction of the receptor with its cognate hormone.
Additionally, this conformational change appears to place helix 12 in a
position to effectively interact with a coactivator protein. Mutation
of either partner in this putative salt bridge in hRAR
(34)
abolishes transactivation and results in a receptor with the dominant
negative phenotype. In preliminary experiments with
hVDR,2 a theoretical hVDR salt bridge
"reversal" double mutant (K264E-E420K), which would be predicted to
preserve the putative electrostatic bond, was found to be inactive
transcriptionally. Thus, Lys-264 in hVDR does not appear to form a salt
bridge with Glu-420, but this does not exclude potential electrostatic
bonds between Glu-420 and other positively charged VDR residues or the
existence of a novel salt bridge capable of positioning the VDR AF-2 to
seal the ligand binding pocket.
The proposed role of the AF-2 region in providing a protein-protein
interaction surface for contacting a coactivator is supported not only
by the three-dimensional crystal structures of hRAR, rTR
, and
hRXR
as discussed above, but also by the recent identification of
candidate coactivator proteins for various nuclear receptors. The ER,
for example, has been shown to interact with a 160-kDa protein in an
estrogen- and AF-2-dependent manner (50). Another protein,
SPT6, appears to physically interact with the C-terminal portion of ER
and to enhance ER-dependent transactivation (51). Moreover,
the nuclear protein RIP140 also associates with and enhances ER
activity in the presence of estrogen but not the antagonist, 4-hydroxytamoxifen (52). In this latter study, the interaction of
hormone-occupied ER with RIP140 was abolished by mutations in the ER
AF-2 that abrogated transcriptional activation (in the absence of
AF-1), including a point mutant at residue 546 (homologous to hVDR
Glu-420). Other candidate coactivators include GRIP1 (53), which
interacts with ER, GR, and the androgen receptor, and steroid receptor
coactivator-1 (54), which stimulates the transcriptional activity of
several nuclear receptors. In each case, the interaction between the
putative coactivator and the receptor protein requires residues in the
AF-2-containing LBD. Preliminary data3 with
hVDR AF-2 mutants indicate that both
1,25-(OH)2D3 ligand and an intact AF-2 are
required for squelching of dexamethasone/GR/glucocorticoid responsive
element-mediated transcription in COS-7 cells. These results suggest
that the AF-2 domain of hVDR (Fig. 1,
-helix 12),
including residues Leu-417 and Glu-420, likely represents a docking
site for a general nuclear receptor coactivator, possibly including one
or more of those proteins described above. In addition, given that some
amino acid variation in this region exists between the nuclear
receptors (Fig. 1), several residues in this domain may constitute a
"coactivator code" that dictates receptor-specific coactivator
interactions in certain cell types.
Analogous to other ligand-activated nuclear receptors, VDR apparently exists in a monomeric, inactive conformation with the C-terminal AF-2 presumably extended away from the hormone binding cavity. We propose that upon binding 1,25-(OH)2D3 in target cell nuclei (55), VDR assumes an active conformation as the AF-2 is repositioned for both ligand retention and coactivator contact, with Leu-417 and Glu-420 being critical for this latter association. In addition, the hormone facilitates interaction of VDR and RXR through a stabilized heterodimerization interface mediated by other regions within the VDR C-terminal hormone binding domain (6, 7). The VDR partner, RXR, also possesses an AF-2 (Fig. 1 and Ref. 25) that participates in transcriptional activation by 1,25-(OH)2D3 through association with a distinct coactivator. This interpretation is supported by the fact that C-terminally truncated RXRs function as dominant negative partners in VDR-mediated transcription (41). Additionally, based upon a previous study (5), it is speculated (56) that the AF-2 in RXR has the ability to be positioned for transactivation in the absence of its 9-cis-retinoic acid ligand, likely via allosteric modulation by 1,25-(OH)2D3-occupied VDR in the RXR:VDR heterodimer. Therefore, the mechanism of VDR action is proposed to be similar to that of TR (57, 58), where liganding of only the primary receptor and not the RXR heteropartner is sufficient to elicit maximal transcriptional stimulation via the AF-2s of both receptors. This differs from the function of RXR:RAR heterodimers (59), in which the liganding of both partner receptors yields responsive element occupation, in vivo, and full transcriptional activation by the cooperation of AF-2s. The availability of VDR AF-2 mutants presented herein will facilitate the isolation of VDR-specific coactivators, which should help to define further the mechanistic diversity of action among the VDR/TR/RAR subfamily.
We acknowledge the technical support of Michael Galligan and Michelle Thatcher.