ABSTRACT
Several members of the nuclear receptor superfamily including RXR (retinoid X
receptor) bind to a specific retinoic acid response element (site A) of the
apoAI promoter. However, transcriptional activation of the apoAI gene by
different homo- and heterodimeric forms of RXR or RAR (retinoic acid receptors) cannot be
evaluated in mammalian cells, which contain endogenous RXR or RAR. In order to
circumvent this limitation, we assessed the DNA-binding activities and transcriptional activation of different homo- and heterodimers of these receptors in yeast. Electrophoretic
mobility shift assays (EMSA) demonstrated that yeast expressed RAR
[alpha]
does not bind to site A of the apoAI promoter, whereas binding of RAR
[beta]
to site A is ligand-dependent. Both RAR
[alpha]
and RAR
[beta]
form heterodimers with RXR
[alpha]
and bind to site A with high affinity. These DNA-binding studies correlate with the transcriptional data, which indicated
that RAR
[beta]
but not RAR
[alpha]
activates transcription from site A in response equally well to 9-
cis
and all-
trans
retinoic acids. 9-
cis
RA is a more potent ligand than all-
trans
RA to activate transcription via RXR/RAR heterodimers. We conclude that this
yeast expression system is a useful tool to elucidate the `transactivaton code'
for apoAI site A via specific combinations of different homo and heterodimeric
versions of RXR and RAR.
All-
trans
and 9-
cis
isomers of retinoic acid are potent modulators for a broad spectrum of
essential biological activities including embryogenesis, cell proliferation and
differentiation (
1
,
2
). Thus, this class of compounds has received considerable attention for their
pharmacological utilities. Although some of these retinoids are potential drugs
for the treatment of dermatological diseases and cancers, their toxicities
including teratogenicity have limited their therapeutic values. The pleiotropic
action of retinoids is mediated by different homo- or heterodimeric versions of retinoic acid/retinoid X receptors, which
belong to the nuclear receptor superfamily of ligand activated transcription
factors (
3
,
4
). Each of these receptors has three distinct subtypes namely RAR[alpha],[beta],[gamma] and RXR[alpha],[beta],[gamma], which form heterodimers with each
other (
5
). Furthermore, these receptor subtypes also have isoforms, which can generate
theoretically 48 different RXR/RAR heterodimeric forms (
2
). It is highly suggestive that the undesirable side effects of retinoids are in
part due to the formation of homo- and heterodimers of these receptor subtypes or isoforms, which modulate
the expression of different target genes in response to the same or different
naturally occurring retinoids. Therefore, ligands which are selective for
certain receptor subtypes in conjunction with a specific target gene can be
exploited as beneficial drugs with fewer side effects.
Regulation of the apoAI gene, which encodes apolipoprotein AI, the major protein
constituent of HDL, is controlled by synergistic interactions between several
transcription factors bound to three distinct sites (sites A, B and C) of a
liver specific enhancer located between nucleotides -222 to -110 upstream of the transcription start site (+1) (
6
). In mammalian cells, RXR[alpha] homodimers activate transcription via site A (-214 to -192) of the apoAI gene in response to 9-
cis
RA but not to all-
trans
RA, whereas RXR[alpha]/RAR[alpha] heterodimers activate transcription via this DNA element in
response to both 9-
cis
RA and all-
trans
RA (
7
). However, the transcriptional activation by retinoic acid receptor (RAR[alpha],[beta],[gamma]) homodimers in response to 9-
cis
RA/all-
trans
RA is still unclear due to the presence of endogenous RXR or RAR subtypes
present in mammalian cells. We have reported previously that RXR[alpha] homodimers bind to site A and function as ligand-dependent transcriptional activators in yeast cells, which are
devoid of endogenous RXR/RAR or enzymes that convert all-
trans
RA to 9-
cis
RA (
8
). In this report, we further utilized this retinoid-responsive transcription unit to assess the DNA-binding properties, differential transcriptional activation and
ligand responsiveness exhibited by different homo- and heterodimeric versions of RXR/RAR in yeast using site A as the
retinoic acid responsive element (RARE). Our results demonstrate that this
yeast expression system is a useful genetic tool to define the `transactivation
code' for site A via specific combinations of homo- and heterodimeric versions of RXR and RAR as well as to unravel
systematically the synergistic interactions between these transcription
factors. Furthermore, this microbial system can be used to identify
unambiguously selective ligands capable of transactivating the apoAI gene via
different RXR/RAR homo- or heterodimers.
The
Saccharomyces cerevisiae
strain used was BJ2168 as described previously (
8
-
10
). Growth and transformation of yeast cells were performed according to standard
procedures (
11
). Double and triple transformant yeast strains were grown in synthetic drop-out media to maintain expression plasmids.
The human RAR[alpha] and RAR[beta] genes were amplified by polymerase chain reaction (PCR) to possess
an
Eag
I site 5' and a
Bss
HII site 3' to their coding sequences and were cloned into the yeast expression
vector, YEp351, which has been modified to carry the CUP1 promoter (
9
,
12
) and a synthetic linker containing
Eag
I and
Bss
HII restriction sites downstream of the ubiquitin gene (
9
,
10
). The resultant expression plasmids, YEp
c
RAR[alpha] and YEp
c
RAR[beta] which contain the inducible CUP1 promoter driving the ubiquitin-receptor fusion genes and the
LEU2
gene as the auxotrophic marker were used to transform yeast strains carrying
the reporter plasmid in the absence or presence of the RXR[alpha] expression plasmids (YEpRXR[alpha]) as described previously (
8
). The reporter plasmid (YEpA) contains two copies of site A upstream of the
CYC1 promoter, which is fused to the
lac-Z
gene of
E.coli
(
8
-
10
). The resultant double transformant yeast strains (YEp
c
RAR[alpha]/YEpA, YEp
c
RAR[beta]/YEpA) and the triple transformant yeast strains (YEpRXR[alpha]/YEp
c
RAR[alpha]/YEpA, YEpRXR[alpha]/YEp
c
RAR[beta]/YEpA) were analyzed for protein expression, DNA-binding, heterodimer formation and transcriptional activation.
Yeast transformants carrying the RAR[alpha] or the RAR[beta] receptor expression plasmid were grown overnight in synthetic drop
out medium in the absence or presence of 100 [mu]M cupric sulfate until the cell density reached late log phase (OD = 1.0 at
600 nm). Cells were harvested and yeast extracts were prepared according to
standard protocols (
8
,
9
,
12
). Protein samples were electrophoresed on 10% SDS-PAGE, electroeluted onto nitrocellulose and probed with a polyclonal
antiserum raised against human RAR[alpha] or RAR[beta] obtained from Santa Cruz Biotechnology, Inc. (CA).
The procedures for EMSA and labeling of probes have been described previously (
8
,
9
). Briefly, double-stranded oligonucleotide corresponding to site A was labeled with [
32
P]dATP by filling in reaction with Klenow enzyme. The labeled oligo site A (30
000 c.p.m./reaction) was incubated with yeast extract in 20 [mu]l of a reaction mixture, which contains 7.5% glycerol, 0.05% NP-40 and 1 [mu]g of poly (dI-dC). Protein concentrations of yeast extracts were
normalized by adding bovine serum albumin. The reaction mixture was incubated
at room temperature for 20 min. For antibody supershift assays, 1 [mu]l of antiserum against RXR[alpha], RAR[alpha] or RAR[beta] was incubated with the reaction mixture at room
temperature for 30 min prior to addition of
32
P-labeled oligo site A. Bound and free DNAs were separated on a 6% non-denaturing gel. In some cases, yeast extracts were incubated with 9-
cis
RA or all-
trans
RA (1 [mu]M) at 4oC for 1 h prior to EMSA. To determine the dissociation constants (
K
d
) of retinoic acid receptor homo- versus heterodimers for site A, a constant amount of yeast extracts
containing these transcription factors were incubated with an increasing
concentration (0.25 to 25 nM ) of
32
P-labeled probe (site A) as described previously (
9
). The bound and free radioactivity were determined from the autoradiogram using
a betascope. The data were converted into Scatchard plot analysis (
13
).
To determine the hormone binding properties of yeast expressed retinoic acid
receptor homodimers or heterodimers, yeast extracts containing these receptors
were incubated at 4oC overnight with 5 nM of [
3
H]9-
cis
RA in the absence (total binding) or presence (non-specific binding) of 100-fold molar excess of radioinert 9-
cis
RA. Specific binding was determined by subtracting non-specific from total binding. Bound and free radioactivity were separated
by dextran-coated charcoal suspension as described previously (
12
). The binding affinity of the receptors for [
3
H]9-
cis
RA was determined by Scatchard plot analysis (
13
).
Double transformant yeast strains were grown in synthetic medium in the absence
of tryptophan and uracil whereas triple transformant yeast strains were grown
in synthetic medium in the absence of tryptophan, uracil and leucine. For
receptor induction, 100 [mu]M of cupric sulfate was added to the medium during cell inoculation. When
cell density reached late log phase, yeast cells were assayed for [beta]-galactosidase induction as described previously (
8
-
10
).
In the present study, the expression of the human RAR[alpha] and RAR[beta] was under the control of a regulated promoter, CUP1, which could
be induced by cupric sulfate, whereas the RXR[alpha] gene was driven by the constitutive promoter, TDH3 as described
previously (
8
,
10
). As shown in Figure
1
, Western blot analysis revealed a distinct 55 kDa protein in yeast extracts
prepared from the yeast strain carrying the RAR[alpha] expression plasmid under induced conditions (+Cu) (lane 3). This
immunoreactive band was not detected in the same yeast strain without the
expression plasmid (lane 1) or without copper induction (lane 2). The other
band with a molecular mass of ~48 kDa is a non-specific immunoreactive protein, which is present in the parent yeast
strain. Using a polyclonal antibody against human RAR[beta], an immunoreactive band with a molecular mass of 52 kDa was detected in
yeast extracts prepared from the yeast strain carrying the RAR[beta] expression plasmid only under induced conditions (+Cu) (data not shown).
One important observation in our previous studies is that yeast expressed RXR[alpha] homodimers bind to site A of the apoAI enhancer in the absence of 9-
cis
RA. In the present study, the DNA-binding properties of homo- or heterodimeric versions of yeast expressed RXR[alpha]/RAR[alpha] were examined by electrophoretic mobility shift
assays (EMSA). As shown in Figure
2
(left panel), when 5 [mu]l of yeast extracts (7.0 [mu]g protein) prepared from yeast strains carrying expression plasmids for
RXR[alpha] (yRXR[alpha]) or RAR[alpha] (yRAR[alpha]) were incubated with a
32
P-labeled double-stranded DNA probe containing site A, a distinct retardation complex
formation, which could be supershifted with an anti-RXR[alpha] polyclonal antibody was observed only for yRXR[alpha] (lanes 1 and 2) but not for yRAR[alpha] (lane 3). We have previously shown that yeast
extracts prepared from parental yeast strain carrying no expression plasmid did
not have any endogenous factor that binds to site A (
8
). In order to rule out the possibility of receptor inactivation at low protein
concentrations, bovine serum albumin (BSA) was added in the subsequent binding
mixtures to normalize protein concentrations and we have found that the degree
of retardation complex formation was the same in the presence or absence of
BSA. When 1 [mu]l (0.6 [mu]g protein) of yeast extracts containing either RXR[alpha] alone (yRXR[alpha]) or RAR[alpha] alone (yRAR[alpha]) was incubated with the site A probe, there
was no retardation complex formation (Fig.
2
: right panel, lanes 1 and 2). A distinct protein-DNA complex was detected when these two extracts were mixed with the
probe (lane 3). This heterodimeric complex formation could be displaced by 100-fold molar excess of radioinert oligo A (lane 4) and could be supershifted
with an anti-RXR[alpha] polyclonal antibody (lane 5).
In vitro
binding of RAR[beta] homodimers to site A was examined by EMSA. When 5 [mu]l (15 [mu]g protein) of yeast extracts (yRAR[beta]) containing RAR[beta] was incubated with
32
P-labeled site A probe, there was no complex formation (Fig.
4
, lane 5). However, retardation complex formations were apparent when the yeast
extracts were preincubated with 9-
cis
RA or all-
trans
RA at 4oC for 30 min prior to EMSA (Fig.
4
, lanes 6 and 7). This ligand-dependent DNA-binding property of RAR[beta] homodimers is in contrast to the ligand-independent DNA-binding properties of yeast expressed RXR[alpha] homodimers (
8
) and hepatic nuclear factor 4 (HNF-4) (
9
). To demonstrate RXR[alpha]/RAR[beta] heterodimer formations, we performed mixing experiments as
described previously in Figure
2
. When 0.2 [mu]l (0.6 [mu]g protein) of yeast extracts containing either RXR[alpha] (yRXR[alpha]) or RAR[beta] (yRAR[beta]) was analyzed by EMSA, there was no
retardation complex formation (Fig.
4
, lanes 1 and 2). However, a distinct protein-DNA complex, which could be supershifted with the RXR[alpha] antiserum was observed when the same volume of extracts
containing RXR[alpha] or RAR[beta] were mixed with the probe (Fig.
4
, lanes 3 and 4). These complexes could also be supershifted with the RAR[beta] antiserum (data not shown). This heterodimer formation was not affected
by 9-
cis
RA or all-
trans
RA (data not shown). Thus, these DNA-binding studies demonstrate that RAR[beta] homodimers bind to apoAI site A only in the presence of 9-
cis
RA or all-
trans
RA, whereas RXR[alpha]/RAR[beta] heterodimers bind to site A in the absence of ligands.
One major complexity of the retinoid signaling pathway is that RXR forms
heterodimers not only with RAR subtypes, but also with vitamin D receptor,
thyroid hormone receptor, the orphan receptors such as peroxisome proliferator
activated receptor (PPAR) and the COUP transcription factor (
16
,
17
). In fact, the heterodimer formation with RXR is a prerequisite for these
transcription factors to bind to their target DNA sequences efficiently (
18
,
19
). Whether these heterodimers can exhibit target gene activation in response to
the same ligands as for the homodimers is an area of intense interest. Since
the relative potency of target gene activations exhibited by these
transcription factors is determined partly by their affinities for target DNA
sequences, we examined the differential binding affinities of homo- and heterodimeric versions of retinoic acid receptors for site A by
Scatchard analysis. As shown in Figure
6
A and B, the binding affinities (
K
d
) of RXR[alpha]/RAR[alpha], RXR[alpha]/RAR[beta] heterodimers for site A are 3.1 and 4.0 nM
respectively, whereas the
K
d
obtained for RAR[beta] homodimers is 10.4 nM (Fig.
6
C), which is similar to that reported for RXR[alpha] homodimers (
9
). Although it has been shown that the efficiency of DNA binding of RXR[alpha]/RAR[alpha] heterodimers is much higher than that of RXR[alpha] homodimers (
16
,
17
), this is the first time that we quantitatively demonstrate RXR/RAR
heterodimers bind to site A with a higher affinity than RXR or RAR homodimers.
To examine the transcriptional activation of apoAI site A by homo- and heterodimeric versions of RXR and RAR in yeast, we transformed the
yeast strains expressing these receptors with a reporter plasmid containing two
copies of apoAI site A as the enhancer and analyzed the reporter enzyme ([beta]-gal) induction in response to 9-
cis
RA or all-
trans
RA. For RXR[alpha] homodimers, we used the double transformant yeast strain, which
expressed RXR[alpha] constitutively as described previously (
8
) whereas RAR[alpha] or RAR[beta] homodimers were expressed under the CUP1 promoter. As shown in
Figure
7
, the three homodimers exhibited distinct hormone responsiveness to the two
naturally occurring retinoids. 9-
cis
RA is very specific for the transcriptional activation of RXR[alpha] homodimers, whereas RAR[beta] activates transcription in response equally well to both 9-
cis
RA and all-
trans
RA. In contrast, RAR[alpha] did not exhibit any transcriptional activity in the presence of these
two ligands. The [beta]-gal activities detected for RAR[alpha] homodimers are similar to those observed for the yeast
strain carrying only the reporter plasmid (YEpA) (data not shown).
It has been shown that
in vitro
synthesized RAR does not bind to retinoic acid response elements (RAREs) even
in the presence of 9-
cis
RA (
7
). However, transient cotransfection studies using CV1 cells indicated that RAR[alpha] homodimers could activate transcription from RAREs of the CRBP1, RAR[beta]2 and apoAI genes in response both to 9-
cis
RA and all-
trans
RA (
20
). It is now apparent that this discrepancy is due to the presence of endogenous
RXR in mammalian cells, which form heterodimers with RAR leading to ligand-dependent transcriptional activation. The apoAI site A is a complex RARE,
which binds to several members of the steroid/thyroid receptor superfamily
including RXR, RAR, and the orphan receptor, hepatic nuclear factor 4 (HNF-4) (
18
,
19
). One important observation in our previous studies is that yeast expressed RXR[alpha] homodimers bind to site A of the apoAI enhancer in the absence of 9-
cis
RA, but transactivates a yeast basal promoter linked to site A only in the
presence of 9-
cis
RA (
8
). As an on going investigation of the `transactivation code' for specific
combinations of RAR homo- and heterodimers on site A of the apoAI gene, we expressed different homo- and heterodimers of RAR/RXR in yeast strains carrying the site A
reporter plasmid and examined their DNA-binding properties and transcriptional activity.
Our EMSA studies clearly indicate that RAR[alpha] does not bind to site A, whereas RAR[beta] binds to site A only in the presence of 9-
cis
RA or all-
trans
RA. These DNA-binding studies were further supported by the transcription activation
experiments which demonstrated that RAR[alpha] did not exhibit ligand-dependent transcriptional activation, whereas RAR[beta] transactivated site A equally well in response to 9-
cis
RA or all-
trans
RA. These observations are in contrast to mammalian cell (CV1) cotransfection
studies, which indicated that RAR[alpha] but not RAR[beta] transactivated site A of the apoAI gene in response to both 9-
cis
RA or all-
trans
RA (
20
). Whether this discrepancy is due to the presence of tissue specific factors
that heterodimerize with RAR[alpha] or RAR[beta] in CV1 cells remains to be determined. However, it has been
demonstrated quantitatively by immunoprecipitation, the presence of RXR[alpha], RAR[alpha] and RAR[gamma] in HeLa cells, HepG2 cells and MCF-7 cells (
21
). RAR[beta] has also been shown to transactivate the thyroid response element
(TREpal) carrying an inverted repeat of AGGTCA in yeast cells in response
primarily to all-
trans
RA and weakly to 9-
cis
RA (
22
). Both RAR[alpha] and RAR[beta] form heterodimers with RXR[alpha] very efficiently. Furthermore, saturation analyses
indicated that these heterodimers bind to site A with an affinity two to three
times higher than those observed for RAR[beta] or RXR[alpha] homodimers (
9
). These differential DNA-binding affinities correlate quite well with the transcriptional
activation potency exhibited by different homo- and heterodimeric versions of RXR/RAR. 9-
cis
RA is a more potent ligand than all-
trans
RA for both heterodimers. Similar hormone responsiveness was observed for RXR[gamma]/RAR[gamma] heterodimers in yeast cells carrying the reporter plasmid
containing a retinoid response element ([beta]RE) derived from the promoter of the RAR[beta] gene (
23
). In contrast, all-
trans
RA is the potent ligand for RXR/RAR heterodimers when TREpal is used as the
cis
-acting element in yeast cells (
22
). Thus, 9-
cis
RA and all-
trans
RA can activate a wide variety of target genes via different combinations of
homo- and heterodimeric versions of RAR/RXR, leading to diversified biological
responses. Although RXR[alpha]/RAR[alpha] and RXR[alpha]/RAR[beta] heterodimers exhibit higher affinity binding for
apoAI site A, we cannot rule out the possibility that the hormone
responsiveness to different retinoids in this system is contributed by both
homo- and heterodimers. However, we have detected several synthetic compounds
(non-retinoids) from our chemical library that can activate RXR[alpha]/RAR[alpha] heterodimers but not RXR[alpha]/RAR[beta] heterodimers or RXR[alpha] homodimers in this yeast expression
system (unpublished observation) indicating the usefulness of this yeast-based assay to detect selective ligands for the retinoid receptor
superfamily.
Although yeast expressed RAR[alpha] does not bind to site A, it binds to the radioactive ligand, [
3
H]9-
cis
RA with an affinity (
K
d
= 10
-10
M) similar to that observed for RXR[alpha] and RAR[beta] produced by yeast cells. This observation indicates that binding
of a specific ligand to a receptor and target gene activation by the same
ligand are two distinct molecular events. It has also been shown that the
ecdysone receptor, another member of the nuclear receptor superfamily, does not
bind to its natural ligand, 20-OH ecdysteroid or muristerone A unless it heterodimerizes with another
nuclear factor, Ultraspiracle (usp), the insect homologue of RXR (
24
). Taken together, the reconstitution of RXR/RAR and other nuclear receptor
functions in yeast (
8
-
10
,
12
,
15
,
22
) has allowed us to redefine the complexity of signal transduction pathways
exhibited by the nuclear receptor superfamily. Thus, target gene activation can
be achieved by four different classes of nuclear receptors. The first class
belongs to the classical steroid hormone receptors (androgen, progesterone or
estrogen), as well as RXR[alpha] and RAR[beta] homodimers, which are ligand-activated transcription factors. In contrast, RAR[alpha] homodimers bind ligands but can not transactivate a
target gene unless they heterodimerize with RXR[alpha]. The third class of nuclear receptor is the ecdysone receptor, which
requires the presence of its heterodimeric partner, usp to bind hormone and
subsequently for target gene activation (
24
). The last group of nuclear receptor is the ill-defined `orphan receptors', which have not been associated with a specific
ligand. These orphan receptors can function as ligand-independent transactivators such as hepatic nuclear factor 4 (HNF-4) (
9
,
19
) or repressors such as apolipoprotein regulatory protein 1 (ARP-1) (
25
).
In conclusion, the observation that site A is also a target for RAR[beta] homodimers suggests an important role of all-
trans
RA in the regulation of the apoAI gene as opposed to our initial belief that
its isomeric form, 9-
cis
RA is the major active ligand to upregulate the apoAI gene (
25
,
26
). Furthermore, the present studies clearly indicate that this yeast expression
system is a powerful tool to identify selective ligands for different homo- and heterodimeric versions of RXR/RAR in combination with different
RAREs.
The authors would like to thank Dr Chris Glass (UCSD) for the RXR[alpha] antibody.
REFERENCES
Return