Molecular Endocrinology 14 (11): 1849-1871
Copyright © 2000 by The Endocrine Society
Function of N-Terminal Transactivation Domain of the Estrogen Receptor Requires a Potential
-Helical Structure and Is Negatively Regulated by the A Domain
Raphaël Métivier,
Fabrice G. Petit1,
Yves Valotaire and
Farzad Pakdel
Equipe dEndocrinologie Moléculaire de la Reproduction
UPRES-A CNRS 6026 Université de Rennes I 35042 Rennes
cedex, France
 |
ABSTRACT
|
|---|
Transcriptional activation by the estrogen
receptor (NR3A1, or ER) requires specific ligand-inducible activation
functions located in the amino (AF-1) and the carboxyl (AF-2 and AF-2a)
regions of the protein. Although several detailed reports of ER
structure and function describe mechanisms whereby AF-2 activates
transcription, less precise data exist for AF-1. We recently reported
that the rainbow trout and human estrogen receptors (rtERs and hERs,
respectively), two evolutionary distant proteins, exhibit comparable
AF-1 activities while sharing only 20% homology in their N-terminal
region. These data suggested that the basic mechanisms whereby AF-1 and
the ER N-terminal region activate transactivation might be evolutionary
conserved. Therefore, a comparative approach between rtER and hER could
provide more detailed information on AF-1 function. Transactivation
analysis of truncated receptors and Gal4DBD (DNA binding domain of the
Gal4 factor) fusion proteins in Saccharomyces cerevisiae
defined a minimal region of 11 amino acids, located at the beginning of
the B domain, necessary for AF-1 activity in rtER. Hydrophobic cluster
analysis (HCA) indicated the presence of a potential
-helix within
this minimal region that is conserved during evolution. Both rtER and
hER sequences corresponding to this potential
-helical structure
were able to induce transcription when fused to the Gal4DBD, indicating
that this region can transactivate in an autonomous manner.
Furthermore, point mutations in this 11-amino acid region of the
receptors markedly reduced their transcriptional activity either within
the context of a whole ER or a Gal4DBD fusion protein. Data were
confirmed in mammalian cells and, interestingly, ERs with an
inverted
-helix were as active as their corresponding
wild-type proteins, indicating a conserved role in AF-1 for these
structures. Moreover, using two naturally occurring rtER N-terminal
variants possessing or not the A domain (rtERL
and rtERS, respectively), together with A
domain-truncated hER and chimeric rtER/hER receptors, we demonstrated
that the A domain of the ER plays an inhibitory role in
ligand-independent activity of the receptor. In vitro and
in vivo protein-protein interaction assays using both rtER
and hER demonstrated that this repression is likely to be mediated by a
ligand-sensitive direct interaction between the A domain and the
C-terminal region of the ER.
 |
INTRODUCTION
|
|---|
Steroid hormones such as estrogen trigger and govern essential
functions such as growth, differentiation, and the functioning of many
target tissues. They also have dramatic influences on proliferative and
metastatic states of breast cancer cells (1). Therefore, it is
important to understand the precise mechanisms underlying the action of
estradiol (E2). An important advance in this
research was the demonstration that estrogenic hormones exert their
effects on the regulation of gene transcription through interaction
with at least two predominantly nuclear-localized (2) specific
receptors [NR3A1 and NR3A2 (3), or estrogen receptor-
(ER
) and
estrogen receptor ß (ERß), respectively (4)]. ER is a
trans-acting factor, binding specific
cis-elements (estrogen responsive elements, or ERE) located
within the regulatory regions of genes (5). Upon binding of ligand, a
conformational change allows stabilization of receptor homodimers and
interaction with EREs (6, 7). The importance of estrogen signaling is
evident from ER genes knockout mice, which exhibit some striking
phenotypes including infertility and defects in bone (for review see
Ref. 8 and references therein).
Comparison of steroid hormone receptor sequences permitted the
emergence of a superfamily of nuclear receptors (NRs) (9) that includes
receptors for steroid and other hormones, as well as orphan receptors
for which no ligand is known. There is now some evidence that ligand
binding in this family was acquired during evolution (10). Sequence
comparisons between ERs from different species have shown that these
proteins can be divided into six functionally distinct domains, denoted
A to F (11). This division was extended to the entire superfamily.
While the C domain containing two zinc fingers is responsible for the
specific interaction with steroid-responsive elements (12, 13) and has
a constitutive dimerization property (7), the E region located in the C
terminus is involved in hormone binding (11) and hormone-dependent
dimerization (14) and contains a transactivation function (AF-2) (15, 16). Recently, an activation function called AF2a residing in the
boundary region between the hinge (D) and the E domains of the human ER
(hER) was reported (17). Although the N-terminal region of ERs is the
less conserved domain of the protein during evolution, it contains an
additional AF-1 that could function in an hormone-independent manner
(11, 18, 19, 20). In the context of a full-length receptor, AF-1 is only
active after ligand binding. Full activity of the ER is thought to be
due to a synergistic effect between the AFs (20, 21). While no clear
role is yet ascribed for the F domain in transcriptional activation by
the receptor, it may be important for discrimination between agonists
and antagonists (22) and for E domain-mediated dimerization signal
(23).
Many studies have focused on AF-2 function in ER, leading to a better
understanding of the sequences and mechanisms involved in both ligand
recognition and discrimination (24, 25, 26, 27) and in transcriptional
activation by the liganded receptor (for review see Refs. 28, 29, 30). The present study focuses on the N terminus of ER, since this
region is critical for the antiestrogenic properties of some compounds
(18, 31) and may be responsible for interaction with coactivators
(32, 33, 34) or corepressors as shown for the thyroid receptor (35). The
transactivation function AF-1 begins to be considered as the most
important one of the receptor, directing its activity depending upon
the cell context (36, 37). Different regions of the ER were previously
defined as sufficient for AF-1 activity according to the cell type
(19). However, to date, the AF-1 region in ER has not been precisely
defined since, in yeast as well as in mammalian cells, regions
identified overlap the entire B domain. We have previously shown that
rtER, like hER, possesses a functional AF-1 (38), despite poor sequence
homology (20%) between the two proteins. Therefore, we postulated that
the molecular basis of AF-1 might be conserved in the two species, and
we focused our study in a comparative manner, by identification of the
regions implicated in rtER AF-1. A minimal sequence of 11 residues
possessing intrinsic transactivation potency was identified to be
necessary for rtER AF-1. A predictive method for detecting secondary
structure indicates that this region could adopt an
-helix
conformation. Such a structure was also found in hER, which
exhibited similar characteristics, and the two
-helices were
perfectly interchangeable in terms of transactivation.
Additionally, comparison of the transcriptional activity of a naturally
occurring rtER truncated for its A domain (rtERS)
with the full-length form (rtERL) and hER
(39, 40), indicated that the A domain could be involved in repression of
ligand-independent activity. In this study, we verified this hypothesis
by generation of an hER truncated for its A domain, and interestingly,
direct protein-protein interaction assays with ER domains demonstrated
that the A domain interacts with the C-terminal region. This
interaction appeared to be stronger in the absence of
E2, and was weakened by E2
treatment. These data could therefore explain how the A domain
represses the hormone-independent activity of the ER. This paper thus
reports the characterization of an evolutionary conserved mechanism by
which the ER N-terminal region functions, i.e. the
requirement for a potential
-helix, and an indirect repression by
the A domain in the absence of ligand.
 |
RESULTS
|
|---|
Comparative Analysis of rtER and hER Transactivation Potency in
Yeast
Although a large number of reports describe the role and function
of ER domains or subdomains, considerably less attention has been
focused on the role of the N-terminal region. Previous reports have
implicated large sequences in the N terminus in AF-1 activity in
different cell contexts (19, 21). A major influence of this
transactivation function is in antiestrogens activity (18, 31).
Therefore, we decided to analyze in detail the function of the ER
N-terminal region, by employing a comparative approach between the hER
and the rainbow trout (rt) ER. We have recently identified and
characterized another rtER isoform, called rtERL,
from trout ovary produced by the same gene (40), which possesses an
additional sequence of 45 amino acids (a.a) (Fig. 1A
). The earlier isolated rtER
form, now called rtERS, showed no homology in its
N terminus with the A domain present in mammalian ERs. This observation
together with multiple sequences alignments and the recent isolation
and characterization of two similar isoforms in chicken (41) led us to
the conclusion that this upstream sequence is the rtER A domain (Fig. 1A
). Note that homology between rtER and hER A domains does not exceed
15%. In yeast, both AF-1 and AF-2 of ER are functional, but with an
acute sensitivity to AF-1 (19, 36, 38), providing a powerful model with
which to study the ER AF-1. To compare transcriptional activity, the
cDNAs corresponding to these proteins were subcloned in the YEpucG
yeast expression vector (39, 42). BJ2168 yeast host strain was
cotransformed with the pLG
178/3EREc reporter plasmids (Lac
Z gene controlled by the minimal Cyc1 yeast promoter and three
consensus EREs, respectively) and the YEpucG constructs. Transformants
were selected for auxotrophy on the appropriate media, and then
ß-galactosidase activity was measured after a 4-h incubation
with or without 10-6 M
E2. This comparative analysis of transcriptional
activation between the two rtER isoforms and hER in the absence of
hormone in yeast showed a differential behavior. Indeed, the shorter
isoform (rtERS) lacking the A domain exhibited a
consistent E2-independent activity (Fig. 1B
and
Ref. 40), whereas hER and rtERL, which both
possess such a region, did not exhibit a similar characteristic (Fig. 1B
). This ligand-independent activity was ascribed to the AF-1 region
since AF-2 is not exposed to the transcriptional machinery in the
absence of ligand and thus cannot be active (22, 43). Interestingly, a
chimeric rtER containing the hER A-B region instead of its own B domain
[named rtER(hAB)] did not exhibit any hormone-independent activity
(Fig. 1B
). However, after E2 treatment,
transcriptional activity of this chimera was similar to wild-type
receptors. This indicates that the human AF-1 could function in the
rtER protein context and that a region in this hAB sequence represses
the ligand-independent activity of rtERS. Since
the only difference between rtERS and
rtERL is the presence of an A domain, we
postulated that this domain could be responsible for a repressive
effect on AF-1. To test this hypothesis, a truncated hER lacking its A
domain (hER
137) was constructed, and, interestingly, it exhibited
a basal activity representing 17% of the total activity (Fig. 1C
). To
study directly the AF-1, truncated receptors for N- and C-terminal
regions were constructed in both ER species (namely rtBD and hAD), and
their transactivation abilities were assayed on the yeast
3EREc-LacZ reporter. As previously shown on other reporter
plasmids (38), rtERS possesses, as does hER, a
functional AF-1 (Fig. 1C
), as the two rtBD and hAD constructs
constitutively activate transcription, to 7580% of the maximum
induction obtained with the full-length ER. In contrast, the activity
of the C-terminal region (rtCF and hCF) represented about 10% the
activity of the full-length ER. These data show that the N-terminal
region of rtER, like that of the hER, functions in a
hormone-independent manner when isolated from its C-terminal
region.

View larger version (29K):
[in this window]
[in a new window]
|
Figure 1. Comparative Analysis of Transcriptional Activities
of Two Phylogenically Distant ERs in Yeast Saccharomyces
cerevisiae
A, Amino acid representation of the divergence between the two rtER
isoforms. The arrow represents the point where sequences
begin to be identical. B, BJ2168 yeast cells were cotransformed with
the pLG178/3EREc (3 ERE-Cyc-Lac Z) reporter plasmid
together with the YEpucG constructs expressing the two rtER isoforms,
hER, hER 137, or the chimera rtER(hAB). After auxotrophy
selection, clones were cultured and treated during 4 h with 1000
nM E2 or ethanol (EtOH). ß-Galactosidase
activity was quantified by liquid assays, and values from at least four
separate experiments in triplicate were expressed as percent of
E2 stimulation. C, Yeast cells were transformed with
expression vectors containing cDNA encoding different parts of the hER
and rtER, whereas a control was accomplished by using the empty YEpucG.
Results are expressed as percent of E2 stimulation and
represent the mean ± SD from at least four separate
experiments.
|
|
This primary analysis highlighted two interesting facts: the ER A
domain has a repressive effect on the AF-1 activity of ER, since when
it is deleted, a consistent basal activity is detected; the AF-1 of two
evolutionary distant ERs are functional in yeast, suggesting a common
mechanism of action. This led us to determine precisely which sequences
are required for ER AF-1, using a comparative approach to identify
evolutionary conserved mechanisms.
Identification of Sequences Involved in rtER AF-1 by
Transcriptional Assay in Yeast
To find evolutionary conserved sequences and define the mechanism
of AF-1 transactivation function, the shorter isoform of rtER
(rtERS) was an ideal starting point, due to its
profound ligand-independent activity in yeast. To determine sequences
implicated in AF-1, successive deletions in the N-terminal portion of
rtERS were constructed either in the full-length
receptor context or in the isolated N-terminal region, and their
transactivation potencies were analyzed using the yeast 3EREc-Lac
Z reporter. Data illustrated in Fig. 2A
showed that in contrast to the entire
rtER B domain, none of the deleted rtBD for the 111, 50, 34, or 17
first amino acids were able to induce the transcription of the reporter
gene, indicating that an important sequence for AF-1 resides at the
beginning of the B domain. Previous studies have shown that in some
cellular contexts, AF-1 activates transcription poorly on its own but
could synergize with AF-2 (20, 21). Whereas yeast is well characterized
as an AF-1-dependent cell context, it is possible to explain the lack
of activity of the truncated receptors used in Fig. 2A
by a
nondetectable AF-1 activity. Thus, to detect these residual AF-1
activities through synergism with AF-2, N-terminally truncated
full-length receptors were constructed and expressed in yeast. As shown
in Fig. 2B
, total or successive deletions of N-terminal residues
generated receptors that were able to induce the yeast reporter gene in
a hormone-dependent fashion. However, the maximum activity obtained
with these truncated receptors in the presence of
10-6 M
E2 represented only 1012% of the wild-type
receptor activity. Thus, these mutants are transcriptional activators
similar to the rtCF protein lacking the A and B domains. To determine
more precisely the amino acids necessary for AF-1 activity, receptors
with either a.a 517 or a.a. 817 deleted were constructed and
introduced together with the 3EREc-LacZ reporter into yeast.
Data illustrated in Fig. 2C
show that
rtERS
517 and
rtERS
817 were able to induce reporter gene
activity in a hormone-dependent fashion, with a maximal
E2 stimulation that only represented 11% of the
full-length ER activity. To confirm that the absence of AF-1 activity
of these constructs was specific of the amino acids deleted, and not
dependent upon the simple fact of deletion, a
rtERS lacking the seven first a.a was constructed
(rtERS
17). This truncated receptor retained
8090% of the wild-type ER activity (Fig. 2C
) and therefore provided
a good control for these experiments. These results indicate that a.a.
817 of the rtERS isoform are necessary for its
AF-1 activity and that they might play an essential role in the full
activity of the rtER resulting from the synergistic effect of AF-1 and
AF-2 transactivation functions.

View larger version (31K):
[in this window]
[in a new window]
|
Figure 2. Determination of the rtER Regions Implicated in
AF-1 in Yeast
Yeast cells containing the 3EREc-Cyc-Lac Z reporter
plasmid were transformed with either no vector or plasmids expressing
the different constructs described on the left side of
each panel. After growth on medium containing either 10 or 1000
nM E2, or ethanol (EtOH), ß-Galactosidase
activity (Miller Units) was measured. Values represent mean ±
SD from at least four separate experiments. These
experiments demonstrate that a.a. 717 of rtERS are
necessary for AF-1.
|
|
Deletion of the rtERS 18 First Amino Acids
Is Sufficient to Abolish the Activity of a Gal4DBD Fusion
To confirm the presence of a transactivation function in the rtER
N-terminal domain in an alternative system, a series of constructs
encoding different parts of the receptor linked to the Gal4DBD were
generated. Transactivation potency of these chimeric proteins was
tested in the Y190 yeast cells containing the Lac Z reporter
gene under control of three responsive elements for the yeast Gal4
transcription factor (UASG) (Fig. 3
). The rtERS fused
to the Gal4DBD did not exhibit any activity in the absence of
E2. However, addition of the hormone induced
transcriptional activity of this fusion protein, as described for hER
(44). The fusion of the rtER N-terminal domain to the Gal4DBD generated
a constitutively active protein
(rtERS
153575/Gal4DBD), confirming the fact
that this region contains a transactivation function that could operate
independently of the remainder of the receptor (Fig. 1
). In contrast,
the rtER C-terminal region fused to the Gal4DBD (rtCF/Gal4DBD) was not
functional with or without hormone, indicating that the AF-2 of rtER is
not active by itself in this context. This is in agreement with
previous work that has proposed that binding of
E2 to full-length hER fused to Gal4DBD induces
a conformational change in the chimeric protein enabling AF-1
to mediate transcriptional activation (44). Strikingly, neither
the rtERS
1111 nor the
rtERS
117 truncated receptors fused
with the Gal4DBD were able to induce any Lac Z activity,
while the rtERS
17 receptor was
E2-dependent active (Fig. 3).
However, this latter construct exhibited only a weak activation of
about 10 Miller Units, although it possessed 8090% of the
full-length receptor transcriptional ability (Fig. 2C
). This could be
explained by the fact that we used a low-level expression vector
(pGBT10) in the case of the rtERS
1111,
rtERS
117, and rtER
17 inserts.
Therefore, the lack of transcriptional activity of the
rtERS
1111/Gal4DBD or
rtERS
117/Gal4DBD could be due to a weak
expression of these proteins, resulting in an activity below the
sensitivity of the system. However, the use of a filter-lift assay for
Lac Z activity, known to be more sensitive than a liquid
assay, did not enable detection of ß-galactosidase activity,
even over a very long incubation. Western experiments shown in Fig. 3B
indicated that the truncated receptors expressed using the pGBT10
plasmid were produced at equivalent levels in yeast. Because these
constructs possess both LBD and DBD, they should be able to
dimerize, since the LBD allows ligand-dependent dimerization and
the DBD allows a ligand-independent dimerization ability (7, 14).
Then, to test the functionality of these mutant receptors, their
interaction with the C-terminal region of rtER fused to the Gal4AD
(rtERS
1223/Gal4AD) was assessed, in the
presence of ligand. Results shown in Fig. 3C
confirmed that these
constructs were functional. All these data indicate that the
rtERS
1111/Gal4DBD or
rtERS
117/Gal4DBD have no intrinsic
transcriptional activity. Although the rtER AF-2 fused to the Gal4DBD
was not active in this system, this domain might influence the
potential AF-1 activity of the two truncated fusion proteins. Thus,
constructs devoid of AF-2 were generated (rtB
1111/Gal4DBD and
rtB
117/Gal4DBD) and introduced in this one-hybrid system. As
illustrated in Fig. 3
, neither of the two fusions were
transcriptionally active, whereas they were expressed at similar levels
in yeast. Therefore, it seems that the rtERS
first 17 a.a. are necessary for AF-1 function in this system,
since the activity of the rtERS/Gal4DBD in the
presence of E2 is thought to be mediated through
AF-1.

View larger version (34K):
[in this window]
[in a new window]
|
Figure 3. Transactivation Potency of Truncated rtER Fused to
the Gal4DBD
A, Schematic representation of the different Gal4DBD fusion proteins
used (numbers refer to the amino acid positions). Y190
yeast cells containing the Lac Z reporter gene placed
under control of three responsive elements for the yeast Gal4 activator
(UASG) were transformed with plasmids expressing Gal4DBD
alone (control) or the different fusion proteins indicated. The level
of expression of these constructs was assayed by Western blot using
anti-Gal4DBD antibodies After growth on selective media lacking
histidine, transformants were subjected to 1000 nM
E2 or ethanol (EtOH) treatment, and ß-galactosidase
activity was measured (B). Values represent the mean ±
SD from at least four separate experiments. C, To test the
functionality of these constructs, their homodimerization property was
assayed in two-hybrid experiments using the rtER protein fused to the
Gal4 activation domain (Gal4AD). Transformants were selected by growth
and treated with 1000 nM E2. ß-Galactosidase
activity was measured and values represent the mean ±
SD from three experiments.
|
|
Basic Functionality of ER AF-1 Relies on an
-Helix Possessing
Intrinsic Transactivation Potency
The C-terminal tridimensional structure of NRs is now well
characterized since the crystal structure of holo-retinoid X receptor,
liganded retinoic acid receptor, thyroid hormone receptor, or ER
(45, 46, 47) were defined as a canonical structure of 12
-helices (48).
Similar structural data are not yet available for the N-terminal region
and, therefore, no structure can be ascribed for AF-1. However, a
performant predictive method such as the hydrophobic cluster analysis
(HCA), based upon the hydrophobicity property of the amino acids (49, 50), could be used to define probabilities for the existence of
secondary structures. We followed this predictive approach, and the
rtER HCA plot represented in Fig. 4
, A
and B, revealed that an
-helix (existence probability vs.
a ß-sheet of 2.2) could be formed between a.a 14 and 22, overlapping
the region defined as being required for AF-1 activity (a.a 817).
Interestingly, the deletion of the first 4 a.a of this
-helix
(FNYL), occurring in the rtERS
117 construct,
leads to the disintegration of the structure by the loss of three
hydrophobic amino acids. This could indicate that this structure is
required for rtER AF-1. In this case, AF-1 being a ligand-independent
transactivation function, a major condition for this structure is to
activate, on its own, the transcription state of a target gene.
Therefore, we linked a.a 926 or 1123 of rtERS
to the Gal4DBD (respectively, rtERS926/Gal4DBD
and rtERS1123/Gal4DBD, Fig. 4C
) and introduced
these constructs in Y190 yeast cells containing the 3
UASG-Lac Z reporter gene. While the
basal activity of the reporter yeast strain did not exceed 0.15 Miller
Units, both fusion proteins were able to stimulate the transcription
status of the reporter by almost 5 Miller Units. Such an activity was
specific to the sequences fused to Gal4DBD since rtB
1111/Gal4DBD
did not stimulate the reporter (Fig. 4D
). To demonstrate the link
between transactivation potency and the presence of this potential
structure, we introduced point mutations in the corresponding sequence
fused to the Gal4DBD (a.a 926), converting the hydrophobic amino
acids of the chain to proline, an amino acid known to disrupt secondary
structure without perturbing the chain charge and polarity. Thus,
either the phenylalanine or leucine of the
-helix was substituted. A
70% reduction in transactivation was observed by mutating the first
amino acid of the
-helix (phenylalanine:
rtERS926F15P/Gal4DBD),
whereas no Lac Z activity was detected with the construct
mutated for the central amino acids of the structure (leucine:
rtERS926L18P/Gal4DBD)
(Fig. 4D
). The residual activity obtained with the
rtERS926F15P/Gal4DBD
fusion protein can be explained by the fact that this sequence could
still present an
-helix conformation, but shorter than the original
and with a lower probability. As a control, an alanine external to the
structure was mutated (Fig. 4C
). This mutation
(rtERS926A12P/Gal4DBD
construct) induced a slight increase of approximately 50% of the
reporter activation compared with the wild-type sequence, which may be
due to a better exposition of the structure to the transcription
machinery. In fact, this
-helix is preceded by a succession of
glycines and alanines that could give flexibility to the protein chain.
Substitution of one of the alanines by a proline may create a bend,
perhaps more propitious to an interaction of the structure with the
transcriptional apparatus. Western blot experiments carried out with
yeast extracts demonstrated an equivalent and correct expression of the
fusion proteins (Fig. 4C
).

View larger version (34K):
[in this window]
[in a new window]
|
Figure 4. Identification of a Potential -Helix in
rtERS N-Terminal Region Possessing Intrinsic
Transactivation Potency
A, Illustration of the HCA plot analysis for the rtERS N
terminus. The nomenclature used is classical: hydrophobic amino acids
are circled, proline is represented by *, glycine by
, serine by , and threonine by . B, Simplified
one-dimensional HCA representation of the HCA plot (1
represents an hydrophobic amino acid while the star is
used for prolines and 0 for the others). Numbers reflect
the statistical relevance of the secondary structure
boxed. C, Gal4DBD fusion proteins (rtERS
926 and rtERS 1123) containing the potential -helix,
with or without substitution of some residues in proline (A12P, F15P
and L18P) were constructed and introduced in Y190 yeast cells. Correct
and equivalent expression of these constructs was checked by Western
blot using an anti-Gal4DBD antibody. As controls, vectors expressing
Gal4DBD alone or the rtB 1111/Gal4DBD protein were also included.
D, After selection, ß-galactosidase activity was measured. Values
represent the mean ± SD from at least four separate
experiments.
|
|
Since our data identified a potential
-helix located at the
beginning of the B domain to be important for rtER AF-1, we
investigated whether or not this structure was conserved during
evolution. An HCA-based structural study performed on the hER
N-terminal region revealed a striking conservation of many potential
structures (compare Fig. 5
, A and B, with
Fig. 4
, A and B). The most evident are the two ß-strands separated by
a rich proline sequence and preceded by a coil (a.a. 106133 and a.a.
6396 for hER and rtERS, respectively).
Interestingly, the hER HCA plot displayed an
-helix in the
beginning of the B domain, located within a.a 3944 (Fig. 5
, A and B).
Previous reports characterized a region between a.a. 29 and 63 to be
involved in hER AF-1 (19, 31). To examine the transactivation potency
of this potential
-helical structure in this region, a.a 3547 were
linked to the Gal4DBD (hER3547/Gal4DBD), with or without point
mutations resulting in substitution of either the leucine or tyrosine
residues to proline (hER3547L39P or
hER3547Y43P, respectively). A control was also
generated by substitution of a tyrosine to phenylalanine
(hER3547Y43F), which conserved the basis of
structure, charge, polarity, and sterical properties of the chain.
These constructs were expressed in Y190 yeast strain, and liquid
ß-galactosidase assays were performed. Results illustrated in Fig. 5D
demonstrate that the hER 3547 sequence does possess an intrinsic
ability to activate the transcription state of the
3UASG-LacZ reporter nearly 6 to 7 Miller units.
Moreover, disrupting the structure with point mutations to substitute a
proline for either the leucine or the tyrosine abolishes this activity,
whereas substitution of the tyrosine to phenylalanine did not affect
the constitutive activity of the GAl4DBD fusion protein (Fig. 5D
).
Since the expression of these fusions in yeast is similar (Fig. 5C
),
this indicates that, as in rtER, a small sequence with a specific
spatial organization could activate transcription. To confirm the
implication of this structure in AF-1 in the whole receptor context, we
analyzed the transcriptional ability of an hER deleted for this
potential secondary structure (hER
144). As illustrated in Fig. 6
, the deletion of this region produces a
55% loss of transcriptional activity of the receptor. Note also that
this receptor mutant exhibited a more significant reduction (
80%)
of its ligand-independent activity, compared with the hER
137 (4.7
Miller units vs. 25.3 for the hER
137). As performed for
Gal4DBD fusion proteins, we introduced point mutations in the
-helix of the hER minus its A domain (hER
137) for which a
ligand-independent activity is detectable. More precisely, we mutated
the first amino acid of the structure (leucine at position 39) and the
central amino acid (tyrosine at position 43) into prolines, and
examined the transactivation abilities of the resulting mutant proteins
in yeast (hER
137L39P and
hER
137Y43P, respectively) (Fig. 6
). This
analysis showed that the mutated receptors have a reduced ability to
activate gene transcription, similarly to hER
144, confirming that
the structure of this small region is important for AF-1 activity.
As shown with the rtERS (hAB) chimera (Fig. 1D
),
the hER AF-1 could function within the rtER context, suggesting that
the two AF-1 sequences are interchangeable. To answer the question of a
conserved role of these structures in the whole receptor context, two
chimeras were constructed by exchanging the two
-helices either
containing or not a crucial point mutation (Fig. 7A
). Note that one or two flanking amino
acids were conserved in these transpositions to preserve the
environment of the helical structure. Transactivation ability of these
receptor chimeras was tested in yeast containing the 3
ERE-Cyc-Lac Z reporter. As illustrated in Fig. 7B
, the
replacement of the human helix by its rainbow trout counterpart, or
vice versa, did not affect the activity of the receptors, either in the
absence or presence of E2. Indeed, the
hER
137(rt
) as well as the hER
137 or the rtER
17(h
)
activated the reporter gene state from 150 to 250 Miller Units. Point
mutations within the potential helix of these chimeric receptors
dramatically reduced their transcriptional activity. This reduction
occurred to a greater extent in the case of the
rtER
17(h
Y43P), confirming the absolute
requirement of such a structure for activity of rtER in yeast.
Together, these data suggest that a small region located in the
beginning of the B domain that could adopt an
-helix conformation
is necessary for ER AF-1 in yeast.

View larger version (43K):
[in this window]
[in a new window]
|
Figure 7. The -Helices of Both Receptors Are Perfectly
Interchangeable
A, To determine whether the -helices identified within the rtER and
hER are playing the same role in their AF-1, transposed chimeric
receptors containing the wild-type or point mutant sequences were
constructed. Human sequence is underlined, and sequences
transposed are boxed. Point mutations are indicated with
an asterisk. B, Transcriptional activities of these
chimeric receptors were tested in a yeast strain containing the
3-ERE-Cyc-LacZ reporter, and ß-galactodidase was
measured on stable transformants treated either with ethanol (EtOH) or
10 to 1000 nM E2. Bars represent
the mean ± SD of values obtained in three independent
experiments.
|
|
Involvement of the
-Helix for ER AF-1 in Mammalian Cells
Although yeast is a powerful model with which to study the
transactivation abilities of NRs, it remains a simple cellular model.
To verify the implication of
-helices in AF-1 in both ER species in
more complex models, transient transfection experiments were conducted.
Three cell lines were chosen for this study in accordance with their
dependence or not on ER AF-1: HeLa cells in which AF-2 is the dominant
or even the only activation function driving ER activity; HepG2 cells,
which constitute the opposite cell context (17, 18, 21); and the CHO
cell line in which both AFs are functional (our unpublished
observations). These three cell lines were transiently transfected by a
classical calcium phosphate/DNA coprecipitation method. Transient
expression was performed during 36 h in steroid-free media
containing 10-6 M
E2 or ethanol as vehicle control. The
transcriptional activity of the various mutants was assessed by
cotransfection of a luciferase reporter driven by the thymidine kinase
promoter under control of one ERE (ERE-TK-Luc). Luciferase activities
were normalized using the SV/ß-galactosidase reporter (pCH110) as an
internal control for transfection efficiency. Results expressed as fold
induction of the reporter are shown in Fig. 8
. The activity of these constructs was
identical in HeLa Cells. The luciferase reporter was activated 8- to
10-fold in the presence of ligand, indicating that each of the
truncated receptors is functional in mammalian cells. Importantly,
activation by the hCF construct was similar to the full-length
receptor, confirming that AF-2 is the only transactivation function
used in HeLa cells. Moreover, these results show that all of the
deletions or point mutations performed in the two ER proteins do not
affect their AF-2 activity. In the AF-1-dependent cell context,
truncated receptor activities were strikingly different, and differed
to a greater extent in HepG2 than CHO cells, illustrating their
relative relevance to this transactivation function. For instance, in
CHO and HepG2 cells, the two rtER isoforms were similar activators in
the presence of ligand, stimulating the reporter up to 10-fold.
However, the shorter isoform exhibited a basal activity representing
25% or 45% of the E2-stimulated activity in CHO
or HepG2 cells, respectively. Since the same observation could be made
in these contexts when comparing hER to the A domain truncated form
hER
137, this confirms the hypothesis of a repressive function for
the A domain of the ER. In HepG2 cells, deletion of the 111 first amino
acids of the rtER resulted in a 80% reduction of reporter induction,
whereas disruption of the
-helix occurring within the rtER
817
construct generated a 50% loss in transcriptional efficiency. This
indicates that in mammalian cells, as opposed to yeast, other regions
are implicated in rtER AF-1, as it was demonstrated for hER (19). The
hER
144, hER
137L39P, and
hER
137Y43P activated the reporter from
approximately 4- to 7-fold in CHO or HepG2 cells, respectively,
demonstrating that the human helical homolog of the rtER is also
important for AF-1 in mammalian cells. On the other hand, these latter
constructs were more potent activators than the hCF receptor, which
activated 2-fold the reporter in HepG2 cells. This also confirms that
other regions in the N terminus of hER are implicated in AF-1. The
transposed hER including the rtER helical structure was as
active as the wild-type ER in the two AF-1 sensitive cell contexts,
whereas the transposed hER, including the mutated version of the
structure, stimulated the reporter to the same level as the hER
144
or point mutant constructs. The inverse transposed receptors (rtER with
the human structures) gave rise to similar results. These data obtained
in transient transfection experiments confirm the implication of the
-helical structures in their cognate AF-1, but also in the context
of evolutionary distant protein. These results also demonstrate that
regions other than this structure are involved in ER AF-1 in mammalian
cells.

View larger version (38K):
[in this window]
[in a new window]
|
Figure 8. The ER AF-1 Requires the -Helix in Mammalian
Cells
To confirm the importance of the rtER and hER -helices in a more
complex cellular context than yeast, transient transfection experiments
were performed using different cell lines depending upon their relative
sensitiveness to AF-1: HeLa (A), CHO (B), and HepG2 cells (C). The
three cell lines were transfected at 6070% confluence with a
classical calcium phosphate/DNA precipitation protocol using an
ERE-TK-Luc reporter and pCH110 as internal control. CHO cells were
transfected in 24-well plates with 25 ng of expression vector, whereas
250 ng of expression vector were used in 6-well plates for HeLa and
HepG2 cells. After 36 h of transient expression in steroid-free
media containing ethanol (EtOH) or 10-8 M
E2, luciferase activity was quantified using a luminometer
and ß-galactosidase assay was performed. Luciferase activities were
normalized for transfection efficiency with the ß-galactosidase
activity and expressed as the fold induction vs. the
activity obtained with the promoter alone. Results from at least three
experiments in triplicates are expressed as the mean ±
SEM.
|
|
Inhibitory Effect of the A Domain on ER Ligand-Independent Activity
Could Be Mediated by a Direct Interaction between the A Domain and the
C-Terminal Region
As previously shown, the comparison of the
rtERS and hER transcriptional activity in the
absence of hormone both in yeast and mammalian cells clearly revealed
that the A domain of ER could repress the ligand-independent activity
of the receptor (Figs. 1
and 8
and Ref. 40). This was further
substantiated by truncation of the hER A domain (hER
137) and
isolation of another rtER isoform, rtERL,
possessing this domain (Fig. 1B
). As the aim of this study is to better
comprehend ER N-terminal functions, we attempted to characterize how
the A domain could act. As illustrated in Fig. 1C
, the hAD construct
was active independent of E2, sharing the same
pattern of transactivation as the rtBD construct devoid of the A
domain. Moreover, constructs possessing the entire A+B domains of the
rtER or hER fused to the Gal4DBD showed similar activities to the B
domain alone (data not shown). Thus, we hypothesized that other domains
of the ER could be implicated in the repressive effect of the A domain,
by a process implying a direct interaction between the A domain and
other parts of the receptor.
Pull-down assays were used first to study the in vitro
interaction between glutathione-S-transferase fusion with
the ER A domain (GST/rtA and GST/hA, Fig. 9
) and
[35S]-labeled ER constructs translated in
rabbit reticulocyte lysate that overlap the entire sequence of the two
receptors, namely rtERS,
rtERS
1220,
rtERS
157575, hER, hER
1178, and
hER
272595 (Fig. 9
, A and D). As shown in Fig. 9B
, the three rtER
constructs were expressed in the reticulocyte system and migrated at
the expected molecular sizes (e.g. 66, 35.5, and 17.8 kDa
for rtERS, rtERS
1220,
and rtERS
157575, respectively). In the
pull-down experiment, the A domain interacted with both
rtERS and rtERS
1220,
whereas no interaction was found with the B domain
(rtERS
157575). The effect of
E2 on this interaction was examined (Fig. 9B
, lanes 4, 8, 9, and 17), using lysates treated either with ethanol or
with 10 or 50 µM E2,
concentrations known to induce a conformational change in rtER
detectable by protease digest analysis (data not shown). This
experiment showed that E2 significantly reduces
this interaction process. Indeed, quantification with a phosphoimager
showed that in the absence of E2, 10% of
rtERS labeled protein was retained vs.
5% after E2 treatment, and 7% vs.
2.5% for rtERS
1220. As illustrated in lanes
8 and 9, the effect of E2 was maximal at the two
E2 doses tested. Effect of two antiestrogens on
this interaction was also tested, by treating
rtERS
1220 lysate with 100 or 500
µM of 4-hydroxytamoxifen (OHT) or
ICI164,384. As shown in lanes 10 and 11, OHT was
able to compete the interaction between rtER A domain and
rtERS
1220, but with a lower efficiency than
E2 (6 to 4% vs. 2.5%), irrespective
of the dose. In contrast, the pure antiestrogen
ICI164,384 was unable to affect this interaction,
suggesting that the conformational change induced by this ligand is not
appropriate to expose AF-1. Similar results were observed with
rtERS treated by antiestrogens (data not shown).
Similarly, the three hER constructs were expressed at the correct size
(66, 46, and 29.8 kDa), and only the hER and hER
1178 labeled
proteins were retained on the GST/hA beads (Fig. 9E
). Moreover,
E2 was able to reduce this interaction from 10 to
7% or 9.8 to 6.5% in the case of labeled hER or hER
1178,
respectively. The OHT was also able to reduce this interaction from
19.8% to 9 or 8.7%, depending on the dose, whereas
ICI164,384 had no effect. The validity of the
experiments was confirmed by performing a Coomassie blue staining of
the gels (Fig. 9
, C and F), demonstrating equivalent loading on the
SDS-PAGE gels.

View larger version (36K):
[in this window]
[in a new window]
|
Figure 9. In Vitro Evidence for an
Evolutionary Conserved Interaction between ER A Domain and C-Terminal
Region
A and D, Schematic representation of the ER constructs used for
in vitro [35S]methionine-labeled
translation, and the GST fusion protein with the A domains from rtER
(GST/rtA) or hER (GST/hA). B and E, The GST fusion proteins were
analyzed by pull-down assays for their binding to the different ER
constructs. Pull-down assays were performed in the absence (lanes 3, 7,
and 16) or in the presence (lanes 4, 9, and 17) of 50 µM
or 10 µM E2 (lane 8), 100 or 500
µM of OHT (lanes 10 and 11), or
ICI164,384(ICI, lanes 12 and 13). Input lanes (1 5 14 ) represent 25% of the amount of labeled proteins used in the assay.
C and F, Stability of the GST fusion proteins and equal loadings were
checked by Coomassie blue staining. Positions of standard markers
(S.M.) are indicated.
|
|
To confirm these interactions in vivo, two-hybrid assays
were performed. For this purpose, the A domain from
rtERL and hER were fused to the Gal4DBD
(rtA/Gal4DBD and hA/Gal4DBD), while the C-terminal
(rtERS
1223 and hER
1178) or N-terminal
(rtERS
149575 and hER
181595) domains
were fused to the activation domain of the yeast Gal4 protein (Gal4AD),
as well as the rtERS and full-length hER
(Fig. 10
, A and
C). Y190 yeast strain was transformed with these
fusion constructs alone or in association, and correct expression of
the proteins was confirmed by Western blot of whole yeast extracts
using anti-Gal4DBD or anti-HA epitope antibodies (data not shown).
Positive clones from ß-galactosidase filter-lift assays were cultured
and treated or not with 1 µM
E2, 50 µM
OHT, or 50 µM
ICI164,384 for 4 h, and Lac Z
reporter gene activation was then quantified in liquid assay. Results
illustrated in Fig. 10B
show that the rtER A domain interacts with the
full-length and rtERS
1223 constructs, but
not with rtERS
149575.
E2 reduced these protein-protein contacts by
approximately 70%. The mixed antiestrogen OHT also reduced this
interaction by approximately 45%, whereas
ICI164,384 had no effect, as seen in pull-down
experiments. Interactions were specific since no transcriptional
activation was seen when transforming the Gal4DBD fusions alone or in
association with the Gal4AD. Two-hybrid assays using the hER fusions
gave similar results. Indeed, the hA/Gal4DBD fusion did not activate
transcription alone, or when cotransformed either with the Gal4AD as
control or with hER
181595 (Fig. 10D
). On the other hand,
significant activity was detected by cotransformation of the hA/Gal4DBD
with either hER/Gal4AD or hER
1178/Gal4AD constructs. These
interactions were not affected by ICI164,384
treatment, but E2 and OHT inhibited this physical
interaction by 40% or 25%, respectively. Occurrence of these physical
interaction reductions in the presence of E2 and
OHT were statistically significant at P < 0.001, as
assessed by Students t test. From these experiments, we
suggest that the ER A domain can, possibly through an intramolecular
folding, interact with its own C-terminal region to repress AF-1 in the
absence of ligand. The occurrence of this interaction in two
phylogenically far-distant species receptors suggests that sequences
and/or structures within the A domain and the C-terminal regions of the
receptors were conserved during evolution. To confirm this hypothesis,
we performed transposed experiments both in vitro and
in vivo, by testing the interaction of the rtER A domain
with hER subdomains and vice versa. Results of these experiments are
shown in Fig. 11
. The rtER A domain was
able, in both the two-hybrid assay (Fig. 11A
) and pull-down experiments
(Fig. 11B
) to interact with the human receptor and its C-terminal, but
not N-terminal, region. The opposite was also verified with the
hER A domain. These physical interactions were also reduced by
E2 treatment in two-hybrid and pull-down
experiments (compare lanes 3 and 4; 7 and 8; 11 and 12; and 15 and 16,
Fig. 11B
). The ratio of this reduction was similar to when we performed
the interaction of ER A domain with their C-terminal counterparts. The
results of these experiments demonstrate that sequences or structures
implicated in the interaction between ER A domain and the C-terminal
region are potentially conserved during the evolution.

View larger version (17K):
[in this window]
[in a new window]
|
Figure 10. In Vivo Interaction of A Domain and
C-Terminal Region
A and C, Illustration of the Gal4AD and Gal4DBD fusions used in the
two-hybrid assay. B and D, Yeast strain containing
3UASG-LacZ and 2UASG -His
reporter genes were transformed with the rtA/Gal4DBD or hA/Gal4DBD in
combination with the Gal4AD fusions shown in panel A. Controls were
performed by expressing A domain/Gal4DBD fusions alone or with Gal4AD.
Transformants were subjected to growth selection (see Materials
and Methods) and filter-lift for ß-galactosidase assays, before quantification. For this
purpose, cells were grown in liquid media and treated with ethanol
(EtOH) or E2, OHT, or ICI164,384 (ICI). Values
represent the mean ± SD from five separate
experiments. Double asterisks indicate a
significant difference between controls (EtOH) and hormonal treatments
at P < 0.001 by Students t test.
|
|

View larger version (33K):
[in this window]
[in a new window]
|
Figure 11. Physical Interaction between ER A Domain and
C-Terminal Region Implies Conserved Sequences or Structures
A, Y190 yeast strain was transformed either with the rtA/Gal4DBD or
hA/Gal4DBD as probes for the two-hybrid assays. Full-length or deleted
receptors fused to Gal4AD were used as baits and cotransformed with the
A domain of the other species construct. After selection, transformants
were treated with ethanol (EtOH) or 1,000 nM E2
and ß-galactosidase activity was assessed. Results shown are
expressed as the mean ± SD from three separate
experiments. Double asterisks indicate a
significant difference between controls (EtOH) and hormonal treatments
at P < 0.001 by Students t test.
B, Pull-down assays using in vitro
[35S]methionine-labeled full-length or C-terminal region
of the two receptors were performed with the opposite A domain fused to
the GST fusion or the GST alone (lanes 2, 6, 10, and 14). Incubations
were in the presence of ethanol (EtOH, lanes 3, 7, 11, and 15) or 50
µM E2 (lanes 4, 8, 12, and 16). Equal loading
of GST fusion proteins was checked by Coomassie blue staining.
Positions of standard markers (S.M.) are indicated.
|
|
 |
DISCUSSION
|
|---|
E2 receptor (NR3A1) is a modulatory
transcriptional activator that contains three transactivation
functions, namely AF-1 in the N-terminal portion of the protein and
AF-2a and AF-2 in the C terminus. For several years, AF-2 was the most
studied function since ligand directly regulates the exposition of this
activation function, providing a useful explanation for the ligand
dependency of the protein transcriptional property (22, 43). Since no
precise regions or structures were defined for ER AF-1 function, we
decided to adopt a comparative approach between evolutionary
far-distant proteins for studying ER AF-1, postulating that the
molecular basis of this functional domain might be conserved. We used
as a model two naturally occurring N-terminal variants of rtER and the
hER
expressed in a yeast system in which AF-1 and AF-2a appear the
dominant receptor activation functions (17). Results were then
confirmed in the more complex system of mammalian cells.
To define regions of rtER implicated in AF-1, several deletions in the
N-terminal domain of either full-length or C-terminal truncated
receptors and a series of plasmid constructs containing various
portions of the rtER N-terminal region fused to the DBD of the yeast
Gal4 transcription factor were generated. From these analyses, a
segment of 11 a.a (residues 717) responsible for
hormone-independent transcriptional activation by the receptor in yeast
was delineated. In agreement with reports describing steroid receptor
activation functions (16, 51), this segment of rtER has no homology
with the acidic transactivation domains of herpes simplex VP16 or yeast
Gal4 activators. However, using the HCA plot prediction, an amphipathic
-helix could be formed between residues 14 to 22 of
rtERS (rtERS
-H1422),
overlapping amino acids required for transactivation. The activity of
this region was further substantiated by linking specific peptides
encompassing the potential
-helix to the Gal4DBD. These chimeric
proteins exhibited constitutive transcriptional activity at least
25-fold higher than Gal4DBD either alone or linked to 40 a.a of
rtER N-terminal lacking the potential structure defined above.
Additionally, since introduction of prolines in this sequence abrogated
transcriptional activation, this reinforced the idea that the presence
of an
-helical structure is required for transcriptional
activation. Thus, alone, this
-helix can transactivate, as has been
demonstrated for VP16-derived or totally artificial acidic structures
(52, 53). A comparative analysis of the two receptor HCA plots (Figs. 4
and 5
) clearly showed several conserved hydrophobic clusters, not at
the sequence level, but in the general organization of the two
N-terminal domains. Similar results were obtained with ER
from other
species such as salmon, tilapia, Xenopus, rat, or mouse
(data not shown). Importantly, a hER N-terminal HCA plot also revealed
an
-helix located at the beginning of the B domain, between
residues 3944 (hER
-H3944). This analysis indicated that the
presence of this structure was conserved among the different ER
cloned. More precisely, the sequence of the potential rtER
-helix
corresponding to the minimal region essential for the rtER AF-1 was
conserved during evolution in fish species, and an
-helical
structure lies in a similar position at the beginning of the ER
B
domain in all tetrapod species. This structure seems to be specific for
ER
AF-1 since ERß shows a different organization. It is also worth
noting that neither the fish nor other species sequences corresponding
to the
-helix were found in other NRs (data not shown).
Interestingly, Metzger et al. (21) characterized a.a 29 to
63 of hER, a region encompassing the hER
-H3944 as an activating
domain in yeast. By linking the minimal region required to form an
-helix in this sequence to the Gal4DBD, we demonstrated that this
structure could activate the transcription in an autonomous manner.
Moreover, the transcriptional activity was dependent upon the helical
structure, since substitution of the key hydrophobic amino acids into
proline eliminated transactivation potency. The importance of this
potential
-helix was also tested in the full-length hER context,
using either truncated (hER
144) or point mutant
(hER
137L39P and
hER
137Y43P) receptors. Compared with the A
domain truncated receptor (hER
137), this led to a 50% decrease of
E2-induced activity, and approximately 80% of
E2-independent reporter activation. Therefore,
this conserved structure seems to be, at least in part, implicated in
hER AF-1 in yeast. The suggestion of two structures in ER AF-1 was also
demonstrated in mammalian cells by transient transfection experiments.
Deletion or point mutations either in the rtER or hER
-helices
abrogated, at least partially, their AF-1 activity in AF-1 sensitive
cell contexts, whereas it had no effect in AF-2-sensitive cells such as
HeLa. All these results, plus the finding of potential structures in
similar positions in two far-distant ERs, led us to propose that this
-helix is the functioning core of ER AF-1. This hypothesis was
further substantiated by the use of rainbow trout and human receptors
in which their
-helices were transposed. Indeed, such chimeric
receptors were as active as their corresponding wild-type forms, both
in yeast and in mammalian cells. The two point mutant hER receptors
also particularly enlightened the link between ER ligand-independent
activity and AF-1, since mutation of a sequence implicated in AF-1
dramatically reduces E2-independent activity.
Thus, this helical structure seems to be more involved in
E2-independent transactivation than
E2-dependent activity.
A study on the VP16 protein found that an
-helical structure could
be formed from a random coil organization only upon interaction with a
component of the basal transcription machinery,
hTAFII 31 (54). A related mechanism was very
recently observed in the context of a member of the NR
superfamily, the glucocorticoid receptor (GR or NR3C1). Indeed, in that
study the authors showed that the GR N-terminal domain, which appears
to have little intrinsic spatial organization, acquires helicity and
tertiary structures upon binding to DNA when linked to the GR DBD (55).
This kind of interdomain signaling mechanism could also be considered
with the ER
-helix. In fact, one could speculate that the rtER and
hER
-helical structures could appear only in certain
conditions.
Recent findings showed that coactivators such as SRC-1 (32), GRIP1,
RAC3, pCAF (33), and p300/CBP (32, 34) could enhance the AF-1 of ER as
well as that of other NRs. These coactivators are potential linking
factors as they were first identified to interact with the NR
C-terminal region through their signal motif NR box LXXLL (56)
exhibiting an
-helical structure. Moreover, these partners could
physically interact with the N-terminal region of steroid receptors.
The most important region defined in hER for these interactions was Box
1 (33), a region encompassing the
-helix that we have identified.
The direct implication of this structure in the interaction with these
coactivators is one of the next questions that we must answer. In
respect to the comparison with yeast and mammalian cell contexts, this
could lead to an interesting fact: yeast has been shown to possess no
homologs to NR box coactivators. Then, if SRC-1 or another coactivator
of this family is able to interact with these structures, it could
imply that yeast possess functional homologs of these coactivators,
which remain to be determined. It should also be remembered that
physical interaction between these proteins and the ER N terminus lies
on regions other than the NR boxes (33). Studies on VP16 have
implicated
-helical structures in transcriptional activation (57, 58) and shown that targets of a VP16-derived
-helix could be TFIIB
(59). Consequently, interaction of hER
-3944 with components of the
basal transcriptional machinery such as TFIID, TFIIH, or TFIIB should
be tested since these factors are also able to interact with NRs
(60, 61, 62).
Studies on hER AF-1 have shown that in yeast, two regions (a.a 162
and a.a 118149) are able to efficiently activate transcription in an
autonomous manner, while a third one between a.a 80 and a.a 113 is
involved with the other two in synergism between AFs (19). However,
rtER constructs containing these regions, namely the
rtERS
117 (corresponding to a.a 4562 in
rtERL), rtERS
134 (a.a
4579), and rtERS
150 (a.a 4595) did not
possess any activation potency in yeast, when expressed alone or fused
to the Gal4DBD. This may suggest, that in contrast to the hER
-H3944, the rtER
-H1422 is absolutely required in yeast for a
general conformation and possibly suitable folding of the other regions
implicated in AF-1. Alternatively, one could speculate that during
evolution, the ER N-terminal transactivation function has evolved in
multiple subdomains to provide several surfaces for the recruitment of
evolutionary novel coactivator or adaptor proteins. However, our
transfection experiments in mammalian cells have also shown that other
regions are required for AF-1 and/or total activity of the rainbow
trout receptor. Indeed, deletion or point mutations within the
potential structure defined by HCA plot only reduced the fold induction
by 2- to 3-fold depending upon the AF-1 sensitive cell-context. This
fits well with earlier experiments defining three sequences for AF-1 in
mammalian cells (19). Other studies on ER AF-1 have shown that a major
regulation of this transactivation function in mammalian cells proceeds
through phosphorylation of key serine residues, namely the
S118, S167,
S154, S104,and
S106 one (63, 64, 65, 66, 67, 68), linking the AF-1 activity with
other signaling pathways (69). Since the rtER N-terminal sequence
exhibits some serine residues, and for some, in comparable positions to
those in hER, the contribution of these residues to rtER AF-1 could be
tested. The p68 helicase protein was recently identified as a specific
AF-1 coactivator (70) and was shown by pull-down assays to associate
with hER in a region encompassing Ser118,
providing a link between phosphorylation and transactivation. Testing
the interaction of this protein with rtER should provide information
about the importance of some of these residues for AF-1
functioning.
Deletion of the A domain in hER (hER
137) did not modify
ligand-dependent transcriptional activity of the receptor, confirming
that sequences implicated in AF-1 are located in the ER B domain.
Unexpectedly, this deletion markedly increased the ligand-independent
activity of the receptor. Comparison with a new rtER isoform possessing
an A domain (rtERL), which does not exhibit any
ligand-independent activity (Ref. 38 , and see Fig. 1
) led to the
hypothesis that the A domain of ERs exerts an inhibitory role on AF-1.
The existence of an inhibitory function, repressing both AF-1 and AF-2,
most probably by an intramolecular process, was recently reported for
the progesterone receptor (71). Deletion of rtER C-terminal (rtAD
construct) led to a substantial increase in the ligand-independent
activity of the receptor (75% of the full-length ER activity),
confirming that AF-1 is markedly reduced by the presence of the
C-terminal domain. This observation might indicate a functional
interaction between the ligand-binding and N-terminal domains of ER.
Such an interaction was suggested for ER (72), and demonstrated for the
androgen and progesterone receptors by pull-down and two-hybrid assays,
but this was related to the synergistic effect between the two AFs
contained in these regions (73, 74). Here, we demonstrate that the ER A
domain specifically interacts both in vitro and in
vivo with the C-terminal region of the receptor, masking
hormone-independent activity of the full-length receptor. This
interaction was conserved during evolution since both rtER and hER
displayed this property, and since the A domain of one receptor could
interact with the C-terminal region of the other. This highlights the
importance of such a mechanism for the ER overall activity and suggests
that these interactions rely on conserved residues or structures in the
C-terminal domain of ERs. These regions need now to be identified.
However, it is of importance to note that this kind of interaction,
inhibited by the ligand, which is not required for DNA binding, could
absolutely not define a dimerization process. In the absence of ligand,
the A domain could therefore function as a negative regulatory domain
for ER AF-1 activity by a direct interaction with the C-terminal
region. Stoichiometry of the interaction remains also to be determined.
Indeed, since ER dimerizes through its DBD in the absence of
E2, it is not clear whether the A domain of one
monomer interacts with the C-terminal region of the same monomer or the
other dimer partner. However, completion of these two kind of
interactions in the absence of ligand led to an inactive conformation
of the receptor (Fig. 12
, left, to simplify, hypothesis of an intramolecular process
is shown). However, one cannot exclude indirect mechanisms involving
eventual cofactors: 1) the A domain may enhance direct interaction
between ER and a corepressor such as SMRT (silencing mediator for
retinoid and thyroid hormone receptors) or NCoR (nuclear
receptor corepressor) (75, 76) (Fig. 12
, middle); 2) the A
domain could act via a steric prevention of AF-1 coactivator
recruitment (Fig. 12
, right). It was also suggested that a
synergistic effect between AF-1 and AF-2 of NRs may be a consequence of
direct interactions of these activation domains with a common
coactivator, as shown for AR with the CREB-binding protein
complex and for GR and ER with DRIP150 and
DRIP250 (73, 77, 78). The A domain may prevent
interaction of the AFs with such coactivators in the absence of ligand.
An alternative hypothesis relates to a corepressor pathway. A recent
study described the AF-1 regulating role of the yeast Hsp40Ydj1 protein
involved in the maturation of heat shock proteins (Hsp) (79). The Hsp,
and especially the Hsp90 chaperone pathway, is thought to inhibit
steroid receptor transcriptional activities in the absence of ligand
(80, 81). Since Ydj1 seems to be involved in an N-terminal/C-terminal
interaction resulting in the inhibition of AF-1 in the absence of
E2 (79), it would be interesting to investigate
whether the A domain is involved in contacts between ER and Ydj1.

View larger version (12K):
[in this window]
[in a new window]
|
Figure 12. Models for the Repressive Action of the ERs A
Domain
Hypothetical mechanism by which the A domain exerts an inhibitory
effect on AF-1 activity. The A domain could physically repress the
AF-1, in cooperation with the C-terminal region (left
side of the panel). An inactive conformation induced by binding
of the A domain to the C-terminal region may recruit a repressor,
inhibiting the ligand-independent activity of the receptor
(middle). Finally, this conformation may prevent the
receptor from interacting with a specific activator (on the
right side).
|
|
In all these hypotheses, ligand has to induce a conformational change
permitting exposure of the AFs (6, 7, 44). Interestingly, our results
showed that the interaction between the A domain and the C-terminal
region of the ER is reduced in the presence of ligand. More precisely,
this reduction was observed in the presence of total or partial
agonists (E2 and OHT, respectively) of the ER,
but not in the presence of the pure antagonist
ICI164,384. This is highly relevant in the view
of a key role of the A domain in the regulation of AF-1 since activity
of OHT is thought to proceed through AF-1. On the other hand,
ICI164,384, which is known to behave as an ER
total antagonist, did not affect the A domain/C-terminal region
interaction, suggesting an inactive conformation of the ER bound to
this antiestrogen. Since ICI induces a C-terminal structural
organization distinct from E2, one could
speculate that ICI may not be able to derepress the AF-1 function.
Thus, this could be another explanation for the potency of ICI as an
estrogen antagonist. A previous report suggested that interaction
between whole N-terminal and C-terminal regions of ER occurs only after
ligand binding, allowing synergism between the AFs leading to ER full
activity (72). Since this study was performed using transfection
experiments with reporter gene and both AF-containing regions of ER as
separate polypeptides, this approach could not detect interactions that
are transcriptionally unproductive. Therefore, our data are not
inconsistent with the study of Kraus et al. (72), since the
interaction of the C-terminal region with A domain does not result in
transcriptional activation. However, the way to reconcile a physical
interaction enhanced by ER agonists, at least resulting in
transcriptional activation, with an inter