Molecular Endocrinology 14 (10): 1627-1648
Copyright © 2000 by The Endocrine Society
The LIM/Homeodomain Protein Islet-1 Modulates Estrogen Receptor Functions
Frédérique Gay,
Isabelle Anglade,
Zhiyuan Gong and
Gilles Salbert
Equipe Information et Programmation Cellulaire (F.G., G.S.) and
Equipe Endocrinologie Moléculaire de la Reproduction (I.A.)
UMR 6026 Centre Nationale de la Recherche Scientifique
Université de Rennes I, Campus de Beaulieu 35042 Rennes
cedex, France
Department of Biological Sciences (Z.G.)
National University of Singapore Singapore 119260
 |
ABSTRACT
|
|---|
LIM/Homeodomain (HD) proteins are
classically considered as major transcriptional regulators which, in
cooperation with other transcription factors, play critical roles in
the developing nervous system. Among LIM/HD proteins, Islet-1 (ISL1) is
the earliest known marker of motoneuron differentiation and has been
extensively studied in this context. However, ISL1 expression is not
restricted to developing motoneurons. In both embryonic and adult
central nervous system of rodent and fish, ISL1 is found in discrete
brain areas known to express the estrogen receptor (ER). These
observations led us to postulate the possible involvement of ISL1 in
the control of brain functions by steroid hormones. Dual
immunohistochemistry for ISL1 and ER provided evidence for ISL1-ER
coexpression by the same neuronal subpopulation within the rat
hypothalamic arcuate nucleus. The relationship between ER and ISL1 was
further analyzed at the molecular level and we could show that 1) ISL1
directly interacts in vivo and in vitro with
the rat ER, as well as with various other nuclear receptors; 2) ISL1-ER
interaction is mediated, at least in part, by the ligand binding domain
of ER and is significantly strengthened by estradiol; 3) as a
consequence, ISL1 prevents ER dimerization in solution, thus leading to
a strong and specific inhibition of ER DNA binding activity; 4) ISL1,
via its N-terminal LIM domains, specifically inhibits the ER-driven
transcriptional activation in some promoter contexts, while ER can
serve as a coactivator for ISL1 in other promoter contexts. Taken
together, these data suggest that ISL1-ER cross-talk could
differentially regulate the expression of ER and ISL1 target genes.
|
INTRODUCTION
|
|---|
The LIM/HD proteins are a subset of homeodomain (HD)-containing
transcription factors defined on the basis of a common LIM domain,
which consists of a conserved cysteine- and histidine-rich structure of
two tandemly repeated zinc fingers. The acronym of LIM is derived from
the first identified members of this family, namely Lin-11 from
Caenorhabditis elegans (1), Islet-1 from rat (2), and Mec-3
from C. elegans (3). Up to date, LIM/HD transcription
factors have been extensively studied for their critical roles during
development in invertebrates and vertebrates. Most, if not all, the
LIM/HD-encoding genes are expressed in the developing nervous system
and have been implicated, in cooperation with other factors, in
regulating crucial events during nervous system development such as
neuronal survival, neuronal fate specification, axonal pathway and
target selection, as well as neurotransmitter identity (for review, see
Ref. 4).
Among the LIM/HD transcription factors, Islet-1 (ISL1) has been
extensively studied for its regulatory role in motoneuron
differentiation. Indeed, in chick embryo, ISL1 is expressed in
motoneurons of the embryonic spinal cord soon after their final
mitosis, even before the appearance of other differentiated properties
(5, 6). Similarly, in zebrafish embryo, this LIM/HD protein expression
is initiated in many primary neurons in the brain and spinal cord,
including primary motoneurons, at the end of the gastrulation and
precedes that of any other known neural marker (7, 8). These results
therefore established ISL1 as the earliest marker of developing
motoneurons. Subsequent studies of LIM/HD proteins in embryonic
motoneurons have led to the identification of a combinatorial code of
several LIM/HD proteins, including ISL1, that controls various aspects
of motoneuron identity (6, 9, 10, 11, 12, 13).
Aside from its expression in embryonic motoneurons, ISL1 is broadly
expressed in other cell types during embryogenesis, but also in adult.
Consistent with its identification as an insulin enhancer binding
protein (2), ISL1 is expressed in a variety of cell lines of pancreatic
endocrine origin, as well as in normal adult rat islet cells (2).
However, ISL1 expression is not restricted to pancreatic cells since
this protein is widely distributed in endocrine cells of the thyroid
and the anterior lobe of the pituitary (14, 15). In addition, ISL1 is
also present in many neurons within the central nervous system of the
embryonic and adult rat (15). In this last case, ISL1 is expressed in
neurons of the caudate-putamen, the septal nuclei, and in the nuclei of
the basal telencephalon. Finally, ISL1 is expressed in a subset of
neurons of the diencephalon, a particularly high expression being
detected in the pineal gland, in the thalamic reticular nucleus, and in
neurons of the hypothalamus. The hypothalamic arcuate nucleus exhibits
a particularly dense population of ISL1-expressing neurons (15). A
similar expression pattern of ISL1 can be observed in salmonids showing
that, in both rodent and fish central nervous systems, ISL1 expression
maps brain nuclei involved in the control of the endocrine system (15, 16), the expression pattern observed in embryo being similar to that
observed in adult. These data could suggest the involvement of ISL1 in
both the development and/or the maintenance of the differentiated
phenotypes of the ISL1-expressing neurons. In this respect, ISL1 may
function in a similar way to Mec-3, a LIM/HD protein identified in
C. elegans. This transcription factor is expressed in early
embryo and throughout adulthood and controls both development of
mechanoreceptor sensory cells and maintenance of their differentiated
properties (3, 17). Previous studies have also reported the involvement
of LIM/HD proteins in both acquisition and maintenance of appropriate
neurotransmitter secretory phenotypes in distinct subsets of neurons. A
role in neuronal identity specification has indeed been demonstrated
for the Drosophila Islet protein, which is expressed in
discrete subsets of interneurons including dopaminergic and
serotoninergic neurons (18). Loss of Islet function causes dramatic
defects in both axon pathfinding and targeting as well as loss of
dopamine and serotonin synthesis, showing that in Drosophila
embryo, Islet controls two distinct aspects of neuronal identity
including secretion of the appropriate neurotransmitter (18).
Similarly, the Drosophila Apterous LIM/HD protein is
involved in a direct and cell-specific regulation of the
Drosophila FMRFa neuropeptide gene and thereby contributes
to both initiation and maintenance of FMRFa expression in a subset of
neurons (19). Nevertheless, the ISL1-expressing neurons identified in
the rat central nervous system do not share any obvious neuropeptide
secretory phenotype (15). Thus, ISL1 is unlikely to control
neurotransmitter identity of these neurons. However, particularly
striking are the parallel expression patterns of ISL1 and of the
estrogen receptor 
(referred below as ER) in the brain of both
rodent and fish (15, 16, 20, 21, 22, 23, 24, 25).
The evolutionary conservation of the highly parallel expression
patterns of these two transcription factors in restricted brain areas
led us to investigate the functional consequences of ISL1-ER
coexpression. In this report, we give evidence for a complex and
promoter-specific cross-talk between these two transcription factors.
Such a transcriptional interference is likely to play a major role in
the modulation of estrogen signaling in neurons.
|
RESULTS
|
|---|
The Nuclear ER and the LIM/HD Protein ISL1 Colocalize in the Rat
Hypothalamic Arcuate Nucleus
In the rat central nervous system, ISL1 has been previously shown
to be expressed in brain nuclei controlling the endocrine system (15).
Such a distribution is reminiscent of ER-containing neurons that are
also found in nuclei involved in the control of the endocrine system
(21). One of these, the arcuate nucleus, is densely populated by both
ER and ISL1 expressing neurons (15, 21). We thus performed
immunofluorescence studies to verify whether ER and ISL1 were
effectively coexpressed in the same neurons of the arcuate nucleus
(Fig. 1
). We found that, in this nucleus,
with very few exception (Fig. 1E, open arrowheads), all
ER-positive (ER+) cells were also positive for
ISL1 (Fig. 1, A and B), whereas some ISL1+ cells
were not revealed by the ER-specific antibodies (Fig. 1, C and D,
white arrowheads). This was also true for
ER+ cells of the ventromedial hypothalamic
nucleus since they also expressed ISL1 (data not shown). These results
imply that, at least in the ventral hypothalamus, ER expression can be
systematically associated with ISL1 expression. Such an extensive
colocalization was not due to an interference between the detection
systems since control sections that did not receive the biotinylated
goat-antimouse immunoglobulins (see Materials and Methods)
did not display any green fluorescence (data not shown).
View larger version (15K):
[in this window]
[in a new window]
|
Figure 1. ISL1 and ER Colocalize in the Rat
Hypothalamic Arcuate Nucleus
Dual immunofluorescence studies were run on rat brain cryostat
sections. Both ER (red) and ISL1 (green)
showed a strictly nuclear localization and were coexpressed in most
cells of the arcuate nucleus. Panels A/B, C/D, and E/F show different
regions of the arcuate nucleus (magnification: x180). The white
arrowheads (D and F) point out ISL1-positive cells that do
not express ER (C and E), while open arrowheads show
ER-positive cells lacking ISL1 immunoreactivity (E).
|
|
ISL1 and ER Interact in Vivo
The functional relevance of the ISL1-ER colocalization in the rat
central nervous system was assessed by in vivo
coimmunoprecipitation studies, performed with homogenate from female
rat hypothalamus, known to express both transcription factors. Animals
were injected with 17ß-estradiol (E2) 1 h
before the protein extract was prepared, half of the extract being
incubated with a rabbit polyclonal antibody (ER715), directed against
the hinge region of the rat ER. The specificity of the interaction was
assessed by incubation of the remaining extract with rabbit preimmune
serum. Immunoprecipitation was followed by immunoblotting with the
mouse monoclonal anti-ISL1 antibody, and the results are presented in
Fig. 2. As shown in lane 2, a single band
of the expected molecular mass (39 kDa) was detected with the
anti-ISL1 antibody when proteins extracted from rat hypothalamus had
been incubated with the anti-ER antibody. This band was moreover
specific since it was not present when the protein extract had been
incubated with preimmune serum (lane 1). Thus, ER can specifically
interact with ISL1 in vivo. Moreover, the fact that ISL1 can
be immunoprecipitated with ER from rat hypothalamus extracts strongly
suggests that the striking colocalization between ISL1 and ER observed
in some rat hypothalamus nuclei could be physiologically relevant.
View larger version (16K):
[in this window]
[in a new window]
|
Figure 2. ISL1 and ER Interact in Vivo
Protein extracts from female rat hypothalamus were incubated with
rabbit preimmune serum (lane 1) or anti-ER antibodies (lane 2), as
indicated at the top of the autoradiogram.
Immunocomplexes were collected, washed, and resolved by SDS-PAGE
analysis. After Western blotting, the membrane was probed with
anti-ISL1 antibodies, and detection of the immunocomplexes was
performed with a chemiluminescent detection kit. The molecular mass
markers (in kilodaltons) are shown on the left.
|
|
ISL1 Interacts with a Subset of Nuclear Receptors in
Vitro
We next wished to find out whether the ISL1-ER interaction
observed in vivo could also occur in vitro. We
therefore used the matrix-bound fusion protein of
glutathione-S-transferase (GST) combined with ISL1
(GST-ISL1) for in vitro protein-protein interaction studies
to test various nuclear proteins for their ability to interact with the
zfISL1 protein in vitro. The majority of the
35S-methionine-labeled in
vitro-translated proteins tested interacted with GST-ISL1 (Fig. 3), as shown by the bands observed when
these 35S-methionine-labeled proteins were
incubated with GST-ISL1. It was the case for the wild-type rat ER (rER;
Fig. 3A, lane 9), the truncated ER protein, corresponding only to the
ligand-binding domain (LBD) of the rER (TERP; Fig. 3A, lane 12), the
rainbow trout ER (rtER; Fig. 3B, lane 3), the human nuclear orphan
receptor, chicken ovalbumin upstream promoter (COUP)-TFI (hCOUP; Fig. 3B, lane 6), as well as for the zebrafish steroidogenic factor 1
(zfSF1; Fig. 3B, lane 9). In contrast, the Xenopus Cyclin E
was not or very weakly retained on the GST-ISL1 bound-beads (xCyclin;
Fig. 3B, lane 12). The GST-ISL1 fusion protein did not significantly
retain the luciferase control protein (Luc; Fig. 3A, lane 6). Moreover,
the protein-protein interactions involving ISL1 and the nuclear
receptors were specific since the GST moiety of the fusion protein did
not interact with any of the
35S-methionine-labeled in
vitro-translated proteins. These gels were submitted to
phosphoimage screen, and the relative amounts of
35S-methionine-labeled in
vitro-translated proteins retained on the GST-ISL1 bound-beads
were assessed by comparing the intensity of the corresponding band to
that of the input lane. This analysis showed that 7% of the input was
retained on the GST-ISL1-bound beads for the zfSF1-programmed lysate,
while 5.71% of the input was retained for the in
vitro-translated rER. The percentage of the input retained was
quite similar for TERP, rtER, and hCOUP-TFI, ranging from 3% to 4%.
Finally, only 0.8% and 0.4%, respectively, of the input was retained
for the xCyclin E and luciferase proteins. Taking the specific activity
of the labeled proteins into account, we thus could determine that,
among the tested proteins, ISL1 presented a similar and maximal
relative affinity for zfSF1 and hCOUP-TFI, its relative affinity for
TERP, rER and rtER being 1.3- to 2.3-fold weaker. Finally, ISL1 showed
only weak or negligible affinity for xCyclin E and luciferase. As ISL1
seemed to associate directly and specifically with various nuclear
receptors including ER, we next studied the potential effect of the
ligand on ISL1-rER interaction. In order to reach this goal, we
performed GST pull-down assays with rER-programmed lysate in the
absence (Fig. 3C, lanes 1–3) or in the presence of 17ß-estradiol
(Fig. 3C, lanes 4–6). The amount of rER protein retained was
significantly higher in the presence of ligand. As confirmed by
phosphoimage screen, a 2-fold induction in the amount of
35S-methionine-labeled rER bound by GST-ISL1 in
the presence of ligand was observed compared with that retained in the
absence of ligand (data not shown). Thus, ISL1 binding to rER is
affected by the presence of ligand in an
E2-stimulated manner and is likely to involve the
C-terminal LBD of the nuclear receptor since ISL1 also interacts
in vitro with the natural truncated rER variant, TERP.
View larger version (32K):
[in this window]
[in a new window]
|
Figure 3. ISL1 Interacts with Various Nuclear Receptors
in Vitro
Panels A and B show SDS-PAGE analysis of the
35S-methionine-labeled in vitro-translated
proteins retained by the GST-ISL1 fusion protein. Glutathione-agarose
beads containing GST (panels A and B, lanes 2, 5, 8, and 11) or
GST-ISL1 (panels A and B, lanes 3, 6, 9, and 12) proteins were
incubated with various 35S-methionine-labeled in
vitro-translated proteins indicated at the
bottom of the gels. Panel C shows SDS-PAGE analysis of
the 35S-methionine-labeled in
vitro-translated rER retained by the GST-ISL1 fusion protein,
in the absence or presence of ligand. The rER-programmed lysate was
treated in the absence (panel C, lanes 1–3) or in the presence (panel
C, lanes 4–6) of E2 before incubation with
glutathione-agarose beads containing GST (panel C, lanes 2 and 5) or
GST-ISL1 (panel C, lanes 3 and 6). For each lysate, the input lane
corresponds to the amount of 35S-methionine-labeled
in vitro-translated proteins used for the GST pull-down
assay.
|
|
ISL1-ER Interaction Inhibits ER Dimerization in
Vitro
Since ISL1 and ER interacted in vivo and in
vitro, it was of interest to examine a potential effect of this
protein-protein interaction on ER dimerization function. GST pull-down
competition assays were set up to analyze rER binding to a GST-hER LBD
fusion protein in the absence or in the presence of zfISL1 (Fig. 4). These competition assays were
performed in the presence of limiting amounts of
E2-treated rER-programmed lysate (1.5 µl) and
GST-hER LBD fusion protein (500 ng), as determined in preliminary
experiments. Increasing amounts (1.5–15 µl) of unlabeled
zfISL1-programmed lysate were added, such that ISL1 was present from a
3- to a 30-fold molar excess, as assessed by in vitro
translations of rER and zfISL1, performed in the presence of
35S-methionine and taking the specific activity
of both 35S-methionine-labeled proteins into
account. In the absence of competitor, rER was able to interact with
the LBD of hER, as shown by the band observed when
35S-methionine-labeled rER was incubated with the
GST-hER LBD fusion protein (Fig. 4, lane 3). The relatively low
intensity of this band could probably reflect the difficulty for hER
LBD to destabilize the dimers formed in solution by the
E2-treated in vitro-translated rER.
Even though a significant amount of rER interacted with the
GST-hER LBD-bound beads in the absence of zfISL1, this protein-protein
interaction was moreover specific since the GST moiety did not
retain the 35S-methionine-labeled rER (Fig. 4, lane 2). Addition of increasing amounts of in
vitro-translated zfISL1 led to a gradual decrease in the retention
of 35S-methionine-labeled rER by GST-hER LBD
(Fig. 4, lanes 4–7), reflecting a competition between rER and zfISL1
for binding to the GST-hER LBD fusion protein. Competition assays
performed in the absence of E2 gave similar
results (data not shown). Finally, in contrast to the marked inhibition
of rER binding to the fusion protein observed in the presence of the
highest dose of cold ISL1-programmed lysate, addition of an equivalent
dose of unprogrammed lysate did not modify the amount of rER-programmed
lysate retained on the GST-hER LBD-bound beads (Fig. 4, lane 8),
showing that the inhibitory effect observed previously was specific for
ISL1. These data suggest that ISL1 binding to the ER precludes ER dimer
formation, thus giving evidence for a specific and dose-dependent
inhibition of ER dimerization function by ISL1 in vitro.
View larger version (53K):
[in this window]
[in a new window]
|
Figure 4. ISL1-ER Interaction Inhibits rER Dimerization
SDS-PAGE analysis of the 35S-methionine-labeled in
vitro-translated rER retained by the GST-hER LBD fusion protein
in the absence or presence of unlabeled ISL1-programmed competitor
lysate. The rER-programmed lysate (1.5 µl) was incubated with
limiting amounts (500 ng) of glutathione-agarose beads containing
GST (lane 2) or GST-hER LBD (lanes 3–8) proteins, as indicated at the
bottom of the gel. Both rER-programmed lysate and
GST-hER LBD fusion protein were treated by E2 before
incubation. The input lane (lane 1) corresponds to 20% of the total
amount of 35S-methionine-labeled in
vitro-translated rER used for the GST pull-down competition
assay. The amount of labeled in vitro-translated rER
retained by the GST-hER LBD protein in the absence of any competitor
lysate is shown in lane 3. Increasing amounts (1.5–15 µl) of
ISL1-programmed lysate were added, such that ISL1 was present in a
molar excess ranging 3- to 30-fold, as indicated at the
top of the gel (lanes 4–7). Unprogrammed competitor
lysate (15 µl) was used as a control (lane 8).
|
|
ISL1-ER Interaction Specifically Decreases rER Binding to DNA
Since ISL1 destabilized ER dimers in solution, we next wished to
determine whether this inhibitory effect could occur in the presence of
ER DNA target sequence and affect the binding of the receptor to its
recognition site. We performed electrophoretic mobility shift assay
(EMSA) experiments to analyze the effect of addition of increasing
amounts of GST-ISL1 fusion protein on the rER binding to a
32P-labeled estrogen responsive element (ERE)
consensus sequence. Preliminary experiments using GST and GST-ISL1
proteins extracted from total insoluble bacterial lysates revealed that
increasing amounts of unpurified GST-ISL1 fusion protein failed to
significantly affect the DNA binding activity of the in
vitro translated rER. Since the GST-ISL1 was highly unstable and
only represented a small proportion of the total insoluble proteins, we
thus repeated these EMSA experiments using increasing amounts of fusion
proteins enriched from the insoluble fraction by affinity
chromatography with glutathione-agarose beads. As shown in Fig. 5A, incubation of rER-programmed lysate
with labeled double-stranded EREc in the absence of fusion
protein (lanes 2 and 4) led to the formation of a preponderant shifted
ER-ERE complex (indicated by an arrowhead). The major
retarded complex reflected the specific binding of rER to the EREc
probe since no similar shifted band was observed when unprogrammed
lysate was incubated with the probe under the same conditions (Fig. 5A, lanes 1 and 3). The intensity of the shifted band corresponding to the
specific ER-ERE complex was not affected when rER-programmed lysate was
coincubated with 450 ng of purified GST before addition of the probe
(Fig. 5A, lane 6), indicating that the GST moiety of the GST-ISL1
fusion protein did not affect the ER binding to its DNA recognition
site. Neither the GST moiety (Fig. 5A, lane 5) nor the GST-ISL1 fusion
protein (Fig. 5A, lanes 7, 9, and 11) alone interacted with the labeled
double-stranded ERE oligonucleotide. Interestingly, addition of
increasing amounts (150 ng, 300 ng, and 450 ng) of purified GST-ISL1 in
the presence of receptor (Fig. 5A, lanes 8, 10, and 12, respectively)
resulted in a progressive decrease in the ER-ERE complex intensity. In
contrast, the intensity of the nonspecific bands was modified neither
by GST nor by GST-ISL1. Similar gel shift experiments were
performed with rER-programmed lysate treated with 17ß-estradiol,
but no effect of the ligand was observed (data not shown). Thus,
ISL1-ER interaction inhibits the specific ER-DNA interaction in a
dose-dependent manner, both in the absence and in the presence of
ligand. Requirement of relatively high amounts of GST-ISL1 protein to
significantly inhibit rER binding to DNA could be related to a very
weak stability of the fusion protein (data not shown) and/or to a
partial renaturation of the GST-ISL1 protein during dialysis (see
Materials and Methods). Since ISL1 also interacted with
hCOUP-TFI in vitro, we next tested the potential effect of
increasing amounts of purified GST-ISL1 on hCOUP-TFI binding to DNA.
Previous work has shown that the orphan receptor hCOUP-TFI is able to
bind a DR24-type element within the rtER gene promoter (26). EMSA
experiments using a 32P-labeled double-stranded
DR24 oligonucleotide as a probe were performed rigorously in the same
conditions as those mentioned above (Fig. 5B). Incubation of
hCOUP-TFI-programmed lysate with labeled probe in the absence of fusion
protein (Fig. 5B, lanes 2 and 4) led to the formation of a major
hCOUP-TFI-DR24 shifted complex (indicated by an arrowhead),
giving evidence of the specific binding of hCOUP-TFI to the DR24 type
element since no similar shifted band was observed when unprogrammed
lysate was incubated with the probe under the same conditions (Fig. 5B, lanes 1 and 3). Coincubation of the hCOUP-TFI programmed lysate with
450 ng of purified GST protein did not modify the intensity of the
shifted band (Fig. 5B, lane 6). Similarly, coincubation of hCOUP-TFI
with increasing amounts of purified GST-ISL1 fusion proteins (Fig. 5B, lanes 8, 10, and 12) did not significantly modify the shifted band
intensity. Thus, although ISL1 interacts in vitro with both
the rER and the human nuclear orphan receptor COUP-TFI, these
protein-protein interactions differentially modulate DNA binding by the
nuclear receptors.
View larger version (70K):
[in this window]
[in a new window]
|
Figure 5. ISL1-ER Interaction Specifically Decreases rER
Binding to DNA
Effects of ISL1-ER interaction on the rER binding to an EREc probe
(panel A) and effects of ISL1-COUP-TFI interaction on the hCOUP-TFI
binding to a DR24 probe (panel B) were assessed by EMSA analysis.
Unprogrammed (lanes 1, 3, 5, 7, 9, and 11), ER-programmed lysate (panel
A, lanes 2, 4, 6, 8, 10, and 12) or hCOUP-TFI-programmed lysate (panel
B, lanes 2, 4, 6, 8, 10, and 12) were coincubated with either 450 ng of
purified GST (lanes 5–6) or 150 ng, 300 ng, and 450 ng of purified
GST-ISL1 (lanes 7 and 8, 9 and 10, and 11 and 12, respectively) before
addition of the 32P-labeled double-stranded oligonucleotide
indicated at the bottom of the gel. Incubations of
unprogrammed (lanes 1 and 3), ER-programmed lysate (panel A, lanes 2
and 4) or hCOUP-TFI-programmed lysate (panel B, lanes 2 and 4) alone
with the labeled probe were performed either in binding buffer alone
(lanes 1 and 2) or in the same buffer conditions (i.e.
binding buffer plus dialysis buffer) as in lanes 5–12 (lanes 3 and 4).
The position of the main ER-ERE and COUP-TFI-DR24 complexes is
indicated by an arrowhead while nonspecific bands are
indicated by asterisks.
|
|
ISL1 Inhibits rER Transactivation Function in a Dose-Dependent
Manner
The functional relevance of the ISL1-ER interaction was tested in
transient transfections of Chinese hamster ovary (CHO) cells, using an
artificial estrogen-responsive reporter gene ERE-SV-Luciferase (Fig. 6, A and B). This reporter gene contains
the FP1 sequence of the rtER gene (bp -85 to -56), which encompasses
an estrogen-responsive element (27, 28). Cells were cotransfected with
or without rER, zfISL1, and zfISL2 expression vectors and were either
untreated or treated with E2 after transfection.
Cotransfection of the ERE-SV-Luc reporter construct with rER expression
vector alone only led to a slight increase in the basal luciferase
activity but allowed induction in the luciferase activity when
E2 was present (Fig. 6A). In contrast,
cotransfection with zfISL1 expression vector alone affected neither the
basal nor the E2-induced transcriptional activity
of the ERE-SV-Luc reporter construct (data not shown). When cells were
cotransfected with rER expression vector and increasing amounts of
zfISL1 expression vector, no significant modification of the basal
luciferase activity was observed, while ISL1 led to a gradual decrease
in the E2-induced luciferase activity (Fig. 6A).
Even the lowest dose (5 ng per well) of zfISL1 expression vector led to
a 35% inhibition of the ligand-dependent activity of the reporter gene
(Fig. 6, A and B). Finally, the transcriptional induction by
E2 was almost completely abolished in the
presence of 100 ng of zfISL1 expression vector (Fig. 6, A and B). These
data are consistent with the previous EMSA results and show that ISL1
is able to repress the ligand-dependent transcriptional activation
mediated by rER on an artificial estrogen-responsive promoter. In
contrast, when cells were cotransfected with rER expression vector and
increasing amounts (5–100 ng per well) of zfISL2 expression vector, no
such inhibition of the estrogen-dependent reporter gene activation was
observed (Fig. 6A). These results clearly demonstrate that inhibition
of rER transactivation function is specific for ISL1 since the related
LIM/HD protein ISL2 had no effect on the ligand responsiveness of the
reporter gene. The specificity of ISL1 inhibitory effect for rER was
also investigated using a COUP-TFI-dependent reporter construct (G.
Salbert, unpublished results). While cotransfection of 50 ng of
hCOUP-TFI expression vector alone allowed a strong activation of the
reporter construct transcriptional activity, cotransfection of
hCOUP-TFI with increasing amounts (5–100 ng per well) of zfISL1
expression vector had no effect on the induction mediated by
hCOUP-TFI (data not shown). These data are consistent with our EMSA
results showing that ISL1 does not affect hCOUP-TFI DNA binding and
further attest to the specificity of ISL1-ER functional
interaction. To map the functional(s) domain(s) of ISL1 required for
the repression of rER transactivation function, the ERE-SV-Luc
reporter plasmid was cotransfected with or without rER,
zfISL1, and zfISL1
HD/AD expression vectors in CHO cells. The
CMV-zfISL1HD/AD plasmid encodes for a truncated protein
corresponding only to the N-terminal LIM domains of zfISL1, thus
lacking both the homeodomain and the C-terminal transactivation domain.
Estradiol treatment of the cells cotransfected with rER alone led to a
strong increase in the reporter gene transcriptional activity (Fig. 6B). As previously shown, cotransfection of both rER and increasing
amounts (5–100 ng per well) of zfISL1 expression vectors resulted in a
strong and progressive decrease in the induction mediated by
E2 (Fig. 6B). Similarly, cotransfection of rER
expression plasmid with increasing amounts of zfISL1HD/AD expression
vector strongly reduced the E2-induced luciferase
activity but did not affect the basal luciferase activity (Fig. 6B).
These results suggest the involvement of the N-terminal LIM domains of
ISL1 in ISL1-ER interaction and in the subsequent modulation of rER
functions: the LIM domains of ISL1 indeed appear to be sufficient to
mediate the full inhibition of rER transactivation function, suggesting
that the homeo- and transactivation domains are dispensable for ISL1
effect. We then examined whether ISL1 could inhibit ER function through
exclusion of ER from the nucleus. This was done by cotransfection of
GFP-rER with ISL1 in COS-7 cells. No difference in the subcellular
distribution of ER was observed upon cotransfection with ISL1 (data not
shown).
View larger version (22K):
[in this window]
[in a new window]
|
Figure 6. ISL1 Inhibits rER Transactivation Function in a
Dose-Dependent Manner
The ERE-SV-Luc reporter was cotransfected in CHO cells with a constant
amount of CMV-rER expression vector and with increasing amounts of
CMV-zfISL1, CMV-zfISL2, and CMV-zfISL1HD/AD expression vectors.
Cells were either treated or untreated with E2
(10-8 M) as indicated. Panel A shows relative
luciferase activities corresponding to luciferase activities corrected
for transfection efficiency by the ß-galactosidase values. Untreated
cells are shown by gray bars while
E2-treated cells are shown by black bars.
Panel B shows "fold inductions" of the reporter gene activities
(luciferase activities normalized with ß-galactosidase activities) in
the presence of E2 compared with the activity in the
absence of ligand for each expression vector combination. All results
are expressed as the means ± SEM of three values
obtained in a representative experiment.
|
|
ISL1 Modulates the Ligand-Dependent Transcriptional Activity of the
rtER Gene Promoter
We next wished to confirm the previous results in a natural
estrogen-responsive promoter context. With this goal in mind, we used
the rtER gene promoter previously isolated and characterized in our
laboratory since it contains an estrogen-responsive unit (27, 28), as
well as three putative ISL1 DNA binding sites, I1, I2, and I3 (this
study). Two reporter plasmids, namely the 0.2-Basic reporter gene and
the 2.0-Basic reporter gene were used (26). These plasmids contain
either bp -207 to +28 or bp -2,079 to +28, respectively, of the rtER
gene inserted in front of the luciferase coding sequence. The 0.2-Basic
reporter plasmid only contains the estrogen-responsive unit of the rtER
gene promoter, composed of an imperfect functional ERE and a consensus
half -binding site for nuclear receptors (27, 28). The consensus
half-site together with the downstream ERE half-site form a DR24-type
element, which is bound by hCOUP-TFI (26). The 2.0-Basic reporter
construct contains, in addition to the estrogen-responsive unit, three
distal putative ISL1-binding sites identified in the rtER gene promoter
(I1, I2, and I3), which are absent in the 0.2-Basic reporter plasmid.
CHO cells were cotransfected with either 0.2-Basic or 2.0-Basic
reporter constructs with or without rER, hCOUP-TFI, and zfISL1
expression vectors (Fig. 7). When cells
were cotransfected with the 0.2-Basic construct and rER expression
vector, a 2-fold induction in the luciferase activity was observed in
the presence of E2. When both rER and hCOUP-TFI
expression vectors were cotransfected, a marked increase in the fold
induction by E2 was observed (Fig. 7), showing
that the orphan receptor hCOUP-TFI was able to enhance rER activation
properties as it was the case for rtER (26). Transfection of zfISL1
expression vector alone did not significantly affect the 0.2-Basic
reporter gene activity, suggesting that ISL1 had on its own neither
activatory nor inhibitory effect on the transcriptional activity of the
0.2- kb rtER promoter fragment lacking the ISL1 potential target
sequences (Fig. 7). In contrast, when both rER and zfISL1 expression
vectors were cotransfected with the same reporter construct, the
E2-induced luciferase activity decreased
significantly, the induction by E2 being 55% of
that observed with rER alone (Fig. 7). When rER, hCOUP-TFI, and zfISL1
expression plasmids were cotransfected, ISL1 was still able to inhibit
the ligand-dependent transcriptional activation, the fold induction by
E2 representing only 45% of the induction
mediated by rER and hCOUP-TFI in the presence of ligand (Fig. 7). When
cotransfected in CHO cells, the 2.0-Basic reporter construct had a
quite similar responsiveness to rER, although the ligand-dependent
induction in the transcriptional activity was slightly weaker than that
observed with the shorter 0.2-Basic construct. As above, hCOUP-TFI was
able to enhance the E2-stimulated luciferase
activity, therefore leading to an increase in the fold induction by
E2 when compared with the value obtained with rER
alone (Fig. 7). Despite the presence of ISL1-binding sites in this
reporter construct, ISL1, on its own, had no significant effect on the
transcriptional activity of the rtER promoter (i.e. did not
repress or activate the reporter gene). Moreover, in this promoter
context, ISL1 did not affect the ligand-dependent transcriptional
activation mediated by either rER alone or rER and hCOUP-TFI (Fig. 7).
Thus, while ISL1, on its own, has no regulatory effect on the activity
of the tested reporter genes, this transcription factor is able to
modulate the ligand-dependent activation of the estrogen-responsive
rtER gene proximal promoter. However, the ISL1 inhibitory effect occurs
only in the 0.2-Basic reporter gene context, a construct that, in
contrast to the 2.0-Basic reporter gene, lacks the ISL1 target DNA
sequences identified in the rtER gene promoter.
View larger version (28K):
[in this window]
[in a new window]
|
Figure 7. ISL1 Modulates the Ligand-Dependent rtER
Transcriptional Activity of the 0.2-Basic Reporter Construct but Has No
Effect on the 2.0-Basic Reporter Construct
The 2.0-Basic reporter plasmid contains the estrogen-responsive unit of
the rtER gene promoter composed of a functional ERE (bp -80 to -65)
and a half-binding site for nuclear receptor together forming a DR24
type element that is bound by hCOUP-TFI. This construct also contains
the three putative ISL1-binding sites identified in the rtER gene
promoter (I1, I2, and I3), which are absent in the 0.2-Basic reporter
plasmid. The 0.2-Basic and the 2.0-Basic reporter genes were
cotransfected in CHO cells with 50 ng of various expression vectors,
CMV-rER, CMV-hCOUP-TFI, and CMV-zfISL1. Cells were either untreated or
treated with E2 (10-8 M) after
transfection. Results are shown as "fold inductions" of the
reporter gene activities in the presence of E2 compared
with the activity observed in the absence of ligand. Results are shown
as the means ± SEM of 12 values obtained in five
independent experiments.
|
|
The rtER Gene Promoter Contains ISL1-Binding Sites
We then performed EMSA experiments to investigate whether the
zfISL1 protein could bind to the three putative ISL1-binding sites, I1,
I2, and I3, identified in the rtER gene promoter. These sequences were
defined by sequence analysis of the rtER gene promoter, based on their
sequence identity with the DNA consensus sequences previously described
for the chinook salmon Islet-2-related protein (29). As shown in Fig. 8, GST (lanes 1, 3, 5, 7, and 9) or
GST-ISL1 (lanes 2, 4, 6, 8, and 10) fusion proteins were incubated with
various 32P-labeled double-stranded
oligonucleotides, namely P1 (lanes 1 and 2), I1 (lanes 3 and 4), I2
(lanes 5 and 6), I3 (lanes 7 and 8), or EREc (lanes 9 and 10), as
indicated at the bottom of the gel. The P1 oligonucleotide
was synthesized based on the P1 element of the rat insulin I gene
enhancer that is bound by ISL1 (2). As expected, incubation of GST-ISL1
fusion protein with labeled double-stranded P1 oligonucleotide led to
the formation of a major retarded complex (Fig. 8, lane 2), which
corresponds to the specific binding of ISL1 to the P1 sequence. The
shifted band was indeed not observed when GST alone was incubated under
the same conditions with labeled P1 oligonucleotide (Fig. 8, lane 1).
Incubation of GST-ISL1 fusion protein with labeled I1, I2, or I3
oligonucleotides resulted in the formation of a shifted complex (Fig. 8, lanes 4, 6, and 8, respectively) reflecting specific binding of the
GST-ISL1 fusion protein. Finally, incubation of the fusion protein with
a labeled double-stranded EREc oligonucleotide did not result in the
formation of any retarded complex, demonstrating that, as expected,
ISL1 does not bind to the EREc sequence (Fig. 8, lane 10). Competition
EMSA carried out with labeled P1 oligonucleotide in the presence of
increasing amounts of unlabeled P1, I1, I2, and I3 oligonucleotides
confirmed the binding of ISL1 to the three rtER gene promoter sequences
and showed that ISL1 has a quite similar affinity for both P1 and I1
sequences, its affinity for I2 or I3 being weaker (data not shown).
According to these EMSA results, we can therefore conclude that, even
though ISL1 has, on its own, no regulatory effect on the full-length
rtER gene promoter, this transcription factor can bind the three
ISL1-binding sites identified within the rtER gene promoter.
View larger version (66K):
[in this window]
[in a new window]
|
Figure 8. The rtER Gene Promoter Contains Functional
ISL1-Binding Sites
The binding of ISL1 to the I1, I2, and I3 potential target DNA
sequences identified in the rtER gene promoter was assessed by EMSA.
Fifty nanograms of unpurified GST (lanes 1, 3, 5, 7, and 9) or GST-ISL1
(lanes 2, 4, 6, 8, and 10) fusion proteins were incubated with
32P-labeled P1 (lanes 1 and 2), I1 (lanes 3 and 4), I2
(lanes 5 and 6), I3 (lanes 7 and 8), or EREc (lanes 9 and 10)
double-stranded oligonucleotides. A black arrowhead
indicates the position of the ISL1-DNA complex.
|
|
ISL1 Does Not Affect rER Transactivation Function in the Presence
of ISL1-Binding Sites
A new set of transfection experiments was then performed to
determine whether the presence of ISL1-binding sites could interfere
with the ISL1-mediated inhibition of ER activity in an artificial
promoter context. The ERE-SV-Luc artificial estrogen-responsive
plasmid, which contains no ISL1-binding sites and whose ER-driven
transcriptional activity was previously shown to be inhibited by ISL1
(Fig. 6
), was used as a reporter construct. As the previous results
suggested that the inhibitory effect of ISL1 would not occur in a ISL1
binding site-containing promoter context, we generated two recombinant
reporter plasmids based on the ERE-SV-Luc construct. The ERE/I1-SV-Luc
reporter gene contains one of the ISL1-binding sites identified in the
rtER promoter, and the ERE/I2/I3/I1-SV-Luc reporter gene contains the
three ISL1 target sequences downstream of the estrogen-responsive
element. CHO cells were transfected with these reporter constructs,
with or without rER and zfISL1 expression vectors, and cells were
treated or not with E2. When cells were
cotransfected with the ERE-SV-Luc reporter gene and rER expression
vector (Fig. 9), an
E2 treatment markedly increased the luciferase
activity. Cotransfection of zfISL1 expression vector alone did not lead
to any significant modification of the reporter gene activity (Fig. 9).
In contrast, when rER and zfISL1 expression vectors were cotransfected
with the ERE-SV-Luc reporter construct, the induction mediated by
E2 decreased markedly compared with that observed
with rER expression vector alone (Fig. 9). The two recombinant
ERE/I1-SV-Luc and ERE/I2/I3/I1-SV-Luc reporter genes had a similar
estrogen responsiveness than the parental ERE-SV-Luc plasmid. The
transcriptional activation mediated by rER alone in the presence of
E2 were about 2.5- and 2.3-fold, respectively,
compared with the basal activities (Fig. 9). Despite the presence of
ISL1-binding sites in these reporter constructs, cotransfection of
zfISL1 expression vector alone did not affect (i.e. neither
increased nor decreased) the reporter genes’ transcriptional
activities (Fig. 9). In contrast to the inhibitory effect of ISL1 on
the ER-driven transcriptional activation of the ERE-SV-Luc construct,
no significant inhibition of the estrogen responsiveness of the two
recombinant reporter genes was observed in the presence of both rER and
ISL1. The transcriptional activation mediated by rER in the presence of
E2 and of ISL1 were indeed about 2.6- and
2.4-fold, respectively, for the ERE/I1-SV-Luc and ERE/I2/I3/I1-SV-Luc
reporter genes, compared with the basal activities (Fig. 9). We can
therefore conclude that ISL1 is able to inhibit the ER-driven
transcriptional activation in an artificial estrogen-responsive
promoter context, this inhibitory effect occurring only in the absence
of ISL1-binding sites. The presence of one or more ISL1-binding sites
in the reporter gene seems to be sufficient to restore the full
transcriptional activation of the estrogen-responsive promoter mediated
by rER in the presence of E2 and ISL1.
View larger version (32K):
[in this window]
[in a new window]
|
Figure 9. ISL1 Represses rER Transactivation Function in an
Artificial Estrogen-Responsive Promoter Context Devoid of ISL1-Binding
Sites
The ERE/I1-SV-Luc and ERE/I2I3/I1-SV-Luc reporter genes contain,
respectively, one (I1) or a combination (I2/I3/I1) of the three ISL1
DNA target sequences identified in the rtER gene promoter downstream of
the ERE. These reporter genes were cotransfected in CHO cells with 50
ng of the indicated expression vectors CMV-rER and CMV-zfISL1. Cells
were either treated or untreated with E2 (10-8
M) after transfection. Results are expressed as "fold
inductions" of the reporter gene activities in the presence of
E2 compared with the activity in the absence of ligand for
each expression vector combination. Results are shown as the means
± SEM of 12 values obtained in five independent
experiments.
|
|
rER Can Act as a Coactivator for ISL1 in the Absence of an ERE
In regard to the previous data, distinct but not mutually
exclusive hypotheses could be envisaged to explain the results obtained
in natural and artificial promoter contexts containing both an ERE and
ISL1-binding sites. First, it could be postulated that ISL1-binding
sites and rER compete with each other for interaction with ISL1;
binding of ISL1 to its DNA target sequences may decrease or prevent
ISL1-rER interaction, thus abolishing ISL1 inhibitory effect on rER
functions. Alternatively, rER may interact with the DNA-bound ISL1 and
serve as a coactivator of the LIM/HD protein, the estrogen-dependent
induction in the reporter constructs containing both an ERE and
ISL1-binding sites being thus mediated via the ISL1-binding sites. To
test this last hypothesis, we generated two artificial reporter
constructs, I1-SV-Luc and I2/I3/I1-SV-Luc, containing, respectively,
one or a combination of the three ISL1-binding sites identified in the
rtER gene promoter but lacking an ERE. CHO cells were transfected with
these reporter constructs as well as with the control SV-Luc plasmid,
with or without zfISL1 and rER expression vectors, and were treated or
not with E2 (Fig. 10). In the absence as well as in the
presence of ISL1-binding sites, cotransfection of ISL1 expression
vector alone did not modify the transcriptional activity of any of the
reporter constructs, neither in untreated nor in estradiol-treated
cells (Fig. 10). Similarly, when rER expression vector was
cotransfected alone, we did not observe any activation or repression of
the reporter construct activities, even in the presence of
E2 (Fig. 10). Cotransfection of the I1-SV-Luc
reporter construct with zfISL1 expression plasmid and increasing
amounts (50–100 ng per well) of rER expression vector did not affect
(neither increased nor decreased) the basal transcriptional activity,
but gradually enhanced the reporter gene activity observed in the
presence of E2. This resulted in a ligand- and
rER dose-dependent induction in the I1-SV-Luc reporter construct
activity (Fig. 10). Similar results were obtained with the
I2/I1/I3-SV-Luc reporter plasmid, the
E2-dependent induction observed in the presence
of both ISL1 and increasing amounts of rER expression vectors being
even more pronounced than that observed under the same conditions with
the I1-SV-Luc reporter construct (Fig. 10). In contrast, when similar
cotransfection experiments were performed using the zfISL1HD/AD
expression vector in spite of the full-length zfISL1 encoding vector,
no such E2-dependent induction in the I1-SV-Luc
and I2/I1/I3-SV-Luc transcriptional activities was observed in the
presence of increasing amounts of rER expression vector (data not
shown). These data strongly suggest that ER can serve as a coactivator
for ISL1 in certain promoter contexts, since ER is able to confer
estrogen dependency to artificial reporter constructs containing
ISL1-binding sites but lacking ERE. This effect is dependent not only
on the presence of ISL1-binding sites within the promoter but also on
the number of these sites. The coactivation requires ISL1 binding to
DNA through the homeodomain and may involve ER monomer tethering to the
promoter by the LIM domains of ISL1, this binary complex allowing
transactivation by E2.
View larger version (30K):
[in this window]
[in a new window]
|
Figure 10. rER Acts as a Coactivator for ISL1 in an
Artificial ISL1-Binding Sites-Containing Promoter Context Devoid of ERE
The I1-SV-Luc and I2I3/I1-SV-Luc reporter genes contain, respectively,
one (I1) or a combination (I2/I3/I1) of the three ISL1 DNA target
sequences identified in the rtER gene promoter inserted in the SV-Luc
reporter construct. These reporter genes were cotransfected in CHO
cells with the indicated amounts of expression vectors CMV-zfISL1 and
CMV-rER. Cells were either treated or untreated with E2
(10-8 M) after transfection. Results are
expressed as "fold inductions" of the reporter gene activities in
the presence of E2 compared with the activity in the
absence of ligand. Results are shown as the means ±
SEM of three values obtained in a representative
experiment.
|
|
ISL1 Binding to Its DNA Target Sequences Modulates ISL1-rER
Interaction
In the presence of both the ISL1-binding sites and the ERE within
a promoter, one would expect, according to the preceding results, a
higher estrogen-dependent activation when ISL1 and rER are present than
that mediated by rER alone. Furthermore, in these promoter contexts, a
more pronounced E2-dependent transcriptional
induction by ISL1 and rER would also be expected in the presence of
three ISL1-binding sites than in the presence of a single ISL1-binding
site. Nonetheless, Fig. 9 shows that the estrogen-induced
transcriptional activation was identical in the presence or absence of
ISL1 and occurred irrespective of the number of ISL1-binding sites
(from 1 to 3) present in the ERE-containing promoters. This observation
could suggest that, once bound to DNA, ISL1 has a reduced affinity for
ER, and that the estrogen response of promoters containing both type of
recognition sites is mainly due to rER interaction with the ERE. To
check this hypothesis, we set up GST pull-down assays using the
matrix-bound GST-hER LBD fusion protein to study its ability to retain
the 35S-methionine-labeled in
vitro-translated zfISL1, either in the absence or in the presence
of ISL1-binding sites. Consistent with our previous results, the
in vitro-translated zfISL1 protein was able to interact with
the LBD of ER in the absence of DNA, as revealed by the band observed
when 35S-methionine-labeled zfISL1 lysate was
incubated with the E2-treated GST-hER LBD fusion
protein (Fig. 11A, lane 3). This
protein-protein interaction was moreover specific, since no similar
band was present when the zfISL1-programmed lysate was incubated with
the GST fusion protein bound-beads (Fig. 11A, lane 2). The presence of
increasing amounts of I1 double-stranded oligonucleotide, corresponding
to the most distal ISL1-binding site characterized within the rtER gene
promoter, led to a strong and dose-dependent decrease in the amount of
35S-methionine-labeled zfISL1 protein retained by
the GST-hER LBD (Fig. 11A, lanes 4–7), showing that ISL1 displayed a
weaker affinity for ER in the presence of ISL1-binding sites than that
observed in the absence of DNA. In contrast, coincubation of the
35S-methionine-labeled zfISL1 lysate with the
GST-hER LBD fusion protein in the presence of increasing amounts of
double-stranded AP1 oligonucleotide did not influence ISL1-ER
interaction (Fig. 11A, lanes 8–11), showing that the inhibitory effect
of DNA on ISL1-ER interaction is specific for the I1 sequence. These
data not only confirm that ISL1 can interact in vitro with
ER via the LBD of the nuclear receptor, but also establish that the
presence of ISL1-binding sites specifically interfere with the
protein-protein interaction, binding of ISL1 to its DNA target
sequences decreasing its affinity for ER.
View larger version (27K):
[in this window]
[in a new window]
|
Figure 11. ISL1 Binding to DNA Decreases ISL1-rER Interaction
Panel A shows SDS-PAGE analysis of the
35S-methionine-labeled in vitro-translated
zfISL1 retained by the GST-hER LBD fusion protein in GST pull-down
assays, performed in the absence or in the presence of double-stranded
I1 and AP1 oligonucleotides. The zfISL1-programmed lysate was incubated
with limiting amounts of glutathione-agarose beads containing GST (lane
2) or GST-hER LBD (lanes 3–11) proteins, as indicated at the
bottom of the gel. The input lane (lane 1) corresponds
to 50% of the total amount of 35S-methionine-labeled
in vitro-translated zfISL1 used for the GST pull-down
assay. The amount of labeled in vitro-translated ISL1
retained by the GST-hER LBD protein in the absence of any
oligonucleotide is shown in lane 3. Increasing amounts (25, 50, 100, or
200 ng) of double-stranded I1 and AP1 oligonucleotides (lanes 4–7 and
lanes 8–11, respectively), were added as indicated at the
top of the gel. Panel B shows transfection experiment
results, expressed as relative luciferase activities, corresponding to
luciferase activities corrected for transfection efficiency by the
ß-galactosidase values. The ERE-SV-Luc reporter construct contains
the FP1 proximal region of the rtER promoter overlapping the imperfect
ERE. This reporter gene was cotransfected in CHO cells with the
indicated amounts of expression vectors CMV-rER and CMV-zfISL1. Cells
were either treated or untreated with E2 (10-8
M) after transfection. Each value corresponds to the
mean ± SEM of three values obtained in a
representative experiment. Untreated cells are shown by gray
bars while E2-treated cells are shown by
black bars. Note that in some wells, double-stranded
(ds) oligonucleotides (I1, AP1, and EREc) were cotransfected with the
different plasmids.
|
|
A last set of transfection experiments was then performed to confirm
that the presence of ISL1-binding sites could interfere with the
ISL1-mediated inhibition of ER activity in an artificial promoter
context. The ERE-SV-Luc artificial estrogen-responsive plasmid, which
contains no ISL1-binding sites and whose ER-driven transcriptional
activity was previously shown to be inhibited by ISL1 (Fig. 6), was
used as a reporter construct. Untreated or
E2-treated CHO cells were transfected with this
reporter construct, with or without rER and zfISL1 expression vectors,
and in the absence or presence of various double-stranded
oligonucleotides (I1, AP1, and EREc). As previously shown, in the
absence of oligonucleotides, rER activation of the reporter gene was
counteracted by ISL1 (Fig. 11B). Cotransfection of I1 oligonucleotide
(50 ng or 100 ng) with rER expression vector alone did not lead to any
significant modification of either the basal or the
E2-stimulated luciferase activity of the
ERE-SV-Luc construct (Fig. 11B). By contrast, when increasing amounts
of I1 oligonucleotide were cotransfected in the presence of rER and
ISL1, a gradual, dose-dependent reversion of the inhibitory effect of
ISL1 on the estrogen responsiveness was observed. The highest dose of
I1 oligonucleotide (100 ng) restored a fold induction by
E2 similar to that observed in the presence of
rER expression alone (Fig. 11B). Cotransfection of AP1 oligonucleotide
with rER expression vector alone led to a slight decrease in both the
basal and the E2-stimulated luciferase activity,
the fold induction by E2, however, being
unaffected. In cells cotransfected with both rER and zfISL1 expression
vectors, the presence of AP1 oligonucleotide did not modify the
inhibitory effect of ISL1 on the ER-driven transcriptional activation
that was still observed (Fig. 11B). Finally, cotransfection of EREc
oligonucleotide with rER expression vector alone led to a complete loss
of induction by E2 of the ERE-SV-Luc reporter
construct (Fig. 11B). We can therefore conclude that, while ISL1 is
able to inhibit the ER-driven transcriptional activation in an
artificial estrogen-responsive promoter context, this inhibitory effect
is reversed when ISL1 binds to DNA. These data suggest that binding of
ISL1 to DNA decreases its affinity for rER, thus allowing the
restoration of a full ligand-dependent transcriptional activation of an
ERE-containing promoter by rER.
|
DISCUSSION
|
|---|
Focusing on the adult rat hypothalamic arcuate nucleus, which has
been previously shown to be densely populated with ISL1-expressing
neurons (15), we report here that ISL1 and ER colocalize. ISL1 and ER
showed a strictly nuclear localization and were coexpressed in most
cells of this hypothalamic nucleus, very few neurons expressing only
one of these proteins. This result raises the possibility that ISL1
could be involved in the maintenance of the differentiated phenotype of
the ER-expressing neurons. To date, little is known about the molecular
mechanisms of LIM/HD protein functions in specification and maintenance
of neuronal properties since no or few LIM/HD target genes have been
identified. Based on the colocalization of ISL1 and ER in the central
nervous system, we hypothesized that ISL1 could functionally interact
with ER in vivo and, subsequently, modulate
estrogen-dependent target gene expression. Using in vivo
coimmunoprecipitation, we first demonstrated the existence of a
specific ISL1-ER interaction in vivo, suggesting that the
striking colocalization between ISL1 and ER observed in rat
hypothalamic nuclei could be physiologically relevant. GST pull-down
assays confirmed that these two transcription factors are able to
physically interact in vitro. Despite the fact that a number
of interacting partners have already been identified for nuclear
receptors, as well as for some LIM/HD transcription factors, this is,
to our knowledge, the first evidence for direct protein-protein
interaction between members of these two families of transcription
factors in vitro and in vivo. We show here that
ISL1 interacts in vitro with several nuclear receptors,
including orphan receptors such as COUP-TFI and SF-1. In addition, ISL1
interacts with the rat and rainbow trout ERs, but also with the natural
ER variant TERP, which consists only in the C-terminal LBD of rER (30).
This last observation is particularly interesting since it shows that
the ISL1-rER interaction is, at least in part, mediated by the LBD of
rER. The fact that estradiol enhances this protein-protein interaction
is in accordance with this idea. It can indeed be envisaged that ligand
binding to ER induces a conformational change within the LBD that
facilitates ISL1-ER interaction. Focusing on the LBD, we give in
vitro evidence that ISL1 may interact with ER residues required
for the generation of the LBD dimerization interface, ISL1 precluding
ER dimer formation in solution. Moreover, ISL1 is able, probably
through inhibition of ER dimerization, to specifically inhibit ER-ERE
complex formation. No similar inhibition of COUP-TFI DNA binding
activity was observed, although ISL1-COUP-TFI interaction was
demonstrated in pull-down assays, suggesting that ISL1 does not
negatively interfere with this orphan receptor. As a consequence, the
LIM/HD protein ISL1 specifically represses ER transactivation function.
The observation that ISL1 interacts with TERP may moreover suggest that
TERP could relieve ISL1-mediated inhibition of ER functions. Previous
works have shown that TERP exerts a biphasic action on ligand-activated
ER-dependent transcription, depending on the relative amounts of both
proteins (31). The inhibitory effect of TERP on ER transcriptional
activity observed at TERP to ER ratios equal to or greater than 1:1
have been shown to occur by formation of inactive TERP-ER heterodimers
and by competition for SRC-1 coactivator (32). Although little is known
about the molecular mechanisms underlying the stimulation of ER
transactivation function by lower TERP levels, one could envisage that,
under these conditions, TERP might titer a repressor of ER activity,
thus enhancing ER-dependent transcription. Such a repressor could be
ISL1.
Several previous studies have already shown that ER-mediated
transcriptional activation can be modulated by interaction with other
transcription factors. Indeed, in addition to its ability to associate
with other nuclear receptor superfamily members (33, 34, 35), ER has
already been described as interacting with homeodomain-containing
proteins, these interactions resulting in synergy between the two
transcription factors. For example, Pit1 increases the estrogen
responsiveness of the PRL promoter, the integrity of the DNA binding
POU-specific domain being required (36). This effect is likely to be
due to a cooperative binding of Pit1 and ER, the binding of Pit1 to its
DNA target site promoting interaction of ER with the nonpalindromic ERE
within the PRL promoter (36). Such cooperative interactions and
subsequent modulation of target gene expression have also been reported
between ER and the POU family member Brn3b (37). In this case, the POU
domain protein directly interacts with ER in vitro and
in vivo, and this interaction seems to occur independently
of DNA and of estradiol. This physical association requires the POU
domain of Brn3b and the DNA binding domain of the steroid receptor and
significantly enhances the affinity of ER for DNA. As a consequence,
this POU domain protein modulates the transcriptional activation of
estrogen-dependent promoters mediated by ER (37). Our results provide
the first evidence for a functional interaction between ER and a member
of the LIM/HD protein family. We report here that inhibition of ER
transactivation function by ISL1 is likely to be mediated by the
N-terminal LIM domains of ISL1: a truncated ISL1 protein consisting
only of the LIM motifs is indeed able to repress, in a dose-dependent
manner, ER-mediated transcriptional activation in a similar extent to
the full-length ISL1. A wide range of protein-protein interaction
involving LIM/HD proteins have already been reported, most of which are
mediated by the LIM domains (for review, see Ref. 38). The LIM
interacting proteins include cytoskeletal components (39, 40, 41), bHLH
proteins (42, 43, 44, 45), the insulin receptor (46, 47), as well as various
POU-HD transcription factors (48, 49, 50, 51), and other members of the LIM/HD
family (38, 40, 52, 53). In addition, LIM domains have recently been
shown to interact with a family of cofactors, including NLI/LBD1 or its
Drosophila homolog Chip (53, 54, 55, 56, 57, 58). As a result of their
ability to homodimerize, these cofactors allow indirect association of
several LIM/HD proteins, such as ISL1 and ISL2, which absolutely
require NLI/LBD1 to homo- and heterodimerize (53). Several data suggest
that LIM protein cofactors regulate LIM/HD proteins activity in
vitro and in vivo, but the mechanism by which these
cofactors act remains unclear (45, 54, 55, 56, 57, 58, 59, 60, 61, 62). Although NLI/LBD1/Chip have
been suggested to act, at least in part, by relieving the interference
exerted by LIM domains on LIM/HD proteins activity (54, 56, 59), recent
data argue against such a mechanism (61, 63). Aside from their role in
intermolecular interactions, LIM domains have indeed been proposed to
mediate intramolecular inhibition of LIM/HD activity (49, 50, 51, 56, 60, 64, 65). This hypothesis was first proposed based on the inhibitory
effect of the LIM domains of the rat ISL1 protein on its DNA binding
activity (64). Consistent with this result, the full-length ISL1 is
unable to activate transcription of a reporter gene containing the
"TAAT" motif, and both DNA binding and transactivation activities
are restored by deletion of the LIM domains (64). Another study reports
that overexpression of Islet 3 (ISL3) LIM domains in zebrafish embryo
causes dramatic cerebellar and ocular defects (60). These morphological
abnormalities can be rescued by coexpression of full-length ISL3,
suggesting that the LIM domains act as a specific dominant-negative
variant of ISL3 (60). Binding of coactivator proteins to the LIM
domains would be required to relieve the intramolecular inhibition they
exert and to allow activation of ISL3 (60). Consistent with this
hypothesis, mutation or deletion of the LIM domains activates xLIM1 in
a Xenopus axis induction assay (65), while binding of the
LBD1 cofactor to the LIM domains of the intact xLIM1 activates the
LIM/HD protein in vivo (54). Likewise, NLI/LBD1 enhances
transactivation properties of P-Lim in a similar manner to that
subsequent to the deletion of LIM domains (56). However, this model of
intramolecular inhibition is still controversial. Indeed, in both the
proglucagon and the somatostatin gene promoter context, no inhibitory
effect of the LIM domains was observed for the rat ISL1 protein (66, 67). Consistent with these last results, the LIM domains of the chinook
salmon ISL2 protein affect neither the specificity nor the affinity of
DNA binding (29). Furthermore, recent data show that deletion of the
LIM domains of Apterous results in a loss of biological activity of
this LIM/HD protein (63), which argue against a putative repressor role
for the LIM domains. Moreover, overexpression of the
Drosophila cofactor Chip results in a decrease in Apterous
activity in vivo, instead of the enhancement of the LIM/HD
protein function that would be expected according to the intramolecular
inhibition model (61).
In the present study, we give evidence that the full-length zebrafish
ISL1 protein specifically binds the P1 high-affinity target sequence
identified in the rat insulin enhancer (2), as well as the three
ISL1-binding sites, defined in the rtER gene promoter in regard to
their sequence identity with the consensus DNA targets characterized in
fish (29). Although ISL1 efficiently binds to these DNA sequences
in vitro, this transcription factor has, on its own, neither
any positive nor any negative transcriptional regulatory effect.
However, in artificial promoter contexts containing one or more
ISL1-binding sites but lacking an ERE, ER is able to serve as a
coactivator for ISL1, leading to a strong ligand-induced
transcriptional activation. This effect depends both on the presence of
ISL1-binding sites within the promoter and on the number of these sites
and requires the homeo- and/or transactivation domains of ISL1.
Coactivation is likely to involve ISL1 binding to DNA through the
homeodomain, and interaction of ER with the LIM domains, the DNA-bound
ISL1 thus tethering an ER monomer to the promoter. Ligand-induced
recruitment of coactivators by ER may potentiate transcriptional
activation. However, our results strongly suggest that coactivation is
unlikely to occur in an ERE-containing promoter context. Indeed, in an
artificial promoter context containing both the ISL1-binding sites and
the ERE, a similar ligand-induced transcriptional activation was
observed in the presence of ER and ISL1 than that mediated by ER alone.
If ER acted as a coactivator for ISL1 in this promoter context, one
would expect a higher estrogen-dependent activation when ISL1 and ER
are present than that mediated by ER alone. Moreover, in an artificial
ERE-containing promoter, a more pronounced
E2-dependent transcriptional induction by ISL1
and ER would also be expected in the presence of three ISL1-binding
sites than in the presence of a single ISL1-binding site, which was not
the case. Thus, it seems more likely that the estrogen response of
promoters containing both type of recognition sites is mainly due to ER
interaction with the ERE, inhibition of ER functions by ISL1 being
alleviated by binding of the LIM/HD protein to its DNA target
sequences. In accordance with this hypothesis, we demonstrate that,
once bound to DNA, ISL1 has a reduced affinity for ER and does no
longer inhibit ER transactivation function. We thus propose a model for
the regulation of ER functions by ISL1 in which, in the context of
estrogen-responsive promoters lacking ISL1-binding sites, ISL1 inhibits
ER-driven transcriptional activation (Fig. 12A). This specific repressor effect of
ISL1 on ER function is likely to be mediated through direct
protein-protein interaction involving the LIM domains of ISL1 and the
LBD of the nuclear receptor. Ligand-induced ISL1-ER interaction would
inhibit ER dimerization and subsequent DNA binding, decreasing
estrogen-dependent target gene expression. In promoter contexts
containing both types of recognition element, binding of ISL1 to its
DNA binding sites would weaken ISL-ER interaction and relieve
interference with ER transactivation function (Fig. 12B). Finally, in
the case of promoters containing ISL1-binding sites but lacking EREs,
monomeric ER may acts as a coactivator for ISL1. Interaction of ER with
the DNA-bound ISL1 may recruit coactivators and promote transcription
initiation, thus leading to the estrogen-dependent activation of genes
lacking EREs (Fig. 12C).
View larger version (41K):
[in this window]
[in a new window]
|
Figure 12. A Model for the Functional Cross-Talk between ISL1
and ER
Panels A and B show the putative underlying molecular mechanisms of the
modulation of ER functions by ISL1 in a promoter context containing an
ERE, in the absence (panel A) or presence (panel B) of ISL1-binding
sites (IBS). Panel C shows a putative molecular basis for the
enhancement of ISL1-driven transcriptional regulation by ER in a
promoter context containing only ISL1-binding sites (IBS).
|
|
Based on the striking colocalization of ISL1 and ER in some
hypothalamus nuclei, and on the in vivo ISL1-ER interaction,
we postulate the existence of a physiologically relevant cross-talk
between these proteins: ISL1 could modulate the expression of
estrogen-dependent genes in a promoter-specific manner. Targeted
mutation of the mouse ISL1 gene has been shown to perturb the
differentiation of many cell types (68). However, ISL1-/- embryo are
arrested in their development (E9.5) and die too early (E11.5) to allow
the analysis of a putative role of ISL1 function in modulating the
neuronal response to estrogen (68). Thus, the in vivo
functional relevance of ISL1-ER interaction and the regulatory
mechanisms modulating this cross-talk remain to be elucidated. Since in
any given cell nucleus both ISL1- and ER-binding sites are likely to be
present, under physiological conditions in which both transcription
factors bind DNA, one can hypothesize that these proteins are unlikely
to interfere with each other. Nevertheless, there might be particular
situations in which the inhibitory cross-talk occurs. Mechanisms
regulating the DNA binding activity of ISL1 would be particularly
relevant to the inhibitory activity of ISL1 on ER functions, since ISL1
binding to DNA modulates ISL1-ER interaction.
Phosphorylation events have been extensively described to
modulate transcription factors activity, affecting the structure, the
dimerization, DNA binding or transactivation activities, as well as the
interaction with coactivators/corepressors or with components of the
transcriptional machinery of various transcription factors. The target
proteins include basic-helix-loop-helix proteins, members of the
basic-leucine zipper family, and protooncogene products, as well as a
vast variety of nuclear receptor superfamily members (for review, see
Refs. 69, 70, 71, 72). Although no data are to date available concerning
LIM-homeodomain proteins, a putative role for phosphorylation events in
the regulation of ISL1 biological activity cannot be excluded. Such a
regulatory role on protein-protein interaction involving the
Antennapedia homeodomain protein have already been described (73).
Phosphorylation has also been reported to affect the DNA binding
ability of POU- and HD-containing transcription factors including
Pit-1, Oct-1, Oct-2, Engrailed, and Cut (74, 75, 76, 77, 78, 79). Other regulatory
mechanisms could be involved in the modulation of ISL1 functions and,
more particularly, in the regulation of the DNA binding activity of
ISL1. Although the hypothesis of intramolecular inhibition of the
HD-mediated DNA binding activity is highly controversial, one cannot
exclude that such an inhibition would occur in some particular cases,
still increasing the complexity of the mechanisms underlying ISL1
inhibition of ER functions. Although the precise function of the LIM
domains is not completely elucidated, these domains have been clearly
demonstrated to bind zinc and iron (29, 80, 81, 82, 83, 84, 85). The fact that the
modulation of the redox status of ISL1 abrogates DNA binding by the
ISL1 homeodomain suggests an additional mechanism for regulating ISL1
DNA binding activity (86). Finally, the activity of various LIM/HD
proteins has been shown to be regulated by the NLI/LBD/Chip cofactors
both in vitro and in vivo. Recent data
demonstrate that NLI is found in the nuclei of developing embryonic
neuronal cells and is coexpressed with ISL1 early in motoneuron
differentiation (55). The synchronous expression of both proteins in
such cells during the initial stages of differentiation strongly
suggests a role for this cofactor in the ISL1-dependent development of
motoneurons (55). Coexpression and modulation of ISL1 biological
activity by NLI in other neuronal populations during development and in
the adult could provide an additional mechanism for the regulation of
ER function by the LIM/HD protein ISL1.
Altogether, these data and our results give insight into the highly
complex mechanisms directing cell type- and promoter-specific gene
expression, which are due not only to the number and diversity of
transcription factors binding to regulatory sequences, but also to a
complex cross-talk between these transcription factors and to the
variety of regulatory events that modulate their biological
activities.
|
MATERIALS AND METHODS
|
|---|
Chemicals and Materials
The affinity-purified rabbit anti-ER antibody (ER MC-20),
Texas Red (TR)-labeled streptavidin, and protein G-agarose beads were
from TEBU (Le Perray-en-Yvelines, France). The anti-Islet1 monoclonal
antibody (39.4D5) and the rabbit anti-ER antibody (ER715) were
provided, respectively, by the Developmental S
TOP
ABSTRACT
INTRODUCTION
