(Received for publication, May 8, 1996, and in revised form, November 1, 1996)
From the Department of Pathology and Laboratory Medicine,
Molecular Biology Institute, and
Jonsson
Comprehensive Cancer Center, School of Medicine, UCLA, Los Angeles,
California 90095 and the § Department of Anatomy, School of
Medicine, University of California,
San Francisco, California 94143
Drosophila single-minded, which acts as a positive master gene regulator in central nervous system midline formation in Drosophila, its two mouse homologs SIM1 and SIM2, and the mammalian aryl hydrocarbon receptor (AHR) and aryl hydrocarbon receptor nuclear translocator (ARNT) proteins are members of the basic-helix-loop-helix·PAS family of transcription factors. In the yeast two-hybrid system, we demonstrate strong constitutive interaction of ARNT with SIM1 and SIM2 and fully ligand-dependent interaction of ARNT with AHR. Both the helix-loop-helix and the PAS regions of SIM1 and of ARNT are required for efficient heterodimerization. SIM1 and SIM2 do not form homodimers, and they do not interact with AHR. We also failed to detect homodimerization of ARNT.
The interaction of ARNT with SIM1 was confirmed with in vitro synthesized proteins. Like AHR, in vitro synthesized SIM1 associates with the 90-kDa heat shock protein. SIM1 inhibits binding of the AHR·ARNT dimer to the xenobiotic response element in vitro. Introduction of SIM1 into hepatoma cells inhibits transcriptional transactivation by the endogenous AHR·ARNT dimer. The mouse SIM1·ARNT dimer binds only weakly to a proposed DNA target for the Drosophila SIM·ARNT dimer. In adult mice mRNA for SIM1 was expressed in lung, skeletal muscle, and kidney, whereas the mRNA for SIM2 was found in the latter two. ARNT is also expressed in these organs. Thus mouse SIM1 and SIM2 are novel heterodimerization partners for ARNT in vitro, and they may function both as positive and negative transcriptional regulators in vivo, during embryogenesis and in the adult organism.
Basic-helix-loop-helix (bHLH)1 proteins constitute a
major group of transcription factors involved in
neurogenesis, myogenesis and other morphogenic processes (1). These
proteins act as homo- or heterodimers. The basic region and the
helix-loop-helix region are responsible for DNA binding and for
dimerization, respectively. The DNA-binding consensus motif for most
bHLH proteins, 5-CANNTG-3
, is referred to as the E-box. A subgroup of
bHLH proteins is characterized by the presence of a juxtaposed stretch
of approximately 300 amino acids, termed the PAS region, which contains
two degenerate ~50 amino-acid direct repeats, termed PAS-A and PAS-B.
The PAS region, initially identified in two Drosophila
proteins, period and single-minded, and the mammalian aryl hydrocarbon
receptor nuclear translocator (ARNT), was later found also to be
present in the aryl hydrocarbon receptor (AHR) and the
hypoxia-inducible factor HIF-1
(2-5). The PAS region can function
as a homodimerization interface, as a heterodimerization domain for
other bHLH·PAS proteins, and as an interaction domain with non-PAS
regions (6-10).
AHR is the only known ligand-activated bHLH transcription factor (11).
It binds a variety of carcinogenic and toxic environmental chemicals,
including polycyclic aromatic hydrocarbons (e.g.
benzo(a)pyrene and halogenated aromatic hydrocarbons (e.g.
2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD)), and mediates
most if not all of the pathological effects of these compounds. In the
cytosol of the mouse hepatoma cell line, Hepa-1, the unliganded
receptor is found in a complex with two molecules of the 90-kDa heat
shock protein (HSP90) and an unknown 43-kDa protein (12). Ligand
binding triggers release of the HSP90 and the 43-kDa molecules,
translocation of AHR to the nucleus, its heterodimerization with ARNT,
and subsequent DNA binding. The DNA target for the AHR·ARNT dimer,
the xenobiotic response element (XRE), present in the 5-flanking
sequences of TCDD-inducible genes, acts as a transcriptional enhancer.
The consensus XRE sequence for binding the AHR·ARNT dimer,
5
-TNGCGTG-3
, is nonpalindromic (13, 14).
Heterodimerization of ARNT with HIF-1 is required for
transcriptional up-regulation of several genes in response to hypoxia (5, 15). At high concentrations, ARNT has been demonstrated to form
homodimers and to bind the E-box motif 5
-CACGTG-3
, suggesting a
potential novel mode of regulation within the bHLH·PAS family of
proteins (16, 17).
Drosophila single-minded (dSIM) is a positive master gene
regulator of central nervous system midline formation. Two murine homologs of the dSIM gene have recently been isolated (18,
19). In addition to their expression in the central nervous system, both proteins are also expressed in a variety of other tissues during
development. The expression of SIM2 in adult mice appears to be
restricted to muscle, kidney, and lung (19). The human homolog of
SIM2 maps to the Down's syndrome critical region of chromosome 21 and has been implicated as a candidate gene responsible for the abnormal morphological features observed in Down's syndrome (20-22). A DNA-binding site (central nervous system midline element (CME), 5-(G/A)(T/A)ACGTG-3
), which resembles the XRE, was postulated for dSIM when putatively dimerized with another protein, based on
sequence analysis of enhancers of genes involved in midline formation
and subject to regulation by dSIM (23). A DNA target consensus
(5
-GT(G/A)CGTG-3
) for the dSIM·ARNT heterodimer, slightly different
from the CME, was identified by a binding site selection and
amplification strategy (24).
Here we investigate the properties of the two murine SIM proteins, SIM1 and SIM2, in conjunction with mouse ARNT and AHR. We demonstrate that both SIM1 and SIM2 can heterodimerize via their helix-loop-helix·PAS regions with ARNT, but not with AHR, and that they do not form homodimers. Furthermore, SIM1 may have a dual role, both negatively affecting AHR·ARNT binding to the XRE and also acting in concert with ARNT as a novel DNA-binding heterodimer.
The Gal4 activation domain vector pGAD425 and the
LexA DNA binding domain vector pBTM117C were generous gifts from J. Colicelli (UCLA) (25, 26). pBTM117C was generated by insertion of a 3.5-kb fragment carrying the CAN1 gene (27) in
pBTM117.2 PCR primers were synthesized on a
ABI PCR-Mate model 391 (Applied Biosystems) or purchased from IDT
(Coralville, IL). The yeast two-hybrid fusion constructs ARNTbHLHAB,
ARNTbHLH, ARNTbHLHA, ARNTAB, and AHRC were generated by PCR using
Amplitaq (Perkin-Elmer) with the corresponding ARNT and AHR cDNAs
contained in pcDNA1/Neo (9, 10) as templates. For ARNTHLHAB,
pcDNA1/NeoARNT
b was used as template. The SIM1 and SIM2
constructs SIM1
C, SIM1bHLH, SIM1bHLHA, SIM1AB, and SIM2
C were
PCR-amplified with full-length SIM1 and SIM2 cDNAs as templates.
The primer sequences are available upon request. The PCR products were
cloned into pCR2 (Invitrogen), excised with SalI and
NotI, and transferred into pGAD425 and pBTM117C.
Full-length SIM1 was PCR-amplified using Ultma Polymerase
(Perkin-Elmer) and cloned into pcDNA3 (Invitrogen). The addition of
the -globin 5
-untranslated region from pT7
HER, essentially as
described previously for AHR (28), resulted in 3-4-fold higher levels
of protein expression in the TNT T7-coupled rabbit reticulocyte system
(Promega). pcDNA3/SIM1
bH1 was generated in the same
fashion.
Strain L40C (MATa trp1 leu2
his3 can1 LYS::lexAopHIS3 URA3::lexAop-LacZ)
was a generous gift of J. Colicelli. For two-hybrid interaction assays,
cells were transformed with the combinations of the indicated plasmids
by standard procedures (29) and were patched onto selective medium to
ensure uptake of both plasmids. Single colonies were picked and grown
individually in selective medium. To test for ligand dependent
dimerization, aliquots were grown from the initial culture in the
presence of 105 M
-naphtoflavone in
Me2SO, (final concentration of 0.1% Me2SO) or
vehicle alone. Cells were resuspended in 500 µl of modified Z-buffer
(40 mM NaH2PO4, 60 mM
Na2HPO4, 10 mM KCl, pH 7.0). Fifty µl of 0.1% SDS and three drops of CHCl3 were added, and
the mixtures were vortexed for 1 min. The chromogenic substrate
o-nitrophenyl-
-D-galactopyranoside (Sigma) was added at 4 mg/ml in 200 µl of sodium
phosphate buffer (pH 7.0) and incubated at 37 °C. The reaction was
stopped by the addition of 500 µl of 1 M sodium
carbonate. Cell debris was spun down, and the absorbance of the clear
supernatant was measured at 420 nm (29). Reaction conditions were
linear with regard to both protein concentration and incubation
time.
The XRE
oligonucleotide was synthesized and labeled as described previously
(9). Eight different double-stranded CMEs (23) were generated by
annealing CME1-CME8
(5-TCCGGCTCTTC(T/G)CACGT(A/T)(C/T)CTCCGAGCTCA-3
) with
CME1
-CME4
(5
-TGAGCTCGGAG(A/G)(A/T)ACGTG-3
). Oligonucleotides were labeled with [
-32P]dATP (specific activity 6000 Ci/mmol, DuPont NEN), diluted 1:1 with H2O and purified on
a Chromaspin-10 (Clontech) and a Centrex 0.45-µm nitrocellulose spin
column (Schleicher & Schuell). Proteins were synthesized in the TNT T7
coupled systems with either rabbit reticulocyte or wheat germ lysate as
translation agents (Promega). EMSA was performed as described
previously using binding buffer A (25 mM HEPES, 200 mM KCl, 50 µg/ml poly(dI·dC)(dI·dC), 10 mM dithiothreitol, 1 mM EDTA, 10% glycerol, pH
7.5) (9). In some experiments, the KCl and the poly(dI·dC)(dI·dC)
concentration of the binding buffer was lowered to 50 mM
and 10 µg/ml, respectively (binding buffer B). Unlabeled,
double-stranded CME2 (5
-TGAGCTCGGAGGAACGTGAGAAGAGCCGGA-3
or the
cAMP-response element (5
-GGATCCATGACGTCATGGATC-3
) was used as a
competitor. In brief, equimolar amounts (unless otherwise stated) of
in vitro synthesized proteins were incubated in the presence
of 10
8 M TCDD in Me2SO (final
concentration of 1% Me2SO) or vehicle alone at room
temperature for 90 min. Incubation was continued for a further 20 min
after the addition of binding buffer and for a further 20 min after the
addition of the radiolabeled XRE (15,000 or 100,000 cpm/reaction). The
protein-DNA complexes were resolved on nondenaturing polyacrylamide
gels (4.5% acrylamide) with 1 × HTE running buffer (200 mM HEPES, 100 mM Tris, 5 mM EDTA, pH 8.0), and the gels were fixed in 10% acetic acid, 30% MeOH, dried,
and exposed to Biomax autoradiography film (Eastman Kodak Co.).
Quantitation of protein-DNA complexes was performed on an AMBIS
radioanalytic system (Ambis).
Proteins were synthesized in the TNT T7 coupled transcription/rabbit reticulocyte lysate translation system in the presence or absence of [35S]methionine as indicated. Coimmunoprecipitation was carried out as described previously (9). For the interactions with HSP90, equimolar amounts of in vitro synthesized proteins were diluted in HSP90 immunoprecipitation buffer and SIM1·HSP90 or AHR·HSP90 complexes precipitated with anti-HSP90 antibodies as described previously (10).
Transient TransfectionsHepa-1c1c7 cells were cultured and
maintained in nucleoside-free -minimal essential medium (Irvine
Scientific) supplemented with 10% heat-inactivated fetal calf serum
(Sigma). 5 × 105 cells in 60-mm
dishes were transfected with the plasmids as indicated, according to
the transfection method of Chen and Okayama (30), except that the
CO2 concentration was maintained at 5% throughout the
procedure. Plasmids for transfection were prepared by the Qiagen
maxiprep procedure according to the supplier's protocol. 16-24 h
after transfection, cells were subjected to a 15% glycerol shock for
1.5 min, washed extensively with phosphate-buffered saline and then
refed media with or without 10 nM TCDD. 24-36 h later,
cells were harvested and washed in phosphate-buffered saline, and
lysates were prepared by repeated freezing and thawing. Endogenous
acetyltransferases were inactivated by incubating at 65 °C for 10 min. Protein concentrations were measured according to the method of
Bradford using bovine serum albumin as a standard (Bio-Rad). CAT
activities were determined and expressed as nmol of
N-acetylchloramphenicol formed/30 min/mg of protein.
A mouse multiple tissue Northern
blot (Clontech) was probed with a random primed,
[-32P]dATP-labeled cDNA fragments of the
C-terminal parts of SIM1 and SIM2 (starting at the codons corresponding
to amino acid 392 of SIM1 and to amino acid 393 of SIM2, respectively).
Prehybridization, hybridization, and washing conditions were according
to the manufacturer's protocol. The blot was exposed to Biomax
autoradiography film (Kodak).
This system for assessing protein-protein interactions
takes advantage of the modular nature of eukaryotic transcription
factors. One protein of interest is fused to a DNA binding domain, and the other is fused to a transcriptional transactivation domain. Dimerization of the two fusion proteins reconstitutes a functional transcription factor that can activate several reporter genes incorporated into the genome of the host cells. Previously, we demonstrated that full in vitro heterodimerization and XRE
binding activity was retained by ARNT and AHR constructs lacking their C-terminal halves, indicating that the bHLH·PAS regions were
sufficient for ligand-dependent heterodimerization (9, 10).
Here we generated constructs such that the LexA DNA binding domain and the Gal4 transactivation domain were fused with the bHLH·PAS regions of the mouse proteins ARNT, AHR, SIM1, and SIM2 and with several mutant
forms of ARNT and SIM1. A full-length mouse SIM1 mammalian expression
plasmid and an N-terminal deletion construct (SIM1bH1) were also generated (Fig. 1).
The interactions among ARNTbHLHAB, AHRC, SIM1
C, and SIM2
C were
quantitated by assaying
-galactosidase activity in the transformed
yeast cells (the
-galactosidase gene in the host cells contains
several LexA binding sites in its 5
-flanking sequence) (Table
I). The cells were either untreated or treated with the AHR agonist BNF. Heterodimerization of ARNTbHLHAB with AHR
C was ligand-dependent. BNF treatment resulted in a 40- or
60-fold increase in dimerization of these proteins (as measured by
-galactosidase reporter activity after subtracting the background
values of the empty vector controls, and depending on the orientation
of ARNTbHLHAB and AHR
C constructs in the two hybrid vectors). Such a
marked effect of ligand on AHR·ARNT dimerization has previously only been observed in mammalian cells expressing the endogenous AHR and ARNT
proteins and not in yeast (31-34). ARNTbHLHAB interacted with SIM1
C
and less strongly with SIM2
C, the addition of BNF having no effect
in either case. No ARNT homodimerization was observed with the bHLHAB
construct. Similar results were obtained with an ARNT derivative HLHAB,
identical to bHLHAB except for the absence of the basic region (data
not shown). No evidence for homodimerization of AHR
C or for it
interacting with SIM1
C or SIM2
C was obtained. Finally, no
homodimerization of either SIM1
C or SIM2
C and no
heterodimerization of SIM1
C with SIM2
C was observed. Thus, in the
two-hybrid system, AHR, SIM1, and SIM2 each interacted exclusively with
ARNT, and each failed to form homodimers, as did ARNT. Similar results
were obtained with each pair of proteins regardless of which protein
was expressed as a fusion with the Gal4 transactivation domain and
which was expressed as a fusion with the LexA DNA binding domain.
|
The contribution of the bHLH and the PAS region of SIM1 toward
heterodimerization with ARNT-bHLHAB was also investigated (Table II). No interactions were observed between ARNTbHLHAB
and SIM1 constructs containing only the bHLH, bHLHA, or AB regions.
Identical results were obtained for SIM1C with the equivalent ARNT
constructs. These results indicate that the entire bHLH·PAS regions
of SIM1 and ARNT are required for efficient heterodimerization in
yeast.
|
To confirm the observed interaction of ARNT
with SIM1, the heterodimerization between ARNT and SIM1 was assessed
using in vitro expressed proteins. The proteins were
synthesized separately in rabbit reticulocyte lysate, either in the
presence (SIM1) or absence (ARNT) of [35S]methionine.
They were then incubated together, subjected to immunoprecipitation
with antibodies to ARNT, and subjected to SDS-polyacrylamide gel
electrophoresis (Fig. 2A). SIM1 was
coimmunoprecipitated with ARNT, both in the presence or absence of
TCDD. SIM1 was absent from the supernatant, indicating that the
coimmunoprecipitation was quantitative (data not shown). A small amount
of SIM1 was precipitated in an incubation mixture containing preimmune
serum, which was comparable with the amount precipitated in the absence of ARNT protein. ARNT antibodies did not coprecipitate significant amounts of the SIM1bH1 construct (data not shown). In a
separate experiment, in vitro synthesized SIM1 failed to
bind the AHR photoaffinity ligand
2-azido-3-[125I]iodo-7,8-dibromodibenzo-p-dioxin
under conditions in which binding to AHR is clearly discernible (Ref.
10 and data not shown). Thus SIM1·ARNT heterodimerization in
reticulocyte lysate is constitutive and TCDD-independent and, as
expected, requires helix 1 of SIM1.
SIM1 Associates with HSP90
In the absence of ligand, cytosolic AHR is found in a complex with two molecules of HSP90 (12). When AHR is translated in rabbit reticulocyte lysate, it associates with HSP90 present in the lysate (10, 35). When AHR and SIM1 were synthesized in the presence of [35S]methionine and then treated with HSP90 antibodies, AHR and SIM1 were precipitated, whereas the corresponding control IgM did not precipitate the labeled proteins (Fig. 2B). Thus, like AHR, SIM1 complexes with HSP90.
Weak Binding of the ARNT·SIM1 Heterodimer to the CMEThe
CME has been proposed as the DNA binding site for dSIM putatively
associated with another (ARNT-like) protein (23). We analyzed the
different forms of the consensus CME core sequence, 5-(G/A)(T/A)ACGTG(A/C)-3
in separate EMSAs. No ARNT·SIM1·CME complex formation was observed using ARNT and SIM1 synthesized in
vitro in either rabbit reticulocyte or wheat germ lysate, with our
routine EMSA conditions using buffer A, which contains 200 mM KCl and 50 µg/ml poly(dI·dC)(dI·dC). Under less
stringent conditions using buffer B, which contains 50 mM
KCl and 10 µg/ml poly(dI·dC)(dI·dC), and using increased amounts
of radiolabeled oligonucleotide, weak complex formation was observed
for CMEs 1-4
(5
-TGAGCTCGGAG(A/G)(A/T)ACGTGAGAAGAGCCGGA-3
), with
minor differences observed between them. No binding was
observed for CMEs 5-8
(5
-TGAGCTCGGAG(A/G)(A/T)ACGTGCGAAGAGCCGGA-3
). SIM1·ARNT binding on
CME2 was directly compared with AHR·ARNT dimer binding to the XRE
using binding buffer B for both EMSA assays (Fig. 3). Binding to the XRE required both ARNT and AHR, and binding to CME2
required both ARNT and SIM1. SIM1·ARNT·CME2 complex formation was
abolished with a 200-fold molar excess of unlabeled CME2 but was not
affected by a 200-fold molar excess of a c-AMP response element
oligonucleotide; it could furthermore be supershifted by an ARNT
antibody (data not shown). The amount of AHR·ARNT·XRE complex
formed was 60 times greater than the amount of SIM1·ARNT·CME2, when
equimolar amounts of ARNT, SIM1, and AHR were utilized in the EMSA.
SIM1 Inhibits the DNA Binding Activity of the AHR·ARNT Heterodimer in Vitro
To investigate the effect of SIM1 on the
formation of the AHR·ARNT·XRE complex, EMSA was performed under the
high stringency binding conditions using buffer A (Fig.
4A). No binding to the XRE was observed using
equimolar amounts of ARNT and SIM1, either in the presence or absence
of TCDD. As demonstrated previously, formation of the AHR·ARNT·XRE
complex was dependent on ligand. Addition of SIM1 reduced
AHR·ARNT·XRE complex formation. A 50% reduction was achieved with
an amount of SIM1 equimolar to that of ARNT and AHR, indicating that
SIM1 and ligand-activated AHR have similar affinities for ARNT. The
addition of SIM1bH1, which as indicated above, does not
bind ARNT, did not inhibit AHR·ARNT·XRE complex formation (Fig.
4B).
SIM1 Inhibits Ligand-dependent AHRC Activity in Mouse Hepatoma Cells
The reporter plasmid pMC6.3k contains the upstream
regulatory promoter/enhancer region of the rat CYP1A1 gene
linked to the CAT reporter gene. It directs TCDD-dependent
CAT activity when transfected into a cell line possessing functional
ARNT and AHR. When the SIM1 expression vector pcDNA3/SIM1 was
cotransfected along with pMC6.3k into Hepa-1c1c7 cells, which express
both ARNT and AHR (9), it abolished TCDD-dependent CAT
activity. Neither SIM1bH1 nor the parental vector,
pcDNA3, affected TCDD-induced CAT activity (Fig.
5A). Thus, SIM1 negatively modulates
transcriptional activation of the AHR·ARNT dimer in vivo.
SIM1 did not impair the constitutive reporter activity of pSV2CAT in
Hepa-1 cells (Fig. 5B), illustrating the specificity of the
inhibitory effect of SIM1 on AHR·ARNT and indicating that inhibition
is not due to an effect on the general transcriptional machinery.
In the Adult Mouse, SIM1 and SIM2 mRNAs Are Expressed in Kidney and Skeletal Muscle, and SIM1 Is Also Found in Lung (Fig. 6)
A multiple-tissue Northern blot of adult mouse
was first hybridized with a probe specific for SIM1 (i.e.
with its C-terminal half) and exposed to film. This revealed an
exceptionally large 9-kb transcript in kidney, skeletal muscle and at a
low level in lung, consistent with the mRNA size found in mouse
embryo (18). The blot was then hybridized with a probe specific for
SIM2. The SIM2 transcript (4 kb in length) was found to be expressed in kidney and muscle. The SIM2 transcript size is consistent with that
found by Ema et al. (19).
bHLH transcription factors can be divided into two major
functional groups. Class A proteins are expressed fairly ubiquitously and form heterodimers with class B proteins. Class B proteins form
functionally active heterodimers with class A proteins but do not
interact with other class B proteins (1). Analogous to E12 and E47,
which act as ubiquitous class A heterodimerization partners for many
other bHLH proteins, ARNT should be classified as a mammalian "class
A" bHLH·PAS protein, which can interact with the "class B"
bHLH·PAS proteins AHR, HIF-1, SIM1, and SIM2. These last proteins
behave like conventional class B bHLH proteins in that they only form
functional heterodimers with ARNT. However, unlike conventional class B
bHLH proteins, which have limited tissue or lineage specificity, some
at least of the bHLH·PAS class B proteins appear not to be restricted
to certain tissues and/or lineages but are activated by specific
exogenous agents (ligand for AHR, hypoxic conditions for HIF-1). ARNT
appears to be absolutely required for the signal-transducing activity
of all class B bHLH·PAS proteins, since the latter appear to be
capable of binding to DNA only when dimerized in a complex with ARNT
(the recently described ARNT2 protein (33) might possibly substitute
for ARNT in these interactions). The basic region of ARNT conforms well
to the consensus for the basic region of other bHLH proteins. In
DNA-binding bHLH·PAS protein complexes studied so far, ARNT binds to
the invariant 5
-GTG-3
motif (24, 36). This motif, reminiscent of an
E-box half-site, is present in all of the known DNA targets for
bHLH·PAS dimers. The more divergent and variable basic regions of the
class B bHLH·PAS proteins thus dictate binding to the non-E-box-like region of bHLH·PAS DNA targets.
Previously, we observed no dimerization of ARNT with ARNT bHLHAB in vitro (9). Here we found no significant homodimerization of the bHLHAB or HLHAB constructs of ARNT in yeast. Thus, ARNT appears to be unable to homodimerize, at least at moderate concentrations (9). However, other investigators have reported homodimer formation for full-length ARNT partially purified from insect and mammalian cells. Furthermore, E-box-mediated transcriptional activation was demonstrated in vivo, when ARNT expression was driven by strong promoters (16, 17). The homodimerization of ARNT observed by these investigators thus occurred at high nonphysiological concentrations and may not occur under normal in vivo conditions. Hirose et al. (34) also failed to detect homodimerization of ARNT in the yeast two-hybrid system. Since we used constructs truncated at the N terminus and at the C terminus, another possible explanation for our results is the requirement of additional sequence(s) outside of the bHLH·PAS region for ARNT homodimerization.
AHR·ARNT heterodimerization in our LexA-based yeast two-hybrid system is fully ligand-dependent and highly inducible. Other groups have also reconstituted the AHR·ARNT signaling pathway with full-length or with chimeric proteins in yeast and found considerably higher background and lower induction levels (32-34), probably due to use of different yeast expression vectors and/or host strains of different genetic backgrounds.
Treatment with ligand for AHR did not increase the SIM1/ARNT or SIM2/ARNT interactions; nor did ligand increase the SIM1/ARNT interaction in the in vitro coimmunoprecipitation assay. The result for the SIM1·ARNT dimer is consistent with the observation that in vitro synthesized SIM1 did not bind an AHR agonist. The signal(s), if any, promoting association of ARNT with SIM1 and SIM2 remain unknown. It is possible, however, that SIM1/ARNT and SIM2/ARNT interactions are ligand-dependent processes and that the relevant ligand(s) is present in yeast, Hepa-1c1c7 cells, and reticulocyte lysate. The observation that SIM1 binds HSP90 in reticulocyte lysate does not preclude the possibility that reticulocyte lysate contains ligand for SIM1, since AHR can bind HSP90 in reticulocyte lysate even when ligand for AHR is added (10). The highly efficient association of SIM1 with ARNT upon expression in reticulocyte lysate suggests that SIM1 probably binds ARNT more efficiently than it binds HSP90. Other investigators have demonstrated binding of Drosophila SIM to human ARNT and rabbit HSP90 (37). Our observations involve interactions between bHLH·PAS proteins that are all mammalian in origin and are therefore of greater biological relevance.
dSIM, a master gene regulator in central nervous system midline formation, acts as a positive regulator of transcription in Drosophila during development (38). It contains in its C-terminal part several potent transactivation domains. Similar transactivation domains have been identified in ARNT and AHR (39, 40). However, the C-terminal halves of both SIM1 and SIM2 do not show any similarity to the Drosophila homolog. Assuming that the proposed CMEs or the consensus sequence identified by the binding site selection strategy are the DNA targets for the SIM1·ARNT heterodimer, the discrepancy between the strong SIM1·ARNT heterodimerization and its very weak binding activity on the CME could indicate that DNA binding but not heterodimerization of SIM1 with ARNT requires a factor(s) or a post-translational modification activity nearly absent from reticulocyte lysate. Alternatively, there may be cooperative binding of SIM1·ARNT dimers to different CME sites or cooperative binding of SIM1·ARNT with transcription factors that bind to other sites in the flanking region of SIM1-regulated genes. Finally, it is possible that the weak DNA binding in vitro reflects the true in vivo situation. The basic region of SIM1 (and SIM2) starts with its amino-terminal amino acid. AHR requires additional amino acids amino-terminal to its basic region for strong binding to the XRE (28, 41). The absence of equivalent amino acids in SIM1 could explain its very poor DNA binding activity.
Our demonstration that SIM1 can inhibit AHR·ARNT binding to the XRE and can inhibit expression from an XRE-driven reporter gene indicates that SIM1 may act as negative regulator of transcription as well as a positive regulator. The above inhibitory effects may result from SIM1 competing with AHR for binding to ARNT, although we cannot exclude the possibility that SIM1 may also have other inhibitory effects of AHR·ARNT activity. In a similar fashion, SIM1 may act as a negative regulator of all ARNT-dependent genes.
In the adult mouse, the mRNAs for SIM1 and SIM2 are found in tissues, where ARNT and AHR mRNA expression have previously been detected (42). Studies on the embryonic expression patterns of ARNT, AHR (43, 44), SIM1, and SIM2 (17, 18) in the mouse indicate that these proteins are also coexpressed in a number of developing tissues, including portions of the forebrain, embryonic muscle, and facial cartilage. Importantly, during embryogenesis, the distribution of SIM2 expression was always accompanied with that of ARNT (19). This apparent coexpression implies that the interaction between ARNT and the two mouse SIM proteins demonstrated here is likely to be physiologically relevant and that SIM1 and SIM2 are tissue-specific modulators of AHR·ARNT activity.
K. Chalvardijan provided expert technical assistance. We thank Drs. S. Elledge, S. Hollenberg, and J. Colicelli for yeast strains and plasmids and S. Crews for helpful discussions.
While this manuscript was being prepared, Ema
et al. (19) reported on the isolation and characterization
of a cDNA for mSIM. The mSIM cDNA sequence was identical to
SIM2 over the bHLH·PAS region but differed in its C-terminal half
from it. The corrected SIM2 cDNA sequence has meanwhile been
published as an erratum (18). The results and the conclusions drawn
from the current study are unaffected, since all expression constructs
in our experiments used only the bHLH·PAS region of SIM2
(i.e. SIM2C), the cDNA sequence of which is identical
to that of mSIM.