The thyroid hormone receptor (TR) regulates the transcription of its target
genes by interacting with specific hormone response elements consisting usually of directly repeated half-sites with the consensus sequence AGGTCA. To investigate the role of the spacer sequences
separating the half-sites, heterodimers formed by TR
[alpha]
and the retinoid-X receptor (RXR) were used in a PCR based selection and amplification assay. The TR
[alpha]
/RXR heterodimer selected for elements with directly repeated half-sites having a spacer of 4 nucleotides (DR4). Preferences for nucleotides
in the TR binding half-site motif as well as for the 4 nucleotides separating the two half-sites were found. DNA binding and transfection studies using DR4
elements with different spacer sequences showed the importance of these
nucleotides for the activity of the response element: some spacer sequences allowed little or no transactivation from the element, whereas other
sequences supported strong transactivation. A pyrimidine nucleotide in position
three of the spacer enhanced TR
[alpha] binding and transactivation. Additional experiments showed that heterodimers between RXR and other putative receptors exhibited a similar but distinct specificity for the spacer sequence.
Our results thus suggest that the four nucleotides separating the two half-sites in hormone response elements have a major role in determining
induction of hormone responsive genes.
In vertebrate organisms a multitude of cellular events such as growth, development, differentiation and homeostasis are regulated by thyroid hormones, which affect their target cells by interacting with
intracellular thyroid hormone receptors (TRs) (
1
). The thyroid hormone receptor was identified as a ligand-dependent transcription factor which belongs to the steroid/thyroid
hormone receptor superfamily of nuclear receptors (
1
-
3
). The members of this gene family share high structural similarities,
particularly in their DNA binding domains (
4
-
6
). Due to their DNA binding domains, these receptors interact with specific
hormone response elements in their target genes. These elements contain two different core recognition motifs (half-sites) which were previously identified for the glucocorticoid (AGAACA) (
7
) and the estrogen receptor (AGGTCA), respectively (
8
). Since the recognition motifs for all members of the nuclear receptor family are highly conserved, the specificity of hormone mediated signal transduction is
regulated by the arrangement of these recognition motifs in dimeric or
multimeric receptor-specific binding sites (
9
).
Thyroid hormone response elements (TREs) contain two half-sites of the AGG TCA motif (
9
), which can be arranged as direct repeats (DR), inverted repeats (IR) or
everted repeats (ER) (
10
). TR binds to and transactivates from these elements as a heterodimer together
with the retinoid-X receptor (RXR) (
11
-
15
). RXR heterodimerises not only with TR, but serves as the partner for various
nuclear receptors including those for retinoic acid, vitamin D3, and a number
of orphan receptors for which no ligands have yet been identified (
16
). For response elements of the direct repeat type, the number of nucleotides
separating the two half-sites specifies receptor specificity (
17
,
18
). Accordingly, TR-RXR heterodimers bind preferentially to elements spaced by 4 nucleotides
(nt) (DR4). In these complexes RXR occupies the 5' half-site while TR, located at the 3' half-site of the element, determines the specificity (
19
-
21
). RXR augments thyroid hormone mediated transactivation by increasing the
affinity of TRs for binding to the cognate response elements.
The analysis of promoters from TR target genes led to the identification of
natural TREs, of which most elements are of the DR4 type (
18
,
22
-
26
). However, none of these elements contains two copies of the consensus AGGTCA
motif. In addition, in some elements, like in the rat growth hormone promoter,
more than two half-sites were found which are arranged as a direct and an inverted repeat (
27
-
30
). The principles by which these complex elements govern transcription of target
genes are not yet fully understood. The variations in the sequence and
structure of response elements and the possibility of TR isoforms to
heterodimerise with different forms of RXRs offer flexibility in modulating
ligand-dependent transactivation. To characterise the structural requirements
within the spacer sequence of a TRE, we applied a PCR-based approach to select for an optimal binding site for TR[alpha]-RXR heterodimers. The identified sequences shows that TR[alpha] in this assay preferentially binds to direct repeats
separated by 4 nt, and that the nucleotide composition of the sequence separating the repeats influences DNA binding affinity as well
as transactivation properties of TRs. Our data also suggest that other, still
unidentified proteins, have similar but distinct requirements on the
composition of the spacer.
cTR[alpha] and mRXR[gamma] were expressed as histidine-tagged fusion proteins in a Vaccinia virus expression
system. The his-tagged cTR[alpha] was described earlier (
31
). A mRXR[gamma] cDNA was cloned into the
Eco
RI site of pATA-18-His (
32
) and used for recombination into the Vaccinia virus genome as described (
33
). Nuclear extracts from infected HeLa cells were prepared as described by
Vivanco Ruiz
et al
. (
34
). The his-tagged receptors were purified from nuclear extracts on a Ni
2+
-NTA matrix (Diagen, Hilden, Germany) as described by Janknecht
et al
. (
35
). After chromatography, the receptors were concentrated by ammonium sulphate
precipitation [0.295 g (NH
4
)
2
SO
4
/ml], dialysed against a buffer containing 20 mM HEPES (pH 7.9), 20% glycerol, 0.1 M KCl, 0.2 mM EDTA, 0.5 mM PMSF and 0.5
mM DTT, and stored at -70oC. For preparation of His-cTR[alpha]-His-mRXR[gamma] heterodimers HeLa cells were co-infected with viruses
encoding His-cTR[alpha] and His-mRXR[gamma]. The heterodimers were extracted from these cells and
purified as described for the other receptors. These extracts contained almost
exclusively heterodimerised His-cTR[alpha] and His-mRXR[gamma] (data not shown). The expression of unmodified
receptors (cTR[alpha] and mRXR[alpha]) was described in reference (
32
).
For the preparation of whole cell extracts from JEG cells, cells from confluent
10 cm dishes were washed with PBS, scraped off the plate and collected by
centrifugation. The cell pellet was then extracted on ice with 150 [mu]l of a buffer containing 50 mM Tris-HCl (pH 7.5), 440 mM NaCl, 5 mM MgCl
2
, 0.5% Triton X-100, 10 mM DTT, 0.2 mM PMSF and 200 U/ml trasylol (Bayer, Leverkusen,
Germany). After 30 min the extracts were cleared by centrifugation and stored
at -70oC.
Gel retardation assays were performed and quantified as described by (
36
)-for supershifts 1 [mu]l antiserum was added to the reaction mixture and then incubated for
further 30 min on ice before loading the gel. The anti-TR[beta] antiserum was described by (
37
); the anti-RXR antiserum was a gift from P. Chambon.
For the selection of an optimal binding site the following oligonucleotides were
synthesised (SGS, Ksping, Sweden):
Template:
5'-CCC GAA TTC CCC ACT GCT TAT
The primers were labelled at their 5'-ends using [[gamma]-
32
P]ATP (Amersham, Buckinghamshire, UK) and T4 polynucleotide kinase (Promega,
Madison, USA) and then applied to PCR. In the first reaction, 5 pmol template
were amplified with 30 pmol of each labelled primer during 20 cycles of 1 min
at 94oC, 1 min at 62oC and 1 min at 72oC (
Taq
polymerase and 10* buffer were obtained from Promega, Madison, USA). The PCR product was
purified in a 10% non-denaturing polyacrylamide gel and extracted with TE buffer. The labelled
PCR product was then incubated with purified His-cTR[alpha]-His-mRXR[gamma] heterodimers. Bands co-migrating with His-cTR[alpha]-His-mRXR[gamma] bound
to a DR4 element were excised, and the DNA extracted with TE buffer. The DNA
was used as template in a second PCR reaction, for which again 30 pmol of each
labelled primer was used (25 cycles of 15 s at 94oC, 15 s at 62oC and 15 s at 72oC). After four rounds of selection and amplification a strong gel
shift with the PCR product was observed. DNA from this complex was amplified
and cloned into the
Eco
RI and
Hin
dIII sites of pUC19. Positive clones from the colour selection were picked and
used for preparation of plasmid DNA. DNA sequence analysis was performed with a
M13 universal primer (New England Biolabs, Berverly, USA) and a Pharmacia
sequencing kit (Sollentuna, Sweden).
Synthetic oligonucleotides encoding the different DR4 elements were obtained
from SGS (Ksping, Sweden) and cloned into the
Hin
dIII site of the reporter gene vector pBLcat (
38
). The double-stranded oligonucleotides had the following sequences:
For an extended half-site:
5'
agcttcAGGTCAagcttca
3'
3'
agTCCAGTtcgaagttcga
5'
For AGG TCA atca AGG TC
The sequences of the spacer regions (nnnn) are indicated in the text and
figures. The clone for cTR[alpha] in the expression vector pSG5 was described in reference (
36
), VP16-V3 and VP16-V3-DBD were described in reference (
32
), mRXR[alpha] and [gamma] in pSG5 were gifts from P. Chambon, VP16-hRXR[alpha] and VP16-hTR[beta] were gifts from T. Perlmann.
JEG cells were grown in Dulbecco's modified Eagle's medium (DMEM) with 8% fetal
calf serum and antibiotics. For transfection 2 * 10
5
cells were seeded onto 35 mm dishes and transfected the next day by a calcium phosphate procedure (
25
) with 100 ng expression vector, 500 ng reporter gene vector and 250 ng of a RSV-[beta]-galactosidase control plasmid. The medium was replaced with T
3
depleted medium, or medium containing 25 nM 3,3',5-triiodothyronine (T
3
) 12 h after addition of precipitated DNA. For transfection with the VP16-constructs only T
3
depleted medium was used. Cells were harvested after 24 h and the CAT as well
as the [beta]-galactosidase assays were done using standard protocols (
39
). The induction rates were normalised for the transfection efficiency by using the [beta]-galactosidase activities (
40
). All transfections employed duplicate samples, and were done at least twice.
Representative experiments are shown in the figures.
To characterise an optimal binding site for TR[alpha], we applied a PCR-based selection and amplification approach which would allow
selection of an optimal 3' half-site and a spacer sequence (
41
). An oligonucleotide was synthesised in which one half-site of the AGGTCA motif was present in the 5'-position followed by 12 random base pairs (bp). Since RXR has
been shown to interact exclusively with the 5'-half-site of response elements bound by TR[alpha]-RXR heterodimers, this template was expected to
select for a TR[alpha] specific binding site in the 3'-position. cTR[alpha] and mRXR[gamma] used in the selection were expressed as
histidine-tagged fusion proteins in a Vaccinia virus expression system and purified
as heterodimers by affinity chromatography.
The heterodimers formed an easily detectable complex with the amplified PCR
product after four rounds of selection that co-migrated in a gel retardation assay with a heterodimer complex bound to a
DR4 element. The resulting PCR products were cloned into pUC19 and sequenced.
The sequences from 12 individual clones established a DR4-like binding site containing the consensus sequence atcaAGGTCC 3' to the fixed AGGTCA motif. This sequence differed in the last
position from the AGGTCA half-site previously identified as a core recognition site for TR. However, the
role in receptor binding of the C in position 10 of the PCR selected sequence
remained unclear. We therefore synthesized two oligonucleotides containing the sequences AGGTCAatcaAGGTCC and AGGTCAatcaAGGTCA, respectively, and compared their binding to that of TR[alpha] in a gel retardation assay. The results show that the C in the last
position of the second half-site markedly diminishes TR[alpha] homodimer binding (Fig.
1
), whereas binding of RXR-TR[alpha] heterodimers was unaffected in this assay (data not shown).
The DNA binding data described above showed that DR4 elements with different
spacer sequences had distinct affinities for TR[alpha]. To determine the importance of the sequence of the spacer and to show
the influence of the adenine/cytidine exchange in the last position of the
second half site, oligonucleotides representing the different elements were
cloned into a reporter plasmid and tested in transactivation experiments in JEG
cells. The results (Fig.
3
A) show that the cytidine in the last position strongly reduced hormone
induction as compared to the element with an adenosine. The data also show that
the element with the sequence AGGTCA CTTC AGGTCA yielded a >2-fold higher induction as compared to a similar element containing an ATCA
spacer, whereas the spacer sequence CCAT, representing the `anti-consensus' of the PCR selection, was inactive.
The cotransfection experiments with JEG cells demonstrated the importance of the
nucleotides in the spacer region in DR4 type of elements. To examine whether
the reduced transactivation correlated with receptor binding to these elements,
gel retardation assays were done. Vaccinia virus expressed TR[alpha] was incubated with the elements with low activity (spaced by CTCA and
CCAT), or with the element spaced by CTTC and its derivatives. The result (Fig.
4
A) shows that elements with a reduced activity in the transfection studies had
also a reduced ability to bind to the receptor: nearly no TR[alpha] homodimer binding was detected to elements that gave background levels
of transactivation. Parallel gel retardation experiments (data not shown) with
TR[alpha]-RXR heterodimers showed no or little difference in binding.
Next we investigated if the spacer sequence in DR4 elements also determine transactivation by other receptors. Several orphan receptors have been
shown to heterodimerize with RXR or differently spaced response elements of the
direct repeat type (
42
-
44
,
58
). We therefore tested the transactivation by VP16-RXR chimera from the DR4 elements with AGGTCA half-sites. For this, we compared the VP16-mediated transactivation by the VP16-V3 and the VP16-RXR constructs in JEG cells. The experiments
show (Fig.
5
) that VP16-RXR exhibited a similar but yet distinct selectivity for the spacer
sequence as compared to the control VP16-V3 construct. The AGTG spacer gave the strongest activation with VP16-RXR, VP16-V3 was the most active with the CTTG sequence, whereas the
ATTC and CTCA spacer gave poor activation for both receptor constructs. In addition, the `anti-consensus' CCAT gave an appreciable activation with VP16-RXR and was inactive with VP16-V3.
To understand better the structure and function of TREs, we applied a PCR-based selection and amplification approach (
41
) for identification of an optimal TR[alpha]-RXR binding site. TR[alpha]-RXR selected a direct repeat separated by 4 nt with
an AGGTCC consensus binding site for TR[alpha]. This confirmed the earlier results from a similar selection approach (
21
), that identified a spacer length of 4 nt as specific for TR[alpha]. The AGGTCC sequence found in our selection has also been found in some
natural elements (
18
,
25
) and it is very similar to the AGGTCA motif, which is often found in natural response elements that confer
high induction by TR and related nuclear receptors. We also observed a high
selectivity for a cytidine instead of an adenosine in the last position of the
half-site motif (10 out of 12 clones). DNA binding and cotransfection
experiments with DR4 elements only distinct by an adenosine or a cytidine in
this position showed a clear reduction in binding and transactivation due to
the cytidine.
Obviously, the cytidine in the tested DR4 element (AGGTCA ATCA AGGTCC)
destabilised DNA binding and by that also reduced the transactivation from this
element. The natural TRE from the MoMLV also contains a cytidine in that
position (GGGTCA TTTC AGGTCC) (
59
). Transfection experiments with the VP16-activated chimera VP16-V3 showed for this TRE a 4-fold reduced binding as compared to the similar element CTTC
(AGGTCA CTTC AGGTCA). These data suggest that a cytidine in the last position
of the second half-site is not optimal for TR binding, although, a TRE with an optimal spacer
sequence and an optimal first half-site can be functional as was shown for the MoMLV element. In fact, a
previous report demonstrated receptor interaction with an adenosine in the last
position of the 3' half-site (
45
). We conclude that since the cytidine in the last position of the recognition
motif clearly neither favours DNA-binding, nor transactivation, its selection is likely to reflect
constraints inherent in the PCR-based selection approach.
Using the PCR based selection we detected not only preferences for certain
nucleotides in the second recognition motif but also in the 4 bp separating the
two half-sites. To study further the importance of the specific nucleotides in the
spacer sequence for the binding and transactivation from DR4-like response elements, we changed each nucleotide one by one to an adenosine in the synthetic
CTTC spacer sequence, and tested the activity of the elements in both DNA
binding and transactivation experiments. Both the PCR selection and the DNA
binding experiments showed that adenosines were preferred in position one of
the spacer. However, the high binding affinity was contrasted by a poor
transactivation in JEG cells. The second and the fourth positions in the spacer
were of less importance, since changes of a thymidine or cytidine into an
adenosine did not change, or had only minor effects, on the activity of these
elements in any of the assays used. The results demonstrated, however, a strong
dependence on pyrimidines in position 3. Only pyrimidine nucleotides were found
in position 3 during the PCR selection (with the exception of one clone),
indicating that purines in this position destabilize TR[alpha] binding. Similar conclusions were reached from gel shift experiments in
which a thymidine to adenosine exchange in position 3 dramatically reduced TR
binding
in vitro
. Furthermore, transactivation studies using elements with an adenosine in
position three (CCAT, CTAC and ATAC) showed very low activity, thus supporting our suggestions.
The data from our different binding and transactivation experiments indicate
that TR[alpha] binds probably to an extended half-site motif which overlaps with the spacer region. In previous
publications other groups reported that TR binds as a monomer to a 8 bp binding
site (
46
,
47
). According to these data, a 5'-flanking TA (
46
) or TG (
47
) motif stabilises binding so that these elements can be activated by TR
monomers. In our cotransfection experiments that compare the activity of DR4
elements spaced by CT
Despite our testing of a very large number of elements with different spacer
sequences (Figs
3
and
6
, and data not shown), no simple rule could be established for the preferred
nucleotides. For instance, the sequence ATTC gave lower transactivation as
compared to CTTC, whereas the sequence ATCA was more efficient than CTCA. This
indicates that the structure or overall composition is of importance, possibly
for allowing proper bending of the element. Alternatively, it is possible that
some spacer sequences together with their flanking sequences form a recognition
motif that binds other, interfering transcription or chromatin associated
factors
in vivo
. Such a mechanism may be responsible for the low transactivation from the
element with the selected ATCA spacer.
Several reports on orphan receptors have demonstrated that nucleotides upstream
of the binding site greatly influence receptor binding. Wilson
et al
. (
48
,
49
) showed that the `A-box' domain of NGFI-B, located in between the DNA and ligand binding regions, conferred
specific interaction with adenosines upstream of the AGGTCA half-site. For these reasons we replaced the A-box region of TR[alpha] with the corresponding domain of NGFI-B (data not shown). The chimeric receptor showed
increased transactivation form DR4 elements with a preference for adenosine
rich spacers like NGFI-B, however, it was still able to transactivate efficiently from an element
with a CTTC spacer (data not shown). It is therefore likely that the
specificity of TRs for the spacer sequence is conferred not only by the A-box region. Other experiments characterising NGFI-B demonstrated that amino acids C-terminal to the A-box (the `T-box') supports the A-box in contacting DNA (
49
). However, a second chimeric TR[alpha] with both the T and A-box regions from NGFI-B completely failed to transactivate (data not shown).
Mutations in the T-box of NGFI-B and TR have been shown to inhibit or reduce DNA binding, which may
explain the inactivity of our chimeric receptor (
49
,
21
).
Recent reports have suggested that also the N-terminal domains of nuclear receptors can modulate DNA binding. Mutations
in the N-terminal region of the oncogenic form of TR[alpha], v-erbA, have been shown to reduce the binding to response
elements with a AGG
Our experiments demonstrated that the sequence separating the two half-sites in a directly repeated response element influences the ability of
thyroid hormone receptors to bind to and to transactivate a target gene.
However, the influence of the spacer sequence was diminished by RXR both in the
DNA binding and the transactivation experiments. The transactivation directly
correlated with the DNA binding ability
in vitro
of the involved receptor proteins, suggesting that the spacer sequence also
in vivo
determines receptor binding. Our data also indicate that limiting
concentrations of TR[alpha]-RXR heterodimers in a cell, such as in JEG cells transfected with
only a TR[alpha] expressing plasmid, increases the selectivity for thyroid hormone
response by allowing receptor binding preferentially to high affinity response
elements. In the RXR co-transfected cells the heterodimer concentration would be higher, allowing
receptor heterodimers to bind also to elements with low affinity. Our results
with JEG cells were confirmed by experiments using HeLa cells with the VP16-activated TR[alpha] fusion protein VP16-V3: the VP16-mediated transactivation from the different DR4
elements showed a similar pattern as observed in JEG cells, but the differences
between the elements were less pronounced in the HeLa cells (data not shown). It is possible that the different results obtained with the two cell lines are due to their distinct amounts of RXR (
13
,
53
).
The binding to an extended binding site in some DR4 elements can support
especially the binding by TR[alpha] homodimers. Our
in vitro
binding
studies and the transfection experiments with TR[alpha] showed good correlation between DNA binding
in vitro
and ligand-dependent transactivation in TR[alpha] transfected JEG cells. In these experiments high transactivation
from the CTTC element was already maintained in none RXR transfected cells and cotransfection of additional RXR did not increase the transactivation from this element (Fig.
3
). Since JEG cells contain endogenous RXR, however, we could not demonstrate
that transactivation from this element involved TR[alpha] homodimers
in vivo
. In addition, we and others have shown earlier, that TR[alpha] homodimers are not stable under ligand treatment on DR4 elements (
54
-
56
). The disruption of the homodimeric complexes was also observed in gel
retardation assays using a strong TR[alpha] homodimer binding element like CTTC (data not shown). The instability of
TR[alpha] homodimers in the presence of ligand, however, does not exclude a
biological function of TR[alpha] homodimers
interacting with strong binding sites
in vivo
. Under RXR limiting conditions, TR[alpha] homodimers could function as repressors on such response elements, which
are relieved from their response elements in presence of ligand. Such a
mechanism was demonstrated for TR mediated repression and activation on an IR0
element in an
in vitro
transcription system (
57
).
Some elements, for instance the one with the spacer CCAT, were unable to allow
TR-mediated transactivation despite appreciable TR[alpha] or TR[alpha]-RXR binding
in vitro
. This indicates that even in the presence of RXR some DR4 elements cannot
function as response elements for TR[alpha]. Indeed, our experiments that showed high transactivation with VP16-RXR chimeric protein via DR4 elements with the CCAT spacer,
suggests that other orphan receptors heterodimerize with RXR on this type of
elements. While orphan receptors like rev-erbA and ROR bind to their response elements as monomers, two recently
described orphan receptors were shown to bind as heterodimers with RXR to DR4
elements (
42
,
44
,
58
). However, additional experiments have shown that the factors in the JEG cells
that heterodimerize with RXR are distinct from the LXR and OR orphan receptors
(
32
). Taken together, the data suggest that receptors other than TR that
heterodimerize with RXR on DR4 elements, exhibit a preference for the spacer
sequence that is similar to but distinct from that of TR.
We are grateful to Dr Pierre Chambon for anti-RXR antibodies as well as cDNAs encoding mouse RXR[alpha] and [gamma], and to Dr Thomas Perlmann for the gift of expression
vectors for VP16-hRXR[alpha] and VP16-hTR[beta]. We would also like to thank Drs Thomas Perlmann and
Marco Tini for critically reading the manuscript, and Dr Martin Privalsky for
sharing data prior to publication. MH was supported by a grant from the
Deutsche Forschungsgemeinschaft; GW was supported from the Swedish Society for
Medical Research. This project was funded by the Swedish Cancer Society, The
Beijer Foundation, Gsran Gustafsson's Foundation, The Swedish Society for Medical Research, Lars Hierta Foundation and funds at the Karolinska Institute.
REFERENCES
Return