Signalling by small molecules, such as retinoic acid, is mediated by heterodimers comprising a class II nuclear receptor and an RXR (retinoid X receptor) subunit. The receptors bind to DNA response elements and act as ligand-dependent transcription factors, but, in the absence of signal, the receptors bind the co-repressors SMRT [silencing mediator for RAR (retinoic acid receptor) and TR (thyroid hormone receptor)] and NCoR (nuclear receptor co-repressor) and repress gene expression. Alternative splicing of the SMRT transcript in mammals generates six isoforms containing 1, 2 or 3 CoRNR (co-repressor for nuclear receptor) box motifs which are responsible for the interactions with nuclear receptors. We show that human cell lines express all six SMRT isoforms and then determine the binding affinity of mouse SMRT isoforms for RAR/RXR and three additional class II nuclear receptor–DNA complexes. This approach demonstrates the importance of the full complement of CoRNR boxes within each SMRT protein, rather than the identity of individual CoRNR boxes, in directing the interaction of SMRT with nuclear receptors. Each class of SMRT isoform displays a distinct feature, as the 1-box isoform discriminates between DNA response elements, the 2-box isoforms promote high-affinity binding to TR complexes and the 3-box isoforms show differential binding to nuclear receptors. Consequently, the differential deployment of SMRT isoforms observed in vivo could significantly expand the regulatory capacity of nuclear receptor signalling.
The class II NRs (nuclear receptors) mediate signalling by small molecule ligands, such as thyroid hormone and retinoic acid [1,2]. They act as heterodimers, comprising a class II receptor and an RXR (retinoid X receptor) subunit. The heterodimer binds a DNA element, adjacent to responsive genes, that comprises a 6-bp DR (direct repeat) which is separated by a variable number of nucleotides. RAR (retinoic acid receptor) response elements, for example, can be DR-5, in which the DRs are separated by 5-bp, DR-2 or DR-1, whereas TR (thyroid hormone receptor) response elements are usually organized as a DR-4 [3,4]. Binding of the small molecule ligand to its receptor promotes association with co-activators, but, in the absence of ligand, the co-activator dissociates to be replaced by a co-repressor [5,6]. The co-repressor paralogues SMRT (silencing mediator for RAR and TR) and NCoR (nuclear receptor co-repressor) [7–9] can each assemble a multicomponent repressor complex that includes histone deacetylases [10,11], converting chromatin into a closed inactive state . The SMRT/NCoR N-terminal region assembles the repressor complex [7–9], whereas the C-terminal region encodes two domains that interact with NRs. The interaction domains include three motifs, or CoRNR (co-repressor for nuclear receptor) boxes, with the consensus sequence (I/L)XX(I/H)IXXXI [13–15]. SMRT and NCoR interact with a wide range of class II NRs, an unusual ability that is even more remarkable given that the NRs bind DR DNA response elements, varying in the separation from 1 to 5 bp, which will alter the arrangement of the NRs around the DNA helix.
In previous studies, short peptides corresponding to individual CoRNR boxes have been used to demonstrate and quantify the different affinities of the CoRNR boxes for specific NRs [15–17]. However, an additional complexity is provided by alternative splicing of the primary transcript which generates multiple co-repressor isoforms [9,18–22]. For NCoR, one outcome is the generation of RIP13Δ1 (receptor interacting protein Δ1), an isoform lacking part of the C-terminal interaction region , whereas the alternative splicing of SMRT generates 16 C-terminal isoforms in Xenopus and six in humans and mice . The exclusion of exon 45b varies the peptide sequence adjacent to the most C-terminal CoRNR box, whereas the complete exclusion of exon 45 removes the entire C-terminal CoRNR box. The alternative splicing of exon 38b regulates the inclusion of another CoRNR box and, consequently, mouse and human cells contain three classes of SMRT isoform that contain three, two or only one CoRNR box .
To determine whether the diversity of SMRT isoforms contributes to its function as a co-repressor, it will be necessary to resolve a number of questions. The first is whether SMRT isoform expression is regulated. We have previously shown that this is the case, as the alternative splicing of SMRT in Xenopus generates isoforms expressed in developmental and tissue-specific combinations . The second question is whether the isoforms can perform specific roles in vivo. This has been addressed using an antisense oligonucleotide to eliminate the inclusion of exon 38b and therefore restrict the range of available SMRT isoforms in Xenopus. The resulting embryos have head and neuronal defects accompanied by changes in the expression levels of thyroid hormone responsive genes . A remaining question is whether the isoforms differ in their affinity for NRs and in the present study we have analysed the interactions of mouse SMRT isoforms with a range of DNA-bound class II NRs in vitro. The results suggest that the complement of CoRNR box motifs within the C-terminal interaction domains plays a significant role in the selectivity of SMRT–NR interactions. These findings are likely to impact on the regulatory capacity of NR signalling.
MATERIALS AND METHODS
Generation of SMRT ID (interaction domain) region isoform clones
cDNA derived from mouse brain was amplified with primers for exons 38 and 46, corresponding to amino acid residues 2023–2426 of mouse SMRT (GenBank® accession number NM_011424), to generate six bands. These were cloned into TA vectors (Invitrogen) and the inserts transferred into pGEX-4T-2 (GE Healthcare) to produce N-terminal GST (glutathione transferase)-fusion proteins (Figures 1A and 1B).
SMRT C-terminal isoform expression
in vivo and in vitro
Expression and purification of SMRT and NR proteins
Protein expression was induced by 1 mM IPTG (isopropyl β-D-thiogalactoside) and cell pellets lysed by sonication in 10 mM NaCl, 1 mM EDTA and 10 mM Tris/HCl (pH 8.2) (2-box and 3-box SMRT isoforms), or 10 mM sodium phosphate buffer (pH 7.4), 140 mM NaCl and 2.7 mM KCl (1-box SMRT isoform). Proteins were purified from the supernatant on the AKTA system by a HiTrap heparin HP column (GE Healthcare), followed by SP sepharose HP ion-exchange chromatography. The concentration of each protein was determined by UV spectroscopy using a molar absorption coefficient derived from each protein sequence using the application ProtParam (http://www.expasy.ch/tools/protparam.html), and the purity of the protein was analysed by SDS/polyacrylamide gel.
The coding region of clones for mouse TRα, RXRβ, VDR (vitamin D receptor) and PPARγ (peroxisome-proliferator-activated receptor γ) (from Professor Ronald M. Evans, Gene Expression Laboratory, Salk Institute, La Jolla, CA, U.S.A.) and RARα (retinoic acid receptor α) (Professor Pierre Chambon, Institut de Génétique et de Biologie Moléculaire et Cellulaire, Illkirch, Strasbourg, France) were transferred on to pET-21b (Novagen) and the proteins purified by HisTrap™ chromatography.
Generation of response-element probes
Equimolar amounts of complementary oligonucleotides were annealed by heating to 90°C for 15 min in 10 mM Tris/HCl (pH 8.2), 150 mM NaCl and 10 mM MgCl2, cooled to room temperature (23 °C) and then stored overnight at 4 °C. The short DNA duplexes were isolated with a Superdex 75 10/30 column to generate probes containing the DR-5, -4, -3, -2 and -1 response elements corresponding to the elements found in characterized genes: DR-5, 5′-GGGTAGGGTTCACCGAAAGTTC- CTCGCCATGAA-3′, βRAR ; DR-4, 5′-GATCCTACTAT- AGGTCACATGAGGTCAAGTTAC-3′, CMHCα (cardiac specific myosin heavy chain α) ; DR-3, 5′-GGGTAGAGG- TCA-GGAAGGTCACTCGCCATG-3′, osteocalcin ; DR-2, 5′-AGCTTGGGGGTCAGCAGGTCAGCTTGCA-3′, FGF8 (fibroblast growth factor 8) ; DR-1, 5′-TTCCGTACAG- GTCACAGGTCACTCGAGATAT-3′, acylCoA oxidase  (response elements are in bold). One oligonucleotide of each pair was supplied labelled with the fluorescent tag hexachlorofluorescein (Eurogentec).
EMSA (electrophoretic mobility-shift assay) gels and analysis
Each class II NR was combined on ice with bacterially expressed RXRβ in 15 mM Tris/HCl (pH 7.6), 200 mM KCl, 2 mM dithiothreitol, 1 mM EDTA and 4.5% (v/v) glycerol, before the addition of 1 ng of hexachlorofluorescein-labelled DNA probe. The complex was allowed to form for 15 min at room temperature before the addition of SMRT isoforms for 15 min at 4 °C. The samples were resolved on a 7% polyacrylamide gel in 0.5×Tris/borate/EDTA (1×TBE=45 mM Tris/borate and 1 mM EDTA) at 4°C, at 100 V for 60 min, and the fluorescence was quantified with a Fujifilm FLA5000 phosphorimager using ImageGuage software.
Each experiment was performed in triplicate and the ‘fraction bound’ values plotted using GraphPad Prism (v4.0) software (GraphPad Software, San Diego, CA, U.S.A.). For each assay the points fitted to a non-linear regression curve describing the binding of a ligand to a receptor where:
in which Y is the fraction of NR–DNA complex bound, X is the molar concentration of the SMRT isoform, and Bmax, the maximal binding, is fixed at 1.0. Equal weighting was given to each point. The apparent Kd was taken to be the concentration of SMRT required for half-maximal binding, and the apparent Ka is the inverse of Kd. Statistically significant differences in Ka were calculated using a pair-wise Students' t test. This approach is likely to provide a simplified version of the actual binding model, but provides a means of comparing the interactions between SMRT isoforms and NR heterodimers.
Note on the nomenclature of SMRT isoforms
Both human and mouse SMRT comprise a core of 47 coding exons. The isoforms are named systematically after the missing exons. This annotation of the SMRT gene results in a one exon shift in numbering, exons 37 and 44 in Malartre et al.  are now exons 38 and 45. The isoform in which all exons are retained is denoted SMRTct.
Human cell lines exhibit different SMRT isoform profiles
RT-PCR (reverse transcription-PCR) using primers specific against the 3′-coding region of the SMRT transcript detected six isoforms in mouse brain (Figure 1A), corresponding to the alternative splicing of exons 38b, 45b and 45 [19,21,22] (Figure 1B). Equivalent primers used with cDNA from four human tumour cell lines showed quantitative differences in SMRT isoform expression (Figure 1A). The six mouse SMRT isoforms, comprising a single 1-box isoform, three 2-box isoforms and two 3-box isoforms, were cloned and the bacterially expressed and purified proteins (Figure 1C) used for subsequent analyses.
SMRT isoforms show differential binding to RAR–RXR complexes on DR-5, DR-2 and DR-1 response elements
NRs bind to DRs of the consensus A/GGGTCA. In these experiments, we used probes derived from the characterized response elements of identified responsive genes (see the Materials and Methods section). On both DR-5 and DR-2, the RXR component binds to the upstream and RAR to the downstream half of the bipartite response element, but on a DR-1 response element the order is reversed . To analyse the contribution of response-element spacing and NR orientation to binding, the affinity of each of the six SMRT isoforms for RAR–RXR on DR-5, DR-2 and DR-1 response elements was assessed by EMSA (Figure 2A).
Binding of SMRT isoforms with RARα–RXRβ heterodimers on a DR-5 response element
The 1-box SMRT isoform (Δ38b/45) binds the RARα–RXRβ DR-5 complex with a Ka of 3.9 μM−1 (Figure 2B). This is comparable with 2-box SMRT Δ38b, which shows a standard binding curve with a Ka of 5.8 μM−1 (Figure 2B). In contrast, the 2-box SMRT isoforms Δ38b/45b and Δ45 show sigmoidal binding that may indicate co-operative binding to the RARα–RXRβ DR-5 complex (Figure 2B). Compared with both 1-box and 2-box SMRT isoforms, the affinity of 3-box isoforms for the complex is decreased by an order of magnitude (Ka=0.1 μM−1 and 0.4 μM−1) (Figure 2B).
On DR-2 and DR-1 response elements the 2-box isoforms show the greatest affinity for RARα–RXRβ complexes (Figure 3), whereas 3-box isoforms bind only weakly. The major difference is for 1-box SMRT (Δ38b/45), which binds the RARα–RXRβ DR-5 response-element complex approx. 5-fold more than it binds the RARα–RXRβ DR-2 or DR-1 complexes (Figure 3). As RARα occupies the downstream half-site on a DR-2, but the upstream half-site on a DR-1 sequence, it suggests that it is the spacing of the response element, rather than NR order, that determines interaction with the 1-box SMRT isoform.
Comparison of the binding constants for the SMRT isoforms on DR-5, DR-2 and DR-1 response elements
Binding of SMRT isoforms to a range of class II NR–DNA complexes
Although the general structure of the SMRT-binding site is conserved across the NR family , the NRs clearly differ in their ability to interact with individual CoRNR box peptides . However, the affinities of the complete interaction domains of the six SMRT isoforms for a range of NRs have yet to be compared systematically. We therefore extended the EMSA analysis to three additional class II NRs, TRα, VDR and PPARγ, each combined with RXRβ and added to the DR-4, DR-3 and DR-1 response elements respectively.
The 1-box SMRT isoform binds both the TRα–RXRβ DR-4 and the VDR–RXRβ DR-3 complexes with an apparent Ka greater than 3 μM−1, an affinity comparable with that determined for the RARα–RXRβ DR-5 complex (Figure 4A). In contrast, the affinity for the PPARγ–RXRβ DR-1 complex was less than 1 μM−1, which is in agreement with the suggestion that spacing of the repeat element half sites is a major determinant of the 1-box SMRT interaction with NR–DNA complexes (Figure 4A).
Comparison of the binding of SMRT isoforms to type II NRs
All 2-box SMRT isoforms bind most strongly to TRα–RXRβ heterodimers (Figure 4B). In addition, the presence of exon 45b in isoform Δ38b causes an almost 3-fold increase in the affinity of SMRT for the TRα–RXRβ–DNA complex. A similar result has been observed previously for the binding of 2-box SMRT isoforms to TRβ–RXRα . However, this sequence does not generally increase the affinity of 2-box SMRT for a NR–DNA complex, since isoforms lacking exon 45b (Δ38b/45b and Δ45) show equal or even higher affinity than SMRT Δ38b for the other NR combinations tested (Figure 4B).
Compared with the 2-box and 1-box SMRT isoforms, the 3-box isoforms bind NR–DNA complexes weakly, with two exceptions. In the first, the SMRTct isoform preferentially binds a TRα–RXRβ complex, and, in the second, the Δ45b isoform preferentially binds a VDR–RXRβ complex (Figure 4C). This demonstrates the differential activity of two splice isoforms that share identical CoRNR boxes. Together with the data from the 2-box SMRT isoforms, this suggests that the exon 45b encoded peptide includes sequences that specifically promote interaction with TRα.
Alternative splicing often generates multiple proteins from a single gene . A current task is to understand the contribution that isoforms make to the efficient working of an organism. Ultimately, this involves generating a catalogue of isoforms, detailing their biochemical properties and addressing their biological function. SMRT is a good candidate for this type of analysis, as it is composed of discrete regions that can be analysed in isolation. We have characterized isoforms of the C-terminus, the region that interacts with NRs.
Individual CoRNR box peptides exhibit differential binding to NRs [13,14,16,23,33,34]. Cohen et al.  showed that the centrally located SMRT CoRNR 2-box, but not NCoR CoRNR 2-box, preferentially binds RARα, suggesting that co-repressor paralogues target different NR complexes. In other studies on NCoR, Hu et al.  found that CoRNR 2-box binds strongly to RAR and that CoRNR 1-box binds RXR and TRβ. In contrast Webb et al.  demonstrated strong binding of NCoR CoRNR 3-box to TRβ and suggested that the presence of the third box determines the specificity of NCoR for TRβ. The subsequent finding that SMRT also contains a third CoRNR box  argues against this conclusion. It has also been suggested that CoRNR box 1 interacts preferentially with RXR, whereas CoRNR 2- and 3-boxes confer specificity for the class II NR in the heterodimer [6,16,18]. However, the 2-box SMRT isoform, Δ45, lacking CoRNR 1-box, still binds TRα–RXRβ heterodimers with high affinity, indicating that CoRNR 1-box is not essential for the binding of SMRT to class II/RXR heterodimers. Although individual CoRNR boxes clearly show differential binding, less is known about how the complete SMRT ID region interacts with NRs.
A comparison of 1-, 2- and 3-box SMRT isoforms shows that their binding characteristics are determined by different features of the NR–response-element complex
The 1-box isoform binds to NR complexes on the DR-5, DR-4 and DR-3 response elements, but shows reduced affinity when these are on the DR-2 and DR-1 response elements. Although the repeats of the response element are nominally closer on a DR-1 than a DR-5, the helical twist of DNA places the midpoints of each repeat on the same side of the DNA helix in a DR-5, but on almost opposite sides in a DR-1 response element (Figure 4A). Retinoid signalling is unusual in that it can activate gene expression through both DR-5 and DR-2 response elements. The selective repression of the DR-5 promoters by the 1-box SMRT isoform may provide an additional means to distinguish between genes regulated by DR-5 and DR-2 response elements in tissues, such as muscle, in which the 1-box SMRT isoform is the predominant co-repressor .
The 2-box SMRT isoforms bind most strongly to the TRα–RXRβ heterodimers and with lower affinity to the other NR–DNA complexes investigated (Figure 4B). The depletion of 3-box SMRT isoforms in Xenopus embryos causes increased repression of thyroid hormone responsive genes , and this is consistent with the replacement of low-affinity 3-box isoforms with 2-box SMRT co-repressor isoforms that have high affinity for TRα–RXRβ.
The 3-box SMRT isoforms bind NR–DNA complexes with the least affinity. This is somewhat surprising as the addition of the third box reduces the affinity of the 2-box isoform, suggesting that it either alters the conformation of SMRT or competition between boxes destabilizes binding.
Having quantified differences in the binding of SMRT isoforms to NR–DNA complexes, this information will now provide a framework for further functional experiments to determine the relationship between binding affinity in vitro and co-repressor activity in vivo.
In addition to class II NR heterodimers, SMRT also binds class I steroid hormone homodimers , and it will be interesting to compare the binding activity of SMRT isoforms in this context. This may have clinical significance, given the association between SMRT expression and poor patient outcome in breast cancer . SMRT binding is not limited to NRs, and interaction sites for a number of other transcription factors, including Oct-1 and NF-κB (nuclear factor κB), also map to the C-terminal region . Nuclear co-repressor diversity, both through the expression of NCoR and SMRT and alternatively spliced SMRT isoforms, may significantly expand the capacity to regulate NR signalling.
co-repressor for nuclear receptor
electrophoretic mobility-shift assay
nuclear receptor co-repressor
peroxisome-proliferator-activated receptor γ
retinoic acid receptor
retinoid X receptor
silencing mediator for retinoic acid receptor and thyroid hormone receptor
thyroid hormone receptor
vitamin D receptor
We thank Professor Ronald Evans, Professor Pierre Chambon and Professor Malcolm Maden (Developmental Neurobiology, MRC Centre for Biomedical and Health Sciences, King's College London, London, U.K.) for clones. We are grateful to Matt Guille, Marianne Malartre and Sarah Thresh for informative comments, and members of the Kneale and Sharpe laboratories for advice and assistance.
F. F. was a University of Portsmouth, Institute of Biomolecular and Biomedical Sciences Bursary PhD student; S.S. was funded by a Biotechnology and Biological Sciences Research Council Quota studentship.
Present address: Astellas Pharma, 114 rue Victor Hugo, F-92686, Levallois-Perret Cedex, France.
Present address: Marine Biology Research Division, Scripps Institution of Oceanography, La Jolla, CA 92093-0202, U.S.A.