An Extended LXXLL Motif Sequence Determines the Nuclear Receptor Binding Specificity of TRAP220*

The interaction of coactivators with the ligand-binding domain of nuclear receptors (NRs) is mediated by amphipathic α-helices containing the signature motif LXXLL. TRAP220 contains two LXXLL motifs (LXM1 and LXM2) that are required for its interaction with NRs. Here we show that the nuclear receptor interaction domain (NID) of TRAP220 interacts weakly with Class I NRs. In contrast, SRC1 NID binds strongly to both Class I and Class II NRs. Interaction assays using nine amino acid LXXLL core motifs derived from SRC1 and TRAP220 revealed no discriminatory NR binding preferences. However, an extended LXM1 sequence containing amino acids −4 to +9, (where the first conserved leucine is +1) showed selective binding to thyroid hormone receptor and reduced binding to estrogen receptor. Replacement of either TRAP220 LXXLL motif with the corresponding 13 amino acids of SRC1 LXM2 strongly enhanced the interaction of the TRAP220 NID with the estrogen receptor. Mutational analysis revealed combinatorial effects of the LXM1 core and flanking sequences in the determination of the NR binding specificity of the TRAP220 NID. In contrast, a mutation that increased the spacing between TRAP220 LXM1 and LXM2 had little effect on the binding properties of this domain. Thus, a 13-amino acid sequence comprising an extended LXXLL motif acts as the key determinant of the NR binding specificity of TRAP220. Finally, we show that the NR binding specificity of full-length TRAP220 can be altered by swapping extended LXM sequences.

The nuclear hormone receptors (NRs) 1 are a family of structurally related, ligand-regulated transcription factors that exert both positive and negative control of gene expression in metazoans (1). The NRs are subdivided into classes depending on their DNA binding and dimerization properties. Class I comprises the steroid hormone receptors, including the estrogen (ER), androgen (AR), and progesterone (PR) receptors, which function as homodimers. The largest group (Class II) function as heterodimers with the 9-cis retinoic acid receptor (RXR), and includes receptors for retinoic acid (RAR), vitamin D (VDR), thyroid hormone (TR), and peroxisome proliferators (PPARs). Binding of cognate ligand induces a conformational change in the ligand binding domain (LBD) of NRs, which influences their function with respect to subcellular localization, dimerization, cofactor binding, and transcriptional activity (2).
A wide range of NR cofactors have been identified which perform distinct functions at target promoters, including chromatin modification and remodeling and recruitment of the RNA polymerase II holoenzyme (3). A number of cofactors have been shown to bind the LBD directly via short amphipathic ␣-helices containing the LXXLL sequence (4,5). Structural studies have demonstrated that a hydrophobic channel (AF2) is exposed on the surface of the LBD as a consequence of ligand binding (6 -11). This channel accommodates the LXXLL ␣-helix, which is held in place by hydrophobic interactions and a charged clamp involving two amino acids (lysine and glutamate) that are conserved throughout the NR family (12,13). The minimal sequence that can bind the AF2 surface (the LXXLL core motif) is contained within 8 amino acids (Ϫ1 to ϩ7) (14).
Different cofactors have been shown to have variable numbers of functional LXXLL motifs. The p160 coactivators SRC1, TIF2/GRIP1, and ACTR/AIB1/pCIP have homologous NR interaction domains (NID) containing three LXXLL motifs (4,5,15). These motifs are highly conserved both in sequence and spacing, and it has been shown that at least two (preferably adjacent) motifs are required for high affinity binding of SRC1 to Class I receptors (4,5,15,16). Thus, the presence of multiple LXXLL motifs in coactivators facilitates cooperative binding to the AF2 surfaces of NR dimers. The mammalian mediator complex TRAP/DRIP/SMCC/ARC/CRSP contains a single subunit capable of binding to NR LBDs (17,18). This protein, termed TRAP220, DRIP205, or PBP, contains two LXXLL motifs within the NID that are required for ligand-dependent binding to NRs (19 -21). Other cofactors including TIP60 (22), TIF1␣ (23), PGC1 (24), Fushi tarazu (25), and ASC-2/RAP250/ TRBP/PRIP/NRC/AIB3 (26), contain a single functional LXXLL motif, although the stoichiometry of binding of these proteins to NR dimers has yet to be determined. The transcriptional repressor RIP140 contains nine functional LXXLL motifs (4), and a similar number have been identified in the ER␣-binding protein PELP1 (27).
Several cofactors have been found to display preferences for NR subclasses. For example, TIP60 has been reported to bind the Class I receptors, but displayed little interaction with VDR, TR, or RXR (22). The p160s interact with a wide range of NRs, although individual LXXLL motifs derived from these proteins display differential binding to NRs (14,38). In addition, we have shown that LXXLL sequences derived from CBP and RIP140 show selectivity that is at least partly determined by the LXXLL core sequences (14). Similarly, LXXLL motifs (or variants such as LXXIL and FXXLL) from other cofactors, such as PERC (31), NRIF3 (32), and NSD1 (33) have been reported to display selectivity in their interactions with NRs. TRAP220 was isolated on the basis of its interaction with Class II NRs such as TR, VDR, and PPARs, and it has been shown that ablation of either of its two LXXLL motifs had differential effects on binding to NRs. RXR binding was disrupted by mutation of LXM1, whereas TR, RAR, and VDR showed a preference for binding the distal motif, LXM2 (20,28). TRAP220 has also been reported to interact with ER␣, although this appears to be relatively weak (18,19,29,30). Recent reports have demonstrated that TRAP220 interacts more strongly with ER␤ (30,50). In addition, data from other studies have suggested that sequences flanking the core motif also influence NR binding specificity (30, 34 -39).
In this study we assessed the interaction of TRAP220 NID, LXXLL peptides, and full-length proteins with a panel of NR LBDs. We show that TRAP220 displays stronger interactions with certain Class II receptors in comparison to steroid receptors. We describe mutational analyses and mapping experiments that shed light on the molecular basis of the NR binding specificity of TRAP220.

MATERIALS AND METHODS
Plasmid Constructions-The following plasmids used in transient transfection experiments have been described previously: pSG5-SRC1e (16) and p3ERE-TATA-LUC (40). The reporter plasmid pMAL-TK-LUC (41) and hTR␤-RSV (42) were gifts from K. Chatterjee. The expression construct pCIN4-TRAP220 was a gift from R. Roeder. A cDNA-encoding full-length TRAP220 with an N-terminal HA tag was subcloned into the modified vector pSG5(PT) using unique XmaI and NotI sites.
GST Pull-down Assays-Recombinant cDNAs in pSG5 or pSG5(PT) expression vectors were transcribed and translated in vitro in the presence of [ 35 S]methionine in reticulocyte lysate (Promega) according to the manufacturer's instructions. GST fusion proteins were expressed in Escherichia coli DH5␣ using isopropyl-␤-D-thiogalactopyranoside induction, purified on glutathione-Sepharose beads (Amersham Biosciences), and normalized amounts were incubated with 35 S-radiolabeled protein in NETN buffer (20 mM Tris, pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, 10 mM dithiothreitol) containing complete protease inhibitors (Roche Molecular Biochemicals) in the presence or absence of 10 Ϫ6 M cognate ligand (E 2 , 9c-RA or T 3 ), as described previously (16). Samples were washed three times, and bound proteins were separated on SDS-polyacrylamide gels. Radiolabeled proteins were visualized by autoradiography.

TRAP220 Enhances the Transcriptional Activity of TR␤ but
Not ER␣-Previous studies have shown that ectopic expression of the mammalian mediator complex protein TRAP220/ DRIP205/PBP modestly enhances TR␣/␤, VDR, and PPAR␥mediated transcription in a ligand-dependent manner (19 -21). We assessed the ability of TRAP220 to stimulate TR␤-mediated activation of the T 3 -responsive reporter gene, MAL-TK-LUC. This reporter contains a single thyroid response element (TRE) consisting of a direct repeat spaced by four nucleotides (DR4) within a thymidine kinase promoter and driving a firefly luciferase gene (41). As shown in Fig. 1A, increasing amounts of transiently transfected TRAP220 expression vector enhanced TR␤-mediated transcription of the MAL-TK-LUC reporter ϳ2fold in the presence of ligand. This is comparable to previously observed levels of TRAP220 enhancement of TR␣-driven expression of T 3 -responsive reporters (19). Under similar conditions, the p160 nuclear receptor coactivator, SRC1e, also modestly enhanced TR␤ activation of MAL-TK-LUC (Fig. 1A). Coexpression with RXR␣ did not significantly increase the levels of reporter activity achieved in the presence or absence of coactivators (data not shown).
We also assessed the ability of TRAP220 to enhance the activity of the Class I receptor ER␣. As shown in Fig. 1B, TRAP220 did not enhance ER␣-mediated transcription of the 3ERE-TATA-LUC reporter gene (40), under similar transfection conditions to those used for TR␤. In contrast, co-transfection of SRC1e strongly enhanced ER␣-mediated transcription both in the presence and absence of ligand, as reported previously (4,16,44,45). The expression and nuclear localization of both TRAP220 and p160 coactivators in transiently transfected cells was confirmed by indirect immunofluorescence using anti-HA antibodies (data not shown).
TRAP220 NID Exhibits Nuclear Receptor Binding Specificity-The ability of TRAP220 to enhance transcription by TR␤ but not ER␣ suggested that TRAP220 exhibits NR binding preferences. We therefore decided to compare the ability of SRC1 and TRAP220 NIDs to bind to a panel of NR LBDs in yeast two-hybrid experiments. Fusion proteins (shown schematically in Fig. 2), consisting of the LexA-DNA binding domain (DBD) fused to amino acids 431-761 of SRC1 encompassing LXM1, LXM2, and LXM3 (SRC1 NID; Fig. 2A), or amino acids 335-667 of TRAP220, encompassing LXM1 and LXM2 (TRAP220 NID; Fig. 2B), were assessed for binding to AAD-NR-LBD fusion proteins. Western blots using an antibody recognizing LexA confirmed that these constructs were expressed to similar levels in L40 (see Fig. 6B), and neither LexA-SRC1 NID nor LexA-TRAP220 NID was able to activate transcription of the reporter either alone or in the presence of VP16-AAD (data not shown). As shown in Fig. 2A, strong ligand-dependent reporter activation was observed between SRC1 NID and the LBDs of the PR, ER␣, RAR␣, RXR␣, and full-length TR␤, (ranging from 80 -160 units of reporter activity) and also a weaker but significant interaction with the AR LBD (25 units reporter activity). We also observed a significant binding to retinoid receptors in the absence of ligand (50 and 30 units of reporter activity for RAR␣ and RXR␣, respectively). The TRAP220 NID fusion protein also displayed strong ligand-dependent reporter activation when co-expressed with RXR␣ LBD, TR␤, and PPAR␥ LBD (140, 110 and 300 units of reporter activity respectively, Fig. 2B), and an intermediate level of reporter activation when co-expressed with RAR␣-LBD (40 units of reporter activity). Although no ligand-independent interactions were observed between TRAP220 NID and retinoid or thyroid receptors, the reporter activity indicated a strong interaction with PPAR␥-LBD in the absence of exogenous ligand (120 units of reporter activity), consistent with our previous observation that yeast cells may contain endogenous ligands for PPAR␥. 2 In contrast to SRC1, the TRAP220 NID showed a greatly reduced ability to interact with Class I receptors, showing relatively weak ligand-dependent reporter activities when co-expressed with ER␣ LBD (20 units), PR LBD (3 units), and no detectable interaction with AR LBD. Thus, our results indicate that while the SRC1 NID interacts strongly with both steroid and Class II NRs, the TRAP220 NID shows a marked preference for binding to the Class II NRs such as TR␤, RXR␣, and PPAR␥.
To determine whether both LXXLL motifs of TRAP220 are required for binding the panel of LBDs used in this study, we generated LexA-TRAP220 NID constructs in which either LXM1 or LXM2 was mutated, by replacing the leucines ϩ4 and ϩ5 with alanines. As shown in Fig. 2C, interaction of the TRAP220 NID with TR␤ and PPAR␥ LBD was dependent on the presence of two functional LXXLL motifs, as mutation of either motif resulted in a significant reduction in reporter activation. Similarly, the weak interaction with ER␣ LBD was abrogated by mutation of either LXM1 or LXM2, indicating a requirement for both motifs. In contrast, interaction of TRAP220 NID with RXR␣ LBD was severely reduced by mutation of LXM1 but largely unaffected by mutation of LXM2. This indicates that the RXR␣ LBD preferentially interacts with LXM1, consistent with a previous study that showed selective binding of RXR␣ LBD (albeit weak) to TRAP220 LXM1 in GST pull-down assays (28).
The NID and Core LXXLL Motifs of TRAP220 Exhibit Distinct NR Binding Specificities-It has been shown previously that 8 -10 amino acid sequences encompassing the signature motif LXXLL are sufficient to bind to liganded NR LBDs, and this has been referred to as the core LXXLL motif (4,14). To explore the nature of the specificity exhibited by the TRAP220 NID in more detail, we assessed the ability of LXXLL core motifs derived from SRC1 and TRAP220 to bind NRs. We generated ER-DBD fusion proteins containing sequences corresponding to amino acids Ϫ1 to ϩ8 of TRAP220 LXM1 and LXM2, and SRC1 LXM2 (Fig. 3A) and confirmed their expression in W303-1b cells by Western blots (Fig. 3B). As shown in Fig. 3C, TRAP220 LXM1 core motif was able to bind all NR LBDs tested in a ligand-inducible manner. Notably, similar levels of reporter activation were achieved due to interactions with Class I and Class II NRs, with the exception of the RXR␣ LBD, for which the reporter activity was consistently 2-3-fold higher. This potentially reflects the preference of RXR␣ for TRAP220 LXM1 shown here (Fig. 2C) and in other studies (28). Similarly, the TRAP220 LXM2 core motif was also able to bind all NR LBDs tested in a ligand-dependent manner. Binding of TRAP220 LXM2 to the Class I receptors ER␣ and PR was 2-4-fold greater than binding to AR and the Class II receptors TR␤, PPAR␥, RAR␣, and RXR␣. In contrast to TRAP220 LXM1, LXM2 did not exhibit any enhanced ligand-induced interaction with RXR␣. As expected and shown for comparison, the SRC1 LXM2 core motif induced high levels of ligand-de-2 V. H. Coulthard and D. Heery, unpublished results.

FIG. 1. Effects of ectopically expressed TRAP220 on ligandinduced reporter activation by TR␤ and ER␣.
A, HeLa cells were transiently transfected as described under "Materials and Methods" with pMAL-TK-LUC reporter plasmid (2 g) and expression vectors for TR␤ (200 ng), SRC1e (500 ng), or TRAP220 (2, 3, and 4 g) as indicated. Luciferase activity was measured 24 h post addition of ligand (10 Ϫ7 M T 3 , black bars) or vehicle (white bars), and the data were normalized to a ␤-galactosidase internal control. Reporter activity was expressed relative to that obtained for reporter in the absence of ligand (set at 1). The data represent the mean of triplicate samples, and error bars are shown to indicate S.D. Similar results were obtained in replicate experiments. B, reporter assays as in A except that HeLa cells were transiently co-transfected with the p3ERE-TATA-LUC reporter (1 g), together with expression vectors for ER␣ (100 ng), SRC1e (500 ng), or TRAP220 (0.25, 0.5, 0.75, 1, 2, and 3 g) as indicated. In these experiments the ligand was E 2 (10 Ϫ8 M). pendent reporter activity with all NR LBDs tested (Fig. 3D). The three core motifs displayed considerable ligand-independent interaction with PPAR␥ and RAR␣ LBDs (and to a lesser degree with RXR␣ LBD in the case of TRAP220 LXM2). Taken together, our results show that while the TRAP220 and SRC1 NIDs appear to have quite distinct NR binding preferences, LXXLL core motifs derived from these domains show little selectivity for NRs. This suggests that LXXLL core motifs are only partial determinants of the specificity of coactivator interactions with NRs, and that other sequences in the NID contribute to specificity.
To investigate the potential influence of amino acids flanking TRAP220 core LXXLL motifs in determining NR binding specificity, we generated an additional ER-DBD fusion protein containing the sequence Ϫ4 to ϩ9 of TRAP220 LXM1 (extended, Fig. 4A). As shown in Fig. 4B, in the presence of AAD-TR␤, the TRAP220-extended motif was able to activate the reporter 3.5fold above the level observed for the TRAP220 LXM1 core motif, in a ligand-dependent manner, suggesting that flanking sequences stabilize the interaction with TR␤. In contrast, when co-expressed with ER␣ LBD the reporter activation by TRAP220 extended motif was ϳ10-fold lower than that achieved by the TRAP220 LXM1 core motif, in the presence of ligand. These results indicate that amino acids immediately flanking TRAP220 LXM1 core motif stabilize TR␤ interaction and reduce ER␣ binding, demonstrating that residues flanking LXXLL core motifs are key determinants of NR binding specificity.
The NR Binding Specificity of TRAP220 NID Can Be Altered by Exchange of Extended LXXLL Motifs-The crystal structures of NR LBDs in complex with LXXLL peptides (12,34) or a polypeptide containing part of the SRC1 NID (13) revealed that the conserved leucine residues make intimate contacts with a hydrophobic groove on the LBD surface, whereas amino acids ϩ2, ϩ3, ϩ6, and ϩ7 are solvent-exposed. Sequences flanking the core motif were undefined in these structures (12,13), thus it remains unclear as to how the LBD makes contacts with NID sequence outside the core motif. However, there is evidence that sequences immediately flanking core LXXLL motifs play a role in determining NR binding specificity (14, 28, 30, 34 -39). For example, a number of LXMs contain a cluster of positively charged residues at positions Ϫ4, Ϫ3, and Ϫ2, that are potentially involved in contacting the conserved glutamic acid in helix 12 of the LBD, which has been proposed to be part of a charged clamp (13). To investigate the contribution of amino acids within extended LXXLL motifs to NR binding specificity, we made a series of TRAP220 NID mutants (A-H; Figs. 5A and 6A). All of the mutations involve an ex-FIG. 2. Interaction of SRC1 and TRAP220 NIDs with NR subclasses. A, yeast two-hybrid interactions of the SRC1 NID with NR LBDs. Schematic representation of the LexA-SRC1 NID bait protein in which the LexA-DBD is fused in-frame with amino acids 431-761 of human SRC1. The LXM1, LXM2, and LXM3 sequences are indicated. Prey vectors consisted of the acidic activation domain of VP16 (AAD) fused in-frame with fulllength TR␤, or the LBDs of AR, PR, ER␣, RAR␣, RXR␣, and PPAR␥. Transformants of the yeast reporter strain L40 co-expressing bait and prey proteins were cultured overnight in the presence or absence of 10 Ϫ6 M cognate ligand; T 3 , mibolerone, R5020, E 2 , AT-RA, 9C-RA, or rosiglitazone. Reporter activity in cell-free extracts is expressed as units of ␤-galactosidase activity. Shaded bars and black bars represent reporter activity in the absence or presence of cognate ligand, respectively. A representative experiment is shown, and similar results were obtained in replicate experiments. B, yeast two-hybrid interactions of the TRAP220 NID with NR LBDs. The LexA-TRAP220 NID fusion protein is shown schematically, and consisted of LexA-DBD fused to amino acids 335-667 of TRAP220. The LXM1 and LXM2 sequences are indicated. Reporter assays to assess interactions with VP16-NR proteins were performed as in A. C, yeast two-hybrid interactions of wild-type and mutant TRAP220 NID constructs with VP16-NR-LBDs. LexA-TRAP220 NID wild-type and mutant constructs are shown schematically. Mut 1 contains the mutation LL-607/8-AA, which inactivates the LXM1 motif, whereas Mut 2 contains the mutation LL-648/9-AA resulting in loss of LXM2 binding to NRs. Reporter activation was determined as in A.
change of amino acids in the extended TRAP220 LXM1 sequence for the corresponding amino acids of SRC1 LXM2. We chose SRC1 LXM2 as this sequence shows strong interactions with both Class I and Class II NRs (14). Western blotting confirmed that wild type and mutant proteins were expressed at similar levels in the yeast reporter strain (Fig. 5B). Exchange of amino acids ϩ2 and ϩ3 of TRAP220 LXM1 for those of SRC1 LXM2 (mutant E) had little effect (Ͻ2-fold) on TRAP220 NID interaction with the Class I receptors PR and ER␣ (Fig. 5C) or the Class II receptors RAR␣, TR␤ and RXR␣ (Fig. 5D). Similarly, the replacement of a proline at position Ϫ2 that has been suggested to define subclasses of coactivators (37) with a lysine residue as found in SRC1 LXM2 (mutant G), had little effect on TRAP220 NID interactions with the Class I receptors, PR and ER␣ (Fig. 5C) or the Class II receptor TR␤ (Fig. 5D). However the ability of mutant G to bind RAR␣ and RXR␣ LBDs in the presence of ligand was slightly reduced (ϳ2-fold), and curiously we observed an increase in ligandindependent interaction with RXR␣ LBD (Fig. 5D). We also generated a mutant in which the entire TRAP220 extended LXM1 sequence (Ϫ4 to ϩ9) was replaced with the corresponding SRC1 LXM2 sequence (mutant F), incorporating a total of eight amino acid exchanges at positions Ϫ4, Ϫ3, Ϫ2, ϩ2, ϩ3, ϩ7, ϩ8, and ϩ9. Remarkably, this mutant displayed a strongly enhanced binding to Class I receptors PR and ER␣ (4-and 10-fold respectively, Fig. 5C). Importantly, the interaction of TRAP220 mutant F with the Class II receptors RAR␣, TR␤, and RXR␣ was unaffected and was similar to that displayed by wild-type TRAP220 (Fig. 5D). Taken together, our results indicate that replacement of the 13 amino acids incorporating FIG. 3. Interaction of TRAP220 and SRC1 LXXLL core motifs with NRs. A, schematic representation of LXXLL core sequences from SRC1 or TRAP220 fused in-frame with the ER␣ DBD, including TRAP220 LXM1 (amino acids 603-611), TRAP220 LXM2 (amino acids 644 -652), and SRC1 LXM2 (amino acids 689 -697). B, Western blot detection of the DBD fusion proteins in equivalent amounts of cell free extracts from yeast transformants, using the antibody recognizing the ER␣ F domain epitope. C-E, yeast two hybrid experiments assessing interaction of DBD-TRAP220 LXM1, DBD-TRAP220 LXM2 or DBD-SRC1 LXM2 with VP16-AAD-NR fusion proteins. Reporter activation assays were carried out on cell-free extracts of W303-1b clones carrying the 3ERE-lacZ reporter and co-expressing bait and prey fusion proteins, as described in the legend to Fig. 2. Shaded and black bars represent reporter activity in the absence or presence of cognate ligand, respectively. A representative experiment is shown, and similar results were obtained in triplicate experiments. Fig. 3 showing the interaction of TRAP220 LXM1 core and extended motifs with TR␤ and the ER␣ LBD. Reporter activities were determined as described in the legend to Fig. 2. Shaded and black bars represent reporter activity in the absence and presence of 10 Ϫ6 M receptor cognate ligand, respectively. TRAP220 LXM1 with the corresponding SRC1 LXM2 sequence is sufficient to change the NR binding properties of the TRAP220 NID. In contrast, amino acid exchange at position Ϫ2 or ϩ2 and ϩ3 is not in itself sufficient to permit strong interactions with Class I receptors.

FIG. 4. The extended TRAP220 LXM1 core motif exhibits selective NR binding. Yeast two-hybrid experiments as in the legend to
The spacing between LXM core motifs in the p160 family is highly conserved, with 51 amino acids between LXM1 and LXM2, and between LXM2 and LXM3, of SRC1 (4). By comparison, the spacing between LXM1 and LXM2 of TRAP220 is 36 amino acids. Previous studies have shown that reducing the spacing between the GRIP1 or TRAP220 motifs can negatively influence NR binding properties (28,35), suggesting there may be an optimal spacing requirement for specific NR interactions.
To examine whether the spacing between LXM1 and LXM2 is a determinant of the TRAP220 preference for NR classes, we generated the mutant (LexA-TRAP220 NID spacer) in which the spacing between TRAP220 LXM1 and LXM2 core motifs was increased to 51 amino acids, as in p160 proteins. To achieve this, a 15 amino acid sequence taken from a corresponding region of SRC1 (amino acids 710 -724, located between LXM2 and LXM3) was inserted between LXM1 and LXM2 of TRAP220 (Fig. 5A, spacer). As shown in Fig. 5C, this mutation did not significantly alter the binding of TRAP220 FIG. 5. Effects of mutations altering LXM sequence and spacing on binding of the TRAP220 NID to NRs. A, schematic representation of the LexA-TRAP220 NID mutants. Amino acid exchanges between TRAP220 and SRC1 sequences are highlighted by the shaded boxes. The additional SRC1-derived sequence used to generate the TRAP220 spacer mutant is indicated. B, Western blot showing the expression of the LexA-TRAP220 NID fusion proteins in cell-free extracts used in reporter assays. The antibody used was a mouse monoclonal antibody recognizing the LexA DBD. C, yeast two-hybrid experiments, performed as in Fig. 2, showing the interaction of LexA-TRAP220 NID wt and mutant proteins with AAD-PR and AAA-ER␣ fusion proteins, in the absence (white bars) or presence (black bars) of cognate ligand (R5020 and E 2 , respectively). D, yeast two-hybrid experiment as in C using AAD-RAR␣, AAD-TR␤, and AAD-RXR␣ constructs with or without AT-RA, T 3 , and 9C-RA, respectively. The data represent the mean reporter activity of two transformants, and error bars indicate S.D. NID to PR or ER␣, suggesting that increased spacing between the TRAP220 LXXLL motifs is not sufficient to allow a strong interaction with Class I receptors. Moreover, this sequence insertion did not adversely affect the interaction of the NID with TR␤ or RXR␣ (Fig. 5D), although reporter activation due to interaction with AAD-RAR␣ was slightly reduced (Fig 5D). Thus, the difference in spacing between LXMs in TRAP220 and SRC1 does not appear to be a critical determinant of the distinct NR binding preferences of these proteins.
Combinatorial Effects of Mutations in LXM1 on the NR Binding Specificity of the TRAP220 NID-Having determined that exchange of the 13 amino acids comprising LXM1 can alter the NR binding specificity of the TRAP220 NID (mutant F), we used an expanded panel of TRAP220 NID mutants (A-D) to investigate the specificity determinants in more detail (Fig.  6A). Yeast two hybrid experiments were carried out to assess the ability of these mutants to interact with the ER␣ LBD. Western blots confirmed that the wild-type and mutant proteins were expressed at similar levels in yeast (Fig. 6B). Mutant A is similar to mutant F with the exception that amino acids ϩ2, ϩ3, are wild type. Mutant B contains SRC1 sequence at positions ϩ2, ϩ3, ϩ7, ϩ8, and ϩ9 and thus assesses the contribution of N-terminal flanking sequence. Mutant C contains SRC1 sequence at positions Ϫ4, Ϫ3, Ϫ2, ϩ2, and ϩ3, to assess the contribution of the C-terminal flanking sequence. Finally, mutant D contains SRC1 sequence at positions Ϫ2, ϩ2, and ϩ3. As shown in Fig. 6C, all these TRAP220 NID mutants showed increased ER␣ binding compared with wild type NID. However, none were as efficient at binding ER␣-LBD as mutant F, which contained the entire SRC1 LXM2 extended motif. Mutant A displayed 8-fold enhanced reporter activation due to binding ER␣-LBD, compared with a 19-fold enhancement seen for mutant F under similar conditions (Fig. 5B). This suggests that the amino acids at positions ϩ2 and ϩ3 are important in the context of extended LXXLL motif, although alone (mutant E) they have only a minimal effect on NR binding (Fig. 5C). Similarly, enhanced ER␣-LBD interaction was observed for mutant C (6-fold) (Fig. 6C), suggesting that the N-terminal flanking sequences (Ϫ4, Ϫ3, Ϫ2) in combination with the ϩ2, ϩ3 amino acids, have an important influence on TRAP220 NID NR binding specificity. Moreover, a similar increase was observed for mutant D (7-fold), which has a single amino acid exchange at the Ϫ2 position (Pro to Lys) in combination with the ϩ2 and ϩ3 amino acid exchange. This is in contrast to the results obtained for mutant G, which contained the Pro to Lys change only, and mutant E, which has the ϩ2, ϩ3 change only, neither of which showed a strong increase in binding to ER␣ (Fig. 5C). Thus, in combination these mutations influence the NR binding specificity of TRAP220 LXM1. The relatively weak interaction of AAD-ER␣ with mutant B, which contained SRC1 LXM2 sequence at the C-terminal flank (ϩ7, ϩ8, ϩ9), coupled with the change at the ϩ2 and ϩ3 position suggests that this combination of amino acids is less important for interaction with ER␣. However, the difference in ER␣ binding of mutants F and C (Fig. 6C) suggests that the LXM1 C-terminal flanking sequence does influence NR binding specificity of the TRAP220 NID.
To examine whether mutation of the distal LXXLL motif would also alter TRAP220 NR binding specificity, we generated mutant H, which replaces the LXM2 sequence Ϫ4 to ϩ9 with the corresponding SRC1 LXM2 sequence (Fig. 7A). As shown in Fig. 7B, this mutation resulted in a 8-fold increase in reporter activity due to ER␣ binding, but had little effect on the ability of the TRAP220 NID to bind TR␤. Thus replacement of either LXM1 or LXM2 in the TRAP220 NID with the 13 amino acid sequence, RHKILHRLLQEGS, results in a strong increase in binding to the Class I receptors ER␣ and PR, without affecting binding to Class II NRs. In contrast, altering spacing between LXM1 and LXM2 to that seen in SRC1, did not alter NR binding. Thus, we conclude the extended LXXLL motif is the principle NR binding specificity determinant of TRAP220.
Altering the NR Binding Specificity of Full-length TRAP220 -To support the conclusions from our yeast twohybrid data, we introduced mutations similar to that in mutant F into the LXM1 motif of full-length TRAP220 protein, and assessed its ability to bind NRs in vitro (Fig. 8). Bacterially expressed GST, GST-ER␣-LBD, GST-TR␤-LBD, and GST-RXR␣-LBD fusion proteins were tested for binding to in vitro translated, radiolabeled full-length SRC1e and TRAP220 wildtype proteins, or a TRAP220 LXM1 mutant F protein. Equal amounts of GST fusion proteins and radiolabeled proteins were used in each experiment. As expected, SRC1e showed strong ligand-dependent binding to ER␣, TR␤, and RXR␣ LBDs. Under the same conditions, TRAP220 wild-type protein showed strong ligand-dependent or ligand-stimulated binding to RXR␣ and TR␤, respectively. In contrast, TRAP220 wild-type protein showed little, if any, detectable binding to ER␣ LBD, consistent with our yeast two-hybrid experiments. Remarkably, the TRAP220 LXM1 mutant F displayed strong interactions with TR␤, RXR␣ LBDs, and also the ER␣ LBD in the presence of ligand. Thus, swapping of a single extended LXM sequence is sufficient to change of specificity of the TRAP220 coactivator, allowing it to bind the Class I receptor ER␣. DISCUSSION TRAP220/DRIP205/PBP was identified as a consequence of its strong ligand-dependent binding to Class II NRs such as TR␣/␤, VDR, and PPAR␥. In contrast, several groups have shown that the interaction of TRAP220 with ER␣ is comparatively weak (18,19,30,51). In this study we compared the interactions of TRAP220 proteins with Class I and Class II NRs using yeast two-hybrid and GST pull-down experiments. Consistent with this, our results show that TRAP220 NID (Figs. 2,   FIG. 7. Mutation of the extended LXM2 sequence, and its effect on the interaction of TRAP220 NID with ER␣. A, schematic representation of highlighting sequence exchanges in LexA-TRAP220 mutant H. B, yeast two-hybrid experiment showing the binding of LexA-SRC1 NID and LexA-TRAP220 NID wt and mutant H proteins to AAD-ER␣ and AAD-TR␤ as in Fig. 2. 5, 6, and 7) or full-length TRAP220 proteins (Fig. 8) displayed only weak ligand-induced interactions with ER␣ (or other Class I NRs) in yeast two-hybrid and GST pull-down assays. While the SRC1 NID showed no apparent preference for binding to the panel of NRs used here ( Fig. 2A), TRAP220 NID bound preferentially to TR␤, RXR␣, PPAR␥, and to a lesser extent RAR␣, whereas its binding to ER␣, AR, and PR was dramatically reduced (Fig. 2B). To understand the molecular basis of these differential interactions, we undertook an indepth analysis of TRAP220/NR binding.
The presence of multiple LXXLL motifs in p160s and other coactivators is thought to facilitate co-operative binding to NRs and may be important for selective interactions through differential usage of LXXLL motifs (15,16,35,52). Similarly, mutagenesis of TRAP220/DRIP205 LXXLL motifs revealed preferential interaction of RXR␣ LBD with LXM1, whereas LXM2 showed preferential binding to the LBDs of the heterodimeric partners of RXR such as TR␣, VDR, and PPAR␣ (28). Our yeast two-hybrid experiments indicate that mutation of the conserved leucines in either LXM1 or LXM2 results in greatly reduced binding to Class II receptors (Fig. 2C), suggesting that NR LBD dimers contact both TRAP220 motifs. This is consistent with the crystal structures of agonist-bound LBD homodimers or RXR␣ LBD heterodimers in combination with short LXXLL peptides, or a partial SRC1 NID, which show that both LBDs in the NR dimer are occupied with a LXXLL core ␣-helix (12,13,34,53). An exception in our experiments was the RXR␣ LBD, which required only a functional LXM1 for strong binding (Fig. 2C).
To investigate whether the LXXLL motifs from TRAP220 display the same NR specificity as demonstrated by its NID, we examined their NR binding properties. Remarkably, TRAP220 LXM1 and LXM2 core motifs displayed strong interactions with all NR LBDs tested (Fig. 3, C and D), in contrast to the NID (Fig. 2B). Previous studies have shown that the NR binding specificity of other coactivators is determined in part by LXXLL core motifs (4,14,35,52) but also involves sequences immediately flanking the core motif (35,38). As shown in Fig.  4, an extended LXM1 sequence, containing additional amino acids on the N-and C-terminal flanks, showed increased interaction (3.5-fold) with TR␤, but a 10-fold reduced interaction with ER␣ LBD. Thus, the extended 13 amino acid LXM1 sequence displays selective NR binding properties similar to the TRAP220 NID (Fig. 2B), or full-length protein (Fig. 8).
To determine which residues in the extended LXM1 are important for its NR subclass selectivity, we used a panel of TRAP220 NID mutants (A-H). Initially, we confirmed that the extended LXM1 sequence is a major determinant of TRAP220 NR binding specificity, by exchange of the LXM1 for the corresponding sequence of SRC1 LXM2 (Mutant F, Fig. 5A). This resulted in strongly enhanced interaction of the TRAP220 NID with ER␣ and PR ( Fig. 5C) but had little or no effect on binding to Class II NRs (Fig. 5D). A similar result was recently reported in which replacement of TRAP220 LXM1 (residues Ϫ5 to ϩ9) with the corresponding sequence from TIF2 LXM2, enhanced binding to ER␣, in contrast to wild-type TRAP220 (30). However, this effect is not restricted to LXM1, as our data show that exchange of the extended TRAP220 LXM2 sequence for SRC1 LXM2, (Mutant H) also results in enhanced binding to ER␣ LBD, without affecting binding to TR␤ (Fig. 7B).
Other studies have shown that a reduction in the spacing between LXXLL motifs in p160s (35) or TRAP220 (28) has adverse affects on NID/NR interactions. This suggest that a minimal sequence length is required to generate a folded or flexible domain which can permit docking of both LXM1 and LXM2 with both AF2 surfaces on NR dimers. However, we noted that the spacing between p160 LXMs (51 amino acids) is highly conserved, even across species, and differs from that found between TRAP220 LXMs (36 amino acids). To investigate whether this differential spacing is a feature of the NR selectivity of TRAP220, we generated the mutant designated spacer in which the distance between LXM1 and LXM2 in the TRAP220 NID is increased to 51 amino acids, using a sequence derived from the SRC1 LXM2/LXM3 spacer region (Fig. 5A). However, the spacer mutant showed no enhanced interaction with steroid receptors (Fig. 5C), nor was binding to Class II NRs adversely affected (Fig. 5D). Thus, while a minimal spacer sequence may be required to allow contact with both AF2 surfaces, the exact spacing does not appear to be critical. Moreover, the absence of any effect of the spacer mutation on binding to Class II NRs is consistent with the hypothesis that rather than folding into a rigid domain, the NID may be a flexible or largely unstructured sequence, accommodating interactions with AF2 surfaces on different NR dimers.
Other TRAP220 NID mutants were used to examine the importance of different residues within the extended LXM1 sequence for NR binding specificity. A previous study used phage display to identify subclasses of LXXLL sequences that show differential interactions with ER␣ LBD (37). The LXMs of TRAP220 fall into the subclass typified by having a conserved proline residue at the Ϫ2 position, which shows relatively weak interactions with ER␣. In p160s, three lysine residues flanking the LXM1 motif, including a lysine at the Ϫ2 position, have been shown to be targets for acetylation by CBP/p300, and this event assists the dissociation of NR/p160 complexes (47). This lysine residue at the Ϫ2 position potentially makes electrostatic contact with Glu-380 in helix 5 of the ER␣ LBD (47). Our data show that exchange of the proline at Ϫ2 in LXM1 for lysine (Mutant G) did not significantly alter the binding properties of the TRAP220 NID (Fig. 5, C and D). Similarly, substitution of proline for alanine at position Ϫ2 of LXM2 had little effect on TRAP220 binding to TR␣, VDR, or PPAR␣ (28). Thus, mutation of this residue in LXM1 is not sufficient on its own to enhance TRAP220 binding to steroid receptors.
Several studies have highlighted the importance of the ϩ2 and ϩ3 amino acids in NR/cofactor interactions. Mutation of the ϩ2 and ϩ3 amino acids of the variant FXXLL motif of NSD1 to alanines (Ser-Thr to Ala-Ala) abolished binding to both Class I and Class II NRs (33). Similarly, a ϩ3 mutation in TIF2 LXM3 has been reported to reduce its interaction with ER␣ (30). However, in our experiments, exchange of the ϩ2 and ϩ3 amino acids of TRAP220 LXM1 for the positively charged equivalents in SRC1 LXM2 (Mutant E) did not by itself result FIG. 8. Interaction of wild type and mutant TRAP220 proteins with GST-NRs in vitro. Normalized amounts of GST, GST-ER␣-LBD, GST-RXR␣-LBD, and GST-TR␤-LBD proteins were immobilized on glutathione-Sepharose beads and incubated with 35 S-labeled full-length TRAP220 wild-type (wt), TRAP220 LXM1 mutant F, or SRC1e wildtype protein (as control), in the presence or absence of 10 Ϫ6 M cognate ligand as indicated (E 2 , 9C-RA, or T 3 , respectively). Bound proteins were analyzed by SDS-PAGE and autoradiography, as described under "Materials and Methods." One-tenth of the total 35 S-labeled protein used in the pull-down is shown for comparison (10% input). in a strong increase in binding of TRAP220 NID to Class I NRs (Fig. 5C), or alter interaction with Class II NRs (Fig. 5D).
Mutants A-D (Fig. 6A) were used in NR binding assays to allow us to investigate the effect of combinatorial changes in the extended LXM1 sequence. All of these mutants displayed enhanced interaction with ER␣ compared with wild-type NID (Fig. 6C). For example, ER␣ binding to Mutant F was greater than to Mutant A, indicating that the ϩ2, ϩ3 positions are important for optimal binding. Similarly, exchange of residues ϩ2, ϩ3 resulted in increased ER␣ binding only when combined with exchange of N-terminal (Ϫ4, Ϫ3, Ϫ2), or to a lesser extent the C-terminal (ϩ7, ϩ8, ϩ9) flanking sequences (compare Mutant E in Fig. 5C with Mutants B and C in Fig. 6C). The similar extent of ER␣ interaction with Mutants C and D also suggests that the Ϫ2 position plays a key role in the interaction of ER␣ with the N-terminal flank, possibly via the Glu-380 residue in helix 5 of ER␣ (47). Taken together, our results indicate that exchange of the entire extended LXM1 motif for the SRC1 LXM2 sequence is required for optimally enhanced binding to ER␣, and that residues within the core motif and flanking sequence cooperate to determine NR binding specificity.
The LXXLL core motif forms a two-turn ␣-helix that is clamped in position on the LBD surface by electrostatic interactions with conserved lysine and glutamate residues (13). Unfortunately, the crystal structures available to date offer little structural insight into how flanking sequences in extended LXXLL motifs might determine receptor specific contacts with LBDs. A recent study has revealed the existence of a second charged clamp in the GR LBD that appears to anchor the third LXXLL motif of TIF2 (LXM3). This involves interactions with the ϩ2 and ϩ6 amino acids, further highlighting the importance of the ϩ2 position amino acid and C-terminal flank in coactivator interactions with Class I NRs (58). Further structural studies involving extended LXXLL peptides, and preferably stably folded NIDs will be required to visualize selective interactions of different NRs with TRAP220 and other cofactors such as TIP60, NRIF3, and ASC-2.
Purified TRAP/DRIP/mediator complexes have been shown to enhance the activity of NRs in cell-free or purified in vitro transcription assays (17,50,54). In comparison, only very modest enhancement of TR␣/␤, VDR, and PPAR␥-mediated transcription has been observed due to ectopic expression of the NR binding subunit TRAP220/DRIP205/PBP in transiently transfected cells (17, 19 -21). Consistent with this, we observed that exogenously expressed TRAP220 induced a very moderate enhancement of TR␤ activation of a DR4-driven reporter gene (Fig. 1A). The relatively weak coactivator activity of TRAP220 in transfection experiments might be due to inefficient assembly of exogenous TRAP220 proteins into functional mediator complexes. By contrast, TRAP220 failed to enhance ER␣-mediated transcription of the 3ERE-TATA-LUC reporter gene ( Fig 1B). Nonetheless, chromatin immunoprecipitation experiments have suggested that TRAP/DRIP proteins are recruited to NR-regulated promoters, including ER␣-and AR-responsive promoters (51,55,56), and microinjection of HeLa cell nuclei with anti-TRAP220/PBP antibodies down-regulated ER␣ activation of a reporter gene (57). In addition, a recent report has shown that the TRAP complex can be purified from HeLa cell nuclear extracts using GST-ER␣ LBD in the presence of agonist ligand (50). In contrast to reports suggesting that TRAP170 interacts with steroid receptors, GST-ER␣ AF1 did not retain TRAP complex in HeLa cell nuclear extracts (50). Thus, recruitment of TRAP complexes to the promoters of estrogen-regulated genes may involve additional interactions with other components of this multiprotein complex.
In conclusion, we have demonstrated that TRAP220 exhibits preferential binding to Class II NRs and only weak interactions with steroid receptors. This binding specificity is determined by an extended LXXLL motifs of 13 amino acids in length, which when exchanged is sufficient to alter the specificity of fulllength TRAP220 protein. It will be of interest to undertake mutagenesis of NR LBDs to further explore the molecular mechanism of selective interactions between NRs and their cofactors.