Biochemical and NMR mapping of the interface between CREB-binding protein and ligand binding domains of nuclear receptor: beyond the LXXLL motif.

CBP, cAMP-response element-binding protein (CREB)-binding protein, plays an important role as a general cointegrator of various signaling pathways and interacts with a large number of transcription factors. Interactions of CBP with ligand binding domains (LBDs) of nuclear receptors are mediated by LXXLL motifs, as are those of p160 proteins, although the number, distribution, and precise sequences of the motifs differ. We used a large N-terminal fragment of murine CBP to map by biochemical methods and NMR spectroscopy the interaction domain of CBP with the LBDs of several nuclear receptors. We show that distinct zones of that fragment are involved in the interactions: a 20-residue segment containing the LXXLL motif (residues 61-80) is implicated in the interaction with all three domains tested (peroxisome proliferator-activated receptor gamma-LBD, retinoid X receptor alpha-LBD, and estrogen-related receptor gamma-LBD), whereas a second N-terminal well conserved block of around 25 residues centered on a consensus L(40)PDEL(44) motif constitutes a secondary motif of interaction with peroxisome proliferator-activated receptor gamma-LBD. Sequence analysis reveals that both zones are well conserved in all vertebrate p300/CBP proteins, suggesting their functional importance. Interactions of p300/CBP coactivators with the LBDs of nuclear receptors are not limited to the canonical LXXLL motifs, involving both a longer contiguous segment around the motif and, for certain domains, an additional zone.

Nuclear receptors (NRs) 1 constitute a superfamily of ligandregulated transcription factors, implicated in the regulation of diverse metazoan functions, such as development, cell differentiation, reproduction, metabolism, and homeostasis (1,2). Promoter-bound NRs control target gene expression by recruiting diverse cofactor complexes that act on local chromatin structure in an activating or repressive manner. NRs share a common structural and functional architecture that generally consists of five to six domains including a ligand binding domain (LBD) ( Fig. 1A; for review, see Ref. 3). Structural studies of various NR LBDs have characterized and analyzed extensively a recognition interface between NR LBDs and transcriptional coactivators (4 -9).
The best characterized NR coactivators belong to the p160 family and interact with LBDs as well as with A/B domains of NRs. Several studies suggest that p160s play a central role in the recruitment of other coactivators, including the 300-kDa cAMP response element-binding protein (CREB)-binding protein (CBP), which belongs to the p300/CBP family (10 -12). CBP plays an important role as a general cointegrator of various signaling pathways and interacts with a large number of transcription factors and cofactors, through numerous protein binding domains spread along its primary sequence (Fig. 1A) (12). Both CBP and p160s possess a histone acetyltransferase domain (13,14). Synergistic effects of p160, p300, and other associated coactivators on the transactivation mediated by NRs suggest a combined recruitment to the promoter of a histone acetyltransferase-methyltransferase complex, which participates in the activation of transcription (Fig. 1B) (15)(16)(17).
CBP possesses three signature sequences, designated "NR boxes," each containing an LXXLL motif (where L is leucine and X represents any amino acid), common to most coactivators. Functional studies performed in vitro as well as in vivo have highlighted the essential role of NR boxes in the interaction of coactivators with NR LBDs (18, 26 -32; for a recent review, see Ref. 33). Both functional and structural studies now suggest a general mechanism for the assembly of NRs and , and a C-terminal conserved LBD (E domain), which contains the ligand-dependent activation function (AF2) and is sometimes followed by a nonconserved F domain. The A/B and E domains interact with members of both the p160 and p300/CBP protein families, whereas the C domain interacts with the PCAF. Members of the p160 family possess an N-terminal basic helix-loop-helix motif and a Per-Arnt-Sin (PAS) homology region, a central NR interacting domain encompassing three LXXLL motifs (NR boxes), and a C-terminal histone acetyltransferase (HAT) domain. Interactions with the A/B domain and LBD of NRs are mediated by a C-terminal glutamine-rich region and by the NR interacting domain, respectively. The C-terminal extremity also contains motifs of interaction with PCAF, coactivator-associated arginine methyltransferase 1 (CARM1), and the p300/CBP family. Members of the p300/CBP protein family contain two homologous zinc binding domains, TAZ1 and TAZ2, a KIX domain identified for its interaction with CREB, a bromodomain (Br) that interacts with acetylated protein targets, a histone acetyltransferase domain, and a steroid receptor coactivator interacting domain (SID). p300/CBP proteins contain three potential NR boxes. Interaction domains for PCAF and p160 are indicated. B, recruitment of cofactors by NRs at the promoter of target genes. Once bound to specific DNA sequences, NR dimers recruit cofactors in a concerted manner to inhibit (corepressors) or enhance (coactivators) gene transcription. The putative coactivator complex consisting of p300/CBP, p160, PCAF, and CARM1 enhances transcription by decompacting the chromatin structure through a histone acetyltransferase activity (carried by p300/CBP, PCAF, and p160), and by recruiting components of the transcriptional machinery. coactivators and tend to limit the interactions between coactivators and NR LBDs to the binding of NR boxes on the surface of the LBD. The number of LXXLL motifs, their sequences, and the nature of adjacent amino acids vary among the coactivators and are likely to account for the observed differences in binding to selected NRs. Although p160 NR boxes are clustered in a small central region spanning ϳ200 amino acids, the three LXXLL motifs of CBP are spread across its sequence (motif I: L 69 XXL 72 L 73 ; motif II: L 357 XXL 360 L 361 ; motif III: L 2068 XXL 2071 L 2072 in murine CBP, Fig. 1A). Motif I has been shown to mediate strongly the interaction with NR LBDs (10,11,29), but the two other motifs belong to well defined structural domains (34,35). Although polypeptides containing any of the three LXXLL motifs of CBP were shown to interact weakly with NR LBDs (11,29), it seems unlikely that motifs II and III mediate interactions with NR LBDs.
Because p160 and p300/CBP share no sequence homology and differ in the distribution of LXXLL motifs along the sequence, we wished to investigate whether these differences might reflect important differences in the mode of interaction of CBP with NR LBDs compared with p160. Moreover, because studies carried out using fragments of p160s indicate that interactions with NR LBDs are limited essentially to the LXXLL motifs, we wished to establish whether this was also true for CBP. We have used biochemical techniques and NMR spectroscopy to map the interaction of CBP fragments with a number of NR LBDs. Assignments were achieved using a 13 C, 15 N-labeled sample, enabling changes in the 1 H-15 N HSQC spectra of 15 N-labeled samples in the presence of NR LBDs to be interpreted in site-specific terms. Interactions with apo and holo forms of human PPAR␥-LBD, apo and holo forms of mouse RXR␣-LBD, as well as the heterodimers fully apo and fully holo PPAR␥-LBD/RXR␣-LBD were probed in this way. Finally, the interaction with human ERR␥-LBD in the absence and presence of the antagonist, 4-hydroxytamoxifen, was investigated.
Purification-Purified ERR␥-LBD was a kind gift from H. Greschik (36). For all other proteins, cells were resuspended in buffer A (10 mM Tris, 500 mM NaCl at pH 8.0) with a protease inhibitor mixture (leupeptin, chymostatin, pepstatin, aprotinin, antipain A, final concentration 2.5 g/ml of each) and sonicated before ultracentrifugation (45,000 rpm, 1 h). The supernatant was loaded on a Talon cobalt affinity column (Clontech), via a Bio-Rad Biologic FPLC system. The column was washed extensively with buffer A until the UV base line returned to 0. Proteins were eluted with buffer a containing 0.5 M imidazole.
With the exception of the shorter fragments of CBP (CBP 1-116 , CBP 1-127 , CBP  ), all proteins were purified by gel filtration (GF), using an Amersham Biosciences Superdex S200(16/90) GF column, at a flow rate of 1 ml/min, in 10 mM Tris, 50 mM NaCl at pH 8.0 (for further biochemical analyses and dynamic light scattering (DLS)) or 50 mM sodium phosphate buffer at pH 6.9 (for NMR and DLS). In particular, the GF step allowed monomeric RXR␣-LBD to be separated from the tetrameric species also present in the affinity-purified fraction. The GF step was also crucial for separating the PPAR␥-LBD/ RXR␣-LBD heterodimer from traces of monomeric PPAR␥-LBD, or tetrameric and monomeric RXR␣-LBD variably present in affinitypurified fractions of various culture preparations.
Analytical GF experiments were performed on an Amersham Biosciences Superdex S200(10/30) GF column (10 mM Tris, 50 mM NaCl, pH 8.0, flow rate 0.5 ml/min). Typically, analytical GF experiments were performed by injecting affinity-purified CBP fragments mixed with a large excess (10 -20-fold) of purified holoPPAR␥-LBD (see below). The mass in solution of purified species was estimated by comparing the elution volume with those of GF standard markers (Bio-Rad). Samples collected after GF were concentrated (Microcon 10, Millipore) and run on SDS-PAGE.
Preparation of HoloNR LBDs-NR LBD species bound to agonist ligand were obtained by mixing diluted purified apoNR LBDs with a 2-fold molar excess of high affinity, specific ligands (tyrosine-based agonist BMS953 for PPAR␥ (37); 9-cis retinoic acid for RXR␣), before concentration and GF. ERR␥-LBD bound to the antagonist 4-hydroxytamoxifen was prepared by mixing diluted GF-purified ERR␥-LBD with a 1.5-fold molar excess of antagonist, followed by a centrifugation step (10,000 rpm, 30 min) to remove traces of aggregates, before concentration. The purity of proteins was checked by SDS-PAGE. The quality of purified proteins was assessed by native PAGE experiments on a Phast-system (Amersham Biosciences), using 8 -25% gradient gels, with a 250 mM Tris, 880 mM L-alanine buffer at pH 8.8.
Mild Trypsinolysis Experiment-Affinity-purified His 6 -CBP 1-442 (1 mg/ml) was mixed with increasing concentrations of trypsin, from 0.4 to 200 ng/ml. Samples were incubated at 4°C for 1 h, and 10 l of each was run on SDS-PAGE (12.5%) and revealed by Coomassie Blue staining.
Mass Spectrometry-GF-purified CBP 1-116 ⅐PPAR␥-LBD⅐BMS953 complex was dialyzed against 50 mM ammonium acetate buffer at pH 7.0. The protein concentration was estimated to be around 1 mg/ml. ESI measurements were performed on a ESI-TOF instrument (LCT, Micromass, UK) fitted with a Z-spray source. Samples were diluted to 10 M in 50 mM ammonium acetate and infused continuously into the ESI ion source at a flow rate of 3 l/min. Great care was taken to ensure that noncovalent interactions preformed in solution survived the ionization/ desolvation process. In particular, atmospheric pressure/vacuum interface parameters were optimized to obtain the best sensitivity and spectrum quality without affecting the noncovalent complex stability.
DLS Experiments-DLS experiments were performed with a Dynapro-MS/X instrument (Proterion) and analyzed using Dynamics 5.25.44 software. Experiments were performed at temperatures between 4 and 37°C, with proteins concentrated to around 1 mg/ml (50,000 -150,000 cps), centrifuged, and filtered on 0.02-m filters to remove dust.
Nuclear Magnetic Resonance Experiments-Typically, samples were 0.3 mM protein concentration in 300 or 500 l of 50 mM sodium phosphate buffer at pH 6.9. For measurements in the presence of NR LBDs, 3 mM solutions of CBP fragments were diluted 10-fold into solutions containing Ͼ0.5 mM NR LBDs. NMR spectra were recorded at 500 or 600 MHz ( 1 H frequency) on Bruker DRX500 and DRX600 spectrometers equipped with triple resonance probes with z-gradients. All spectra were recorded at 15°C. One-dimensional 1 H spectra were recorded with appropriate spectral widths for the sample, using a WATERGATE sequence (38) for water suppression. Data were processed in the manufacturer's software XWINNMR (Bruker) with a 90°shifted sine bell applied prior to Fourier transformation. 1 H-15 N HSQC were recorded with spectral widths set to avoid folding of 15 N resonances. Typically, 256 t 1 increments were acquired, and quadrature was achieved using the method of States et al. (39). Spectra were processed applying a 90°shifted sine bell and zero-filling at least to the next power of 2 in each dimension prior to Fourier transformation. Spectra were processed using NMRPipe (40) and analyzed in XEASY (41). 15 N-Edited TOCSY and NOESY were recorded with 50 increments in the indirect 15 N dimension and 320 increments in the indirect 1 H dimension. Quadrature was achieved using the States-TPPI method (42) in both dimensions. Spectra were processed as above with linear prediction in the indirect 15 N dimension. 13 C experiments for assignment of backbone resonance were based on standard pulse sequences provided by the manufacturer with only minor modification. Spectra were processed as above with linear prediction in the indirect 13 C dimension.

Search for a Stable Complex between CBP and NR LBDs-To
characterize the mode and specificity of interaction of p300/ CBP coactivators with NRs, we mapped the interaction domain of CBP with various NR LBDs by biochemical methods and NMR spectroscopy.
The purified N-terminal fragment of CBP encompassing the two first LXXLL motifs and the transcriptional adaptor zinc binding (TAZ1) domain (CBP 1-442 ) revealed polydisperse behavior in DLS experiments. Moreover, this fragment was prone to aggregation and displayed high sensitivity to proteolysis. Mild trypsinolysis rapidly yielded stable degradation products ( Fig. 2A), and the use of thrombin to remove the His tag resulted in cleavage at Arg 217 , Arg 218 , and Arg 219 (Fig. 2B). Proteolysis also occurred during the purification process, despite the use of protease inhibitors. N-terminal sequencing of degradation products allowed identification of the proteolytic sites (Fig. 2B). The central portion of CBP 1-442 is particularly susceptible to proteolysis and may therefore be less structured than either the TAZ1 domain or the N-terminal extremity encompassing the first LXXLL motif.
To stabilize CBP 1-442 , we attempted copurification with PPAR␥-LBD, which was reported to interact with CBP with high affinity (21,22,24,25). CBP 1-442 was mixed with an excess of PPAR␥-LBD, in the presence of BMS953 (a PPAR␥ agonist) (37) and applied on GF (Fig. 2C). PPAR␥-LBD⅐BMS953 coeluted with CBP 1-442 , whereas excess PPAR␥-LBD eluted later. Because PPAR␥-LBD⅐BMS953 eluted exclusively with PPAR␥-LBD when injected alone, its coelution with CBP 1-442 confirms the formation of a complex. The estimated mass of CBP 1-442 ⅐PPAR␥-LBD⅐BMS953 in analytical GF experiments was 210 kDa. Because the calculated mass is 82 kDa, the complex appeared to be at least dimeric or further aggregated, suggesting that the propensity of CBP 1-442 alone to aggregate is maintained in the complex. Moreover, spontaneous proteolysis of CBP 1-442 still occurred after copurification with PPAR␥-LBD, indicating that complex formation does not diminish the intrinsic instability of CBP 1-442 . We therefore focused on shorter N-terminal fragments of CBP and determined the quality of copurified CBP⅐PPAR␥-LBD⅐BMS953 complexes by DLS measurements, native PAGE, and mass spectrometry experiments under native conditions. Complex formation between individually purified N-terminal CBP fragments and PPAR␥-LBD⅐BMS953 was assessed by analytical GF. All fragments containing the first LXXLL motif copurified with PPAR␥-LBD⅐BMS953, whereas a fragment lacking the first LXXLL motif (CBP 83-425 ) did not (Fig. 3, A and  B). Complexes involving CBP 1-220 or larger fragments remained sensitive to proteolysis and showed a multimeric elution pattern in GF. In contrast, complexes involving CBP  or CBP 1-116 were more stable to proteolysis and displayed monodisperse behavior in DLS experiments.
Native ESI-TOF mass spectrometry experiments performed at low cone tension (50 V) with copurified CBP 1-116 ⅐PPAR␥-LBD⅐BMS953 confirmed the existence of a ternary complex (Fig.  3C, species C: 47,934.85 Ϯ 2.20 Da). Interestingly, only the apo complex was observed at higher cone tension (120 V) (species B: 47,388.9 Ϯ 1.36 Da), suggesting that, under these conditions, the agonist is dissociated from the NR LBD, whereas the CBP fragment remains associated. This is consistent with the observation that CBP 1-116 ⅐PPAR␥-LBD could be copurified in the absence of any ligand (data not shown). Interestingly, purified apo-and holoCBP 1-116 ⅐PPAR␥-LBD complexes show different patterns of migration on native PAGE (Fig. 3D). The holo complex migrates more slowly than the apo complex, whereas apo-and holoPPAR␥-LBD show similar patterns. This observation suggests further conformational changes in the CBP 1-116 ⅐PPAR␥-LBD complex induced by binding of BMS953.
The 1 H-15 N HSQC spectrum of the slightly longer construct 15 N-CBP  in the presence of PPAR␥-LBD shows a number of significant changes from that of 15 N-CBP 1-127 alone (Fig.  4A). In the absence of interaction partner, the 1 H NMR spectrum of CBP 1-127 suggests that it is largely unstructured. There are neither upfield-shifted methyl proton resonances nor any downfield-shifted amide protons, both characteristic features of the spectrum of a folded protein. Similarly, in the 1 H-15 N HSQC spectrum of 15 N-CBP 1-127 , all amide protons lie between 8.7 and 7.7 ppm, and there is little dispersion in the 15 N dimension (Fig. 4). In contrast, in the presence of PPAR␥-LBD, around 25 cross-peaks either disappear from the spectrum or undergo changes in chemical shift. The remainder of the spectrum is unchanged, suggesting a specific interaction. After assignment of the spectra of a smaller CBP fragment (fragment CBP 31-90 , see below) it was possible to identify affected residues as spanning the segment of the sequence between Ala 62 and Ser 80 , encompassing the L 69 XXL 72 L 73 motif. Binding of BMS953 to PPAR␥-LBD led to only few differences in the 1 H-15 N HSQC spectrum of 15 N-CBP 1-127 with respect to that recorded in the absence of agonist (Fig. 4, B and C). These small changes were mapped to residues Glu 37 , Leu 40 , and Gly 48 .
The effects of PPAR␥-LBD on the spectrum of 15  In contrast to the results obtained with apoPPAR␥-LBD, the 1 H-15 N HSQC spectrum of 15 N-CBP 1-127 in the presence of apo RXR␣-LBD is unchanged from that of 15 N-CBP 1-127 alone (Fig.  4D). The presence of 9-cis-retinoic acid on the RXR␣-LBD in the copurified complex led to attenuation or disappearance of a set of resonances similar to those affected by PPAR␥-LBD (Fig. 4,  E and F). The segment Glu 71 -Gly 78 is markedly perturbed, whereas cross-peaks of the preceding two residues (Leu 69 , Ser 70 ) lie in crowded regions and therefore cannot be identified with confidence, although cross-peaks are indeed affected in those regions. The intensities of cross-peaks of residues Ser 64 , His 66 , and Gly 76 are attenuated but still observable, possibly suggesting faster exchange.
CBP  Contains the Minimal Domain of Interaction with NR LBDs-15 N-Edited NOESY spectra of 15 N-CBP 1-127 are notably poor in information (data not shown). Very few resonances could be assigned, and most of these only tentatively. Nonetheless, the identification of short motifs and comparison with 15 N-edited NOESY spectra of 15 N-CBP 1-127 in the presence of PPAR␥-LBD and PPAR␥-LBD⅐BMS953 allowed the segment of 15 N-CBP 1-127 whose cross-peaks were affected upon binding to be delimited and a shorter construct to be designed. This shorter construct spanned residues Gly 31 to Ala 90 of murine CBP.
Resonance assignment of the majority of backbone 1 H, 15  ). In addition to the disappearance of a number of cross-peaks, several new cross-peaks appear. These could not be assigned unambiguously but, assuming identical sets of chemical shift differences between resonances of free and complexed 15 N-CBP 31-90 , suggest an altered exchange rate for the smaller construct. Effects on the cross-peaks of Asp 61 , Gln 68 , and Leu 69 cannot be determined precisely because of spectral overlap, but every other cross-peak between Ala 62 and Ser 80 is affected. For the last four residues, cross-peaks appear slightly shifted, whereas all others have disappeared. Some of these may be shifted to the positions of new peaks, although the identity of these could not be confirmed, and line broadening may explain the remainder. Of particular note is the effect on the side chain NH 2 cross-peaks of Gln 68 , the only glutamine residue in the sequence which could be assigned on the basis of chemical shifts of scalar coupled 13 C nuclei in HN(CO)CA and CBCA-(CO)NH experiments. Cross-peaks of residues following Ser 80 are not affected in the complex. Effects on a small number of cross-peaks of residues preceding this stretch indicate further interaction involving sequences outside the continuous segment Asp 61 -Ser 80 . Chemical shift changes for Arg 26  ceding Gly 31 and may therefore be artifactual, although a similar small shift on the cross-peak of Phe 34 in the complex 15 N-CBP 1-127 ⅐PPAR␥-LBD suggests that this is not the case. The changes in the 1 H-15 N HSQC spectrum 15 N-CBP  in the presence of PPAR␥-LBD⅐BMS953 differ from those observed with apoPPAR␥-LBD in that no new cross-peaks are detected (Fig. 5, B and C). Effects on residues in both the N-terminal (Arg 26 -Phe 34 ) and central (Asn 47 -Glu 49 ) regions identified above are diminished.
The 1 H-15 N HSQC spectrum of 15 N-CBP  in the presence of apoPPAR␥-LBD/RXR␣-LBD is essentially identical to that recorded in the presence of only PPAR␥-LBD, suggesting that a similar interaction occurs, possibly involving solely PPAR␥-LBD (Fig. 5D). The presence of BMS953 on PPAR␥-LBD and 9-cisretinoic acid on RXR␣-LBD leads to the disappearance of a more extensive set of cross-peaks than observed with PPAR␥-LBD alone (Fig. 5, B and E). No such extensive effects were observed in the spectrum of 15 N-CBP 1-127 recorded in the presence of RXR␣-LBD/9-cis-retinoic acid (Fig. 4E), indicating that the changes in the spectrum cannot be considered as the sum of two effects. The disappearance of cross-peaks from residues N-terminal to Gly 31 , which above experienced only small changes in chemical shift, suggests a difference in exchange rate between the complexes 15 N-CBP 31-90 ⅐PPAR␥-LBD⅐BMS953⅐RXR␣-LBD⅐9-cis retinoic acid and either 15 N-CBP 31-90 ⅐PPAR␥-LBD/ RXR␣-LBD or 15 N-CBP 31-90 ⅐PPAR␥-LBD⅐BMS953.
Changes in the 1 H-15 N HSQC spectrum of 15 N-CBP  in the presence of ERR␥-LBD, in the absence of ligand, are strictly limited to the segment following Pro 60 (Fig. 5G). In contrast to changes observed in complexes involving PPAR␥-LBD, crosspeaks of Ala 62 and Ala 63 experience changes in chemical shift, as do those of Ser 77 and Gly 78 , whereas all N-terminal residues and both Ser 79 and Ser 80 are unaffected. These results map a more restricted sequence of CBP  involved in the interaction with the LBD for this receptor. Of particular note is the effect of the antagonist, 4-hydroxytamoxifen. The 1 H-15 N HSQC spectrum of 15 N-CBP  in the presence of ERR␥-LBD/ 4-hydroxytamoxifen (Fig. 5, H and I) is characterized by the recovery of all cross-peaks, indicating that the interaction has been abolished.

DISCUSSION
Of the three LXXLL motifs in the sequence of CBP, motifs II and III both belong to well defined structural domains that we judged unlikely to mediate interactions with NR LBDs. Several studies suggested, however, a potential interaction of motif II with NR LBDs (11,29). We therefore examined whether motif II could indeed mediate an interaction with PPAR␥-LBD but found that only fragments of CBP encompassing motif I interact strongly enough to permit copurification. The NR interacting domain of CBP defined here is limited to an N-terminal fragment containing only LXXLL motif I, which therefore appears to be the unique NR box of CBP.
Although larger N-terminal fragments of CBP are prone to aggregation and highly sensitive to proteolysis, even in the presence of PPAR␥-LBD, shorter N-terminal fragments show more stable behavior and form monodisperse complexes with PPAR␥-LBD which can be characterized under native conditions by mass spectrometry and gel electrophoresis. Despite these favorable characteristics, NMR spectra show that these polypeptides do not adopt a well defined tertiary fold. Upon complex formation, it is reasonable to expect an essentially unstructured polypeptide chain to become more structured to some extent, yet the observation of the bound structure depends on the kinetics characterizing the binding event. In favorable cases, a relatively unstructured polypeptide may be observed to fold upon binding (43). Here, complex formation between CBP fragments and NR LBDs manifested itself only in chemical shift changes and line broadening, characteristic of low affinity interactions in which exchange between free and bound species occurs on the s-ms time scale. The interactions are therefore sufficiently strong to allow copurification but too low to provoke the spectral changes that are commonly associated with folding upon binding. Such observations are consistent with the partial dissociation of CBP 1-116 ⅐PPAR␥-LBD during mass spectrometry experiments, whereas the same complex shows a unique migration pattern on native PAGE, and no free species are detected. For comparison, it is interesting to note that the NR interacting domains of p160 coactivators have affinities for NR LBDs in the micromolar range (44,45). Under these conditions, no structural information could be FIG. 4. Effects of binding to NR LBDs on the 1 H-15 N HSQC spectrum of 15 N-CBP 1-127 . In pairs of superposed spectra, the spectrum of 15 N-CBP 1-127 is in black, that in the presence of apoNR LBD is in red and that of holo NR LBD is in green. Right panels (C and F) allow the effects of ligands to be assessed. Plots show the effects of PPAR␥-LBD (A, red), PPAR␥-LBD⅐BMS953 (B, green), RXR␣-LBD (D, red), RXR␣-LBD/9-cis-retinoic acid (E, green). Affected residues are labeled, using assignments from CBP 31-90 . extracted from NMR experiments, and attempts to find more favorable experimental conditions were unsuccessful. Nevertheless, assignment of cross-peaks in the 1 H-15 N HSQC spectra of fragments allowed perturbations observed upon complex formation to be interpreted in site-specific terms.
Not surprisingly, LXXLL motif I is implicated in the binding of N-terminal fragments of CBP to all NR LBDs studied here. The primary role of the LXXLL sequence is supported by the almost complete abolition of binding of CBP 1-101 to the LBDs of ER, RXR, and retinoic acid receptor by mutation of the LSELL motif into LSEAL (29). The interacting segment is, however, considerably larger than the 5-residue motif, involving a continuous 20-residue sequence, well conserved among vertebrates (Fig. 6A) and, in some cases, additional residues quite remote from the central binding motif. In all complexes involving PPAR␥-LBD and that with holoRXR␣-LBD, the 20-residue sequence following Pro 60 is implicated, with the LXXLL motif located centrally within this element. Only in the case of ERR␥ is the extent of this interacting sequence shorter by a pair of residues at each end. Although for complexes involving PPAR␥-LBD additional N-terminal residues are implicated, the interaction with ERR␥-LBD is strictly limited to residues around the LXXLL motif.
The antagonist 4-hydroxytamoxifen induces a conformation of ERR␥-LBD in which helix H12 is dissociated from the core of the domain (46). Abolition of the interaction of CBP  with ERR␥ LBD by 4-hydroxytamoxifen suggests that LXXLL motif I of CBP contacts NR LBDs in a canonical manner and interacts with helix H12 in an agonist conformation. The hydrophobic cleft formed by this helix was identified previously as the site of interaction on PPAR␥ LBD by NMR studies using the peptide A 63 SHKQLSELLRGG 75 from CBP (16). The 20-residue sequence of CBP containing the LXXLL motif may be aligned with NR box II of SRC1, which has been cocrystallized with PPAR␥-LBD (5). 8 residues (CBP residues His 66 , Lys 67 , Leu 69 , Leu 72 , Leu 73 , Gly 76 , Ser 77 , and Ser 79 ) are found to be identical in the SRC1 peptide (Fig. 6B). Among the 9 other nonconserved residues, only 1 (CBP residue Gln 68 , SRC1 residue Ile 689 ) is involved in the interface between PPAR␥-LBD and the bound peptide (Fig. 6C) and may therefore be expected to play a significant role in defining the specificity of CBP toward NR LBDs. The intensities of the side chain amide resonances of Gln 68 in NMR spectra are systematically attenuated on interaction, suggesting an important role for this side chain.
The extent of the sequence around the LXXLL motif of CBP defined here differs slightly from that of NR box II of SRC1 observed in the structure of the complex with PPAR␥-LBD. 2 additional residues on the C-terminal side of the LXXLL motif of SRC1 were observed in the crystal structure, whereas 3 additional residues on the N-terminal side of the motif were defined for CBP. Interestingly, residues C-terminal to the LXXLL motif are involved in defining the specificity of interaction of p160s with NR LBDs (30). In the case of CBP, it appears that residues N-terminal to the LXXLL motif play a similar role: indeed, residues Asp 61 -Ser 64 signals are affected more strongly in complexes involving PPAR␥-LBD than that with ERR␥-LBD.
The portion of the sequence of CBP containing the additional residues involved in interactions with PPAR␥ LBD forms a well conserved block of around 25 residues centered on a consensus L 40 PDEL 44 motif (Fig. 6A), which itself is apparently not directly involved in the interaction. This degree of conservation suggests a functional role for this portion, and anchoring by residues on either side of this 5-residue motif may present it in a manner required for further interaction. Of interest is the presence in this block of a FGSLF sequence, which matches the (F/W)XXL(F/W) motif recently identified by phage display as an AR-binding peptide (47). Remarkably, an FXXLF motif is also present in the N-terminal sequence of AR and mediates FIG. 6. Sequence analysis of CBP homologs and comparison to hSRC1 NR box II. A, sequence alignment of the N-terminal domains of p300/CBPs from various species. The zone surrounding the LXXLL motif (CBP residues 61-80, black line) is well conserved among vertebrates. A second conserved zone (CBP residues 30 -51) contains additional residues (* ϭ Gly 31 , Phe 34 , Glu 37 , Leu 40 , Asn 47 , Gly 48 , Glu 49 ) affected upon complex formation with PPAR␥-LBD. mCBP, murine CBP; hCBP, human CBP; hP300, human p300; rCBP, rat CBP; tetCBP, Tetraodon nigroviridis p300/CBP; molCBP, Aplysia californica p300/CBP; dCBP, Drosophila melanogaster CBP; caeCBP, Caenorhabditis elegans CBP. B, alignment of residues 685-703 of hSRC1 (NR box II), ordered in the crystal structure of the PPAR␥-LBD⅐hSRC1 623-710 complex (5), with residues 61-80 of CBP. Of the 17 aligned residues, 8 are conserved (highlighted in black). C, model for the docking of residues 64 -80 of CBP to the LXXLL binding site of PPAR␥ illustrating the residues of CBP involved in the interface (based on the crystal structure of the PPAR␥-LBD⅐SRC1 complex (5) (PPAR␥-LBD is shown in green and CBP in yellow). Of the 9 other nonconserved residues between CBP 64 -80 and SRC1 685-701 , only one (CBP residues Gln 68 ) is involved in the interface between PPAR␥-LBD and the bound peptide and therefore may be expected to play a role in defining the specificity of CBP toward NR LBDs.
interactions with LBDs (48,49). Because the N-terminal extremity of CBP was shown to mediate direct interaction with both the LBD and the AF1 of PPAR␥ (50), it would be interesting to test whether the FGSLF box or other parts of this conserved block are involved in communication between the AF1 and the LBD.
If the LXXLL motif may be considered to contact LBDs in a canonical manner, it is less straightforward to predict the site of interaction of the N-terminal conserved block of CBP. It could interact with a new surface created by the interaction of the LXXLL motif with the LBD: such an interaction was observed recently in the crystal structure of farnesoid X-receptor, in which two LXXLL peptides interact with the LBD, the first in the hydrophobic groove and the second with the first in a hydrophobic manner (51). On the other hand, the N-terminal conserved block could interact with a different part of the LBD surface: this was observed in the case of AR, where two peptides interact at distinct sites, one with the AF2 and the other elsewhere on the surface of AR-LBD (49). It would now be of interest to investigate the interactions of labeled PPAR␥-LBD with longer peptides, encompassing the full 20-residue segment around the LXXLL motif and the N-terminal conserved block, to map the interaction site on PPAR␥-LBD more completely.
In the absence of agonist, PPAR␥-LBD was shown to be partially disordered, with considerable line broadening affecting the NMR spectrum, suggesting the presence of extensive conformational exchange processes (52). In the crystal structure, helix H12 is observed in an agonist position, showing that this conformation is nonetheless stable enough to be formed in the absence of ligand (5). Binding of a ligand quenches the conformational exchange, suggesting that ligand binding may serve to reduce the entropic cost of binding to coactivators such as CBP, while preserving the mode of interaction.
Finally, because PPARs require heterodimerization with RXR to fulfill their physiological functions, the interaction observed here between CBP and the PPAR␥-LBD/RXR␣-LBD heterodimer in the absence of agonists is intriguing and raises the question as to whether this interaction might play a role in vivo. Interactions of members of the p300/CBP family with PPARs in the absence of ligands have been reported previously. Schulman et al. demonstrated (22) a significant interaction between full-length PPAR/RXR and an N-terminal construct from CBP, by both electromobility shift assay and far Western blotting. Using similar techniques, Dowell et al. (21) showed that p300 interacts with full-length PPAR/RXR bound to a PPAR response element. The same authors also show that a glutathione S-transferase-fused N-terminal construct from p300 allows the pull-down of endogenous apoPPARs from mammalian cells, whereas corepressors such as SMRT or N-CoR do not appear to interact with the heterodimer in vivo (53). Here we observed that although the interaction of CBP with apoRXR␣-LBD is undetectable by NMR, as well as by copurification, N-terminal fragments of CBP most probably contact the apo heterodimer via PPAR␥-LBD.
In conclusion, p160 proteins and CBP interact with NR LBDs in quite distinct ways, despite the use of a common motif. CBP utilizes a more extensive segment of protein sequence around the LXXLL motif than p160 proteins in interacting with the NR LBDs studied here, suggesting a binding mode governed by the LXXLL motif but modulated by flanking residues and possibly fine tuned by additional interactions involving the N-terminal conserved block.