Structural Diversity in p160/CREB-binding Protein Coactivator Complexes*

Ligand-induced transcription by nuclear receptors involves the recruitment of p160 coactivators such as steroid receptor coactivator 1 (SRC1), in complex with histone acetyltransferases such as CREB-binding protein (CBP) and p300. Here we describe the solution structure of a complex formed by the SRC1 interaction domain (SID) of CBP and the activation domain (AD1) of SRC1, both of which contain four helical regions (Cα1, Cα2, Cα3, and Cα3′ in CBP and Sα1, Sα2′, Sα2, and Sα3 in SRC1). A tight four-helix bundle is formed between Sα1, Cα1, Cα2, and Cα3 that is capped by Sα3. In contrast to the structure of the AD1 domain of the related p160 protein ACTR in complex with CBP SID, the sequences forming Sα2′ and Sα2 in SRC1 AD1 are not involved in the interface between the two domains but rather serve to position Sα3. Thus, although the CBP SID domain adopts a similar fold in complex with different p160 proteins, the topologies of the AD1 domains are strikingly different, a feature that is likely to contribute to functional specificity of these coactivator complexes.

The recruitment of CBP to target gene promoters/enhancers facilitates acetylation of histone N-terminal tails, leading to chromatin remodeling and enhanced gene expression. This has been demonstrated for nuclear receptors, which activate transcription of their target genes in response to ligand binding (3,4). Ligand-bound receptors undergo a conformational change that stimulates their interaction with cofactors that contain functional LXXLL motifs, such as the p160 coactivators SRC1, TIF2, and ACTR (5,6). Studies of steroid-regulated gene promoters have revealed that p160s and HAT proteins are among the first cofactors recruited in response to ligand (7,8). Such temporally ordered recruitment of coactivators to promoters/enhancers is crucial for the sequential chromatin modification and remodeling events preceding transcription (4).
The efficacy of nuclear receptor/cofactor interaction is influenced by a number of determinants, including the precise sequence and number of LXXLL motifs, the sequences flanking core motifs, and other distal sequences (9 -11). CBP contains three LXXLL motifs, although they mediate only weak direct interactions with estrogen, androgen, and progesterone receptors (9,10). Thus, at least in the case of the steroid receptors, efficient recruitment of CBP/p300 and associated factors is achieved indirectly via the p160 proteins (12).
A number of studies have shown that recruitment of CBP/p300 proteins is facilitated by the p160 activation domain AD1, which acts as a potent transcriptional activator in mammalian and yeast cells (3,(12)(13)(14)(15)(16)(17). The AD1 domain docks with the SRC1 interaction domain (SID) of CBP, located within the sequence 2058 -2130 (12), and this interaction is essential for ligand-dependent transcription mediated by steroid receptors (12,18). In addition to binding p160s, the SID also facilitates interactions of CBP with activation domains in other nuclear proteins, including the transcription factors Ets1, Ets2, p53, and IRF3, and viral activators such as E1A, KSHV IRF1, and Tax (19 -22). Indeed, it has been shown that competition between such proteins for binding to the SID contributes to the negative cross-talk observed between different signaling pathways (22). Similarly, binding of viral proteins to the SID and other CBP/p300 domains not only facilitates viral gene transcription but may also down-regulate expression of host defense genes, through exclusion of host factors from CBP-p300 complexes.
The AD1/SID interaction is also important in MOZ-TIF2-associated leukemogenesis. MOZ-TIF2 is a fusion protein expressed in acute myeloid leukemia blasts containing the inv8(p11;q13) translocation (23). This fusion protein contains the C-terminal sequence of TIF2, including the AD1 domain, facilitating its interaction with CBP/p300, as demonstrated by in vitro interactions and in vivo by using fluorescence resonance energy transfer experiments (24). As a consequence of this interaction, CBP is mislocalized from promyelocytic leukemia bodies, and cellular levels of CBP are depleted, leading to a reduced transcriptional activity of CBP-dependent activators such as nuclear receptors and p53 (24). Consistent with this, AD1 integrity was found to be essential for transformation of hematopoietic progenitors by MOZ-TIF2 in vitro (24,25) and for induction of acute myeloid leukemias by MOZ-TIF2 in mice (25). Therefore, understanding the structure of the CBP-p160 complexes has relevance to the etiology of acute myeloid leukemia.
Phenotypic differences in knock-out mice indicate that p160s have tissue-specific functions (reviewed in Ref. 26). Similarly, CBP and p300 proteins appear to have distinct roles in vivo, for example in hematopoiesis (27). Thus, the existence of different p160-CBP or p160-p300 complexes in vivo suggests they may have specific albeit partially redundant functions. For example, SRC1 and ACTR interact with thymine DNA glycosylase, whereas TIF2 does not. This is because of the presence of a tyrosine repeat motif (YXXY) in ACTR and SRC1, which is not conserved in the TIF2 sequence (28). Chromatin immunoprecipitation assays investigating cofactor recruitment at the pS2 gene promoter have indicated that the presence of one p160 can exclude recruitment of others (7), although the molecular basis of this is unknown. Another study provided evidence that CBP-p160 complexes are functionally distinct, as it was observed that different p160 combinations can be detected on androgen receptor target gene promoters, possibly through the formation of specific p160 heterodimer pairs (29). Thus, determination of the structures of different complexes will be essential to understand how such selectivity is achieved. In this study, we describe a high resolution solution structure of the complex formed between the SID domain of CBP and the AD1 domain of SRC1. By comparison to the CBP SID/ACTR AD1 structure (30), we show that although the structure of CBP SID is strikingly similar in both complexes, p160 proteins adopt distinct conformations that are likely to be important for their different biological functions.

EXPERIMENTAL PROCEDURES
Protein Expression/Purification-To express the CBP/SRC1 interaction domain complex in Escherichia coli, we obtained a modified pET22b dual expression vector, containing sequences encoding the ACTR AD1 and CBP SID domains, preceded by independent ribosome-binding sites (a gift from Peter Wright, Scripps Institute). To generate the CBP SID/ SRC1 AD1 expression vector, the ACTR AD1 sequence was removed by restriction digestion with NcoI and HindIII and replaced with an NcoI/ HindIII-digested PCR fragment containing the coding region for residues 920 -970 of SRC1 followed by a thrombin cleavage site (underlined) and a polyhistidine tag (KLVPRGSLEHHHHHH). The mouse CBP SID-(2059 -2117) and human SRC1 AD1-(920 -970) dual expression vector was transformed into E. coli strain B834, which was used to produce either unlabeled, 15 N-labeled, or 15 N/ 13 C-labeled samples of the CBP SID-SRC1 AD1 complex. For labeled samples, cells were grown in minimal media containing 0.6 g/liter [ 15 N]ammonium sulfate and 2 g/liter [ 13 C]glucose as the sole nitrogen and/or carbon sources. The medium was also supplemented with 50 mg/liter unlabeled methionine, as B834 is a methionine auxotroph. Transformants were grown at 37°C, and expression of the two proteins forming the complex was induced in mid-log phase by the addition of 1 mM isopropyl-1-thio-␤-D-galactopyranoside. The cells were harvested 3 h post-induction, resuspended in 20 mM sodium phosphate, 100 mM sodium chloride, and 5 mM imidazole buffer (pH 7.0). Lysis was achieved by sonication. Insoluble material was removed by high speed centrifugation, and the cleared lysate was filtered through a 0.2-m filter prior to chromatography. The histidinetagged CBP SID-SRC1 AD1 complex was affinity-purified on a HiTrap chelating HP column (Amersham Biosciences) charged with Ni 2ϩ and eluted with a linear gradient of imidazole (5-250 mM). Fractions con-taining the complex were pooled and subjected to a final purification by gel filtration on a Superdex 75 16/60 pre-packed column (Amersham Biosciences). Typical yields of the Ͼ95% pure complex were 10 mg/liter.
Reverse Phase HPLC-Reverse phase HPLC analysis was performed on a C4 column. The individual polypeptides were resolved by applying a two-step linear gradient of acetonitrile (1.6 -80%), and the relative amount of each was quantified by absorbance at 215 nm. Eluted peaks were identified by mass spectrometry.
Circular Dichroism Spectroscopy-Far UV CD spectra were obtained from 0.4 mM samples of the CBP SID-SRC1 AD1 complex in a 20 mM sodium phosphate, 100 mM sodium chloride buffer (pH 7.0). Samples were placed in a 0.1-mm path length cell, and spectra were recorded from 180 to 250 nm at a resolution of 1 nm and a scan speed of 20 nm/min, with each spectrum representing the average of 10 accumulations. Prior to analysis, CD spectra were corrected for buffer absorbance, and the raw data were converted to molar CD per residue.
NMR Spectroscopy-NMR spectra were acquired from 0.35-ml samples of 1.5 mM CBP SID-SRC1 AD1 complex in a 20 mM sodium phosphate, 100 mM sodium chloride, 10 M EDTA, and 0.02% (w/v) sodium azide buffer (pH 7.0), containing either 10% D 2 O, 90% H 2 O, or 100% D 2 O as appropriate. All NMR data were acquired at 25°C on either an 800-MHz Varian Inova or a 600-MHz Bruker Avance spectrometer. The two-dimensional and three-dimensional spectra recorded to obtain sequence-specific assignments for CBP SID and SRC1 AD1 in complex were as follows: 1 H TOCSY with mixing times of 40 and 55 ms (31) and NOESY with an NOE mixing time of 100 ms (32); 15 N/ 1 H HSQC; TOCSY-HSQC with a mixing time of 50 ms and NOESY-HSQC with an NOE mixing time of 100 ms (33); 13 C/ 1 H HCCH-TOCSY with a mixing time of 20 ms (34), HMQC-NOESY with an NOE mixing time of 100 ms (35); and 15 N/ 13 C/ 1 H HNCACB (36) and CBCA(CO)NH (37). Typical acquisition times in F 1 and F 2 for the three-dimensional experiments were 11-13 ms for 15 N, 7.5-9.5 ms for 13 C, and 15 ms for 1 H, and an acquisition time of 75 ms in F 3 ( 1 H). The majority of the three-dimensional spectra were collected over ϳ88 h, two-dimensional 1 H experiments over 8.5-24 h, and 15 N/ 1 H HSQC spectra over about 30 min. Typical acquisition times in two-dimensional experiments were either 70 ( 15 N) or 35 ms ( 1 H) in F 1 and 250 ms in F 2 ( 1 H). The WATERGATE method (38) was used to suppress the water signal when required. The three-dimensional NMR data were processed using NMRPipe (39) with linear prediction used to extend the effective acquisition times by up to 1.5-2-fold in F 1 and F 2. The spectra were analyzed using the XEASY package (40).
Structure Calculations-A family of converged CBP SID/SRC1 AD1 structures was determined in a two-stage process using the program CYANA (41). Initially, the combined automated NOE assignment and structure determination protocol (CANDID) was used to automatically assign both the intra-and intermolecular NOE cross-peaks identified in three-dimensional 15 N-and 13 C-edited NOESY spectra. This approach provides a completely unbiased assignment of the NOE peaks, in which all peaks are evaluated as either intra-or intermolecular in origin. Subsequently, several cycles of simulated annealing combined with redundant dihedral angle constraints were used to produce the final converged CBP SID/SRC1 AD1 structures (42). The input for the CANDID stage consisted of essentially complete 15 N, 13 C, and 1 H resonance assignments for the nonexchangeable groups in the CBP SID-SRC1 AD1 complex, two manually picked three-dimensional NOE peak lists corresponding to all NOEs involving amide protons (1179) and all NOEs between aliphatic protons (2371), and one manually picked twodimensional NOE peak list corresponding to all NOEs involving aromatic side chain protons (127). In addition, the CANDID stage included 80 torsion angle constraints for CBP SID and 64 torsion angle constraints for SRC1 AD1 obtained from analysis of chemical shift data using TALOS (43). Hydrogen bond constraints were also added for regions predicted to contain regular ␣-helical secondary structure by both the NOE and chemical shift data, which corresponds to residues Ala 2067 -Leu 2075 , Gln 2082 -Lys 2092 , and Ala 2099 -Ala 2107 of CBP SID and Glu 929 -Ser 941 of SRC1 AD1. The peak lists were prepared using XEASY, and the intensities of peaks were obtained using the "interactive integration" routine (40). CANDID calculations were carried out using the default parameter settings in CYANA 1.0.6, with the upper limit for NOE-derived distance constraints set at 5.5 Å and chemical shift tolerances set to 0.02 ppm (direct and indirect 1 H) and 0.3 ppm ( 15 N and 13 C). The final converged CBP SID/SRC1 AD1 structures were produced from 100 random starting coordinates using a standard torsion angle-based simulated annealing protocol combined with 5 cycles of redundant dihedral angle constraints (44,45). The calculations were based upon 1855 nonredundant NOE-derived upper distance limits (maximum value 6.0 Å), assigned to unique pairs of protons using CAN-DID, 142 ⌽ and ⌿ torsion angle constraints derived from TALOS, and 104 hydrogen bond constraints for helical residues. Analysis of the family of structures obtained was carried out using the programs CYANA and MOLMOL (41,46) Sequence-specific Assignments-Sequence-specific backbone resonance assignments (N, NH, C␣, and C␤) were obtained for the CBP SID-SRC1 AD1 complex from the identification of intra-and interresidue connectivities in HNCACB, CBCA(CO)NH, and 15 N/ 1 H NOESY-HSQC spectra. Assignments were then extended to the side chain signals using correlations observed primarily in 15 N/ 1 H TOCSY-HSQC and 13 C/ 1 H HCCH-TOCSY, with additional supporting evidence provided by 15 N/ 1 H NOESY-HSQC and 13 C/ 1 H HMQC-NOESY spectra where required.

RESULTS AND DISCUSSION
Expression and Purification of the SRC1-CBP Complex-We previously used yeast two-hybrid and in vitro pulldown experiments to define the minimal sequences required for interaction of the CBP SID and SRC1 AD1 domains (12). A modified pET22b dual expression vector (30), containing sequences encoding the SRC1 AD1-(920 -970) and CBP SID-(2059 -2117) domains, was used to coexpress these polypeptides in E. coli. The addition of a polyhistidine tag at the C terminus of the AD1 domain facilitated purification of the complex, as described under "Experimental Procedures." A highly purified complex containing polypeptides of the expected size was obtained, as determined by SDS-PAGE (Fig. 1A). Far UV CD spectra acquired for the CBP SID-SRC1 AD1 complex indicated that the complex was predominantly helical, having characteristic negative ellipticity peaks at ϳ209 and 221 nm and a large positive peak at 195 nm (Fig. 1B). Analysis of the spectra using the CDPro software package (47) indicated that the complex con- tained 54% (Ϯ7.3%) helical, 20% (Ϯ6.1%) turn, and 25% (Ϯ5.7%) random coil secondary structure.
Reverse phase HPLC analysis was performed to determine the stoichiometry of the purified complex. The eluted peaks were identified by mass spectrometry (data not shown). Quantification of the peptide bond absorbance from both peaks confirmed the formation of a 1:1 CBP SID-SRC1 AD1 complex (Fig. 1C).
Sequence-specific Assignments and Structure Calculation for the CBP-SRC1 Complex-Very comprehensive sequence-specific resonance assignments were obtained for the CBP SID-SRC1 AD1 complex despite the relatively poor dispersion observed in spectra, which is illustrated by the HSQC spectrum (Fig. 1D). For example, backbone amide assignments were obtained for all non-proline residues in the complex except as follows: Asn 2060 , Arg 2061 , and Gln 2117 in CBP SID; Asn 927 in SRC1 AD1 (96%); and for all C␣ and C␤ signals apart from the two unlabeled methionines and two residues in CBP SID (Pro 2059 and Gln 2117 ) (96%).
The CANDID protocol was effective in determining unique assignments for the NOEs identified in the three-dimensional 15 N-and 13 Cedited NOESY and the aromatic to aliphatic region of the two-dimensional NOESY. At the end of the final cycle, unique assignments were obtained for 89.8% (1059/1179) of the NOE peaks picked in the 15  The CBP SID and SRC1 AD1 domains are intimately associated in the complex, with C␣1, C␣2, C␣3, and S␣1 forming a four-helix bundle (Fig. 2B). The overall interface between the two domains is substantial, corresponding to a solvent-inaccessible surface of 1019 Å 2 on CBP SID and 1088 Å 2 on SRC1 AD1. Helices S␣2Ј and S␣2 pack together on one side of the complex to stabilize the corner region of the "L"-shaped domain and are not involved in the interface with CBP SID. In contrast, S␣3 serves to cap the four-helix bundle region and forms contacts with C␣2, C␣3, C␣3Ј, and the two serine residues in the PSSP turn of CBP SID. The first two helices of CBP SID (C␣1 and C␣2) lie almost antiparallel to each other and are separated by a well defined five-residue turn consisting of the sequence SPSSP. The N-terminal portion of C␣2 contains a stretch of five glutamines preceding the more hydrophobic C-terminal half of the helix. C␣3 is inclined away from C␣1 and C␣2 exposing a hydrophobic groove between C␣1 and C␣3, which accommodates S␣1 of the SRC1 AD1 domain. Helices C␣3 and C␣3Ј are separated by a short region of irregular structure and resemble a single bent helix. The stabilizing interactions between C␣1 and C␣2 and between C␣2 and C␣3 appear to involve no ionic or hydrogen bonds but rely on favorable van der Waals contacts, primarily involving residues Ala 2067 , Leu 2071 , and Thr 2074 from C␣1; Gln 2083 , Val 2087 , Ile 2090 , and Leu 2091 from C␣2; and Leu 2097 , Phe 2101 , and Thr 2106 from C␣3. Asn 2094 also makes van der Waals interactions with both C␣2 and C␣3.
The SRC1 AD1 domain also appears to be primarily stabilized by van der Waals interactions, involving residues Ala 931 and Gln 935 from S␣1, The fourth helix of SRC1 AD1 (S␣3) fits into a second hydrophobic groove in CBP SID located between the well defined SPSSP turn linking C␣1 and C␣2 and the C terminus of C␣3. This interaction is stabilized by hydrophobic contacts involving Leu 955 , Ile 957 , Leu 960 , Val 961 , Gln 962 , and Gly 964 of SRC1 AD1 and Ser 2079 , Ser 2080 , Gln 2084 , Arg 2105 , Lys 2108 , and Tyr 2109 of CBP SID. The five C-terminal residues of SRC1 AD1 appear to loop back toward S␣3 and C␣3Ј, leading to some van der Waals contacts between Leu 969 of SRC1 AD1 and Tyr 2109 of CBP SID. However, this loop is primarily stabilized by intramolecular van der Waals interactions involving Gln 962 and residues within a C-terminal linker sequence originating from the expression vector. In the absence of the linker sequence, contacts between Leu 969 and Tyr 2109 may not occur, and the C-terminal region of SRC1 AD1, including S␣3, may make additional contacts within the complex, or with other domains/proteins. Some features of the NMR data obtained for the CBP SID-SRC1 AD1 complex clearly show the presence of a second minor conformational state, which is in exchange with the structure reported here (Fig. 1D). This is clearly indicated by the presence of cross-peaks between signals from backbone amide groups in 15 N/ 1 H TOCSY-HSQC spectra, which arise through chemical exchange processes (48). This also results in exchange broadening of a significant number of backbone amide signals. Interestingly, only a few of the TOCSY exchange peaks involve residues in CBP SID, whereas almost all of the residues from Leu 936 -Val 961 in SRC1 AD1 are affected, which corresponds to the six C-terminal residues of S␣1 through to the end of S␣3. This suggests that the SRC1 AD1 domain has a degree of conformational instability, even when bound to CBP SID, which may facilitate the rapid dissociation of the complex. However, we do not exclude the possibility that other protein/protein interactions may stabilize complexes of the full-length proteins.
Comparison of CBP-SRC1 and CBP-ACTR Complexes-The solution structure of CBP SID in complex with the AD1 domain from ACTR has been reported previously (30). Comparison of the two complexes revealed that they have both conserved and distinct structural features (Fig. 3, A and B). The SID adopts similar secondary and tertiary structures, when bound to different AD1 domains (Fig. 3, A and B), as highlighted by the superposition of the SID polypeptide backbone atoms (residues Ile 2063 , Pro 2065 -Ala 2067 , Asp 2070 -Ser 2079 , and Gln 2085 -Lys 2108 ), which yields an r.m.s.d. value of 1.79 Å (Fig. 3C). The most notable difference between the CBP SID polypeptides in the two complexes is the conformation of the sequence composed of five glutamine residues (2082-2086). In the CBP-SRC1 complex this region forms the N terminus of an extended C␣2 helix (Fig. 3, A and C), whereas in the CBP-ACTR complex the three proximal glutamine residues were pro- posed to form part of a solvent-exposed loop (designated poly(Q); Fig. 3, B and C). This difference may be due to the paucity of NMR signal assignments for residues in this region in the CBP-ACTR complex (30), resulting in fewer NMR constraints on the conformation rather than a true structural difference. Another difference between the CBP SID domains is C␣3, which is 17 residues long (Pro 2095 -Ala 2111 ) in the CBP/ ACTR structure, although in the CBP-SRC1 complex this region forms two shorter helices (C␣3, Pro 2095 -Thr 2106 ; C␣3Ј, Tyr 2109 -Asn 2112 ) resembling a kinked helix. The C␣3Ј helix is inclined away from the rest of the four-helix bundle, which is not observed in the ACTR-AD1 complex.
Structural similarity between p160 AD1 domains in the two complexes is confined to the N-terminal helices A␣1 and S␣1, which form four-helix bundles with the CBP SID (Fig. 3, A and B). The remaining sequences in the AD1 domains adopt strikingly distinct topologies as highlighted by superimposing the AD1 polypeptide structures (Fig. 3D). This is also reflected in the significant chemical shift differences for NMR signals from equivalent residues in the two AD1 domains and also for CBP SID residues that make distinct contacts with AD1 in the two complexes, in particular, residues in the C␣3/C␣3Ј region (data not shown). Although SRC1 AD1 contains three further helical regions (residues S␣2Ј, Glu 945 -Leu 948 ; S␣2, Glu 950 -Ser 954 ; and S␣3, Ile 957 -Gln 962 ), ACTR AD1 contains only two additional helices (A␣2, Leu 1064 -Leu 1071 , and A␣3, Ile 1073 -Gln 1079 ) (Fig. 3D). These distinct topologies are somewhat unexpected given that SRC1 and ACTR share 53% sequence identity and 67% sequence similarity within the AD1 domain. In the CBP SID-ACTR AD1 complex, A␣2 and A␣3 wrap around C␣3 making extensive contacts with this helix, whereas A␣2 occupies a hydrophobic groove between the proposed poly(Q) loop and C␣3 (Fig. 3B). In the CBP SID-SRC1 AD1 complex, only the C-terminal residue (Leu 955 ) of S␣2 (corresponding to Leu 1071 in A␣2) makes any contact with the SID domain (Fig. 3A). Instead, S␣3 fills the groove between the PSSP turn and C␣3, effectively capping the four-helix bundle (Fig. 3A). This contrasts with the position of corresponding helix of ACTR (A␣3), which packs against the adjacent hydrophobic face of C␣3 (Fig. 3B).
The CBP SID-ACTR AD1 complex was reported to contain a salt bridge between residues Asp 1068 in A␣2 and Arg 2105 in C␣3, coordinating a hydrogen-bonding network involving Asp 1060 and Arg 2105 and Asp 1068 and Tyr 2109 . This buried charged interaction has been proposed to be important for the specificity of the CBP SID/ACTR AD1 interaction, although it does not contribute directly to the stability of the complex (49). Both of these aspartate residues are conserved in SRC1 AD1 (Asp 944 and Asp 952 ); however, they are spatially distant from the C␣3 helix. Asp 944 is present in the loop between S␣1 and S␣2Ј, and Asp 952 is contained within the S␣2 helix (Fig. 4C). However, it is possible that Asp 944 forms a salt bridge with Lys 2076 from C␣1. Similarly, an aspartate residue (Asp 965 ) near the C terminus of SRC1 AD1 may form a salt bridge with Arg 2105 , which is involved in van der Waals contacts between C␣3 and S␣3. As would be expected from the tertiary structures of their respective complexes, the solvent-inaccessible surface areas of CBP SID (1681 Å 2 ) and ACTR AD1 (1839 Å 2 ) domains are greater than those observed for the complex of CBP SID (1019 Å 2 ) and SRC1 AD1 (1088 Å 2 ).
In summary, the backbone structure of the CBP SID and the first helix of AD1 are very similar in both complexes, as reflected in a backbone atom r.m.s.d. value of 2.05 Å (calculated by superimposition of residues Ile 2063 , Pro 2065 -Ala 2067 , Asp 2070 -Ser 2079 , and Gln 2085 -Lys 2108 of CBP SID and residues Asp 928 -Ser 941 and Asp 1044 -Ser 1057 of SRC1 AD1 and ACTR-AD1, respectively). The remainder of the p160 AD1 domains adopt very distinct folds, although the C-terminal helices of AD1 appear to be essential to stabilize the complex.
Structural Flexibility of the CBP SID Permits Complex Formation with Multiple Partners-The NMR structure of the CBP SID in isolation (also termed IBiD because of its interaction with IRF3) revealed it to have significant ␣-helical content (19), although the conformation of the isolated CBP SID has been proposed to be consistent with that of a molten FIGURE 4. Critical residues in SRC1 AD1 required for CBP binding. A, sequence alignment of p160 AD1 domains and CBP-binding sequences from viral proteins E1A and TAX. For the viral proteins, low homology sequences are represented in gray, and regions showing homology to helices S␣1/A␣1 and S␣3/A␣3 are in black. SRC1 residues, which were subjected to replacement with alanine and which are required for CBP binding/transcriptional activity of AD1, are highlighted in red. SRC1 residues highlighted in blue had little effect on AD1 transcriptional activity when replaced with alanine (see C, below). The black boxes over the alignments indicate regions of helical secondary structure observed in complexes for SRC1 (S␣1, S␣2Ј, S␣2, and S␣3) and ACTR (A␣1, A␣2, and A␣3). B, reporter (luciferase) assay showing the relative transcriptional activities of GAL4-AD1 (SRC1) or mutants thereof containing alanine replacements for the residues indicated. Reporter activity is presented as fold induction over control (GAL4-DBD) and was normalized to a cotransfected ␤-galactosidase reporter. The data presented are averages from triplicates, and the error bars indicate the S.D. DBD, DNA binding domain; WT, wild type. C, ribbon representation of the backbone topology of CBP SID-SRC1 AD1 complex. The side chains of residues involved in two potential inter-molecular salt bridges (Lys 2076 of CBP SID and Asp 944 of SRC1 AD1; Arg 2105 of CBP SID and Asp 967 of SRC1 AD1) are highlighted in purple. Note that Arg 2105 and Asp 952 are spatially distant and cannot form a salt bridge equivalent to that observed between Arg 2105 and Asp 1086 in the CBP-ACTR complex. The side chains of residues Asp 952 , Asp 958 , and Asp 960 , which are required for AD1 transcriptional activity, are represented in green.
globule (50). Isolated AD1 polypeptides have little if any intrinsic structure (50) 6 but appear to undergo induced folding in complex with CBP SID (30). An important question is whether short conserved structural motifs promote similar modes of binding of different proteins to CBP, or whether different complexes adopt drastically different conformations that influence the overall structure of CBP/p300 proteins, which in turn influences combinatorial complex formation.
This study reveals structural similarity within the ␣1 helix of two p160 AD1 domains, which contributes to a four-helix bundle in both the CBP-SRC1 and CBP-ACTR complexes. We have shown previously that the LLXXLXXXL sequence motif within ␣1 is conserved in other SID-binding proteins (see Fig. 4A), and that disruption of this sequence in Ets2 and E1A abrogates their binding to the CBP SID (22). Fusion of the SRC1 AD1 sequence to a GAL4 DNA binding domain permits very potent activation of a GAL reporter gene in mammalian cells, through recruitment of endogenous CBP and p300 (12). As shown in Fig. 4B, replacement of the Leu 932 , Leu 933 , and Leu 936 residues in S␣1 with alanines resulted in loss of AD1 transcriptional activity. This highlights the requirement of the S␣1 for interaction with full-length CBP/p300 in vivo.
The C-terminal helix (␣3) of the SRC1 and ACTR AD1 domains also plays a critical role in formation of the two complexes with CBP SID as discussed above. Interestingly, the sequence of the S␣3 helix of SRC1 AD1 ( 957 IDKLV 961 ) is divergent from that in ACTR and TIF2 (IPELV) (Fig. 4A). The IPELV sequence in ACTR and TIF2 resembles the LPXL motif that has been shown to mediate the interaction of CITED and HIF1␣ transcription factors with the CH1 domain of CBP (51). Mutagenesis of the SRC1 AD1 domain in the context of GAL4-AD1 revealed that the D958A or L960A mutations resulted in almost complete loss of reporter activation, whereas K959A retained almost full transactivation function (Fig. 4B). In addition, reporter activation by the V961A mutant was reduced by 70%, suggesting it is important but not essential for CBP/p300 recruitment (Fig. 4B). These results confirm the importance of S␣3 and key residues within that helix for CBP/p300 recruitment.
Replacement of the conserved aspartate residues (D944A and D952A) also resulted in loss of reporter activation by GAL4-AD1 (Fig. 4B). Thus, as suggested for the CBP-ACTR complex (30,49), salt bridge formation may be important in stabilizing the higher order folding of CBP-p160 complexes. There is evidence to suggest that SRC1 sequences outside of the AD1 domain may influence the interaction with CBP. The signaling molecule 8-bromo-cAMP induces mitogen-activated protein kinasedependent phosphorylation of SRC1 at Thr 1179 and Ser 1185 , which leads to increased ligand-independent transcriptional activity of the progesterone receptor and enhanced CBP recruitment (52,53). Similarly, the interaction of ACTR with CBP was abrogated by mutation of residues outside of the AD1 domain, which are known to be phosphorylated in vivo (Thr 24 , Ser 543 , Ser 857 , Ser 860 , and Ser 867 ) (53). This may reflect binding between additional regions of SRC1 and CBP and/or conformational changes in SRC1 that effect the affinity of the AD1 region for CBP.
We have indicated previously that several CBP SID-binding proteins contain low level sequence homology in the regions of SRC1 AD1 corresponding to S␣1, and to a lesser extent S␣3, that may indicate conserved modes of binding to CBP. An alignment of the sequences of interaction domains of several CBP SID-binding proteins is shown in Fig. 4A. The transactivation domain of human T-cell leukemia Tax protein contains a short sequence ( 312 YTNIPISL 319 ) that is required for binding to CBP SID (21). This sequence may form an amphipathic ␣-he-lix analogous to S␣3 of SRC1 AD1 and play an equivalent role in stabilizing CBP/Tax interactions (Fig. 4A). The adenoviral oncoprotein E1A revealed also sequences essential for CBP binding, which may be functionally and structurally equivalent to SRC1 AD1 S␣1 and S␣3 (22) (Fig.  4A). IRF3 contains a sequence termed the IRF association domain (IAD), which mediates its interactions with CBP SID, and the crystal structure of this complex was recently reported (54). Little structural similarity is evident between the CBP-IRF3 complex and the CBP-p160 structures. Contacts between the IAD and SID are mediated by hydrophobic contacts involving C␣1, C␣2, and C␣3 in CBP SID with two amphipathic ␣-helices (H3 and H4) in the IAD containing IXXLI and LXXLVXXXV motifs, respectively. Interestingly, the secondary structure of the SID is similar to that observed in p160 complexes, and we note that the poly(Q) region of CBP is comprised within the C␣2 helix, as in the CBP-SRC1 complex. Thus the SID appears to be able to fold into distinct conformations to accommodate binding to different transcription factors.
In conclusion, the NMR structure reported here reveals both similar and diverse structural features in CBP SID-p160 AD1 complexes, which may reflect functional differences between p160 proteins. Structural versatility of the CBP SID is likely to underpin its ability to assemble different transcription factor-cofactor complexes in a promoterdependent context.