Structure of p300 bound to MEF2 on DNA reveals a mechanism of enhanceosome assembly.

Transcription co-activators CBP and p300 are recruited by sequence-specific transcription factors to specific genomic loci to control gene expression. A highly conserved domain in CBP/p300, the TAZ2 domain, mediates direct interaction with a variety of transcription factors including the myocyte enhancer factor 2 (MEF2). Here we report the crystal structure of a ternary complex of the p300 TAZ2 domain bound to MEF2 on DNA at 2.2Å resolution. The structure reveals three MEF2:DNA complexes binding to different sites of the TAZ2 domain. Using structure-guided mutations and a mammalian two-hybrid assay, we show that all three interfaces contribute to the binding of MEF2 to p300, suggesting that p300 may use one of the three interfaces to interact with MEF2 in different cellular contexts and that one p300 can bind three MEF2:DNA complexes simultaneously. These studies, together with previously characterized TAZ2 complexes bound to different transcription factors, demonstrate the potency and versatility of TAZ2 in protein-protein interactions. Our results also support a model wherein p300 promotes the assembly of a higher-order enhanceosome by simultaneous interactions with multiple DNA-bound transcription factors.

INTRODUCTION

CBP and p300 are transcription co-activators of diverse gene expression programs (1). Their essential roles have been demonstrated by genetic studies and diseases associated with mutations in CBP/p300. Mice deficient in p300 lack cardiomyocyte proliferation and muscle-specific gene expression (2), whereas elevated p300 levels lead to myocardial hypertrophy (3). Mutations in CBP cause Rubinstein–Taybi Syndrome, a neurological disorder characterized by severe mental retardation (4). Disruptions of CBP and p300 regulation and function have also been linked to numerous cancers (5). CBP/p300 enhances transcription by facilitating the assembly of activating transcription complexes at the promoter and by modifying chromatin structure through its histone acetyltransferase (HAT) activity that can also modify non-histone proteins. These large nuclear proteins do not bind DNA but contain multiple domains capable of interacting with a variety of transcription factors (6), including the KIX domain that binds CREB and a number of cysteine–histidine rich regions (CH1–3), each of which binds directly to a large number of proteins (7). The CH3 region, which contains the TAZ2 domain, has been shown to interact with more than 20 proteins including viral protein E1A and host transcription factor p53 (8–18). Thus, CBP/p300 is extremely versatile in engaging numerous transcription factors to regulate gene expression (19–21). However, the molecular basis for this remarkable versatility is not well understood.

Extensive studies have established MEF2 as a major transcription factor partner of CBP/p300 in muscle, neurons and T cells (22–26). MEF2 is a class of transcription factors highly conserved in eukaryotes (27,28). Earlier genetic studies have demonstrated a central role of MEF2 in myogenesis (29–32). It is now clear that the MEF2 proteins (MEF2A-D) have broad roles in the differentiation, proliferation, and survival/apoptosis of a wide range of cell types (28,33,34). MEF2 also serves as a key regulator of stress responses and adaptive programs in animals, including fiber-type switch of skeletal muscle, hypertrophic growth of heart and activity-dependent remodeling of neuronal circuitry (35–39).

MEF2 turns on and off gene expression in a calcium-dependent manner (40). In the resting state, MEF2 recruits co-repressors such as Cabin1/Cain and histone deacetylases (HDAC) to specific loci of the genome to inhibit the expression of target genes (41–48). Upon activation, the co-repressors dissociate from MEF2 via calcium-dependent mechanisms (33,40,49–52); the DNA-bound MEF2 subsequently interacts with other calcium-activated transcription factors (e.g. NFAT and CREB) and recruits co-activators such as CBP/p300 and myocardin to activate transcription (1,2,22–26,53,54). The HDACs and CBP/p300 also regulate the transcriptional activity of MEF2 directly by controlling the acetylation state of specific lysine residues of MEF2 (55,56). The overall transcription state of MEF2-bound promoters is tightly controlled by signal-dependent protein–protein interactions between MEF2 and different co-regulators. Establishing how MEF2 interacts with these co-regulators will be a key to understanding the mechanisms of MEF2-regulated transcription.

The MEF2 family of transcription factors shares a highly conserved domain at the N-terminus, known as the MADS-box/MEF2 domain that mediates DNA binding, dimerization and protein–protein interactions with a variety of protein partners, including transcriptional co-repressors. Systematic structural studies reveal that a short amphipathic helix conserved in Cabin1 and class IIa HDACs (HDAC4, 5, 7 and 9) (referred to as MEF2-binding motif hereafter) binds a hydrophobic groove of MEF2 (53,57,58). These studies established a structural model for co-repressors to bind to MEF2 through their core interaction domains. Here, we characterized the interactions between MEF2 and its co-activator p300 by structural and biochemical analyses. These studies reveal the first atomic model of sequence-specific recruitment of transcription co-activator p300 by a DNA-bound transcription factor and provide new insights into the p300-mediated enhanceosome assembly.

MATERIALS AND METHODS

Materials

DNA plasmids of human MEF2A, Gal4-luciferase reporter, Gal4-MEF2D were kindly provided by Dr Xiangjiao Yang (McGill University). pET28a, pET30b and Rosetta BL21(DE3) pLysS competent cells were purchased from Novagen. Primers (for cloning) and oligo (for crystallization) were ordered from Integrated DNA Technologies (IDT). Glutathione-Sepharose beads, Sp sepharose, MonoS and Superdex 200HR were purchased from GE healthcare. In vitro transcription/translation kit was purchased from Promega and 35S-Met was from Pierce. The mutations were generated using Quickchange mutagenesis kit (Stratagene). Lipofectamine 2000 was purchased from Invitrogen. Dual-luciferase reporter assay kit was from Promega. HeLa cells (ATCC) were grown in DMEM containing 10% fetal bovine serum. Other chemicals and reagents were purchased from Sigma.

Protein expression and purification

Human MEF2A (1–95) was expressed and purified as described previously (53). Briefly, pET30b-MEF2A (1–95) was transferred into BL21(DE3) pLysS Rosetta cells. The cells were induced at room temperature for 5 h. The induced cells were lysed by sonication in the lysis buffer (50 mM HEPES, pH 7.6, 300 mM NaCl, 1 mM EDTA, 1 mM DTT, 10% glycerol, protease inhibitors cocktails). The lysate was cleared by high-speed centrifugation and applied to Sp Sepherase column. Fractions containing MEF2 were pooled and further purified by superdex 200HR column using the buffer (10 mM HEPES, pH 7.6, 250 mM NaCl, 1 mM EDTA, 1 mM DTT).

The TAZ2 domain of human p300 (1726–1837) was cloned into pET28a. Based on the NMR structure of the CBP TAZ2 domain (59), the four cysteines (Cys1738, 1746, 1789 and 1790) not involved in zinc coordination were mutated to serine for better stability. The construct was transferred into Rosetta BL21(DE3) pLysS and induced at room temperature for 5 h with additional 150 µM ZnSO4 in the medium. The cells were lysed by sonication. The p300 TAZ2 domain was first purified by Ni-NTA Agarose (QIAGEN) and eluted in 15 mM HEPES7.6, 1 M NaCl, 15 mM β-mercaptoethanol and 100 mM Imidazole. The imidazole was removed by dialysis and His-tag was cleaved off by thrombin. The p300 domain was further purified by MonoS, followed by superdex 200 HR in the running buffer of 10 mM HEPES pH 7.6, 25 mM NaCl, 0.4 M NH4OAc, 2 mM DTT and 10 µM ZnSO4.

Crystallization

The complex was prepared by mixing purified DNA, MEF2 dimer and the p300 TAZ2 domain in the superdex 200 HR running buffer (see above). The initial stoichiometry was 1:1:1 but after several rounds of optimization with various crystallization conditions and molar ratios of the complex composition, the best crystals came from complexes prepared by mixing 0.15 mM double stranded DNA, 0.3 mM MEF2 dimer and 0.9 mM p300 TAZ2 domain. The complex crystals were grown by the hanging drop method against a reservoir solution of 50 mM BTP pH 6.3, PEG 1K 16% and 2 mM DTT at 17°C.

Data collection and structure determination

The crystals were transferred into the cryo-buffer of 32% PEG 1K and 50 mM BTP pH 6.3 and flash frozen in liquid nitrogen. The data were collected in Advanced Light Source (ALS 8.2.1, Berkeley). Data were processed by HKL2000 (60). The p300:MEF2:DNA complex structure was solved by the molecular replacement method using the MEF2A:DNA crystal structure as a partial search model in Phaser (61,62). The initial search by molecular replacement located three MEF2:DNA complexes in the asymmetric unit. Electron density calculated using phases generated from the partial model revealed clearly the position of the p300 TAZ2 domain. Model building was carried out in O and Coot (63,64). The refinement was carried out using CNS (65,66) and the model was analyzed using programs from CCP4 (67). The Ramachadran plot of the final model has 96.7% of residues in the favorite region and 3.3% of residues in the allowed region and no residue in the disallowed region. All figures were prepared in Pymol (The Pymol Molecular Graphics System, DeLano Scientific LLC).

GST-pulldown

The TAZ2 fragments were expressed as GST fusion proteins. The proteins were purified by GST affinity column and gel filtration in the S200 buffer (50 mM HEPES, pH 7.6, 300 mM NaCl, 1 mM DTT, 10% glycerol and protease inhibitor cocktail). The quality and concentration of the proteins were checked by SDS–PAGE. The full-length MEF2A was prepared using an in vitro translation kit and labeled with 35S-MET. The mixtures of labeled MEF2A and purified GST-p300 fusion proteins were incubated with GST beads for 1 h at 4°C, followed by three times wash using the same S200 buffer. The beads were collected for SDS–PAGE analysis followed by exposure to X-ray film.

Luciferase reporter assay

A plasmid was constructed to express the p300 TAZ2 domain (1721–1837) fused to the N terminal end of VP16 (Figure 5A). MEF2-Gal4 and Gal4 reporter plasmids were described by Gregoire and Yang (68). All these expression constructs were co-transfected to HeLa cells using lipofectamine 2000 or calcium phosphate. The experiments were performed in triplicate, and luciferases were measured using the dual-luciferase reporter assay kit according to manufacturer’s instructions. Briefly, 20 µl of cell lysate was transferred to a luminometer tube predispensed with 100 µl of Luciferase Assay Reagent II (LARII), and luminescence of firefly luciferase was measured by a Luminometer (Berthold). Then, 100 µl of Stop & Glo Reagent was added to the same tube and the light emission of renilla luciferase was recorded. Results are presented as activity ratios of firefly over renilla luciferases.

Figure 5.

Structure-guided biochemical analyses of p300:MEF2 interactions. (A) Diagram of the biochemical assay. (B) Luciferase reporter assay of p300 TAZ2 mutants. Columns 1–3 are controls. Column 4 is from wild-type p300 and MEF2 interaction. Columns 5–8 are IF-I mutants, 10–14 are IF-II mutants and 16–19 are IF-III mutants. Column 6 is a combined mutant of IF-I and IF-II.

RESULTS

p300 TAZ2 domain binds the MADS-box/MEF2 domain

The interaction between MEF2 and p300 was analyzed previously by several reports (25,26,69). These studies revealed that the MADS-box/MEF2 domain of MEF2 (amino acids 1–95) and a C-terminal region of p300 (amino acids 1572–1868) were necessary and sufficient to mediate p300 and MEF2 interaction. We confirmed this result by GST pull-down assay using individually purified proteins. By deletion analyses, we narrowed down the MEF2-binding region in p300 to a fragment that encompasses residues 1721–1837 (Figure 1, lane 4). This fragment corresponds to the TAZ2 domain of the CH3 region (amino acids 1725–1812) and a C-terminal extension (amino acids 1813–1837) with unknown function. We will refer this fragment (amino acids 1721–1837) as the TAZ2 domain hereafter. We previously hypothesized that the C-terminal extension might bind MEF2 analogously to class IIa HDACs and Cabin1 (58). However, further deletion studies revealed that while p300 (1805–1837) did bind MEF2 (Figure 1, lane 5), its affinity for MEF2 appeared to be much lower than that of p300 (1721–1837) (Figure 1, lanes 4 and 5). Moreover, the core TAZ2 domain of p300 (1721–1812) retained partial binding activity for MEF2 (Figure 1, lane 6). We also analyzed the interaction between p300 (1721–1837) and MEF2A (1–95) using electrophoresis mobility shift assay (EMSA), surface plasmon resonance (SPR), multi-angle light scattering (MALS), mass spectrometry and fluorescence anisotropy. While these experiments confirmed the binding of p300 (1721–1837) and MEF2A (1–95) in solution, quantitative analysis of binding constant and stoichiometry was complicated by the fact that the complex appears to be in equilibrium between several species (data not shown). Moreover, the p300 TAZ2 domain alone showed high background binding to DNA, making the full titration of the MEF2:DNA complex by high concentration of TAZ2 difficult. Nevertheless, our biochemical analyses suggest that p300 binds MEF2 significantly differently from Cabin1 and class IIa HDACs in that p300 uses a larger domain instead of a short helix.

Figure 1.

In vitro binding assay of the interaction between the p300 TAZ2 domain and MEF2A. Top panel: binding of the 35S-labeled MEF2A to various GST-TAZ2 fusion proteins. Lane 2 was MEF2A with beads only. Lane 3 was GST only protein. Bottom panel: GST-TAZ2 fusion proteins used in the top panel were analyzed by SDS–PAGE. The non-degraded fusion protein bands are indicated by stars.

Crystallographic analysis of the p300:MEF2:DNA complex

In order to analyze the details of p300:MEF2 interaction, we crystallized a complex between p300 (1726–1837) and MEF2A (1–95) on DNA. Our initial crystallization was hindered by the instability of p300 (1726–1837) due to the large number of cysteine residues in this region. Based on the NMR structure of the CBP TAZ2 domain (59), we mutated four cysteine residues (Cys1738, 1746, 1789 and 1790) that do not coordinate with zinc to serine. The cysteine-mutated p300 TAZ2 domain showed much improved stability and retained the same ability to bind MEF2 as the wild-type protein (data not shown). Using this engineered p300 (1726–1837) together with MEF2A (1–95) and a double stranded oligonucleotide containing the consensus MEF2 binding element (see Figure 2A below), we successfully crystallized the p300:MEF2:DNA complex. The crystals diffracted to 2.2Å. The statistics of data collection and structure refinement are summarized in Table 1.

Figure 2.

Structure of the p300 TAZ2 domain bound to MEF2 on DNA. (A) Overall structure in ribbon style. The MEF2 dimers are colored in yellow, magenta and blue, respectively; p300 is in green; the DNA backbone is shown in gold and its sequence is listed below. The secondary structural elements of one MEF2 dimer are labeled. (B) Protein binding interfaces: the proteins in the complex are shown in surface model and colored as in A. The three protein–protein interaction interfaces are labeled as IF-I, IF-II and IF-III. (C) Zoom-in view of the p300 TAZ2 domain. The four helices (α1–α4) and three zinc fingers (red spheres) are indicated. Part of the second zinc binding loop is disordered and labeled in a dashed line. (D) Superposition of the p300 TAZ2 domain in the MEF2-bound complex (green) with the unbound form (blue, 3IO2) (70).

Table 1.

Data collection and refinement statistics

Space group	P 21 21 21
Unit cell parameters (Å)	a = 74.85Å, b = 90.93Å, c = 144.8Å
Resolution (Å)	30.0–2.09 (2.16–2.09)a
Reflections (Total/unique)	1317602/59524
I/σ(I)b	27.3 (1.95)
Rsym (%)c	8.7 (70.9)
Completeness (%)	100 (100)
Refinement
Resolution (Å)	29.666–2.192 (2.217–2.192)
Reflections	49213 (4536)
Completeness (%)	95.698
R (%)d	22.13 (25.62)
Rfree (%)	27.16 (35.15)
Model quality
RMSD bond length (Å)e	0.008
RMSD bond angles (°)	1.400
Overall B-factor (Å2)	36.0
Number of total atomsf	7089
Protein atoms	5302
Nucleic acid atoms	1787
Ions	3
H2O	238

aThe data for the highest resolution shell are shown in brackets.

bI/σ(I)—ratio of mean intensity to a mean standard deviation of intensity.

cRsym = Σ|I−|/ΣI, where I is the observed intensity, is the statistically weighted average intensity of multiple observations of symmetry-related reflections.

dRfactor = Σ||Fo|−|Fc||/Σ|Fo|, where Fo and Fc are observed and calculated structure factor amplitudes, respectively. Rfree is calculated for a randomly chosen 10% of reflections.

eRMSD—root mean square deviation

fNumber of protein atoms and nucleic acid atoms—the ordered region.

Structure of the p300 TAZ2 domain bound to MEF2 on DNA. (A) Overall structure in ribbon style. The MEF2 dimers are colored in yellow, magenta and blue, respectively; p300 is in green; the DNA backbone is shown in gold and its sequence is listed below. The secondary structural elements of one MEF2 dimer are labeled. (B) Protein binding interfaces: the proteins in the complex are shown in surface model and colored as in A. The three protein–protein interaction interfaces are labeled as IF-I, IF-II and IF-III. (C) Zoom-in view of the p300 TAZ2 domain. The four helices (α1–α4) and three zinc fingers (red spheres) are indicated. Part of the second zinc binding loop is disordered and labeled in a dashed line. (D) Superposition of the p300 TAZ2 domain in the MEF2-bound complex (green) with the unbound form (blue, 3IO2) (70).

Data collection and refinement statistics

aThe data for the highest resolution shell are shown in brackets.

bI/σ(I)—ratio of mean intensity to a mean standard deviation of intensity.

cRsym = Σ|I−|/ΣI, where I is the observed intensity, is the statistically weighted average intensity of multiple observations of symmetry-related reflections.

dRfactor = Σ||Fo|−|Fc||/Σ|Fo|, where Fo and Fc are observed and calculated structure factor amplitudes, respectively. Rfree is calculated for a randomly chosen 10% of reflections.

eRMSD—root mean square deviation

fNumber of protein atoms and nucleic acid atoms—the ordered region.

Overall structure of the p300:MEF2:DNA complex

The overall shape of the complex resembles a trefoil with the p300 TAZ2 domain in the center and the MEF2 dimer:DNA complexes at the leaf positions (Figure 2A). The core of the p300:MEF2:DNA complex is assembled through extensive helical packing that involves a total of 10 helices from the TAZ2 domain and the MADS-box/MEF2 domain (Supplementary Figure S1). The three MEF2 dimer:DNA complexes bind the p300 TAZ2 domain via three distinct and non-overlapping interfaces: Interface I (IF-I), Interface II (IF-II) and Interface III (IF-III), respectively (Figure 2B). The buried solvent accessible surface areas for these interfaces are: IF-I = 1213.8A2, IF-II = 1223.1A2 and IF-III = 1117.0A2. The p300 TAZ2 domain in the complex is composed of four helices, α1 (amino acids 1727–1746), α2 (amino acids 1758–1766), α3 (amino acids 1782–1793) and α4 (amino acids 1806–1833) that pack against each other (Figure 2C). The structure is further stabilized by three zinc ions that are chelated between α1–α2 (Finger 1), α2–α3 (Finger 2) and α3–α4 (Finger 3), respectively. Part of the second finger (amino acids 1772–1778) is disordered in the crystal and labeled in a dash line. Each zinc ion is bound by one histidine and three cysteines in an HCCC-type coordination.

The MADS-box/MEF2 domain in the complex forms an intertwined dimer; each monomer consists of helix H1 (amino acids 15–37), β strand S1 (amino acids 41–49), β strand S2 (amino acids 53–59), helix H2 (amino acids 62–70), β strand S3 (amino acids 76–80) and helix H3 (amino acids 82–90) (Figure 2A). For each MEF2 dimer, helix H1 and the N-terminal tail form the major DNA-binding surface that binds the double stranded DNA at the periphery of the complex, while the pair of H2 helices form the protein-binding interface with the TAZ2 domain in the center of the complex. All three MEF2 dimers show fully folded MADS-box/MEF2 domain (Figure 2A). The three MEF2A dimer:DNA complexes bound to the TAZ2 domain are nearly identical to each other and highly similar to the ‘apo’ MEF2A:DNA complex (Supplementary Figure S2). There is no interaction between the individual MEF2 dimer:DNA complexes.

Rigid p300 TAZ2 domain

Our TAZ2 structure reveals a protein fold that is very similar to the homologous TAZ2 domain of CBP first solved by NMR (59). The crystal structure of the unbound p300 TAZ2 domain has recently been reported (70). Superposition of the p300 TAZ2 domain in the unbound state with that in the MEF2-bound complex shows that the two structures are very similar, although the N-terminal end of α1 and the C-terminal end of α4 have slightly different trajectories (Figure 2D). The α2-α3 loop also displays significant differences due to different crystal packing interactions. The spatial arrangement of the four helices and the structure of the three zinc binding sites are highly conserved in the unbound state and in complexes with different protein partners, although the loops between the helices show a certain degree of variability due to different crystal packing and protein–protein interactions (9,14,16,59). A significant difference from the solution structure of the TAZ2 domain is that we observed an extended helix α4 in the crystal structure of the p300 TAZ2 domain (Figure 2C). This long helix α4 was also observed in a ligand-free TAZ2 structure (70). Overall, the structural comparisons suggest that the formation of the p300:MEF2:DNA complex involves largely rigid body docking interactions of the preformed MEF2A:DNA complex and the p300 TAZ2 domain.

Protein–protein interactions at Interface I

Interface I is formed between helix α4 of the TAZ2 domain and one of the MEF2 dimers. Here helix α4 of the TAZ2 domain packs diagonally between helices H2 of the MEF2 dimer. This arrangement is very similar to the binding of the amphipathic helix of Cabin1 and HDAC9 to MEF2, however, the exact positions of the helices are slightly different in different complexes (Figure 3A). Previously we showed that the binding of Cabin1 and HDAC9 to MEF2 induced a shift of helix H2 and conformational changes of the L2 loop between helix H2 and strand S3 (53,57,58). The L2 conformations of MEF2 bound to the p300 TAZ2 domain are indeed notably different from that seen in the Cabin1:MEF2 and HDAC9:MEF2 complexes (Figure 3A). The detailed interactions of IF-I share many features observed at the Cabin1:MEF2 and HDAC9:MEF2 interfaces (Figure 3B and C). Leu1818 and Leu1822 of p300 bind the hydrophobic groove of MEF2 similarly to homologous leucine residues in Cabin1 and HDAC9 (residues of p300 are italicized throughout the text) (57,58). Surrounding the hydrophobic interactions are numerous salt bridges and hydrogen bond networks. At the N-terminal end of TAZ2 α4, Arg1814 engages in electrostatic interactions with Asp61 and Asp63 of MEF2, whereas Gln1811and Gln1815 form hydrogen bonds with Asp63 and the main chain carbonyl of Thr70 of MEF2, respectively. Both Gln1811and Gln1815 are also in van der Waals contact with Tyr69 of MEF2. At the C-terminal end of TAZ2 α4, Arg1829 engages in electrostatic interactions with Asp61 and Asp63 of the other MEF2 monomer. Along the helix of TAZ2 α4, the long aliphatic side chains of Arg1814, Gln1815, Gln1819 and Arg1821 make extensive van der Waals contacts with residues of helix H2 of the MEF2 dimer. Here, Gln1819 also forms a hydrogen bond with Thr70 while Arg1821 forms a salt bridge with Glu71. Overall, IF-I shows a remarkably similar binding mode to Cabin1 and HDAC9 as well as new features of protein–protein interactions.

Figure 3.

Protein–protein interactions at Interface I. (A) Structural comparison of Interface I (p300 in green and MEF2 in gold) with the Cabin1:MEF2:DNA complex (red) and the HDAC9:MEF2:DNA complex (cyan). The structures are superimposed by the Cα backbone of the β strands of the MEF2 core. (B) View from the side of helix α4 of the p300 TAZ2 domain. Interacting residues are shown in stick model. Throughout the illustrations, residues from p300 are italicized and labeled with black fonts while residues from MEF2 are labeled according to the protein color. (C) View from the C-terminal end of helix α4.

Protein–protein interactions at Interface I. (A) Structural comparison of Interface I (p300 in green and MEF2 in gold) with the Cabin1:MEF2:DNA complex (red) and the HDAC9:MEF2:DNA complex (cyan). The structures are superimposed by the Cα backbone of the β strands of the MEF2 core. (B) View from the side of helix α4 of the p300 TAZ2 domain. Interacting residues are shown in stick model. Throughout the illustrations, residues from p300 are italicized and labeled with black fonts while residues from MEF2 are labeled according to the protein color. (C) View from the C-terminal end of helix α4.

Protein–protein interactions at Interface II

Interface II is formed between helix α1 and the first zinc loop of the TAZ2 domain and a second MEF2 dimer. Here, helix α1 of the TAZ2 domain lies above the two H2 helices of the MEF2 dimer at a cross angle of ∼60° (Supplementary Figure S3). The interface is composed of two separated foci of binding interactions; the first being between the N-terminal part of the TAZ2 α1 helix and helix H2 of one MEF2 monomer (Figure 4A). Met1725 and Leu1733 of p300 and Leu67 of MEF2 form a central hydrophobic patch that is surrounded by a number of electrostatic and hydrogen bond interactions, including salt bridges between Arg1737 and Asp63 and between Arg1732 and Glu71. At the second binding site of IF-II, the C-terminal part of TAZ2 α1 helix and part of the first zinc loop interact with helix H2 of the other MEF2 monomer. Leu1755 of p300 and Leu67 of MEF2 form the hydrophobic core. Leu67 also makes van der Waals contacts with Gln1740 and His1744, while Glu71 makes hydrogen bonds with the main chain and side chain of Ser1757. The side chain of Arg1737 and Pro1756 also make van der Waals contacts with Thr70 and Glu71, respectively.

Figure 4.

Protein–protein interactions at interfaces II and III. (A) The interaction network at Interface II. (B) Interactions at Interface III viewed from the side of helix α1. (C) Interactions at Interface III viewed from the side of helix α3.

Protein–protein interactions at interfaces II and III. (A) The interaction network at Interface II. (B) Interactions at Interface III viewed from the side of helix α1. (C) Interactions at Interface III viewed from the side of helix α3.

Protein–protein interactions at Interface III

Interface III is formed between helices α1 and α3 of the TAZ2 domain and a third MEF2 dimer. Here, α1 and α3 of the TAZ2 domain pack against helix H2 of each of the two MEF2 monomers, both in an approximately parallel orientation (Supplementary Figure S4). Similar to IF-II, IF-III is also composed of two separated foci of binding interactions. One is between helix α1 of the TAZ2 domain and helix H2 of one of the MEF2 monomers (Figure 4B). Here, Pro1727, Gly1728, Ser1730 and Arg1731 of p300 and Asp63, Lys64, Leu66, Leu67 and Thr70 of MEF2 engage in a pseudo knobs-into-holes type helix packing interaction. Arg1731 also forms a hydrogen bond with Thr70. At the second binding site (Figure 4C), Lys1783, Gln1784, Ala1787, Cys1790, Tyr1791 and Lys1794 on the exposed face of p300 α3 form an extensive interface with Asp61, Asp63, Lys64, Leu67 and Glu71 on helix H2 of one of the MEF2 monomers. Most notably, Cys1790 and Tyr1791 of p300 and Leu67 of MEF2 form a hydrophobic patch, Lys1783 and Lys1794 engage in electrostatic interactions with Asp61 and Glu71, respectively. Overall, the binding modes of IF-II and IF-III are substantially different from that seen in the Cabin1:MEF2 complex and HDAC9:MEF2 complex, but both interfaces bury large solvent accessible surface areas and display good shape and chemical complementarity.

Biochemical analysis of the p300:MEF2 interaction

To test if the protein–protein interfaces seen in the crystal structure contribute to p300:MEF2 interaction, we analyzed a series of structure-guided mutations using a mammalian two-hybrid assay specifically designed for detecting the interaction between the p300 TAZ2 domain and the MADS-box/MEF2 domain in mammalian cells. Initially, we transfected a MEF2-responsive luciferase reporter with full-length p300 and MEF2 to detect MEF2-p300 interaction. These studies did not yield interpretable results due to high background, presumably caused by the endogenous MEF2 and p300. To overcome this problem, we adopted a mammalian two-hybrid assay by fusing MEF2 with the GAL4 DNA-binding domain (GAL4-MEF2) and the p300 TAZ2 domain with VP-16 (TAZ2-VP16) (Figure 5A). This setup allows us to detect the interaction between TAZ2 and MEF2 with minimal interference from endogenous factors. A potential caveat of this approach is that MEF2 is not directly bound to DNA whereas under physiological conditions p300 is recruited by DNA bound MEF2. However, we have previously shown that free MEF2 and DNA-bound MEF2 bind co-factors with similar affinities, suggesting that the structure of MEF2, at least in the MADS-box/MEF2 domain, is conserved between the free and DNA-bound forms (57). HeLa cells co-transfected with constructs of GAL4-MEF2 and TAZ2-VP16 together with a GAL4-driven reporter plasmid produced a strong signal comparable to that generated by the positive control of GAL4-VP16 (data not shown). At Interface I, the double mutation Leu1818Ala/Leu1822Ala caused a significant reduction of the luciferase signal, whereas mutation of Gln1815Ala and Gln1815Tyr had little effect (Figure 5B). These results suggest that the hydrophobic residues at the center of IF-I are important for binding, whereas residues at the periphery are more tolerable to mutations. At Interface II, double mutation Leu1733Arg/Arg1737Ala and triple mutation Leu1733Ala/Arg1737Ala/Gln1740Ala reduced the luciferase signal substantially, whereas the single mutation Gln1740Ala had only partial effect. Interestingly, the mutation Gln1740Tyr enhanced the luciferase signal, suggesting that the Tyrosine residue introduced at position 1740 may interact better with MEF2, presumably by inserting into the hydrophobic groove between helices H2. These mutational data are consistent with the structural features of IF-II and suggest that IF-II contributes to p300:MEF2 interaction in solution. At Interface III, mutation Arg1731Ala and Tyr1791Ala partially reduced the luciferase signal, which is consistent with the structural roles of these residues at IF-III. However, the double mutation Arg1731Ala/Tyr1791Ala enhanced the luciferase signal. The mechanism of this gain-of-function mutation is not clear. It is possible that the Arg1731Ala/Tyr1791Ala double mutant adopts a new way to bind MEF2 that is different than seen in the crystal structure. Taken together, these data suggest that protein–protein interactions observed at interfaces I, II and III contribute to the binding of the p300 TAZ2 domain to MEF2 in solution. For some of the mutations that showed strong effects in the luciferase assay, such as the double mutant Leu1818Ala/Leu1822Ala and Leu1733Arg/Arg1737Ala, our GST-pulldown also confirmed that the mutant proteins lost the binding to MEF2 in vitro (data not shown).

Structure-guided biochemical analyses of p300:MEF2 interactions. (A) Diagram of the biochemical assay. (B) Luciferase reporter assay of p300 TAZ2 mutants. Columns 1–3 are controls. Column 4 is from wild-type p300 and MEF2 interaction. Columns 5–8 are IF-I mutants, 10–14 are IF-II mutants and 16–19 are IF-III mutants. Column 6 is a combined mutant of IF-I and IF-II.

DISCUSSION

The crystal structure of the p300 TAZ2 domain reported by Miller et al. (59,70) reveals a new feature that is not seen in the previous NMR structure of the CBP TAZ2 domain, namely an extended helix α4. This unusually long helix makes no contacts with the rest of the TAZ domain but engages in extensive interactions with symmetry related molecules in crystal (70). It was not observed in the NMR structure of the same domain bound to p53 because a shorter p300 fragment (1723–1812) was used (14). However, the extended helix α4 was involved in extensive crystal packing interactions in the structure reported by Miller et al. (59,70), a question remains if the observed structure of helix α4 is a result of crystal packing. The p300 TAZ2 domain in our crystal structure reveals a similar extended helix α4, which, instead of making crystal contacts, engaged in extensive interactions with a MEF2 dimer at IF-I. This observation suggests that the structure of helix α4 is an intrinsic property of the p300 TAZ2 domain. Consistent with this view, CD analysis indicated that the C-terminal region of p300 TAZ2 domain forms alpha helix in solution (15). Our studies further suggest that a potential function of this long helix is to mediate protein–protein interactions with other proteins, including a variety of transcription factors, such as MEF2.

Based on the crystal structure of the Cabin1:MEF2:DNA complex and sequence homology, we have previously hypothesized that the region of p300 corresponding to residues Leu1807 to Gln1819 could form an amphipathic helix to bind MEF2 in a way similar to Cabin1 (58). Our structure reveals that this region of p300 indeed forms an alpha helix as part of α4, but it is packed against helix α3 and sterically blocked from MEF2 binding. Instead, MEF2 at IF-I binds to a hydrophobic patch on helix α4 that is centered around Leu1818 and Leu1822. This hydrophobic patch has been noted in the unbound structure of the p300 TAZ domain, which mediates the majority of crystal packing interactions (70). Despite the shift of the predicted MEF2-binding site, the binding mode between MEF2 and the TAZ2 domain at IF-I is remarkably similar to that seen in the Cabin1:MEF2:DNA complex and the HDAC9:MEF2:DNA complex, whereas those at IF-II and III are significantly different from the canonical mode defined previously (57).

A frequently observed mode of co-factor recruitment in transcription complexes is binding-induced folding. Usually one protein with a stably folded domain presents a template for the binding and folding of its partner with intrinsically disordered regions. The TAZ2 domain serves as the binding site for the transcription activation domains of a number of transcription factors, including STAT1 and p53, which are disordered in solution and adopt a short helix to bind the hydrophobic surface between helices α1, α2 and α3 (14,16). The conserved region-1 (CR1) of the adenoviral protein E1A, which is unstructured in solution, also binds this region in a helical conformation (9). The binding sites for different factors on TAZ2 partially overlap, such that E1A can compete with p53 for p300 binding (9). The hydrophobic surface between helices α1, α2 and α3 is also a MEF2-binding site in the p300:MEF2 complex (IF-III). However, the binding of MEF2 to TAZ2 involves mostly rigid body docking without significant structural changes of either MEF2 or the TAZ2 domain.

CBP/p300 can acetylate specific lysine residues in MEF2 to enhance its DNA binding and transcriptional activity (56,71). The fact that the TAZ2 domain (amino acids 1721–1837) is adjacent to the HAT domain in p300 (amino acids 1066–1707) suggests that p300 could acetylate MEF2 within the same complex. It should be noted that the transcription activation domain of MEF2, which is located beyond the MADS-box/MEF2 domain at the C-terminal end, might interact with p300 at additional regions. Thus in the context of full-length proteins, p300 and MEF2 may bind each other through multiple contacts and a variety of binding modes.

CBP/p300-mediated enhanceosome assembly has long been hypothesized (19–21). Our structure provides the first example where a p300 domain is directly recruited to DNA by a DNA-bound, sequence-specific transcription factor. A surprising finding is that MEF2 binds the p300 TAZ2 domain at three distinct and non-overlapping surfaces, leading to an assembly of a higher-order enhanceosome structure wherein three DNA-bound MEF2 dimers are ‘attracted’ to different faces of the TAZ2 domain. This has two implications for the transcriptional regulation by p300 and MEF2. First, the TAZ2 domain from one p300 molecule could interact with three MEF2 dimers bound to DNA, thereby promoting the assembly of higher-order structures in the promoter and long-range interactions of DNA-bound MEF2 complexes (Figure 6). Our structure does not impose any spatial constraint between the different MEF2 sites as long as they are well separated so the intervening sequences could be looped out. In fact, the MEF2 binding sites could even come from different chromosomes. In this regard, recent ChIP-on-chip studies reveal that MEF2 bind a much larger number of sites in the genome than previously thought, and that most MEF2-targeted promoters contain multiple MEF2 binding sites that are separated far from each other (often more than several kilobases) (31). Second, as discussed above, the TAZ domain can interact with a variety of other proteins, including different transcription factors. The fact that MEF2 can bind three distinct sites on TAZ2 opens up the possibility that MEF2 may bind one of the three sites in a given promoter context, leaving room for the binding of other transcription factors to the TAZ2 domain. This versatility of p300:MEF2 interaction may facilitate combinatorial control of transcription regulation between MEF2 and other transcription factors. Future studies are needed to further address these questions, for example, by analyzing the effect of mutations at different interfaces on the expression of specific subsets of MEF2-targeted genes.

Figure 6.

A hypothetical enhanceosome model. Multiple MEF2 dimers bind to numerous sites in the enhancer/promoter region of a targeted gene. Brown balls represent the nucleosomes. p300 is recruited to one MEF2 dimer:DNA complex to initiate the assembly of the activation complex. Subsequent binding of p300 to other two MEF2 dimer:DNA complexes induces a looping of DNA and the assembly of an enhanceosome.

A hypothetical enhanceosome model. Multiple MEF2 dimers bind to numerous sites in the enhancer/promoter region of a targeted gene. Brown balls represent the nucleosomes. p300 is recruited to one MEF2 dimer:DNA complex to initiate the assembly of the activation complex. Subsequent binding of p300 to other two MEF2 dimer:DNA complexes induces a looping of DNA and the assembly of an enhanceosome.

ACCESSION NUMBER

Coordinates and structure factors have been deposited in the Protein Data Bank with accession number 3P57.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

National Institutes of Health, USA (HL076334, GM064642 and GM077320 to L.C. and GM83172 to X.J.); National Science Foundation of China (30840027 and 90919036 to A.H.); Fujian Science Advancing Program (2009J1010 and Project 111). Funding for open access charge: Fujian Science advancing Program (2009J1010).

Conflict of interest statement. None declared.