doi:10.1016/j.jmb.2004.01.038 J. Mol. Biol. (2004) 337, 521–534

Solution Structure of the KIX Domain of CBP Bound to the of c-Myb

Tsaffrir Zor, Roberto N. De Guzman, H. Jane Dyson and Peter E. Wright*

Department of Molecular The hematopoietic factor c-Myb activates transcription of Biology and the Skaggs target genes through direct interactions with the KIX domain of the Institute for Chemical Biology co-activator CBP. The solution structure of the KIX domain in complex The Scripps Research Institute with the activation domain of c-Myb reveals a helical structure very simi- 10550 N. Torrey Pines Road, La lar to that adopted by KIX in complex with the phosphorylated kinase Jolla, CA 92037, USA inducible domain (pKID) of CREB. While pKID contains two helices, aA and aB, which interact with KIX, the structure of bound c-Myb reveals a single bent amphipathic helix that binds in the same hydrophobic groove as the aB helix of pKID. The affinity of c-Myb for KIX is lower than that of pKID, and relies more heavily on optimal interactions of the single helix of c-Myb with residues in the hydrophobic groove. In particular, a deep hydrophobic pocket in KIX accounts for more than half the inter- actions with c-Myb observed by NMR. A bend in the a-helix of c-Myb enables a critical leucine side-chain to penetrate into this pocket more deeply than the equivalent leucine residue of pKID. The components that mediate the higher affinity of pKID for KIX, i.e. the phosphate group and the aA helix, are absent from c-Myb. Results from isothermal titration calorimetry, together with the structural data, point to a key difference between the two complexes in optimal pH for binding, as a result of differential pH-dependent interactions with histidine residues of KIX. These results explain the structural and thermodynamic basis for the observed constitutive versus inducible activation properties of c-Myb and CREB. q 2004 Elsevier Ltd. All rights reserved. Keywords: CREB-binding protein; transcriptional activation; constitutive *Corresponding author activation; LXXLL motif

Introduction general RNA polymerase II complex that binds the .1 The key role of CBP and its homolog Transcriptional activation in eukaryotes involves p300 in transcription regulation is underscored by interactions between DNA-bound activators, the findings that mutations in the CBP gene have co-activators and components of the basal been described in various types of cancer and in transcription complex. The general co-activator the Rubinstein–Taybi syndrome, a haplo-insuffi- CREB-binding protein (CBP) functions as a bridge ciency disorder characterized by skeletal abnorm- between a number of transcription factors that alities, growth retardation and high incidence of bind specific DNA elements and the tumors.2 The first interaction to be described was between CBP and the kinase-inducible activation domain Present address: T. Zor, Genomics Institute of the (KID) of the cAMP-regulated Novartis Foundation, La Jolla, CA 92121, USA. CREB. Phosphorylation of KID at Ser133 was Abbreviations used: CBP, CREB-binding protein; KID, shown to be essential for binding the KIX domain kinase-inducible activation domain; pKID, of CBP and for subsequent transcriptional phosphorylated KID; HSQC, heteronuclear single 3,4 quantum coherence; ITC, isothermal titration activation. The NMR structure of the phosphoryl- calorimetry; LPE, linked protonation effect. ated KID (pKID)–KIX complex showed that KIX E-mail address of the corresponding author: forms a helical bundle structure that is bound by [email protected] the two mutually perpendicular helices of pKID.5

0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved. 522 Solution Structure of KIX:c-Myb Complex

The phosphoserine residue is located at the was attributed mainly to favorable intermolecular N terminus of the aB helix of the KIX-bound interactions involving the phosphate moiety. Bio- pKID. This helix makes multiple hydrophobic physical and biochemical results indicated that interactions with the shallow hydrophobic groove unphosphorylated KID binds specifically to KIX formed by the a1 and a3 helices of KIX. NMR with a sevenfold lower affinity than that of chemical-shift mapping6 as well as mutagenesis c-Myb:KIX.6 This difference in binding affinity studies7 showed that the same hydrophobic groove enables c-Myb to be a constitutive transcription is the docking site for c-Myb, a constitutive tran- factor, while CREB has a very low basal transcrip- scriptional activator regulating cell growth and tional activity and has to be phosphorylated on differentiation of hematopoietic cells,8 whose aber- the KID domain in order to bind the KIX domain rant amplification or truncation has been observed of CBP with significant affinity.3,4 However, the in several types of leukemia.9 structural basis for the participation of KIX in both Although pKID and c-Myb share a common phosphorylation-dependent and phosphorylation- binding site on KIX, there is no obvious sequence independent interactions is unclear. Specifically, similarity in their binding regions. In the unbound the overall similarity in the binding mode of KIX state, the activation domains of both c-Myb and to c-Myb and to KID in the basal unphosphoryl- KID (either phosphorylated or unphosphorylated) ated state appears to be contradictory to the are only partly structured, and binding to KIX is different affinities and transcriptional outcome. coupled with folding to form an amphipathic To evaluate the structural basis for the different helix that binds the hydrophobic groove of KIX.6,10 binding affinities of c-Myb and CREB for KIX that The affinity of KIX for the phosphorylated KID is result in different regulation of transcription, we 20-fold higher than for c-Myb.6 This difference have determined the solution structure of the KIX

Figure 1. Three-dimensional structure of the c-Myb:KIX complex. A, Stereoview of the backbone trace of a best-fit superposition of the family of 20 NMR structures. The backbone of KIX is shown in blue and that of c-Myb in red. Only the ordered parts of KIX (residues 589–665) and c-Myb (residues 291–310) are shown. B, Ribbon diagram of the lowest-energy structure, colored as in A. Solution Structure of KIX:c-Myb Complex 523 domain of CBP in complex with the activation Table 1. NMR structure statistics domain of c-Myb. The structure reveals optimal KIX: interactions of the single helix of c-Myb with NMR constraints KIX c-Myb c-Myb residues in the hydrophobic groove of KIX. In par- ticular, Leu302 of c-Myb is inserted deeply into a Distance constraints 1289 125 100 hydrophobic pocket in KIX, accounting for more Intra-residue 622 0 Sequential 252 75 than half of the interactions between the two Medium-range 226 50 proteins. A bend in the a-helix of c-Myb enables Long-range 189 0 Leu302 of c-Myb to penetrate into this pocket Ambiguous constraints 52 0 more deeply than the equivalent leucine of pKID. Torsion angle constraints f 58, 16 c The differences in the sequence and helical struc- 57 16 ture of the two transcription factors result in Structure statistics (20 structures) modulation of complementarity and enable the Violations statistics /structure KIX domain to distinguish between the con- NOE violations . 0.1 A˚ 7 ^ 2 stitutive c-Myb that activates transcription and the Maximum NOE violation (A˚ ) 0.30 ^ unphosphorylated form of CREB that should not Torsion angle constraint 0.3 0.4 violations . 08 (deg.) activate transcription. Maximum torsion angle 2.2 violation (deg.) Energies Results Mean constraint violation 9.7 ^ 1.7 (kcal mol21) ^ Structure determination Mean AMBER 21700 21 (kcal mol21) Multi-dimensional NMR experiments were used Mean deviations from ideal to assign the chemical shifts for the KIX domain of covalent geometry Bond lengths (A˚ ) 0.0058 ^ 0.0001 CBP (residues 586–672) complexed with a 25 Bond angles (deg.) 1.92 ^ 0.02 residue peptide derived from the c-Myb activation 1 15 domain (residues 291–315). H– N correlated PROCHECK statistics heteronuclear single quantum coherence (HSQC) Residues in most favored 90.3 spectra of both 15N-labeled KIX bound to unlabeled regions (%) c-Myb and 15N-labeled c-Myb bound to unlabeled Residues in allowed 9.1 regions (%) KIX showed that the domains studied adopt a Residues in generously 0.5 6 unique folded conformation. In addition, the mini- allowed regions (%) mal region of c-Myb retains binding affinity for Residues in disallowed 0.1 KIX similar to that of larger peptides.6 Solution regions (%) structures were calculated using torsion angle and inter-proton distance restraints. The 20 lowest- RMSD deviations from KIX c-Myb KIX: average structure c-Myb energy structures (Figure 1A) form a tight family Backbone atoms 0.60 0.63 0.61 with low RMSD values, good backbone confor- (N,Ca,C0)a (A˚ ) mations and no significant constraint violation All heavy atomsa (A˚ ) 1.10 1.01 1.08 (Table 1). a-Helices, backbone 0.40 0.63 0.46 atoms (N,Ca,C0)b (A˚ ) a-Helices, all heavy 1.01 1.01 1.01 b ˚ Structure of the KIX domain in the KIX:c-Myb atoms (A) complex a In ordered regions: KIX residues 589–665, c-Myb residues 293–309. b KIX residues 597–611, 623–640, 646–662, c-Myb residues The KIX domain of CBP is composed of three 293–309. mutually interacting helices, a1 (residues 597–611), a2 (residues 623–640) and a3 (residues 646–664), and two short 310 helices, G1 (residues 591–594) and G2 (617–621) (Figure 1B). The C ter- Structure of c-Myb minus of helix a3 of KIX is stabilized and extended in the c-Myb complex relative to the pKID complex,11 probably because a slightly longer KIX In the unbound state, residues 295–309 of c-Myb construct (87 residues, versus 81 residues) was populate a partially helical conformation, esti- used. A shallow hydrophobic groove is formed by mated to have 25–30% helical content.6 Binding of helices a1 and a3, which pack at an angle of KIX to the 25-mer c-Myb peptide leads to a signifi- ^ 17( 2)8, while helices a1 and a2 pack at an angle cant stabilization of the c-Myb helix. In the bound of 56(^2)8. The secondary structural elements state, the helix extends from Lys293 to Leu309 or define a compact structural domain with an exten- Lys310. Inter-molecular interactions with the sive hydrophobic core. The core interactions hydrophobic groove of KIX serve to stabilize the reported for KIX in complex with pKID were all helical conformation of c-Myb. Outside the helix observed in the KIX:c-Myb complex.5 boundaries, residues 291–292 and 311–315 are 524 Solution Structure of KIX:c-Myb Complex unstructured, as indicated by the lack of NOEs (C- affinity is approximately threefold lower at pH 5.5 and N termini) and sharp resonances (C terminus). compared to pH 7.0. These differences can be While the backbone f and c angles indicate a con- related directly to changes in the enthalpy tinuous helix that spans 17–18 residues, analysis (Table 2), which is more favorable for c-Myb at of intra-helical distances and angles required for pH 5.5 and for pKID at pH 7. i,i þ 4 hydrogen bonds reveals a bend with an Rather than being related directly to differences angle of 50(^9)8 that is centered at Ser304. Strong in the mechanics of binding, the observed enthalpy i,i þ 4 hydrogen bonds, characteristic of a helices, differences may arise from a linked protonation are observed in the segment 291–304 as well as in effect (LPE), which may occur due to a change in the segment 304–311. Only a single i,i þ 3 hydro- the pKa of a protein side-chain upon complex gen bond crosses the bend from the backbone formation.12 An experimental test for LPE is to carbonyl group of Leu302 to the backbone amide measure the observed enthalpy change (DHobs)in group of Thr305. the presence of buffers with very different enthal- pies of ionization (DHion), and to calculate the The KIX:c-Myb interface change in the number of bound protons (DNHþ) upon complex formation, according to: The single a helix of c-Myb binds in the shallow DHobs ¼ DH0 þ DNHþDHion ð1Þ hydrophobic groove formed by the a1 and a3 helices of KIX (Figure 2A). This groove is formed ITC measurements were made for the c-Myb:KIX by the side-chains of Leu599, His602, Leu603, the and pKID:KIX complex formation at pH 7.0 using aliphatic region of Lys606, Leu607 and Ala610, 21 Tris buffer (DHion ¼ 11:35 kcal mol ) or phosphate all located on the a1 helix, together with Tyr650, 21 buffer (DHion ¼ 0:8 kcal mol ) (1 cal ¼ 4.184 J). Leu653, Ala654, Ile657, Tyr658, Gln661 and the The results are shown in Figure 3B. The calculated a aliphatic portion of Lys662, on the 3 helix thermodynamic parameters of binding are shown (Figure 2B). The borders of the hydrophobic groove in Table 2. As expected, the affinity was not depen- are formed by the charged side-chains of Asp598 dent on buffer identity, but the enthalpy of binding and Arg646 on one side and Lys662 and Glu665 was affected. A large LPE was observed for c-Myb: on the other side. The groove serves as the docking KIX interaction: binding was significantly more surface for the non-polar face of the amphipathic enthalpy-driven in the presence of the phosphate helix of c-Myb, formed by the side-chains of buffer (with its negligible ionization enthalpy) Ile295, Leu298, Leu301, Leu302, Met303, Thr305 relative to Tris buffer (which has a strong unfavor- and Leu309. The charged side-chain of Arg294 able ionization enthalpy). In contrast, a small LPE is Glu298 are located in the N-terminal side of in the opposite direction was observed for pKID: the binding region and those of Glu306 and KIX complex formation. Calculation according to Glu308 are in the C-terminal part. The orientation equation (1) yields DNHþ of þ 0.75 for KIX binding of c-Myb in the hydrophobic groove is determined to c-Myb and 20.07 for KIX binding to pKID. The by the electrostatic interactions of Arg294 with buffer-independent enthalpy ðDH0Þ at pH 7.0, Glu665 on one side and Glu306 with Arg646 on obtained from these values according to equation the other side. In the center of the shallow hydro- (1), is 212.2 kcal mol21 for KIX binding to c-Myb phobic groove, there is a deep hydrophobic pocket and 214.0 kcal mol21 for KIX binding to pKID. lined by the side-chains of Leu603, Lys606, Leu607, Tyr650, Leu653, Ala654 and Ile657 of KIX. The side-chain of Leu302 of c-Myb is inserted into Discussion the center of the pocket, where it is completely buried. Comparison of the c-Myb:KIX and pKID:KIX complexes pH-dependence of the interaction between KIX and c-Myb Structures of the KIX (residues 586–666) complex with the KID domain of CREB (residues The NMR experiments for the structure determi- 101–160) phosphorylated at Ser133 have been pre- nation were recorded at pH 5.5 in order to increase viously determined in this laboratory.5 Our current the quality of the spectra by lowering amide data show that the structure of KIX is virtually proton exchange with the solvent. The previous identical in the c-Myb and pKID complexes. This thermodynamic analysis was carried out at a more is illustrated in Figure 4A, where the two structure physiological pH 7.0.6 To correlate between the families are overlaid. The backbone RMSD structural and thermodynamic data, we performed between the mean of the two sets of KIX structures an isothermal titration calorimetry (ITC) study at is 0.502 A˚ (0.447 A˚ if only the three helices are the two pH values. The results are shown in superimposed). A representative pair of structures Figure 3A, and the affinities and enthalpy from the overlay is shown in Figure 4B. It is clear changes are shown in Table 2. Both of the inter- that the hydrophobic groove formed by the a1 and actions show a significant pH-dependence. For a3 helices of KIX can accommodate either the helix c-Myb, the affinity for KIX is about sixfold higher of the bound c-Myb or the aB helix of pKID, but at pH 5.5 compared to pH 7.0, while for pKID, the the orientation of these helices in the groove is Solution Structure of KIX:c-Myb Complex 525

Figure 2. Complementarity between KIX and c-Myb. A, KIX is shown as a surface. The backbone of c-Myb is shown as a red ribbon and the side-chains of c-Myb that interact with KIX are shown in yellow. B, Stereoview of interacting side-chains along the hydrophobic groove: c-Myb (red) and KIX (blue). The side-chains at the ends of the hydrophobic groove are shown in black. The side-chain of A654 is shown but not labeled.

significantly different. The minor binding site on dominantly hydrophobic. The size of the surface is ˚ 2 KIX for the aA helix of pKID is solvent-accessible comparable to that of KIX:pKID (1460 A ) but all in the c-Myb:KIX complex, since c-Myb does not of it resides in the hydrophobic groove, while only have a second helix that can wrap around KIX. about 80% of the pKID:KIX interface is in the Upon complex formation, KIX and c-Myb bury hydrophobic groove; the aA helix of pKID 1480 A˚ 2 of solvent-accessible surface, which is pre- accounts for the remaining 20%. 526 Solution Structure of KIX:c-Myb Complex

The location of the bend in the structure of c-Myb bound to KIX is precisely at the position of Leu302, the side-chain of which is buried in a deep hydrophobic pocket on KIX. This is shown in close-up in Figure 4C. The equivalent Leu141 of pKID is inserted into the same pocket, but at a less optimal orientation. The bend of the c-Myb helix enables the side-chain of Leu302 to project into the pocket about 1 A˚ deeper than the side- chain of Leu141 of pKID. Therefore, Leu302 of c-Myb (but not Leu141 of pKID) is able to make hydrophobic contacts with the side-chain of Leu607 at the bottom of the deep pocket. This structural difference between the two complexes is consistent with differences in KIX chemical shifts upon complex formation in each case.6 Optimal packing of Leu302 in the deep pocket is helped by the interactions of the adjacent Leu301 and Met303 (Figure 2). Leu301 interacts with the hydrophobic portion of Lys606 and with Ala610, while the equivalent Asp140 of pKID can interact only with Lys606. Met303 makes extensive hydrophobic con- tacts with His651, Ala654 and Tyr650. The equival- ent Ser142 of pKID is too far from His651 and Ala654, and its polar hydroxyl group is close to Figure 3. Sensitivity of KIX:c-Myb and KIX:pKID the hydrophobic aromatic ring of Tyr650, presum- complex formation to pH and buffer variation. A, Sensi- ably a relatively unfavorable interaction. tivity to pH. ITC measurements of KIX binding to In the c-Myb:KIX complex, the side-chains of pKID29 (circles) and to c-Myb (squares) at pH 7.0 Leu298 and Glu299, located one helical turn from (filled) and at pH 5.5 (open). Concentrations of KIX Leu302, pack against a shallower region of the were 0.145 mM and 0.125 mM for the pKID29 titrations hydrophobic groove, making extensive van der at pH 7.0 and at pH 5.5, respectively, and 0.225 mM and Waals contacts with Ile657 and Tyr658, as well as 0.170 mM for the corresponding c-Myb titrations. Ala654 (Glu299 only) or Ala610 (Leu298 only). In B, Sensitivity to buffer. ITC measurements at pH 7.0 of KIX binding to pKID34 (circles) and to c-Myb (squares) addition, Leu298 interacts with Gln661 and in half in Tris buffer (filled) and in phosphate buffer (open). of the structures also with Lys662. Leu298 and Concentrations of KIX were 0.050 mM and 0.040 mM for Glu299 are replaced in the pKID:KIX complex by the pKID34 titration in Tris buffer and in phosphate Ile137 and Leu138 of pKID, respectively. The inter- buffer, respectively. Concentrations of KIX were actions of these side-chains with KIX are in general 0.225 mM and 0.215 mM for the c-Myb titration in Tris similar, except for some differences that result from buffer and in phosphate buffer, respectively. Thermo- the bend in the c-Myb helix. The charged side- dynamic parameters of complex formation are in Table 2. chain of Glu299 (c-Myb) is more accessible to solvent than the equivalent Leu138 (pKID). Thr305 is also located one helical turn from Leu302 and is completely buried in a secondary pocket of KIX, formed by the side-chains of Leu599, His602, Table 2. Thermodynamic parameters of complex Leu603 and Lys606. The hydroxyl group of Thr305 formation is able to hydrogen-bond to the side-chain of Complex pH Buffer K a (mM) DHb (kcal mol21) Lys606. In contrast, the equivalent Asp144 of d pKID, although it interacts also with Lys606, c-Myb–KIX 7.0 Tris 15.0 23.8 remains largely solvent-exposed in the pKID:KIX c-Myb–KIX 5.5 Tris 2.5 29.4 complex. Glu306 of c-Myb makes an electrostatic pKID29c –KIX 7.0 Tris 3.2 214.3 interaction with Arg646 of KIX and van der Waals pKID29–KIX 5.5 Tris 8.3 212.0 contacts with Leu599 and Tyr650, while the equiv- alent Ala145 of pKID is unable to complement the c-Myb–KIX 7.0 Tris 15.0 23.8 charge of Arg646. c-Myb–KIX 7.0 Phosphate 12.5 211.6 Electrostatic interactions between Glu308 of pKID34–KIX 7.0 Tris 1.3 214.7 c-Myb, located on the second helical turn from pKID34–KIX 7.0 Phosphate 1.0 214.0 Leu302, and His602 were observed in most struc- a The variation in Kd in duplicate measurements was tures and with Lys606 of KIX in about half of the typically 15% or better. structures. In contrast, the equivalent Gly147 of b The reproducibility of DH for duplicate measurements was pKID is disordered, since its size and uncharged 21 typically 0.2 kcal mol or better. nature do not support these interactions. Finally, c pKID29 ¼ residues 119–147; pKID34 ¼ residues 116–149. further hydrophobic contacts are made between Solution Structure of KIX:c-Myb Complex 527

Figure 4. Superposition of pKID:KIX and c-Myb:KIX complexes. A, Stereoview of a superposition of pKID:KIX5 and c-Myb:KIX families of structures. The backbone of KIX in complex with pKID (yellow) is colored dark blue, and the KIX backbone in complex with c-Myb (red) is colored light blue. B, Ribbon diagram of representative structures from A. C, Close-up view of hydrophobic pocket interactions. The c-Myb:KIX complex is superimposed on the pKID:KIX complex. The KIX backbone and side-chains are shown in dark blue for the pKID complex and in light blue for the c-Myb complex. The backbone of c-Myb is in red and that of pKID is in yellow. The side-chain of the penetrating leucine residue is shown in the corresponding color. the side-chain of Leu309 from c-Myb and Leu599 and the C-terminal end of the bound helix of and His602 from KIX. In the pKID:KIX complex, c-Myb in the KIX complex are optimized for bind- these two KIX residues interact with Pro146 of ing, consistent with the constitutive binding of pKID. The different spacing of Leu309 (c-Myb) c-Myb. In contrast, the equivalent interactions in and Pro146 (pKID) from the deep pocket-inserted the pKID:KIX complex are not optimal, consistent leucine residue (Leu302 and Leu141, respectively) with the inducible binding mode and the low is the result of the bend in the a helix of c-Myb. affinity of unphosphorylated KID for KIX. The interactions of the conserved leucine residue Another major difference in the interactions of 528 Solution Structure of KIX:c-Myb Complex c-Myb and pKID with KIX occurs at the N-terminal the hydrophobic groove of KIX.6 The c-Myb:KIX end of the helix. Arg294 of c-Myb is located on the structure provides insight into the differences that second helical turn from Leu302, and makes van are mandated in a constitutive transcriptional der Waals contacts with Gln661 and Lys662, and activator by the absence of the phosphate group. an electrostatic interaction with Glu665 of KIX, Analysis of the structure indicates that the inter- while the adjacent Ile295 interacts with Tyr658 and actions at the common binding surface are more Lys662. In the pKID:KIX complex, these two KIX extensive for c-Myb. In particular, insertion of residues interact with the pSer133 of pKID, and Leu302 of c-Myb into the deep hydrophobic pocket Tyr658 interacts additionally with Tyr134 and on KIX provides a major driving-force for binding. Leu128. These favorable interactions between KIX Mutation of Leu302 of c-Myb abrogates binding, and the phosphorylated form of KID are absent and mutations of hydrophobic pocket residues in for c-Myb, which lacks the highly charged phos- KIX have a larger effect on binding of c-Myb than phoserine residue. The interactions would, of on binding of pKID.4 The bend in the helix of course, be absent from unphosphorylated KID. c-Myb enables Leu302 to penetrate into the pocket These differences in specific interactions at the more deeply than the equivalent Leu141 of KID, amino acid side-chain level account for the thereby increasing the number and strength of observed difference in transcriptional activation interactions with KIX residues. Notably, Leu607, mode (constitutive for c-Myb and inducible for which defines the bottom of the hydrophobic CREB), and can explain differences between CREB pocket, interacts only with Leu302 of c-Myb and and c-Myb in affinity for KIX and tolerance of not with Leu141 of KID. Consistent with this find- mutations in KIX.4 Thus, differences in the C-term- ing, Shaywitz et al. showed that the L607F KIX inal half of the helix explain the difference in the mutant interacted more efficiently than wild-type basal levels of binding between c-Myb and KID, KIX with unphosphorylated KID, leading to a sig- while differences in the N-terminal half explain nificant increase in both basal and induced tran- the inducible nature of the pKID–KIX interaction. scriptional activity of CREB.14 No such increase was observed for c-Myb binding to the mutant Structural basis for transcriptional KIX. This mutation appears to rearrange the deep activation mechanism hydrophobic pocket in a way that specifically enhances the imperfect binding of KID but not the The KIX domain of the co-activator CBP utilizes more optimized binding of c-Myb. Additional the hydrophobic groove to bind both the constitutive c-Myb residues that complement the hydrophobic activator c-Myb and the kinase inducible domain surface of KIX better than the corresponding KID (KID) of CREB. The minimal requirement for inter- residues are Leu301, Met303 and Thr305. The latter action with the hydrophobic groove of KIX is bind- side-chain is accommodated perfectly in a second- ing-coupled stabilization of an amphipathic helix, ary pocket on KIX (Figure 2). In addition, the ends which we have shown to be sufficient for the gener- of the hydrophobic groove of KIX are lined by ation of a low-affinity complex. The affinities of KIX three charged residues, His602, Arg646 and for unphosphorylated KID, c-Myb and pKID are Glu665, which can make electrostatic interactions 108 mM, 15 mM and 0.7 mM, respectively.6 The results with c-Myb residues Glu308, Glu306 and Arg294, presented here explain the structural basis for these but not with KID residues. We suggest that the large differences in affinity. Induction of CREB by differences in the interface of the c-Myb and KID phosphorylation increases its affinity for KIX by two complexes are crucial to enable constitutive bind- orders of magnitude, mainly via formation of inter- ing of c-Myb to KIX and at the same time to molecular interactions between the phosphoserine achieve only low-level basal interactions between residue of pKID and Tyr658 and Lys662 of KIX. The CREB and KIX. constitutive activator c-Myb does not have a residue Thus, the lack of specific electrostatic inter- that can substitute for the highly charged phospho- actions in the N-terminal part of the bound c-Myb serine residue of pKID. Glu299 of c-Myb is relatively helix, in contrast to those present at the N terminus close to Tyr658 (Figure 2B), but its position does not of pKID helix aB can explain the inducible nature allow formation of a hydrogen bond with Tyr658 or of the CREB:CBP binding reaction and the higher electrostatic interaction with Lys662. Substitution of affinity of pKID for KIX. Conversely, the presence the phosphate-acceptor Ser133 in CREB by the nega- of the central leucine residue that binds into the tively charged residues aspartate or glutamate does deep pocket, together with additional strong, not lead to constitutive CREB activity,13 further indi- specific interactions in the C-terminal part of the cating that both the specific position and charge bound helix, can explain the constitutive binding density of the phosphoserine residue are crucial for of c-Myb and its higher affinity for KIX compared high-affinity complex formation. Consistent with to that of unphosphorylated KID. this notion, mutation of either Tyr658 or Lys662 of KIX had a drastic effect on its binding to pKID but The LXXLL motif in transcriptional only a modest effect on c-Myb binding.4 activation complexes The affinity of KIX for c-Myb is almost an order of magnitude higher than its affinity for unphos- Leu302 of c-Myb is part of an LXXLL motif, phorylated KID, although both specifically bind to which appears commonly in complexes formed Solution Structure of KIX:c-Myb Complex 529 between nuclear receptors and their co-activators. accompanied by a significantly more favorable Structures of complexes between nuclear receptors enthalpy, while for the pKID–KIX interaction the and coactivators from the p160 family15 – 18 resemble change of enthalpy is in the opposite direction the c-Myb:KIX complex in the binding of an (modestly more favorable at neutral pH) amphipathic helix containing the LXXLL sequence (Figure 3A; Table 2). The significant change of in a hydrophobic groove of the second protein. In thermodynamics in the pH range 5.5–7.0 suggests addition, while the LXXLL motif contributes most that a histidine residue is mediating the effect via of the binding energy, adjacent residues increase an electrostatic interaction that differs for each the affinity and confer specificity.17 In the complex complex. Since there is no histidine in the c-Myb between the thyroid hormone receptor and the and pKID sequences, the putative histidine residue p160 co-activator GRIP1, all three leucine residues must reside in KIX. An alternative explanation for of the motif bind in relatively shallow hydrophobic the pH effect in the pKID–KIX interaction could pockets within the hydrophobic groove and be the ionization of the phosphoserine residue, mutation of either one is detrimental for binding.15 resulting in a stronger electrostatic interaction In contrast, the contribution of the three leucine with Lys662. This possibility is consistent with residues for the c-Myb:KIX interaction varies: 31P-NMR experiments showing titration of the Leu302, which binds in the deep hydrophobic phosphoserine residue in the same pH range.19 pocket, buries 98% of its solvent-accessible surface, However, we have observed a similar increase in Leu298, which binds in a shallow hydrophobic affinity of KIX for unphosphorylated pKID at pH pocket in the hydrophobic groove, buries 80%, 7.0 compared to pH 5.5 (data not shown), and Leu301, which binds in a shallow amphipathic suggesting that the pH effect is unlikely to be pocket outside the hydrophobic groove, buries associated with the phosphoserine residue. only 33%. Consistently, Leu302 has a small toler- The buffer-effect data (Figure 3B) indicate that ance for conservative mutations, Leu298 has a at neutral pH there is uptake of almost a complete larger tolerance for conservative mutations and proton by a KIX residue, probably histidine, upon Leu301 can be mutated even to alanine without a binding to c-Myb. This binding reaction is signifi- significant reduction in binding or transcriptional cantly more favorable at acidic pH, since the pH- activity.4 Indeed, sequence alignment of c-Myb sensitive electrostatic interaction can take place and CREB according to the structures of their KIX without the cost of proton transfer. On the other complexes (Figure 5) reveals the identity of only hand, the proton release required to generate an Leu302. There is a conservative change of Leu298 uncharged histidine residue upon binding of KIX to to isoleucine, and a non-conservative change of pKID causes this binding reaction to be more favor- Leu301 to aspartate. Thus, the c-Myb binding site able as the pH increases from acidic to neutral pH, contains a sequence motif in common with a where the energy cost of proton transfer is avoided. number of other domains involved in binding There are five histidine residues in the KIX con- interactions in transcriptional activation. Although struct used in the structural studies of the pKID5 the specific interactions of the c-Myb LXXLL motif and the c-Myb complexes. His592 and His594 in the KIX complex do not appear to have much form part of the G1 310 helix at the N terminus of in common with such sequences in other com- KIX; their side-chains are rather disordered, and plexes, the presence of this binding “signature” in are far from the bound c-Myb. His602 and His605 c-Myb but not in KID may be correlated with the are in the middle of the a1 helix, and His651 is constitutive binding of c-Myb versus the low basal towards the N terminus of a3. The positions of binding of unphosphorylated KID. these side-chains are shown in the family of c-Myb: KIX structures in Figure 6. The solvent-accessible pH-Dependence of pKID and c-Myb affinities area of His592, His594 and His605 is not changed upon complex formation with either c-Myb or Binding of c-Myb at the more acidic pH of 5.5 is pKID. His651 also appears far from the c-Myb

Figure 5. Sequence alignment of pKID and c-Myb. Sequences of various KID constructs referred to here, aligned with the sequence of the 25-mer c-Myb peptide according to the similarity in the three-dimensional structures of their complexes with the KIX domain of CBP. Sequence KID60 corresponds to the construct used to obtain the solution structure of the pKID:KIX complex.5 Sequence pKID34 was used for the buffer sensitivity ITC measurements, and sequence pKID29 was used for the pH-dependence measurements and for all previous thermodynamic measurements.6 Asterisks ( p ) indicate sites of contact with KIX in the three-dimensional structure of each complex. Helices observed in the complexes are outlined in blue (pKID complex) and red (c-Myb complex). Ser133, which is phosphorylated in the pKID proteins, is outlined in orange. Leu302 of c-Myb, and the analogous Leu141 of pKID, which is buried in the same hydrophobic pocket of KIX, is outlined in yellow. The LXXLL sequence of c-Myb is shown in green letters. 530 Solution Structure of KIX:c-Myb Complex helix, but the side-chain interacts in half of the stitutive c-Myb. Lower affinity may be an advan- pKID:KIX structures with Arg125 from the aA tage for c-Myb, since it is regulated mainly at the helix of pKID. This interaction is electrostatically level of its expression and degradation8,20 and is unfavorable, and therefore the affinity of KIX for required to maintain the proliferative state of pKID should be higher when His651 is deproto- immature hematopoietic cells.21,22 Therefore, the nated and uncharged. The only intermolecular lower binding affinity of the c-Myb activation interaction of His651 in the c-Myb:KIX complex is domain is consistent with the kinetic profile of its between its Ha and the H1 of Met303, which constitutive activity, which is turned on and off would be pH-independent. The most likely histi- slowly and has a cumulative effect over time. In dine residue to mediate the pH-dependence of contrast, CREB is turned on more rapidly by an c-Myb:KIX complex formation is His602, which extra-cellular signal leading to PKA-mediated makes an electrostatic interaction with Glu308 of phosphorylation on Ser133,23 and it is turned off c-Myb. Consistent with the ITC results, proton rapidly following dephosphorylation by the Ser/ uptake would be required for this interaction to Thr phosphatase PP-1.24 Activation of the inducible occur at neutral pH; binding would be more favor- pathway mediated by CREB may require high- able at acidic pH. The residue equivalent to Glu308 affinity binding to CBP in order to compete effec- of c-Myb in the pKID–KIX complex is Gly147, tively with constitutive pathways converging at which is uncharged and therefore unable to make CBP. The fact that c-Myb binds CBP with a lower such an electrostatic interaction. (It is possible relative affinity might also be an advantage for that an electrostatic interaction occurs for one of regulation by other transcription factors, to either the pKID constructs, where Gly147 forms the increase (C/EBPb, Ets-1 and AML-1) or attenuate C-terminal residue; in this case, the C-terminal car- (GATA-1) its activity in a synergistic fashion.1,9,25 boxyl group could possibly substitute for the side- The distinct physiological requirements of the two chain of Glu308.) The structures5 show that His602 transcriptional activators correlate well with the is close to Ala145 and Asp144 of pKID, but the dis- pH optima of co-activator recruitment, physio- tance and angle between the His602 side-chain and logical pH for pKID and lower pH for c-Myb. the latter residue do not support an electrostatic interaction or a hydrogen bond. The non-polar interactions of His602 in the pKID–KIX complex Inducible versus constitutive: a common predict that binding would be stronger in the binding pattern? uncharged state, possibly contributing to the increased affinity of KIX for pKID as the pH is In addition to the KIX domain of CBP, serveral increased from acidic to neutral. other systems in the cell utilize ligand binding that may be inducible or constitutive. For example, the 14-3-3 family of proteins mediates various Physiological rationale for observed signal transduction pathways by binding to a binding affinities number of phosphoserine-containing proteins.26 The structure of a 14-3-3 protein bound to its target The intermolecular interactions of the phospho- phosphopeptide27 reveals many similarities to that serine residue of pKID increase its affinity for KIX of the KIX:pKID complex. The serine-phosphoryl- at neutral pH by 20-fold over that of the con- ated protein binds in an amphipathic groove on

Figure 6. Positions of KIX histidine side-chains. Stereoview of the family of 20 structures of the KIX:c-Myb complex, showing the positions of the five histidine residues of KIX, His592 (orange), His594 (yellow), His602 (green), His605 (blue) and His651 (purple). The backbone of KIX is shown in light blue, and that of c-Myb in red. Solution Structure of KIX:c-Myb Complex 531 the 14-3-3 surface, with an electrostatic interaction helix that binds along the groove and a critical between the phosphoserine and three basic resi- leucine residue that is inserted into the deep dues, and a hydrogen bond between the phospho- hydrophobic pocket of KIX. Optimal comple- serine and a tyrosine residue. Other residues in mentarity of the c-Myb:KIX interface, in particular addition to the phosphoserine, both hydrophobic within the deep hydrophobic pocket, results from and polar, are essential for binding. In spite of the the sequence of c-Myb and from a bend in the significant contribution of non-phosphate inter- helix of bound c-Myb. Modulation of that comple- actions, dephosphorylation results in loss of mentarity enables the KIX domain to distinguish binding.27 In addition, several non-phosphorylated between the constitutive c-Myb that activates tran- proteins are able to bind with high affinity to the scription and the unphosphorylated form of CREB amphipathic groove of 14-3-3 proteins.28,29 The that should not activate transcription. Induction of 14-3-3 proteins are therefore similar to KIX in their CREB by phosphorylation converts the low-affinity ability to bind the inducible protein only after complex into a specific high-affinity complex, phosphorylation and yet to bind constitutive non- mainly via formation of additional inter-molecular phosphorylated proteins with high affinity, using interactions involving the phosphate moiety, the same surface. As it is in the KIX:c-Myb com- which compensate for the imperfect comple- plex, binding of a constitutive non-phosphorylated mentarity of the pKID–KIX interface. peptide to 14-3-3 depends on good comple- mentarity of both hydrophobic and polar residues. However, in contrast to c-Myb, the constitutive Materials and Methods 14-3-3 binding peptide has two acidic residues that substitute for the doubly charged phospho- Protein expression and purification serine by making electrostatic interactions in the same basic pocket that accommodates the phos- The KIX domain (residues 586–672) of mouse CBP (amino acid sequence identical with that of human) was phoserine residue, thereby bringing the affinity of expressed in either unlabeled or uniformly 15Nor the constitutive peptide for 14-3-3 to the level of 13 15 30,31 C, N-enriched forms, in BL21(DE3) Escherichia coli and inducible peptides in the phosphorylated state. purified to homogeneity as described.5 Uniformly 15Nor By contrast, the absence of such acidic residues in 13C,15N-labeled c-Myb (residues 291–315 from mouse) c-Myb keeps its affinity for KIX lower than that of were overexpressed in E. coli using a ubiquitin fusion pKID in spite of the excellent interface protein system (a generous gift from Dr Toshiyuki complementarity. Kohno) and purified to homogeneity as described.6 The functional similarity between KIX and 14-3-3 Unlabeled c-Myb and phosphorylated KID peptides proteins suggests a general pattern of recognition (residues 119–147 (pKID29) or residues 116–149 for proteins that competitively bind both phos- (pKID34) of mouse CREB, amino acid sequence identical with that of human) were synthesized chemically using a phorylated and non-phosphorylated targets. The Perseptive Biosystems peptide synthesizer (Perkin– binding of an inducible protein would involve Elmer) and purified to homogeneity by reverse-phase strong interaction with the phosphate moiety by HPLC. The identity and integrity of all proteins and electrostatic and hydrogen bonds and imperfect peptides were confirmed by mass spectrometry. but essential additional interactions along the groove. Those interactions would not be sufficient NMR spectroscopy for binding in the absence of the phosphate but ensure specificity and higher affinity of binding NMR samples were prepared in the concentration 2 following the phosphorylation event. The binding range of 0.5–1.0 mM in 90% (v/v) H2O/10% (v/v) H2O of a constitutive partner would involve optimal buffer (20 mM Tris d11-acetate d4 (pH 5.5), 50 mM NaCl, complementarity along the groove to ensure 2 mM NaN3). NMR spectra were recorded at 27 8Con specificity and reasonable affinity and, if higher Bruker AMX500, DRX600 and DMX750 spectrometers, affinity is required, it may dictate the presence of equipped with triple axis gradient probes. Titration of acidic residues substituting for the phosphate labeled KIX with unlabeled c-Myb or labeled c-Myb with unlabeled KIX was monitored by 2D 1H–15N interactions. HSQC spectra. The unlabeled component was added in 10–20% excess. NMR data processing and analysis were performed using Felix97 (Molecular Simulations Inc. Conclusions San Diego) or NMRPipe32 and NMRView.33 Nearly com- plete backbone assignments for c-Myb-bound KIX were The structure of the KIX:c-Myb complex and our accomplished using 3D HNCACB,34,35 CBCA(CO)NH,36 thermodynamic analysis provide new insights into HNCO,37 HCACO,38 15N-edited NOESY-HSQC and 15N- 39 ð ¼ Þ the mechanism of recognition of various transcrip- edited TOCSY-HSQC tm 51 ms spectra. Only the tional activator domains by the highly conserved backbone amides of Arg623 and Lys667 were not assign- KIX domain of CBP and p300, and illustrate the able due to exchange broadening. Aliphatic side-chain resonances were assigned using 3D 15N-edited TOCSY, structural and thermodynamic basis for con- HCCH-TOCSY, and HCCH-COSY40 spectra. Aromatic stitutive activation by c-Myb and inducible side-chain resonances were assigned from 3D CB(CG)CD activation by CREB. We suggest that the minimal and CB(CG)CE41 spectra. Although backbone resonances requirements for low-affinity interaction with the could be assigned, only partial assignments could be hydrophobic groove of KIX are an amphipathic a made for the side-chain protons of Arg600, Lys606, 532 Solution Structure of KIX:c-Myb Complex

Lys621 and the histidine residues. Complete backbone all peptides and proteins against the ITC buffer, followed assignments for KIX-bound c-Myb were accomplished by filtration and degassing was performed to minimize using 3D HNCACB and CBCA(CO)NH spectra. 3D background noise. The ITC buffer was prepared from 15 13 N-edited TOCSY-HSQC ðtm ¼ 43:2msÞ and C-edited 50 mM Tris titrated to pH 7.0 with HCl or to pH 5.5 HCCH-TOCSY, HCCH-COSY, CCH-COSY, and CCH- with acetic acid or 50 mM sodium phosphate buffer at TOCSY42 spectra were acquired for side-chain assign- pH 7.0. All ITC buffers included 50 mM NaCl. KIX and ments. Several exchange-broadened resonances could be pKID concentrations were determined by measuring assigned only by using sequential NOE connectivities absorbance at 280 nm. The stoichiometric ratio obtained observed in 3D 15N and 13C-edited NOESY spectra. The from the curve fit was consistently 1:1, within 5% error. 13C chemical shifts were referenced relative to 2,2- The concentration of c-Myb was determined according dimethyl-2-silapentane-5-sulphonate (DSS). to the stoichiometric ratio obtained from the curve fit. The concentration of KIX in the ITC cell was 40–225 mM Generation of restraints with higher concentrations used for the lower affinity and/or lower enthalpy complexes, while the concen- Intra-molecular distance constraints for KIX and tration of the peptide in the syringe was 12-fold over m c-Myb protons were obtained from 15N-edited NOESY that of KIX. Typically, two injections of 5 l were fol- m (t ¼ 80 ms and 100 ms, respectively) and 13C-edited lowed by 28 injections of 10 l until a molar ratio of 2.5 m was obtained. Integration of the thermogram and sub- NOESY (tm ¼ 120 ms and 80 ms, respectively). Intensi- ties were calibrated against known inter-proton distances traction of the blanks yielded a binding isotherm that in regular structural elements (a helices) or between was fit to a model of one-site interaction (ITC data protons of fixed separation. Upper limits of distance con- analysis software in Origin 2.3 of MicroCal Inc.). straints were 3 A˚ ,4A˚ and 5 A˚ , while all lower limits were set to the van der Waals contact distance (1.8 A˚ ). Data bank accession numbers Due to the longer mixing time, the upper limits for the constraints derived from 15N-edited NOESY, were ˚ The assigned chemical shift list has been deposited in increased by 0.5 A. Appropriate pseudo-atom corrections the BioMagResBank with accession number 6095 and to the upper bounds were applied to constraints involv- the coordinates in the with accession ing methyl and methylene groups, and aromatic ring number 1SB0. protons. No explicit hydrogen bonding constraints were imposed. Inter-molecular distance constraints were 13 13 derived from a 3D C(v2)-edited, C(v3)-filtered 43 NOESY (tm ¼ 120 ms) experiment. A scaling factor was determined by comparing the intensities of well- resolved peaks with those of the corresponding peaks in Acknowledgements the 13C-edited NOESY spectrum acquired for KIX. On that basis, upper bounds of 3 A˚ ,4A˚ and 5 A˚ were This work was supported by grant CA96865 assigned. Backbone f and w angles in helical regions, as from the National Institutes of Health and by the indicated by the chemical-shift index,44 were restrained Skaggs Institute for Chemical Biology. T.Z. is an ^ ^ to 250( 30)8 and 240( 30)8, respectively. EMBO fellow and R.N.D. is supported by the Leukemia and Lymphoma Society. We thank Dr Structure calculations and analysis Toshiyuki Kohno for providing us with the ubiquitin fusion construct. We are grateful to Initial structures were calculated separately with the 45 Linda Tennant and Dr Maria Martinez-Yamout for program DYANA for KIX and c-Myb, using intra-mol- help with sample preparation, and Drs Eduardo ecular distance and torsion angle constraints determined Zaborowski, Gerard Kroon and John Chung for in the bound state. Intermolecular restraints were then added to dock the c-Myb:KIX complex, and 200 expert advice regarding NMR experiments. We DYANA structures were generated. The 30 structures are particularly grateful to Dr Natalie Goto for with the lowest target functions were refined by simu- critical reading of the manuscript and we thank lated annealing using the AMBER 6 software package.46 Dr Ishwar Radhakrishnan and Gabriela Perez- The final structures were calculated by one cycle of Alvarado for helpful discussions. DYANA, generating 100 structures with lowest target functions (out of 200) for refinement by two cycles of AMBER. The 20 structures with the lowest constraint violation energies were selected for analysis. The struc- References tures show excellent convergence, low AMBER energies 1. Goodman, R. H. & Smolik, S. (2000). CBP/p300 in (mean of 21700 kcal mol21) and no NOE violation ˚ cell growth, transformation, and development. Genes greater than 0.3 A. Statistics are given in Table 1. Dev. 14, 1553–1577. Analysis of the structures was performed using 47 2. Giordano, A. & Avantaggiati, M. L. (1999). p300 and PROCHECK, and in-house programs. Graphics images CBP: partners for life and death. J. Cell. Phys. 181, were generated and interface area was calculated using 48 218–230. MOLMOL. 3. Chrivia, J. C., Kwok, R. P., Lamb, N., Hagiwara, M., Montminy, M. & Goodman, R. H. (1993). Isothermal titration calorimetry (ITC) Phosphorylated CREB binds specifically to nuclear protein CBP. Nature, 365, 855–859. The titration of a pKID or c-Myb peptide into a KIX 4. Parker, D., Ferreri, K., Nakajima, T., LaMorte, V. J., protein was performed at 27 8C using an MCS titration Evans, R., Koerber, S. C. et al. (1996). Phosphorylation calorimeter from MicroCal Inc. Simultaneous dialysis of of CREB at Ser133 induces complex formation with Solution Structure of KIX:c-Myb Complex 533

CPB via a direct mechanism. Mol. Cell. Biol. 16, p160 nuclear receptor coactivators. Nature, 415, 694–703. 549–553. 5. Radhakrishnan, I., Pe´rez-Alvarado, G. C., Parker, D., 19. Mestas, S. P. & Lumb, K. J. (1999). Electrostatic contri- Dyson, H. J., Montminy, M. R. & Wright, P. E. bution of phosphorylation to the stability of the (1997). Solution structure of the KIX domain of CBP CREB-CBP activator– complex. Nature bound to the transactivation domain of CREB: a Struct. Biol. 6, 613–614. model for activator:coactivator interactions. Cell, 91, 20. Bies, J. & Wolff, L. (1997). Oncogenic activation of 741–752. c-Myb by carboxyl-terminal truncation leads to 6. Zor, T., Mayr, B. M., Dyson, H. J., Montminy, M. R. & decreased proteolysis by the ubiquitin-26S protea- Wright, P. E. (2002). Roles of phosphorylation and some pathway. Oncogene, 14, 203–212. helix propensity in the binding of the KIX domain 21. Westin, E. H., Gallo, R. C., Arya, S. K., Eva, A., Souza, of CREB-binding protein by constitutive (c-Myb) L. M., Baluda, M. A. et al. (1982). Differential and inducible (CREB) activators. J. Biol. Chem. 277, expression of the amv gene in human hematopoietic 42241–42248. cells. Proc. Natl Acad. Sci. USA, 79, 2194–2198. 7. Parker, D., Rivera, M., Zor, T., Henrion-Caude, A., 22. Gonda, T. J. & Metcalf, D. (1984). Expression of , Radhakrishnan, I., Kumar, A. et al. (1999). Role of myc and fos proto-oncogenes during the differen- secondary structure in discrimination between con- tiation of a murine myeloid leukaemia. Nature, 310, stitutive and inducible activators. Mol. Cell. Biol. 19, 249–251. 5601–5607. 23. Hagiwara, M., Brindle, P., Harootunian, A., 8. Weston, K. (1998). Myb proteins in life, death and Armstrong, R., Rivier, J., Vale, W. et al. (1993). differentiation. Curr. Opin. Genet. Dev. 8, 76–81. Coupling of hormonal stimulation and transcription 9. Oh, I. H. & Reddy, E. P. (1999). The myb gene family via the cyclic AMP-responsive factor CREB is rate in cell growth, differentiation and apoptosis. limited by nuclear entry of protein kinase A. Mol. Oncogene, 18, 3017–3033. Cell. Biol. 13, 4852–4859. 10. Radhakrishnan, I., Pe´rez-Alvarado, G. C., Dyson, 24. Hagiwara, M., Alberts, A., Brindle, P., Meinkoth, J., H. J. & Wright, P. E. (1998). Conformational Feramisco, J., Deng, T. et al. (1992). Transcriptional preferences in the Ser133-phosphorylated and non- attenuation following cAMP induction requires phosphorylated forms of the kinase inducible trans- PP-1-mediated dephosphorylation of CREB. Cell, 70, activation domain of CREB. FEBS Letters, 430, 105–113. 317–322. 25. Blobel, G. A. (2000). CREB-binding protein and p300: 11. Radhakrishnan, I., Pe´rez-Alvarado, G. C., Parker, D., molecular integrators of hematopoietic transcription. Blood, 95, 745–755. Dyson, H. J., Montminy, M. R. & Wright, P. E. (1999). 26. Tzivion, G. & Avruch, J. (2002). 14-3-3 proteins: Structural analyses of CREB-CBP transcriptional active cofactors in cellular regulation by serine/ activator-coactivator complexes by NMR spec- threonine phosphorylation. J. Biol. Chem. 277, troscopy: implications for mapping the boundaries 3061–3064. of structural domains. J. Mol. Biol. 287, 859–865. 27. Yaffe, M. B., Rittinger, K., Volinia, S., Caron, P. R., 12. Baker, B. M. & Murphy, K. P. (1996). Evaluation of Aitken, A., Leffers, H. et al. (1997). The structural linked protonation effects in protein binding basis for 14-3-3: phosphopeptide binding specificity. reactions using isothermal titration calorimetry. Cell, 91, 961–971. Biophys. J. 71, 2049–2055. 28. Zhang, L., Wang, H., Masters, S. C., Wang, B., 13. Gonzalez, G. A. & Montminy, M. R. (1989). Cyclic Barbieri, J. T. & Fu, H. (1999). Residues of 14-3-3 AMP stimulates somatostatin gene transcription by zeta required for activation of exoenzyme S of phosphorylation of CREB at serine-133. Cell, 59, Pseudomonas aeruginosa. Biochemistry, 38, 675–680. 12159–12164. 14. Shaywitz, A. J., Dove, S. L., Kornhauser, J. M., 29. Masters, S. C., Pederson, K. J., Zhang, L., Barbieri, J. T. Hochschild, A. & Greenberg, M. E. (2000). Magni- & Fu, H. (1999). Interaction of 14-3-3 with a non- tude of the CREB-dependent transcriptional phosphorylated protein ligand, exoenzyme S of response is determined by the strength of the inter- Pseudomonas aeruginosa. Biochemistry, 38, 5216–5221. action between the kinase-inducible domain of 30. Wang, B., Yang, H., Liu, Y. C., Jelinek, T., Zhang, L., CREB and the KIX domain of CREB-binding protein. Ruoslahti, E. & Fu, H. (1999). Isolation of high- Mol. Cell. Biol. 20, 9409–9422. affinity peptide antagonists of 14-3-3 proteins by 15. Darimont, B. D., Wagner, R. L., Apriletti, J. W., phage display. Biochemistry, 38, 12499–12504. Stallcup, M. R., Kushner, P. J., Baxter, J. D. et al. 31. Petosa, C., Masters, S. C., Bankston, L. A., Pohl, J., (1998). Structure and specificity of nuclear receptor- Wang, B. C., Fu, H. I. & Liddington, R. C. (1998). coactivator interactions. Genes Dev. 12, 3343–3356. 14-3-3 zeta binds a phosphorylated Raf peptide and 16. Nolte, R. T., Wisely, G. B., Westin, S., Cobb, J. E., an unphosphorylated peptide via its conserved Lambert, M. H., Kurokawa, R. et al. (1998). Ligand amphipathic groove. J. Biol. Chem. 273, 16305–16310. binding and co-activator assembly of the peroxisome 32. Delaglio, F., Grzesiek, S., Vuister, G. W., Guang, Z., proliferator-activated receptor-gamma. Nature, 395, Pfeifer, J. & Bax, A. (1995). NMRPipe: a multidimen- 137–143. sional spectral processing system based on UNIX 17. McInerney, E. M., Rose, D. W., Flynn, S. E., Westin, S., pipes. J. Biomol. NMR, 6, 277–293. Mullen, T. M., Krones, A. et al. (1998). Determinants 33. Johnson, B. A. & Blevins, R. A. (1994). NMRView: a of coactivator LXXLL motif specificity in nuclear computer program for the visualization and analysis receptor transcriptional activation. Genes Dev. 12, of NMR data. J. Biomol. NMR, 4, 604–613. 3357–3368. 34. Grzesiek, S. & Bax, A. (1992). An efficient experiment 18. Demarest, S. J., Martinez-Yamout, M., Chung, J., for sequential backbone assignment of medium- Chen, H., Xu, W., Dyson, H. J. et al. (2002). Mutual sized isotopically enriched proteins. J. Magn. Reson. synergistic folding in recruitment of CBP/p300 by 99, 201–207. 534 Solution Structure of KIX:c-Myb Complex

35. Wittekind, M. & Mueller, L. (1993). HNCACB, a 42. Lo¨hr,F.&Ru¨ terjans, H. (1996). Novel pulse high-sensitivity 3D NMR experiment to correlate sequences for the resonance assignment of aromatic amide-proton and nitrogen resonances with the side-chains in 13C-labeled proteins. J. Magn. Reson. alpha- and beta-carbon resonances in proteins. ser. B, 112, 259–268. J. Magn. Reson. 101, 201–205. 43. Zwahlen, C., Gardner, K. H., Sarma, S. P., Horita, 36. Grzesiek, S. & Bax, A. (1992). Correlating backbone D. A., Byrd, R. A. & Kay, L. E. (1998). An NMR amide and side-chain resonances in larger proteins experiment for measuring methyl–methyl NOEs in by multiple relayed triple resonance NMR. J. Am. 13C-labeled proteins with high resolution. J. Am. Chem. Soc. 114, 6291–6293. Chem. Soc. 120, 7617–7625. 37. Bax, A. & Grzesiek, S. (1993). Methodological 44. Wishart, D. S. & Sykes, B. D. (1994). Chemical shifts advances in protein NMR. Acc. Chem. Res. 26, as a tool for structure determination. Methods 131–138. Enzymol. 239, 363–392. 38. Kay, L. E., Ikura, M., Tschudin, R. & Bax, A. (1990). 45. Gu¨ ntert, P., Mumenthaler, C. & Wu¨ thrich, K. (1997). Three-dimensional triple-resonance NMR spec- Torsion angle dynamics for NMR structure calcu- troscopy of isotopically enriched proteins. J. Magn. lation with the new program DYANA. J. Mol. Biol. Reson. 89, 496–514. 273, 283–298. 39. Zhang, O., Kay, L. E., Olivier, J. P. & Forman-Kay, J. D. 1 15 46. Pearlman, D. A., Case, D. A., Caldwell, J. W., Ross, (1994). Backbone H and N resonance assignments W. S., Cheatham, T. E., III, DeBolt, S. et al. (1995). of the N-terminal SH3 domain of drk in folded and AMBER, a package of computer programs for apply- unfolded states using enhanced-sensitivity pulsed ing molecular mechanics, normal mode analysis, field gradient NMR techniques. J. Biomol. NMR, 4, molecular dynamics and free energy calculations to 845–858. simulate the structural and energetic properties of 40. Bax, A., Clore, G. M. & Gronenborn, A. M. (1990). molecules. Comp. Phys. Commun. 91, 1–41. 1H–1H correlation via isotropic mixing of 13C 47. Laskowski, R. A., Rullmann, J. A. C., MacArthur, magnetization, a new three-dimensional approach M. W., Kaptein, R. & Thornton, J. M. (1996). AQUA for assigning 1H and 13C spectra of 13C-enriched and PROCHECK-NMR: programs for checking the proteins. J. Magn. Reson. 88, 425–431. quality of protein structures solved by NMR. 41. Yamazaki, T., Forman-Kay, J. D. & Kay, L. E. (1993). J. Biomol. NMR, 8, 477–486. Two-dimensional NMR experiments for correlating ¨ 13Cb and 1Hd/1 chemical shifts of aromatic residues 48. Koradi, R., Billeter, M. & Wuthrich, K. (1996). in 13C-labeled proteins via scalar couplings. J. Am. MOLMOL: a program for display and analysis of Chem. Soc. 115, 11054–11055. macromolecular structures. J. Mol. Graph. 14, 51–55.

Edited by M. F. Summers

(Received 21 January 2004; accepted 23 January 2004)