STD and TRNOESY NMR Studies on the Conformation of the Oncogenic Protein β-Catenin Containing the Phosphorylated Motif DpSGXXpS Bound to the β-TrCP Protein*

β-TrCP is the F-box protein component of an Skp1/Cul1/F-box (SCF)-type ubiquitin ligase complex. Biochemical studies have suggested that β-TrCP targets the oncogenic protein β-catenin for ubiquitination and followed by proteasome degradation. To further elucidate the basis of this interaction, a complex between a 32-residue peptide from β-catenin containing the phosphorylated motif DpSGXXpS (P-β-Cat17–48) and β-TrCP was studied using Saturation Transfer Difference (STD) Nuclear Magnetic Resonance (NMR) experiments. These experiments make it possible to identify the binding epitope of a ligand at atomic resolution. An analysis of STD spectra provided clear evidence that only a few of the 32 residues receive the largest saturation transfer. In particular, the amide protons of the residues in the phosphorylated motif appear to be in close contact to the amino acids of the β-TrCP binding pocket. The amide and aromatic protons of the His24 and Trp25 residues also receive a significant saturation transfer. These findings are in keeping with a recently published x-ray structure of a shorter β-catenin fragment with the β-TrCP1-Skp1 complex and with the earlier findings from mutagenesis and activity assays. To better characterize the ligand-protein interaction, the bound conformation of the phosphorylated β-catenin peptide was obtained using TRansfer Nuclear Overhauser Effect SpectroscopY (TRNOESY) experiments. Finally, we obtained the bound structure of the phosphorylated peptide showing the protons identified by STD NMR as exposed in close proximity to the molecule surface.

␤-TrCP is the F-box protein component of an Skp1/ Cul1/F-box (SCF)-type ubiquitin ligase complex. Biochemical studies have suggested that ␤-TrCP targets the oncogenic protein ␤-catenin for ubiquitination and followed by proteasome degradation. To further elucidate the basis of this interaction, a complex between a 32residue peptide from ␤-catenin containing the phosphorylated motif DpSGXXpS (P-␤-Cat 17-48 ) and ␤-TrCP was studied using Saturation Transfer Difference (STD) Nuclear Magnetic Resonance (NMR) experiments. These experiments make it possible to identify the binding epitope of a ligand at atomic resolution. An analysis of STD spectra provided clear evidence that only a few of the 32 residues receive the largest saturation transfer. In particular, the amide protons of the residues in the phosphorylated motif appear to be in close contact to the amino acids of the ␤-TrCP binding pocket. The amide and aromatic protons of the His 24 and Trp 25 residues also receive a significant saturation transfer. These findings are in keeping with a recently published x-ray structure of a shorter ␤-catenin fragment with the ␤-TrCP1-Skp1 complex and with the earlier findings from mutagenesis and activity assays. To better characterize the ligand-protein interaction, the bound conformation of the phosphorylated ␤-catenin peptide was obtained using TRansfer Nuclear Overhauser Effect SpectroscopY (TRNOESY) experiments. Finally, we obtained the bound structure of the phosphorylated peptide showing the protons identified by STD NMR as exposed in close proximity to the molecule surface.
The ubiquitin-proteasome pathway of protein degradation is essential for various important biological processes including cell cycle progression, gene transcription, and signal transduction (1,2). This work is based on the study of the oncogenic protein ␤-catenin (␤-Cat), 1 which plays an essential role in the Wingless/Wnt signaling pathway and is an important component of cadherin cell-adhesion complexes (Fig. 1A). The abundance of ␤-catenin in the cytoplasm is regulated by ubiquitindependent proteolysis (3), and Wnt signaling is regulated by the presence or absence of the intracellular protein ␤-catenin. When Wnt signal is absent, the signal transduction pathway is off because ␤-catenin is rapidly destroyed. A large multiprotein machine normally facilitates the addition of phosphate groups to ␤-catenin by glycogen synthase kinase-3␤ (GSK3␤). Phosphorylated ␤-catenin binds to a protein called ␤-TrCP and is then modified by the covalent addition of a small protein called ubiquitin. Proteins tagged with ubiquitin are degraded by the 26 S proteasome, the protein-recycling center of the cell. When cells are exposed to the Wnt signal, it binds to cell surface receptors. Receptor activation blocks ␤-catenin phosphorylation, and its subsequent ubiquitination by an unknown mechanism that requires the intervention of the Disheveled protein (Dsh). ␤-Catenin is thus diverted from the proteasome. It accumulates and enters the nucleus, where it finds a partner, a DNA-binding protein of the TCF/LEF family. Together, they activate new gene expression programs. In human colon cancer cells, inappropriate activation of the Wnt pathway initiates cell proliferation by activating genes encoding oncoproteins and cell cycle regulators (4,5). Thus, the abnormal accumulation of ␤-catenin as a result of mutations, either in the adenomatous polyposis coli tumor suppressor protein or in ␤-catenin itself, is thought to initiate colorectal neoplasia and has been implicated in various human cancers. The formation of ubiquitin-protein conjugates requires three enzymes that participate in a cascade of ubiquitin transfer reactions: a ubiquitin-activating enzyme (E1), a ubiquitin-conjugating enzyme (E2), and a ubiquitin ligase (E3). ␤-Catenin contains a DSGXXS sequence that includes the phosphorylation site required for ubiquitination (3,6). Then proteins polyubiquitinated by these enzymes undergo degradation by the 26 S proteasome (7,8). Skp1/Cul1/F-box (SCF) complexes constitute a class of E3 enzymes that are required for the degradation of many key cellular proteins. ␤-TrCP is the F-box protein that functions as a receptor for target molecules such as phosphorylated ␤-catenin, phosphorylated IB␣ (the inhibitor of the master transcription factor NFB), and phosphorylated HIV-1 protein Vpu. The substrate specificity of SCF complexes is thus thought to be determined by the F-box protein (Fig. 1B).
It has been shown that ␤-TrCP is a member of the WD40 repeat-containing family and specifically only recognizes ␤-catenin, IB␣, and Vpu as substrates when they are phosphorylated at both serine residues in the conserved DSGXXS motif (6, 9 -11). Human ␤-TrCP is composed of two main domains: an F-box near the N terminus involved in interaction with Skp1, which is a connecting factor to the other subunits of the SCF E3 ligase complex for ubiquitin-mediated proteolysis; and at the C terminus, the seven WD repeats necessary and sufficient for binding phosphorylated protein substrates of ␤-TrCP. Hence, all substrates of ␤-TrCP, either viral like Vpu or cellular like IB␣, share this consensus motif, which is also present in many cellular signaling proteins. The structure of the SCF ␤-TrCP ubiquitin-ligase complex and its function are depicted in Fig. 1B. The phosphorylation of the motif on the two serine residues is required for interaction with ␤-TrCP, and the subsequent targeting of these substrates leads to their degradation by the 26 S proteasome. Thus, the elucidation of the mechanism involved in substrate recognition by ␤-TrCP upon phosphorylation of this DSGXXS motif is essential.
In a previous study, we examined the structural influence of phosphorylation at the two sites Ser 33 and Ser 37 of the free 32-residue phosphorylated peptide (12) (hereafter referenced as P-␤-Cat 17-48 , Fig. 1A). We now report the ␤-TrCP-bound conformation of P-␤-Cat 17-48 and its binding, studied by NMR methods. An array of one-dimensional and two-dimensional NMR spectroscopic techniques was investigated for the structural characterization and spectral assignment of the ligand.
To drive rationally new therapeutic approaches, an understanding of the underlying molecular mechanism of the interaction of phosphorylated ␤-catenin with ␤-TrCP is essential. NMR-based screening methods have been developed in the last few years to characterize the binding process (13). For binding ligands, an NMR approach (14,15), usually referred to as TRansferred Nuclear Overhauser Enhancement (TRNOE) spectroscopy, is successfully applied to investigate the conformation of a ligand in its bound form. With regard to the structural analysis of the ␤-TrCP-bound conformations, TRNOESY experiments were recorded and were further completed by molecular modeling studies to determine the structural properties of the P-␤-Cat 17-48 peptide when bound to ␤-TrCP.
A further technique, the Saturation Transfer Difference (STD) NMR method, allows determination of binding epitopes (13,16). This method is based on magnetization transfer by protein signals and their relayed effect to ligand. During the saturation period, progressive saturation transfers from the protein to the ligand protons when the ligand binds to the target. The ligand protons nearest to the protein are most likely to be saturated to the highest degree and therefore have the strongest signal in the STD spectrum; whereas the ligand protons located further away are saturated to a lower degree, and their STD intensities are weaker. Therefore, the degree of saturation of individual ligand protons reflects their proximity to the protein surface and can be used as an epitope method to describe the target-ligand interactions (13,17,18). Thus, STD NMR experiments were performed to define the binding epitopes of the P-␤-Cat 17-48 peptide.
A crystal structure of the complex involving a peptide derived from the ␤-catenin sequence and ␤-TrCP was recently published (19). It was interesting to compare the two structures. Comparing the x-ray complex with the experimental NMR data could lead to a selection of a particular model, especially in the absence of crystallographic information on the different models of the ␤-TrCP complexes. : peptides ␤-Cat 17-48 with the amino acid sequence DRKAAVSHWQQQSYLDSGIHSGATTTAPSLSG and P-␤-Cat 17-48 containing the phosphorylated sites 33 and 37, Ser(PO 3 H 2 ), DRKAAVSHWQQQSYLDpSGIHpSGATTTAPSLSG were purchased from Neosystem Laboratories. The purity of the peptides (95%) was tested by analytical HPLC and by mass spectrometry.

Reagents-␤-Cat fragments
Purification of the WD Repeat Region from Human Protein ␤-TrCP-The ␤-TrCP protein was expressed as a fusion protein with glutathione FIG. 1. A, schematic representation of ␤-catenin. a, three-dimensional structure of a protease-resistant fragment of ␤-catenin containing the armadillo repeat region. The core region of ␤-catenin is composed of 12 copies of a 42-amino acid sequence motif known as an armadillo repeat. The 12 repeats form a super helix of helices that features a long positively charged groove of the proteolysis-resistant fragment (42). The structure of the N-and C-terminal domains remains unresolved. b, primary structure sequence of the full ␤-catenin protein. The 12 armadillo repeats are shown in green. The phosphorylation site containing the consensus motif DpSGXXpS is shown in yellow. c, sequence (32 residues) of the phosphorylated ␤-catenin fragment, which was investigated in the present work. B, structure of the SCF ␤-TrCP ubiquitinligase complex. A model of regulation of ␤-catenin degradation is depicted: ␤-TrCP recruits the Skp1/Cul1/E2 ubiquitination apparatus (shown respectively in gray, yellow, and blue) to the complex, leading to multi-ubiquitination of phosphorylated ␤-catenin. The multi-ubiquitinated ␤-catenin protein is then recognized by 26 S proteasome and degraded (8).
The expression, separation, and purification of GST-␤-TrCP were accomplished following a standard GST fusion protein protocol that was previously described (21). This amount of purified recombinant protein was then used to prepare the NMR samples.
NMR Spectroscopy-1 H NMR and STD spectra were recorded with a Bruker AMX-500 spectrometer. Standard Bruker software was used to acquire and process the NMR data. NMR samples contained nonphosphorylated ␤-Cat 17-48 or phosphorylated P-␤-Cat 17-48 with or without the GST-␤-TrCP protein. They were prepared in phosphate-buffered saline solution, pH 7.1 (20 mM phosphate, 5% D 2 O, 0.02% NaN 3 ). 1 H NMR spectra were recorded at constant temperature (278 K). The NMR samples were adjusted to a protein concentration of 0.02 mM based on the visible absorption at 595 nm. A 150-fold ligand excess (3 mM) over binding sites were used throughout the studies. One sample control was prepared containing the ␤-Cat 17-48 nonphosphorylated peptide with the GST-␤-TrCP protein at the same ratio 150:1 as that used for the P-␤-Cat 17-48 /GST-␤-TrCP sample. A second negative control was done using P-␤-Cat 17-48 with the fusion protein GST-Nef (22).
Resonances of the peptide were assigned, based on two-dimensional TOCSY and NOESY experiments. Water suppression was achieved by WATERGATE (23). Chemical shifts were referred to internal TSPD 4 , 3-(trimethylsilyl) [2,2,3,3-d 4 ] propionic acid, sodium salt (Tables I and  II). Two-dimensional TRNOESY spectra were recorded to determine the most favorable ratio for TRNOE effects, which was found to be 150:1. TRNOESY experiments were performed with mixing times of 100, 200, and 400 ms for molar ratios between 600:1 and 25:1. After optimization of the P-␤-Cat 17-48 peptide/protein ratio, the final sample was prepared with 0.5 mg of GST-␤-TrCP protein (0.02 mM protein) and 5.2 mg of the P-␤-Cat 17-48 peptide (3 mM; 150:1 peptide/binding site ratio) in 500 l of buffer. TRNOESY experiments were acquired in phase-sensitive mode using the States-TPPI method, with 4k points and 512 t 1 increments, a relaxation delay of 1 s. The mixing time of the transferred NOE experiment was set to 200 ms according to the TRNOE build-up curves of P-␤-Cat 17-48 peptide in the presence of the GST-␤-TrCP protein.
For STD experiments, the ligand to protein ratio was set to 150:1 (3 mM P-␤-Cat 17-48 peptide, 0.02 mM GST-␤-TrCP protein). STD NMR spectra were acquired using a series of 40 equally spaced 50-ms gaussian-shaped pulses for selective saturation, with 1-ms delay between the pulses, and a total saturation time of ϳ2 s. With an attenuation of 50 dB, the radio frequency field strength for the selective saturation pulses in all STD NMR experiments was 190 Hz. The frequency of the protein (on-resonance) saturation was set to the protein 1 H NMR signals in the low frequency region Ϫ3 ppm. The off-resonance saturation frequency was set at 30.0 ppm. Subtraction of FID values with on-and off-resonance protein saturation was achieved by phase cycling. As no baseline distortion was observed, no T 1 filter was applied to eliminate the background resonances of the GST-␤-TrCP protein. A total relaxation delay (Aq ϩ d1) of 3.4 s and 8 dummy scans were employed to reduce subtraction artifacts. 1k total scans (or 10k for a better signal to noise ratio) were collected. One-dimensional spectra were multiplied with exponential functions (LB ϭ 2 Hz) and zero-filled two times.
Relative STD values were calculated by dividing STD signal intensities by the intensities of the corresponding signals in a one-dimensional 1 H NMR reference spectrum of the same sample recorded with 512 scans and similarly processed.
For STD experiments, several control samples were prepared to verify the specificity of the interaction. We performed one-dimensional STD NMR experiments with P-␤-Cat 17-48 and other fusion proteins such as GST-Nef (22) or GST-N-Ter, which corresponds to the GST protein fused with the first 260 residues of the full-length ␤-TrCP protein, including the F-box, but not the 7 WD domains. These experiments resulted in no observable signal, proving the specificity of the interaction observed between P-␤-Cat 17-48 and the GST-␤-TrCP fusion protein.
Investigation of the time dependence of the saturation transfer with saturation times from 0.2 to 4.0 s showed that 2 s was leading to efficient transfer of saturation from the protein to the ligand protons.
Two-dimensional STD inverse correlation ( 1 H-13 C)-phase-sensitive HSQC experiments were recorded using echo/anti-echo gradient selection with decoupling during acquisition (24). The two-dimensional 1 H/ 13 C correlation were obtained via double INEPT transfer using TRIM pulse of 0.2-ms duration. One millisecond half-sinusoid echo/anti-echo gradients of 40, 10 G/cm with recovery delay of 200 s were used. To select protons attached to carbon, a delay of 1.78 ms was optimum for C-H one-bond couplings: delay ϭ 1/(4 1 J C-H ). This technique is optimized to obtain C ␣ -H ␣ correlations with a sensitivity enhancement. The spectra were recorded with 512 experiments of 2048 data points and 48 scans per t 1 experiment. Subtraction of the on-and off-resonance spectra was made by phase cycling. The acquisition times for the twodimensional experiments were typically around 20 h with a relaxation delay (Aq ϩ D 1 ) of 2.2 s. All spectra were processed with Bruker XWINNMR software. Two-dimensional spectra were multiplied with 90°-shifted squared sine bells in both dimensions and zero-filled two times in the F1 dimension. 512 and 2k points of linear prediction were added in the F1 dimension for NOESY or TRNOESY and for twodimensional STD spectra, respectively.
Structure Calculation-NMR spectra were analyzed with the FELIX software (Biosym Technologies). P-␤-Cat 17-48 chemical shifts were assigned using standard technique (25). Distance restraints used in the structure calculations were derived from TRNOESY experiments performed with mixing times of 200 ms. As for the free ligand we obtained a NOESY spectrum at 100 ms with very few peaks and only intraresidue and sequential correlations at 200 ms. They were based on TRNOE peak volumes and were classified as strong, medium, weak, or very weak restraints, corresponding to distance restraints of 1.8 -2.7, 1.8 -3.6, 1.8 -5.0, and 1.8 -6.0 Å, respectively. The distance between the two Tyr aromatic protons (2.42 Å) was used as a reference for calibration. The final list of distance restraints containing 455 restraints was incorporated for structure calculation with the standard protocol of ARIA 1.2 (26 -28). This program was used to compute the solution structure starting from an extended structure with random side chain conformation. All NOE cross-peaks from the TRNOESY spectra were used as input to ARIA. By default, ARIA calibrates distance restraints by computing the relaxation matrix from the NOE peak volumes and the chemical shift assignments (29). A rotational correlation time of 8 ns for the peptide bound to the GST-␤-TrCP protein was used for computation of the relaxation matrix. The simulated annealing protocol consisted of four stages: a high temperature torsion angle simulated annealing phase at 10,000 K (30 ps), a first torsion angle dynamics cooling phase from 10,000 K to 50 K (15 ps), a second Cartesian dynamics cooling phase from 2000 K to 50 K (27 ps), and a final minimization phase at 50 K. ARIA runs were performed using the default parameters with eight iterations. Twenty structures were generated each round, and the 10 lowest energy structures were carried to the next iteration. The peptide structures were determined and further refined using molecular dynamics with the PARALLHDG 5.3 parameters file. Modifications to the force field for the phosphorylated residues (pSer) were made with CHARMM (30) and reported for molecular dynamics simulations. The residues, except the N-terminal Asp 17 , Gly 34 , Gly 38 , and Pro 44 , were constrained using 28 dihedral angles constraints between Ϫ20°and Ϫ180°(usually the only range considered in NMR-derived structures) (31), the region of a Ramachandran plot populated by nearly all non-glycyl residues. While ARIA allows the final structures to undergo a procedure in a simulated water box, it was decided that because of the presence of the bound peptide in the hydrophobic recognition sites this step was not applicable. After the final iteration, with ARIA software (OPLS force field) 20 structures were generated, and the 10 best with lower energies were retained as the final structures. The final list of distant restraints contained 455 restraints divided into 213 intra-residue, 150 sequential, 85 medium range, and 7 long range NOEs. Analysis of the ARIA results (Table III) was carried out using the aria.overview script (32). The final PDB structures were analyzed and visualized with Procheck-NMR (33) and MOLMOL (34) programs. The set of P-␤-Cat 17-48 structures was selected for structural analysis according to good geometry and few NOE violations, and by eliminating those structures with , values in the disallowed regions of the Ramachandran plot specified with Procheck.

RESULTS
NMR Assignments for the P-␤-Cat 17-48 /␤-TrCP Protein Complex-Proton chemical shifts and resonance assignments were established using two-dimensional 1 H, 1 H TOCSY, and NOESY experiments and are reported in Table I. Sequential assignments of 1 H resonances were based on characteristic sequential NOE connectivities observed between the ␣-proton of residue i and the amide proton of residue iϩ1, i.e. d ␣N (i,iϩ1), in the NOESY data set. The Arg 18 (residue 2 in the peptide numbering) broad amide signal was not identified for the P-␤-Cat 17-48 peptide. In a similar way, it could not be observed in the study of the free peptide (12). This could be attributed to the fast proton exchange with water. Carbon chemical shifts were assigned using two-dimensional inverse correlation ( 1 H-13 C)phase-sensitive HSQC experiments (Table II). The ability of the nonphosphorylated ␤-Cat 17-48 and of the phosphorylated P-␤-Cat 17-48 peptides to bind to the GST-␤-TrCP protein in solution was then tested by NMR spectroscopy. To identify the structural and conformational features of the binding process, STD experiments and TRNOESY spectra were recorded.
TRNOESY NMR Experiments-To get more information on the bound conformation of the peptide, TRNOESY experiments were recorded for different ratios near 25:1, 50:1, 100:1, 150:1, 300:1, and 600:1. The number and the intensity of the correlations increased when the peptide/GST-␤-TrCP molar ratio reached 150:1 and then decreased at the molar ratio 300:1. Thus, the structural properties of the peptide bound to GST-␤-TrCP were determined using two-dimensional TRNOESY experiments in aqueous solution (H 2 O/D 2 O, 95:5, v/v, pH 7.1) with a molar ratio equal to 150:1. The P-␤-Cat 17-48 peptide gives negative NOEs in the free state. The observation of supplementary TRNOE cross-peaks, when GST-␤-TrCP was added, attested the peptide-protein interaction. Because of the low protein/peptide molar ratio, free and bound peptide molecules were in rapid chemical exchange, and only a single set of broadened ligand resonances was observed. In particular, the addition of GST-␤-TrCP caused a soft line broadening especially observable on the amide protons of residues pSer 33 , His 36 , pSer 37 (belonging to the DpSGXXpS motif) and on the amide and aromatic 4H protons of the Trp 25 . Chemical shift differences relative to the free peptide were also observed, in particular for two different groups of protons ( Fig. 2A): the first one, corresponding to the DpSGXXpS motif included the amide protons of the pSer 33 , Gly 34 , His 36 , pSer 37 , and Gly 38 ; the second one corresponding to the amide protons of His 24 (and also its aromatic protons 2H and 4H), Trp 25 , and Gln 26 . These two effects (line broadening and shifts), which occurred on the protons of the same residues, gave us an indication of the existence of binding.
The perturbations of the chemical shifts of amide protons of P-␤-Cat 17-48 are shown in Fig. 2A. The interaction caused environmental changes on the peptide protein interfaces and hence, affected the chemical shifts of the nuclei in this area. We followed the NH resonances to their bound position and we extracted the binding constant by fitting the fractional shift against an equation depending on total GST-␤-TrCP protein and P-␤-Cat 17-48 concentrations (35,36). Even if GST fusion proteins use to be dimeric in solution, the calculations were done assuming that the protein was present as a monomer. Fast exchange was measured for the interaction of the phosphorylated P-␤-Cat 17-48 peptide to the GST-␤-TrCP protein with a dissociation constant estimated around 500 M.
Significant differences were observed between the NOESY spectra of the P-␤-Cat 17-48 peptide with and without GST-␤-TrCP, indicating that the cross-peaks observed in presence of GST-␤-TrCP are in fact real TRNOEs. These differences between the two experiments were the most visible with short mixing time such as 200 ms. Thus, the TRNOESY spectra confirmed the results of the STD NMR spectra: P-␤-Cat  does have binding activity to GST-␤-TrCP. By comparing build-up rates of the P-␤-Cat 17-48 peptide NOEs with and without GST-␤-TrCP, the mixing time was fixed at 200 ms for the structural calculations, because there was no observable spin diffusion at this mixing time. Furthermore, we could observe much stronger intensities for the TRNOES peaks (up to seven times) than for the NOES peaks of the free peptide. A two-dimensional NOESY NMR experiment of the nonphosphorylated ␤-Cat 17-48 peptide in the presence of the GST-␤-TrCP protein was also performed as a control experiment at the same ratio as that used for the P-␤-Cat 17-48 /GST-␤-TrCP sample, and resulted in the absence of any supplementary peak relative to the free nonphosphorylated peptide. A second negative control was done to ensure that the observed TRNOES represent specific interaction, using P-␤-Cat 17-48 with the fusion protein GST-Nef (22), and led to similar results. STD NMR Experiments-The one-dimensional STD NMR experiments were performed on the P-␤-Cat 17-48 peptide in the presence of GST-␤-TrCP (ca. from 50 -600-fold excess of the ligands). The only signals present were those resulting from the transfer of saturation from bound to free ligand, thus permitting their immediate identification. This technique has the great advantage that it can be combined with any NMR pulse sequence, and in the one-dimensional mode, the method is fast and reliable (37).
Specific affinity was investigated by STD NMR spectroscopy experiments (38) to study the influence of the serine phosphorylation on binding and to describe the region responsible for the binding interaction of the P-␤-Cat 17-48 peptide with GST-␤-TrCP. Without any GST-␤-TrCP protein present, STD spectra did not contain ligand signals, because saturation transfer does not occur without the protein.
The sample containing GST-␤-TrCP protein was the only one to show saturation transfer from the protein to the ligand in the STD spectra. We showed that a 2-s saturation time (40 gaussian pulses, each during 50 ms) was sufficient for efficient transfer of saturation from the protein to the ligand protons. The analysis of the relative STD intensities of P-␤-Cat 17-48 was accomplished by measuring the intensity of the dispersed proton resonances and referencing them to the most intensive proton signals. The one-dimensional NMR spectrum shown in Fig. 3A revealed that a direct analysis of the intensity of individual proton resonances in the aliphatic region is impeded because of severe signal overlap of the corresponding methyl groups. Nevertheless, the spectral region corresponding to the amino protons is well resolved and can be used to classify the amino acid residues relevant for interaction with ␤-TrCP. The signals observed in the one-dimensional spectra for the amide protons of the whole peptide are summarized in Fig. 2B. The strongest signals correspond to protons which can be in intimate contact with the protein. This figure highlights the specificity of the interaction between the peptide and the GST-␤-TrCP protein.
The STD spectrum (Fig. 3A) clearly demonstrated the involvement of the HN of residues such as Leu 31 , Asp 32 , pSer 33 , Ile 35 , His 36 , and pSer 37 , which have similar larger STD intensities, ranging from 70 to 100% (Fig. 2B). The saturation transfer also presents maximum intensity for the aromatic protons of residues such as His 24 , Trp 25 , and Tyr 30 (Fig. 3A). This indicated the proximity of these protons to the protein surface. For the side chains of Gln 26 , Gln 27 , and Gln 28 , the signal of the N⑀H protons have similar large STD intensities. Two-dimensional STD TOCSY NMR spectra were also acquired but resulted in poor additional information; we mainly observed signals of the aromatic protons of Trp 25 , which were best amplified in the one-dimensional STD spectrum.
The same sample was also used for a two-dimensional STD HSQC experiment. Fig. 3B shows a region of the HSQC spectrum and Fig. 3C shows the very same region on the STD HSQC spectrum. Again, we could observe how signals corresponding to residues in a close proximity to the receptor are saturated to a higher degree, resulting in more intense crosspeaks. The analysis of the relative two-dimensional HSQC-STD volumes of P-␤-Cat 17-48 was also accomplished by inte- grating the proton resonances and referencing them to the most intensive proton signals, as we did for the one-dimensional STD analysis. The peaks corresponding to Leu 31 , Ile 35 , and Ala 39 are present in the spectrum Fig. 3C, which means that the aforesaid residues interact with the ␤-TrCP protein.
Conversely, it is interesting to notice, that signals from Ala 21 , Val 22 , and Leu 46 present in the HSQC spectrum (Fig. 3B) have nearly all disappeared in the STD-HSQC spectrum (Fig. 3C) because of their low degree of saturation. Surprisingly, signals of the H␥ of Gln 27 and Gln 28 are weak in the STD-HSQC spectrum, whereas their HN were quite strong in the onedimensional STD spectrum. But to a large extent, the results are in agreement with those of the one-dimensional STD experiment.
To summarize, two groups of residues seem to be important for binding and seem to be closer to the protein surface than the other protons. The first one is the group of residues of the DpSGXXpS motif (without Gly 34 , but including Leu 31 ) and those of the 24 -28 region. This is consistent with the fact that these two regions have shown the most significant chemical shifts ( Fig. 2A).
A one-dimensional STD NMR experiment of the nonphosphorylated ␤-Cat 17-48 peptide in the presence of the GST-␤-TrCP protein was also performed as a reference experiment, at the same ratio as that used for the P-␤-Cat 17-48 /GST-␤-TrCP sample. The addition of the ␤-Cat 17-48 peptide to the GST-␤-TrCP protein resulted in the absence of any broadening of the peaks in the control spectrum. Recording STD NMR experiments of the nonphosphorylated ␤-Cat 17-48 peptide in the presence of the GST-␤-TrCP protein provided a negative control. This control experiment was one experimental way to distinguish here between specific effects of binding peptide to its target and nonspecific interactions between ligand and macromolecular complex. In this case, STD spectra do not contain ligand signals. No effects can be observed as enhancements of the signals of protons, because of the absence of contact between ␤-Cat 17-48 and the ␤-TrCP protein. Thus, saturation transfer does not occur without the Asp-pSer 33 -Gly-XX-pSer 37 motif. This showed that the large number of the enhancements observed for the P-␤-Cat 17-48 peptide in the presence of the GST-␤-TrCP protein was caused by the areas of the ligand actually in contact with the ␤-TrCP binding site. Other negative controls were made to verify the specificity of the interaction. We performed one-dimensional STD NMR experiments with P-␤-Cat 17-48 and other fusion proteins such as GST-Nef (22)  ␤-TrCP protein). STD spectra did not contain ligand signals, hereby showing that the saturation transfer does not occur without the presence of the 7 WD domains.
Three-dimensional Structure Calculations-A summary of sequential d(i, iϩ1) and short range d(i, iϩ2), d(i, iϩ3) and d(i, iϩ4) 1 H-1 H NOE connectivities of the peptide in the bound state is shown in Fig. 4A. A great number of NOEs (455) were observed when the phosphorylated peptide was bound to GST-␤-TrCP, suggesting a rather rigid conformation in the bound state. Multiple TRNOESY connectivities between side chain protons of the aromatic residues (His 24 , Trp 25 , Tyr 30 , and His 36 ) and the main backbone, as well as the side chain protons of other residues, appeared upon addition of GST-␤-TrCP (for instance, between 2H of Trp 25 and 2H of His 24 , or between 2H of His 36 and H␣ of Gly 38 ) suggesting that the aromatic side chains rings are constrained. The numerous NOE connectivities observed for the side chain protons indicate that binding of the peptide by GST-␤-TrCP tends to freeze the rotation of the peptide side chains.
The TRNOE spectrum exhibits a great number of NOEs including intense and medium (i, iϩ1) connectivities suggesting the presence of secondary structures. The presence of mul- tiple ␣N(i, iϩ2), ␣N(i, iϩ3), ␣N(i, iϩ4) (21-25 and 23-27), and ␣␤(i, iϩ3) connectivities attests that the phosphorylated pep-tide adopts a well defined and folded structure in the presence of GST-␤-TrCP. In particular, in region involving residues 21-32, the six ␣N(i, iϩ3) (Fig. 4A) connectivities are indicative of a helical conformation located just before the Asp-pSer 33 -Gly-Ile-His-pSer 37 -Gly motif. In this motif, the three ␣N(i, iϩ2) (32-34, 33-35, and 37-39) and ␣N(i, iϩ3) (32-35) connectivities argue in favor of a bend conformation, while the C-terminal part seems to be less structured. It should be noted that the spatially constrained aromatic rings of His 24 and Trp 25 (i.e. NOEs between 4H of Trp 25 and H␣ of Val 22 , or between 4H of His 24 and H␤* of Ala 21 ) are both incorporated within the helical part of the phosphorylated peptide.
The structures generated using ARIA resulted in TRNOE restraint files consisting of 350 unambiguous and 105 ambiguous distance constraints (Table III). Among these TRNOEs, those that were possibly ambiguous were assigned with the most logical assignment, determined by the sequence. Various runs were performed with ARIA to utilize as many unambiguous and ambiguous restraints as possible from the 500 MHz TRNOESY spectra, and the lowest energy structures were retained. Twenty solution structures were generated using the combined ARIA/CNS protocol described above. The structural statistics for the ten lowest energy structures, generated by ARIA in the final iteration, are shown in Table III. The ensem- ble of the ten lowest energy structures exhibits a backbone r.m.s.d. of 5.6 Ϯ 0.9 Å, but for the 21-36 region, this value decreases to 2.00 Ϯ 0.9 Å (Table III). This is confirmed by the plot of the r.m.s.d. of the backbone dihedral ⌽ and ⌿ angles values for the family of the 10 conformers of the bound structure of P-␤-Cat   (Fig. 4B). This graphic highlights a better definition of residues 20 -37 (except for Gly 34 ). This is consistent with the total number of NOE constraints observed per residue as illustrated in Fig. 4C. The reduced number of interresidue NOEs observed in the C-terminal part of the peptide suggests a less well structured region. Interestingly, the well ordered part of the peptide (20 -37) corresponds to the residues that were shown to be the most involved in the binding surface as evidenced in the STD-NMR experiments.

DISCUSSION
Structure Description-The ensemble of the seven lowest energy structures is shown in Fig. 5A. The structures are superimposed from residues 21-36. The bound peptide showed a helical structure in its N-terminal part for residues 20 -31 and a large bend from residue Asp 32 to residue Gly 38 in the Asp-pSer 33 -Gly-Ile-His-pSer 37 -Gly motif. At low temperatures, the NMR structures of peptides can indicate the important structural features responsible for the ability of a ligand to bind a receptor (39). We observe that the overall shape of the bound structure of ␤-catenin resembles that of the free peptide at 278 K (Fig. 5B). At low temperatures, the NMR structures of peptides can indicate the important structural features responsible for the ability of a ligand to bind with a receptor (39). The N-terminal helical structure is followed by a first turn in the Asp-pSer 33 -Gly motif, the two XX residues, and finally, a second turn due to the pSer 37 . As previously observed (12), the phosphorylation of ␤-catenin strongly modifies the peptide structure, which results in a negative charge surface (Fig. 5C, red), which could interact with the positively charged groups of ␤-TrCP. However, the way the ␤-catenin protein interacts with ␤-TrCP cannot be summed up to a simple electrostatic interaction.
Hydrogen bonds and hydrophobic interaction may play a fundamental role in the stabilization of the complexes with the ␤-TrCP protein. As the ␤-catenin peptide only binds when it is in its phosphorylated state, it seems that hydrophobic interactions can play a secondary role in the stability of the complex. The two amino acids (His 24 and Trp 25 ) create a surface up-stream of the DpSGXXpS motif that could make additional hydrophobic and/orinteractions (Fig. 5C, white).
Epitope Mapping-The STD amplification factor of the amide protons (Fig. 2B) were first classified (Fig. 5D). The strongest ones have been labeled in red, on the TRNOE-derived P-␤-Cat 17-48 -bound structure. Nearly all of them are found in the DpSGXXpS motif where the phosphate groups of pSer 33 and pSer 37 are pointing out of the structure. The Trp 25 and His 36 aromatic amino acids of P-␤-Cat 17-48 make contacts with ␤-TrCP as evidenced by STD-NMR experiments. The STD intensities observed for aromatic protons, which classically possess larger T1 relaxation rates, could be overestimated (18). We have shown in our experiments that even when the relaxation delay was increased up to seven times T1 (the average value of T1 was around 500 ms) signals of Trp 25 and essentially those coming from the aromatic chain were always the strongest we observed in STD experiments. Tryptophan is a rarely found aromatic residue (1.2%) but Trp 25 from ␤-catenin is very well conserved in other organisms. It is present in ␤-catenins from human, mouse, rat, sea urchin, and clawed frog. Another interesting fact is that a mutation on this single amino acid in ␤-catenin has been found to be involved in some human intestinal gastric cancers (40). Trp 25 is also conserved in the protein Armadillo from house fly and in the protein Plakoglobin from human, rat, pig, and zebrafish. Hence the position of Trp 25 , upstream of the DSG motif, could reflect the particular role that tryptophan can play in the interaction of ␤-catenin with ␤-TrCP.
Comparison between the Bound and the Free P-␤-Cat  Structures-The phosphorylated free peptide is characterized by a bend in the motif region and a N-terminal helix (12). In the bound structure of ␤-catenin, the DpSGXXpS motif and Nterminal part are well defined regions characterized by low r.m.s.d. for their ⌽ and ⌿ backbone dihedral angles, and these regions are involved in the binding surface contact of ␤-catenin with the ␤-TrCP protein. As evidenced by STD experiments, the C-terminal Ala 39 -Gly 48 region of the P-␤-Cat 17-48 peptide has only few contacts with ␤-TrCP. These results indicate that the region after the DSGXXS motif is less important for binding. Both in the free and in the bound structures of the P-␤-Cat 17-48 peptide, the C-terminal region (starting from Ala 39 ) appears to have fewer NOEs constraints and a higher r.m.s.d. for ⌽ and ⌿ angles. These could be consistent with a higher mobility of pSer 37 that could contact any accessible positively charged residues spanning the ␤-TrCP surface (Fig. 6A).
A further indication was obtained by observing the possible role of the Trp 25 region. The short helix observed for the residues around Trp 25 was not observable in the crystal structure of ␤-catenin. But this region, just upstream of the DSGXXS motif, is found both in the free and the bound states of ␤-catenin and Vpu peptides. As shown by the intense STD response of some residues in the helix part, this region could be determinant in the stabilization of the complex with the peptide containing the consensus DpSGXXpS motif to the ␤-TrCP protein.
The main structural changes are observed in the motif region of the P-␤-Cat 17-48 peptide, which is modified upon binding, and becomes more compact in the bound state than in the free one as reflected by the C␣-C␣ interatomic distance of the two serines (mean distance around 11 Å for the free structure, and around 8 Å for the bound structure). We also observe different orientations for the side chain of these residues. Hence, in the free P-␤-Cat 17-48 peptide structure, the phosphate groups have a nearly coplanar orientation whereas in the bound structure they have an opposite coplanar orientations closer to that of the crystal structure (Fig. 6A). Interestingly, the different orientation of the phosphorylated residue pSer 33 is the main differ- ence observed between free and bound structures of the ␤-Cat peptide interacting with ␤-TrCP (Fig. 6A). The area downstream of the motif is probably less relevant in the binding with low STD response and fewer TRNOEs constraints.
Comparison between the NMR and the X-Ray Bound Structures-A crystal structure (19) of the complex between ␤-TrCP and ␤-catenin was recently published (30 -40 corresponding residues). As the region upstream of the DSG motif is not present in the crystal structure, these contacts were not ob-servable. We superimposed residues 33-36 of the lowest energy structure of the bound peptide and of the recent crystal structure (Fig. 5E). A good agreement is found between the two structures, essentially for the DpSGX region of the consensus DpSGXXpS motif. In both cases, the side chain of the first phosphoserine pSer 33 points in the same direction, and the phosphate groups are found close together. However, in the consensus DpSGXXpS motif, the XpS region containing the second phosphoserine is found to be different in the two structures. Interestingly, this result highlights the strong implication that the first phosphoserine pSer 33 of ␤-catenin can have with ␤-TrCP. In the crystal structure, the phosphate group of pSer 37 may also interact with many positively charged amino acids. This is not the case for residue pSer 33 , the phosphate group of which can only contact few Arg or Lys residues. Thus, the different positions of the phosphorylated residue pSer 37 represent the main difference observed between all the known bound structures of peptides interacting with ␤-TrCP (Fig. 6B).
We reported our STD results on the surface representation of the top face of the ␤-TrCP WD40 domain with the bound DpSGXXpS motif from the crystal structure of human ␤-TrCP-SkP1 complex bound to a ␤-catenin substrate peptide (Fig. 5F). We decided to look at the smallest interatomic distances found between the ␤-catenin substrate peptide residues and those of the ␤-TrCP WD40 domain in the crystal structure. We measured the distances between the amide protons of P-␤-Cat and the nearest atoms of the ␤-TrCP protein. The shortest distances are 1.9 Å and 2.4 Å for the distances HN pSer 33 -OH Tyr 271 and HN His 36 -OD1 Asn 394 respectively, which correspond to the residues with the strongest STD responses (Fig.  2B). In the case of the second phosphorylated serine, the STD response of which is only in the medium range, we measured a large distance of 5.9 Å between the ligand and the receptor (HN   FIG. 5. TRNOE-derived structures of P-␤-Cat 17-48 . A, twenty structures of P-␤-Cat 17-48 were generated after eight iterations with ARIA software. The best seven are displayed. The molecules are superimposed from residue 21 to residue 36. B, comparison between the lowest energy conformers of the free peptide (down, in yellow) and the bound peptide (top, in gray). C, surface representation of the best conformer of bound P-␤-Cat  . The surface is colored according to the electrostatic potential, calculated with the MOLMOL software. D, amide protons with the strongest STD amplification factor are displayed as colored spheres, in red for the five strongest, and in blue for the eight others. The phosphate groups of pSer 33 and pSer 37 are shown, and point out of the structure. E, superimposition of the lowest energy structure of the bound peptide (in gray) and the 30 -40 corresponding residues obtained in a recent crystal structure (in green) (19). The two peptides are superimposed from residues 33-36. F, surface representation of the top face of the ␤-TrCP WD40 domain with the bound DpSGXXpS motif from the crystal structure of the human ␤-TrCP-SkP1 complex bound to a ␤-catenin substrate peptide (19). The protein surface is colored (red, negative region and blue, positive surface) according to the electrostatic potential calculated with the MOLMOL software. STD results have been added to the crystal structure: the amide protons with the strongest STD amplification factors are displayed as colored spheres (green for the five strongest and light green for the rest).
FIG. 6. A, superimposition of the 32 DpSGXXpS 37 fragment for the P-␤-Cat 17-48 -bound peptide (in gray), the free P-␤-Cat 17-48 peptide (in yellow) (12), and the crystal structure (in green) (19). The three structures are superimposed from residues 33-36. B, superimposition of the 32 DpSGXXpS 37 fragment of the P-␤-Cat 17-48 -bound peptide (in gray), the crystal structure (in green) (19), and the Vpu-bound structure (in purple) (21). The three structures are superimposed from residues 33-36 (in ␤-catenin numbering) and from residues 52-55 (in Vpu numbering). pSer 37 -C␣ Gly 408 ). Thus, the STD results are in keeping with the crystal structure. An excellent agreement is found between the amide protons with the strongest STD amplification factors, and the corresponding shortest distances in the crystal structure. Thus, either by using STD NMR experiments or by inspecting the crystal structure, the DSG motif was found to make the closest contacts with ␤-TrCP.
The bound structure of the P-Vpu peptide, which binds ␤-TrCP and contains the DSGXXS motif, has been recently published (21). We followed it up by superimposing on the DpSGXXpS motif, which is proposed to make close contact with the ␤-TrCP protein, the three known peptide-bound structures i.e. the P-␤-Cat 17-48 -bound peptide, the ␤-catenin fragment of the crystal structure, and the P-Vpu-bound structure (Fig. 6B). Structural similarities were observed for the three structures. Interestingly, the DpSGXX motif structure is conserved with a bend in the DSG domain that exposes the first phosphoserine to make contacts with the ␤-TrCP protein, whereas the second phosphoserine was seen to possess different orientations. These results are in excellent agreement with the fact that the atoms of the peptide detected by STD experiments make close contact with the macromolecular protein receptor and are essential in the binding of the ligand studied so far. Given that the ␤-catenin peptide only binds in its phosphorylated state, it seems that hydrophobic interactions may play a role in the stability of the complex. His 24 , Trp 25 , and Tyr 30 involved in this helical segment define a hydrophobic cluster, which could fit into a hydrophobic binding pocket localized on the ␤-TrCP surface and has been proposed to be located at the interface between the WD1 and WD2 domains.
A recent publication (41) suggests that Vpu is a strong competitive inhibitor of ␤-TrCP that impairs the degradation of SCF-␤-TrCP substrates as long as Vpu has an intact phosphorylation motif and can bind to ␤-TrCP. These results indicate that the ligands of ␤-TrCP have common features for interaction and may even have identical epitopes. CONCLUSION In this study, we demonstrated that the doubly phosphorylated ␤-Cat 17-48 peptide, containing the phosphorylation site DpSGXXpS, efficiently binds to GST-␤-TrCP. When bound to the GST-␤-TrCP protein, the phosphorylated peptide adopts a folded structure, close to those observed in the free peptide; it reflects the local rearrangement of the DpSGXXpS phosphorylation site at the interface with ␤-TrCP. Our results suggest that two regions of the peptide seem to be required for the interaction with ␤-TrCP: the 32 DpSGXXpS 37 phosphorylated motif, as previously suggested (19), and also the helical region located just before the phosphorylated motif. The side chains of the three aromatic residues His 24 , Trp 25 , and Tyr 30 involved in this helical segment define a hydrophobic cluster which could fit into a hydrophobic binding pocket localized on the ␤-TrCP surface. The residues that are the most relevant to binding were identified: Leu 31 , Asp 32 , pSer 33 , Ile 35 , and His 36 in the bend region including the DpSGXXpS motif, and His 24 , Trp 25 , Gln 26 , Gln 27 , and Gln 28 in the helical region located just before. Interestingly, we observed that the residues located in these two regions had upon addition of ␤-TrCP the most important shift of their amide proton resonance, and also had the largest STD responses. This indicates that the binding process significantly modifies the chemical environment of these residues. Finally, our results show that the first phosphorylated serine is reoriented upon binding and must be important for the interaction of peptides containing the consensus motif with ␤-TrCP. The second one is mobile and could accommodate different positively charged residues on the ␤-TrCP binding surface. The study of shorter peptides will elucidate the role of the different region in the interaction. It is clear that the epitope mapping drawn from STD experiments and the bound structure obtained from TRNOESY experiment are in keeping with the crystallographic data.