Structure of the C-terminal Phosphotyrosine Interaction Domain of Fe65L1 Complexed with the Cytoplasmic Tail of Amyloid Precursor Protein Reveals a Novel Peptide Binding Mode*

Fe65L1, a member of the Fe65 family, is an adaptor protein that interacts with the cytoplasmic domain of Alzheimer amyloid precursor protein (APP) through its C-terminal phosphotyrosine interaction/phosphotyrosine binding (PID/PTB) domain. In the present study, the solution structures of the C-terminal PID domain of mouse Fe65L1, alone and in complex with a 32-mer peptide (DAAVTPEERHLSKMQQNGYENPTYKFFEQMQN) derived from the cytoplasmic domain of APP, were determined using NMR spectroscopy. The C-terminal PID domain of Fe65L1 alone exhibits a canonical PID/PTB fold, whereas the complex structure reveals a novel mode of peptide binding. In the complex structure, the NPTY motif forms a type-I β-turn, and the residues immediately N-terminal to the NPTY motif form an antiparallel β-sheet with the β5 strand of the PID domain, the binding mode typically observed in the PID/PTB·peptide complex. On the other hand, the N-terminal region of the peptide forms a 2.5-turn α-helix and interacts extensively with the C-terminal α-helix and the peripheral regions of the PID domain, representing a novel mode of peptide binding that has not been reported previously for the PID/PTB·peptide complex. The indispensability of the N-terminal region of the peptide for the high affinity of the PID-peptide interaction is consistent with NMR titration and isothermal calorimetry data. The extensive binding features of the PID domain of Fe65L1 with the cytoplasmic domain of APP provide a framework for further understanding of the function, trafficking, and processing of APP modulated by adapter proteins.

Alzheimer disease is a neurodegenerative disorder characterized by senile plaques and neurofibrillary tangles. The predominant constituent of the amyloid is the 39 -43-residue amyloid ␤ peptide (A␤), 3 a proteolytic cleavage product of the amyloid precursor protein (APP) (1). APP is an integral transmembrane glycoprotein composed of a large extracellular domain, a single membrane-spanning region, and a short cytoplasmic domain (see Fig. 1A). The APP gene encodes several different isoforms of APP as a consequence of alternative splicing. The three major isoforms are APP 770 , APP 751 , and APP 695 , which all possess the same 47-residue cytoplasmic domain. The functional role of APP remains unclear; however, several lines of evidence suggest that APP is part of a diverse protein-protein interaction network, which is centered on the short cytoplasmic domain. The cytoplasmic domain participates in the important cellular processes of intracellular trafficking and secretion of APP and signal transduction via interactions with adaptor and signaling proteins, respectively. Several proteins reportedly interact with the cytoplasmic domain of APP. These include heterotrimeric G protein G o (2), the 59-kDa ubiquitously expressed protein APP-BP1 (3), the neuron-specific X11 protein (4), Fe65 family proteins (4 -9), mammalian Disabled (Dab) protein (10,11), and c-Jun N-terminal protein kinaseinteracting protein (JIP) (12,13) as well as the microtubulebinding protein PAT1 (14). Some of these proteins are phosphotyrosine interaction/phosphotyrosine binding (PID/PTB) domain-containing proteins, including X11 (4), Fe65 (4 -9), Dab (10,11), and JIP (12,13). They recognize the NPTY sequence within the cytoplasmic domain of APP.
The Fe65 family proteins, Fe65, Fe65L1, and Fe65L2, are adaptor proteins that possess three protein-protein interaction domains: one WW domain and two PID/PTB domains (see Fig.  1B) (15). The N-and C-terminal PID/PTB domains are referred to as PID1 and PID2, respectively. The three protein-protein interaction domains are well conserved among the Fe65 family proteins, sharing 50 -60% amino acid sequence identity, whereas most of the remaining parts of the proteins are unrelated. Among the three Fe65 family proteins, the most significant difference is their tissue distribution: Fe65 mRNA is neuron-specific (16), whereas Fe65L1 mRNA is ubiquitously expressed (7), and Fe65L2 mRNA significantly accumulates in the brain and testis (9). The Fe65 gene was originally isolated as a neuron-specific gene, and it has some characteristics of a transcription factor (16,17). The interaction between the Fe65 family proteins and the APP cytoplasmic domain has been confirmed both in vitro (5,7) and in vivo (8). For all three Fe65 family proteins, the C-terminal PID/PTB domain (PID2) was demonstrated to be sufficient for their binding to the cytoplasmic domain of APP (4). Furthermore phosphorylation of the tyrosine in the NPTY motif is not required (4). On the other hand, a yeast two-hybrid screening study (5) and a peptide competition experiment (8) showed that a 32-residue-long peptide, DAAVTPEERHLSKMQQNGYENPTYKFFEQMQN, located at the extreme C terminus of the cytoplasmic domain of APP, was necessary for binding PID2 of Fe65. The 32-residue-long peptide (termed APP-32mer in Fig. 1A), including the NPTY motif, is much longer than the peptide that was reported previously to be recognized by the PID/PTB domain. The YENPTY sequence is also a sorting motif, or an internalization motif, required for trafficking of APP into the endocytic pathway (18). Studies have revealed that Fe65 family proteins can alter the processing of APP by influencing APP trafficking (19 -21). Fe65 has been shown to increase ␣-secretase-cleaved APP and A␤ production (19). Fe65L1 also promotes ␣-secretase-cleaved APP secretion and APP maturation (20). These regulatory activities of Fe65L1 require the binding of Fe65L1 to APP C-terminal fragments (21).
Although the interaction between APP and the Fe65 family proteins is highly significant, in terms of the biology of Alzheimer disease, little is known about the molecular basis of the interaction. Here we report the solution structure of PID2 of mouse Fe65L1 in the free form and in complex with APP-32mer and the characterization of their interaction by isothermal titration calorimetry (ITC) and NMR spectroscopy. Among the three Fe65 family proteins, we selected PID2 of Fe65L1 as a target because the mRNA encoding Fe65L1 is ubiquitously expressed. To facilitate the structure determination of the complex of PID2 of Fe65L1 and APP-32mer, we designed several different chimeric proteins in which the corresponding regions of PID2 and APP-32mer were integrated. The complex structure reported here reveals a novel peptide binding mode as compared with those of the canonical PID domains that recognize the NPX(p)Y motif ((p) indicates that tyrosine may be phosphorylated).

EXPERIMENTAL PROCEDURES
Expression and Purification of PID2 of Fe65L1-The DNA fragment encoding PID2 of mouse Fe65L1 (amino acid residues Pro 582 to Cys 704 ; Swiss-Prot accession number Q9DBR4) was amplified via PCR from the RIKEN full-length enriched mouse cDNA library (Clone ID 1200015I07) (22,23) and was cloned into the plasmid vector pCR2.1 (Invitrogen) as a fusion with an N-terminal His tag and a tobacco etch virus protease cleavage site. The 13 C, 15 N-labeled protein was synthesized by the cellfree protein expression system (24 -26). The cell-free reaction solution was first absorbed to a TALON affinity column, which was washed with 20 mM Tris-HCl buffer (pH 8.0) containing 1 M NaCl, and was eluted with 20 mM Tris-HCl buffer (pH 8.0) containing 300 mM NaCl and 500 mM imidazole. The solution was then desalted on a HiPrep 26/10 Desalting column with 20 mM Tris-HCl buffer (pH 8.0) containing 300 mM NaCl. The His tag was cleaved by an incubation at 30°C for 1 h with tobacco etch virus protease. The sample was then loaded on a TALON affinity column, which was washed with 20 mM Tris-HCl buffer (pH 8.0) containing 300 mM NaCl, and was eluted with 20 mM Tris-HCl buffer (pH 8.0) containing 300 mM NaCl and 500 mM imidazole. The flow-through fraction was desalted on a HiPrep 26/10 Desalting column with 20 mM Tris-HCl buffer (pH 8.0) containing 1 mM EDTA. The sample was then loaded on a HiTrap Q column, which was washed with 20 mM Tris-HCl buffer (pH 8.0) containing 1 mM EDTA, and was eluted with a concentration gradient of 20 mM Tris-HCl buffer (pH 8.0) containing 1 mM EDTA and 1 M NaCl. Finally the sample was purified on a HiLoad 16/60 Superdex75 column, which was washed with 20 mM Tris-HCl buffer (pH 7.0) containing 200 mM NaCl, 1 mM EDTA, and 1 mM DTT. The fraction containing the PID domain was concentrated to 1.49 mg/ml in 20 mM Tris-HCl buffer (pH 7.0) containing 200 mM NaCl, 1 mM EDTA, 1 mM DTT, and protease inhibitor mixture (Complete, EDTA-free (Roche Applied Science)). For the NMR structure determination, a 0.84 mM sample of free PID2 was prepared in 20 mM deuterated Tris buffer (pH 7.0) containing 100 mM NaCl, 2 mM deuterated DTT, 0.02% NaN 3 , and 10% 2 H 2 O, 90% 1 H 2 O. The protein sample for the NMR measurements consisted of 136 amino acid residues. The first 7 amino acid residues at the N terminus (GSSGSSG) and the last 6 residues at the C terminus (SGPSSG) were derived from the linker sequence used in the expression and purification system. NMR Sample Preparation of the PID2 and APP-32mer Complex-To prepare the sample of the PID2 and APP-32mer complex for NMR measurement, a solution of non-labeled APP-32mer (Toray Research Center, Tokyo, Japan), dissolved in buffer A (20 mM deuterated Tris buffer (pH 7.0) containing 100 mM NaCl, 1 mM deuterated DTT, 0.02% NaN 3 , and 10% 2 H 2 O, 90% 1 H 2 O) with pH adjustment, was gradually added to 13 C, 15 N-labeled PID2, which had been buffer-exchanged in buffer A, until the free form of PID2 was completely converted into the peptide-bound form as confirmed by the 1 H-15 N HSQC spectra. The final concentration of PID2 was 0.37 mM.
Design and Expression of the PID2-APP-32mer Chimera-To prepare the PID2-APP-32mer chimera, the gene encoding PID2 was fused to the cDNA of APP-32mer (amino acid residues Asp 739 * 4 to Asn 770 *; Swiss-Prot accession number P12023), and the resulting gene was inserted within the plasmid vector pCR2.1 (Invitrogen), incorporating an N-terminal His tag and a tobacco etch virus cleavage site. The 13 C, 15 N-labeled chimera was synthesized by the cell-free protein expression system (24 -26). Three kinds of chimeras, differing in the relative positions of PID2 and APP-32mer in the protein, were designed for the structure determination (Fig. 1C). The three chimeras were designated as I, II, and III, respectively. The construct of Chimera I included, from the N to the C terminus, a His tag, APP-32mer, a 14-residue linker (SGSSGSSGSSGSSG), and PID2. The Chimera II construct consisted of a His tag, APP-32mer, a 23-residue linker containing a protease Factor Xa cleavage site (SGPSS-GIEGRGSSGSSGSSGSSG), and PID2; the construct of Chimera III included a His tag, PID2, the 23-residue linker, and APP-32mer. The 13 C, 15 N-labeled chimeras were synthesized and purified as described below. After the His tag cleavage in the purification procedure, Chimera I consisted of 176 amino acid residues; both Chimeras II and III consisted of 185 amino acid residues. The first 7 amino acid residues at the N terminus (GSSGSSG) in each sample were derived from the linker sequence used in the expression and purification system.
Purification of the PID2-APP-32mer Chimeras-The cell-free reaction solution of each chimera was adsorbed onto a HiTrap chelating column (Amersham Biosciences), which was washed with 20 mM Tris-HCl buffer (pH 8.0) containing 1 M NaCl and 20 mM imidazole, and was eluted with 20 mM Tris-HCl buffer (pH 8.0) containing 500 mM NaCl and 500 mM imidazole. To cleave the His tag, the eluted sample was incubated with tobacco etch virus protease at 30°C for 3 h for Chimeras I and II and overnight for Chimera III. The sample was then loaded on a HiPrep 26/10 Desalting column, which was eluted with 20 mM Tris-HCl buffer (pH 8.0) containing 300 mM NaCl and 20 mM imidazole. The fractionated dialysate was applied to a HiTrap chelating column, which was equilibrated with 20 mM Tris-HCl buffer (pH 8.0) containing 300 mM NaCl and 20 mM imidazole, and was eluted with a concentration gradient of imidazole (500 mM final concentration). The eluted fraction of each chimera sample was concentrated in 20 mM Tris-HCl buffer (pH 8.0) containing 300 mM NaCl, 1 mM DTT, 0.5 mM EDTA, and protease inhibitor mixture (Complete Mini). Each NMR sample contained ϳ1.0 mM of uniformly 13 C, 15 13 15 C, N-labeled 13 15 Non-labeled , is shown here. The 32-residue-long peptide derived from the cytoplasmic domain of APP (APP-32mer) used in the structure determination and interaction study is shown in magenta. The 14-residue-long peptide (APP-14mer) used in the comparative interaction study is indicated above the amino acid sequence. The NPTY motif is highlighted in cyan. B, domain structures of the three Fe65 family proteins from mouse. The full length of each protein and the domain boundary defined by the Simple Modular Architecture Research Tool are indicated. The actual region of PID2 whose structure was determined in this study is indicated above the domain structure of Fe65L1. The Swiss-Prot accession codes of the amino acid sequences of Fe65L1, Fe65, and Fe65L2 are Q9DBR4, Q9QXJ1, and Q8R1C9, respectively. C, domain structures of the chimeras and the complex samples used in this study. APP-32mer, PID2 of Fe65L1, and the linkers in between are shown in magenta, black, and yellow, respectively. Linker_S stands for the 14-residue linker (SGSSGSSGSSGSSG); Linker_L stands for the 23-residue linker (SGPSSGIEGRGSSGSSGSSGSSG). N and C represent the N-and C-terminal tags derived from the expression vector (N, GSSGSSG; and C, SGPSSG), respectively. res., residues. ). HEPES buffer (20 mM HEPES buffer (pH 7.0) containing 100 mM NaCl and 1 mM DTT) was used for the ITC measurements. PID2 was buffer-exchanged in the ITC buffer, and APP-32mer was dissolved in the same buffer with pH adjustment. The concentrations of PID2 and APP-32mer used for the ITC measurement were 0.109 and 1.640 mM, respectively, as determined using UV absorbance values measured at 280 nm. A degassed sample of PID2 was kept at room temperature (25°C) and was stirred at 307 rpm in a 1.4-ml reaction cell. For each titration, 5-l aliquots of peptide were delivered into the PID2 solution at 240-s intervals to allow complete equilibration. The final ratio of peptide to protein reached 2:1 at the end of the titration. Heat transfer was measured as a function of elapsed time. The heat of dilution, obtained by titrating the identical peptide solution into the reaction cell containing only the HEPES buffer, was subtracted prior to analysis. The corrected titration curve was fitted with a one-site model, and the thermodynamic parameters were calculated using the Origin software (version 7.0) provided by MicroCal.
Titration Experiments by NMR-NMR titration experiments were carried out by adding the unlabeled APP-32mer solution to 13 C, 15 N-labeled PID2. In total, four NMR samples were prepared for the titration experiment. In the four NMR samples, the concentration of PID2 was maintained at 0.050 mM, and the concentration of APP-32mer was varied to generate a series of different PID2:APP-32mer molar ratios (1:0, 1:0.5, 1:1, and 1:1.5). The buffer used for the APP-32mer titration was the same as that used for the structure determination. 1 H-15 N HSQC spectra were measured after each titration step, and the signals were monitored by observing the changes in the chemical shifts of the amide signals in 1 H-15 N HSQC spectra. The weighted chemical shift change (in ppm units) of the amide proton (⌬␦ HN ) and nitrogen (⌬␦ N ) was calculated according to the following equation: where W HN ϭ 1 and W N ϭ 0.154 (27).
NMR Spectroscopy-The NMR data for PID2 in the free form were recorded at 298 K on Varian Inova 600-and 900-MHz spectrometers equipped with pulsed field gradient probes. The NMR data for the three chimeras, Chimera I, Chimera II, and Chimera III, and PID2 complexed with non-labeled APP-32mer were recorded at 296 K on a Bruker AVANCE 700-MHz spectrometer, which was equipped with a triple resonance cryoprobe, and on AVANCE 800-and 900-MHz spectrometers. Sequence-specific resonance assignments were made using the standard triple resonance techniques (28). For PID2 complexed with the non-labeled APP-32mer, two-dimensional [F2] 13 C, 15 N-filtered total correlation spectroscopy and twodimensional [F1,F2] 13 C, 15 N-filtered NOESY (29) were measured for the chemical shift assignments of the non-labeled APP-32mer, and two-dimensional [F2] 13 C, 15 N-filtered NOESY (29), three-dimensional 13 C, 15 N F1-filtered, 13 C F3-edited NOESY and three-dimensional 13 C, 15 N F1-filtered, 15 N F3-edited NOESY (30) with mixing times of 80, 100, and 100 ms, respectively, were recorded for the detection of intermolec-ular NOEs. The filter-related spectra were measured on an 800-MHz spectrometer. All of the spectra were processed using the NMRPipe software package (31). The programs KUJIRA (32) and NMRView (33,34) were used for visualization of the NMR spectra and chemical shift assignments.
Structure Calculations-For the structure calculations of PID2 in the free form, Chimera I, Chimera II, and Chimera III, 15 N-edited NOESY and 13 C-edited NOESY with 80-ms mixing times were used to determine the distance restraints. Dihedral angle restraints were derived using the program TALOS (35). The stereospecific assignments of the Val and Leu methyl groups were determined when they were distinguishable from their NOE patterns. Automated NOE cross-peak assignments and structure calculations with torsion angle dynamics were performed using the program CYANA (36 -39).
For the structure calculation of PID2 in complex with the non-labeled APP-32mer, the stereospecific assignments and the dihedral angle restraints for the PID2 part were added in a way similar to that described above. The dihedral angle restraints for APP-32mer were only added in its secondary structural region by checking the corresponding NOE patterns. In addition, a total of five NOE peak lists were used in the complex structure calculation. Among the five peak lists, two were obtained from three-dimensional 13 C-edited NOESY-HSQC and 15 N-edited NOESY-HSQC, and three were obtained from two-dimensional [F2] 13 C, 15 N-filtered NOESY, three-dimensional 13 C, 15 N F1-filtered, 13 C F3-edited NOESY, and threedimensional 13 C, 15 N F1-filtered, 15 N F3-edited NOESY. Dihedral angle restraints and stereospecific assignments of the Val and Leu methyl groups were obtained in a similar manner as described above. The peak list from two-dimensional [F2] 13 C, 15 N-filtered NOESY provided information about the intramolecular NOEs of APP-32mer and the intermolecular NOEs between PID2 and APP-32mer, whereas the peak lists from three-dimensional 13 C, 15 N F1-filtered, 13 C F3-edited NOESY and three-dimensional 13 C, 15 N F1-filtered, 15 N F3-edited NOESY provided information about the intermolecular NOEs between PID2 and APP-32mer. The cross-peaks along the F2 axis in two-dimensional [F2] 13 C, 15 N-filtered NOESY and those along the F1 axes in three-dimensional 13 C, 15 N F1-filtered, 13 C F3-edited NOESY and three-dimensional 13 C, 15 N F1-filtered, 15 N F3-edited NOESY are the signals coming only from the non-labeled APP-32mer. Therefore, in principle, the information about the APP-32mer assignment can be integrated into the three filter-related peak lists in advance of the structure calculation. In the CYANA peak list, each line contains the following information: the peak number, the chemical shifts, the peak volume, the atom numbers that identify the atoms in the corresponding chemical shift list, and so on. The default value for the atom number in the peak list is zero, indicating a missing assignment. CYANA makes the assignment automatically for each peak according to its structure calculation algorithms. If the assignment of a certain atom is known, then an atom number, instead of zero, can be given for that atom. CYANA then starts the structure calculation using the assignments of these atoms as inputs. In the filterrelated peak list, one column of the atom number is known to correspond to that of APP-32mer; therefore, instead of zero, the atom number of APP-32mer is used. In this way, the assignment of that column will be restricted only to APP-32mer. The atom number of APP-32mer was given according to the chemical shift assignment of that atom and the chemical shift tolerance, which was set to the same value as that used in the CYANA calculation. In the end, automated NOE cross-peak assignments and structure calculations with torsion angle dynamics were performed using the program CYANA (36 -39).
In the structure calculations described above, a total of 100 structures were calculated, and the 20 structures in the final calculation cycle with the lowest target function values were selected. The stereochemical quality of the structures was assessed using the program PROCHECK-NMR (40). Molecular models were generated using the programs MOLMOL (41) and PyMOL (42). The statistics of the structures, as well as the distance and torsion angle constraints used for the structure calculation, are summarized in Table 1.
Protein Data Bank Accession Numbers-The structures of PID2 in the free form and in the complex form and Chimera I, Chimera II, and Chimera III have been deposited in the Protein Data Bank with the accession codes 1WGU, 2ROZ, 2YT0, 2YT1, and 2YSZ, respectively.

Structure Determination of PID2 of Mouse Fe65L1 in Its Free
Form-The domain boundary of PID2 of mouse Fe65L1 is defined as residues Glu 587 -Pro 717 according to the Simple Modular Architecture Research Tool. Before the structure determination, several constructs with an elongated or short-ened N or C terminus were prepared to obtain an optimized sample (data not shown). As a result, the construct encompassing residues Pro 582 -Cys 704 was obtained as a suitable sample with sharp and well dispersed signals in the 1 H-15 N HSQC spectrum. The NMR structure determination was performed for 13 C, 15 N-labeled PID2, encompassing residues Pro 582 -Cys 704 . The ensemble consisting of 20 structures with the fewest violations is shown in the wire model in Fig. 2A, and the ribbon representation is shown in Fig. 2B. As shown in Fig. 2, PID2 of Fe65L1 consists of seven ␤-strands (␤1, Gln 590 -Pro 601 ; ␤2, Ser 629 -Ala 635 ; ␤3, Thr 638 -Ile 642 ; ␤4, Val 650 -Arg 655 ; ␤5, Leu 659 -Gly 663 ; ␤6, Phe 671 -Asp 676 ; and ␤7, Phe 682 -Trp 688 ) and two ␣-helices (␣1, Met 608 -Ser 621 ; and ␣2, Ala 694 -Cys 704 ). The seven ␤-strands are folded into two antiparallel ␤-sheets that are orthogonally arranged to form a ␤-barrel structure, exhibiting a canonical PID/PTB fold. Strands ␤2-␤4 form the first ␤-sheet, and ␤5-␤7 form the second ␤-sheet, whereas the ␤1 strand is part of both the first and second ␤-sheets. Helix ␣1 is located between strands ␤1 and ␤2, and helix ␣2 is at the C terminus.
The Binding of PID2 of Fe65L1 to the Cytoplasmic Domain of APP-The binding properties between PID2 of Fe65L1 and the cytoplasmic domain of APP were investigated by ITC and NMR spectroscopy. Two peptide fragments with different lengths derived from the cytoplasmic domain of APP (Fig. 1A), a 14-residue peptide APP-14mer (sequence, QNGYENPTYKFFEQ) and a 32-residue peptide APP-32mer (sequence, DAAVT-PEERHLSKMQQNGYENPTYKFFEQMQN), were used in the  and APP-32mer. Using APP-14mer at a concentration similar to that of APP-32mer as the titrant, the exothermic nature of the protein-peptide binding reaction was also observed from the titration curve, but it was quite weak (data not shown). It was difficult to obtain the thermodynamic parameters, including K D , from the curve fitting because the titration curve did not reach complete saturation under these conditions. Therefore, it can be inferred that the binding affinity between PID2 and APP-14mer is much weaker than that between PID2 and APP-32mer.
The peptide binding properties and the binding site were further investigated by NMR using a chemical shift perturbation analysis. A 13 C, 15 N-labeled PID2 sample was titrated with non-labeled APP-32mer and APP-14mer, and the residues showing chemical shift changes were monitored on 1 H-15 N HSQC spectra. The addition of APP-32mer to PID2 resulted in peaks disappearing and reappearing at many positions in the spectra (Fig. 3B). This observation of slow exchange on the NMR time scale is indicative of high affinity binding. In the case of APP-14mer, fewer peaks disappeared and reappeared, and at the same time, peak broadening was also observed (data not shown  Fig. 3C, the weighted chemical shift changes of the amide 1 H and 15 N, obtained from NMR titration experiments with APP-32mer, are plotted as a function of the residue number. The residues with large weighted chemical shift changes are mapped onto the ribbon drawing of PID2 (Fig. 3D). These residues mainly reside on helix ␣2, strand ␤5, and the ␤1-␣1, ␤4 -␤5, and ␤6 -␤7 loops.
Overall Structure of PID2 of Fe65L1 in Complex with APP-32mer-Based on its high binding affinity for PID2, APP-32mer was chosen for the complex structure study. To prepare the sample for the complex structure determination, non-labeled APP-32mer was added to 13 C, 15 N-labeled PID2. To facilitate the chemical shift and NOE assignments of APP-32mer, three kinds of chimeras containing the corresponding region of PID2 and APP-32mer were also designed as shown in Fig. 1C. Each chimera with amino acid residues that were uniformly 13 C, 15 N-labeled behaved like a single chain protein; therefore, the structure determination of the chimeric protein was performed in a conventional manner. With the help of the chemical shift assignment of the APP-32mer portion in the chimeras, the assignment of the non-labeled APP-32mer in the complex sample was readily achieved. Most of the chemical shifts of the protons in APP-32mer were assigned except for the protons from the 2 residues at the N terminus (Asp 739 * and Ala 740 *) and the 2 residues at the C terminus (Gln 769 * and Asn 770 *). The structures of the three chimeras are shown in Fig. 4, A-C, and the complex structure of PID2 with APP-32mer is shown in Fig.  4, D-F. The overall structures of the three chimeras and the complex are very similar. In all of the structures, the N-terminal part of APP-32mer (Pro 744 *-Gln 753 *) forms a 2.5-turn ␣-helix, the middle region of APP-32mer (Tyr 757 *-Glu 758 *) forms an antiparallel ␤-sheet with strand ␤5 of PID2, and the region Asn 759 *-Tyr 762 * of APP-32mer forms a type-I ␤-turn.
Interaction of PID2 of Fe65L1 with APP-32mer-APP-32mer interacts with PID2 of Fe65L1 using its residues ranging from Glu 745 * to Tyr 762 *. A detailed view of the intermolecular contacts is given in Fig. 5, and a schematic drawing of the intermo- lecular contacts is shown in the left panel of Fig. 6B. The NPTY motif, Asn 759 * to Tyr 762 *, forms a type-I ␤-turn (Fig. 5B). The amide group of Asn 759 * forms a hydrogen bond with the main chain carbonyl group of Ser 660 , whereas the side chain of Asn 759 * forms two hydrogen bonds with the main chains of Val 656 and Leu 659 (Fig. 6B). Tyr 762 *, whose aromatic ring stacks with the C␤ portion of Ser 660 , interacts with Pro 605 , Phe 658 , and Asp 676 (Fig. 5B). Tyr 757 * and Glu 758 *, which are located immediately N-terminal to the NPTY motif, form an antiparallel ␤-sheet with Phe 661 and Met 662 on the ␤5 strand of PID2. In addition, the side chain of Tyr 757 * is surrounded by Met 662 , Gln 701 , and Cys 704 of PID2, whereas the hydrophobic part of the Glu 758 * side chain interacts with Val 606 (Fig. 5B). With regard to Gly 756 *, intermolecular NOEs with Phe 661 and Met 608 of PID2 are observed. In the N-terminal region of APP-32mer, the region from Pro 744 *-Gln 753 * forms a 2.5-turn ␣-helix and interacts with PID2 mainly via Leu 749 * and Met 752 * (Fig. 5C). The main intermolecular contacts involving Leu 749 * and Met 752 * are hydrophobic. Leu 749 * binds at a pocket lined with the hydrophobic residues Val 668 , Ala 693 , and Ala 694 and the   (Fig. 5C). Intermolecular interactions were also observed for other residues in the N-terminal region of APP-32mer. Glu 745 * interacts with Ala 694 , and His 748 * interacts with Glu 698 and Gln 701 . The hydrophobic region of the Gln 753 * side chain interacts extensively with the hydrophobic residues of Val 664 , Val 668 , and Met 608 of PID2 (Fig. 5C). Moreover the side chain of Gln 753 * forms two hydrogen bonds with the main chains of Gly 665 and Lys 666 . Consequently the molecular interface covered by APP-32mer includes ␤5, ␣2, and the ␤1-␣1, ␤4 -␤5, ␤5-␤6, and ␤7-␣2 loops of PID2.
After examination of the intermolecular interface, a structure comparison was made between PID2 in the free and complex forms. The fact that the structured regions of PID2 in the free form fit well with those in the complex and chimeric forms with root mean square deviation values of 0.8 and 1.4 Å for the backbone and heavy atoms, respectively, indicates that the structural changes of PID2 induced by APP-32mer binding were negligible.
Several results indicated that Thr 743 * of APP has a regulatory role in the interaction with Fe65 as its phosphorylation prevents the interaction (43). On the basis of the present structural data, it still remains unclear how the phosphorylated Thr 743 * prevents the interaction with Fe65, and further studies will be necessary in the future.

DISCUSSION
Fe65L1, an adaptor protein, interacts with the cytoplasmic domain of APP (6, 7). Here we performed binding assays with PID2 of Fe65L1 and the APP peptide using ITC and NMR chemical shift perturbation and then solved the structures of PID2 of Fe65L1 in the ligand-free form and in complex with APP-32mer. In addition, we described three structures of PID2-APP-32mer chimeras. These structures provided the first view of a PID/PTB domain belonging to the Fe65 family. In the complex structure, the extensive binding features, characterized by the involvement of the N-terminal region of APP-32mer, revealed a novel mode of peptide binding  A, B, and C, PID2, APP-32mer, and the linkers are shown in gray, magenta, and yellow, respectively. D, side-by-side stereoview of the wire models of the 20 conformers of the PID2 and APP-32mer complex with the fewest violations with PID2 in black and APP-32mer in magenta. E, ribbon representation of the PID2 and APP-32mer complex with PID2 in green and APP-32mer in magenta. F, surface presentation of PID2 in complex with APP-32mer shown by a stick model. In the surface presentation of PID2, the hydrophobic residues (Ala, Val, Leu, Ile, Met, Phe, and Trp) are mapped in green, positively charged residues (Arg and Lys) are blue, and negatively charged residues (Asp and Glu) are red. Other residues are mapped in gray. that has not been observed in PID/ PTB⅐peptide complexes. The binding site of APP-32mer, identified from the complex structure study, covers the ␤5 strand, the ␣2 helix, and the ␤1-␣1, ␤4 -␤5, ␤5-␤6, and ␤7-␣2 loops of PID2.
Besides the Fe65 family proteins, some other PID/PTB domain-containing proteins also reportedly bind the cytoplasmic domain of APP, including X11, Dab, and JIP family proteins. The structures of the PID/PTB domains of X11 and Disabled proteins (Dab1 and Dab2) in complex with a short APP peptide have been solved, and the mode of protein-peptide recognition has been investigated (44,45). In the following, a comparison of these PID/ PTB⅐peptide complex structures, especially the protein-peptide binding features, will be discussed.
Comparison with the X11 PID/ PTB⅐APP Peptide Complex-Biochemical characterization of the interaction between X11 and APP indicated that a 14-residue peptide with the sequence encompassing the NPTY motif of APP (QNGY-ENPTYKFFEQ, residues 754*-767*) competes efficiently with the full-length APP in binding with the PID/PTB domain of X11 (4). The surface plasmon resonance binding assay gave a K D value of 0.32 M for the PID/PTB domain of X11 and the 14-residue APP peptide (44). However, a much longer peptide sequence (residues 739*-770*) of APP is required for high affinity binding to PID2 of Fe65L1. To clarify the different modes of APP peptide recognition by the PID/PTB domains of Fe65L1 and X11, the structures of the two complexes were compared (Fig. 6A).
A schematic drawing of the protein-peptide interaction network is shown in Fig. 6B. In both complexes, the NPTY motif in the peptide forms a type-I ␤-turn, and the region N-terminal to the NPTY motif forms an antiparallel ␤-sheet with the ␤5 strand of each PID/PTB domain. However, Pro 760 *, Thr 761 *, Lys 763 *, Phe 764 *, and Phe 765 * from the APP peptide interact further

Novel Peptide Binding Mode by PID2 of Fe65L1
with the C-terminal ␣-helix of the X11 PID/PTB domain; this prolonged region of the helix was omitted from the Fe65L1 PID2 construct to obtain a feasible NMR sample. It may be argued that the shorter length of the C-terminal ␣-helix in the Fe65L1 PID2 construct is the reason for the fewer protein-peptide contacts and hence the lower binding affinity. However, this is probably not the critical reason. In the peptide competitive binding study performed by Borg et al. (4), the 14-residue peptide was able to compete with X11 for APP binding with submicromolar affinity, but it had no effect on Fe65 binding even though their Fe65 PID2 construct was 10 residues longer at the C terminus than our Fe65L1 PID2 construct reported here. Moreover in their site-directed mutagenesis study (4), three of the APP mutations were at positions in/around the NPTY motif, corresponding to Tyr 757 *, Asn 759 *, and Tyr 762 * of APP-32mer. The Y757*G mutation severely impaired the binding of both X11 and Fe65, whereas the Y762*A mutation had almost no effect. In contrast, the N759*A mutation exhibited a differential effect; it abolished the X11 binding but did not affect the Fe65 binding. Therefore, we closely examined the protein-peptide interaction network involving the 3 residues (Fig. 6B). Consistent with the mutagenesis results (4), Tyr 757 * forms two main chain-main chain hydrogen bonds with its side chain involved in many van der Waals interactions, whereas Tyr 762 * only makes weaker van der Waals interactions in both PID/PTB⅐peptide complexes. On the other hand, Asn 759 * forms one main chain-main chain hydrogen bond and two side chainmain chain hydrogen bonds with both PID/PTB domains and also makes van der Waals interactions with Cys 704 in Fe65L1 PID2 and Phe 479 in the X11 PID/PTB domain, respectively. Both Cys 704 of Fe65L1 and Phe 479 of X11 are buried in the hydrophobic core of the PID/PTB domains. As Phe 479 is more bulky than Cys 704 , the van der Waals interaction between Asn 759 * and Phe 479 of X11 was evaluated as being much stronger than that between Asn 759 * and Cys 704 of Fe65L1 PID2 by LIGPLOT (46). Therefore, the differential effect of the N759*A mutation probably resulted from this variation in the van der Waals interaction. Comparison with the Dab2 PID/PTB⅐APP Peptide Complex-The structures of the binary complex of the Dab2 PID/ PTB domain with a 9-residue peptide (NGYENPTYK) derived from residues 755*-763* in the cytoplasmic domain of APP and the ternary complex of the Dab1 PID/PTB domain with the same 9-residue peptide and inositol 1,4,5-trisphosphate have been reported (45). The binding assay using fluorescence polarization for the Dab1 PID/PTB domain and a 17-residue APP peptide (acetyl-QNGYENPTYKAFFEQGK) gave a K D value of 0.55 M (11), although the K D value for the complex between the Dab1/Dab2 PID/PTB domain and the 9-residue APP peptide has not been reported. In the two complex structures, the PID/PTB domains of Dab1 and Dab2 bind the 9-residue peptide in a similar fashion; therefore, the binary complex of the Dab2 PID/PTB domain with the 9-residue peptide was used in the structure comparison (Fig. 6A). The accessible surface areas buried upon binding were compared for the three complexes. The buried surface areas are on a similar order, ϳ1800, ϳ2000, and ϳ1500 Å 2 for the Fe65L1 PID2⅐APP-32mer, X11 PID/ PTB⅐APP peptide (14 residues), and Dab2 PID/PTB⅐APP peptide (9 residues) complexes, respectively. Previous studies reported that the K D values of the binding of the three PID/PTB domains with the APP peptides are on a similar order, ranging from 0.32 to 0.79 M.
A schematic drawing of the protein-peptide interaction network of the Dab2 PID/PTB complex is shown in the right panel of Fig. 6B. The van der Waals interaction between Asn 759 * and Phe 166 in Dab2, which is similar to that between Asn 759 * and Phe 479 in the X11 PID/PTB domain, was also evaluated as being stronger than that between Asn 759 * and Cys 704 in PID2 of Fe65L1. Therefore, Asn 759 * is again important for the stronger binding to the Dab2 PID/PTB domain. In addition, the hydroxyl group of Tyr 762 * in the 9-residue peptide forms a side chain-main chain hydrogen bond with Gly 139 of Dab2 that may also contribute somewhat to the stronger binding. In PID2 of Fe65L1, the loop between ␤6 and ␤7 is 1 residue longer than the corresponding region of the Dab2 PID/PTB domain where Gly 139 is located (Fig. 6C), providing a space wide enough to accommodate a phosphorylated tyrosine. In this sense, PID2 of Fe65L1 belongs to the kind of PID/PTB domain in which the NPXY motif binding is independent of the tyrosine phosphorylation state, like the X11 PID/PTB domain.
The crucial role of Cys 704 in the interaction with APP was demonstrated by the functional impairment of the Cys to Phe mutation (4). It is interesting to discuss the consequences of the Cys to Phe transition, considering that in X11 and Dab2, this residue is Phe (Phe 479 and Phe 166 ). In the structure of the free Fe65L1 PID2, Cys 704 fits snugly into the hydrophobic pocket formed by Val 656 , Leu 659 , and Met 662 , and thus the replacement of Cys 704 by a bulky Phe would inevitably cause steric hindrance with these neighboring residues. Therefore, the Cys to Phe transition would disrupt the structure of Fe65L1 PID2, and hence the APP binding ability would be lost. In the X11 and Dab2 PID/PTB domains, the residues surrounding Phe 479 and Phe 166 are different from those in PID2 of Fe65L1, and thus the spaces are large enough to accommodate the bulky Phe. APP-32mer is shown in magenta, and the APP peptides in complex with X11 and Dab2 PID/PTB domains are orange. The NPTY motif in the APP peptide is shown in cyan with the tyrosine and asparagine residues depicted by stick models. B, schematic drawing of the intermolecular contacts between the APP peptide and the PID/PTB domains of Fe65L1 (left), X11 (middle), and Dab2 (right). The color scheme of the residues represented by the ovals is the same as that in A. The solid red line represents the main chain-main chain hydrogen bond, and the solid cyan line represents the main chain-side chain or the side chain-side chain hydrogen bond. One solid line represents one hydrogen bond. The dotted line represents the van der Waals interaction. The van der Waals interaction involving Asn 759 * is highlighted by thick dotted lines, and the side chains of the related residues in the PID/PTB domains, Cys 704 , Phe 479 , and Phe 166 , are shown in red stick models. The intermolecular interaction was evaluated by the program LIGPLOT (46). C, structure-based sequence alignment of the PID/PTB domains of Fe65L1, X11, and Dab2. The secondary structure elements of PID2 of Fe65L1 are shown at the top of the sequence. The ␤-strands and ␣-helices are highlighted in cyan and pink, respectively. The residues involved in the intermolecular contacts are indicated in red, and the residues involved in main chain-main chain and other kinds of hydrogen bonds are labeled by red and cyan ovals, respectively. OCTOBER 3, 2008 • VOLUME 283 • NUMBER 40

JOURNAL OF BIOLOGICAL CHEMISTRY 27175
Comparison with the Cytoplasmic Domain of APP in the Free Form-The solution structure of the cytoplasmic domain of APP, spanning all 47 residues, has been examined by NMR spectroscopy (47). The study showed that although the peptide does not adopt a stably folded structure regions with transient structure exist. It is informative to compare the free and bound conformations of the peptide to gain insight into the energetics of the protein-peptide interaction. The features of the transient structure in the free APP-32mer include an N-terminal helix capping box formed by Val 742 *-Arg 747 *, a type-I ␤-turn at Asn 759 *-Tyr 762 *, and nascent helices for Asp 739 *-Ala 741 *, Ser 750 *-Glu 758 *, and Lys 763 *-Met 768 *. When APP-32mer binds PID2 of Fe65L1, Asn 759 *-Tyr 762 * remains as a type-I ␤-turn, and Pro 744 *-Gln 753 *, covering part of the helix capping box and the nascent helix in the free form, is folded into a 2.5-turn ␣-helix. Therefore, the regions of the type-I ␤-turn (Asn 759 *-Tyr 762 *) and the short segment of the nascent helix (Ser 750 *-Gln 753 *) are preordered and stabilized in the bound state. On the other hand, Tyr 757 * and Glu 758 *, featured as a nascent helix in the free form, form an antiparallel ␤-sheet with ␤5 of PID2 of Fe65L1, suggesting that a conformational rearrangement occurred for these 2 residues.
Binding thermodynamics studies are useful to characterize the molecular mechanism of binding, and they reflect the vari- ous types of molecular forces that drive binding. The thermodynamic parameters obtained from ITC in this study revealed a highly favorable change in enthalpy and an unfavorable change in entropy. Although the significant folding of the peptide upon binding contributes to the loss of entropy (T⌬S Ͻ 0), additional hydrophobic and enthalpic interactions, accounting for the favorable change in enthalpy (⌬H Ͻ 0), are gained upon binding, leading to a favorable change in the free energy (⌬G Ͻ 0) and stable binding (47).
The Complex of PID2 of Fe65L1 and APP-32mer Shows a Novel Mode of Peptide Binding-Given their functions as adaptors or scaffolds to regulate and organize signaling networks, structural and functional studies of PID/PTB and PID/PTB-like domains and their complexes have been gaining much attention recently (48,49). In most of the PID/PTB⅐peptide complexes studied, the NPX(p)Y motif, the consensus sequence recognized by PID/PTB domains, forms a type-I ␤-turn, whereas the residues N-terminal to the NPX(p)Y motif form an antiparallel ␤-sheet with the ␤5 strand of the PID/PTB domains. In addition, this common binding mode does not depend on the phosphorylation state of the peptide, although the phosphorylation state of the peptide greatly influences the protein-peptide binding affinity for some PID/PTB domains. This typical binding mode is seen, for example, in the complexes of X11 PTB and APP peptide (Fig. 7B) and IRS-1 PTB and IL-4 peptide (Fig. 7C) (50). On the other hand, some non-canonical peptide binding modes have also been reported. In the complex of the Numb PTB domain and GPpY-containing peptide (Fig. 7D), the peptide binds the protein in a helical turn conformation (51). In the complex of the SNT-1 PTB domain and the peptide derived from fibroblast growth factor receptor 1 (Fig. 7E), the N-terminal region of the peptide wraps around the ␤-sheets formed by strands ␤5-␤7, whereas the C-terminal region of the peptide forms antiparallel ␤-sheets with the ␤5 strand and a unique C-terminal ␤-strand of the protein (52). Another example is the complex of the talin F3 domain, a PTB-like domain, and the membrane-proximal integrin ␤3 cytoplasmic domain (Fig. 7F) (53). The membrane-proximal region of ␤3 forms an ␣-helix at its N terminus and lies across strands ␤1, ␤2, ␤6, and ␤7 of the talin F3 domain. Although the APP-32mer and integrin ␤3 peptides, which bind PID2 and the talin F3 domain, respectively, both possess ␣-helices at their N termini, they interact with different regions of the PID/PTB domains.
Biological Implications-Previous studies have shown that the interaction between Fe65L1 and APP plays an important role in APP trafficking and processing (20,21). A pulse/chase investigation revealed that the ratio of the mature to total cellular APP increased after the induction of Fe65L1 (20). The secreted A␤40 level was increased upon Fe65L1 overexpression, and this effect required the binding of Fe65L1 to APP (21). The complex structure presented here provides a molecular basis for the interaction between PID2 of Fe65L1 and APP-32mer. This structural information may facilitate the development of an agent that inhibits the interaction between PID2 and the cytoplasmic domain of APP to reduce A␤ production and to retard or prevent Alzheimer disease.
In addition to the Fe65 family proteins, some other PID/PTB domain-containing adaptor proteins, like the X11, Dab, and JIP family proteins, were also found to bind the cytoplasmic domain of APP. Through these adaptors, other molecules may be recruited to the cytoplasmic domain of APP, leading to APPcentered molecular machinery. The complex structural studies of the cytoplasmic domain of APP with the PID/PTB domains of these adaptors will provide detailed views of the intermolecular interactions. Differential sequence-specific recognition of the cytoplasmic domain of APP by the PID/PTB domains of Fe65L1, X11, Dab1, and Dab2 will offer valuable insights into the competitive binding of APP by these proteins if it exists, although only one case has been reported about the competitive binding of APP by Fe65 and X11␤ at this time (54).
In conclusion, our structural data and binding studies have elucidated the molecular mechanism by which PID2 of Fe65L1 recognizes the cytoplasmic domain of APP. The protein-peptide recognition mode is unique in that the N-terminal region of the peptide forms an ␣-helix and interacts extensively with PID2 of Fe65L1. These studies provide a framework for better understanding of the function, trafficking, and processing of APP and may make it possible to develop new therapeutic strategies toward Alzheimer disease through the regulation of the interactions between APP and APP-binding adapter proteins.