Crystal structures of the Dab homology domains of mouse disabled 1 and 2.

Disabled (Dab) 1 and 2 are mammalian homologues of Drosophila DAB. Dab1 is a key cytoplasmic mediator in Reelin signaling that controls cell positioning in the developing central nervous system, whereas Dab2 is an adapter protein that plays a role in endocytosis. DAB family proteins possess an amino-terminal DAB homology (DH) domain that is similar to the phosphotyrosine binding/phosphotyrosine interaction (PTB/PI) domain. We have solved the structures of the DH domains of Dab2 (Dab2-DH) and Dab1 (Dab1-DH) in three different ligand forms, ligand-free Dab2-DH, the binary complex of Dab2-DH with the Asn-Pro-X-Tyr (NPXY) peptide of amyloid precursor protein (APP), and the ternary complex of Dab1-DH with the APP peptide and inositol 1,4,5-trisphosphate (Ins-1,4,5-P3, the head group of phosphatidylinositol-4,5-diphosphate (PtdIns-4,5-P2)). The similarity of these structures suggests that the rigid Dab DH domain maintains two independent pockets for binding of the APP/lipoprotein receptors and phosphoinositides. Mutagenesis confirmed the structural determinants specific for the NPXY sequence and PtdIns-4,5-P2 binding. NMR spectroscopy confirmed that the DH domain binds to Ins-1,4,5-P3 independent of the NPXY peptides. These findings suggest that simultaneous interaction of the rigid DH domain with the NPXY sequence and PtdIns-4,5-P2 plays a role in the attachment of Dab proteins to the APP/lipoprotein receptors and phosphoinositide-rich membranes.

The Drosophila disabled gene product (DAB) 1 was identified as the result of a genetic screen designed to isolate modifier genes of the abl tyrosine kinase (1)(2)(3). Defects in abl and disabled prevent the formation of proper axonal connections in the central nervous system and cause the death of flies during embryonic development. DAB is tyrosine-phosphorylated by the sevenless receptor kinase and functions as an adaptor protein to recruit SH2-SH3 domain proteins to the signaling complex in Drosophila (4). An evolutionarily conserved family of disabled proteins was revealed by the discovery of a 96-kDa growth factor-responsive phosphoprotein possessing an aminoterminal region of ϳ150 amino acids that is highly similar to the amino terminus of DAB (5). DAB family members contain an amino-terminal DAB homology (DH) domain that acts as a protein-protein and protein-phospholipid interaction module and is fused to diverse carboxyl-terminal sequences.
The DH domain has been identified in two mammalian proteins, Dab1 (also called mDab) and Dab2 (also known as p96 or Doc-2) (5, 6). Dab1 plays a key role in migration of neurons during brain development as a component of the Reelin signaling pathway (7)(8)(9)(10)(11)(12)(13)(14). Reelin, a large glycoprotein secreted by pioneer cell populations early in development (10,11), directs the positioning of neurons during brain development by binding to very low density lipoprotein receptors or apolipoprotein E type 2 receptors (apoER2) (9,12). Engagement of these receptors triggers tyrosine phosphorylation of Dab1 (9,(13)(14)(15), creating a scaffold that recruits signaling proteins possessing SH2 domains. Fyn, an Src tyrosine kinase family member, is directly involved in Reelin-induced Dab1 phosphorylation (16,17). Dab1 is also known to bind to the amyloid precursor family of proteins (18,19), to a novel protocadherin (20,21), and to integrin ␤ (22), although no direct connection between these interactions and specific signaling events has been established. Dab2 participates in a distinct subset of receptor-mediated events. The requirement of Dab2 for visceral endoderm development (23) is consistent with its proposed function as an adapter in the transforming growth factor transforming growth factor-␤/SMAD2 pathway (24). Ablation of Dab2 results in reduced numbers of clathrin-coated pits in kidney proximal tubule cells and the secretion of plasma proteins into the urine (23). This phenotype is consistent with the formation of Dab2-megalin complexes (25) and the binding of Dab2 to clathrincoated pits (26). In addition, Dab2 specifically associates with PtdIns-4,5-P 2 , clathrin (26,27), and cytoskeletal components (28,29) and, therefore, may participate in both endocytic trafficking of lipoprotein receptors and cell adhesion/spreading.
The DH domain belongs to the Pleckstrin homology (PH) superfamily of structures (30), and it most closely resembles the phosphotyrosine binding/phosphotyrosine interaction (PTB/PI) domains present in many proteins involved in protein-protein interaction and signal transduction (31)(32)(33)(34). The PTB/PI domains interact with target proteins and lipids; their binding specificities are crucial to specific functions revealed by high resolution structural studies (32)(33)(34). There are three functional groups of NPXY binding PTB/PI domains (32). One group, typified by Shc and IRS-1, interacts with NPXpY motifs (pY is phosphotyrosine). The second group is defined by the DH domain, which binds only to the unphosphorylated NPXY sequence (8,27). In the third group, the PTB/PI domains of X-11 and Fe65 bind to NPXY motifs independent of their phosphorylation status (32). The Shc PTB/PI domain also binds to phosphoinositides at a site that overlaps with the peptide binding region (35), whereas the DH domain preferentially binds PtdIns-4,5-P 2 independent of the protein interaction site (19). The structures of four NPXY-binding PTB/PI domains are known: Shc, IRS-1, SNT, and X-11 (36 -39). All contain a PH domain-like fold (30) consisting of a seven-stranded ␤ sandwich flanked by a carboxyl-terminal helix. The Shc, IRS-1, and SNT structures illustrate specific binding modes for NPXpY sequences, whereas the PTB/PI domain structure of X-11 explains its ability to bind both phosphorylated and unphosphorylated peptides.
We have investigated the structures of the DH domains of Dab1 and Dab2 to elucidate the structural basis for recognition by the DH domain of the unphosphorylated NPXY sequence motif and PtdIns-P 2 . We report three structures, the Dab2-DH structures in the ligand-free state and the binary complex with the NPXY peptide of APP and the Dab1-DH structure in the ternary complex with the APP peptide and Ins-1,4,5-P 3 . The Dab2-DH structures are the first high resolution model of Dab2, showing the structural similarity to Dab1-DH. We present a comparative analysis of the peptide and the phosphoinositide binding pockets of Dab1 and Dab2. While this manuscript was in preparation, the ternary complex structure of Dab1-DH with the apoER2 peptide and Ins-1,4,5-P 3 was reported (40). This recent ternary complex structure is very similar to our ternary complex structure in that it shows two independent binding sites for the peptide and phosphoinositide. A major difference between the two ternary complex structures is found at the orientation of Ins-1,4,5-P 3 in the binding site. The recent ternary complex structure, which was solved using crystals obtained by soaking with PtdIns-4,5-P 2 , describes the binding of Ins-1,4,5-P 3 in two different orientations. In contrast, our ternary complex structure, which was solved using crystals obtained by co-crystallizing with Ins-1,4,5-P 3 , resolves the uncertainty about the orientation of Ins-1,4,5-P 3 in the binding site. Taken together these structural insights clarify the protein and phospholipid recognition modes that are crucial to the regulation of biological processes by the DHdomain proteins.

EXPERIMENTAL PROCEDURES
Crystallography-We cloned cDNAs encoding Dab1-DH (residues 25-183) and Dab2-DH (residues 33-191) into a pET bacterial overexpression vector (Novagen). The proteins were expressed in Escherichia coli BL21(DE3) cells and purified by three column steps (SP-Sepharose, resource S, and Superdex 75 column;, all from Amersham Biosciences). The purified proteins were placed in a solution of 20 mM HEPES, pH 7.5, 1 mM dithiothreitol, and 1 mM EDTA (protein concentration, 20 mg/ml). We obtained the binary complex of Dab2-DH with the 9-mer peptide of APP ( Ϫ7 NGYENPTYK ϩ1 ) by the vapor diffusion hangingdrop method. The crystallization conditions were 3.8 M sodium formate, 50 mM HEPES, pH 8.5, at 18°C; crystals appeared after 3-5 days. To obtain a mercury derivative, the peptide-bound crystals were soaked in a solution containing 10 mM thimerosal for 3 days. We also obtained ligand-free Dab2-DH by using a crystallization condition slightly different from that used for the binary complex (3.5 M sodium formate, 50 mM HEPES, pH 8.5 at 18°C). We could not obtain the crystals of the ternary complex of Dab2-DH with the APP peptide and Ins-1,4,5-P 3 using either the crystallization conditions of the binary complex or the ligand-free Dab2-DH. A search for new crystallization conditions for the Dab2-DH ternary complex also failed. Instead, the ternary complex crystals of Dab1-DH with the same ligands were obtained by using different crystallization conditions containing 20% (w/v) polyethylene glycol 3350, 0.2 M tri-lithium citrate at 18°C; crystals appeared after 1 month. For data collection, we transferred the crystals to a cryoprotectant solution containing 50% mineral oil and 50% of Paraton-N light hydrocarbon oil. Native and derivative data of the binary complex of Dab2-DH were measured at 100 K on dual image plate detectors (Nonius) at St. Jude, and native data of ligand-free Dab2-DH and the ternary complex of Dab1-DH were collected at Southeast Regional Collaborative Access Team 22-ID beamline at the Advanced Photon Source, Argonne National Laboratory. Crystals of the binary complex and ligand-free Dab2-DH belonged to space group R32 with unit cell dimensions of a ϭ 127.3 Å, c ϭ 269.5 Å and a ϭ 128.2 Å, c ϭ 272.3 Å. The ternary complex crystals of Dab1-DH belonged to space group P2 1 2 1 2 1 with unit cell dimensions of a ϭ 59.1 Å, b ϭ 66.5 Å, c ϭ 90.5 Å. Data reduction, merging, and scaling were done with the DENZO and SCALEPACK programs (41). We used the single isomorphous replacement method with one heavy atom derivative (thimerosal) to determine the protein phases of the binary complex of Dab2-DH with the APP peptide. A difference Patterson map calculation, heavy atom search, electron density calculation, and density modification were done in the program CNS (42). An atomic model was built in the program O (43) and refined in the program CNS (42). The ligand-free Dab2-DH structure was determined by the difference Fourier method because its space group and unit cell dimensions were similar to those of the binary complex of Dab2-DH. The ternary complex structure of Dab1-DH was determined by the molecular replacement method using one molecule of the binary complex structure of Dab2-DH as a search model. The cross-rotation and translation functions were calculated in the program EPMR (44). Model building in the program O (43) and refinement in the program CNS (42) were iterated several times. The data collection and refinement statistics are summarized in Table I.
Nuclear Magnetic Resonance-The NMR samples contained 1 mM uniformly 15 N-labeled sample (wild-type Dab1-DH or one of its two mutants, L45A and K82A, in a solution of 40 mM phosphate, 6 mM deuterated dithiothreitol, pH 6.8, and 5% D 2 O. Ins-1,4,5-P 3 (Sigma) in the same solution was titrated into the protein sample during the course of the NMR experiments. Two-dimensional 15 N heteronuclear single quantum coherence spectroscopy (HSQC) spectra were measured with a Varian INOVA 600-MHz NMR spectrometer at 25°C. Chemical shift perturbation experiments were performed by measuring the chemical shift differences in the HSQC spectra as wild-type Dab1-DH was titrated with Ins-1,4,5-P 3 . In addition, the HSQC chemical shifts of the two mutants were measured and compared with that of the wild-type measured before Ins-1,4,5-P 3 titration. Data were processed and displayed by the program packages NMRpipe and NMRDraw (45). The resulting spectra were analyzed using SPARKY (46).
Mutagenesis and Peptide Binding Assay-The construction of a Dab1-hemagglutinin expression plasmid has been described previously (47). Site-directed mutagenesis was conducted using the QuikChange site-directed mutagenesis kit (Stratagene). A synthetic peptide corresponding to the NPXY-containing cytoplasmic domain of amyloid precursor-like protein 1 ( Ϫ12 CELQRHGYENPTYRFLEE ϩ5 ) and a random peptide used as a negative control (CFEYRNRHQETPELLGEY) have been described previously (18). The peptides were coupled to Sepharose by using the SulfoLink Immobilization kit (Pierce). HEK293T cells were transiently transfected with expression plasmids for either wild-type Dab1 or mutant Dab1 carrying single amino acid substitutions (S114T, H136R, and F158V). After 24 h, cells were lysed in cell lysis buffer (25 mM Tris-HCl, 1% Nonidet P-40, 150 mM NaCl, 5 mM EDTA, 1 mM sodium orthovanadate, 20 mM sodium fluoride, and 20 g/ml each of aprotinin and leupeptin). The lysates were incubated with either the amyloid precursor-like protein 1 NPXY peptide or the control peptide at 4°C for 2 h. The peptide precipitates were washed three times with the cell lysis buffer, and the proteins were separated by SDS-PAGE. After transfer to nitrocellulose membrane, bound Dab1 was detected by Western blotting with anti-Dab1 antibodies. To confirm the expression of Dab1 and Dab1 mutants, 10% of the lysate input was analyzed on a separate gel. Super Signal West Dura Extended Duration Substrate (Pierce) was used for detection.
Mutagenesis and PtdIns-P 2 Binding Assay-DNA fragments encoding Dab1-DH and Dab2-DH were cloned into pET28a vector between the NdeI and SalI sites. Single and double mutations at the proposed phosphoinositide-binding site were introduced by using the QuikChange site-directed mutagenesis kit (Stratagene). The mutations were K45A and K85A of Dab1-DH and K53A and K90A of Dab2-DH. All clones and mutations were checked by automated sequencing on the ABI Prism 3700 DNA Analyzer. His-tagged wild-type and mutant proteins were expressed in BL21(DE3) cells and separated in nickel nitrilotriacetic acid-agarose gel (Qiagen, Valencia, CA). The proper folding of the mutant proteins was confirmed by circular dichroism spectroscopy. The PtdIns-4,5-P 2 binding ability of Dab1-DH and Dab2-DH were assessed by protein-lipid binding assay with the Histagged fusion proteins and PtdIns-4,5-P 2 membrane strips (Echelon Biosciences Inc.). The membrane spotted with PtdIns-4,5-P 2 was incubated with TBS-T (20 mM Tris-HCl, pH 7.6, 150 mM NaCl, and 0.1% (v/v) Triton X-100) containing 3% fatty acid-free BSA for 1 h at room temperature to block nonspecific reactions and was then incubated overnight at 4°C with 500 ng/ml indicated His-tagged protein in TBS-T with 3% fatty acid-free BSA with gentle shaking. The membrane was washed 3 times over 30 min in TBS-T containing 3% BSA and then incubated for 1 h with a 1:500 dilution of rabbit anti-His polyclonal IgG (Santa Cruz Biotechnology, CA) in TBS-T with 1% fatty acid-free BSA. The membrane was washed 4 times over 1 h in TBS-T and then incubated for 2 h with a 1:5000 dilution of anti-rabbit polyclonal IgG linked to alkaline phosphatase in TBS-T with 1% fatty acid-free BSA. Finally, the membrane was washed 4 times over 1 h in TBS-T, and the amount of His-tagged protein bound to PtdIns-4,5-P 2 was detected by the addition of enzyme-catalyzed fluorescent substrate, which was converted to a highly fluorescent product (Amersham Biosciences). The membrane was scanned with a Typhoon 9200 system (Amersham Biosciences), and the protein bound to the membrane was quantitated with the software ImageQuant 5.2 (Molecular Dynamics).

RESULTS
Overall View of the DH Domain-We solved the structures of the DH domain in three different ligand forms: 1) ligand-free Dab2-DH; 2) the binary complex of Dab2-DH with the APPderived NPXY peptide ( Ϫ7 NGYENPTYK ϩ1 ); 3) the ternary complex of Dab1-DH with the same NPXY peptide and Ins-1,4,5-P 3 . The ligand-free and binary complex structures of Dab2-DH were crystallized in the same space group and have three molecules in the asymmetric unit, two of which are related by a noncrystallographic 2-fold rotation and one of which forms a dimer with a symmetrically related molecule. The ternary complex structure of Dab1-DH has two molecules in the asymmetric unit related by a noncrystallographic 2-fold rotation. Molecules in each of these three structures are essentially identical (average r.m.s.d. values in the ligand-free Dab2-DH and its binary complex are 0.26 and 0.1 Å, respectively, whereas the r.m.s.d. value was 0.45 Å for C␣ atoms between molecules in the ternary complex of Dab1-DH).
Each molecule of the DH domain consists of seven ␤-strands and three ␣-helices. The seven ␤-strands are folded into two antiparallel ␤-sheets that are orthogonally arranged to form a ␤ barrel structure, typically known as the PH superfold (30, 34) (Fig. 1). Strands ␤2-␤4 form the first ␤-sheet (one-half of the ␤ barrel) and strands ␤5-␤7 form the second ␤-sheet (the second half of the ␤ barrel), whereas strand ␤1 is part of both the first and the second ␤-sheets. Helices ␣1 and ␣3 are located at the amino and carboxyl termini, and helix ␣2 is found between strands ␤1 and ␤2. The carboxyl-terminal helix ␣3 blocks one end of the ␤ barrel structure. The APP-derived NPXY peptide binds to the end of the ␤ barrel involving the secondary structures of helix ␣3, strand ␤5, and the connecting loop of ␤1/␣2. A shallow basic cleft is located at the other end of the ␤ barrel, is b Phasing power is the ratio of the root mean square value of the calculated heavy atom structure factor to the root mean square value of the difference between calculated and observed derivative structure factors, where it is averaged not only over all reflections but over all phases for each reflection, weighted by the phase probability. c R work ϭ ⌺͉͉F obs ͉ Ϫ ͉F calc ͉/⌺͉F obs ͉, where ͉F obs ͉ and ͉F calc ͉ are observed and calculated structure factor amplitudes, respectively. d R free is equivalent to R work except that 5% of the total reflections were set aside for an unbiased test of the progress of refinement.
lined by a cluster of positively charged residues, and serves as a binding pocket for Ins-1,4,5-P 3 (Fig. 1). Structural Rigidity of the Ligand-binding Site in the DH Domain-Structural comparison of ligand-free Dab2-DH and its binary complex shows subtle differences in the APP peptidebinding site (r.m.s.d. value of 0.30 Å calculated and averaged on C␣ pairs). For example, the side chains of Glu-141, Lys-163, and Phe-166 adopt different conformations in the ligand-free Dab2-DH and the binary complex of Dab2-DH ( Fig. 2A). In the binary complex of Dab2-DH, Glu-141 (Glu-133 in Dab1) interacts with the hydroxyl group of Tyr-0 of the APP peptide, whereas Lys-163 and Phe-166 are part of the hydrophobic binding pocket for Tyr-5 of the APP peptide ( Fig. 2A). Phe-166 (Phe-158 in Dab1) also interacts with the side chain of Asn-3 of the APP peptide (Fig. 3A), and mutation of the corresponding residue in Dab1 (F158V) abolished binding affinity for the peptide (Fig. 4). Structural differences between the binary complex of Dab2-DH and the ternary complex of Dab1-DH are limited to the connecting loop of helix ␣2 and strand ␤2 at the phosphoinositide-binding site (Fig. 2B). Binding of Ins-(1,4,5)-P 3 changed the conformation of the ␣2-␤2 connecting loop, causing Arg-76 and His-81 (Arg-84 and His-89 in Dab2) lining the phosphoinositide-binding site to form hydrogen bonds with the 1-and 5-phosphate groups of Ins-1,4,5-P 3 , respectively (Figs. 2B and 5A). Superposition of these structures suggests that the rigidity of the main chain of the DH domain is crucial for maintaining two independent recognition interfaces favorable for the lipoprotein receptor/APP proteins containing the NPXY sequence and for the PtdIns-4,5-P 2 -rich membranes.
Interaction of the DH Domain with the NPXY Sequence-The APP peptide-binding site on the DH domain is located between the secondary structures of helix ␣3, strand ␤5, and the connecting loop of ␤1/␣2 (Fig. 1). At this binding site, the NPXY motif of the APP peptide ( Ϫ3 NPTY 0 ) forms a ␤ turn, whereas the tri-sequence ( Ϫ5 YEN Ϫ3 ) forms an antiparallel ␤ strand to strand ␤5 (Fig. 3). Similar conformations of the NPXY-containing peptides were previously reported for other PTB/PI struc- tures (36,37,39) and recently for a Dab1-DH structure (40).
The NPXY sequence with unphosphorylated Tyr-0 is preferred in binding to Dab1 and Dab2 (19). The structures we have determined show that the benzene group of Tyr-0 of the APP peptide is stacked against the C␤ portion of Ser-114 in Dab1 (Ser-122 in Dab2), whereas its hydroxyl group is hydrogen-bonded to the backbone oxygen of Gly-131 (Gly-139 in Dab2) and forms van der Waals interactions with the side chain of Glu-133 (Glu-141 in Dab2) and the imidazole ring of His-136 (His-144 in Dab2) (Fig. 3). Similar interactions have been observed in the recently solved structure of Dab1-DH (40). These extensive interactions exclude binding of phosphorylated Tyr-0, consistent with the biochemical finding that tyrosine phosphorylation of the NPXY motif inhibits binding by the DH domain (19,27).
The type of residue at the (Ϫ5) position from Tyr-0 of the NPXY sequence also contributes to Dab1 and Dab2 binding. For example, NPXY sequences require either tyrosine or phenylalanine at the (Ϫ5) position to bind efficiently to Dab1 and Dab2 (19,27). The complex structures we resolved indicate that the side chain of Tyr-5 of the APP peptide binds at a pocket lined with hydrophobic residues, although the pocket-lining residues differ slightly between Dab1-DH and Dab2-DH because of two distinct side-chain conformations of Tyr-5 in the APP peptide structures (Fig. 3). The hydrophobic nature of the Tyr-5 binding pocket explains why the (Ϫ5) residue in the NPXY sequence should be either tyrosine or phenylalanine. For example, the benzene ring of Tyr-5 in Dab2-DH is surrounded by the hydrophobic side chains of Ile-124, Val-159, Lys-163, Phe-166, and Arg-126 (Ile-116, Ile-151, Arg-155, Phe-158, and Lys-118 in Dab1), whereas the hydroxyl group of Tyr-5 interacts with the carbonyl oxygen of Val-159 (Ile-151 in Dab1), which is also hydrogen-bonded to the amide nitrogen of Lys-163 in helix ␣3 (Arg-155 in Dab1) (Fig. 3A). In Dab1-DH, the benzene ring of Tyr-5 is stacked between the hydrophobic portion of Arg-155 (Lys-163 in Dab2) and the carbonyl oxygen of Glu-4 of the APP peptide and interacts with the side chain of Phe-158 (Phe-166 in Dab2), whereas the hydroxyl group of Tyr-5 is not engaged in interaction with Dab1 (Fig. 3B). In the recent report describing the Dab1-DH structure (40), Phe-5 of an apoER2 NPXY peptide also bound to a hydrophobic pocket lined by residues similar to those delineated in our Dab1-DH and Dab2-DH structures.
To examine the relative importance of protein residues for binding to the NPXY sequence, we constructed three Dab1 mutants based on the Dab1-DH structure: S114T in strand ␤ 5 (S122T in Dab2), F158V in helix ␣ 3 (F166V in Dab2), and H136R in strand ␤ 7 (H144R in Dab2) (Fig. 3B). Ser-114 is quite invariant in three of four PTB/PI domain structures. In the Shc PTB/PI domain, the S114A mutation reduces peptide binding by about 40% (36). In our S114T substitution in Dab1, the additional methyl group of Thr should disrupt the hydrophobic interaction with the benzene group of Tyr-0 of the APP peptide. The second Dab1 mutation, F158V, resulted in a 25fold decrease in affinity for the NPXY motif (19), consistent with the observation that Phe-158 of Dab1-DH interacts with the side chain of Asn-3 of the NPXY motif (Fig. 3B). Third, we designed the H136R mutant of Dab1 because His-136 corresponds to Arg-175 in Shc, which plays a crucial role in interaction with phosphorylated tyrosine of the peptide ligand (36). When peptide binding analysis of the three Dab1 mutants was performed, we detected no significant binding of the F158V and S114T mutants to the APP NPXY peptide (Fig. 4). The H136R mutant bound to the NPXY peptide very weakly (Fig. 4), but it did not bind to the tyrosine-phosphorylated NPXY peptide (data not shown). Western blot analysis of the lysates confirmed that the mutant forms of Dab1 were expressed at similar levels (Fig. 4).
Interaction of the Dab DH Domain with Inositol 1,4,5-Trisphosphate-The Dab DH domain preferentially binds to PtdIns-4,5-P 2 , one of the most abundant inositol lipids in cells (19,26). The Shc PTB/PI domain also binds phosphoinositides, and this binding is necessary for tyrosine phosphorylation of Shc by activated receptors (35). Unlike the Shc PTB/PI domain, which has overlapping binding sites for proteins and phosphoinositides, the Dab DH domain is thought to have two independent binding sites for proteins and phosphoinositides (19,26). Ins-1,4,5-P 3 is known to compete with large unilamellar vesicles containing PtdIns-4,5-P 2 for binding to Dab1-DH (19). The binding of Ins-1,4,5-P 3 to Dab1-DH was confirmed by the chemical shift perturbation of peaks in the 15 N HSQC NMR spectra (Fig. 5B). In Fig. 5B, the upper inset shows that as Dab1-DH is titrated with Ins-1,4,5-P 3 , the peaks move across the spectra, indicating that the binding event is in fast exchange on the NMR time scale (apparent K d ϭ ϳ52 M). The lower inset shows that many of the peaks do not move, and therefore, that the majority of the Dab1-DH structure is not changed on binding of Ins-1,4,5-P 3 . Our structure of Dab1-DH shows the binding site for Ins-1,4,5-P 3 , which is separated from that of the APP peptide by the core ␤-barrel structure (Fig. 1). The Ins-1,4,5-P 3 -binding site is lined by the basic residues Lys-45 (strand ␤ 1), Arg-76 (helix ␣2), His-81, and Lys-82 (the loop of ␣2/␤2), Arg-124 (the loop of ␤5/␤6), and Lys-142 (strand ␤7). These basic residues compensate for the negative charges of the three phosphate groups of Ins-1,4,5-P 3 (Figs. 1 and 5A). Briefly, hydrogen bonds form between each of the 4-and 5-phosphate groups of Ins-1,4,5-P 3 and both Lys-45 and Lys-82 and between the 1-phosphate group of Ins-1,4,5-P 3 and Arg-76. The 4-phosphate group also forms hydrogen bonds with Arg-124 and Lys-142, and the 5-phosphate group forms hydrogen bonds with His-81. The Ins-1,4,5-P 3Ϫ binding residues are conserved in Dab2-DH ( Figs. 1 and 5A), suggesting that Dab2-DH can bind to Ins-1,4,5-P 3 in the same manner as does Dab1-DH. The recently reported structure of Dab1-DH also showed that the binding site of the polar head group of PtdIns-4,5-P 2 (i.e. Ins-1,4,5-P 3 ) is lined by the same basic residues as those found in our structure of Dab1-DH, but that report described the binding of Ins-1,4,5-P 3 in two different orientations; the first orientation is the same as that shown in our structure of Dab1-DH, whereas in the second orientation, the position of the 1-phosphate group is rotated from the latter orientation (40). These two binding orientations may have been an artifact caused by incomplete soaking of the crystals in a solution containing PtdIns-4,5-P 2 .  Fig.  6. B, Ins-1,4,5-P 3 binds to Dab1-DH. NMR chemical shift perturbation experiments were performed by comparing 15 N HSQC spectra of 15 N-labeled Dab1-DH before (red) and after the addition of increasing amounts of Ins-1,4,5-P 3 (green-purple-blue). Insets show the perturbation of peaks that are associated with residues involved (upper inset) or not involved (lower inset) in binding.
In contrast, our approach to structure determination of Dab1-DH involved co-crystallization with Ins-1,4,5-P 3 . As observed in the Dab DH domain, the PH domains bind phosphatidylinositol lipids via their positively charged regions, centered on the ␤1/␤2 loop (30,48). Moreover, two of three positively charged residues of the Shc PTB/PI domain, which are crucial for the binding of phospholipids, are located at the rim of the ␤ barrel between the ␤1/␤2 loops (35).
To further test the contribution of Dab1-DH Lys-45 and Lys-82 (Lys-53 and Lys-90 in Dab2) to the binding of PtdIns-4,5-P 2 , we created alanine substitutions for one or both residues. In a binding assay, the K82A mutant of Dab1-DH had a reduced binding affinity for PtdIns-4,5-P 2 , whereas the other mutants did not bind at all (Fig. 6). These results suggest that the mutated lysine residues significantly contribute to binding of PtdIns-4,5-P 2 in Dab proteins. Inclusion of the APP peptide did not affect binding of the wild-type Dab1-DH to PtdIns-4,5-P 2 (data not shown), confirming that the peptide-binding and phosphoinositide-binding sites of the DH domain are separate. When we compared the HSQC spectra of the two 15 Nlabeled Dab1-DH mutants K45A and K82A with the spectra of wild-type Dab1-DH bound to Ins-1,4,5-P3, the peaks that were shifted by the mutations were almost identical to the peaks perturbed by Ins-1,4,5-P3 binding to wild-type Dab1-DH. This finding indicates that residues Lys-45 and Lys-82 in Dab1 are directly involved in binding to PtdIns-4,5-P 2 , as shown in the Dab1-DH structure (Fig. 5A). DISCUSSION We have solved the structures of Dab2-DH in a ligand-free state and in complex with the APP peptide. These structures provide the first view of Dab2-DH and identify the Dab2-DHbinding site for the NPXY sequences. We also elucidated the ternary complex structure of Dab1-DH with the APP peptide and Ins-1,4,5-P 3 . Recently, the ternary complex structure of Dab1-DH with the apoER2 peptide and Ins-1,4,5-P 3 was reported (40). These two ternary complex structures are very similar and show two independent binding sites for the peptide and phosphoinositide.
The DH domain structures reported here explain the preference of Dab1 and Dab2 for either tyrosine or phenylalanine at the Ϫ5 position of the NPXY sequence and the acceptance by Dab1 of tryptophan at this position (19,27). The Ϫ5 residue binding pocket of Dab2 is lined by hydrophobic residues, with the exception of Arg-126. Although the hydrophobic residues form tight van der Waals interactions with the benzene ring of Tyr-5 of the APP peptide, the guanidinium group of Arg-126 interacts with the benzyl electrons of Tyr-5 (Fig. 3A) in a manner considered to be a weakly polar interaction between two hydrophobic atoms that is stronger than van der Waals interactions (49,63). These two types of interactions explain the specificity of the Ϫ5 residue binding pocket for benzyl residues, either phenylalanine or tyrosine. However, the Ϫ5 residue binding pocket may not permit the binding of large tryptophan residues, which may sterically hinder residues lining the Ϫ5 residue binding pocket or the binding of non-aromatic hydrophobic residues, which is likely to be impeded by the charge on Arg 126 at the Ϫ5 residue binding pocket. This premise is supported by the results of our mutagenesis experiments, in which replacement of Tyr-5 of the APP peptide with amino acids other than phenylalanine or tyrosine substantially reduced binding to Dab2-DH (27). On the other hand, the Ϫ5 residue binding pocket of the Shc PTB domain accommodates bulky hydrophobic residues, although phenylalanine is not preferred over other residues, including isoleucine, leucine, and valine (50 -52). Superimposed structures of the binary complex of Dab2-DH and the Shc PTB/PI domain show that Arg-126 in Dab2 is replaced by Gly in Shc (data not shown); therefore, Arg-126 of Dab2-DH plays a role in preferential binding to Phe or Tyr at the Ϫ5 position of the NPXY sequences. In Dab1-DH, in contrast, the Tyr-5 benzene ring of the APP peptide is sandwiched between Arg-155 and the carbonyl oxygen of Glu-4 of the APP peptide, and one edge of the Tyr-5 benzyl ring interacts with Phe-158, whereas the other edge is exposed to solvent (Fig. 3B); thus, any aromatic residue (not only tyrosine and phenylalanine but also tryptophan) may fit into this (Ϫ5) residue binding pocket. This finding is consistent with the mutagenesis results, in which replacement of Tyr-5 of the APP peptide with tryptophan was tolerated in binding to Dab1-DH (19). Binding of other hydrophobic residues to the (Ϫ5) residue binding pocket may occur but will have less affinity because of the inadequate size of the pocket. In the X-11 PTB/PI domain structure, a similar pocket accommodates Tyr-5 of the APP peptide (39).
The DH domain structures we have solved explain the binding specificity of Dab1 and Dab2 for NPXY sequences of lipoprotein receptors and APP family proteins in which Tyr-0 is not phosphorylated. The DH domain has a polar pocket that extensively interacts with the Tyr-0 hydroxyl group, preventing the binding of phosphorylated tyrosine (Figs. 1D and 3) (40). In contrast, the IRS-1, Shc, and SNT PTB/PI domains have a basic pocket with two arginine residues that compensate for the negative charges of phosphorylated Tyr-0 of the NPXY peptide (Fig. 1D) (36 -38, 53). The X-11 PTB domain does not interact with the Tyr-0 hydroxyl group of the NPXY peptide (Fig. 1D) (39) and, therefore, may not detect the phosphorylation state of Tyr-0 in the NPXY peptide; this feature would allow phosphorylated and unphosphorylated peptides to bind equally well (39).
Both Dab1 and Dab2 can bind to phosphoinositides (preferably PtdIns-4,5-P 2 ) in the plasma membrane via the DH domain (19,26). The Dab1-DH ternary structure shows that the 1-, 4-, and 5-phosphates of Ins-1,4,5-P 3 interact with the side chains of five positively charged residues to form 11 direct and water-mediated hydrogen bonds (Fig. 5A). This large number of interactions may explain the relatively high selectivity of Dab1 for PtdIns-4,5-P 2 (apparent K d ϭ ϳ2.0 M) (19). An equally large number of interactions has been observed in the PH domains of phospholipase C, Bruton's tyrosine kinase, general receptors for phosphoinositides-1, and the dual adaptor for phosphotyrosine and 3-phosphoinositides, all of which have high affinity for their respective phosphoinositides (54 -57). Single or double substitution of lysine residues with alanine (Lys-45 and Lys-82 of Dab1-DH; Lys-53 and Lys-90 of Dab2-DH) abolishes or substantially diminishes the affinity of Dab for PtdIns-4,5-P 2 (Fig. 6). In the ternary complex structure of Dab1-DH, each of these two lysine side chains forms two hydrogen bonds with the 4-and 5-phosphates. The phospholipase C PH domain binds to PtdIns-4,5-P 2 with high affinity, and its structure shows that two lysine residues form hydrogen bonds to both the 4-and 5-phosphates (54). The strong interaction of the Dab DH domain with PtdIns-4,5-P 2 targets Dab1 and Dab2 to the area of the plasma membrane where PtdIns-4,5-P 2 is abundant.
Simultaneous interactions of the Dab DH domain with both the NPXY peptide and Ins-1,4,5-P 3 are implicated in Dab functions. For example, the interaction of Dab1 with the NPXY sequences of various lipoprotein receptors brings Dab1 proximal to the inner leaflet of membranes, and its interaction with the head of PtdIns-4,5-P 2 recruits Dab1 to the membranes. During mammalian brain development, Reelin binds to the lipoprotein receptors apoER2 and very low density lipoprotein receptors and induces tyrosine phosphorylation of Dab1 (14,15,47). An Src family tyrosine kinase, Fyn, mediates Dab1 hyperphosphorylation (16); the dependence of Fyn kinase activation on the presence of Dab1 (16,17) suggests that Dab1 acts as an allosteric activator as well as a substrate of Fyn. Because the activities of Src family tyrosine kinases, including Fyn, are membrane-associated (58,59), membrane localization of Dab1 via the DH domain is likely to bring substrate/activator into contact with the kinase, thus facilitating phosphorylation of Dab1 during Reelin signaling. Dab2, on the other hand, is thought to function in sorting the lipoprotein receptors into clathrin-coated pits for endocytosis (23,26,27). PtdIns-4,5-P 2 is likely to be clustered in raft-like regions of the membrane (60) and is thought to be essential for endocytosis (61). Dual interaction of Dab2-DH with both the NPXY sequences and PtdIns-4,5-P 2 may associate lipoprotein receptors with the PtdIns-4,5-P 2 -rich membrane compartment. Furthermore, the carboxylterminal region of Dab2 interacts with both clathrin and the clathrin adapter molecule AP-2, which suggests that Dab2 plays a role in the delivery of lipoprotein receptors to clathrincoated pits for endocytosis. Additionally, because transforming growth factor-␤ receptors are frequently found in the endosomal compartment (62), localization of Dab2 in the same compartment brings Dab2-bound SMAD proteins into proximity with transforming growth factor-␤ receptors for signaling.
In conclusion, our structural data and mutagenesis studies have elucidated the molecular mechanism by which the DH domain recognizes NPXY sequences containing an unphosphorylated Tyr-0 and a benzyl residue at the Ϫ5 position. These studies also identified the phosphoinositide binding pocket lined by six basic residues to which Ins-1,4,5-P 3 is bound and showed that two lysine residues in the pocket are essential for binding.