A Novel Interaction between Atrophin-interacting Protein 4 and β-p21-activated Kinase-interactive Exchange Factor Is Mediated by an SH3 Domain*

Cross-talk between G protein-coupled receptors and receptor tyrosine kinase signaling pathways is crucial to the efficient relay and integration of cellular information. Here we identify and define the novel binding interaction of the E3 ubiquitin ligase atrophin-interacting protein 4 (AIP4) with the GTP exchange factor β-p21-activated kinase-interactive exchange factor (βPIX). We demonstrate that this interaction is mediated in part by the βPIX-SH3 domain binding to a proline-rich stretch of AIP4. Analysis of the interaction by isothermal calorimetry is consistent with a heterotrimeric complex with one AIP4-derived peptide binding to two βPIX-SH3 domains. We determined the crystal structure of the βPIX-SH3·AIP4 complex to 2.0-Å resolution. In contrast to the calorimetry results, the crystal structure shows a monomeric complex in which AIP4 peptide binds the βPIX-SH3 domain as a canonical Class I ligand with an additional type II polyproline helix that makes extensive contacts with another face of βPIX. Taken together, the novel interaction between AIP4 and βPIX represents a new regulatory node for G protein-coupled receptor and receptor tyrosine kinase signal integration. Our structure of the βPIX-SH3·AIP4 complex provides important insight into the mechanistic basis for βPIX scaffolding of signaling components, especially those involved in cross-talk.

In addition to mediating the ubiquitination and sorting of CXCR4, a number of recent studies detail the regulatory role AIP4 plays in developmental, immunological, and oncogenic signaling pathways (12)(13)(14)(15)(16)(17)(18). The AIP4/Itch protein is composed of an N-terminal C2 domain, followed by a proline-rich region, four WW domains, and a C-terminal catalytic HECT domain (3,4,12,13). AIP4 is abundantly expressed in most human tissues and displays tissue specificity similar to that of Cbl (19).
␤-PAK-interactive exchange factor (␤PIX also referred to as Cool-1) is composed of modular domains, including an N-terminal SH3 domain followed by DH (Dbl, diffuse B-cell lymphoma homology) domain, PH (pleckstrin homology) domain, and a leucine zipper (20,21). Functionally, ␤PIX acts as a guanine nucleotide exchange factor for Rac/Cdc42 and was first identified as an exchange factor for PAK (20,22,23). Recent studies illustrate a role for ␤PIX in breast cancer pathogenesis as well as aspects of CXCR4-induced cellular chemotaxis (24 -27).
Like other SH3 domains, ␤PIX-SH3 binds to proline-rich sequences with a polyproline II helix (PPII) conformation. Although most SH3 domain ligands contain a PXXP motif, ␤PIX-SH3 binds ligands with a non-canonical PXXXPR sequence and serves as a scaffolding point for a number of protein-protein interactions as demonstrated by the direct coupling of ␤PIX to p21-activated kinase (PAK) family members (20). The ␤PIX-SH3 domain also binds to the PXXXPR motif of Cbl family members to facilitate the clustering of proteins involved in the down-regulation of EGF receptor signaling (26,28). Interestingly, the ␤PIX-SH3 domain also binds proteins, including Rac1 and SAP (signaling lymphocyte activation molecule-associated protein) (29,30) that possess neither the PXXXPR nor the canonical PXXP motifs. The structural basis for these interactions and the functional ramifications of competition among multiple ligands for overlapping binding sites on ␤PIX remain to be elucidated.
Here we identify the E3 ubiquitin ligase AIP4 as a binding partner for the ␤PIX scaffolding protein. We demonstrate an endogenous ␤PIX⅐AIP4 complex in breast cancer cell lines and show that complex formation is mediated in part by ␤PIX-SH3 domain binding to a proline-rich stretch of AIP4. We show using isothermal titration calorimetry (ITC) that the interaction is heterotrimeric in solution, with one AIP4-derived peptide binding to two ␤PIX-SH3 domains. Fluorescence assays suggest that the mode of interaction for the ␤PIX-SH3⅐AIP4 complex is unique compared with known ␤PIX-SH3 interactions. We determined the crystal structure of the ␤PIX-SH3⅐AIP4 complex to 2.0-Å resolution. AIP4 binds the ␤PIX-SH3 domain as a canonical Class I ligand with an additional PPII helix that makes extensive contacts with another face of ␤PIX. Our structure of the ␤PIX-SH3⅐AIP4 complex provides important insight into the mechanistic basis for ␤PIX scaffolding of signal components. The novel interaction between ␤PIX and AIP4 represents a new regulatory node for GPCR and RTK signal integration.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-Except where noted, all buffers and chemicals were purchased from either Fisher or Sigma. Monoclonal anti-Myc and monoclonal anti-AIP4 were purchased from Santa Cruz Biotechnology. Polyclonal anti-␤PIX antibody was from Chemicon International. Polyclonal anti-HA antibody was purchased from Covance. Monoclonal anti-FLAG M2 antibody was purchased from Sigma. Monoclonal anti-Itch antibody was purchased from BD Biosciences. Goat anti-mouse IgG and goat anti-rabbit IgG, horseradish peroxidase-conjugated secondary antibodies were from Upstate Biotechnology. SDF1␣ was purchased from R&D Systems, and EGF was purchased from Calbiochem.
Cell Culture, Plasmids, and Transfection-HEK293T, MCF7, MDA-MB-231, and NIH 3T3 cells were all cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. The pcDNA3-FLAG-CIN85 mammalian expression construct was a kind gift from Dr. Sachiko Kajigaya (National Institutes of Health, Bethesda, MD). The pEBB-␤PIX-HA mammalian expression construct was a kind gift from Dr. Bruce Mayer (University of Connecticut, Farmington, CT). The pRK5-Myc-AIP4 mammalian expression construct was a kind gift from Dr. Tony Pawson (University of Toronto, Toronto, Ontario, Canada). The pGEX-6P1-␤PIX-SH3 domain expression construct was a kind gift from Dr. André Hoelz (Rockefeller University, New York, NY). All truncation and mutant constructs were generated by site-directed mutagenesis using high fidelity thermostable DNA polymerase Pfu (Stratagene). All transfection studies were performed using DNA-Lipofectamine (Invitrogen), and cells were assayed or harvested ϳ48 h after transfection. Harvested cells were either used immediately or snap frozen in liquid nitrogen and kept at Ϫ80°C.
Immunoblotting, Co-immunoprecipitation, and Pulldown Assays-Harvested cells were lysed in lysis buffer (50 mM Tris-HCl, pH 8, 150 mM NaCl, 5 mM EDTA, 0.5% sodium deoxycholate (w/v), 1% Nonidet P-40 (v/v), 0.1% SDS (w/v)), plus protease inhibitor tablets (Roche Applied Science). Cell lysates were subsequently cleared by centrifugation, and the supernatant was incubated with the appropriate antibody at 4°C for 1 h to overnight followed by incubation with protein A-or G-agarose for 2 h at 4°C. Immunoprecipitates were then washed three to five times with cold lysis buffer, eluted by boiling in SDS buffer, and separated by SDS-PAGE. Proteins were subsequently transferred to nitrocellulose membranes and probed with the appropriate antibody of interest. To probe for endogenous AIP4 levels a monoclonal antibody to mouse Itch was used. This antibody readily detects AIP4, because Itch and AIP4 are highly homologous. In overexpression studies, all AIP4 constructs encode an N-terminal Myc tag, ␤PIX constructs encode a C-terminal HA tag, and CIN85 constructs encode an N-terminal FLAG tag to aid in immunoblot detection.
For GST pulldown assays, lysates of transformed Escherichia coli BL21(DE3) cells expressing either GST-␤PIX-SH3 or GST alone were bound to glutathione-Sepharose beads (GE Amersham Biosciences) in a Tris buffer (50 mM Tris, 150 mM NaCl, 2 mM dithiothreitol, pH 7.3) for 1 h at 4°C. The beads were subsequently washed in the same buffer and incubated with HEK 293T cellular lysate expressing either Myc-tagged AIP4 or an empty vector. The complexes were pulled down, washed extensively, and resolved via SDS-PAGE. Proteins were subsequently transferred to nitrocellulose and probed for AIP4 binding using an anti-Myc antibody.
Protein Expression and Purification-The SH3 domain ␤PIX was expressed and purified essentially as described previously (31).
Peptide Synthesis and Preparation-All peptides were synthesized by the Proteomics Resource Center at the Rockefeller University (New York, NY), purified by high-performance liquid chromatography, and verified by mass spectrometry. Peptides that correspond to regions of AIP4 were as follows: AIP4 206    ) and that of human Cbl-b (Cbl-b 899 -914 , 899 SQAPARPPKPRPRRTA 914 Y). Note that a tyrosine residue was added on to the N terminus of peptides AIP4 206 -229 and AIP4-R217A or the C terminus of Cbl-b 899 -914 to aid in concentration determination. All protein and peptide concentrations were determined by UV spectroscopy and/or amino acid analysis.
Fluorescence Spectroscopy-Fluorescence measurements were performed essentially as described previously (31). All experiments were carried out in 50 mM Hepes, 150 mM NaCl, pH 7.3, at 20°C in a 4-ϫ 4-mm quartz cuvette. Intrinsic Trp fluorescence was measured by exciting samples at 295 nm (1 nm bandpass) and monitoring emission from 315 to 450 nm (10 -15 nm bandpass). The EC 50 values were determined by titrating ␤PIX-SH3 domain samples with increasing amounts of WT AIP4 206 -229 peptide and monitoring both the shift in Trp emission wavelength and overall fluorescence intensity. Competition experiments were performed in a similar manner using the PAK2 176 -199 and Cbl-b 899 -914 peptides. The PAK2 176 -199 peptide sequence contains a tyrosine residue, and it is possible that some energy transfer to this residue from the Trps in the SH3 domain takes place upon binding, thereby reducing the fluorescence emission intensity. However, a PAK2 peptide in which this Tyr residue is converted to a Phe did not significantly increase the fluorescence emission intensity (data not shown). Data were analyzed in Sigma-Plot using nonlinear regression and fitted to a sigmoidal doseresponse (variable slope) equation four-parameter logistic equation: Y ϭ minimum ϩ (maximum Ϫ minimum)/1 ϩ 10(log EC 50 Ϫ x)*Hill slope.
ITC-All ITC measurements were performed at 20°C using a MicroCal (Northampton, MA) VP-ITC calorimeter essentially as described previously (31). Typically, the sample cell contained 50 M solution of the ␤PIX-SH3 domain and was titrated at equal intervals with 10-l aliquots of 500 M AIP4 peptide solution for a total of ϳ290 l. The heat generated due to dilution of the titrants (peptide) was subtracted for baseline correction. Baseline corrected data were analyzed with MicroCal ORIGIN version 6.0 software. All experiments were performed at least twice.
Crystallography, Data Collection, and Structure Determination-The SH3 domain (3 mM in 20 mM Tris, pH 8.0, 100 mM NaCl, 5 mM dithiothreitol) was mixed at a 1:2 molar ratio with AIP4 206 -229 peptide dissolved in the same buffer and crystallized at 20°C by the vapor diffusion method. Drops containing 1 l of the complex were mixed with 1 l of the well solution (0.1 M Tris-HCl, pH 7.1-7.9, 0.2 M ammonium sulfate, 32-38% (w/v) polyethylene glycol-MME 5000). The crystals were allowed to grow at 20°C and appeared as long rectangular rods, with typical dimensions of 40 m ϫ 30 m ϫ 600 m. The crystals were snap-frozen in a stream of liquid nitrogen in the mother liquor supplemented by 10% glycerol. Data were collected using a Raxis IV (Rigaku) with 1°oscillations. Data were integrated and scaled using the program HKL2000 (32). The phases were determined by molecular replacement, using the model from 2AK5 from the Protein Data Bank and the program Phaser (33). The model was manually modified with the correct peptide residues with iterative building and refinement using the programs COOT, REFMAC5, and ARP/WARP (34). Superposition of the C␣ positions was calculated using Lsqman (35). Electrostatic surfaces were calculated using Grasp (36). All molecular graphics in the figures were made using the program PyMOL (DeLano Scientific, San Carlos, CA).
Accession Codes-Structural coordinates and structure factors have been deposited in the Protein Data Bank and have been assigned the ID code 2P4R.

RESULTS
The Proline-rich Region of AIP4 Binds to the SH3 Domain of ␤PIX-AIP4 regulates the stability and degradation of p73 and p63 oncogenic transcription factors, and both AIP4 and ␤PIX have been shown independently to be involved in breast cancer metastasis (15-18, 26, 37, 38). With this in mind we first assayed for the endogenous association of ␤PIX and AIP4 in two breast cancer cell lines as well as a fibroblast line also known to express both proteins (11,13). Our co-immunoprecipitation experiments reveal that endogenous ␤PIX forms a stable complex with endogenous AIP4 in MDA-MB-231 and NIH3T3 cell lines (Fig. 1A). Interestingly, we could not detect an endogenous interaction between these proteins in the MCF7 line (Fig. 1A).
To test whether AIP4 and ␤PIX interact directly to form a complex, we conducted a series of pulldown and co-immunoprecipitation experiments using affinity-tagged constructs. We hypothesized that AIP4 interacts directly with ␤PIX through binding of the proline-rich region of AIP4 to the SH3 domain of ␤PIX. First we found that ␤PIX-SH3 alone was sufficient to pull down AIP4 in a GST pulldown assay with HEK293T lysateexpressing Myc-tagged AIP4 (Fig. 1B). We further confirmed the importance of the proline-rich region of AIP4 using coimmunoprecipitation assays in HEK293T cells. HA-tagged ␤PIX was co-expressed with full-length or truncated forms of Myc-tagged AIP4; immunoprecipitation was performed with anti-Myc antibody, and immunoblots were performed to test for the association of ␤PIX. We found that deletion of the proline-rich region of AIP4 abolished the interaction with ␤PIX in this assay (Fig. 1C).
AIP4 Binds to the CIN85 Scaffolding Protein-The ␤PIX-SH3 domain is phylogenetically similar to the three SH3 domains of CIN85, an adaptor protein involved in Cbl-mediated downregulation of RTKs (9,26). Like ␤PIX, the SH3 domains of CIN85 recognize and bind the non-canonical PXXXPR motifs in Cbl and PAK proteins (26,28,39). When both FLAG-tagged CIN85 and Myc-tagged AIP4 were co-expressed in HEK293T cells we found that CIN85 pulled down AIP4, albeit to a lesser extent than ␤PIX (Fig. 1D). Notably, this interaction dissipated as cells co-expressed increasing amounts of ␤PIX (Fig. 1D).
AIP4 Forms a Heterotrimeric Complex with ␤PIX-SH3 Domains in Solution-Guided by our cellular binding assays we designed a series of peptides corresponding to the proline-rich region of AIP4 shown to bind the ␤PIX-SH3 domain. We synthesized two WT peptides AIP4 206 -229 and AIP4 211-226 that contain the proline-rich stretch, but vary in the number of residues flanking these regions, and three peptides AIP4 201-216 , AIP4 216 -232 , and AIP4 216 -226 that contain only the N-or C-terminal PXXXPR motif (Fig. 2B). Our ITC studies reveal that both peptides AIP4 206 -229 and AIP4 211-226 bind to the SH3 domain of ␤PIX with high affinity, 7.38 Ϯ 0.9 M and 10.1 Ϯ 0.6 M, respectively (Fig. 2C). The C-terminal portion of the sequence is required in the binding interaction, because peptide AIP4 216 -232 bound to the SH3 domain with a K d of 72.2 Ϯ 5.0 M, whereas peptide AIP4 201-216 did not show any detectable binding (Fig. 2C). The significant decrease in affinity observed between peptide AIP4 216 -232 relative to AIP4 206 -229 and AIP4 211-226 further indicates that residues adjacent to the C-terminal proline-rich region are necessary to impart high affinity binding. Importantly, a binding stoichiometry for AIP4 peptide to ␤PIX-SH3 domain of 1:2 was consistently observed for all of these peptides (Table 1). This suggests that binding the proline-rich region of AIP4 by ␤PIX may induce the formation of heterotrimeric complexes in solution, similar to what has been previously reported for CIN85-SH3⅐Cbl-b and ␤PIX-SH3⅐Cbl-b (28).
Alternate Mode of Binding for the AIP4⅐␤PIX-SH3 Complex in Solution-We next investigated the role that residue Arg-217 of AIP4 plays in formation of ␤PIX-SH3⅐AIP4 heterotrimeric complexes. The analogous Arg in Cbl-b (Arg-904) is necessary for the formation of heterotrimeric complexes of Structural Insight into the ␤PIX⅐AIP4 Complex ␤PIX-SH3⅐Cbl-b (28). We find that the R217A mutant of AIP4 associated with ␤PIX when each were co-expressed in HEK 293T cells and subjected to immunoprecipitation ( Fig. 2A). As expected, our ITC tests on an AIP4-R217A peptide indicated that this mutation reduced the affinity of the ␤PIX-SH3⅐AIP4 interaction with an apparent K d of 43.4 Ϯ 3.5 M (Fig. 2C). To our surprise, this mutation did not alter the binding stoichiometry for the ␤PIX-SH3⅐AIP4 complex, n ϭ 0.452 Ϯ 0.04, in contrast with the effect of an analogous Arg mutation on the stoichiometry observed for both the CIN85⅐Cbl-b and ␤PIX⅐Cbl-b complexes (28).
The observation that the AIP4-R217A mutation does not alter the binding stoichiometry to a 1:1 complex suggests that AIP4 may bind differently to the ␤PIX-SH3 domain than either PAK or Cbl proteins. To test this hypothesis we employed a fluorescence assay that monitors ␤PIX-SH3 domain intrinsic Trp fluorescence emission during ligand binding. Similar to our previously reported results for PAK peptides (31), titration of ␤PIX-SH3 with increasing amounts of peptide AIP4 206 -229 results in a blue shift in the fluorescence emission max (Fig. 2, D and E). However, unlike the binding of PAK or Cbl-b peptides, binding of AIP4 206 -229 also resulted in a large increase in fluorescence emission intensity (Fig. 2F). This effect was titratable, and fitting the fluorescence emission max generated an EC 50 of 6.3 M, which was similar to the binding affinity observed for the same AIP4 206 -229 peptide in our ITC studies (Fig. 2E). Fitting the increase in fluorescence emission intensity also yielded similar values (data not shown). We used this assay to test the preferential binding specificity for the ␤PIX-SH3 domain and found that the binding of the AIP4 206 -229 peptide could be displaced by a PAK-derived peptide (Fig. 2G). Taken together, the observations that the emission spectra for the ␤PIX-SH3 domain differ significantly for the binding of AIP4 206 -229 than either the PAK-or Cbl-b-derived peptides suggests that these peptides may be interacting differently with the SH3 domain, possibly exhibiting alternate modes of binding. We decided to pursue the crystal structure of AIP4⅐␤PIX-SH3 to better understand the molecular basis for these differences.
Structure Determination-Purified recombinant ␤PIX-SH3 was mixed with AIP4 206 -229 and crystallized using polyethylene glycol-MME 5000 by the vapor-diffusion method. The phases were solved by molecular replacement. The crystals were of space group P6 1 22, with one SH3 domain per asymmetric unit. The final model contained all of the residues of ␤PIX-SH3; there was an additional residue at the N terminus arising from the remnant of the GST fusion after cleavage by PreScission protease. In addition, the model contained residues 209 -224 of the AIP4 synthetic peptide; the C-terminal residue was modeled as Ala where density supporting the side chain was missing. There was one sulfate molecule and one glycerol molecule, as well as 69 solvent molecules in the model after multiple rounds of refinement to a final resolution of 2.0 Å (R free ϭ 23.5%). Refinement statistics are presented in Table 2.
Overall Structure of ␤PIX-SH3⅐AIP4 Peptide Complex-The structure of ␤PIX-SH3 with the AIP4 peptide showed a 1:1 complex of peptide ligand to protein. The structure of the SH3 domain itself is mainly composed of a five-stranded antiparallel ␤-sheet consistent with previous reports. The root mean squared deviation of the C␣ positions compared with previous ␤PIX-SH3 structures was ϳ0.55 Å. The model clearly showed the core residues that constitute the binding site in AIP4 are residues 217-223 (Fig. 3A), which are in a left-handed PPII helix configuration in the binding groove of ␤PIX-SH3 in a Class I orientation.
The binding surface of SH3 domains is characterized by several shallow pockets, which have been referred to as P Ϫ3 , P Ϫ1 , P 0 , P ϩ2 , and P ϩ3 (40). SH3 ligands adopt the secondary structure of a PPII helix, which has a triangular cross-section arising from the perfect helical repeat every three residues: residues along two of the edges interact with the surface of SH3 domains. Non-proline side chains on the first interacting edge point away from the surface, resulting in a poorer fit to the SH3 domain surface, and are referred to as external packing sites: the unique substitution of the backbone nitrogen with the delta carbon favors Pro at these positions (41). Residues on the second edge point toward the surface, so many residues can fit into the shallow pockets of the SH3 domain, and positions on the second edge are referred to as internal packing sites. The third edge points away from the SH3 domain and thus can normally accommodate any residue, although Pro residues are often found at these positions, likely stabilizing the PPII secondary structure. Thus the characteristic PXXP motif of SH3 ligands arises from the presence of Pro residues at two subsequent external packing sites along the PPII helix. The P 0 and P ϩ3 positions of Class I ligands are external packing sites and strongly favor Pro residues (41) as shown in our model of AIP4 FIGURE 1. The proline-rich region of AIP4 binds preferentially to the SH3 domain of ␤PIX. A, interaction of endogenous AIP4 and ␤PIX. ␤PIX was immunoprecipitated (IP) from the indicated cell lines with antibody specific for ␤PIX and subjected to SDS-PAGE followed by immunoblot (IB) analysis using antibody specific for Itch/AIP4 or ␤PIX (upper panels). Total cell lysate (TCL) was subjected to direct IB analysis to check for expression levels and control for gel loading (lower panels). B, GST pulldown assay. Lysates from HEK293T cells expressing either Myc-AIP4 or a control vector were subjected to GST pulldown with the SH3 domain of ␤PIX or GST alone, and the filter was blotted with anti-Myc to probe for AIP4 binding. The relative levels of GST and GST fusion protein are revealed following Coomassie staining of the same filter. The expression levels of AIP4 in the TCL are shown in the lower panel. C, the proline-rich region of AIP4 is required to co-immunoprecipitate ␤PIX in HEK293T cells. Top, schematic domain representations of the C-terminally HA-tagged ␤PIX, ␤PIX-SH3 GST fusion, and N-terminally Myc-tagged AIP4 constructs used in the pulldown and co-immunoprecipitation assays. For ␤PIX, the SH3 domain, Dbl-homology domain (DH), pleckstrin-homology domain (PH), G protein-coupled receptor kinase interactor binding domain (GBD), and leucine zipper (LZ) region are shown. For AIP4, the calcium binding C2 domain, the four WW domains, and the ubiquitin ligase HECT domain are shown. The deletion mutant Myc-⌬PolyPro-AIP4 lacks the proline-rich stretch of residues from 206 to 229. Bottom, HEK 293T cells were transfected with HA-tagged ␤PIX and Myc-tagged AIP4 or mutants as indicated, and the cell lysates were immunoprecipitated with anti-Myc antibody and subjected to SDS-PAGE followed by IB analysis. The filter was probed with anti-HA or anti-Myc antibodies to detect ␤PIX and AIP4, respectively (upper panels). TCL was subjected to direct IB analysis to check for expression levels and control for gel loading (lower panels). D, AIP4 binds to CIN85 possibly in competition with ␤PIX. HEK 293T cells were transfected with HA-tagged ␤PIX, Myc-tagged AIP4, and FLAG-tagged CIN85 as indicated, and the cell lysates were immunoprecipitated with anti-FLAG antibody and subjected to SDS-PAGE followed by IB analysis. The filter was probed with anti-Myc or anti-FLAG antibodies to detect AIP4 and CIN85, respectively (upper panels). TCL was subjected to direct IB analysis to check for expression levels and control for gel loading (lower panels). In cells transiently co-expressing AIP4 and CIN85, increasing amounts of ␤PIX expression abrogates the ability of CIN85 to co-immunoprecipitate AIP4.
bound to ␤PIX-SH3 (Figs. 3A and 4C). In addition there are three hydrogen bonds characteristic of SH3-ligand interactions observed between ␤PIX-SH3 side chains and three backbone carbonyl atoms from AIP4: at positions P 1 with Tyr-59, at P Ϫ3 with Trp-43, and a watermediated bond at P 0 with Asn-58 (Fig. 3, A and D). The B factor of that water molecule is 11.4, comparable to the main-chain atoms of ␤PIX-SH3. An Arg residue (Arg-217) interacts with a negatively charged "specificity pocket" at P Ϫ3 (Fig. 3, B  and D).
The N-terminal region of the AIP4 peptide contains a second PPII helix encompassing residues 209 -215 immediately adjacent to core SH3 ligand (Fig. 3A). The two helices are joined by a turn of 107°t hrough Ser-216. One anchor of this second PPII helix is Pro-215 of AIP4, which binds a hydrophobic cleft between Trp-43 and Trp-54 of ␤PIX-SH3. Pro-215 is on an external packing site of this PPII helix and would be uniquely favored for binding (41). Furthermore, two potential hydrogen bonds, between the backbone carbonyl atom of Pro-212 of AIP4 and Trp-54 of ␤PIX-SH3, and between the guanido group of Arg-211 of AIP4 and the carboxylate group of Glu-45 of ␤PIX-SH3, may help stabilize the ligand interactions. The total buried surface area of 1042 Å 2 is ϳ30% larger than is typical for other ligand⅐SH3 complexes (28). The extent of the interactions between the N-terminal PPII helix of AIP4 and ␤PIX-SH3 may contribute to the higher affinity observed for peptides AIP4 206 -229 and AIP4 211-226 , in the ITC experiments (Fig. 2, B and C).
Comparison with ␤PIX-SH3⅐Cbl-b and ␤PIX-SH3⅐PAK2 Complex Structures-The structure of ␤PIX-SH3 has previously been determined in a complex with two other peptide ligands; however, unlike AIP4, neither ligand contained a canonical PXXP motif, resulting in a kink in the PPII helix for these ligands. The structure of ␤PIX-SH3 with a peptide derived from Cbl-b is a heterotrimeric complex formed like a sandwich with the peptide in between two ␤PIX-SH3 domains in opposite orientations (28). We shall limit our comparison to the Class I configuration. The Cbl-b peptide did not contain a PXXP motif, and one of the external packing sites, P 0 , had a Lys residue instead of Pro (Fig. 4, A and C). Consequently, the PPII secondary structure, which ideally contains and angles of Ϫ78°and 149°, respectively, was perturbed at this position resulting in the C␣ and C␤ atoms of this Lys residue occupying positions similar to the backbone nitrogen and C␣ of the proline found in the AIP4 complex (Fig. 4C) (42). A further difference was found at the specificity pocket P Ϫ3 site, which for Cbl-b is occupied by Arg-904 on another face of a local surface protrusion, partly to accommodate the positive charge of the lysine at P 0 (Fig. 4A).
In comparing the present structure to that of the ␤PIX-SH3⅐PAK2 complex, the PAK2 peptide was in the opposite (Class II) orientation in the binding groove (Fig. 4, B and C) (31). In this orientation, the external packing sites are at P Ϫ1 and P 2 (41). Like the Cbl-b b motif, PAK2 lacks the expected PXXP motif, with an Ala in the P Ϫ1 position instead of Pro. Despite this, PAK2 is found to have an affinity to ␤PIX-SH3 similar to that of AIP4 (31). Once again there was a perturbation of the PPII helix at P 0 that appeared to allow the isoleucine residue at that position to interact more fully with the surface of ␤PIX-SH3 (Fig. 4, B and C). In this case the high affinity may also be attributed to interactions of an additional 11 residues of PAK2 that appear to be loosely coiled on the adjacent surface of ␤PIX-SH3 (31).

DISCUSSION
A growing body of evidence indicates that signaling cascades initiated by either GPCRs or RTKs are not mutually exclusive of one another. Cross-talk between receptor classes via downstream adaptors and effectors allows the cell to fine-tune and integrate signals (43). Modular protein domains mediate transient protein-protein interactions that are crucial to relaying and regulating signaling cascades (44). The ␤PIX-SH3 domain exemplifies a protein module at the nexus of many cellular processes. These include the down-regulation of EGF receptor through binding and sequestering Cbl as well as regulation of GPCRs through formation of ␤PIX-SH3⅐PAK complexes during leukocyte chemotaxis in response to CXCR4 activation (27). In the present study, we have identified the E3 ubiquitin ligase AIP4 as a binding partner of ␤PIX. Because AIP4 regulates CXCR4 signaling and ␤PIX is involved in EGF receptor downregulation, their association represents a potential new regulatory node for GPCR and RTK signal integration.
Our cellular binding assays demonstrate that ␤PIX can indeed interact with AIP4 and further show that this interaction is mediated by binding of the ␤PIX-SH3 domain to a proline-rich stretch of AIP4. Interestingly, while this interaction is readily detectable in MDA-MB-231 and NIH3T3 cells, it was undetectable in MCF7 cells under our assayed conditions. This raises the possibility that the localization and regulation of AIP4 and ␤PIX might be cell-type-specific as has been shown for PAK and PIX in different breast cancer cell lines (38). We found that AIP4 could also interact with CIN85 and that this interac-FIGURE 2. The ␤PIX-SH3⅐AIP4 complex is trimeric in solution. A, the AIP4-R217A mutant immunoprecipitates ␤PIX in HEK293T cells. HEK 293T cells were transfected with HA-tagged ␤PIX, and Myc-tagged AIP4 or AIP4-R217A as indicated, and the cell lysates were immunoprecipitated with anti-Myc antibody and subjected to SDS-PAGE followed by IB analysis with anti-HA to probe for ␤PIX binding as well as anti-Myc (upper panels). TCL was subjected to direct IB analysis to check for expression levels and control for gel loading (lower panels). B, AIP4-derived peptides used in ITC binding studies. Note that a Tyr residue has been added to the N terminus of the WT and R217A peptides to aid in concentration determination. The symbols preceding each peptide name are the same as those used on the ITC graph, and amino acid numbers are given as subscripts.   tion is abrogated upon increasing the cellular concentration of ␤PIX. Although we attribute this interaction to CIN85 SH3 domains, we cannot rule out the possibility that other parts of the proteins are involved in recognition and binding. Thus the ␤PIX-SH3 serves as a scaffold for numerous signal transduction proteins, including PAK, Cbl, Rac, and SAP and as the present report now shows, AIP4. How ligands may compete among themselves for the same SH3 domain as well as how all of these processes are regulated in the context of the full-length proteins merits further investigation.
Our ITC experiments demonstrate that AIP4 residues 206 -229, which encompass the proline-rich region, bind to ␤PIX with micromolar affinity and furthermore form heterotrimeric complexes, with one AIP4-derived peptide binding to two ␤PIX-SH3 domains. These experiments also reveal that residues outside the core PXXP motif are important for high affinity, and the crystal structure of the complex shows that an additional PPII helix outside the core is wrapped around another face of ␤PIX-SH3. This additional PPII helix is a novel feature for an SH3 ligand and to our knowledge has never been described by any of the SH3⅐ligand complex structures in the Protein Data Bank. Previous studies on ␤PIX-SH3⅐Cbl-b complexes showed that mutation of an Arg residue (analogous to Arg-217 on AIP4) reduced the complex to a monomeric interaction with a dramatically lower affinity. In contrast, our ITC results indicate that the AIP4-R217A mutant peptide does not alter complex stoichiometry. We speculate that additional interactions provided by the secondary PPII helix (including specific hydrophobic as well as salt-bridge interactions) with ␤PIX-SH3 are sufficient to overcome this substitution.
Although we have no direct structural evidence for the configuration of a heterotrimeric complex, we would propose that it might be similar to that observed for the Cbl-b⅐␤PIX-SH3 complex (28). In the case of Cbl-b, the peptide is the center of pseudosymmetry with the matching ␤PIX-SH3 laying on top of the Class I complex in the opposite orientation. Thus the same peptide becomes a Class II-oriented ligand with respect to the pseudosymmetry mate. The interactions between the two SH3 domains in that case were minimal. In our case, Arg-224 hypothetically would occupy the P Ϫ3 specificity pocket of another ␤PIX-SH3 molecule similarly arranged to SH3B of the Cbl-b complex (28). As in the case the Cbl-b, one of the pockets normally occupied by Pro would instead accommodate another residue: for our hypothetical heterotrimer, the P 1 pocket would be occupied by Thr-222 (Fig. 5), whereas P 2 would be occupied by Pro. In such an arrangement, Arg residues would occupy both specificity pockets, and Pro residues would occupy three of four Pro-favoring pockets. In examining the AIP4 and Cbl-b peptide sequence, we would propose a new heterotrimeric motif for ␤PIX-SH3 and other closely related SH3 domains such as those from CIN85. Given the inherent symmetry of PPII helices that allow for the Class I and II orientations, what appears to be distinct about the two peptides are the arginine residues anchoring the ends: the motif could thus be considered RXP(P/Z)X(P/Y)PR.
To prove this model, future experiments will be needed to further characterize the complex of AIP4 peptide and ␤PIX-SH3 and to resolve the discrepancy in the stoichiometry of the complex between the crystal structure and that which is consistent with the ITC results. One possible reason for this discrepancy might be due to the inherent asymmetry of the proposed complex. The proposed partner ␤PIX-SH3 domain of the heterotrimer would be limited to interactions with the core ligand, whereas the ␤PIX-SH3 domain observed in the crystal structure has additional contacts with the N-terminal PPII helix resulting in higher affinity to the peptide. Under our crystallization conditions with excess AIP4 peptide, it may be that the observed monomeric complex formed to a significant degree and that, due to favorable crystal packing interactions (Fig. 3C), it led to the formation of a lattice of the monomeric complex to the exclusion of the heterotrimeric complex. We aim to carry out further studies, including other biophysical methods to further characterize the complex of ␤PIX-SH3⅐AIP4 peptide, as well as crystallization trials under different conditions and with different AIP4-derived peptides. The full molecular understanding of how ␤PIX andAIP4 interact to influence their respective pathways will require eventual structural studies of the full-length proteins.
CXCR4 signaling is attenuated in part by AIP4, which mediates the ubiquitination and lysosomal sorting of the activated receptor (6). Notably, AIP4 itself contains four modular WW domains, and these domains often serve as platforms for assem- FIGURE 3. AIP4 binds to ␤PIX-SH3 as a Class I ligand. A, AIP4 peptide binds to a largely hydrophobic surface of ␤PIX between the RT (depicted in magenta) and n-src (depicted in cyan) loops. In the left-hand view the SH3 ligand-binding surface is seen from above with the peptide lying on the surface. In the right-hand view, the model has been rotated at on oblique angle to give a better view of the second PPII helix. Electron density for the peptide was seen for residues 209 -224, although an Ala was placed for the C-terminal Arg, because the side-chain density was not observed. There are two left-handed PPII helices, from residues 209 -215 and from 217-223. The second helix represents the core SH3 ligand with Arg-217 occupying the P Ϫ3 specificity pocket, and Pro-220 and Pro-223 occupying the obligate Pro-preferring P 0 and P 3 pockets. A better view of the first PPII helix is seen in the right-hand view. One Arg residue, Arg-214, lies near a negatively charged surface, Arg-211 appears to form a salt bridge with ␤PIX-SH3, and again two Pro residues, Pro-212 and Pro-215, are buried in shallow pockets on this surface of ␤PIX-SH3. In the asymmetric unit the N terminus of the peptide appears to point into empty space, but D shows that this end of the peptide interacts closely with a symmetry mate. The surface is colored to show the electrostatic potential as calculated in GRASP, with positive charge represented in blue and negative charge shaded in red. Underlying the surface is a schematic representation of the SH3 backbone. B, a representative section of the calculated electron 2f o Ϫ f c density map is shown at 1.2 in a stereo view. AIP4 peptide residues are represented with yellow carbons, whereas those from ␤PIX-SH3 are shown in green. Arg-217 of AIP4 is shown in this view in close proximity to Glu-24 of ␤PIX-SH3. C, the N-terminal region of the AIP4 peptide is shown in gray in this view with the symmetry mates as packed in the crystal. The corresponding ␤PIX-SH3 is shown as molecular surface also in gray with the protein atoms shown also in stick representation. The various symmetry mates are colored so that the corresponding peptide carbon atoms correspond to the color of the molecular surface. Several of the crystal contacts between ␤PIX-SH3 domains appear to be mediated by electrostatic interactions, whereas this region of AIP4 peptide appears to lie across a hydrophobic surface of a ␤PIX-SH3 symmetry mate. D, a schematic representation of ligand binding as generated by LIGPLOT. Hydrogen bonds with three AIP4 backbone carbonyl atoms, one of them water-mediated, are shown in the core SH3 ligand (residues 217-223). In addition, the second PPII helix also has one hydrogen bond with a backbone carbonyl, and Arg-211 forms a charge-stabilized hydrogen bond with ␤PIX-SH3. The peptide is shown with bonds in purple. ␤PIX-SH3 residues, which form hydrogen bonds with the peptide, are shown with yellow bonds, whereas Van der Waals interactions are depicted by half circles. There is one water-mediated hydrogen bond shown with a cyan sphere representing the water molecule.
bly of multiprotein networks (45). AIP4 is capable of binding Cbl family members via these WW domains, and importantly, this interaction facilitates AIP4-catalyzed ubiquitination and proteasomal degradation of Cbl (19,46). In doing so, AIP4 can impede Cbl-mediated ubiquitination of activated EGF receptors (19). In addition, it is also possible that AIP4 may, like Cbl, directly ubiquitinate ␤PIX or exert a cooperative effect on ␤PIX ubiquitination by regulating Cbl ligase activity. Indeed, others have shown that E3 ligases of the RING and HECT families may interact cooperatively with a shared common substrate to regulate signaling cascades (47,48). This process is exemplified by regulation of the NOTCH protein that can be ubiquitinated and regulated by both AIP4 and Cbl (49,50).
The association of AIP4 with ␤PIX may also regulate GPCR signaling. Following activation of C5a receptors, released G␤␥ proteins bind to PAK, which in turn forms a complex with ␣PIX that further initiates cascades resulting in cellular chemotaxis (51). We speculate that a similar process may occur following activation of CXCR4, with AIP4 being displaced from ␤PIX by G␤␥⅐PAK complexes. Liberated AIP4 could in turn regulate CXCR4 ubiquitination and endocytotic sorting. Our fluorescence experiments demonstrate competition between PAKand AIP4-derived peptides for the ␤PIX-SH3 domain and would support such a mechanism; however, further work is needed to examine this potential regulatory loop. Finally, we note that PIX is a binding partner for GPCR-kinase interacting proteins (GITs). GIT proteins serve as signaling scaffolds and FIGURE 4. The canonical binding of AIP4 to ␤PIX compared with PAK2 and Cbl-b. A, the binding of Cbl-b peptide (in magenta) in the Class I orientation is shown in comparison with AIP4 peptide (in yellow) after a least squares fit of ␤PIX-SH3 domains from the two models. Residues from ␤PIX that appear to form hydrogen bonds are shown under the molecular surface. The alignment of the respective core binding sequences from the peptides is shown in C. A Pro residue usually occupies the P 0 and P 3 positions in canonical Class I SH3 ligands. There is a distortion in the PPII helix at the Lys-907 of Cbl-b, shown in the center of the figure, resulting in a displacement of the backbone so that the C␣ and C␤ atoms of that residue occupy similar positions to the backbone N and C␦ positions of P220 of AIP4. B, PAK2 peptide (in cyan) binds to ␤PIX in the opposite Class II orientation, but again contains a perturbation from a PPII helix in a similar position to the Cbl-b peptide. Unlike typical Class II ligands, the PAK2 peptide contains an Ala residue in the P Ϫ1 position. In contrast to the Cbl-b peptide, this perturbation does not result in a residue making up for the lack of a Pro, but rather appears to allow extensive contacts between Ile-183 and the surface of ␤PIX. The high affinity observed for PAK2 by ITC may be due to the residues in the peptide lying outside of the SH3 ligand-binding surface. C, an alignment of the core binding regions of the three peptides is shown relative to the shallow hydrophobic binding pockets as sites P Ϫ3 to P ϩ3 as defined previously. The core sequences are residues 217-223, 904 -910, and 180 -186 for AIP4, Cbl-b, and PAK2, respectively. The sequences are color-coded to match panels A and B, with AIP4 in yellow, Cbl-b in magenta, and PAK2 in cyan. Underneath each residue are the respective / angles, with numbers in red where the PPII helix in the case of Cbl-b and PAK2 are distorted. The pockets, which favor the presence of Pro residues, are indicated by gray squares drawn around the residues. FIGURE 5. Proposed schematic arrangement of dimer interface of ␤PIX with AIP4 peptide. An alignment of the core binding regions of the AIP4 and Cbl-b peptides is shown relative to the shallow hydrophobic binding pockets as sites P Ϫ3 to P ϩ3 as defined previously in two orientations. Below the peptide sequences are the positions of the residues with respect to the structure reported in this report for the AIP4⅐␤PIX complex and for the Class I orientation for the Cbl-b⅐␤PIX complex. Above the peptide sequences are the same positions in the Class II orientation as described in the Cbl-b⅐␤PIX complex, and the proposed Class II orientation of an additional ␤PIX-SH3 domain based on the ITC data and an analysis of the sequence for AIP4. The proposed interaction of AIP4 in a class II orientation would result in Thr-222 occupying the P Ϫ1 site, which normally favors the presence of a Pro residue. The core sequences are residues 217-224 for AIP4 and 904 -911 for Cbl-b, respectively. Underneath each residue are the respective / angles, which show how the PPII helix in the case of Cbl-b and PAK2 are distorted. The pockets that favor the presence of Pro residues are indicated by gray squares drawn around the residues.
regulate numerous cellular processes, including cytoskeletal dynamics, membrane trafficking, and clathrin-dependent endocytosis of GPCRs (52). GIT and PIX form large multiprotein complexes within the cell, and this raises the possibility that a GIT⅐PIX⅐AIP4 complex may transiently exist in the cell perhaps regulating CXCR4 endocytosis and trafficking.
In conclusion, we have identified and defined the binding interaction between the E3 ubiquitin ligase AIP4 and the signaling scaffold protein ␤PIX. Given the extensive involvement of these proteins in GPCR and RTK signaling, their interaction represents a new node in GPCR/RTK signal integration. Future studies will aim to elucidate the role this interaction plays in receptor down-regulation and cancer tumor progression and metastasis.