Identification of VCP/p97, Carboxyl Terminus of Hsp70-interacting Protein (CHIP), and Amphiphysin II Interaction Partners Using Membrane-based Human Proteome Arrays*S

Proteins mediate their biological function through interactions with other proteins. Therefore, the systematic identification and characterization of protein-protein interactions have become a powerful proteomic strategy to understand protein function and comprehensive cellular regulatory networks. For the screening of valosin-containing protein, carboxyl terminus of Hsp70-interacting protein (CHIP), and amphiphysin II interaction partners, we utilized a membrane-based array technology that allows the identification of human protein-protein interactions with crude bacterial cell extracts. Many novel interaction pairs such as valosin-containing protein/autocrine motility factor receptor, CHIP/caytaxin, or amphiphysin II/DLP4 were identified and subsequently confirmed by pull-down, two-hybrid and co-immunoprecipitation experiments. In addition, assays were performed to validate the interactions functionally. CHIP e.g. was found to efficiently polyubiquitinate caytaxin in vitro, suggesting that it might influence caytaxin degradation in vivo. Using peptide arrays, we also identified the binding motifs in the proteins DLP4, XRCC4, and fructose-1,6-bisphosphatase, which are crucial for the association with the Src homology 3 domain of amphiphysin II. Together these studies indicate that our human proteome array technology permits the identification of protein-protein interactions that are functionally involved in neurodegenerative disease processes, the degradation of protein substrates, and the transport of membrane vesicles.

As protein-protein interactions (PPIs) 1 are central to most biological processes, their systematic identification is considered a key strategy for uncovering the complex organization principles of functional cellular networks (1)(2)(3). A number of experimental and computational techniques have been developed to determine all the potential and actual PPIs in selected model organisms (4 -8) and humans (9 -11).
Because protein arrays allow the parallel detection of addressable elements in a single experiment (12)(13)(14)(15)(16)(17)(18), they have been recognized as a promising technology for the identification of PPIs (19,20) or other specific biochemical activities (20 -23), and much effort has been devoted to their development in recent years. Zhu et al. (20) e.g. printed 5,800 purified yeast proteins onto coated glass slides and used them successfully in proof-of-principle experiments for the detection of novel calmodulin-and phospholipid-interacting proteins. Protein arrays were also used efficiently for identifying novel targets of protein kinases (24 -26). In principle, the production of protein arrays and their utilization for systematic large scale interaction and activity screens has proven viable. One major drawback, however, has been the necessity to use purified proteins. Protein purification is usually difficult, time-consuming, and expensive. For array-based studies, especially high throughput systematic screening (19), technologies are needed that permit the use of crude protein extracts.
Here we describe the design and application of an arraybased technology that allows the identification of PPIs with crude bacterial cell extracts containing recombinant human proteins. Using this approach, novel partners for the proteins valosin-containing protein (VCP), carboxyl terminus of Hsp70interacting protein (CHIP), and amphiphysin II were detected and confirmed with different independent binding assays, validating our strategy.
Sequence Analysis-Clones were subjected to DNA tag sequencing, starting from the 5Ј-ends of the cDNA inserts. After base calling and quality clipping of sequencer trace data (PHRED, trim_cutoff ϭ 0.05), 13,964 sequences were translated into protein sequences. BLASTP searches in the NCBI and TrEMBL (Swiss-Prot) protein databases were performed.
Overexpression of His-tagged Proteins for Array Production-For expression of His-tagged human fusion proteins, 384-well microtiter plates were filled with 40 l of TB medium/well (100 mM glucose, 100 g/ml ampicillin, and 15 g/ml kanamycin) and inoculated with 13,824 Escherichia coli clones from the hEx1 cDNA library. Cells were grown for 3 h at 37°C followed by induction of protein expression by addition of 40 l of 1 mM isopropyl thiogalactoside and overnight incubation at 30°C. Then cells were pelleted by centrifugation for 10 min at 4,000 rpm at 4°C and stored at Ϫ80°C.
Native Cell Lysis and Generation of Protein Arrays-Cell pellets were resuspended in 50 l of lysis buffer (1 mg/ml lysozyme, 0.5% Nonidet P-40, 1% Triton X-100, 1 mM PMSF, 25 units/ml benzonase, and 150 mM NaCl in PBS) and incubated for 90 min on ice. During the incubation, cell suspensions were mixed carefully every 10 min using a 384-pin replicator. Then the plates were centrifuged at 4,000 rpm at 4°C for 1 h to pellet cell debris. Supernatants were spotted onto 225 ϫ 225-mm nitrocellulose membranes (Protran BA 83, Schleicher & Schuell) using the K2-003 Gridder (KBiosystems). Each extract was spotted 20 times on each position in a 5 ϫ 5 duplicate gridding pattern.
Overlay Assay with GST-tagged Fusion Proteins-GST fusion proteins were expressed in E. coli as described previously (30). Crude E. coli cell extracts were used for overlay screens. Membranes with high density spotted His-tagged fusion proteins were blocked for 4 h at 4°C with PBS-T (PBS supplemented with 0.05% Tween 20) containing 0.5 mM DTT and 5% nonfat dry milk powder. Then they were rinsed with PBS-T and incubated overnight at 4°C in 100 ml of overlay buffer (PBS-T supplemented with 5% fetal calf serum, 100 M ATP, and 100 M GTP) to which crude bacterial cell extract containing the GST fusion protein (ϳ0.5 mg) had been added prior to use. After incubation, membranes were washed once for 10 min with PBS-TT (TT ϭ 0.05% Tween 20 ϩ 0.2% Triton X-100) and three times for 10 min each with PBS-T. Subsequently filters were incubated for 1 h at room temperature in 100 ml of PBS-T containing anti-GST HRPconjugated antibody (5,000-fold diluted, Amersham Biosciences). Membranes were rinsed twice with PBS-T, washed once with PBS-T for 15 min, and then washed three times for 5 min. Subsequently filters were incubated for 3 min in CHEMIGLOW solution (Alphainnotech) according to the manufacturer's instructions. Signals were detected using a Fuji luminescent image analyzer (LAS 1000 CH).
The resulting images were analyzed with Aida Image Analyzer version 3.21.001 (Raytest GmbH). The intensity of spots was determined using a spot diameter of 2,280 m if both spots of a respective duplicate gave a distinct signal in the image. Background correction of the average signal intensities of duplicates was performed blockwise. Interactions were considered positive when the relative signal of respective duplicates was greater than 3 times the standard deviation of average background signal.
In Vitro Protein-Protein Binding Assays-For pull-down experiments, ϳ20 g of GST fusion protein was immobilized on 100 l of glutathione-agarose beads (Sigma) and incubated with bacterial protein extracts containing 20 g of His-tagged fusion protein in IP buffer (50 mM Hepes, pH 7.4, 150 mM NaCl, 1.5 mM MgCl 2 , 1 mM EGTA, 20 mM NaF, 10% glycerol, 1% Nonidet P-40, and protease inhibitors) at 7°C for 4 h. Then beads were washed four times with IP buffer and incubated with 100 l of 4ϫ SDS gel loading buffer at 95°C for 5 min. Proteins were analyzed by SDS-PAGE and Western blotting using anti-GST (HRP conjugate, Amersham Biosciences) and anti-His (penta-His HRP conjugate, Qiagen) antibodies.
Mass Spectrometry-For identification, proteins of interest were separated by SDS-PAGE. Protein bands were excised from the gels and digested with trypsin. The peptide mixture was analyzed by peptide mass fingerprinting using a MALDI TofSpec2E mass spectrometer (Micromass). In case of equivocal results, the identification was performed with MS/MS sequencing. The MS/MS measurement was achieved with a nanoelectrospray hybrid quadrupole mass spectrometer (Q-TOF, Micromass). The Mascot software package (Matrix Science) was used for protein identification.
Yeast Two-hybrid (Y2H) Analysis-L40ccU MATa and L40cc␣ MAT␣ yeast strains were transformed with plasmids encoding bait and prey proteins, respectively (10). For interaction mating, the MAT␣ and MATa strains were individually mixed in 96-well microtiter plates, transferred onto YPD agar plates using the K2-003 spotting robot (KBiosystems) and incubated for 36 h at 30°C. After mating, the clones were transferred onto SDII (Leu Ϫ Trp Ϫ ) agar plates and grown for 50 h at 30°C to select for diploid yeast strains. For selection of interactions, diploid yeasts were spotted onto SDIV (Leu Ϫ Trp Ϫ Ura Ϫ His Ϫ where Ura is uracil) agar plates as well as on nylon membranes placed on SDIV agar plates. After 7 days of incubation at 30°C, the agar plates and nylon membranes were assessed for growth and ␤-galactosidase activity, respectively (10).
Ubiquitination Assays-Ubiquitination of caytaxin in vitro was performed essentially as described previously for Raf-1 (31). All reactions contained 5% of protein extract from bacteria expressing His-tagged caytaxin and 1.25 mg/ml of purified ubiquitin (Sigma). When indicated , and 3 M Hsc70 were added. Samples were incubated for 2 h at 30°C and subsequently analyzed by SDS-PAGE and immunoblotting using an anti-His antibody (hexa-His, Roche Applied Science).
Peptide Arrays-Two identical arrays were synthesized on a cellulose-(3-amino-2-hydroxy-propyl)-ether (CAPE) membrane using the Spot technology as described previously (32). Each array represents four scans of overlapping 15-mer peptides with a shift of three amino acids representing the interaction partners (Table I). Probing the array toward binding to amphiphysin II was performed as described previously (32) except that the peroxidase-labeled anti-GST antibody (RPN1236, Amersham Biosciences) was used for detection. To exclude false positive results in the binding experiment, one array was pre-examined with GST/anti-GST antibody. All spot signal intensities (Boehringer light unit) were measured on a Lumi-Imager TM (Roche Diagnostics) and evaluated by using the software Genespotterா (Microdiscovery).

Protein Arrays and Overlay
Assays-To produce protein arrays for interaction studies, a human fetal brain cDNA library (hEx1) allowing the expression of His-tagged fusion proteins was used (27). cDNA inserts of 13,824 expression plasmids were sequenced at their 5Ј-ends, and the sequence information was stored in a database. BLASTP searches in protein databases (33) and bioinformatic analysis of sequence data revealed that about 64% of the cDNA fragments encoded human proteins in the correct reading frame, 19% of which were full-length.
For production of protein arrays, the E. coli clones were expressed in 384-well microtiter plates by addition of isopropyl thiogalactoside to the culture medium, and cells were lysed under non-denaturing conditions using lysozyme. After removal of insoluble cell debris by centrifugation, crude bac-terial protein extracts were double gridded onto membrane filters using a robot. The resulting high density spotted membranes (225 ϫ 225 mm in size) containing 13,824 protein extracts were dried and stored at 4°C prior to protein interaction screens.
To determine how many of the protein samples on each membrane contained a recombinant His-tagged human fusion protein, the filter membranes were first probed with an anti-His antibody. Image analysis revealed that ϳ22% of the expression clones produced a detectable fusion protein. This is in agreement with previous studies demonstrating that ϳ20% of the hEx1 library clones express a soluble human protein (34). In total, about 3,000 His-tagged recombinant proteins were accessible for interaction studies on each high density spotted filter membrane.
To identify PPIs, the protein arrays were used for proof-ofprinciple overlay experiments with the human proteins VCP, CHIP, and amphiphysin II (Table I). GST-tagged fusions were expressed in E. coli, and after native lysis with lysozyme, crude protein extracts were prepared and incubated overnight with the high density spotted filter membranes (Fig. 1). After extensive washing, PPIs were detected by ELISA using an anti-GST antibody. Specific double spots that gave signals above background were detected with the relevant GST fusion protein (Supplemental Figs. 1-3) but not with the control protein GST (data not shown). For each GST fusion protein three independent overlay experiments were performed, all of which essentially gave the same results (data not shown). The identity of the His-tagged proteins detected by overlay screening was determined using image analysis tools directly linked to our sequence database. In addition, interacting His- tagged proteins expressed in E. coli were affinity-purified with nickel-nitrilotriacetic acid-agarose beads and analyzed by MS (Table I).
Identification and Characterization of VCP Protein-Protein Interactions-VCP, a member of the AAA ATPase family, is involved in a variety of cellular processes including membrane fusion (35), the degradation of proteins by the ubiquitin-proteasome system (36), and the pathogenesis of protein misfolding diseases (37,38). Functionally active VCP molecules assemble into hexameric barrels (39) and bind to cofactors, such as p47 (35), which determine the specific role of VCP in the cell. Although adaptors and cofactors have been found (35,40,41), identification of additional VCP binding partners is critical to elucidate its function in distinct subcellular processes.
Overlay experiments with VCP (Supplemental Fig. 1) led to the identification of the well known interaction partner p47 (42) ( Table I). Also an association with AMFR, only found very recently to interact with VCP (43), was detected (Table I). Therefore, we focused on the validation of the VCP/AMFR interaction in more detail using binding experiments, co-immunoprecipitations, and co-localization studies (Fig. 2). Fig.  2A shows the selective enrichment on the GST-VCP affinity matrix of a 35-kDa carboxyl-terminal His-tagged AMFR fragment from a crude E. coli protein extract. In strong contrast, no AMFR protein was bound to the matrix with the control protein GST. The identity of the affinity-purified AMFR protein was determined by mass spectrometry (Fig. 2B) as well as immunoblotting using specific anti-AMFR or anti-His antibodies (data not shown).
The interaction between VCP and AMFR was also detected by co-immunoprecipitation experiments using transiently transfected COS-1 cells co-expressing full-length VCP and an amino-terminally truncated HA-tagged AMFR fragment. As shown in Fig. 2C, a protein complex containing VCP/HA-AMFR was precipitated with the anti-HA antibody but not with the anti-Ku70 antibody (negative control), indicating that both proteins associate with each other in mammalian cells.
The interaction between VCP and HA-AMFR was also confirmed by indirect immunofluorescence microscopy in transiently transfected COS-1 cells (Fig. 2D). Co-localization of VCP, labeled with CY3 (red) and AMFR, labeled with Alexa 488 (green), is indicated by the yellow color when the cell is observed under a triple filter. Both proteins co-localized in the cytoplasm as well as in membranous structures in the perinuclear region of COS-1 cells (Fig. 2D), supporting the results of the pull-down and co-immunoprecipitation assays.
Finally we mapped the regions critical for the protein-protein interaction in VCP and AMFR using GST pull-down assays (Fig. 2, E and F). We found that the amino-terminal region of VCP (aa 1-199) is necessary and sufficient for the association with His-AMFR in vitro, whereas the carboxyl-terminal AAA domains D1 and D2, important for ATP binding and hydrolysis (39), are not required for the protein-protein interaction (Fig. 2, E and F). The VCP binding region in AMFR was mapped downstream of the Cue domain to the carboxyl terminus of the protein (aa 498 -643) (Fig. 2, G and H). Thus, an amino-terminal region in VCP associates with the carboxyl terminus of AMFR in vitro.
Identification and Verification of CHIP Protein-Protein Interactions-The co-chaperone CHIP, which is known to interact with Hsp70, Hsc70, and Hsp90 (29,30), was also selected as a probe for the overlay experiments because recent studies have demonstrated that it also functions as an E3 ubiquitin ligase and facilitates the polyubiquitination of chaperone substrates. However, the mode of action as well as the interaction partners of CHIP in this process are largely unclear.
As foreseen, the well characterized interaction partners Hsp70, Hsc70, and Hsp90 were detected by the GST-CHIP overlay assay (Table I and Supplemental Fig. 2). We identified nine different expression clones for Hsc70 and four for Hsp70. Four clones expressing fragments of Hsp90 were detected. In agreement with previous studies, the CHIP binding region was mapped to the carboxyl terminus in Hsp70, Hsc70, and Hsp90 (data not shown). In addition to the known interactions, an association of CHIP with the microtubule-associated protein Tau was identified by the overlay screening (Table I).
Previous studies have demonstrated that CHIP forms a protein complex with Tau in mammalian cells (44). However, a direct association has not been described. Here we show that CHIP interacts with a domain at the carboxyl terminus of Tau (aa 685-756) that is located downstream of the tubulin-bind- ing domains (aa 559 -684). We also found that the uncharacterized ubiquitin-conjugating enzyme E2Q as well as the disease protein caytaxin (45) interact with CHIP in the overlay screen (Table I).
To confirm the CHIP interactions, pull-down as well as Y2H assays were performed (Fig. 3, A and B). For pull-down experiments GST-CHIP fusion protein was expressed in E. coli and bound to agarose beads. Then the immobilized beads were incubated with bacterial protein extracts containing the respective His-tagged interaction partners. After extensive washing, bound His-tagged proteins were detected by SDS-PAGE and immunoblotting using an anti-His antibody. As shown in Fig. 3A the recombinant proteins Tau, Hsc70, Hsp70, Hsp90, E2Q, and caytaxin interact with GST-CHIP in pull-down assays but not with the control protein GST or the affinity matrix.
For the two-hybrid analysis, the cDNA fragments encoding His-and GST-tagged fusions were subcloned into DNA-binding domain and activation domain vectors, and the resulting plasmids were transformed into MATa and MAT␣ yeast strains for interaction mating (10). Diploid yeast clones were then selected for growth and ␤-galactosidase activity (10). Except the Tau/CHIP interaction, all PPIs could be confirmed by the two-hybrid assays (Fig. 3B), indicating that protein complexes not only form in vitro but also in yeast. We suggest that failure to detect the Tau/CHIP PPI could be due to insufficient translocation of the microtubule-binding protein Tau into the nucleus, which is a prerequisite for reporter gene activation.
Functional Analysis of the CHIP/Caytaxin Interaction-Previous studies have shown that CHIP can function as an E3 ligase (31,44). Using a cell-free assay (Fig. 3C), we tested whether the protein can stimulate the ubiquitination of poten-tial interaction partners such as caytaxin. Caytaxin was incubated for 2 h with CHIP, UbcH5b (E2), and E1 in vitro, and the formation of higher molecular weight ubiquitinated caytaxin molecules was monitored by Western blotting. As shown in Fig. 3C, CHIP efficiently stimulated the ubiquitination of caytaxin, indicating that the protein is a substrate for the E3 ligase activity of CHIP and that the interaction between both proteins could be important for regulation of caytaxin turnover in vivo. Surprisingly caytaxin ubiquitination proceeded with similar efficiency in the absence or presence of Hsc70 (Fig. 3C). This is in marked contrast to findings obtained for other known substrates of the CHIP ubiquitin ligase, the ubiquitylation of which was highly dependent on the presence of Hsc70 (31,44). Caytaxin is the first substrate of CHIP identified to be directly recognized by the ubiquitin ligase without an involvement of Hsc70.
Identification of Amphiphysin II Interaction Partners and Mapping of Potential Binding Domains-Using the conserved carboxyl-terminal Src homology 3 (SH3) domain (aa 497-593) as a probe for overlay screens, we next searched for amphiphysin II interaction partners (Supplemental Fig. 3). Amphiphysin II belongs to a family of adaptor proteins that regulates synaptic vesicle endocytosis in neurons and binds e.g. to the protein dynamin (46). In addition, experimental evidence has been presented showing isoforms of amphiphysin II to associate with nuclear proteins and exhibit tumor suppressor and proapoptotic properties (47). Hence amphiphysin II may have multiple functions in distinct subcellular processes and tissues.
Overlay experiments with the SH3 domain of amphiphysin II allowed the identification of the known interaction partner dynamin (Table I). In addition, the proteins Discs large-associated protein 4 (DLP4), XRCC4 (DNA repair protein), and

FIG. 2-continued
Interaction Partners for VCP/p97, CHIP, and Amphiphysin II fructose-1,6-bisphosphatase (FBP) ( Table I) were identified as novel amphiphysin II interaction partners and verified by in vitro pull-down and two-hybrid assays (Fig. 4, A and B).

DISCUSSION
Here we designed and applied a protein array-based technology for the parallel screening of potential human PPIs. The method has several advantages. Unpurified recombinant human proteins, produced in bacteria under non-denaturing conditions in 384-well plates, can be used for interaction screens. The screening procedure is simple and sensitive, does not require radioactivity or expensive equipment, and is therefore particularly useful for systematic high throughput interaction studies of human proteins.
We applied our technology in poof-of-principle experiments to identify novel interaction partners for the human proteins VCP, CHIP, and amphiphysin II ( Table I). The hexameric VCP protein mainly functions as an unfoldase in membrane fusion events (51) as well as in the transport and degradation of were selected on SDIV (Trp Ϫ Leu Ϫ His Ϫ Ura Ϫ ) agar plates. In parallel, yeast colonies were spotted onto nylon membranes for detection of ␤-galactosidase activity (blue staining). C, caytaxin is a substrate for the CHIP ubiquitin ligase. Lysates from caytaxin-expressing bacteria were subjected to in vitro ubiquitination reactions in the presence of purified components of the CHIP ubiquitination machinery. Detection of caytaxin and its ubiquitinated forms was performed by immunoblotting using a hexahistidine-specific antibody. The asterisk denotes a bacterial polypeptide that cross-reacts with the anti-His antibody. ub, ubiquitin.

FIG. 3. Verification of CHIP PPIs identified by overlay screens.
A, in vitro binding assays. GST-tagged proteins bound to glutathione-Sepharose beads were incubated with bacterial protein extracts containing His-tagged human fusion proteins. After SDS-PAGE and immunoblotting, bound proteins were detected with anti-His antibody. B, confirmation of interactions by Y2H assays. MATa and MAT␣ yeast strains expressing pairwise combinations of LexA and GAL4 fusions were mated on YPD plates, and yeast strains containing interactions ubiquitinated proteins (40,43). Our approach allowed the identification of two known VCP interactors, p47 (35) and AMFR (43), which were both validated using a comprehensive set of in vitro and in vivo binding experiments. Using pulldown assays, the binding regions for the proteins VCP and AMFR were mapped. Interestingly the carboxyl terminus of AMFR is necessary and sufficient for the interaction with the amino terminus of VCP. Further studies are needed to address the question whether AMFR targets VCP to endoplasmic reticulum membranes in vivo and modulates the degradation of polyubiquitinated substrates.
CHIP is a co-chaperone that associates with heat shock proteins such as Hsp70 and Hsp90 (52,53). Moreover it functions as an E3 ligase to polyubiquitinate target proteins (31,44). The overlay screen revealed that CHIP interacts with Hsp70, Hsc70, and Hsp90 as well as with the microtubuleassociated protein Tau, confirming previous studies (44,52). Moreover novel interactions with the proteins caytaxin and E2Q were identified. Recent studies have demonstrated that mutants in caytaxin cause Cayman disease, an inheritable recessive ataxia restricted to the area of Grand Cayman Island (45). Currently the molecular mechanism causing Cayman ataxia is completely unclear. Our findings that caytaxin associates with CHIP suggests that it might be a substrate for CHIP-mediated ubiquitination. To test this hypothesis in vitro, ubiquitination assays were performed indeed showing that CHIP stimulates the polyubiquitination of caytaxin. Interestingly the in vitro ubiquitination was not dependent on the presence of Hsc70. This finding substantiates that the identified CHIP/caytaxin interaction is direct and not mediated by a chaperone. These results suggest that CHIP may influence the ubiquitination rate and turnover of caytaxin in neurons of healthy individuals as well as Cayman disease patients. The novel CHIP-interacting protein E2Q is a potential ubiquitinconjugating enzyme (E4) not yet characterized. To date, the functional relevance of the CHIP/E2Q interaction is unknown. However, it seems reasonable to speculate that CHIP may use E2Q for efficient ubiquitination of certain target proteins. FIG. 4. Verification of amphiphysin II PPIs identified by overlay screens. A, in vitro binding assays. B, yeast twohybrid assays. C, screening of proteinpeptide interactions. Overlapping 15-mer peptides with a shift of three amino acids representing the proteins DLP4, XRCC4, and FBP were synthesized on membranes and incubated with the GST-SH3 domain of amphiphysin II (aa 497-593) to identify interacting amino acid sequences. Protein-peptide interactions were detected with anti-GST antibody (black spots). D, schematic representation of the binding regions in the proteins DLP4, XRCC4, and FBP (boxes; aa positions are indicated by numbers). The known interaction motif PXXP (49) and Arg/Lys-rich sequences are indicated. E, sequences of the GST-amphiphysin IIbinding peptides in the indicated regions of the proteins DLP4 and XRCC4. Sequences with the PXXP motif (49) in DLP4 as well as Arg/Lys-rich sequences in XRCC4 are assigned in red.
Using the SH3 domain (aa 497-593) as a probe, we identified three novel partner proteins for the synaptic protein amphiphysin II by protein array overlay screening. These interactions could be validated by pull-down and Y2H assays. For example, an association between the SH3 domain with the Discs large-associated protein DLP4 was established in this study. DLP4 has been shown to function in membrane morphogenesis and endocytosis in the postsynaptic density of neurons. As amphiphysin II co-localizes with the marker Discs large (Dlg) in Drosophila (54,55) and functions in vesicle transport processes in postsynaptic densities, we suggest that a functional DLP4⅐amphiphysin II protein complex influencing synaptic activities could form in neurons.
Two other novel interactions of the SH3 domain of amphiphysin II, namely with the DNA double strand break repair protein XRCC4 and FBP, were detected by array screening. XRRC4 is localized in the nucleus and critical for non-homologous end joining of broken chromosomal DNA fragments, whereas FBP is crucial for glyconeogenesis, the production of glucose from non-carbohydrate precursors such as pyruvate in the cytosol (55). Whether amphiphysin II, via interactions, can modulate XRCC4 or FBP activity needs to be analyzed.
It is generally thought that SH3 domains associate with short proline-rich peptide motifs in their target proteins (56). Therefore, we mapped the binding regions in DLP4, XRCC4, and FBP using peptide arrays (32). Interestingly we found that the SH3 domain of amphiphysin II not only interacts with proline-rich sequences of the known core PXXP element (49) but also with Arg/Lys-rich sequences or peptides that do not contain an obvious interaction motif. This suggests that different types of amino acid sequences can be recognized by the SH3 domain of amphiphysin II; this might be critical for the formation of PPIs in vitro.
Jointly these results indicate that our array-based screening approach permits the identification of novel PPIs potentially involved in endocytosis, protein folding, and neurodegeneration. However, this technology might also be applicable for the systematic screening of human protein interactions in general as well as for the detection of protein-DNA and protein-drug associations.