A Novel AP-2 Adaptor Interaction Motif Initially Identified in the Long-splice Isoform of Synaptojanin 1, SJ170*

Phosphoinositides play a fundamental role in clathrin-coat assembly at the cell surface. Several endocytic components and accessory factors contain independently folded phosphoinositide-binding modules that facilitate, in part, membrane placement at the bud site. As the clathrin-coat assembly process progresses toward deeply invaginated buds, focally synthesized phosphoinositides are dephosphorylated, principally through the action of the phosphoinositide polyphosphatase synaptojanin 1. Failure to catabolize polyphosphoinositides retards the fission process and endocytic activity. The long-splice isoform of synaptojanin 1, termed SJ170, contains a carboxyl-terminal extension that harbors interaction motifs for engaging several components of the endocytic machinery. Here, we demonstrate that in addition to DPF and FXDXF sequences, the SJ170 carboxyl terminus contains a novel AP-2 binding sequence, the WXXF motif. The WXXF sequence engages the independently folded α-subunit appendage that projects off the heterotetrameric AP-2 adaptor core. The endocytic protein kinases AAK1 and GAK also contain functional WXX(FW) motifs in addition to two DPF repeats, whereas stonin 2 harbors three tandem WXXF repeats. Each of the discrete SJ170 adaptor-interaction motifs bind to appendages relatively weakly but, as tandemly arrayed within the SJ170 extension, can cooperate to bind bivalent AP-2 with good apparent affinity. These interactions likely contribute to the appropriate targeting of certain endocytic components to clathrin bud sites assembling at the cell surface.

At the nerve terminal, compensatory endocytosis preserves synaptic architecture in the face of ongoing calcium-dependent exocytotic release of neurotransmitter from synaptic vesicles (19). Synaptic vesicle membrane is re-internalized en masse within clathrin-coated vesicles (20) and, as clathrin-coated intermediates upon the plasma membrane progress toward deeply invaginated buds, PtdIns(4,5)P 2 is focally hydrolyzed. The hydrolysis is partly necessary to promote release of the phosphoinositide binding factors such as AP180 and epsin from the budded vesicle (21). A major inositol 5Ј-phosphatase in brain is synaptojanin 1 (21,22). The protein, which is subject to alternative splicing, is abundant at the synapse where it colocalizes significantly with clathrin-associated components (16,23). Targeted disruption of the synaptojanin 1 gene in mice is lethal shortly after birth (21). Brain slices and cultured hippocampal neurons derived from synaptojanin nullizygous animals are slow to recover from high-frequency stimulation, and ultrastructural analysis shows that massed, deeply invaginated clathrin-coated buds and coated vesicles accumulate at the periphery of the presynaptic active zone (21,24). It thus appears that in the absence of phosphoinositide hydrolysis, vesicle release is delayed and the pool of synaptic vesicles becomes greatly diminished. A generally analogous phenotype is seen in mutant Caenorhabditis elegans lacking functional expression of the synaptojanin orthologue UNC-26 (25). In neither case, however, is there a complete arrest of clathrin-dependent endocytosis (21,25). Indeed, disruption of synaptojanin-like phosphatases in S. cerevisiae leads to the inappropriate appearance of PtdIns(4,5)P 2 upon intracellular structures (26), showing that hydrolysis is not a prerequisite for internalization. Synaptojanin 1 is composed of two tandemly arrayed catalytic phosphatase domains, the SacI and inositol 5Ј-phosphatase homology domains, followed by a carboxyl-terminal proline-rich domain (23,27). The central inositol 5Ј-phosphatase domain dephosphorylates PtdIns(4,5)P 2 to phosphatidylinositol 4-phosphate. The proximal SacI homology domain is also a phosphatidylinositol phosphatase (28) that preferentially hydrolyzes phosphatidylinositol 3-phosphate or phosphatidylinositol 4-phosphate. Thus, synaptojanin 1 is a phosphoinositide polyphosphatase that can regenerate phosphatidylinositol by sequential dephosphorylation reactions. In humans, the gene for synaptojanin 1 is located on chromosome 21 (29). There is some evidence for increased expression in Down's syndrome (30,31) and possible connection to 21q22-linked bipolar disorder (32).
The relatively abundant 145-kDa synaptojanin 1 isoform (SJ145) is thought to be recruited to the clathrin lattice on the presynaptic plasma membrane in the brain by the interaction of the carboxyl-terminal proline-rich sequences with the Src homology 3 (SH3) domains of endocytic proteins like endophilin or amphiphysin (33,34). In this study, we show that the nonneuronal 170-kDa long splice isoform of synaptojanin 1, termed SJ170, harbors a novel AP-2 adaptor-binding element absent from SJ145. This interaction determinant engages the ␣-subunit appendage and likely promotes efficient targeting of the polyphosphatase to the assembling clathrin lattice in non-neuronal cells. A functionally analogous WXX(FW) sequence is also found in the endocytic protein kinases AAK1 and GAK/auxilin 2, and multiple copies in stonin 2 promote AP-2 association.
Protein and Tissue Extract Preparation-GST and the various GST fusion proteins were produced in Escherichia coli BL21 cells. The standard induction protocol entails shifting log-phase cultures (A 600 ϳ 0.6) from 37°C to room temperature and then adding isopropyl-1-thio-␤-Dgalactopyranoside to a final concentration of 100 M. After 3-5 h at room temperature with constant shaking, the bacteria were recovered by centrifugation at 15,000 ϫ g max at 4°C for 15 min and stored at Ϫ80°C until used. Bacteria were lysed on ice in 50 mM Tris-HCl, pH 7.5, 300 mM NaCl, 0.2% (w/v) Triton X-100, 10 mM ␤-mercaptoethanol with sonication or in B-PER reagent (Pierce). Insoluble material was removed by centrifugation at 23,700 ϫ g max at 4°C for 15 min and then the GST fusions were collected on GSH-Sepharose. After extensive washing in phosphate-buffered saline, GST fusions were eluted with 10 mM Tris-HCl, pH 8.0, 10 mM GSH, 5 mM dithiothreitol on ice and dialyzed into phosphate-buffered saline, 1 mM dithiothreitol before use in binding assays. Several of the purified fusion proteins were cleaved from the GST with thrombin (Amersham Biosciences) while still immobilized upon GSH-Sepharose. Digestion was as recommended by the manufacturer, followed by addition of the irreversible thrombin inhibitor D-Phe-Pro-Arg chloromethyl ketone (Calbiochem) to a final concentration of 25 M.
Cytosol was prepared from frozen rat brain (PelFreez) by sequential differential centrifugation after homogenization in 25 mM Hepes-KOH, pH 7.2, 250 mM sucrose, 2 mM EDTA, and 2 mM EGTA supplemented with 1 mM phenylmethylsulfonyl fluoride and Complete (Roche Diagnostics) protease inhibitor mixture. The 105,000 ϫ g max supernatant is defined as cytosol and was stored in small aliquots at Ϫ80°C. Undifferentiated PC12 cells were grown at 37°C in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and 2 mM L-glutamine. Lysates were prepared after collecting the cells by trypsinization and washing with phosphate-buffered saline. Pelleted PC12 cells were solubilized in 25 mM Hepes-KOH, pH 7.2, 250 mM sucrose, 2 mM EDTA, and 2 mM EGTA supplemented with 1% Triton X-100 on ice for 30 min in the presence of Complete (Roche) protease inhibitor mixture and 1 mM phenylmethylsulfonyl fluoride. Following centrifugation at 20,000 ϫ g max at 4°C for 15 min, aliquots of the lysate were stored frozen at Ϫ80°C. Before use, thawed samples of either rat brain cytosol or PC12 cell lysates were adjusted to 25 mM Hepes-KOH, pH 7.2, 125 mM potassium acetate, 5 mM magnesium acetate, 2 mM EDTA, 2 mM EGTA, and 1 mM dithiothreitol (assay buffer) by addition of a ϫ10 stock and then centrifuged at 245,000 ϫ g max (TLA-100.4 rotor) at 4°C for 20 min to remove insoluble particulate material.
Binding Assays-Pull-down type assays, in 300 l total volume, were as described (36,39). Typically, 50 -400 g of GST and the GST fusion proteins were first each immobilized upon ϳ25 l of packed GSH-Sepharose by incubation at 4°C for 2 h with continuous mixing. The Sepharose beads containing the required immobilized proteins were then washed and resuspended to 50 l in assay buffer. Clarified rat brain cytosol, PC12 cell lysates, or purified, thrombin-cleaved ␣ C or ␤2 appendage (in the presence of 0.1 mg/ml carrier bovine serum albumin) were added and the tubes were incubated at 4°C for 60 min with continuous gentle mixing. For the competition assays, thrombincleaved proteins were added directly into the assay mixture to a final concentration of 20 M in the presence of 25 M D-Phe-Pro-Arg chloromethyl ketone, an irreversible thrombin inhibitor. The GSH-Sepharose beads were then recovered by centrifugation at 10,000 ϫ g max at 4°C for 1 min and an aliquot of each supernatant was removed and adjusted to 100 l with SDS sample buffer. After washing the GSH-Sepharose pellets 4 times each with ϳ1.5 ml of ice-cold phosphatebuffered saline by centrifugation, the supernatants were aspirated and each pellet resuspended in SDS sample buffer.
Electrophoresis and Immunoblotting-Samples were resolved on polyacrylamide gels prepared with an altered acrylamide:bis-acrylamide (30:0.4) ratio stock solution. The decreased cross-linking generally improves resolution but also affects the relative mobility of several proteins, most noticeably AP180 and epsin 1. After SDS-PAGE, proteins were either stained with Coomassie Blue or transferred to nitrocellulose in ice-cold 15.6 mM Tris, 120 mM glycine. Blots were usually blocked overnight in 5% nonfat milk in 10 mM Tris-HCl, pH 7.8, 150 mM NaCl, 0.1% Tween 20, and then portions were incubated with primary antibodies as indicated in the individual figure legends. After incubation with horseradish peroxidase-conjugated anti-mouse or antirabbit IgG, immunoreactive bands were visualized with enhanced chemiluminescence.

Selective Binding of SJ170 to the AP-2 ␣ Appendage-SJ170
is expressed at much lower levels in peripheral tissues than is SJ145 in the central nervous system. Using antibodies against an epitope common to both isoforms, SJ145 is easily detectable in rat brain extracts, whereas the abundance of SJ170 is below the level of detection in extracts of adult brain, liver, lung, testis, or heart (40). The SJ170 isoform only becomes detectable after enrichment with an SH3 domain-binding partner like Grb2 or amphiphysin (40). Consequently, it has been proposed that the alternatively spliced carboxyl-terminal extension in SJ170 facilitates efficient recruitment to endocytic buds forming at the cell surface (41). Indeed, in extracts from undifferentiated PC12 cells, which contain both SJ145 and SJ170, only the SJ170 isoform associates with immobilized GST-␣ C appendage in pull-down type assays (Fig. 1, lane d). Importantly, SJ170 does not sediment with immobilized GST alone (lane b), whereas the more abundant SJ145 isoform remains in the soluble fraction under both conditions (lanes a and c). Similar pull-down experiments utilizing rat brain cytosol, with a much higher SJ145 concentration, still shows no interaction with the ␣ C appendage (data not shown). The enrichment of SJ170 in the GST-␣ C appendage pellet fraction is similar to other known endocytic accessory proteins, including amphiphysin, AP180, and epsin (lane d). By contrast, both SJ145 and SJ170 bind to the SH3 domain of Grb2 (lane f), as reported (40). Therefore, the failure to detect SJ145 in the GST-␣ C appendage pellet fraction rules out the possibility that SJ170 is being recruited by the SH3 domain of appendage-bound amphiphysin. These results underscore the importance of the unique SJ170 extension in precise intracellular targeting.
A Third AP-2 Binding Sequence within the SJ170 Extension-The alternatively spliced segment unique to SJ170 contains two known types of AP-2 interaction motif, two DPF triplets ( 1323 DPF and 1557 DPF) and a 1463 FXDXF motif. We have shown previously that a carboxyl-terminal portion of SJ170 devoid of both DPF motifs (GST-SJ170C2, Fig. 2) binds to AP-2 with good apparent affinity (37). Yet mutation of the FXDXF sequence to AXAXF only reduces AP-2 association marginally (37). Therefore, to determine whether an additional AP-2 adaptor-binding element is present within the SJ170C2 (residues 1454 -1530), we sequentially truncated the carboxyl end in the context of a GST fusion protein. Constructs containing only residues 1454 -1497 (GST-SJ170C5) or 1454 -1489 (GST-SJ170C5.5) bind to AP-2 efficiently (Fig. 2). However, further removal of 7 amino acids (GST-SJ170C6, residues 1454 -1482) abolishes the stable interaction with the AP-2 adaptor complex near completely. This identifies residues 1482-1489 as a region that contributes to AP-2 adaptor interactions.
This localized tract of SJ170 (residues 1482-1489) could either harbor an independent AP-2 binding motif or contribute structurally to the optimal presentation of the proximal FX-DXF sequence. To distinguish between these two possibilities, we fused a restricted region of the carboxyl terminus of SJ170 to GST and assessed adaptor engagement in vitro using rat brain cytosol. Importantly, appending only 12 amino acids of the SJ170 sequence, 1478 SNPKGWVTFEEE, to GST (GST-SJ170M1) facilitates a substantial AP-2 interaction (Fig. 3A, lane d compared with lane b). The extent of adaptor binding to the GST-SJ170M1 (lane d) is lower than that seen with a larger fusion that also contains the FXDXF sequence (lane j) but is clearly above that seen with the FXDXF sequence alone (lane l). No AP-2 associates with GST alone under the same assay conditions (lane b). This clearly validates that the mapped region houses a novel AP-2 interaction motif that is both autonomous and transplantable. For comparison, when similarly fused to GST, a YXXØ-type endocytic internalization sequence, ASSYKYSKVNKE (derived from the cation independent mannose 6-phosphate receptor), is completely incapable of associating with AP-2 (Fig. 3A, lane h). This is likely because of the fact that AP-2 cycles through active and inactive conformations (42)(43)(44) and the cytosolic pool of the adaptor complex likely has the 2 subunit in a closed, binding-incompetent conformation (6). Together with the truncation data, our experiments firmly localize a novel AP-2 interaction sequence to within residues 1482-1489 of the SJ170 sequence.
A Dominant Tryptophan/Phenylalanine-based Motif-Within the region necessary for AP-2 binding, the primary sequence of SJ170 contains one Trp and one Phe residue (Fig. 2). Because phenylalanine side chains contribute an important hydrophobic binding component to several characterized AP-2 appendage interaction motifs (37), we separately mutated to Ala the two Phe residues close to the putative binding region (F1486A and F1492A). The proximal Phe is essential for productive AP-2 binding; only residual AP-2 interaction remains upon substitution with Ala (Fig. 2), which we attribute to the intact proximal intact FXDXF motif in this model protein. Phe 1492 , however, is not required for AP-2 engagement, in agreement with the capacity of the GST-SJ170C5.5-(1454 -1489) fusion protein to associate with AP-2. Like the required Phe 1486 , substitution of the Trp for Ala (W1483A) abolishes the adaptor binding capacity of the GST-SJ170M1 protein (Fig. 3A, lane f ). We therefore tentatively term this interaction sequence the WXXF motif. Experiments with longer SJ170 carboxyl-terminal fusions rule out the possibility that the observed AP-2 binding is artifactually because of expression of small isolated segments out of context of the native protein. The entire alternatively spliced region found in SJ170, when immobilized as a GST fusion, affinity isolates AP-2 (41) (Fig. 3B, lane d). A single W1483A substitution in the protein almost totally abolishes the interaction with AP-2 (lane f compared with lane d). This reveals that in this type of assay, the WXXF motif is a dominant endocytic interaction sequence within SJ170; we attribute the residual binding in the W1483A mutant to the intact 1556 DPF and 1463 FXDXF motifs. A similar effect is seen with the smaller SJ170C2 (residues 1454 -1530) fusion (Fig. 2). We also confirm that the entire SJ170 carboxyl-terminal segment binds weakly to clathrin as well (Fig. 3, lanes d and f ) (41), but the W1483A substitution does not perturb this interaction detectably.
In other experiments with the GST-SJ17C5 fusion we find that the proximal Trp side chain cannot be replaced with Tyr, Phe, His, or Leu and still function in AP-2 interactions as does the wild-type fusion protein (data not shown). The distal aromatic, Phe 1486 , can be altered to a Trp without a discernible effect on adaptor binding, however (data not shown). The results of these experiments clearly establish the existence of a novel WXXF AP-2 interaction determinant within SJ170 and argue strongly against the possibility that this association is an atypical form of YXXØ interaction with the AP-2 2 subunit.
Binary SJ170-AP-2 ␣ C Appendage Interactions-Previously, we showed that the SJ170C2 protein, when cleaved from GST, could compete with soluble endocytic accessory proteins for binding to GST-␣ C appendage (37) (Fig. 4, lane h compared with lane d). When added into brain cytosol to a concentration of 20 M, the inhibition of AP180 and amphiphysin I and II binding approximates that seen with the 20 M epsin 1 DPW domain (residues 229 -407) (lane h compared with lane f ). However, unlike the DPW domain of epsin 1, which contains 8 tandemly arrayed DPW triplets (45), the SJ170C2 is unable to prevent binding of soluble brain epsin 1 to GST-␣ C (lanes g and h compared with lanes e and f ). Because the WXXF motif appears to be a dominant AP-2 interaction motif in pull-down type assays (Fig. 3), we tested the ability of SJ170C2 mutated at either the 1463 FXDXF or 1483 WXXF site to inhibit binding to GST-␣ C . Unexpectedly, mutation of either sequence dramatically reduced the inhibitory capacity of the protein (Fig. 4,  lanes j and l compared with lane h). With the exception of limited inhibition of AP180 association, the remainder of the endocytic interaction partners bind similarly to that observed in the absence of an added SJ170C2 competitor (lanes j and l  compared with lane d). Analysis of the different SJ170C2 pro-  and j), or GST-SJ170C6 (lanes k and l) immobilized on GSH-Sepharose was incubated with rat brain cytosol. After centrifugation, aliquots corresponding to 1/75 of each supernatant (S) and 1/6 of each washed pellet (P) were resolved by SDS-PAGE and either stained with Coomassie Blue or transferred to nitrocellulose. Portions of the blots were probed with the anti-AP-2 ␣-subunit mAb 100/2 or anti-AP-2 2-subunit antiserum. B, approximately 100 g of GST (lanes a and b), GST-SJ170WT (residues 1305-1575; lanes c and d), or SJ170WT (W1483A) (lanes e and f) immobilized on GSH-Sepharose was incubated with rat brain cytosol. After centrifugation, aliquots corresponding to 1/75 of each supernatant (S) and 1/6 of each washed pellet (P) were resolved by SDS-PAGE and either stained with Coomassie Blue or transferred to nitrocellulose. Portions of the blots were probed with the anti-AP-2 ␣-subunit mAb 100/2, anti-AP-2 2-subunit antiserum, a mixture of the anti-clathrin heavy chain mAb TD.1 and the anti-␤1/2-subunits mAb 100/1, or the anti-clathrin light chain mAb Cl57.3. teins by SDS-PAGE (not shown) shows that they were all added to equivalent concentrations, a 5-fold molar excess over the immobilized ␣ C appendage. Also, secondary structure prediction algorithms indicate that this region of SJ170 is likely to be disordered, as are the AP-2-and clathrin-binding regions of both epsin 1 and AP180 (17,37), making misfolding improbable in our view. Our interpretation of the data is that both the WXXF and the FXDXF motifs are required for effective competition at the platform interaction surface of the ␣ C appendage in the configuration of this type of assay.
Appendage Specificity of the WXXF Motif-If the WXXF and FXDXF sequences can each associate physically with the ␣ C appendage, why is there such a marked difference in the AP-2 binding capacity of these sequences in pull-down assays (Figs. 2 and 3)? One explanation could be that unlike the FXDXF motif, which cannot bind to the structurally and functionally related AP-2 ␤2 appendage (37), the WXXF motif might be capable of binding to both the ␣ and ␤2 appendages that project off the heterotetrameric adaptor core. This is certainly not unprecedented, as the ␣ C and ␤2 appendages are structurally analogous and epsin, AP180, and eps15 are each able to bind to either appendage (46). Alternative, although not necessarily mutually exclusive, possibilities are that the WXXF motif could engage a separate site on the ␣ C appendage and this, through avidity and proximity effects, could promote the FXDXF interaction with the adjacent platform subdomain, or that the WXXF sequence has a higher apparent affinity for the ␣ appendage than the FXDXF sequence.
To begin to distinguish between these possibilities, we first tested whether the GST-SJ170C5 model protein can associate with both AP-2 and AP-1 adaptors in vitro because other endocytic accessory factors that are capable of binding to the ␤1/2 appendage can affinity purify both adaptor types from cytosol. In contrast to the epsin 1 DPW domain (36) (Fig. 5A, lane d) (lanes d, f, and h). Furthermore, whereas immobilized GST-␣ C appendage associates with SJ170, AP180, and epsin 1 present in a PC12 cell lysate (Fig. 5B, lane f ), the GST-␤2 appendage binds to epsin and AP180 but not to SJ170 (lane d). These experiments suggest that the WXXF motif does not interact effectively with ␤ appendages, and this finding is validated by in vitro binding assays using purified proteins (Fig. 5C). Thrombin-released, monomeric ␣ C appendage binds robustly to both GST-SJ170C5 (lane d) and the minimal GST-SJ170M1 fusion protein (lane f ) but not to GST alone (lane b). In contrast, neither of these SJ170 fusion proteins associates with the monomeric ␤2 appendage (lanes j and l compared with lane h). Instead, the ␤2 appendage remains in the supernatant fraction (lanes g, i, and k, solid arrowheads), as does the ␣ C appendage in the presence of GST (lane a, open arrowhead). We conclude that the WXXF motif harbored by SJ170 is highly selective for the ␣ appendage domain of AP-2.
Next, to examine the relative affinities of the WXXF and FXDXF motifs for AP-2 we compared the GST-SJ170C2 fusion protein (Fig. 2) to the GST-SJ170C2 (Phe-Asp 3 Ala) and GST-SJ170C2 (Trp 3 Ala) mutants. From the binding profile of cytosolic AP-2 to increasing amounts of the wild type SJ170C2 (Fig. 6) we estimate that half-maximal adaptor binding occurs at ϳ20 g (ϳ2 M). Inactivation of the FXDXF motif in the context of the SJ170C2 segment causes a modest (ϳ5fold) shift in the amount required for half-maximal binding (to ϳ100 g). By contrast, altering the WVTF sequence to AVTF has at least a 25-fold effect, increasing the amount needed for 50% binding to Ͼ Ͼ500 g. These values substantiate the dominant role that the WXXF sequence plays in binding the AP-2 ␣ appendage in pull-down assays and are in general agreement with the competition data presented in Fig. 4.
Finally, to resolve whether the WXXF motif engages the same interaction surface upon the ␣ C appendage utilized by epsin, eps15, amphiphysin, and AP180, we compared the binding of soluble SJ170 to immobilized wild type GST-␣ C or GST-␣ C (W840A), GST-␣ C (R905A), or GST-␣ C (R916A) mutant appendages. The Trp 840 side chain contributes a major portion of the platform interaction surface (35,47) and a W840A substitution cases severe and general disruption of partner interactions (Fig. 5B, lane h compared with lane b and f) (37,47) including inhibition of SJ170 binding (lane h). These data sug-  lanes c and d), SJ170C5 (lanes e and f), or GST-ARHC1 (residues 180 -308 of ARH; lanes g and h) immobilized on GSH-Sepharose was incubated with rat brain cytosol. After centrifugation, aliquots corresponding to 1/75 of each supernatant (S) and 1/6 of each washed pellet (P) were resolved by SDS-PAGE and either stained with Coomassie Blue or transferred to nitrocellulose. Portions of the blots were probed with the anti-AP-2 ␣-subunit mAb 100/2, anti-AP-2 2-subunit antiserum, the anti-AP-1 ␥-subunit antibody AE/1, or anti-AP-1 1-subunit serum RY/1. B, approximately 50 g of GST (lanes a and b), GST-␤2 appendage (lanes c and d), GST-␣ C appendage (lanes e and f), or the GST-␣ C appendage point mutants W840A (lanes g and h), R905A (lanes i and j), or R916A (lanes k and l) immobilized on GSH-Sepharose were incubated with PC12 cell lysate. After centrifugation, aliquots corresponding to 1/60 of each supernatant (S) and 1/8 of each washed pellet (P) were resolved by SDS-PAGE and either stained with Coomassie Blue (left panel) or transferred to nitrocellulose (right panels). Portions of the blots were probed with the anti-synaptojanin 1 antibody AR/1, an anti-AP180 mAb, or anti-epsin 1 antibodies. The asterisk indicates a likely degradation product of epsin 1. C, approximately 400 g of GST (lanes a, b, g, and h), GST-SJ170C5 (lanes c, d, i, and j), or GST-SJ170M1 (lanes e, f, k, and l) immobilized on GSH-Sepharose was incubated with either thrombin-cleaved, monomeric ␣ C appendage (lanes a-f) or ␤2 appendage (lane g-l) in the presence gest that the WXXF sequence binds physically to the platform of ␣-subunit appendage. Furthermore, an R916A substitution both abolishes AP180 and SJ170 binding while leaving the epsin interaction intact, in agreement with previous observations (35,47). Although the GST-␣ C (R905A) mutant fails to bind AP180 (lane j) the protein still associates with SJ170 (lane j), albeit at a lower level than wild type appendage (lane e).
A Functional WXXF Motif in Other Endocytic Proteins-To justify terming WXXF an AP-2 interaction motif, the functional sequence should necessarily be found in other AP-2 binding partners. We noticed that adaptor-associated kinase 1 (AAK1), a Ser/Thr protein kinase that phosphorylates the 2 subunit of the AP-2 complex (43,44), has the sequence 695 WNPF. This tract lies within a likely unstructured region of AAK1 known to engage AP-2. Indeed, AAK1 was originally identified on the basis of binding to the AP-2 ␣ appendage (43). Two overlapping regions of AAK1 (residues 547-862 and residues 671-862, Fig.  7A) fused to GST efficiently bind AP-2 in pull-down assays (Fig.  7B, lanes d and f). Recovery of AP-2 with the GST-AAK1 pellet does not depend on eps15 binding to the embedded NPF triplet as no detectable eps15 is evident in the pellet fractions (data not shown). Mutation of either the Trp 695 or Phe 698 in the WNPF sequence to Ala inhibits this ability to bind AP-2 (lanes g-l). The effect of a F698A mutation is more severe (lane j) and is essentially the same as a Trp-Phe 3 Ala double mutant (lane l). Although the interaction of AAK1 with the ␣ appendage was originally attributed to the two DPF triplets positioned distal to the WNPF motif (Fig. 7A), we find that this (DPF) domain alone (GST-AAK1C3; residues 741-862) does not bind to AP-2 in comparable pull-down assays (Fig. 7B, lane n). This is in good agreement with our previous observations that the capacity of the DPW region of epsin 1 to bind AP-2 depends upon the number of tandemly arrayed DPW repeats and that reduction from 8 to 3 almost totally abolishes AP-2 binding (36). By contrast, a short region (residues 671-703; GST-AAK1C2 3 stop) harboring WNPF but lacking the DPF repeats does bind AP-2 (data not shown). Taken together, the AAK1 results are in accord with the SJ170 data presented above showing that the WXXF sequence has a higher apparent affinity for AP-2.
The AAK1 693 STWNPFDD sequence is invariant in the presumptive rodent AAK1 orthologues (rat accession number XM_232172 and partial mouse clone AAH43125) and a nearly identical sequence (SGWNPFGE) is also found within the carboxyl-terminal segment of mouse (NM_080708) and human (Q9NSY1) BMP2-inducible protein kinase (BIKe) (48). The kinase domains of AAK1 and BIKe are 77% identical and the rodent and human BIKe sequences each also contain two DPF triplets. Interestingly, another related Ser/Thr kinase, GAK/ auxilin 2 contains the sequence 1037 WAAW (Fig. 7C). The anchor aromatic side chains are conserved between the rodent (WDTW) and human (WAAW) GAK/auxilin 2 orthologues, and we have shown for the SJ170 WXXF that WXXW functions comparably. We therefore determined whether the GAK WAAW participates in AP-2 interactions. In a manner analogous to the AAK1 fusion, a GAK fusion (GST-GAKC1, residues 945-1311; Fig. 7D, lane d) binds AP-2 from brain cytosol in pull-down assays, as reported by others (49). This interaction is almost completely abrogated by a single W1036A mutation within the WAAW sequence (Fig. 7D, lane f); again we attribute the residual binding to the intact DPF sequences. GAK is involved in clathrin coat disassembly following fission and also binds physically to clathrin (49, 50) (possibly via clathrin binding DLL sequences (51), Fig. 7C) but the W1036A substitution has no effect on this interaction (lane f compared with lane d).
We conclude that in SJ170, AAK1, and GAK, the WXX(FW) motif participates directly in AP-2 binding although residues outside the WXX(FW) motif probably contribute toward optimal engagement. Interestingly, each of these proteins contains 2 DPF triplets located within regions of the polypeptide likely to be unstructured (Figs. 2 and 7, A and C). The combination of several discrete binding motifs and conformational plasticity could allow simultaneous engagement of multiple adaptor molecules within an assembled clathrin lattice.
We also note that three copies of the WXX(FW) motif are tandemly arrayed within the first 250 residues of the brain isoform of the endocytic protein stonin 2/stoned B (Table I), a polypeptide tract predicted to be essentially disordered (not shown) and with no significant homology to other known proteins. An orthologue of the Drosophila stoned B protein, stonin 2 also contains NPF triplets (52,53) and, like GAK (49,50), has been suggested to participate in clathrin uncoating events (53). Available EST sequences indicate that the WXX(FW) sequences in stonin 2 are phylogenetically conserved from zebrafish to mammals. When fused in-frame with GST, the amino-terminal 426 amino acids of human stonin 2 bind to AP-2 in a pull-down assay (Fig. 8, lane d). The extent of binding is similar to that observed with the GST-SJ170C2 fusion (lane b). Although the association between stonin 2 and AP-2 is thought to be indirect, being mediated by eps15 (52), we still observe AP-2 binding following near-quantitative removal of eps15 from cytosol by preincubation with GST-␣ C appendage. The pellet obtained after preincubating the cytosol with GST-␣ C appendage shows the recovery of eps15 together with the sedi-   a and b) or GST-stonin 2 (residues 1-426) (lanes c-f) immobilized on GSH-Sepharose was incubated with either mock (GST)-depleted (lanes a-d) or GST-␣ C appendage-depleted (lanes e and f) rat brain cytosol (which was prepared by preincubation with either 150 g of immobilized GST or GST-␣ C appendage; the resulting pellets (lanes g and h) demonstrate capture of AP-2 binding partners, including eps15). After centrifugation, aliquots corresponding to 1/60 of each supernatant (S) and 1 /8 (lanes b, d, and f) or 1/12 (lanes g and h) of each washed pellet (P) were resolved by SDS-PAGE and either stained with Coomassie Blue or transferred to nitrocellulose. The position of the bound AP-2 ␣and ␤2-subunits are indicated with open arrowheads, whereas the intact GST-stonin 2 fusion protein is indicated with filled arrowheads. Portions of the blots were probed with the anti-AP-2 ␣-subunit mAb 100/2, anti-AP-2 2 subunit, or anti-eps15 antiserum. *, cross-reactivity of the anti-eps 15 antiserum with the intact GST-stonin 2 fusion protein.
n) immobilized on GSH-Sepharose was incubated with rat brain cytosol. After centrifugation, aliquots corresponding to 1/60 of each supernatant (S) and 1/8 of each washed pellet (P) were resolved by SDS-PAGE and either stained with Coomassie Blue or transferred to nitrocellulose. The blot was probed with the anti-AP-2 ␣-subunit mAb 100/2. D, approximately 150 g of either GST (lanes a and b), GST-GAK1C1 (lanes c and d), or GST-GAK1C1 (W1036A) (lanes e and f) immobilized on GSH-Sepharose was incubated with rat brain cytosol. After centrifugation, aliquots corresponding to 1/60 of each supernatant (S) and 1/8 of each washed pellet (P) were resolved by SDS-PAGE and either stained with Coomassie Blue or transferred to nitrocellulose. The blot was probed with a mixture of the anti-clathrin heavy chain mAb TD.1 and the anti-␤1/2-subunits mAb 100/1.  red asterisk. B, approximately 200 g of GST (lanes a and  b), GST-AAK1C1 (lanes c and d), GST-AAK1C2 (lanes e and f), GST-AAK1C2 (W695A) (lanes g and h), GST-AAK1C2 (F698A) (lanes i and j), lanes k and l), or GST-AAK1C3 (lanes m and Hs ␥-synergin a The 0 position residue, corresponding to the first anchor aromatic side chain, and the ϩ4 position aromatic are indicated. Conserved residues used in defining a tentative consensus sequence are also underlined. b The human (Hs) or rat (Rn) sequence accession numbers are: SJ170, NM_003895; AAK1, NM_014911; GAK, NM_005255 and NM_031030; brain stonin 2, NM_033104; brain NECAP 1, NP_056324; p56, AY289196; epsinR, BC004467; rabaptin-5, NM_004703; and ␥-synergin, NM_080550.
d Asterisk indicates the extreme carboxyl-terminal residue of the protein.
stonin 2 prevents internalization of transferrin, low density lipoprotein, and epidermal growth factor (52). The overexpressed GFP-stonin 2 alters the intracellular localization of AP-2 causing the adaptor complex to cluster in large aggregates that prevent proper placement at the cell surface (52). Importantly, there are no other recognizable ␣-appendage binding sequences within stonin 2. Altogether, these experiments show that the WXX(FW) motif is functional in several endocytic proteins, and the compiled sequences (Table I) indicate that WXX(FW)X(DE) is a predictive consensus for this third type of AP-2 interaction motif. DISCUSSION We have defined a third sequence type that engages the AP-2 adaptor, WXX(FW)X(DE), which is characterized by key anchor aromatic side chains, as are the DP(FW) and FXDXF motifs. Very recently, a WDWH sequence was also shown to bind directly to the AP-2 ␣ C appendage (54) although the interaction surface was not delineated. The WVTFEE sequence is evolutionarily conserved between rat and human SJ170 but is not found in the related polyphosphatase synaptojanin 2 (55). Whereas synaptojanin 2 is clearly implicated in endocytic uptake (56,57), it appears to act at a different step in clathrincoated vesicle formation, earlier than SJ170 (57). In addition, synaptojanin 2, unlike synaptojanin 1, is targeted to membranes via the small GTPase Rac1 (55,56), and this might account for the difference in the representation of AP-2 interaction determinants in these two enzymes.
Our data clearly show the high selectivity of the SJ170 WVTF sequence for the AP-2 ␣ appendage. Whereas it appears that the WXX(FW)X(DE) motif binds to the platform subdomain of the ␣ C -appendage, the results of the competition studies, the motif deletion analysis in the GST-SJ170C2 model protein, and the capacity of some of the ␣ C appendage mutants to still bind intact SJ170 indicate that the WXXF binding site does not overlap the DP(WF)/FXDXF interaction site completely. At present, we cannot rule out that the effects of the ␣ C appendage platform mutants we tested are because of conformational perturbation of an adjacent WXX(FW)X(DE) binding surface. During revision of this paper, an analogous WVQF sequence was identified in two clathrin-associated proteins, designated NECAP 1 and 2, that mediates binding to AP-2 (58) ( Table I). The NECAP WVQF sequence does not compete with either DP(FW) or FXDXF motifs for ␣ C appendage binding (58). In this regard, it is interesting to note that several groups have shown recently that the AP-1 ␥-subunit appendage, as well as the structurally/functionally related GGA ␥-adaptin ear (GAE) domain, bind a DFXXØ sequence (where Ø represents a bulky hydrophobic side chain) (59 -65) (Table I). This superficially related AP-1/GGA interaction motif binds to the sandwich domain of the ␥ appendage. The overall 8-stranded ␤-sandwich fold of the ␥ appendage is analogous to the amino-terminal sandwich subdomain of the ␣ appendage that supports the platform subdomain (35,47). In the GGA1 GAE⅐DFGGF (p56; see Table I) interaction, the proximal Phe packs into a cavity created partly by GGA1 appendage residues Pro 565 , Arg 607 , and Arg 609 (61). Analogous residues are required for the GGA3 GAE⅐DFGPL (rabaptin-5) (64) and DFXXØ⅐␥ appendage associations (59,60,62). Yet none of the side chains necessary for the DFXXØ interaction are conserved at the equivalent position of the ␣ C appendage sandwich subdomain (59) making it unlikely in our view to be the binding site for the WXX(FW) motif. Furthermore, close inspection of the WXX(FW)X(DE) sequences delineated here shows obvious side chain differences from the DFXXØ-type sequence (Table I). It is clear that there is no strong preference for acidic residues before the proximal Trp (0 position) in the WXX(FW) but, instead, conservation of an acidic residue at the ϩ5 position relative to the anchor Trp (0). In NECAP 1 and 2, the WXXF sequence is positioned at the extreme carboxyl terminus (Table I) and is similarly positioned in the Xenopus NECAP orthologue (AAH54244). Here, the terminal carboxyl group may replace a distal acidic side chain, as is seen in the truncated LLDLD-type clathrin box sequences in the S. cerevisiae epsin orthologues Ent1p and Ent2p (66). The favored Gly in the ϩ1 position of the DFXXØ sequence is not present in any of the identified AP-2 ␣ appendage binding sequences. Indeed, we do not detect any interaction between a minimal SJ170 AP-2-binding fusion (GST-SJ170C5) and AP-1 (Fig. 5A).
Systematic analysis of the contextual rules and side chain preferences for the WXX(FW)X(DE)⅐␣ appendage interaction should begin to explain the molecular basis for selectivity compared with the DFXXØ ␥-binding motif. Intriguingly, the minimal AP-2 binding region in numb, a known endocytic protein (67,68), has been mapped to the sequence 613 VD-PFEAQWAALENKSKQRTNPSPT (67). Given our failure to detect robust binding of AP-2 to model proteins containing only one or two DPF triplets, it will also be interesting to determine whether a 620 WAALEN sequence present in both numb and numb-like contributes at all to AP-2 appendage binding. Ultimately, however, unambiguous appreciation of the precise orientation of a SJ170, AAK1, GAK, and/or stonin 2 WXX (FW)X(DE) sequence bound to the ␣ appendage must await structural studies.
One important question that is currently unresolved is why multiple discrete adaptor interaction motifs are necessary biologically? From our current studies, it is clear that the DP(WF), FXDXF, and WXX(FW)X(DE) sequences each differ in apparent affinity for the ␣ C appendage. One interesting possibility for the presence of multiple interaction motifs within the carboxyl-terminal extension of SJ170 is that they might function cooperatively to promote release of other accessory factors late in the budding process. Secondary structure predictions suggest that the SJ170 extension is probably unstructured so the embedded DPF, FXDXF, and WXX(FW)X(DE) sequences could bind to multiple appendages within the clathrin lattice (in a random coil conformation, the AP-2 interaction harboring region can extend over 60 nm). Concomitant enzymatic dephosphorylation of PtdIns(4,5)P 2 would terminate ENTH/ANTH or PTB domain associations with the lipid bilayer. Binding to a distinct ␣ C appendage surface via the WXX(FW)X(DE) motif would allow SJ170 (and AAK1, GAK, stonin 2, and NECAP) to engage AP-2, whereas the principal platform site is occupied by other endocytic adaptors or accessory proteins. Once docked, SJ170 could potentially compete other endocytic accessory proteins from the bud site, a model that would explain why accessory proteins like epsin and amphiphysin are not enriched within purified clathrin-coated vesicles, and begin to provide a clue to the temporal regulation of accessory protein presence at the clathrin-coated bud.