Type Iγ661 Phosphatidylinositol Phosphate Kinase Directly Interacts with AP2 and Regulates Endocytosis*

Clathrin-coated vesicles mediate sorting and intracellular transport of membrane-bound proteins. The formation of these coats is initiated by the assembly of adaptor proteins (AP), which specifically bind to membrane cargo proteins via recognition of endocytic sorting motifs. The lipid signaling molecule phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2) is critical for this process, as it serves as both a targeting and regulatory factor. PI(4,5)P2 is synthesized by type I phosphatidylinositol phosphate kinases (PIPKI). We have discovered a direct interaction between the μ2-subunit of the AP2 complex and PIPKIγ661 via a yeast two-hybrid screen. This interaction was confirmed using both the μ2-subunit in glutathione S-transferase pulldowns and via coimmunoprecipitation of endogenous PIPKIγ661 with the AP2 complex from HEK293 cells. The interaction is mediated, in vivo, by a tyrosine-based motif in the 26-amino acid tail of PIPKIγ661. Because AP2 regulates endocytosis of transferrin receptor from the plasma membrane, we also examined a role for PIPKIγ661 using a flow cytometry endocytosis assay. We observed that stable expression of wild type PIPKIγ661 in Madin-Darby canine kidney cells enhanced transferrin uptake, whereas stable expression of kinase-dead PIPKIγ661 had an inhibitory effect. Neither condition affected the overall cellular level of PI(4,5)P2. RNA interference-based knockdown of PIPKIγ661 in HeLa cells also had an inhibitory effect on transferrin endocytosis using the same assay system. Collectively, this evidence implies an important role for PIPKIγ661 in the AP2-mediated endocytosis of transferrin.

The surface expression level of plasma membrane receptors is highly regulated via balanced vesicular trafficking to and from the plasma membrane. An important aspect of this regulation is the internalization of receptors by the clathrin-mediated endocytosis pathway. This process not only internalizes and delivers extracellular ligands to the endosomal transport system, but it also serves to recycle receptors or target them for degradation.
Assembly of the clathrin coat is a highly orchestrated process that requires the action of several distinct proteins (1), including G-proteins, protein and lipid kinases and phosphatases, and various adaptor proteins, such as the adaptin protein (AP) 3 complexes. These complexes specifically bind to integral membrane cargo proteins and serve both as protein docking sites and as nucleation points for assembly of the clathrin lattice. Several AP complexes have been identified, with each complex having a distinct function (2). For example, AP1B is thought to modulate sorting to the basolateral membrane in epithelial cells, whereas AP2 is thought to regulate receptor-mediated endocytosis from the plasma membrane (3)(4)(5)(6).
The AP complexes consist of four distinct subunits. The crystal structure of the AP2 complex has recently been solved and consists of two large subunits, ␣ and ␤2, a medium subunit, 2, and a small subunit, 2 (7). The 2-subunit is critical for the function of the complex as it specifically binds to membrane cargo proteins via tyrosine-based sorting motifs within their cytoplasmic domains (8). This subunit consists of an N-terminal domain, which is buried within the core of the AP2 complex, joined by a flexible linker to a solvent exposed C-terminal domain. In the AP2 crystal, the 2-subunit C-terminal domain was found to be buried within a pocket formed by the other three subunits of the complex (7). In this "closed" conformation, the -subunit is thought be unable to interact with membrane-bound cargo proteins. The current model for the AP2 cargo binding suggests that the complex shifts into an "open" conformation upon binding to the plasma membrane in the presence of cargo (7,9). The displaced -subunit can then bind to cargo and dock to the membrane.
Assembly of the AP2 complex onto membranes is mediated by a specific phospholipid, phosphatidylinositol 4,5-bisphosphate (PI(4,5)P 2 ) (10). PI(4,5)P 2 also regulates numerous proteins that participate in clathrin-mediated endocytosis, including AP180, epsin, and dynamin (11). In addition, PI(4,5)P 2 is an important regulatory factor for modulation of the actin cytoskeleton and subsequently vesicular transport (12,13). Two PI(4,5)P 2 binding sites have been identified within the AP2 complex. One binding site is located on the ␣-subunit and is positioned to dock the AP2 complex to membranes (5,7). The other binding site is located on the 2-subunit and formed by a cluster of conserved lysine residues (14). The AP2 structure suggests that the 2-subunit binds PI(4,5)P 2 only when AP complex is docked to the membrane and the 2-subunit is bound to cargo (7,15). This model has recently been demonstrated using an in vitro reconstituted system coupled with surface plasmon resonance (16). This collection of evidence emphasizes the critical role for PI(4,5)P 2 in both the initial targeting of AP2 to endocytic sites and for stabilizing the -subunit/cargo interaction at the plasma membrane upon transition of the complex into the open conformation. PI(4,5)P 2 is synthesized at the plasma membrane by type I phosphatidylinositol phosphate kinases (PIPKI). By generating PI(4,5)P 2 , PIPKIs are important not only for regulation of vesicular trafficking, but also have functions in cytoskeletal assembly, cell proliferation, apoptosis, signal transduction, and nuclear functions (15,17,18). All three of the known PIPKI isoforms (␣, ␤ and ␥; using the human nomenclature) have been implicated in clathrin-mediated endocytosis. Previous work has shown that expression of PIPKI␣ in NR6 cells resulted in increased endocytosis of the epidermal growth factor receptor, whereas expression of kinase inactive PIPKI␣ inhibited endocytosis (19). Another study found that PIPKI␤ expression had a similar effect on transferrin receptor endocytosis in HeLa cells and that RNAi-based knockdown of PIPKI␤ expression had an inhibitory effect (20). PIPKI␥ has been implicated as the major producer of PI(4,5)P 2 in synaptic nerve terminals and this role has been supported by observed phenotypes in PIPKI␥ knockout mice, where synaptic vesicle endocytosis is generally impaired (21,22). However, the mechanism by which PIPKIs are specifically targeted to sites of endocytosis for localized generation of PI(4,5)P 2 remains unclear.
The PIPKI␥ isoform is alternatively spliced in cells, resulting in two major variants, PIPKI␥635 and PIPKI␥661, which differ only by a 26-amino acid C-terminal extension. We have discovered a direct interaction between PIPKI␥661 and the 2-subunit of AP2 via a yeast two-hybrid approach employing the C-terminal 178 amino acids of PIPKI␥661 as bait. Here we confirm that this interaction is direct and occurs in vivo. Using cell lines stably expressing PIPKI␥661 and siRNAs specifically knocking down expression of PIPKI␥, we have also found that this interaction has direct implications on the endocytosis of the transferrin receptor from the plasma membrane. These combined results suggest that PIPKI␥661 may be an important regulatory factor in clathrin-mediated endocytosis.

MATERIALS AND METHODS
Expression Constructs-PIPKI␥661, PIPKI␥635, and PIPKI␥661KD mammalian and bacterial expression vectors were described previously (23,24). PIPKI␥ C-terminal truncation constructs were also described previously (24). The following PIPKI␥661 mutants were generated using the QuikChange TM mutagenesis kit (Stratagene) and the following mutagenic primers and their complements: I␥S645F, 5Ј-GGA-GCTGGGTGTACTTCCCGCTTCACTATAGC; I␥P646F, 5Ј-GGAGCTGGGTGTACTCCTTCCTTCACTATAGCGCG; I␥L647V, 5Ј-GGGTGTACTCCCCGGTTCACTATAGCGC; and I␥P646R, 5Ј-GCTGGGTGTACTCCCGGCTTCACTAT-AGCGC. The I␥Y644F mutant was described previously (24). The full-length murine 2-subunit yeast two-hybrid clone was obtained from a murine brain cDNA library (Molecular Interaction Facility, University of Wisconsin). A soluble truncation of this subunit was generated by digestion of the 2-subunit open reading frame with an internal EcoRI site and an external 5Ј XhoI site. The resulting fragment was cloned into pET28 and pET42 bacterial expression vectors (EMD Biosciences).
Cell Cultures and Transfection-HEK293, HeLa, and Madin-Darby canine kidney (MDCK) cells were cultured using Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. HEK293 cells were transfected via calcium phosphate with 2 or 5 g of each expression vector. The cells were used for immunoprecipitation 48 h after transfection.
MDCK Stable Cell Lines-MDCK cells were stably transfected with various PIPKI␥661 constructs, which are under the control of the tetracycline responsive promoter. The transfected MDCK cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 1 g/ml puromycin, and 100 g/ml hygromycin B to select for stable transfection. The medium was supplemented with 2 g/ml doxycycline to suppress transgene expression, as doxycycline withdrawal results in expression of transfected PIPKI␥.
Antibodies-Monoclonal mouse anti-human transferrin receptor was purchased from BD Bioscience. Monoclonal anti-␣-adaptin (C-8) antibody was obtained from Santa Cruz Biotechnology. Monoclonal mouse anti-HA (16B12) was obtained from Covance and rabbit polyclonal anti-HA antibody was purchased from Upstate. Horseradish peroxidase-conjugated anti-GST antibody was purchased from Amersham Biosciences and horseradish peroxidase-conjugated and monoclonal anti-T7 antibodies were obtained from Novagen. Monoclonal antitalin antibody (8d4) was purchased from Sigma. Monoclonal anti-phosphotyrosine antibody (4G10) was purchased from Upstate. Polyclonal PIPKI␥ antiserum was generated and purified as described previously (23). Polyclonal PIPKI␣ antiserum was generated and purified as described previously (25). Secondary antibodies were obtained from Jackson ImmunoResearch Laboratories.
Immunoprecipitation and Immunoblotting-Cells were washed twice with ice-cold PBS and subsequently resuspended and lysed in IP buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.2% Nonidet P-40, 2 mM Na 3 VO 4 , 1 mM EDTA, 1 mM EGTA, and 1 mM MgCl 2 ). Cell lysates were incubated with 50 l of 1:1 diluted protein A-Sepharose TM and 2 g of the specified antibody as indicated at 4°C overnight. The immunocomplexes were separated by 7.5 or 10% SDS-PAGE, and transferred to polyvinylidene difluoride (Millipore Corp.). Chemiluminescent substrate (Pierce) was used for visualization on x-ray film (RPI Corp.).
Protein Expression and Purification in Escherichia coli-Constructs in pET28 or pET42 expression vectors were transformed into BL21(DE3) competent cells (Novagen). Proteins were expressed and purified using His-Bind TM resin following the manufacturer's instructions (Novagen) or using glutathione-Sepharose TM 4B Fast Flow as per the manufacturer's instructions (Amersham Biosciences). Tyrosine-phosphorylated recombinant PIPKI␥661 was generated via coexpression with Src and purified as described previously (26).
GST Pulldown Assays-Recombinant T7-tagged PIPKI␥ was incubated with GST-2 together with glutathione-Sepharose TM 4 Fast Flow (Amersham Biosciences) in 500 l of buffer B (PBS, 1% bovine serum albumin, 0.4% Triton X-100, and 2 mM dithiothreitol) for 4 h or overnight at 4°C. The beads were washed with 1 ml of buffer B four times, resolved by SDS-PAGE, and analyzed via Western blot. GST was used as a negative control for nonspecific binding. All other GST pulldowns were performed with the proteins indicated in the same manner.
Transferrin Uptake Assays-Stable MDCK cells were grown in 10-cm dishes. Expression of exogenous PIPKI␥ was induced for 72 h by withdrawing doxycycline from medium. For transferrin uptake assays, cells were incubated with serum-free medium for 2 h. The serum-starved cells were then incubated with Alexa Fluor 647 transferrin (5 g/ml, Invitrogen) in binding medium at 37°C for 20 min. After incubation, the cells were washed three times with PBS, three times with ice-cold acid (0.2 M acetic acid and 0.5 M NaCl, pH 4.1), and three times with PBS again. Cells were finally collected by trypsinization and washed once with PBS. Half of the cells were used to determine PIPKI␥ expression by flow cytometry. Briefly, cells were first incubated with primary anti-HA antibody and then with fluorescein isothiocyanate-labeled secondary antibody. The fluorescence intensities of fluorescein isothiocyanate staining were detected from ϳ10,000 cells and used to determine PIPKI␥ expression. The other half of the cells was suspended in 0.5 ml of PBS for determination of transferrin uptake using a FACS Calibur TM (BD Biosciences) flow cytometer. Fluorescence intensities of internalized Alexa Fluor 647 transferrin were quantified from ϳ10,000 cells.
RNA Interference-HeLa cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. The cells were passed into 60-mm dishes 1 day prior to transfection. The cells were transfected with a PIPKI␥-specific siRNA oligo (5Ј-AAGGACCUGGACUUCAUGCAG) using Oligofectamine TM (Invitrogen) transfection reagent. Scrambled control siRNA (5Ј-AAGUACCUGUACUUCAUGCAG) or PIPKI␣-specific siRNA (5Ј-AAGAAGUUGGAGCACUCU-UGG) were used as controls. After 24 h, the cells were transfected again in the same manner. The cells were then used for transferrin uptake assays 72 h post-transfection.
Metabolic Labeling and Determination of Cellular PI(4,5)P 2 Levels-MDCK cells were metabolically labeled with 20 Ci/ml myo-[ 3 H]inositol (PerkinElmer Life Sciences) and the lipids were extracted and deacylated as described previously (27). The deacylated glycerophosphoinositol phosphates were resuspended in water prior to analysis by HPLC. The deacylated lipids were separated using a Zorbax SAX column and a gradient of 1.3 M ammonium phosphate (pH 3.85). The level of cellular PI(4,5)P 2 was determined with a Packard in-line liquid scintillation flow detector using deacylated [ 3 H]PI(4,5)P 2 (PerkinElmer Life Sciences) as a standard.
Immunofluorescence and Microscopy-Cells were washed with PBS at room temperature, fixed with 4% paraformaldehyde in PBS at room temperature for 15 min, and permeabilized with 0.2% Triton X-100 in PBS at room temperature for 10 min. The cells were blocked with 3% bovine serum albumin in PBS at room temperature for 1 h, incubated with the primary antibody for 1 h at 37°C, and washed with 0.1% Triton X-100 in PBS. The cells were then incubated with fluorophore-labeled secondary antibody at room temperature for 45 min and washed with 0.1% Triton X-100 in PBS. The coverslips were mounted to glass slides in Vectashield (Vector Laboratories) mounting medium. Fluorescent images were captured using a Nikon Eclipse TE2000-U microscope with a CoolSNAP CCD camera (RS Photometrics) or using a ϫ60 Plan oil immersion lens on a confocal laser-scanning microscope (model MR-1000; Bio-Rad) mounted transversely to an inverted microscope (Diaphot 200, Nikon; W. M. Keck Laboratory for Biological Imaging, Madison, WI). Images were processed as described previously using Photoshop TM CS (23). To visualize the colocalization between AP2 and PIPKI␥661 constructs, MDCK cells in 6-well plates were transfected with 1 g of each expression vector. After 16 h of expression, the cells were incubated in serumfree Dulbecco's modified Eagle's medium for 2 h and then treated with 50 g/ml of transferrin (Invitrogen) for 20 min at 37°C. The cells were subsequently fixed and stained as described above.

PIPKI␥661 Directly Interacts with the -Subunit of AP2 in
Vitro and in Vivo-Our laboratory performed a yeast two-hybrid screen using the C-terminal 178 amino acids of PIPKI␥661 as bait (23). This screen indicated a direct association with the 2-subunit of AP2. The prey, obtained from a murine brain library, contained the entire mouse 2-subunit with the exception of the start codon. To confirm that this interaction was indeed direct, a GST pulldown approach was used.
Because the full-length 2-subunit is primarily insoluble in E. coli, a truncation mutant was generated. This was accomplished by deleting of the bulk of the N-terminal domain and replacing it with GST. The resulting soluble construct contained the complete linker domain and the C-terminal domain. The GST pulldown was then performed by incubating PIPKI␥661 with GST-2 in the presence of glutathione-conjugated Sepharose TM beads. As shown in Fig. 1A, PIPKI␥661 directly associated with GST-2, but did not associate with GST alone.
In vivo, AP2 serves as an adaptor for binding to several other proteins involved in clathrin-mediated endocytosis. Consequently, to ensure that this interaction was relevant in vivo, we immunoprecipitated endogenous AP2 from HEK293 cells using a monoclonal antibody specific for ␣-adaptin. The precipitated complexes were washed extensively and resolved by SDS-PAGE. Using a polyclonal antibody described previously, PIPKI␥ was detected in the ␣-adaptin lane but not in the normal mouse IgG lane (Fig. 1B) (23). The reciprocal experiment, using the monoclonal antibody specific for ␣-adaptin, yielded concurrent results (Fig. 1B). It was surprising to observe that only the highest molecular weight band was retained by immunoprecipitated AP2. This band corresponds to PIPKI␥661, as the PIPKI␥ polyclonal antibody detects both PIPKI␥661 and PIPKI␥635 splice variants, as seen in the lysate lane.
Because PI(4,5)P 2 has extensively been shown to be an integral component of the endocytic process, we sought to determine whether PIP kinase activity was necessary for this interaction in vivo. Wild type and kinase inactive PIPKI␥661 were transfected into HEK293 cells via calcium phosphate. AP2 was then immunoprecipitated using monoclonal antibodies specific for the ␣-subunit. As shown in Fig. 1C, both wild type (WT) and kinase-dead (KD) PIPKI␥661 associated with AP2 in vivo. In addition, KD PIPKI␥661 appears to associate with the AP2 complex with slightly higher affinity as compared with WT PIPKI␥661.
The endogenous coimmunoprecipitation data indicated that AP2 preferentially associates with the highest molecular weight species of PIPKI␥ detected by the PIPKI␥ polyclonal antibody.
A coimmunoprecipitation approach was used to confirm which of the two splice variants could associate with AP2 in vivo. HEK293 cells were transfected with PIPKI␥661 or PIPKI␥635 and endogenous AP2 was immunoprecipitated with the ␣-adaptin-specific antibody. As seen in Fig. 1D, only PIPKI␥661 was capable of binding to AP2 in vivo. This result indicated that the in vivo binding site for 2 was localized in the C-terminal 26 residues of PIPKI␥661. To narrow the specific binding site, three previously generated truncations of PIPKI␥661 were employed in the same coimmunoprecipitation approach (24). As shown in Fig. 1D, truncation at Trp 642 resulted in reduction of associated PIPKI␥ to background levels observed in the normal mouse IgG control.
A Tetrapeptide Motif on PIPKI␥661 Mediates the Interaction with AP2-Upon closer inspection of the sequence contained between the Trp 642 and Tyr 649 truncations, there is a putative tyrosine sorting motif ( 644 YSPL 647 ). These motifs generally consist of a tetrapeptide sorting sequence YXX, where corresponds to a bulky hydrophobic residue (8). Through an exhaustive peptide library screen, Ohno et al. (28,29) has presented extensive evidence documenting the tyrosine sorting motif specificity for each of the -subunits of the AP complexes. Using these results as a guide, several point mutations were generated with the intention of weakening PIPKI␥661 binding to AP2. A mutation intended to strengthen binding was also generated as a positive control. Each of these HA-tagged constructs was transfected into HEK293 cells and endogenous AP2 was immunoprecipitated with a ␣-adaptin-specific antibody. Associated PIPKI␥661 was then detected by immunoblot with an HA-specific monoclonal antibody.
The observed results, shown in Fig. 2A, followed the affinities predicted by Ohno et al. (28). Mutation of Tyr 644 , Pro 646 (Tyrϩ2), or Leu 647 (Tyrϩ3) to the most disfavored residues resulted in disruption of the interaction. The Tyrϩ1 position was previously shown to have little contribution to binding affinity in the Ohno et al. (28) peptide screen, and mutation of Ser 645 to the least favored residue, phenylalanine, had little effect on AP2 binding. Likewise, mutation of Pro 646 (Tyrϩ3) to the most favored residue, arginine, did not alter PIPKI␥661 binding to AP2. These in vivo results were confirmed via GST pulldown experiments. As in Fig. 1A, GST-2 was used to pull down PIPKI␥661, PIPKI␥635, PIPKI␥ Y644F, or PIPKI␥ L647V. The incubation buffer was supplemented with 1% bovine serum albumin to inhibit nonspecific interactions. As shown in Fig. 2B, PIPKI␥661 was specifically retained by GST-2, whereas none of the PIPKI␥ constructs were associated with GST alone. These combined results suggest that PIPKI␥661 contains a tyrosine sorting motif that is recognized by the 2-subunit of AP2 and mediates the direct interaction between these two proteins.
Previous work performed by Ohno et al. (28) has also demonstrated that phosphorylation of the tyrosine residue within YXX sorting motifs disrupts the association of such motifs with the 2-subunit. We have previously shown that Tyr 644 of PIPKI␥661 is phosphorylated by Src in a focal adhesion kinasedependent manner (23,24). Consequently, tyrosine phosphorylation of this residue might disrupt the association between PIPKI␥661 and the 2-subunit. To address this possibility, FIGURE 1. PIPKI␥661 directly interacts with the -subunit of AP2 in vitro and in vivo. A, GST pulldown of PIPKI␥661 and the C-terminal domain of the 2-subunit. T7-tagged PIPKI␥661 was incubated with GST-2 or GST alone. Bound PIPKI␥661 was analyzed by Western blot using an antibody specific for the T7 epitope tag. The GST-2 was blotted with an antibody specific for GST. B, HEK293 cells were lysed and endogenous AP2 was immunoprecipitated (IP) with a monoclonal antibody specific for ␣-adaptin. The precipitated complexes were then resolved by SDS-PAGE and immunoblotted (IB) as indicated. The reciprocal experiment was performed in the same manner using a polyclonal antibody specific for PIPKI␥. Normal mouse and rabbit IgGs were used as negative controls. C, HEK293 cells were transfected with 5 g of wild type (WT ) or kinase-dead (KD) PIPKI␥661 via calcium phosphate. Endogenous AP2 was then immunoprecipitated using an antibody specific for ␣-adaptin. The precipitated complexes were resolved by SDS-PAGE and immunoblotted as indicated. D, HEK293 cells were transfected with 5 g of each cDNA as in C and immunoblotted as indicated. The residues shown in gray were converted to stop codons via site-directed mutagenesis, resulting in truncation of the PIPK⌱␥661 C terminus from that residue (I␥E, truncation at Glu 658 ; I␥Y, truncation at Tyr 649 ; and I␥W, truncation at Trp 642 ). The PIPKI␥661 C-terminal extension is shown for reference. JULY 21, 2006 • VOLUME 281 • NUMBER 29 we performed in vitro GST pulldowns by incubating GST-2 with PIPKI␥661 or tyrosine-phosphorylated recombinant PIPKI␥661, generated as described previously (26). As shown in Fig. 2C, tyrosine phosphorylated PIPKI␥661 associated with much lower affinity as compared with nonphosphorylated PIPKI␥661. This result is consistent with the requirement of an unphosphorylated tyrosine within the YXX sorting motif and also might serve as a regulatory mechanism for the interaction between AP2 and PIPKI␥661.

PIPKI␥661 Associates with the AP2 Complex
PIPKI␥661 and AP2 Partially Colocalize in MDCK Cells-Because we had determined that PIPKI␥661 interacts with the AP2 complex both in vivo and in vitro, we next examined whether PIPKI␥ shared a similar subcellular localization with AP2. Endogenous PIPKI␥ and AP2 were immunostained with antibodies specific for PIPKI␥ and ␣-adaptin, respectively, in MDCK cells and examined by confocal microscopy. As shown in Fig. 3A, PIPKI␥ is primarily targeted to the plasma membrane and to sites of cell-cell contacts in MDCK cells. This localization also overlaps with the punctuate plasma membrane staining of endogenous AP2, and can be observed not only at the plasma membrane, but also at discrete locations within the cytoplasm (as indicated by the arrows in the merged images). This partially overlapping staining at the plasma membrane was not surprising, as PIPKI␥661 has been shown to serve several roles at the plasma membrane, including regulation of focal adhesion assembly via direct interaction with talin (23,30). It is important to note that the PIPKI␥ polyclonal antibody employed here detects multiple PIPKI␥ splice variants expressed in MDCK cells, and we have shown that the interaction with AP2 is specific for only the PIPKI␥661 splice variant.
To further examine the specificity of the interaction between PIPKI␥661 and AP2, HA-tagged WT PIPKI␥661, PIPKI␥661 T644F, and PIPKI␥661 S645F constructs were expressed in MDCK cells. Expressed under normal conditions, all three HA-PIPKI␥661 constructs were targeted to the plasma membrane in a similar manner (data not shown). However, upon stimulation of clathrin-mediated endocytosis via treatment with transferrin, distinct colocalization patterns were observed. As shown in Fig. 3B, both WT and S645F PIPKI␥661 colocalized with AP2 in internalized vesicular structures upon treatment with transferrin. However, in cells expressing PIPKI␥661 T644F, the HA-PIPKI␥ signal remained at the plasma membrane and was not significantly internalized under identical conditions. These data collectively support both the specificity of the PIPKI␥661/ AP2 interaction demonstrated in vivo and in vitro and also reinforce the functional implications observed in transferrin uptake experiments described below.
The interaction between PIPKI␥661 and AP2 may be a transient event, occurring primarily to facilitate targeting of PIPKI␥661 to sites of endocytosis for localized generation of PI(4,5)P 2 . This dynamic association would be necessary for recognition of cargo proteins by the 2-subunit upon assembly of the AP2 complex onto the plasma membrane, because PIPKI␥661 would occupy the cargo binding site when directly associated with AP2. Consequently, the vesicular cytoplasmic colocalization patterns observed in Fig. 3, A and B, may be PIPKI␥661 directly associated with AP2 during the cycling of the endocytic machinery.
PIPKI␥661 Regulates AP-2-dependent Transferrin Endocytosis-There is considerable evidence that endocytosis from theplasmamembranemediatedbytheAP2complexisaPI(4,5)P 2dependent process (11,31). The observed endogenous colocalization and the direct interaction between the -subunit of AP2 and PIPKI␥661 collectively imply that this interaction may have implications on AP2 function. To address this possibility we generated stable MDCK cell lines inducibly expressing either wild type or kinase-dead PIPKI␥661. PIPK expression was FIGURE 2. PIPKI␥661 binds AP2 via a tyrosine sorting motif. A, HEK293 cells were transfected with 5 g of the indicated HA-tagged PIPKI␥661 construct via calcium phosphate. Endogenous AP2 was immunoprecipitated using an antibody specific for ␣-adaptin. The precipitated complexes were resolved by SDS-PAGE and immunoblotted (IB) as indicated. B, the YSPL motif of PIPKI␥661 is require for the direct association with the 2-subunit. T7-tagged PIPKI␥661, PIPKI␥635, PIPKI␥ Y644F, or PIPKI␥ L647V were coincubated with GST-2 or GST alone. Bound PIPKI␥ was analyzed by Western blot using an antibody specific for the T7 epitope tag. GST-2 was blotted with an antibody specific for GST. C, Src-mediated tyrosine phosphorylation of PIPKI␥661 disrupts its association with the 2-subunit. Phosphorylated or nonphosphorylated T7-tagged PIPKI␥661 was incubated with GST-2 or GST alone. T7and GST-tagged proteins were analyzed and blotted as described in B. Tyrosine phosphorylation (pY) of PIPKI␥661 was assessed using a phosphotyrosine-specific antibody.

PIPKI␥661 Associates with the AP2 Complex
induced by withdrawal of doxycycline from the growth media. After 72 h of expression, the cells were subjected to an endocytosis assay via treatment with Alexa Fluor 647 transferrin and the amount of internalized transferrin was then assessed via flow cytometry. The results from these assays, shown in Fig. 4, A and B, demonstrate that stable expression of wild type PIPKI␥661 resulted in an over 40% average increase of mean fluorescence intensity relative to nonexpressing cells. The opposite effect was observed in cells expressing kinase-dead PIPKI␥661, with a 25% average decrease of intensity.
To confirm that these effects on transferrin endocytosis were due to a specific interaction with PIPKI␥661, we also generated MDCK stable cell lines inducibly expressing wild type or kinase-dead PIPKI␥635. As shown in Fig. 4B, induced expression of either wild type or kinase-dead PIPKI␥635, under the same conditions, had no appreciable effect on transferrin endocytosis. In addition, similar results were also obtained using an MDCK stable cell line inducibly expressing PIPKI␥661 with a T644F mutation (data not shown). These results were consistent with both our in vitro and in vivo interaction data, which indicated that only the PIPKI␥661 splice variant containing the YSPL motif is capable of direct interaction with the -subunit of the AP2 complex.
Previous work by Morgan et al. (32) has also proposed a link between trafficking of the AP2 complex and the focal adhesion protein talin. Additionally, the established binding site for talin on PIPKI␥661 does overlap with the YSPL motif necessary for the direct interaction with AP2 (23,30). To uncouple these two distinct interactions for PIPKI␥661, we generated a stable cell line expressing the S645F mutant. As shown in Fig. 2A, this mutation does not affect binding to the AP2 complex in vivo; however, this mutation does disrupt the interaction between PIPKI␥661 and talin as observed by coimmunoprecipita- FIGURE 3. PIPKI␥ partially colocalizes with AP2 in MDCK cells. A, PIPKI␥ and AP2 were visualized by confocal microscopy using antibodies specific for PIPKI␥ (green) and ␣-adaptin (red). A merge of the two channels is shown to the far right. The arrows indicate sites of strong colocalization of endogenous PIPKI␥ and AP2, which was observed both at discrete sites on the plasma membrane and within the cytoplasm in MDCK cells. B, MDCK cells were transfected with HA-tagged wild type (wt), PIPKI␥661 T644F (Y644F), or PIPKI␥661 S645F (S645F) and treated with transferrin as described under "Materials and Methods." PIPKI␥ and AP2 were visualized by confocal microscopy using antibodies specific for HA (green) and ␣-adaptin (red). A merge of the two channels is shown to the far right. Scale bar, 10 m. tion (Fig. 4C). This phenotype was further confirmed by indirect immunofluorescence. As shown in Fig. 4D, expression of wild type PIPKI␥661 resulted in a distinct colocalization pattern with talin at the plasma membrane in MDCK cells. Expression of PIPKI␥661 S645F, on the other hand, showed little colocalization with a more diffuse talin staining pattern, similar to that observed upon expression of PIPKI␥661 T644F.
In subsequent endocytosis assays, shown in Fig. 4B, expression of the S645F mutant also had a stimulatory effect in MDCK cells, but to a greater extent than that of wild type PIPKI␥661. This higher response may simply be the consequence of a lack of competition with talin for binding to PIPKI␥661. Mutation of this residue might also inhibit Src-mediated phosphorylation of Tyr 644 . Lee et al. (33) have recently shown that both phosphorylation and mutation of Ser 645 results in diminished phosphorylation of Tyr 644 . Consequently, the S645F mutant might not be susceptible to the disruptive effect of Tyr 644 phosphorylation we have observed in vitro.
PIPKI␥661 Expression Does Not Alter Cellular PI(4,5)P 2 Levels in MDCK Cells-Because potential changes in PI(4,5)P 2 levels alone might be attributed to the observed effects, we quantified total cellular PI(4,5)P 2 levels via metabolic labeling and HPLC. The PIPKI␥661 WT and KD stable MDCK cell lines were cultured in the presence of myo-[ 3 H]inositol and protein expression was induced for the same duration as in the transferrin uptake assays. The cellular lipids were extracted, deacylated, and resolved by anion exchange. The PI(4,5)P 2 peak was identified using a deacylated PI(4,5)P 2 standard and quantified by total counts using an in-line flow scintillation counter (Fig.  5C). Induction of expression was verified by immunoblot of unlabeled cells (Fig. 5B). As seen in Fig. 5A, expression of either PIPKI␥661WT or PIPKI␥661KD did not have a significant effect on the total cellular PI(4,5)P 2 level. Because the total PI(4,5)P 2 levels are not significantly affected by increased PIPKI␥661WT or PIPKI␥661KD expression, it is likely a modulation of highly localized pools of PI(4,5)P 2 is responsible for the observed effects on transferrin endocytosis. The relatively stable level of cellular PI(4,5)P 2 is also consistent with previously reported observations for increased expression of PIPKI␥661 in other cell lines (34).
Transferrin Endocytosis Is Inhibited as a Result of Reduced PIPKI␥ Expression-An RNAi based approach was also employed as an alterative method for addressing a possible role for PIPKI␥661 in transferrin receptor endocytosis. Using an siRNA oligo specific for the human PIPKI␥, we knocked down PIPKI␥ expression in HeLa cells. These cells were then used in ϪDOX, unfilled peak). B, the mean fluorescence intensity of cells expressing each of the indicated PIPKI␥ constructs were normalized to the mean observed for cells with expression repressed by DOX and the average change in transferrin uptake is shown from three independent experiments. Error is shown as 1 S.D. C, HEK293 cells were transfected with HA-tagged wild type or S645F PIPKI␥661 (S645F) as described under "Materials and Methods." Reciprocal immunoprecipitations (IP) were performed using anti-HA, anti-talin, and normal mouse IgG (mIgG) antibodies. The precipitated complexes were resolved by SDS-PAGE and blotted as indicated. D, MDCK cells were transfected with wild type PIPKI␥661, PIPKI␥661 Y644F (Y644F), or PIPKI␥661 S645F (S645F) as in C. Talin and PIPKI␥ were visualized by confocal microscopy using antibodies specific for talin (green) and PIPKI␥ (red). A merge of the two channels is shown to the far right. Scale bar, 10 m.
the same transferrin uptake assay as utilized with the MDCK PIPKI␥661 stable cell lines. A nonspecific siRNA oligo was used as a control for normalizing transferrin uptake. As an additional control, we knocked down PIPKI␣ with an siRNA oligo specific for this isoform. As shown in Fig. 6, A and B, transferrin uptake was inhibited on average by 50% in cells transfected with PIPKI␥ siRNA. However, no significant effect was observed with either nonspecific control siRNA or with siRNA specific for PIPKI␣. The results observed for knockdown of PIPKI␣ via siRNA was also consistent with results reported previously in HeLa cells (34). The level of knockdown for each protein via RNAi is shown in Fig. 6C. A change in the expression level of the transferrin receptor could also contribute to the observed inhibition of transferrin endocytosis upon knockdown of PIPKI␥. To rule out this possibility, we assessed the transferrin receptor expression level under these conditions by Western blot. As shown in Fig. 6D, knockdown of PIPKI␥661 does not affect the overall expression level of the transferrin receptor, implying that the observed effects on endocytosis may be a direct consequence of reduced expression of PIPKI␥.

DISCUSSION
PI(4,5)P 2 is well established as a major regulatory factor for clathrin-mediated endocytosis (11). Not only does it serve to directly modulate the activity of several proteins necessary for clathrin coat assembly and vesicle formation, but it also functions as an important targeting molecule. The AP2 complex is an excellent example, as it contains two documented PI(4,5)P 2 binding sites, with one site located on the ␣-subunit and the other on the 2-subunit (7,14). However, the specific mechanism responsible for targeted generation of PI(4,5)P 2 at sites of endocytosis remains unclear.
Here we have shown that PIPKI␥661 directly interacts with the 2-subunit of the AP2 adaptor complex both in vitro and in vivo. This direct interaction is dependent on a tetrapeptide sequence in the PIPKI␥661 C terminus that conforms to a YXX consensus sorting motif. We have shown that mutation of any of the key residues within this motif results in disruption of the interaction with the 2-subunit both in vivo and in vitro. The direct interaction between AP2 and PIPKI␥661 provides a mechanism for targeting PIPKI␥661 to sites of endocytosis at the plasma membrane (Fig. 7). Consequently, this would result in generation of a highly concentrated pool of PI(4,5)P 2 at these sites.
The structure for the AP2 core provides some insight to a possible mechanism for regulation of this interaction. In the AP2 crystal, the 2-subunit is buried in a grove formed by the ␣and ␤2-subunits (7). In this closed conformation, the 2-subunit YXX docking site is positioned away from the membrane docking site of the complex. It has been proposed that phosphorylation of the linker domain of 2 might trigger a conformational change that would allow the subunit to swing out of the pocket into an open conformation and bind to cargo motifs at the plasma membrane (7,(35)(36)(37). This structural shift allows for enhanced AP2 membrane association via a direct interaction between 2 and PI(4,5)P 2 , which is not possible in the nonphosphorylated, closed conformation (16). Because PIPKI␥661 binding would occupy the cargo binding site, PIPKI␥661 may bind to AP2 in this inactive conformation.
Upon docking to the plasma membrane, PIPKI␥661 could be displaced from 2 by either a conformational change in 2 or by competition of sorting motifs of higher affinity. This dis- placement could also be facilitated by tyrosine phosphorylation of PIPKI␥661. Ohno et al. (29) has previously demonstrated that phosphorylation of the tyrosine within the YXX motif inhibits the interaction with 2 in cargo peptide binding studies (29). We have previously demonstrated that Tyr 644 of PIPKI␥661 is preferentially phosphorylated by Src (24). We have also demonstrated that tyrosine phosphorylation of PIPKI␥661 disrupts the association with the 2-subunit in vitro. Additionally, several tyrosine kinase receptors trigger Src activation upon binding to extracellular ligands (38). Conse-quently, PIPKI␥661 would likely become phosphorylated and dissociate from AP2 upon targeting to activated tyrosine kinase receptors or to sites where Src may be active. Therefore, phosphorylation of Tyr 644 on PIPKI␥661 could potentially serve as an important regulatory mechanism for the interaction between PIPKI␥661 and AP2 at the plasma membrane.
This putative mechanism for targeting of PIPKI␥661 to endocytic sites is highly reminiscent of the model previously proposed by our laboratory for targeting of PIPKI␥661 to focal adhesions. Upon targeting to focal adhesions via direct interaction with talin, PIPKI␥661 may be competitively displaced from talin by ␤1-integrin (24). Thus, both models involve a mechanism where PIPKI␥ is delivered to the plasma membrane and then competitively displaced from its binding partner, while concurrently generating a highly localized pool of PI(4,5)P 2 in a tightly regulated fashion.
Additional targeting factors may also recruit PIPKI␥661 to endocytic sites in a concerted manner. The small G-protein ADP-ribosylation factor 6 has previously been shown to enhance PIPKI␥661 activity at endocytic sites and thereby stimulate recruitment of AP2 to the plasma membrane (39). ADP-ribosylation factor 6 has also been shown to associate with the AP2 complex in coimmunoprecipitation experiments (40). PIPKI␥661 might serve as the bridging factor linking ADP-ribosylation factor 6 to the AP2 complex, thereby promoting AP2 assembly via enhanced PI(4,5)P 2 production by PIPKI␥661.
It is not surprising that several previous studies have linked various type I PIPK isoforms to clathrin-mediated endocytosis. Barbieri et al. (19) has demonstrated in a previous study that expression of wild type PIPKI␣ enhanced endocytosis of epidermal growth factor receptor, whereas expression of a kinase-dead mutant had an inhibitory effect in NR6 cells. Alternatively, Padron et al. (20) recently reported that either expression or RNAi-based knockdown of PIPKI␤ had a significant impact on transferrin endocytosis, whereas expression or knockdown of PIPKI␣ or PIPKI␥ had no effect in CV-1 and HeLa cells. The authors concluded, however, FIGURE 6. Decreased PIPKI␥ expression inhibits transferrin receptor endocytosis. A, HeLa cells were transfected with control siRNA, siRNA specific for PIPKI␥, or siRNA specific for PIPKI␣. These cells were then used in transferrin uptake assays and fluorescence intensity was quantified as described in the legend to Fig. 5A. The following colors correspond to each RNAi treatment: solid black, control siRNA; solid gray, PIPKI␣ siRNA; and broken gray, PIPKI␥ siRNA. Untransfected control cells not treated with transferrin are shown as a filled peak. B, the level of transferrin uptake was measured by the difference in the mean fluorescence obtained from ϳ10,000 cells by the fluorescence-activated cell sorting assay as described in the legend to Fig. 5. The mean intensities of PIPKI␥ and PIPKI␣ siRNA-treated cells from three independent experiments were normalized to the mean fluorescence observed in cells treated with control siRNA for each experiment. Error is shown as 1 S.D. C, immunoblot demonstrating the level of knockdown mediated by each siRNA. Cell lysates were resolved by SDS-PAGE and immunoblotted (IB) as indicated. D, HeLa cells were transfected with control or siRNA oligos specific for PIPKI␥ as in A. The cells were then lysed with sample buffer, resolved by SDS-PAGE, and immunoblotted as indicated.
that these observed effects were the result of large perturbations of cellular PI(4,5)P 2 levels upon modulation of PIPKI␤ expression.
Our results for the effects of PIPKI␣ knockdown were consistent with this study, as we also observed no significant change in transferrin endocytosis under these conditions in HeLa cells. We also observed no change in PI(4,5)P 2 levels in MDCK cells when stably expressing either wild type or kinasedead PIPKI␥661 or PIPKI␣ (data not shown). However, our results directly conflict with the results reported by Padron et al. (20), both for the endocytosis of transferrin receptor and cellular PI(4,5)P 2 levels. This might be attributed to the difference in the cell lines used (CV-1 versus MDCK) or the fact that our methods utilize stable cell lines, whereas Pardon et al. (20) used a transient transfection method for assessing a role for PIPKI␥ in transferrin receptor endocytosis. High overexpression levels of PIPKIs obtained via transient transfection often cause a nonspecific accumulation of PI(4,5)P 2 in the cytosol, resulting in the cells rounding-up or even cell death, which complicates delineation of the phenotypic effects mediated by expression of the PIPKI. Still, our results imply that the direct interaction between AP2 and PIPKI␥661 likely results in modulation of highly localized pools of PI(4,5)P 2 , rather than significant alterations of the overall cellular level of PI(4,5)P 2 in the context of elevated or knockdown protein expression levels.
Additionally, we have discovered that the 1␤ subunit of AP1B directly interacts with PIPKI␥661 and this subunit binds specifically to PIPKI␥661 in vitro (data not shown). 4 The 1␤and 2-subunits both share similar affinities for YXX sorting motifs, and this has been documented previously (28). We have observed that mutation of the YSPL motif within PIPKI␥661 to unfavorable residues has similar disruptive effects on its association with the 1-subunit in vivo (data not shown). Consistent with these observations for 1␤, PIPKI␥ knock-out mice have also been shown to have severe defects in both synaptic vesicle trafficking and exocytosis (22,41). Collectively, these data imply that PIPKI␥661 may have multiple roles in trafficking, both in endocytosis and exocytosis, via direct interactions with AP2 and AP1 complexes.
An accumulation of recently reported evidence has led to the proposal that AP2 functions in cargo recruitment, rather than serving as the primary mechanism for clathrincoated pit assembly (42). It, therefore, is possible that this direct interaction with PIPKI␥661 might only have implications on the subset of cargos recognized by AP2 (e.g. transferrin receptor). Consequently, the role for PIPKI␥661 in endocytosis may be cell-line or tissue specific. Still, the direct interaction between AP2 and PIPKI␥661 provides an attractive mechanism for spatial and temporal generation of PI(4,5)P 2 at sites of AP2-mediated endocytosis. Further examination of precisely how this interaction is modulated will lead to a greater understanding of how clathrin-mediated endocytosis is regulated in cells.