Binding of AP2 to sorting signals is modulated by AP2 phosphorylation.

The two clathrin-associated adaptor complexes AP1 and AP2 are known to participate in the formation of clathrin-coated vesicles at the trans-Golgi network and at the plasma membrane. During this process adaptors are involved in the sequestration of vesicle cargo by binding to the sorting signals within the cytoplasmic domains of the cargo proteins and in the recruitment of the clathrin coat. After budding of the clathrin-coated vesicles, the clathrin and adaptors dissociate from the vesicles. Here we show that in vitro binding of AP2 to sorting signals, which is one of the initial steps in receptor-mediated endocytosis, is modulated by adaptor phosphorylation. AP2 was phosphorylated by incubating purified AP2 in the presence of ATP and dephosphorylated by incubation with alkaline phosphatase. Affinity for tyrosine-, leucine-based and noncanonical sorting motifs was 15-33 times higher for phosphorylated than for dephosphorylated AP2. Also the binding of AP2 to membranes was regulated by adaptor phosphorylation/dephosphorylation and was about 8-fold higher for phosphorylated than for dephosphorylated AP2. Moreover, AP2 isolated from cytosol is higher phosphorylated than membrane-extracted and exhibits a 5-fold higher binding affinity than AP2 extracted from membranes. Taken together these data point to a cycle of phosphorylation/dephosphorylation as a mechanism for regulating the reversible association of AP2 with membranes and sorting signals during the process of receptor-mediated endocytosis.

Trafficking of membrane proteins within the secretory and endocytic route of eukaryotic cells is characterized by the sorting of the membrane proteins at sites where transport pathways diverge and the packaging of the membrane proteins into vesicles occurs. The adaptors AP1 and AP2 appear to play a key role in the sorting and packaging of membrane proteins in clathrin-coated vesicles (CCVs). 1 AP1-containing CCVs are formed at the TGN and facilitate the transport of cargo from the TGN to endosomes, whereas AP2-containing CCVs function in receptor-mediated endocytosis at the plasma membrane. The adaptors are heterotetrameric complexes composed of two 100-kDa subunits (designated ␥ and ␤1 in AP1 and ␣ and ␤2 in AP2; for review see Ref. 1). The ␣-, ␤1-, and ␤2-subunits have been implicated in clathrin binding (2)(3)(4). Recognition of the membrane proteins is mediated by the medium subunits 1 and 2 (5), but also the ␤-subunits have been implicated in binding to membrane targets (6,7). Tyrosine-and leucinebased sorting motifs in cytoplasmic tails of membrane proteins can serve as adaptor binding sites, and the tyrosine and leucine residues have been shown to be critical for the sorting of the membrane proteins, as well as for the adaptor binding (for review see Ref. 8). For noncanonical sequences mediating either sorting or adaptor binding, this correlation, however, remains to be established. After formation of the CCVs, the clathrin coat is removed by the uncoating ATPase (hsc70) in a reaction depending on ATP hydrolysis (9 -11). The adaptors are removed by a yet uncharacterized reaction.
Thus adaptors associate with specific membranes such as the TGN or the plasma membrane and dissociate from the vesicles budding from these membranes. The mechanisms that control the association of adaptors with and their dissociation from membranes are unknown. In principle three different mechanisms can be envisaged. First, the adaptors may become specifically modified before they bind to or dissociate from membranes. Second the targets in membranes, to which adaptors bind, are subject to a covalent modification prior to binding or to dissociation. Third, a change of the internal milieu of CCVs, e.g. by an ion pump, may translate into a conformational change of the target e.g. by altering its oligomeric state.
Several components of CCVs are subject to phosphorylation, including clathrin, adaptors, and cytoplasmic tails of cargo proteins, which serve as membrane targets (12)(13)(14). Furthermore several kinases are associated with CCVs and purified adaptors (15)(16)(17). The phosphorylation of membrane-associated and cytosolic AP2 has been shown to be different, with cytosolic AP2 being the higher phosphorylated species (13). These observations point to a role of adaptor phosphorylation for regulating its binding to membranes. In the present study we analyzed the effect of phosphorylation and dephosphorylation of purified adaptors on the binding to peptides representing adaptor binding sequences, and we also analyzed AP2 binding to membranes. We observed that phosphorylation of AP2 enhances the association constant for in vitro binding to various AP2 binding motifs and the recruitment of AP2 to membranes, whereas dephosphorylation has the opposite effect. The phosphorylation of the ␣-, ␤2-, and 2-subunit is catalyzed by kinases copurifying with AP2. Furthermore, cytosolic AP2 is higher phosphorylated and has a higher binding affinity as compared with membrane-extracted AP2. Taken together these data suggest that a cycle of phosphorylation and dephosphorylation of AP2 is regulating the binding of AP2 to membranes.

MATERIALS AND METHODS
Cells and Antibodies-NRK and MDBK cells were obtained from ATCC (Manassas, VA) and cultured in Dulbecco's modified Eagle's medium, 10% fetal calf serum. Antibodies specific for subunits of AP2 were obtained from Sigma (anti-␥ and anti-␣) and from Transduction Laboratories (anti-␣ and anti-␤). The antibodies specific for 2 and 2 were kindly provided by M. S. Robinson (Cambridge, United Kingdom).
Preparation of AP1 and AP2-Clathrin-coated vesicles were prepared from porcine brain according to Keen et al. (18). Coat proteins were extracted from CCVs with 0.5 M Tris-HCl, pH 7.5, for 1 h under rotation. After centrifugation at 100,000 ϫ g for 30 min, the material was applied to a Superose-6 column (1.6 ϫ 55 cm) connected to a fast protein liquid chromatography system (Amersham Pharmacia Biotech) at a flow rate of 0.3 ml/min. The column was equilibrated in 0.5 M Tris-HCl, pH 7.0, 0.2 mM dithiothreitol. 1.5-ml fractions were collected, and adaptor-containing fractions were identified by SDS-PAGE. For the separation of AP1 from AP2, the adaptor-containing fractions were pooled and applied to hydroxyapatite chromatography according to Manfredi and Bazari (19). The purity of AP1 and AP2 was analyzed by SDS-PAGE and Western blotting using monoclonal antibodies to the ␣-subunit of AP2 or the ␥-subunit of AP1.
Peptide Synthesis-Peptides were synthesized using amino acids protected with Fmoc (N-(9-fluorenyl)methoxycarbonyl) and activated with benzotriazol-1-yl-oxytripyrolidinophosphonium hexafluorophosphate and a 9050 peptide synthesizer (Millipore). After cleavage from the resin and the protecting groups, peptides were purified by reverse phase HPLC using Delta Pac C-18 column (Millipore) and an elution from 0 -50% acetonitrile in 0.1% trifluoroacetic, water for 50 min. Purity was confirmed by HPLC, UV spectrometry, and mass spectrometry. Peptides corresponding to the tail sequences of the following membrane proteins were used in this study: from lysosomal acid phosphatase (RMQAQPPGYRHVADGQDHA), from the 67 residues comprising MPR46 tail the residues 2-16 (RLVVGAKGMEQFPHL) and 49 -67 (GDDQLGEESEERDDHLLPM), and from invariant chain (MDDQ-RDLISNNEQLP-MLGRRPGAPESKCSR). The latter one was produced as a glutathione S-transferase fusion protein and was used as described (20).
In Vitro Phosphorylation of AP2 with [␥-32 P]ATP-AP2 was dialyzed against 50 mM Hepes-KOH, pH 7.4, 5 mM MgCl 2 , 2 mM MnCl 2 and incubated with 1 l of [␥-32 P]ATP (3000 Ci/mmol; Amersham Pharmacia Biotech) for 15 min at room temperature. The reaction was stopped by the addition of sample buffer and boiling for 5 min at 95°C. The probes were analyzed by SDS-PAGE, autoradiography, and Western blotting with antibodies against the ␣-, ␤2and 2-subunits. For experiments including kinase inhibitors, a mixture of threonine and serine kinase inhibitors was used (catalog number 539572; Calbiochem). For phosphorylation with unlabeled ATP, AP2 was dialyzed into Buffer A and incubated for 30 min at 37°C in the presence of 2 mM ATP. To test for phosphatases copurifying with AP2, the adaptor was first incubated with [␥-32 P]ATP for 30 min at 37°C, was then treated with kinase inhibitors, and further incubated for up to 2 h at 37°C. Subsequently, the samples were resolved on SDS-PAGE and analyzed after autoradiography. AP2 that was incubated in parallel with unlabeled ATP was analyzed for tail binding by using the biosensor. For phosphorylation with unlabeled ATP, AP2 was dialyzed into Buffer A and incubated for 30 min at 37°C in the presence of 2 mM ATP.
Preparation of NRK Membranes-NRK cells were grown to confluency on a 15-cm dish. The cells were washed two times with phosphatebuffered saline, scraped in Buffer B (25 mM Hepes-KOH, pH 7.2, 125 mM KAc, 5 mM MgAc), and lysed by passing the suspension through a 22ϫg syringe 10 times. The cell lysate was centrifuged for 10 min at 1000 ϫ g to remove intact cells and nuclei. The supernatant was removed and centrifuged for 30 min at 100,000 ϫ g to separate membranes (pellet) from the cytosol (supernatant). The membrane pellet was incubated for 1 h with 0.5 M Tris-HCL, pH 7.5, to remove endogenous AP2. Subsequently the membranes were recovered by centrifugation at 100,000 ϫ g, solubilized in Buffer B, and stored in aliquots at Ϫ70°C until use.
Recruitment of AP2 onto NRK Membranes-AP2 was dialyzed in Buffer B, pre-incubated with alkaline phosphatase or ATP, and incubated with NRK membranes for 30 min on ice. The membranes were recovered by centrifugation at 100,000 ϫ g for 15 min, solubilized in 2ϫ sample buffer, resolved by SDS-PAGE, and analyzed by Western blotting with an antibody against the ␣-subunit of AP2.
Surface Plasmon Resonance Interaction Analysis-The interaction between AP1, AP2, or MDBK cytosolic and membrane fractions and the cytoplasmic tail peptides was analyzed in real time by surface plasmon resonance using a BIAcore 2000 biosensor (BIAcore AB). The peptides were coupled to a CM5 sensor chip via their primary amino groups following the manufacturer's instructions (immobilized ligands). Peptides with an isoelectric point below 3.5 could not be immobilized to the CM5 sensor chip via amino coupling because of their low pI. These peptides were immobilized to the sensor surface via the thiol group of a cysteine residue that was synthesized at the amino terminus (for details see Ref. 21). AP1 and AP2 (analytes) were injected at a flow rate of 20 l/min unless stated otherwise. The adaptors were used at concentrations ranging from 50 -350 nM to avoid mass transport effects that can occur if the analyte concentration is low. For cytosolic and membraneextracted adaptors, concentrations were used ranging from 5-11 nM because of the limited amount of adaptors. Association (1-2 min) was followed by dissociation (2 min) during which Buffer A was perfused. Subsequently bound APs were removed by a short pulse injection (15 s) of 10 mM NaOH, 0.5% SDS.
Determination of Kinetic Rate Constants-The rate constants (k a for association and k d for dissociation) of the interaction between tail peptides and the adaptor complexes were calculated by using the evaluation software of the BIAcore 2000. Association was determined 15-20 s after switching from buffer flow to adaptor solution to avoid distortions because of injection and mixing. The dissociation rate constants were determined 5-10 s after switching to buffer flow. The association rate constant k a , the dissociation rate constant k d , and the calculation of the equilibrium rate constant, K D ϭ k d /k a , were determined by assuming a first order kinetic A ϩ B ϭ AB. Further details are described elsewhere (22,23).
Fractionation of MDBK Cells-MDBK cells were harvested in 100 mM Mes, pH 6.8, 1 mM EGTA, 0.5 mM MgCl 2 , and 0.2 mM dithiothreitol and lysed by passing the suspension through a 22ϫg syringe 10 times. The cell lysate was centrifuged at 1000 ϫ g to remove intact cells and nuclei. The supernatant was removed and centrifuged for 30 min at 100,000 ϫ g to separate membranes from the cytosol. The pellet was extracted with 0.5 M Tris, pH 7.5, and centrifuged for 30 min at 100,000 ϫ g.
Gel Filtration-For gel filtration the cytosol and the membrane extract (50-l aliquots) were passed over a Superdex-200 column connected to a SMART™ system (Amersham Pharmacia Biotech), equilibrated, and eluted with Buffer A at a flow rate of 40 l/min. Fractions of 50 l were collected and analyzed by Western blotting using an antibody against the ␣-subunit of AP2. The AP2-positive fractions were combined and stored at Ϫ70°C.
Isoelectric Focusing-AP2 was dialyzed against Buffer A and incubated with alkaline phosphatase for 30 min at 37°C. AP2 was precipitated with methanol/chloroform (24), subjected to isoelectric focusing according to Braun et al. (25), transferred onto nitrocellulose, and analyzed by Western blotting with antibodies against the four subunits of AP2.

Dephosphorylation of AP2 Reduces in Vitro
Binding to the Cytoplasmic Tail of MPR46 -AP1 and AP2 are phosphoproteins (13,16,17), and both kinases and phosphatases copurify with adaptor complexes (15,17,18). To test whether dephosphorylation of adaptors affects their binding to the cytoplasmic tails of transmembrane cargo proteins of CCVs, AP1 and AP2 purified from pig brain were treated with alkaline phosphatase. Following this treatment, the binding of adaptors to a peptide derived from the cytoplasmic tail of MPR46 was analyzed by surface plasmon resonance. MPR46 is a cargo molecule of AP1 CCVs at the TGN and of AP2 CCVs at the plasma membrane. The peptide representing residues 49 -67 of the cytoplasmic tail of MPR46 is known to bind AP1 and AP2 with high affinity (22). A corresponding peptide in which the critical dileucine motif was mutated to a pair of alanines served as a control for AP2 binding. Treatment with phosphatase had only a minor effect on the affinity of AP1, whereas the affinity of AP2 was decreased about 20-fold from a K D of 16 nM to a K D of 366 nM. The decrease in affinity was because of a lower asso-ciation rate constant (see Fig. 1 and Table I). The effect of alkaline phosphatase treatment was dependent on the concentration of alkaline phosphatase (Fig. 2) and sensitive to the inhibition by 1 mM pyrophosphate or 1 mM vanadate (data not shown).
In bovine kidney cells all subunits of AP2 except for the 2-subunit are known to exist as phosphoproteins (13). Isoelectric focusing of AP2 prior to and after phosphatase treatment confirmed these data for AP2 subunits from pig brain. The ␣-, ␤2-, and 2-subunit exist as multiple forms, whereas the 2subunit exists as a single form. Phosphatase treatment shifted the pattern of the ␣-, ␤2-, and 2-subunits toward more basic forms, whereas it did not affect the isoelectric point of 2 (shown for ␣, ␤2, and in Fig. 3). We conclude from these data that three of the four subunits of AP2 are phosphorylated and that dephosphorylation of AP2 reduces its high affinity in vitro binding to a peptide derived from a transmembrane cargo protein of CCVs.
Dephosphorylation of AP2 Reduces Binding to Tyrosine-and Leucine-based and Noncanonical Sorting Motifs-AP2 recognizes in cytoplasmic tails of transmembrane proteins tyrosineand leucine based-and noncanonical sorting motifs (8,26). The tail peptide 49 -67 of MPR46 contains a leucine-based and a noncanonical AP2 binding motif that can substitute each other (22). To determine whether dephosphorylation of AP2 reduces binding to either of these motifs, binding to the lysosomal acid phosphatase (LAP) tail peptide, the invariant chain tail peptide, and the MPR46 tail peptide 2-16 was determined. These peptides are known to bind AP2 via a tyrosine-based signal (LAP, 27), a leucine-based signal (invariant chain, 20), or a noncanonical sorting motif MPR46, 22). Alkaline phosphatase treatment reduced the affinity to any of the three classes of AP2 binding motifs by 4-to 15-fold (Table II). As for the MPR46 peptide 49 -67 this decrease in affinity resulted almost exclusively from lower association rate constants (not shown).
To test the influence of AP2 phosphorylation and dephospho- Binding of AP1 and AP2 purified from pig brain to the MPR46 tail peptide 49 -67 was recorded in real time using a biosensor as described under "Materials and Methods." A mutant tail peptide in which the critical dileucine motif was substituted for a pair of alanines served as a control for binding specificity. The adaptors were used at a concentration of 350 nM and injected for 2 min at a flow rate of 20 l/min. Prior to binding to the tail peptide, the adaptors were incubated in the presence or absence of 2 units of alkaline phosphatase (30 min at 37°C). It is notable that the incubation with phosphatase did only significantly decrease the affinity of AP2 for the MPR46 tail peptide. For calculation of the binding constants see Table I. FIG. 2. Dose-dependent effect of alkaline phosphatase on AP2 binding. AP2 was incubated with various amounts of alkaline phosphatase prior to recording AP2 binding to the MPR46 tail peptide 49 -67 as described in the legend to Fig. 1. The equilibrium rate constants for the AP2 binding to the tail peptide was determined (K D ϭ k d /k a ), and the values are plotted against the amount of phosphatase. Purified AP1 and AP2 were incubated for 30 min at 37°C in the absence or presence of 2 units of alkaline phosphatase. Subsequently, binding of AP1 and AP2 to the MPR46 tail peptide 49 -67 was recorded with a biosensor as described under Fig. 1. The rate constants for association (k a ), dissociation (k d ), and the equilibrium rate constant (K D ϭ k d /k a ) were determined from the sensorgrams shown in Fig. 1 Peptides corresponding to the cytoplasmic tails of LAP, invariant chain, and MPR46 were immobilized on a sensor surface and probed for AP2 binding as described above. Prior to tail peptide binding, AP2 was kept untreated or incubated with ATP (2 mM) or alkaline phosphatase (2 units) for 30 min at 37°C. The equilibrium rate constants (K D ) for the AP2 interaction with the various tail peptides were calculated from the BIAcore sensorgrams. The decrease or increase of the K D caused by preincubation with ATP or alkaline phosphatase resulted from changes of the association rate constants (not shown).  rylation more precisely, we purified AP2 and devided the preparation into three aliquots. One aliquot was treated with alkaline phosphatase, one was incubated in the presence of ATP to allow phosphorylation by copurified kinases, and one aliquot served as a nontreated control. One-third of each sample was subsequently analyzed for binding to the MPR46 tail peptide 49 -67 (Fig. 5B), whereas the remaining sample was subjected to isoelectric focusing, followed by Western blotting and detection of the AP2 ␤-subunit (Fig. 5A). As shown in Fig. 5A, the purified AP2 ␤-subunit (mock) exists as two major forms with pIs of 6.5 and 7.2 (Fig. 5A, lane 1). Incubation with alkaline phosphatase resulted in a shift of almost all ␤2 to the basic form (lane 2). When incubated in the presence of ATP, ␤2 was converted to more acidic forms (lane 3). A similar result was obtained for the ␣-subunit (data not shown). This indicates that the bulk of isolated AP2 represents a partially phosphorylated form (dash), which is converted by dephosphorylation or phosphorylation into more basic (open triangle) or acidic (filled triangle) forms, respectively. In conclusion, dephosphorylation of AP2 with alkaline phosphatase or phosphorylation by copurifying kinases changes the phosphorylation state of the entire AP2 pool and not just of a fraction of it. When the same aliquots were analyzed for binding to the MPR46 tail peptide 49 -67 (see Fig. 5B and Table II), the phosphorylation of AP2 by endogenous kinases led to a 2.5-fold enhanced association rate to the peptide 49 -67 of MPR46 (Fig.  5B, ϩATP) as compared with untreated AP2 (mock). On the other hand, the phosphatase-treated AP2 sample exhibited a 20-fold decreased affinity for the tail peptide, clearly demonstrating that the phosphorylation status of AP2 is a critical determinant for the binding to sorting signals.
Phosphorylation also enhanced the affinity of AP2 to peptides representing tyrosine-and leucine-based and noncanonical AP2 binding motifs 2-to 4-fold (Table II). The presence of 150 g/ml polylysine during the phosphorylation reaction affected neither the incorporation of 32 P into nor the affinity of FIG. 6. Recruitment of AP2 to membranes. AP2 (3 g) was incubated with ATP (2 mM; upper panel) or alkaline phosphatase (2 units; lower panel) for 30 min at 37°C. As a control, AP2 was kept at 37°C without any additives. Subsequently, AP2 was incubated with crude membranes (10 g) prepared from NRK cells for 30 min on ice. The membranes had been treated with 0.5 M Tris to remove endogenous AP2 prior to the incubation with bovine AP2. Following the incubation, the membranes were pelleted, subjected to SDS-PAGE, and analyzed for the amount of bound AP2 by Western blotting using an ␣-adaptinspecific antibody. Whereas preincubation of AP2 with ATP stimulated the binding to membranes, preincubation with phosphatase decreased the amount of AP2 bound to membranes. Quantification is given in "Results. "   FIG. 3. Effect of alkaline phosphatase on AP2 subunits. 20 g of purified pig brain AP2 were incubated in the presence (ϩ) or absence (-) of 2 units of alkaline phosphatase for 30 min at 37°C. Subsequently the samples were precipitated and subjected to isoelectric focusing, followed by transfer onto nitrocellulose. The different AP2 subunits were analyzed by Western blotting using subunit-specific antibodies. The solid triangles point to the subunit isoforms that decrease or increase in concentration because of shifting from a more acidic pH value to a more basic pH value upon incubation with phosphatase. The open triangle indicates the position of the 2-subunit, which is the only AP2 subunit resistant to alkaline phosphatase. The asterisk (*) indicates a nonspecific signal present in the ␤2 blot. FIG. 5. Effect of AP2 phosphorylation/dephosphorylation on binding to MPR46 peptide 49 -67. AP2 was incubated with 2 mM ATP for 30 min at 37°C with 2 units of alkaline phosphatase or kept untreated (mock). Subsequently, two-thirds of the sample was subjected to isoelectric focusing followed by Western blotting and detection of the AP2␤-subunit (A). The remaining sample was analysed for binding to the MPR46 tail peptide 45-67 as described above (B). Note that tail binding of AP2 was of higher affinity upon preincubation with ATP, whereas dephosphorylation of AP2 reduced its binding affinity by more than 20-fold (see Table II for the numbers). AP2 (not shown). We conclude from these data that in vitro phosphorylation of AP2 by associated kinases enhances the affinity of AP2 to the different classes of AP2 binding motifs.
When phosphorylated AP2 was incubated in the presence of general protein kinase inhibitors for 30 min at 37°C, no dephosphorylation was detectable (not shown). This suggests that in contrast to kinases, endogenous phosphatases are not associated with purified AP2.
Phosphorylation and Dephosphorylation of AP2 Modulates the Recruitment of AP2 to Membranes-The experiments described above had demonstrated the dependence of in vitro binding of AP2 to immobilized peptides by the phosphorylation status of AP2. Next we assayed whether phosphorylation or dephosphorylation of AP2 affects its binding to membranes. A membrane fraction was prepared from NRK cells and stripped from endogenous adaptors by treatment with 0.5 M Tris, pH 7.5. The purified AP2 was incubated for 30 min at 37°C in the presence of ATP, and subsequently the AP2 was placed on ice and incubated for 30 min in the presence of membranes. Membrane-associated AP2 was then pelleted and quantified by Western blotting (Fig. 6). Incubation of AP2 in the presence of ATP increases the amount of membrane-bound AP2 as compared with a control in which AP2 was incubated in the absence of ATP by 4.1 Ϯ 1.0-fold (n ϭ 5). Under the experimental conditions used, the ATP that was used to stimulate the phosphorylation of AP2 may also affect the phosphorylation status of some membrane components even at 4°C. To rule out this possibility, AP2 was cleared from ATP by centrifugation through a high molecular weight cut-off filter and then incubated with the membranes. The stimulation of AP2 binding was similar as before (data not shown), indicating that it is the phosphorylation of AP2 that increases the AP2 association with membranes. The amount of membrane binding of AP2 was lowered by 2.0 Ϯ 1.1-fold (n ϭ 4) when the complex was treated with alkaline phosphatase for 30 min at 37°C and then incubated on ice 30 min with membranes in the presence of 1 mM pyrophosphate/1 mM vanadate as compared with control AP2. This clearly indicates that recruitment of AP2 to membranes is controlled by the phosphorylation of AP2.
Cytosolic and Membrane-associated AP2 Differ in Phosphorylation and Affinity to Cytoplasmic Tail Peptides-In MDBK cells the ␤2-subunit of cytosolic AP2 is higher phosphorylated than the ␤2-subunit of membrane-associated AP2. In addition the incorporation of 32 P into the ␣and ␤2-subunits is higher for cytosolic than for membrane-associated AP2 (13). Isoelectric focusing of cytosolic and membrane-extracted AP2, followed by Western blot analysis for the ␣-subunit, showed that cytosolic AP2 is enriched in acidic ␣-subunit forms. This indicates that the steady state phosphorylation of the ␣-subunit is higher for cytosolic AP2 than for membrane-extracted AP2 (Fig. 7).
The cytosolic and membrane-derived extracts contain both AP1 and AP2. To distinguish AP2 binding from that of AP1, we used the MPR46 tail peptide 2-16, which binds AP2 but not AP1 as an affinity matrix (22). Binding was determined after adjusting the extracts to 10 nM AP2 (Fig. 8). The affinity of the cytosolic AP2 (K D ϭ 11 nM) was about 5-fold higher than that of the membrane-derived AP2 (K D ϭ 61 nM). As observed for the in vitro binding of phosphorylated and dephosphorylated AP2, the difference in AP2 affinity resulted mainly from different association rate constants (K a ϭ 2.8 ϫ 10 5 ϫ M Ϫ1 ϫ s Ϫ1 for cytosolic AP2 versus a K a ϭ 0, 23 ϫ 10 5 ϫ M Ϫ1 ϫ s Ϫ1 for membrane-extracted AP2). DISCUSSION The results reported in this study support a cycle of phosphorylation and dephosphorylation of AP2 as a mechanism to regulate the cycle of association and dissociation of AP2 with membranes during receptor-mediated endocytosis. Phosphorylation increases the affinity of AP2 for sorting signals in cargo proteins segregating into CCVs. The affinity of phosphorylated and dephosphorylated AP2 for different sorting signals differed by a factor of 15 to 33. The recruitment of phosphorylated and dephosphorylated AP2 to membranes differed by a factor of 8. Thus, the affinity to sorting signals does not exactly parallel that to membranes. Although phosphorylation increased the affinity of AP2 for sorting signals (2-to 4-fold) and to membranes (4-fold) to a similar extent, dephosphorylation decreased the binding to membranes only by 2-fold, whereas the affinity to sorting signals decreased 3-to 15-fold. This suggests that lowering the affinity of AP2 for sorting signals below a certain threshold does not further decrease AP2 recruitment to membranes. The residual recruitment of dephosphorylated AP2 to membranes (see Fig. 6) is therefore likely to be because of the interaction of AP2 with other membrane components such as membrane-associated proteins or lipids.
AP2 binds at the cytoplasmic face of the plasma membrane to membrane proteins that contain the appropriate signals for sorting into CCVs. In addition, AP2 interacts with a number of FIG. 8. Binding of cytosolic and membrane-extracted AP2 to the MPR46 peptide 2-16. AP2 was prepared from the cytosol and from membranes of MDBK cells as described in the legend to Fig. 7. Subsequently, equal amounts of the cytosolic and the membrane-extracted AP2 were passed over a sensor surface to which the AP2 binding peptide 2-16 from MPR46 was immobilized. For clarity, AP2 binding to the mutated tail peptide 2-16, which was not detectable, is not shown. Note that binding of AP2 prepared from cytosol was of higher affinity as compared with membrane-extracted AP2. Binding constants are given in "Results." FIG. 7. Isoelectric focusing pattern of the ␣-subunit of cytosolic and membrane-extracted AP2. Cytosol and crude membranes were prepared from MDBK cells in the presence of phosphatase inhibitors. The membranes were then incubated with Tris to extract AP2. This membrane-derived extract and the cytosol was passed over a Superdex-200 gel-filtration column connected to a SMART™ system. AP2-enriched fractions were adjusted to equal concentrations of AP2, subjected to isoelectric focusing, and analyzed by Western blotting for the AP2 ␣-subunit. Note that the isoelectric focusing pattern of the ␣-subunit from the bovine kidney cell line differs from that of the ␣-subunit from pig brain (see Fig. 3). The triangles indicate the two major ␣-subunits isoforms, whose relative abundance is different in cytosolic and membrane-extracted AP2.
proteins involved in the coat assembly, the vesicle fission, and coat disassembly, including clathrin (2, 3), amphiphysin (28), synaptotagmin (29), Eps15 (30,31), Shc (32), and epsin (33). Furthermore, AP2 interacts with polyphosphoinositides, which supports the recruitment of AP2 to sites of CCV formation (34). Phosphorylation of AP2 has been shown to inhibit its binding to clathrin (13). Most of the proteins involved in clathrin-mediated endocytosis are known to exist as phosphoproteins, and their assembly and disassembly has been shown to be controlled by phosphorylation. Phosphorylation of amphiphysin inhibits its binding to AP2 and to clathrin, whereas the assembly of amphiphysin, dynamin1, and synaptojanin into complexes that include clathrin and AP2 from brain extracts is promoted by dephosphorylation (35). This has led to the concept that dephosphorylation regulates the assembly of cytosolic proteins to endocytic coat complexes (35).
The data of this study suggest a modification of the concept that dephosphorylation promotes coat assembly in general. Phosphorylation of AP2 was found to promote its binding to the sorting signals present in the cytoplasmic domains of membrane proteins to be integrated in endocytic CCVs and its recruitment to membranes. As AP2 binding to the cargo proteins of CCVs is considered to represent one of the initial steps in formation of endocytic CCVs (36), the following sequence of phosphorylation and dephosphorylation of AP2 is proposed. Phosphorylated AP2 is recruited to the cytoplasmic face of the plasma membrane. After binding to the membrane, it is dephosphorylated. This allows recruitment of clathrin and of other components assisting coat assembly and CCV fission, such as amphiphysin, synaptojanin, and dynamin1. Their assembly is supported by dephosphorylation. The uncoating of endocytic CCVs is initiated by the removal of the clathrin lattice, which is catalyzed by the hsc70 uncoating ATPase and auxilin (37,38). The dissociation of AP2 and presumably also of the other accessory proteins assisting coat assembly and vesicle fission is promoted by their phosphorylation.
This model resembles that which has been proposed for the initial steps of clathrin-mediated endocytosis of ligand-activated G protein-coupled receptors, where the nonvisual arrestins replace AP2 and link the receptors to the clathrin coat (39 -41). The cytoplasmic nonvisual ␤-arrestin1 is constitutively phosphorylated. Its binding to agonist-occupied receptors is promoted by the phosphorylation of the receptors by receptor kinases, and the subsequent binding of clathrin requires the dephosphorylation of ␤-arrestin1 (42).
It remains to be determined whether phosphorylation of cytoplasmic AP2 is constitutive as that of cytoplasmic ␤-arres-tin1. Kinases that convert AP2 into its high affinity form for membrane binding were found to be closely associated with AP2. Although known for many years (15)(16)(17)43), these kinases remain to be identified. Although adaptor-phosphorylating kinases that are influenced by polylysine have been reported for than 10 years (16,44), the kinases associated with AP2 purified from pig brain CCVs were found to be polylysineinsensitive. If the kinases are constitutively active in the cytoplasm, AP2 recruitment to the membrane would probably need to be controlled by a mechanism that modifies its membrane targets as has been established for the binding of ␤-arrestin1 to phosphorylated G protein-coupled receptors.
With respect to AP2-specific phosphatases, it has recently been shown that dephosphorylation of the AP2 ␤-subunit is mediated by a PP2A phosphatase (43). Endocytosis was blocked if the phosphatase was inhibited. Interestingly, the authors observed that in cells where PP2A was blocked, still a large fraction of AP2 was membrane-associated. This is in line with our results, showing that the initial membrane associa-tion is phosphorylation-dependent. In this context one can speculate whether the AP2-specific phosphatase is activated by the membrane binding of AP2, by a conformational change of AP2, or by recruiting AP2 to the site of the phosphatase. Moreover we still do not know whether a single phosphatase is able to dephosphorylate the ␣-, ␤-, and -subunits of AP2 simultaneously or whether the dephosphorylation is subunit-specific. Additionally, the phosphorylation, as well as the dephosphorylation of the individual subunits, is likely to be dependent on the functional status of AP2 and can be influenced by other known or unknown AP2 binding factors, offering the possibility of a tight regulation of the adaptor function.
The identification and characterization of the kinases and the phosphatases controlling the phosphorylation of AP2 and other proteins involved in formation of endocytic CCVs will deepen our understanding of how the assembly and disassembly of the endocytic machinery is regulated.