Arrestin/Clathrin Interaction - The Journal of Biological Chemistry

THE JOURNAL OF BIOLOGICAL CHEMISTRY © 1997 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 272, No. 23, Issue of June 6, pp. 15017–15022, 1997 Printed in U.S.A.

Arrestin/Clathrin Interaction LOCALIZATION OF THE ARRESTIN BINDING LOCUS TO THE CLATHRIN TERMINAL DOMAIN* (Received for publication, February 12, 1997)

Oscar B. Goodman, Jr., Jason G. Krupnick, Vsevolod V. Gurevich‡, Jeffrey L. Benovic, and James H. Keen§ From the Departments of Biochemistry and Molecular Pharmacology, and Microbiology and Immunology, Kimmel Cancer Institute, Thomas Jefferson University, Philadelphia, Pennsylvania 19107

Previously we demonstrated that nonvisual arrestins exhibit a high affinity interaction with clathrin, consistent with an adaptor function in the internalization of G protein-coupled receptors (Goodman, O. B., Jr., Krupnick, J. G., Santini, F., Gurevich, V. V., Penn, R. B., Gagnon, A. W., Keen, J. H., and Benovic, J. L. (1996) Nature 383, 447– 450). In this report we show that a short sequence of highly conserved residues within the globular clathrin terminal domain is responsible for arrestin binding. Limited proteolysis of clathrin cages results in the release of terminal domains and concomitant abrogation of arrestin binding. The nonvisual arrestins, b-arrestin and arrestin3, but not visual arrestin, bind specifically to a glutathione S-transferase-clathrin terminal domain fusion protein. Deletion analysis and alanine scanning mutagenesis localize the binding site to residues 89 –100 of the clathrin heavy chain and indicate that residues 1–100 can function as an independent arrestin binding domain. Site-directed mutagenesis identifies an invariant glutamine (Glu-89) and two highly conserved lysines (Lys-96 and Lys-98) as residues critical for arrestin binding, complementing hydrophobic and acidic residues in arrestin3 which have been implicated in clathrin binding (Krupnick, J. G., Goodman, O. B., Jr., Keen, J. H., and Benovic, J. L. (1997) J. Biol. Chem. 272, 15011–15016). Despite exhibiting high affinity clathrin binding, arrestins do not induce coat assembly. The terminal domain is oriented toward the plasma membrane in coated pits, and its binding of both arrestins and AP-2 suggests that this domain is the anchor responsible for adaptor-receptor recruitment to the coated pit.

Receptor-mediated endocytosis is a pathway by which plasma membrane receptors are internalized selectively into cells through clathrin-coated pits (for review, see Refs. 1 and 2). Receptors are either internalized constitutively (e.g. some nutrient receptors) or internalized preferentially following ligand binding (e.g. some hormone receptors). Among those that belong to the latter category are several G protein-coupled recep-

* This work was supported by National Institutes of Health Grants GM-28526 (to J. H. K.), GM-44944 (to J. L. B.), and 5-T32-DK07705-03 (to O. B. G., Jr. and J. G. K.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ Present address: Sun Health Research Institute, Sun City, AZ 85372. § To whom correspondence should be addressed: Kimmel Cancer Institute, Thomas Jefferson University, 233 S. 10th St., BLSB 915, Philadelphia, PA 19107. Tel.: 215-503-4624; Fax: 215-923-1098; E-mail: [email protected]. This paper is available on line at http://www-jbc.stanford.edu/jbc/

tors (GPRs).1 Many GPRs are phosphorylated rapidly upon agonist activation, and it is this phosphorylation event that leads to the selective binding of a family of proteins termed arrestins. To date there are four known arrestins: two visual arrestins (arrestin and cone arrestin) and two nonvisual arrestins (b-arrestin and arrestin3; for review, see Ref. 3). One role of arrestin binding is to uncouple the GPR from its cognate heterotrimeric G protein (desensitization), probably through direct competition with the G protein for the activated GPR (4). Following desensitization, the receptor is believed to be translocated into an intracellular compartment (5), a process commonly referred to as sequestration. Interestingly, sequestration also appears to be required for resensitization of receptors by allowing for receptor dephosphorylation and recycling (6, 7). Studies of the m2 muscarinic acetylcholine receptor (8) and a sequestration-deficient mutant form of the b2-adrenergic receptor (9) have implicated GPR kinases and arrestins in GPR sequestration. Recent work from our laboratories has provided a plausible molecular mechanism for this internalization process, in which arrestins function as clathrin adaptors in the uptake of b2-adrenergic (10) and other2 GPRs. We found that nonvisual arrestins bind to clathrin with a Kd of 10 – 60 nM, comparable to that of the clathrin associated protein AP-2 (12, 13). Moreover, b2-adrenergic receptor, b-arrestin, and clathrin colocalize in vivo upon agonist addition in an arrestin-dependent manner, indicating that the arrestin/clathrin interaction observed in vitro likely occurs in vivo in the presence of an activated receptor. These results suggest that the arrestin/ clathrin interaction is of central importance in regulating GPR trafficking. Clathrin, the major structural component of the coated pit lattice, consists of three heavy chains (Mr ' 192,000) and three light chains (Mr ' 26,000) (for review, see Ref. 14). By limited proteolysis of preformed clathrin cages, the heavy chain has been shown to comprise two distinct domains (15). An '50-kDa amino-terminal globular region, the terminal domain (TD), lies at the distal end of each triskelion leg. The remaining '140 kDa of the heavy chain forms the triskelion core, consisting of the vertex, proximal leg, and a portion of the distal leg of the clathrin trimer. One clathrin light chain is associated with each proximal leg. Based on electron microscopic studies and models of the polygonal lattice, a clathrin hub lies at each vertex, and each lattice edge is formed by the superposition of two distal and two proximal clathrin heavy chain legs (16 –18). In the coated pit, the TDs are believed to be located beneath 1 The abbreviations used are: GPR(s), G protein-coupled receptor(s); TD, terminal domain; MES, 4-morpholineethanesulfonic acid; GST, glutathione S-transferase; PAGE, polyacrylamide gel electrophoresis. 2 F. Santini, R. B. Penn, J. L. Benovic, and J. H. Keen, manuscript in preparation.

15017

15018

Arrestin-Clathrin Terminal Domain Interactions

each vertex of the surface lattice and to project inward toward the plasma membrane (16). It has been shown that the TD contains a binding site for plasma membrane-localized AP-2 and that this site is required for AP-2-mediated coat assembly (19). Indeed, cryoelectron microscopy analysis suggests that AP-2 lies between the membrane and the shell of clathrin TDs (16), suggesting that this interaction may be important in the AP-2 adaptor function as well as AP-2 assembly activity. Here we show that the TD also contains a distinct arrestin binding site. By deletion and alanine replacement mutagenesis of recombinant TDs, we identify clathrin heavy chain residues 89 –100 as critical for arrestin binding, but not required for AP-2 binding. Site-directed mutagenesis experiments indicate that several highly conserved and invariant residues contribute to this binding. Finally, we show that arrestins are not sufficient to support clathrin assembly, consistent with previous assembly models. These observations suggest that a general function of the clathrin TD is to interact with multiple adaptors, targeting their cognate surface receptors to coated pits. EXPERIMENTAL PROCEDURES

Materials—Glutathione-agarose was purchased from Sigma. Clathrin and AP-2 were prepared from bovine calf brains as described previously (20). L-1-Tosylamido-2-phenylethyl chloromethyl ketone-trypsin was from Worthington Biochemical, Inc., and HEPES and isopropyl-1thio-b-D-galactopyranoside were from Boehringer Mannheim. All other chemicals were reagent grade or better. Buffers utilized are as follows. Buffer A is 0.1 M Na-MES, pH 6.8, 0.1 mM dithiothreitol, 0.1 mg/ml soybean trypsin inhibitor. Buffer B is 20 mM K-HEPES, pH 7.3, 120 mM potassium acetate, 0.1 mM dithiothreitol, 0.1 mg/ml soybean trypsin inhibitor, 1 mg/ml each leupeptin, pepstatin, and antipain and 0.1% Triton X-100. Buffer C is 0.1 M Na-MES, pH 6.8, 0.1 mM phenylmethylsulfonyl fluoride. Construction of Bacterial Expression Vectors—Glutathione S-transferase (GST)-TD fusion proteins were constructed from polymerase chain reaction fragments derived from a bovine brain cDNA library. Amino acid residues 1–579 were amplified and cloned into the BamHI and SmaI sites of pGEX-2T (GST-TD, Pharmacia). GST-TD(1–389) was generated by subcloning the BamHI-EcoRI fragment of pGST-TD into pGEX-2T. TD deletion and site mutants were subsequently obtained from pGST-TD template by polymerase chain reaction-based mutagenesis (primer sequences available upon request). All constructs were confirmed by DNA sequencing (Nucleic Acids Facility, Kimmel Cancer Institute). Expression and Purification of Recombinant Proteins in Escherichia coli—Proteins were expressed in E. coli BL21 cells. A 5.0-ml overnight culture grown in LB medium supplemented with 100 mg/ml ampicillin was diluted into 250 ml of medium and grown 2 h at 37 °C (A600 ; 0.6). Cultures were then induced with 75 mM isopropyl-1-thio-b-D-galactopyranoside for 2 h at 30 °C. Bacterial pellets were resuspended in 5 ml of phosphate-buffered saline, 1 mM phenylmethylsulfonyl fluoride, 2 mg/ml lysozyme, incubated at room temperature for 15 min, and then supplemented with Triton X-100 (1% v/v). Lysates were obtained by two rapid freeze-thaw cycles, treated with 300 units of DNase for 15 min on ice, and centrifuged at 50,000 rpm for 10 min. Supernatants were supplemented with 2 mM dithiothreitol and gently agitated in the presence of 200 ml of glutathione-agarose beads overnight at 4 °C. Beads were washed three times with 10 ml of phosphate-buffered saline, 1% Triton X-100, and three times with 1.4 ml of phosphatebuffered saline in the absence of detergent. Protein content of the beads was determined by densitometric analysis of Coomassie-stained SDSPAGE gels (Molecular Dynamics). Yields were typically 1.0 mg/250 ml of original culture volume. Recombinant arrestins were expressed and purified as described previously (10). Cage Binding Assays—Preparation of intact and clipped clathrin cages reconstituted with light chains has been described elsewhere (21). Arrestin binding to cages was performed as described previously (10). Briefly, arrestins (300 nM) were incubated in the presence of clathrin cage preparations (100 nM trimers) in buffer A for 10 min at room temperature. Complexes were centrifuged through a 75-ml 0.2 M sucrose cushion prepared in buffer A in a TLA100 rotor (Beckman) at 75,000 rpm for 5 min at 4 °C and analyzed by 7.5% SDS-PAGE. Coomassiestained arrestin bands were quantified by densitometry.

FIG. 1. Binding of arrestins to intact but not truncated clathrin cages. Arrestins (300 nM) were incubated with 150 nM intact clathrin cages (lanes 2 and 6), trypsinized clathrin cages (lanes 3 and 7), trypsinized clathrin cages reconstituted with light chains (lanes 4 and 8), or alone (lanes 1 and 5) in buffer A (final volume of 100 ml) for 10 min at 22° C. Cage pellets were analyzed by SDS-PAGE. The data are representative of four independent experiments. HC, heavy chain; LC, light chain. GST-TD Binding Assays—Fusion proteins ('5 mg in a 5-ml bed volume of glutathione-agarose) and arrestins ('0.4 mg) were incubated in 100 ml of buffer B for 1 h at 4 °C on a rotator. Deletion mutant fusion protein additions were adjusted as necessary to attain equimolar inputs. The beads were centrifuged and the unbound fraction aspirated. The beads were washed subsequently with 500 ml of buffer B and eluted either with buffer B: 2 M Tris-HCl, pH 7 (1:1, v/v), for 10 min or with SDS-PAGE sample buffer. Binding was quantified by immunoblotting 5% of each fraction with the monoclonal anti-arrestin antibody F4C1 (L. Donoso, Thomas Jefferson University) and comparison with a 10-fold range of arrestin standards. Coat Assembly Studies—Clathrin assembly by AP-2 has been detailed elsewhere (20). Briefly, 100 mg of clathrin was dialyzed overnight against buffer C in the presence of '40 mg AP-2 and/or 8 mg of arrestin. Retentates were microcentrifuged at 10,000 3 g for 2 min (generating a low speed pellet fraction) and then at 75,000 rpm for 5 min in a TLA100 rotor (generating high speed pellet and supernatant fractions). Five percent of each fraction was analyzed by 7.5% SDS-PAGE, and the clathrin heavy chains were quantified by densitometry of Coomassie Blue staining. RESULTS

Localization of the Arrestin Binding Site to the Clathrin TD—To determine which region of the clathrin triskelion contains a binding site for nonvisual arrestins, we subjected clathrin cages to limited tryptic digestion. This treatment hydrolyzes clathrin light chains and cleaves the 192-kDa heavy chains, resulting in release of the globular '50-kDa TD (15). The remainder of the heavy chains, comprising the hub, proximal leg, and a portion of the distal leg, remains assembled and readily sedimentable (20), and we refer to these structures as truncated cages. As shown in Fig. 1, binding of b-arrestin and arrestin3 to truncated cages (lanes 3 and 7) was greatly reduced relative to that observed for intact cages (lanes 2 and 6). Reconstitution of truncated cages with intact clathrin light chains (23) did not restore arrestin binding (lanes 4 and 8). Although we cannot rule out the possibility that other portions of the clathrin heavy chain may also contribute to arrestin binding, these data indicate that the clathrin TD, corresponding approximately to residues 1– 479 (24), contains an arrestin binding site. To assess directly the interaction between TD and arrestin, we expressed and affinity purified recombinant bovine clathrin TD as a GST-fusion protein (GST-TD). Binding assays were carried out by incubating arrestins with GST-TD beads or, as a negative control, GST beads. After washing, bound arrestins were eluted completely with either boiling SDS-PAGE sample buffer or high concentrations of Tris-HCl ($0.5 M). In previous studies (10), we demonstrated that although visual arrestin fails to interact appreciably with clathrin, the nonvisual arrestins, b-arrestin and arrestin3, bind to clathrin cages with high affinity (Kd ' 10 – 60 nM). This arrestin selectivity is recapitulated in the GST-TD binding assay (Fig. 2). As expected, none of the arrestins bound to GST beads (lanes 1, 4, and 7). Although visual arrestin also failed to interact appreciably with GST-TD (lane 8), both arrestin3 and b-arrestin


15019

FIG. 2. Binding of arrestin3 and b-arrestin, but not visual arrestin, to full-length GST-TD. Arrestins (75 nM) were incubated with glutathione-agarose beads containing GST alone (lanes 1, 4, and 7) or GST-TD (clathrin heavy chain residues 1–579) (lanes 2, 5, and 8) for 1 h at 4° C in buffer B (GST:arrestin ;8 mol/mol). Beads were centrifuged, washed twice with buffer B, and bound proteins eluted with SDS sample buffer. Eluant fractions (5% of each) were run on 7.5% SDSPAGE, transferred to nitrocellulose membrane, and analyzed by immunoblotting with the F4C1 antibody. Lanes 3, 6, and 9 denote 5% of the total input protein (in equal proportion to that analyzed). The data are representative of four independent experiments.

FIG. 3. Binding of arrestin3 and b-arrestin to truncated GST-TD constructs. Arrestins (75 nM) were incubated with glutathione-agarose beads containing GST fused to the clathrin terminal domain residues indicated (GST:arrestin ;8 mol/mol). SDS eluants were analyzed by immunoblotting and the data normalized to the arrestin binding to clathrin heavy chain residues 1–579. The data are representative of four independent experiments.

bound GST-TD beads (lanes 2 and 5, respectively). In general, arrestin3 bound more extensively (60 –90% of input) than b-arrestin (30 –50% of input). Localization of the Arrestin Binding Site within the Clathrin TD—To localize more precisely the arrestin binding site within the clathrin TD, we initially employed an in vitro transcription/ translation system and a GST-fusion protein containing the COOH-terminal half of arrestin3, the region of nonvisual arrestins known to be involved in clathrin binding (10, 25). In vitro translated clathrin heavy chain residues 1–389 effectively bound to this fusion protein (data not shown), localizing the binding region to the NH2-terminal moiety of the clathrin TD. Unfortunately, TD constructs containing greater COOH-terminal deletions were not expressed stably in the in vitro translation system. Accordingly, we next generated a series of COOH-terminaldeleted GST-TD constructs and assessed their ability to bind arrestins (Fig. 3). Interestingly, the first 100 residues of TD supported arrestin binding to an extent comparable to that observed for the full-length TD (1–579). Deletion of an additional fifteen residues (1– 85) resulted in the complete abrogation of binding of both arrestin3 and b-arrestin, suggesting that a crucial arrestin binding determinant is contained within residues 86 –100 of the clathrin heavy chain (Fig. 3). To identify residues in this region involved in arrestin binding, we initially performed triplet alanine replacement mu-

FIG. 4. Binding of arrestin3 and b-arrestin to alanine triplet mutants in clathrin terminal domain residues 86 –100. Arrestin3 (panel A) and b-arrestin (panel B), each at 75 nM concentration, were incubated with glutathione-agarose beads containing GST-TD mutants (GST:arrestin ;8 mol/mol). SDS eluants were analyzed by immunoblotting and the data normalized to the arrestin binding of 1–579. The data are the mean of three independent experiments; wt, wild-type. Error bars denote 6S.E.

tagenesis of residues 86 –100, expressed in the 1–579 context (Fig. 4). For both arrestins, the most potent mutant was the KMK triplet (K98A, M99A, K100A; Table I). With the exception of the KTL construct, all alanine triplet replacements exhibited reduced arrestin binding, with the effects on arrestin3 (Fig. 4A) generally being somewhat greater than on b-arrestin (Fig. 4B). These findings focused attention on residues 89 –100 as contributing to arrestin binding. Site-directed Mutagenesis—To assess the contribution of individual clathrin residues to arrestin binding, we produced fusion proteins containing the complete clathrin TD (residues 1–579) with site-specific mutations in the 89 –100 sequence. This region contains a number of invariant and highly conserved residues, and these were the targets for alteration (Table I). Charged residues were systematically inverted, whereas Ile-90 and Phe-91 were each converted individually to alanine to assess the contribution of hydrophobic interactions. The invariant glutamine residue at position 89, which may participate in hydrogen bonding, was converted to either methionine (approximately isosteric, non-hydrogen bonder) or alanine. In general, both arrestin3 and b-arrestin exhibited the same trends in binding to the TD site-specific mutants, indicating that both arrestins interact similarly with residues in this

15020


TABLE I Conservation of clathrin heavy chain residues 86 –100 among multiple species Clathrin heavy chain residues 86 –100 from the indicated sources are aligned and tabulated. Shaded residues denote identities in six of eight species, the 1 symbol is present above residues which are identical in seven of the eight heavy chains, and invariant residues are boldface.

region. However, the inhibition of arrestin3 binding was, in almost all cases, more complete than that for b-arrestin (Fig. 5, A and B). Of the charge inversion mutants, K98E and K96E were most deficient with respect to arrestin3 and b-arrestin binding. The K100E mutant exhibited only a modest loss of binding, whereas E94K showed no loss whatsoever. Of the hydrophobic residue mutants, F91A but not I90A was significantly defective in binding to both arrestins, suggesting that an aromatic moiety may be important. Both alterations of the invariant glutamine residue, Q89A and Q89M, were also significantly defective. To show that the loss of arrestin binding was not due to a global disruption of TD structure, we subjected the wild-type and mutant GST-TD fusion proteins most defective in arrestin binding to limited tryptic proteolysis. In all cases, a major '50-kDa band was obtained whose electrophoretic mobility was identical to TD derived from bovine brain preparations (data not shown), confirming that the TD was a folded structure much more resistant to protease than the linker peptide connecting it to the GST moiety. Furthermore, we tested binding of all site mutants to AP-2, which has been shown previously to bind clathrin TD (19). In contrast to the arrestin binding results, all mutants retained substantial AP-2 binding. Although two of the mutants (F91A and K98E) exhibited some reduction, this decrease was significantly less than that observed for arrestin binding. Can Arrestins Induce Clathrin Assembly?—Arrestins bind clathrin with affinity similar to AP-2, a protein originally identified based on its ability to promote clathrin lattice formation under physiological solution conditions (22, 26). Like the arrestins, AP-2 binds to clathrin TDs (Ref. 19 and Fig. 5C), but AP-2 also binds stably to the truncated light chain-free hubs of clathrin generated either proteolytically (21) or by recombinant methods (27). It was therefore important to determine whether arrestins can support clathrin coat assembly. Whereas an equimolar input of AP-2 promoted clathrin assembly, arrestins did not (Fig. 6): even a 10-fold molar excess of arrestin to clathrin did not yield assembled lattices (data not shown). Furthermore, although both proteins bind to clathrin TDs, the presence of arrestins in the assembly reaction did not alter the ability of AP-2 to stimulate coat formation. DISCUSSION

Characteristics of the Arrestin Binding Site—Fusion proteins containing progressive COOH-terminal truncations within the TD along with alanine scanning mutagenesis demonstrated that clathrin heavy chain residues 89 –100 within the TD are an important determinant of arrestin binding. Examination of clathrin heavy chain sequences (Table I) indicates that these residues are highly conserved among diverse species, suggesting that this region is involved in a critical function. GST-TD fusion proteins containing either Q89M or K96E

FIG. 5. Binding of arrestin3, b-arrestin, and AP-2 to GST-TD site-directed mutants. Arrestin3 (panel A), b-arrestin (panel B), and AP-2 (panel C), each at 75 nM concentration, were incubated with glutathione-agarose beads containing the indicated GST-TD mutants (GST:soluble protein ;8 mol/mol). SDS eluants were analyzed by immunoblotting and the data normalized to binding to the wild-type (WT) 1–579 construct. The data are the mean of three independent experiments; error bars denote 6S.E.

mutations had almost undetectable arrestin3 binding and greatly reduced b-arrestin binding. Interestingly, these are invariant residues in all known clathrin heavy chains (Table I), suggesting that arrestins or arrestin-like proteins are ubiqui-


FIG. 6. Arrestin does not promote clathrin lattice assembly. Clathrin (500 nM) was dialyzed overnight against buffer C in the presence of an equimolar quantity of AP-2 and/or arrestins as indicated. Equal proportions of the low speed pellet, high speed pellet, and high speed supernatant were analyzed and the clathrin heavy chains quantified by densitometry of Coomassie Blue-stained SDS-PAGE gels (for details, see “Experimental Procedures”). Values are expressed as a fraction of total input clathrin. The data are representative of two independent experiments.

tous and play important roles in cellular function. The Gln-89 residue is likely to be exposed on the surface of the TD and participate in hydrogen bonding as replacement by isosteric non-hydrogen-bonding methionine more extensively diminished arrestin binding than did replacement by the smaller alanine residue. The F91A and K98E mutants had greatly diminished arrestin binding, but both also exhibited some loss of AP-2 binding. This may reflect some alteration in global structure not revealed by the protease sensitivity assay. As the charge of Glu-94 can be inverted without noticeably affecting arrestin binding, this residue likely lies on the surface of the TD but is not involved directly in arrestin binding. However, its conservation throughout all species for which sequence data are available suggests that it is required for other important interactions. TD as an Arrestin and Adaptor Binding Domain—The globular '50-kDa region at the end of each extended clathrin leg has been visualized as a distinct globular region and has accordingly been designated the terminal domain. There is increasing evidence for specific structural and functional attributes within this region. Electron microscopy studies suggest that the morphology of the TD has a variable shape capable of being extended to a greater or lesser extent in a scroll-like manner (18). Based on proteolytic studies, the TD has been considered to be composed of residues 1– 479, but sequences further into the linker region, at least 480 – 497, appear to be associated with the NH2-terminal region as they remain physically bound following cleavage (24). Finally, our data indicate that residues 1–100 comprise a distinct structural domain that exhibits functional arrestin binding essentially indistinguishable from the intact TD. Based on its spatial orientation within the coated pit, the clathrin TD is an apt binding site for arrestin and other candidate adaptor molecules. In the assembled clathrin coat structure, the hubs and distal legs form the cytoplasmic surface of the lattice, a 10-nm thick structure that lies about 15 nm from the membrane surface (16). The TD is believed to be joined to

15021

the distal leg by a short, flexible linker region (15, 18). Cryoelectron microscopic images suggest that the TDs project inward at varied angles, forming an inner shell concentric with the surface lattice (16). Therefore, TDs would be in an appropriate position to contact the exposed tails of receptors and the cytoplasmic proteins that bind to them, such as arrestins. Arrestins as Adaptors for GPR-mediated Endocytosis—The nonvisual arrestins, by virtue of their demonstrated ability to bind tightly to both clathrin and agonist-activated receptor, and their involvement in receptor sequestration at the plasma membrane, constitute a new class of adaptors distinct from the heterotetrameric APs (10, 28). Although both bind clathrin with similar affinities, arrestins and the plasma membraneassociated AP-2 differ in several important respects. Arrestins bind tightly only to the TD region of clathrin. In contrast, AP-2 binds stably to both the TD and the clathrin triskelion core, with each of its large subunits independently recognizing clathrin (29, 30). Arrestin binding to clathrin does not support coat assembly (Fig. 6); interestingly, binding of AP-2 to recombinant clathrin hubs also does not induce lattice formation, although the hubs independently can undergo a polymerization reaction to generate a lattice (27). We note that these observations support the proposal that the AP-2-mediated coat assembly reaction requires bivalent interaction of AP-2 with two clathrin triskelia (20). In intact cells, AP-2 is a structural component of clathrincoated pits and vesicles: it is colocalized with clathrin and can be recovered as a stoichiometric component of coated vesicles isolated from tissues (14). In contrast, arrestins seem to be recruited to coated pits only in response to agonist activation of GPRs (10) and are detected only at trace levels ('1 arrestin:50 triskelia) in purified bovine brain coated vesicles (data not shown). These observations are consistent with the model that preformed coated pits bind receptors, or in this case arrestinreceptor complexes, rather than being formed de novo in a receptor-induced assembly reaction (31). In contrast, the properties of AP-2 indicate that it participates both structurally and functionally in the lattice assembly reaction. Finally, as potential adaptors, the nonvisual arrestins and AP-2 have remarkably different receptor binding affinities despite possessing similar affinities for clathrin. Whereas arrestins bind cognate GPRs with low or subnanomolar affinity (32), corresponding affinities of APs for cytoplasmic receptor tails appear several orders of magnitude weaker (33). Perhaps the arrestins are sufficient as GPR adaptors, whereas AP-2 works in concert with other potential receptor-binding proteins, such as shc (34) and eps15 (11). Acknowledgments—We thank C. Carman and Drs. Z. Huang, A. Gagnon, and L. Kallal for helpful discussions. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

Schmid, S. L. (1992) BioEssays 14, 589 –596 Robinson, M. S. (1994) J. Cell Biol. 108, 833– 842 Sterne-Marr, R., and Benovic, J. L. (1995) Vitam. Horm. 51, 193–235 Kuhn, H., Hall, S. W., and Wilden, U. (1984) FEBS Lett. 176, 473– 478 von Zastrow, M., and Kobilka, B. K. (1994) J. Biol. Chem. 269, 18448 –18452 Yu, S. S., Lefkowitz, R. J., and Hausdorff, W. P. (1993) J. Biol. Chem. 268, 337–341 Barak, L. S., Tiberi, M., Freedman, N. J., Kwatra, M. M., Lefkowitz, R. J., and Caron, M. G. (1994) J. Biol. Chem. 269, 2790 –2795 Tsuga, H., Kameyama, K., Haga, T., Kurose, H., and Nagao, T. (1994) J. Biol. Chem. 269, 32522–32527 Ferguson, S. S. G., Menard, L., Barak, L. S., Koch, W. J., Colapietro, A.-M., and Caron, M. G. (1995) J. Biol. Chem. 270, 24782–24789 Goodman, O. B., Jr., Krupnick, J. G., Santini, F., Gurevich, V. V., Penn, R. B., Gagnon, A. W., Keen, J. H., and Benovic, J. L. (1996) Nature 383, 447– 450 Benmerah, A., Gagnon, J., Begue, B., Megarbane, B. Dautry-Varsat, A., and Cerf-Bensussan, N. (1995) J. Cell Biol. 131, 1831–1838 Schroder, S., and Ungewickell, E. (1991) J. Biol. Chem. 266, 7910 –7918 Peeler, J. S., Donzell, W. C., and Anderson, R. G. W. (1993) J. Cell Biol. 120, 47–54 Keen, J. H. (1991) Annu. Rev. Biochem. 59, 415– 438 Kirchhausen, T., and Harrison, S. C. (1981) Cell 23, 755–761

15022


16. Vigers, G. P. A., Crowther, R. A., and Pearse, B. M. F. (1986) EMBO J. 5, 529 –534 17. Crowther, R. A., and Pearse, B. M. F. (1981) J. Cell Biol. 91, 790 –797 18. Heuser, J., and Kirchhausen, T. (1985) J. Ultrastruct. Mol. Struct. Res. 92, 1–27 19. Murphy, J.-E., and Keen, J. H. (1992) J. Biol. Chem. 267, 10850 –10855 20. Keen, J. H. (1987) J. Cell Biol. 105, 1989 –1998 21. Keen, J. H., Beck, K. A., Kirchhausen, T., and Jarrett, T. (1991) J. Biol. Chem. 266, 7950 –7956 22. Keen, J. H., Willingham, M. C., and Pastan, I. H. (1979) Cell 16, 303–312 23. Ungewickell, E., Unanue, E. R., and Branton, D. (1982) Cold Spring Harbor Symp. Quant. Biol. 46, 723–731 24. Kirchhausen, T., Harrison, S. C., Chow, E. P., Mattaliano, R. J., Ramachandran, K. L., Smart, J., and Brosius, J. (1987) Proc. Natl. Acad. Sci. U. S. A. 84, 8805– 8809 25. Krupnick, J. G., Goodman, O. B., Jr., Keen, J. H., and Benovic, J. L. (1997)

J. Biol. Chem. 272, 15011–15016 26. Zaremba, S., and Keen, J. H. (1983) J. Cell Biol. 97, 1339 –1347 27. Liu, S., Wong, M. L., Craik, C. S., and Brodsky, F. M. (1995) Cell 83, 257–267 28. Ferguson, S. S. G., Downey, W. E., III, Colapietro, A. M., Barak, L. S., Menard, L., and Caron, M. G. (1996) Science 271, 363–365 29. Goodman, O. B., Jr., and Keen, J. H. (1995) J. Biol. Chem. 270, 23768 –23773 30. Gallusser, A., and Kirchhausen, T. (1993) EMBO J. 12, 5237–5244 31. Santini, F., and Keen, J. H. (1996) J. Cell Biol. 132, 1025–1036 32. Gurevich, V. V., Dion, S. B., Onorato, J. J., Ptasienski, J., Kim, C. M., SternMarr, R., Hosey, M. M., and Benovic, J. L. (1995) J. Biol. Chem. 270, 720 –731 33. Pearse, B. M. F. (1988) EMBO J. 7, 3331–3336 34. Okabayashi, Y., Sugimoto, Y., Totty, N. F., Hsuan, J., Kido, Y., Sakaguchi, K., Gout, I., Waterfield, M. D., and Kasuga, M. (1996) J. Biol. Chem. 271, 5265–5269