The alpha chain of the AP-2 adaptor is a clathrin binding subunit.

We have utilized a rabbit reticulocyte lysate coupled transcription-translation system to express the large subunits of the clathrin associated protein-2 (AP-2) complex so that their individual functions may be studied separately. Appropriate folding of each subunit into N-terminal core and C-terminal appendage domains was confirmed by limited proteolysis. Translated β2 subunit bound to both assembled clathrin cages and immobilized clathrin trimers, confirming and extending earlier studies with preparations obtained by chemical denaturation-renaturation. Translated αa exhibited rapid, reversible and specific binding to clathrin cages. As with native AP-2, proteolysis of αa bound to clathrin cages released the appendages, while cores were retained. Further digestion revealed a ≈29-kDa αa clathrin-binding fragment that remained tightly cage-associated. Translated αa also bound to immobilized clathrin trimers, although with greater sensitivity to increasing pH than the translated β2 subunit. Clathrin binding by both the α and β subunits is consistent with a bivalent cross-linking model for lattice assembly (Keen, J. H.(1987) Cell Biol. 105, 1989). It also [Abstract] raises the possibility that the α-clathrin interaction may have other consequences, such as modulation of lattice stability or shape, or other α functions.

We have utilized a rabbit reticulocyte lysate coupled transcription-translation system to express the large subunits of the clathrin associated protein-2 (AP-2) complex so that their individual functions may be studied separately. Appropriate folding of each subunit into Nterminal core and C-terminal appendage domains was confirmed by limited proteolysis. Translated ␤2 subunit bound to both assembled clathrin cages and immobilized clathrin trimers, confirming and extending earlier studies with preparations obtained by chemical denaturation-renaturation. Translated ␣ a exhibited rapid, reversible and specific binding to clathrin cages. As with native AP-2, proteolysis of ␣ a bound to clathrin cages released the appendages, while cores were retained. Further digestion revealed a Ϸ29-kDa ␣ a clathrin-binding fragment that remained tightly cage-associated. Translated ␣ a also bound to immobilized clathrin trimers, although with greater sensitivity to increasing pH than the translated ␤2 subunit. Clathrin binding by both the ␣ and ␤ subunits is consistent with a bivalent cross-linking model for lattice assembly (Keen, J. H. (1987) Cell Biol. 105,1989). It also raises the possibility that the ␣-clathrin interaction may have other consequences, such as modulation of lattice stability or shape, or other ␣ functions.
Receptor-mediated endocytosis is a multi-step process involving membrane invagination, coated pit formation, and budding of these pits to form coated vesicles (2). A major protein implicated in endocytosis is clathrin, a triskelion-shaped protein that forms the structural basis for the regular polygonal lattice of coated pits and vesicles (1,3). These coated membranes also contain additional protein components that have been referred to as assembly, adaptor, or associated proteins (APs). 1 One probable function of APs is to promote polymerization of the clathrin lattice at defined sites and times. APs are also likely to interact with receptor cytoplasmic tails resulting in the selective inclusion of various receptors into coated pits (reviewed in Refs. 2,4,5).
APs vary in structure and intracellular localization. The best characterized examples include AP-1, a Golgi-associated heterotetramer consisting of ␥, ␤1, AP47, and AP19 polypeptides; AP-2, a plasma membrane-associated heterotetramer of ␣, ␤2, AP50, and AP17 polypeptides; and AP-3/AP180, a neuronspecific monomer (2,4). This study concerns the AP-2 complex and focuses on the interactions of its ␣ subunit with clathrin. Two genetically distinct isoforms of ␣ subunit exist: ␣ c , an isoform which is expressed ubiquitously, and ␣ a , an isoform believed to be expressed primarily in neurons. The isoforms are 84% identical and differ predominantly in their C-terminal portions. The ␣ a isoform contains a unique 42 amino acid insert beginning at position 704 (6).
Although AP-2-clathrin interactions have been studied in detail (1,(7)(8)(9)(10)(11)(12)(13), it has been difficult to ascertain the contributions of individual AP-2 subunits. Fractionation of AP-2 polypeptides with urea and guanidinium chloride was used to study these interactions, indicating that the ␣ and ␤2 subunits alone were necessary and sufficient for coat assembly activity (13). Ahle and Ungewickell (7), using mild denaturation to purify ␤2 subunit from AP-2, demonstrated that the former was capable of competitively inhibiting AP-2 binding to preassembled clathrin cages. This work was extended by Gallusser and Kirchhausen (14) who demonstrated that recombinant ␤2 subunit purified by denaturation-renaturation from Escherichia coli inclusion bodies was capable of promoting clathrin assembly. Collectively, these results support the hypothesis that the ␤2 subunit plays an important role in AP-2-driven clathrin assembly in vivo, but the role of the ␣ subunit remains undefined.
We have previously reported that an ␣/AP50/AP17 complex prepared by mild denaturation-renaturation was capable of binding to preformed clathrin cages, suggesting that one or more of the other subunits, most likely the ␣ subunit, also recognizes and binds clathrin (13). We adopt the approach of in vitro translation of the individual large AP-2 subunits to further explore this issue. The findings reported here indeed demonstrate that the ␣ subunit can bind tightly to clathrin, consistent with a role in coat assembly or other coat-associated functions.

MATERIALS AND METHODS
The T N T rabbit reticulocyte lysate transcription-translation kit and pSP65 cloning vector were purchased from Promega. Translabel was obtained from ICN Biomedicals, Inc. Sepharose CL-4B and Superose 6B resins were from Sigma, and CN-Br activated Sepharose CL-4B was purchased from Pharmacia Biotech Inc. Clathrin and assembly proteins were prepared from calf brains as described previously (1,15). L-1-tosylamido-2-phenylethyl chloromethyl ketone-trypsin was from Worthington Biochemical, Inc. HEPES was purchased from Boehringer Mannheim. All other chemicals were reagent grade or better.
The cDNA template for the C-terminal deletion mutant ␣ a(1-605) was produced by restricting the pSP65␣ a sense construct with AvaI (␣ a nucleotide 2021). The mutant was then generated by runoff transcription-translation. Full-length ␤2 subunit were transcribed from the pBluescript SKϩ T3 promotor, using a construct kindly provided by T. Kirchhausen (Harvard University) (16). Luciferase was translated with the sp6 T N T control construct (Promega). ␤-Galactosidase was produced using a pSP65-based construct provided by V. Gurevich (Thomas Jefferson University). Transcription-translation reactions were assembled according to the manufacturer's instructions and incubated 90 -120 min. Proteins were translated in the presence of [ 35 S]Translabel: although the reagent contained both [ 35 S]L-methionine and -cysteine, a large excess of unlabeled cysteine in the translation kit blocked the latter's incorporation. Prior to use, translation reactions were centrifuged at 100,000 revolutions/min for 20 min at 2°C in a Beckman TLA 100 rotor to clear ribosomes and aggregates. All experiments utilized freshly translated subunits.
Proteolysis of Translated ␣ a and ␤2 Subunits-A fresh 1.0 mg/ml trypsin stock solution was prepared in 1 mM HCl. An aliquot of the stock solution was diluted at least 10-fold into 50 mM Tris-HCl, pH 7.4, immediately before use. A 5-l translation reaction aliquot and an appropriate amount of diluted trypsin stock were combined with 50 mM Tris-HCl, pH 7.4, in a final volume of 100 l, according to the protocol of Matsui and Kirchhausen (12). Reactions were incubated 15 min at room temperature and quenched with either SDS sample buffer or a 3-fold molar excess of soybean trypsin inhibitor.
Digests of ␣ a bound to cages were carried out in Buffer A. Cage high speed pellets (see below) with bound ␣ a were resuspended in Buffer A and an aliquot of this suspension combined with trypsin stock freshly diluted in buffer A. Reactions were incubated 15 min at room temperature and quenched with a 3-fold molar excess of soybean trypsin inhibitor.
Cage Binding Assays-Preparation of clathrin cages and binding of APs to preformed clathrin cages has been described elsewhere (10,11). All steps were performed at 4°C except where noted. In this work, cage aliquots were supplemented with 100 mM Tris-HCl and spun 2 min at 13,000 ϫ g just prior to use. Clathrin concentrations were determined spectrophotometrically in 0.5 M Tris-HCl using the calculated extinction coefficient of 1.0 mg ml Ϫ1 cm Ϫ1 . Typically, 20 -60 g of cages and 5 l of translation reaction were incubated for 30 min in Buffer A supplemented with 100 mM Tris-HCl, pH 6.5, in a final volume of 300 l. Samples were centrifuged at 75,000 revolutions/min for 7 min in a TLA 100.2 rotor. Pellets were resuspended in 60 l of sample buffer, and supernatants were precipitated with 10% (w/v) trichloroacetic acid and similarly resuspended. Equal aliquots of pellets and supernatant were analyzed on 7.5% SDS-polyacrylamide minigels.
Clathrin-Sepharose Binding Assays-Preparation of clathrin-Sepharose and its binding by APs have been described previously (1). We modified the protocol to accommodate a small scale binding assay. Into Bio-Rad BioSpin Chromatography columns, 300 l of resin was aliquoted. Columns were equilibrated with several column volumes of Buffer A containing 0.05 mg/ml bovine serum albumin, leupeptin, antipain, and pepstatin (10 g/l). A 3-l aliquot of translation reaction was mixed with 97 l of the equilibration buffer, loaded, and incubated on the column for 30 min, rinsed four times with one column volume each of equilibration solution, and eluted with four column volumes of Buffer B supplemented with 0.05 mg/ml bovine serum albumin. Unbound and eluted fractions were pooled separately, precipitated in 10% trichloracetic acid, and analyzed by SDS-polyacrylamide gel electrophoresis (PAGE).
SDS-PAGE Gel Band Analysis-SDS-PAGE gels were Coomassie stained, destained, and treated with EN 3 HANCE (DuPont NEN), dried, and subjected to fluorography for 0.5-16 h on Kodak X-0MAT film. Radioactive bands were excised from the dried gels, dissolved in H 2 O 2 , and counted in a Wallack LKB scintillation counter.

RESULTS
In Vitro Translated Polypeptides-Two different ␣ a polypeptides were synthesized in vitro ( Fig. 1): the full-length polypeptide consisting of 977 amino acids (lane 1) and a C-terminal truncation mutant (␣ a(1-605) ) possessing only the N-terminal 605 amino acids (lane 2). The full-length ␣ a polypeptide has a calculated molecular mass of 107,655 Da but migrated on SDS-PAGE as a band of Ϸ114,000 M r , coincident with the Coomassie-stained ␣ a band of bovine brain AP-2. The yield of the deletion mutant (M r ϭ 67,500) was substantially lower than that of the full-length product, as approximately a 10-fold longer autoradiographic exposure was required for comparable detectability on film. As a negative control, we translated ␣ a in the presence of the antisense ␣ a construct, obtaining a greatly attenuated yield as expected (lane 3). The expressed ␤2 subunit (lane 4, M r ϭ 105,133) was readily resolved from translated ␣ a by gel electrophoresis. Like translated ␣ a , the ␤2 polypeptide comigrated with a corresponding Coomassie-stained band of AP-2. Finally, in some experiments, translated luciferase (lane 5, M r ϭ 62,000) and ␤-galactosidase (lane 6, M r ϭ 116,000) were utilized as controls.
Proteolysis of in Vitro Translated ␣ a and ␤2 Polypeptides-Limited proteolysis of AP-2 reveals two major stable protein domains: one containing 60 -66-kDa N-terminal core domains of the ␣ and ␤2 subunits associated with intact AP50 and AP17 intact subunits, and the other consisting of 30 -40-kDa domains corresponding to smaller C-terminal appendages of the ␣ and ␤ subunits (10, 12, 17, 18, and data not shown).
Similarly, proteolysis of the in vitro translated ␣ a or ␤2 subunits alone yielded fragments of Ϸ58 -66 kDa and Ϸ40 kDa (Fig. 2). Identical results were obtained when translated proteins were cleaved in the presence of carrier AP-2, as assessed by comigration of Coomassie Blue-stained and radiolabeled bands on SDS-PAGE (data not shown). Quantitative analysis of the changes in the full-length ␣ a product and the appendage domain confirm a precursor-product relationship (Fig. 3). The coincidence of the two curves in Fig. 3 demonstrates that at all trypsin concentrations the fraction of full-length ␣ a cleaved is virtually identical to that of appendage generated. Furthermore, from the published sequence of the ␣ a cDNA (6), the C-terminal 40 kDa portion of ␣ a (i.e. amino acids 610 -977) is predicted to contain five of the 17 methionine residues of ␣ a , or 27% of the total ␣ a radiolabel. In close agreement with this prediction, 24% of the undigested full-length counts were found in the 40-kDa proteolytic product upon complete digestion of full-length ␣ a . Hence, we conclude that the initial cleavage of translated ␣ a by trypsin occurs in a region corresponding to the relatively exposed linker of an ␣ subunit in the intact AP-2 complex.
The translated ␤2 subunit was slightly more resistant to proteolysis initially than the ␣ a subunit, as has been observed by others (17). The initial proteolytic susceptibility of translated ␣ a , resulting in the generation of cores and appendages (Fig. 2, lanes 2 and 3) was comparable to that of bovine brain ␣ a in an AP-2 complex as assessed by digestion of translated ␣ a in the presence of AP-2 (data not shown). However, further digestion of both translated subunits revealed significant differences from those in brain AP-2. The 58 -66-kDa core domains of translated ␣ a , and especially of ␤2, were much more labile than those of AP-2. While virtually no 60 -66-kDa fragments of the translated subunits remained at trypsin concentrations greater than 170 ng/ml (Fig. 2, lanes 4 -6 and 10 -12), AP-2 derived N-terminal products were stable at much higher trypsin concentrations, in excess of 840 ng/ml (21 and data not shown). In contrast, the 40-kDa C-terminal appendages of both the translated ␣ and ␤ polypeptides were similar to AP-2 in their relative stability, resistant even at trypsin concentrations of 9,600 ng/ml (Fig. 2, lanes 6 and 12).
Elsewhere, we have reported that a high affinity inositol polyphosphate-binding site exists near the N terminus of the ␣ subunit. 2 Inclusion of 1 mM phytic acid (1,2,3,4 5,6-IP 6 ) did not alter the proteolytic pattern of ␣ cleavage described above (data not shown).
Binding of Full-length Translated Proteins to Preformed Clathrin Cages-To examine the interaction of translated polypeptides with assembled clathrin, we utilized a cage binding assay. In preliminary experiments, we ascertained that the addition of the translation reaction had no effect on the sedimentability of clathrin cages (data not shown). Furthermore, we determined that both translated luciferase, a standard for the in vitro expression system, and ␤-galactosidase, a polypeptide of similar size to the AP-2 large subunits, showed essentially no affinity for the assembled clathrin cages in the binding assay (Fig. 4, lanes 1-8). As expected from previous work (7), translated ␤2 subunit did bind to clathrin cages (data not shown). As shown in Fig. 4, the majority of translated ␣ polypeptide cosedimented with the clathrin cages (lane 13), while a small amount of translated and apparently aggregated ␣ polypeptide was sedimented by low speed centrifugation (lane 12). Typically, sedimentation of translated ␣ a in the absence of cages, taken as a measure of nonspecific binding, was 10 -15% of the total binding (lane 10).
By incubating the in vitro translation mixture with increasing concentrations of clathrin cages we obtained a dissociation constant of 1.1 ϫ 10 Ϫ7 M (Fig. 5). From the asymptote of the binding curve, the maximal ␣ fraction bound was 0.71, implying that not all of the translated protein was capable of binding clathrin. The binding was not inhibited by 1 mM phytic acid, a potent inhibitor of AP-2 self-association (9), further evidence that binding was specific and not due to aggregation or selfassociation (data not shown).
Clathrin cage binding by translated ␣ a subunit was inhibited by saturating quantities of AP-2, confirming that the interaction was specific (Fig. 6). From the apparent IC 50 and the published K d for the AP-2-clathrin interaction of 10 Ϫ8 M (17, 20), we calculate a K d for the ␣ a -clathrin interaction of 0.7 ϫ 10 Ϫ7 M, in reasonable agreement with the value estimated by direct binding.
Both clathrin-clathrin and AP-clathrin interactions are readily reversed by high concentrations of protonated amines such as Tris-HCl (1, 6, 15). The ␣-clathrin interaction was also reversible. Brief (5 min) treatment of the sedimented cages, to which translated ␣ a was bound, with 500 mM Tris-HCl, pH 7, followed by a high speed spin released most (Ͼ80%) of the ␣ a into the supernatant. Furthermore, the solubilized ␣ a again cosedimented with the clathrin cages reformed by stepwise dialysis of the dissociated preparation into Buffer C and then into Buffer A (data not shown).
Binding of Translated Proteins to Clathrin Trimers Immobilized on Sepharose CL-4B Beads-We used clathrin-Sepharose to assess the ability of ␣ and ␤ subunits to bind clathrin triskelia (Fig. 7). Underivatized Sepharose CL-4B bound little if any of either of the translation products. The translated ␤2 polypeptide bound tightly to clathrin-Sepharose, providing the first direct demonstration of its ability to bind disassembled clathrin as well as cage structures. Binding of the ␣ a polypeptide was also observed, although it was not as complete as that of the ␤2 polypeptide under these conditions. Interestingly, while ␤2 bound with relatively little sensitivity to pH, ␣ a binding to clathrin trimers was considerably more sensitive to increasing pH throughout the range 6.5-7.5.
Functional Domains of Translated Proteins-Previous studies have shown that following controlled proteolysis, only the large core fragments of AP-2 retain an ability to bind to clath-2 I. Gaidarov, Q. Chen and J. H. Keen, manuscript in preparation.
Alternatively, to assess the ability of the isolated N-terminal region of translated ␣ a to bind to preformed clathrin cages, we also produced a truncated polypeptide, designated ␣ a(1-605) , by runoff transcription-translation (Fig. 1). This protein did bind in a saturable manner to clathrin cages, although the apparent binding affinity (K d ϭ 3 ϫ 10 Ϫ7 M) was somewhat lower that of the full-length protein (Fig. 5). Only 36% of the total protein was capable of binding, suggesting either that a greater proportion of the translated protein was misfolded or that a binding equilibrium had not been established.
In contrast to the failure of either core or appendage domain of translated and digested ␣ a to bind, when cages with bound ␣ a were incubated with trypsin the core fragments were preferen-tially retained by the cages while the appendage domain was released into the supernatant (Fig. 8). A similar result has been obtained with native AP-2 (12). Interestingly, on more vigorous proteolysis of the cage-bound translated ␣ a a discrete 29-kDa cage-associated fragment became evident and was prominent only in the presence of clathrin cages (compare Fig. 8, lane 2,   FIG. 6. Binding of translated ␣ a to clathrin cages is blocked by brain AP. Cages (20 g) were incubated with a bovine brain AP preparation containing the indicated concentration of AP-2 in Buffer T for 90 min, followed by incubation with 4 l of translated ␣ a for an additional 30 min. Cage-associated ␣ a was determined after centrifugation. The data were corrected for nonspecific sedimentation and are expressed relative to binding in the absence of exogenous brain AP.  2 and 4) or absence (lanes 1 and  3) of trypsin (100 ng/ml). Following sedimentation, cage-associated and released ␣ a fragments were analyzed by electrophoresis and autoradiography. Lanes 1 and 2, high speed pellets; lanes 3 and 4, supernatants.
FIG . 5. Binding of translated full-length ␣ a and ␣ a(1-605) 1, 6, 7, and 12) which had been pre-equilibrated with buffer A adjusted to the indicated pH (see below). Columns were washed with 1200 l of the equilibration buffer (panel W) and the remaining bound protein eluted with 1200 l of Buffer B supplemented with 0.05 mg/ml bovine serum albumin (panel E). All fractions were precipitated with 10% trichloroacetic acid, and equal proportions of washes and eluants were analyzed by electrophoresis and autoradiography. Lanes 1, 2, 7, and 8, pH 6.5; lanes 3 and 9, pH 6.8; lanes 4 and 10, pH 7.2; lanes 5, 6, 11, and 12, pH 7.6. B, data from panel A for each polypeptide are quantified and plotted as the percentage of maximal binding. Maximal binding for each polypeptide was observed at pH 6.5 and was 80% for ␣ a and essentially 100% for ␤2.
with Fig. 2). This fragment likely corresponds to the clathrinbinding domain of the ␣ subunit. DISCUSSION AP-2 is capable of binding clathrin trimers with high affinity, an interaction representing an initial step in the coat assembly process. Following treatment with urea or guanidinium chloride, the AP-2 complex has been fractionated by gel filtration or hydroxylapatite chromatography, yielding partially purified ␣, ␤, and 50 kDa/17 kDa subunits. Of these, only the large ␣ and ␤2 subunits of AP-2 were required for clathrin coat assembly in vitro (13). Dissociated ␤2 subunits from such preparations were shown to bind to clathrin cages but could not alone sponsor clathrin assembly (7). Recent studies using recombinant protein have reported that ␤2 alone is capable of inducing clathrin assembly (14), and ␤1 has been implicated in clathrin recruitment in the trans-Golgi network (22). Previous work from this laboratory suggested that ␣ subunits could bind to clathrin trimers and cages (23). However, these experiments suffered from potential limitations in that the ␣ fractions contained small quantities of ␤ and 50-kDa/17-kDa polypeptides, preventing an unambiguous assignment of clathrin binding activity to the ␣ subunit.
To further examine the issue of clathrin-binding subunits, we have translated the ␣ a and ␤2 subunits of AP-2 in vitro in a rabbit reticulocyte lysate system to assess their respective clathrin binding capabilities de novo. The approach of in vitro translation has several important advantages. The individual subunits are generated without resort to the strong denaturants that make it extremely difficult to be certain that the native state has been reattained. In contrast, the translation system produces polypeptides in a physiological environment with the appropriate folding factors, more closely resembling the intracellular milieu. Further, readily detectable radioactive polypeptides are generated that can be used at tracer levels (Յ10 Ϫ10 M) in functional assays. This is a major advantage in the study of AP-2 structure and function because the protein and its subunits are prone to aggregation and self-association (8) even at relatively low protein concentrations (Նg/ml or 10 Ϫ8 M). Finally, the study of individual polypeptides of multisubunit proteins by in vitro transcription-translation (24) may be particularly appropriate for the APs. Although these proteins have been assumed to function only as intact tetrameric complexes, there is recent evidence that the AP50 functions independently of the AP-2 complex as an activator of the vacuolar proton pump (25,26). The structural and functional attributes of the isolated ␣ and ␤ polypeptides reported here and previously (7,14) suggest that they, too, could have independent roles.
Our results show that readily detectable amounts of AP-2 ␣ a and ␤2 polypeptides can be expressed in a functional form in vitro. Both appear to assume the proper secondary and tertiary conformation by folding into the core and appendage domains that are characteristic of the intact AP-2 protein. The fragments obtained on limited proteolysis correspond well to those expected from the large subunits of bovine brain AP-2, though there are differences. While the C-terminal appendage fragment is resistant to further proteolysis, the core fragments appear to be more heterogenous and extremely labile. This is consistent with the proposed quaternary structure of isolated AP-2 (10,21). The ␣ and ␤2 C-terminal appendages do not display stable intermolecular contacts with other subunits of AP-2 and likely function as independently folded and stable domains. In contrast, in native AP-2 protein the N-terminal core domains of ␣ and ␤ are in proximity to each other and to the AP50 and AP17 polypeptides. These interactions do not occur with the translated polypeptides. Consequently, the in vitro translated subunits may be relatively unprotected and more prone to proteolysis, yielding the results seen in Fig. 2.
Upon proteolysis of cage-bound ␣ a , the appendage fragments are preferentially released, while the core is almost entirely retained. Of particular interest is the appearance of a novel 29-kDa fragment when cage-bound ␣ a is proteolyzed (Fig. 8). The appearance of this band correlates well with the disappearance of the core 58 -66-kDa fragments, suggesting that further digestion is blocked by tight association and stabilization by clathrin. Conversely, if dissociated this fragment may be more rapidly degraded: in the absence of clathrin, heterogeneous bands of this size are barely detectable (Fig. 2). It seems likely that this fragment comprises part of a discrete clathrin-binding domain within the ␣ subunit.
In contrast to the tight retention of the core and 29-kDa fragment when the full-length protein is proteolyzed, ␣ a(1-605) and proteolytic core fragments generated prior to cage binding interact with much lower affinity. This may indicate that the appendage and/or C-terminal linker regions of ␣ a are required to maintain the free core domain in a conformation in which it is able to interact with clathrin. Alternatively, the appendage or linker regions may interact with clathrin directly.
The observation that both ␣ a and ␤2 subunits of AP-2 have clathrin-binding sites supports the concept that coat assembly proceeds by bivalent binding and stabilization of overlapping clathrin triskelia in a conformation that leads to polygon formation, essentially the cross-linking model proposed earlier (1,2). Whether this hypothetical mechanism extends to other proteins such as AP-3/AP180, auxillin (27,28) and a novel AP-20 (29) that have been reported to promote clathrin assembly in vitro remains to be determined. In any case, the expanding group of proteins capable of promoting polymerization suggests that coat assembly may be invoked by different effectors under a variety of different circumstances in vivo.
Though both ␣ a and ␤2 subunits bind strongly to assembled clathrin lattices, ␣ a subunit binding to clathrin trimers is much more sensitive to pH in the physiological range than is the ␤ subunit. This seems unlikely to be a consequence of lability of the isolated ␣ conformation in solution, as we detect no change in either the proteolysis pattern or susceptibility of translated ␣ subunit over this pH range (data not shown). The increased affinity of the ␣ subunit for clathrin with decreasing pH correlates with the increased ability of AP-2 to drive coat formation with decreasing pH (21), arguing for a role of the ␣ subunit in lattice assembly. In addition, cytoplasmic acidification to pH 6.3-6.5 also results in "freezing" of clathrin lattices with increased curvature (32) thereby arresting receptor-mediated endocytosis (19,21,30,31). These observations suggest that the ␣-clathrin interaction may also be involved in lattice shape changes during vesiculation and endocytosis, or conceivably, that through this binding clathrin may affect other ␣ functions.