β3A-adaptin, a Subunit of the Adaptor-like Complex AP-3*

Recent studies have described a widely expressed adaptor-like complex, named AP-3, which is likely involved in protein sorting in exocytic/endocytic pathways. The AP-3 complex is composed of four distinct subunits. Here, we report the identification of one of the subunits of this complex, which we call β3A-adaptin. The predicted amino acid sequence of β3A-adaptin reveals that the protein is closely related to the neuron-specific protein β-NAP (61% overall identity) and more distantly related to the β1- and β2-adaptin subunits of the clathrin-associated adaptor complexes AP-1 and AP-2, respectively. Sequence comparisons also suggest that β3A-adaptin has a domain organization similar to β-NAP and to β1- and β2-adaptins. β3A-adaptin is expressed in all tissues and cells examined. Co-purification and co-precipitation analyses demonstrate that β3A-adaptin corresponds to the ∼140-kDa subunit of the ubiquitous AP-3 complex, the other subunits being δ-adaptin, p47A (now called μ3A) and ς3 (A or B). β3A-adaptin is phosphorylated on serine residues in vivo while the other subunits of the complex are not detectably phosphorylated. β3A-adaptin is not present in significant amounts in clathrin-coated vesicles. The characteristics of β3A-adaptin reported here lend support to the idea that AP-3 is a structural and functional homolog of the clathrin-associated adaptors AP-1 and AP-2.

A major function of AP-1 and AP-2 is to link clathrin lattices to the corresponding membranes. This role is fulfilled by the ␥and ␤1-adaptin subunits of AP-1 and the ␣and ␤2-adaptin subunits of AP-2 (6 -9). The adaptors are also responsible for the recognition of sorting signals present in the cytosolic domains of integral membrane proteins (10 -21), an event that leads to the concentration of these proteins within clathrincoated areas of the trans-Golgi network and the plasma membrane. Recent evidence suggests that the 1 subunit of AP-1 and the 2 subunit of AP-2 are directly involved in signal recognition (16,18,22).
In the past few years, it has become clear that the structure and function of clathrin coats serve as paradigms for other protein coats. Indeed, some subunits of the non-clathrin coat, COPI, are structurally related to subunits of AP-1 and AP-2 (23)(24)(25)(26). In addition, recent studies have identified other mammalian proteins that display significant homology to AP-1 and AP-2 subunits and are components of previously unknown coats. Pevsner et al. (27) isolated cDNAs encoding two proteins, named p47A and p47B, that exhibit ϳ80% identity to each other and ϳ30% identity to 1 and 2. Newman et al. (28) described another protein, named ␤-NAP, that is ϳ30% identical to ␤1and ␤2-adaptins. Finally, Watanabe et al. (29) and Dell'Angelica et al. (30) reported the identification of two proteins named 3A and 3B; these proteins are 84% identical to each other and ϳ30% identical to 1 and 2. Some of these subunits are expressed in a wide variety of tissues and cell lines (p47A, 3A, and 3B), while others are only expressed in brain, spinal cord, and neuronal cell lines (p47B and ␤-NAP).
Not surprisingly, these homologs of AP-1 and AP-2 subunits were found to exist as heterotetrameric assemblies resembling adaptor complexes (30,31). One of these complexes, expressed in neuronal cells, contains ␤-NAP and either p47A or p47B (31). A similar complex, which was named AP-3, is expressed in all cells examined to date and contains subunits of ϳ160, ϳ140, ϳ47, and ϳ22 kDa (30). The ϳ47-kDa protein corresponds to p47A while the ϳ22-kDa protein is either 3A or 3B. The ϳ140-kDa subunit of this complex is immunologically related to the neuronal protein ␤-NAP but distinct from it based on its expression in various non-neuronal cell lines (30).
In this paper we report the cloning of a novel cDNA encoding the ϳ140-kDa subunit of the ubiquitous AP-3 complex. This protein, referred to as ␤3A-adaptin, is closely related to ␤-NAP and displays significant homology to ␤1and ␤2-adaptins. Consistent with it being a subunit of the ubiquitous AP-3 complex, ␤3A-adaptin is expressed in a wide variety of tissues and cell lines. Biochemical analyses show that ␤3A-adaptin is phosphorylated on serine residues in vivo and is absent from clathrincoated vesicles. The similarity of ␤3A-adaptin to ␤1and ␤2adaptins lends support to the idea that AP-3 is structurally and functionally related to AP-1 and AP-2, and that it is likewise involved in the regulation of intracellular protein trafficking.

EXPERIMENTAL PROCEDURES
Cloning of ␤3A-adaptin cDNA-Two EST clones (GenBank accession codes R02669 and T98538) from human fetal liver/spleen (Washington University-Merck EST Project) were found to encode portions of a novel protein with significant homology to ␤-NAP. Based on these partial sequences, a full-length cDNA encoding this protein, named ␤3A-adaptin, was isolated from a Marathon-Ready human pancreas cDNA library (CLONTECH, Palo Alto, CA) by a combination of 5Ј-and 3Ј-RACE 1 PCR, using the Advantage cDNA PCR kit (CLONTECH). The specific primers used in a first and second (nested) 5Ј-RACE reactions were complementary to nucleotides 3181-3205 and 2955-2981 of the ␤3A-adaptin cDNA, respectively. The first and second (nested) primers used for 3Ј-RACE PCR corresponded to nucleotides 1773-1800 and 1898 -1922 of the full-length cDNA, respectively. Both 5Ј and 3Ј nested PCR products were cloned into the pNoTA/T7 shuttle vector (5 Prime 3 3 Prime, Inc., Boulder, CO). Several independent clones were isolated and sequenced to guard against errors introduced by the DNA polymerase during PCR amplification. Both strands were sequenced by the dideoxy method.
Cells-The sources and culture conditions for all the human cell lines used in this study are described elsewhere (30).
Northern Blot and RT-PCR Analyses of mRNA Expression-Northern blot analysis was carried out as described before (30). The ␤3Aadaptin probe was obtained by PCR using a 5Ј primer corresponding to nucleotides 1898 -1922 and a 3Ј primer complementary to nucleotides 2955-2981 of the full-length cDNA, respectively. The probe used for detecting the ␤-NAP mRNA consisted of a 450-base pair fragment which was PCR-amplified from the EST clone 165789 (Washington University-Merck EST Project) using 5Ј and 3Ј primers corresponding to nucleotides 2000 -2028 and 2424 -2449 of the full-length ␤-NAP cDNA (GenBank accession number: U37673), respectively. The above sets of primers were also used for RT-PCR analysis of the expression of ␤3A-adaptin and ␤-NAP mRNAs in cell lines; the analysis was performed by using the Gene Amp XL RNA PCR kit (Perkin-Elmer, Branchburg, NJ) according to the manufacturer's instructions.
Production of GST Fusion Proteins-To prepare a series of GST fusion proteins bearing different segments of ␤3A-adaptin, the corresponding cDNA fragments were engineered by PCR to be cloned inframe into the pGEX-5X-1 vector (Pharmacia Biotech, Uppsala, Sweden). The ␤3A-adaptin cDNA segments corresponding to amino acids 1-287, 1-642, and 810 -1094 were cloned into the EcoRI-NotI cloning sites of the vector, while the segment encoding amino acids 643-809 was cloned into the BamHI-NotI sites of the vector. The DNA sequence of all the constructs was confirmed by manual sequencing. Fusion proteins were expressed in Escherichia coli cells and then affinitypurified using glutathione-Sepharose 4B beads (Pharmacia Biotech Inc.) following the manufacturer's instructions.
Antibodies-Monoclonal anti-␣-adaptin antibody 100/2 was obtained from Sigma. The preparation and purification of polyclonal rabbit antibodies to p47 (A and B) and 3 (A and B) were described previously (30). The preparation of an antiserum against a GST fusion protein that bears the hinge region of ␤-NAP (GST-␤-NAP 647-796 ) has also been described previously (30). Here, this antiserum was passed through a column containing GST coupled to Affi-Gel 15 (Bio-Rad), to remove anti-GST antibodies, and then affinity-purified (32) using as a ligand a peptide comprising residues 647-796 of ␤-NAP. The purified antibody, herein referred to as ␤3H1, recognizes the hinge region of both ␤-NAP and ␤3A-adaptin, as inferred from immunoprecipitation experiments with the corresponding fragments translated in vitro. The ␤3H7 antibody was raised in rabbits by immunization with the GST-␤3A 643-809 fusion protein. The antibody was affinity-purified using as a ligand the same fusion protein coupled to Affi-Gel-15 beads, and was subsequently immunoabsorbed with GST, followed by GST-␤-NAP 647-796 to obtain a monospecific antibody to the hinge region of ␤3A-adaptin. The lack of cross-reactivity between the ␤3H7 antibody and ␤-NAP was corroborated by immunoprecipitation experiments and by immunoblotting. The ␤3A1 and ␤3C1 antibodies were obtained by immunizing rabbits with GST-␤3A 1-642 and GST-␤3A 810 -1094 fusion proteins, followed by affinity purification on immobilized GST-␤3A 1-287 and GST-␤3A 810 -1094 , respectively. Both antibodies were absorbed with GST. Polyclonal rabbit antibodies to BSA (Cappel, Cochranville, PA) or to the FLAG epitope (Santa Cruz Biotechnology, Santa Cruz, CA) were used as irrelevant antibody controls. The preparation of an antibody to human ␦-adaptin is described elsewhere. 2 Immunoprecipitation and Immunoblotting-Metabolic labeling of M1 cells with [ 35 S]methionine, immunoprecipitation-recapture experiments, and immunoblot analysis were performed as described previously (30).
Alkaline Phosphatase Treatment-The subunits of the AP-3 complex were isolated from [ 35 S]methionine-labeled M1 cells by immunoprecipitation-recapture (30). The immunoprecipitates were resuspended in 50 mM Tris-HCl (pH 8.5), 1 mM EDTA and divided into two aliquots. One of the aliquots was treated with 1 milliunit of calf intestinal alkaline phosphatase (Boehringer Mannheim) for 1 h at 37°C; the other one was mock-treated. Samples were analyzed by SDS-PAGE (33) followed by fluorography.
Other Materials and Methods-The preparation of cytosol from M1 cells and its fractionation by gel filtration were performed as described previously (30). For isolation of the complex by affinity chromatography, the cytosol was passed through a Protein A-Sepharose column (1.6 ml) and then loaded onto a 0.4-ml column containing 0.7 mg of purified anti-3 antibody which had been covalently coupled to Protein A-Sepharose (32). Bound proteins were eluted with 0.1 M glycine (pH 2.5). A crude membrane fraction and purified clathrin-coated vesicles from bovine brain were the kind gift of Lois Greene and Evan Eisenberg (National Heart, Lung, and Blood Institute, National Institutes of Health).

Molecular Cloning and Sequence Analyses of a cDNA Encod-
ing ␤3A-adaptin-Our previous work demonstrated the existence in non-neuronal cells of a protein immunologically related to the neuronal ␤-NAP (30). To identify this protein, we searched EST data bases for ␤-NAP homologs. Two EST clones from human fetal liver/spleen were found to encode a novel protein with significant homology to ␤-NAP. The complete cDNA was obtained using 5Ј-and 3Ј-RACE procedures based on partial sequences of the EST clones. Analysis of the sequence of this 3950-base pair cDNA revealed a long open reading frame encoding a protein of 1094 amino acids with a predicted molecular mass of 121,350 Da. The protein displayed significant homology not only to ␤-NAP (Fig. 1), but also to the Saccharomyces cerevisiae chromosome VII open reading frame YGR261c and to mammalian ␤1and ␤2-adaptins (Fig. 2). The new human protein was named ␤3A-adaptin.
The homology of ␤3A-adaptin to ␤-NAP extends over the entire length of their polypeptide chains (Fig. 1), although the degree of sequence similarity varies in the three regions (Fig.  2B). The percentage of conserved amino acid residues shared by the two proteins is highest in the A region; this region also exhibits the highest degree of similarity to the S. cerevisiae YGR261c protein and to the mammalian ␤-adaptins (Fig. 2B). In the ␤1and ␤2-adaptins, the amino-terminal segment corresponds to a "head" or "core" domain of the proteins where interactions with the other adaptor subunits are thought to take place. The head domains of ␤1-adaptin, ␤2-adaptin, and ␤-NAP have been shown to contain up to 14 Arm repeats (35, 36), a degenerate motif that probably functions as a proteinprotein interaction element. Analysis of the ␤3A-adaptin sequence reveals that this protein also contains 12 or 13 Arm repeats in the A region (not shown).
The H region of ␤3A-adaptin is strongly hydrophilic and rich in acidic residues (31%) and serine residues (26%) (Fig. 1). This region contains many potential sites for phosphorylation by the serine/threonine kinases casein kinase I and casein kinase II (21 and 25 sites, respectively; Refs. 37 and 38). The H region of ␤3A-adaptin is 35 and 19% identical to the analogous regions in ␤-NAP and YGR261c, respectively (Fig. 2B). Although the percentage of identical residues is lower in this region, as com-pared with the A region, the general characteristics of the sequence (i.e. high content of acidic and serine residues) are conserved among these proteins. No significant homology to mammalian ␤-adaptins was observed in this region, although the analogous segment in ␤1and ␤2-adaptins is also hydrophilic and displays a hinge-like structure.
Finally, the C region of ␤3A-adaptin is 50% identical to the homologous segment of ␤-NAP but shows no significant homology to ␤1and ␤2-adaptins and is absent from YGR261c (Fig.  2B). In the ␤-adaptins, this segment corresponds to the "ear" or "appendage" domain.
A philogenetic tree constructed with the above sequences and that of the more distantly related COPI subunit, ␤-COP, (Fig. 2C) shows that ␤3A-adaptin, ␤-NAP, and YGR261c all belong to a group that may have diverged from the others early in the evolution of the family. The mammalian ␤1 and ␤2adaptins, a D. melanogaster ␤-adaptin and two other S. cerevisiae adaptins (Sc ADB1 and Sc ADB2) cluster together in a separate branch. The ␤-COPs from rat, fly, and yeast are distantly related to the members of the other two groups. Thus, these sequence analyses define a group of proteins (␤3A-adaptin, ␤-NAP, and YGR261c) that are more closely related to clathrin-associated ␤-adaptins than to COPI components.
Tissue and Cell Expression of the ␤3A-adaptin mRNA-The pattern of expression of the ␤3A-adaptin mRNA in various human tissues and cell lines was examined by Northern blot analysis (Fig. 3). A single ϳ4.2-kb ␤3A-adaptin message was detected in all tissues examined (Fig. 3A), as well as in both non-neuronal (M1, HeLa, and RD4) and neuronal cell lines (H4, SK-N-SH, SK-N-MC, and Ntera-2) (Fig. 3B). In contrast to the ␤3A-adaptin mRNA, the ␤-NAP mRNA is known to be expressed only in brain (Ref. 28 and data not shown). Northern analyses of various cell lines detected expression of the ␤-NAP mRNA only in the neuronal precursor line Ntera-2 (Fig. 3B). Additional analyses by RT-PCR, which is more sensitive than Northern analysis, revealed the presence of small amounts of ␤-NAP mRNA in other neuronal cells but not in non-neuronal cells (Fig. 3C). Thus, the ␤3A-adaptin mRNA is widely expressed, whereas expression of the ␤-NAP mRNA is restricted to brain and cell lines of neuronal origin.
␤3A-adaptin Is a Component of the AP-3 Complex-To characterize the biochemical properties of the ␤3A-adaptin protein, we raised antibodies to different regions of the molecule. The antibodies were affinity-purified and used to identify the protein by Western blot analysis of the human fibroblast cell line M1, which expresses the ␤3A-adaptin mRNA but not the ␤-NAP mRNA (Fig. 3). Antibodies to both the H and C regions of ␤3A-adaptin (␤3H7 and ␤3C1, respectively) were found to recognize a protein that migrated as a ϳ140-kDa polypeptide on SDS-PAGE and that co-eluted on a gel filtration column with the ϳ22-kDa protein 3 (Fig. 4A), a known component of the AP-3 complex (30). The peak elution of ␤3A-adaptin and 3 corresponded to a complex with Stokes radius of ϳ85 Å, as previously reported (30).
To address directly the question of whether ␤3A-adaptin is a component of AP-3, this complex was affinity-purified from M1 cells on an anti-3 column and the eluate was analyzed by Western blotting with anti-3 and anti-␤3A-adaptin antibodies (Fig. 4B). The anti-3 antibody recognized minor and major species of 20 -22 kDa, as previously shown (30), whereas the different antibodies to ␤3A-adaptin recognized a ϳ140-kDa species (Fig. 4B). Neither band was observed when the same procedure was performed on a control column without anti-3 antibody (data not shown). Thus, ␤3A-adaptin co-purifies with 3 on affinity chromatography.
The association of ␤3A-adaptin with 3 was also analyzed using an immunoprecipitation-recapture technique (30). This technique consisted of immunoprecipitating the AP-3 complex from [ 35 S]methionine-labeled M1 cells using an antibody to 3 and, after dissociation in the presence of SDS, isolating the individual subunits of the complex by re-precipitation with specific antibodies. Re-precipitation with the ␤3H1 antibody (Fig. 5A) and with the antibodies ␤3H7, ␤3A1, and ␤3C1 (not shown) further confirmed that ␤3A-adaptin is indeed a component of the AP-3 complex. Fig. 5A also illustrates that, in  (27) and another protein known as ␦-adaptin. 2 ␤3A-adaptin Is a Phosphoprotein-Because of the many potential phosphorylation sites predicted from the sequence (see above), we were interested in determining whether the protein was phosphorylated in vivo. To this end, we treated [ 35 S]methionine-labeled ␤3A-adaptin isolated from the cytosol of M1 cells with alkaline phosphatase and determined the migration of the untreated and treated samples on SDS-PAGE. The alkaline phosphatase treatment resulted in decreased migration of the entire population of labeled ␤3A-adaptin molecules, as can be seen in Fig. 5A (lanes 5 and 6), and with better resolution in Fig. 5B. This experiment suggested that ␤3A-adaptin is a phosphorylated protein at steady state. In contrast to ␤3Aadaptin, none of the other subunits of AP-3 changed their migration upon treatment with alkaline phosphatase (Fig. 5A).
The fact that the AP-3 complex exists both in soluble and membrane-bound pools (30) led us to investigate whether phosphorylation of ␤3A-adaptin correlates with association to membranes. We found that both the cytosolic and membrane-bound forms of ␤3A-adaptin were equally sensitive to alkaline phosphatase (Fig. 5B), thus deeming it unlikely that the observed phosphorylation regulates membrane association.
To confirm that ␤3A-adaptin is a phosphoprotein and to identify the amino acids that are phosphorylated, M1 cells were metabolically labeled with [ 32 P]orthophosphate and the AP-3 complex and each of its subunits were isolated by immunoprecipitation with specific antibodies (Fig. 6A). We observed that ␤3A-adaptin was the only subunit of the complex that incorporated [ 32 P]orthophosphate under the conditions of the experiment (Fig. 6A). Phosphoamino acid analyses of 32 P-labeled ␤3A-adaptin revealed that the phosphorylation was on serine residues (Fig. 6B). From these experiments, we concluded that ␤3A-adaptin exists as a serine-phosphorylated protein under basal conditions. Although we did not detect phosphorylation of the other subunits of AP-3, it is possible that they are phosphorylated less extensively or that they are only phosphorylated under certain physiologic conditions. AP-3 Is Not Enriched in Clathrin-coated Vesicles-Because of the structural similarities of AP-3 to the clathrin-associated adaptors AP-1 and AP-2, it was of interest to examine if AP-3 was associated with clathrin-coated vesicles. To this end, we determined the amount of ␤3A-adaptin associated with bovine brain-coated vesicles in comparison to a crude membrane fraction. This analysis was done by Western blotting using a monospecific antibody (␤3H7) to the ␤3A-adaptin hinge that does not recognize ␤-NAP. We performed a similar analysis using anti-bodies to two other components of the AP-3 complex, p47 (A and B) and 3 (A and B), and to a component of AP-2, ␣-adaptin, as a control. Two forms of ␣-adaptin (␣ a and ␣ c ; Ref. 39) were highly enriched in the clathrin-coated vesicle preparation relative to the crude membrane fraction (Fig. 7). This was in contrast to ␤3A-adaptin which was not detected in the clathrincoated vesicles, although it was present in high amounts in the crude membrane fraction (Fig. 7). Similarly, p47 and 3 were not detected in the clathrin-coated vesicle fraction (Fig. 7). Thus, these experiments suggest that the AP-3 complex is not associated with clathrin-coated vesicles. DISCUSSION In this study we describe a novel human protein named ␤3A-adaptin. The ␤3A-adaptin mRNA is expressed in a wide variety of tissues and cell lines and the protein itself has been shown to exist in various cell lines (this study; Ref. 30). Biochemical analyses demonstrate that ␤3A-adaptin is a subunit of the ubiquitous adaptor-like protein complex AP-3, which has been previously shown to exist in association with trans-Golgi network and/or endosomal compartments (30). The other subunits of the AP-3 complex are ␦-adaptin, 2 p47A (Ref. 27; now called 3A), and 3A or 3B (29, 30) (Fig. 5). The AP-3 complex is not associated with clathrin-coated vesicles, suggesting that it may be a component of a different coat. In this regard, the ubiquitous AP-3 complex resembles the complex containing the brain-specific ␤-NAP, which is also absent from clathrin-coated vesicles (28,31).
The existence of a closely-related homolog of the brain-specific ␤-NAP was first evidenced by the immunoprecipitation of a ϳ140-kDa protein from non-neuronal cells using an antibody to ␤-NAP (30). Since ␤-NAP is not expressed in non-neuronal cells (Refs. 28 and 31, this study), this protein had to be an immunologically cross-reactive homolog. Sequence analyses now show that ␤3A-adaptin shares 61% overall identity and 75% overall similarity with ␤-NAP. ␤-NAP itself is part of a complex with either the brain-specific p47B (now called 3B) or the ubiquitous p47A (3A) subunits, and with two other proteins that are also similar to subunits of AP-3 (31). This suggests that both the ubiquitous ␤3A-adaptin-containing complex and the brain-specific ␤-NAP-containing complex have a similar structure and may even share some common subunits. Moreover, in cells that express both the ubiquitous and brainspecific subunits (i.e. neurons), the subunits could combine to generate several different complexes.
In addition to ␤-NAP, other coat proteins display a lower but significant degree of homology to ␤3A-adaptin. This includes the S. cerevisiae gene product, YGR261c, also known as APL6/ YKS5. This protein is likely to be the yeast counterpart of ␤3A-adaptin and/or ␤-NAP and, in fact, has been shown to be part of a complex with three other proteins (APL5/YKS4, APM3/YKS6, and APS3/YKS7) which are closely related to AP-3 subunits. 3 ␤3A-adaptin also has a significant degree of homology to the clathrin-associated adaptor subunits ␤1and ␤2-adaptins and to the COPI subunit ␤-COP. The fact that ␤3A-adaptin is related to all of these organellar coat proteins and the ability of p47A (3A) to bind tyrosine-based sorting signals (30) strongly suggest that the AP-3 complex may be similarly involved in the regulation of intracellular protein trafficking.
An examination of the ␤3A-adaptin sequence reveals at least three distinct regions, named A (amino-terminal), H (hinge), and C (carboxyl-terminal) (Fig. 2B). The three regions are likely analogous to structural and functional domains that have been well defined in ␤1and ␤2-adaptins (40,41). The A region is homologous to the head or core domains of ␤1and ␤2-adaptins, which are involved in interactions with the other subunits of AP-1 and AP-2 (9,40,42). The H region is analogous to the hinge or "stalk" domains of ␤1and ␤2-adaptins, which mediate interactions with clathrin (7,43). Finally, the C region of ␤3A-adaptin might be analogous to the ear or appendage domains of ␤1and ␤2-adaptins. The function of the ␤1and ␤2-adaptin ear domains is still unclear, although the analogous domain of ␣-adaptin has been shown to bind regulatory molecules such as dynamin (44) and Eps15 (45). The domain organization of ␦-adaptin is also thought to resemble those of ␣and ␥-adaptin. 2 These similarities suggest that AP-3 may have an overall structure analogous to AP-1 and AP-2.
A salient feature of ␤3A-adaptin is that it is phosphorylated on serine residues (Figs. 5 and 6). Moreover, ␤3A-adaptin is the only subunit of AP-3 that is detectably phosphorylated under the conditions of our experiments. Although the sites of phosphorylation within ␤3A-adaptin have not been located, the presence of a large number of consensus sequences for phosphorylation by casein kinase I and II in the H region suggests that this region might be highly phosphorylated. While the degree of identity of ␤3A-adaptin to ␤-NAP and S. cerevisiae YGR261c is lower in the H region as compared with the A region, it is noteworthy that the overall characteristics of the segment are conserved among these proteins, including the presence of many potential sites of phosphorylation by casein 3 S. Lemmon and L. Robinson, personal communication. A crude membrane fraction and purified clathrin-coated vesicles from bovine brain were analyzed for the presence of ␣-adaptin, ␤3Aadaptin, p47, and 3 by Western blotting. The antibody to ␣-adaptin recognizes both the ␣ a (ϳ105 kDa) and ␣ c (ϳ102 kDa) forms of the protein (39). Similarly, the antibodies to 3 and p47 react with both the A and B forms of these proteins (30). The antibody to ␤3A-adaptin (␤3H7), on the other hand, is specific for this protein and does not cross-react with ␤-NAP. Notice the enrichment of ␣-adaptin in the clathrin-coated vesicle fraction and the absence of p47, 3, and ␤3Aadaptin from that fraction. Notice that ␤3A-adaptin was the only AP-3 subunit that labeled with 32 P under these conditions. B, 32 P-labeled ␤3A-adaptin was isolated from a polyacrylamide gel such as that shown in A, and subjected to two-dimensional phosphoamino acid analysis as described under "Experimental Procedures." The migration of phosphoserine (pS), phosphothreonine (pT), and phosphotyrosine (pY) standards was visualized with ninhydrin. ϩ, anode; Ϫ, cathode. kinase I and II. Indeed, ␤-NAP is also heavily phosphorylated both in vivo and in vitro (28) and YGR261c displays a genetic interaction with casein kinase I in yeast cells. 3 Components of the clathrin-associated adaptors (47)(48)(49)(50) and of COPI (51) have also been shown to be phosphorylated, suggesting that phosphorylation might be a common mechanism for regulating coat function within cells.
The findings presented here add to the growing evidence that AP-3 is a structural and functional homolog of AP-1 and AP-2. Indeed, all three complexes are capable of binding reversibly to membranes and have a similar heterotetrameric structure. Moreover, the analogous subunits of the three complexes are structurally related to each other and may even exhibit a similar domain organization. Finally, the analogous subunits might play similar roles. For instance, the p47A (3A) subunit of AP-3 is capable of binding tyrosine-based sorting signals (22,30), like its relatives 1 and 2 (16,18,22). The comparable domain organization of ␤1-, ␤2-, and ␤3A-adaptins suggests that they might fulfill similar roles in the interaction of the adaptor complexes with scaffolding proteins. Further studies of AP-3 should establish not only the extent to which it resembles AP-1 and AP-2 but also the functional differences that account for the existence of the three complexes in all cells.