Interaction of Insulin Receptor Substrate-1 with the ς3A Subunit of the Adaptor Protein Complex-3 in Cultured Adipocytes*

Signaling through the insulin receptor tyrosine kinase involves its autophosphorylation in response to insulin and the subsequent tyrosine phosphorylation of substrate proteins such as insulin receptor substrate-1 (IRS-1). In basal 3T3-L1 adipocytes, IRS-1 is predominantly membrane-bound, and this localization may be important in targeting downstream signaling elements that mediate insulin action. Since IRS-1 localization to membranes may occur through its association with specific membrane proteins, a 3T3-F442A adipocyte cDNA expression library was screened with non-tyrosine-phosphorylated, baculovirus-expressed IRS-1 in order to identify potential IRS-1 receptors. A cDNA clone that encodes ς3A, a small subunit of the AP-3 adaptor protein complex, was demonstrated to bind IRS-1 utilizing this cloning strategy. The specific interaction between IRS-1 and ς3A was further verified by in vitro binding studies employing baculovirus-expressed IRS-1 and a glutathione S-transferase (GST)-ς3A fusion protein. IRS-1 and ς3A were found to co-fractionate in a detergent-resistant population of low density membranes isolated from basal 3T3-L1 adipocytes. Importantly, the addition of exogenous purified GST-ς3A to low density membranes caused the release of virtually all of the IRS-1 bound to these membranes, while GST alone had no effect. These results are consistent with the hypothesis that ς3A serves as an IRS-1 receptor that may dictate the subcellular localization and the signaling functions of IRS-1.

Insulin exerts its specific biological effects in fat and skeletal muscle by binding to and activating its tyrosine kinase receptor present on the cell surface. Subsequent signaling from this receptor involves its autophosphorylation and tyrosine phosphorylation of substrate proteins that then act as docking sites for various SH2 domain-containing proteins. A major substrate of the insulin receptor tyrosine kinase is the insulin receptor substrate-1 (IRS-1). 1 This protein was first identified as a 185-kDa phosphoprotein from insulin-stimulated Fao hepatoma cells (1), and it was subsequently cloned (2) and purified from rat liver (3) and mouse 3T3-L1 adipocytes (4,5). In unstimulated rat and 3T3-L1 adipocytes, a significant fraction of IRS-1 is found associated with an intracellular membrane fraction (6 -8). Upon insulin stimulation, tyrosine-phosphorylated IRS-1 recruits PI 3-kinase to these intracellular membranes (9), one mechanism hypothesized to account for insulin's specific effects on increased glucose uptake (7,10). Subsequently, and in a time-and dose-dependent manner, insulin causes the translocation of IRS-1⅐PI 3-kinase complexes from these intracellular membranes into the cytoplasm (6 -8), an event that is accompanied by apparent serine/threonine phosphorylation of IRS-1 as well as other membrane components (6,8,11). Major questions that remain are how IRS-1 is localized to intracellular membranes, what the function(s) of such localization is, and how the membrane association of IRS-1 is regulated by insulin.
In order to answer these questions, more information is needed to establish the specific intracellular localization of IRS-1 within insulin-responsive cells. One mechanism by which proteins translocate between different subcellular compartments is via transport vesicles (for reviews, see Refs. [12][13][14][15][16][17][18]. These vesicles possess specific structural coat proteins on their surfaces that selectively and efficiently drive and direct protein trafficking. For example, clathrin-coated vesicles (reviewed in Ref. 19) contain triskelia of clathrin proteins as well as heterotetrameric adaptor protein (AP) complexes (20 -24). Clathrin-coated vesicles bud from two distinct membrane compartments, the plasma membrane and the trans-Golgi network (TGN), and the composition of the adaptor protein complex is dependent upon the source of the budding vesicle. AP-1 is present in TGN-derived vesicles, while AP-2 complexes associate with the plasma membrane (22,23). AP-1 and AP-2 complexes consist of a ␤-adaptin subunit (␤1 or ␤2), an ␣-(AP-2) or ␥-adaptin (AP-1) subunit, a subunit (1 or 2), and a subunit (1 or 2) (22,23). Functionally, ␣ and ␥ subunits specify associated vesicular proteins and the contents of the vesicle, while ␤ subunits mediate clathrin attachment to the membrane (reviewed in Refs. 14 and 19). Functional roles for the smaller subunits (, ) are less well defined, although results from previous studies demonstrate an interaction between subunits and tyrosine-based sorting signals (25)(26)(27). Additionally, deletion of the 1 yeast homolog in a Saccharomyces cerevisiae mutant with a defective clathrin heavy chain allele enhances the effects of certain clathrin-mediated processes (28).
Recently, a clathrin-independent complex was described that co-localizes with markers for both the TGN and endosomes (29 -32). This complex, AP-3, contains a ␤-adaptin subunit with * 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.
The isoforms designated ␤3A and B (32,33), which is homologous to ␤-NAP, a neuron-specific, non-clathrin-associated phosphoprotein related to ␤-COP (34). AP-3 also possesses a ␦-adaptin subunit, designated p160, which is homologous to the ␣and ␥-adaptins, a 3 subunit, isoforms that are designated p47A and B, and 3 A and B subunit isoforms (29,32). Functionally, each of the four subunits in the AP-3 complex of S. cerevisiae appears essential for cargo-selective transport to the yeast vacuole, since deletion of any of these subunits results in the mislocalization of alkaline phosphatase and the vacuolar t-SNARE, Vam3p, to a nonvacuolar compartment (35,36). Additionally, it was demonstrated that the eye pigment defect in the Drosophila melanogaster garnet mutant might be caused by compromised function of the ␦ subunit (37).
In the present studies, we sought to identify proteins that bind non-tyrosine-phosphorylated IRS-1 and hence might be candidates for directing its membrane localization. We report here that a cDNA clone encoding the adaptin subunit 3A was isolated from a cultured adipocyte cDNA expression library based on its ability to bind radiolabeled, baculovirus-expressed IRS-1. The interaction between 3A and IRS-1 was verified by in vitro binding studies using tagged, recombinant fusion proteins and by membrane receptor binding studies in which exogenous 3A effectively competed for endogenous IRS-1 bound to adipocyte intracellular membranes. Finally, immunoblot analysis of 3T3-L1 adipocyte subcellular fractions revealed that 3A and IRS-1 co-fractionate in detergent-insoluble low density membranes (LDMs). The identification of the adaptin subunit, 3A, as a potential receptor for IRS-1 in adipocyte intracellular membranes may facilitate our understanding of the intracellular targeting of IRS-1 and its role in insulin signaling.

EXPERIMENTAL PROCEDURES
Materials-Baculovirus expression vector pAcHLTA and linearized BaculoGold DNA were from Pharmingen. Nickel-NTA resin was from Qiagen. [␥-32 P]ATP (6000 Ci/mmol) and enhanced chemiluminescence detection reagents were from NEN Life Science Products. The catalytic subunit of protein kinase A and glutathione-agarose were from Sigma. Prepacked PD-10 columns and pGEX2TK were purchased from Amersham Pharmacia Biotech. Isopropyl-␤-D-thiogalactopyranoside (IPTG) was from Jersey Lab Supply. Horseradish peroxidase-conjugated goat anti-rabbit antibody was from Boehringer Mannheim. Rabbit anti-IRS-1 was purchased from Upstate Biotechnology, Inc. (Lake Placid, NY). TRIzol reagent was obtained from Life Technologies, and the multiple tissue Northern blot was from CLONTECH. The 3T3-F442A murine adipocyte cDNA expression library made in the ZapII system was generously provided by Dr. Bruce Spiegelman (Dana Farber Cancer Center, Boston, MA), and affinity-purified anti-3 was generously provided by Dr. Esteban Dell'Angelica (Cell Biology and Metabolism Branch-NICHD, National Institutes of Health, Bethesda, MD).
Construction and Purification of His-IRS-1 Protein-Murine IRS-1 cDNA in Bluescript SKII Ϫ was digested with SalI and NotI to yield a 4.9-kb cDNA fragment. This fragment was digested with BspE1 to remove 5Ј-untranslated sequence, yielding a 4.5-kb fragment. A 5Ј double-stranded oligonucleotide containing EcoRI and BspE1 ends was ligated, along with the 4.5-kb IRS-1 fragment, into EcoRI/NotI-digested pAcHLTA. The resulting plasmid, when introduced into Sf9 cells with linearized BaculoGold DNA, produced recombinant baculovirus particles. Infection of Sf9 cells with these virii induced expression of IRS-1 protein containing at its amino terminus a polyhistidine tag, a phosphorylation site for protein kinase A, and a thrombin cleavage site. The plasmid was verified by restriction mapping and sequencing.
A high titer baculovirus stock was amplified according to Pharmingen's specifications. Large scale cultures of Sf9 cells were grown in spinner flasks, and cells were harvested 3 days post infection. Cells were lysed on ice in buffer containing 10 mM Tris-HCl, pH 7.5, 130 mM NaCl, 1% Triton X-100, 10 mM NaF, 10 mM NaPP i , 10 mM NaP i , 10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride (PMSF), 100 M vanadate, and 16 g/ml benzamidine. Nickel-NTA resin equilibrated in wash buffer containing 50 mM NaP i , pH 8.0, 300 mM NaCl, 10% glycerol, and 10 mM imidazole was incubated with clarified lysates for 1 h at 4°C. The resin was washed five times with 10 volumes of wash buffer, and the protein was eluted with buffer containing 50 mM NaP i , pH 6.0, 300 mM NaCl, 10% glycerol, and 500 mM imidazole. Peak protein fractions were pooled and dialyzed at 4°C against buffer containing 25 mM Hepes and 150 mM NaCl. The protein concentration was determined by the method of Bradford (38), and aliquots were frozen at Ϫ80°C.
Labeling of His-IRS-1 Protein-100 g of purified IRS-1 were incubated in kinase buffer (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 12 mM MgCl 2 , 1 mM dithiothreitol (DTT)), 1 unit/l protein kinase A catalytic subunit, and 1 mCi of [␥-32 P]ATP at 4°C for 30 min in a final volume of 150 l, essentially according to a previously published procedure (39). The reaction was quenched by the addition of 350 l of buffer containing 10 mM NaP i , pH 8.0, 10 mM NaPP i , and 10 mM EDTA. Labeled IRS-1 was separated from unincorporated [␥-32 P]ATP on a PD-10 column equilibrated in buffer containing 10 mM Tris-HCl, pH 7.5, and 150 mM NaCl. Peak fractions were determined by scintillation counting and SDS-polyacrylamide gel electrophoresis (PAGE) (40) and were pooled. The probe was kept at 0 -4°C and was used the same day for incubation with nitrocellulose filters in expression cloning procedures.
Expression Cloning of 3A from a 3T3-F442A Murine Adipocyte cDNA Expression Library-Approximately 10 6 plaques from a 3T3-F442A murine adipocyte cDNA expression library made in the ZapII system were screened with baculovirus-expressed IRS-1 protein labeled with protein kinase A catalytic subunit as described above. The library was plated out according to the manufacturer's specifications and allowed to grow for 6 h at 37°C. The agar was overlaid with nitrocellulose filters impregnated with 10 mM IPTG, and plaques were left to grow overnight at 37°C. Plates were then chilled at 4°C for 1 h before the filters were removed from the agar. Filters were washed three times in buffer 1 (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.05% Tween 20, 1 mM DTT) at room temperature for 15 min/wash. The filters were then blocked for 4 -5 h at 4°C in buffer 2 (10 mM Tris-HCl, pH 7.5, 250 mM NaCl, 5% nonfat dry milk, 1 mM DTT) and subsequently incubated overnight at 4°C in fresh buffer 2 containing ϳ10 6 cpm/ml 32 P-labeled IRS-1. Filters were washed several times in buffer 1 at room temperature, air-dried, and subjected to autoradiography. Positive plaques were identified, isolated, and purified by further rounds of screening. Excision of Bluescript phagemids was according to the manufacturer's protocol, and sequencing from both strands was performed using an automated sequencer (Applied Biosystems model 373). Nucleotide and deduced amino acid sequences were then subjected to data base searches to determine homology to known cDNAs and proteins.
Northern Blot Analysis-Total cellular RNA was isolated from murine fat pads, rat fat pads, rat adipocytes prepared from fat pads, 3T3-L1 fibroblasts, and 3T3-L1 adipocytes utilizing TRIzol reagent. Equal amounts of RNA were run on an agarose-formaldehyde gel and transferred to Hybond N (Amersham Pharmacia Biotech) as described by the manufacturer. The filter was baked for 2 h at 80°C and was prehybridized overnight at 42°C in buffer containing 50% formamide, 5ϫ standard saline citrate (SSC), 5ϫ Denhardt's solution, 0.5% SDS, and 200 g/ml denatured salmon sperm DNA. The filter was then hybridized overnight at 42°C in hybridization buffer (prehybridization buffer containing 5% dextran sulfate) and 3 ϫ 10 6 cpm/ml 3A cDNA 32 P-labeled by random priming (41). The filter was rinsed briefly at room temperature; washed with 0.1ϫ SSC, 0.1% SDS for 30 min at 65°C; and then subjected to autoradiography. A Northern blot of various murine adult tissues was probed with 3 ϫ 10 6 cpm/ml 3A cDNA 32 P-labeled by random priming and was washed under the same high stringency conditions described above. Additionally, a Northern blot of total cellular RNA isolated from murine fat pads, rat fat pads, rat skeletal muscle, 3T3-L1 fibroblasts, and 3T3-L1 adipocytes was hybridized at 42°C with 3 ϫ 10 6 cpm/ml 3A cDNA 32 P-labeled by random priming. The filter was rinsed briefly at room temperature, washed with 0.2ϫ SSC, 0.1% SDS for 30 min at 45°C, and then subjected to autoradiography.
Construction and Affinity Purification of 3A Fusion Protein-In order to subclone 3A into pGEX2TK, full-length 3A cDNA was amplified by polymerase chain reaction using EcoRI-digested full-length 3A cDNA as template and an oligonucleotide primer to introduce a 5Ј BamHI site. The resultant construct was verified by sequencing. The glutathione S-transferase (GST)-3A fusion protein was purified from bacterial cultures following previously described methods. Briefly, after a 3-h incubation at 30°C with 0.4 mM IPTG, the bacterial culture was centrifuged, and the pellet was washed once with STE buffer (50 mM Tris-HCl, pH 7.2, 150 mM NaCl, 1 mM EDTA), frozen in ethanol/dry ice, and stored at Ϫ70°C. The pellet was thawed rapidly and resuspended in STE containing 1 mM PMSF, 10 g/ml aprotinin, and 10 g/ml leupeptin. Lysozyme was then added to a final concentration of 0.5 mg/ml, and the suspension was left on ice with occasional agitation for 20 min. The cell suspension was sonicated until it was no longer viscous, Triton X-100 was added to a final concentration of 1%, and the suspension was incubated with gentle agitation on ice for 30 min. The suspension was centrifuged for 10 min at 12,000 ϫ g at 4°C. The pellet was resuspended in 100 mM Tris-HCl, pH 7.5, containing 2 M NaCl and 1 mM PMSF, 10 g/ml aprotinin, and 10 g/ml leupeptin, and recentrifuged as above. This step was repeated; the pellet was resuspended in 6 M urea with 1 mM DTT and 1 mM PMSF, 10 g/ml aprotinin, and 10 g/ml leupeptin; and the suspension was recentrifuged. The supernatant was dialyzed to remove urea and was then incubated with glutathione-agarose beads for 1-2 h at 4°C. The beads were washed extensively with STE buffer containing protease inhibitors, and bound fusion protein was eluted with an equal volume of buffer consisting of 50 mM Tris-HCl, pH 8.0, with 10 mM reduced glutathione. Aliquots were frozen at Ϫ20°C.
In Vitro Binding of IRS-1 and 3A-Equal amounts of purified baculovirus-expressed, histidine-tagged IRS-1, and either purified GST-3A or the negative controls, GST-Rab5C or GST alone, were incubated in TBS (10 mM Tris-HCl, pH 7.5, 150 mM NaCl) for 2 h at 4°C. Nickel-NTA beads, washed in TBS with 0.1% bovine serum albumin (BSA), were added to precipitate the IRS-1, and the beads were then washed extensively in buffer containing 10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM ␤-mercaptoethanol, 0.5% Tween 20, 0.1% BSA, 100 mM imidazole. The washed beads were boiled in SDS sample buffer, and bound proteins were resolved by SDS-PAGE and then transferred to nitrocellulose filters. Filters were blocked in TBS containing 0.1% Triton X-100, 3% dry milk, and 0.5% BSA and then incubated with anti-GST antibodies. Filters were incubated with horseradish peroxidase-conjugated antibodies to rabbit IgG, followed by detection by chemiluminescence.
Additionally, 32 P-labeled IRS-1, prepared as described above, was incubated for 2 h at 4°C with either GST-3A or GST alone in equal amounts. In order to precipitate the GST fusion proteins, glutathioneagarose beads were added, and the beads were washed sequentially with wash buffer alone, wash buffer containing 0.5 M NaCl, and finally with wash buffer alone. The beads were then boiled in SDS sample buffer, and bound proteins were resolved by SDS-PAGE. The gels were fixed, dried, and subjected to autoradiography for detection of bound, labeled IRS-1.
Cell Fractionation of 3T3-L1 Adipocytes-3T3-L1 fibroblasts were differentiated into adipocytes as described previously (42). Cells were serum-starved overnight in Dulbecco's modified Eagle's medium plus 0.5% BSA, and cellular fractions were prepared exactly as described (6). Protein concentration of the resuspended membrane fractions and the concentrated cytosol was determined by the method of Bradford (38). For preparation of the Triton X-100-insoluble fraction, equal amounts of LDMs were resuspended in [2-(N-morpholino)ethanesulfonic acid] (MES) buffer containing 50 mM MES, pH 6.5, 1 mM EDTA, 150 mM NaCl, 50 mM NaF, 100 M vanadate, 30 mM NaPP i , 1 mM PMSF, 10 g/ml aprotinin, and 10 g/ml leupeptin. An equal volume of MES buffer containing 2% Triton X-100 was added to the LDMs so that the final concentration of Triton X-100 was 1%. The membranes were incubated on ice for 30 min and then centrifuged at 350,000 ϫ g in a TL-100 ultracentrifuge (Beckman). The pellet was resuspended in MES buffer. Equal protein (30 g) from each fraction was subjected to SDS-PAGE and transferred to nitrocellulose filters, and the filters were blocked as described above. Filters were then incubated with either anti-IRS-1 or anti-3 antibodies. Filters were incubated with horseradish peroxidase-conjugated antibodies to rabbit IgG, followed by detection by chemiluminescence.
Release of Membrane-bound IRS-1 by GST-3A-10 g of LDM were resuspended in buffer containing 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 50 mM NaF, 100 M vanadate, 30 mM NaPP i , 1 mM PMSF, 10 g/ml aprotinin, 10 g/ml leupeptin. Resuspended LDMs were incubated with buffer alone or increasing amounts (5-75 g) of GST-3A or 75 g of GST alone for 2 h at 4°C. The samples were centrifuged at 350,000 ϫ g for 30 min in a TL-100 ultracentrifuge, and supernatants were transferred to fresh tubes. The pellets and equal aliquots of the supernatants were boiled in SDS sample buffer and subjected to SDS-PAGE. Proteins were transferred to nitrocellulose filters, and the filters were incubated with anti-IRS-1 antibodies. Filters were incubated with horseradish peroxidase-conjugated antibodies to rabbit IgG, followed by detection by chemiluminescence.

RESULTS
It was previously demonstrated that in basal 3T3-L1 adipocytes, IRS-1 is predominantly associated with intracellular membranes (6 -8) and that following insulin stimulation of these cells, PI 3-kinase is recruited to this membrane-bound IRS-1 (9). These findings are significant in that they suggest a discrete role for membrane-bound IRS-1 to deliver PI 3-kinase to a specific intracellular locale (7,10,43), a step that is perhaps critical to the observed functional role that PI 3-kinase plays in insulin-stimulated glucose transport (44 -49). Evidence indicates that the interaction of IRS-1 with intracellular membranes does not occur through the pleckstrin homology domain of IRS-1 (50,51), suggesting that interactions between IRS-1 domains and membranes may be protein/protein-as well as lipid/protein-mediated. In order to identify putative membrane-localized receptors for IRS-1 in cultured adipocytes, a 3T3-F442A murine adipocyte expression library was screened for proteins that could bind 32 P-labeled non-tyrosine-phosphorylated IRS-1. DNA prepared from a purified positive plaque obtained using this screening strategy was sequenced, and the nucleotide and deduced amino acid sequences were subjected to data base searches to determine homology to known cDNAs and proteins. The 1.12-kb clone contains 76 nucleotides of 5Ј noncoding sequence, an open reading frame of 582 nucleotides, and 466 nucleotides of 3Ј noncoding sequence (Fig. 1). The open reading frame exhibits 95% homology at the nucleotide level and 100% identity at the amino acid level with human 3A (Fig. 1), a protein with a predicted M r of 21,732 and an isoelectric point of 5.2 (30,31). 3A was recently shown to be a small subunit of AP-3, an adaptin complex present in endosomes and the TGN (31,32). Amino acid sequences that are underlined and designated as domains 1 and 2 are motifs found in the small subunits of AP-1, AP-2, and AP-3.
Northern blot analyses were performed to determine whether the expression pattern of 3A in murine tissues differs from that in human tissues. Results presented in Fig. 2A show a single hybridizing mRNA species with an apparent size of approximately 1.5 kb in all murine tissues, a size consistent with that reported for human 3A (30,31). The tissue distribution of 3A in murine tissues differs substantially from the human profile in that the highest mRNA levels are present in testis, liver, lung, kidney, and brain with nondetectable levels present in heart, skeletal muscle, and spleen. In contrast, the abundance of 3A in different human tissues showed the highest mRNA levels to be present in heart and testis, intermediate levels to be present in brain, liver, skeletal muscle, and spleen, and lower levels to be present in kidney and lung (31). We next examined the abundance of 3A mRNA in different tissues and cells used as fat cell models. Fig. 2B illustrates that a 1.5-kb 3A mRNA species is present in high levels in cultured 3T3-L1 fibroblasts and that this mRNA species reproducibly increases 2-fold following differentiation of the fibroblasts into adipocytes. In contrast, 3A mRNA is not detected in total cellular RNA prepared directly from rat fat pads or mouse fat pads or in total cellular RNA from adipocytes prepared from rat fat pad. In order to verify that equal amounts of RNA were loaded in each lane of the Northern blot presented in Fig. 2, A and B, the blots were stripped and reprobed with ␤-actin cDNA (data not shown).
The results presented in Fig. 2, A and B, demonstrate that under high stringency conditions, 3A mRNA is not detectable in primary adipocytes or skeletal muscle prepared from adult mouse or rat. However, when a filter similar to that shown in Fig. 2B was hybridized with 32 P-labeled 3A cDNA followed by low stringency washes, it was found that additional abundant mRNA species that are recognized by our probe exist in these cell types (Fig. 2C). Most notably, a transcript of slightly slower mobility than the 1.5-kb 3A mRNA is the major hybridizing species in rat skeletal muscle under low stringency conditions. In addition, as shown in Fig. 2C, approximately equal amounts of both transcripts are present in primary adipocytes from rat and mouse using the lower stringency washing conditions.
In order to confirm that 3A binds IRS-1, the interaction between IRS-1 and 3A was examined in in vitro binding experiments as depicted in Fig. 3. Equal amounts of purified baculovirus-expressed histidine-tagged IRS-1 and either purified GST-3A or the negative controls, GST-Rab5C or GST alone (Fig. 3A), were incubated together, and then nickel-NTA beads were added to precipitate the IRS-1. Anti-GST immunoblot analysis of the beads (Fig. 3B) and the supernatants (data not shown) demonstrates that approximately 5-fold more GST-3A than negative control proteins (GST and GST-Rab5c) binds immobilized histidine-tagged-IRS-1 . Results from control experiments (data not shown) indicate that approximately 10% of the total GST-3A binding to IRS-1 immobilized on the beads is nonspecific and occurs with the beads alone, while 70% of the total GST binding to the immobilized IRS-1 occurs nonspecifically. The results of the converse experiment are presented in Fig. 3, C and D. 32 P-Labeled-IRS-1 was incubated with either GST-3A or the negative control GST immobilized onto glutathione-agarose beads as described under "Experimental Procedures," and bound and eluted IRS-1 was detected by autoradiography. 3-Fold more IRS-1 bound to GST-3A than to GST alone (Fig. 3D), consistent with the results shown in Fig. 3B. Taken together, the data shown in Fig. 3 demonstrate that a specific interaction occurs between IRS-1 and 3A.
To examine the localization of 3A and IRS-1 in cultured adipocytes, membrane and cytosolic fractions were prepared from 3T3-L1 adipocytes, and these subcellular fractions were immunoblotted with affinity-purified anti-peptide antibodies recognizing IRS-1 and both the A and B isoforms of 3. As shown in Fig. 4 and as previously published (6 -8), under basal conditions, most of the IRS-1 protein in 3T3-L1 adipocytes is present in the LDM fraction. Insulin stimulation causes a doseand time-dependent decrease of IRS-1 in the LDMs with a concomitant increase in IRS-1 in the cytosol. When the IRS-1containing LDMs are further fractionated by incubation in 1% Triton X-100 followed by centrifugation to pellet detergentresistant membrane structures, IRS-1 is recovered in the detergent-resistant pellet as shown in Fig. 4. Immunodetectable 3 co-fractionates with membrane-bound IRS-1, and our data do not indicate that any 3 translocates with IRS-1 into the cytosol (Fig. 4). These data presented are similar to observations published by Clark et al. (52,53) that IRS-1 is present in detergent-resistant LDMs. The presence of both IRS-1 and 3A in the LDM fraction of cultured adipocytes is consistent with the demonstrated interaction of IRS-1 and 3A in the expression cloning format and the in vitro binding experiments, sug-gesting the hypothesis that 3A functions as the IRS-1 receptor in intracellular membranes of cultured adipocytes.
To investigate whether 3A functions as an IRS-1 receptor in vivo, a receptor competition binding assay was performed utilizing LDMs prepared from basal 3T3-L1 adipocytes. We hypothesized that if IRS-1 is bound to 3A in these membranes, then it should be possible to compete with endogenous 3A for binding to endogenous IRS-1 by adding exogenous GST-3A. Equal amounts of LDM protein were incubated with increasing amounts of GST-3A, the membranes were pelleted, and both the membranes and the supernatants were immunoblotted for the presence of IRS-1. Fig. 5 shows that incubating LDMs with increasing amounts of GST-3A resulted in increasing amounts of immunodetectable IRS-1 in the supernatants, while similar amounts of GST (Fig. 5) or another negative control, GST-Rab5c (data not shown), did not increase the IRS-1 detected in the supernatants. Additionally, corresponding decreases in immunodetectable IRS-1 were observed in the membrane pellets with increasing GST-3A. The addition of exogenous GST-3A to the LDMs did not result in selective  3. 3A specifically associates with IRS-1 in vitro. A and B, baculovirus-expressed, polyhistidine-tagged IRS-1 was incubated with GST alone, GST-Rab5c, or GST-3A for 2 h at 4°C and precipitated with nickel-NTA resin, and the bound proteins were resolved by SDS-PAGE, transferred to nitrocellulose, and detected with anti-GST antibodies followed by horseradish peroxidase-conjugated secondary antibody. Detection was by chemiluminescence. A, control blot illustrating that equivalent amounts of GST, GST-Rab5c, or GST-3A were used to bind IRS-1. Each lane contains 1 g of fusion protein. B, comparative binding of 10 g of IRS-1 to 10 g of GST, GST-Rab5c, or GST-3A. The bands corresponding to the various fusion proteins are labeled. Results are representative of four independent experiments. C and D, purified baculovirus-expressed IRS-1 was 32 P-labeled as described under "Experimental Procedures" and was incubated with GST or GST-3A for 2 h at 4°C. Complexes were precipitated with glutathione-agarose, washed extensively, and then resolved by SDS-PAGE. The gel was fixed, dried, and subjected to autoradiography. C, control anti-GST blot illustrating that equal amounts of GST and GST-3A were used to bind IRS-1. Each lane contains 1 g of fusion protein. D, comparative binding of 32 P-labeled-IRS-1 to GST alone or GST-3A. A representative autoradiogram is shown. IRS-1 bands in each lane from four independent experiments were quantitated using a scanning densitometer and are presented in the bar graph with IRS-1 binding to GST alone given an arbitrary value of 1. bulk protein release from these membranes compared with GST or GST-Rab5c addition, since a portion of the supernatants from each condition were run on a gradient gel and silver-stained (data not shown), and the protein profile for all three conditions appeared the same. These results indicate that 3A binding to IRS-1 releases it from its endogenous receptor in adipocyte intracellular membranes. DISCUSSION Several observations underscore the importance of subcellular localization of key elements in insulin signaling pathways. It was initially observed by Kelly (9) that insulin stimulation of adipocytes results in the redistribution of cytosolic PI 3-kinase to an intracellular membrane fraction. This evidence was later followed by the observations that tyrosine-phosphorylated residues within IRS-1 bind SH2 domains present in PI 3-kinase (54 -56) and that IRS-1 in basal 3T3-L1 adipocytes is localized to intracellular membranes (6 -8). Thus, it is the tyrosine phosphorylation of this membrane-bound IRS-1 that recruits PI 3-kinase to intracellular membranes upon insulin stimulation. Novel findings presented in this study indicate IRS-1 may be localized to the intracellular membranes of basal 3T3-L1 adipocytes through an interaction with 3A, the small subunit of the AP-3 adaptin complex. The interaction between IRS-1 and 3A was initially demonstrated by screening a cultured adipocyte cDNA expression library, and it was subsequently confirmed by in vitro binding studies using purified fusion proteins (Fig. 3). Both IRS-1 and 3A are present in detergent-resistant structures present in LDMs (Fig. 4), thus localization of the two proteins is consistent with 3A acting as a receptor for IRS-1. Finally, the results presented in Fig. 5 demonstrate that exogenous 3A can specifically displace IRS-1 from its endogenous intracellular membrane receptor in 3T3-L1 adipocytes, providing strong evidence that 3A possesses structural features either the same as or very similar to those of the IRS-1 receptor present in these membranes.
Other examples exist whereby substrates of receptor tyrosine kinases exhibit specific interactions with proteins present in adaptin complexes. The ubiquitous Eps15 protein, initially described as a substrate of the epidermal growth factor receptor, is constitutively associated with AP-2 (57). Additionally, Okabayashi et al. (58) found that Shc, another substrate of the activated insulin receptor, interacts with AP-2 complex adaptins in vitro (58). Although neither study identified which specific adaptin subunit was responsible for the direct interaction with its respective protein, it was conclusively demonstrated in the former study that GST-Eps15 precipitated AP-2 complexes from lymphocytes, epithelial cells, and fibroblasts, while in the latter study, GST-Shc associated tightly with intact AP-2 holocomplexes from bovine brain lysates.
The Northern blot results presented in Fig. 2, A and B, in conjunction with the Northern analyses of human 3 subunits presented in the study by Dell'Angelica et al. (31) demonstrate that various forms of 3 mRNA are expressed at moderate to high levels in a variety of cultured cells and tissues, including 3T3-L1 cultured adipocytes. In contrast, high stringency Northern blot analysis of mRNAs from primary rodent adipocyte and skeletal muscle, two well described systems in which to study insulin's major effects, showed undetectable levels of 3A mRNA (Fig. 2B). If binding to 3A ensures proper membrane localization of IRS-1 so that it can mediate its biological effects, a stoichiometric association between 3A and IRS-1 should exist. Since both primary rodent adipocytes and skeletal muscle contain abundant levels of IRS-1, we would expect at least detectable levels of 3A in these cell types. Probing and washing a similar filter to that shown in Fig. 2B using lower stringency conditions revealed the presence of the 1.5-kb 3A transcript as well as an abundant, slower migrating transcript in primary rodent adipocytes and skeletal muscle (Fig. 2C). We are currently pursuing the identification of additional, possibly novel, isoforms of 3 present in insulin-responsive tissues.
The exact structure and function for AP-3, as well as the identification of the cellular compartments to which it localizes, are not yet defined. Similar to the clathrin-associated adaptin complexes, AP-1 and AP-2, the AP-3 complex is composed of four distinct subunits: ␤3A, which is more related to the neuronal ␤-NAP than to ␤1 or ␤2; ␦; 3; and 3 (29, 31-33). While electron microscopy studies have not been performed on AP-3, FIG. 5. GST-3A causes release of IRS-1 from its endogenous 3A receptor in low density membranes. 10 g of low density membranes were incubated with buffer only or with increasing amounts of GST-3A (5-75 g) or 75 g of GST alone for 2 h at 4°C. The membranes were then pelleted by centrifugation. Pellets and supernatants were resolved on a 6% SDS-PAGE gel and transferred to nitrocellulose. The filter was blocked and incubated with anti-IRS-1 followed by horseradish peroxidase-conjugated secondary antibody. Detection was by chemiluminescence. Bands corresponding to IRS-1 are indicated. The results are representative of three independent experiments. sequence homologies of AP-3 complex subunits with the subunits of AP-1 and AP-2 suggest that all three complexes share overall similarity in their structural features (32). Differences do exist, however, and probably it is these differences that most contribute to the distinct subcellular localizations of AP-1, AP-2, and AP-3 and to their purported functions.
Immunofluorescence microscopy demonstrates that the 3containing complex is predominantly concentrated in a juxtanuclear area, suggesting that it is associated with the TGN, yet 3 immunostaining also extends far into the periphery of the cell (31). 2 An identical immunostaining pattern has also been observed for ␦ (32). Extensive co-localization of peripheral AP-3 staining with endosomal markers such as the transferrin receptor suggests that the AP-3 complex functions in peripheral regions of the cell but at sites that are clearly distinct from the plasma membrane where AP-2 is localized. Since IRS-1 is specifically associated with LDMs in nonstimulated 3T3-L1 adipocytes, while very little IRS-1 (ϳ3%) protein is detected in the plasma membrane fraction (6,8), the observed interaction between IRS-1 and 3A is consistent with AP-3, and not AP-2, involvement in the localization of IRS-1. Immunoblot analyses indicate that ␤3A and 3 are not detectable in purified clathrin-coated vesicles (29,(31)(32)(33), further distinguishing the 3containing AP-3 complex from AP-1 and AP-2, both established clathrin-associated complexes, although a more recent report indicates that an in vitro interaction between clathrin and AP-3 may exist (59). Utilizing affinity-purified anti-IRS-1, we were unable to detect IRS-1 in purified clathrin-coated vesicles prepared from 3T3-L1 adipocytes, primary rat adipocytes, or rat liver (data not shown), a finding that strengthens the notion that the interaction between subunits and IRS-1 is specific for a 3and not a 1and/or 2-containing complex.
It has been shown that in 3T3-L1 adipocytes, membranebound IRS-1 is tyrosine-phosphorylated by insulin receptors present at the cell surface, yet under conditions where the endocytosis of activated insulin receptors is inhibited, IRS-1 still becomes tyrosine-phosphorylated (6). This finding, as well as results from others (8,53), suggests that in basal adipocytes, IRS-1 is tethered to components that are proximal to, but that lie beneath, the plasma membrane. In support of this idea, the pleckstrin homology domain and the phosphotyrosine binding domain of IRS-1 have been shown to function in the efficient coupling between IRS-1 and the activated insulin receptor in the plasma membrane (60 -63), yet there is no direct evidence to suggest that these domains are responsible for IRS-1 localizing to intracellular membranes. Although our data are consistent with the findings of Clark et al. (52,53) that IRS-1 present in the LDM fraction is in a detergent-resistant compartment, the authors of those studies concluded that IRS-1 was possibly interacting with the cell cytoskeleton. The evidence underlying their conclusion was based on similarities between structures present in electron micrographs of their IRS-1-containing fraction and 5-15 nm intermediate filament cytoskeletal elements. However, a detailed investigation of whether IRS-1 specifically associates with cytoskeletal elements was not provided in this study. Besides the cytoskeleton, other subcellular compartments, including caveolae and clathrin-containing vesicles, have been demonstrated to be detergent-resistant. Thus, the precise intracellular localization of IRS-1 remains to be determined. The observed interaction between IRS-1 and 3A and the immunolocalization of AP-3 at the cell periphery are consistent with IRS-1 localizing to AP-3-containing structures in the cell.
The localization of IRS-1 to AP-3 may provide the specificity for subsequent signaling events, such as the insulin-stimulated recruitment and activation of PI 3-kinase and its downstream targets. It has been demonstrated that epidermal growth factor and platelet-derived growth factor also stimulate PI 3-kinase, yet in the case of these other growth factors, PI 3-kinase is recruited directly to the respective transmembrane receptor. Activated epidermal growth factor receptors have been shown to bind AP-2 complexes (64), an interaction mediated by the ability of the 2 subunit to recognize the tyrosine endocytic signal within the cytoplasmic tail of the receptor (12,65). Both 3A and 3B show a high degree of homology to the subunits of AP-3 (p47A and B), in the region of the subunit that recognizes and binds tyrosine-based sorting signals. Using the yeast two-hybrid system, however, this region in both subunits was shown not to interact with this motif, indicating that subunits have their own distinct function (31). The adaptin complexes are thought to be part of the cellular machinery involved in protein sorting (12,17). It was recently reported that the 3Ј phosphorylated lipid products of PI-3 kinase enhance the in vitro recognition of tyrosine-based sorting signals for 2, suggesting that these lipid products may function in vivo to regulate AP-2-mediated protein trafficking events (65). Substantial evidence exists to indicate that PI 3-kinase is a critical component of protein trafficking in eukaryotes, including reports that missorting of platelet-derived growth factor and CSF receptors occurs when PI 3-kinase binding sites are deleted from the receptor structure (66,67). Also, specific inhibitors of PI 3-kinase cause alterations in the targeting of proteins to their proper cellular destinations (68). Hence, it is notable that inhibition of insulin-stimulated PI 3-kinase by both wortmannin and LY29004 abolishes the insulin-stimulated translocation of GLUT4 from its intracellular storage compartment within the endosomal system to the cell surface (44,45). It will be interesting to determine whether the lipid products generated by PI-3 kinase bound to IRS-1 also function in vivo to enhance AP-3-mediated trafficking events, perhaps contributing to the regulation of insulin-stimulated glucose transport. In addition, insulin treatment of 3T3-L1 adipocytes causes the translocation of membrane-bound IRS-1 into the cytosol (6 -8), an event that recently was shown to be accompanied by phosphorylation of component(s) in the intracellular membranes (8). We are currently investigating whether insulin-stimulated phosphorylation regulates the association of IRS-1 with 3A.
The data in this present study indicate that IRS-1 binds 3A, and they are consistent with the notion that 3A may function as the IRS-1 receptor in intracellular membranes of cultured adipocytes. Defining a specific role for 3 subunits and for the AP-3 holocomplex in intact insulin-responsive cells in future experiments may therefore provide new insights into mechanism(s) of insulin-stimulated bioeffects including the glucose transport.