Differential Intracellular Localizations of GDP Dissociation Inhibitor Isoforms INSULIN-DEPENDENT REDISTRIBUTION OF GDP DISSOCIATION INHIBITOR-2 IN 3T3-Ll

a strikingly level of GDI-2 fractionates with total of cells, and adipocytes compared to GDI-1, which is virtually totally cytosolic. In 3T3-Ll adipocytes, most of the membrane-bound GDI-2 was present in a low density, intracellular membrane fraction. Immunodeple-tion of GLUT4-enriched vesicles from this membrane fraction also depleted significant amounts of GDI-2 proteins. Localization of both GDI-2 and GLUT4 in the same perinuclear regions of these cells was established by im- munofluorescence microscopy, whereas GDI-1 displayed a diffuse, cytoplasmic distribution. Insulin acutely de- creased both GLUT4 and GDI-2 protein levels in the low density microsomes by about 50%. Concomitantly, GLUT4

DK30898. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. transporter GLUT4 (3)(4)(5)(6)(7)(8). Whereas both GLUTl and GLUT4 translocate to the adipocyte cell surface in response to insulin, the bulk of insulin-stimulated hexose flux is mediated by GLUT4 (9)(10)(11). Recent studies indicate that insulin modulates both the exocytosis and endocytosis pathway of GLUT4 trafficking in fat cells (12)(13)(14). The molecular mechanism(s1 underlying GLUTl and GLUT4 protein movements in response to insulin is still obscure. Rab proteins, a family of at least 30 Ras-related small GTPbinding proteins, function as specific regulators of cell membrane trafficking and fusion (reviewed in Refs. [15][16][17][18]. By alternating between two distinct conformations in a GDP/GTPdependent fashion, these proteins are hypothesized to function as molecular switches, ensuring fidelity in the process of vesicle targeting to their correct acceptor compartment. Rab protein shuttling between cytosolic and membrane locations appears to be regulated by GDP dissociation inhibitor (GDI)' proteins (GDI-1 and GDI-2) (15)(16)(17)(18)(19)(20)(21). The two GDI isoforms display a high degree of deduced amino acid sequence identity (86%) and similar abilities to bind and solubilize membrane-associated forms of Rab4 and Rab5 proteins in a GDP/GTP-dependent manner (21). Both GDI isoforms are ubiquitously expressed in various rat and human tissues and cultured cells, including insulin-sensitive cells and tissues (21)(22)(23). No functional differences between GDI-1 and GDI-2 have yet been identified (21).
Several considerations suggest that Rab and GDI proteins may also be involved in the membrane movements of GLUT4 transporter proteins. Thus, the nonhydrolyzable GTP analog, GTP-& stimulates GLUT4 protein translocation in permeabilized fat cells and 3T3-Ll adipocytes (24,25). Furthermore, increased expression of a Rab3 isotype (Rab3D), as well as the GDI-1 and GDI-2 proteins, appears to be associated with differentiation of 3T3-Ll fibroblasts to insulin-responsive cells (21,26). Finally, the acute redistribution of Rab4 protein from fat cell low density microsomes to cytosol in response to insulin has recently been observed (27). The aim of the present studies was to characterize the cellular localizations of GDI isoforms in insulin-sensitive 3T3-Ll adipocytes and to determine whether insulin may modulate the distribution of these proteins as well. The results demonstrate a strikingly higher GDI-2 content associated with intracellular membrane compartments compared to that of GDI-1, and an acute effect of insulin to decrease GDI-2 levels in low density, intracellular membrane fractions that contain GLUT4.
GST-GDI-1 and GST-GDI-2 fusion proteins were produced as described previously (21). Their purity and concentrations were estimated electrophoretically by Coomassie Blue staining and, after electrotransfer on nitrocellulose membranes, by colloidal gold stain (Bio-Rad) and were used as references for the calculation of GDI-1 and GDI-2 cytosolic levels.
Cell 7keatment and Subcellular Fractionation-3T3-Ll adipocytes, serum-starved for 4 h prior to the experiment, were incubated in the absence or presence of insulin (10" M ) for different time periods, as specified in the figure legends. After washes, the cells were scraped in the homogenization buffer (20 mM Tris/HCI, pH 7.4,255 mM sucrose, 1 mM EDTA, and 1 x protease inhibitor mixture (1 mM phenylmethylsulfonyl fluoride, 5 pg/ml leupeptin, 5 pg/ml aprotinin, 1 pg/ml pepstatin, and 1 mM benzamidine)) a t 4 "C, and fractionated according to previously characterized procedures (6,28,30). Briefly, adipocytes were homogenized in a motor-driven Teflon/glass homogenizer and centrifuged at 16,000 x g for 20 min in a Beckman JA-17 rotor. The subcellular fractions, high density microsomes (HDM), low density microsomes (LDM), plasma membranes (PM), and cytosol were then obtained either following the conventional procedure or by a modified protocol adapted in our laboratory for more rapid fractionation of smaller amounts of cells. In the latter case, HDM and LDM were obtained by centrifuging the 16,000 x g supernatant for 5 min at 36,000 x g (HDM) and then for 24 min a t 200,000 x g (LDM) in a Beckman TLA 100.3 rotor. The PM fraction was collected from the interface of a 1.12 M sucrose cushion following centrifugation of the 16,000 x g pellet a t 70,000 x g for 10 min. PM were then resuspended in homogenization buffer and pelleted a t 200,000 x g for 4 min. All particulate fractions were resuspended in 10 mM TridHCI buffer, pH 7.5, containing 1 mM EDTA and 1 x protease inhibitor mixture to 1-2 mg/ml protein (bicinchoninic protein assay kit, Pierce). The average yield of HDM, LDM, PM, and cytosol from a 100-mm dish (6.7 x 10' 3T3-Ll adipocytes) was 50 pg, 160 pg, 450 pg, and 1.6 mg, respectively. Aliquots of the fractions were analyzed for GDI-1, GDI-2, or GLUT4 contents by Western blotting (triplicate gels).
In some experiments the cultured cells were fractionated in total membranes and cytosol. In this case the cells were serum-starved, washed, and homogenized as above. Total membrane preparations and cytosol were obtained by sequential centrifugations: 5 min at 800 x g and 15 min a t 200,000 x g in a Beckman TL-100 Ultracentrifuge. Pellets from the second spin were resuspended as above to 1 mg/ml protein concentration. Aliquots of the total membrane and cytosolic fractions were analyzed for GDI-1 and GDI-2 contents by Western blotting (duplicate gels).
Zmmunopurification of GDI-2-and GLUT4-containing Vesicles-Immunopurification of GDI-2-and GLUT4-containing vesicles was performed with purified anti-GDI-2 (R3362) or preimmune immunoglobulins, and with purified anti-GLUT4 (R1288) or nonspecific rabbit immunoglobulins (Pierce). LDM were resuspended to a protein concentration of 1-3 mg/ml in PBS, containing 1 x protease inhibitor mixture. LDM, typically 800 pg for GDI-2-containing vesicle immunopurification and 200 pg for GLUM, were then incubated separately with each of the specific and nonspecific antibodies, preadsorbed on protein A-Sepharose CL4B (Pharmacia Biotech Inc.). Following immunoadsorbtion (16 h at 4 "C), the pellets were separated from the supernatants by gravitation, then washed twice with PBS containing 0.4% bovine serum albumin and three times with PBS only. Both pellets and supernatants were treated with Tris/HCI buffer, pH 8.3 (final concentration 20 mid, containing 3% SDS and 3 mM dithiothreitol for 30 min at 55 "C. The samples were then alkylated in the presence of 25 mM iodoacetic acid for 1 h a t 55 "C (31), and the proteins were analyzed by SDS-PAGE and Western blotting.  (30 pg; M ) , fractionated from the indicated cells, were resolved by SDS-PAGE (duplicate gels). The proteins were transferred to nitrocellulose membranes and immunoblotted with anti-GDI-1 or anti-GDI-antisera (1:20,000 dilution each) as indicated. 4-6fold more total membranes than cytosol were used when normalized per cell number. A representative experiment from two to four independent fractionations of each cell line is shown.
Zmmunoblotting-For Western blot analysis, proteins (20-100 pg) solubilized in Lammli sample buffer (32) were separated on 10.5% SDS-PAGE. After transfer onto nitrocellulose filters, the blots were saturated with blocking buffer under previously specified conditions (33) and blotted (16 h at 4 "C) with the different antibodies. After washings the bound antibodies were detected either with '2511-Protein A (DuPont NEN; 1 h at 25 "C, 0.2 pCilm1) and autoradiography (Kodak X-Omat A R ) or with horseradish peroxidase-bound anti-rabbit immunoglobulins and enhanced chemiluminescence (ECL, Amersham Corp.). Immunoreactivity was quantified by cutting rectangular strips, corresponding to the band of interest, and analyzing their radioactivity (Beckman 5500)~-counting system. Alternatively the intensity of the visualized protein bands was evaluated by densitometry (Ultrascan XL densitometer, LKB). Two exposures of each blot were quantified to ensure that the exposures were within the linear range of the film.
For the double staining, anti-P-COP or anti-GLUT4 monoclonal antibodies were added. The cells were washed and exposed (30 min at 25 "C) to fluorescein isothiocyanate-coupled goat anti-rabbit IgG, for immunodetection of anti-GDIs antibodies, and to rhodamine-coupled goat antimouse IgG, for immunodetection of anti-P-COP or anti-GLUT4 antibodies. The cells were thoroughly washed, first in PBS, 1% FBS, 0.5% Triton X-100, then only in PBS, and fixed in 4% formaldehyde. The coverslips were mounted on slides in 90% glycerol + 2.5% DABCO.
Immunofluorescence analysis was performed with a digital imaging microscope (Nikon Diaphot 200), using Nikon Apo 60/1.4 immersion lens. Fig. 1 shows the relative amounts of immunoreactive GDI-1 and GDI-2 proteins in the cytosolic uersus total membrane fractions of several cultured cell types from various species. GDI-1 appears to reside virtually exclusively in the soluble fractions derived from COS-1 cells, CHO-K1 cells, and 3T3-Ll adipocytes. Only 0.05-0.1% of the total cellular GDI-1 appears to be associated with total membranes. Surprisingly, GDI-2 content in total membranes of these cell types greatly exceeds that of GDI-1 and accounts for 5-8% of the total cellular GDI-2 (Fig. 1). When equivalent amounts of total membrane and cytosolic protein from the different cell types were compared, the ratios of GDI in membranes:GDI in cytosol were calculated to be between 1:19 and 1:30 for GDI-1 and between 1:2 and 1:3 for GDI-2. Using quantitative immunoblotting with recombinant GST-GDI-1 and GST-GDI-2 fusion proteins as standards (not shown), we also estimated the GDI protein concentrations in 3T3-Ll adipocyte cytosol to be 0.5 and 0.35 pg/lOO pg of cytosolic protein (-5 x lo5 3T3-Ll adipocytes) for GDI-1 and GDI-2, respectively.

RESULTS AND DISCUSSION
Fractionation of insulin-sensitive 3T3-Ll adipocytes was performed to further characterize the association of GDI-2 with cellular membranes. As shown in Fig. 2, the LDM fraction, composed of intracellular membranes containing most of the cellular GLUT4 (28,30), contained relatively high levels of GDI-2 protein. Significant levels of GDI-2 could also be detected in PM (plasma membrane-enriched) and HDM fractions (Golgi-and endoplasmic reticulum-enriched). When equivalent amounts of protein from the four cellular fractions were analyzed, 25-30% of the total immunoreactive GDI-2 detected in the samples was found in the LDM. In contrast, GDI-1 was barely detectable in the LDM and HDM fractions and was not detectable in PM (Fig. 2). The LDM-bound GDI-2 could be released from the membranes by high salt treatment (0.8 M NaCl or 0.6 M KCl), consistent with the concept that it is a peripheral membrane protein (not shown).
The apparent differential cellular localizations of GDI-1 uersus GDI-2 observed in Figs. 1 and 2 were confirmed by immunofluorescence microscopy of 3T3-Ll adipocytes (Fig. 3). Immunoreactive GDI-1 displayed a diffuse pattern of staining (Fig.  3A) throughout the cytoplasm of 3T3-Ll adipocytes, similar to that observed by Ullrich et al. (20) in Madin-Darby canine kidney cells. In contrast, GDI-2 exhibited a vesicular pattern that was concentrated in the perinuclear region of these cells in addition to a cytoplasmic distribution (Fig. 3, C and E). This perinuclear GDI-2 distribution coincided with that observed for PCOP (Fig. 3, B and D), a coat protein associated with the Golgi apparatus (34). Furthermore, the perinuclear localization of GDI-2 also coincided with the fluorescence associated with immunoreactive GLUT4 in 3T3-Ll adipocytes (Fig. 3, E and F).
These data are consistent with the concept that significantly higher levels of GDI-2 are associated with perinuclear membrane compartments compared to GDI-1.
The apparent colocalization of GDI-2 and GLUT4 in the perinuclear region of 3T3-Ll adipocytes observed in Fig. 3 suggests the possibility that these proteins might reside in the same membrane structures. We tested this hypothesis by immunoadsorption of isolated LDM preparations with either anti-GDI-2 or anti-GLUT4 antibodies, and we probed for coimmunoadsorption of the GLUT4 or GDI-2 proteins, respectively. Fig.   4A shows that vesicles from the LDM fraction immunoisolated with anti-GDI-2 immunoglobulins contained significantly higher immunoreactive GLUT4 levels than vesicles immunoadsorbed with preimmune immunoglobulins. Nonetheless, most of the GLUT4 (-98%) remained in the vesicles that were not immunoisolated with anti-GDI-2 (not shown). Conversely, when anti-GLUT4 immunoglobulins were used to immunoisolate GLUT4-containing vesicles, a significant depletion (22 2 8% x f. S.E., three experiments) of GDI-2 in the remaining vesicles was observed (Fig. 4B) concomitant with depletion of GLUTccontaining vesicles (Fig. 4C). Immunoreactive GDI-2 could not be analyzed in the vesicles isolated with anti-GLUT4 antibodies due to the interference from immunoglobulin heavy chains that migrate in the same region of the electrophoresis gel. Taken together, these data indicate that a significant portion of the LDM-associated GDI-2 resides in GLUTCenriched membranes from insulin-sensitive 3T3-Ll adipocytes. We next tested whether insulin action might modulate the cellular localization of GDI proteins in 3T3-Ll adipocytes. The of GDI-1 and GDI-2 PM, LDM, and soluble fractions derived from cells incubated with or without insulin for up to 30 min, were subjected to SDS-PAGE and immunoblotting with anti-GDI or anti-GLUT4 antisera. As expected, immunoreactive GLUT4 increased in the PM fraction and decreased in the LDM in response to insulin over this time course (Fig. 5B). Immunoreactive GDI-2 in the LDM fraction also rapidly decreased with insulin treatment of the cultured adipocytes, while no change was detected in GDI-2 content of PM (Fig. 5A). No increase in the cytosolic levels of GDI-2 in response to insulin could be observed (not shown). However, since only about 5% of the cellular GDI-2 resides in the LDM with over 90% in the cytosol, its movement to the cytosol would not result in more than a few percent increase in this fraction. The GDI-1 protein levels in cytosol and LDM showed no change in response to insulin (not shown). These experiments demonstrate that insulin decreases selectively the association of GDI-2 with an intracellular membrane fraction of 3T3-Ll adipocytes.
The ability of insulin to modulate the GDI-2 content of 3T3-Ll adipocyte LDM suggests a role for this protein in insulin-regulated membrane cycling. This hypothesis is consistent with the recent demonstration that Rab4 is similarly decreased in the LDM fraction of fat cells by insulin (27). GDI-2 can release Rab4 from membranes in vitro (21), suggesting important interactions of these proteins in intact cells. Recent reports suggest that Rab proteins associate with specific membrane compartments by a mechanism that involves binding of GDI.Rab complexes to membrane targets for such complexes (35, 36). The higher membrane content of GDI-2 compared to GDI-1 reported here may relate to this function of GDI proteins to present Rab proteins to membranes. Whether this hypothe-sis is correct requires further testing, as does the possibility that insulin action might regulate this or other function of GDI-2. In any case, the present results demonstrate a striking difference between GDI-1 and GDI-2 cellular localizations and a specific regulation of the GDI-2 protein by insulin.