Evidence that functional erythrocyte-type glucose transporters are oligomers.

In this study we tested the hypothesis that functional erythrocyte-type glucose transporters (GLUT1) exist as oligomeric complexes by expressing chimeric transporter proteins in Chinese hamster ovary cells harboring endogenous GLUT1 transporters. The chimeric transporters were GLUT1-4c, in which the 29 C-terminal residues of human GLUT1 were replaced by the 30 C-terminal residues of rat skeletal muscle glucose transporter (GLUT4), and GLUT1n-4, containing the N-terminal 199 residues of GLUT1 and the 294 C-terminal residues of GLUT4. Endogenous GLUT1 was quantitatively co-immunoprecipitated by using an anti-GLUT4 C-terminal peptide antibody from detergent extracts of Chinese hamster ovary cells expressing either of the chimeric proteins, as detected by immunoblotting the precipitates with an anti-GLUT1 C-terminal peptide antiserum. No co-immunoprecipitation of native GLUT1 with native GLUT4 from extracts of 3T3-L1 adipocytes, which contain both these transporters, was observed with the same antibody. These data are consistent with the hypothesis that GLUT1 transporters exist as homodimers or higher order oligomers and that a major determinant of oligomerization is located within the first 199 residues of GLUT1.

The facilitated diffusion of glucose in mammalian cells is mediated by a family of integral membrane glycoproteins composed of a t least five isoforms identified by cDNA cloning (1)(2)(3)(4)(5)(6)(7)(8)(9). All glucose transporters share deduced primary structures that predict 12 transmembrane domains and cytoplasmic locations for the N-terminal and C-terminal segments as well as for a large hydrophilic loop connecting the sixth and seventh transmembrane regions (1)(2)(3)(4)(5)(6)(7)(8)(9). Little is known about the molecular mechanism by which D-glucose transport occurs, due in part to a lack of detailed information describing the three-dimensional structures of glucose transporter proteins. Previous studies on the size of the native human erythrocyte glucose transporter, measured by irradiation inactiva-* This work was supported by National Institute of Health Grants DK 30898 (to M. P. C.) and DK 36081 (to A. C.). 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.
Recipient of a postdoctoral fellowship from the Juvenile Diabetes Foundation International.

07302.
Y Supported by National Institutes of Health Training Grant DK tion of either cytochalasin B binding (10,12) or D-glucose flux (11) and by freeze-fracture electron microscopy (13), suggested molecular sizes compatible with either a dimer (12,13) or a tetramer (10,11). However, these findings were also consistent with the possibility that the glucose carrier is assembled with one or more heterologous proteins. Such physical associations with the erythroid glucose transporter have been recently proposed for hexokinase (14) and glyceraldehyde-3-phosphate dehydrogenase (15).
The aim of the present studies was to test directly whether multiple erythrocyte-type glucose transporters (GLUTl)' are present in oligomeric complexes in cell membrane preparations. We engineered cDNA constructs encoding two chimeric transporter proteins which contain N-terminal regions of GLUTl and C-terminal regions of the adipocyte-type transporter (GLUT4) and expressed these proteins in stable CHO cell lines. Immunoprecipitation of these chimeric transporters from membrane detergent extracts with an anti-GLUT4 Cterminal peptide antibody also precipitated native GLUTl molecules, endogenously expressed in CHO fibroblasts. These data strongly indicate that GLUTl transporter proteins exist as homooligomeric structures.

EXPERIMENTAL PROCEDURES
Production of the Chimeric Constructs-GLUT1-4c chimera is composed of the coding region for the first 463 codons of the human GLUTl cDNA linked to the coding region for the last 30 codons of the rat GLUT4 cDNA. To obtain a compatible restriction fragment containing the first 463 codons of GLUT1, a non-unique Sau3Al site a t position 1565 on the cDNA from the plasmid pGem3Z-GLUT1 (16) was mutated to a unique BglII site by site-directed mutagenesis. The chimeric cDNA was generated by ligating a 1565-bp (BamHI-BglII) fragment from GLUTl cDNA to a 3' 763-bp (BglII-BamHI) fragment of the GLUT4 cDNA at the BglII site. GLUTln-4 chimera has the first 215 codons of rat GLUT4 cDNA replaced by the first 199 codons of GLUTl cDNA. The chimeric cDNA was assembled by ligating a 799-bp BamHI-PstI fragment from GLUTl cDNA to a 1549-bp PstI-BamHI fragment spanning codons 216-509 of GLUT4 cDNA, including 670 bp of its 3"untranslated region. Both chimeras were initially assembled in the bacterial plasmid vector pUC18. The identities of the chimeric cDNAs were verified by restriction analysis and DNA sequencing. ln-4, each expressing the corresponding type of chimera, were chosen Cell Culture Conditions-CHO-K1 fibroblasts and 3T3-LI preadipocytes were obtained from American Type Culture Collection. CHO-K 1 control fibroblasts, and CHO 1-4c and CHO ln-4 cells were cultured in Ham's F-12 medium containing 10% fetal calf serum. 3T3-Ll cells were grown in DMEM containing 10% calf serum. Two days after confluence, 3T3-Ll fibroblasts were induced to differentiate basically as described in Ref. 17. Briefly, cells were incubated in DMEM supplemented with 10% fetal calf serum, 5 pg/ml insulin, 0.25 mM dexamethasone, and 0.5 mM 3-isobutyl-I-methylxanthine. After 2 days, the medium was replaced by the same lacking dexamethasone and 3-isobutyl-l-methylxanthine, and the incubation was prolonged for another 2 days. Cells were then cultured in DMEM containing 10% fetal calf serum and used from day 8 to day 15 after the beginning of the differentiation.
["5ZlZAPS-Forskolin Labeling of Total Membranes-Total membranes from CHO-K1, CHO 1-4c, and CHO In-4 cells were subjected to photoaffinity labeling with [12"I]IAPS-forskolin, essentially as described by Wardzinski et al. (18). Briefly, cells were grown to confluence in 150-mm dishes and then harvested, homogenized in 20 mM Hepes, pH 7. with a 1000-watt UV lamp. Subsequently, 2-mercaptoethanol was added to a final concentration of 1%. Membranes were then collected by ulracentrifugation at 200,000 X g for 90 min, solubilized in sample buffer, and subjected to SDS-polyacrylamide gel electrophoresis (10% resolving gel). Gels were dried under vacuum and autoradiographed by using Kodak X-OMAT AR films, and the intensity of the bands was determined by laser densitometric scanning. Assay of 2-Deoxyglucose Uptake-CHO-K1, CHO 1-4c, and CHO 111-4 cells were assayed for 2-deoxyglucose uptake as described in Ref.

16.
Zmmunoblotting of Total Cell Membranes-Immunoblotting of total membranes from CHO 1-4c, CHO In-4, and parental cells to detect endogenous GLUTl was performed by using a 1:lOOO dilution of R-480 antibody, an anti-GLUT1 C-terminal antiserum directed against residues 475-492 of human GLUTl (17). Membrane proteins (100 pg for each sample) were resolved by SDS-polyacrylamide gel electrophoresis (10% gel) and then electophoretically transferred to a nitrocellulose filter a t 200 mA for 3 h. Filters were incubated with R-480 antiserum (2 h at 23 "C), and antibody binding was detected by [Iz5I] Protein A (Du Pont-New England Nuclear). Autoradiography of the filters and quantification of the intensity of the bands were performed as described previously.
Immunoprecipitation with Anti-GLUT4 C-terminal Peptide Antiserum-Total membranes from CHO 1-4c, CHO 111-4 cells (100 pg for each sample), and 3T3-Ll adipocytes (200 pg for each sample) were solubilized in a buffer containing 50 mM Hepes, pH 7.4,150 mM NaC1, 2% Nonidet P-40, and 1 mM phenylmethylsulfonyl fluoride (0.3 mg/ml final protein concentration) for 30 min at either 0 or 23 "C and subjected to centrifugation at 100,000 X g for 1 h at 4 "C. The supernatant and the insoluble pellet were assayed for glucose transporter immunoreactive proteins by immunoblot. In these lysis conditions more than 95% of glucose transporters were recovered in the supernatant, which was therefore used as the source of immunoprecipitable transporters. The anti-GLUT4 C-terminal peptide antiserum R-1288 (19) was added to aliquots of the lysate a t 1:250dilution. Nonspeciflc immunoprecipitation was determined by addition of the same dilution of antiserum to lysis buffer only, in order to obtain a quantitative estimate of the nonspecific background due to immunoglobulin heavy chains that are located in the same range of molecular weight as the transporters.
Analogous experiments performed by adding rabbit non-immune serum to membrane extracts as a negative control showed identical results.
After incubation with the antiserum (6-14 h a t 4 "C), Protein A-Sepharose CL-4B (20 p1 of packed resin for each sample) was added for an additional 2 h at 4 "C. The beads were washed twice with lysis buffer and twice with 50 mM Hepes, p H 7.4. The pellets were then resuspended in 20 mM Tris, pH 8.3, 10% SDS, 10 mM dithiothreitol, incubated for 30 min a t 23 "C, and finally alkylated in the presence of 25 mM iodoacetic acid for another 30 min a t 23 "C. After reductive alkylation, sample buffer was added, each sample was divided in two equal aliquots, and SDS-polyacrylamide gel electrophoresis in duplicate 10% gels was performed. Immunoblotting was performed as described above, and immunoreactive glucose transporters were detected by incubation of filters with 1:lOOO dilutions of either R-480 or R-1288 antisera, followed by exposure to ['251]Protein A and autoradiography. The yield of the immunoprecipitation was estimated by laser densitometric scanning of the bands corresponding to immunoprecipitated glucose transporters compared with the bands resulting from the same amount of membrane proteins in the absence of immunoprecipitation. The total membrane samples were subjected to solubilization in 10% SDS and to reductive alkylation as described above prior to gel electrophoresis.

RESULTS AND DISCUSSION
A schematic representation of the chimeric transporters constructed for this study is shown in Fig. la. According to the predicted bidimensional structure of glucose transporters, the GLUT4 partial sequence in GLUT1-4c molecule is entirely located in the cytoplasmic domain. The switch from GLUTl to GLUT4 sequence in the GLUTln-4 chimera oc- curs in the predicted sixth transmembrane region, providing this construct with the middle loop of GLUT4. Fig. l b shows that stable CHO cell lines transfected with cDNAs encoding either of these chimeric proteins exhibit 5-6-fold increases in rates of 2-deoxyglucose uptake compared with the parental CHO-K1 cells. Similar increases in 2-deoxyglucose transport rates were observed in previous studies in which CHO cells were transfected with cDNA encoding native human GLUTl (16). A 10-and 13-fold increase in [12sI]IAPS-forskolin photolabeling of total membranes from CHO cell lines expressing the GLUTln-4 and GLUT1-4c proteins, respectively, was observed (Fig. lb). This reagent (18) effectively labels both endogenous GLUTl and the expressed chimeric transporters, thereby providing an estimate of total transporter proteins present in these cells. The amounts of endogenous GLUTl are similar in all transfected and parental cell lines, as assessed by immunoblotting total membranes with an anti-GLUTl C-terminal peptide antibody (R-480), raised against amino acids 475-492 of GLUTl (17) (Fig. lb). It is interesting to note that stable CHO cell lines transfected with the rat GLUT4 cDNA expressed GLUT4 protein, as detected by immunoblot (not illustrated). However, no elevations of 2deoxyglucose uptake rates or [1251]IAPS-forskolin labeling were observed, indicating very low GLUT4 expression in these transfected cells. The results described above indicate that both chimeric transporters can be expressed at high levels in CHO cells. Furthermore, the chimeric carriers are capable of catalyzing glucose transport when expressed in CHO cells.
In order to assay whether endogenous GLUTl and the expressed chimeras in CHO cells are physically associated, chimeric transporters in Nonidet P-40 extracts of total membranes from both CHO 1-4c and CHO ln-4 cells were immunoprecipitated with an anti-C-terminal peptide antibody (R-1288) directed against amino acids 498-509 of rat GLUT4 (19). The immunoprecipitates were resolved by electrophoresis, and transporter proteins were detected by immunoblot analysis (Fig. 2) using the R-1288 antiserum, which recognizes both GLUT1-4c (lanes 1-3) and GLUTln-4 transporters (lanes 7-9), as well as using the anti-GLUT1 C-terminal peptide antibody R-480 (lanes 4-6 and 10-12). Both antibodies are isoform-specific in their ability to immunoprecipitate and immunoblot the respective native GLUTl and GLUT4 transporters. In particular, R-1288 antiserum fails to either immunoprecipitate or immunoblot GLUTl from total membrane extracts from parental CHO-K1 cells (not shown). Immunoprecipitation performed with R-1288 antiserum provided a quantitative recovery of both GLUT1-4c and GLUTln-4 transporter proteins (lunes 2 and 8, respectively), compared with the total amount of chimeric transporters present in the membrane extracts (lunes 1 and 7, respectively). Nonspecific immunoprecipitation was assessed as described in detail under "Experimental Procedures' ' (lanes 3, 6, 9, and  12). Importantly, analysis of these immunoprecipitates by immunoblotting with the anti-GLUT1 C-terminal antiserum R-480 also showed quantitative recovery of endogenous GLUTl (lunes 5 and 11 compared with lunes 4 and 10, respectively). Therefore, GLUTl was entirely co-immunoprecipitated with the chimeric molecules. These data indicate that nearly all native GLUTl from CHO 1-4c and CHO ln-4 cells is physically associated with the chimeric transporters in the detergent extracts. Immunoprecipitations with the anti-GLUTl C-terminal antiserum R-480 were not performed due to low efficiency of GLUTl recovery using this antibody.
In an effort to evaluate the specificity of the co-immunoprecipitation of GLUTl with chimeric transporters, similar experimental protocols were performed with membrane ex-  6 ) and CHO ln-4 cells (lanes 7-12) were either directly subjected to electrophoresis (lunes I , 4, 7, and IO) or immunoprecipitated with R-1288 antiserum, an antipeptide antibody raised against residues 498-509 of rat GLUT4 (7) (lanes 2, 5, 8, and 11). Lanes 3, 6, 9, and 12 show the immunoprecipitation performed in the presence of lysis buffer only. The bands corresponding to the glucose transporters are indicated by arrows. The band of apparent molecular mass 75,000, better visible in lanes 8, 9, 11,  Immunoblol: Anti-GLUT4 Anti-GLUT1 -1288 (lanes 1-3) and R-480 antisera (lanes 4-6) were used to perform immunoblotting, as described in the legend of Fig. 2. Total membranes from 3T3-Ll adipocytes (200 pg for each sample) were either directly resolved by electrophoresis (lanes I and 4 ) or immunoprecipitated with R-1288 antiserum (lunes 2 and 5 ) . Lanes 3 and 6 show the immunoprecipitation performed with R-1288 in the presence of lysis buffer only. The same experiment was performed three times on independent membrane preparations with no change in the results. tracts from 3T3-Ll adipocytes. These cells express both native GLUTl and native GLUT4 transporter isoforms in a ratio of about 3:l (20). GLUT4 was quantitatively immunoprecipitated from total membrane extracts of 3T3-Ll adipocytes using the R-1288 anti-GLUT4 C-terminal peptide antibody (Fig. 3, lune 2). However, no GLUTl could be detected in the same immunoprecipitate when probed with the R-480 antiserum (Fig. 3, lune 5 ) . Similar results were obtained when detergent extracts of either plasma membranes or low density microsomes from 3T3-Ll adipocytes were studied in this manner (not illustrated). These results, demonstrating that GLUTl and GLUT4 fail to co-immunoprecipitate, are consistent with the previous findings of Calderhead et ul. (20), using 3T3-Ll adipocytes, and Zorzano et al. (21), using primary rat fat cells. It is noteworthy that the latter cell type expresses at least 10-fold more GLUT4 transporter protein than GLUT1. If hetero-oligomeric complexes of GLUTl and GLUT4 occurred in these extracts, quantitative precipitation of GLUTl should have been observed. The weight of the results presented here and those previously published indicate that GLUTl molecules are not physically associated with GLUT4 transporters under the conditions of these experiments. Furthermore, the absence of co-precipitation of GLUTl and GLUT4 in rat adipocytes indicates that the 10fold greater expression of GLUT4 alone is not sufficient to drive a physical association of this isoform with GLUTl (21). Therefore, it is unlikely that the expression of chimeric transporters in CHO cells to a level severalfold greater than that of endogenous GLUTl causes an artifactual oligomerization of the two molecules. Moreover, we observed co-precipitation of endogenous GLUTl in cells where the expression of the chimeric molecules was only 3-4-fold over the parental level (not shown).

FIG. 3. Immunoblot analysis of immunoprecipitates obtained with anti-GLUT4 C-terminal peptide antiserum from membrane extracts of 3T3-Ll adipocytes. R
In order to better compare the results of Figs. 2 and 3, the relative amounts of GLUTl and GLUT4 immunoreactivity in CHO 1-4c cells and in 3T3-Ll adipocytes were determined. The epitope recognized by R-1288 is conserved between mouse GLUT4 and the chimeric molecule. Since the hamster GLUTl has not been cloned to date, we assumed that R-480 antibody would bind GLUTl molecules from the two species with comparable affinity, given the high degree of sequence homology in this isoform from different species. Nevertheless, GLUTl quantitation must be considered approximate. Based on these considerations, the amount of GLUT4 immunoreactivity in CHO 1-4c cells was found to be only 1.8-fold greater than in 3T3-Ll adipocytes, when normalized per cell number. On the other hand, GLUTl immunoreactive protein is approximately 4.5-fold greater in 3T3-Ll adipocytes than in CHO 1-4c fibroblasts (not shown). These observations combined with the report that the former cell line expresses more GLUTl than GLUT4 (20) imply that the total amount of glucose transporter proteins harbored by chimera-expressing CHO cells is significantly lower than that expressed by 3T3-L1 adipocytes. These results further argue against the possibility of artificial glucose transporter physical association driven by elevated expression of chimeric transporters.
The results in Figs. 2 and 3 suggest that co-immunoprecipitation of GLUTl transporter with GLUT1-4c and GLUTln-4 proteins is due to specific interaction and not to random association mediated by their hydrophobic properties. Moreover, it would be difficult to attribute the transporter protein association described here to an artifact due to solubilization. Nonidet P-40 is a nonionic detergent with a hydrophilelipophile balance value in the lipophilic range. This type of detergent appears to provide a hydrophobic environment most closely matched to that of the membrane, thereby maintaining membrane proteins in their native conformation (22). Proteins that appear to exist as monomers in the membrane are found as monomers following detergent solubilization (see, for example, rhodopsin (23)). Membrane proteins existing as oligomers in situ are found as oligomers in detergents (see, for example, Ca*+-ATPase (24) and the erythrocyte anion transporter (25)). It is noteworthy that the co-immunoprecipitation of GLUTl with GLUT1-4c chimera was still observed when CHO 1-4c total membranes were solubilized in radioimmune precipitation buffer containing 1% Nonidet P-40,0.5% sodium deoxycholate, and 0.1% sodium dodecyl sulfate (not illustrated). Furthermore, we performed experiments where membranes from CHO 1-4c cells were mixed with membranes from CHO GT3 fibroblasts, a cell line overexpressing human GLUTl (16) prior to solubilization at 0 "C in 2% Nonidet P-40 and immunoprecipitation with R-1288 antiserum. GLUTl immunoreactivity expressed by CHO GT3 cells is 17-20-fold greater than that expressed by parental cells (16). If the physical association of GLUTl and the chimeric transporters occurred in the detergent solution, the amount of GLUTl immunoreactive protein co-precipitating with GLUT1-4c molecules should be much greater from the mixed membrane lysate than from a lysate containing CHO 1-4c membranes alone. We found that the same amount of GLUTl immunoreactive protein coprecipitates with the chimeric molecules under both conditions, indicating that glucose transporter physical association is present prior to membrane solubilization (results not illustrated).
The finding that GLUTl does not associate with GLUT4 in these experiments indicates that GLUTl oligomerization involves sequences intrinsic to the GLUTl primary structure not present in GLUT4. Since GLUTl co-immunoprecipitates with the GLUTln-4 chimera, the site of interaction between the two molecules (or between the two molecules and a binding protein) must be located within the first 199 residues of GLUT1. Interestingly, within this GLUTl segment, a "leucine zipper-like'' motif is present (26). This structure is clearly a candidate for driving the association between glucose transporter molecules. However, this sequence is partially conserved in GLUT4 and thus may not be the only sequence required for interaction. The potential involvement of this "leucine zipper-like'' motif suggests the hypothesis of a dimeric transporter structure, since leucine repeats have been shown to play a major role in dimerization of several DNAbinding proteins (27).
The data in Fig. lb, showing that expressed chimeric glucose transporters exhibit catalytic activity, combined with the evidence that they are oligomeric (Fig. 2) indicate that transporter oligomers are functional. A multimeric assembly of glucose carrier proteins could provide a rational explanation for the observed catalytic properties of the human erythrocyte sugar transport system. The human erythrocyte sugar carrier exposes both influx and efflux sites to the substrate simultaneously (28,29). When these sites are occupied by nontransported, competitive inhibitors of transport, they interact with negative cooperativity (28,29). Cooperative interactions are frequently observed in multimeric enzymes and ligandbinding proteins, where binding of a ligand to one subunit of the complex affects the affinity of the remaining subunitb) for substrate (30, 31). Sugar transport by most cells is consistent with a homodimer model in which sugar influx and efflux sites are arranged in an antiparallel fashion and translocation of substrate by one subunit promotes the complementary translocation of sugar by the second subunit (32). A recent study performed by hydrodynamic techniques showed that purified GLUTl from human erythrocytes exists as a mixture of dimers and tetramers in a cholate solution (33). In addition, the ligand binding properties of this molecule were found to be dependent upon GLUTl oligomeric structure (33). The finding of homooligomeric GLUTl by an independent but complementary method strongly reinforces the major conclusion of this study. Further work will be necessary to