GLUT1 Transmembrane Glucose Pathway AFFINITY LABELING WITH A TRANSPORTABLE D -GLUCOSE DIAZIRINE*

We synthesized a transportable diazirine derivative of D -glucose, 3-deoxy-3,3-azi- D -glucopyranose (3-DAG), and studied its interaction with purified human erythrocyte facilitative glucose transporter, GLUT1. 3-DAG was rap- idly transported into human erythrocytes and their resealed ghosts in the dark via a mercuric chloride-inhib- itable mechanism and with a speed comparable with that of 3- O -methyl- D -glucose (3-OMG). The rate of 3-DAG transport in resealed ghosts was a saturable function of 3-DAG concentration with an apparent K m of 3.2 m M and the V max of 3.2 (cid:109) mol/s/ml. D -Glucose inhibited the 3-DAG flux competitively with an apparent K I of 11 m M . Cytochalasin B inhibited this 3-DAG flux in a dose-depend- ent manner with an estimated K I of 2.4 (cid:51) 10 (cid:50) 7 M . Cytochalasin E had no effect. These findings clearly establish that 3-DAG is a good substrate of GLUT1. UV irradiation of purified GLUT1 in liposomes in the pres- ence of 3-DAG produced a significant covalent incorporation of 3-DAG into GLUT1, and 200 m M D -glucose abol- ished this 3-DAG incorporation. Analyses of trypsin and endoproteinase Lys-C digestion of 3-DAG-photolabeled GLUT1 revealed that the cleavage SDS on a TSK 3000 column (21). Peptides produced by this digestion were separated electrophoretically using Tricine/SDS-poly-acrylamide gel according to Schagger and Von Jagow (25). Extensive pepsin digestion was carried out using 3 mg of pepsin/ml (mass ratio of GLUT1 to pepsin was 1:2) and after sonication for 30 s followed by a 30-min incubation at 37 °C with shaking. This digestion is known to digest all the nonmembrane portion of the protein (26). Other Analytical Methods— Protein was determined by the method of Lowry et al. (27). Separation of proteins and peptides by electrophoresis was carried out using SDS-polyacrylamide or Tricine gels as described elsewhere (26). Radioactivities of [ 3 H]3-OMG and [ 3 H]NaBH 4 were measured using a liquid scintillation counter (LKB Rackbeta, Pharma-cia Biotech Inc.). Partial amino acid and N-terminal sequence analysis of peptide fragments was performed by Dr. Audree V. Fowler (UCLA School of Medicine).

A family of structurally related intrinsic membrane proteins known as facilitative glucose transporters catalyzes the movement of glucose and other selected sugars across the plasma membrane diffusion barrier in mammalian cells (1). Six isoforms have been identified in this family (2); these include GLUT1 or erythrocyte type (3), GLUT2 or liver/pancreatic beta cell type (4), and GLUT4 or muscle/adipose cell type (5). Hydropathy analyses of cDNA-deduced amino acid sequences together with the biochemical information obtained from purified GLUT1 (1)(2)(3) suggest that this protein family shows the common transmembrane topology composed of a highly conserved, large (amounting to approximately 50% of protein mass) transmembrane domain, with less conserved, grossly asymmetric, cytoplasmic and exoplasmic (nonmembrane) domains. Evidence (1, 3, 6 -8) further suggests that the transmembrane domain is made of 12 transmembrane ␣-helices (TMHs, 1 Refs. [1][2][3][4][5][6][7][8][9][10][11][12] and accommodates a water-filled channel through which glucose and other substrates move (glucose pathway). The cytoplasmic domain includes a short N-terminal segment, a large cytosolic loop between TMHs 6 and 7 (central loop), and a large C-terminal segment. The exoplasmic domain includes a large loop between TMHs 1 and 2 bearing a single N-linked oligosaccharide moiety. The fact that isoform-specific amino acid sequences are found almost exclusively at the cytoplasmic and exoplasmic domains suggests that they are responsible for the tissue-specific regulation of the transporter functions. The conserved transmembrane domain primary structure, on the other hand, suggests that the substrate channel is basically identical in its structure among isoforms of this protein family.
A number of recent observations have shown that the intrinsic activity of glucose transporters in vitro can be modulated by hormones and metabolites (9,10). Evidence also indicates that impaired glucose transporter intrinsic activity is in part responsible for insulin resistance seen in human diabetes and obesity (11). The identification of possible molecular defects for the reduced glucose transporter intrinsic activity in these diseased states would be facilitated once we understand how transporters work at the molecular level. Detailed information on the tertiary structure of the transporter, particularly that of the putative glucose channel, would be an essential first step toward such an understanding.
Information currently available on the transmembrane domain structure of GLUT1, the best studied isoform in this protein family, is minimal. Circular dichroism studies with purified GLUT1 reconstituted in liposomes (6) have revealed that more than 80% of the protein mass is in the ␣-helical structure. Since the transmembrane domain accounts for not more than 50% of the protein mass, this finding strongly suggests that the 12 transmembrane segments are largely if not totally ␣-helices. For the putative glucose channel structure, Mueckler et al. (3) have first noted the possible significance of the presence of amphipathic TMHs, which may line an aqueous channel for glucose movement and provide hydroxyl and amide hydrogens for hydrogen bond formation with glucose. Conspicuously lacking in the current literature is an effort to identify such a channel residue or residues that interacts with glucose or other substrates (substrate binding sites) in this protein.
Based on these findings, Holman and his co-workers (15) have proposed a model for GLUT1 structure suggesting that the Nand C-terminal halves are two separate domains and that only the C-terminal half is directly involved in substrate binding and translocation. This model implies that the C-terminal half alone forms a putative glucose channel.
The aim of the present study is to identify experimentally the channel-forming transmembrane ␣-helices in GLUT1 by affinity labeling using a photoactive substrate analog. We synthesized a diazirine derivative of D-glucose, 3-DAG, and demonstrated that this photoreactive glucose analog is a good substrate of GLUT1. The analog was rapidly transported by intact human erythrocytes and their ghosts, which was inhibited by cytochalasin B but not by cytochalasin E. 3-DAG upon UV irradiation was covalently incorporated into purified GLUT1 in liposomes. Analyses of the trypsin and endoproteinase Lys-C cleavage products of 3-DAG-labeled GLUT1 clearly revealed that both the N-and C-terminal halves were affinitylabeled by this substrate analog. Based on these findings, we propose that the putative glucose channel is formed between the N-and C-terminal halves of the transporter rather than exclusively within the C-terminal half of the transmembrane domain.

Materials-Cytochalasins B and E were from Sigma. 3-O-[methyl-
3 H]D-glucose was purchased from Amersham Corp. [ 3 H]NaBH 4 was obtained from DuPont NEN. Trypsin and Lys-C were from Wako Chemicals (Richmond, VA). N-Glycosidase F was purchased from Boehringer Mannheim. Diacetone glucose, hydroxylamine O-sulfonic acid, molecular sieve, and 2-methoxyethyl ether were from Aldrich.
Assay for 3-DAG and 3-DAG Transport Measurement-Human erythrocytes were isolated free of white cells and cell debris from freshly drawn blood, and hemoglobin-free, resealed erythrocyte ghosts were prepared as described elsewhere (18). The rates of 3-DAG transport into erythrocytes and their resealed ghosts were measured by following the time course of 2 mM mercuric chloride-inhibitable net influx and efflux of 3-DAG, in the absence and in the presence of additives as specified in each figure legend. For the determination of the amount of 3-DAG uptake, ghosts or erythrocytes (200 l) were vigorously vortexed with chloroform (200 l) and water (200 l). The aqueous layer was separated, and the organic layer was extracted with water (200 l) once more. The combined aqueous extracts were made up to 1 ml. The glucose analog was quantitated in this solution according to Park and Johnson (19) by assaying glucose through oxidation at C-1, with a slight modification (the time of the incubation at 70°C was increased to 30 min). In experiments measuring 3-DAG transport in the presence of Dglucose, 3-DAG was measured after derivatization with 1-phenyl-3methyl-5-pyrazolone, followed by capillary zone electrophoretic separation in the presence of 100 mM borate buffer, pH 9.5, detected at 245 nm (e ϭ 29,000) according to the procedure described by Honda et al. (20). We used an uncoated silica capillary 72 cm long with a 50-m inner diameter, 50-cm effective length, at 20 kV (82 A). The electrophoresis separated sugars at 14.7 min for 3-DAG, 24.9 min for D-glucose, and 16.5 min for 3-OMG under the conditions used.
Photoaffinity Labeling of GLUT1 with 3-DAG-Purified human erythrocyte GLUT1 proteoliposomes were prepared exactly as described (21) and used for the affinity labeling. Typically, a 0.5-ml suspension of purified GLUT1 (500 g of protein) proteoliposomes in 20 mM phosphate buffer, pH 7.5, containing 3-DAG (1 or 5 mM) was introduced in a quartz cell (1 ϫ 1 cm). Photolysis was carried out in 3-10 sessions at 320 nm for 3 min for each session, with constant stirring at 20°C in the sample compartment of the fluorescence spectrophotometer (Perkin-Elmer 512). When specified, photolysis was also carried out in the presence of 200 mM 3-OMG. The suspension was centrifuged after each session (in a SW 50 rotor at 45,000 ϫ g for 30 min with Beckman L5-50) to decant unbound ligand. The pellet of the final photolysis session was washed three times with 2.5 ml of DAG-free borate buffer (100 mM, pH 9.5). Assay for Reacted 3-DAG-The photolytically incorporated 3-DAG was quantitated through reducing property of the unprotected monosaccharide using radioactive sodium borohydride, [ 3 H]NaBH 4 (22). A suspension of the labeled protein in water (0.5 ml) was made alkaline with 50 l of 0.05 N NaOH, and a solution of sodium borotritide of known specific activity (700 mCi/mmol) in freshly distilled dimethylformamide (in 5 M excess) was added. The reaction mixture was incubated at room temperature for 4 h. The reaction was stopped by adding 100 l of acetone followed by centrifugation as above. The pellet was washed three times, dissolved in 1% SDS, and counted. In a control experiment, a small amount of radioactivity became associated with protein, which was subtracted after normalization.
Separation and Isolation of TMHs by Trypsin and Endoproteinase Lys-C Digestion-Mild trypsin digestion and N-glycosidase F treatment of purified GLUT1 protein in liposomes and subsequent separation of cleavage products by SDS-PAGE were carried out as described (23). Digestion with bacterial endoproteinase Lys-C was performed in 200 mM Tris-HCl, pH 8.5, containing 0.1% SDS at 37°C for 5 h at 100 g/ml protein concentration and at 1:20 enzyme/substrate mass ratio as described by Riviere et al. (24). Prior to digestion with Lys-C, lipids and unbound ligands were removed by size exclusion chromatography in 0.1% SDS on a TSK 3000 column (21). Peptides produced by this digestion were separated electrophoretically using Tricine/SDS-polyacrylamide gel according to Schagger and Von Jagow (25). Extensive pepsin digestion was carried out using 3 mg of pepsin/ml (mass ratio of GLUT1 to pepsin was 1:2) and after sonication for 30 s followed by a 30-min incubation at 37°C with shaking. This digestion is known to digest all the nonmembrane portion of the protein (26).
Other Analytical Methods-Protein was determined by the method of Lowry et al. (27). Separation of proteins and peptides by electrophoresis was carried out using SDS-polyacrylamide or Tricine gels as described elsewhere (26). Radioactivities of [ 3 H]3-OMG and [ 3 H]NaBH 4 were measured using a liquid scintillation counter (LKB Rackbeta, Pharmacia Biotech Inc.). Partial amino acid and N-terminal sequence analysis of peptide fragments was performed by Dr. Audree V. Fowler (UCLA School of Medicine).

3-DAG Is Transported by GLUT1 in Human Erythrocytes-A
60-s time course of the mercuric chloride-sensitive net uptake of 3-DAG by human erythrocytes was followed in the dark at 20°C by measuring time-dependent uptake of 3-DAG by cells (Fig. 1). The uptake of this glucose analog in the dark by erythrocytes was very fast, being complete within 1 min. This mercuric chloride-sensitive 3-DAG uptake was completely inhibited in the presence of 10 M cytochalasin B (Fig. 1). The uptake was not affected at all in the presence of 10 M cytochalasin E (not illustrated).
Resealed human erythrocyte ghosts were suspended in 1:10 BSS (18) at a cytocrit of approximately 65% and incubated in the presence of 0.2 mM 3-DAG in the dark. The concentration of 3-DAG inside the ghosts measured after 10 min of this incubation was essentially identical to that in the medium (not illustrated). The time course of 3-DAG net exit from these preequilibrated ghosts suspended (10% cytocrit) in 1:10 BSS buffer was then followed in the dark at 20°C (Fig. 2). The 3-DAG exit from ghosts was also fast, being virtually completed within 10 s. This 3-DAG exit was sensitive to the presence of cytochalasin B; the exit rate was reduced approximately 50% by 0.2 M cytochalasin B and more than 80% by 1 M cytochalasin B but not at all significantly by 10 M cytochalasin E (Fig. 2). More systematic assessment of this inhibition of 3-DAG exit by increasing concentrations of cytochalasin B (not illustated) revealed an apparent K I (the concentration which produced 50% of the maximal inhibition) of 2.4 ϫ 10 Ϫ7 M for this inhibition.
The rate of 3-DAG transport as a function of 3-DAG concentration was studied by measuring the initial 3-s net 3-DAG uptake by intact erythrocytes suspended in medium containing varying concentrations (0.2-5.0 mM) of 3-DAG. The experiments were otherwise similar to those of Fig. 1. Analysis of the results of these initial velocity measurements indicated that the rate of 3-DAG uptake is a simple, saturable function of 3-DAG concentration (Fig. 3) with an apparent K m (the 3-DAG concentration at which the rate of uptake was 50% of maximal) of 3.2 mM and the V max (maximal rate of uptake) of 3.2 mol/ s/ml. The rate of [ 3 H]3-OMG uptake by intact erythrocytes as a function of 3-OMG concentration measured in parallel (not illustrated) revealed an apparent K m of 18 mM with the V max of 3.1 mol/s/ml.
The rate of 3-DAG uptake by erythrocytes was competitively inhibited by D-glucose (Fig. 4) but not by L-glucose (not illustrated). In these experiments, erythrocytes were incubated in BSS containing a fixed concentration (either 0.3 or 3.0 mM) of 3-DAG and an increasing concentration (0 -40 mM) of D-glucose, and the rates of 3-DAG uptake were assessed by measuring the initial 3-s 3-DAG uptake as described under "Experimental Procedures." Analysis of the data (Fig. 4) shows that D-glucose inhibited the 3-DAG uptake as a saturable function of D-glucose concentrations with an apparent inhibition constant (D-glucose concentration at which 3-DAG uptake was inhibited 50%; K I Ј). Two distinct values for K I Ј of 11.2 and 18.4 mM were calculated for the inhibition for the uptake measured at 0.3 and 3.0 mM 3-DAG, respectively (Fig. 4). Further analysis of the data indicated that D-glucose increases the apparent K m Ј of 3-DAG uptake without changing the V max . These findings clearly demonstrate that 3-DAG and D-glucose compete for the same transporter, namely GLUT1. 1:10 BSS buffer containing 2 mM HgCl 2 was quickly added, and the ghosts were separated as pellets by centrifugation as above at 7°C. 3-DAG content was assayed for each ghost pellet as described under "Experimental Procedures" and expressed in quantity relative to that measured at t ϭ 0 (the 3-DAG content in ghosts incubated in the presence of 2 mM HgCl 2 ). The data were plotted as a function of incubation time. Each data point represents an average of duplicate measurements.

FIG. 3. Kinetic analysis of the rate of 3-DAG uptake by human erythrocytes as a function of 3-DAG concentrations.
Cells were suspended in BSS containing specified concentrations of 3-DAG at t ϭ 0 as in the experiments of Fig. 1. The initial velocities for 3-DAG uptake were measured by quantitating the first 3-s uptake of 3-DAG by cells by arresting the flux with HgCl 2 and separating cells free of medium by centrifugation. The results were plotted as k(S)/v versus (S) according to the relationship, (S)/v ϭ K m /V max ϩ (S)/V max , where v and (S) are initial velocity (in mol/s) and 3-DAG concentration in ghosts at the start (t ϭ 0) of the flux measurement, respectively, K m and V max are the Michaelis-Menten constant and maximal velocity, respectively, and k is a composite constant and equal to 1.81 ϫ V max. Each data point represents an average of triplicate measurements whose S.D. were less than 10%.
An extensive pepsin digestion (see "Experimental Procedures") of 3-DAG-photolyzed GLUT1 did not cause any significant reduction in the 3-DAG incorporation of GLUT1 protein (not illustrated). Since this pepsin digestion cuts all individual TMHs free of nonmembrane segments (26), the finding indicates that 3-DAG was incorporated into TMH residues with little incorporation into nontransmembrane domain including loops.
A mild trypsin digestion of GLUT1 after photolysis with 3-DAG followed by SDS-PAGE and protein staining (Fig. 6a) revealed the two well known tryptic fragments (23), namely a broad faintly stained band around 30 kDa of the glycosylated N-terminal half and a sharp 19-kDa band of the C-terminal half of the protein, respectively. A broad band at around 55 kDa was rather evident (Fig. 6a), which was identified in immunoblot (not illustrated) as an undigested GLUT1 monomer. Quantitation of 3-DAG labeling after blank subtraction (Fig. 6b) revealed a significant 3-DAG incorporation to each of these protein-staining bands. For mild trypsin-treated sample, consistent with incomplete digestion revealed by protein staining (Fig. 6a), the 30-and 19-kDa tryptic fragments accounted for approximately 50% of total protein label. Although the labeling of the 19-kDa fragment was sharp and noticeable, the overall level associated over the broad 30-kDa fragment was significant; it amounted to 64 Ϯ 9% (n ϭ 3) of the label associated with the 19-kDa fragment in each of three independent estimations. Treatment with N-glycosidase F sharpened the broad 30-kDa band to a 21-kDa band for protein staining and 3-DAG labeling (Fig. 6, a and b). This finding clearly demonstrates that both the N-and C-terminal halves of GLUT1 are labeled by 3-DAG, although labeling is significantly greater at the C-terminal half than at the N-terminal half. It is also important to note that 3-DAG labels the protein at more than one site. This would indicate that the transport process involves physical proximity of glucose to the channel residues at multiple sites. This is entirely possible as the length of the channel approximated by the thickness of the lipid bilayer is more than 3 times that of D-glucose.
Digestion of purified GLUT1 with endoproteinase Lys-C produced six cleavage fragments (fragments A-F in Table I), all of which except fragment B were separately identifiable on SDS-PAGE (Fig. 7a). These fragments were individually eluted from the gel, and their relationship to individual TMHs was established based on partial N-terminal amino acid sequence determination (Table I). Similar experiments using 3-DAG-incorporated GLUT1 (Fig. 7b) revealed at least three major 3-DAGlabeled peaks. Electrophoretic mobilities unequivocally identified the first two peaks as the Lys-C fragments A (residues 301-451) and C (residues 118 -183), respectively. Assignment of the third labeled peak to Lys-C fragments was equivocal. The mobility of this labeled peak failed to match with either of the fragments D and E, although it was significantly closer to the fragment D, indicating that the label is largely at fragment D (residues 256 -300), with a slight if any label at fragment E (residues 184 -225). Although exact quantitation was not possible, relative intensities of 3-DAG label among identifiable Lys-C fragments were approximately 20, 30, and 50% for the fragments A, C, and D/E, respectively. The 3-DAG incorporation to the fragments A and D, which includes TMH8 -12 and TMH7, respectively, demonstrates that the Cterminal half of the protein participates in glucose channel formation. Of particular interest is the significant 3-DAG incorporation in the fragment C, the peptide corresponding to the N-terminal half of the protein containing TMHs 4 and 5. This further supports the conclusion based on the results of tryptic digestion discussed above and demonstrates that the N-terminal half of the GLUT1 transmembrane domain, more specifically TMH4 and/or TMH5, is also directly involved in glucose channel formation. The intense labeling found over the fragments D and E (which includes TMH7 and TMH6, respectively) ( Table I) is mostly due to the C-terminal half or TMH7 and only in small part, if any, due to the N-terminal half or TMH6 labeling. A broad 3-DAG label peaked at an apparent molecular mass of 25 kDa (designated as O in Fig. 7b) may represent a mixture of incompletely cut fragments whose identity is yet to be established.
Definitive assignment of the 3-DAG labeling to individual TMHs and the amino acid residues was not possible in the present study. The findings discussed above nevertheless strongly indicate that the N-terminal half of the transmembrane domain is also directly involved in the channel formation in GLUT1 and argue against the model that only the C-terminal half of the protein is directly involved in the putative glucose transport pathway (15). The findings, on the other hand, are consistent with an alternative model of GLUT1 structure, where we (32) proposed that both the N-and C-terminal halves of GLUT1 transmembrane domain form a glucose channel at their interface. This alternative model emphasizes required physical dimension and solvent accessibility of the channel in GLUT1 as well as the ␣-domain (28)  Photolysis of purified GLUT1 with 3-DAG was identical to that in Fig. 6, except that 500 g of GLUT1 protein was applied in streak. Protein molecular weight markers (Promega, low range) are indicated (from the top): carbonic anhydrase, soybean trypsin inhibitor doublet, horse heart myoglobin, lysozyme, and myoglobin F1, -2, and -3, respectively. Lys-C digestion was carried out as described under "Experimental Procedures." SDS-PAGE, Coomassie Blue staining, gel slicing, and radioactivity counting were as in Fig. 6. The protein staining pattern (panel a) and the 3-DAG incorporation pattern (panel b) are shown. The results were reproduced in two other independent sets of experiments. of covalently reactive, transportable substrate analogs for the facilitated glucose transport in human erythrocytes, Midgley et al. (29) have studied two diazirine derivatives of D-glucose, namely 4-deoxy-3,3-azi-D-glucopyranose (4-DAG) and 6-deoxy-3,3-azi-D-glucopyranose (6-DAG) and shown that both of these analogs induce counterflow (or transient accumulations) of Dgalactose in human erythrocytes, suggesting that they are substrates of GLUT1. We 2 have previously synthesized 4DAG and found that, although this compound is transportable by GLUT1 in human erythrocytes, its low synthetic yield limits the usefulness of this analog as a molecular probe for glucose channel structure mapping by affinity labeling. Synthesis of 3-DAG, however, gave a much higher yield. Our data in the present study clearly demonstrate that 3-DAG is a good substrate of GLUT1, allowing one to affinity photolabel the putative glucose binding site or glucose channel in GLUT1.
Evidence indicates that all of the isoforms of the facilitative glucose transporter family have a common transmembrane domain structure, including that of the putative glucose channel. GLUT1 of human erythrocytes is the only glucose transporter isoform currently available as a pure and functional protein with which the transmembrane domain structure can be studied. The detailed description of this protein structure would provide insight into the basic molecular and subcellular mechanisms underlying the intrinsic transport activity and its regulation of not only this isoform but probably also of other isoforms.
Detailed protein structure may be best studied by x-ray crystallography. To obtain high quality crystals of the intrinsic membrane protein such as glucose transporters, however, is extremely difficult and not likely to be forthcoming in the near future. The biochemical approach to the structural determination of GLUT1 described here, on the other hand, is quite promising although the protocol for the separation of individual transmembrane segments requires further optimization. Preliminary results obtained in our laboratory already indicate that labeled residues can be individually identified by biochemical and biophysical methods. Improved methodology on protein chemistry together with the availability of covalently reactive GLUT1 substrates such as 3-DAG, 4-DAG, and 6-DAG would allow one to map the transmembrane glucose channel structure of not only facilitative glucose transporters but also many other hexose transporters including sodium-glucose co-transporter (30) and bacterial phosphotransferase-linked hexose transporters (31).