Reconstitution and “Transport Specificity Fractionation” of the Human Erythrocyte Glucose Transport System A NEW APPROACH FOR IDENTIFICATION AND ISOLATION OF MEMBRANE TRANSPORT PROTEINS*

A stereospecific n-glucose transport system was reconsti- tuted from human erythrocyte ghosts by hollow fiber dialysis of a cholate-solubilized supernatant protein fraction (20 to 25% of the ghost protein) in the presence of added phospholipid and cholesterol. D-Glucose transport was inhibited by cytochalasin B (K, = 3 to 5 x 10m7 M), HgZ+, and phloretin. o-Glucose uptake exhibited saturation kinetics, with a K, of 20 to 25 mM. A more rapid, saturable n-glucose exchange process had a K m of 40 to 45 mM. This reconstituted glucose transport system was associated with single bilayer vesicles, primarily 450 to 650 A in diameter, and was reconstituted under conditions such that 15 to 25% of the vesicles contained the transport system. At least 63,000 transport sites/red cell have been reconstituted. When the entire vesicle population was preloaded with 0.8 M o-glucose and subsequently incubated in glucose-free medium, most of the glucose leaked out of specifically those vesicles containing the sugar transport system; the concomitant reduction in intravesicular density of the sugar-transporting tion permitted the separation of this fraction from the rest of the vesicle

From the Department of Pharmacology, Harvard Medical School, Boston, Massachusetts 02115 A stereospecific n-glucose transport system was reconstituted from human erythrocyte ghosts by hollow fiber dialysis of a cholate-solubilized supernatant protein fraction (20 to 25% of the ghost protein) in the presence of added phospholipid and cholesterol. D-Glucose transport was inhibited by cytochalasin B (K, = 3 to 5 x 10m7 M), HgZ+, and phloretin.
o-Glucose uptake exhibited saturation kinetics, with a K, of 20 to 25 mM. A more rapid, saturable n-glucose exchange process had a K m of 40 to 45 mM. This reconstituted glucose transport system was associated with single bilayer vesicles, primarily 450 to 650 A in diameter, and was reconstituted under conditions such that 15 to 25% of the vesicles contained the transport system. At least 63,000 transport sites/red cell have been reconstituted.
When the entire vesicle population was preloaded with 0.8 M o-glucose and subsequently incubated in glucose-free medium, most of the glucose leaked out of specifically those vesicles containing the sugar transport system; the concomitant reduction in intravesicular density of the sugar-transporting vesicle fraction permitted the separation of this fraction from the rest of the vesicle population on density gradients ("transport specificity fractionation").
This vesicle fraction was profoundly enriched only in Coomassie-staining protein(s) of sodium dodecyl sulfate gel electrophoretic mobility corresponding to "Band 4.5" (nomenclature of Steck, T. L. (1974) J. Cell Biol. 62, 1-19). These results, together with cytochalasin B binding data from other laboratories which indicate that the sugar transport system is a major protein component of the erythrocyte membrane, provide strong evidence that a component of Band 4.5 is part of the sugar transport system.
The kinetics and specificity of the human erythrocyte sugar transport system have been extensively studied in erythrocytes and resealed erythrocyte ghosts (l-3). Inhibitor binding * This work was suunorted bv the Milton Fund of Harvard University and Grant 'PCM 72-2"0695 from the National Science Foundation.
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. $ Junior Fellow of the Society of Fellows, Harvard University; to whom reprint requests should be addressed. (4)(5)(6) and differential labeling (7,8) evidence has implicated several different polypeptides as possible components of the sugar transport system; the conflict among these findings may be attributable to the lack of specificity for the sugar carrier of the ligands employed in these studies.
The molecular identity of the membrane protein(s) responsible for this facilitated diffusion process has yet to be unambiguously established. The objectives of the study reported here were to identify and purify the polypeptide(s) comprising the erythrocyte sugar transport system. Toward those ends, a novel approach, termed "transport specificity fractionation," was developed. This approach, which may prove to be of general utility for identification and purification of certain membrane transport proteins, involves the processes discussed in the following sections: Reconstitution of Transport System of Interest into Vesicles before Purification -A crude membrane fraction containing the transport system is solubilized with cholate in the presence of a large excess of added lipid; undissolved material is pelleted. Under appropriate conditions, the supernatant fraction contains the transport activity (along with impurities). Vesicles containing the transport activity are formed by a dialysis technique (9) that is a modification of that of Racker (10).
Identification of that Fraction of Vesicles Containing Transport System -Recent studies on the process of reconstitution of purified (NaK)-ATPase by this technique (9) indicate that one obtains the apparently random distribution of functional monomer units of the transport system among dimensionally homogeneous (400 to 600 A) unilamellar vesicles. Based on these findings, one might expect that, when the starting material employed for reconstitution is a heterogeneous membrane fraction (rather than a purified membrane protein), the cholate-solubilizable components of this membrane fraction might be randomly distributed among the vesicle population upon reconstitution; thus, at a sufficiently high lipid/protein ratio, each vesicle will contain only a few protein molecules; just a fraction of the vesicles will contain the membrane transport protein of interest.
Zsolation by Use of Specific Permeability Properties of Transport System -Vesicles containing the transport system are fractionated from the rest of the vesicle population. This is done by exploiting some physical property of the vesicles that can be specifically modulated according to a given vesicle's

") in a Beckman
Ti-50 rotor. One to several simultaneously prepared batches of this supernatant were used immediately for hollow fiber dialysis.
Two types of reconstituted n-glucose transporting vesicle prepa-rations were formed: "9:l phospholipid/protein" vesicles and "18:l phospholipid/protein" vesicles. (This nomenclature was based on the ratio of the total exogenous phospholipid to total ghost protein employed in the reconstitution.) The 9:l vesicles were prepared by dialyzing the above supernatant for 16 h in Bio-Fiber 50 Minitubes, as previously described for reconstitution of (NaKl-ATPase (9); each Minitube was connected to Tygon tubing in series with the interior of the hollow fibers to form a unit of 0.7 ml total internal volume. As required, preparations of 0.7 to 2.8 ml of vesicles were produced by dialysis in 1 to 4 of these hollow fiber units connected in series. glucose transport activity as follows. A 125-~1 aliquot of each set of pooled fractions was eluted on Sephadex G-50M in a Pasteur pipette; this column was equilibrated and eluted with Buffer A + 800 mM glycerol + 2 mM glucose as described under "Transport Assays." The void fractions, containing the vesicles, were pooled. This manipulation resulted in removal of all extravesicular glucose and its replacement by iso-osmotic amounts of glycerol. These pooled void fractions were then assayed for uptake of 1 mM D-versus @HIglucose as described under "Transport Assays." These pooled void fractions were also assayed for lipid phosphate.
Recovery of transport activity was 60 to 110%. Fraction-The total volume of aqueous medium enclosed in the vesicles is determined by forming the vesicles in the presence of a soluble radioactive marker, such as sucrose or glucose, and assaying for the fraction of this marker enclosed in the vesicles; these vesicles enclose 1.8 to 2.2 ml of aqueous medium/g of phospholipid. As shown in Fig.  1, we consistently see that about one-fourth of the aqueous compartment equilibrates rapidly with D-ghCOSe. Vesicles incubated at 23" in the presence of 1 mM cold n-glucose for 1 h subsequently exhibited the same uptake curve for D-[3Hlglucose, demonstrating that there was insignificant decay of the transport activity during the incubation. Vesicles Sodium Dodecyl Sulfate Gel Electrophoresis -Sodium dodecyl sulfate-a&amide gels (7.5% acrylamide, 0.1% bisacrylamide) were prepared in 6-mm (inner diameter) tubes, electrophoresed at pH 8.5, and stained with Coomassie blue according to Davies and Stark (16). The high ratio of lipid to protein in the vesicles imposed severe limitations on the amount of protein that could be loaded on a gel without clogging and fouling of the gel by the excess lipid. To circumvent this problem, the following procedure was developed for delipidation of the samples before they were subjected to SDS' gel electrophoresis.

2.5
The elution profile of the vesicles on a Bio-Gel A-150m . column (Fig. 2) indicates that they are primarily 450 to 650 A in diameter. Fig. 2 also shows that the stereospecific D-glUCOSe transport activity parallels the elution profile of the phospholipid, demonstrating that those vesicles containing the transport activity are similar in size to the bulk population of vesicles. The A-150m profiles, together with the demonstration that one-fourth of the aqueous compartment rapidly equilibrates with n-glucose, show that about one-fourth of these vesicles contain the transport activity. Reconstitution performed at double the initial lipid-protein ratio (18:l uersus 9:1, see "Methods") results, by these criteria, in rapid Dglucose transport in 13 to 17% of the vesicles (data not illustrated). The experimentally determined trapping of aqueous medium (1.8 to 2.2 ml/g of phospholipid) correlates well with what one would expect for vesicles of mean diameter 550 A surrounded by a single bilayer of (1.9 mol/mol) lecithin/cholesterol. One can calculate (9) this theoretical quantity (1.82 ml/ g) from the x-ray diffraction data for hydrated lecithinlcholesterol bilayers (18,19). This agreement indicates that the vesicles must be primarily of single bilayer structure rather than multilamellar.
Vesicles formed by an identical procedure but without added ghost protein have the same size distribution and unilamellar structure. Vesicles were formed in the absence of added ghost protein, in the presence of r@H]glucose; Fig. 3 shows that most of this internally trapped n-glucose remains inside the vesicles over several hours at either 24" or 0" after dilution into either isoosmotic or hypo-osomotic glucose-free medium.

Kinetics and Inhibition
of Glucose Transport -The concentration dependency of the initial rate of stereospecific n-glucose uptake shows saturability, with a K,,, of 20 to 25 mM (Fig.  4); the low level of uptake of L-glucose is not saturable over the L-glucose concentration range studied. Our most active preparations have a V,,,,, for n-glucose uptake about 2-fold greater than that derived from Fig. 4. The concentration dependency for stereospecific D-glUCOSe uptake under conditions which allow exchange diffusion was determined by forming these vesicles in the presence of 800 mM D-glUCOSe, removing the external glucose on Sephadex, and measuring uptake of externally added n-[3Hlglucose (Fig. 5). This latter process is also saturable, with a K,,, of 40 to 45 mM; the maximum velocity is 4-to d-fold higher than under the former conditions, which allow only unidirectional movement of Dglucose. and stereospecific o-glucose transport activity (A---A) on a Bio-Gel A-150m column (20 x 0.9 cm). Elution was with Buffer A + 800 rnM glycerol at 4" at a lo-cm hydrostatic head; vesicles were formed by the same procedure as those of Fig. 1. Column fractions were assayed for phospholipid and stereospecific D-glucose uptake (40 s of incubation at 23.5") with 1 m&f D-[3H]glucose as described under "Methods." The column was calibrated with 0, 850 A diameter monodisperse latex beads; @ Sindbis virus, 700 A diameter (17); 0, phosphatidylcholine vesicles, 490 A mean diameter (9); @, 380 A diameter monodisperse latex beads. 0, depicts mean elution position of vesicle phospholipid (400-~1 vesicles applied to column). V,, denotes void volume, V, denotes column volume.
more permeable to n-glucose than the rest of the vesicles. By   Fig. 6 depicts the inhibition of n-glucose uptake by externally added cytochalasin B (K, = 4 x 1O-7 M); cytochalasin B, in the concentration range studied, had no significant effect on L-glucose uptake. Table I demonstrates asymmetric, thiol reagent-reversible inhibition of D-glUCOSe uptake by internally incorporated Hg2+ and internal phloretin. Externally added Hg*+ or phloretin are much less effective in inhibiting Dglucose uptake. Purification of D-Glucose-transporting Vesicle Fraction -The transport data presented so far serve to compare the kinetics of  uptake of the vesicle fraction containing the sugar transport system with the passive D-glUCOSe permeability inherent to the lipid portion of vesicles formed by this technique. The data indicate that, even at saturating D-&Icase concentrations as high as 800 mM, the vesicle fraction containing the sugar transport system should be severalfold "Transport

Specificity
Fractionation" of Sugar Carrier glucose-free iso-osmotic glycerol medium for 2 h at 24" only slightly alters the peak position; this is expected because most of the glucose should remain in such vesicles. Vesicles were formed with added ghost protein in the presence of 800 mM glucose at the lipid/protein ratio (l&l) demonstrated to insert the n-glucose transport system into exploiting this D-ghCOSe permeability difference, conditions were achieved to selectively reduce the intravesicular fluid density of the n-glucose-transporting vesicle fraction. This enables that fraction to be purified on a density gradient as follows.
Vesicles are preloaded with 800 mM glucose by substitution of 800 mM glucose for the 800 mM glycerol normally employed in the dialysis medium (see "Methods").
Bio-Gel A-150m chromatography and intravesicular aqueous compartment determinations showed that the substitution of 800 mM n-glucose for 800 mM glycerol in the dialysate does not alter the vesicle size distribution or unilamelar structure. The particle density of single bilayer vesicles of 550 a diameter containing 800 mM glucose would be expected to differ by 0.024 g/ml from that of vesicles containing, instead, 800 mM glycerol. As shown in Fig. 7, vesicles formed in the presence of 800 mM glycerol can be separated from vesicles containing 800 mM glucose on a glucose/glycerol density gradient kept iso-osmotic with the intravesicular medium. The peak positions of the two classes of vesicles on this isopycnic gradient correspond well to their respective particle densities calculated for single bilayer vesicles of this size. Incubation of the glucose-containing protein-free vesicles in [Cytochalosin Bl, uM FIG. 6. Inhibition of initial rate of n-glucose uptake by externally added cytochalasin B. Cytochalasin B was added in l-p1 aliquots in ethanol to loo-p1 vesicles incubated for 2 min at 23.5". The vesicles were then chilled on ice, diluted with an equal volume of 2 mM D-[3H]glucose solution containing the same [cytochalasin. Bl, and were incubated for 40 s at 23.5". Vesicles formed by same procedure as described in Fig. 1. about 15% of the vesicles. Vesicles were exchanged on Sephadex into lso-osmotic glucose-free medium, and incubated for 2 h at 24". After concentration by ultrafiltration, the preparation was subjected to the density gradient procedure described above (Fig. 8). Assay of the gradient for stereospecific Dglucose transport activity indicated that the low density portion of the gradient denoted Region A (about 25% of the vesicles) reproducibly contained most of the transport activity. This clearly demonstrates separation of the vesicles containing the sugar transport system from the rest of the population on the basis of transport specificity. Polypeptide Composition of Reconstituted Sugar Transport System -The inclusion of 2-mercaptoethanol as an antioxidant in the buffers used to prepare ghosts results in ghosts with polypeptide patterns on SDS gels (Fig. 9) that are qualitatively similar to those of ghosts prepared by the conventional procedure of Dodge et al. (12). Relative to the "standard" pattern (201, these 2-mercaptoethanol-washed ghosts are somewhat depleted in Bands 5 and 6 (nomenclature of Steck et al. (20,21)). Fig. 9 also shows the polypeptide pattern of the vesicles before their fractionation on a density gradient. Fig. 10 shows densitometer tracings of Coomassie bluestained SDS gels prepared from vesicles after density gradient fractionation.
These gels compare the polypeptide pattern of the three gradient regions corresponding to A, B, and C in Fig. 8 enormous excess of phospholipid present in the vesicle fractions (see "Methods"). A large amount of Band 6 which was present in the vesicle preparation before density gradient centrifugation ( Fig. 9) does not co-sediment with the vesicles on the gradient.

DISCUSSION
The results clearly demonstrate reconstitution of a stereospecific n-glucose transport system from human red blood cell membranes, and are qualitatively consistent with findings recently reported (XI), employing a sonication rather than a dialysis technique to reconstitute n-glucose transport. The data reported here, however, generate a more detailed comparison of the reconstituted n-glucose transport system with carrier-mediated glucose transport in the red cell.

Number of Transport
Sites Reconstituted per Red Cell Ghost-Based on the mean vesicle diameter of 550 A, their single bilayer structure, and x-ray diffraction data on lecithin/ cholesterol bilayers (18,19), each vesicle should contain an average of -1.5 x lo7 daltons of phospholipid.
When the reconstitution was performed by processing 1 mg of red cell ghost protein for every 18 mg of phospholipid employed to form vesicles, what resulted was the insertion of at least one copy of the sugar transport system into 15% of the vesicles; this means that we have reconstituted at least one sugar carrier for every 10.2 x 10' daltons of phospholipid, hence one tarrier/5.74 x 10" daltons of ghost protein processed. Dodge et al. (12) have determined the mass of protein per red cell ghost to be 6 x lo-l3 g (3.6 x 10" daltons/ghost).
Thus, we have unambiguously demonstrated the reconstitution of (3.6 x 10" daltons/ghosts) + (one carrierV5.74 x lo6 daltons of protein) = 0.63 x loj (+20%) carriers per red cell ghosts. This is a minimum estimate of the number of sugar carriers per red cell for two reasons: some of the vesicles may contain more than one carrier, and not all the carriers in the red cell may have been reconstituted.
Our findings unambiguously demonstrate that the sugar carrier comprises at least -2% of the estimated 3.5 x lo6 polypeptides per red cell ghost (derived from the data of Steck et al. (20,22)).
Glucose Transport and Cytochalasin B-The Ki (3 to 5 x 10m7 M) for inhibition of reconstituted n-glucose transport by cytochalasin B agrees quantitatively with its Ki for inhibition of glucose transport in the red cell (23)(24)(25) and is similar to its K, for n-glucose-competitive cytochalasin binding to red cell membranes (1 to 5 x 10m7 M) (25,26). Recent results indicate that 1.5 x lo" (25) to 3 x lo" (26) high affinity cytochalasin B binding sites appear to be intimately associated with the glucose transport system. The number of reconstituted sugar carriers we obtain per ghost (63,000) is substantially lower than the above estimate of cytochalasin B sites. Because our estimate of sugar carrier number by the reconstitution approach is a minimum one, and because more than 1 cytochalasin B molecule may bind for each functional monomer of carrier, our findings do not inherently conflict with the cytochalasin B binding data.
Inhibition of o-Glucose Uptake by Phloretin and Hg'+-Phloretin inhibits carrier-mediated glucose flux in erythrocytes (27). Benes et al. (28) have shown asymmetry in the action of phloretin; externally added phloretin selectively inhibits glucose efBux, but internal phloretin is required to inhibit uptake of glucose in the red cell ghost. Our finding that internal phloretin is required to inhibit influx of Dglucose agrees with this study.
Externally added Hg'+ inhibits n-glucose transport in the red cell (29); this inhibition is reversed by thiol reagents. We find marked asymmetry in our sulfhydryl reagent-reversible inhibition of n-glucose uptake by Hg2+; only about one-half of the transport is inhibitable by externally added Hg2+, complete inhibition requiring internal

Specificity
Fractionation" of Sugar Carrier reason for this is that, as has been found for reconstituted (NaK)-ATPase (9), the vesicles contain populations of reconstituted transport protein that are oriented both "inside out" as well as "right side out" with respect of its normal in uiuo orientation; thus, internal Hg'+ is required to inhibit glucose uptake by this inside out fraction of reconstituted sugar carrier. The fact that virtually all of the n-glucose uptake is inhibited by external cytochalasin B may be because the relatively hydrophobic cytochalasin B molecule can rapidly enter the vesicle.

Maximum
Velocity of Reconstituted Glucose Transport - The K,,, values of the reconstituted transport system for Dglucose uptake (20 to 25 mM) and for exchange diffusion (40 to 45 mM) are similar to those values (25 and 38 mM, respectively) for these processes as observed in the erythrocyte (30,31). Under saturating conditions, eMux of glucose at 20" from resealed ghosts depletes the internal concentration of glucose at the rate of about 100 mM/min (32). Assuming that the maximum estimate of 3 x lo5 transport-associated cytochalasin B binding sites (26) represents the number of sugar carriers per red cell and given a volume per cell of 110 pm" (12), the rate of glucose efllux at saturation is 0.35 x 10mL9 mol/site/min.
Based on the maximal rate for exchange diffusion of glucose at 20" of 260 mM/min (30), a calculation along the above lines yields 0.87 x lo-l9 mol/site/min as maximal Dglucose exchange rate. Assuming one carrier/vesicle shown to contain the transport system, our measurements of V,,, for unidirectional influx and diffusion exchange of n-glucose yields fluxes per carrier that, in our most active preparations, are only 5 and 13%, respectively, of these values as calculated above for the red cell. The process of reconstitution of the (NaKl-ATPase has been shown to result in a substantial lowering ,f the turnover number of that enzyme (9); there is evidence of a "coupling factor" requirement for optimal transport activity by reconstituted Ca-ATPase (33). Perhaps the sugar carrier has such stringent lipid environment or cofactor requirements for optimal transport activity. This problem is under investigation.

Evidence
for Sugar Carrier as a Component of The cytochalasin B data (25,26) indicate that there are 1.5 to 3 x lo5 copies of the sugar carrier per red cell ghost. This number is comparable to the number of copies per cell estimated for several of the major Coomassie-staining protein bands of SDS gels of ghosts (201, and suggests that there is enough of this transport protein (particularly if the functional unit is a dimer or multimer) to comprise one of these bands.
Our approach to the isolation of the sugar transport system is based on the use of the transport properties of the sugar carrier as a physical tool for its isolation. By specifically reducing the intravesicular density of that fraction of vesicles containing the sugar carrier, the carrier-containing vesicles are purified on a density gradient; the sugar carrier-containing portion of the gradient is enriched in only one major Coomassie-staining region, Band 4.5. Band 4.5 has been described as a "diffuse, ill-defined region" of SDS gels (20); the reason for this may be that it contains several protein components, or contains a glycoprotein whose heterogeneity in sugar content causes it to migrate diffusely on SDS gels.
The fact that the number of carrier sites per red cell indicates that the carrier is a significant membrane protein component, together with the observation that only Band 4.5 is enriched by transport specificity fractionation of the carrier, provide strong evidence that the sugar carrier contains a component of Band 4.5. These findings are consistent with differential labeling of Band 4.5 by impermeable maleimides that interact with the sugar carrier (7). While this manuscript was in preparation, two abstracts (34, 35) of reconstitution studies came to our attention, both implicating a component of Band 4.5 as involved in n-glucose transport.
Although the glucose-transporting vesicle fraction is not enriched in Band 3, the presence of substantial amounts of this protein in that fraction allows for the possibility that a component of Band 3 is also a subunit of the sugar carrier. Conclusive identification of the sugar carrier requires further fractionation and characterization; it is conceivable that a component of the sugar carrier may be a protein that does not bind Coomassie stain. It is expected that further utilization of the transport specificity fractionation approach will facilitate the attainment of this objective. This novel approach may be of utility in the identification and isolation of other membrane transport components.