Relative Proximity and Orientation of Helices 4 and 8 of the GLUT1 Glucose Transporter*

A structure has been proposed for glucose transporter-1 (GLUT1) based upon homology modeling that is consistent with the results of numerous mutagenesis studies (Mueckler, M., and Makepeace, C. (2004) J. Biol. Chem. 279, 10494–10499). To further test and refine this model, the relative orientation and proximity of transmembrane helices 4 and 8 were analyzed by chemical crosslinking of di-cysteine mutants created in a reporter GLUT1 construct. All six native cysteine residues of GLUT1 were changed to either glycine or serine residues by site-directed mutagenesis, resulting in a functional Glut1 construct with Cys mutated to Gly/Ser (C-less). The GLUT1 reporter molecule was engineered from C-less GLUT1 by creating a unique cleavage site for factor Xa protease within the central cytoplasmic loop and by eliminating the site of N-linked glycosylation. Fourteen functional di-cysteine mutants were then created from the C-less reporter construct, each mutant containing a single cysteine residue in helix 4 and one cysteine residue in helix 8. These mutants were expressed in Xenopus oocytes, and the sensitivity of each mutant to intramolecular crosslinking by two homo-bifunctional, thiol-specific crosslinking reagents, bismaleimidehexane and 1,4-phenylenedimaleimide, was ascertained by protease cleavage followed by immunoblot analysis. Four pairs of cysteine residues, Cys148/Cys328, Cys145/Cys328, Cys148/Cys325, and Cys145/Cys325, were observed to be in close enough proximity to be susceptible to crosslinking by one or both reagents. All five of the cysteine residues susceptible to crosslinking are predicted to lie on the same face of helix 4 or 8 and to reside close to the cytoplasmic face of the membrane. These data indicate that the cytoplasmic ends of helices 4 and 8 lie within 6–16 Å of one another and that the two helices twist or tilt such that they are further than 16 Å apart toward the center and the exoplasmic side of the membrane. An updated model for the clustering of the transmembrane helices of GLUT1 is presented based on these data.

The passive exchange of glucose across the membranes of animal cells is mediated by members of the glucose transport-er-1 (GLUT1) 1 protein family (SLC2a) (reviewed in Refs. [1][2][3]. The GLUT family belongs to the major facilitator superfamily, the largest category of proteins that catalyze the transport of small molecules across membranes (4). GLUT1, the prototype member of the GLUT family and the first eucaryotic member of the major facilitator superfamily to be identified and cloned (5), is one of the most extensively studied of all membrane transporters (2,6,7). Kinetic and biophysical studies of glucose transport in the human red blood cell are mostly consistent with a simple alternating conformation mechanism (8 -10), a conclusion that seems to be consistent with recent high-resolution structural studies of two bacterial major facilitator superfamily proteins (11,12).
GLUT1 was the first transporter predicted to possess 12 transmembrane helices (5), a characteristic that it seems to share with most if not all major facilitator superfamily proteins (4). This prediction was confirmed by glycosylation-scanning mutagenesis experiments (13) and other biochemical analyses (reviewed in Ref. 14). The 12 transmembrane helix model for GLUT1 is also consistent with the deduced structures of the lac permease (12) and glycerol-3-P antiporter (11). Several of the 12 proposed transmembrane segments were predicted to form amphipathic ␣-helices, an observation which led to the hypothesis that these helices form the walls of a water-filled cavity involved in the binding and transfer of glucose across the membrane (5). It was also suggested that hydroxyl-and amidecontaining amino acid side chains within the transmembrane helices form the sugar-binding site of GLUT1 by hydrogen bonding with glucose hydroxyl groups.
Considerable experimental support has accumulated for this basic structural model. Cysteine-scanning mutagenesis and substituted cysteine accessibility studies implicate transmembrane segments 1, 2 2 (15), 5 (16), 7 (15,17), 8 (29), 10 (18), and 11 (19) of GLUT1 in the formation of a water-accessible cleft within the membrane. Gln 161 within helix 5 (20) and Gln 282 within helix 7 (21) seem to participate in forming the exofacial substrate-binding site. Val 165 , which is positioned one helical turn distant from Glu 161 , is accessible to aqueous sulfhydryl reagents and, based upon mutagenesis and inhibitor studies (22), seems to lie near the exofacial substrate-binding site. An aromatic side-chain at position 412 within helix 11 seems to be essential for transport activity (23). Finally, hydrogen exchange studies demonstrate that 30% of peptide hydrogen atoms are exposed to water in purified, reconstituted GLUT1, which is consistent with the formation of an aqueous cleft in the membrane (24).
In this study, we utilized chemical crosslinking of di-cysteine (di-C) GLUT1 mutants constructed in a reporter molecule to determine the relative orientation and proximity of transmembrane helices 4 and 8, both of which are predicted to comprise a part of the inner helical bundle that forms the aqueous translocation pathway. The results are consistent with this prediction and permit modeling of the orientation of both helices in the membrane. General Procedures-Procedures for the site-directed mutagenesis and sequencing of human GLUT1 cDNA and the in vitro transcription and purification of GLUT1 mRNAs (25), isolation, microinjection, and incubation of Xenopus oocytes (26), preparation of oocyte membranes (23), SDS-PAGE and immunoblotting with GLUT1 C-terminal antibody (20), and 2-deoxyglucose uptake measurements (27), have been described in detail previously.

Materials-Imported Xenopus laevis
Construction of di-Cysteine GLUT1 Mutants-C-less GLUT1 cDNA subcloned into the oocyte expression vector pSP64T was subjected to site-directed mutagenesis to produce aglyco-C-less GLUT1 containing the amino acid motif for factor Xa protease sensitivity (IEGR). The following changes were introduced into C-less GLUT1 cDNA: Asn 45 3 Thr, Glu 246 3 Ile, and Ser 248 3 Gly. This construct (aglyco-C-less Xa) was then used as a template to individually change each amino acid residue to a Cys from 141-150 in TMS4 and 321-329 in TMS8. To produce the di-C constructs (one Cys in TMS4 and the other Cys in TMS8), restriction fragments produced by Blp I and StuI digests of the single-C mutant cDNAs were ligated together.
Treatment with Homo-bifunctional Maleimide Crosslinking Reagents-Stage 5 Xenopus oocytes were injected with 50 ng of each mutant GLUT1 mRNA. Two days after injection, groups of 15-20 oocytes were incubated for 2 h with 0.15 mM BMH, 0.15 mM o-PDM, or 0.15 mM N-ethylmaleimide (NEM) in 1 ml of Barth's saline at 22°C. Fifteen millimolar stock solutions of the maleimide reagents were prepared in 100% Me 2 SO. After incubation, the reactions were quenched by the addition of 2 mM cysteine, incubated for 10 min, and then the oocytes were washed twice with 250 mM sucrose, 10 mM HEPES-NaOH, pH 7.4, containing a protease inhibitor mixture. In other experiments, 3.0 g of freshly isolated total membranes from injected oocytes were incubated with 0.1 mM o-PDM or BMH, or 0.2 mM NEM in 50 mM NaCl, 20 mM HEPES-NaOH, pH 6.8, for 20 min at 22°C. The reaction was quenched by the addition of 2 mM cysteine followed by incubation for 10 min. Total membranes from either treatment protocol were adjusted to 1% decylmaltoside, 1.5 mM CaCl 2, and then digested overnight at 4°C with 1 g of factor Xa protease (New England Biolabs). The digested membranes were then analyzed by SDS-PAGE and immunoblot analysis using rabbit polyclonal antibody raised against a peptide corresponding to the C-terminal 15 residues of human GLUT1. In some experiments, whole oocytes were treated with crosslinking reagents for 1 h as described above, and then [ 3 H]-2- DOG uptake (50 M, 30 min at 22°C) was measured after quenching the reactions with cysteine.
Statistical Analysis-Uptake data were analyzed for statistical significance using the two-tailed, unpaired Student's t test.

RESULTS
A reporter GLUT1 molecule (aglyco-C-less Xa) was engineered to facilitate a determination of the relative proximity and orientation of pairs of transmembrane helices using chemical crosslinking of di-cysteine (di-C) mutants, an experimental approach that has been used successfully to analyze the structure of the Escherichia coli lac permease (28). A factor Xa protease recognition site was introduced into the large, central cytoplasmic loop of C-less GLUT1 (GLUT1 lacking its six native cysteine residues), permitting the analysis of pairs of cysteine residues residing in opposite halves of the molecule after chemical crosslinking. The site of N-linked glycosylation was also eliminated in the reporter GLUT1 construct by site-directed mutagenesis to simplify the analysis by preventing the appearance of multiple GLUT1 bands on SDS gels.
We have demonstrated previously that C-less GLUT1 exhibits close to wild-type transport activity when expressed in Xenopus oocytes (22). Fig. 1A shows that C-less GLUT1 containing the factor Xa site (C-less Xa) and the aglyco reporter construct (aglyco-C-less Xa) both exhibit robust transport activity in Xenopus oocytes, although aglyco-C-less Xa activity was reduced ϳ30% relative to that of the parental C-less molecule. Fig. 1C shows that this reduced activity was due to a lower level of expression of the reporter construct in oocyte membranes relative to C-less GLUT1. These data indicate that the reporter construct retains functional activity, eliminating the possibility that the mutations introduced into the molecule may have significantly altered its structure.
According to homology modeling of the GLUT1 structure using the lac permease as a template, helices 4 and 8 of GLUT1 are a part of the inner helix bundle that forms the aqueous substrate-binding cavity and may lie in close proximity to one another. However, homology modeling in this instance might be very imprecise because of the extremely weak sequence similarity between the lac permease and GLUT1 (ϳ10%). Therefore, to test the model, we constructed di-C mutants in aglyco-C-less Xa in which a single residue in helix 4 and one residue in helix 8 were both changed to cysteines. Fourteen di-C mutant cDNAs were then constructed from the singlecysteine mutants by restriction enzyme digestion and ligation of the appropriate fragments (see Table I). Fig. 1B shows that all 14 di-C mutants exhibited significant transport activity above background levels when expressed in Xenopus oocytes, suggesting that the mutations did not produce major structural changes in the transporter. As we have observed for single-cysteine GLUT1 mutants (16 -19), most of the variability in transport activity among the different di-C mutants could be attributed to differences in the level of protein expression (data not shown). Total membranes were isolated from oocytes expressing each of the 14 di-C mutants. The membranes were incubated for 20 min in the presence or absence of 0.1 mM of each of two different homo-bifunctional, sulfhydryl-specific, crosslinking agents: BMH, a flexible molecule ϳ16 Å in length, and o-PDM, a rigid molecule ϳ6 Å in length. Control reactions were conducted in the presence of NEM to replicate the sulfhydryl reaction in the absence of crosslinking. The reactions were quenched by the addition of cysteine, the membranes were solubilized in 1% decylmaltoside, digested for 16 h with factor Xa protease, and then subjected to SDS-PAGE and immunoblot analysis. Analyses of aglyco-C-less Xa and 7 of the 14 di-C mutants are shown in Fig.   2A. Protease cleavage should result in the production of an ϳ20-kDa C-terminal fragment of aglyco-C-less Xa that can be detected with a polyclonal antibody raised against the Cterminal 16 residues of GLUT1. If the two cysteine residues in a di-C mutant are in the proper orientation and close enough to one another in proximity, the mutant transporter should show pseudo-protease resistance proportional to the efficiency of the crosslinking reaction. Fig. 2A shows that, as expected, aglyco-C-less Xa was equally susceptible to protease cleavage in the presence of NEM, BMH, and o-PDM, indicating that the parental reporter molecule was not subject to crosslinking. Four di-C mutants, Cys 141 /Cys 321 , Cys 145 /Cys 325 , Cys 146 /Cys 328 , and Cys 148 /Cys 326 , were also not subject to crosslinking by either BMH or o-PDM. However, mutants Cys 145 /Cys 328 , Cys 148 / Cys 325 , and Cys 148 /Cys 328 were clearly susceptible to crosslinking by both BMH and o-PDM. None of the other seven di-C mutants listed in Table I exhibited detectable crosslinking  (data not shown).
Because o-PDM and BMH are permeable to membranes, we next examined whether crosslinking could be achieved in intact oocytes using the protocol described under "Experimental Procedures." Fig. 2B shows that aglyco-C-less Xa and three di-C mutants, Cys 141 /Cys 321 , Cys 146 /Cys 328 , and Cys 148 /Cys 326 , were not subject to crosslinking by either o-PDM or BMH, but that Cys 145 /Cys 325 , Cys 145 /Cys 328 , Cys 148 /Cys 325 , and Cys 148 /Cys 328 were all susceptible to crosslinking in intact oocytes by BMH only. None of the other di-C mutants listed in Table I were subject to crosslinking by either reagent in intact oocytes (data not shown).
Given that crosslinking seemed to occur in intact oocytes, we tested the ability of NEM and the two crosslinking reagents to inhibit activity of the four mutants that were subject to crosslinking, along with four representative di-C mutants that were not susceptible to crosslinking either in vitro or in vivo. Fig. 3 shows that 2-deoxyglucose uptake activity of the parental aglyco-C-less Xa transporter was not affected by any of the three reagents. However, transport activity of all eight di-C TABLE I Construction of di-cysteine mutants in a C-less GLUT1 reporter molecule cDNA encoding C-less human GLUT1 was subjected to oligonucleotide-mediated, site-directed mutagenesis, changing Asn 45 to Thr, Glu 246 to Ile, and Ser 248 to Gly, to produce aglyco C-less containing the factor Xa protease recognition motif (aglyco-C-less Xa). This construct was then used as the template to individually change amino acid residues in TMS4 or TMS8 to a cysteine. Di-C mutants were then constructed from the single-C mutants by ligation of restriction fragments. These data indicate that at least one of the two cysteine residues in all eight of the di-C mutants was accessible to at least two of the three reagents tested, suggesting that the lack of crosslinking in the majority of the 14 di-C mutants was most likely not because of inaccessibility of the cysteine residues to the reagents, but rather, it reflects the relative proximity and orientation of the two residues.

DISCUSSION
The data presented in this study indicate that 4 of 14 di-C mutant GLUT1 transporters were susceptible to intramolecular chemical crosslinking by homo-bifunctional, thiol-specific crosslinking reagents (see Fig. 4). The crosslinking data are consistent with our model for the structure of GLUT1 based on studies employing the substituted cysteine accessibility method (SCAM) (15)(16)(17)(18)(19) and homology modeling (see Fig. 5).
Our data indicate that Gly 145 and Ser 148 in helix 4 lie within FIG. 2. Chemical crosslinking of di-C mutants. Stage 5 Xenopus oocytes were injected with 50 ng of mRNA encoding the parental reporter construct (aglyco-C-less Xa) or the indicated di-C mutant. After incubation of oocytes for 2 days, crosslinking analysis was conducted on either purified oocyte membranes or intact oocytes, as described under "Experimental Procedures." The reactions were quenched by the addition of 2 mM cysteine, and total oocyte membranes were then isolated from the intact oocyte reactions. The total oocyte membranes from both groups were solubilized in 1% DM, digested with 1 g factor Xa protease overnight at 4°C, and then subjected to SDS-PAGE, followed by immunoblotting with rabbit polyclonal antibody raised against the C-terminal 16 residues of human GLUT1. A, autoradiogram of immunoblots from crosslinking conducted on purified membranes. B, autoradiogram of immunoblots from crosslinking reaction conducted on intact oocytes. The position of the ϳ20-kDa C-terminal GLUT1 cleavage product is indicated. ϳ6 -16 Å of Leu 325 and Val 328 in helix 8. These data are consistent with Leu 325 and Val 328 being positioned along the side of helix 8 that faces the aqueous cavity in one or both orientations of the unloaded transporter, in agreement with the results of SCAM analyses (29). Helix 4, which has not yet been analyzed by SCAM, can be roughly oriented in the membrane based upon the crosslinking data, with Gly 145 and Ser 148 facing the aqueous cavity in direct apposition to Leu 325 and FIG. 5. Cross-sectional model of GLUT1 in the membrane as viewed from the cytoplasmic face. Amino acids are represented by the single letter code. Lines between residues Gly 145 -Leu 325 , Gly 145 -Val 328 , Ser 148 -Leu 325 , and Ser 148 -Val 328 represent observed crosslinking with a homo-bifunctional maleimide reagent in the corresponding di-C mutants. *, residues that are accessible to p-chloromercuribenzenesulfonate in SCAM studies. shown as colored spheres. Lines connecting residues Gly 145 -Leu 325 , Gly 145 -Val 328 , Ser 148 -Leu 325 , and Ser 148 -Val 328 represent observed crosslinking with a homo-bifunctional maleimide in the corresponding di-C mutants. The orientation of the helices is based on homology modeling using the lac permease as the template.