Model of the Exofacial Substrate-Binding Site and Helical Folding of the Human Glut1 Glucose Transporter Based on Scanning Mutagenesis^†

EXPERIMENTAL PROCEDURES

RESULTS AND DISCUSSION

Eleven of the 12 predicted transmembrane segments in the human Glut1 glucose transporter have previously been analyzed by cysteine-scanning mutagenesis and the substituted cysteine accessibility method (26–36). The former analysis permits identification of amino acid residues that are critical for transport activity, and the latter analysis identifies residues that are accessible to the external solvent, enabling a determination of the relative orientation of the transmembrane helices. The cDNA region of Glut1 encoding helix 9, the final Glut1 transmembrane segment to be analyzed by these procedures, was subjected to site-directed mutagenesis to create a series of 21 mutants with single cysteine substitutions (single-C mutants) along the helix. The mutations were generated using a parental Glut1 cDNA template in which all six native cysteine codons were altered to encode either glycine or serine residues (30, 53). The resulting cysteineless (C-less) Glut1 protein is fully functional when expressed in Xenopus oocytes. Helix 9 is presented in helical wheel format in Figure 1. It is the second most hydrophobic helix in Glut1. The specific mutations introduced into C-less Glut1 to produce the 21 helix 9 single-C mutants are listed in Table 1.

Synthetic mRNAs encoding each of the single-C mutants and the parental C-less Glut1 molecule were injected into X. laevis oocytes, and the expression and function of the mutant transporters were assessed 48 h later. Figure 2A shows representative confocal immunofluorescent micrographs of sections of oocytes expressing each of the mutants and stained with an antibody that recognizes the cytoplasmic C-terminus of Glut1. All of the mutants were detected in the oocyte plasma membrane and in internal membrane structures. As we have reported previously (50), Glut1 expressed in oocytes appeared as two broad bands by immunoblot analysis (Figure 2B). The lower band represents core glycosylated protein that had not yet traversed the Golgi apparatus. The upper band represents fully glycosylated transporter protein that had largely matured to the plasma membrane. The relative prominence of the lower bands is due to the extremely slow transit time of Glut1 through the early biosynthetic compartments of oocytes (50). As observed in many previous experiments (26–36), the expression levels of the different mutants were somewhat variable. The upper bands were quantified by comparison to the previously determined Glut1 content of red blood cell membranes loaded as standards on the same gels. The absolute amount of Glut1 detected in the oocyte plasma membranes was used to normalize the mutant transporter activity levels (see below).

The function of the mutant transporters was assessed by measuring the uptake of [³H]-2-deoxyglucose. Figure 3 shows uptake data after normalization to account for different levels of expression of the single-C mutants. The uptake values are compared in each case to that of the C-less parental protein analyzed using the same batches of oocytes. Only a single mutant, G340C, exhibited a statistically significant reduced specific activity relative to the C-less control. G343C protein was expressed at very low levels relative to C-less Glut1 in oocytes. Therefore, to accurately determine the activity of this mutant relative to the control, we injected different levels of C-less Glut1 and G343C mRNAs into oocytes to obtain varying levels of expression of the two proteins. The specific transport activities of G343C and the C-less control were then compared directly in independent experiments at very similar levels of protein expression. These are the results shown in Figure 3 for G343C.

The accessibility of the helix 9 residues to the external solvent was assessed by measuring [³H]-2-deoxyglucose in oocytes after incubation in the presence or absence of the sulfhydryl-specific reagent, p-chloromercuribenzene sulfonate (pCMBS) (Figure 4). Reaction of pCMBS adds a bulky charged aromatic group to accessible cysteine side chains and inhibits transport activity to varying degrees, depending on the residue (37). The activity of only one mutant, T352C, was inhibited after incubation of oocytes in pCMBS, and the magnitude of the inhibition was very modest (~10%). V165C, a helix 5 mutant, is a well-characterized positive control for the pCMBS reaction (37). Transmembrane helices that appear to comprise the inner helical bundle and are thus predicted to have faces that are in direct contact with the external solvent contain between four and eight pCMBS reactive residues (26–36). These results are consistent with helix 9 acting as an outer stabilizing helix that has minimal contact with the exoplasmic fluid.

With the analysis of helix 9 described in this study, all 252 amino acid residues comprising the 12 predicted transmembrane segments of human Glut1 have been subjected to cysteine-scanning mutagenesis and SCAM analysis. These data are summarized schematically in Figures 5 and 6. Forty-five of the 252 predicted transmembrane residues were accessible to pCMBS from the external solvent (purple residues in Figure 5 and numbered residues in Figure 6). The results permit a crude but reliable approximation of the relative orientation of most of the transmembrane helices in the exofacial conformation of the transporter. The data are mostly consistent with a basic model for the helical packing of Glut1 (32, 47) that is similar to the known crystal structures of the Escherichia coli lac permease (48), glycerol-3-P antiporter (49), and EmrD multidrug transporter (54) (see Figures 5 and 6) in which a central aqueous cavity formed by eight inner helices is stabilized by four outer helices. Helix 4 exhibited no pCMBS-sensitive residues, but the orientation shown in Figure 6 is consistent with the results of chemical cross-linking experiments involving dicysteine mutants (ref (55); see the discussion below). The orientation of helices 1 and 12 could not be assigned unambiguously because their pCMBS sensitivity exhibited no periodicity and their reactive residues probably lie outside of the membrane within the exoplasmic fluid (see the discussion below). They were oriented in the most thermodynamically favorable way such that their polar side chains face either the aqueous cavity or an adjacent helix and that only hydrophobic side chains directly face the fatty acyl core of the bilayer.

It is important to emphasize that the crystal structures appear to represent the cytoplasmic-facing orientations of the two bacterial transporters and that the SCAM results summarized in Figures 5 and 6 apply to the exoplasmic-facing orientation of Glut1. The latter assertion follows from the fact that pCMBS is membrane impermeant (56) along with the strong preference for reactivity with Glut1 residues that are predicted to reside in the exofacial halves of the 12 transmembrane segments. The crystal structures of the three bacterial transporters indicate that their prominent cytoplasmic substrate-binding cavities are tightly sealed off at the periplasmic face of the membrane, strongly suggesting a carrier model for transport function in which a substrate-binding site (or two distinct sites) are alternately exposed at opposite faces of the membrane in two mutually exclusive transporter conformations (57). An alternating conformation/access model is also strongly suggested by the presence of a bound high-affinity disaccharide substrate within the cytoplasmic cavity of crystallized lac permease and the absence of bound ligand at the periplasmic face of the membrane (48, 57). An alternating access model for glucose transport is supported by a large body of kinetic and biophysical data collected over the course of several decades (discussed in refs (10), (58), and (59)). Although kinetic anomalies have been observed for Glut1 (12), it is unclear to what extent these reflect the basic transport mechanism as opposed to regulatory phenomena or experimental limitations. It is also noteworthy that Glut4 exhibits simple symmetric carrier kinetics in rat adipocytes (60), consistent with an alternating access carrier model. It seems unlikely that two closely related mammalian glucose transporters would exhibit fundamentally different mechanisms.

Three of the four predicted outer helices of Glut1 (helices 3, 6, and 9) possess only a single pCMBS-sensitive residue, consistent with very limited accessibility to the external solvent. The exception is helix 12, which contains five pCMBS-sensitive residues. The five reactive residues in helix 12 are contiguous and are predicted to lie very near the exoplasmic end of the helix (see Figures 5 and 6). It is therefore probable that these residues reside outside the lipid bilayer in direct contact with the external fluid when the transporter is in its exoplasmic configuration. Seven of the eight predicted inner helices possess between four and eight residues that are reactive with pCMBS, most of which lie in the exoplasmic halves of their respective helices. These data are consistent with solvent-accessible helical faces that line a putative exoplasmic substrate-binding cavity analogous to the cytoplasmic substrate binding cavities evident in the crystal structures of the bacterial transporters. The sole exception is helix 4, which does not possess any pCMBS-sensitive residues. This observation strongly suggests that helix 4 has little if any exposure to the external solvent. The four pCMBS-reactive residues in helix 1 are contiguous and, like the pCMBS-sensitive residues in helix 12, are predicted to lie at the extreme exoplasmic end of the helix (see Figures 5 and 6). It is therefore probable that these residues are positioned in the extracellular fluid outside of the membrane when the transporter is in its exofacial configuration and that the bulk of helix 1 may be shielded from the exoplasmic fluid in the external orientation.

Helix 4 lines the aqueous cavity present in the cytoplasmic configurations of the lac permease (48) and glycerol-3-P anti-porter (49) and participates directly in substrate binding in the case of the lac permease. If we assume that the basic cytoplasmic folding pattern of the two bacterial transporters is conserved in Glut1 (32, 47), then the lack of solvent accessibility of helix 4 in the exofacial configuration suggests that the exoplasmic helical bundling patterns of the MFS transporters may differ from that of their cytoplasmic structures. Additional evidence in support of this possibility is provided by chemical cross-linking experiments conducted on dicysteine mutant Glut1 molecules expressed in oocytes (55). This study demonstrated that Leu325 and Val328 near the cytoplasmic end of helix 8 are positioned within ~6 Å of Gly145 and Ser148, both of which are located near the cytoplasmic end of helix 4. These two pairs of residues are predicted to be positioned much further apart (~26 Å) in the Glut1 structure modeled on the cytoplasmic configuration of the glycerol-3-P antiporter (49). Thus, either this Glut1 model is not accurate, or the cytoplasmic and exoplasmic structures of the transporter differ considerably. If the modeling is accurate with respect to the positioning of these two pairs of residues, then the cross-linking data presumably reflect the exoplasmic configuration of Glut1 in which an exofacially exposed cavity is sealed off at the cytoplasmic end, thus explaining the close apposition of the cytoplasmic ends of helices 4 and 8. A similar conclusion was reached in a comparison of chemical cross-linking data obtained on dicysteine mutants of the lac permease and the cytoplasmic-oriented crystal structures (61); i.e., the cross-linking data indicate the existence of a conformation of the permease that is distinct from the crystal structure.

What experimental evidence exists regarding the exofacial substrate-binding site of Glut1? Elegant studies using a series of deoxy, epimeric, halogeno, and O-alkyl glucose analogues suggest that the transporter forms hydrogen bonds with glucose at the C-1, C-3, C-4, and possibly the C-6 and ring oxygen positions (62, 63). Additionally, glucose binding appears to be stabilized by a hydrophobic interaction between C-6 and an aromatic side chain of the transporter (62). It is not known whether all of the interactions between the sugar and Glut1 occur simultaneously in the exofacial cavity. It is possible that some interactions occur sequentially as relative movement occurs between glucose and the transmembrane helices. Movement of glucose in the cavity would presumably be accompanied by solvation and desolvation of the sugar and protein at specific sites as sugar–protein interactions form and break. Several amino acid residues in Glut1 whose possible interaction with glucose is consistent with experimental mutagenesis and SCAM results are shown in diagrammatic form in Figure 6. Note that not all potential substrate-binding residues in Glut1 have been analyzed, so that the residues shown represent an incomplete subset of candidates. The experimental constraints used to identify these residues were as follows. (1) The residue must be exposed on a helical face in contact with the external solvent as determined by pCMBS sensitivity. (2) Alteration of the side chain by mutagenesis must inhibit the rate of substrate influx. (3) The residue must possess a side chain that is capable of donating or accepting a hydrogen bond or possessing an aromatic side chain (in the case of interaction with the C-6 region of glucose). Most of the putative substrate-binding residues lie within the middle third of their respective helix. This is consistent with the cytoplasmic substrate-binding sites in the lac permease (48) and the glycerol-3-P antiporter (49) but was not used as an absolute constraint because of the uncertain location of the ends of the transmembrane segments. Additionally, as discussed above, the precise molecular interactions between the transporter and glucose may be altered with the commencement of the conformational change that ensues after initial substrate binding. This possibility is not addressed by the static crystal structures and represents a major challenge in deducing the mechanism of membrane transport. Because of the failure to meet the first constraint, no residues in helix 1 or 4 are predicted to be involved in exofacial substrate binding. As discussed above, each of the four contiguous outermost residues in helix 1 reacts with pCMBS, but this is likely because they actually reside in the exoplasmic fluid outside of the bilayer. Thus, the bulk of helices 1 and 4 may lie outside of the inner helical bundle in the exoplasmic configuration, as indicated by the rightward arrows in Figure 6, suggesting a difference between the cytoplasmic and exoplasmic folding patterns.

The biochemical data summarized here cannot unequivocally identify substrate-binding residues. However, the experimental evidence supporting a role for Q161, Q282, and W412 is reasonably compelling. Q161 is highly conserved among members of the Glut family, and substrate influx is dramatically inhibited by mutations at this site (38). Even the conservative substitution of an asparagine residue at this position reduces transport activity by 1 order of magnitude, indicating that the precise spatial positioning of the carbonyl and amide side chain groups is critical for function. Substitution of an asparagine at Q161 reduces the apparent affinity for the exofacial substrate-binding site of a nontransported glucose analogue by 18-fold. However, kinetic analysis is also consistent with Q161 residing at a critical spatial position such that mutations at this site inhibit the inward to outward conformational change of the unloaded transporter and thus tend to lock Glut1 into an inward-facing orientation (38). Similar kinetic and inhibitor analyses support a role for Q282 in exofacial substrate binding (40). However, the caveat concerning mutants at this site possibly being locked into the cytoplasmic configuration does not apply, because a leucine substitution at Q282 has no effect on transport activity, despite decreasing transporter affinity for exofacial ligands. This would not be the case if amino acid substitutions at this site inhibited the rate-limiting step involving the inward to outward conformational change. There is evidence that the QLS motif at positions 279–281 is involved in discriminating between glucose and fructose as substrates, perhaps by affecting hydrogen bond formation at the C-1 hydroxyl group (64). However, none of the QLS motif residues appears to be accessible to the external solvent, and thus, the motif may function to orient other residues in its vicinity that are directly involved in substrate binding. Because Q282 lies adjacent to the QLS motif in helix 7, it is possible that this residue forms a hydrogen bond with glucose at C-1 as indicated in Figure 6. Studies conducted more than 30 years ago with glucose analogues suggested that a Glut1 residue interacts with the C-6 position of the pyranose ring in a complex manner involving both polar and aromatic interactions (62), with a tyrosine or tryptophan residue being the most likely candidate. Of the tryptophan or tyrosine residues that are predicted to possess side chains that are facing the external aqueous cavity, W412 is the most likely to fulfill this function. Glycine, leucine, and cysteine substitutions at this site impair transport activity (29, 39), whereas mutation of W412 to other aromatic residues (phenylalanine and tyrosine) has a minimal impact on transport activity (M. Mueckler, unpublished observations). W412 has been directly implicated in the photolabeling of Glut1 by the inhibitory ligand cytochalasin B (41). These observations suggest that an aromatic side chain is important at position 412 and are consistent with a role for W412 in substrate binding via interaction with the C-6 position. The seven putative substrate-binding residues are colored red in Figure 5. Interestingly, all of the side chains lie within the middle third of their respective transmembrane helices.

There is clearly a need to determine the structure of glucose transporters by X-ray diffraction analysis, a goal that has not yet been achieved for any eukaryotic MFS protein. Toward that end, we have developed a simple procedure that permits the purification of milligram quantities of monomeric recombinant mammalian glucose transporters expressed in Pichia pasteuris. Attempts at crystallization are ongoing.