Five transmembrane helices form the sugar pathway through the Na+/glucose cotransporter.

To test the hypothesis that the C-terminal half of the Na+/glucose cotransporter (SGLT1) contains the sugar permeation pathway, a cDNA construct (C5) coding for rabbit SGLT1 amino acids 407-662, helices 10-14, was expressed in Xenopus oocytes. Expression and function of C5 was followed by Western blotting, electron microscopy, radioactive tracer, and electrophysiological methods. The C5 protein was synthesized in 20-fold higher levels than SGLT1. The particle density in the protoplasmic face of the oocyte plasma membrane increased 2-fold after C5-cRNA injection compared with noninjected oocytes. The diameters of the C5 particles were heterogeneous (4.8 +/- 0.3, 7.1 +/- 1.2, and 10.3 +/- 0.8 nm) in contrast to the endogenous particles (7.6 +/- 1.2 nm). C5 increased the alpha-methyl-D-glucopyranoside (alphaMDG) uptake up to 20-fold above that of noninjected oocytes and showed an apparent K0.5alphaMDG of 50 mM and a turnover of approximately 660 s-1. Influx was independent of Na+ with transport characteristics similar to those of SGLT1 in the absence of Na+: 1) selective (alphaMDG > D-glucose > D-galactose >> L-glucose approximately D-mannose), 2) inhibited by phloretin, KiPT = approximately 500 microM, and 3) insensitive to phlorizin. These results indicate that C5 behaves as a specific low affinity glucose uniporter. Preliminary studies with three additional constructs, hC5 (the human equivalent of C5), hC4 (human SGLT1 amino acids 407-648, helices 10-13), and hN13 (amino acids 1-648, helices 1-13), further suggest that helices 10-13 form the sugar permeation pathway for SGLT1.

Cotransporters are a major class of membrane proteins that use cation (Na ϩ , H ϩ , and Li ϩ ) gradients to drive the uphill transport of solutes and water into cells. Although many of these proteins have been cloned and subjected to structure/ function analyses, it has been problematic to draw definitive conclusions about their structure or to identify ligand binding sites and permeation pathways. This is largely due to the difficulty in obtaining structural information at the atomic level. Today, most of our understanding comes from indirect methods such as the comparison of the primary and secondary structure and function of homologous proteins, and functional analysis of chimeric proteins. Here we have studied the functional properties of protein truncations to probe the pathway for sugar transport through the Na ϩ -dependent glucose transporter, SGLT1 (1,2). SGLT1 is a protein of 662 residues that contains 14 transmembrane helices (3). The structure and function of cloned SGLT1 isoforms and homologs, such as SMIT1 and SNST1, have been compared (4,5), and the major differences were found to be concentrated toward the C termini. Functional analysis of a SGLT2-SGLT1 chimera (6) suggested that a Cterminal fraction of SGLT1 (residues 410 -662, helices 10 -14) determines both the sugar affinity and selectivity. To test if the terminal five helices of SGLT1 actually form the pathway for sugar transport, we expressed and functionally analyzed the truncated protein (residues 407-662) in Xenopus oocytes. It behaves as a specific low affinity glucose uniporter with properties comparable with those of native SGLT1 in the absence of Na ϩ .

MATERIALS AND METHODS
Molecular Biology Methods-The recombinant construct of plasmid pGEM3Zfϩ containing the full-length coding sequence of rabbit SGLT1 (7) was digested with NcoI to remove a ϳ1220-base pair DNA fragment. The fragment encodes amino acids 1-406, which cover the nine Nterminal ␣-helices of the SGLT1 transporter in the secondary structure model. Recirculation of the remaining large fragment (C 5 construct) recreated the methionine start codon in frame. The truncated protein consisted of amino acids 407-662, which include transmembrane helices 10 -14. A human equivalent of C 5 (hC 5 ) was made by the same strategy. A second human SGLT1 truncation lacking only the 14th transmembrane span (hN 13 ) was made by mutagenic incorporation of a stop codon after Asn-648, which thereby deleted the 16 C-terminal hydrophobic residues, and the mutated domain was verified by sequencing. A BstXI fragment of hN 13 bearing the new stop codon was swapped into construct hC 5 to produce a third truncated human construct, hC 4 , amino acids 407-648 (see Fig. 1). Template DNA was linearized with EcoRI (C 5 ) or XbaI (hC 5 , hC 4 , hN 13 ) and used for in vitro transcription and capping with SP6 RNA polymerase (MEGAscript transcription kit, Ambion, Austin, TX).
Oocyte Preparation, Western Blot Analysis, Freeze Fracture, and Functional Assays-Mature Xenopus laevis oocytes were manipulated as described (7). All experiments were repeated 2-3 times on oocytes from different donors, usually 3-5 days after cRNA injection. Solubilization of the oocytes for Western blot analysis, SDS-polyacrylamide gel electrophoresis, and immunoblotting were done as described (8), using antibody 8821 (9) at a 1:3000 dilution. Immunoreactive proteins were detected by chemiluminescence (SuperSignal Kit, Pierce) and exposure on Hyperfilm ECL (Amersham Life Science Inc.). Freeze fracture of oocyte plasma membranes and sugar influxes were performed by standard procedures (10,11). 7-10 control or cRNA-injected oocytes were incubated in the presence of [ 14 C]␣MDG 1 (specific activity, 293 mCi/mmol; Amersham Life Science Inc.) alone or in the presence of different nonradioactive substrates/inhibitors. Sugars, phlorizin, and phloretin were from Sigma. Two-electrode voltage clamp measurements were performed as described (7).

RESULTS
C 5 Expression-The truncated rabbit SGLT1 transporter (C 5 ) includes the last five transmembrane domains of wild-type SGLT1 (Fig. 1). Western blot analysis ( Fig. 2A) showed that oocyte expression of C 5 was 20-fold higher than that of SGLT1. C 5 appeared as a major band at ϳ29 kDa and a minor band at ϳ50 kDa. Full-length SGLT1 appeared as two bands; the lower band at ϳ55 kDa corresponds to the core glycosylated protein, and the higher band corresponds to the complex glycosylated protein. All signals were completely blocked with the antigenic peptide.
Freeze fracture and morphometric techniques were used to quantify the proteins in the plasma membranes of C 5 -expressing oocytes and noninjected controls. Fractured plasma membranes expose complementary protoplasmic and external faces. Intrinsic membrane proteins appeared as particles and complementary pits in the fracture faces. We found that the density of particles in the protoplasmic face of C 5 -expressing oocytes (726 Ϯ 292 m Ϫ2 ) was twice as high as in noninjected oocytes (353 Ϯ 54 m Ϫ2 , Fig. 2B). In contrast, particle density in the plasma membrane of oocytes expressing SGLT1 was ϳ5000 m Ϫ2 (10). Electron micrographs (Fig. 2B) showed heterogeneity in the size of C 5 intramembrane particles: analysis of 916 particles revealed three populations: one with a diameter of 7.1 Ϯ 1.2 nm (mean Ϯ S.D.), representing 90% of all C 5 particles, a second with a diameter of 4.8 Ϯ 0.3 nm (2%), and a third with a diameter of 10.3 Ϯ 0.8 nm (8%). In contrast, noninjected oocytes showed a homogeneous distribution with diameter of 7.6 Ϯ 1.2 nm (n ϭ 875). Sugar Transport and Specificity by C 5 -Can the truncated protein function as an active sugar transporter? Electrical properties of C 5 expressed in oocytes were examined by the two-electrode voltage clamp technique (7). In experiments over several months on C 5 -expressing oocytes from different donors, none of the following currents characteristic of full-length SGLT1 (12) were observed: 1) transient currents in the voltage range Ϫ150 to ϩ50 mV; 2) phlorizin-sensitive "leak" currents in the absence of sugar; or 3) Na ϩ -specific sugar-induced currents, even at high (100 mM) sugar concentrations. The sugar uptake in 50 M [ 14 C]␣MDG mediated by C 5 was Na ϩ -independent; there was no difference in transport between C 5 -expressing oocytes incubated for different time periods in 100 mM Na ϩ or choline chloride. Therefore all following experiments, unless indicated, were in 100 mM choline chloride. To estimate the initial rates of ␣MDG transport, noninjected and C 5 -expressing oocytes were incubated for different times in 100 mM choline chloride (Fig. 3A). Whereas the 50 M [ 14 C]␣MDG sugar uptake by noninjected oocytes increased linearly with time (3 Ϯ 1 pmol/oocyte/h), C 5 uptake increased linearly for 5-10 min, plateaued in 15-20 min, and amounted to 27 Ϯ 3 pmol/oocyte after 1 h. Therefore, sugar uptakes of C 5 were measured after 7 min in subsequent experiments. C 5 transported 15 Ϯ 1.4 pmol/oocyte in choline chloride and 13 Ϯ 1.6 pmol/oocyte in Na ϩ , where control uptakes were 1.7 Ϯ 0.1 pmol/oocyte in choline and 2.4 Ϯ 0.1 pmol/oocyte in Na ϩ . In general, C 5 initial uptake rates in Rabbit and human SGLT1 hemi-transporters C 5 and hC 5 (amino acids 407-662, helices 10 -14), human hC 4 (amino acids 407-648, helices 10 -13), and human hN 13 (amino acids 1-648, helices 1-13) proteins were expressed in oocytes.

FIG. 2. Expression of C 5 .
A, Western blot. Proteins from oocyte homogenates corresponding to one control oocyte, one oocyte expressing rbSGLT1, or 1 ⁄16 of a C 5 -expressing oocyte were separated on an 8% polyacrylamide gel electrophoresis gel and transferred to nitrocellulose. Visualization of the immunoreactive proteins was achieved by a 5-s exposure to Hyperfilm ECL. The arrows indicate molecular mass standards in kDa (Novex, San Diego, CA). The fine band in the C 5 lane at ϳ50 kDa represents less than 10% of the intense signal at 29 kDa. B, freeze fracture. Electron micrographs of the protoplasmic face of the plasma membrane of a control (NI) and a C 5 -expressing oocyte. Intramembrane particle density of the control was 353 Ϯ 54 particles/m 2 (mean Ϯ S.D., n ϭ 2990), from 8.5 m 2 plasma membrane), consistent with previous results (10). Intramembrane particle density in the C 5expressing oocyte increased to 726 Ϯ 292 particles/m 2 (mean Ϯ S.D., n ϭ 4578, from a total area of 7.3 m 2 ). The calibration bar is 100 nm.

SGLT1 Hemi-transporter 20325
choline were 10 -20-fold greater than rates in noninjected oocytes (Figs. 3, C and D, and 4). The C 5 -mediated influx of 50 M [ 14 C[␣MDG was reduced 50% by the addition of 50 mM ␣MDG (Fig. 3C), suggesting an apparent K m ␣MDG of about 50 mM. The maximal velocity (V max ) for C 5 was estimated from the net initial rates (v) of sugar uptake in the presence of 50 mM cold substrate ([S]). At any substrate concentration the initial velocity is related to V max by the equation: v/V max ϭ [S]/(K m ϩ [S]). The calculated V max was ϳ1500 pmol/oocyte/min. Knowing the number of C 5 particles in the plasma membrane and the maximal velocity for ␣MDG, we estimated the C 5 turnover number (V max /number of particles). The number of C 5 plasma membrane particles is approximately the product of the oocyte surface area (6 ϫ 10 7 m 2 ; Ref. 10) and the C 5 particle density (375 m Ϫ2 ; Fig. 2B). The difference in the particle density between C 5 -injected oocytes and noninjected oocytes (legend of Fig. 2B) was ϳ373 m Ϫ2 . Therefore the C 5 turnover number is ϳ660 s Ϫ1 (V max /number of parti-cles ϭ 1500 ϫ 10 Ϫ12 ϫ 6.02 ϫ 10 23 /6 ϫ 10 7 ϫ 375/oocyte ϫ 60 s/min). D-galactose reduced the uptake by 20% at 100 mM and by 90% at 200 mM (not shown). 100 mM D-glucose inhibited the C 5mediated uptake by 44%, whereas L-glucose and D-mannose were without effect (Fig. 3D). The apparent affinity of C 5 for D-glucose is therefore about 100 mM. The SGLT1-mediated uptake in the absence of Na ϩ was blocked by 100 mM ␣MDG and D-glucose and to a lesser extent by L-glucose and D-mannose (Fig. 3B).
Inhibition-Phlorizin is a potent high affinity competitive inhibitor of Na ϩ -dependent sugar transport by SGLT1 (K i PZ ϭ ϳ0.03-1.4 M; Ref. 5). The ␣MDG uptake mediated by rbSGLT1 in the absence of Na ϩ was unaffected by 100 M phlorizin but was inhibited 47 and 60% by 500 and 1000 M phlorizin, respectively (Fig. 4A). The apparent K i PZ for SGLT1 in the absence of Na ϩ is therefore between 500 and 1000 M. Sugar transport by C 5 was unaffected by phlorizin (Fig. 4A). Phloretin, the aglycone of phlorizin, had a strong inhibitory effect on sugar uptake mediated by C 5 and by SGLT1 in the absence of Na ϩ (Fig.  4B). In both cases the apparent phloretin K i was ϳ500 M.

FIG. 4. Inhibition.
A, phlorizin. Control oocytes (NI) and oocytes expressing C 5 or rbSGLT1 were incubated in 50 M [ 14 C]␣MDG (100 mM choline chloride) in the absence or the presence of 0.1, 0.5, and 1 mM phlorizin. Sugar transport by C 5 (7.0 Ϯ 0.65 pmol/oocyte) was not inhibited by phlorizin, whereas the transport by rbSGLT1 was reduced by 1 mM phlorizin from 8 Ϯ 0.7 pmol/oocyte to 3.0 Ϯ 0.3 pmol/oocyte. Similar results were obtained in four additional experiments with C 5 and two with rbSGLT1. B, phloretin. Sugar uptakes were performed as described in A but in the presence of phloretin. The addition of 1 mM phloretin reduced the C 5 -mediated ␣MDG influx from 7 Ϯ 1 pmol/oocyte to 2 Ϯ 0.5 pmol/oocyte, whereas in SGLT1-expressing oocytes this reduction was from 8 Ϯ 1 pmol/oocyte to 2.2 Ϯ 0.1 pmol/oocyte. Similar inhibition was measured in two additional experiments.

SGLT1 Hemi-transporter 20326
pmol/oocyte/h) was 30% above the noninjected level of transport (3.4 Ϯ 0.1 pmol/oocyte/h, p Ͻ 10 Ϫ4 ). DISCUSSION Expression of the C 5 protein in Xenopus oocytes resulted in a 20-fold increase in the rate of Na ϩ -independent glucose uptake; the half-time of sugar equilibration with the cell was 5-7 min for C 5 and 180 min for control oocytes (Fig. 3A), and the initial rate of sugar uptake was 10 -20-fold greater than in controls (Figs. 3, C and D, and 4). Sugar uptake by C 5 was independent of the cation species (Na ϩ or choline) and was equilibrative, i.e. the steady-state uptake, 20 pmol/oocyte, was that expected for the equilibration of 50 M sugar with an intracellular space of 0.5 l. The apparent affinity of C 5 for ␣MDG (K m ␣MDG ) was ϳ50 mM), and sugar transport was similar to the uptake mediated by SGLT1 in the absence of Na ϩ . This includes: 1) the higher initial rate of sugar uptake than in control oocytes (Figs. 3 and  4); 2) the sugar specificity, ␣MDG Ͼ D-glucose Ͼ D-galactose Ͼ Ͼ L-glucose Ϸ D-mannose (Fig. 3); 3) the low sensitivity to phlorizin (Fig. 4), which is not unexpected as phlorizin inhibition of SGLT1 is Na ϩ -dependent; and 4) the similar sensitivity to phloretin, K i PT ϳ500 M (Fig. 4). Because the facilitated sugar transport mediated by the GLUT family members also shows sensitivity to phloretin (13), it is possible that the GLUT family members and C 5 interact by a common mechanism with the sugar substrate. Our results suggest that the C 5 truncated protein retains sufficient tertiary structure to transport sugar in a fashion similar to the wild-type protein in the absence of Na ϩ Ϫ consistent with the previous report (6) that affinity and substrate specificity of Na ϩ /sugar cotransport are determined by the C-terminal half of the protein. Although it is perhaps surprising that the truncated protein retains function, it should be noted that a truncated lactose permease (helices 7-12) mediates downhill lactose transport (14), suggesting the localization of essential structures for high affinity binding and substrate translocation within the last six helices of lac permease.
Injection of C 5 cRNA into oocytes resulted in high levels of accumulation (10 -20 times higher than SGLT1) of a protein with the expected molecular mass of ϳ30 kDa ( Fig. 2A). This higher synthesis may be due to a prolonged half-life of the cRNA or changed secondary structures of the cRNA favoring easier initiation of C 5 translation, or it could simply be a reflection of a more effective transfer to nitrocellulose of the smaller C 5 protein. Freeze fracture electron microscopy of the plasma membrane (Fig. 2B) showed that only a small fraction of the protein was inserted into the plasma membrane and that the density of the protoplasmic face intramembrane particles in C 5 expressing oocytes was about twice that of noninjected oocytes. The lower amount of C 5 molecules in the plasma membrane is probably due to inefficiency of C 5 trafficking to the plasma membrane (8). The estimated V max for C 5 was 1500 pmol/oocyte/min, and the turnover of ␣MDG by C 5 was 660 molecules/s/particle. In contrast, the turnover of wild-type SGLT1 is 30 -60 s Ϫ1 (5), an order of magnitude lower. This lower turnover of Na ϩ -driven sugar transport by SGLT1 presumably reflects constraints imposed on the conformational changes in the C 5 domain, which normally accompany sugar transit through intact SGLT1. These constraints are likely the natural consequences of the coupling of sugar transport with Na ϩ transport. Interestingly, the C 5 turnover number is in the range of the turnover of facilitated glucose carrier proteins (GLUTs), like the human erythrocyte glucose transporter (330 -660 s Ϫ1 ; Ref. 15). Although C 5 functions as a facilitated sugar transporter, it is not clear if the functional protein is monomeric or oligomeric. Our freeze fracture electron microscopic analysis of C 5 in the oocyte plasma membrane suggests that the protein exists as three populations of intramembrane particles with diameters of 4.8, 7.1, and 10.3 nm. The smallest particle is probably a C 5 monomer, and the larger particles are probably oligomers.
Analysis of glucose/galactose malabsorption SGLT1 mutants provides additional evidence that sugar affinity and transport are determined by the last five C-terminal helices of SGLT1. In two naturally occurring mutations in patients with glucose/ galactose malabsorption, sugar transport is compromised in the absence of major effects on Na ϩ binding. In one, R499H in helix 12, the sugar affinity was markedly reduced (16); in the other, Q457R in helix 11, sugar binding was reduced by an order of magnitude, and there was no sugar translocation (17). Moreover, mutations in the N-terminal half of SGLT1, e.g. K321A in helix 7, dramatically decreased the apparent Na ϩ affinity without changes in sugar affinity (18). These results, together with our observations that C 5 , unlike SGLT1 (12), does not exhibit the Na ϩ leak pathway or voltage-induced fast current transients, may be taken as evidence that Na ϩ binding and translocation occur through the N-terminal half of the protein. Previous energy transfer measurements between extrinsic fluorescent probes covalently attached to the putative Na ϩ and sugar binding sites of the cotransporter indicated that the distance between the active sites was 35 Å (19).
Our conclusion is that five transmembrane helices (10 -14) of SGLT1 retain sufficient tertiary structure to transport sugar downhill in a stereospecific, selective, phloretin-sensitive manner. Given that the 14th helix is absent from a number of functional SGLT family members (3,20), e.g. the Na ϩ /iodide and H ϩ /proline cotransporters, and that partial function was retained by the human SGLT1 when the 14th helix was deleted (hN 13 ), the present study suggests that the critical requirement for sugar transport pathway perhaps is formed by just four helices (10 -13). The N-terminal region of SGLT1 (helices 1-9) may be required to couple Na ϩ and sugar transport, and studies are now in progress to test this hypothesis.