Topology of a Human Equilibrative, Nitrobenzylthioinosine (NBMPR)-sensitive Nucleoside Transporter (hENT1) Implicated in the Cellular Uptake of Adenosine and Anti-cancer Drugs*

The human equilibrative nucleoside transporter hENT1, the first identified member of the ENT family of integral membrane proteins, is the primary mechanism for the cellular uptake of physiologic nucleosides, including adenosine, and many anti-cancer nucleoside drugs. We have produced recombinant hENT1 in Xenopus oocytes and used native and engineered N -glyco-sylation sites in combination with immunological approaches to experimentally define the membrane architecture of this prototypic nucleoside transporter. hENT1 (456 amino acid residues) is shown to contain 11 transmembrane helical segments with an amino terminus that is intracellular and a carboxyl terminus that is extracellular. Transmembrane helices are linked by short hydrophilic regions, except for a large glycosylated extracellular loop between transmembrane helices 1 and 2 and a large central cytoplasmic loop between transmembrane helices 6 and 7. Sequence analyses suggest that this membrane topology is common to all mam-malian, BSA and 1.5% donkey serum. After further washes in PBS, samples were mounted in Vectashield® mounting medium and exam- ined using a Leica TCS NT laser scanning confocal microscope and LEICA software. For immunoblotting, oocyte membranes (2 (cid:1) g of total protein) were resolved on 12% SDS-polyacrylamide gels (25). The electrophoresed proteins were transferred to polyvinylidene difluoride membranes and probed with affinity-purified anti-hENT1 254–272 (22). Blots were then incubated with horseradish peroxidase-conjugated anti-rabbit antibody (Amersham Pharmacia Biotech) and developed with enhanced chemi- luminescence reagents (Amersham Pharmacia Biotech). Blots were performed at least twice to ensure reproducibility.

Nucleoside transporters play key roles in physiology and pharmacology (1). Uptake of exogenous nucleosides, for example, is a critical first step of nucleotide synthesis in tissues such as bone marrow and intestinal epithelium and certain parasitic organisms that lack de novo pathways for purine biosynthesis (2,3). The same transport mechanisms function as drug transporters and mediate uptake of many synthetic nucleoside analogs used in cancer (and viral) chemotherapy (2). Nucleoside transporters also control the extracellular concentration of adenosine in the vicinity of its cell surface receptors and regulate processes such as neurotransmission and cardiovascular activity (1)(2)(3). Adenosine itself is used clinically to treat cardiac arrhythmias, and nucleoside transport inhibitors such as dipyridamole, dilazep, and draflazine function as coronary vasodilators. In mammals, plasma membrane transport of nucleosides is brought about by members of the concentrative, Na ϩ -dependent (CNT) 1 and equilibrative, Na ϩ -independent (ENT) nucleoside transporter families (1)(2)(3). CNTs are expressed in a tissue-specific fashion; ENTs are present in most, possibly all, cell types.
Two ENT isoforms have been identified in human and rat tissues (4 -7). Human (h) and rat (r) ENT1 and ENT2 (456 -457 amino acid residues) transport both purine and pyrimidine nucleosides, including adenosine, and are distinguished functionally by a difference in sensitivity to inhibition by NBMPR: hENT1 and rENT1 are potently inhibited by NBMPR (K d 1-10 nM) and have the functional designation equilibrative-sensitive (es), while hENT2 and rENT2 are unaffected by micromolar concentrations of NBMPR and have the functional designation equilibrative-insensitive (ei) (4 -7). They also differ in sensitivity to inhibition by vasodilator drugs (hENT1 Ͼ hENT2 Ͼ rENT1 ϭ rENT2) and by the ability of hENT2 and rENT2 to transport nucleobases as well as nucleosides (1,3,7,8). ENTs are widely distributed in other eukaryotes, including insects, nematodes, protozoa, yeast, and plants and do not appear to be present in prokaryotes (3).
The predicted membrane topology of hENT1, the first identified member of the ENT family, contains 11 putative transmembrane helices (4). Binding domains for NBMPR and vasodilator drugs, which compete with permeant for the substrate-binding site at the extracellular surface, comprise a region of the h/rENT1 protein (amino acid residues 100 -231) encompassing putative transmembrane helices 3-6 (9, 10). A residue in the same structural domain of rENT2 (Cys 140 ) is responsible for substrate-protectable inhibition of the transporter by PCMBS (11). In protozoa, mutations of Gly 183 in transmembrane helix 5 of Leishmania donovani LdNT1.1 result in altered substrate specificity and drug resistance to tubercidin (12). A mutant form of TbAT1 from Trypanosoma brucei brucei that confers resistance to melaminophenyl arsenicals contains six amino acid substitutions in different transmembrane helices and loops of its sequence (13). While these studies have been successful in identifying functionally important roles for transmembrane helices 3-6 and other re- gions, there is currently no experimentally based model of ENT topology. Such information is essential to provide a structural basis for further molecular and mechanistic studies of ENT transporter function.
One approach to test the two-dimensional orientation of integral membrane proteins is to identify sites of N-glycosylation. During protein synthesis, attachment of N-linked oligosaccharides to nascent polypeptide occurs on Asn residues in the motif Asn-X-Ser/Thr, where X can be any amino acid except Pro (14,15). Due to strict compartmentalization of enzymes, N-glycosylation is carried out exclusively on the luminal side of the endoplasmic reticulum, which is topologically equivalent to the extracellular side of the protein. Thus, it is possible to identify exofacial and endofacial segments of the protein from the Nglycosylation profile. Here, we have produced recombinant hENT1 in Xenopus oocytes and used native and engineered N-glycosylation sites in combination with immunological studies of native h/rENT1 in erythrocyes and ventricular myocytes to experimentally define the topology of this prototypical nucleoside transporter protein. When combined with computerbased sequence analyses, the results provide a unified model of membrane architecture for all eukaryotic ENT nucleoside transporters.

EXPERIMENTAL PROCEDURES
Computer Predictions of Membrane Topology-The locations of transmembrane helices in hENT1 were predicted by analysis of the amino acid sequence using the hidden Markov model procedure of Sonnhammer et al. (16) as implemented in the computer program TMHMM (version 2.0). In addition, multiple sequence alignments were analyzed for putative transmembrane helices using the TMAP procedure of Persson and Argos (17) and the neural network approach (PHDhtm) of Rost et al. (18). Analyses were performed on the 34 members of the ENT protein family listed in Table I Mutagenesis of Native hENT1 N-Glycosylation Sites-The single mutants N48Q, N277Q, and N288Q (putative loops A and H in Fig. 1) were generated in plasmid pKS-hENT1 (4,9) by the PCR-based megaprimer method (19) using Pfu DNA polymerase.
Introduction of Novel Glycosylation Sites into hENT1-Glycosylation-defective N48Q was used as PCR template to engineer N-glycosylation sites into various regions of hENT1 (Fig. 1). The double mutant N48Q/Q246N, containing a new potential N-glycosylation acceptor site in loop H, was generated by the megaprimer method (19). N-Glycosylation tags from loop A of hENT1 and from the carboxyl terminus of rCNT1 (20) were separately fused to the 5Ј-and 3Ј-ends of N48Q cDNA by the overlap extension PCR method (21), to give constructs N48Q/Ntail and N48Q/C-tail, respectively. A modification of the inverse PCR method (19) was used to introduce N-glycosylation tags from loop A into loops B, D, and E of N48Q, yielding constructs N48Q/loop-B, N48Q/ loop-D, and N48Q/loop-E, respectively. All mutations and insertions were confirmed by DNA sequencing.
Production in Xenopus Oocytes-hENT1 cDNAs were transcribed with T3 polymerase and expressed in Xenopus laevis oocytes according to standard protocols (4,5,22). Healthy defolliculated stage VI oocytes were microinjected with 20 nl of water or 20 nl of water containing RNA transcript (1 ng/nl) and incubated in modified Barth's medium at 18°C for 72 h prior to the assay of nucleoside transport activity and isolation of oocyte membranes. Transport was measured as described previously (4,22) on groups of 12 oocytes at 20°C using [ 14 C]uridine (Amersham Pharmacia Biotech) (1 Ci/ml) in 200 l of transport buffer containing 100 mM NaCl, 2 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , and 10 mM HEPES, pH 7.5. Initial rates of uridine uptake (10 M) were determined using an incubation period of 5 min for hENT1, N48Q, N227Q, N288Q, N48Q/ Q246N, N48Q/loop-D, N48Q/N-tail, and N48Q/C-tail or 30 min for N48Q/loop-B and N48Q/loop-E. The influx values shown are corrected for endogenous uridine uptake measured in control water-injected cells (typically 0.01-0.03 pmol/oocyte⅐5 min Ϫ1 and 0.06 -0.2 pmol/oocyte⅐30 min Ϫ1 ) and represent the means Ϯ S.E. of 10 -12 oocytes. Experiments were performed at least twice with different batches of cells. To prepare membranes (22), groups of 100 oocytes were lysed at 0°C by repeated pipetting in 1 ml of ice-cold 5 mM sodium phosphate, pH 8.0, 0.5 mM phenylmethylsulphonyl fluoride, and 0.5 mM EDTA, followed by centrifugation (1000 ϫ g for 5 min at 4°C) to remove nuclei, yolk granules, and melanosomes. Membranes (plasma membrane ϩ intracellular membranes) in the supernatant were collected and washed twice by centrifugation at 16,000 ϫ g for 30 min. Protein was determined by the bicinchoninic acid protein assay (Pierce) using BSA as standard.
Enzymic Deglycosylation-Oocyte membranes (10 g of total protein) were heated to 65°C for 5 min in 0.5% SDS, cooled to room temperature, and digested with N-Glycosidase F (Roche Molecular Biochemicals) according to the manufacturer's instructions. Identically treated samples omitting enzyme were used as controls.
Competitive ELISA-Antibodies (anti-hENT1 254 -272 ) for competitive ELISA were raised in rabbits against a synthetic peptide corresponding to hENT1 residues 254 -272 by established methods (23). Samples of the antiserum (diluted 1:500) were incubated for 90 min with serial dilutions of intact erythrocytes or of unsealed peripheral protein-depleted erythrocyte membranes (24), as previously described for topological studies on GLUT1 (23). The amount of free antibody remaining in 100-l samples of the supernatants after centrifugation was then assessed by ELISA using microtitre plates coated with 2 g/well proteindepleted membranes.
Immunocytochemistry and Immunoblotting-Antibodies (anti-rENT1 227-290 ) for immunocytochemistry were raised against a glutathione S-transferase fusion protein bearing rENT1 residues 227-290 and affinity-purified on a column of immobilized cellulose binding domain fusion protein bearing the same rENT1 fragment. The antibodies labeled a single band of ϳ65 kDa on immunoblots of rat heart membranes (data not shown). Freshly isolated rat ventricular myocytes fixed with 2% paraformaldehyde in PBS were incubated without or with 0.1% Triton X-100 in PBS to permeabilize the cell membranes, followed by blocking with 10% donkey serum in PBS. Cells were next incubated overnight at 4°C with 10 g/ml affinity-purified antibodies in PBS containing 2% BSA. Where indicated, antibodies were pretreated with an equal concentration by weight of cellulose binding domain fusion protein for 1 h before use. After washing in PBS, samples were incubated for 1 h at 20°C with a 1:100 dilution of an fluorescein isothiocyanate conjugate of donkey anti-rabbit IgG (Chemicon) in PBS containing 1% BSA and 1.5% donkey serum. After further washes in PBS, samples were mounted in Vectashield® mounting medium and examined using a Leica TCS NT laser scanning confocal microscope and LEICA software.
For immunoblotting, oocyte membranes (2 g of total protein) were resolved on 12% SDS-polyacrylamide gels (25). The electrophoresed proteins were transferred to polyvinylidene difluoride membranes and probed with affinity-purified anti-hENT1 254 -272 (22). Blots were then incubated with horseradish peroxidase-conjugated anti-rabbit antibody (Amersham Pharmacia Biotech) and developed with enhanced chemiluminescence reagents (Amersham Pharmacia Biotech). Blots were performed at least twice to ensure reproducibility.

RESULTS
The topology of hENT1 predicted by hydropathy analysis of the amino acid sequence using the algorithm of Eisenberg et al. (26) contains 11 putative transmembrane helices with an amino terminus (residues 1-12) that is intracellular and a carboxyl terminus (residues 452-456) that is extracellular ( Fig. 1) (10). Transmembrane helices are linked by short (Յ16 residue) hydrophilic regions, except for those connecting transmembrane helices 1 and 2 (loop A), and transmembrane helices 6 and 7 (loop H), which contain 41 and 66 residues, respectively. Loop A is predicted to be extracellular and contains an N-glycosylation acceptor site at Asn 48 . Loop H is proposed to be intracellular and contains the remaining two hENT1 N-glycosylation acceptor sites at Asn 277 and Asn 288 . In the present study, we have combined computer predictions of membrane topology with glycosylation scanning mutagenesis and immunological approaches to provide an in depth analysis of the membrane topology of hENT1 and other ENT family members.
Computer Predictions of Membrane Topology-Application of the TMHMM algorithm to the hENT1 amino acid sequence predicted an 11-transmembrane helix topology, with cytosolic and extracellular amino and carboxyl termini, respectively (Fig. 2). This method employs a seven-state hidden Markof model for membrane proteins, and although it is applied to individual sequences, it performs with an accuracy comparable with that for multiple sequence alignment methods, correctly predicting the entire topology for 77% of the sequences in a standard dataset of 83 proteins with known topology (16). The predicted locations of the 11 transmembrane helices were almost identical to those previously deduced from hydropathic analysis (4) (Fig. 1). Similarly, application of the TMAP (17) and PHDhtm (18) algorithms to the aligned sequences of hENT1 and 33 other mammalian, insect, nematode, protozoan, yeast, and plant ENT family members (3) also led to the prediction of a common 11-transmembrane helix topology (Fig. 2). TMAP scans alignments for peaks in propensity curves for the hydrophobic and terminal regions of the transmembrane sequence spans, whereas PHDhtm uses a neural network trained with experimentally determined structures to predict transmembrane helices. Additional support for the locations of the proposed transmembrane helices was provided by the observation that both insertions and deletions were present in the aligned sequences in each of the proposed loops linking the transmembrane helices, except for loop C between transmembrane helices 4 and 5 (Fig. 2). The predicted shortness of loop C, the fact that it never naturally contains an insertion or deletion, and the observation that it contains a highly conserved Pro, suggest that it may play an important structural role in the ENT family. ENTs predicted to have an 11-transmembrane helix topology included T. brucei brucei TbAT1, which previously has been suggested to have a 10-transmembrane helix membrane architecture (13).
hENT1 Glycosylation Variants-Initially, hENT1 Asn 48 , Asn 277 , and Asn 288 were mutated to Gln to create the three single-site mutants N48Q, N277Q, and N288Q. Glycosylationdefective mutant N48Q was then used as the recipient for introduction of novel N-glycosylation acceptor sites into the different locations of hENT1 shown in Fig. 1. For optimal glycosylation, each novel site was well spaced (greater than 11 residues) from its nearest transmembrane helix, and engineered loops were greater than 30 residues in size (27). Thus, artificial glycosylation sites were introduced into loops B, D, E, and H, and at the amino and carboxyl termini, resulting in constructs N48Q/loop-B, N48Q/loop-D, N48Q/loop-E, N48Q/ Q246N, N48Q/N-tail, and N48Q/C-tail, respectively. The remaining external loop (loop C) was not targeted, for reasons described above. RNA transcripts for hENT1 and each of the glycosylation variants were expressed in Xenopus oocytes and assayed for nucleoside transport activity (10 M uridine influx) as described under "Experimental Procedures" (4,22). N48Q/ loop-B and N48Q/loop-E displayed lower levels of transport activity than hENT1, N48Q, N227Q, N288Q, N48Q/Q246N, N48Q/loop-D, N48Q/N-tail, or N48Q/C-tail and required a longer incubation period (30 min versus 5 min) to obtain comparable levels of cellular uptake. Nevertheless, all nine engineered transporters were functional, demonstrating that the native conformation was generally retained in all constructs and that the mutated transporters reached the cell surface.
Immunoreactivity of Wild-type Recombinant hENT1-Blots of membranes from oocytes producing wild-type hENT1 exhibited two characteristic immunobands: a diffuse major band centered around 55 kDa (glyco form) and a sharp minor band at about 50 kDa (aglyco form) (Fig. 3A). Similar to the native glyco form of the human erythrocyte es transporter (29,30), treatment with N-glycosidase F eliminated the 55-kDa band and increased the amount of immunostaining at 50 kDa. Since intracellular membranes (endoplasmic reticulum and Golgi) as well as plasma membrane were present in the preparation, the small amount of aglyco hENT1 seen in the absence of Nglycosidase F possibly reflects newly synthesized transporter not yet modified by sugar addition. The apparent molecular mass of aglyco hENT1 was in close agreement with that predicted from its amino acid sequence (50.2 kDa), and blots of control membranes from water-injected oocytes produced no immunoreative bands, demonstrating the specificity of the antibodies used in the study.
Identification of the Native N-Glycosylation Site Defines Loop A as Extracellular-To determine which of the three potential sites in hENT1 is actually glycosylated, the electrophoretic The results of analyses of the aligned sequences by the TMAP and PHDhtm methods are illustrated as black and dotted rectangles, respectively. Segments predicted by the three algorithms to be intracellular or extracellular are indicated by "i" and "o," respectively. mobilities of mutants N48Q, N277Q, and N288Q were compared with that of wild-type hENT1. As shown in Fig. 3A, mutants N277Q and N288Q each gave two immunobands indistinguishable from those of wild-type hENT1, indicating that both Asn 277 and Asn 288 were not glycosylated. Mutant N48Q, in contrast, displayed a single 50-kDa immunoband whose mobility was unaltered by treatment with N-glycosidase F. Thus, disruption of the glycosylation sequence at Asn 48 abolished N-glycosylation of the protein, identifying Asn 48 as the only endogenous glycosylation site of wild-type hENT1. Previous studies have shown that the glycosylation site of the native human erythrocyte es transporter is located near one of its ends (30). We have now identified this site as Asn 48 , which is close to the amino terminus, a result that agrees with our topology model in Fig. 1, which places loop A extracellularly.
N-Glycosylation Tagging Reveals an Intracellular Amino Terminus-We fused to the amino terminus of N48Q a sequence of 13 amino acids (MTNRLDMSQNVSM) from loop A encoding the native N-glycosylation site of hENT1. The new acceptor site was 14 amino acids away from transmembrane helix 1. On immunoblotting, this fusion construct, N48Q/N-tail, gave a single band similar to that of N48Q. No detectable increase in protein mass was observed, and the mobility of the protein was unaffected by N-glycosidase F digestion, indicating that the amino-terminal tag had not been glycosylated (Fig.  3B). This suggested that the amino terminus is in the cytoplasm, consistent with the orientation predicted by our topographical model. With the demonstration that Asn 48 is extracellular, a cytosolic location for the amino terminus supports the prediction that the intervening hydrophobic region spans the lipid bilayer only once, giving rise to transmembrane helix 1.
The Orientation of Loop B Is Extracellular-To test the orientation of loop B (between transmembrane helices 3 and 4), the 19-amino acid residue loop A glycosylation tag TNRLDMSQNVSLVTAELSK was engineered near the middle of this region, increasing its size to 34 residues. The resulting construct N48Q/loop-B was moderately glycosylated (Fig. 3D), suggesting that loop B faces the extracellular side of the membrane. Support for this interpretation comes from studies of rENT2 where we have identified a unique Cys residue (Cys 140 ) that is responsible for inhibition of this transporter by extracellular PCMBS (11). Because Cys 140 is located in the exofacial half of transmembrane helix 4, it is likely that the loop between transmembrane helices 3 and 4 in rENT2 is also extracellular. Since all ENTs share a common membrane topology, loop B of hENT1 must also be exofacial.
The Central Hydrophilic Region (Loop H) Faces the Cytoplasm-Asn 277 and Asn 288 , located in the carboxyl-terminal half of loop H, are not utilized by the glycosylation machinery (Fig. 3A). To verify that the first half of this large hydrophilic region is also intracellular, a new glycosylation acceptor site was created in this region of N48Q by replacing Gln 246 with Asn 246 , therby generating the sequence N-E-T at that position. The electrophoretic mobility of the immunoband of this double mutant N48Q/Q246N was very similar to that of N48Q both before and after N-glycosidase F treatment (Fig. 3B), indicating the absence of glycosylation at the engineered site (Asn 246 ). While Asn 288 is probably too close to the membrane to be glycosylated, regardless of which way the loop is oriented, Asn 277 and the new site Asn 246 meet all known requirements for optimal glycosylation (27). Therefore, the absence of glycosylation of these two residues suggested that both halves of loop H are intracellularly oriented. If loop H faces the cytoplasm, the preceding loop linking transmembrane helices 5 and 6 (loop C) must be extracellular.
Immunological Studies of Loop H-Anti-hENT1 254 -272 serum (which recognizes a middle portion of loop H in a region between mutated residues Gln 246 and Asn 277 ) was tested in competitive ELISA experiments using unsealed human erythrocyte membranes and intact erythrocytes as the competing antigen. As shown in Fig. 4, incubation of antiserum with unsealed membranes inhibited antibody binding to plate- bound antigen, reaching Ͼ75% inhibition at the highest concentration of membranes employed (50 g). In contrast, intact erythrocytes containing equivalent amounts of integral membrane proteins had no effect. The loop H epitope for the antibodies is therefore exposed on the cytoplasmic side of the human erythrocyte membrane.
The location of loop H was also assessed by immunofluorescence microscopy of fixed rat ventricular myocytes before and after permeabilization of the plasma membrane. Myocytes were chosen for study because the heart contains an abundance of the rat homolog of hENT1, rENT1. Fig. 5A shows that fusion protein antibodies against loop H of rENT1 (anti-rENT1 227-290 ) strongly stained the surface membranes of detergent-permeabilized rat ventricular myocytes, with particularly strong staining of the t-tubules (which are continuous with the plasmalemma) being apparent. The specificity of the staining was demonstrated by the almost total inhibition of staining caused by preincubation of the antibodies with a cellulose binding domain-fusion protein bearing a fragment of rENT1 corresponding to residues 227-290 (Fig. 5B). Lack of staining was also evident when nonspecific rabbit IgG was employed (data not shown). In contrast to permeabilized cells, little or no staining was evident when nonpermeabilized cells were incubated with anti-rENT1 227-290 (Fig. 5C). These findings confirm the cytosolic orientation of the central hydrophilic region and are consistent with our topology model, which predicts loop H to be intracellular. Previous peptide mapping experiments with the es transporter in human erythrocyte and rat liver membranes found a common trypsin cleavage site situated in an intracellular domain at the approximate center of both proteins (30). It is likely that this site is located in loop H.
Loops D and E Are Extracellular-We then used glycosylation tags derived from loop A to investigate the orientations of loop D (between transmembrane helices 7 and 8) and loop E (between transmembrane helices 9 and 10). Accordingly, the sequences TNRLDMSQNVSLVTAELSKD (20 amino acids long) and FTNRLDMSQNVSLVTAELSKDA (22 amino acids long) were independently fused to the middle loops D and E of N48Q, respectively, expanding their sizes to 34 residues. Constructs N48Q/loop D and N48Q/loop E were glycosylated (Fig.  3D), confirming that both loops face the lumen of the endoplasmic reticulum and that these regions are therefore extracellular as indicated by the topology model.
N-Glycosylation Tagging of the Carboxyl Terminus Places It Extracellularly-The carboxyl terminus of hENT1 is predicted to be extracellular, with only four amino acids extending outside the lipid bilayer. To verify this, the 58-amino acid carboxyl terminus of rCNT1, containing two endogenous sites of glycosylation (20), was fused to that of N48Q, creating the construct N48Q/C-tail. Consistent with an extracellular orientation, the fusion protein was glycosylated (Fig. 3B). Both the glyco-and N-glycosidase F-treated forms of N48Q/C-tail migrated with sizes smaller than that expected, considering that the construct had acquired 58 extra amino acids during tag fusion. Immunoblots produced with antibodies specific for the rCNT1 carboxyl terminus (20) confirmed that the glycosylation tag from rCNT1 was present in N48Q/C-tail (Fig. 3C). The size discrepancy must therefore have resulted from anomalous migration of the fusion protein. DISCUSSION Glycosylation scanning mutagenesis was used to determine experimentally the orientation of seven key topographical landmarks of recombinant hENT1 membrane architecture (three in the amino-terminal half of the protein, the central cytoplasmic loop, three in the carboxyl-terminal half of the protein) and, in so doing, define the overall topology of the transporter as a whole. Together with computational approaches, the results obtained for the amino terminus (intracellular), loops A, B, D, and E (extracellular), loop H (intracellular), and the carboxyl terminus (extracellular) validated the topology model of hENT1 shown in Fig. 1 and were consistent with antibody studies, which independently established that loop H of native h/rENT1 was cytoplasmic. Sequence analyses predicted a common 11-transmembrane helix membrane architecture for hENT1 and other eukaryote ENT family members, setting the stage for detailed analysis of ENT structure and function in regions of the transporter implicated in inhibitor binding and nucleoside and nucleoside drug translocation. Since some protozoan ENTs are likely to be proton-dependent (3), our topology model will also have relevance to molecular studies directed toward identification of transporter domains and amino acid residues involved in cation coupling.
In their membrane topology, ENTs are strikingly similar to members of the GLUT family of sugar transporters, although the latter are predicted to have an additional transmembrane segment at the carboxyl terminus. For example, the human glucose transporter GLUT1 has a cytoplasmic amino terminus, a large glycosylated loop connecting transmembrane helices 1 and 2, and a large cytoplasmic loop connecting transmembrane helices 6 and 7 (23). Although the ENTs and the GLUTs, which belong to the major facilitator superfamily of proteins (31), show no obvious sequence similarities, they do share some functional similarities. For example, in human erythrocytes that possess both hENT1 and GLUT1, the potent sugar transport inhibitors cytochalasin B and phloretin weakly inhibit uridine fluxes, while conversely the nucleoside transport inhibitor dipyridamole weakly inhibits sugar transport (32). Moreover, adenosine inhibits hexose transport in these cells in a concentration-dependent manner (32). These observations suggest similarities between the substrate/inhibitor-binding sites of the two transporters. However, it remains to be seen whether these functional and topological similarities reflect similar helix-packing arrangements and molecular mechanisms in the two groups of transporters and a possible common evolutionary origin.