Identification of Amino Acid Residues Responsible for the Pyrimidine and Purine Nucleoside Specificities of Human Concentrative Na+ Nucleoside Cotransporters hCNT1 and hCNT2*

hCNT1 and hCNT2 mediate concentrative (Na+-linked) cellular uptake of nucleosides and nucleoside drugs by human cells and tissues. The two proteins (650 and 658 residues, 71 kDa) are 72% identical in sequence and contain 13 putative transmembrane helices (TMs). When produced inXenopus oocytes, recombinant hCNT1 is selective for pyrimidine nucleosides (system cit), whereas hCNT2 is selective for purine nucleosides (system cif). Both transport uridine. We have used (i) chimeric constructs between hCNT1 and hCNT2, (ii) sequence comparisons with a newly identified broad specificity concentrative nucleoside transporter (systemcib) from Eptatretus stouti, the Pacific hagfish (hfCNT), and (iii) site-directed mutagenesis of hCNT1 to identify two sets of adjacent residues in TMs 7 and 8 of hCNT1 (Ser319/Gln320 and Ser353/Leu354) that, when converted to the corresponding residues in hCNT2 (Gly313/Met314and Thr347/Val348), changed the specificity of the transporter from cit to cif. Mutation of Ser319 in TM 7 of hCNT1 to Gly enabled transport of purine nucleosides, whereas concurrent mutation of Gln320 to Met (which had no effect on its own) augmented this transport. The additional mutation of Ser353 to Thr in TM 8 converted hCNT1/S319G/Q320M, from cib to cif, but with relatively low adenosine transport activity. Additional mutation of Leu354 to Val (which had no effect on its own) increased the adenosine transport capability of hCNT1/S319G/Q320M/S353T, producing a full cif-type transporter phenotype. On its own, the S353T mutation converted hCNT1 into a transporter with novel uridine-selective transport properties. Helix modeling of hCNT1 placed Ser319 (TM 7) and Ser353 (TM 8) within the putative substrate translocation channel, whereas Gln320(TM 7) and Leu354 (TM 8) may exert their effects through altered helix packing.

Specialized nucleoside transporter (NT) 1 proteins are required for uptake or release of purine and pyrimidine nucleosides from cells (1,2). Most nucleosides, including those with antineoplastic and/or antiviral activity (3,4), are hydrophilic, and transportability across plasma membranes is a critical determinant of metabolism and, in the case of nucleoside drugs, pharmacologic actions (5). NTs also regulate adenosine concentrations in the vicinity of its cell surface receptors and have profound effects on neurotransmission, vascular tone, and other processes (6,7). In human and other mammalian cells, seven nucleoside transport processes 2 that differ in their cation dependence, permeant selectivities, and inhibitor sensitivities have been observed. The major (cit and cif) and minor (cib, csg, and cs) concentrative NTs are inwardly directed Na ϩ -dependent processes and have been demonstrated functionally in specialized epithelia such as intestine, kidney, liver, and choroid plexus, in other regions of the brain, and in splenocytes, macrophages, and leukemic cells (1,2). Concentrative NT transcripts have also been found in heart, skeletal muscle, placenta, pancreas, and lung. The equilibrative (bidirectional) transport processes (es and ei) have generally lower substrate affinities and occur in most, possibly all, cell types (1,2). Epithelia (e.g. intestine and kidney) and some nonpolarized cells (e.g. leukemic cells) therefore coexpress both concentrative and equilibrative NTs, whereas other nonpolarized cells (e.g. erythrocytes) exhibit only equilibrative NT (1,2). Molecular cloning studies have isolated cDNAs encoding the human and rat proteins responsible for each of the major NT processes (cit, cif, es, and ei) operative in mammalian cells (8 -17). These proteins comprise two previously unrecognized families of integral membrane proteins (CNT and ENT) with quite different predicted architectural designs (1,2). The relationships of these NT proteins to the processes defined by functional studies are CNT1 (cit), CNT2 (cif), ENT1 (es), and ENT2 (ei). Although the NT proteins responsible for the minor mammalian concentrative processes (cib, cs, and csg) remain to be identified, we have cloned a cDNA encoding a CNT protein with cib-like transport activity from the ancient marine vertebrate the Pacific hagfish (Eptatretus stouti). The CNT family also includes the Escherichia coli proton/nucleoside cotransporter NupC (18).
Human and rat CNT1 (650 and 684 residues, 71 kDa), designated hCNT1 and rCNT1, respectively, are 83% identical in amino acid sequence (8,11) and contain 13 putative TMs (one less than predicted in earlier models (8)) with an exofacial glycosylated tail at the carboxyl terminus. 3 hCNT2 (658 residues) (12,13) is 83% identical to rCNT2 (659 residues) (9,10) and 72% identical to hCNT1 (11). Recombinant hCNT1 and rCNT1 produced in oocytes mediate saturable Na ϩ -dependent transport of uridine (apparent K m 40 M), with a Na ϩ /uridine coupling stoichiometry of 1:1. Transport is inhibited by pyrimidine nucleosides (thymidine and cytidine) and adenosine but not by guanosine or inosine (8,11). Adenosine is transported by rCNT1 with a similar K m (25 M) as uridine but with a substantially reduced V max (10). The nucleoside specificity of hCNT2 and rCNT2 is complementary to that of h/rCNT1, showing a preference for adenosine, other purine nucleosides, and uridine (10,13). Although hCNT2 has a higher K m (40 M) for uridine than adenosine (8 M), V max values for the two nucleosides are similar. Thus, "purine nucleoside-selective" CNT2 shows a greater tolerance for uridine as a permeant than does "pyrimidine nucleoside-selective" CNT1 for adenosine. The difference in substrate specificity between CNT1 and CNT2 is reflected in their capabilities to transport different pyrimidine and purine antiviral and anticancer nucleoside drugs. For example, h/rCNT1 transport 3Ј-azido-3Ј-deoxythymidine(zidovudine) and ddC, but not ddI, whereas hCNT2 transports only ddI (8,11,20). Gemcitabine, an anticancer cytidine analog, is a good hCNT1 permeant but is not transported by hCNT2 (22). 4 Recent chimeric studies between the CNT1 and CNT2 proteins of rat have identified TMs 7 and 8 as potential determinants of substrate selectivity (23). When a point mutation (S318G) was introduced into TM 7 of rat CNT1, it was converted from being pyrimidine nucleoside-selective into an apparently broad specificity transporter (24). Here, we present a comprehensive and independent study of the structural features responsible for the substrate specificities of the human CNT1 and CNT2 proteins, focusing not only on TM 7 but also TM 8 and combinations thereof. We have used information derived from chimeric constructs between hCNT1 and hCNT2, sequence comparisons between mammalian CNTs and the hagfish cib transporter (hfCNT) (which is broadly selective for both pyrimidine and purine nucleosides), and site-directed mutagenesis to identify two sets of adjacent residues in TMs 7 and 8 (including the human counterpart of rCNT1 Ser 318 ) that, when converted to the corresponding residues in hCNT2, dramatically alter the substrate selectivity of hCNT1. We report that mutation of the two adjacent residues in TM 7 alone convert hCNT1 into a protein with cib-like activity. We also show that the concurrent mutation of two adjacent residues in TM 8 convert the latter protein with cib-type characteristics into one with purine nucleoside-selective, cif-like characteristics. Mutations in TM 8 of hCNT1 alone produced a novel uridine-selective transport phenotype. Molecular modeling studies have identified possible roles for each of the four identified hCNT1 residues.

Nomenclature and Construction of Chimeric hCNT1 and hCNT2
Transporters-Chimeras between hCNT1 and hCNT2 were created using the three junction points (arrows A, B, and C) illustrated in Fig. 1. A four-character numerical nomenclature was chosen to represent each chimera. The numbers 1 and 2 in the name indicate the approximate percentage of each wild-type cDNA in a particular construct, where "1" represents the DNA and encoded amino acid sequence of hCNT1 and "2" denotes that of hCNT2. For instance, C2211 is a 50:50 chimeric transporter whose amino-terminal half is hCNT2 and whose carboxyl-terminal half is hCNT1; C2221 is a 75:25 chimeric transporter whose aminoterminal three-quarters is hCNT2 and whose carboxyl-terminal onequarter is hCNT1. hCNT1 and hCNT2 cDNAs (GenBank TM accession numbers AF036109 and HSU62968) used to construct the chimeras were cloned in this laboratory as described previously (11,13) into the pBluescript II KS(ϩ) (Stratagene) vector. All chimeras were produced in two steps by the overlap extension polymerase chain reaction method (25) using high fidelity Pyrococcus furiosus DNA polymerase. All chimeras were sequenced in both directions to ensure that the correct splice sites had been introduced.
Nomenclature and Construction of Site-specific Mutated hCNT1 Transporters-Sequence comparisons between the TM 7-9 regions of h/rCNT1 (cit), hfCNT (cib), and h/rCNT2 (cif) were used to identify residue differences between the cit, cib, and cif transport proteins (Fig.  4). The nine residues of hCNT1 selected for mutagenesis are shown by arrows as follows: three in TM 7, five in TM 8, and one in TM 9. In each case the residue in hCNT1 was converted to the corresponding residue at that position in hCNT2 and are designated M1-M9 (Table I). For example, mutant M1 has the single substitution S311A, whereas M1/ 2/3 is a combination mutant with three substitutions in TM 7 corresponding to S311A, S319G, and Q320M. All hCNT1 point mutations were produced in two steps by a modified overlap extension polymerase chain reaction method (26). All constructs were sequenced in both directions to confirm that the correct mutations had been introduced. The final combination mutant M2/3/6/7 was sequenced in its entirety to ensure that no additional mutations had been introduced.
In Vitro Transcription and Expression in Xenopus Oocytes-Plasmid DNAs were linearized with NotI and transcribed with T3 polymerase using the mMESSAGE mMACHINE TM (Ambion) transcription system. Defolliculated stage VI Xenopus oocytes (11) were microinjected with 20 nl of water or 20 nl of water containing capped RNA transcript (20 ng) and incubated in modified Barth's medium (changed daily) at 18°C for 72 h prior to the assay of transport activity.
Transport Assays-Transport assays were performed as described previously (7, 10) on groups of 12 oocytes at 20°C using 14 C-labeled nucleosides (Moravek Biochemicals or Amersham Pharmacia Biotech) (1 Ci/ml) in 200 l of transport buffer containing either 100 mM NaCl or 100 mM choline chloride and 2 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , and 10 mM HEPES, pH 7.5. Except where otherwise indicated, nucleoside uptake was determined at a concentration of 20 M using an incubation period of 30 min (8). Each experiment was performed at least twice on different batches of cells and included hCNT1 and hCNT2 as controls to eliminate transport variability between batches of oocytes. The flux values shown are means Ϯ S.E. of 10 -12 oocytes.
Molecular Modeling-Predictions of the possible orientations of putative TMs 7, 8, and 9 in hCNT1 and its homologs, with respect both to the lipid bilayer and to other helices, were made by analysis of the patterns of residue substitution in these regions of the aligned sequences of the following 18 members of the CNT transporter family: rCNT1 (rat CNT1, GenBank TM accession number U10279); hCNT1 (human CNT1, GenBank TM accession number U62968); pkCNT1 (pig kidney CNT1, GenBank TM accession number AF009673); rCNT2 (rat CNT2, GenBank TM accession number U25055); mCNT2 (mouse CNT2, GenBank TM accession number AF079853); hCNT2 (human CNT2, Gen-Bank TM accession number AF036109); hfCNT (hagfish cib transporter, GenBank TM accession number AF132298); F27E11.1 (Caenorhabditis elegans, GenBank TM accession number AF016413); F27E11. The patterns of residue substitution in the aligned sequences were investigated by the graphical method of Baldwin (27). Two approaches were used to predict the location of buried and lipid-accessible residues: the identification of TM positions able to accommodate polar residues and the identification of positions of restricted variability. Residues at these locations would be predicted to be buried within the bundle of TMs, either forming helix-helix contacts or lining the substrate translocation pathway. Conversely, positions of high variability, and where polar residues are never found, might be predicted to be exposed to the hydrophobic core of the lipid bilayer. Polar residues that would not be expected to be in contact with the lipid acyl chains were defined as charged residues and those capable of forming more than one hydrogen bond. Because of their potential involvement in substrate recognition, serine and threonine were included in the category of polar residues, although their side chains can hydrogen bond to the main chain of an ␣-helix and so can be on its lipid-facing surface. Non-polar residues, which could be in contact with the lipid acyl chains, were defined as those normally classed as hydrophobic, but also include tyrosine, which has been found on the lipid-facing surface of TMs in other membrane proteins including bacteriorhodopsin (27).

RESULTS AND DISCUSSION
hCNT1 (650 residues) and hCNT2 (658 residues) belong to different CNT sub-families and exhibit strongest residue similarity within TMs of the carboxyl-terminal halves of the proteins (Fig. 1). Functionally, hCNT1 and hCNT2 display cit-and cif-type Na ϩ -dependent nucleoside transport activities (11,13). Therefore, although both hCNT1 and hCNT2 transport uridine, they are otherwise selective for pyrimidine (hCNT1) and purine (hCNT2) nucleosides (except for modest transport of adenosine by hCNT1). Below, we describe a series of chimeric and site-directed mutagenesis experiments aimed at identifying hCNT1/2 domains and amino acid residues responsible for the marked differences in permeant selectivity between the two transporters.
hCNT1/hCNT2 Chimeras-Splice sites between hCNT1 and hCNT2 were engineered at the beginning or end of putative extramembranous domains as predicted by the topology model in Fig. 1, thereby minimizing disruption of native TMs and loops. To increase further the probability of obtaining functional hCNT1/2 chimeras, we concealed splice sites within regions of identical amino acid sequence in the two proteins. These splice sites divided the proteins into four unequal quarters ranging from 85 to 261 residues, each containing 2-4 TMs (Fig. 1).
RNA transcripts for each chimeric cDNA were synthesized by in vitro transcription and were microinjected into Xenopus oocytes, which were then assayed for (i) functionality (using uridine as a universal hCNT1/2 permeant) and (ii) substrate selectivity (using thymidine and inosine as diagnostic hCNT1 Residues corresponding to insertions in the sequence of hCNT1 or hCNT2 are indicated by circles containing "ϩ" and "Ϫ" signs, respectively. Arrows A, B, and C represent splice sites used for the construction of chimeras. and hCNT2 permeants, respectively). Shown in Fig. 2 is a representative transport experiment for the two wild-type transporters hCNT1 and hCNT2 and for chimeras C2211, C2221, C1221, and C1121 (where 1 is the hCNT1 sequence and 2 is the hCNT2 sequence). The first in the series, C2211, was a 50:50 chimera incorporating the amino-terminal half (TMs 1-6) of hCNT2 and the carboxyl-terminal half (TMs 7-13) of hCNT1. Functionally, C2211 exhibited pyrimidine nucleosideselective characteristics similar to hCNT1 (marked thymidine uptake and low inosine transport), indicating that the regions conferring substrate selectivity were located largely within the carboxyl-terminal half of the transporter. The second chimera, C2221, increased the hCNT2 portion of the transporter by 3 TMs, leaving the 4 remaining TMs at the carboxyl terminus as hCNT1. This 75:25 hCNT2/hCNT1 construct displayed inosine and thymidine transport characteristics similar to hCNT2, implicating residues 303-387 (incorporating TMs 7-9) as the determinant of substrate specificity. The hCNT2-like transport profile of chimera C1221 (incorporating the middle 5 TMs of hCNT2 (TMs 5-9) into hCNT1) was consistent with this conclusion. Chimera C1121 (incorporating only residues 303-387 of hCNT2 into hCNT1) directly confirmed involvement of the TM 7-9 region by also exhibiting hCNT2-like transport properties.
Although these studies implicated TMs 7-9 as the primary region responsible for substrate specificity, the finding that chimeras C2221, C1221, and C1121 showed modestly increased uptake of thymidine relative to wild-type hCNT2 (Fig. 2) suggested secondary involvement of other regions of the protein.
Similarly, Fig. 2 shows that chimera C2211 exhibited significantly increased uptake of inosine relative to wild-type hCNT1. The chimeras transported uridine to similar extents as hCNT1 and hCNT2, suggesting that the native conformations were retained in all constructs.
Complementary reciprocal chimeras were also prepared (C1122, C1112, C2112, and C2212). C1122 displayed low uri- dine transport but maintained purine nucleoside selectivity with an ϳ5-fold increase in inosine uptake compared with water-injected oocytes (2.02 Ϯ 0.63 pmol/oocyte⅐30 min Ϫ1 versus 0.42 Ϯ 0.04 pmol/oocyte⅐30 min Ϫ1 ) and no detectable thymidine transport. The other chimeras were non-functional, perhaps because of altered helical packing or improper plasma membrane targeting. A structural feature shared by these chimeras (and by low activity C1122) was the presence of hCNT2 sequence at the carboxyl terminus.
As shown in Fig. 3, the cif-like transport characteristics of chimera C1121 was confirmed by testing the transportability of a panel of six physiological purine and pyrimidine nucleosides (adenosine, uridine, inosine, thymidine, guanosine, and cytidine). Fluxes were similar in profile and magnitude to those exhibited by wild-type hCNT2 (adenosine, uridine, inosine, guanosine Ͼ Ͼ thymidine, cytidine). Furthermore, uridine uptake was strongly Na ϩ -dependent, providing additional evidence that the native conformation of the transporter had been retained.
Identification of Candidate Residues for Mutation in hCNT1-Our analysis of hCNT1/hCNT2 chimeras located an 85-residue segment in the carboxyl-terminal half of hCNT1 (residues 303-387) that when substituted by corresponding hCNT2 sequence resulted in a purine nucleoside-selective transporter. hCNT1 and hCNT2 are 80% identical and 85% similar within this region. When aligned with rCNT1 and rCNT2, which are functionally similar to their human counterparts, there is 98% identity between hCNT1 and rCNT1 and 92% identity between hCNT2 and rCNT2 in this 85-residue domain. With so few sequence differences between the CNT1 and CNT2 subfamilies, it seemed likely that introduction of point mutations into hCNT1 would identify individual residues contributing to hCNT1/2 substrate specificity. These amino acids would be expected to be located within transmembrane helices.
Comparison of sequences of h/rCNT1 and h/rCNT2 in TMs 7, 8, and 9 ( Fig. 4) identified nine residues that were conserved in CNT1 and CNT2 transporter subtypes, respectively, but differed between the subtypes and might therefore contribute to permeant selectivity (Table I). Some were common to h/rCNT1 and hfCNT, a native broad specificity cib-type transporter (i.e. transports both pyrimidine and purine nucleosides), which we have identified from the Pacific hagfish (see Fig. 4). Others were common to hfCNT and h/rCNT2. In subsequent experiments, these nine residues in hCNT1 were mutated singly and in combination to the corresponding residues in hCNT2 ( Table  I). Three of the mutations were in TM 7 (M1-3), five were in TM 8 (M4 -8), and one was in TM 9 (M9). Each mutant protein was assayed for uridine, inosine, and thymidine transport activity. Representative transport data for each of the hCNT1 mutants investigated in our study are presented in Table II. The results (described below) are presented as expressed fluxes corrected for endogenous uptake in control water-injected oocytes.
Characteristics of TM 7 Mutants of hCNT1-Simultaneous mutation of the three candidate residues in TM 7 of hCNT1 into the corresponding residues of hCNT2 (mutant M1/2/3) altered the substrate selectivity of hCNT1 from being pyrimidine nucleoside-selective to non-selective (broad specificity), allowing uptake of inosine in addition to thymidine and uridine. Mediated uptake of inosine was similar to that of chimera C1121 ( Fig. 2 and Table II), suggesting that TM 7 is largely responsible for allowing transport of purine nucleoside substrates. Similarly, introduction of TM 7 from rCNT2 into rCNT1 produced a chimeric transporter with inosine transport capability (23). To explore which of the three mutated residues in hCNT1 contributed this change, we systematically changed each residue separately to create mutants M1, M2, and M3. The single mutations at positions 311 (mutant M1) and 320 (mutant M3) had no apparent effect on transport, whereas the Ser to Gly shift at position 319 of hCNT1 (mutant M2) allowed for marked inosine uptake. Ratios of inosine:thymidine and inosine:uridine uptake for this mutant were, however, consistently lower than those for mutant M1/2/3. This would suggest that although mutant M2 allowed transport of inosine, other residues also contributed to the higher inosine flux evident with mutant M1/2/3.
The combination TM 7 mutants M1/2, M2/3, and M1/3 were therefore constructed and tested for inosine, thymidine, and uridine transport (Table II). Mutant M1/2 exhibited a transport profile similar to mutant M2, suggesting that the substitution of Ala for Ser at position 311 did not contribute to inosine transportability. In contrast, substitution of glutamine by methionine at position 320 in TM 7 in combination with the change of Ser to Gly at position 319 (mutant M2/3) resulted in a transport profile resembling that of mutant M1/2/3. Mutant M1/3 showed little apparent uptake of inosine, confirming that  7, 8, and 9. Positions of conserved residues that differ between hCNT1/rCNT1 (8, 11) and hCNT2/rCNT2 (12,13) and selected for mutation in hCNT1 are indicated by arrows (M1-M9). At each of these positions, amino acids in h/rCNT1 and h/rCNT2 in common with hfCNT (GenBank TM accession number AF132298) are shown as open boxes and solid boxes, respectively. Circles indicate amino acids in hfCNT that differ from conserved residues in either h/rCNT1 or h/rCNT2. the S319G mutation is required for purine nucleoside transport. Mutation of the corresponding residue in CNT1 of rat (Ser 318 ) also produced an increase in inosine transport activity (24). The ratio of inosine:thymidine uptake was enhanced by additional mutation of rCNT1 Gln 319 (the rat equivalent of hCNT1 Gln 320 ) but resulted in a combination rCNT1 mutant (S318G/Q319M) with low overall transport activity (24). Compared with rCNT1S318G, rCNT1S318G/Q319M exhibited a decreased apparent K m for inosine influx and an increased K m for thymidine (24).
To test whether the hCNT1 mutant M2/3 was truly a broad specificity transporter, oocytes producing the recombinant protein were assayed using the full panel of purine and pyrimidine nucleosides. As shown in Fig. 3, all nucleosides were transported. As well, the Na ϩ dependence of uridine uptake was maintained. Identification of M2 (S319G) in the phenotype change between cit and cib is consistent with the sequence comparisons in Fig. 4 between h/rCNT1, h/rCNT2, and hfCNT. The latter protein exhibits a very similar transport profile to mutant M2/3 when produced in oocytes, 5 and also has a glycine residue at this position (Fig. 4).
Characterization of TM 8 and TM 9 Mutants of hCNT1-Mutations in TM 7 did not modify uptake of thymidine (Table  II). Substitutions in TM 8 or 9 (Table I) were therefore predicted to result in changes in pyrimidine nucleoside selectivity. Simultaneous mutation of the five candidate residues in TM 8 of hCNT1 into the corresponding residues in hCNT2 (mutant M4/5/6/7/8) led to a substantial decrease in thymidine transport, while leaving uridine uptake unaffected. A reduction in thymidine uptake has also been reported for introduction of TM 8 of rCNT2 into rCNT1 but was associated with very low uridine transport activity (23). As shown in Fig. 3, the loss of thymidine transportability extended also to cytidine, creating a recombinant protein with a unique uridine-selective transport profile. Interestingly, the M4/5/6/7/8 combination mutant ex-hibited a modest increase in inosine transport ( Fig. 3 and Table  II), suggesting that TM 7 is not exclusively responsible for purine nucleoside selectivity. Mutation of hCNT1 Ala 370 , the only candidate residue in TM 9 ( Fig. 4 and Table I), did not alter uridine, inosine, or thymidine transport, either alone (mutant M9) or in combination with M2/3 (mutant M2/3/9) (Table II), and was not investigated further. TM9 residues are also potentially excluded by chimeric studies between rCNT1 and rCNT2, where incorporation of TMs 7-8 of rCNT2 into rCNT1 was sufficient to change the transporter from cit to cif (23). Unlike hCNT1 C121 (Fig. 2), the rat chimera was only partly Na ϩ -dependent (23).
Two strategies were employed to identify individual residues or combinations of residues in TM 8 that might contribute to the specificity profile of mutant M4/5/6/7/8. First, two double mutants were constructed. One (M6/8) was suggested by the sequence alignment between h/rCNT1, rh/rCNT2, and hfCNT in Fig. 4 which identified only two residues in TM8 (Ser 353 and Tyr 358 ) that were common to h/rCNT1 and hfCNT but different in h/rCNT2 (and might therefore be involved in loss of thymidine/cytidine transportability). The other (M6/7) was suggested by the two adjacent residues (Ser 353 and Leu 354 ) in TM 8 that were different between h/rCNT1 and h/rCNT2 (Fig. 4). Like Ser 319 and Gln 320 in TM 7, one of the residues was conserved between h/rCNT1 and hfCNT and the other between hfCNT and h/rCNT2. Second, each of the five candidate residues was mutated individually to generate mutants M4 -9.
Combination Mutants between TMs 7 and 8 -We next combined mutations M2/3 and M6 to generate the composite TM7/8 mutant M2/3/6. This recombinant protein, when screened for uridine, inosine, and thymidine transport (Table II), exhibited properties similar to chimera C1121 (uridine, inosine Ͼ Ͼ thymidine). However, when assayed with the full panel of physiological nucleosides, it was discovered that mutant M2/3/6 exhibited relatively low transport of adenosine compared with mutant M2/3, chimera C1121, and hCNT2 (Fig. 3). We therefore tested the combination mutants M2/3/6/7 and M2/3/6/8 (Table II and Fig. 3). Whereas mutant M2/3/6/8 was indistinguishable from M2/3/6, mutant M2/3/6/7 showed a marked increase in adenosine transport, while maintaining the other cif-like characteristics of mutant M2/3/6. Uridine uptake by mutants M2/3/6/7 and M2/3/6 was confirmed to be Na ϩ -dependent and was similar in magnitude to that of wild-type hCNT1 and hCNT2. Kinetic Properties of M2/3/6/7, M2/3/6, and M6 - Fig. 5 shows representative concentration dependence curves for initial rates of transport (3-min flux) of uridine, thymidine, inosine, and adenosine by the combination mutants M2/3/6/7 and M2/3/6. Calculated kinetic parameters (apparent K m and V max ) from these influx data are presented in Table III. M/2/3/6/7mediated transport of uridine was saturable and conformed to Michaelis-Menten kinetics with an apparent K m value (29 M) in the range reported previously for hCNT1, rCNT1, and hCNT2 (37-45 M) (8,11,13). Inosine and adenosine influx were both CNT2-like, with V max values similar to uridine and apparent K m values of 20 and 18 M, respectively (cf. 15 M for inosine transport by rCNT2 and 8 M for adenosine transport by hCNT2) (13,23). In contrast, h/rCNT1 also mediates high affinity transport of adenosine but with a very much reduced V max relative to uridine (resulting from a low rate of conversion of the CNT1-adenosine complex from outward-facing to inward-facing conformations) (10). Mutant M2/3/6/7 retained some thymidine transport activity (see also Table II), but both the apparent affinity and maximum velocity were reduced relative to those of the other three permeants. The ratio V max :K m was 0.3 for thymidine compared with 7.5, 6.5, and 11.3 for adenosine, uridine, and inosine, respectively, a difference of ϳ20-fold (Table III). Human and rat CNT1 transport thymidine and uridine to similar extents (8), 5 with a reported apparent K m of 5 M for rCNT1 (cf. 170 M for M2/3/6/7 in Table III), whereas wild-type rCNT2 has been reported to mediate low fluxes of thymidine (9). Therefore, mutant M2/3/6/7 showed hCNT2 (cif)-type transport characteristics for all four permeants. Mutant M2/3/6 exhibited very similar kinetics to M2/3/6/7, except for a reduced V max of adenosine influx. In agreement with the 20 M adenosine uptake data shown in Fig. 3, V max :K m ratios for M2/3/6/7 and M2/3/6 were, respectively, 7.5 and 2.5, a difference of 3.1-fold. In contrast, corresponding ratios for uridine transport by the two mutants were similar (6.5 and 7.9, respectively). Thus, the M7 mutation increased the maximum velocity of adenosine transport while having no effect on adenosine apparent affinity or the kinetics of other permeants.
Mutant M6 was characterized using a 10-min flux because of its relatively low transport activity and, consistent with the results presented in Table II, exhibited a reduced V max :K m   FIG. 5. Kinetic properties of hCNT1  mutants M2/3/6/7, M2/3/6, and M6. Initial rates of transporter-mediated nucleoside uptake were measured in NaCl transport buffer at 20°C (3-min flux for M2/3/6/7 and M2/3/6, 10-min flux for M6). Mediated transport was calculated as the difference in uptake between RNA-injected oocytes and control oocytes injected with water alone. All of the fluxes were performed on the same batch of oocytes. A-D, M2/3/6/7 (solid circles) and M2/3/6 (open circles): E and F, M6 (triangles). Kinetic parameters calculated from these curves are presented in Table III. ratio for thymidine (0.8) relative to uridine (2.8). The apparent K m for thymidine influx was 48 M, compared with 160 M for mutant M2/3/6 and 167 M for mutant M2/3/6/7, suggesting an interaction of mutations in the two TMs.
Molecular Modeling-Examination of the aligned sequences of putative TMs 7, 8, and 9 in the CNT family of transporters revealed the presence of a number of positions where residue variability was very restricted. Conservation of these characteristic residues suggests that they are involved either in maintaining the structure of the transporters or in the binding of nucleoside substrates, these being features of the family members that are held in common. They are thus likely either to face the putative substrate translocation channel or another helix. The positions of the conserved residues are fairly symmetrically distributed around the circumference of TMs 7, 8, and 9, although TM 8 shows a slightly more asymmetric distribution (Fig. 6A). The distributions of positions that can accommodate polar residues in one or more of the transporters, or at which no polar residue is found, are likewise fairly symmetrical for TM 9. In contrast, TMs 7 and 8 exhibit a more amphipathic character, with predominantly polar residues clustered on one face of the helix and predominantly hydrophobic residues clustered on the other. These distributions of conserved residues, and the existence of conserved residues in the apolar faces of the TMs, suggest that all three TMs are largely sequestered from contact with membrane lipids, presumably by interactions with other transmembrane segments of the protein. The nature of these TMs can be contrasted with, for example, TM 4 which has a much more asymmetric distribution of both conserved and polar residues, and which is likely to occupy a position in the transporter structure that is much more exposed to the membrane lipids (Fig. 6A).
Because the loops connecting putative TMs 7, 8, and 9 in the transporter are predicted to be very short (5 and 13 residues), it is likely that these three putative helices are adjacent in the tertiary structure of the protein. The pattern of conserved polar residues within the helices, together with the results of sitedirected mutagenesis, allow a model to be proposed for their arrangement in the transporter structure, which is shown in Fig. 6B. Although tentative, this model aids interpretation of the experimental results and more importantly may be used to make predictions that can be tested by future site-directed mutagenesis experiments. Hydrophilic portions of the surfaces of TMs 7, 8, and 9 are proposed to contribute to the substrate translocation channel or binding site. In the case of TM 7, this surface would include the highly conserved residues Glu 308 , Asn 315 , and Glu 322 (present in 100, 61, and 94% of the CNT family members, respectively), one or more of which might form hydrogen bonds with the nucleoside substrate. Ser 319 , located close to Glu 308 on this surface, would likewise be lo-cated in the substrate translocation channel. However, the fact that changing this residue in mutant M2 to glycine (which is found at this position in 78% of the other family members) allows hCNT1 to transport inosine suggests that it sterically hinders transport of the purine nucleoside in the wild-type molecule rather than contributing to substrate binding. Similarly Ser 311 , mutation of which to alanine in mutant M1 had no effect on transport activity, is located near the hydrophobic surface of TM 7 and presumably plays no part in substrate binding. Potentiation of the effect of the M2 mutation by simultaneous mutation of Gln 320 to methionine (mutant M2/3) may reflect an alteration of helix packing resulting from the predicted location of this residue at the interface with an adjacent helix, suggested to be TM 8 in the model shown in Fig. 6B.
A similar alteration of helix packing may account for the effect of mutating Leu 354 in TM 8 to valine on the ability of the combination mutant M2/3/6 to transport adenosine. The lack of effect of mutating Val 341 to Ala, and of Tyr 347 or Tyr 358 to phenylalanine (mutants M4 and M8, respectively), probably reflects the location of these residues on the surface of TM 8 distant from the substrate channel and other channel-forming helices. In contrast, Ser 353 is predicted to lie on the surface of TM 8 that faces the translocation channel. The reduction in thymidine uptake activity produced by mutation of this residue in hCNT1 to threonine (M6 mutation) suggests that it might be directly involved in substrate recognition via hydrogen bonding, a suggestion strengthened by the observation that this position is occupied by either a serine or a threonine residue in all members of the CNT family except for the putative transporter of H. pylori, where a proline residue is found.
Mutation of Ala 370 in TM 9 to serine (M9 mutation) was without effect on the transport activity of hCNT1, and so it is not possible to conclude whether or not this helix contributes to the substrate translocation channel. However, it does bear a number of highly conserved hydrophilic residues that might contribute to solute recognition, in particular at position 372 which is occupied by a serine residue in 83% of the CNT family members. Because of the conservation of these residues, and the fact that TM 9 is likely to be adjacent to TM 8 in the transporter tertiary structure, it has therefore been included as a channel-lining helix in the model shown in Fig. 6B, oriented such that Ser 372 faces the channel and Ala 370 is located on the helix surface at greatest distance from the channel. This proposed involvement of TM 9 in the translocation channel should be readily testable by site-directed mutation of Ser 372 and the adjacent residue Ser 383 .
Conclusions-hCNT1 and hCNT2 have cit and cif transport activity for pyrimidine and purine nucleosides, respectively. We have identified four residues (Ser 319 , Gln 320 , Ser 353 , and Leu 354 ) in the TM 7-9 region of hCNT1 that, when mutated together to the corresponding residues in hCNT2, converted hCNT1 (cit) into a transporter with cif functional characteristics. An intermediate broad specificity cib-like transport activity was produced by mutation of the two TM 7 residues alone: mutation of Ser 319 to Gly allowed for transport purine nucleosides and this was augmented by mutation of Gln 320 to Met. Mutation of Ser 353 in TM 8 to Thr converted the cib-like transport of the TM 7 double mutant into one with cif-like characteristics but with relatively low adenosine transport activity. Mutation of Leu 354 to Val increased the adenosine transport capability of the TM 7/8 triple mutant, producing a full cif transport phenotype. On its own, mutation of Ser 353 converted hCNT1 into a transporter with novel uridine-selective transport properties.
A cib-type transport activity has been described in human colon and myeloid cell lines (28 -30), in rabbit choroid plexus (31), and in Xenopus oocytes injected with rat jejunal mRNA (32). A candidate cib-type transporter SNST1 that is related to the Na ϩ -dependent glucose transporter SGLT1 was identified in 1992 in rabbit kidney (33). There is no sequence similarity between SNST1 and either the CNT or ENT protein families. Although recombinant SNST1, when produced in oocytes, stimulates low levels of Na ϩ -dependent uptake of uridine that is inhibited by pyrimidine and purine nucleosides (i.e. cib-type pattern), its function remains unclear because (i) the rate of uridine transport in oocytes is only 2-fold above endogenous (background) levels, whereas a Ͼ500-fold stimulation is observed with h/rCNT1 (8,11), and (ii) cib-type transport activity Residues mutated in the present study are indicated in white letters on a black background; those that affected the transport activity of the protein are in capital letters, and those which were without effect are shown in lowercase letters.
has not been observed in the tissues (kidney and heart) in which SNST1 message was reported (19,21). From the experiments reported here and our cDNA cloning of a broad specificity CNT for hagfish (hfCNT), we hypothesize that mammalian cib is a member of the CNT protein family.
Information from the aligned sequences of TMs 7-9 in CNT family members produced a model for their possible arrangement in the transporter structure, in which Ser 319 lies within the substrate translocation channel and sterically hinders purine nucleoside transport in wild-type hCNT1. Mutation of the other residue in TM 7, Gln 320 , which is predicted to interface with an adjacent helix, may potentiate purine nucleoside transportability through an alteration in helix packing. Altered helix packing may also account for the augmentation of adenosine transport caused by mutation of Leu 354 , since this residue is also predicted to be located on a surface of TM 8 distant from the substrate channel. In contrast, the other TM 8 residue Ser 353 is predicted to face the translocation channel and may directly participate in substrate recognition via hydrogen bonding.