Functional Identification of the Hypoxanthine/Guanine Transporters YjcD and YgfQ and the Adenine Transporters PurP and YicO of Escherichia coli K-12*

Background: Four putative purine transporter genes represent a distinct homology cluster within family NCS2 in Escherichia coli. Results: The four genes encode high affinity transporters for adenine or hypoxanthine/guanine with essential residues at the same consensus sites as in xanthine/uric acid-transporting homologs. Conclusion: Distantly related purine transporters use topologically similar selectivity determinants. Significance: Our study provides the first structure-function insight on the cluster of hypoxanthine-guanine-adenine transporters of family NCS2. The evolutionarily broad family nucleobase-cation symporter-2 (NCS2) encompasses transporters that are conserved in binding site architecture but diverse in substrate selectivity. Putative purine transporters of this family fall into one of two homology clusters: COG2233, represented by well studied xanthine and/or uric acid permeases, and COG2252, consisting of transporters for adenine, guanine, and/or hypoxanthine that remain unknown with respect to structure-function relationships. We analyzed the COG2252 genes of Escherichia coli K-12 with homology modeling, functional overexpression, and mutagenesis and showed that they encode high affinity permeases for the uptake of adenine (PurP and YicO) or guanine and hypoxanthine (YjcD and YgfQ). The two pairs of paralogs differ clearly in their substrate and ligand preferences. Of 25 putative inhibitors tested, PurP and YicO recognize with low micromolar affinity N6-benzoyladenine, 2,6-diaminopurine, and purine, whereas YjcD and YgfQ recognize 1-methylguanine, 8-azaguanine, 6-thioguanine, and 6-mercaptopurine and do not recognize any of the PurP ligands. Furthermore, the permeases PurP and YjcD were subjected to site-directed mutagenesis at highly conserved sites of transmembrane segments 1, 3, 8, 9, and 10, which have been studied also in COG2233 homologs. Residues irreplaceable for uptake activity or crucial for substrate selectivity were found at positions occupied by similar role amino acids in the Escherichia coli xanthine- and uric acid-transporting homologs (XanQ and UacT, respectively) and predicted to be at or around the binding site. Our results support the contention that the distantly related transporters of COG2233 and COG2252 use topologically similar side chain determinants to dictate their function and the distinct purine selectivity profiles.

modeling on the template of the single structurally known member of the family (the uracil permease UraA) (2) have indicated that their key binding site determinants are similar even though the overall sequence identity is low, ranging from 22 to 28%. On the other hand, the COG2233 homologs retain characteristic sequence motifs that are different in transporters of the poorly studied COG2252 cluster of the family (Fig. 2).
In this work, we provide a first insight on the structure-function relationships of COG2252 members of family NCS2 using the homologs of E. coli K-12 as study paradigms. With respect to the nucleobase uptake-related coding potential, the E. coli K-12 genome includes 10 members of family NCS2 and two members of NCS1. The NCS1 members CodB and YbbW (AllP) are predicted as a cytosine permease and allantoin permease, respectively, from genomic and/or genetic evidence (9,10). The NCS2 members that belong to COG2233 have been identified functionally as uracil (UraA) (11), xanthine (XanQ and XanP) (12), uracil and xanthine (RutG), 3 or uric acid (UacT) permeases (7). The NCS2 members of cluster COG2252 (YgfQ, YjcD, YicO, and PurP) are related in sequence with the fungal and plant AzgA-like adenine-guanine-hypoxanthine transporters (13,14), whereas PurP is annotated as a high affinity adenine transporter based on genetic (15,16) and systems biology evidence (17). Here, we cloned and overexpressed the four COG2252 genes of E. coli and showed that PurP and YicO are high affinity transporters specific for adenine, whereas YjcD and YgfQ are high affinity transporters for hypoxanthine and guanine. Then we subjected PurP and YjcD to site-directed mutagenesis at positions of residues that are conserved and functionally important in homologs of the COG2233 cluster (7, 18 -23) but differ in the coding sequences of COG2252 genes. Our data provide support to the contention that the distantly related purine transporters of the two homology clusters use distinct but topologically equivalent side chains to dictate the binding site function and selectivity.

EXPERIMENTAL PROCEDURES
Materials- [2, H]Adenine (31.8 Ci mmol Ϫ1 ), [2, H]hypoxanthine (27.7 Ci mmol Ϫ1 ),  H]guanine (21.2 Ci mmol Ϫ1 ),  H]xanthine (28 Ci mmol Ϫ1 ), [5, H]uracil (44.9 Ci mmol Ϫ1 ), and [8-14 C]uric acid (57.8 mCi mmol Ϫ1 ) were purchased from Moravek Biochemicals. Non-radioactive nucleobases were from Sigma. Oligodeoxynucleotides were synthesized from BioSpring GmbH. High fidelity DNA polymerase was from Kapa Biosystems. Restriction endonucleases used were from Takara. Horseradish peroxidase (HRP)-conjugated avidin was from Amersham Biosciences. All other materials were reagent grade and obtained from commercial sources. 3 M. Botou, P. Lazou, K. Papakostas, and S. Frillingos, manuscript in preparation. FIGURE 1. Phylogenetic tree of functionally known members of NCS2 family. Multiple protein sequence alignments were performed with ClustalW, and the phylogenetic unrooted tree was constructed by neighbor joining based on the amino acid pairwise distance with the Poisson correction method and bootstrap test of inferred phylogeny using MEGA 4.1. Bootstrap numbers are given in each node. Accession numbers of the NCS2 homologs are given under "Experimental Procedures." Functionally known homologs are denoted by a star that is either filled (information derived directly from transport assays) or empty (information deduced from genetic or genomic studies). Information on RutG, SmLL9, SmLL8, SmVC3, and AcS4X6 is from unpublished data of our research group, 3 and information on the functional profile of YjcD, YgfQ, PurP, and YicO is from the current study. Ura, uracil; Xan, xanthine; UA, uric acid; HX, hypoxanthine; Gua, guanine; Ade, adenine.
DNA Manipulations-Construction of expression plasmids and biotin acceptor domain (BAD)-tagged versions of COG2252 homologs was essentially as described previously for XanQ and XanP (12). The coding sequences of genes were amplified by PCR on the template of genomic DNA prepared from E. coli T184 and transferred to plasmid vector pT7-5 by restriction fragment replacement; BAD-tagged versions were prepared using two-stage (overlap extension) PCR (27) on the templates of pT7-5/purP (or other NCS2 or NCS1 gene as indicated) and pT7-5/xanQ-BAD. For construction of mutants, two-stage PCR was performed on the template of PurP-BAD or YjcD-BAD. The entire coding sequence of all constructs was verified by double strand DNA sequencing (MWG-Biotech).
Growth of Bacteria-E. coli Keio strain JW3692 (⌬purP) or JW4025 (⌬yjcD) harboring the given plasmids was grown aerobically at 37°C in Luria-Bertani medium containing kanamycin (0.025 mg/ml) and ampicillin (0.1 mg/ml). E. coli T184 harboring the given plasmids was grown at the same conditions except that streptomycin (0.01 mg/ml) was used instead of kanamycin. Fully grown cultures were diluted 10-fold, allowed to grow to midlogarithmic phase, induced with isopropyl 1-thio-␤-D-galactopyranoside (0.5 mM) for an additional 1.5 h at 37°C, harvested, and washed with appropriate buffers (see below).
Transport Assays and Kinetic Analysis-E. coli cells were washed twice in potassium phosphate buffer (0.1 M), pH 7.5; normalized to an A 420 of 10 (35 g of total protein/50 l) in the same buffer; and assayed for active transport of radiolabeled  Fig. 1 and of YbbY were aligned with ClustalW, and the part of this alignment referring to the 10 E. coli NCS2 members is presented. Each COG2252 was also analyzed with homology structure prediction using HHpred and threaded on the template of the x-ray structure of UraA. The structure-based alignment of the COG2252 homologs did not differ significantly from the result of the ClustalW alignment except in TM3, TM10, and the C-terminal bundle of the gate domain (␣12-␣14). In these regions, the structure-based alignment was used; the shifts of the initial ClustalW alignment are denoted with forward slashes. Secondary structural elements (␣-helices and ␤-strands) of the UraA template (2) are indicated above the sequence alignment. Highly conserved amino acids are indicated in red, and amino acids that are invariant (or invariant with one exception) in the functionally known COG2252, COG2233, or purine-transporting COG2233 members are shaded in orange. Strictly invariant COG2252 or COG2233 residues are indicated with red asterisks on top or on bottom of the alignment, respectively. The boxed sequence regions represent highly conserved motifs of the COG2233 cluster (E-value Ͻe Ϫ39 ) and of the COG2252 cluster (E-value Ͻe Ϫ29 ) revealed with the program MEME.  (12). For kinetic uptake measurements, initial rates were assayed in JW3692 or JW4025 cells at 5-15 s, and data were fitted to the Michaelis-Menten equation using Prism4 to determine K m and V max values. For ligand competition experiments with PurP, YicO, YgfQ, YjcD, and selected mutants, uptake of [ 3 H]adenine or [ 3 H]hypoxanthine was assayed in JW3692 cells in the absence or presence of unlabeled analogs at the indicated concentration range. Data were fitted to the equation y ϭ B ϩ (T Ϫ B)/(1 ϩ 10 ((log IC50 Ϫ logx)h) ) for sigmoidal dose response (variable slope) where x is the concentration variable, y (activity) values range from T (top) to B (bottom), and h is the Hill coefficient using Prism4 to obtain IC 50 values; in all cases, h was close to Ϫ1, consistent with the presence of one binding site. K i values were calculated from the Cheng-Prusoff equation, where L is the permeant concentration and K m is the value obtained for this permeant, assuming a simple model of competitive inhibition with the binding site of the transporter (17).
Immunoblot Analysis-E. coli cells were washed twice in Tris-HCl (0.05 M), pH 8.0 containing NaCl (0.1 M) and Na 2 EDTA (1 mM); normalized to an A 420 of 30 (2.1 mg of total protein/ml) in the same buffer supplemented with 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride (0.2 mM); and used to prepare membrane fractions by osmotic shock, treatment with EDTA/lysozyme, and sonication as described (12). Membrane fractions prepared from 10-ml cultures of E. coli T184, JW3692, or JW4025 harboring the given plasmids were harvested by ultracentrifugation in an Optima MAX-XP ultracentrifuge (Beckman Coulter), normalized to a protein concentration of 100 g/40 l in Laemmli sample buffer, and subjected to SDS-PAGE (12%) as described (12). Proteins were electroblotted to poly(vinylidene difluoride) membranes (Immobilon-PVDF, Pall Corp.). The BAD-tagged permeases were probed with avidin-HRP. Signals were developed with enhanced chemiluminescence (ECL).

RESULTS
Functional Identification of PurP, YicO, YjcD, and YgfQ-The COG2252 genes ygfQ, yjcD, yicO, and purP were mobilized from the E. coli K-12 genome; transferred to transcriptional control of the lacZ promoter/operator in plasmid vector pT7-5; and induced for overexpression in E. coli host cells. Several K-12 strains were used as E. coli host in these experiments to establish optimal conditions for assaying the uptake of each purine nucleobase. At first, aerobically grown T184 was used at conditions of negligible endogenous activity of xanthine, hypoxanthine, or uric acid uptake (12). However, T184 cells displayed a high background of adenine and guanine uptake. This may be due in part to endogenous expression of PurP, which is known to be linked with the ability of E. coli to utilize adenine as a nitrogen source (15,16). We analyzed the RNA expression profile of the endogenous NCS2 and NCS1 genes and found that aerobically grown T184 cells express both purP and yjcD transcripts, which may account for the high adenine/ guanine background (data not shown). Then we tested a range of single gene knock-out strains (Keio collection) and established that JW3692 (purP knock-out) can be used as a host for adenine uptake assays and JW4025 (yjcD knock-out) can be used for guanine uptake assays, whereas other strains, including JW2850 (⌬xanQ), JW5470 (⌬uacT), JW5636 (⌬yicO), JW5467 (⌬ygfQ), and JW0327 (⌬codB), display significant background for both adenine and guanine uptake. With respect to hypoxanthine uptake, several strains were tested, including T184 (12), JW4025, JW5467, and JW3692, and found to be suitable as a host, but JW3692 was preferred for direct comparisons between adenine and hypoxanthine uptake in the same cell system.
Using the appropriate host strain in each case, we found that all COG2252 constructs can be expressed in the E. coli membrane at high levels and display highly significant purine uptake activities with distinct substrate profiles (Fig. 3). PurP and YicO transport [ 3 H]adenine but not guanine, hypoxanthine, xanthine, uric acid, or uracil, and YjcD and YgfQ transport [ 3 H]guanine and [ 3 H]hypoxanthine but not adenine, xanthine, uric acid, or uracil. No other NCS2 or NCS1 homolog of E. coli can transport adenine, guanine, or hypoxanthine to a significant extent as tested in parallel assays ( Fig. 3 and data not shown). Kinetic analysis revealed that PurP transports adenine with very high affinity (K m 1.0 M), YicO is also a high affinity transporter for adenine (K m 6.5 M), and YjcD and YgfQ are high affinity transporters for guanine (K m 1.6 and 1.8 M, respectively) and hypoxanthine (K m 11.2 and 22.5 M, respectively) ( Table 1).
Analysis of the purine selectivity profiles with competitive inhibition experiments showed that the adenine transporters PurP and YicO can also recognize hypoxanthine albeit with very low affinity (K i 370 and 330 M, respectively) but do not recognize guanine, xanthine, uric acid, uracil, cytosine, or thymine, whereas the hypoxanthine uptake activity of YjcD and YgfQ is inhibited with high affinity by guanine (K i 4 and 3 M, respectively) but not at all by adenine, xanthine, uric acid, or any of the pyrimidine nucleobases (Fig. 4). The specificity profiles were investigated further by assaying transport of the appropriate substrate in the presence or absence of a series of analogs ( Fig. 4 and Table 2). PurP and YicO cannot recognize guanine, 6-mercaptopurine, or 3-methyladenine and are competed with low affinity by 7-methyladenine or 9-methyladenine, but they do recognize with high affinity N 6 -benzoyladenine (K i 2 and 19 M, respectively), 2,6-diaminopurine (K i 9 and 71 M, respectively), or purine (K i 3 and 21 M, respectively). YjcD and YgfQ recognize a broader range of guanine and purine analogs, including 1-methylguanine (K i 33 and 105 M, respectively), 6-thioguanine (K i 4 and 4 M, respectively), 8-azaguanine (K i 44 and 154 M, respectively), and 6-mercaptopurine (K i 18 and 57 M, respectively). Overall, there is a clear distinction in specificity between the two adenine transporters and the two guanine/hypoxanthine transporters (no analog is recognized with high affinity by both transporter types), whereas PurP and YjcD display higher affinity than their iso-functional paralogs with respect to substrate or analogs tested ( Table 2).
Delineation of Mutagenesis Targets for Structure-Function Analysis of PurP and YjcD-The four transporters characterized here are structurally homologous to the well studied NCS2 proteins UraA (2), XanQ (6), UacT (7), and UapA (8) but belong to a separate cluster of orthologs for which studies on the role of individual side chains are not yet available. To initiate a rationally designed mutagenesis study in homologs of this cluster, we selected as replacement targets residues that are functionally important in the well studied NCS2 proteins and fall in characteristic sequence motifs of the family. Highly conserved sequence motifs of the functionally known COG2233 homologs fall in TM1, TM3, TM5, TM8, TM9, TM10, and TM12 (Fig. 2). Amino acid residues delineated as important from the previous mutagenesis studies of purine-transporting COG2233 members are at the motifs of TM1, TM3, TM8, TM9, and TM10. These functionally linked positions are either invariably conserved or falling to distinct conservation patterns (   DECEMBER 27, 2013 • VOLUME 288 • NUMBER 52 side chains are generally not conserved in the COG2252 homologs. Although different, the COG2252 side chains at the corresponding positions are strongly conserved or invariant (Fig. 5) and fall at highly conserved sequence regions as well (Fig. 2).

Characterization of the COG2252 Transporters of E. coli
The relevant amino acid side chains of PurP and YjcD (Thr-38/35, Ala-91/88, Asp-267/271, Thr-271/275, Asp-298/302, Ile-317/321, Glu-318/322, and Ser-319/323) were subjected to site-directed replacements with Ala, same character or similar character amino acids, and/or amino acids occupying the corresponding positions of COG2233 members (see below). Homology modeling of PurP and YjcD on the structural template of UraA (2) showed that the selected side chains fall at the putative binding site region or at the periphery (Fig. 6). Notably, the COG2252 homologs appear to form a more relaxed helical structure at the beginning of the ␣-helix of TM10 (Fig. 6) so that the 3-amino acid helical turn of Ile-317/321, Glu-318/322, and Ser-319/323 corresponds to a 4-amino acid helical turn in the topology of XanQ (Ala-323, Gln-324, Asn-325, and Asn-326) or other COG2233 transporters. The significance of this observation is discussed in the Discussion section.
The specificity profile of each active mutant was determined and compared with wild-type PurP in adenine uptake inhibition assays ( Table 2). It was found that mutants A91S, T271S, and D298E displayed significantly lower affinities than wild type (higher K i values) for both the high affinity ligands N 6 -benzoyladenine, purine, and 2,6-diaminopurine and the low affinity ligand 9-methyladenine, whereas mutant T271A displayed significantly lower affinity than wild type for N 6 -benzoyladenine and purine but nearly wild-type affinity for the other analogs. Less significant differences were observed with the other E. coli JW4025 or JW3692 expressing the corresponding constructs was assayed for initial rates of ͓ 3 H͔guanine (0.1-40 M) or ͓ 3 H͔adenine (0.04 -40 M) and ͓ 3 H͔hypoxanthine (0.1-100 M) uptake at 5-15 s at 25°C. Negative control values obtained from JW4025 or JW3692 harboring vector pT7-5 alone were subtracted from the sample measurements in all cases. Kinetic parameters were determined from non-linear regression fitting to the Michaelis-Menten equation using Prism4; values represent the means of three determinations with standard deviations shown. Significant K m differences of mutants from wild type are underlined and indicated in bold. ND, assays were performed, but kinetic values were not determined because of very low uptake rates.
The specificity profile of each active mutant was examined using hypoxanthine uptake inhibition assays (Table 2). It was found that mutants T35A, I321E, A88G, and A88S deviate from the wild-type profile with T35A displaying lower affinity (higher K i values) for guanine and its 1-methyl, 6-thio, or 8-aza derivatives; A88S displaying lower affinity for guanine, 1-methylguanine, and 8-azaguanine; A88G showing lower affinity for guanine and 1-methylguanine but higher affinity for 6-thioguanine and 6-mercaptopurine; and I321E displaying higher affinity for guanine but lower affinity for 1-methylguanine, 8-azaguanine, and 6-mercaptopurine ( Table 2). The low affinity of mutants T35A, A88G, and A88S for guanine as revealed in the hypoxanthine inhibition assays is corroborated by the high K m values for guanine uptake found for these mutants on guanine uptake kinetic analysis (Table 1). It is clear from our data that the replacement of Ala-88 with Gly reverses selectivity of YjcD from a guanine-preferring profile (wild-type K m is 1.8 M for guanine versus 11.2 M for hypoxanthine) to a hypoxanthinepreferring profile (A88G K m is 50.1 M for guanine versus 4.2 M for hypoxanthine), whereas the amino acid replacement in mutants T35A and A88S results in lower affinity (higher K m ) for both guanine and hypoxanthine.
Taken together, the YjcD mutagenesis data revealed that Asp-271, Thr-275 (TM8), Glu-322, and Ser-323 (TM10) are essentially irreplaceable for function (all conservative sidechain changes at these positions inactivate or result in very low activity that does not permit kinetic analysis), and Thr-35 (TM1) and Asp-302 (TM9) are replaceable but subject to constraints with respect to the allowed changes (at least one conservative side-chain change inactivates). Replacement of Thr-35 (TM1) with Ala resulted in decreased affinity for substrates and/or ligands, and replacements of Ala-88 (TM3) and Ile-321 (TM10) with Gly and Glu, respectively, altered the selectivity profile resulting in higher affinity for hypoxanthine and lower affinity for guanine (A88G) or, reciprocally, higher affinity for guanine and lower affinity for hypoxanthine (I321E).

DISCUSSION
We have shown in this study that the genome of E. coli K-12 encodes two homologous but functionally distinct pairs of high affinity transport proteins for the uptake of adenine versus guanine/hypoxanthine, namely the two adenine-specific permeases PurP and YicO and the two guanine/hypoxanthine permeases YjcD and YgfQ. The four proteins belong to cluster COG2252 of the evolutionarily broad family NCS2 and are closely related to each other, ranging in sequence identity from 81% (between YjcD and YgfQ) to 34% (between YjcD and PurP). Previous reports on the functional profile of these proteins are FIGURE 5. Conservation pattern at sequence motifs of NCS2 family that contain important residues. The full-length sequences of NCS2 members were aligned using ClustalW and HHpred (Fig. 2), and the part of this alignment referring to motifs of TM1, TM3, TM8, TM9, and TM10 is shown. Residues that are invariably conserved in COG2233 are highlighted in red. Residues that display distinct conservation patterns are shown in orange (functionally irreplaceable in the purine-transporting members XanQ, UacT, and UapA) (6 -8), purple (implicated in substrate selectivity from the studies of XanQ and UacT) (7, 23), or light blue. The corresponding residues of COG2252 transporters are shown in blue (invariably conserved) or light blue. The residue numbers on top and on bottom of the sequences refer to PurP and XanQ, respectively. The three columns on the right show the total number of residues of each homolog, the sequence identity score between each homolog and the structural prototype of the family (UraA) (based on HHpred), and the substrate profile of each homolog (square brackets are used when information is derived only from genetic/genomic studies). Information on RutG and SmLL9 is from unpublished data of our research group 3 , and information on the functional profile of YjcD, YgfQ, PurP, and YicO is from the current study. Ura, uracil; Xan, xanthine; UA, uric acid; HX, hypoxanthine; Gua, guanine; Ade, adenine.
Values represent the means of three determinations with standard deviations Ͻ20%. High affinity competitors are indicated in bold, and significant differences from wild type are underlined. L, the permeant concentration; G, guanine; HX, hypoxanthine; 3-MA, 3-methyladenine; 7-MA, 7-methyladenine; 9-MA, 9-methyladenine; N 6 -bA, N 6 -benzoyladenine; 2,6NP, 2.6-diaminopurine; 2-mP, 2-mercaptopurine; 6-mP, 6-mercaptopurine; A, adenine; 1-MG, 1-methylguanine; 6-SG, 6-thioguanine; 7-MG, 7-methylguanine; 8azaG, 8-azaguanine. In addition to the data shown in the table, the hypoxanthine uptake activity of the YjcD mutants was found not to be inhibited by xanthine, 2-thioxanthine, or N 6 -benzoyladenine. not available except for PurP. The purP gene product has already been annotated in the databases as an adenine transporter based on genetic studies with E. coli mutants that were defective in the uptake and utilization of adenine (15,16). Those genetic studies have also indicated that E. coli uses a different protein system for the uptake of guanine and hypoxanthine (34), but such a system remains unidentified to date. In a very recent report that appeared almost concurrently with the present study, Kozmin et al. (35) suggested that YjcD is a primary importer for modified purine bases, and its physiological role might, in fact, be related with the uptake of guanine and hypoxanthine. However, their suggestions were based on growth kinetics and suppression of the toxic effect of analogs in E. coli mutants and not on direct measurements of transport activities. Our current results clearly indicate that the guanine/ hypoxanthine uptake system might be represented by either YjcD or YgfQ, whereas the adenine uptake system might be represented by either PurP or YicO. Analysis of the endogenous uptake activity in various single gene knock-out Keio strains implies that E. coli K-12 may use only YjcD and PurP, respectively, under the applied aerobic growth conditions. Both PurP (17,36) and YjcD (36) have been shown to belong to the PurR regulon, which plays a critical role in the transcriptional regulation of purine metabolism in enterobacteria (17).
With respect to the functional distinction between YjcD (or YgfQ) and PurP (or YicO), the two pairs of purine transporters not only differ in their basic substrate preferences but also display no significant overlap in their purine recognition profiles (Fig. 4). Of 25 purines or analogs tested, the two adenine trans-porters were found to recognize only analogs modified at position 2 or 6 of the adenine ring as high affinity competitors (N 6benzoyladenine, purine, and 2,6-diaminopurine). The two guanine/hypoxanthine transporters were found to recognize a broader range of structures, including modifications at positions 1, 2, and 6 (1-methylguanine, 6-thioguanine, and 6-mercaptopurine) and at the imidazole moiety (8-azaguanine). None of the analogs tested can serve as a high affinity ligand for both types of transporters. In view of the striking sequence similarity of the two types of transporters at critical sites (Fig. 2), it would be interesting to examine the structure-functional basis of this selectivity split. It is most interesting that the highly homologous AzgA-like transporters of the fungus A. nidulans (13) and the plant A. thaliana (14) do not display such a selectivity split but are able to use both adenine and guanine (or hypoxanthine) as substrates.
The detailed kinetic analyses of the four COG2252 transporters revealed some significant differences in affinity and specificity between the two isofunctional homologs in each pair (Tables 1 and 2). YicO displays 5-10-fold lower affinity for adenine and for any of the competing high affinity analogs relative to PurP. YgfQ displays 2-3.5-fold lower affinity for hypoxanthine and for the majority of competing analogs relative to YjcD but not for guanine and 6-thioguanine, which are recognized with almost equal affinities by both transporters. The latter observation indicates that YgfQ has a different specificity profile from YjcD in which the amino group of guanine at purine position 2 plays a more important role for the binding affinity. From the comparison of the sequences of YjcD and YgfQ (81%  DECEMBER 27, 2013 • VOLUME 288 • NUMBER 52 identical) and of PurP and YicO (73% identical) it is apparent that the observed specific variations in substrate preference or affinity are associated with a limited set of amino acid changes.

Characterization of the COG2252 Transporters of E. coli
In summary of the first part of our results, we have identified the analytical functional profile of four purine transporters of E. coli that belong to homology cluster COG2252. Based on our current experimental evidence, we suggest alternative, function-based designations for these genes other than those that currently exist. 4 The existing names do not reflect function or reflect only the initial functional observations in the case of the purine permease PurP (37). YjcD may be renamed to GhxP (guanine/hypoxanthine permease), and PurP may be designated with the alternative name AdeP (adenine-specific permease) to emphasize the functional split. The isofunctional paralogs YgfQ and YicO may be renamed to GhxQ and AdeQ, respectively.
The second part of our results refers to the identification of functionally important residues of PurP (AdeP) and YjcD (GhxP) by mutagenesis. The mutagenesis study was designed on the basis of the structural homology between COG2252 and COG2233 transporters. Despite the low sequence identity between the two clusters (ranging from 11 to 16% on ClustalW analysis), functionally important residues of the previously studied COG2233 members that are predicted to fall at conserved motifs of the binding site region correspond to distinctive but highly conserved amino acids in the COG2252 homologs (Figs. 5 and 6). The site-directed replacement analysis of these amino acids in PurP and YjcD revealed key functional similarities with topologically equivalent residues of the COG2233 transporters as summarized in Figs. 9 and 10.
Amino acid residues delineated as functionally irreplaceable in both PurP and YjcD (Asp-267/271 and Glu-318/322) correspond to functionally irreplaceable residues in the topologies of the purine-transporting COG2233 homologs UacT, XanQ, and UapA (Glu-270/272/356 and Asn-319/325/409). In addition, one of these two amino acid positions is implicated directly with substrate binding in the structurally known UraA (Glu-241) (2) and, based on docking analysis, in UapA (Glu-356) (8). The other of the two positions has been proposed to be a binding site residue in XanQ (Asn-325) based on site-directed alkylation analysis (21) and appears to play an irreplaceable role implicated directly or indirectly (8) in binding in purine-transporting homologs but not in UraA (2).
On the other hand, an adjacent Gln/Glu residue of the TM10 motif that appears to be irreplaceable and directly associated with binding in all COG2233 members studied thus far (2, 6, 8) 4 K. E. Rudd, personal communication. is absent from the COG2252 homologs based on the structuretopology predictions (Fig. 9). This residue is invariably conserved as a Gln in the group of xanthine and/or uric acid (2-oxy purine) transporters of the family and as a Glu in uracil transporters. The relevant residue of UraA (Glu-290) is irreplaceable for binding and has been shown to bind uracil in the x-ray structure of UraA via two hydrogen bonds, one of which is with the carbonyl oxygen at position 2 of uracil (2). In the xanthinetransporting homologs XanQ (18) and UapA (37), replacement of the relevant Gln residue leads to impairment of xanthine binding probably due to disruption of an essential hydrogen bond between Gln-324/408 and the carbonyl oxygen at position 2 of xanthine as indicated recently from substrate docking analysis (8). Because the carbonyl oxygen at purine position 2 is characteristic of the structures of 2-oxy purines (xanthine and uric acid) and the xanthine-selecting COG2233 transporters do not recognize hypoxanthine (2-non-oxy xanthine), it is tempting to speculate that the lack of the essential Gln residue of TM10 in COG2252 transporters is associated with their preference for 2-non-oxy purines and their inability to recognize xanthine. It is also relevant that none of the YjcD mutants tested in this study displayed any ability to bind xanthine ( Table 2), and that, inversely, none of the many mutants tested in the course of the systematic analysis of XanQ (6) display any significant ability to bind hypoxanthine. It would be interesting to investigate whether engineering of appropriate Gln insertions/ deletions at the crucial ␣10 region might shift the transporter function from a hypoxanthine-to a xanthine-selective profile or vice versa.
Amino acid residues found in this study to be associated with selectivity mutants in YjcD (Ala-88) or in both PurP and YjcD (Ile-317/321) correspond to residues that have also been linked with selectivity mutants in the purine-transporting COG2233 homologs UacT, XanQ, and/or UapA (Thr-100/Asn-93/Ser-156 and Ser-317/Ala-323). These two amino acid positions are predicted to be at the vicinity of the purine binding site and may be associated with binding through hydrogen bonding of their side chain with a purine-binding residue (Asn 93 -Glu 272 in XanQ) (23), hydrogen bonding of a main-chain atom with substrate (Ala-407 in UapA) (8), or a more remote indirect interaction depending on the side-chain occupancy. Interestingly, the selectivity-associated replacements of COG2233 transporters at these positions affect primarily the imidazole moiety of the ring (Fig. 9) and allow selectivity shifts between xanthinetransporting or uric acid-transporting homologs and more promiscuous, dual substrate transporters. Thus, the relevant  DECEMBER 27, 2013 • VOLUME 288 • NUMBER 52 mutants of the uric acid permease UacT allow highly efficient transport of xanthine (T100A and S317A) (7), mutants of the xanthine-specific XanQ allow transport of uric acid (N93S and N93A) (23) or high affinity recognition of the non wild-type ligand 8-methylxanthine (N93S, N93A, and A323S) (7,23); and mutant S156A of the dual substrate xanthine and uric acid permease UapA shifts the profile in a xanthine-selective direction (38). In contrast, selectivity-associated mutants of COG2252 transporters affect primarily the amino group of guanine position 2 (YjcD mutants A88G and I321E) or the amino group of adenine position 6 (PurP mutant I317A) at the pyrimidine moiety (Fig. 9). It is notable as well that none of the selectivity mutants of YjcD or PurP allow significant promiscuity with respect to the recognition of different purines or relaxation of the selectivity split between the adenine (PurP and YicO) and the guanine/hypoxanthine recognition profile (YjcD and YgfQ) ( Table 2).

Characterization of the COG2252 Transporters of E. coli
A number of amino acid residues at the periphery of the presumed binding site of YjcD or PurP are not irreplaceable for activity or for the proper selectivity profile but are subject to constraints with respect to the allowed side-chain changes; several mutations at these positions yield low substrate affinity (YjcD T35A and PurP A91S, T271A, T271S, and D298E) or impairment of transport (YjcD T35H and D302N and PurP  T38H, T271D, T271N, and D298N). Furthermore, two peripheral residues of YjcD (Thr-275 and Ser-323) were irreplaceable ( Fig. 9). Similar constraints have been found with corresponding residues of the COG22333 transporters albeit to various extents of significance. For example, the invariant His of COG2233 homologs (corresponding to Thr-35/38) in TM1 was delineated as important for high affinity binding of substrate in XanQ (20), strictly irreplaceable for transport in UacT (7), and important for the proper folding and targeting to the plasma membrane in UapA (39). The less well conserved Asp/Asn/Glu of TM9 (corresponding to Asp-302/298) has been identified as irreplaceable for transport in XanQ, UacT, and UapA, but its role is obviously less critical in the COG2252 homologs (Fig.  10). The weak consensus residue Asp/His/Met of TM8 (corresponding to Thr-271) has been linked with constraints in the allowed side-chain replacements that are more stringent in XanQ (Asp-276) where a carboxyl group is needed for high uptake activity and an Asp is essential for both the selectivity and the pH profile (22). In COG2252 transporters, the corresponding residue is either subject to severe replacement constraints (PurP) or irreplaceable (YjcD). Finally, the weak consensus Asn/Ile/Val of TM10 (corresponding to Ser-323) is subject to side-chain volume constraints in XanQ (18) and UacT (7) that are probably associated with steric hindrance between TM8 and TM10 at the periphery of the binding site (7), whereas the corresponding Ser residue of COG2252 transporters is either non-essential for transport (PurP) or functionally irreplaceable (YjcD). This heterogeneity in the roles of residues  6) were displayed with PyMOL. Shown are ␣-helical parts of the core domain that are predicted to form the substrate coordination shelter (2). Important residues of YjcD or PurP delineated in this study as well as the corresponding residues of XanQ (6) and UacT (7) are shown as thick sticks and color-coded (red, irreplaceable for function; purple, replacements lead to decreased substrate affinities; green, replacements lead to change of specificity; blue, constraints with respect to the allowed changes; grayscale, remaining positions). Residues delineated as irreplaceable in all four transporters are highlighted with red ovals. Residues delineated as crucial for the selectivity profile in at least three of the transporters are highlighted with green ovals. The chemical structures of the major transported substrates in each case are also shown. The light green circles highlight sites of the purine ring associated with the observed selectivity changes and the corresponding selectivity-associated mutants in each case.
that are peripheral to the binding site should be attributed to elaborate differences in the mechanism of the individual purine transporters, which remain to be elucidated.
Although YjcD and PurP recognize different substrates, their functionally important residues identified in this study ( Fig. 9) are almost identical. In addition, our mutagenesis data did not reveal any mutations leading to change in selectivity from the one transporter type to the other. These observations imply that other, as yet unidentified COG2252 residues that differ between the two transporters might have an important contribution to the substrate recognition preferences. Given that key amino acids of the presumed binding site are identical and play a similar role in both transporters (Fig. 9), it might be assumed that residues that contribute decisively to the selectivity split are mostly at the periphery and affect the recognition of substrate through indirect side-chain interaction effects. On the other hand, the coordination of the 2-non-oxy purine bases in the binding site of COG2252 transporters might involve additional interactions that differentiate between PurP and YjcD but are not evident from the current data or play a minor role in the COG2233 homologs. To address such alternatives, further mutagenesis studies using a more thorough investigation of conserved motifs of the family (Fig. 2) in conjunction with substrate docking analyses will be needed. The ongoing Cysscanning analysis of XanQ and other systematic studies of COG2233 transporters (6 -8) will probably provide a comprehensive data set of candidate mutagenesis targets to this end.
In conclusion, we have identified several amino acid positions with important roles in the adenine or guanine/hypoxanthine transporters of E. coli and shown that functionally irreplaceable or selectivity-linked residues of these transporters correspond to residues with similar roles in the distantly related xanthine or uric acid transporters of family NCS2. This line of evidence provides support to the contention that purine transporters of the separate, weak sequence homology clusters COG2233 and COG2252 of family NCS2 may use a set of topologically similar side chains to dictate their function and selectivity preferences, reflecting an evolutionarily "deep" homology in their binding site structures.
To our knowledge, this is one of the first reports showing functional homologies of individual side chains in distantly related transport proteins that share weak similarity in linear sequence alignments. Structure-based alignments are more prompt to reveal targets for informative site-specific mutagenesis in such cases as shown both in the current study and in a recent seminal study on the structural motif homologies between the fucose permease FucP and the lactose permease LacY of the major facilitator superfamily (40).
Acknowledgment-We thank Dr. Kenneth Rudd for suggestions of appropriate function-based designations for the transporter genes characterized in this study. FIGURE 10. Roles of important amino acid residues in various NCS2 transporters. Information was derived from the x-ray structure analysis of UraA (2), previous mutagenesis studies of the COG2233 homologs UapA (8,37,38), UacT (7) and XanQ (18 -23) and the current study of YjcD and PurP. Major substrates of each transporter are displayed on top. Shown are the positions of amino acid residues analyzed in the current study and corresponding positions of the COG2233 transporters. Important amino acids of UapA, UacT, XanQ, YjcD, and PurP are color-coded (red, irreplaceable for function; purple, replacements lead to decreased substrate affinities; green, replacements lead to change of specificity; blue, constraints with respect to the allowed changes; orange, essentially irreplaceable with replacements leading to inactivation or very low activity; grayscale, remaining positions) (see text for further details). Side chains of UraA that form direct hydrogen bonds with substrate are encircled and shown in red (Glu-241 and Glu-290). Interrupted line circles indicate His-245, which was proposed to hydrogen bond with substrate indirectly through a water molecule, and Gly-289, which hydrogen bonds with substrate through the main-chain nitrogen (2). Asn-291 and His-24, which hydrogen bond to each other at the periphery of the binding site, are shown in boxes. Positions of residues shown to be irreplaceable in both COG2233 and COG2252 homologs are shaded in orange. Positions of residues shown to be important for substrate specificity/selectivity in both COG2233 and COG2252 homologs are shaded in green.