Resculpting the binding pocket of APC superfamily LeuT-fold amino acid transporters

Amino acid transporters are essential components of prokaryote and eukaryote cells, possess distinct physiological functions, and differ markedly in substrate specificity. Amino acid transporters can be both drug targets and drug transporters (bioavailability, targeting) with many monogenic disorders resulting from dysfunctional membrane transport. The largest collection of amino acid transporters (including the mammalian SLC6, SLC7, SLC32, SLC36, and SLC38 families), across all kingdoms of life, is within the Amino acid-Polyamine-organoCation (APC) superfamily. The LeuT-fold is a paradigm structure for APC superfamily amino acid transporters and carriers of sugars, neurotransmitters, electrolytes, osmolytes, vitamins, micronutrients, signalling molecules, and organic and fatty acids. Each transporter is specific for a unique sub-set of solutes, specificity being determined by how well a substrate fits into each binding pocket. However, the molecular basis of substrate selectivity remains, by and large, elusive. Using an integrated computational and experimental approach, we demonstrate that a single position within the LeuT-fold can play a crucial role in determining substrate specificity in mammalian and arthropod amino acid transporters within the APC superfamily. Systematic mutation of the amino acid residue occupying the equivalent position to LeuT V104 titrates binding pocket space resulting in dramatic changes in substrate selectivity in exemplar APC amino acid transporters including PAT2 (SLC36A2) and SNAT5 (SLC38A5). Our work demonstrates how a single residue/site within an archetypal structural motif can alter substrate affinity and selectivity within this important superfamily of diverse membrane transporters. Electronic supplementary material The online version of this article (doi:10.1007/s00018-017-2677-8) contains supplementary material, which is available to authorized users.


Introduction
For any eukaryotic or prokaryotic cell to remain viable, it must express a large and diverse complement of membrane transport proteins to enable import and export, between the cell and the local environment, of all material vital for life. Carrier-mediated, transmembrane amino acid transport is essential in neurotransmission, nutrient absorption from diet, osmoregulation, and in the supply of components for protein synthesis, nitrogen metabolism, cell growth, energy production, and conversion. Thus, each cell type possesses a unique array of amino acid transporters to permit optimal physiological performance within any given milieu.
The largest collection of amino acid transporters across all forms of life is found within the Amino acid-Polyamine-organoCation (APC) superfamily [Transporter Classification DataBase (TCDB)] [1,2]. Substrates include the 20 proteinogenic amino acids (which differ in size, shape, hydrophobicity, polarity, and charge on the α carbon sidechain), non-proteinogenic α amino acids (e.g., betaine and ornithine), and unbranched chain amino acids and analogues where the amino group is in the β or γ position (e.g., taurine, GABA). Carriers vary greatly in substrate specificity with some transporting a single type of amino acid (with extreme selectivity), others accepting almost all amino acids (with varying levels of discrimination), with most falling somewhere in between. This commonality in general function (amino acid transport), but heterogeneity in substrate selectivity, provoked this investigation into the molecular basis of carrier diversity.
LeuT-like structures contain a core of ten transmembrane (TM) spans, organised into a 5 + 5 inverted structural repeat, with TM1, 3, 6, and 8 forming the central binding pocket of each carrier [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17]. In the original LeuT structure, the aliphatic side-chain of the substrate sits within a hydrophobic pocket formed from the side-chains of residues in TM3, 6, and 8 [3]. Comparative modelling of LeuT with more than 300 prokaryotic and eukaryotic NSS sequences identified that the residues that interact with the substrates' side-chains in the deeper regions of the binding pockets are not conserved [26]. Bioinformatics analyses, using functional site prediction strategies, anticipated key functional sites within the NSS family and correctly predicted 31/34 substrateinteracting residues in the LeuT structure [27]. Residues in the three "non-predicted" positions in LeuT all form van der Waals' contacts with the substrate side-chain [3]. LeuT was subsequently crystallised bound to a series of amino acids with increasing side-chain size [28]. When LeuT is locked in the outward-open substrate-bound conformation, by interaction with the large indole ring of the non-transported inhibitor tryptophan [28], V104 in TM3, one of the non-conserved residues identified in the bioinformatics analyses [26,27], and the focus of the current investigation, occupies a deep position below the indole ring.
An ambition of global industry is to use in silico methodology to predict drug delivery, action of novel pharmaceuticals, and utilisation and efficiency of new agrichemicals. Ultimately, to achieve such an understanding of the roles of individual transporters in these essential functions, it is necessary to determine how the identity of amino acid residues coordinating the substrates within the binding pocket of each APC superfamily carrier defines substrate specificity. Here, we investigated the hypothesis that amino acid residues occupying the equivalent position to V104 in the LeuT-fold of APC carriers are critical in governing substrate specificity. The basis of substrate selectivity in LeuT-fold APC superfamily amino acid transporters was investigated using a series of wild-type and mutated transporters from the important Amino Acid/Auxin Permease (AAAP, 2.A.18) family [2,29], which are expressed ubiquitously in plants, animals, yeast, and fungi. The relationship between different amino acid transporter families within the APC superfamily was investigated by computational phylogenetic methodology. Structural models were constructed based upon the outward-occluded substrate-bound conformations of the APC superfamily members LeuT, AdiC, and Mhp1 [3,9,25]. Site-directed mutagenesis and functional measurements of transporter activity were used to validate the structural models in multiple AAAP transporters. Excellent agreement was observed between model predictions and functional activity. In addition, re-evaluation of published data on non-amino acid transporting APC carriers suggests that the site investigated has an importance in defining substrate specificity beyond amino acid transporters. Taken together, our results demonstrate how a single residue/site within an archetypal    Table S1). Values at branches represent posterior probabilities (scale bar: average number of substitutions per site). The majority of APC families are well supported as clans. Three sequences branch separately from their annotated family but are located close to the base of the tree with very weak support, suggesting that their position is an artefact due to long-branch attraction structural motif alters substrate affinity and selectivity in an extensive, widely distributed and important superfamily of cellular transport proteins.

Phylogeny
Sequences for the APC superfamily tree were retrieved from the TCDB [1]. An initial selection of sequences included TCDB entries (denoted TC#) with structural data plus TCDB entries with strong functional characterisation. Additional sequences were added to improve both the taxonomic diversity of the sampling and to cover additional APC families. Human sequences for SLCs 5, 6, 7, 11, 12, 32, 36, and 38 were retrieved via the Bioparadigms database [20]. Additional sequences (rat, mouse, rabbit, Drosophila melanogaster, Aedes aegypti, and Acyrthosiphon pisum) were from NCBI and were included for reference. All sequences were aligned using MUSCLE [30]. The alignment was trimmed using TrimAl v1.4 [31], with trimming parameters defined by the automated1 option. Phylogenies were generated in PhyloBayes [32] using the CAT20 model [33]. Trimmed (and untrimmed) sequence alignments associated with the phylogenies are available at figshare using the link: https://figshare.com/s/378479b6958df7816b1b.

Plasmid constructs and site-directed mutagenesis
The use of plasmid constructs for rat PAT2 (SLC36A2) and human SNAT5 (SLC38A5) in pSPORT1, and mouse PAT1 (SLC36A1) in pCRII-TOPO, has been described previously [39][40][41]. The Drosophila transporter CG1139 was purchased from the Drosophila Genomics Resource Centre (Indiana University, USA) and expressed in pGH19 (gift from G. Robertson, University of Wisconsin, USA). Site-directed mutagenesis was performed using the QuikChange Lightning kit, according to the manufacturer's instructions. The PCR cycling parameters used were an initial 2 min incubation at 95 °C, followed by: 18 cycles at 95 °C denaturation (20 s); 68 °C annealing (10 s); 68 °C extension (30 s kb −1 plasmid); and a final extension at 68 °C for 5 min. Parental plasmid DNA was digested with DpnI and the PCR reaction product used to transform XL-10 Gold cells. Oligonucleotides were designed using the QuikChange primer design tool. Mutations were verified by sequencing (GATC Biotech, London, UK) of the entire open reading frame.

Radiolabelled amino acid transport assays
Radiolabelled amino acid transport (uptake) assays were performed, as described previously [42]. In brief, oocytes were incubated at room temperature (approximately 22 °C) in transport solution [100 mM choline chloride (or 100 mM NaCl for solutions requiring Na + ), 2 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM MES, or HEPES adjusted to the required pH with Tris base] containing [ 3 H]-or [ 14 C]-labelled compound (1-5 μCi ml −1 ). Assays were performed at pH 5.5, in the absence of extracellular Na + , unless stated otherwise. Oocytes were then washed three times in ice-cold transport solution and lysed in 10% SDS, and the associated radioactivity determined by liquid scintillation counting.

Two-electrode voltage clamp (TEVC)
Individual oocytes, which had been injected with either CG1139 cRNA or water (control), were clamped at − 60 mV and superfused with Na + -free, pH 5.5 transport solution (see above), as described previously [42]. Various amino acids were added for 1 min and the associated inward positive current measured using a Geneclamp 500 amplifier, Digidata 1200, and pClamp software (Molecular Devices, Sunnyvale, CA). Compound-associated currents were calculated by averaging the current during the last 15 s of exposure and subtracting the average current recorded in the 15 s preceding exposure (baseline). Data were analysed using Clampfit 8.2.

Data and statistical analysis
Data are mean ± SEM. and are typically expressed as pmol. oocyte −1 (uptake duration) −1 . Transporter-specific uptake was calculated as uptake into transporter-expressing oocytes after subtraction of uptake into water-injected oocytes (measured under identical conditions). Michaelis-Menten kinetics were fitted using GraphPad Prism 6. Comparisons  of mean values were made by one-way or two-way analysis of variance (ANOVA), as appropriate, with Tukey's or Sidak's multiple comparisons post-tests (GraphPad Prism 6). ANOVAs are two-way unless stated otherwise.

Results
LeuT-fold transporters are highly divergent in their overall amino acid sequences The APC superfamily consists of 18 transporter families [2], 14 of which are predicted to possess the LeuT-fold 5 + 5 inverted structural repeat. Phylogenetic sequence analysis of prokaryotic and eukaryotic representative members of the fourteen families indicates that most are well supported as putative monophyletic groups or clans [43] (Fig. 1, supplementary Table S1). The relationships between individual families, however, are generally poorly supported highlighting the divergent nature of the superfamily as a whole (which is also reflected in their differing functions  [3-12, 14, 15] are denoted in bold in Fig. 1 (see also supplementary Table S1). There appears to be no strong relationship between function and phylogeny underlining the importance of investigating sequence, structure and function, in an integrated manner to understand transport specificity.
To investigate the importance of the LeuT V104-equivalent residue as a molecular determinant of substrate specificity in the APC superfamily we chose, therefore, an exemplar amino acid carrier with which to begin our analyses. The mammalian proton/amino acid cotransporter PAT2 (SLC36A2) [39,40] (asterisk in Fig. 1, see also Table 1) is a member of the Amino Acid/Auxin Permease (AAAP, 2.A.18) family [2,29] of transporters which are found in plants, animals, yeast, and fungi. PAT2 was chosen to sample an area of the phylogenic tree which has, to date, been underexplored and to investigate the generality of the observations in relation to the APC superfamily as a whole. PAT2 is a tractable transport protein, amenable to mutagenesis and functional measurements. It has a narrow and well-defined substrate selectivity that appears to be restricted severely by side-chain size [39,40], identifying it as a suitable candidate for functional and mutational analyses. In humans, PAT2 contributes to amino acid transport in diverse cell types such as renal proximal tubule cells, neurones, and adipocytes. Mutations in PAT2 leading to defective function contribute to the human disorders of iminoglycinuria (Online Mendelian Inheritance in Man (OMIM) 242600) and hyperglycinuria (OMIM 138500) [44].

The Drosophila melanogaster transporter CG1139 has broader substrate specificity than mammalian SLC36 carriers
The transport of proteinogenic α amino acids via PAT2 is limited to proline, glycine, and alanine [39,40,42,47,48]. CG1139 was previously shown to transport alanine (inhibited by glycine and proline) [49]. When modelled upon the LeuT amino acid-bound structure, CG1139 I149 occupies the same position as PAT2 F159 and LeuT V104 (Fig. 2b,  c). The reduced volume [50] of the I149 side-chain (relative to F159) identifies CG1139 as a suitable model transporter for comparison with PAT2 for the investigation of substrate selectivity. The difference in volumes occupied by the two residues led us to predict that CG1139 would transport α amino acids with larger side-chains than those able to access PAT2.
CG1139 transports the prototypical SLC36 proteinogenic substrates (alanine, proline, and glycine) and shares other SLC36-like characteristics (Fig. 3). The striking difference shown here is that CG1139 also accepts l-amino acids with longer side-chains including serine, cysteine, and α-aminobutyric acid (α-ABA) (Fig. 3d, f-h). To test whether F159 in PAT2 limits access of extended substrate side-chains compared to the less bulky CG1139 I149 (Fig. 2c), we replaced F159 in PAT2 with isoleucine.

The residue occupying position 159 in PAT2 determines accessibility of the substrate side-chain within the binding pocket
The large aromatic phenylalanine in PAT2 was replaced with the equivalent but smaller isoleucine from CG1139 (Figs. 2, 4). Like CG1139, but not wild-type PAT2, competition experiments demonstrate that serine, α-ABA and cysteine can now access the binding pocket and inhibit PAT2-F159I transport (Fig. 4a-e). Larger side-chains are excluded from both PAT2 and PAT2-F159I (Fig. 4b). The selectivity change in PAT2-F159I is due to improved affinity for serine, α-ABA, and cysteine (all P < 0.001), whereas there is a consistent but insignificant decrease (P = 0.054) in affinity for proline (Fig. 4c-f). PAT2-F159I not only binds Ser and α-ABA but efficiently translocates these amino acids (Fig. 4g). In PAT2, proline, glycine, and alanine have Km values in the range 120-700 μM [39,40,42], whereas serine is a very weak substrate. In contrast, the PAT2-F159I Km for serine (823 ± 202 μM) is close to that of CG1139 (Km 1.31 ± 0.12 mM) (Fig. 4h). This change in PAT2 selectivity, following F159I mutation to become CG1139-like (Figs. 3,  4), demonstrates clearly that residue size at position 159 is a key determinant of the substrate side-chain that can fit within the SLC36 family-binding pocket. A series of PAT2 mutants, where residue 159 was systematically reduced in size (phenylalanine 191.9 Å 3 , isoleucine 163.9 Å 3 , threonine 121.5 Å 3 , and cysteine 103.3 Å 3 ) [50], retained SLC36-like characteristics, transporting proline (the gradual decrease in uptake reflecting a decrease in affinity), glycine, and alanine (Fig. 5a-e). Methionine is not transported by CG1139, PAT2, or PAT2-F159I (Fig. 5a, f). The increase in hydrophobic-binding pocket volume in PAT2-F159T allows methionine to inhibit amino acid transport (Fig. 5b) without undergoing transport (Fig. 5a), whereas the additional space in PAT2-F159C creates a gain-of-function phenotype with excellent methionine transport (Fig. 5a,  f). Titration of the binding pocket volume versus substrate side-chain size was accomplished using a series of hydrocarbon side-chain extended amino acid derivatives from the simplest amino acid glycine (no side-chain) to those where the side-chain terminal carbon atom is in the beta (Ala), gamma (α-ABA), delta (norvaline, NVal), epsilon (norleucine, NLeu), and zeta (2-aminoheptanoic acid, AHA) positions (Fig. 5c). Glycine and alanine interact with all five carriers (Fig. 5c). α-ABA is excluded from PAT2 (Fig. 5c). NVal and NLeu are excluded from CG1139, PAT2, and PAT2-F159I, but they inhibit amino acid transport via PAT2-F159T and PAT2-F159C. AHA, containing the longest sidechain, can only inhibit amino acid transport by the largest binding pocket (PAT2-F159C) (Fig. 5c). For natural proteinogenic amino acids, the F159C mutation converts PAT2 from a carrier with limited space, within the binding pocket region associated with the substrate side-chain, to one that can transport longer amino acids such as methionine, glutamine and leucine (Fig. 5a, e-f). The general SLC36/PAT2 pocket mitigates against branching on the β carbon and this is retained in PAT2-F159C (relatively weak interaction with isoleucine, valine, and threonine) (Fig. 5d, e). PAT2-F159C allows access of lysine (with an epsilon-amino group and a nitrogen atom in the zeta position) into the binding pocket (Fig. 5d). However, the severely reduced rate of transport, compared to proline and methionine (Fig. 5f), suggests that the charged side-chain is incompatible with translocation. A visual summary of the comprehensive transport measurements described (Figs. 3, 4, 5) is presented in Fig. 6.
Our results suggest that the residue occupying the equivalent position to F159 (PAT2) is key to determining substrate selectivity in the SLC36 family and related invertebrate transporters (Figs. 3, 4, 5). To test whether the role of this residue is a common feature across APC superfamily amino acid transporters, the investigation was broadened. The basis of substrate selectivity was investigated in a distinct human AAAP transporter family focussing upon SNAT5 (SLC38A5), a carrier with very different substrate specificity to PAT2. Fig. 4 Substitution of F159 in PAT2 with isoleucine, the equivalent residue in CG1139, enables PAT2-F159I to transport amino acids with longer side-chains. Amino acid uptake was measured in oocytes expressing wild-type PAT2 (PAT2-WT), PAT2 with the F159I mutation (PAT2-F159I), or CG1139. Uptake was also measured in oocytes injected with water as a control. a Proline (10 μM) uptake in the absence (control) or presence of serine or α-ABA (both 5 mM). n = 19-20; NS, P > 0.05; **, P < 0.01; ***, P < 0.001 vs. control (ANOVA, Tukey's multiple comparisons test). b Proline (10 μM) uptake in the absence (control) or presence of various unlabelled amino acids (all 10 mM). Data are expressed as % control (that in the absence of unlabelled amino acid). Uptake into water-injected oocytes is expressed as % PAT2-F159I control. n = 17-20; NS, P > 0.05; *, P < 0.05; ***, P < 0.001 vs. control (ANOVA, Tukey's multiple comparisons test).
Comparison of a human SNAT5 homology model with the outward-occluded substrate-bound crystal structure of LeuT (Fig. 7a) reveals that SNAT5 A138 occupies the equivalent position to LeuT V104 (and thus PAT2 F159) (Fig. 7b,  c). Serine is transported well by SNAT5, whereas alanine is transported poorly (Fig. 7d) [41]. In a reversal of the protocol used with PAT2 (Fig. 5), SNAT5 residue 138 was mutated and systematically increased in size from alanine to phenylalanine (90.0-191.9 Å 3 ) [50], producing A138T, A138I, and A138F. Reducing the SNAT5-binding pocket volume decreases serine transport in A138T and A138I and abolishes transport in A138F (Fig. 7d). The reduction in affinity for serine and asparagine (Fig. 7e, f) in A138T and A138I indicates that they do not fit as well within the smaller binding pocket. In contrast, the A138T and A138I mutants gain function and become excellent alanine transporters (Fig. 7d) 6 Substitution of PAT2 F159 with smaller side-chain residues is predicted to enlarge the substrate-binding pocket and concomitantly increase access of substrates with elongated side-chains. Predicted substrate-binding pocket of PAT2 modelled against the outward-open substrate-bound LeuT structure (2A65). Parts of TM1 (grey), TM6 (grey), and TM3 (blue) are shown as ribbons with potential substrates presented as spheres. The side-chain of F159 and related mutants are shown as blue spheres (with the oxygen atom of threonine in red and the sulphur atom of cysteine in yellow). a, b Prototypical PAT2 substrate alanine (Ala) and the amino acid analogue α-ABA, which is excluded from PAT2, are shown in the binding pocket of wild-type PAT2. Alanine fits within the pocket, but extension of the side-chain by the single methylene group in α-ABA produces a clash with the large volume aromatic ring of F159 thus limiting access. c Substitu-tion of F159 with the smaller side-chain of isoleucine (F159I) permits α-ABA access to the PAT2 substrate-binding pocket. d, e Substitution of F159 with threonine (F159T) is predicted to further increase the PAT2 substrate-binding pocket to accommodate NVal but not the side-chain extended AHA which clashes with the threonine sidechain. f Access of AHA to the PAT2 substrate-binding pocket is permitted by a further reduction in the residue side-chain volume by substitution of F159 with cysteine (F159C), although some rearrangement of the flexible side-chain is likely required to allow a comfortable fit. Note that the view of the PAT2 substrate-binding pocket in a-d has been rotated by 180º in the horizontal plane in e and f to highlight the predicted clash between the side-chains of AHA and F159T. All amino acid substrates and analogues were inserted into the binding pocket upon the leucine backbone (2A65) using PyMOL (Fig. 7g). The affinities for serine and alanine change, such that SNAT5 favours serine, the smaller binding pocket of A138T appears to take both substrates with similar affinity, whereas the even smaller pocket of A138I prefers alanine ( Fig. 7e-g). Predicted changes in the binding pocket are visualised in Fig. 7h- Mutation of the LeuT V104 equivalent residue (A138) in the System N transporter SNAT5 (SLC38A5) has a striking effect on substrate selectivity. a Human SNAT5 modelled against the crystal structure of LeuT (2A65). TM1-TM10 of LeuT are coloured light grey. The substrate-binding pocket of SNAT5 is predicted to be formed by TM1 (green), TM3 (blue), TM6 (orange), and TM8 (magenta). The remaining six SNAT5 TMs modelled are coloured dark grey (left-hand figure, within the plane of the membrane; righthand figure, orientation above the plane of the membrane). For clarity, an extracellular loop (residues 227-247) in SNAT5 has been omitted due to a lack of predicted structural homology to LeuT. b Partial HHPred alignment of TM3 in LeuT and SNAT5. c Crystal structure of LeuT (light grey) with the predicted structure of SNAT5 TM3 superposed (blue). The side-chains of V104 (LeuT) and A138 (SNAT5) are shown as sticks. Serine is shown as a substrate (spheres) and was inserted into the binding pocket upon the leucine backbone using PyMOL. d Serine and alanine (both 50 μM) uptake (at pH 8.5, Na + -containing solution) were measured in wild-type SNAT5 (SNAT5-WT) and the SNAT5 mutants A138T, A138I, A138F. n = 20; NS, P > 0.05; ***, P < 0.001 vs. SNAT5-WT (ANOVA, Tukey's multiple comparisons test). e-g Serine uptake (50 μM, pH 8.5, Na + ) by SNAT5-WT, SNAT5-A138T, and SNAT5-A138I, measured in the presence of e unlabelled serine, f asparagine, or g alanine (all 0-10 mM). Data are expressed as % control (absence of competitor) after the subtraction of uptake into water-injected oocytes. n = 9-10; NS, P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001 vs. SNAT5-WT (ANOVA, Tukey's multiple comparisons test) whereby the upper symbols relate to SNAT5-A138I and the lower symbols to SNAT5-A138T. h-k Predicted substrate-binding pocket of SNAT5 modelled on LeuT with serine as a substrate. The sidechains of A138 (TM3), A138I and A138F are represented by light blue spheres (with the oxygen atom of threonine in red). Substitution of A138 with residues with sequentially larger side-chains [threonine, A138T (i); isoleucine, A138I (j); phenylalanine, A138F (k)] results in a concomitant decrease in the volume of the predicted substratebinding pocket which reduces affinity for serine (i, j) (and asparagine) or abolishes transport (k)

Discussion
In all forms of life, transmembrane transfer of nutrients, micronutrients, and excretory products is crucial to enable homeostasis, adaptation, and optimal cellular performance. Amino acid carriers are numerous in nature (see TCDB) [1], possess distinct functions, and exhibit dramatic differences in substrate specificity. Each cell type expresses a unique array of amino acid carriers. For example, the mammalian small intestinal epithelial cell expresses at least seven amino acid transport systems at the luminal surface, eight distinct mechanisms at the basolateral membrane, with many others being expressed in intracellular organelles. More than half of the 71 human LeuT-fold carriers demonstrate amino acid transport with that number likely to rise as orphan transporters are characterized functionally. Thus, the LeuT-fold is particularly efficient at transmembrane movement of amino acids. Each of the 37 human LeuT-fold amino acid carriers characterized thus far demonstrates a unique selectivity, varying greatly in substrate specificity and relative affinity. The heterogeneity in selectivity, and the currently limited understanding of its underlying molecular basis, was the motivation for this study. Rather than focussing simply on well-conserved residues, we thought it judicious to seek evidence for conservation of a discrete position within the LeuT-fold with the capacity to influence substrate recognition across the APC superfamily. The properties (spatial, steric, chemical, and electrical) of the binding pocket of each transporter are determined by main-chain hydrogen bonding partners in the unwound regions of TM1 and TM6, along with the side-chains of various residues contained primarily within TM1, 3, 6, and 8 [18]. The principal means by which proteinogenic amino acid substrates are differentiated is by side-chain recognition. The side-chain of the V104 residue forms part of the hydrophobic pocket within the LeuT binding site and makes van der Waals' contacts with the aliphatic substrate sidechain [3]. We hypothesised that the conserved function of LeuT V104, and of different amino acid residues occupying the equivalent site in other APC superfamily transporters (Table 1), is to shape the binding pocket and thus influence substrate selectivity.
The equivalent residue to F159 is conserved within mammalian SLC36 transporters but varies in the arthropod expansion. The model arthropod carrier CG1139 (which is important in fly growth) [49] shares many functional characteristics with mammalian PATs but notably transports amino acids with larger side-chains, consistent with having an isoleucine rather than phenylalanine at position 149 [the LeuT V104-equivalent (Figs. 23,4,5)]. In PAT2-F159I, substrate selectivity broadened to become more CG1139-like with the key-determining factor being the space available within the hydrophobic-binding pocket (Figs. 4 and 6). Mutation to F159T and F159C decreased residue size, resulting in further changes in substrate selectivity, consistent with an increase in binding pocket volume enabling access and translocation of amino acids with longer side-chains (Fig. 5). Notably, a threonine is present at the equivalent position in the Aedes aegypti carrier AaePAT1 (Fig. 2, Table 1) which accepts amino acids with longer side-chains than alanine and glycine [53]. AaePAT1 is highly upregulated in the midgut following a blood meal and is responsible for amino acid uptake in the yellow-fever mosquito [53]. Similarly, a cysteine is found in the aphid Acyrthosiphon pisum transporter ApGLNT1 (Fig. 2, Table 1) (which sits at the bacteriocyte membrane at the symbiotic interface where it supplies glutamine [transported here by PAT2-F159C (Fig. 5)] to the proteobacterium Buchnera aphidicola) [54].
Thus, the homologous residue to LeuT V104 is a key determinant of substrate recognition in mammalian SLC36 (and SLC38) transporters and in more remotely related invertebrate carriers.
Furthermore, interpreting published data in the light of our work suggest that this site has a much broader and general significance for substrate selectivity across the APC transporter superfamily. Despite being only distantly related (Fig. 1), PAT2 (from the AAAP family) superposes on APC superfamily structures from the NSS, APC, and NCS1 families (Fig. 2). We find evidence for a role of this equivalent residue in substrate selectivity from functional and mutational studies of, mainly non-amino acid transporting, members of the SLC6, SLC7, and SLC12 families [22,24,[55][56][57][58][59][60][61][62]. In SLC6 transporters from within the NSS (2.A.22) family (Fig. 1, Table 1), I172 in the serotonin transporter SERT, V152 in the dopamine transporter DAT (equivalent to V120 in the Drosophila DAT crystal structure, 4XP4), V148 in the noradrenaline transporter NET, and C144 in the creatine transporter CT1 are equivalent to LeuT V104 and are predicted to occupy sites close to the binding pockets. Even subtle mutations of these residues can modify substrate selectivity, affinity, and inhibitor (e.g., selective serotonin reuptake inhibitors) binding [22,24,[55][56][57][58][59]61]. These functional observations are confirmed in the crystal structures of human SERT and Drosophila DAT where the V104-equivalent residues define regions of the binding pockets associated with binding of substrates and antidepressants [12,15,45]. Similar observations are made in the APC (2.A.3) family which includes the structurally resolved AdiC and ApcT as well as the mammalian SLC7 transporters (Fig. 1, Table 1). Mutation of N133 in mouse LAT2 (slc7a8), by introduction of the LAT1 (SLC7A5)-equivalent residue, to produce N133S (corresponding to V104 in LeuT, N134 in human LAT2 and S144 in human LAT1), increases 3,3-diiodothyronine (T2) transport [62]. In the CCC (2.A.30) family, mutation of the V104 equivalent residue (A379) in the bumetanide-sensitive NKCC1 (SLC12A2) (Fig. 1, Table 1), a Na + /K + /2Cl − cotransporter important in human fluid and electrolyte secretion and homeostasis, demonstrates that side-chain size was inversely related to 86 Rb + flux, and affinities for sodium and chloride were reduced compared to wild-type [60]. In the NCS1 (2.A.39) family (Fig. 1, Table 1), W117 in Mhp1 is conserved among all other members [5]. The indole ring of tryptophan forms a pi-stacking interaction with the hydantoin moiety of the Mhp1 substrate and presumably performs a similar function in other family members as all substrates contain ring structures [25]. Thus, the data reported here, supported by published data from a variety of eukaryote and prokaryote transporters, confirm that the residue occupying the equivalent position to LeuT V104 is important in determining substrate selectivity, in both amino acid transporters and other carriers, across the APC superfamily (Fig. 1, Table 1).
Amino acid transporters and other solute carriers (SLCs) are involved in many key physiological processes and, as such, are drug targets for treatment of numerous disease states [18,20,63]. In addition, SLCs are integral determinants of drug disposition as therapeutic agents can hijack transporters [64]. Thus, numerous prokaryotic and eukaryotic transporters are potential targets in overcoming diseasecausing mutations, targeting disease-causing vectors, and improving drug delivery and agricultural yield. Realistically, we cannot determine the structure and function of all transporters throughout the kingdoms of life, particularly as the number of potential targets increases daily. Rather, accurate in silico modelling and predictive methods of function are something of a "holy grail". Such methodologies are being used currently with some success in the identification of novel substrates for specific SLCs [65]. However, for rational approaches to novel drug design and treatment of disease states to be developed, and for such predictive modelling strategies to be successful, extensive knowledge and understanding of both structure and function of archetypal membrane transporters are required. Judging by the large number of distinct amino acid transporters in the APC superfamily (see TCDB) [1], the LeuT-fold appears particularly well adapted for the translocation of amino acids. Here, we show, through comprehensive functional studies, that a single divergent residue position is a principal molecular determinant of substrate specificity in LeuT-fold amino acid transporters. The V104-equivalent residue is an important piece of the puzzle which, along with future studies of both dynamic (functional) and static (structural) states, will enable construction of an accurate 3D functional map of the APC superfamily LeuT-fold.