The Aspergillus nidulans Proline Permease as a Model for Understanding the Factors Determining Substrate Binding and Specificity of Fungal Amino Acid Transporters*

Background: PrnB, different from Put4p, its Saccharomyces cerevisiae orthologue, is a highly specific l-proline transporter of Aspergillus nidulans. Results: Identification of 12 residues important for PrnB activity and/or specificity. Put4p-mimicking mutants of PrnB recognize other Put4p substrates. Conclusion: Residues variable in the yeast amino acid transporter (YAT) family determine transporter specificity. Significance: Understanding the molecular mechanisms underlying amino acid recognition and translocation through YATs. Amino acid uptake in fungi is mediated by general and specialized members of the yeast amino acid transporter (YAT) family, a branch of the amino acid polyamine organocation (APC) transporter superfamily. PrnB, a highly specific l-proline transporter, only weakly recognizes other Put4p substrates, its Saccharomyces cerevisiae orthologue. Taking advantage of the high sequence similarity between the two transporters, we combined molecular modeling, induced fit docking, genetic, and biochemical approaches to investigate the molecular basis of this difference and identify residues governing substrate binding and specificity. We demonstrate that l-proline is recognized by PrnB via interactions with residues within TMS1 (Gly56, Thr57), TMS3 (Glu138), and TMS6 (Phe248), which are evolutionary conserved in YATs, whereas specificity is achieved by subtle amino acid substitutions in variable residues. Put4p-mimicking substitutions in TMS3 (S130C), TMS6 (F252L, S253G), TMS8 (W351F), and TMS10 (T414S) broadened the specificity of PrnB, enabling it to recognize more efficiently l-alanine, l-azetidine-2-carboxylic acid, and glycine without significantly affecting the apparent Km for l-proline. S253G and W351F could transport l-alanine, whereas T414S, despite displaying reduced proline uptake, could transport l-alanine and glycine, a phenotype suppressed by the S130C mutation. A combination of all five Put4p-ressembling substitutions resulted in a functional allele that could also transport l-alanine and glycine, displaying a specificity profile impressively similar to that of Put4p. Our results support a model where residues in these positions determine specificity by interacting with the substrates, acting as gating elements, altering the flexibility of the substrate binding core, or affecting conformational changes of the transport cycle.

Amino acid supply in cells depends on specialized transmembrane transporter proteins. The majority of these proteins belong to the ubiquitous amino acid polyamine organocation (APC) 5 superfamily, one of the largest families of secondary transporters (1,2). The APC includes members that function as solute/cation symporters and solute/solute antiporters and are found in bacteria, archeae, fungi, protists, plants and animals (1,3). Fungal APC members belong to the yeast amino acid transporter (YAT) family (4). Numerous YAT transporters can be identified in all sequenced genomes of fungi. Characterized members include 15 of the 16 characterized amino acid transporters of Saccharomyces cerevisiae (5,6), two proteins of Aspergillus nidulans, PrnB and AgtA (7,8), as well as transporters from Candida glabrata, Neurospora crassa, and Penicillium chrysogenum (9 -11). These proteins display different specificities and are subject to tight transcriptional and post-translational regulation (12)(13)(14).
The prnB gene encoding the main proline transporter of A. nidulans is part of the prn gene cluster mediating transport and utilization of proline as a sole carbon and/or nitrogen source (7,15,16). The prnB gene has been extensively used as a model for transcriptional and post-transcriptional regulation of gene expression (15)(16)(17)(18)(19)(20).
The PrnB transporter has been used as a model to address structure-function relationships in YATs, by mutational analysis and cysteine scanning mutagenesis (21)(22)(23). Earlier computational studies suggested that APC members possess the so called "5 ϩ 5"-fold (24), commonly found in the crystal struc-tures of several prokaryotic transporters belonging to evolutionary distinct protein families with different substrate specificities (25)(26)(27)(28). By using a sensitive homology threading approach, a three-dimensional model of PrnB was previously generated, using as template the crystal structure of the LeuT transporter of Aquifex aeolicus (23). This model predicted that PrnB, as well as other members of the APC superfamilly and the YAT family, share a common structural core (23). This core consists of two intertwined, antiparallel V-shaped repeating units of five transmembrane segments each (TMS1-5 and TMS6 -10) connected by a relatively long loop and is followed by two additional helices (TMS11 and -12). Along with mutational evidence, the model suggested that the substrate binding pocket of APCs would be located in the vicinity of the unwound part of two kinked helices (TMS1 and TMS6) of each V-shaped repeated unit of the transporter. These predictions were confirmed by the subsequent crystal structures of three bacterial APC members: the arginine/agmatine (AdiC) and glutamate/␥aminobutyric acid (GadC) antiporters and the ApcT, a broad specificity proton/amino acid symporter (29 -31).
In this study, using the PrnB transporter, we combine molecular modeling and induced fit docking along with comparative sequence analyses, genetic and biochemical approaches in an attempt to identify substrate binding residues and specificity determinants of YATs. More precisely, we provide further evidence that residues Gly 56 and Thr 57 of TMS1 contribute to substrate binding and also identify new substrate interacting residues in TMS3 (Glu 138 ), TMS6 (Phe 248 ), TMS8 (Trp 351 ), and TMS10 (Thr 414 ). Moreover, we show that PrnB is a highly specific L-proline transporter compared with Put4p, which is also known to transport L-alanine, glycine, GABA, and the toxic proline analog L-azetidine-2-carboxylic acid (AZC) (32)(33)(34)(35)(36)(37). Taking advantage of the high primary sequence similarity and the specificity differences of PrnB and Put4p, we identify residues in TMSs 3 (Ser 130 ), 6 (Phe 252 and Ser 253 ), 8 (Trp 351 ), and 10 (Thr 414 ) that contribute to the determination of the specificity profile of proline transporters. Based on these results, we discuss the potential mechanisms governing specificity determination in fungal amino acid transporters.
Homology Modeling-Multiple threading alignments were created with LOMETS metaserver (38) using the full-length sequence of PrnB from A. nidulans or Put4p from S. cerevisiae (Uniprot accession codes P18696 and P15380, respectively). Fig. 1 shows a multiple alignment of YAT sequences. The alignments of the PrnB and Put4p sequences with that of AdiC (PDB code 3L1L), shown in Fig. 2, were obtained by TM-Coffee (39) and modified manually on the basis of this multiple alignment. The full-length sequences of PrnB and Put4p were submitted to the TOPCONS meta-server (40) for consensus membrane pro-tein topology prediction. 2000 models of PrnB residues 38 -515 and 2000 models of Put4p residues 107-590 were constructed in total using the Rosetta threading protocol with the membrane specific scoring function terms (41)(42)(43). The secondary structure of the protein, as predicted by SAM server (44), as well as the consensus membrane protein topology prediction made by TOPCONS meta-server, were also taken into account in the form of restraints during model construction. 3-mer and 9-mer peptide fragments for modeling were created locally using the SAM-predicted secondary structure of PrnB and Put4p. Unfolded and broken structures were excluded by retaining only the 90% top scores models. A random model among them was superimposed on the crystal structure of AdiC using the combinatorial extension algorithm (45), and the putative substrate binding cavities of PrnB and Put4p were determined by selecting every residue that was poised within 8 Å from the arginine ligand. Trajectories were assembled from the remained low-energy models and were clustered with GROMACS Tools version 4.6.1 (46) using the backbone heavy atoms and C␤ carbon of the putative substrate binding pocket. The algorithm used for clustering was "gromos" with cutoff at 1.5 Å. Root mean square (r.m.s.) flunctuation of the backbone atoms was calculated with g_rmsf module of GROMACS Tools.
Induced Fit Docking-Protein preparation using OPLS2005 force field (47) and molecular docking was performed with the Schrödinger Suite 2014, as described in our previous study (48). Substrates were docked on the most dominant cluster representative receptor structures, using the induced fit docking protocol (Schrödinger Suite 2014 Induced Fit Docking protocol), which is intended to circumvent the inflexible binding site and accounts for the side chain and backbone movements upon ligand binding (49).
Strains, Media, and Growth Conditions-Strains used in this study are summarized in Table 1. All carry the veA1 mutation. Chemicals were from Sigma and AppliChem GmbH. Standard, previously described MM and complete media for A. nidulans, as well as medium supplements, were used (described in the Fungal Genetics Stock Centre). Assessment of the functionality of PrnB alleles was performed in MM using 1% glucose or 1% glycerol as carbon sources. Urea, amino acids, and GABA have been supplemented at a final concentration of 5 or 10 mM, unless otherwise indicated. D-Serine was used at a final concentration of 0.1 mg/ml, using urea as the sole nitrogen source. For membrane-enriched protein extracts, strains were incubated 3-4 days at 37°C in solid complete medium. Conidiospores were harvested, filtered through miracloth, and used to inoculate flasks containing 200 ml of liquid MM, appropriately supplemented, using 0.1% fructose and 10 mM NaNO 3 as carbon and nitrogen sources, respectively. Cultures were incubated 18 h at 25°C and prnB induction was achieved by the addition of 20 mM L-proline for the last 4 h of growth.
Molecular Manipulations and Transformation-prnB alleles have been generated by in vitro site-directed mutagenesis of plasmid pA4 (50), using Kapa-HiFi (Kapa Biosystems) DNA polymerase and primers 11-32 of Table 2. The alleles were confirmed by sequencing (VBC Genomics, Austria). Linearized sequences of these plasmids were used to transform, as previously described (51), strain prn397. Transformants were iso-lated and analyzed by Southern blotting for the sequence of the prnB gene, ensuring that all strains used in this study carry single copy in-locus prnB alleles derived from homologous integration events. A ϳ1100-bp SmaI fragment, including part of the promoter and ORF of prnB, from plasmid pA4 was used as a prnB-specific probe. The ϳ1500 bp 5Ј and 3Ј flanking sequences of gabA were PCR amplified using primers 1-8 of Table 2 digested with KpnI/XbaI and XbaI/NotI and subsequently cloned into a KpnI/NotI linearized pBlueScript II SK(ϩ)vector. In the unique XbaI site of the resulting plasmid, the AfpyrG gene, PCR amplified using primers 9 and 10 from plasmid p1439 (52,53) and XbaI digested was cloned. The resulting cassette was PCR amplified using primer pairs 1 and 4 and ϳ2 g of the products were used to transform strain nkuA⌬, selecting for uracil/uridine prototrophy. Single copy gabA⌬::AfpyrG strains were checked by Southern analysis, using as probes the sequences of AfpyrG and the 5Ј flanking sequence of gabA. Plasmid preparation from Escherichia coli strains and DNA bands purified from agarose gels were performed with the Nucleospin Plasmid kit and the Nucleospin ExtractII kit, respectively, according to the manufacturer's instructions (Macherey-Nagel, Lab Supplies Scientific SA, Hellas, Greece). DNA probes for Southern analysis (54) were labeled with [ 32 P]dCTP (3000 Ci/mmol, Institute of Isotopes Black arrowheads indicate residues conserved in the family, which in PrnB (Gly 56 , Thr 57 , Glu 138 , and Phe 248 ) and Can1 (Gly 103 , Thr 104 , and Phe 295 ) appear to interact with the invariant part of amino acid substrates. Tyr 181 of Bap2, necessary for transport activity, is also highlighted. Black asterisks indicate residues non-conserved in the family, which in PrnB (Trp 351 and Thr 414 ) and Can1 (Ser 176 , Thr 180 , Gln 298 , and Thr 456 ) appear to participate in substrate binding, as well as specificity determination residues of PrnB (Ser 130 , Lys 245 , Trp 351 , and Thr 414 ). Empty arrowheads indicate residues that are close to the binding site and, directly or indirectly, affect the specificity of PrnB (Phe 252 and Ser 253 ). Black circles indicate residues important for PrnB function, which are conserved in the family and do not appear to directly participate in substrate binding (Phe 250 and Glu 255 ). Visualization was done using ESPript 3.0 (78), following annotation using Photoshop CS5 (Adobe). Co. Ltd., Miklós, Hungary) using a random hexanucleotide primer kit following the supplier's instructions (Takara Bio, Lab Supplies Scientific SA) and purified on MicroSpin TM S-200 HR columns following the supplier's instructions (Roche Diagnostics). Restriction enzymes used were from Takara Bio.
Epifluorescence Microscopy-Samples were prepared as previously described (55,56). 20 mM L-proline was used for induction of prnB expression, at 25°C for 12-14 h. Samples were observed using a Plan-Apochromat ϫ100 1.40 NA oil immersion objective lens on an Axioplan 2 fluorescence microscope (Carl Zeiss, Inc.) with appropriate filters and the resulting images were acquired with a Zeiss-MRC5 digital camera using AxioVs40 V4.40.0 software. Images were then processed with ImageJ (NIH) and annotated with Photoshop CS5 software (Adobe).
Radiolabeled Uptake Measurements and Kinetic Analyses-Activity of PrnB mutants was measured by estimating rates of L-[ 3 H]proline (60.0 Ci mmol Ϫ1 , Moravek Biochemicals, Brea, CA) uptake as previously described (21)(22)(23)57). L-[ 3 H]Proline uptake in MM, pH 6.8, at 25°C was assayed in germinating conidiospores of A. nidulans derived from 8 to 9-h cultures at 25°C, induced with 20 mM L-proline and washed 3 times with 50 ml of MM, finally concentrated at 10 7 conidiospores/100 l. Initial velocities were measured at 1-5 min of incubation with 0.25-20 M radioactive substrate because uptake was linear for at least 30 min (21). K m /K i values were obtained directly by performing and analyzing (Prism3) L-[ 3 H]proline uptake in the presence of at least 10 different concentrations of competitor. In all cases background uptake values were corrected by subtracting values measured in a prnB⌬ strain. K i values calculated satisfy the criteria for use of the Cheng and Prusoff equation , in which L is the permeant concentration. IC 50 values were determined from full dose-response curves, and in all cases the Hill coefficient was close to Ϫ1, consistent with the presence of one uptake system.
Protein Extraction and In-gel Fluorescence-Membrane-enriched protein extracts were prepared as previously described (22,55), with the following modifications (58). Following membrane protein precipitation by centrifugation (1 h, 18,000 ϫ g, 4°C), the extraction buffer was removed and the pellet was resuspended in ice-cold In-gel Fluorescence Resuspension Buffer (IFRB: 50 mM Tris-HCl, pH 7.5, 5 mM EDTA, 10% glycerol) freshly supplemented with protease inhibitor mixture (Sigma, P8215, 1:500) and 1 mM PMSF. Cell debris was pelleted by centrifugation at 500 ϫ g for 3 min at 4°C and the supernatant was transferred to a new pre-cooled Eppendorf tube. Protein concentration was measured by the method of Bradford. 30 -50 g of protein samples were mixed with one-third volume of 4ϫ SB for In-gel Fluorescence (4ϫ IFSB: 70 mM Tris-HCl, pH 7.5, 7% glycerol, 7 mM EDTA, pH 8.0, 8% SDS, 100 mM DTT, 0.02% bromphenol blue) and immediately used to load a standard 10% SDS-polyacrylamide gel. Following SDS-PAGE at 110 V, the gel was rinsed with dH 2 O and the fluorescent bands were detected with a CCD camera system (ImageQuant TM LAS 4000, GE Healthcare) by exposure to blue light (EPI source) set at 460 nm, using a cut-off filter of 515 nm, by increasing the exposure time until fluorescent bands were clearly visible (10 -60 s). The gel was subsequently stained with Coomassie staining solution. Images were processed with ImageJ (NIH) and annotated with the Adobe Photoshop CS5 software.

RESULTS
To identify the topology of PrnB and Put4p residues that are essential or crucial for proline binding and transport, we constructed the three-dimensional structural models of the transporters, using a robust comparative modeling approach (Fig. 3). The models were based on the outward-facing occluded structure of AdiC (PDB code 3L1L). In that structure, AdiC was crystallized with its substrate, arginine, in an intermediate conformation, which enables the identification of most of the amino acid residues involved in substrate binding (29).

TABLE 2 Primers used in this study
Primers 7-28 have been used for in vitro site-directed mutagenesis along with a, herein not shown, reverse complement primer.  3A shows a representative structure of the largest cluster of each of the two transporters. The overall models of the two permeases are, as expected, very similar ( Fig. 3B). A notable difference was observed in the positioning of TMS6a and TMS11, which are closer to the pore of the transporter in Put4p. The observed differential positioning of TMS6a could be partially explained by the differential flexibility of the nearby unwound segment of TMS6. Indeed, r.m.s. fluorescence measurements showed that the positioning of TMS6a is substantially more variable in Put4p than in PrnB (Fig. 3C). Our mutational analysis suggests that this could be related to the broader specificity of Put4p (see below). In contrast to TMS6a, the r.m.s. fluctuation of TMS11 between PrnB and Put4p is quite similar.
Identifying Substrate Binding Residues-We performed induced fit docking of L-proline in PrnB as well as L-proline, L-alanine, and L-azetidine-2-carboxilic acid in Put4p ( Fig. 4 and data not shown). Docking of these substrates was performed to the representative structures of at least the three largest clusters of the homology models, which altogether represent more than 50% of the structures created for each transporter. With this strategy we allowed implicitly larger structural movements to occur in the substrate binding regions, than those introduced by the induced fit docking protocol alone. This way we accounted for transporter flexibility, which was necessary (as discussed below) to obtain docking poses that are in agreement with previous studies (23,29,59).
Our docking studies suggested that proline binds almost identically in PrnB and Put4p (Fig. 4, A-D), whereas L-alanine and AZC are bound to Put4 in a way very similar to that of proline (data not shown). More specifically, in the first two clusters, which represent the most abundant populations of the structures generated (45.5% in PrnB and 54.4% in Put4p), proline seems to bind to two distinct sites. In the first site, which is located closer to the extracellular region of the transporters, proline interacts with the side chains of two amino acid residues, which are highly conserved in the YAT family (Fig. 4, A  and B). The first, Thr 57 of PrnB (Thr 126 of Put4p), is located in the unwound segment of TMS1 and is part of the characteristic "G(T/S)G" motif of YATs (Fig. 1). This residue, previously suggested to be part of the proline binding site (23), is found to interact with proline via a hydrogen bond. The second residue, Glu 138 of PrnB (Glu 211 of Put4p), located in TMS3 and also highly conserved in the family, interacts with proline via a salt bridge. The interactions of these two residues with proline via the invariant part of amino acid substrates (the ␣-carboxyl and the imino group) are compatible with their high conservation in several YATs of different specificities (Fig. 1). In addition, the docked substrate forms H-bonds with the backbone of the highly conserved PrnB G56 and PrnB G58 (Fig. 4, C, Put4p G125 , and Put4p G127 in B, respectively). This configuration seems to be stabilized by hydrophobic or -cation interactions with the side chains of PrnB F248 and PrnB W351 (Fig. 4, A and C, Put4p F321 , and Put4p F424 in B and D, respectively). PrnB F248 (Put4p F321 ), located at the vicinity of the unwound part of TMS6, is highly conserved in the family. On the contrary, an aromatic residue in position 351 of PrnB is found only in proline-specific YATs (Fig. 1).
An interesting difference in proline binding was acquired by a significant percentage of the second cluster of Put4p structures (26.4% of the population). In these structures it was found that proline still interacts with the side chains of residues Thr 126 and Glu 211 , but it also interacts with Lys 318 (Fig. 4B). Interestingly, this Lys residue is not conserved in YATs and is present exclusively in proline-specific permeases (Fig. 1). The side chain of Lys 245 , the corresponding residue of PrnB, in all PrnB models is positioned away from the proposed binding site and toward the extracellular surface, similarly to some of the configurations obtained for Put4p (Fig. 4D). However, previous experimental data showed that Lys 245 is important for efficient proline transport by PrnB, whereas its substitution by Leu significantly increases the K m for proline (21). The fact that Lys 245 of PrnB was not predicted to interact with proline could be related to the positioning of TMS6a. This suggests a possible interaction of Lys 245 and the proline substrate in another conformation of the transporter, most probably a more outward facing. Another intriguing difference was the identification of three additional proline binding orientations in PrnB (Fig. 5), which could reflect intermediate positions of the substrate between the two substrate binding sites found in both PrnB and Put4p.
The second binding site, found only by the docking of L-proline in PrnB (Fig. 4E) and L-alanine in Put4p (Fig. 4F), is located deeper into the substrate binding core. Proline, at this site, is found to interact with PrnB via H-bonding with the side chains of Thr 414 and Tyr 127 (Fig. 4E). Similarly, L-alanine in Put4p forms H-bonds with the side chain of Tyr 200 (Tyr 127 of PrnB) and the backbone of Gly 326 and Glu 328 (Ser 253 and Glu 255 of PrnB, respectively) (Fig. 4F). Interestingly, in this configuration, L-alanine, a substrate of Put4p but not PrnB, seems to be close to Leu 325 , which along with Gly 326 differ in the two transporters (Phe 252 and Ser 253 in PrnB, respectively). Thr 414 , part of TMS10, is also different between the two transporters (Ser 487 of Put4p) and other YATs (Fig. 1). Most interestingly, it corresponds to Thr 456 of Can1, a substrate interacting residue of the arginine permease of S. cerevisiae that has been very recently shown to determine the specificity of the basic amino acid transporters of yeast (59). The above residues, along with PrnB W351 , constitute the main differences in the predicted substrate binding pockets of PrnB and Put4p (Fig. 9) and could,  thus, be related to the different specificity profiles of the two orthologues.
Rationale of the Mutational Analysis-The comparative analysis of structural alignments of YAT members of Fig. 1 showed that the substrate interacting residues of PrnB and Put4p fall into two categories. The first includes residues showing high to absolute evolutionary conservation in YATs (Gly 56 , Thr 57 , Glu 138 , and Phe 248 of PrnB). Consistent with their conserved nature, our docking analysis showed that the above PrnB residues interact with proline through the invariable part of amino acid substrates (Fig. 4).
The second category includes residues (Ser 130 , Gly 247 , Phe 252 , Ser 253 , Trp 351 , and Thr 414 of PrnB) predicted to be in the substrate binding pocket of PrnB and Put4p but are variable between the two permeases ( Fig. 9) and other YATs (Fig. 1). Among these residues, Trp 351 and Thr 414 are found to interact directly with proline (Fig. 4), whereas Ser 130 is close to Thr 414 (Fig. 3). Finally, Gly 247 , Phe 252 , and Ser 253 , parts of the unwound segment of TMS6, are close to the substrate binding core of PrnB.
Consequently, conservative substitutions of the conserved proline binding residues of the first category and Put4p-mimicking substitutions of the variable proline binding residues of the second category were carried out (S130C, G247A, F252L, S253G, W351F, and T414S). Moreover, the unwound segments of TMS6 of the proline-specific YATs are uniquely longer by one amino acid (Fig. 1). To investigate the importance of this variation, we constructed PrnB mutants shorter by one amino acid residue lacking either Ile 251 (I251⌬) or Phe 252 (F252⌬).
Study of Conserved Residues-Results concerning the conserved amino acid residues of the binding site in YATs are shown in Fig. 6 and Table 3. As shown in Figs. 6B and 7B, all alleles, except I251⌬ and F252⌬, are normally targeted to the plasma membrane. Consistently with previous Western blots (22), all PrnB-GFP alleles detected by in-gel fluorescence appear as doublets. Moreover, the lower molecular weight fast migrating bands that were also detected most probably correspond to N-terminal degradation products, as it has been observed for AgtA, another YAT of A. nidulans (60). In-gel fluorescence analysis (Figs. 6D and 7D) showed that all mutant proteins, except I251⌬ and F252⌬, are expressed at levels comparable with the wild-type PrnB-GFP, displaying a 2-fold maximal variation. The above results strongly indicate that the dif- ]proline initial uptake rates are well correlated with the growth of strains carrying the corresponding prnB mutants on proline as sole nitrogen source at 25°C. This is in agreement with previous studies on A. nidulans transporters, where only mutations that reduce the uptake rate Ͻ50% of the wild-type lead to a defect reflected by relevant growth tests, whereas mutations retaining Ͻ20% uptake rate are difficult to distinguish from a strain carrying deletion of the transporter

Specificity profile of functional PrnB mutants carrying substitutions in residues conserved in YATs
The abbreviations used in the table are: GABA, ␥-aminobutyric acid; L-AZC, L-azetidine-2-carboxylic acid; L-P-NH3, L-prolinamide; 3,4-D-Pro, 3,4-dehydro-L-proline; L-PCA, L-pipecolinic acid; NI, no inhibition (90 -100% uptake); Ͼ2000, slight inhibition at 2 mM (65-90% uptake); ϳ2000, inhibition at 2 mM (ϳ50 -65% uptake). Results are averages of at least three independent experiments in triplicate for each concentration point. Standard deviation was below 20%.   Fig.  6. 5M, PrnB allele carrying the following five substitutions: S130C/F252L/S253G/W351F/T414S. (48,57,61). Notable exceptions are strains expressing T57N, F248W/F250W, and E255D. T57N and E255D, despite retaining high initial uptake rates (Fig. 6C), show less pronounced growth on proline than expected. This is also the case for F248W/F250W, which despite retaining 33% of the wild type uptake rate, is indistinguishable from a prnB⌬ strain on proline. This apparent inconsistency can be well rationalized by reduced transport capacity, consistent with the 6 -20-fold decreased apparent half-saturation constant (K m ) for proline that these three mutants exhibit (Table 3). Interestingly, conservative substitutions G56P, E138D, and E138Q lead to total loss of PrnB function, as indicated by the lack of detectable levels of PrnB-dependent L-proline uptake, either by physiological growth tests at 25 or 37°C or by L-[ 3 H]proline uptake measurements. The above data are in agreement with results from docking analysis indicating that the side chain of Glu 138 and the backbone of Gly 56 directly interact with the proline substrate (Fig. 4A). T57N, although partially functional, displays up to 6-fold decreased K m /K i values for most substrates/ligands, whereas T57A was previously shown to be completely nonfunctional (23). These results are consistent with the direct interaction of Thr 57 with the ␣-carboxyl group of proline (Fig.  4A). Supportive of a predicted hydrophobic/-cation interaction of Phe 248 with the substrate are also the results from the presence of a tyrosine (F248Y) at position 248, which does not lead to significant changes in proline uptake and the kinetic characteristics of the transporter. However, the bulkier tryptophan (F248W) significantly reduced the functionality of PrnB, simultaneously reducing the K m /K i values for many ligands, especially for L-pipecolic acid, whereas previous data from our laboratory had shown that the presence of Leu at position 248 of PrnB results in 3-fold increased K m for proline and significantly reduces transport capacity (21). The important role of conserved Phe 250 and Glu 255 was reflected by the dramatic reduction of the K m /K i values of the transporter caused even by the very conservative substitutions F250Y and E255D. F250Y is significantly impaired for proline transport at 37°C, whereas it displays 3-fold decreased K m /K i values for many substrates. Most importantly, the combination of F248W with F250W resulted in up to 20-fold lower K m /K i values for most ligands. An analogous reduction was also acquired by the E255D sub-stitution, whereas E255Q was completely non-functional, indicating that a negative charge at this position is necessary for proline transport. These two residues (Phe 250 , Glu 255 ) are not predicted to directly interact with the substrate and could possibly be involved in conformational changes of the transporter during the transport cycle (see "Discussion"). Study of Variable Residues-Substitutions of the variable amino acid residues of the proline binding site are shown in Fig.  7 and Table 4. Epifluorescence microscopy suggests that a substantial proportion of I251⌬ and F252⌬ is driven to the vacuole for degradation. In line with this, reduced amounts of intact GFP-tagged molecules in in-gel fluorescence are observed, suggesting that the structure of the unwound part of TMS6 in proline transporters is different from other YATs and appears to be important for their stability. Most alleles carrying Put4pmimicking substitutions exhibit broadened specificity profiles, displaying significantly reduced K i values for some of the Put4p substrates not recognized by PrnB without displaying significant differences in their apparent K m values for L-proline. More specifically, Put4-ressembing substitutions of residues predicted to directly interact with the substrate (W351F, T414S), although significantly reducing the functionality of PrnB (Fig. 7, A and C), did not affect the K m for proline and displayed significantly lower K i values specifically for L-alanine, AZC, and glycine (Table 4). Contrary, the residually functional T414A, and the partially functional W351E do not show a significant decrease in the K i values of any Put4p substrate, whereas W351A leads to complete loss of PrnB transport activity. Similarly, substitutions F252L and S253G, resembling Put4p in residues close to the substrate binding pore, are largely functional and only decrease K i values for AZC and L-alanine, respectively. Contrary, G247A, another Put4 mimicking substitution of the unwound part of TMS6a, only had a minor effect on the kinetics, specificity profile, and function of PrnB. Finally, substitution of Ser 130 by the corresponding residue of Put4p (S130C) created a thermosensitive allele that exhibited decreased K i values specifically for L-alanine and AZC. Most interestingly, the S130C/T414S double mutant was fully functional, suppressing the thermosensitivity of S130C and the low functionality of T414S (Fig. 7). It possessed K m for proline similar to that of wild-type PrnB, and, most importantly, significantly reduced K i

3,4-D-Pro
Despite the clearly broadened specificity present in the above Put4p-resembling PrnB alleles none showed significantly decreased K i for GABA, another substrate of Put4p. Thus, a strain expressing a PrnB allele combining all five Put4pmimicking substitutions (S130C/F252L/S253G/W351F/T414S) (5M) was constructed and studied. Our results showed that this allele is partially functional and displays a specificity profile impressively similar to that of Put4p, exhibiting K i values for all Put4p substrates (L-proline, AZC, and L-alanine) very similar to the latter. Notably, this is the only PrnB allele where GABA significantly competes the uptake of L-[ 3 H]proline (Table 4).
To examine whether AZC, Ala, Gly, and GABA, apart from inhibiting L-[ 3 H]proline transport, could also be transported by the PrnB mutants, we made use of the genetic tractability of A. nidulans, which can utilize L-alanine, glycine, and GABA as sole nitrogen and/or carbon sources. Given that the permease(s) of A. nidulans involved in L-alanine and glycine transport are not known, we made use of a CkiA (casein kinase I) allele, which has been previously shown to be necessary for the plasma membrane targeting of amino acid transporters (60). The ckiA1919 mutation does not significantly affect prnB-dependent growth on L-proline, but confers severely impaired growth on many other amino acids, including glycine and L-alanine (Fig. 8). The PrnB mutants of Fig. 7 were introduced into the ckiA1919 background by standard genetic crosses, and were found to be targeted at the plasma membrane, as expected (data not shown). Growth tests showed that alleles T414S, the quintuple mutant 5M and, to a lesser extent, S253G, W351F, and W351E can improve the growth of a ckiA1919 strain in L-alanine as the sole nitrogen source (Fig. 8). Additionally, T414S and 5M were also able to improve the growth of ckiA1919 on glycine and ␤-alanine as the sole nitrogen sources. These phenotypes were more prominent at 25°C (Fig. 8), consistent with the partial thermosensitivity of PrnB-GFP (50). Altogether (Fig. 9, Table 5), the above data strongly suggest that these Put4p-mimicking PrnB mutants are indeed able to transport these Put4p substrates.
Contrary, none of the prnB alleles shown in Fig. 7 was able to improve the growth on GABA as the sole nitrogen source (data  (62). At this point it is worth mentioning that the expression of prnB alleles under these conditions is low but significant, because all are under the control of the native prnB promoter, which is significantly induced only in the presence of proline (7,20). In fact, as observed by epifluorescence microscopy (data not shown) and in-gel fluorescence analysis, the protein levels of intact PrnB-GFP are about 5-fold lower at derepressed than at induced conditions (Fig. 6D). Consequently, it cannot be ruled out that other alleles could also possess low efficiency in transporting L-alanine, glycine, and GABA, which, however, cannot become apparent in the absence of induction.
Finally, we tried to explore the possibility of AZC transport by the PrnB mutants. In S. cerevisiae, AZC is toxic and four permeases have been shown to be involved in its uptake, Put4p, Gap1, Agp1, and Gnp1 (33). Consistent with previous reports (63), we observed that AZC is not toxic in A. nidulans. AZC resistance in certain yeast strains is conferred by the product of the MPR1 gene, coding for an L-azetidine-2-carboxylic acid N-acetyltransferase (64 -66). However, deletion of AN2102, an orthologue of MPR1, did not render A. nidulans sensitive to AZC (data not shown). Most interestingly, we found that, rather than being toxic, AZC can be used as a poor nitrogen source by A. nidulans. This growth is independent of PrnB, GabA, and PrnC, 6 whereas no differences in growth on AZC as a sole nitrogen source could be observed in strains carrying any PrnB mutant allele described herein (data not shown).
The Specificity Profile of PrnB Mutants-To further confirm the docking results and gain further insight into the specificity profile of PrnB, apart from Put4p substrates, we calculated the K i values of all the functional PrnB mutants for other prolinerelated compounds (Fig. 10, Tables 3 and 4). The kinetic data obtained were consistent with our induced fit docking calculations. The kinetics of L-prolinamide, a proline analog possessing an amide in place of the ␣-carboxyl group of proline (Fig.  10), supports an interaction of the ␣-carboxyl group with the transporter. More specifically, L-prolinamide shows weak inhibition of L-[ 3 H]proline uptake. In addition, mutants that exhibit significantly improved K m /K i values for many substrates (e.g. E255D, F248W/F250W) do not exhibit an analogous improvement for L-prolinamide, suggesting that this substance is unable to efficiently interact with PrnB in a manner similar to L-proline. Surprisingly, the K i values of another proline analog, 3,4dehydro-L-proline, despite being a very good agonist in the wild-type (K i ϭ 52 Ϯ 5 M) and most mutants, showed no correlation with the observed differences in the K m for proline of most mutants. This suggested that 3,4-dehydro-L-proline also binds differentially to L-proline in PrnB.
Regarding the other substrates/ligands, mutants in residues not involved in specificity (Thr 57 , Gly 247 , Phe 248 , Phe 250 , and Glu 255 ) display similar correlative differences of the K m /K i values for L-proline, L-alanine, and AZC ( Table 3), suggesting that these substances are recognized by the transporter in a similar way. However, this is not the case for glycine and GABA, which possess different positioning of the amino group (Fig. 10), whose inhibition constants did not correlate with the variations in the K m for proline in the above mutants. These data suggest that glycine and GABA are not recognized by PrnB in a manner similar to L-proline. Curiously, the K i values for L-pipecolic acid, a proline analog with a side chain ring by one methyl group longer (Fig. 10), did not relate to the K m for proline. In fact, whereas the wild-type PrnB displays almost equal K i values for L-pipecolic acid and AZC, a proline analog with a smaller side chain ring, most mutants preferentially recognized only one of these two analogs (Table 3). This suggests that the size of the side chain is very critical for substrate recognition by PrnB. Consistently, Put4p-mimicking mutants that better recognize the smaller AZC, exhibit improved recognition for the also smaller L-alanine. Thus, it seems very likely that the substrate binding core of, the highly selective, PrnB is finely adapted to the size of proline, whereas the one of Put4p can also accept smaller molecules. It is not known whether Put4p recognizes L-pipecolic acid; however, our data suggest that it does not.

DISCUSSION
Transmembrane proteins are notoriously difficult to crystallize. As a result, homology modeling has been used as a reasonable alternative in cases where no x-ray structure is available (23,48,59,61,(67)(68)(69)(70)(71)(72)(73). Using this approach, combined with induced fit docking and a detailed characterization of the specificity profile, we identified 9 residues that are important for PrnB function, Gly 56 , Thr 57 , Ser 130 , Glu 138 , Phe 248 , Phe 250 , Glu 255 , Trp 351 , and Thr 414 (summary in Table 5). These residues, with the exception of Phe 250 , are located within the pore of the transporter and seem to affect different aspects of substrate translocation.
The Proline Binding Pocket of PrnB-Our results support the previously proposed model (23) predicting that the main substrate binding core of fungal amino acid transporters is formed by the unwound segments of TMS1 and TMS6. In addition we show that important substrate interacting residues are also found in TMSs 3, 8, and 10. Residues from TMS1 (Gly 56 , Thr 57 ), TMS3 (Glu 138 ), and TMS6 (Phe 248 ), highly conserved in fungal amino acid transporters, seem to directly interact with the 6 A. Biratsi, C. Gournas, and V. Sophianopoulou, unpublished data. invariant part of amino acid substrates, creating the scaffold for amino acid recognition. More specifically, the ␣-carboxyl and amino groups of amino acid substrates seem to be recognized by the side chains of the highly conserved Thr 57 and Glu 138 . This module, impressively similar to the way that amino acids bind to the bacterial transporters AdiC and LeuT (29,74), is also supported by recent studies of induced fit docking and mutational analysis of the arginine permease (59), as well as mutants of the leucine and tryptophan permeases of S. cerevisiae (67,69). Glu 138 was shown to be essential for PrnB transport activity, because even its most conservative substitutions by Asn or Asp were completely nonfunctional. These data suggest that not only a negatively charged residue is needed at this position, but also the length of its side chain is also important. Consistently, E184Q and E184A variants of Can1 are also nonfunctional (59). This glutamate, highly conserved in the YAT family, is not present in AdiC (Fig. 1). AdiC functions as an antiporter, whereas most fungal transporters are proton symporters (3). Interestingly, Ghaddar et al. (59) showed that the orientation of the side chain of the substrate could be influenced by the protonation state of the corresponding Glu 184 of Can1. Thus, it could be possible that this glutamate residue could also be associated with the commitment of the symported proton.
Insights into the Translocation Mechanism-Substitutions of Glu 255 , the other highly conserved in the YAT family glutamate, display a pattern different from that of Glu 138 ; E255Q is totally inactive, whereas E255D retains significant activity, displaying up to 19-fold improved apparent inhibition constants for many substrates/ligands. This situation resembles that of mutants in the Nucleobase Ascorbate Transporter family (48,57) and could be explained by a substrate trapping condition, where the substrate remains more time at the binding core of the transporter. This could be caused by either more efficient binding to a specific position of the translocation pathway or by partially defective transition from one conformational state of the transport cycle to another. Our data are supportive of the second scenario. First, analogous improvement of the K m is obtained for mutants in Phe 250 , which is orientated away from the substrate binding pocket and could not directly contribute to substrate binding. Second, Glu 255 also seems not to directly interact with proline, at least in this conformation of the transporter, whereas the corresponding residue of AdiC, Glu 208 , part of the distal gate, was suggested to control substrate translocation (29). This distal gate seems conserved in YATs, as, the highly conserved in YATs, Glu 255 and Tyr 127 of PrnB correspond to Glu 208 and Tyr 93 of AdiC (Fig. 1, Table 5), whereas the corresponding residues of the Bap2 leucine permease of S. cerevisiae, Glu 305 and Tyr 181 , were recently shown to be important for its  function (67). In fact, in PrnB and Put4p, all three gates of AdiC seem to be conserved. In AdiC, the proximal gate consists of Trp 202 , the middle of Trp 393 , and the distal of Glu 208 , Tyr 93 , and Tyr 365 . The corresponding residues of PrnB are Phe 248 , Trp 351 , Glu 255 , Tyr 127 , and Trp 422 (Fig. 1, Table 5). Like in AdiC, the main substrate binding site of PrnB and Put4p is found between the proximal and middle gates (Figs. 3 and 4A). Interestingly, our predicted second binding site of proline in PrnB and L-alanine in Put4p lies between the middle and distal gates, just upstream of the distal gate (Fig. 4, E and F). A second binding site, between the middle and distal gates, has also been proposed for AdiC (29), whereas disruption of the distal gate formation was suggested to lead to potential major conformational changes of AdiC allowing it to adopt an inward-open conformation (29). It is noteworthy that in both our models, the substrate did not directly interact with the nearby PrnB S130 or Put4p C203 . However our data clearly suggest an interaction between Thr 414 and Ser 130 . Substitutions of these two residues, which are very close at the structural model of PrnB, show allele-specific phenotypes, as the single Put4p-mimicking mutants T414S and S130C, but not the double S130C/T414S, are impaired in proline transport activity. Moreover, only T414S but not S130C or the double mutant could transport L-alanine, glycine, or ␤-alanine (Fig. 8). Despite this interaction that seems to determine both proline transport and specificity, neither mutant displayed significantly affected apparent K m for proline. The above data clearly suggest a physical or functional interaction between Ser 130 and Thr 414 . An intriguing explanation would be that a transient interaction of the substrate with this position would be related to major structural rearrangements of the transporter, thereby promoting substrate translocation. Consistently, Ser 357 , the corresponding residue of Thr 414 in AdiC, undergoes significant displacement upon occlusion of the transporter (29). In addition, our data are compatible with the suggestion that lysine transport by the T456S mutant of Can1, rather than being due to more efficient substrate binding, could result by facilitated structural transitions (59). The above statements are in line with the induced fit transition mechanism, proposed for the function of LeuT (74 -77), suggesting that successful interactions of the substrate in more than one site of the transporter promote the conformational changes of the transport cycle. Interestingly, our docking analysis in PrnB revealed a potential trajectory for proline (Fig. 5) showing putative intermediate interactions of the substrate during transition from one gate to another. Molecular dynamics simulations, along with mutational analysis, could verify the above.
On the Specificity Determination of YATs-Our current study of PrnB mutants substituted in residues not conserved in the YAT family, along with the detailed kinetic analysis of their specificity profile, provides important insight on the specificity determinants of this family (Table 5). Apart from the Ser 130 and Thr 414 residues discussed above, the variable in YATs Trp 351 of TMS8 was also shown to be a potential substrate binding residue, important for PrnB specificity and proline transport activity. Trp 351 seems to have a role analogous to Trp 202 in AdiC, which forms the middle gate (29). Interestingly, the Put4p resembling W351F mutant displayed enlarged specificity (Table 4) and could also transport L-alanine (Fig. 8). Analogous phenotypes were also obtained by Put4p-mimicking mutants in residues close to the potential distal gate (S130C, T414S, as discussed above) and the unwound part of TMS6 (F252L, S253G). The latter could determine specificity directly or indirectly. L-Alanine, docked at the putative second binding site of Put4p, appears to be very close to Leu 325 and Gly 326 , interacting via H-bond with Gly 326 (Fig. 4F). PrnB, that only weakly recognizes and does not transport L-alanine, possesses more bulky (Phe instead of Leu) and hydrophilic (Ser instead of Gly) residues at these positions (Fig. 9), which would pose steric effects and disturb L-alanine binding. Consistently, a S253G mutant displayed an impressive improvement of the K i for L-alanine and could also transport it (Table 5). Alternatively, the bulkier side chains of these residues could affect the local flexibility of the unwound part of TMS6, leading to decreased flexibility of the TMS6a of the highly selective PrnB (Fig. 3C). TMS6a includes important binding residues of the putative proximal gate (Phe 248 ), as well as residues important for the kinetic characteristics of PrnB (Lys 245 ). Interestingly, the corresponding TMS6a of AdiC was shown to undergo the most pronounced structural shift during the formation of the occluded conformation and was suggested to be involved in the closure of the proximal gate (29).
In summary (Table 5), our data show that the specificity of PrnB and Put4p, and by extension of other YATs, is determined by at least three ways. 1) Residues forming the middle gate (Trp 351 ); 2) residues between the middle and distal gates (Ser 130 and Thr 414 ); and 3) residues of TMS6 (Lys 245 ) and the unwound part of TMS6 (Phe 252 and Ser 253 ). These residues (Fig. 9) appear to determine specificity by directly interacting with the substrate, imposing steric constrains, and/or affecting local flexibility. In YATs of known function, even from different fungal species, residues in corresponding positions display interesting patterns (Fig. 1). For example, in TMS6, an Asn residue is present in the corresponding position of Lys 245 in all three basic amino acid permeases, accompanied by a Gln residue in place of Ile 251 . In TMS3, in position of Ser 130 , three fungal bicarboxylic acid permeases from different species possess Lys, whereas in TMS8, in position of Trp 351 , an aromatic residue present in both L-proline permeases is substituted by Asp in bicarboxylic acid permeases and Asn in Can1, Lyp1, and Alp1. The above observations suggest that our results could be directly applicable to other members of the YAT family, whereas our approach of comparing two orthologues from different organisms shows that interspecies comparisons can yield significant results concerning specificity determination.