Distinct specific interactions of the UapA transporter with membrane lipids are critical for dimerization, ER-exit and function

Transporters are transmembrane proteins that mediate the selective translocation of solutes across biological membranes. Recently, we have shown that specific interactions with plasma membrane phospholipids are essential for formation and/or stability of functional dimers of the purine transporter, UapA, a prototypic eukaryotic member of the ubiquitous NAT family. Here, we show that distinct interactions of UapA with specific or annular lipids are essential for ab initio formation of functional dimers in the ER or ER-exit and further subcellular trafficking. Through genetic screens we identify mutations that restore defects in dimer formation and/or trafficking. Suppressors of defective dimerization restore ab initio formation of UapA dimers in the ER. Most of these suppressors are located in the movable core domain, but also in the core-dimerization interface and in residues of the dimerization domain exposed to lipids. Molecular Dynamics suggest the majority of suppressors stabilize interhelical interactions in the core domain and thus assist the formation of functional UapA dimers. Among suppressors restoring dimerization, a specific mutation, T401P, was also isolated independently as a suppressor restoring trafficking, suggesting that stabilization of the core domain restores function by sustaining structural defects caused by abolishment of essential interactions with specific or annular lipids. Importantly, introduction of mutations topologically equivalent to T401P into a rat homologue of UapA, namely rSNBT1, permitted the functional expression of a mammalian NAT in A. nidulans. Thus, our results provide a potential route for the functional expression and manipulation of mammalian transporters in the model Aspergillus system. Author Summary Transporters are proteins found in biological membranes, where they are involved in the selective movement of nutrients, ions, drugs and other small molecules across membranes. Consequently, their function is essential for cell viability, while their malfunction often results to disease. Recent findings have suggested that transporter functioning depends on proper interactions with associated membrane lipids. In this article, using UapA, a very well-studied transporter from a model fungus (Aspergillus nidulans), we show that two types of specific interactions with lipids are essential for tight and specific association of two UapA molecules in a single functional unit (UapA dimer), and for targeting to the cell membrane and transport activity. The first type of interaction concerns specific lipids associating with positively charged amino acids at the interface of the UapA dimer, whereas the other type involves lipids that interact with charged amino acids at the outer shell of the transporter. Most interestingly, defects due to abolishment of UapA-lipid interactions were shown to be restored by mutations that increase UapA stability. Using this information, we genetically manipulated and increased the stability of a mammalian transporter (rSNBT1), and thus achieved its functional expression in the experimentally tractable system of A. nidulans.


Introduction
Transporters are essential transmembrane proteins that catalyze the uptake or efflux of metabolites, nutrients, ions and drugs across biological membranes. Transporter malfunction, due to genetic mutations or metabolic defects, results in significant cellular or organismal disruption [1] (http://www.tcdb.org/). Despite their biological and apparent medical importance, knowledge on structure-function relationships in transporters is limited compared to extramembrane hydrophilic proteins. This is in part due to complexity associated with their translocation and co-translational folding into a membrane lipid bilayer (the ER in eukaryotes or the plasma membrane in prokaryotes). Additionally, in eukaryotes, transporters follow specific membrane trafficking, turnover or recycling routes, which add further complications in understanding their mechanisms regulation of expression, function and turnover [2][3][4][5][6].
A further contributory factor is their functional and structural dependence on specific membrane lipids, something which is only recently being explored in detail [7,8].
Further complications for transporter study arise from difficulties in expressing sufficient quantities for downstream structural studies, functional reconstitution in proteoliposomes, or in measuring their kinetics in intact cells where the presence similar transporters with overlapping specificities complicate the analysis [9].
One of the best studied eukaryotic transporters is the UapA xanthine-uric acid/H + symporter of the filamentous ascomycete Aspergillus nidulans [10,11]. This is due to development of rigorous genetic, biochemical and in vivo cellular approaches, uniquely available in the model system of A. nidulans, and more recently, by structural and biophysical studies. The high-resolution, inward-facing, structure of a conformationally locked mutant of UapA (G411V 1-11 ), together with genetic and 6 functional dimers. The total lack of function of the lipid binding site UapA mutant (R287A/R478A/R489A) further suggests that specific interactions with lipids are also necessary for the mechanism of transport per se. Rather surprisingly, the GFP tagged inactive triple arginine mutant can still properly traffic to the PM, which means that either monomeric UapA translocates normally to the PM, or that UapA dimer is initially formed in the ER and traffics to the PM, but then becomes unstable and dissociates into non-functional monomers. As MD simulations further predicted that lipids could also bind to the outermost, membrane-facing regions of the core domains of the UapA dimer [20] further investigation was needed to fully understand the role of membrane lipids in UapA folding, subcellular traffic and transport function.
Here we investigate further the role of residues Arg287, Arg478 and Arg479 in UapA dimer formation and/or stability, and study the role of additional interactions of UapA with annular lipids. Using mutational analyses and genetic screens for suppressors of mutations affecting putative lipid-binding residues, we confirm that that Arg287, Arg478 and Arg479 are essential for ab initio dimerization in the ER, but redundant for membrane traffic, whereas interactions between distinct residues (Lys73, Arg133 and Arg421) and annular lipids are essential for proper folding, ERexit and membrane traffic. Our results reveal that genetic modification of residues in the core domain of UapA can compensate for the "lost" lipid interactions in the original mutants. Importantly, using information on a specific core residue that proved to have a key role in stabilizing UapA, Thr401, we genetically manipulate and achieve in functionally expressing, for the first time, a mammalian homologue of UapA in A. nidulans. This opens the way for functional expression of mammalian NAT transporters, including those essential for vitamin C transport in humans [22.23], in the model Aspergillus system.

Results
Residues Arg287, Arg 479 and Arg479 are crucial for ab initio dimerization of

UapA in the ER
We have previously shown that arginine residues 287, 478 and 479 are essential for phospholipid-dependent functional dimerization of UapA at the PM [20]. To further understand the basis of the functional defect in the triple R287A/R478A/R479A mutant, here we examined whether loss of dimerization occurs ab initio at the level of the ER, or whether what we have previously observed was due to instability of UapA dimers at the PM. For this, we used a previously described BiFC assay [21], adapted to follow the sorting and subcellular localization of de novo made UapA. This is based on time-course experiments following the subcellular localization of de novo made alcAp-UapA-GFP, which showed that after 1 h of transcriptional derepression UapA-GFP is hardly visible, but at 2-3 h it labels the ER and at 4 h appears mostly in the PM [6]. Using this system, we followed reconstitution of split-YFP, via UapA dimerization, in a strain containing two copies of the alcAp-uapA gene, tagged either the N-or the C-part of the yfp orf [21]. To follow de novo made UapA in young mycelia, we repressed the transcription of alcA p -uapA-yfp n and alcAp-uapA-yfp c overnight (16 h in MM at 25 o C in glucose MM), a period to allow conidiospores germination and young hyphae development, and then shifted the culture to fructosederepression medium for 1-4 h of growth. We performed this assay using wild-type UapA and the triple R287A/R478A/R479A mutant. Fig 1A shows that in the wildtype control strain our assay detects early reconstitution of split-YFP fluorescence at the ER network at 3h, but fails to do so in the triple R287A/R478A/R479A mutant, where only a weak signal is observed. After 4h expression, the totality of wild-type UapA fluorescent signal marks the PM, whereas a weak cytoplasmic fluorescent signal and very low cortical localization is observed in the R287A/R478A/R479A mutant. This shows that Ala substitutions of the three Arg led to significant reduction of apparent UapA dimerization at the ER membrane, and further suggests that specific contacts with ER lipids might be a prerequisite for dimerization. Surprisingly, the interactions of R287, R478 and R479 with lipids and dimerization proved redundant for sorting of the mutant UapA to the PM, as judged by the normal PM localization of R287A/R478A/R479A tagged with intact GFP (Fig 1B). These findings suggest that in the R287A/R478A/R479A mutant non-functional UapA monomers or partially misfolded dimers can still be secreted to the PM. Eight of the 13 suppressor were isolated more than once showing that mutagenesis was fairly saturated and also confirming that the amino acid changes detected are responsible for suppression. Fig 2A shows the topology of suppressor mutations in the UapA structure.
The 13 distinct suppressors grew well, albeit slightly less than an isogenic strain expressing wild-type UapA, on uric acid or xanthine (Fig 2B). Among the suppressors, S119T scored as a thermosensitive mutant, growing very weakly on both UapA substrates at 37 o C, similar to the original R287A/R478A/R479A. Direct uptake assays, measuring the transport rate of radiolabeled xanthine [24], were used to estimate the effect of the suppressor mutation on UapA transport kinetics. Fig 2C shows that in most suppressors UapA transport rates were re-established at ~15-30% of the wild type protein, a level known to be the threshold for conferring visible UapA-mediated growth on uric acid or xanthine. Highest transport rates were obtained in suppressors E286Q (~70%) and E286K (~51%). In fact, the relative uptake differences of some suppressors (L192F, A396P, T401P, T401F or L431F) compared with the original R287A/R478A/R479A mutant were marginal. It should be noted that in growth tests purines are added to concentrations at the 1-2 mM level in order to be used as N sources, while in uptakes radiolabeled xanthine is used at submicromolar range (0.3-0.5 μΜ) for technical reasons. Thus any mutation that causes significant reduction in substrate affinity might score as an apparent loss-of-function mutation in uptakes, but still can allow normal growth on the relative substrate when this is supplied at mM concentration. To test whether the low apparent transport capacity of suppressors is due to reduced affinity for xanthine, we estimated the approximate K m of several of the suppressors relative to the original R287A/R478A/R479A mutation or a wild-type UapA control. We found no significant reduction of affinities for xanthine in all suppressors tests (Fig 2C, on top of histograms).

Genetic suppressors R287A/R478A/R479A re-establish UapA dimerization
As phospholipid binding has been shown to be essential for formation of functional UapA dimers, the suppressors of R287A/R478A/R479A could either restore functional dimerization, or confer transport activity to UapA monomers. To test these two alternatives, we performed Bifluorescence Competitions (BiFC) assays to follow UapA dimerization of de novo made UapA in selected suppressors (I157F, L234M, and T401P) and control strains, as described earlier for the original R287A/R478A/R479A mutant.  (Fig 4 A, B). The Finally, Leu431 is located between with Ala87 (TMS1), Met90 (TMS1) and Leu120 (TMS2). Contrary to the above mentioned residues, Leu234 is located in the dimerization domain, interacting mainly with Ile158 in TMS3. The suppressor mutations and more particularly L192F, T401F, L431F, S119T and V153T seem to enhance the above mentioned interactions and stabilize mainly the core domain. In order to verify this hypothesis, models of UapA including selected suppressor mutations were constructed and subjected to geometry optimization and short MD calculations. As shown in Fig 4C, D, E, F the phenyl moieties of the mutated residues I157F, T401F and L192F are accommodated in the space between the other lipophilic residues increasing hydrophobic interactions between TMS8, TMS3, TMS4 and TMS10. The introduction of a methyl group in the mutation S119T will also enhance, albeit to a lower degree, interactions with Val153 (TMS3) and Met400 (TMS10).   Mutant R133A/R421A scored as an apparent total loss of function mutant, as it did not grow in either uric acid or xanthine. Finally, the triple K73A/R133A/R421A also scored as a total loss of function mutant. Overall, Arg133 and Arg421 proved very important for UapA function, while Lys73 when present in the context of R133A was critical for UapA specificity for xanthine, but not for uric acid. For each mutant version of UapA we also assessed localization to the PM, compared to the wild-type UapA, using the GFP epitope attached to UapA. This analysis (right panel in Fig 5B) showed that most single mutations and Y137A/R138A, which led to no defect in UapA transport activity, allowed normal localization of UapA in the PM, as probably expected. On the other hand, R133A and K73A/R421A mutants showed partial UapA retention in perinuclear ER membranes. Finally, mutants with apparently defective (K73A/R133A) or lost (R133A/R421A and K73A/R133A/R421A) transport activity showed partial or total retention into the ER. Growth tests and subcellular localization were in good agreement with measurements of rates of radiolabeled xanthine accumulation, which confirmed that the simultaneous presence of Arg133 and Arg421 is essential for transport activity, whereas Lys73 is critical for xanthine uptake only when Arg133 is also replaced by an Ala (Fig 5C). For the double mutant K73A/R133A which showed reduced growth on xanthine, we also measured the K m for xanthine and showed that this was very close to that of the wild-type UapA (3.6 versus 5±2 μΜ), suggesting that reduced growth on xanthine is not assigned to reduced substrate binding.
We investigated whether the lack of UapA sorting out of the ER in the triple K73A/R133A/R421A mutant is related to its ability to dimerize. For this, we employed BiFC assays, as described before for the R287A/R478A/R479A mutant. We used a genetic approach to understand how residues Lys73, Arg133 and Arg421 might affect UapA sorting to the PM by isolating suppressor mutations that restored UapA-mediated growth on uric acid in the mutant K73A/R133A/R421A, as described earlier for the isolation suppressors of the R287A/R478A/R479A mutant. We obtained, purified and sequenced the uapA orf from 9 suppressors. Rather surprisingly, all proved to include the same single mutation, namely T401P, in addition to the original K73A/R133A/R421A triple mutation (Fig 6A). Noticeably, T401P was also isolated among the suppressors of the dimerization-defective R287A/R478A/R479A mutant. Growth tests showed that although T401P confers normal growth on xanthine and uric acid in the context of K73A/R133A/R421A (Fig   6B), this occurs by only partial restoration of UapA-mediated transport of these purines (Fig 6C). We also constructed by targeted mutagenesis plasmid vectors carrying K73A/R133A/R421A/T401P and T401P alone, introduced them in the A. context, but also partially restores ER-exit, sorting and function in the K73A/R133A/R421A context. transporters. In contrast, the strain expressing rSNBT1-N390T showed clear sensitivity to 5-FU, and although it could not grow on any purine, 5-FU sensitivity could be competed in the presence of excess purines or pyrimidines that are known rSNBT1 substrates (e.g. hypoxanthine, uracil or thymine) (Fig 7B). Importantly, rSNBT1-dependent 5-FU sensitivity was Na + -dependent (Fig 7C), compatible with the physiological mechanism of functioning of rSNBT1 in rat [22,23].

Manipulation of a residue topologically equivalent to T401P leads to functional expression of a mammalian NAT homologue in
To further confirm that the phenotype observed in the relative transformants is due uniquely to the genetically introduced rSNBT1-N390T protein, we also analyzed the meiotic progeny of an rSNBT1-N390T transformant. A. nidulans undergoing meiosis during a process "selfing" [31] is prone to high recombination rates that often lead to loss of sequences introduced by transformation. showed that in all cases 5-FU sensitivity was conserved, a fluorescent signal from the rSNBT1-N390T protein tagged with GFP was also conserved. In contrast, all selected colonies that acquired resistance to 5-FU lost the fluorescent signal of the rSNBT1-N390T-GFP (not shown).
Finally, we also performed direct measurements of radiolabeled uracil accumulation or competition in the strain expressing rSNBT1-N390T, which further confirmed the functionality of the rat transporter in A. nidulans (Fig 7D). Noticeably, the low apparent transport capacity of rSNBT1-N390T in respect to some of is substrates (e.g. hypoxanthine or uric acid) in A. nidulans is very probably due by the observation that mutation N390T restores sorting of rSNBT1 to the PM only partially (Fig 7E).

Discussion
It is becoming well-established that the physicochemical nature of lipid bilayers and specific lipid composition of membranes affect transporter folding, oligomerization, subcellular trafficking, function and turnover [8,[32][33][34][35][36][37][38]. For transporters conforming to the 5+5 inverted repeat (IR) or LeuT fold, similarities in structurally resolved lipidprotein interactions suggest common ways in which transporter structure and function are supported by lipid interactions [39]. These are likely to include stabilization of the inverted repeat topology, but also mechanistic roles as major determinants of the alternating access mechanism of secondary transporters. and normal sorting to the PM were still evident, suggested that lipid binding may also be directly essential for the mechanism of transport. Here, we studied further the role of Arg287, Arg478 and Arg479 and in parallel investigated the role of binding of annular lipids at distinct residues of UapA. We showed that Arg287, Arg478 and Arg479 are essential for early de novo formation the ER membrane, a process absolutely essential for transport activity, albeit not for sorting to the PM. In parallel, we identified distinct positively charged residues (Lys73, Arg133 and Arg421), exposed to the PM membrane bilayer, which are essential for ER-exit, sorting to the PM and transport activity, but apparently not essential for initial formation of dimers in the ER. Thus, the two sets of positively charged residues define two distinct sites of interaction of UapA with membrane lipids, both essential for function, albeit due to different reasons. The distinct defects caused by Ala substitutions at the dimer interface or those exposed to the inner side of the PM bilayer are well supported by for these last two suppressors is that they replace directly the interactions with lipids of the mutated nearby Arg287 (i.e. in R287A). This is also in line with the fact that these are the strongest isolated suppressors in respect to UapA transport activity.
Thus, our findings, especially those concerning type I and II suppressors, strongly suggest that by stabilizing the core, which is the motile part of the monomeric units that undergoes dynamic up and down elevator-like sliding, the dimer is also stabilized and thus function is restored.
Interestingly, all isolated suppressors of the "trafficking" mutant

Media, strains and growth conditions
Standard complete (CM) and minimal media (MM) for A. nidulans growth were used. nidulans protoplast isolation and transformation was performed as previously described [40]. Growth tests were performed at 25 or 37 o C for 48 h, at pH 6.8.

Standard molecular biology manipulations and plasmid construction
Genomic DNA extraction from A. nidulans was performed as described in FGSC  [44] followed by cloning of the uapA ORF. UapA or rSNBT1 mutations were constructed by oligonucleotide-directed mutagenesis or appropriate forward and reverse primers (S2 Table). Transformants arising from single copy integration events with intact UapA ORFs were identified by PCR analysis.

Uptake assays
Kinetic analysis of UapA or rSNBT1 activity was measured by estimating uptake

Epifluorescence microscopy
Samples for standard epifluorescence microscopy were prepared as previously described [45,46]. software with an energy-based algorithm [47]. A loop refinement routine was also implemented.

Induced Fit Docking of Uracil on rSNBT1
Protein preparation using OPLS2005 force field [48] and molecular docking was Non-bonded interactions were described with a Lennard-Jones potential with a cutoff distance of 1 nm and an integration step of 2 fs was implemented. The system was progressively minimized and equilibrated using the GROMACS input scripts generated by CHARMM-GUI and the temperature and pressure was held at 303.15 K and 1 bar respectively [54]. The resulting equilibrated structures were then used as an initial condition for the production runs of 100 ns with all the constraints turned off. Production runs were subsequently analyzed using GROMACS tools and all images and videos were prepared using VMD software [55].

In silico mutation of I157F, L192F and L431F on UapA
Staring from the crystal structure of UapA manual mutation of I157F, L192F and L431F was performed using "mutation" command on Maestro v11.5 (Schrödinger

Release 2018-1). Each resulting structure was inserted to Protein Preparation
Wizard Workflow as implemented on Maestro v11.5. Restrained minimization was converged when heavy atoms RMSD was greater than 1 Å.
In brief the mutants were individually expressed as C-terminally GFP8His tagged constructs in S. cerevisiae FGY217 cells (12L), using vector pDDGFP2. Yeast cells were incubated at 30 C with shaking at 300 rpm to an OD 600 of 0.    Results are averages of three measurements for each concentration point. SD was 20%.     Homology modeling of the topology of rSNBT1, constructed using, as described in Materials and methods, the inward-facing conformation of the crystal structure of the UapA dimer. The two residues mutated and functionally analyzed, N390 and G391, are shown as black spheres. The location of uracil, the major substrate of rSNBT1, is also depicted, as determined by dynamic docking (left panel). In the right panel, a zoomed-out picture of the substrate binding site depicting the major interactions of uracil with specific residues. (B) Growth tests of isogenic A. nidulans strains expressing single-copy wild-type rSNBT1 or its mutated versions rSNBT1-N390T, rSNBT1-N390P and rSNBT1-G391P. A negative control strain (i.e. the recipient Δ7 strain that has null activity for nucleobase transport; see text) is included for comparison. Growth tests are performed at 37 o C on MM supplemented with Na + (100 mM NaCl). 10 mM NaNO 3 is used as a control N source unrelated to purine transport activities in all tests scoring resistance/sensitivity to 5FU (rows 1-7). Rows 3-7 represent in vivo competion assays scoring the ability of excess purines ( 2mM) to compete with the uptake of 5FU (100 μΜ), and thus revert 5FU sensitivity. Χ is xanthine, A is adenine, T is Thymine, HX is hypoxanthine and U is uracil. Notice that T, HX and U competed 5FU uptake and suppressed sensitivity. Growth was also