Structure of the sodium-dependent phosphate transporter reveals insights into human solute carrier SLC20

See allHide authors and affiliations

Science Advances  07 Aug 2020:
Vol. 6, no. 32, eabb4024
DOI: 10.1126/sciadv.abb4024


Inorganic phosphate (Pi) is a fundamental and essential element for nucleotide biosynthesis, energy supply, and cellular signaling in living organisms. Human phosphate transporter (hPiT) dysfunction causes numerous diseases, but the molecular mechanism underlying transporters remains elusive. We report the structure of the sodium-dependent phosphate transporter from Thermotoga maritima (TmPiT) in complex with sodium and phosphate (TmPiT-Na/Pi) at 2.3-angstrom resolution. We reveal that one phosphate and two sodium ions (Pi-2Na) are located at the core of TmPiT and that the third sodium ion (Nafore) is located near the inner membrane boundary. We propose an elevator-like mechanism for sodium and phosphate transport by TmPiT, with the TmPiT-Na/Pi complex adopting an inward occluded conformation. We found that disease-related hPiT variants carry mutations in the corresponding sodium- and phosphate-binding residues identified in TmPiT. Our three-dimensional structure of TmPiT provides a framework for understanding PiT dysfunction and for future structure-based drug design.


Maintaining phosphate (Pi) balance is essential for the growth and development of all organisms, and phosphate transporters are key factors in sustaining phosphate homeostasis in humans, plants, fungi, and bacteria. Solute carriers (SLCs) in humans constitute a group of more than 400 membrane transporters allocated to 65 families based on sequence homology. They transport a variety of solutes, such as inorganic ions, amino acids, and neurotransmitters (1). In humans, Pi is translocated into cells by two major secondary active transporters, i.e., the sodium-dependent phosphate transporter SLC20 (PiT) and SLC34 (NaPi-II) families (2), which prefer monovalent (H2PO4) and divalent (HPO42−) phosphate, respectively. These families use both the free energy provided by the transmembrane electric field and the Na concentration gradient to cotransport Pi through the membrane with proposed stoichiometries of 2Na:1Pi or 3Na:1Pi (2). The NaPi-II family mainly functions to regulate renal and intestinal Pi transport and comprises three members, named NaPi-IIa, NaPi-IIb, and NaPi-IIc (3). The NaPi-IIa and NaPi-IIb couple the translocation of three Na ions and one divalent Pi (HPO42−) in each cycle as electrogenic transporters, whereas NaPi-IIc is an electroneutral transporter with a transport stoichiometry of two Na to one divalent Pi (HPO42−) (2, 3). NaPi-II transporters share a similar transport mechanism, with two opposed reentrant helical hairpins (HP1/HP2) being responsible for Na/Pi entry (4).

Human PiTs (hPiT1 and hPiT2) were initially identified as the cell surface receptors for Glvr and Ram retroviruses (5), and they were then subsequently found to cotransport Pi and Na. PiTs exhibit a preference for monovalent Pi (H2PO4), with a binding affinity of ~50 μM in a pH range of 6.2 to 6.8, which was shown to be lower at both pH 5.0 and pH 8.0 (6). They are ubiquitously expressed in various organs, including the kidney, liver, and brain (2, 7). Dysfunction of hPiTs causes numerous diseases, including vascular and brain calcification (811) and neuropsychiatric disorders (10, 11), but the molecular mechanism underlying these transporters remains elusive.

PiTs are present in all kingdoms, including sodium-dependent PiT from Saccharomyces cerevisiae ScPHO89 (12), liverwort Marchantia polymorpha MpPTB (13), and Plasmodium falciparum PfPiT (14) or proton-dependent PiT from Arabidopsis thaliana AtPht2 (15) and Escherichia coli EcPitA (16). These PiT proteins have a transmembrane region at both termini linked by a loop or an additional intracellular soluble domain (S domain) in the cellular region (>200 residues) (fig. S1) (17, 18). The two terminal transmembrane regions contain a unique sequence of ~150 amino acids known as ProDom domain 001131 (PD001131), which harbors four highly conserved sequences: GΦNDΦ, GxxxxGxxVxxT, PΦSxT, and IxxxWΦ (x, any amino acid; Φ, hydrophobic residue) (1820).

Thermotoga maritima is a hyperthermophilic bacterium that makes it an ideal model organism for integration of biochemical and structural experimental approaches (21). We show that sodium-dependent PiT from T. maritima (TmPiT) belongs to the SLC20 family, and its transmembrane domain exhibits similar protein characteristics (20) and sequence homology to that of the hPiTs (hPiT1, 62% similarity and 38% identity; hPiT2, 61% similarity and 39% identity) (fig. S1). Structural information for a sodium-dependent PiT had been unavailable. Here, we report the crystal structure of TmPiT, a homology of hPiT. We determined the phosphate-binding affinity and uptake ability of TmPiT. Meanwhile, we solved the crystal structure of TmPiT bound to sodium and phosphate (TmPiT-Na/Pi) at 2.3 Å. The TmPiT-Na/Pi complex is in an inward occluded state and exits as a dimer, with the two subunits showing different conformations.


Crystal structure of the TmPiT-Na/Pi complex

We overexpressed TmPiT in S. cerevisiae and performed x-ray crystallography to determine its structure. In the TmPiT-Na/Pi complex, the transport and scaffold domains are formed by 12 transmembrane (TM) helices (Fig. 1, A and B). The transport domain of TmPiT adopts a five-helix “5 + 5” fold and is arranged into two inverted PD001131 repeats, i.e., N-PD001131 (TM1/TM2/HP1a-HP1b/TM3) and C-PD001131 (TM6/TM7/HP2a-HP2b/TM8) (fig. S2A). This five-helix arrangement is distinct from that of the leucine transporter superfamily (22). HP1a-HP1b (HP1) and HP2a-HP2b (HP2) are reentrant helical hairpins that have been reported previously in the aspartate transporter Gltph (SLC1) (23) and the dicarboxylate transporter VcINDY (SLC13) (24) (Fig. 1A and fig. S2A). The four conserved motifs in the two PD001131 repeats are equally distributed to five helices, and all of them are well aligned structurally (Fig. 1, A and D, and fig. S2B). TmPiT is a dimer with dimensions of 83 Å by 61 Å by 55 Å, with the N and C termini being located in the extracellular region (Fig. 1C). The dimer interface between the two subunits (A and B) is formed by TM2/TM7 from the transport domain and TM4/TM5 from the scaffold domain through hydrophobic interactions (table S2 and fig. S2D), and this interface has a buried area of 1283.7 Å2. The electrostatic surface potential of the Pi-binding pocket is shown in Fig. 1E.

Fig. 1 Structure of the TmPiT-Na/Pi complex.

(A) Topology of TmPiT with 12 transmembrane helices that are divided into a transport domain with two inverted topology repeats, N-PD001131 (TM1 to TM3 and HP1a-HP1b, in magenta) and C-PD001131 (TM6 to TM8 and HP2a-HP2b, in blue), and a scaffold domain (TM4/5, in yellow). (B) Ribbon diagram of the TmPiT-Na/Pi complex consisting of a transport domain with N-PD001131 (in magenta) and C-PD001131 (in blue) and a scaffold domain (in yellow). The Pi and Na ions are shown in CPK mode and as purple spheres, respectively. The transmembrane helices in (B) are colored and numbered based on (A). (C) Ribbon diagram of the TmPiT dimer (in bright blue and cyan for subunits A and B, respectively) with height and width dimensions of 61 and 80 Å, respectively. (D) Sequence alignment between N-PD001131 and C-PD001131 for the four highly conserved motifs (boxed and labeled 1 to 4, respectively): GΦNDΦ, GxxxxGxxVxxT, PΦSxT, and IxxxWΦ (x, any amino acid; Φ, hydrophobic residue) are boxed. (E) The electrostatic surface potential is shown for the TmPiT-Na/Pi complex (red, blue, and white for negative, positive, and neutral potentials, respectively). To show the binding pocket, Pi is shown as a sphere, and loops L7 and LHP2 are displayed as ribbons.

TmPiT is a sodium-dependent, high-affinity PiT

Next, we studied the substrate specificity and inhibition of TmPiT. We used microscale thermophoresis to measure the phosphate-binding affinity (Kd) of TmPiT, which was 57.0 ± 1.1 μM (Fig. 2A). The ligand selectivity of TmPiT was investigated using the phosphate analogs arsenate and sulfate, and we calculated a Kd of 6.3 ± 1.0 mM for the former, and there was no apparent interaction with the latter (Fig. 2A), revealing the specific phosphate binding of TmPiT. We also evaluated inhibition of TmPiT by phosphonoformic acid (PFA), a specific inhibitor of NaPi-II that does not inhibit Pi transport by PiTs (6, 25). As expected, we observed limited binding of PFA to TmPiT, with a calculated Kd of 2.4 ± 1.0 mM (Fig. 2A), supporting that these two types of transporters have distinct binding environments. We further investigated the driving force for phosphate transport in TmPiT using a proteoliposomal Pi uptake assay. In the presence of 120 mM NaCl, Pi uptake was 11.7 ± 0.2 μmol/mg. In the absence of NaCl, Pi uptake was abolished (Fig. 2B), confirming that phosphate transport by TmPiT is driven by sodium.

Fig. 2 Na/Pi binding.

(A) Ligand selectivity of TmPiT was determined by a microscale thermophoresis binding assay. n.d., not determined. (B) Phosphate (32P) uptake activity of TmPiT was measured using proteoliposomes containing TmPiT. The data in (A) and (B) represent the means ± SD of three independent experiments. (C) Pi- and Na-binding pocket. The Fobs-Fcalc (observed and calculated structure factors) electron density maps of Pi and Na are shown at 8σ and 6σ, respectively. The transmembrane helices TM1, TM6, HP1a-HP1b, and HP2a-HP2b (labeled) are involved in Pi and Na binding. The Pi- and Na-binding residues are shown in CPK mode and as purple spheres, respectively. (D) Zoomed-in view of the Pi-2Na–binding pocket, showing interacting residues. (E) Zoomed-in view of Nafore binding, showing the pentacoordination residues.

Na/Pi binding

To further understand the mechanism underlying sodium-driven phosphate transport by TmPiT, we analyzed the TmPiT-Na/Pi crystal structure to identify the main residues involved in Pi and Na binding. We observed one phosphate and three sodium ions in our crystal structure of the TmPiT-Na/Pi complex; two of the sodium ions and the phosphate (hereafter Pi-2Na) were located at the core of TmPiT (Fig. 2C), and the third sodium (termed Nafore) was situated near the inner membrane boundary (Fig. 2C).

The Pi-2Na–binding site is formed by TM1, the HP1a-HP1b loop (HP1 tip), TM6, and the HP2a-HP2b loop (HP2 tip) (Fig. 2C). The phosphate was associated with the two sodium ions as “Na1-Pi-Na2” through two aspartates, D22 and D258, with the phosphate being located 4.8 Å away from each sodium ion (Fig. 2D). This phosphate was tightly bound via 12 interactions with eight conserved residues, including D22 (TM1), D258 (TM6), and six polar residues, i.e., S105/T106/T107 (HP1b) and S345/T346/T347 (HP2b) (Fig. 2D and table S3). Na1 was bound via a pentacoordinated interaction with N21/D22/V104/T347/K314 in subunit A (Fig. 2D), whereas K314 was replaced by a water molecule (Wat) in subunit B (table S3). In addition, we found that Na2 was bound via a pentacoordinated interaction with N257/D258/T107/I344/Wat in both subunits (Fig. 2D and table S3). The coordination of both sodium ions is similar to that of the conserved Asp/Asn/Thr residues. All of these Pi-2Na–binding residues belong to the GΦNDΦ and PΦSxT motifs of PiTs (Fig. 1D and table S1), with D22 and D258 being involved in both Pi and Na binding. The detailed interactions are listed in table S3. The Nafore was located near TM1, TM6, and HP2a, with a pentacoordination through the T29, Q243, S247, and D327 residues (Fig. 2, C and E), which are highly or partially conserved among PiTs (fig. S1).

Functional role of the three sodium ions

The three sodium ions and the phosphate are located along TM1 and TM6, with Pi-2Na facing outward and Nafore facing inward toward the membrane. Our data show that the three Na ions are coordinated by three aspartate residues (D22, D258, and D327) (Fig. 3A), which explains why Pi uptake is abolished in the corresponding mutants of hPiT, i.e., D28N, D506N, and E575Q (11, 20, 26). We conducted a mutational study of TmPiT and found that the D22A and D258A mutants maintained a Pi-binding affinity similar to that of the wild type but both presented Pi uptake defects (Fig. 3, B and C). Moreover, the D22/258A double mutant and the D327Q mutant failed to bind Pi (Fig. 3C). Comparing the three aspartate mutations, D327Q reduced the apparent Pi-binding affinity. It suggests that Nafore shows a different functional role in Na1 and Na2. In addition, Na1 is located ~7 Å away from the inner membrane, Na2 is ~12 Å away from the outer membrane, and Nafore is near the inner membrane. Nafore might bind TmPiT first and reorientate TM1 and TM6 into the correct conformation to promote Pi binding (Figs. 2C and 3A). Therefore, we suggest that Na1 and Na2 act as escorts for phosphate binding and Nafore might regulate the binding.

Fig. 3 Mutational study of the residues involved in Na coordination and conformational change.

(A) Residues D22 (TM1), D258 (TM6), and D327(HP2) are the three residues interacting with Na1, Na2, and Nafore, respectively. (B) D22A and D258A mutant proteins were reconstituted into proteoliposomes to determine their phosphate (32P) uptake activity. (C) Phosphate bindings of the TmPiT mutants D22A, D258A, D22/258A, and D327Q were determined by microscale thermophoresis. The conformational change of subunits A and B around the inner gates is shown in (D) and (E), respectively. Two inverted repeated domains are shown: HP1 from N-PD001131 (in magenta in both subunits) and HP2 and TM8 from C-PD001131 (in bright blue and cyan for subunits A and B, respectively). The structures shown include Pi and Na, and residues K314 and W378 are labeled. The Pi and Na ions are shown in CPK mode and as purple spheres, respectively. Residues K314 and W378 are shown in CPK mode. The accessible volumes of the exit region were calculated using CASTp (Computed Atlas of Surface Topography of proteins) (27) and are shown in brown. (F) Phosphate bindings of the TmPiT mutants W139A and W378A were determined by microscale thermophoresis. The data in (B), (C), and (F) represent the means ± SD of three independent experiments.

Conformational changes

In the TmPiT-Na/Pi complex, we observed notable structural differences between subunits A and B, which are reflected by a root mean square deviation of 1.8 Å for the Cα atoms (fig. S2C), mainly attributable to TM8 and the intracellular loops L7 and LHP2 (Fig. 3, D and E). The accessible volumes of this region, calculated using CASTp (Computed Atlas of Surface Topography of proteins) (27), are 68 and 252 Å3 for subunits A and B, respectively (Fig. 3, D and E, and fig. S3). In subunit A, L7 is within a loosely structured omega loop inside the protein, and Na1 directly interacts with the semiconserved residue K314 (Figs. 2D and 3D and fig. S1). However, in subunit B, L7 is within a three-turn helix that exposes K314 to the solvent, and the interaction with Na1 is eliminated (Fig. 3E). LHP2 in subunit B is longer and more disordered than in subunit A, because there is one less helix turn in HP2b, thereby providing a larger space for Na/Pi release. LHP2 represents the loop that links HP2b and TM8, so helix length and the flexibility of LHP2 may affect TM8 conformation.

Furthermore, there is a notable structural difference between subunits A and B at TM8, reflected by a helix kink in subunit B involving the highly conserved tryptophan W378 that is located in the middle of TM8 and has a distinct orientation (Fig. 3, D and E, and fig. S3). W378 (TM8) and W139 (TM3) are the corresponding residues in the two inverted repeats, C-PD001131 and N-PD001131, respectively (Fig. 1D). They are highly conserved in the SLC20 family and belong to the IxxxWΦ motif. In our mutational studies, both the W139A and W378A mutants lost some Pi-binding ability (Fig. 3F). Patients with primary familial brain calcification disease carry a novel duplication of 626WFVT in hPiT, which corresponds to 378WLLI in TmPiT (28). This mutation does not affect expression of hPiT, but it does alter subcellular protein localization and impairs Pi internalization (28).


The phosphate-binding mode of TmPiT is unique due to the symmetric repeated residues, which are unlike those of other phosphate transporters, such as E. coli glycerol-3-phosphate transporter (GlpT) (29) and a fungal phosphate transporter (PiPT) (30). GlpT [Protein Data Bank (PDB) ID: 1PW4] is a phosphate antiporter of the SLC37 family that couples the accumulation of sugar phosphates to the downhill release of phosphate under physiological conditions but is not primarily responsible for net phosphate movement (29). PiPT (PDB ID: 4J05) is a Pi:H+ symporter in the fungus Piriformospora indica. It is a homolog of human SLC22, which functions as an organic anion and cation transporter (30), and presents low sequence homology to TmPiT. Structures of both GlpT and PiPT reveal folding typical of the major facilitator superfamily, with a rocker-switch transport mechanism. In GlpT, one histidine and two arginine residues are proposed to be involved in the phosphate binding (29). In PiPT, the phosphate is buried at the domain interface and coordinated by the aromatic residues Y150, W320, and Y328 and the polar residues Q177, D324, and N431 (30). However, in TmPiT, the Pi-2Na is trapped in a symmetric and hydrophilic environment and bound by D22/S105/T106/T107 and D258/S345/T346/T347. Thus, TmPiT shows a distinct phosphate-binding environment.

In human, both PiT and NaPi-II are sodium-dependent phosphate transporters. When the TmPiT-Na/Pi crystal structure was compared with the modeled NaPi-II (4, 31), we found that the transmembrane domains of PiT and NaPi-II proteins exhibit similar secondary structure features, especially with regard to a common inverted repeat topology. Both structures are capable of binding three Na ions. In TmPiT, Pi-2Na is trapped in a symmetric and hydrophilic environment by two pairs of D/S/T/T motifs (Fig. 2D). For SLC34 protein, corresponding to functional motifs are “161QSSS” and “417QSSS,” which represent the serine-rich stretches of the HPa-HPb linker (4, 31). Structural modeling of NaPi-IIa predicted that the two QSSS motifs are involved in binding Na (Na2 and Na3) and Pi (corresponding to the Na1-Pi-Na2 in TmPiT) (4, 31). Moreover, a conserved aspartic acid is essential for sodium binding, i.e., D327 for Nafore binding in TmPiT and D224 for Na1 binding by NaPi-IIa/IIb. Because of the 2Na:1Pi stoichiometry and electrogenicity for PiT, where only two Na ions are transported, it is similar to the electroneutral NaPi-IIc in that two of three Na ions are released (2, 3). This finding suggests that PiT and NaPi-II share functional features despite of significant sequence dissimilarity.

TmPiT (SLC20) shares some structural and functional features with substrate/ion transporters such as the glutamate/Na transporter Gltph (SLC1) (23), the dicarboxylate/Na transporter VcINDY (SLC13) (24), and the modeled phosphate/Na transporter NaPi-II (SLC34) (31). Although these sodium-driven transporters do not exhibit sequence homology, they contain transporter domains with two inverted repeated helical hairpins (HP1 and HP2) that perform substrate/Na transport, and they have a stable scaffold domain for oligomerization. Substrate/Na transport in these transporters also operates via an elevator-like mechanism (23, 24, 31).

On the basis of the binding and release of substrates/ions and the closing/opening of inner/outer gates, two gate mechanisms were proposed in Gltph (23). Conformational changes in the TmPiT-Na/Pi complex in loops L7, LHP2, and TM8 between subunits A and B indicate that the subunits may have different functional states. Accordingly, these conformational changes may control the inner gate during phosphate and sodium release by assuming closed or open states in subunits A and B, respectively (Fig. 3, D and E, and fig. S3). Residues W139 and W378 are pointed outward and inward, respectively, and may be involved in the opening and closing of the gates. Moreover, subunits A and B reveal different conformations forming an asymmetric dimer, which may imply a cooperative mechanism relationship between the two subunits in TmPiT. However, more experiments and evidence are needed to support this possibility.

On the basis of the sequence repeats in TmPiT, N-PD001131 and C-PD001131 (Fig. 1D); the inverted structure we observed in the TmPiT-Na/Pi complex, TM1-TM2-HP1-TM3 and TM6-TM7-HP2-TM8 (Fig. 1B); and our functional studies of TmPiT mutants revealing that both HP1 and HP2 affect Na/Pi transport; we suggest a two-gated elevator-like mechanism for TmPiT that controls transport at the inner and outer membrane sides, respectively (23). We propose that there might be at least four sequential states, “outward open,” “outward occluded,” “inward occluded,” and “inward open” in the Na/Pi transport cycle (Fig. 4), which occur via a series of conformational changes.

Fig. 4 A working model for Na/Pi transport in TmPiT.

The proposed elevator-like mechanism includes four sequential states: outward open, outward occluded, inward occluded, and inward open. The TmPiT-Na/Pi complex (this study) exists as the inward occluded state. The structure shows the transport domains of the two inverted repeats (N-PD001131, magenta; C-PD001131, blue), the scaffold domain (gray), and Pi and Na, as well as TM3/8 and the L2/LHP1 and L7/LHP2 loops. “W” represents Trp139 and Trp378 in TM3 and TM8, respectively.

The TmPiT-Na/Pi complex structure is in the occluded state; however, both subunits of the TmPiT dimer reveal different conformations at their inner gates (Fig. 3, D and E). For subunit A, Pi-2Na is completely captured by Na1-Pi-Na2 binding, with Na1 being located near the intracellular membrane and the HP1 tip reaching the inner membrane edge, whereas the HP2 tip is distant from the outer membrane (Fig. 2C). Therefore, we suggest that subunit A exists in the inward occluded state, i.e., the elevator-down position. However, near the inner gate of subunit B, important conformational changes occur, during which TM8 bends and forms a large space that may mimic an open-gate conformation to allow Na/Pi release, the L7 becomes exposed to the solvent, and LHP2 becomes disordered, perhaps reflecting “unlocking” of Na1 (Fig. 3E). Therefore, we speculate that subunit B might adopt the inward open state; when the Na/Pi transport gate opens, Pi and two Na (Nafore and Na1) might release into intracellular, and then Na2 may occupy the Na1 position inside the membrane (Fig. 4).

Since there is a mechanistically symmetrical relationship between the HP1 and HP2 repeats (23), we speculated that in the outward conformation, HP2 may undergo a conformational change similar to that of HP1 in the inward state of TmPiT (Fig. 2C and fig. S4) (32). The HP2 tip may move up and reach the outer membrane edge (Fig. 2C and fig. S4) to form the outward state. In the outward open state, the outer gate will be open to interact with Pi-2Na, and Nafore might be prebound with TmPiT. In the outward occluded state, Na2 of Na1-Pi-Na2 is located near the outer membrane, i.e., the elevator-up position (Fig. 4). However, this outward confirmation we speculated needs additional experiments to be proven.

Certain structural characteristics of TmPiT reflected those reported in disease-associated variants of hPiT. Several variants of hPiT2 are associated with neuropsychiatric disorders and primary familial brain calcification (table S4) (10, 11, 19, 20, 26). To understand how these variants might affect hPiT function, we generated topology and structural models of the hPiT2 membrane domain (Fig. 5A and fig. S5A) using Swiss-Model (33) and based on our TmPiT-Na/Pi complex structure and sequence homology (39% identity and 62% similarity in the membrane domain). The intracellular soluble domain (S domain) of hPiT2 was not modeled because of its low sequence homology and limited secondary structure prediction. The S domain may have additional functions independent of phosphate transport, perhaps acting as an external sensor of extracellular phosphate and/or mediating phosphate signaling in the cell (34).

Fig. 5 Topology and homology modeling of hPiT2 and disease-related variants.

(A) The topology model of hPiT2 was created on the basis of TmPiT. hPiT2 variants linked to brain calcification disease are colored and grouped into six categories: Pi-binding (orange), Na-binding (green), N-PD001131 (pink), C-PD001131 (blue), dimer (yellow), and S domain (brown). (B) The variants were mapped onto the modeled hPiT2 structure and are shown as spheres, colored according to the system in (A). Residue numbers for human/Tm are also labeled (see details in table S4). The Pi- and Na-binding sites are indicated with dashed outlines.

From our modeled structure of hPiT2 (Fig. 5B), we observed that the proposed Pi-2Na-binding site spans the highly conserved residues for Pi binding, D28, D506, S113/G114/T115, and S593/T594/T595, except for G114 (fig. S5A), as well as those for Na binding (N27/D28/I112/T595/K561 and N505/D506/T115/V592/Wat, except for I112 and V592) (fig. S1). We then mapped disease-related hPiT mutations onto our modeled hPiT structure (Fig. 5B and table S4) and found that most reported clinical mutations lie in the transport domains (N-PD001131 or C-PD001131), with a few occupying the dimerization or S domains.

In our mutational studies of TmPiT related to hPiT2 clinical variants (Table 1), we observed that the mutants of the Pi-2Na–binding residues D22A and D258A maintained Pi-binding affinities, but both showed Pi uptake defects. Furthermore, the D22/258A double mutant failed to bind Pi. Mutation of the residue responsible for Nafore binding, D327Q, resulted in an inability to bind Pi. Moreover, the W378A and W139A mutants failed to bind Pi (Fig. 3F). Previously, mutational analysis of hPiT2 residue H502 and sequence mapping predicted it to be a critical residue for the transport function (19, 35). In our structure, the residue in TmPiT corresponding to H502 is H254, which forms hydrogen bonds with the Pi-2Na–binding residues D22 and D258 (fig. S5B). Those interactions are likely abolished in the H502Q mutant of hPiT2, potentially indirectly affecting Pi binding to prevent Pi uptake (fig. S5C). These findings highlight how our TmPiT structural data might inform of the molecular mechanisms underlying human diseases associated with mutations of hPiT.

Table 1 Structural and functional comparisons of the TmPiT and hPiT mutants.

*Values present percentage of wild type. †Pi uptake by hPiT mutants is abolished. ‡These variants are found in human patients with primary familial brain calcification (PFBC) disease.

View this table:

The TmPiT-Na/Pi complex we present here represents a complete structural study of a sodium-dependent PiT. TmPiT contains a phosphate and three sodium ions tightly bound by TM1/6 and HP1/HP2. We report structural differences between subunits A and B near the inner gate, specifically in loops L7/LHP2 and TM8. We propose that TmPiT uses an elevator-like mechanism to transport phosphate and sodium. Our high-resolution three-dimensional structure of TmPiT may help establish therapeutic targets for diseases associated with PiT dysfunction.


Cloning, expression, and microsome preparation

Locus Tm0261 of T. maritima strain MSB8 encodes a PiT (TmPiT). TmPiT was amplified from the genomic DNA of T. maritima MSB8 strain by polymerase chain reaction (PCR) using Pfu DNA polymerase (MdBio Inc.). The PCR primers were designed so that the 5′ primer contained a restriction site for Hind III, and the 3′ primer contained a restriction site for Bam HI. The amplified DNA was digested with Hind III/Bam HI and then ligated into the expression vector pYES2. A His-tag was fused into the C terminus of the protein. TmPiT mutations (D22A, W139A, D258A, D327Q, W378A, and D22/258A) were generated by site-directed mutagenesis using the QuikChange system, and mutants were verified by DNA sequencing.

The S. cerevisiae BJ2168 strain was transformed with the TmPiT gene-carrying plasmids and used as host cells for overexpression. Colonies were inoculated into 4 ml of culture medium with 2% (w/v) glucose as a carbon source and incubated at 30°C and 140 rpm for 1 day. Then, 4 ml of yeast culture was transferred into 50 ml of culture medium and incubated for 1 day at 30°C. After 1-day incubation, 50 ml of yeast culture was transferred into 400 ml of culture medium and incubated for a further 2 days at 30°C. Cells were collected and transferred to 1 liter of culture medium with 2% (w/v) galactose for overexpression. The yeast cells were harvested after induction for 3 days, resuspended in buffer [100 mM tris-HCl (pH 9.4) and 10 mM β-mercaptoethanol (β-ME)], and incubated at 37°C for 10 min before being centrifuged at 4000g for 10 min.

The pellets were resuspended and treated with lyticase at 30°C for 2.5 hours [in 100 mM tris-MES (pH 7.6) with 1% (w/v) yeast extract, 2% (w/v) peptone, 1% (w/v) glucose, 0.7 M sorbitol, and 5 mM β-ME] and then collected by centrifugation. Next, the yeast cells were resuspended in homogenization buffer [10% (w/v) glycerol, 5 mM EGTA-tris (pH 7.6), 30 mM tris base, 50 mM tris-ascorbate, 1.5% (w/v) polyvinyl pyrrolidone 40,000, 1 mM phenylmethylsulfonyl fluoride (PMSF), and pepstatin A (1 μg/ml)]. The microsomes were prepared as described previously (36). The yeast suspensions were homogenized by sonication (Qsonica Sonicator Q700) and then pelleted by centrifugation. The supernatant was removed and ultracentrifuged at 80,000g for 50 min. The supernatant was decanted, and the pellet was homogenized in buffer containing 50 mM MES (pH 6.5), 20% (w/v) glycerol, and 60 mM NaCl before undergoing a second ultracentrifugation for 35 min. The membranes were resuspended in a solution containing 50 mM MES-NaOH (pH 6.5), 20% (w/v) glycerol, 60 mM NaCl, and 1 mM PMSF.


Purification of wild-type and mutant TmPiT was performed using a modified “hot-solve” method (37). The microsomes were heated at 65°C for 20 min and then solubilized with n-dodecyl-β-d-maltopyranoside (DDM) for an additional 2 hours before being immediately centrifuged at 4000g for 5 min at 4°C. The TmPiT supernatants were collected and cooled on ice for 10 min and then centrifuged for 20 min at 4°C. The protein sample was purified in a Ni–nitrilotriacetic acid (NTA) column. TmPiT was eluted with buffer [25 mM MES (pH 6.5), 3% glycerol, 120 mM NaCl, 500 mM imidazole, and 0.03% DDM]. The Ni-NTA–purified TmPiT was dialyzed overnight against buffer without imidazole [25 mM MES (pH 6.5), 3% glycerol, 120 mM NaCl, and 0.03% DDM] and then subjected to size exclusion chromatography using a Superdex 200 Increase 10/300 GL column (GE Healthcare) (fig. S6, A and B). TmPiT protein without phosphate was collected for Pi-binding measurements and proteoliposome reconstitution.

Microscale thermophoresis analysis

The binding affinity of TmPiT for different ligands—phosphate, arsenate, sulfate, and PFA—was measured by microscale thermophoresis using a Monolith NT.115pico instrument (NanoTemper Technologies, Germany). The binding affinity of TmPiT mutants for phosphate was assessed using a Monolith NT.115 instrument (NanoTemper Technologies, Germany). TmPiT (10 nM) was labeled with RED-tris-NTA dye (NanoTemper Technologies) in 25 mM MES (pH 6.5), 3% glycerol, 120 mM NaCl, and 0.03% DDM. A volume of 10 μl of 10 nM labeled TmPiT was mixed with 10 μl of ligand (concentrations ranging from 3.1 μM to 50 mM) in 25 mM MES (pH 6.5), 120 mM NaCl, 3% glycerol, and 0.03% DDM. The TmPiT-ligand mixture (4 μl) was loaded into capillaries (NanoTemper Technologies), and thermophoresis was measured at 25°C for 20 s with 15 to 35% LED power and 60% microscale thermophoresis power. Data from three independent measurements were combined and analyzed using MO Affinity Analysis software version 2.3 (NanoTemper Technologies) and fitted to a single binding site model.

Proteoliposome reconstitution

l-α-Phosphatidylcholine from soybean (type II-S, Sigma-Aldrich) was mixed with buffer [10 mM Mops (pH 7.3)] to a final concentration of 20 mg/ml and vortexed until homogenized. The lipids were flash-frozen in liquid nitrogen and then thawed for a total of eight freeze-thaw cycles. The liposomes were extruded using polycarbonate filters with a pore size of 400 nm (Whatman). For the reconstitution experiments, liposomes (1 ml) were destabilized with 40 μl of sodium cholate [20% (w/v) stock solution], mixed with 200 μg of purified TmPiT and then incubated for 30 min at room temperature. The cholate was removed via a PD-10 gel filtration column (GE Healthcare) equilibrated with the same buffer, and the reconstituted proteoliposomes were collected in 2.3 ml of the same buffer. The liposome sample was dialyzed against 10 mM Mops (pH 7.3) at 4°C for 16 hours. The resulting dialysate was diluted to 20 ml by adding 10 mM Mops (pH 7.3), and the proteoliposomes were collected by ultracentrifugation (150,000g and 4°C for 1 hour) and resuspended in 500 μl of 10 mM Mops (pH 7.3) (fig. S6C).

Proteoliposomal Pi transport assay

Pi uptake activity was measured with reconstituted proteoliposomes containing TmPiT. The control was liposomes alone (i.e., without TmPiT). A total of 2 μl of TmPiT proteoliposomes in a 50-μl reaction solution [10 mM Mops-KOH (pH 6.5) and 120 mM NaCl] and 4 μl of [32P]orthophosphate [25 mCi (925 megabecquerel)/mmol; carrier-free, PerkinElmer] were diluted to a final concentration of 100 μM and incubated for 10 min at 25°C. The suspension was rapidly filtered with a G-25 spin column (GE Healthcare) to remove unincorporated Pi. Radioactivity was determined by liquid scintillation spectrometry. We conducted the uptake measurement experiment three times. Data are presented as means ± SD.

Crystallization and structural determination and refinement

To produce the TmPiT-Na/Pi complex, we subjected 10 mM KH2PO4 to size exclusion chromatography. Crystallization was performed using the hanging drop vapor diffusion method at 20°C over a reservoir of solution containing 50 mM sodium citrate (pH 5.5), 100 mM NaCl, 18 to 20% (w/v) pentaerythritol propoxylate (5/4 PO/OH), and 10 to 12% polyethylene glycol 300 (fig. S6D). TmPiT-Na/Pi complex crystals were obtained within 1 week. The TmPiT heavy-atom derivative was cocrystallized with 1 mM Hg(CH3COO)2 under identical conditions.

Both the native and anomalous dispersion datasets were collected from the Taiwan Photon Source (TPS) 05A beamline at the National Synchrotron Radiation Research Center in Hsinchu, Taiwan. X-ray diffraction data were processed with the HKL2000 program (38). Native TmPiT crystals belonged to the monoclinic space group C2 with the unit cell parameters a = 121.2 Å, b = 112.6 Å, c = 110.7 Å, and β = 119.3° (fig. S6E and table S1). The Matthew’s coefficient was calculated to be 3.643 Å3/Da, with a solvent content of 56.4% and two subunits per asymmetric unit.

Structural phasing was determined from the Hg(CH3COO)2 heavy-atom derivative. The final phase of the resolution extension was calculated by AutoSol using PHENIX (39), resulting in distinguishable protein and solvent regions (fig. S6F). The initial structural model was built using AutoBuild (39), and the programs Coot (40) and PHENIX (39) were used for model building and refinement. Ultimately, residues 1 to 400 of TmPiT were built. X-ray diffraction data collection and structural refinement statistics are shown in table S1. The final structural model had a crystallographic R-factor of 18.8% and a free R-factor of 22.4%. The solvent-accessible surface area was calculated with the program CASTp (27) using a probe radius of 1.4 Å. All structural figures shown in this report were generated using PyMOL (


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank C.-I. Chang for the genomic DNA of T. maritima MSB8. Portions of this research were carried out at the National Synchrotron Radiation Research Center, a national user facility supported by the Ministry of Science and Technology of Taiwan. We thank the National Synchrotron Radiation Research Center (NSRRC) for access to beamline TPS 05A and the in-house x-ray facilities in the Macromolecular X-ray Crystallographic Center of National Tsing Hua University and in the Institute of Molecular Biology, Academia Sinica. Funding: This work was supported by the Ministry of Science and Technology, Taiwan: MOST 106-2311-B-007-007-MY3, 107-2321-B-007-004, and 108-2321-B-007-002 to Y.-J.S.; MOST 107-2311-B-001-032-MY3 to C.-D.H.; and MOST 107-2811-B-007-002 and 108-2811-B-007-504 to J.-Y.T. Author contributions: J.-Y.T. performed the TmPiT purification, crystallizations, data collection, and structure determination and refinement. C.-H.C. carried out cloning of TmPiT mutants and microscale thermophoresis analysis. M.-G.L. performed the Pi uptake experiment. Y.-H.C. assisted in protein purification and proteoliposome reconstitution. R.-Y.H. assisted in protein purification and microscale thermophoresis analysis. C.-Y.Y. assisted with data collection. C.-D.H. and Y.-J.S. supervised the structural and functional studies. All authors participated in discussions on results and in preparing the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The structural coordinate have been deposited in the PDB Japan under the accession number 6L85. All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article