The overall Cryo-EM structure
XPR1 contains a transmembrane domain and a cytoplasmic N-terminal SPX domain. The transmembrane domain is responsible for facilitating Pi efflux, while the SPX domain enhances this function by binding inositol pyrophosphate (PP-InsP) stimuli1,26-29. By preparing XPR1 in the presence of PP-InsP (Supplementary Fig. 1, and Methods), we could collect high-quality cryo-EM images for single particle analysis, which allowed us to reconstruct a density map with an overall resolution of 3.3 Å (Supplementary Fig. 2a). The EM density of SPX domain is poor (Supplementary Fig. 3a), probably owing to its mobility. In contrast, the density map of transmembrane domain is clear (Supplementary Fig. 3a). Although the transmembrane domain is small, with a molecule weight of only approximately 43 kDa, the well resolved EM density enables us to build the atomic model (Supplementary Fig. 3b, c, and Supplementary Table 1).
The transmembrane domain of XPR1 contains 10 transmembrane α-helices (TM1-10) and folds into two structurally distinct sub-domains, with N and C termini on the intracellular side (Fig. 1). We refer to the N-terminal portion as N domain, that is formed by TM1-TM5 and a short amphipathic helix (AH) lying parallel to the membrane. The C-terminal portion harbors the conserved EXS (named for homologous regions found in yeast ERD1 and SYG1 and human XPR1) domain30,31 (Supplementary Fig. 4), that is made up of TM6-TM10. The EXS domain associates with TM5, creating a pore that spans across the membrane (Fig. 1a). Lined the membrane-spanning pore, two phosphate ions were identified (Fig. 1a, Supplementary Fig. 3d), implying that the pore serves to transport Pi. Notably, TM9b positions close to the central pore axis on the extracellular side (Fig. 1a), suggesting its potential role in Pi exporting. A 3-dimensional structural homology search with the program DALI32 found no known structures similar to the XPR1 transmembrane domain, indicating a specific mechanism for phosphate recognition and transport in XPR1.
The structure basis for phosphate recognition and transport
HOLE33 analysis for the solvent-accessible pathway of XPR1 transmembrane domain reveals that it exhibits an ion channel-like architecture for Pi transit (Fig. 2a, b). The intracellular entrance of the translocation channel is formed by TM5a, TM6-TM8 and TM10 (Fig. 2a). Owing to a kink in TM9, the channel is bent and created by TM9b, TM10, TM6, TM5b and the tip of TM2 on the extracellular side. The narrowest point of the channel has a radius of approximately 1.2 Å, which exceeds the water access limit (1 Å)34, indicating that the channel is solvent-permeable. The entrance and the interior wall of the first half of the channel are positively charged, while the second half and the exit are negatively charged (Fig. 2b). Given the anionic property of Pi ions, the presence of polarized electrostatic potential across the translocation channel may serve to move Pi ions in and out. The solvent-accessibility of the translocation channel (Fig. 2a), together with aligned Pi ions along the permeation pathway (Fig. 2b), suggests that XPR1 may export Pi ions through this channel without requiring large conformational changes as observed in Pi importers. Pi importers exhibit an outward-open conformation for Pi uptake from extracellular side and undergo large conformational changes to adopt an inward-open conformation, allowing for Pi release into cytosol35-37. This conformational transition is driven by the energy derived from movements of other co-translocated ions (e.g., Na+ or protons) down their concentration gradients35-37. In contrast, the Pi export activity of XPR1 is independent of pH gradient across the cell membrane3,27, and no co-translocated ions have been found23. These previously observed phenomena are now supported and aligned with by the channel-like architecture of XPR1.
Our structure provides a framework for understanding the mechanism by which XPR1 recognizes phosphate. Two phosphate ions are recognized by a series of polar residues in the translocation channel (Fig. 2c). One of them (Pi1) interacts with the side chains of N401, R570, Q576 and E600, and is coordinated downstream (intracellular-to-extracellular direction) of the other Pi ion (Pi2). The Pi2 is recognized by interacting with D398, K482, Y483, D533, R570, R603, R604 and W607. Additionally, the side chain of R459 positions the orientation of D398 and Y483, facilitating them to coordinate Pi2 in the binding site. Moreover, the side chain of R570 points towards the midpoint between the two Pi ions, bridging them, which implies its potential role in Pi delivery during translocation. There is a discontinuity in TM9 that is kinked at W573. The TM9b is stabilized by an intrahelical hydrogen bond between W573 and Q576, as well as the interaction between Q576 and Pi1. These Pi-recognizing residues are conserved in different species (Supplementary Fig. 4).
To assess the roles of the above-mentioned key residues in Pi transport, we reconstituted XPR1 into liposomes and performed Pi transport assays (Methods). By substituted the Pi2-recognizing residues D398, K482, D533, R603 and R604 with alanine, and Y483 and W607 with phenylalanine, and could purify the R603A or the Y483F in sufficient quantities to perform the proteoliposome-based transport assays. We find that XPR1 carrying either R603A or Y483F substitution have impaired Pi transport activities (Fig. 2d), suggesting crucial roles of Y483 and R603 in transporting Pi. Moreover, consistent with the structure configuration of the Pi1binding site (Fig. 2c), replacing side chain of N401, Q576 or E600 with alanine resulted in a reduced Pi transport activity of XPR1 (Fig. 2d).
Clinically linked residues in XPR1 function in phosphate recognition and transport
XPR1 is a causative gene identified in primary familial brain calcification (PFBC) families5,16,17, and the R459C and R570C missense variants have been found to be pathogenic15,20,38,39. Our structure reveals that R459 is involved in maintaining the binding network of Pi2, and R570 forms salt bridges with Pi2 and Pi1 (Fig. 2c). Substituting the side chains of R459 and R570 might perturb Pi binding and potentially impair XPR1 activity. Indeed, we find that the PFBC families variants, R570C and R459C, exhibit an significant reduction in Pi transport activity (Fig. 2d). Therefore, our structure and function analysis provide mechanistic insights into the correlation between patient mutations and XPR1 function.
The mechanism for linking phosphate recognition and transport
How do the phosphate ions pass through the XPR1 transmembrane channel? A structure comparison between channel alone and bound to phosphate ions would provide mechanistic insights. With the resolved experimental structure of the phosphate-bound channel at hand, we compared it with the unbound state model predicted by AlphaFold40. While the AlphaFold model and our experimental structure share the overall fold, a notable disparity is observed in the TM9 segment (Supplementary Fig. 5). The experimental structure has a kink in this segment, causing a shift of TM9b towards the central membrane-spanning pore axis, while the AlphaFold model predicts an unbent TM9. We reasoned that the TM9 kink in the experimental structure is stabilized by the intrahelical hydrogen bond formed between the side chains of Q576 and W573, and the interaction between Q576 and the bound phosphate (Pi1) (Fig. 2c). We therefore speculated that substituting Q576 side chain with alanine would disrupt the intrahelical hydrogen bond and lead to the Pi1 dissociation. To clarify these envisioned conformational changes, we sought to determine the Cryo-EM structure of the Q576A mutant XPR1.
The transmembrane domain structure of the XPR1 Q576A mutant was resolved at a resolution of 3.2 Å (Supplementary Fig. 2b, and Methods). This mutant presents a similar overall structure to that of wild-type, and also two phosphate ions are observed in the translocation channel (Supplementary Fig. 6). Notably, local conformational changes in the Pi1 binding site are revealed (Fig. 3a). In the mutant structure, the Pi1 ion undergoes a forward movement (intracellular-to-extracellular direction) within the translocation channel. Consequently, the hydrogen bonds between Pi1 and the side chains of N401 and E600 are broken. Moreover, the indole ring of W573 flips to form an intrahelical hydrogen bond with R570. This rearranged intrahelical hydrogen bond disrupts the salt bridge between R570 and Pi1, and thus the Pi1 dissociation and forward movement. Additionally, given the high density of π-electrons in the W573 indole ring, the ring-flipping and closing to the Pi1 ion would push the anion forward. The ring-flipping therefore links Pi recognition to transport. Supporting this, when the indole ring of W573 is substituted with polar or hydrophobic group like W573N, W573Y, W573A and W573L, the Pi transport activity of XPR1 is reduced (Fig. 3b). Together, the translocation channel likely employ a "ring flipping-push-movement" mechanism to propel Pi ions through, facilitated by the rearrangement of two intrahelical hydrogen bonds between W573-Q576 and W573-R570 (Fig. 3c). As Q576 is replaced by alanine in the mutant structure, a possibility cannot be ruled out that the side chain of Q576 also plays role in the Pi forward movement in native protein.