The solute carrier SPNS2 recruits PI(4,5)P2 to synergistically regulate transport of sphingosine-1-phosphate

Abstract

Solute carrier spinster homolog 2 (SPNS2), one of only four known major facilitator superfamily (MFS) lysolipid transporters in humans, exports sphingosine-1-phosphate (S1P) across cell membranes. Here, we explore the synergistic effects of lipid binding and conformational dynamics on SPNS2's transport mechanism. Using mass spectrometry, we discovered that SPNS2 interacts preferentially with PI(4,5)P₂. Together with functional studies and molecular dynamics (MD) simulations, we identified potential PI(4,5)P₂ binding sites. Mutagenesis of proposed lipid binding sites and inhibition of PI(4,5)P₂ synthesis reduce S1P transport, whereas the absence of the N terminus renders the transporter essentially inactive. Probing the conformational dynamics of SPNS2, we show how synergistic binding of PI(4,5)P₂ and S1P facilitates transport, increases dynamics of the extracellular gate, and stabilizes the intracellular gate. Given that SPNS2 transports a key signaling lipid, our results have implications for therapeutic targeting and also illustrate a regulatory mechanism for MFS transporters.

Keywords: HDX-MS; S1P; SLCs; SPNS2; molecular dynamics; native MS; protein-lipid interactions; solute carriers; sphingosine-1-phosphate.

Graphical abstract

Figure 1

A predicted model of SPNS2 and the binding preference of PI derivatives to full-length and truncated SPNS2 (A) The AlphaFold2 predicted model of SPNS2 (blue) with the N terminus of the full-length protein (green) shown schematically, embedded within a phospholipid bilayer. PI derivatives: PI, PI(4)P, PI(4,5)P₂, and PIP₃ are shown as purple, green, orange, and gray spheres, respectively, with binding preference to full-length and truncated SPNS2 arranged via proximity to the transporter (according to the K_D values in Table S1). The structure of S1P is also shown. (B) Mass spectra recorded for delipidated full-length SPNS2 (5 μM) with increasing concentrations from 0 to 20 μM of PI (18:1/18:1). (C) Plot of the mole fraction of full-length SPNS2 and PI binding calculated from titration of PI (18:1/18:1) with a resulting fit (R2 = 0.97) from a sequential ligand-binding model (solid lines). Data are plotted as mean ± standard deviation (SD) (n = 3). (D) Plot of mole fraction of truncated SPNS2 with PI-bound states calculated from titration of PI(18:1/18:1) with a resulting fit (R2 = 0.99) from a sequential ligand-binding model (solid lines). Data are plotted as mean ± SD (n = 3). (E) A single charge state (17+) of full-length SPNS2 following incubation with an equimolar solution containing PI, PI(4)P, PI(4,5)P₂, and PIP₃ confirms preferred binding of PI(4,5)P₂. Data are plotted as mean ± SD (n = 3). ^∗∗p < 0.01. ^∗∗∗p < 0.001. ^∗∗∗∗p < 0.0001. (F) A single charge state (14+) of truncated SPNS2 following incubation with an equimolar solution containing PI, PI(4)P, PI(4,5)P₂, and PIP₃. Data are plotted as mean ± SD (n = 3). ^∗p < 0.05. ^∗∗p < 0.01. ^∗∗∗p < 0.001.

Figure 2

Determination of binding affinities for PI derivatives to full-length and truncated SPNS2 and potential PI(4,5)P₂ binding sites on full-length SPNS2 (A–F) Binding curves for PI(4)P, PI(4,5)P₂, and PIP₃ binding to full-length SPNS2 (A, B, and C, respectively) and to truncated SPNS2 (D, E, and F, respectively). Data are plotted as mean ± SD (n = 3). (G) K_D1 values for PI, PI(4)P, PI(4,5)P₂, and PIP₃ binding to full-length and truncated SPNS2. Data are plotted as mean ± SD (n = 3). (H) MD simulation reveals PI(4,5)P₂ molecule (orange sticks) bound to the N terminus (binding site 1), and further three PI(4,5)P₂ binding sites are shown (sites 2–4). SPNS2 is shown in blue (with the N terminus colored in yellow), the potential PI(4,5)P₂ interacting residues are shown as blue sticks, and phosphorus atoms in the headgroups of membrane lipids are shown as gray spheres.

Figure 3

The effect of PI(4,5)P₂ on S1P transport (A) Mass spectrum of equimolar ratios of full-length and truncated SPNS2 following incubation with substrate, S1P. (B) Relative ratios of S1P-bound peaks to the respective apo peaks confirm that full-length SPNS2 has a higher affinity for S1P. Data are plotted as mean ± SD (n = 3). ^∗∗∗∗p < 0.0001. (C) Structures of the two PIP5K inhibitors ISA-2011B and UNC3230 used in the cell-based transport assay to inhibit synthesis of PI(4,5)P₂. (D) Schematic illustration of the cell-based functional assay. S1P is synthesized in cells and exported across cell membranes by SPNS2. Extracellular S1P is detected by an enzyme-linked immunosorbent assay (ELISA). (E) Potential PI(4,5)P₂ interacting residues corresponding to binding sites 1–4 are highlighted in the full-length SPNS2 model (orange). (F) PI(4,5)P₂ levels of full-length SPNS2 overexpressing HEK293T cells treated with and without PIP5K inhibitors ISA-2011B and UNC3230. Data are plotted as mean ± SD (n = 3). ^∗∗p < 0.01. ^∗∗∗p < 0.001. (G) Extracellular S1P levels of full-length SPNS2 overexpressing HEK293T cells treated with and without PIP5K inhibitors ISA-2011B and UNC3230. Data are plotted as mean ± SD (n = 3). ^∗∗p < 0.01. (H) Transport activity of full-length SPNS2 variants with point mutations at PI(4,5)P₂ binding sites. WT, wild-type SPNS2. Truncated, N-terminal truncated SPNS2. The residues R23-R28 are mutated into alanines (R23-28A). Transport activities are normalized to FLAG expression (Figure S5B) and represented as a ratio to the wild type. Data are plotted as mean ± SD (n = 3).

Figure 4

Conformational dynamics of S1P-bound SPNS2 (A) Heatmap comparing S1P-bound SPNS2 with apo SPNS2 via ΔRFU as a function of time. (B) Deuterium uptake plot for peptide 127–142. Data are plotted as mean ± SD (n = 3). (C) ΔRFU of S1P-bound SPNS2 compared with apo SPNS2 mapped onto the predicted model of SPNS2. Ribbon representation is viewed from membrane. (D) Deuterium uptake plot for peptide 143–147. Data are plotted as mean ± SD (n = 3). (E) Deuterium uptake plot for peptide 117–126. Data are plotted as mean ± SD (n = 3). (F) ΔRFU of S1P-bound SPNS2 compared with apo SPNS2 mapped onto the predicted model of SPNS2. Ribbon representation is viewed from extracellular side. (G) Deuterium uptake plot for peptide 335–340. Data are plotted as mean ± SD (n = 3).

Figure 5

Conformational dynamics of PI(4,5)P₂-bound SPNS2 and coarse-grained MD simulations of PI(4,5)P₂ recruitment to SPNS2 (A) Heatmap representing ΔRFU, comparing PI(4,5)P₂-bound SPNS2 with apo as a function of time. (B) Deuterium uptake plot for peptide 158–164. Data are plotted as mean ± SD (n = 3). (C) ΔRFU of PI(4,5)P₂-bound SPNS2 compared with apo SPNS2 mapped onto the predicted model of SPNS2. Ribbon representation is viewed from membrane. (D) Deuterium uptake plot for peptide 143–147. Data are plotted as mean ± SD (n = 3). (E) Deuterium uptake plot for peptide 115–120. Data are plotted as mean ± SD (n = 3). (F) ΔRFU of PI(4,5)P₂-bound SPNS2 compared with apo SPNS2 mapped onto the predicted model of SPNS2. Ribbon representation is viewed from extracellular side. (G) Deuterium uptake plot for peptide 335–340. Data are plotted as mean ± SD (n = 3). (H) Snapshots from CG simulations of SPNS2 in a PI(4,5)P₂ containing phospholipid bilayer (back-mapped to atomistic resolution). Starting from the initial state, Arg43 or the arginine patch (Arg 23–28) interacts with PI(4,5)P₂ molecule, and the N terminus guides the PI(4,5)P₂ molecule closer to SPNS2. SPNS2 is shown in blue (with the N terminus colored in yellow), phospholipids are shown on a white surface (with the phosphates shown as gray spheres), and PI(4,5)P₂ is shown as orange sticks.

Figure 6

Conformational dynamics of SPNS2 in the ternary complex with both S1P and PI(4,5)P₂. (A) Heatmap representing ΔRFU as a function of time, comparing PI(4,5)P₂: S1P: SPNS2 with apo SPNS2. (B) Deuterium uptake plot for peptide 259–263. Data are plotted as mean ± SD (n = 3). (C) ΔRFU of PI(4,5)P₂: S1P: SPNS2 compared with apo SPNS2 mapped onto the predicted model of SPNS2. Ribbon representation is viewed from membrane. (D) Deuterium uptake plot for peptide 470–478. Data are plotted as mean ± SD (n = 3). (E) Deuterium uptake plot for peptide 117–126. Data are plotted as mean ± SD (n = 3). (F) ΔRFU of PI(4,5)P₂: S1P: SPNS2 compared with apo SPNS2 mapped onto the predicted model of SPNS2. Ribbon representation is viewed from extracellular side. (G) Deuterium uptake plot for peptide 336–340. Data are plotted as mean ± SD (n = 3). (H–K) Changes in fractional deuterium uptake of peptides 117–126 (H), 259–263 (I), 336–340 (J), and 470–478 (K) for full-length and truncated SPNS2 at 60 min labeling. All data are plotted as mean ± SD (n = 3). ^∗p < 0.05. ^∗∗p < 0.01. ^∗∗∗p < 0.001. ^∗∗∗∗p < 0.0001.

Figure 7

Proposed transport model based on time-resolved conformational changes of SPNS2 upon binding of PI(4,5)P₂ and S1P All conformational changes are mapped onto the predicted model of SPNS2 and represent. (A) apo SPNS2 at 0 min. (B) ΔRFU of PI(4,5)P₂-bound SPNS2 compared with apo SPNS2 at 60 min labeling. (C) ΔRFU of S1P-bound SPNS2 compared with apo SPNS2 at 60 min labeling. (D) ΔRFU of PI(4,5)P₂: S1P: SPNS2 compared with apo SPNS2 at 60 min labeling. The helices are colored according to the corresponding HDX results above. N and C domains are shown in green and yellow, respectively, with S1P and PI(4,5)P₂, green and orange spheres. A greater deuteration of the helices on the extracellular side indicates PI(4,5)P₂ synergistically enhances S1P-induced opening of the extracellular gate while stabilizing the intracellular gate.