Introduction

Prostaglandins (PGs, mainly PGE2, PGF, and PGD2) are arachidonic acid-derived bioactive lipid mediators that modulate a broad range of physiological and pathological processes, including pain, inflammation, and fever1,2. After activating prostanoid receptors, PGs can be transported into cells by the prostaglandin transporter SLCO2A1 (also known as OATP2A1, SLC21A2, and PGT) and are rapidly degraded by 15-hydroxyprostaglandin dehydrogenase (15-PGDH)3,4,5,6. Consequently, SLCO2A1, mediates the rate-limiting step of PGs (especially PGE2) inactivation and plays a vital role in regulating PGs distribution7,8. To date, more than fifty SLCO2A1 mutations have been reported8. Loss-of-function mutations in SLCO2A1 lead to impaired PGs degradation, causing a subtype of primary hypertrophic osteoarthropathy (PHO), which is an autosomal recessive inherited disease affecting both the skin and bones. PHO disease is characterized by nail clubbing, skin manifestations, periostosis, acroosteolysis, and painful joint enlargement9,10. Deficiency of SLCO2A1 also causes chronic enteropathy associated with the SLCO2A1 gene (CEAS), which is characterized by multiple small intestinal ulcers of nonspecific histology9,11. In addition, the inhibition of 15-PGDH results in increased tissue PGE2 levels, which has been shown to be effective in accelerating tissue regeneration and repair12,13,14,15, suggesting that the inhibition of SLCO2A1 may be an alternative therapeutic strategy for accelerating the recovery of injuries16. These findings highlight the physiological and pharmacological importance of SLCO2A1.

SLCO2A1 was identified and characterized in 199517. SLCO2A1 belongs to the organic anion transporting polypeptide (OATP) family (also known as the solute carrier 21 (SLC21) family), a subfamily of the major facilitator superfamily (MFS). OATPs mediate the uptake and distribution of amphipathic organic compounds, including bile salts, steroid conjugates, thyroid hormones, organic dyes, anionic oligopeptides, numerous drugs, and other xenobiotic substances18,19,20. SLCO2A1 is predicted to share a similar topology with other OATPs, containing 12 transmembrane (TM) helices with a large extracellular loop between TM9 and TM1018. Interestingly, unlike other OATPs, which have a wide range of substrates, SLCO2A1 primarily transports prostaglandins, including PGE2, PGE1, PGF, PGD2, and thromboxane-B2 (TxB2)17,21. PGE2 metabolites such as 15-keto PGE2, which do not have an OH group at C15, have a much lower affinity for SLCO2A122. In addition, substitution of the carboxyl group with an amine significantly reduces the affinity of PGE2 for SLCO2A122. These findings indicate that the hydroxyl group at C15 and the carboxyl group are important for PGE2 binding to SLCO2A122. Like those of other OATPs, the energy coupling mechanism of SLCO2A1 is not fully understood. PGs uptake by SLCO2A1 is not affected by extracellular Na+, Cl, or H+17. Despite extensive functional studies on SLCO2A1 and recent structural progress in SLCO1B1 (also known as OATP1B1) and SLCO1B3 (also known as OATP1B3)23,24, the structural determinants that recognize PGs and distinguish SLCO2A1 from other OATPs remain elusive.

In this study, we report the cryo-electron microscopy (cryo-EM) structures of human wild-type (WT) SLCO2A1 in apo and PGE2-bound forms, both of which are in an outward-open state. Together with mutagenesis functional studies and sequence alignments, we investigated the detailed binding site of PGE2, offering mechanistic insights into PGE2 recognition and the substrate selectivity of SLCO2A1. Structural comparisons of the outward-open SLCO2A1 structure with the inward-open cognate OATP1B1 structure and the AlphaFold2-predicted SLCO2A1 model suggest a hypothetic model for the conformational changes during transport. Furthermore, we mapped known PHO-related and CEAS-related mutations to the structure of SLCO2A1, providing insights into SLCO2A1 pathology. These findings could advance the future development of potential therapeutics targeting SLCO2A1.

Results

Functional characteristics of SLCO2A1

We heterologously expressed human SLCO2A1 in HEK293F cells and assayed its transport activity via a fluorescence-based uptake assay. 6-Carboxyfluorescein (6-CF) is a fluorescent dye (Fig. 1a) that can be effectively taken up by SLCO2A1 and has been successfully used to evaluate the impact of FDA-approved drugs on SLCO2A125. We first measured the concentration-dependent 6-CF uptake, yielding a Michaelis constant (Km) of 84.10 ± 31.83 μM for 6-CF by SLCO2A1 at pH 6.0 (Fig. 1b), which was similar to the value (269.8 ± 40.7 μM) reported in a previous study25. We subsequently assayed the concentration-dependent inhibition of 6-CF uptake by PGE2 in SLCO2A1-expressing cells. The resulting half-maximal inhibitory concentration (IC50) for PGE2 is 4.38 ± 0.97 μM (Fig. 1c), which is ~13-fold greater than the Ki value (331 nM) of PGE2 in competitively inhibiting radioactive PGE226, presumably because of differences in pH conditions and competing ligands. Suramin is an FDA-approved antiviral agent27,28,29 that has been reported to potently inhibit SLCO2A125. Suramin is a large molecule with multiple aromatic rings and six negatively charged sulfonate groups featuring a twofold symmetry (Fig. 1a). We further investigated the concentration-dependent inhibition of 6-CF uptake by suramin, resulting in an IC50 value of 2.21 ± 0.37 μM for suramin (Fig. 1d), which is 13-fold greater than the IC50 value (0.17 μM) for suramin measured at pH 7.4 in competitively inhibiting radioactive PGE2 instead of 6-CF as a substrate25.

Fig. 1: Functional characterization of human SLCO2A1.
figure 1

a The chemical structure of 6-CF, PGE2, and suramin. b Concentration dependence of 6-CF transport by human SLCO2A1. Data were presented as mean values ± SEM; n = 3 independent replicates. c, d Concentration-dependent inhibition of 6-CF transport by PGE2 (c) and suramin (d) in SLCO2A1. Cells expressing mCherry served as a control. Data were normalized to net 6-CF transport activity of SLCO2A1 WT without PGE2 or suramin and were presented as mean values ± SEM; n = 5 independent replicates for log [PGE2 (μM)] = −3, n = 4 independent replicates for log [PGE2 (μM)] = −2, n = 3 independent replicates for all other conditions. e The chemical structure of estrone 3-sulfate (E3S), thyroxine (T4), and taurocholate. f Inhibition of 6-CF uptake by PGE2, E3S, T4, and taurocholate in HEK293F cells expressing SLCO2A1 WT. Cells expressing mCherry served as a control. Data were normalized to 6-CF and were presented as mean values ± SEM; n = 3 independent replicates. One-way ANOVA was used; nsP = 0.0871 for E3S, nsP = 0.1484 for T4, *P = 0.0132 for taurocholate, ****P < 0.0001. Source Data are provided as a Source Data file.

Interestingly, unlike other OATP family members, which have a wide range of substrates, the substrate of SLCO2A1 seems to be more specific. Only prostaglandins and thromboxanes have been reported to be high-affinity endogenous substrates of SLCO2A117,21. To evaluate the substrate specificity of SLCO2A1, we chose three representative endogenous substrates of the OATP family30, the steroid hormone estrone 3-sulfate (E3S)31, the thyroid hormone thyroxine (T4)32, and the bile acid taurocholate33 (Fig. 1e), and investigated their inhibitory effects on 6-CF uptake by SLCO2A1. The competitive inhibition results revealed that E3S, T4, and taurocholate at a concentration of 50 μM caused only 20–40% reduction in 6-CF uptake by SLCO2A1; in contrast, 6-CF uptake was almost completely abolished by 50 μM PGE2 (Fig. 1f). These results indicate that SLCO2A1 can specifically take up PGE2, which is in line with the findings of previous studies17.

Overall architecture of SLCO2A1

To reveal the structural basis for SLCO2A1 transport, we purified homogenous samples of human full-length WT SLCO2A1 in detergents and determined its structure via cryo-EM single-particle analysis (Supplementary Figs. 1a, b and 2). However, during cryo-EM data processing, SLCO2A1 particles exhibit pronounced preferential orientation, which limits high-resolution map reconstruction. A similar problem has also been reported for other OATPs23. To solve this problem, we screened several types of grids and found that particles spreading on nickel–titanium alloy foil (ANTA foil) grids have more even distributions without obvious angular preference (Supplementary Fig. 2a, b).

Although SLCO2A1 is a relatively small membrane protein of ~65 kDa in size, the large and rigid extracellular domain functions as a fiducial marker and is sufficient for accurate particle alignment (Supplementary Fig. 2b). Finally, a 3.2 Å reconstruction map for SLCO2A1-apo was generated, which permits reliable modelling of SLCO2A1 with the assistance of its Alphafold234 model (Fig. 2a, b). The structure of SLCO2A1 is composed of a transmembrane domain (TMD) comprising 12 TM helices and interval loops connecting adjacent TM helices on either side of the membrane (Fig. 2b, c). The TMD of SLCO2A1 adopts the canonical MFS fold, featuring an N-domain (TM1-6) and a C-domain (TM7-12) with twofold pseudosymmetry between them (Fig. 2c). The N-domain and C-domain are connected by an amphipathic cytoplasmic helix (CH1), which is partly embedded in the membrane (Fig. 2a, c). The N- and C-domains engage in extensive interactions on the intracellular side and separate from each other at a distance of 29.8 Å on the extracellular side, indicating that SLCO2A1 is captured in an outward-open conformation (Fig. 2a, b).

Fig. 2: Cryo-EM structure of SLCO2A1.
figure 2

a, b The EM map of SLCO2A1-apo embedded in LMNG micelle (gray) (a) and the cartoon representation of SLCO2A1-apo (b). The N and C domain of SLCO2A1-apo are colored in green and pink, respectively. c Topology diagram of human SLCO2A1 from N-terminal to the C-terminal. Unsolved regions are shown with dashed lines. d Multiple disulfide bonds within EL9–10 are shown as yellow sticks. A fragment of EL3–4 is attached to EL9–10 via a disulfide bond. e 6-CF transport by HEK293F cells expressing SLCO2A1 WT or SLCO2A1 mutants. Cells expressing mCherry served as a control. Data were normalized to SLCO2A1 WT and are presented as mean ± s.e.m.; n = 3 independent replicates for mutants. n = 6 independent replicates for mCherry and SLCO2A1 WT. One-way ANOVA was used; ****P < 0.0001. f A cut-open electrostatic potential surface of SLCO2A1-apo. The electrostatic potential surface was calculated in PyMoL (red to blue, −50 kT/e to +50 kT/e). Red arrow indicates the central pocket of SLCO2A1. Source Data are provided as a Source Data file.

The largest extracellular loop is located between TM9 and TM10 (EL9–10), which is composed of several β sheets and four pairs of disulfide bonds (Fig. 2b), known as the Kazal-like domain35. The Kazal-like domain is conserved in OATP family, but its functional role in OATPs is not fully understood. Although most of the long extracellular loop between TM3 and TM4 (EL3–4) is invisible, we observed that a short region (A140–Q144) lies in a groove of the Kazal-like domain stabilized by a conserved disulfide bond of C143–C448 (Fig. 2d). The C143A and C448A mutations, which break this disulfide bond, resulted in reduction of the 6-CF uptake activity by approximately 80% and 55%, respectively (Fig. 2e). Similar interactions between EL3–4 and EL9–10 were also found in OATP1B124. A previous study showed that EL3–4 potentially regulates the maximum extent of opening24. Consistent with this, EL3–4 may limit excessive opening of OATPs on the extracellular side by linking to the Kazal-like domain (Fig. 2b).

In this outward-facing SLCO2A1, a large central pocket for recognizing and translocating substrates is formed by the interface between the N-domain and C-domain (Fig. 2f). The interior of the central pocket is relatively hydrophobic with charged patches at the bottom, which are sculpted by K53 and R561 along with hydrophobic residues, including L46, V216, L373, I342, and W565.

PGE2 recognition in SLCO2A1

To understand how SLCO2A1 specifically recognizes prostaglandins, we purified SLCO2A1 in the presence of PGE2, which is the most extensively studied substrate of SLCO2A1, with high affinity ranging from 94 to 331 nM21,26,36. We subsequently solved the cryo-EM structure of SLCO2A1 in complex with PGE2 at 3.07 Å resolution (Supplementary Figs. 1c, d and 3). The SLCO2A1-PGE2 structure opens its central cavity to the extracellular space in an outward-open conformation (Fig. 3a–c). The superposition of SLCO2A1-PGE2 with SLCO2A1-apo demonstrated that the overall structures of SLCO2A1 are essentially identical, with a root mean square deviation (r.m.s.d.) value of 0.68 Å over 482 aligned Cα atoms (Supplementary Fig. 4a, b). In contrast to the empty cavity of SLCO2A1-apo, strong electron microscopy density was observed inside SLCO2A1-PGE2, which fits a PGE2 molecule unambiguously (Fig. 3d). PGE2 is attached mainly to the C-domain of SLCO2A1 (Fig. 3c). The translocation path is lined by several basic residues, which can attract negatively charged moieties of prostaglandins (Fig. 3c), thereby facilitating prostaglandin recognition and binding by SLCO2A1.

Fig. 3: PGE2 recognition by SLCO2A1.
figure 3

a, b The EM map of SLCO2A1 in complex with PGE2 (SLCO2A1-PGE2) embedded in LMNG micelle (gray) (a) and the cartoon representation of SLCO2A1-PGE2 (b). The PGE2 is colored in cyan. The N and C domain of SLCO2A1-PGE2 are colored in green and pink, respectively. c A cut-open electrostatic potential surface of PGE2 binding pocket. PGE2 is depicted in cyan sticks. The electrostatic potential surface was calculated in PyMOL (red to blue, −50 kT/e to +50 kT/e). d The electron microscopy densities of PGE2 and nearby residues. The EM densities for PGE2 and nearby residues were contoured at 5.8 σ. The density of PGE2 and nearby residues are shown in blue and orange meshes, respectively. e, f Detailed interactions between PGE2 and SLCO2A1. Residues involved in PGE2-binding are shown as sticks. Hydrogen bond interactions are indicated in orange dashed lines. Distances are indicated by black dashed lines. g 6-CF transport by HEK293F cells expressing SLCO2A1 WT or SLCO2A1 mutants involved in PGE2 binding. Cells expressing mCherry served as a control. Data were normalized to SLCO2A1 and are presented as mean ± s.e.m.; n = 3 independent replicates for mutants. n = 9 independent replicates for mCherry and SLCO2A1 WT. One-way ANOVA was used; ****P < 0.0001. h Concentration-dependent inhibition of 6-CF transport by PGE2 in HEK293F cells expressing SLCO2A1 mutants, S346A, S346F and W565F. Cells expressing mCherry served as a control. Data were normalized to net 6-CF transport activity of S346A, S346F or W565F without PGE2 and are presented as mean ± s.e.m.; For S346A, n = 6 independent replicates for log [PGE2 (μM)] = −3, n = 7 independent replicates for log [PGE2 (μM)] = −2, n = 3 independent replicates for all other conditions; For S346F, n = 3 independent replicates for log [PGE2 (μM)] = −1 or 0, n = 4 for all other conditions; For W565F, n = 7 independent replicates for log [PGE2 (μM)] = −3, n = 6 independent replicates for log [PGE2 (μM)] = −1, n = 3 independent replicates for all other conditions in W565F. Source Data are provided as a Source Data file.

The complex structure reveals a detailed PGE2 binding site in SLCO2A1. Specifically, the carboxyl group of PGE2 forms hydrogen bond interactions with S339 on TM7 as well as van der Waals interactions with F212 on TM5, M380 on TM8, H533 on TM10 and W565 on TM11; the middle five-membered ring is sandwiched by L46 and W565 from the N- and C-domains, respectively; the 15-position hydroxyl moiety participates in a hydrogen bond interaction with S346 on TM7; and the hydrophobic tail inserts into a small cavity between TM7 and TM8, forming hydrophobic interactions with I342, L345, and L373 (Fig. 3e, f). Interestingly, the strictly conserved Arg residue on TM11 of OATPs (Supplementary Fig. 5), R561, is positioned close to the five-membered ring of PGE2 at a distance of 8.3 Å, suggesting a lack of direct interactions between PGE2 and R561 in this state. Similarly, the structure of OATP1B1 in complex with E3S (PDB code: 8HND) revealed that E3S binds in a pocket 11.3 Å away from the equivalent R580 in OATP1B124. We speculate that this conserved Arg may provide a positively charged strength that facilitates anionic ligand binding to the corresponding site and potentially interacts the anionic ligand in other conformational states.

To validate the observed PGE2 binding site in SLCO2A1, we generated eight structure-guided mutations: L46A, S339A, S346A, S346D, S346F, R561Q, W565A, and W565F. All these mutants exhibited comparable expression in HEK293F cells and similar solution behaviour and cellular localization to that of WT SLCO2A1 (Supplementary Figs. 6 and 7). Our 6-CF uptake assay results demonstrated that L46A, S339A, S346D, R561Q, and W565A almost completely abrogated 6-CF uptake, whereas S346A resulted in a 1.5-fold increase in activity, and S346F and W565F retained approximately 20% and 50% activity compared with that of WT SLCO2A1, respectively (Fig. 3g), suggesting that the binding site for 6-CF partially overlaps with that of PGE2. In addition, we also evaluated the transport activity of key pathogenic mutants, particularly those located in the substrate binding pocket (S339P, R561C, and W565G). As expected, these pathogenic mutants almost completely abrogated 6-CF uptake (Fig. 3g). The structure of SLCO1B1 in complex with 2′,7′-dichlorofluorescein (DCF) (SLCO1B1-DCF, PDB code: 8K6L)24, an analogue of 6-CF, was used to analyse the potential binding pattern of 6-CF in SLCO2A1 (Supplementary Fig. 4c, d). The superposition of the substrate binding pockets of SLCO2A1-PGE2 and SLCO1B1-DCF revealed that the DCF and PGE2 sites are distinct and partly include S339, I342, L345, and W565 (Supplementary Fig. 4e). In addition, DCF clashes with L345 and I342 of SLCO2A1, suggesting that 6-CF may adopts a different binding pose in SLCO2A1. Because the S346A, S346F, and W565F mutants have 6-CF uptake activity (Fig. 3g), we further investigated the concentration-dependent inhibition of 6-CF transport by PGE2 in the S346A, S346F, and W565F mutants (Fig. 3h), yielding IC50 values for PGE2 of 6.55 ± 1.76 μM in S346A, 15.39 ± 3.22 μM in S346F and 1.18 ± 0.33 μM in W565F. Compared the WT SLCO2A1, the S346A mutation slightly increased IC50 value by approximately 1.5-fold, whereas the more aggressive mutation S346F caused a 3.5-fold increase, suggesting that the interactions between S346 and the 15-position hydroxyl group of PGE2 have a role in PGE2 recognition. (Fig. 3e). Unexpectedly, the ability of PGE2 to inhibit 6-CF uptake by the W565F mutant substantially increased (Fig. 3h). We postulate that the phenyl ring of W565F mimics the indole ring of Trp, which can stabilize the five-membered ring of PGE2, as evident by the structural finding that the nitrogen atom in the indole ring of W565 has no direct interaction with PGE2 (Fig. 3e).

Previous studies17 and our results demonstrate that SLCO2A1 has high specificity for prostaglandins (Fig. 1f), which is distinct from other OATPs. Prostaglandins are characterized by a 20-carbon fatty acid backbone that forms a five-membered ring flanked by two side chains (Supplementary Fig. 8). Compared with that of other OATPs24,37, the substrate binding pocket of SLCO2A1 is narrowed by L46 on TM1 and W565 on TM11, making it more suitable for prostaglandin binding instead of larger compounds such as E3S, T4 and taurocholate (Fig. 3e, f and Supplementary Fig. 4f, g). In particular, this L46–W565 pair is not conserved in the OATP family; additionally, S339 and S346, which form two hydrogen bonds with PGE2, are not conserved in the OATP family (Supplementary Fig. 5). These analyses imply that L46, S339, S346, and W565 are presumably the molecular determinants of prostaglandin specificity in SLCO2A1.

State transitions of SLCO2A1

MFS transporters experience outward-open, occluded, and inward-open state transitions during a transport cycle, known as the “rock-switch”-like alternating access mechanism38,39. The structures of OATPs in distinct functional states suggest that OATPs employ this common model during general conformational movements23,24. Both the apo- and PGE2-bound SLCO2A1 structures were resolved in an outward-open conformation (Figs. 2a and 3a). To investigate transport dynamics of SLCO2A1, we have performed MD simulations using the SLCO2A1-apo and SLCO2A1-PGE2 structures as an initial model in a lipid bilayer environment individually (Supplementary Fig. 9). We repeated 4 independent 500-ns simulations for each of the structures, which all reached a steady state and demonstrate that the TM core domains are more stable than the extramembrane domains (Supplementary Fig. 9a–d). Although no state transitions were observed in our simulation systems, we found that the PGE2-bound SLCO2A1 tends to close the opened extracellular gate, as clustering analysis indicates the distances between the N- and C- domains of SLCO2A1-PGE2 are 1–2 Å smaller than those of SLCO2A1-apo (Supplementary Fig. 9e, f). To further reveal detailed conformational changes in SLCO2A1 state transitions, we generated an inward-open model of SLCO2A1 on the basis of the inward-open structure of OATP1B1 (PDB code: 8HND) and AlphaFold2 prediction (Fig. 4a, b). As expected, the two N domains or the two C domains from the two distinct states exhibit marked structural similarity, indicating that the two domains undergo rigid body-like motion (Fig. 4a). Superposition of the outward-open structure with the inward-open model using the N-domain as a reference revealed that the C-domain swung ~30° relative to the N-domain (Fig. 4b), which is in line with the findings of other MFS transporters40. Additionally, the PGE2-binding site remains relatively stable, with subtle side-chain rotations (Supplementary Fig. 10a, b).

Fig. 4: Structural basis of human SLCO2A1 state transitions.
figure 4

a Structural comparisons of the N- and C-domains of SLCO2A1-PGE2 and AlphaFold predicted SLCO2A1-inward (gray). b Structural alignments between SLCO2A1-PGE2 and AlphaFold predicted SLCO2A1-inward (gray). The N-domain was used as a reference for structural alignment. c The intracellular gate of the outward-open SLCO2A1-PGE2 viewed from the cytosolic side. Residues involved in N and C domain interactions are shown in sticks. Hydrogen bonds are indicated with orange dashed lines. d The intracellular gate of SLCO2A1-inward viewed from the cytosolic side. Residues which involved in intracellular gate are shown in sticks and blue dashed lines indicate the distances between them. e, f π–π interactions between F311, F315, F551, and Y90 on the intracellular side of SLCO2A1-PGE2 (e) but not in SLCO2A1-inward (f). g 6-CF transport by HEK293F cells expressing SLCO2A1 WT or SLCO2A1 mutants involved in states transition. Cells expressing mCherry served as a control. Data were normalized to SLCO2A1 WT and are presented as mean ± s.e.m.; n = 3 independent replicates for mutants. n = 9 independent replicates for mCherry and SLCO2A1 WT. One-way ANOVA was used; ****P < 0.0001. Source Data are provided as a Source Data file.

In the outward-open SLCO2A1, the intracellular gate is closed by TM4, TM5, TM10, and TM11 (Fig. 4c). Two conserved salt bridges between D197 and K549 and between D196 and R542 play a role in stabilizing the closed intracellular gate (Fig. 4c) and these salt-bridge interactions were disrupted in the inward-open model (Fig. 4d). Notably, the R542C mutation causes PHO disease41,42. To validate these specific interactions, we generated the D196A, D197A, and R542C mutations and assayed their effects on transport activity. The D197A mutation almost abolished 6-CF uptake activity, whereas D196A and R542C retained approximately 20% 6-CF uptake activity compared to that of SLCO2A1 WT (Fig. 4g). These results suggest that disruption of the salt-bridge interactions could change the equilibrium of state transitions, thereby changing transport activity. Additionally, S550 from the C-domain participates in hydrogen bond interactions with R97 and S193 from the N-domain; N203 from the N-domain forms a hydrogen bond with K388 from the C-domain (Fig. 4c). The R97C and R97H mutations cause PHO disease43,44. Similar to the R542C mutation, R97C greatly impaired the 6-CF transport activity, protein expression, and stability (Fig. 4g and Supplementary Fig. 6e, i). By contrast, these hydrogen bond pairs are apart at long distances in the inward-open model (Fig. 4d). Furthermore, the amphipathic CH1 anchors the N- and C-domains via extensive interactions. Moreover, in the outward-open conformation, Y90 on TM2 is closely embraced by F311 and F315 on CH1, as well as F551 on TM11. Conversely, in the inward-open model, Y90 moves away from this phenylalanine triad (Fig. 4e, f). The Y90A mutant retained only 60% activity of WT SLCO2A1 (Fig. 4g), suggesting that these specific interactions contribute to sealing the intracellular gate in the outward-open state. Consistently, a previous study revealed that substituting the equivalent Y86 of OATP1B1 with an Ala almost completely abolished DCF uptake24.

In the predicted inward-open SLCO2A1, the extracellular gate is closed by multiple polar and nonpolar interactions between the N- and C-domains (Supplementary Fig. 10c). Among these interactions, one salt-bridge formed by E60-K351 is highly conserved in OATPs (Supplementary Fig. 5). Notably, the charge-switch mutation E60K is linked to PHO disease43 and this mutation completely abolished 6-CF uptake activity (Fig. 4g), suggesting an important role in SLCO2A1 function. Moreover, the hydrogen bond interactions of T57-E354, D577-S67-Y597, and K53-S346-Y223 further contributed to the closeness of the extracellular gate (Supplementary Fig. 10c, d). Conversely, these residues are positioned at large distances in our outward-open SLCO2A1 structure (Supplementary Fig. 10e, f).

Pathogenic mutations in SLCO2A1

To date, more than fifty pathogenic mutations in SLCO2A1 have been reported8. We mapped 52 missense mutations onto our SLCO2A1 structure, which demonstrated a broad distribution (Fig. 5a). In particular, many mutations are concentrated along the interface of the N- and C-domains, which probably affects substrate recognition and translocation (Fig. 5a). For example, W565 of TM11 forms essential hydrophobic interactions with the five-membered ring of PGE2, and the W565G mutation abolishes this critical interaction; similarly, S339 forms a H-bond with the hydroxyl group of PGE2, and the S339P mutation disrupts this H-bond, thereby affecting substrate binding in the pocket (Figs. 3f, 5a). Several residues, such as Q556, F557, and R561, although having no direct interactions with PGE2, form essential interactions with residues responsible for PGE2 binding, the mutation of which probably reduces the stability of the central pocket.

Fig. 5: Pathogenic mutations in human SLCO2A1.
figure 5

a Missense mutations in SLCO2A1, which cause PHO and CEAS. Mutations are shown in spheres. Pathogenic mutations in the TM core and the extracellular loops are colored in purple and blue, respectively. Dashed squares represent focused regions shown in (be). be Interactions between presentative mutations and nearby residues. F34, R97, F275, C420, E427, R445, C459, E465, and nearby residues are shown in sticks. G183, G222, and nearby residues are shown in spheres. Hydrogen bonds are indicated with orange dashed lines. f 6-CF transport by HEK293F cells expressing SLCO2A1 WT or pathogenic SLCO2A1 mutants. Cells expressing mCherry served as a control. Data were normalized to SLCO2A1 WT and are presented as mean ± s.e.m.; n = 3 independent replicates for mutants. n = 9 independent replicates for mCherry and SLCO2A1 WT. One-way ANOVA was used. ****P < 0.0001. Source Data are provided as a Source Data file.

Although glycine and proline usually break a typical α-helix, they are common in transmembrane helices45, which could act as a hinge and are often of functional importance in membrane proteins. In SLCO2A1, pathogenic mutations, including G71R, G104R, G181D/A, G183R, G222R/E, G369D, and G554R in the TMD, replace glycine with residues possessing larger side chains, leading to the loss of helical dynamics and potential disruption in helical packing (Fig. 5a, b). To test this hypothesis, we investigated the effects of two representative mutations G183R and G222R, both of which almost completely abolished 6-CF uptake activity and caused great reduction in protein expression and stability (Fig. 5f and Supplementary Fig. 6e, h). In contrast, the mutations L179P and L563P may disrupt their respective α-helices, leading to protein instability46. On the intracellular side, mutations such as F275V and F34S may affect helical packing in the N-domain, and R97C/H may impair interactions between the N- and C-domains (Fig. 5c).

On the extracellular Kazal-like domain, the substitution of R445 with Cys ablated electrostatic interactions with E427 and H471, thereby destabilizing EL9–10 (Fig. 5d). In addition, the C420F, C492Y, and C459R mutations disrupt the conserved disulfide bonds, which are apparently critical for the arrangement of the Kazal-like domain (Fig. 5a, e). Moreover, V458F, G460R, and E465K cause dramatic local changes, which likely affect domain stability. Consistent with the structural observation, the E427K mutant exhibits little 6-CF uptake activity, weak protein expression, and poor protein stability (Fig. 5f and Supplementary Fig. 6f, j). The R445C mutation retained only 60% of the transport activity compared to WT SLCO2A1, whereas the C420F and E465K greatly decreased 6-CF uptake activity by more than 80% (Fig. 5f). Moreover, the C420F and E465K mutants exhibit more oligomer peaks in the fSEC profiles and different cellular localization compared to that of WT SLCO2A1 (Supplementary Figs. 6f and 7), suggesting that C420F and E465K may impair protein stability and trafficking.

Discussion

OATP transporters are known for their ability to carry a wide variety of structurally diverse organic anionic substances across the membrane19. However, consistent with previous results17, our functional assays revealed that SLCO2A1 is specific for prostaglandins (Fig. 1f). We determined the cryo-EM structures of human SLCO2A1 in both apo and PGE2-bound states, revealing that SLCO2A1 possesses a central pocket different from those of other OATPs, which is suitable for prostaglandins. The SLCO2A1-PGE2 complex structure reveals a detailed binding mode for PGE2, which is recognized by specific hydrogen bond interactions from S339 and S346 as well as hydrophobic interactions from L46 and W565 (Fig. 3e, f). More importantly, sequence alignment revealed that the residues participating in these specific interactions are not conserved in OATPs, indicating that L46, S339, S346, and W565 are most likely the molecular determinants of PGE2 specificity in SLCO2A1. Notably, 6-CF is structurally distinct from PGE2, which can be transported by SLCO2A1 (Fig. 1a, b). Although we did not determine the complex structure of SLCO2A1-6-CF, structural alignment of the PGE2-binding pocket with the binding site of DCF in OATP1B1, a close analogue of 6-CF, revealed that PGE2 and DCF share very limited overlapping regions (Supplementary Fig. 4e). This observation suggests that in addition to the primary PGE2 binding site, SLCO2A1 may have a relatively low-affinity binding site. This potentially explains why PGE2 has a much lower potency in competitively inhibiting 6-CF (4.38 μM; Fig. 1c) than in competitively inhibiting radioactive PGE2 (25–331 nM)26.

Like other MFS transporters, SLCO2A1 employs the general alternating access transport mechanism, swinging the N- and C-domains around the central site during a transport cycle; however, the energy coupling mechanism of SLCO2A1 is not fully understood. Although evidence suggests that the transport activity of some OATPs, including OATP1B1, is pH dependent47, PGE2 uptake by SLCO2A1 is not affected by pH17 but instead is reduced by depletion of intracellular ATP or at low temperature48. Further evidence demonstrated that SLCO2A1-mediated PGE2 uptake in Xenopus oocytes was significantly reduced when the cell membrane was depolarized in the presence of external K+, indicating that SLCO2A1-mediated PGE2 uptake is regulated by the membrane potential7. Other evidence suggests that preloading cells with lactate causes dose-dependent increase of SLCO2A1-mediated PGE2 uptake, while external lactate causes increase of SLCO2A1-mediated PGE2 efflux49. However, future investigations are needed to determine the exact driving force that energizes SLCO2A1 transport.

In addition to its role in prostaglandins transport, SLCO2A1 is an essential core component of the ATP-conductive Maxi-Cl channel50,51. However, in the outward-open structure and the inward-open Alphafold2 model of SLCO2A1, no obvious continuous channels were observed. Some research has demonstrated that PGE2 and SLCO2A1 blockers suppress Maxi-Cl currents in cells expressing SLCO2A150, suggesting that this potential chloride pathway at least partly overlaps with the central pocket. We speculate that SLCO2A1 may have a unique intermediate state for conducting chloride, which needs to be activated by binding partner proteins or under certain conditions. In summary, our work presents the detailed PGE2 binding site, substrate selectivity, conformational changes, and pathogenic mutations of SLCO2A1, elucidating the structural basis for the substrate specificity and transport mechanism, which could aid future therapeutic development for SLCO2A1-related diseases.

Methods

SLCO2A1 and mutant transport assays using 6-CF

The uptake activity of SLCO2A1 and its mutants was measured by 6-CF influx using HEK293F cells (Gibco, USA) as adapted from a previously reported method52. The expression of SLCO2A1 and different mutants for its transport activity measurement are similar to the expression for structure determination, except for its N-terminal GFP tag replaced by an mCherry tag. Cells transfected with mCherry alone were used as blank controls. Forty-eight hours post-transfection, 2 mL of cells were collected and washed twice with pre-warmed phosphate-buffered saline (PBS). Subsequently, 0.5 mL of pre-warmed uptake buffer (142 mM NaCl, 5 mM KCl, 1 mM KH2PO4, 1.2 mM MgSO4, 1.5 mM CaCl2, 5 mM d-glucose, and 12.5 mM MES, pH 6.0) containing 50 μM 6-CF (MedChemExpress) was added to each tube. After being incubated for 5 min at 37 °C, the uptake buffer was removed by centrifugation at 100 × g for 1 min, and the cells were washed three times with pre-cooled PBS to terminate transport. Then, 500 μL of uptake buffer containing 2% Triton X-100 was added to each tube and incubated for 1 h at 4 °C to lyse the cells. The supernatants of the lysates were obtained after centrifugation at 12,000 × g for 10 min to remove cell debris, and the protein content in the cell lysates was determined via the Lowry method. During the measurement of intracellular 6-CF accumulation via a microplate reader (PerkinElmer EnVision Multilabel Plate Reader), 200 μL of cell lysate was added to each well to detect the fluorescence intensity of 6-CF (excitation/emission wavelengths at 485 nm/535 nm).

The inhibition of SLO2A1-mediated 6-CF uptake by E3S, PGE2, T4, and taurocholate (MedChemExpress) was performed using the above-mentioned method. In addition, 50 μM E3S, PGE2, T4, or taurocholate was added to the uptake buffer together with 50 μM 6-CF. Significant differences were performed with one-way ANOVA followed by Dunnett’s multiple comparisons test to compare the WT with the mutants via GraphPad Prism 10.4.2. All transport assays were repeated at least three times.

When testing the concentration dependence of 6-CF transport by SLCO2A1 WT, uptake buffers containing 0 μM, 50 μM, 100 μM, 200 μM, and 600 μM 6-CF were used. To evaluate the concentration-dependent inhibition of 6-CF transport by PGE2 or suramin, uptake buffer containing 50 μM 6-CF was mixed with PGE2 or suramin at six concentrations ranging from 0.001 to 100 μM. At each inhibitor concentration, the 6-CF transport activity of mCherry-transfected HEK293F cells was measured as a control. The 6-CF uptake activity of the mCherry control was defined as 0%. The net 6-CF uptake activity of the cells transfected with WT SLCO2A1 or its mutants in the presence of each indicated concentration of inhibitor was calculated by subtracting the 6-CF uptake activity of the mCherry control, and the net activity of WT SLCO2A1 or its mutants in the absence of the inhibitor was defined as 100%. The Km and IC50 were analysed by nonlinear regression via GraphPad Prism 10.4.2.

Expression and purification of human SLCO2A1

The genes encoding full-length human wild-type SLCO2A1 (UniProt accession code: Q92959) were cloned and inserted into a pEG2 BacMam expression vector with a twin-Strep tag, a green fluorescent protein (GFP), followed by an HRV 3C protease site at the N-terminus. All SLCO2A1 mutants were generated via standard site-directed mutagenesis methods using WT SLCO2A1 as a template and were confirmed via DNA sequencing. All the primers are summarized in Supplementary Table 2. Bacmids were generated by transforming the plasmids into DH10bac competent cells (Thermo Fisher). After that, we used the Bac-to-Bac baculovirus expression system (Invitrogen) to produce baculoviruses in Spodoptera frugiperda Sf9 insect cells (Invitrogen, USA), and P3 viruses were collected for transfecting HEK293F cells. HEK293F cells (Gibco, USA) cultured in OPM-293 medium (OPM, China) in a 37 °C incubator with 5% CO2 were transfected with P3 baculovirus at a ratio of 1:150 (v/v) when the cell density reached 2.5 × 106 cells/mL (C100, RWD). After 12 h, 10 mM sodium butyrate (Sigma, USA) was added to the culture mixture. Another 48 h later, the cells were harvested via centrifugation, and the cell pellets were quickly frozen in liquid nitrogen and stored at −80 °C.

For purification of SLCO2A1, cell pellets were resuspended in buffer A containing 20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM β-mercaptoethanol (β-ME), and a protease inhibitor cocktail including 1 mM phenylmethylsulfonic acid acyl fluoride (PMSF), 2 μM pepstatin, 4.2 μM leupeptin, and 0.8 μM aprotinin. The membranes were collected by ultracentrifugation at 158,600 × g for 40 min and solubilized in buffer A supplemented with 2 mM ATP, 5 mM MgCl2, 1% (w/v) lauryl maltose neopentyl glycol (LMNG, Anatrace, USA), and 0.1% (w/v) cholesterol hemisuccinate (CHS, Anatrace, USA) for 1 h at 4 °C. The insoluble part was removed by another 30 min ultracentrifugation at 158,600 × g, and the supernatant was loaded onto Streptactin beads (Smart-Lifesciences, China) at 4 °C. The resin was washed with 10 column volumes of buffer B (buffer A supplemented with 0.01% (w/v) LMNG and 0.001% (w/v) CHS). The protein was subsequently eluted with buffer B supplemented with 5 mM desthiobiotin (Sigma, USA). The GFP and Twin-strep tags were cleaved by PreScission protease (1:50 w/w ratio) on ice overnight. The tag-removed SLCO2A1 was subsequently concentrated and subjected to a Superose 6 increase 10/300 GL column (Cytiva) equilibrated with buffer C containing 20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM β-ME, 0.002% (w/v) LMNG, and 0.0002% (w/v) CHS. The peak fractions were collected and concentrated to 10 mg/mL for cryo-EM grid preparation.

For purification of SLCO2A1 in complex with PGE2, PGE2 (MCE) was added to the purification buffers and the SEC buffer at a concentration of 10 μM. PGE2 (1 mM) was added to the concentrated peak fractions, which were subsequently incubated on ice for 1 h before cryo-EM grid preparation.

To characterize the biochemical properties of SLCO2A1 mutants, two milliliters of cultured HEK293F cells expressing mCherry alone, SLCO2A1 WT or mutants were collected via centrifugation. The cell pellets were solubilized in 250 μL buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 5 mM β-ME, and a protease inhibitor cocktail including 1 mM PMSF, 2 μM pepstatin, 4.2 μM leupeptin, and 0.8 μM aprotinin, 1% (w/v) LMNG, and 0.1% (w/v) CHS for 1.5 h at 4 °C. Insoluble material was removed by ultracentrifugation, and the supernatant containing SLCO2A1 mutants was collected. The total protein amount of the samples was measured by BCA assay. Samples containing same amount of total protein amount were separated with SDS-PAGE and the gels were then subjected to in-gel fluorescence imaging (Ex 587 nm). In addition, the samples were assayed by fluorescence detection size-exclusion chromatography (fSEC) (Ex 587 nm). To confirm the cellular localization of SLCO2A1 WT and mutants, HEK293F cells expressing mCherry alone, mCherry-tagged SLCO2A1 WT, and all the mutants were imaged by the A1R Si laser scanning confocal microscope (CLSM) for mCherry fluorescence (587 nm) and brightfield. The confocal images are presented in Supplementary Fig. 7.

Cryo-EM sample preparation and data collection

The purified samples were centrifuged at 15,600 × g for 20 min at 4 °C before cryo-EM sample preparation. Aliquots of 3.0 μL samples were added to glow-discharged grids (ANTA, 300 mesh, R1.2/1.3, Au). The grids were blotted for 3.5–4.5 s at 4 °C with 100% humidity, plunged into liquid ethane cooled by liquid nitrogen and stored in liquid nitrogen.

All datasets were collected via a Titan Krios G4 microscope (Thermo Fisher Scientific) operated at 300 kV and equipped with a K3 direct electron detector and a 10 eV energy filter (Gatan). All movie stacks were fractioned into 40 frames and automatically collected via EPS at a nominal magnification of 105,000 with a pixel size of 0.85 Å and a defocus range between −2.0 and −1.0 μm. The total dose was 60 e/Å2, and the dose rate was adjusted to 15 e/pixel/s. A total of 2425 and 1788 movie stacks were collected for SLCO2A1-apo and SLCO2A1-PGE2, respectively.

Cryo-EM data processing

The data processing pipelines are shown in Supplementary Figs. 2 and 3. All the data were processed via RELION-353 and CryoSPARC54.

For SLCO2A1-apo, the data were processed in CryoSPARC54. The 2425 movie stacks were motion corrected and binned twofold via patch motion, and contrast transfer function (CTF) parameters were estimated via patch CTF estimation. After that, micrographs were used for blob particle picking. A total of 2,151,834 particles were picked and extracted in bin 2 and a box of 128 pixels. After two rounds of 2D classification, 378,073 particles were selected for ab initio reconstruction. Two rounds of ab initio generation generated three classes of particles, each containing a 3D map. The best map containing clearly resolved transmembrane helical density was then used as a model for hetero refinement. A total of 1,320,750 particles were selected from the last round of 2D classification and subjected to three rounds of hetero refinement. A total of 178,724 particles from the best class of hetero refinement underwent one round of 2D classification, and 175,128 particles were selected and retracted in bin 1 and a box of 320 pixels. Then, the 175,128 particles were subjected to nonuniform refinement with C1 symmetry imposed, yielding a refined map at 3.1 Å resolution. Further local refinement with a mask covering only SLCO2A1 without detergent micelles yielded a refined map at a resolution of 3.2 Å.

For SLCO2A1-PGE2, the 1788 movie stacks were motion corrected, binned twofold, and dose-weighted via MotionCorr255, and the defocus values of each summed micrograph were estimated via Gctf56. After that, micrographs were used for autopicking. A total of 1,746,885 particles were picked and extracted in bin 2 and a box of 128 pixels. The particles were subsequently imported into CryoSPARC54 and subjected to two rounds of 2D classification to remove junk particles, resulting in 919,153 good particles, which were subsequently used for ab initio reconstruction. After two rounds of ab initio reconstruction, the best class with a map containing clearly resolved transmembrane helical density consisted of 208,619 particles. Furthermore, the 919,153 good particles underwent three rounds of hetero refinement via the model generated ab initio. A total of 222,289 particles from the best class of hetero refinement and 208,619 particles from the best class of ab initio refinement were merged to remove the duplicates, resulting in a total of 259,151 particles. These particles were then retracted in bin1 and a box of 320 pixels and subjected to nonuniform refinement with C1 symmetry imposed, yielding a refined map at 3.09 Å resolution. Further local refinement with a mask covering only SLCO2A1 without detergent micelles yielded a refined map at 3.07 Å resolution.

Model building and refinement

The predicted AlphaFold2 model of SLCO2A1 was fitted into the cryo-EM density map of SLCO2A1-apo and SLCO2A1-PGE2 in Chimera57 and was manually inspected and adjusted in Coot58. For SLCO2A1-PGE2, PGE2 was manually fitted into the extra densities within the maps in Coot58. Refinement of the SLCO2A1-apo and SLCO2A1-PGE2 models against the cryo-EM map in real space was performed via phenix.real_space_refine in PHENIX59. The models were subsequently checked and corrected in Coot. The model vs. map FSC curve was calculated via Phenix.mtrage. The statistics of the cryo-EM data collection and model refinement of SLCO2A1-apo and SLCO2A1-PGE2 are summarized in Supplementary Table 1.

All figures were prepared using ChimeraX60 and PyMol (Schrödinger, LLC).

Molecular dynamics simulations

The predicted AlphaFold2 model Molecular dynamics (MD) simulations were conducted to investigate the conformational dynamics of the SLCO2A1. Two systems, SLCO2A1-apo and SLCO2A1-PGE2, were studied. The cryo-electron microscopy (cryo-EM) structures were used as the initial models for these simulations. To configure the protonation states of the protein residues, the PROPKA3 (version 3.4.0)61 was used to empirically estimate the pKa values. The protein termini were capped with ACE (acetyl) and NME (N-methylamide) groups for the N- and C-termini, respectively, to maintain a neutral charge. The details of the configuration of MD simulation were consistent with our previous study62.

All simulations were performed using the GROMACS2021 suite63. For each system, four independent production runs were performed, each with a duration of 500 ns. No constraints were imposed on the protein or the substrate during the simulations.

Following the simulations, extensive analysis was conducted on the generated trajectories. The root-mean-square deviation (RMSD) of the protein backbone was calculated to assess the conformational stability, while the root-mean-square fluctuation (RMSF) was used to quantify the flexibility of individual residues. We also monitored the center-of-mass distance between the N-lobe and C-lobe of the protein to characterize their relative movement. For a comprehensive analysis of the conformational landscape, all four trajectories for each system were combined, and a clustering analysis was performed using a 1.5 Å cutoff. The top five most populated cluster centroids were selected as representative structures for further study.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.