MFSD2A in focus: the molecular mechanism of omega-3 fatty acid transport

Abstract

Omega-3 fatty acids, such as docosahexaenoic acid (DHA), are essential nutrients required to support growth, maintenance, and function of the central nervous system (CNS). While the brain has a high demand for DHA, it cannot synthesize it de novo and thus relies on its uptake from the bloodstream. Circulating DHA is primarily obtained from dietary sources and is transported across the blood-brain barrier (BBB) in the form of lysophosphatidylcholine (LPC-DHA) by the transmembrane transporter Major Facilitator Superfamily Domain containing 2A (MFSD2A), in a sodium-dependent manner. Here we provide a comprehensive analysis of recent insights gained from structural, functional and computational studies of MFSD2A. We focus on the mechanism by which this transporter mediates sodium-dependent uptake of LPC-DHA, and lysolipids more broadly, highlighting different conformational states, substrate entry and release pathways, and the ligand binding sites. This review presents a detailed overview of the molecular mechanism that enables MFSD2A to supply the brain with this essential nutrient, while simultaneously providing biophysical insights into how lysolipids are transported across biological membranes.

The importance of DHA for the brain

Docosahexaenoic acid (DHA; 22:6[n-3]; Fig. 1a) is the most abundant omega-3 fatty acid in the brain, followed by alpha-linolenic acid (ALA; 18:3[n-3]; (1)) and eicosapentaenoic acid (EPA; 20:5[n-3]; (1)). All three are essential nutrients and must be obtained from the diet. While small amounts of DHA and EPA can be synthesized in the liver from diet-acquired ALA (2, 3), most DHA and EPA are obtained from dietary sources such as fatty fish, seaweed, and DHA-fortified foods such as eggs, milk, and infant formula (4, 5).

Figure 1 |. MFSD2A is an LPC-DHA transporter at the blood-brain barrier.

A, Chemical structures of docosahexaenoic acid (DHA) and Lysophosphatidylcholine-DHA (LPC-DHA). B, Schematic representation of the BBB, highlighting MFSD2A (PDB: 7MJS) expression within the luminal membrane of BBB endothelial cells. C, Topology of a canonical MFS transporter showing twelve numbered TM helices. The N-domain (TM1–6) is colored purple, and the C-domain (TM7–12) is colored green. Each domain consists of two 3-TM structurally inverted repeats, represented by the semi-transparent triangles behind the helices. The N- and C-domains are connected by a cytosolic loop. Soluble substrates are represented by the orange circle containing ‘S’. D, Schematic representation of the “rocker-switch” mechanism of transport utilized by canonical MFS transporters, highlighting the OFS (left), OcS (middle) and IFS (right). Substrate is again represented by an orange circle containing ‘S’. E, Simplified schematic of MFSD2A driven LPC-DHA transport. The N-domain is colored blue, and the C-domain is colored yellow. The extracellular disulfide crosslink is shown in purple. Sodium (magenta pentagon) and LPC-DHA (green) are shown in the outer leaflet of the membrane adjacent to the transporter in its OFS prior to import (left). As sodium and LPC-DHA enter MFSD2A, they migrate into the central cavity of the transporter simultaneously, and MFSD2A shifts into an OcS (middle). As MFSD2A shifts into an IFS (right), it releases Na⁺ into the cytoplasm and LPC-DHA into the inner leaflet of the membrane.

DHA is critical for the development and maintenance of the central nervous system (CNS) during development, as well as for memory and cognition in adulthood (6). DHA is also well-studied for its role in healthy aging, having protective effects up on mental health, paired association learning, verbal fluency and lowering the risk of cognitive decline (7). While the precise underlying molecular mechanisms in which DHA provides such protection remain to be fully elucidated, numerous studies have shed light on some promising candidate pathways (6). DHA-containing phospholipids constitute an important structural component of cell membranes, particularly in neurons (8). DHA contains a total of 22 carbons, with six double bonds, which create multiple kinks in its hydrocarbon chain structure (Fig. 1a). These properties make DHA difficult to pack tightly amongst other components of the membrane bilayer, impacting overall membrane fluidity (9). This is suggested to increase the membrane’s resistance to extreme temperatures and lipid peroxidation (10, 11), both of which are likely to be important for extending neuronal lifespan. Furthermore, omega-3 fatty acid enrichment has been shown to reduce inflammatory damage to neurons and increases neuronal dendrite growth dynamics (12–15). Finally, DHA is also an established modulator of gene expression, memory formation, nociception, apoptosis, and chemokine expression (16–19).

Transport of DHA across the blood-brain barrier

To enter the brain, dietary DHA must cross the blood-brain barrier (BBB). The BBB is a semi-permeable barrier composed of tightly packed endothelial cells, pericytes, a basement membrane and astrocyte foot processes that separate the brain from circulating blood (Fig. 1b; (20)). In the first layer of the BBB, adjacent endothelial cells lining the brain’s capillaries are closely joined together by protein complexes called tight junctions, preventing large molecules and most water-soluble substances from passing through the space between the cells (Fig. 1b; (21)). These BBB endothelial cells also have low rates of transcytosis, further reducing the passage of macromolecules (22). Surrounding the endothelial cells is a thick basement membrane made of extracellular matrix proteins, providing additional structural support and restricting substance passage (23). Pericytes are embedded within this basement membrane, where they contribute to structural support, regulate blood flow, and maintain integrity of the BBB (24). Finally, astrocytes have end-foot processes that envelop brain capillaries and release signaling molecules that help maintain and regulate BBB function (Fig. 1b; (25)). Together, these features comprise an impermeable barrier to most molecules while permitting the passage of oxygen and the selective entry of certain molecules, including nutrients, a process which is often facilitated by specific transporters (20).

Major Facilitator Superfamily Domain containing 2A (MFSD2A) is a Na⁺ -coupled lipid transporter expressed on both the luminal and abluminal membranes of BBB endothelial cells, where it mediates the uptake of DHA into the brain (Fig. 1b; (26, 27)). This process is thermodynamically driven by an inwardly directed Na⁺ gradient (27). Importantly, for DHA to be transported by MFSD2A, it must be esterified as lysophosphatidylcholine-DHA (LPC-DHA; Fig. 1a), non-esterified DHA cannot be transported by MFSD2A (27). MFSD2A can also transport a variety of other lysolipids such as LPC-18:1, LPC-ALA, LPC-oleate, LPC-palmitate and lysophosphatidylserine-oleate (LPS-oleate; Supp Tab. 1). The minimum requirements for substrate transport are that the lipid has a hydrocarbon chain containing at least 14 carbons (with a preference for unsaturated lipids) and that the lipid has been esterified to a zwitterionic headgroup, or a zwitterion with an additional anionic group, such as LPS (experimentally determined) or lysophosphatidylethanolamine (LPE; computationally predicted; Supp Tab. 1(28, 29)).

In humans, single nucleotide polymorphisms (SNPs) in MFSD2A and changes in its expression levels have been implicated in several severe neurological disorders including autosomal recessive primary microcephaly (30–32), intracranial hemorrhage (33) and Alzheimer’s disease (34). In addition to having a reduced content of DHA in the brain, Mfsd2a knockout mice exhibit neuronal loss in the cerebellum and hippocampus (27), anxiety, microcephaly (27) and a slow, progressive decline in retinal photoreceptor longevity (35). In addition to its transport function, MFSD2A is involved in the formation of the syncytiotrophoblast, the multinucleated layer of placenta that forms the maternal-fetal interface (36). Syncytiotrophoblast is formed by fusion of cytotrophoblasts, a mechanism driven by syncytins. MFSD2A serves as the receptor of one of the key proteins in these cell fusions, syncytin-2 (SYNC2) and reduced expression of MFSD2A, similar to that of SYNC2, has been reported in pre-eclampsia (37–39).

MFSD2A is believed to regulate BBB permeability by inhibiting caveolin-mediated transcytosis (26, 40–42). Indeed, studies in Mfsd2a knockout mice have demonstrated that while vascular density, branching, and pericyte coverage remain normal, endothelial transcytosis is significantly increased (26, 40–42). In dementia, changes in BBB endothelial cells and pericytes leads to BBB dysregulation and breakdown (43). During normal aging, reduced expression of MFSD2A coincides with a decline in pericyte coverage, underscoring the essential role of pericytes in preserving the integrity of the BBB (44). Furthermore, MFSD2A overexpression can rescue cognitive deficits in a rat model displaying permanent bilateral common carotid artery occlusion by decreasing BBB permeability (45). Interestingly, it has been demonstrated that reduced MFSD2A expression is observed at sites of BBB leakage, particularly where pericytes are lost, suggesting that pericytes may regulate MFSD2A via paracrine signalling (46). However, MFSD2A suppression alone is unlikely to account for the increase in BBB leakage observed in pericyte-deficient models. While Ben-Zvi et al. reported tracer leakage in Mfsd2a knockout embryos and postnatal animals (25), no such leakage was shown in two separate mouse studies by Mäe et al. and Wong et al.(46, 47). This discrepancy could stem from developmental differences, with adult animals potentially compensating through alternative mechanisms, or it could be that early transcytosis results were seen due to BBB immaturity rather than true transcytosis. Nevertheless, the loss of endothelial-pericyte crosstalk contributes to MFSD2A suppression, which may play a role in increased BBB permeability by promoting caveolar transcytosis and altering the expression of genes involved in processes such as neuroinflammation.

MFSD2A is a member of the Major Facilitator Superfamily

MFSD2A is a member of the Major Facilitator Superfamily (MFS) of proteins – one of the largest and most diverse families of membrane transporters found across all kingdoms of life (48). MFS transporters are typically comprised of 12 transmembrane helices (TMs) organized into two pseudo symmetric six-helix bundles: the N-domain (TMs 1–6) and the C-domain (TMs 7–12), connected by a long cytoplasmic loop (Fig. 1c; (49)). The N- and C-domains each consist of two 3-TM structurally inverted repeats, and pack against each other in a clam-shell fashion (Fig. 1c; (49)). They form a large solvent-exposed cavity in the center that is lined by the first TM helix from each of the structural repeats (TM1, TM4, TM7, and TM10). The amino acid residues that line this central cavity are crucial for substrate binding and defining substrate specificity (49).

MFS transporters can function as antiporters, symporters, or uniporters (50). Antiporters and symporters are secondary active transporters that move substrates against their concentration gradient by harnessing energy from the coupled movement of substrates like ions (such as Na⁺, K⁺, or H⁺) down their concentration gradient (50). Symporters move both their substrate and co-transported ions in the same direction across the membrane, while antiporters translocate substrates and co-transported ions in opposite directions. In contrast, uniporters enable the flux of substrates down their own concentration gradients and therefore do not require energy from the movement of ions (50). MFS transporters typically utilize a "rocker-switch" mechanism to mediate transport (Fig. 1d; (51)). According to this, the N- and C-domains undergo rigid-body movements around a centrally located substrate-binding site, alternating its exposure to either side of the membrane. This transport cycle involves a transition between an outward-facing and an inward-facing state (OFS and IFS, respectively), via a short-lived occluded state (OcS) where the substrate-binding site is inaccessible from either side of the membrane ( Fig. 1d; (40)).

Most MFS proteins – such as GlpT, LacY, MelB, GLUTs and many others (53–57) – transport small hydrophilic solutes, but there is a small subset of MFS transporters – including MFSD2A, MFSD2B, Spns1, Spns2, and LplT – that have evolved to transport amphipathic lysolipids that contain a charged headgroup and a hydrophobic tail, such as LPC-DHA (Fig. 1a; Supp. Tab 1; (58–62)). The mechanism by which these atypical MFS transporters enable lysolipid transport is not yet thoroughly characterized, but is thought to involve substrate entry and release via the membrane, and lysolipid flipping within the transporter (Fig. 1e; (29, 63–67)). Here, we review recent structural, functional, and computational studies that have provided substantial insights into the mechanism of MFSD2A-mediated transport.

Structures of MFSD2A

The structural characterization of MFSD2A has been technically challenging. MFSD2A is a 60-kDa membrane-embedded protein that bears no substantial soluble domains and is glycosylated at two sites. These attributes make the formation of crystal lattices required for X-ray crystallography difficult and, in tandem with the small size of the protein, hinders the use of single particle cryogenic-electron microscopy (cryo-EM; (68)). All structures of MFSD2A solved to date have relied on cryo-EM in combination with the use of fiducials such as fragment antibodies (Fabs), nanobodies, or physiologically relevant protein-binding partners to increase particle size and aid in their alignment (69). These structures have captured MFSD2A in various conformational states, offering critical insights into how this transporter works (Tab. 1).

Table 1 |.

Summary of MFSD2A protein structures.

PDB	EMDB	Organism	State	Resolution (Å)	Condition	Reference PMID	Year	% identity to human*
7MJS	EMD-23883	Gallus gallus	Inward facing	3	POPG/MSP1E3D1 nanodisc with FAB	34135507	2021	73
7N98	EMD-24252	Mus musculus	Outward facing	3.5	LMNG/GDN/CHS with scFv	34349262	2021	82
7OIX	EMD-12935	Homo sapiens	Outward facing occluded	3.6	GDN in complex with SYNC2	35710838	2022	100
8D2U 8D2V 8D2W 8D2X 8D2T 8D2S	EMD-27150 27151 27152 27153 27149 27148	Danio rerio	Inward facing	3.3 4.1 3.4 3.4 3.4 2.9	DDM with FAB	37156797	2023	64

^*Protein sequence alignments done with the amino acid sequences with following accession codes: H. sapiens (NCBI- NP_116182.2); G. gallus (NCBI XP_417826.5); M. musculus (NCBI- NP_083938.2); D. rerio (NCBI NP_001003570.1) POPG: 1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylglycerol, MSP: Membrane Scaffold Protein, FAB: Fragment antigen-binding, LMNG: Lauryl Maltose Neopentyl Glycol, GDN: glyco-diosgenin, CHS: Cholesteryl hemisuccinate, scFv: Single-chain Fragment variable, SYNC2: Syncytin-2, DDM: n-Dodecyl-β-D-Maltoside.

The first two structures of MFSD2A were reported in 2021 (Tab. 1). The initial breakthrough was a 3.0 Å structure of Gallus gallus MFSD2A that captured the transporter in the inward-facing state (IFS; (70)). Shortly thereafter, a 3.5 Å structure of Mus musculus MFSD2A showed the transporter in the outward-facing state (OFS; (71)). In 2022, a 3.6 Å structure of Homo sapiens MFSD2A in complex with Syncytin-2 (SYNC2) was published, where SYNC2 trapped the transporter in a partially occluded state (pOcS (39)). Later, several additional structures of Danio rerio MFSD2A in the IFS ranging from 2.9 – 4.1 Å were determined (72). For simplicity, throughout this review, amino acids are numbered according to the sequence of Homo sapiens MFSD2A unless otherwise noted. Numbering of all residues mentioned throughout the text is provided in Supplementary Table 2 for every species that has been structurally characterized.

Overall architecture of MFSD2A

Overall, MFSD2A adopts a classical MFS-fold, comprising 12 TM helices that are separated into two 6-helix bundles connected by a flexible cytoplasmic loop (Fig. 2a, b). Interestingly, MFSD2A has a unique feature not typically seen in the MFS fold which comprises an extracellular domain with two elongated, ordered extracellular loops (ELs), one between TM5 and TM6 (EL3) and the other between TM11 and TM12 (EL6). These loops bridge the N- and C-domains and contain a disulfide crosslink between C212 on EL3 and C460 on EL6, that pins these two domains together (Fig. 2a, b). This crosslink had not been observed in any other MFS transporter structures previously, but interestingly, these cysteine residues are conserved across MFSD2A orthologs as well as its close relative MFSD2B, which is also a lysophospholipid transporter (70). The IFS structure also revealed glycosylation at positions 218 and 227 (Fig. 2b; (70)).

Figure 2 |. The structures of MFSD2A in the OFS and IFS.

The structures of MFSD2A in the A, OFS and B, IFS in ribbon representation. The N-domain is colored in blue, while the C-domain is colored yellow. Glycosylation at two sites in the IFS is shown in stick representation. Inset highlights two cysteine residues that form an extracellular disulfide bond are shown in stick representation. Surface representations of C, the OFS (PDB: 7N98) and D, the IFS (PDB: 7MJS) colored by electrostatic potential (N: negative, P: positive), overlayed with the ribbon representation. The left panels show the transporter in the plane of the membrane, whereas the right panels show the transporter from the extracellular side (top-view) for the OFS structure (C), and the intracellular side (bottom-view) for the IFS structure (D). The lateral openings between TMs 5 and 8, and TMs 2 and 11 are indicated. E, Lipid-like density as surface representation in white, extracted from the density map of OFS. Residues mentioned throughout the manuscript are shown in stick representation, labeled and colored in blue or yellow with heteroatoms highlighted. F, Lipid-binding site in IFS shown in two views. Residues mentioned throughout the manuscript are colored in blue or yellow depending on whether they are located on the N- or C-domain, respectively, with heteroatoms highlighted. Bound lysolipid is shown in aquamarine stick representation overlayed with the cryo-EM density displayed as semi-transparent white surface. All amino acid numberings are according to human sequence.

MFSD2A has a large central cavity that is exposed to either the extracellular space in the OFS, or the intracellular space in the IFS (70, 71). In both states, the cavity is amphipathic, with residues varying from neutral to negatively charged, depending upon their location. However, the spatial arrangement of their charge distribution differs between the states: in the OFS, the innermost part of the cavity is negatively charged while the cavity entrance is more neutral (Fig. 2c; (70, 71)), whereas in the IFS, the N-terminal domain is more charged, and the C-terminal domain is more neutral (Fig. 2d; (70)). These electrostatics appear ideal for accommodating both the charged headgroup and hydrophobic tail of zwitterionic lysolipid substrates.

The outward-facing state and substrate entry

Both the OFS and IFS structures reveal two lateral openings that extend from the central cavity to the membrane: one between TMs 2 and 11, and the other between TMs 5 and 8 (Fig. 2c, d; (70, 71)). The central cavities of MFS transporters are typically less exposed to the membrane bilayer than that observed in MFSD2A, but most of these move soluble substrates (50). In contrast, lateral openings to the membrane are a common structural feature in lipid transporters from other families, such as ABC lipid transporters, RND transporters, and P4 ATPase transporters (73–75). Indeed, several observations have led to the hypothesis that these openings allow substrates to enter and leave their binding sites via the membrane rather than the extracellular milieu or cytoplasm to minimize energy costs associated with lipid transport (28, 52, 70–72). In the OFS structure (3.5 Å), Wood et al. observed lipid-like cryo-EM densities at both the TM2/TM11 and the TM5/TM8 lateral openings. At the TM5/TM8 opening, this density resembled an LPC molecule (Fig. 2e; (71)). Indeed, the authors were able to model one into it with its headgroup oriented towards the cavity’s center and its acyl chain parallel to the TMs, as if it were poised to dive headgroup-first into the central binding site (71). However, this molecule was not included in the final model out of caution, since this density was ambiguous and could potentially be attributed to other non-LPC lipids or detergents present in the sample (71). The density at the TM2/TM11 opening was lower in quality, and thus could not be reliably assigned (71). To further investigate substrate entry, Wood et al. demonstrated that mutating residues T198 and N331, which flank the TM5/TM8 opening, reduced transport by approximately 50%, whereas mutation of residues lining the TM2/TM11 opening had only a minor effect on activity (71). These data suggest that the TM5/TM8 lateral opening likely serves as the LPC entry pathway.

Follow-up studies have further investigated lipid entry into the MFSD2A central cavity via these routes. Chua et al. demonstrated that mutational analysis of residues lining both the TM5/TM8 and TM2/TM11 lateral openings are important for transport (28). However, mutational analysis alone cannot identify whether the effects of specific mutations on transport are mediated by compromised lipid entry or effects on conformational dynamics during transport. In a second study, molecular dynamics (MD) simulations demonstrated that lysolipid substrates can enter the central cavity via both lateral openings (52). However, only substrates that entered via the TM5/TM8 opening were able to become fully engulfed within the transporter (see section on the occluded state below). The opening between TM2/TM11 also allowed entrance to the non-substrate, di-acyl chain, membrane component 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC; (52, 70)). This suggests that the TM5/TM8 opening may play a more specific role in substrate entry, and that membrane lipids such as POPC may enter between the TM2/TM11 opening and perhaps provide a structural/regulatory element throughout the transport cycle (52, 70). Taken together, these studies support the notion that lysolipid substrates most likely enter the MFSD2A central cavity via the TM5/TM8 lateral opening. Further studies are required to confirm this and investigate if bilayer lipids that appear to enter via the TM2/TM11 opening regulate MFSD2A-mediate transport.

Na⁺-binding in the outward-facing state

MFSD2A-mediated transport is Na⁺-dependent (29). Several residues that have been shown to be important for Na⁺ binding in the bacterial melibiose transporter MelB – a prototypical, extensively studied, MFS member – are present in MFSD2A, suggesting that the Na⁺-binding site may be conserved between these two transporters (54). These residues are located in the N-domain and include – for MFSD2A – D93, D97, T96 and T159 (52, 70, 71). In the OFS structure of MFSD2A, the side-chain oxygens of these four residues form a pocket that is well-positioned to coordinate a Na⁺ (71). Interestingly, a non-protein density within this region of the cryo-EM map for the OFS could be observed but was not of high enough resolution to unambiguously assign to a Na⁺ (71). To validate this proposed Na⁺-binding site, Wood et al. performed all-atom MD simulations of MFSD2A with a Na⁺ placed within this extra density (71). In the majority of simulations, this Na⁺ shifted towards TM2 by approximately 2 Å and formed polar interactions with D93, D97, T96 and T159, with coordination distances and geometry typically observed for Na⁺ binding (71). It remained bound there for the remainder of each simulation which led to the designation of this as the Na1 site (71). Additionally, when Na⁺ was present in the extracellular solution, it diffused through the central cavity, interacting briefly with Y56, Q57, E190, T194, and/or E312 in no specific order before reaching the Na1 site (71).

In many simulations, a Na⁺ was also seen to bind to a second site (designate Na2) approximately 4 Å away from Na1, where it formed salt bridge interactions with D93 and E155, and was coordinated by water molecules (52). Na ions were seen to bind simultaneously at Na1 and Na2 on several occasions; however, given the Na2 site only provides two direct polar interactions with the protein (compared to six at Na1), these interactions were more transient in nature (52). Indeed, further mutagenesis analysis showed that substitutions in the Na1 site impaired or abolished transport, while those in the Na2 site had only a moderate effect (27). From these data, the authors concluded that Na1, which shows stable Na⁺ binding in MD simulations, is the key sodium-binding site in MFSD2A (27, 29, 52, 71).

Conformational transitions and substrate occlusion

In accordance with the rocker-switch mechanism of transport, the OFS and IFS of MFSD2A are connected by a transient intermediate occluded state (OcS) in which the substrate binding site cannot be accessed from either side of the membrane (39, 52). Capturing the transporter in a fully occluded state has not yet been achieved experimentally, but the structure of MFSD2A in complex with SYNC2 which traps the transporter in an intermediate state part-way between the OFS and the OcS (39), and MD simulations by Bergman et al. (52) have provided insights into how MFSD2A transitions between the OFS and OcS.

The MFSD2A-SYNC2 complex is critical for the formation of a specialized epithelial layer of placenta called syncytiotrophoblast, which serves as the direct contact site between the fetus and maternal blood making it essential for maternal-fed nutrient transport and placenta formation (76). SYNC2 acts as an allosteric inhibitor of MFSD2A and arrests the transporter in a state part-way between the outward-facing and occluded states. This allowed the authors to determine a structure of MFSD2A in a partially occluded state (pOcS; Fig. 3a; (39)). This pOcS structure of MFSD2A exhibits noteworthy differences from the outward-facing one: the extracellular edges of the N- and C-terminal domains are closer together, which partially occludes the central cavity from the extracellular milieu and shrinks the lateral openings between TM5/TM8 and TM2/TM11 (39). Additionally, in this state, the extracellular end of TM7 moves away from TM8 and TM10 to create a large lateral cavity within the C-terminal domain, which is lined by conserved hydrophobic residues and connected to a cluster of polar/charged residues (Y56, Q57, E155, Y151, D93, R90, K436, and S439) at the base of the central cavity through a constriction formed by F65, E312, L333, M337, and F399 (Fig. 3b; (39)). While this structure was determined in the absence of a substrate, the authors were able to dock LPC-18:3 into their model, with the charged headgroup bound to the cluster of polar/charged residues, the glycerol group close to E312 and G313, and the lipid tail occupying the large C-terminal domain cavity (39).

Figure 3 |. Mechanistic insights into MFSD2A occlusion.

A, The structure of the MFSD2A-SYNC2 complex in the plane of the membrane (PDB: 7OIX). The N- and C-domains are colored in blue and yellow, respectively, whereas SYNC2 is colored in white. The inset highlights two cysteine residues that form an extracellular disulfide bond shown in stick representation. B, Side-view of MFSD2A in the pOcS colored as in Figure 2 showing the amphipathic pocket in the C-domain. Residues mentioned throughout the manuscript are shown in stick representation. SYNC2 is omitted for visual clarity. C, Frame from molecular dynamics simulation trajectories (52) showing the lipid-binding site in the OcS, MFSD2A shown in ribbon representation, in blue (N-domain) and yellow (C-domain). Residues noted throughout the manuscript are highlighted in stick representation. LPC shown in stick representation and colored cyan whereas Na⁺ is shown as a grey sphere. D, Slab views of MFSD2A in different conformational states from the extracellular side highlighting the narrowing of the lateral openings in the pOcS and OcS. The protein is colored as in Figure 2 with the addition of semi-transparent surface caps for the N-domain in blue, and the C-domain in yellow.

Extensive atomistic multi-replicate MD simulations have also shed significant light on how MFSD2A transitions between the OFS and the OcS upon substrate binding (52). As described above, these simulations demonstrated that substrate enters MFSD2A from the outer leaflet of the lipid bilayer via both the TM5/TM8 and TM2/TM11 lateral openings (52). While most substrate insertion events were partial – i.e., only the headgroup would insert into the central region, while the lipid tail would remain in the outer leaflet of the membrane – there were two instances of LPC-18:1 insertion via the TM5/TM8 opening that resulted in full substrate embedding and occlusion in the central cavity (52). In both instances, the substrate headgroup interacted with E312 (52). These interactions were long-lasting and bridged by a Na⁺ that had diffused from the extracellular solution into this site. In this position, the hydrophobic tail of LPC-18:1 was stabilized by a network of aromatic residues including F329, F65, and F66 (Fig. 3c; (52)). This binding event was accompanied by a conformational shift of the transporter to the OcS, instigated by a narrowing of the central cavity and closure of the lateral entrances, which enabled the substrate to be completely occluded by protein (Fig. 3d; (52)). Notably, closure of the TM5/TM8 lateral entrance upon transition to the OcS was predominantly induced by an ~16° kinking of TM8 around conserved P345 (52). Interestingly, previous MD simulations (see section on the IFS below) had already shown that when the transporter is in the IFS, E312 can bind Na⁺ and coordinate substrate headgroups separately, and that the conservative mutation E312D significantly impairs transport (70). These observations could now be explained by Bergman et al., whose analysis revealed how E312 provides the precise length and chemistry required for the Na⁺-bridged interaction it forms with the substrate headgroup when the transporter is in the OcS (Fig. 3c; (52, 70)).

Substrate binding in the inward-facing state

The 3.0 Å resolution IFS structure determined by Cater et al. and their accompanying MD simulations offer key insights into how substrate is bound to MFSD2A in the IFS and released from the transporter directly into the inner leaflet of the membrane (70). In this IFS, the central cavity is open to the intracellular solution (Fig. 2b, d), with access to the extracellular side blocked by a cluster of hydrophobic residues (F65, F66, F329, and L333) that pack together at the domain interface (70). Interestingly, several of these residues (F65, F66, F329) were observed to interact with the substrate’s hydrophobic tail in MD simulations that captured the OcS (52). This suggests that as the transporter transitions from the OcS to the IFS, their engagement may be coupled to the movement of substrate to a distinct binding site in the IFS before being released from the transporter.

This distinct substrate binding site in the IFS is located in a deep pocket within the C-domain portion of the central cavity that is lined by hydrophobic residues: M187, V191, V311 (L311 in Gallus gallus), F315, I336, M337, I344, A391, V395, F399, L400, and W403 (Fig. 2f; (62)). While building the IFS structure of MFSD2A into their cryo-EM map, Cater et al. observed a density within this pocket that could not be attributed to the protein itself (70). Using native mass spectrometry, they identified that LPC-18:3 – a substrate of MFSD2A which was endogenously acquired from the cells used to express the protein for structural studies (Supp Tab. 1; (70)) – remained bound to purified MFSD2A. When modelled into this density, LPC-18:3 is positioned with its hydrophobic tail buried within the uppermost region of the pocket, and the charged headgroup extending out to the inner leaflet of the membrane through a lateral gateway between TMs 5, 8, and 10 that is formed at its most constricted site by residues M187, F399, and W403 (Fig. 2f; (70)). These gating residues, and those that line the hydrophobic pocket were shown to be functionally important by mutagenesis (70), further supporting the validity of this binding site, which appears to hold the lysolipid in a position primed for release from the transporter via the lateral gateway into the inner leaflet of the membrane.

Na⁺-dependent substrate release from the IFS

At the innermost region of the central cavity, adjacent to both this lysolipid binding site and the Na1 site -D93, D97, T159, and K436-, is a cluster of charged and polar residues: Q57, R90, Y130, E155 (Q150 in Gallus gallus), E190, T194, and E312; (70). There was no evidence for bound Na⁺ to the Na1 site in the IFS, perhaps due to limitations in resolution (3.0 Å; (70)). Several residues within this charged central region and Na1 site were shown to be relevant for substrate transport via mutagenesis coupled to cell-based functional assays (70). E312 seemed particularly important, and as discussed above, in the OcS, was subsequently shown to be critical for coordinating the phosphate group of the lysolipid headgroup via an Na⁺-bridged interaction (Fig. 3c; (52, 70)).

To assess if lysolipids move from the binding site observed in the IFS structure through the TM5/8 lateral gateway for release into the inner leaflet of the bilayer, Cater et al. performed atomistic ensemble MD simulations of MFSD2A in a POPC bilayer under a variety of different conditions (70). For simulations performed in the absence of substrate, Na⁺ was seen to enter the central cavity and interact with either E312 or D97 (70). In the apo state and when Na⁺ bound to E312, M187, F399 and W403 appear to move closer together compared to their position in the IFS substrate-bound structure to form a gate between TM5 and TM8, restricting access to the central cavity from the inner leaflet of the membrane. When Na⁺ bound to D97, the three gating residues move apart from each other to open a lateral gateway (70). When the POPC molecule adjacent to the open intracellular gate was replaced with lysolipids, these substrates were seen to insert deeply into and extensively sample the central cavity through their polar headgroup, while the hydrophobic tails remained in the membrane. Notably, this was only permitted for lysolipids (which have only one acyl chain), while POPC (which has two acyl chains) headgroups could only partially enter the central cavity but were unable to explore it to the same extent as MFSD2A substrates. In simulations initiated with LPC-18:3 positioned as observed in the IFS structure, the lysolipid headgroup explored the central cavity and interacted with residues E312 and R90 in the charged central region, while the tail remained engaged in long-lasting contacts with the surrounding hydrophobic residues (70).

Nguyen et al. subsequently reported additional cryo-EM structures of MFSD2A from Danio rerio in the IFS (72). Two of these (72) showed a presumed lysolipid tail bound in the same hydrophobic pocket observed by Cater et al. (70) but with its head group bound either to the central charged cavity – T435, R90, D93, E155 (Q149 in Danio rerio), V158 (Q152 in Danio rerio), K436, Y56, Q57 – in one structure (3.3 Å resolution), or to a second charged region further towards the cytoplasmic edge of the central cavity – F428 (Y425 in Danio rerio), E424, R186, M187 – in the other (4.2 Å resolution). This headgroup positioning within the central cavity appears to be distinct from that observed in the previously determined IFS structure (70). It is important to note that in the study by Nguyen et al., the identity of the molecule contributing to the extra non-protein density was not biochemically confirmed, and in some instances, this density was low resolution (72). Nevertheless, these additional MFSD2A structures suggest that the lysolipid headgroup likely explores the central cavity in a manner similar to that observed in the previously conducted MD simulations (70). The comparison between these two IFS structures showing the lysolipid headgroup still within the central cavity to that of the previously determined IFS structure where the lysolipid headgroup is protruding through the lateral intracellular gate is interesting (52, 70). It reveals that the conformational rearrangement of MFSD2A between these states is almost exclusively limited to the gating residues M187, F399 and W403, supporting the notion that substrate exits via this lateral gateway. These differences could possibly be attributed to the medium in which MFSD2A was reconstituted into for structure analysis: The Gallus gallus IFS structure was obtained with purified protein reconstituted into a POPC-containing nanodisc, whereas Danio rerio MFSD2A was purified and imaged in an n-Dodecyl-β-D-maltoside (DDM) micelle. This could in-turn suggest that bilayer lipids may play a role in promoting the opening of the lateral gateway for substrate release (70, 72).

Taken together, these data suggest that as MFSD2A transitions from the OcS to the IFS, Na⁺ moves from E312 to D97. Also, once the transporter enters the IFS, the tail of the substrate binds within a hydrophobic pocket in the C-domain, while its headgroup moves down, through the central cavity, before extending out between the lateral gate (70). Once Na⁺ binds to D97 this lateral gate opens, and the substrate is released into the inner leaflet of the membrane (52, 70).

"Trap-and-flip" mechanism of MFSD2A-mediated transport

The structures described above, in combination with MD simulations and functional analyses have enabled a fairly comprehensive molecular level understanding of how MFSD2A-mediated transport occurs (Fig. 4; (52, 70–72)). This transport mechanism is described as a "trap-and-flip" variation of the "rocker-switch" mechanism of transport (Fig. 4). This "trap-and-flip" mechanism involves substrate entering the central cavity from the membrane via a lateral gateway, bending to effectively flip and bind, headgroup-first, to a central binding site (Fig. 4). It then moves further down through the protein to a distinct site in the IFS before being released through another lateral gateway into the inner leaflet of the membrane (Fig. 4).

Figure 4 |. Schematic representation of the MFSD2A transport cycle.

The “trap-and-flip” mechanism utilized by MFSD2A begins with the schematic representation in the top left corner of the figure and proceeds clockwise. Representative colors and shapes in schematic are defined in the figure key. LPC-DHA and Na⁺ first come into proximity in the OFS MFSD2A. Na⁺ enters the MFSD2A central cavity from the extracellular solution while LPC-DHA enters from the outer-leaflet of the phospholipid bilayer through a lateral gate between TM5/TM8. LPC-DHA enters deep into the non-charged central cavity of the transporter and begins to flip with the charged head group facing down while Na⁺ forms a salt bridge with E312 and the phosphate of the lysolipid, simultaneously, the TM5/TM8 lateral gate closes and MFSD2A shifts from an OFS to an OcS. As LPC-DHA enters deeper into the hydrophobic pocket, Na⁺ transitions from the salt-bridge interaction at E312 to interact with D97, again forming a salt bridge with the phosphate group of the LPC-DHA headgroup. When Na⁺ moves from E312 to D97 it triggers the opening of the ionic lock and allows the intracellular gate at TM5/TM8 to open, in turn, allowing the transporter to shift from an OcS to an IFS. LPC-DHA shifts deeper again into the transporter, the tail of the substrate binds within a hydrophobic pocket in the C-domain, while its headgroup moves down, through the central cavity, before extending and exiting through the intracellular gate at TM5/TM8. LPC-DHA is thought to then exit MFSD2A and enter the inner-leaflet of the cell membrane. Na⁺ also exits MFSD2A into the cytoplasm via the central cavity as the lysolipid exits the transporter. MFSD2A then transitions back to an OFS, most likely via an OcS.

More specifically, the transport cycle begins with MFSD2A in the OFS (Fig. 4). In this conformation, the substrate enters the central cavity from the membrane’s outer leaflet through the lateral opening formed between TM5/TM8 while Na⁺ enters from the extracellular milieu (Fig. 4). Substrates may also enter MFSD2A via the TM2/TM11 lateral opening, and/or bilayer lipids could also enter and regulate the transport process via the same opening, but further studies are required to validate these hypotheses. Once substrate and Na⁺ are in the central region, the lateral openings narrow and the protein contracts to adopt an OcS (Fig. 4). Here, the phosphate of the substrate’s headgroup appears to be stabilized by a Na⁺-bridged interaction with E312, while the lipid tail is held in place by a network of aromatic residues, including F329, F65, and F66. Notably, in the absence of a substrate and when the transporter is in the OFS, Na⁺ interacts with D97 rather than E312. This suggests that the interaction between Na⁺ and E312 and the subsequent transition of the transporter to the OcS may only be possible in the presence of a substrate, consistent with the lysolipid itself contributing to Na⁺ coordination at this site (Fig. 4). Next – although the precise mechanism is yet to be demonstrated – as MFSD2A transitions from the OcS to the IFS, the Na⁺ shifts from E312 to D97. Upon transition to the IFS, engagement of F329, F65, F66 and L333 with each other displaces the lysolipid tail which transitions to a hydrophobic pocket in the C-domain. In turn, its headgroup moves down through the central cavity, interacting first with the charged central region before being passed through the gating residues M187, F399 and W403 (Fig. 4). Additionally, when Na⁺ binds to D97, the lateral gateway between TM5 and TM8 opens, allowing the lysolipid to be released into the hydrophobic environment of the inner leaflet of the membrane, and the transporter then returns to the OFS, ready for the next transport cycle (Fig. 4). Overall, substrate movement occurs through multistep transitions between functional states, facilitated by coordinated interactions between the substrate's head and tail at different binding sites. Key characteristics of the substrate, such as the earlier mentioned zwitterionic headgroup and 14-carbon tail length, are required to maintain these interactions.

Disease-causing mutations

To a certain extent, structural insights into MFSD2A have also advanced our understanding of disease-associated mutations. SNPs in MFSD2A have been linked to severe neurological conditions, predominantly microcephaly (30–32). These SNPs typically lead to loss-of-function or reduced activity, impairing the ability of MFSD2A to facilitate LPC-DHA transport across the BBB, which is crucial for normal brain development (Tab. 2; (31)). We can now map these mutations onto the structure, in some cases providing a mechanistic explanation for functional phenotype. For example, the homozygous missense mutation p.Thr159Met was identified in siblings with progressive microcephaly, spasticity, and brain imaging abnormalities, ultimately leading to early death (31). Functional studies showed that while the mutant protein reaches the plasma membrane, it has significantly impaired transport activity (31). T159 forms part of the Na1 binding site, and thus this loss of function is likely attributable to the disruption of Na⁺ binding (70, 71).

Table 2 |.

Disease-related MFSD2A single-nucleotide variations reported in patients

Mutation	Disease	Location of mutation/Potential impact on MFSD2A function	Expression/function	Reference PMID
T159M	microcephaly	Sodium binding site	Expression and membrane localization like wildtype. Impaired LPC transport.	26005868 (31)
S166L	microcephaly	Below the cavity at the interface of N and C domains	Expression and membrane localization like wildtype. Impaired LPC transport.	26005868 (31)
S339L	microcephaly	Affect the helical bend of TM8, prevent conformational changes	Expression and membrane localization like wildtype. Impaired LPC transport.	26005865 (30)
T198M	microcephaly	Interfere with substrate entry	Reduced expression, membrane localization like wild type. Comparable transport activity to wild type.	32572202 (32)
P402H	microcephaly	Destabilize the structure	Expression and membrane localization like wildtype. Loss of LPC transport.	32572202 (32) 30043326 (82)
P493L	microcephaly	Destabilize the structure	Reduced expression, membrane localization like wild type. Loss of LPC transport.	32572202 (32)
P164T	microcephaly	Destabilize the structure	Reduced expression, membrane localization like wild type. Comparable transport activity to wild type.	32572202 (32)
V250F	microcephaly	Affect stability or folding by interfering with interaction between TM3 and TM6	Reduced expression, membrane localization like wild type. Comparable transport activity to wild type.	32572202 (32)
R326H	microcephaly	Affect stability or folding by interfering with interaction between TM3 and TM6	Reduced expression, membrane localization like wild type. Comparable transport activity to wild type.	32572202 (32)
V81del	primary microcephaly	This residue is not present in the sequence of the MFSD2A isoform NP_116182.2	Not tested	33186761 (83)

Implications and future directions

Understanding how amphipathic molecules like LPC-DHA cross cellular barriers is a fundamental question in human physiology. The brain's high nutrient demands, coupled with the restrictive nature of the BBB, make this question particularly intriguing. Here, we have summarized our current understanding of the mechanisms underlying LPC-DHA transport via MFSD2A.

Despite notable advances, there are fascinating questions in basic physiology and biophysics on the LPC-DHA uptake pathway that remain open and warrant further investigation. For example, while LPC-DHA is carried by albumin in the bloodstream, the mechanism by which it is delivered to the membrane for uptake by MFSD2A is unclear. Does MFSD2A interact with carrier molecules in a manner similar to receptor-mediated transport systems like retinol? (77). Furthermore, while our insights into MFSD2A mediated transport have been highly informative, MFSD2A is expressed exclusively on the luminal side of brain endothelial cells and thus represents just the first step in the pathway that supplies DHA to the brain (26, 27). Following this, DHA needs to be transported intracellularly towards the abluminal membrane, and then of course across that membrane into the brain parenchyma (78). The intracellular carrier protein Fatty Acid Binding Protein 5 (FABP5) has been proposed to bind unesterified DHA (which is likely generated by an endothelial lipase activity on LPC-DHA) and localize it to the abluminal side of brain endothelial cells (79). There, it is thought to be transported by Fatty Acid Transport Protein 1 (FATP1; which is preferentially expressed at this abluminal membrane) across the abluminal membrane towards the brain parenchyma (78, 80). The precise mechanism and structural basis of this pathway have yet to be determined.

Our structural understanding of how MFSD2A binds and transports lysolipids could inform the design of drug carriers or nanoparticles mimicking MFSD2A substrates, thereby hijacking the transporter as a mechanism to cross the BBB. Furthermore, MFSD2A has been implicated in suppressing endothelial transcytosis, a mechanism that typically limits drug delivery across the BBB (26, 46, 81). Inhibiting MFSD2A could transiently disrupt this suppression, enhancing transcytosis and allowing therapeutics to penetrate the brain more effectively. These distinct strategies, inspired by MFSD2A’s structural and functional properties, represent promising avenues for overcoming the formidable challenge of drug delivery to the central nervous system.

To conclude, the structural and functional insights into LPC-DHA transport via MFSD2A reviewed here have not only deepened our understanding of nutrient regulation at the BBB but also reveal insights into the underlying biophysics of how this atypical MFS transporter moves an amphipathic molecule across a cell membrane. Finally, this work may hold potential for developing novel therapeutic strategies to cross the BBB and efficiently address various neurological conditions.

Supplementary Material

Supplementary Tables S1–S2: https://doi.org/10.6084/m9.figshare.28578401.v1