Structures of SemiSWEET transporters in two distinct conformations

Sugars produced by photosynthesis serve as key energy sources for humans. In both plants and animals, sugars are transported across cellular membranes as a means of distribution throughout the body^10,11. While sugar transporters are essential for translocation in plants, human transporters play critical roles in glucose homeostasis; mutations lead to diseases such as diabetes, glucose malabsorption, or epilepsy^10,11. Striking similarities exist among the sugar transporter proteins used by plants and animals. Animal and human genomes encode three major classes of sugar transporters: Major Facilitator Superfamily (MFS)-type transporters of the GLUT family (SLC2)¹¹, sodium-dependent glucose transporters of the SGLT family (SLC5)¹⁰ and the recently identified SWEET/SemiSWEET sugar transporters (SLC50)^1-3. Plant genomes contain homologs of GLUT transporters, specifically STP glucose/H⁺ symporters and SUT sucrose/H⁺ symporters¹², as well as SWEET transporters¹. Major breakthroughs in understanding how these transporters function resulted from solving atomic structures for the prototype of the MFS transporter family, lactose permease¹³, as well as GLUT¹⁴ and SGLT homologs¹⁵. Both MFS and SGLT transporters are built from 12 transmembrane helices (TM), but form fundamentally different structures: MFS transporters are composed of four structurally related triple helix bundles (THBs) arranged in antiparallel format; whereas SGLTs consist of two 5-TM bundles in antiparallel arrangement.

Until now, there was limited information on the structure of SWEETs and their bacterial SemiSWEET homologs^2,3. Plant SWEETs play crucial roles in intercellular transport and cellular secretion. Specific isoforms are key for cellular efflux as a first step for phloem loading⁴, pollen nutrition⁵ and nectar secretion⁶, and they also play key roles in pathogen susceptibility^2,4,16. The human genome contains a single SWEET homolog, which functions as a glucose transporter². The finding that the Ciona intestinalis SWEET is essential indicates that animal and human SWEETs play important roles in physiology¹⁷. SWEETs are unique in that eukaryotic isoforms are predicted to be heptahelical with an internal THB repeat²; while prokaryotic SemiSWEET contain only 3 TMs³.

To determine the structure and function of SemiSWEETs/SWEETs, two SemiSWEETs were crystallized in different states. The basic unit in both structures is a THB arranged as a 1-3-2 bundle, and two THBs are arranged in parallel to form the actual conduit. Six TMs are thus sufficient to form the pore. Moreover, the detection of two distinct states indicates that SemiSWEETs/SWEETs do not function as sugar channels, but rather as transporters that undergo rocking movements. We suggest that the eukaryotic heptahelical SWEETs form a similar structure in which a SemiSWEET-like dimer is fused via an inversion linker helix (TM4). We also show that pairs of Trp and Asn residues in the pore are essential for SemiSWEET and SWEET function.

The structure of a SemiSWEET from Vibrio sp. N418 (VsSemiSWEET; Extended Data Fig. 1) was determined at 1.7Å resolution from crystals grown in the lipid cubic phase (LCP) (Extended Data Table 1). The protomer of VsSemiSWEET contains 3 TMs and a nonconserved extra N-terminal amphipathic α-helix (AH). Within the protomer, TM3 is sandwiched between TM1 and TM2 and there is little direct contact between TM1 and TM2 (Fig. 1). This arrangement of TM helices shares similarities with the triple-helix repeat of MFS transporters¹⁸ although SemiSWEET and MFS belong to unrelated folds (Supplementary Discussion). The orientation of the protomer relative to the membrane was inferred using the ‘inside positive rule’¹⁹, which is in good agreement with the observation that the cytosolic C-terminus of Arabidopsis SWEET11 was phosphorylated in vivo²⁰.

In the VsSemiSWEET crystal, two molecules that are related by a two-fold axis perpendicular to the membrane tightly interact to form a dimer (Fig. 1b & Extended Data Fig. 2). At the dimer interface, TM1 of one protomer packs against TM2 of the other protomer, and vice versa. When viewed from the extracellular side, the backbone forms a basket-like structure, with an opening to the extracellular side, while the intracellular side was sealed by loops L1-2 (Fig. 1b). Several lines of evidence support the physiological relevance of dimer formation. First, the interface between the subunits is extensive, encompassing ~1,970Ų. Second, the dimer is formed in the lipid bilayer environment of the crystal (Extended Data Fig. 2). The majority of non-packing hydrophilic residues within the membrane points to the center of the dimer interface, compatible with a putative translocation route at the interface of subunits. Third, consistent with the structure, VsSemiSWEET dimerizes in solution and remains as a dimer during SDS-PAGE, indicating stable dimer assembly (Extended Data Fig. 3a). Cross-linked BjSemiSWEET³ products also migrate as dimers (Extended Data Fig. 3b). Together, our structural and biochemical observations strongly suggest formation of the transport pore of VsSemiSWEET by a dimer.

From bioinformatic analyses, e.g. as in PFAM, SemiSWEETs belong to the ‘MtN3 clan’ and are distantly related to the ‘PQ-loop’ family in this clan, defined by a conserved PQ-dipetide motif^7,21,22. Contradictory to the assumption that the PQ-motif is positioned in a loop region, the PQ-motif is embedded in the membrane and part of TM1 in VsSemiSWEET (Fig. 1d). The role of the conserved glutamine in SemiSWEETs is revealed by analyzing the dimer interface: the CO and NH group of the side chain amide group of Gln29 hydrogen bonds with NH and CO from two consecutive backbone amides right before TM2 of the other protomer, not only bringing the L1-2 loop to the dimer interface, but stabilizing the L1-2 loop conformation (Fig. 1d). The proline preceding the glutamine induces a kink in the helix, probably increases the flexibility of the transmembrane helix to allow formation of a glutamine backbone interaction, or potentially facilitates its disruption during the transport cycle.

The crystal structure indicates that the 3-TM protomer cannot form an enclosed compartment for substrate transport. Instead, there is a solvent-filled cavity between the two protomers at the central two-fold axis (Fig. 1e). The cavity transverses approximately halfway across the membrane and is completely separated from the lipid bilayer by the surrounding 6 TMs, but remains accessible from the extracellular side (Fig. 1c). This open cavity measures 9.2Å at the narrowest point and is sufficient to allow small molecules to freely diffuse in or out. Among amino acids lining the cavity, Trp59 and Asn75 (Fig. 1e) are most conserved across species, constituting the only two invariant residues in 66 SemiSWEET sequences analyzed. Both residues strategically sit at a similar level above the bottom of the open cavity. Their side chains surround the center of the cavity, form a putative binding pocket, and are most likely within range to interact with substrates given the size and geometry of the cavity. It is noteworthy that both residues are also highly conserved in 3-TM repeats of the SWEETs^2,4 and MPC2^8,9, supporting their functional importance.

To further investigate the transport mechanism, we focused on a SemiSWEET from Leptospira biflexa serovar Patoc (named LbSemiSWEET) that shares significant homology (44% identity and 63% similarity) with the known sugar transporter BjSemiSWEET³ (Extended Data Fig. 1). The structure of LbSemiSWEET was determined at 2.4Å resolution from crystals grown in LCP (Extended Data Table 1). In one asymmetric unit, two molecules tightly interact with each other, forming a dimer in the lipid bilayer (Fig. 2 & Extended Data Fig. 4). The dimer interface of LbSemiSWEET is highly similar to that of VsSemiSWEET, despite only modest sequence similarity (15% identity), strongly supporting the notion that the dimeric architecture is a common feature of SemiSWEETs.

At the interface of two protomers, LbSemiSWEET contains a large cavity right above its center (Fig. 2b). In contrast to VsSemiSWEET, the cavity of LbSemiSWEET is completely sealed from solvent. Near the extracellular side, Asp57 from one protomer forms hydrogen bonds with Tyr51 from the other protomer (Fig. 2c), shielding the cavity from the extracellular solution. This may explain the high conservation of Asp57 and Tyr51 across SemiSWEET, SWEET and MPC families^2,4,8,9: they form the cap on top of the cavity in the “occluded” state; cross-protomer interactions may facilitate the formation of this conformation. At the center of the cavity, there is a strong non-protein electron density, the identity of which cannot be unambiguously determined at this resolution (Fig. 2d). The flatshaped density is surrounded by Trp48 and Asn64 (equivalent to Trp59 and Asn75 in VsSemiSWEET) from both protomers. The antiparallel aromatic ring of Trp from each protomer is within 4Å from the putative substrate, and may interact with and stabilize the putative substrate in the pocket. The precise mode of the interaction is unclear, possibly involving hydrogen bonds and stacking interactions. The Asn side chains point to the putative substrate and are in close proximity, likely contributing to substrate binding. These structural observations implicate critical roles for Trp48 on TM2 and Asn64 on TM3 in substrate binding and translocation.

To assess the roles of Trp48 and Asn64 in sugar transport, we generated alanine substitution mutants in LbSemiSWEET and tested transport activity. In cell-based radiotracer uptake assays, the glucose uptake deficient E. coli strain²³ expressing wild-type LbSemiSWEET showed significantly higher glucose uptake compared to controls, consistent with the homology-based prediction that LbSemiSWEET transports sugar. When Trp48 or Asn64 were mutated to alanine, glucose uptake was markedly reduced to a level similar as in controls (Fig. 3a). We did not detect significant glucose uptake activity for VsSemiSWEET (data not shown). Furthermore, alanine substitutions of corresponding Trp and Asn in both THB repeats in Arabidopsis SWEET1 (Extended Data Fig.1), a glucose transporter, failed to complement the growth phenotype of a hexose uptake-deficient yeast strain (Fig. 3b). The mutations had no significant effect on plasma membrane localization of SWEET1 in yeast (Extended Fig. 5). These results demonstrate that Trp and Asn play critical roles in sugar transport in both SemiSWEETs and plant SWEETs and further support the notion that SemiSWEETs and SWEETs share the same basic architecture.

Alternating access is the prevailing model to explain substrate transport by transporters^24,25. Our structural observations support an alternating access mechanism in SemiSWEETs. In the crystal, VsSemiSWEET adopts an outward open conformation, while LbSemiSWEET is captured in an occluded state (Fig 1c&2b). Although LbSemiSWEET and VsSemiSWEET share modest sequence identity, the monomer of LbSemiSWEET superimposes well onto that of VsSemiSWEET, with a main-chain Root-mean-square deviation (RMSD) of 1.1Å over 66 aligned residues (Fig. 4a). By contrast, pronounced differences were observed between dimers of LbSemiSWEET and VsSemiSWEET (Fig. 1b, 2a, 4b). The dimer interface of VsSemiSWEET opens more towards the extracellular side compared with LbSemiSWEET and is ~10Å wider at the extracellular surface (Fig. 4b). This is achieved mainly through a ~10-degree rotation of the protomer around an axis near the intracellular membrane surface (Fig. 4a). To a lesser extent, slight bending of the transmembrane helices of LbSemiSWEET towards the center contributes to its more closed conformation. The conformational differences between VsSemiSWEET and LbSemiSWEET indicate a “rocker-switch” mechanism and bear interesting parallels to structurally unrelated transporter families, such as MFS proteins and ATP-binding cassette (ABC) transporters, in which rigid body rocking between transmembrane subdomains provides alternating access to the substrate^18,19,26,27. We propose that a similar rocking-type movement of two SemiSWEET subunits will result in two additional states: an “inward open” conformation and an “occluded empty” state to complete a transport cycle (Extended Data Fig 7a). It still remains to be determined whether the transport cycle is coupled to proton transfer or operates by facilitated diffusion. SWEETs show properties that are consistent with a facilitated diffusion or uniport, including pH-independence and low affinity^2,4. More detailed functional analysis informed by the structure may help to determine the exact transport mechanism.

Eukaryotic SWEETs consist of two SemiSWEET-like units fused via an inversion linker transmembrane helix. It was not clear whether SWEETs could be large enough to form a pore from a single heptahelical subunit or have to form a higher oligomer³. In light of the SemiSWEET structures with putative substrate binding sites at the center of the dimer interface, we propose that the two THBs in a single SWEET are able to form the transport route. Mutagenesis analysis of SWEET is compatible with this hypothesis by showing the functional conservation of key residues between SemiSWEET and SWEET proteins. SWEETs might form multibarrelled oligomers as part of a regulatory mechanism. Such regulation has been observed in GLUT glucose transporters²⁸, AMT1 ammonium transporters²⁹ and NRT1 nitrate transporters³⁰. Finally, since mitochondrial pyruvate carriers (MPC)^8,9 contain a related THB, our data is consistent with the observation that two copies of MPC are required to produce the functional transporter by likely forming a hetero-dimer.