Structural determination and overall structure of the G6PT monomer

To determine the atomic structure of human G6PT (UniProt: O43826), the protein was overexpressed in HEK293S GnTI^- cells and purified in lauryl maltose neopentyl glycol (LMNG) supplemented with cholesteryl hemisuccinate (CHS) (S1A Fig). Due to its relatively small size (~47 kDa) and potential conformational heterogeneity, however, initial single-particle cryo-EM studies on G6PT failed to yield high-resolution reconstructions. To overcome these hurdles, we implemented an optimized computational pipeline centered on five key aspects: (i) isolating of a homogeneous particle subset using seed-guided 2D classification and heterogeneous refinement in CryoSPARC [29]; (ii) increasing the initial and maximum resolutions for ab initio reconstruction in CryoSPARC to 15 Å and 6 Å, respectively, to capture finer structural details in the early stages; (iii) performing 3D auto-refinement in RELION 5.0 with Blush regularization, a deep-learning-based approach specifically designed to improve reconstructions of datasets with low signal-to-noise ratios [30]; (iv) systematically eliminating noncontributory particles from the final stacks with CryoSIEVE [31]; and (v) increasing the extraction box and corresponding mask sizes during the final 3D refinement steps in RELION to optimize signal capture and CTF correction. This integrated workflow ultimately yielded a final reconstruction with an overall resolution of 3.1 Å (S1 Fig and S1 Table). The resulting map confirms a monomeric configuration of human G6PT.

G6PT displays a classic MFS fold, with 12 transmembrane helices (TMs) arranged into two pseudosymmetric bundles: the N-terminal domain (NTD; TMs 1–6) and the C-terminal domain (CTD; TMs 7–12) (Fig 1A and 1B). These domains are connected by a long, partially disordered cytosolic loop (L6–7) between TM6 and TM7. The central segment of L6-7 (residues 199–213) is not visible in the density map, suggesting that G6PT lacks the well-ordered intracellular helical bundle characteristic of sugar transporters in the MFS superfamily [32,33] (S3 Fig). The N-terminus of L6-7 (residues 191–194) forms a short helical turn that packs against the cytosolic face of the NTD. Around the kink of the turn, the pathogenic mutation P191L substantially impairs G6P transport [25], pointing out the functional importance of this structural element in possibly stabilizing G6PT’s conformation.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Overall structures of human G6PT in monomeric and dimeric states.

(A) Cryo-EM density map of the G6PT monomer, viewed parallel to the membrane (left) and from the ER lumen (right). The NTD and CTD are colored blue and green, respectively. Nonprotein densities are shown in plum. (B) Overall structure (left) and topology diagram (right) of the G6PT monomer shown parallel to the membrane. The missing region of L6–7 is represented by dashed lines. (C) Cryo-EM density map (left) and atomic model (right) of the G6PT homodimer, viewed parallel to the membrane and from the ER lumen. Two protomers are colored yellow-green and khaki, with nonprotein densities highlighted in plum.

https://doi.org/10.1371/journal.pbio.3003731.g001

Our structure captures G6PT in an outward-open (ER lumen-open) conformation, in which a tightly packed NTD–CTD interface is formed at the cytosolic side, while the two domains diverge toward the ER lumen. This divergence creates V-shaped openings in the membrane between TM2 and TM11 on one side and TM5 and TM8 on the other side. Viewed perpendicular to the membrane, both crevices are filled with nonprotein densities, presumably lipids or detergents (Fig 1A).

The dimerization interface of G6PT

In addition to the monomeric form, our cryo-EM analysis resolved a G6PT homodimer at a resolution of 3.3 Å (S2 Fig and S1 Table). The dimer exhibits pseudo-C2 symmetry and each protomer adopts the same outward-open conformation seen in the monomeric reconstruction, allowing for the unambiguous docking of the monomeric model (Fig 1C). The G6PT dimeric assembly is characterized by a face-to-face topology similar to that reported for MCT2 [34], but distinct from the side-by-side arrangements observed in NRT1.1 or MCT8 [35,36]. Dimerization is mediated by symmetric interactions between TM2/TM4 of one protomer and TM11 of the opposing protomer. Notably, the dimer interface contains two prominent nonprotein densities likely representing detergent or lipid molecules, which appear to bridge and stabilize the two subunits.

The physiological relevance of this dimeric state remains elusive. Although the bacterial homolog GlpT functions as a monomer, oligomerization is a documented feature of many MFS transporters. In our structure, the dimer interface incorporates the sites of two GSD-Ib-causing mutations, S54R and S55R (S4A Fig). Previous studies indicate that these variants maintain expression levels comparable to the wild type but exhibit severely impaired transport activity [23]. The finding that the interface mutations result in functional defects rather than protein instability may suggest that the integrity of the dimer interface, or specific residues within it, is important for G6P transport.

However, caution is warranted when interpreting the physiological significance of this dimeric assembly. Recent literature suggests that G6PT oligomerization is highly sensitive to experimental conditions. Specifically, the use of detergents such as LMNG and CHS may facilitate the formation of nonnative oligomers [37,38]. While certain interface mutations have been shown to impair transport [39], they may disrupt the local packing of the transmembrane helices or hinder the conformational transitions required for the transport cycle.

Nevertheless, it remains to be elucidated whether such dimerization reflects a functional status under physiological conditions. Given that the conformation of each protomer in the dimer is identical to the monomeric structure, we have used the monomer for all subsequent mechanistic analyses.

Translocation pathway

The outward-open structure of G6PT reveals a large, solvent-accessible cavity open to the ER lumen. A defining feature of this translocation pathway is its electropositive surface located midway through the membrane (Fig 2A), which is ideal for attracting and accommodating negatively charged cargos, such as G6P and inorganic Pi. This positively charged environment is primarily contributed by R28 and K29 on TM1, K64 on TM2, and K240 on TM7 (Fig 2B left). Notably, the ²⁸RK motif is a hallmark of the OPA family (Figs 2C and S5. Two lines of evidence underscore the functional necessity of this motif: first, the pathogenic variants R28C and R28H have been reported to severely impair G6P uptake [23]; second, the equivalent ⁴⁵RK motif in the bacterial homolog GlpT is essential for glycerol-3-phosphate transport [40]. To validate these findings in human G6PT, we performed a G6P uptake assay on G6PT variants with single alanine substitutions at the R28 or K29 positions. Both mutants displayed significantly reduced transport activity (Fig 2D), suggesting that the ²⁸RK motif acts as a key structural determinant for substrate recognition.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Structure and functional characterization of the G6PT translocation pathway.

(A) Electrostatic surface of the human G6PT translocation pathway. The surface is colored according to the electrostatic potential (red, −10 kT/e; blue, 10 kT/e). (B) Detailed views of the translocation pathway with key basic residues (left) and aromatic residues (right) lining the cavity shown as sticks. (C) The sequence alignment of key residues lining the translocation pathway of human SLC37 members and the bacterial homolog GlpT. Residues are colored according to their level of conservation. (D) Transport activities of human G6PT with mutations in the central cavity, normalized to wild type (WT). Data are presented as means ± SD (n = 3 replicates). Statistical significance was determined using a one-way ANOVA with Dunnett’s multiple-comparison test. ****P < 0.0001. The data underlying this figure are available in S1 Data.

https://doi.org/10.1371/journal.pbio.3003731.g002

In this outward-open conformation, while the central cavity is accessible from the ER lumen, the translocation pathway is sealed from the cytoplasm by a gate composed of three aromatic residues, W118 on TM4, W138 on TM5, and H366 on TM11 (Fig 2B right). These residues likely play a dual role in stabilizing the cytoplasmic gate and facilitating sugar moiety binding through aromatic stacking or hydrophobic interactions. A key structural element here is the highly conserved ¹¹⁷GWP motif, which introduces a helical kink in TM4 that directs the bulky sidechain of W118 into the translocation pathway. The indispensable nature of this residue is demonstrated by the GSD-Ib-causing mutation W118R, which has been reported to abolish transport [23]. Our functional analysis confirms the importance of W118, as its alanine substitution reduces G6P transport activity by 40% (Fig 2D). Furthermore, while W138 and H366 are conserved in vertebrate G6PT homologs but divergent among human SLC37A1–A3 (Figs 2C and S6), mutagenesis analysis shows that both the W138A and H366A substitutions result in substantial activity loss (Fig 2D). Together, these findings highlight the coordinated roles of this aromatic cluster in sealing the cytosolic gate and potentially modulating substrate binding.

G6P recognition in G6PT

To investigate the mechanism of substrate recognition, we performed cryo-EM studies on G6PT samples pre-incubated with either 50 mM G6P or Pi. However, the resulting reconstructions resembled the apo outward-open configuration and showed no additional ligand density within the central cavity. We then generated G6P-bound G6PT models in both outward- and inward-open conformations using AlphaFold3 [41] and validated the predicted binding modes through molecular dynamics (MD) simulations.

The predicted outward-open G6P–G6PT complex closely aligns with our experimental apo structure, with the primary difference of a G6P molecule occupying the positively charged cavity (S7A Fig). Across three independent simulations over a 300-ns timescale, the G6P molecule settled into a stable pose at the base of the central cavity. The phosphate group is anchored by R28, Y60, and K64, while the glucose moiety is primarily coordinated by W118, W138, Y233, and H366 (Figs 3A, S7B, and S7C). Conversely, simulations initiated from the inward-open G6P-G6PT model exhibited greater deviations in binding modes, although the putative pocket remained structurally stable (S7D Fig). In two of the three simulations, the G6P phosphate group formed polar interactions with Y25, R28, and K240 (Fig 3B), whereas in the third simulation, it was engaged by R28 and Y60 (S7D Fig). Notably, while the phosphate group of G6P maintains specific contacts with G6PT, the glucose moiety occupies a broader region distal to the positively charged surface of the central cavity. This region appears to provide a hydrophobic environment that accommodates the sugar ring with significant conformational plasticity, as no conserved contacts were observed across the simulations.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Substrate recognition modes of G6PT revealed by MD simulations.

(A) Representative binding pose of G6P in the outward-open state. A structural snapshot highlights the interaction details between G6P and G6PT. G6P and interacting residues are shown as sticks; polar interactions (hydrogen bonds and salt bridges) are indicated by dashed lines. (B) Representative binding pose of G6P in the inward-open state. Details of the substrate-binding pocket are shown in the same way as in (A).

https://doi.org/10.1371/journal.pbio.3003731.g003

Our experimental analysis strongly corroborates these computational findings. Alanine substitutions of key residues predicted to engage G6P, including Y25, R28, and Y60, result in varying degrees of transport defects (Fig 2D). Given that R28 maintains stable interactions with the phosphate group across both major conformational states, and its mutation causes the most profound loss of G6P uptake, we propose that R28 serves as the primary determinant of phosphorylated sugar recognition during the transport cycle.

Comparative sequence analysis further clarifies the substrate specificity within the SLC37 family. While R28 and Y60 are conserved across the OPA family, other cavity-lining residues show significant divergence (Fig 2C). For instance, in SLC37A1–3, the equivalent position of Y25 is substituted with a histidine residue, and the basic residue K64 is replaced by uncharged residues. These variations likely alter the geometry and electrostatic landscape of the binding pocket, fine-tuning substrate specificity and affinity to suit the distinct physiological roles of each transporter. Obtaining experimental structures in substrate-bound states will be essential to validate these predicted binding modes and fully elucidate the mechanisms of substrate selectivity.

Conformational changes

The cryo-EM structure of human G6PT represents an outward-open conformation. This complements the inward-open state previously observed in the bacterial homolog GlpT (PDB:1PW4) and provides a structural template for elucidating the rocker-switch mechanism of the OPA family. To gain insights into the conformational transition, we compared our experimental structure with an AlphaFold-predicted (AF-predicted) model of G6PT, which adopts a canonical inward-open fold homologous to the GlpT structure.

Superposing the NTDs of these two conformational snapshots reveals that the CTD undergoes an ~40° rigid-body rotation relative to the NTD during the transition from the outward-open to the inward-open state (Fig 4A). This movement brings the luminal halves of TM7, TM8, and TM11 of the CTD in close proximity to the corresponding halves of TM1, TM2, and TM5 of the NTD, effectively sealing the central cavity from the ER lumen. An inter-domain salt bridge between K29 (TM1) and D245 (TM7), coordinated with a network of hydrophobic interactions, appears to contribute to the closure of the luminal gate. This salt bridge forms in the inward-open state but breaks in the outward-open structure (Fig 4B). The functional importance of this interaction is supported by our mutational analysis: alanine substitutions of either K29 or D245 reduce transport activity (Fig 2D). This gating mechanism is consistent with molecular dynamics studies of GlpT, which demonstrate that the dynamic formation and breakage of the equivalent salt bridge (K46–D274) facilitates conformational transitions [40]. The gating interface is further reinforced by a cluster of hydrophobic residues—M35, W246, F249, and W393—which likely work in concert with the electrostatic interaction to maintain the integrity of the luminal gate (Fig 4C).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Conformational changes of G6PT.

(A) Structural comparison of the AF-predicted inward-open model (pink) and the experimental outward-open structure (NTD, blue; CTD, green). The alignment is based on the superimposition of NTDs. (B) Interactions between K29 (TM1) and D245 (TM7) during the conformational transition. (C) A close-up view of the luminal gate in the inward-open model. Key residues are shown as sticks.

https://doi.org/10.1371/journal.pbio.3003731.g004

Missense mutations underlying GSD-Ib

To date, 57 missense mutations have been identified in GSD-Ib, most of which lead to abolished or reduced activity [22,23,25,26]. To understand how these mutations may affect G6PT function, we mapped them onto our structure (S4B Fig). Most mutations are found in TM regions with their sidechains pointing to the interior of the NTD or CTD, suggesting that they probably affect the protein’s folding or stability (S2 Table). For instance, introducing a charged aspartate at G20 on TM1 or at G339 on TM10 has been shown to reduce activity and decrease protein levels [23]. Several of the remaining mutations are located along the translocation pathway. Among these, R28 variants and W118R may directly interfere with substrate binding, while G68R and W138R probably disrupt conformational transitions by destabilizing the NTD–CTD interface.

Expression and purification of G6PT

The full-length gene encoding human G6PT was codon-optimized and cloned into the pEGBacMam vector with an N-terminal Flag tag and a C-terminal EGFP. Recombinant G6PT protein was expressed in HEK293S GnTI⁻ cells using the BacMam system. Briefly, upon reaching a density of 2 × 10⁶ cells/mL, cells were transduced with P₂ baculoviruses at a volume ratio of 1:50 (v/v). The cells were then harvested at 48 hours post-infection and stored at −80 °C until further use.

For purification, cell pellets were resuspended and homogenized in a solubilization buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 1% (w/v) lauryl maltose neopentyl glycol (LMNG, Anatrace), 0.1% (w/v) cholesteryl hemisuccinate (CHS, Anatrace), and a protease inhibitor cocktail. The lysate was clarified by centrifugation at 20,000 rpm for 45 min. The resulting supernatant was loaded onto anti-DYKDDDDK affinity resin and incubated for 2 hours at 4 °C. The resin was then washed with 20 column volumes of a wash buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 0.1% (w/v) LMNG, 0.01% (w/v) CHS. The protein was subsequently eluted with an elution buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 0.01% (w/v) LMNG, 0.001% (w/v) CHS, and 0.25 mg/mL Flag peptide. The eluate was concentrated using a 100-kDa cutoff concentrator and further purified by size-exclusion chromatography (SEC) on a Superdex 200 Increase 10/300 GL column (GE Healthcare) using SEC buffer [20 mM HEPES pH 7.5, 150 mM NaCl, 0.005% (w/v) LMNG, 0.0005% (w/v) CHS]. Peak fractions were pooled and stored at −80 °C.

Cryo-EM sample preparation and data acquisition

For cryo-EM grid preparation, 3 μL of the purified G6PT protein was applied to glow-discharged (15 W for 45 s) gold grids with an amorphous alloy film (UltraAuFoil Au 300-mesh R1.2/1.3). The grids were blotted for 3.5 s and then plunge-vitrified into liquid ethane at 4 °C under 100% humidity using a Vitrobot Mark IV (Thermo Scientific, USA). Cryo-EM data were acquired on a 300 kV Titan Krios Gi3 (Thermo Fisher Scientific) equipped with a Gatan K3 Summit detector. Automated data collection was performed using SerialEM 4.1 software. Raw movies were recorded at a nominal magnification of 105,000×, resulting in a pixel size of 0.425 Å. Each movie stack consisted of 50 frames collected over 2.9 s, with a total electron dose of approximately 50–54 e⁻/Ų. The defocus range was set between −1.5 and −2.0 μm.

Cryo-EM data processing and 3D reconstruction

For the G6PT monomer reconstruction, a dataset of 13,581 movie stacks was subjected to motion correction using MotionCorr2 in RELION 5.0 [44]. The micrographs were then imported into CryoSPARC [45] for patch contrast transfer function (CTF) estimation. Particles were picked and extracted for multiple rounds of 2D classification. To eliminate low-quality particles and resolve conformational heterogeneity, iterative rounds of ab initio reconstruction and heterogeneous refinement were performed. A subset of 1,691,620 particles was re-extracted in RELION at a pixel size of 0.85 Å for 3D classification. A 3D class containing 283,090 particles and clearly exhibiting 12 TMs was isolated for 3D auto-refinement. The resulting particle stack was refined further using CryoSIEVE to improve the map quality. After several rounds of 3D auto-refinement, post-processing, and polishing in RELION, the final monomeric reconstruction reached an overall resolution of 3.1 Å.

The G6PT dimer was processed using a similar computational workflow. A total of 4,833 movies were motion-corrected and subjected to Patch CTF estimation, as described above. Seed-guided 2D classification, ab initio reconstruction, and heterogeneous refinement were then performed in CryoSPARC. A subset of 1,044,270 particles was re-extracted in RELION at a pixel size of 0.85 Å, and two rounds of 3D classification reduced the particle number to 371,421. This number was then refined to 152,134 by CryoSIEVE. After CTF refinement and particle polishing, the final particle set was imported back into CryoSPARC for nonuniform and local refinement. This yielded a final map at a resolution of 3.3 Å.

Model building, refinement, and validation

The initial G6PT monomer model was generated by rigid-body fitting the NTD (residues 6–198) and CTD (residues 214–422) from an AlphaFold-predicted structure into the 3.1 Å cryo-EM map with ChimeraX [46]. The model was then manually adjusted in Coot [47] and refined in PHENIX [48]. For the G6PT homodimer, the refined monomeric coordinates were docked into the 3.3 Å dimeric density map. Characteristic nonprotein densities observed at the subunit interfaces were tentatively modeled as LMNG molecules (S2C Fig). The final models were validated using MolProbity within PHENIX to ensure optimal stereochemistry and map-to-model agreement.

The translocation pathway was calculated and analyzed using MOLE [49]. Density maps and structural figures were generated using ChimeraX. Sequence alignments were performed using ClustalW [50] and visualized in Jalview [51].

G6P uptake assay

The G6P transport activity was evaluated using a cell-based uptake assay as described in a recent protocol [42]. Full-length human G6PT (WT and mutants) and human G6PC1 (WT; UniProt No. P35575) were cloned into the pEGBacMam vector, and recombinant baculoviruses were produced using the BacMam system. Expi293F cells were co-infected with P₂ baculoviruses encoding G6PC1 and the respective G6PT (WT or mutant) and harvested at 24 hours post-infection for functional analysis.

For each experimental replicate, 8 × 10⁶ cells (or 4 mL of culture) were harvested and washed once with 2 mL of 1× Hank’s balanced salt solution (HBSS buffer) containing 2% FBS. The cells were permeabilized with 1 mL of 1× HBSS buffer containing 10 μM digitonin for 10 min at 37 °C. The treated cells were subsequently washed with 500 μL of 1× HBSS buffer and collected by centrifugation at 1,000 rpm for 3 min. The transport reaction was initiated by adding 1 mM [¹³C]-G6P to the 200 μL reaction system and then incubated at 37 °C for 10 min. The reaction was terminated by adding ice-cold 1× HBSS buffer, followed by two rapid wash steps to remove extracellular substrates. Intracellular metabolites were extracted with 400 μL of 80% methanol at −80 °C overnight. The resulting supernatant was analyzed by mass spectrometry to quantify the internalized [¹³C]-G6P. All transport activities were normalized to the total protein concentration determined by a BCA assay. Noninfected cells were processed in parallel to serve as a negative control.

G6P-G6PT modeling via AlphaFold3

Structural models of the human G6PT in complex with G6P were generated using AlphaFold3. Two distinct configurations were modeled: (1) G6PT with G6P alone and (2) G6PT with both G6P and Pi. The amino acid sequence of G6PT and the SMILES strings for the ligands were converted to JSON input files. Predictions were carried out on a GPU-accelerated computational node using varying random seeds to ensure broad sampling. The resulting models were manually evaluated based on conformational consistency and confidence metrics, specifically the interface predicted Template Modeling (ipTM) scores. Notably, all G6P–Pi–G6PT models converged to an inward-open state. In contrast, four out of five G6P–G6PT models adopted an outward-open conformation. The configurations with the highest confidence scores—the outward-open G6P-G6PT complex (ipTM = 0.86) and the inward-open G6P–Pi–G6PT complex (ipTM = 0.77)—were selected as starting structures for subsequent MD simulations. Pi was found to be dissociated from the transporter core in the G6P–Pi–G6PT models; therefore, it was omitted from the final model used for structural analysis and interpretation.

MD simulation

Simulations were initiated from high-confidence AlphaFold3 models of G6PT in both the G6P-bound outward-open and inward-open conformations. Each system was oriented using the PPM server [52] and embedded in a pre-equilibrated lipid bilayer using the CHARMM-GUI [53,54] membrane builder. To ensure sampling robustness, three independent replicates were performed for each conformational state.

All simulations were executed using the GROMACS package [55], employing the CHARMM36 force field for the protein [56], G6P, lipids, and ions, and the TIP3P model for water molecules [57]. Constant temperature (300 K) and pressure were maintained by a velocity-rescaling thermostat and a semi-isotropic C-rescale barostat, respectively. Long-range electrostatic interactions were calculated using the Particle Mesh Ewald (PME) method. Van der Waals interactions were computed with a 12 Å cutoff, using a force-switching function applied between 10 and 12 Å.

Following a multi-step equilibration protocol to relax lipids and solvents, three independent 300-ns production runs were performed under the NPT ensemble. A time step of 1 fs was used during equilibration, and 2 fs was used for production runs. Trajectory analysis was performed on the final 100 ns of each production simulation. Binding free energies for the G6P–G6PT interactions were calculated using the MM/GBSA method [58] based on the final 100 ns of the production trajectories.