Characterization of yeast GWT1We purified the full-length yeast GWT1 using Sf9 cells (Supplementary Fig. 3a). Cryo-EM analysis revealed well-dispersed particles. However, small protein size and lack of distinct features in its soluble domains hindered single-particle processing (Supplementary Fig. 3b, c). To overcome this problem, we employed a green fluorescent protein (GFP)-fusion strategy, systematically fusing GFP to the loops between transmembrane helices of GWT1 (Fig. 1b).To screen functional GFP-fused GWT1 protein, we developed a fluorescence-based rescue assay in HEK293T cells (Supplementary Fig. 4a and Supplementary Fig. 15). Disruption of PIGW, the homolog of yeast GWT1 in human, led to the loss of GPI-AP CD59 expression on the cell membrane surface in HEK293T cells (Supplementary Fig. 4b). Overexpression of GWT1 restored CD59 surface localization in PIGW knockout cells (Supplementary Fig. 4b). This result supports the view of functional complementation between the yeast and human orthologs, consistent with a previously study13. Based on these in vivo assays, we identified a construct with a fusion GFP at position Y422 (fGWT1) that rescued CD59 expression in PIGW knockout cells (Fig. 1c, d), demonstrating fGWT1’s physiologically relevant activity.Structure determinationWe purified the fGWT1 protein that exhibits homogeneous behavior, making it suitable for further cryo-EM studies (Supplementary Fig. 5a). Initial micrographs revealed well-defined particles with clearly visible GFP moieties in the soluble domains (Supplementary Fig. 5b). Further stabilization with a specific GFP nanobody facilitated a high-quality data acquisition (Supplementary Fig. 5c–e). Following extensive 3D classification and refinement, 238,678 selected particles enabled the generation of a 3D reconstruction with an overall resolution of 3.47 Å (Supplementary Fig. 6a–e). By applying local refinement on the TM region with a mask, we achieved an improved resolution of 3.37 Å (Supplementary Fig. 6c).Cryo-EM density for the GWT1 polypeptide is well-resolved (residues 13-490, except for S101-K106), enabling the construction of an accurate model (Supplementary Fig. 7a-c). After most GWT1 amino acids were assigned to the cryo-EM density, an elongated density was observed within the GWT1 tunnel (Fig. 2a). Despite no substrates being added during purification, this density likely corresponds to an endogenous ligand. This apparent density perfectly accommodates a complete palmitoyl-CoA molecule, aligning well with functional studies demonstrating it as the primary substrate of GWT1 in yeast and mammals7,13 (Fig. 2a, b and Supplementary Fig. 7c).Fig. 2: Cryo-EM structure of GWT1 bound palmitoyl-CoA.a The density map of the monomeric GWT1 bound to palmitoyl-CoA, front views from within the plane of the membrane (left) and the intracellular side of the membrane (right). The approximate position of the ER membrane is indicated with light blue shading. The elongated density is shown in green. b The structure of the monomeric GWT1 in complex with palmitoyl-CoA monomer is illustrated in cartoon representations, offering perspectives from within the plane of the membrane (left) and the intracellular side of the membrane (right). The approximate position of the ER membrane is highlighted with light blue shading. c. The structural models of GWT1 bound to palmitoyl-CoA monomer are shown in the front view and top view. Helices are shown as cylinders and depicted in rainbow colors. Palmitoyl-CoA is encircled by a transmembrane spiral marked with a blue circle, and a red star indicates palmitoyl-CoA. d, The topology of palmitoyl-CoA-bound GWT1 is represented using colors like in panel (c).The overall structure of GWT1The overall structure of GWT1 measures approximately 71 Å × 63 Å × 34 Å (Fig. 2a). The structure of GWT1 unveiled a previously uncharacterized architecture. Spanning the ER membrane with 13 transmembrane helices (TM1-TM13), the enzyme possesses N and C termini facing the luminal and cytosolic sides, respectively (Fig. 2b, c). Notably, only minimal regions extend into the cytosol or ER lumen. Short lateral α-helices (LH1 to LH2) are mostly embedded in the ER membrane and oriented parallel to it. The bound palmitoyl-CoA is surrounded by TM4-7 and TM12. Therefore, these five highly conserved TMs form a “catalytic tunnel” (Fig. 2c, d). TM5 occupied a central position within the GWT1 protein structure. TM12 exhibited notably shorter lengths than other TMs. The N-terminus of GWT1 was positioned on the luminal side of the “catalytic tunnel”, resembling a lid oriented towards the center. TM1-3 and TM13 reside on one side of the catalytic tunnel, while TM8-11 occupy the opposite side, thereby constituting a scaffold domain enveloping the catalytic tunnel (Fig. 2c). Furthermore, TM1 to TM13 are predicted to be evolutionarily conserved among GWT1 proteins from diverse kingdoms. This architecture demonstrates that GWT1 adopts a unique fold distinct from known protein structures.Comparative structural analysis using Dali server identified heparan-α-glucosaminide N-acetyltransferase (HGSNAT) as the most significant hit to GWT1, despite a low sequence identity of less than 15%14 (Supplementary Fig. 8). Superimposition of GWT1 and HGSNAT yielded a root-mean-square-deviation (RMSD) value of 4.1 Å over backbones of 330 aligned residues, implying a potential evolutionary link (Supplementary Fig. 8a). This unexpected finding contrasts with subsequent structural hits, which lacked significant similarity (RMSD > 5) and belonged to functionally distinct protein families such as lipopolysaccharide transporters and toxin-antitoxin proteins (Supplementary Fig. 8b, c). Moreover, no structural homologs was identified within the MBOAT family. These findings indicate that GWT1 represents a structurally and functionally distinct class of acyltransferases.Palmitoyl-CoA binding pocket of GWT1The bound palmitoyl-CoA molecule adopts a distinctive L-shaped conformation positioned in the cytosolic tunnel with the CoA moiety facing the cytosol and the acyl chain lateral inserting into GWT1 (Fig. 3a-c). The orientation observed is in line with the known process of palmitoyl-CoA synthesis taking place at the cytosolic side of the ER membrane15. Palmitoyl-CoA interacts extensively with the TM4, TM5, TM6, TM7, and TM12 of GWT1, burying a substantial surface area of 956.6 Å2.Fig. 3: Overall structure of GWT1 and palmitoyl-CoA-binding site.a Three significant openings leading to the reaction chamber of palmitoyl-CoA include a cytosolic tunnel (highlighted in green), an ER-luminal tunnel (highlighted in blue), and the overall structure of palmitoyl-CoA (highlighted in orange). The electrostatic surface is illustrated. b The structural models of GWT1 bound to palmitoyl-CoA. Helices are represented as cylinders, and the palmitoyl-CoA molecule is depicted as a green ball and stick. The approximate position of the ER membrane is highlighted with light blue shading. c The cut-away view illustrates the binding pocket of the substrate palmitoyl-CoA within the reaction chamber. The regions corresponding to the “head”, “neck”, and “tail” of palmitoyl-CoA are highlighted with orange, purple, and blue frames, respectively. The palmitoyl-CoA molecule is shown as a ball and stick (depicted in green). d–f Amino acids located at the sites surrounding the head (highlighted in orange frame), neck (highlighted in light purple frame), and tail (highlighted in blue frame) of palmitoyl-CoA are shown as sticks with carbon atoms colored in green. g–i Sequence logos were generated using the WebLogo server (https://weblogo.berkeley.edu/logo.cgi) for a region encompassing conserved binding pocket of palmitoyl-CoA. Amino acids are color-coded based on their chemical properties: polar amino acids (G, S, T, Y, C, Q, N) are represented in green; basic amino acids (K, R, H) are depicted in blue; acidic amino acids (D, E) are shown in red; and hydrophobic amino acids (A, V, L, I, P, W, F, M) are illustrated in black.The cytosolic opening of the palmitoyl-CoA binding site is enriched in basic amino acids, resulting in a positively charged pocket (Fig. 3a, c). This unique composition creates a favorable pocket for binding the 3’, 5’-ADP head group of the palmitoyl-CoA. The adenine ring of palmitoyl-CoA is positioned into the cytosolic pocket of GWT1, stacking between K123 and I126. Hydrogen bonds are formed between the adenine ring of the palmitoyl-CoA and the side chain of T127, as well as with the main chain of K123 (Fig. 3d, Supplementary Fig. 9a). The phosphate groups form a hydrogen bond with R181. Notably, F171 in TM5, F238 and F239 in TM7, and F439 in TM12 contribute to stabilizing the “neck” region of the L-shaped palmitoyl-CoA through hydrophobic interactions (Fig. 3c, e, Supplementary Fig. 9a). The position of the palmitoyl-CoA thioester in the core cavity of GWT1 faces the lumen side. We propose that this cavity might function as the reaction center, where the acylation reaction could cleave the aliphatic chain from this thioester site and transfer it to the substrate GlcN-PI. The hydrophobic tail of palmitoyl-CoA inserts into a binding pocket predominantly constructed by conserved residues from TM4, TM5, and TM11, including T27, Y129, M133, L136, I141, L163, M164, L166, S170, Y400, F404 and Y408 (Fig. 3f–i, Supplementary Fig. 9a). Interestingly, the lipid tail-binding pocket of GWT1 appears wide and deep with minimal restrictions on acyl chain length, allowing it to bind diverse fatty acids such as myristoyl, palmitoyl, stearoyl, and arachidoyl (Supplementary Fig. 10). This observation was largely consistent with prior studies on the acyl-CoA specificity of GWT1 and its homologs7,13.To assess the functional significance of residues involved in palmitoyl-CoA binding, we performed rescue assays in HEK293T cells. Three classes of GWT1 variations were targeted: (1) 3’, 5’-ADP head group binding site, (2) neck interface, and (3) lipid tail interface. Additionally, two established loss-of-function mutants, D145A and K155A, served as controls16. Notably, all three classes of variations (including T127A, R130A, N175Q, V178A, R181A, N432A for the head group; T137A, F171A, F238A, R386A, F439A for neck interface; Y129A, L136A, I141A, L163A, M164A, S170A, Y400A, N401Q for the tail) had minimal impact on CD59 localization, suggesting limited effects by the single amino acid substitution of GWT1 (Supplementary Fig. 9b, c).Structural interpretation of luminal access cavity for GlcN-PIOur structure reveals a luminal cavity designated as the putative GlcN-PI binding pocket (Fig. 3a). Remarkably, additional density is observed within this cavity (Fig. 4a–e). The observed density resides near the palmitoyl-CoA thioester and potentially represents the inositol part of GlcN-PI. Consistently, molecular docking study indicates that GlcN-PI can bind to the proposed luminal access cavity (Supplementary Fig. 11a). This docked pose shows the O2’ atom of GlcN-PI in close proximity to the thioester bond of palmitoyl-CoA (Supplementary Fig. 11a). This spatial arrangement facilitates the transfer of the hydrophobic tail of palmitoyl-CoA to the substrate GlcN-PI. Due to its limited size, we cannot rule out the possibility that the observed density represents a product or detergent molecule.Fig. 4: The proposed luminal access cavity for GlcN-PI.a, d The cryo-EM map of GWT1 is displayed from both side (a) and top (d) perspectives, with the luminal access cavity highlighted by a blue dashed oval. b, e An additional density within the highly conserved luminal access cavity. c Molecular lipophilicity potential (MLP) of the luminal access cavity. The surfaces are colored by lipophilicity potential calculated by Chimera X. The putative entry route for GlcN-PI is indicated by a dashed blue path. f Comparison of catalytic sites of GWT1 and HGSNAT. The predicted catalytic sites of HGSNAT are N286 and H297, while the putative catalytic sites of GWT1 are D145 and K155. g Expressing fGWT1 in HEK293T△PIGW cells restores CD59 staining in the FACS assay. HEK293T△PIGW cells, where GPI-AP biosynthesis is blocked, serve as the staining control (gray). Loss-of-function mutants are highlighted in brilliant blue. Coordination residue mutants of the luminal access cavity are highlighted in purple.The luminal access cavity of GWT1, primarily formed by residues from TM4-7 and connecting loops 1-2 (CL1, CL2), is highly conserved among homologous proteins from Candida, Neurospora, and Aspergillus (Fig. 3g–i and Supplementary Fig. 2). The cavity location implies its critical role in catalysis (Fig. 4b, e). Interestingly, the cavity exhibits a lateral opening towards the hydrophobic membrane, potentially serving as the GlcN-PI entry pathway, resembling the entry mechanism observed in DGAT117,18 (Fig. 4c). The luminal surface of the cavity displays high hydrophilicity, possibly favoring the binding of GlcN-PI’s phosphoinositol and deacetylated glucosamine moieties (Fig. 4c). Structural alignment of GWT1 with the HGSNAT revealed that conserved residues I141 and K155 in GWT1 correspond to the proposed catalytic residues N286 and H297 in HGSNAT, occupying similar positions within the protein structure19,20,21 (Fig. 4f).To investigate the potential role of CL1 and CL2 residues surrounding the GWT1 luminal cavity, we performed fluorescence-based rescue assays. Consistent with previous reports, mutation of K155 to alanine resulted in near-complete abolition of activity16. The D145A mutation led to about a 50% decrease in activity, in agreement with its previously proposed critical role of D145 in GWT1 function (Fig. 4g). While several polar amino acids probably contribute to the architecture of the GlcN-PI binding pocket, individual mutations, including R152A, W159A, R216A, Y225A, E227A, E231A, Y232A, Y308A, N443A and N450A exhibited only moderate decrease in activity (Fig. 4g).Structural basis of GWT1 inhibition by manogepixConsidering the potential of GWT1 as a target for combating invasive fungal infections, we sought to determine the structure of GWT1 in complex with manogepix, the most potent GWT1 inhibitor identified to date (Supplementary Fig. 12). To promote formation of the inhibitor-enzyme complex, manogepix was incorporated with fGWT1 during protein expression and purification (Supplementary Fig. 5f).Subsequent cryo-EM analysis yielded a 3.55 Å resolution map of the GWT1-manogepix complex (Supplementary Fig. 12). The map revealed the absence of palmitoyl-CoA density within the catalytic tunnel, suggesting competitive inhibition by manogepix (Fig. 5a). Notably, a strong additional non-protein density was observed within the central cavity and matched the size and shape of manogepix (Fig. 5a, b). The clear density map allowed the accurate building of the manogepix-bound GWT1 structure (Supplementary Table 1). Our molecular docking and molecular dynamics simulations further validated the proposed binding mode (Pose 1, Supplementary Fig. 13). This binding mode outperforms an alternative mode by the ligand density map (Pose 2) in terms of both pose stability and binding interaction energy (Supplementary Fig. 13).Fig. 5: Cryo-EM structure of GWT1 in complex with manogepix.a–b Density map (a) and structure (b) of GWT1 in complex with manogepix. The density map and structure of GWT1 are colored blue, while the density map of manogepix is shown in yellow and its structure is in lilac. c Superposition of the palmitoyl-CoA-bound GWT1 complex and the manogepix-bound GWT1 complex. The palmitoyl-CoA-bound GWT1 complex and the manogepix-bound GWT1 complex are shown as cylinder with colors of pink and blue. d Sliced view of the molecular surface of fGWT1, highlighting the bound manogepix (blue) and palmitoyl-CoA (pink). e–f Close-up views of the manogepix (depicted in lilac), showing cryo-EM density (highlighted in yellow) and contacting GWT1 residues side chain (rendered in white).Structural comparison of GWT1 in complex with palmitoyl-CoA and manogepix revealed almost identical conformation, with an RMSD of only 0.3 Å over 464 Cα atoms (Fig. 5c). However, manogepix strategically wedged within the palmitoyl-CoA binding pocket, potentially occupying the catalytic center and interfering with lipid tail binding (Fig. 5d). The steric conflict between palmitoyl-CoA and manogepix suggests a competitive inhibition mechanism (Fig. 5d). By occupying the palmitoyl-CoA binding site, manogepix effectively prevents substrate loading into the catalytic center, thereby blocking the initial step of the enzymatic reaction. Meanwhile, the 2-aminopyridine group of manogepix overlaps with the thioester group of palmitoyl-CoA, potentially blocking the cleavage of the lipid tail and interfering with the GWT1 acyltransferase activity (Fig. 5d). The 2-aminopyridine group is positioned near the luminal access cavity, raising the possibility of steric clashes with the secondary substrate GlcN-PI (Fig. 5d). This observation suggests an additional layer of inhibition by restricting substrate access.Manogepix exhibited potent antifungal properties with a minimum inhibitory concentration (MIC) of 0.01361 µg/ml against yeast (Supplementary Fig. 14a). The structure of the GWT1-manogepix complex offers detailed insights into inhibitor recognition (Fig. 5e, f). Manogepix, like many potent GWT1 inhibitors, adopts an extended conformation and consists of four key components: 2-aminopyridine, isoxazole, para-xylene, and 2-pyridone groups (Supplementary Fig. 14c). This suggests a general binding mode for GWT1 inhibitors. Manogepix is stabilized within a central pocket by an extensive network of residues (Fig. 5e, f; Supplementary Fig. 14g). These residues, including T27, A30, Y129, G132, and others, primarily form hydrophobic interactions with the lipophilic manogepix molecule (logP = 3.2) through van der Waals contacts and some hydrogen bonds. Specifically, residues T137, I141, M164, G167, V168, F171, F238, F404, and F439 form a hydrophobic cavity surrounding the 2-aminopyridine group, while M133/L136 and L163/L166 flank the para-xylene moiety (Fig. 5e, f). Additionally, the 2-pyridone group engages with F404 through a CH-π interaction at a distance of 3.3 Å, with T27, A30, Y129, G132, and Y408 also contributing to its binding (Fig. 5e, f; Supplementary Fig. 14b). Interestingly, molecular lipophilicity potential analysis reveals a spatially distinct hydrophilic and lipophilic nature of the binding pocket, corresponding to the distribution of manogepix moieties (Supplementary Figs. 14d, e). The isoxazole group and the oxygen atom of the 2-pyridone group face the hydrophilic side of the pocket, suggesting polar contacts play a crucial role in manogepix binding. Indeed, Y400 forms a hydrogen bond (3.1 Å) with the 2-pyridone nitrogen. Meanwhile, S170 interacts with its oxygen via another hydrogen bond (Fig. 5e, f). These interaction analyses highlight the combined roles of hydrophobic and hydrogen-bond interactions in GWT1 inhibitor recognition.Structural basis for drug resistance of GWT1 mutationsThe elucidated structure of the GWT1-manogepix complex provides a foundation for examining GWT1 resistance mutations within invasive fungi. Notably, all previously identified GWT1 resistance mutations, including G132R (BIQ), G132W and F238C (G884), S170F (aminopyrifen), and V168A (manogepix), cluster around the manogepix binding pocket8,22 (Fig. 6a). This spatial arrangement strongly suggests a shared binding site for these structurally diverse GWT1 inhibitors. Functional assays validated these findings, with V168A and F238C mutations significantly reducing manogepix inhibitory activity (82.7% and 75.1% CD59 localization, respectively) compared to wild-type GWT1 (41%). Additionally, we identified manogepix resistance mutations in L136A, I141A, F171A, Y400A and Y408L suggesting their roles in reducing manogepix binding (Fig. 6b, c). Meanwhile, T137A and also exhibited weak drug resistance (Fig. 6c). Interestingly, we found that the alanine mutation on some substrate-interacting residues, including Y129, R386, N401, and F439, exhibited increased sensitivity to manogepix (Fig. 6c). Collectively, these findings validate the binding mode revealed by the GWT1-manogepix complex structure and deepen our knowledge of the drug resistance mechanisms. Moreover, these new drug resistance mutations offer valuable insights for clinical diagnosis.Fig. 6: Drug resistance mutation in GWT1.a Previously reported drug resistance sites for GWT1 are shown as spheres and colored in gray-blue. b All drug resistance sites, including those previously reported and newly identified, are shown. c Evaluating the sensitivity of manogepix by staining CD59 in the FACS assay, expressing fGWT1 in HEK293T△PIGW cells. Cells treated with manogepix (orange) or without manogepix (blue) were analyzed by flow cytometry using gating strategies, with negative controls marked in red. All experiments were conducted independently three times (n = 3).Structural basis of manogepix selectivity for GWT1Our elucidated GWT1 structure in complex with manogepix sheds light on the selectivity of the inhibitor on GWT1 against PIGW. Consistent with previous report, our cellular experiments confirm the negligible inhibitory effect of manogepix on PIGW at 10 μM, highlighting its selectivity10 (Fig. 6c). Despite sharing 31% sequence identity, superimposing the predicted PIGW AphlaFold structure onto GWT1 revealed high overall structural similarity with an RMSD value of 2.0 over 430 Cα atoms (Supplementary Fig. 2 and Supplementary Fig. 14f). Interestingly, our sequence alignment analysis revealed significant variation in amino acid composition surrounding the P3 and P4 pockets of GWT1 compared to PIGW, while the P1 and P2 pockets remained conserved (Supplementary Fig. 14g and Supplementary Fig. 16). This suggests comparable ligand interactions within the P1-P2 pockets for both enzymes. Notably, within the P3-P4 pocket of GWT1, key residues interacting with manogepix differed significantly from those in PIGW. For example, Y408 (GWT1) corresponds to L388 (PIGW), and similar differences were observed at positions 27, 30, 129, 133, 166, 170, 400 and 404 (Supplementary Fig. 2 and Supplementary Fig. 14g). Changing the GWT1 inhibitor binding pocket residue Y408 to L, resembling those of PIGW, resulted in significant resistance to manogepix, with a fivefold increase in MIC50 (Fig. 6c and Supplementary Fig. 14a). Collectively, our structural observations rationalized the selectivity of manogepix towards GWT1. It is primarily driven by the distinct amino acid composition surrounding the P3-P4 pocket. This knowledge would inspire further drug optimization aiming at enhancing inhibitors’ selectivity.Structural implication for GWT1 inhibitorsThe manogepix-bound GWT1 structure provides a structural basis for interpreting previous structure-activity relationship (SAR) data, illuminating design principles for potent GWT1 inhibitors. First, one study demonstrated that incorporation of the 2-aminopyridine moiety led to notable increases in antifungal efficacy against Candida albicans and Aspergillus fumigatus by approximately 4–8 fold, culminating in the synthesis of benzylthiophene amide derivative 15 with potent activity23. This enhancement likely arises from favorable interactions of the 2-aminopyridine group within the luminal access cavity P1 of GWT1 (Supplementary Fig. 14c, d). Second, substitution of the isobutanoxy group in Gepinacin with a methoxy group resulted in a shift from active to inactive24. Our structure suggests this could arise from weakened hydrophobic interactions between the inhibitor and the P4 cavity, highlighting the importance of the P4 cavity in the inhibitor design (Supplementary Fig. 14c, d). Third, incorporating nitrogen to form a 2-pyridone group significantly boosted the antifungal potency of LCUT-8 against Candida albicans, Candida auris, and Cryptococcus neoformans by 8-32 fold compared to LCUT-925. This is likely due to the introduction of an additional hydrogen bond with Y400 in the P4 pocket, as visualized in the GWT1-manogepix complex. Finally, the crucial hydrogen bond between S170 and the 2-pyridone oxygen in the GWT1-manogepix complex offers a reasonable explanation that the near-ubiquitous presence of an oxygen atom at the P4 position of potent GWT1 inhibitors (Supplementary Fig. 14c). Overall, these findings demonstrate the value of structural insights for guiding GWT1 inhibitor development by interpreting the molecular basis of functional group alterations and informing future optimization strategies.
