The structure of phosphatidylinositol remodeling MBOAT7 reveals its catalytic mechanism and enables inhibitor identification

Wang, Kun; Lee, Chia-Wei; Sui, Xuewu; Kim, Siyoung; Wang, Shuhui; Higgs, Aidan B.; Baublis, Aaron J.; Voth, Gregory A.; Liao, Maofu; Walther, Tobias C.; Farese, Robert V.

doi:10.1038/s41467-023-38932-5

Download PDF

Article
Open access
Published: 14 June 2023

The structure of phosphatidylinositol remodeling MBOAT7 reveals its catalytic mechanism and enables inhibitor identification

Nature Communications volume 14, Article number: 3533 (2023) Cite this article

4042 Accesses
5 Citations
19 Altmetric
Metrics details

Subjects

Abstract

Cells remodel glycerophospholipid acyl chains via the Lands cycle to adjust membrane properties. Membrane-bound O-acyltransferase (MBOAT) 7 acylates lyso-phosphatidylinositol (lyso-PI) with arachidonyl-CoA. MBOAT7 mutations cause brain developmental disorders, and reduced expression is linked to fatty liver disease. In contrast, increased MBOAT7 expression is linked to hepatocellular and renal cancers. The mechanistic basis of MBOAT7 catalysis and substrate selectivity are unknown. Here, we report the structure and a model for the catalytic mechanism of human MBOAT7. Arachidonyl-CoA and lyso-PI access the catalytic center through a twisted tunnel from the cytosol and lumenal sides, respectively. N-terminal residues on the ER lumenal side determine phospholipid headgroup selectivity: swapping them between MBOATs 1, 5, and 7 converts enzyme specificity for different lyso-phospholipids. Finally, the MBOAT7 structure and virtual screening enabled identification of small-molecule inhibitors that may serve as lead compounds for pharmacologic development.

ROS-dependent S-palmitoylation activates cleaved and intact gasdermin D

Article 10 April 2024

Proteome-scale discovery of protein degradation and stabilization effectors

Article 20 March 2024

PROTAC targeted protein degraders: the past is prologue

Article 18 January 2022

Introduction

Phospholipids are synthesized in the endoplasmic reticulum (ER) via a series of enzymatic steps that sequentially add acyl chains to a glycerol backbone^1,2. Glycerol acyltransferase enzymes³ preferentially add a saturated fatty acid to the sn−1 position, and acylglycerol acyltransferase enzymes usually esterify an unsaturated fatty acid to the sn−2 position. To adjust membrane properties according to need and to store specific fatty acids as precursors for bioactive lipids, cells remodel the phospholipid acyl chains in enzymatic reactions known as the Lands cycle⁴. In this process, the sn−2 fatty acyl chain is removed by hydrolysis, and the lysophospholipid product is re-esterified with an acyl-CoA to form a phospholipid with a different acyl chain^4,5.

Re-acylation is carried out by lysophospholipid acyltransferase enzymes, including various membrane-bound O-acyltransferases (MBOAT), such as MBOATs 1, 2, 5, and 7 in humans⁵. MBOAT1 and MBOAT5 remodel phosphatidylserine (PS) and phosphatidylcholine (PC), respectively^6,7. MBOAT2 remodels phosphatidylethanolamine (PE) and phosphatidic acid (PA), and to a lesser extent PC and PS⁷. MBOAT7 uniquely remodels phosphatidylinositol (PI), catalyzing preferentially the esterification of arachidonyl CoA to lyso-PI^5,7 (Fig. 1a).

**Fig. 1: Cryo-EM structure of human MBOAT7.**

MBOATs form a large class of ER enzymes (11 in humans) that acylate small molecules, lipids, or proteins⁸. The structures of several MBOAT enzymes have been determined by cryo-electron microscopy (cryo-EM)^{9,10,11,12,13,14,15,16,17,18}. Some MBOATs, such as porcupine and hedgehog acyltransferase, acylate polypeptides in the ER lumen^10,11,18. Other MBOATs catalyze neutral lipid synthesis (e.g., DGAT1 and ACAT1/ACAT2) and act as multimers with a catalytic histidine residue that is buried deeply within the membrane^{12,13,14,15,16}. For these MBOATs, the acyl-CoA substrate accesses the catalytic center through a tunnel that reaches from the cytoplasmic to the lumenal side of the ER. A lateral gate allows for entrance of the acyl acceptor substrate and exit of the neutral lipid product into the membrane.

There are fewer insights into the molecular mechanisms for MBOATs involved in the Lands cycle. The cryo-EM structure of chicken MBOAT5 (cMBOAT5) reveals a dimer, with each protomer having a fatty acyl–binding tunnel and a T-shaped chamber for esterification of arachidonyl-CoA with lyso-PC⁹. It remains unclear, however, how different MBOATs recognize different lysophospholipid substrates.

We focused our studies on MBOAT7 because of its unique function in PI metabolism and because understanding its catalytic mechanism has important implications for human physiology and disease. MBOAT7 deficiency in humans leads to developmental disorders of the central nervous system, with intellectual disability, epilepsy and autism spectrum disorder^19,20,21. Deficiency of MBOAT7 function in liver is associated with increased lipogenesis in humans, likely by activating SREBP-1c and de novo lipogenesis²², and is an important genetic risk factor for developing non-alcoholic fatty liver disease and other liver diseases^{23,24,25,26,27,28}. Conversely, increased MBOAT7 expression is correlated with detrimental outcomes in cancers, such as hepatocellular carcinoma and clear cell renal carcinomas^29,30, suggesting that inhibitors of MBOAT7 may be useful therapeutic agents. Indeed, deletion of MBOAT7 inhibited tumor formation in clear cell carcinoma animal models²⁹.

Here we determined MBOAT7’s molecular structure by single-particle cryo-EM. Combining structural insights with molecular dynamics (MD) simulations and biochemical data, our findings suggest a model for catalysis and substrate specificity of phospholipid remodeling MBOATs. The MBOAT7 structure further enabled us to identify, by virtual screening, small-molecule lead compounds that specifically inhibit MBOAT7.

Results

Cryo-EM study of human MBOAT7 reveals characteristic structural features

To enable biochemical and structural analyses of human MBOAT7, we purified recombinantly expressed enzyme in detergent and reconstituted it into amphipol PMAL-C8 (Fig. 1b and Supplementary Fig. 1a). Purified MBOAT7 was similarly active in detergent or PMAL-C8 (Supplementary Fig. 1b). MBOAT7 in PMAL-C8 exhibited PI synthesis activity with a Vmax of ~300 pmol/min/μg protein, and Kms for arachidonyl-CoA and lyso-PI of 46 µM and 26 µM, respectively, which are similar to those of other MBOATs^10,12,31 (Supplementary Fig. 1c, d). MBOAT7 was most active with arachidonyl-CoA (C20:4) as the acyl donor, followed by other unsaturated acyl-CoAs, and strongly preferred lyso-PI as the acyl acceptor (Supplementary Fig. 1e, f).

We used single-particle cryo-EM to obtain a density map of MBOAT7 at an overall resolution of ~3.7 Å (Fig. 1c, Supplementary Fig. 2), allowing us to build a model of the entire protein except the C-terminal region (442–472), which was not visible and presumably disordered (Fig. 1d, Supplementary Fig. 2). Overall, the structure of MBOAT7 was similar to the model predicted by AlphaFold^32,33 with root-mean-squared distance (RMSD) ~ 1 Å on the backbones and ~2.2 Å on the entire protein (residues 1–441, Supplementary Fig. 3). MBOAT7 comprises 11 transmembrane (TM) helical segments, connected by short intervening helices (IH) and loops. The monomeric protein occupies a space of ~36 × 58 × 28 Å, embedded in the membrane (Fig. 1c), resembling the shape of a saddle.

The fold of MBOAT7 is similar to that of other MBOATs^{9,10,12,14,17,18}. Two TM bundles (TM1–5, TM9–11) tilt away from each other, generating space in the center for the catalytic chamber. The chamber between these bundles is sealed by two long, diagonal TM segments (TM6 and TM7) on one side and TM8 on the other. The diagonal helix TM6 is broken into two short helices (TM6a and TM6b) (Fig. 1d); in contrast, a corresponding helix in cMBOAT5 is continuous⁹ (Supplementary Fig. 4e). TM8, containing the putative catalytic His356 residue, resides in the cavity between the two TM bundles. TM8 exhibits weaker density than the other TMs, indicating flexibility or multiple conformations in the apo state of the enzyme (Supplementary Fig. 2f, TM8 density shown in two σ values).

Besides the “MBOAT core” (TM3–10, Supplementary Fig. 4a), other regions of MBOAT7 are distinct from other MBOAT structures (Supplementary Fig. 4b and c). For example, TM1-2 have features that are different from DGAT enzymes. MBOAT7’s N-terminal region contains only a short sequence before the first TM helix (TM1) and lacks sequences homologous to the longer N-terminal region of DGAT1, which mediates enzyme dimerization^12,13. Also, DGAT1 enzymes have a large lateral gate that opens into the plane of the membrane, which presumably provides access for diacylglycerol substrates and an exit path for triacylglycerol products^12,13. In contrast, TM helices 1 and 2 of MBOAT7 occlude the equivalent lateral gate region (Supplementary Fig. 4d).

A tunnel twists through MBOAT7, connecting the cytoplasmic and the ER lumenal sides of the enzyme, with the active site located in the middle of this tunnel (Fig. 1e). This suggests a model similar to those proposed for other MBOATs^9,10,11, in which acyl-CoA accesses the active site from the cytosolic side of the membrane and the acyl-acceptor, lyso-PI, enters the enzyme from the lumenal side. In MBOAT7, the lateral opening of the tunnel has an extended channel that branches off to the lumenal leaflet of the ER membrane.

Our model predicts that the N- and C-termini of MBOAT7 are on opposite sides of the ER membrane, with the N-terminus in the ER lumen and the C-terminus in the cytoplasm. To experimentally test this model, we performed fluorescence protease protection assays³⁴ with MBOAT7 fused to EGFP at either the N- or C-terminus (Fig. 1f). Controls included expressed mEmerald-Sec61β (fluorescent protein cytosol-exposed) and mCherry-ER3 (a lumenal protein). We used low concentrations of digitonin to selectively permeabilize the plasma membrane, but not the ER membranes. Upon proteinase K treatment, the fluorescent signal of mEmerald-Sec61β was rapidly attenuated, whereas the signal of mCherry-ER3 was protected (Fig. 1f). For MBOAT7, with proteinase K treatment, signal was lost for the C-terminally, but not the N-terminally tagged protein, consistent with the topology predicted by the cryo-EM model.

MBOAT7 model reveals how substrates access the enzyme active site

We were unable to obtain a high-resolution cryo-EM structure of MBOAT7 bound to either substrate, possibly due to increased sample heterogeneity. However, we were able to utilize the structure of cMBOAT5 bound to arachidonyl-CoA (PDB 7F40) to model MBOAT7 substrate binding⁹. Docking analyses revealed that arachidonyl-CoA fits well into the cytosolic part of the enzyme tunnel (Fig. 2 and Supplementary Fig. 5), positioning the thioester bond of the substrate near the catalytic His356 and Asn321 (Fig. 2a–c). Asn321 forms a hydrogen bond with the thioester carboxyl bond, implying it stabilizes the carbonyl group during the nucleophilic attack of His356 on the thioester bond. The adenine ring of the CoA moiety protrudes into the cytosol without strong interactions with the protein, similar to the acyl-CoA binding mode of other MBOAT enzymes^10,13,18. The ribose group of CoA forms hydrogen bonds with Tyr333 (Fig. 2c), and the phosphate groups of CoA form salt bridges with Arg318. The pantetheine group contacts Gln325, Tyr341, Met349 and Ser352, and mutation of Gln325, Tyr341, Ser352 to alanine reduced the acylation activity of MBOAT7 (Fig. 2d). Arg339 faces the entrance of the cytosolic tunnel but does not directly interact with the modeled arachidonyl-CoA. Indeed, mutation of this arginine to alanine did not affect the activity of MBOAT7 (Fig. 2d).

**Fig. 2: Arachidonyl-CoA accesses the MBOAT7 catalytic chamber via a tunnel from the cytosolic leaflet of the membrane.**

MBOAT7 strongly prefers arachidonyl-CoA as a substrate over other unsaturated fatty acyl CoAs (Supplementary Fig. 1e). The structure provides insight into this preference. The tunnel through MBOAT7 connects to a large cavity next to the catalytic residues, lined with bulky hydrophobic residues, including Phe241, Phe453 and Trp320 (the fatty acid–binding pocket in Fig. 2b, c, colored in green). Trp320 appears to form a hydrophobic gate to exclude water from the catalytic center, and mutating Trp320 to alanine strongly reduced MBOAT7 activity (Fig. 2d). The long, kinked acyl chain of arachidonyl-CoA fits well into this side cavity, interacting with Val313, Tyr407, Met410, Lys403 and Ile369 (Fig. 2c, residues colored in green). In contrast, straight and saturated acyl-CoAs do not fit this bent cavity well, likely explaining the substrate preference of MBOAT7 for more flexible unsaturated acyl-CoAs.

MBOAT7’s tunnel provides a route for lyso-PI to enter the catalytic chamber from the lumenal side of the ER membrane. To gain insight into how specifically lyso-PI binds MBOAT7, we modeled lyso-PI (C18:0) binding to MBOAT7 based on analogy to lyso-PC binding to cMBOAT5 (PDB 7F3X), and analyzed this with all atom-MD simulations. In the resultant energy-minimized structure (Fig. 3 and Supplementary Fig. 6), lyso-PI bound MBOAT7 within the ER lumenal opening of the tunnel. The inositol headgroup occupied the solvent-accessible part of the tunnel. The sn-1 acyl chain extended into a hydrophobic side-channel formed by TM1, TM6b, and TM7 (Fig. 3a colored in cyan) that opens into the membrane. This channel is lined primarily with hydrophobic residues (e.g., Leu9, Phe205, Val247, Phe245) that align along the acyl chain of lyso-PI (Fig. 3d). In the apo-MBOAT7 structure, the width of the hydrophobic channel is narrow (~7 Å in diameter) (Fig. 3b, c, hydrophobic channel is highlighted by cyan lines), suggesting that it cannot fit more than one acyl chain. This may explain how MBOAT7 preferentially binds lyso-PI with one acyl chain, but not the vasty more abundant PI with two acyl chains, which could result in product inhibition of the enzyme.

**Fig. 3: Lyso-PI accesses the MBOAT7 catalytic center through an ER lumenal opening.**

To examine the stability of lyso-PI binding to MBOAT7 and to discover alternative binding modes, we repeated 500-ns MD simulations (Supplementary Fig. 7). Lyso-PI did not leave the binding pocket in each of three simulations. The average RMSDs of MBOAT7 and lyso-PI in the simulations were ~3.5 Å and 7 Å, respectively (Supplementary Fig. 7a), indicating that lyso-PI is stably bound in the pocket. Our simulations also revealed two lyso-PI binding states that appear transiently stable with minor differences to the initial state (Supplementary Fig. 7b, c).

In our structures, the MBOAT7 catalytic residues His356 and Asn321 are close to the lyso-PI glycerol backbone (Fig. 3d). His356 can form hydrogen bonds with the hydroxyl group at the sn-2 position of the lyso-PI glycerol backbone and the backbone carbonyl group of Ser352 (Fig. 3d). The His356-Ser352 interaction likely increases the electronegativity of the imidazole Nε atom to facilitate the deprotonation of the sn-2 hydroxyl group of lyso-PI, as found in other MBOAT members^11,12. Asn321 may also form a hydrogen bond with the sn-2 hydroxyl group of lyso-PI. Since it also hydrogen bonds with the arachidonyl-CoA thioester carbonyl group (Fig. 2c), Asn321 may assist in the acyltransferase reaction by stabilizing the two substrates close to catalytic His356.

A discontinuous TM6 helix is important for MBOAT7 activity

Compared with other MBOAT structures, our cryo-EM structure of MBOAT7 is distinct, inasmuch as TM6 is divided into two short helices (Fig. 1d and Supplementary Fig. 4e) linked by two evolutionarily conserved proline residues that are not found in MBOATs 1, 2, or 5 (Fig. 3f and Supplementary Fig. 8). Mutation of the MBOAT7 prolines resulted in a continuous helix in AlphaFold2 predictions and markedly reduced lyso-PI acylation activity (Fig. 3g), suggesting that the discontinuous TM6 is required for optimal MBOAT7 activity.

We obtained a cryo-EM density map of MBOAT7 with lyso-PI and 18:1-ether CoA, a non-hydrolysable acyl-CoA analog (Supplementary Fig. 9). This map provided adequate resolution ( ~ 6 Å) to determine the overall architecture and TM arrangement of MBOAT7 with substrates but not the molecular detail. Docking this map with the apo-structure of MBOAT7 identified conformational changes in TM6: in the map obtained in the presence of substrates, TM6b moved away from the lateral portal, and TM6a moved down and became more horizontal. As a result, the lateral hydrophobic channel enlarged, as compared with the apo-structure (Supplementary Fig. 10). We hypothesize that this conformational change facilitates the release of the larger PI product with two acyl chains from the lateral portal, providing a possible explanation for how TM6 mutations reduced the enzyme activity.

The region between TM4 and TM5 confers lysophospholipid substrate-specificity on MBOATs

The energy minimized structure from MD simulation suggests a unique recognition mode for the inositol headgroup. The inositol ring forms hydrogen bonds with two charged residues, Arg87 and Asn102, and the backbone carbonyl groups of Pro99 and His356 (Fig. 3d). Two charged residues, Arg244 and Tyr8, are close to the phosphate and inositol groups and can transiently interacted with these groups during MD simulations (Supplementary Fig. 7c). Mutation of these residues decreased MBOAT7 catalytic activity (Fig. 3e). Besides Arg244, the phosphate group of lyso-PI is surrounded by a group of aromatic residues, including Tyr362, Phe414 and Phe245 and is stabilized by a combination of hydrogen bond and anion-pi interactions.

How MBOATs 1, 2, 5, and 7 specifically acylate lyso-PS, lyso-PE, lyso-PC, or lyso-PI is unclear. To investigate this, we docked three phospholipid headgroups as ligands into their corresponding MBOAT enzyme structures (obtained from this study or from AlphaFold predictions^32,33). Each headgroup fits best into a pocket created by the converging ends of two highly variable TM bundles, containing parts of TM4, TM5, and the connecting loop (Fig. 4a and b), and a more conserved region spanning partial TM9 and TM10 (Fig. 4b). Important residues for recognizing the different headgroups are unique between different MBOATs and almost all clustered on the TM4 and TM5 variable regions (Fig. 4a–c). Specifically, the amino and carboxyl groups in the serine, representing the headgroup of lyso-PS, form salt bridges with three unique charged residues within the variable region in MBOAT1, including His122, Arg125 and Asp137. Choline, representing the headgroup of lyso-PC, is bound by Tyr129 and Cys146 in MBOAT5, a pattern that is similar to the interaction between cMBOAT5 and lyso-PC⁹. For MBOAT7, docking recapitulated Arg87 and Asn102 hydrogen bonding with the inositol headgroup (Fig. 3). Other important residues for headgroup recognition in TM9 and 10 are highly conserved and almost identical in the three MBOAT proteins analyzed (Fig. 4b, c).

**Fig. 4: A variable region of MBOAT7 TM segments 4 and 5 determines the selectivity towards lysophospholipid substrates.**

To test whether variable regions of TM4 and TM5 determine substrate specificity, we swapped these regions between MBOATs 1, 5 and 7, and tested the activities of the chimeric MBOAT proteins (Fig. 4d, e). Upon replacing the variable regions of MBOAT1 and MBOAT5 with that of MBOAT7, these mutant enzymes gained acyltransferase activity towards lyso-PI (Fig. 4d). Similarly, replacing this region of MBOAT1 or MBOAT7 with that of MBOAT5 yielded enzymes that gained acyltransferase activity towards lyso-PC (Fig. 4d). Thus, these variable regions of MBOATs (corresponding to TM4 and TM5 of MBOAT7) constitute a “specificity-determining region” for different lyso-phospholipids.

A structure-based screen identified MBOAT7 inhibitors

Increased MBOAT7 expression correlates with detrimental outcomes in some cancers, including hepatocellular carcinoma and renal clear cell carcinomas, suggesting a dependency on the enzyme^29,30. Moreover, genetic deletion of MBOAT7 in clear cell renal carcinoma-derived cells induces cell-cycle arrest and prevents the cells from forming tumors in vivo²⁹, suggesting MBOAT7-specific inhibitors may be valuable. To identify such inhibitors, we used VirtualFlow³⁵, a structure-based virtual screening platform, to screen ~5.7 million commercially available compounds (Fig. 5a). We selected candidate compounds, based on differences in free energy–docking scores, molecular scaffolds, and different substructures (Supplementary Fig. 11). Our selection criteria yielded 12 candidates that covered a range of chemical space (Supplementary Fig. 11). We then tested these compounds for inhibition of purified MBOAT7 by measuring activity at two concentrations of compounds, 50 and 5 µM (Fig. 5b). Two compounds, which we term Sevenin-1 and Sevenin-2, inhibited MBOAT7 activity. The IC50s for Sevenin-1 and Sevenin-2 were 0.87 µM [0.56–1.22 µM, 95% confidence interval (CI)] and 5.60 µM (4.61–6.82 µM, 95% CI), respectively (Fig. 5c). Notably, ATR-101³⁶, a selective ACAT1 inhibitor with multiple reported off-target side effects³⁷, also inhibited MBOAT7 (Fig. 5b).

**Fig. 5: Structure-based discovery of MBOAT7 inhibitors.**

To screen for the specificity of Sevenin-1 and Sevenin-2, we measured their effects on purified human DGAT1¹³. Two DGAT1 inhibitors, T863 and DGAT1-IN-1, inhibited DGAT1 activity at 10 µM, whereas Sevenin-1 and −2 showed no inhibition even at 50 µM, suggesting that these two compounds may be MBOAT7-specific (Fig. 5d). According to docking results, both Sevenin-1 and −2 bind to MBOAT7 via occupying the acyl-CoA binding tunnel with the extended side pocket (Fig. 5e, f).

Discussion

Here, we provide models for the structure of monomeric MBOAT7 and its catalytic mechanism to synthesize PI specifically from arachidonyl-CoA and lyso-PI substrates, and we identify chemical inhibitors of this reaction (Fig. 6). Because it was not possible to obtain a high-resolution structure of substate-bound MBOAT7, we elucidated how MBOAT7 recognizes its two substrates through a combination of molecular modeling, simulations, and biochemical experiments. In this model, the arachidonyl-CoA substrate enters the enzyme tunnel from cytoplasmic leaflet of the ER. An extended lateral side pocket can accommodate the flexible acyl chain of unsaturated acyl-CoAs (Fig. 6a), but is not optimal for a straight, saturated acyl chains. Lyso-PI, enters the tunnel from the lumenal leaflet with a side channel harboring its single acyl chain and positions the sn−2 hydroxyl group of the glycerol backbone near the catalytic His356 residue. In the putative catalytic mechanism, His356 deprotonates the hydroxyl group of the glycerol backbone, enabling the oxygen to undergo a nucleophilic attack on the carbonyl group of the thioester bond (Fig. 6b). After forming an intermediate, electron transfers would result in formation of the CoA-SH and PI products. We hypothesize that binding of the substrates causes conformational changes in the TM6 helix to widen the hydrophobic, lumenal opening of the enzyme, allowing release of PI, with two acyl chains, into the membrane.

**Fig. 6: A model for MBOAT7 catalytic function and comparison of general MBOAT classes.**

A key question for MBOAT enzymology is how substrate specificity is achieved. Particularly, the phospholipid remodeling MBOAT enzymes have a high degree of evolutionary sequence conservation, are very similar in structure, and yet are able to distinguish between different lyso-lipid substrates. Our data suggest that TM segments 4–5 of MBOAT7 (or equivalent sequences of other Lands cycle MBOATs) constitute a lyso-phospholipid “specificity-determining region”. Key residues in this region are conserved for each different MBOAT and appear to interact with each headgroup (Fig. 4). Indeed, our analysis of chimeric expressed proteins indicate that these regions determine specificity for PI versus PC for MBOATs 5 and 7.

Beyond Lands cycle enzymes, phylogenetic analyses of MBOATs clusters them into groups according to their substrates (e.g., hydrophobic neutral lipids, hydrophilic polypeptides, or amphipathic lyso-phospholipids) (Fig. 6c). Whereas the acyl-CoA binding regions of MBOATs are relatively conserved, different paths appear to allow acyl-acceptor substrates to access the catalytic site (Fig. 6c). For MBOATs that acylate proteins (e.g., hedgehog acyltransferase, porcupine), the protein substrate reaches the catalytic center from the ER lumen. For MBOATs that catalyze neutral lipid synthesis (e.g., DGAT1, ACAT1, ACAT2), the substrate accesses the catalytic center from a lateral gate in the plane of the membrane. For the MBOATs that mediate phospholipid remodeling, such as MBOAT7, the lysophospholipid substrates access the enzyme tunnel from the lumenal side.

Loss-of-function mutations in human MBOAT7 result in liver disease, intellectual disability, early onset seizures, and autism spectrum disorders^{19,20,21,25,26,27,28,30}. Mapping the pathogenic MBOAT7 mutations responsible for these disorders on to our MBOAT7 structure model enables some predictions on how they disrupt MBOAT7 function (Supplementary Fig. 12a). For example, the E253K disease-causing mutation¹⁹, though located in TM7 away from the catalytic center, may destabilize the protein by disrupting hydrogen bonds of Glu253 with the backbones of Asn301 and Ile 302 (Supplementary Fig. 12b). The Arg384 mutation, found in patients with intellectual developmental disorders, is close to the entrance of acyl-CoA binding pocket and predicted to interfere with the recognition or binding of acyl-CoA substrates (Supplementary Fig. 12c).

High expression of MBOAT7 has been detected in multiple types of cancer^29,30, and studies of clear cell renal carcinoma suggest MBOAT7 inhibition as a possible therapy²⁹. We therefore thought to utilize the MBOAT7 structure to identify lead inhibitory compounds. Indeed, a structure-aided virtual screen identified two potential inhibitors, Sevenin-1 and −2 (Fig. 5). Although these compounds require further characterization and optimization, they provide a foundation for developing selective MBOAT7 inhibitors as therapeutic agents for aggressive cancers with MBOAT7 dependency, where their use may be valuable despite potential adverse effects such as causing transient fatty liver. Interestingly, similar to other MBOAT inhibitors^18,38,39, the structures of Sevenin-1 and −2 contain two bulky aromatic groups connected by an amide bond. This could be a common scaffold for the MBOAT inhibitors, which may help guide the design of future inhibitors for these enzymes.

Methods

Protein expression and purification

MBOAT7 from Homo sapiens (UniProt ID: Q96N66) was expressed in HEK293 freestyle cells (ThermoFisher), using the inducible stable cell line system⁴⁰. Briefly, the MBOAT7 cDNA sequence was cloned into the pSBtet vector with a C-terminal GFP-strepII sequence. The stable cell line was generated by co-transfection of the pSBtet-MBOAT7 and the SB-100X transpose vector, followed by a week of puromycin selection. The stable cells were grown in suspension at 37 °C in FreeStyle 293 Expression Medium (ThermoFisher). When the cells reached a density of 2.0–2.5 × 10⁶ cells per ml, 20 µM doxycycline and 5 mM sodium butyrate were added to the cells to induce the protein production and boost the expression, respectively. Temperature was set to 30 °C. The cultures were harvested after ~60 h. The cells were washed once in ice-cold PBS, then flash frozen in liquid nitrogen and stored in –80 °C or placed on ice for immediate use.

All protein purification steps were performed at 4 °C. Thawed cell pellets were re-suspended in the lysis buffer containing 50 mM Tris-HCL, pH 8.0, 150 mM NaCl, 1 mM EDTA and supplemented with 1× complete protease inhibitor cocktail (Roche). DDM/CHS detergent (1%/0.1%) (Anatrace) was added to lyse the cells for 2 h. Insoluble debris was removed by centrifugation 50,000 × g for 45 min, and the supernatant was incubated with prewashed Strep-Tactin Sepharose beads (IBA) for 1 h. The resin was then washed with ~20 ml of washing buffer (lysis buffer containing 0.1% digitonin) before the elution of GFP-fused MBOAT7 protein by washing buffer containing 2 g/L desthiobiotin. TEV proteases were then added to the eluted proteins to remove the GFP and strep tag before the elution was concentrated by a 50-kDa MWCO spin concentrator and injected onto a Superose 6 column (GE Healthcare) equilibrated in 100 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 0.05% (w/v) digitonin. The peak fractions containing MBOAT7 were pooled and concentrated. To reconstitute MBOAT7 proteins in PMAL-C8 (Anatrace), MBOAT7 in detergent was mixed with PMAL-C8 at a 1:3 (w:w) ratio, followed by gentle agitation overnight in the cold room. Detergent was then removed by mixing with Bio-Beads SM-2 (Bio-Rad) for 1 h. Bio-beads were removed by passing through a disposable polyprep column. The MBOAT7 protein was further purified by injecting onto a Superdex 200 increase 10/300 GL column equilibrated with 100 mM Tris-HCl, pH 7.5, and 150 mM NaCl.

To prepare the protein sample with lyso-PI and acyl-CoA, 10 µM lyso-PI was added at each step after cell solubilization and before PMAL-C8 reconstitution, whereas 10 µM acyl-CoA or nonhydrolyzable analogs was added at each step before the final size-exclusion purification.

Cryo-EM imaging and data processing

Grids for cryo-EM imaging were prepared on a Vitrobot Mark IV system (ThermoFisher Scientific) by applying 3 µL of PMAL reconstituted MBOAT7 at 5–6 mg/ml to Quantifoil holey gold grids (Au R1.2/1.3; 400 mesh) that were glow discharged for 30 s at 15 mA right before use. Grids were blotted for 5 s at 4 °C, 100 % humidity, and a blot-force of 12 N before being plunge frozen in liquid ethane cooled by liquid nitrogen. Cryo-EM data were collected on a Titan Krios electron microscope with a K3 direct electron detector. Movies were collected in counting mode with a pixel size of 0.825 Å and a total does 50.4 electron/Å2.

Movies were corrected for beam-induced motion by MotionCor2⁴¹, and contrast transfer function parameters were determined using CTFFIND4⁴². Template particle picking and 2D classification were performed using Simplified Application Managing Utilities of EM Labs (SAMUEL) similarly to previous dataset processing⁴³. To identify particles with rare angles, the deep-learning-based particle picker Topaz⁴⁴ was used and followed by multiple rounds of 2D classifications in cryoSPARC⁴⁵. Selected particles from SAMUEL and cryoSPARC were merged, and the redundant particles were removed before 3D processing with Relion3.0⁴⁶, including 3D classification, 3D refinement, and domain masking. 3D classification on binned particles at 3.3 Å pixel size were first performed to remove bad particles, and additional rounds of 3D classification were performed on less binned and unbinned particles. 3D classes were combined or discarded based on the quality of cryo-EM density of the transmembrane helices. Overall resolution of cryo-EM map was computed according to the gold standard Fourier shell correlation (FSC) method. The local resolution was calculated with the local resolution estimation function in Phenix⁴⁷. The directional resolution of the cryo-EM map was evaluated by the 3D FSC spherical score using the web server (https://3dfsc.salk.edu/)⁴⁸. For the 3D FSC calculation, a protein mask was provided in addition to the final map and two half maps. A flow-chart summary on the data processing can be found in Supplementary Fig. 2. The statistical details related to data processing and map validations are summarized in Supplemental Table 1.

Model building and refinement

The AlphaFold 2 predicted MBOAT7 model was used as the initial model. The model was docked into the map and refined using phenix.real_space_refine in Phenix⁴⁷ and manually adjusted in COOT⁴⁹. The resulting map was further put back through the real-space refinement procedure to undergo further refinement. The final model was validated with MolProbity⁵⁰. The map-model fitness was evaluation based on the Q score⁵¹, which is summarized in Supplemental Table 1. The comparison between the initial AlphaFold model and final MBOAT7 apo model is shown in Supplementary Fig. 3.

Mass spectrometry

To confirm substrate binding with purified MBOAT7 during protein preparation, lipid contents were extracted using the tert-butyl-methyl ether and methanol method⁵² from purified MBOAT7 proteins and lipid controls (Avanti Polar Lipids). Extracted lipids were resolubilized in 300 µL of isopropanol, methanol and chloroform (4:2:1, v:v:v) containing 7.5 mM ammonium acetate. Mass spectrometry analysis of lipid mixture was performed using an Orbitrap ID-X Tribrid mass spectrometer (ThermoFisher Scientific), equipped with an automated Triversa Nanomate nanospray interface (Advion Bioscience) for direct infusion lipid solution delivery. All full scan mass spectra were acquired in negative mode using mass resolution of 500,000 (FWHM at m/z 200). Visualization of mass spectrometry data was performed using Qual Browser on Xcalibur 4.3.73.11 software (Thermo Scientific).

Molecular dynamics simulation and docking

MBOAT7 was placed in a POPC bilayer by using the orientations of proteins in membranes (OPM) database (https://opm.phar.umich.edu/) as a reference and an in-house protocol. Briefly, proteins and lipids were prepared at Martini resolution, followed by short equilibrium. The Martini protein was then replaced with the original MBOAT7 apo structure and the Martini lipids were back-mapped into the membrane. Lyso-PI was modeled based on lyso-PC bound to cMBOAT5 (PDB 7F3X). The topology of lyso-PI was made based on the topology of PI and diacylglycerol and is available at git@github.com:ksy141/TG.git. One round of energy minimization resulted in the energy-minimized structure. Simulations were carried out three times, each of which was 500 ns long. Simulations were performed using the GROMACS (version 2018)⁵³ simulation engine with the CHARMM36m lipid and protein force field^54,55. Simulations were integrated with a 2-fs timestep. Lennard-Jones pair interactions were cut off at 12 Å with a force-switching function between 10 and 12 Å. The Particle Mesh Ewald algorithm was used to evaluate long-range electrostatic interactions⁵⁶. The LINCS algorithm was used to constrain every bond involving a hydrogen atom⁵⁷. A temperature of 310 K and a pressure of 1 bar were maintained with the Nose-Hoover thermostat and the Parrinello-Rahman barostat, respectively^58,59,60. The coupling time constants of 1 and 5 ps were used, respectively. A compressibility of 4.5 × 10⁻⁵bar⁻¹ was used for semi-isotropic pressure coupling. MDAnalysis was used for analyzing the trajectories⁶¹.

The constraint docking was performed using the MedusaDock 2.0 program⁶². The headgroup search space were estimated based on cMBOAT5 and MBOAT7 structures. Constraints were set between the phosphate group and conserved interacting residues.

Fluorescence protease protection assay

Membrane topology of MBOAT7 constructs was determined by fluorescence protease protection assay, as described previously³⁴. In brief, MBOAT7 and control (mCherry-ER3 and mEmerald-Sec61β) constructs was transfected into the SUM159 cells using the FuGENE® Transfection Reagent according to the manufacturer’s protocol. Two days post transfection, SUM159 cells were washed three times with 2 ml of KHM buffer (110 mM potassium acetate, 20 mM HEPES-KOH pH 7.2, 2 mM MgCl₂) at room temperature. During the imaging acquisition, 40 μM digitonin and 100 μg/ml proteinase K (Sigma) were added to the media at the indicated time points for plasma membrane permeabilization and fluorescent protein degradation, respectively. SUM159 breast cancer cells were obtained from the laboratory of Tomas Kirchhausen at Harvard Medical School (ATCC, CRL-1504) and were maintained in DMEM/F-12 GlutaMAX (Life Technologies) with 5 µg/ml insulin (Cell Applications), 1 µg/mL of hydrocortisone (Sigma), 5% FBS (Life Technologies 10082147, Thermo Fisher), 50 µg/mL of streptomycin, and 50 U/mL of penicillin.

MBOAT7 mutagenesis and activity assays

The Michaelis-Menten and substrate specificity assays were performed using purified and PMAL reconstituted MBOAT7 with a fluorescence-based coupled-enzyme assay, as described^12,14. Essentially, the activity was monitored by detecting the release of the free CoA-SH with the fluorescence probe 7-diethylamino-3-(4-maleimidophenyl)−4-methylcoumarin (CPM). A 10-µL reaction mixture contains 2 mg/ml BSA, 75 mM Tris, pH 7.5, 100 nM methyl arachidonyl fluorophosphonate (MAFP), 150 mM NaCl, and 0.025 g/L of MBOAT7 protein and the substrates acyl-CoA and lyso-PI with varied concentrations. For Michaelis-Menten characterization, either arachidonyl-CoA or lyso-PI was maintained at 100 µM, and the concentrations other substrate were varied. For characterizing substrate specificity, different acyl-CoA and lyso-phospholipids were added to a final concentration of 100 µM. The reaction was initiated by adding the enzymes to the reaction. The reaction proceeded for 3 min at 37 °C before being stopped by adding 1.3% SDS and incubating at room temperature for 5 min. 90 µL of CPM working solution (50 µM in 75 mM Tris, pH 7.5 and 150 mM NaCl) was added to the reaction, and the mixture was kept at room temperature for 30 min and followed by fluorescence detection (excitation 355 nm; emission 460) using the Tecan i-control infinite 200 plate reader. The background was determined from reactions without adding enzymes and subtracted to remove the background noise. MBOAT7 activity was stable over 15 min in the reaction buffer. Nonlinear regression to the Michaelis-Menten equation and allosteric sigmoidal analysis were performed using GraphPad Prism 9.

MBOAT7 mutants were generated by site-directed mutagenesis on the pFBM construct expressing MBOAT7 with a C-terminal GFP and strep tag using the Q5 Mutagenesis Kit (New England Biolabs), according to the manufacturer’s protocol. The primers used for mutagenesis are in Supplementary Data 1. The sequenced constructs were transfected into the HEK293 freestyle cells with PEI, and the cells were harvested after 70 h incubation at 30 °C. The cells were lysed and proteins were purified as described above with the modification that DDM/CHS was used for all solubilization, purification, and elution steps, and 10% glycerol was added in the final elution buffer. The initial levels of the wild-type and mutant MBOAT7 proteins were measured by immunoblotting against the GFP tag fused to the C-terminal of MBOAT7 and adjusted to the same quantity for assays. The acyltransferase activity was determined by measuring the incorporation of radiolabeled oleoyl-CoA into the product phosphatidylinositol⁶. In brief, a total 100 µL reaction mixture was prepared containing 75 mM Tris-HCl, pH 7.4, 2 g/L delipidated BSA, 150 mM NaCl, 100 nM MAFP, 50 µM 18:0 lyso-PI, 100 µM oleoyl-CoA containing 0.2 µCi [¹⁴C]-oleoyl-CoA as a tracer (American Radiolabeled Chemicals), and proteins (free GFP control, WT or mutant MBOAT7, ~100 ng) to initiate the reaction. The reaction was started by adding proteins to the reaction mixture and incubated for 15 min at 37 °C. The reactions were quenched by adding chloroform/methanol (2:1 v:v), followed by 2% phosphoric acid for phase separation. The organic phase was harvested, dried, resuspended in chloroform and loaded on a silica gel TLC plate (Analtech). Lipids were resolved in a solvent system consisting of chloroform, methanol, acetic acid, and water (65:35:8:5 v:v:v:v). The radioactivity of the phosphatidylinositol bands was revealed by Typhoon FLA 7000 phosphor imager (GE Healthcare Life Sciences) and quantified by Quantity One (V4.6.6).

VirtualFlow drug screening

The structure-based drug screen was performed essentially according to the instructions on the VirtualFlow website (https://virtual-flow.org/) and github page (https://github.com/VirtualFlow/VFVS). Firstly, the searching space was determined by covering both the cytosolic and ER lumenal tunnels of MBOAT7 apo structural model using the AutoDockTools⁶³. 5.7 million ligands were selected from the REAL library based on the properties of available small-molecule inhibitors developed for other human MBOAT members (DGAT1, ACAT1/2, Porcupine, HHAT) with the following criteria: molecular masses 375–425 Daltons; partition coefficient (SlogP) 1–2.5, hydrogen bond acceptors (HBA) 3–5; hydrogen bond donor (HBD) 1–2. The dockable forms of the 5.7 million compounds were directly downloaded from the website (https://virtual-flow.org/real-library). The VirturalFlow screen was performed using QuickVina program and each compound was docked once. The process took ~6 weeks using 100 cpu cores. The final 50 compounds with the highest docking scores (the lowest free energy) were analyzed based on their fingerprint similarity using the python RDKit package (https://www.rdkit.org/) and classified by signature substructures (Supplementary Fig. 11). Then, 12 compounds were further demanded from the Enamine company and tested in vitro. For some compounds, the chemical structures of the synthesized compounds and original docked compounds in VirtualFlow are slightly different possibly due to the file conversion during library preparation or library update in Enamine.

Reporting summary

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

Data availability

The three-dimensional cryo-EM density maps have been deposited into the Electron Microscopy Data Bank under accession numbers EMD-28552. The coordinates are deposited into the Protein Data Bank with accession number 8ERC. This study also cited published structures 7F3X (chicken MBOAT5 in complex with lyso-PC) and 7F40 (chicken MBOAT5 in complex with arachidonyl-CoA). The raw TLC, gel and blot data generated in this study are provided in the Source Data file. Source data are provided with this paper.

References

Vance, J. E. Phospholipid synthesis and transport in mammalian cells. Traffic 16, 1–18 (2015).
Article CAS PubMed Google Scholar
Kennedy, E. P. & Weiss, S. B. The function of cytidine coenzymes in the biosynthesis of phospholipides. J. Biol. Chem. 222, 193–214 (1956).
Article CAS PubMed Google Scholar
Gimeno, R. E. & Cao, J. Thematic review series: glycerolipids. Mammalian glycerol-3-phosphate acyltransferases: new genes for an old activity. J. Lipid Res. 49, 2079–2088 (2008).
Article CAS PubMed Google Scholar
Lands, W. E. Metabolism of glycerolipides; a comparison of lecithin and triglyceride synthesis. J. Biol. Chem. 231, 883–888 (1958).
Article CAS PubMed Google Scholar
Shindou, H. & Shimizu, T. Acyl-CoA:lysophospholipid acyltransferases. J. Biol. Chem. 284, 1–5 (2009).
Article CAS PubMed Google Scholar
Zhao, Y. et al. Identification and characterization of a major liver lysophosphatidylcholine acyltransferase. J. Biol. Chem. 283, 8258–8265 (2008).
Article CAS PubMed Google Scholar
Gijon, M. A., Riekhof, W. R., Zarini, S., Murphy, R. C. & Voelker, D. R. Lysophospholipid acyltransferases and arachidonate recycling in human neutrophils. J. Biol. Chem. 283, 30235–30245 (2008).
Article CAS PubMed PubMed Central Google Scholar
Catherine, C. Y., Chang, J.S. & Ta-Yuan Chang, T. -Y. Membrane-bound O-acyltransferases (MBOATs). Front. Biol. 6, 177–182 (2011).
Zhang, Q. et al. The structural basis for the phospholipid remodeling by lysophosphatidylcholine acyltransferase 3. Nat. Commun. 12, 6869 (2021).
Jiang, Y., Benz, T. L. & Long, S. B. Substrate and product complexes reveal mechanisms of Hedgehog acylation by HHAT. Science 372, 1215–1219 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Coupland, C. E. et al. Structure, mechanism, and inhibition of Hedgehog acyltransferase. Mol. Cell 81, 5025–5038.e10 (2021).
Article CAS PubMed PubMed Central Google Scholar
Wang, L. et al. Structure and mechanism of human diacylglycerol O-acyltransferase 1. Nature 581, 329–332 (2020).
Sui, X. et al. Structure and catalytic mechanism of a human triacylglycerol-synthesis enzyme. Nature 581, 323–328 (2020).
Qian, H. et al. Structural basis for catalysis and substrate specificity of human ACAT1. Nature 581, 333–338 (2020).
Long, T., Sun, Y., Hassan, A., Qi, X. & Li, X. Structure of nevanimibe-bound tetrameric human ACAT1. Nature 581, 339–343 (2020).
Guan, C. et al. Structural insights into the inhibition mechanism of human sterol O-acyltransferase 1 by a competitive inhibitor. Nat. Commun. 11, 2478 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Ma, D. et al. Crystal structure of a membrane-bound O-acyltransferase. Nature 562, 286–290 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Liu, Y. et al. Mechanisms and inhibition of Porcupine-mediated Wnt acylation. Nature 607, 816–822 (2022).
Lee, J. et al. Functional and structural changes in the membrane-bound O-Acyltransferase family member 7 (MBOAT7) protein: the pathomechanism of a novel MBOAT7 variant in patients with intellectual disability. Front Neurol. 13, 836954 (2022).
Article PubMed PubMed Central Google Scholar
Sun, L. et al. Phenotypic characterization of intellectual disability caused by MBOAT7 mutation in two consanguineous Pakistani families. Front Pediatr. 8, 585053 (2020).
Article PubMed PubMed Central Google Scholar
Johansen, A. et al. Mutations in MBOAT7, encoding lysophosphatidylinositol acyltransferase I, lead to intellectual disability accompanied by epilepsy and autistic features. Am. J. Hum. Genet. 99, 912–916 (2016).
Article CAS PubMed PubMed Central Google Scholar
Xia, M., Chandrasekaran, P., Rong, S., Fu, X. & Mitsche, M. A. Hepatic deletion of Mboat7 (Lpiat1) causes activation of SREBP-1c and fatty liver. J. Lipid Res. 62, 100031 (2020).
Helsley, R. N. et al. Obesity-linked suppression of membrane-bound O-acyltransferase 7 (MBOAT7) drives non-alcoholic fatty liver disease. Elife 8, e49882 (2019).
Thabet, K. et al. The membrane-bound O-acyltransferase domain-containing 7 variant rs641738 increases inflammation and fibrosis in chronic hepatitis B. Hepatology 65, 1840–1850 (2017).
Article CAS PubMed Google Scholar
Thabet, K. et al. MBOAT7 rs641738 increases risk of liver inflammation and transition to fibrosis in chronic hepatitis C. Nat. Commun. 7, 12757 (2016).
Article ADS CAS PubMed PubMed Central Google Scholar
Mancina, R. M. et al. The MBOAT7-TMC4 variant rs641738 increases risk of nonalcoholic fatty liver disease in individuals of European descent. Gastroenterology 150, 1219–1230 e6 (2016).
Article CAS PubMed Google Scholar
Luukkonen, P. K. et al. The MBOAT7 variant rs641738 alters hepatic phosphatidylinositols and increases severity of non-alcoholic fatty liver disease in humans. J. Hepatol. 65, 1263–1265 (2016).
Article CAS PubMed Google Scholar
Buch, S. et al. A genome-wide association study confirms PNPLA3 and identifies TM6SF2 and MBOAT7 as risk loci for alcohol-related cirrhosis. Nat. Genet 47, 1443–1448 (2015).
Article CAS PubMed Google Scholar
Neumann, C. K. A. et al. MBOAT7-driven phosphatidylinositol remodeling promotes the progression of clear cell renal carcinoma. Mol. Metab. 34, 136–145 (2020).
Article CAS PubMed PubMed Central Google Scholar
Donati, B. et al. MBOAT7 rs641738 variant and hepatocellular carcinoma in non-cirrhotic individuals. Sci. Rep. 7, 4492 (2017).
Lu, X., Lin, S., Chang, C. C. Y. & Chang, T.-Y. Mutant acyl-coenzyme A: cholesterol acyltransferase 1 devoid of cysteine residues remains catalytically active. J. Biol. Chem. 277, 711–718 (2002).
Article CAS PubMed Google Scholar
Tunyasuvunakool, K. et al. Highly accurate protein structure prediction for the human proteome. Nature 596, 590–596 (2021).
Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 590–596 (2021).
Lorenz, H., Hailey, D. W., Wunder, C. & Lippincott-Schwartz, J. The fluorescence protease protection (FPP) assay to determine protein localization and membrane topology. Nat. Protoc. 1, 276–279 (2006).
Article CAS PubMed Google Scholar
Gorgulla, C. et al. An open-source drug discovery platform enables ultra-large virtual screens. Nature 580, 663–668 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
LaPensee, C. R. et al. ATR-101, a selective and potent inhibitor of Acyl-CoA acyltransferase 1, induces apoptosis in H295R adrenocortical cells and in the adrenal cortex of dogs. Endocrinology 157, 1775–1788 (2016).
Article CAS PubMed Google Scholar
Langlois, D. K. et al. ATR-101, a selective ACAT1 inhibitor, decreases ACTH-stimulated cortisol concentrations in dogs with naturally occurring Cushing’s syndrome. BMC Endocr. Disord.18, 24 (2018).
You, L. et al. Development of a triazole class of highly potent Porcn inhibitors. Bioorg. Med. Chem. Lett. 26, 5891–5895 (2016).
Article CAS PubMed PubMed Central Google Scholar
Reed, A. et al. LPCAT3 inhibitors remodel the polyunsaturated phospholipid content of human cells and protect from ferroptosis. ACS Chem. Biol. 17, 1607–1618 (2022).
Article ADS CAS PubMed Google Scholar
Kowarz, E., Loscher, D. & Marschalek, R. Optimized sleeping beauty transposons rapidly generate stable transgenic cell lines. Biotechnol. J. 10, 647–653 (2015).
Article CAS PubMed Google Scholar
Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy. Nat. Methods 14, 331–332 (2017).
Article CAS PubMed PubMed Central Google Scholar
Rohou, A. & Grigorieff, N. CTFFIND4: Fast and accurate defocus estimation from electron micrographs. J. Struct. Biol. 192, 216–221 (2015).
Article PubMed PubMed Central Google Scholar
Thelot, F. A. et al. Distinct allosteric mechanisms of first-generation MsbA inhibitors. Science 374, 580–585 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Bepler, T. et al. Positive-unlabeled convolutional neural networks for particle picking in cryo-electron micrographs. Nat. Methods 16, 1153–1160 (2019).
Article CAS PubMed PubMed Central Google Scholar
Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 (2017).
Article CAS PubMed Google Scholar
Scheres, S. H. RELION: implementation of a Bayesian approach to cryo-EM structure determination. J. Struct. Biol. 180, 519–530 (2012).
Article CAS PubMed PubMed Central Google Scholar
Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D. Biol. Crystallogr. 66, 213–221 (2010).
Article CAS PubMed PubMed Central Google Scholar
Tan, Y. Z. et al. Addressing preferred specimen orientation in single-particle cryo-EM through tilting. Nat. Methods 14, 793–796 (2017).
Article CAS PubMed PubMed Central Google Scholar
Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D. Biol. Crystallogr. 60, 2126–2132 (2004).
Article PubMed Google Scholar
Chen, V. B. et al. MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr. D. Biol. Crystallogr. 66, 12–21 (2010).
Article CAS PubMed Google Scholar
Pintilie, G. et al. Measurement of atom resolvability in cryo-EM maps with Q-scores. Nat. Methods 17, 328–334 (2020).
Article CAS PubMed PubMed Central Google Scholar
Matyash, V., Liebisch, G., Kurzchalia, T. V., Shevchenko, A. & Schwudke, D. Lipid extraction by methyl-tert-butyl ether for high-throughput lipidomics. J. Lipid Res. 49, 1137–1146 (2008).
Article CAS PubMed PubMed Central Google Scholar
Van Der Spoel, D. et al. GROMACS: Fast, flexible, and free. J. Comput. Chem. 26, 1701–1718 (2005).
Article PubMed Google Scholar
Klauda, J. B. et al. Update of the CHARMM all-atom additive force field for lipids: validation on six lipid types. J. Phys. Chem. B 114, 7830–7843 (2010).
Article CAS PubMed PubMed Central Google Scholar
Best, R. B. et al. Optimization of the additive CHARMM all-atom protein force field targeting improved sampling of the backbone ϕ, ψ and side-chain χ1 and χ2 dihedral angles. J. Chem. Theory Comput. 8, 3257–3273 (2012).
Article CAS PubMed PubMed Central Google Scholar
Essmann, U. et al. A smooth particle mesh Ewald method. J. Chem. Phys. 103, 8577–8593 (1995).
Article ADS CAS Google Scholar
Hess, B. P-LINCS: A parallel linear constraint solver for molecular simulation. J. Chem. Theory Comput. 4, 116–122 (2008).
Article CAS PubMed Google Scholar
Nosé, S. A unified formulation of the constant temperature molecular dynamics methods. J. Chem. Phys. 81, 511–519 (1984).
Article ADS Google Scholar
Hoover, W. G. Canonical dynamics: equilibrium phase-space distributions. Phys. Rev. A 31, 1695–1697 (1985).
Article ADS CAS Google Scholar
Parrinello, M. & Rahman, A. Polymorphic transitions in single crystals: a new molecular dynamics method. J. Appl. Phys. 52, 7182–7190 (1981).
Article ADS CAS Google Scholar
Michaud-Agrawal, N., Denning, E. J., Woolf, T. B. & Beckstein, O. MDAnalysis: A toolkit for the analysis of molecular dynamics simulations. J. Comput. Chem. 32, 2319–2327 (2011).
Article CAS PubMed PubMed Central Google Scholar
Wang, J. & Dokholyan, N. V. MedusaDock 2.0: Efficient and accurate protein–ligand docking with constraints. J. Chem. Inf. Model. 59, 2509–2515 (2019).
Article CAS PubMed PubMed Central Google Scholar
Morris, G. M. et al. AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility. J. Comput. Chem. 30, 2785–2791 (2009).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We thank S. Sterling and M. Mayer at the Harvard cryo-EM center, and C. Xu and K. Song at the University of Massachusetts cryo-EM facility, for electron microscopy data collections and Z. W. Lai for help with mass spectrometry experiments. The MD simulations were performed on the high-performance GPU cluster (GM4) at the University of Chicago Research Computing Center (supported by NSF grant DMR-182869). The VirtualFlow drug screen was performed using the high-performance CPU clusters at the Harvard Medical School O2 system. We are grateful for discussions with the members of the Liao and Farese & Walther laboratories, and we thank Gary Howard for editorial assistance. S.K. and G.A.V. were supported by R01GM063796. This work was supported by R01GM141050 (to R.V.F.) and the Howard Hughes Medical Institute (T.C.W).

Author information

Xuewu Sui
Present address: Department of Biochemistry and Biophysics, College of Agriculture and Life Sciences, Texas A&M University, College Station, TX, USA
Tobias C. Walther & Robert V. Farese Jr
Present address: Cell Biology Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center, New York, NY, USA
These authors contributed equally: Maofu Liao, Tobias C. Walther, Robert V. Farese Jr.

Authors and Affiliations

Department of Molecular Metabolism, Harvard T.H. Chan School of Public Health, Boston, MA, USA
Kun Wang, Chia-Wei Lee, Xuewu Sui, Aidan B. Higgs, Aaron J. Baublis, Tobias C. Walther & Robert V. Farese Jr
Department of Cell Biology, Harvard Medical School, Boston, MA, USA
Kun Wang, Chia-Wei Lee, Xuewu Sui, Shuhui Wang, Aidan B. Higgs, Aaron J. Baublis, Maofu Liao, Tobias C. Walther & Robert V. Farese Jr
Pritzker School of Molecular Engineering, University of Chicago, Chicago, IL, USA
Siyoung Kim
Harvard T.H. Chan Advanced Multi-Omics Platform, Harvard T.H. Chan School of Public Health, Boston, MA, USA
Aaron J. Baublis & Tobias C. Walther
Department of Chemistry, Chicago Center for Theoretical Chemistry, James Franck Institute, and Institute for Biophysical Dynamics, The University of Chicago, Chicago, IL, USA
Gregory A. Voth
School of Life Sciences, Southern University of Science and Technology, Shenzhen, China
Maofu Liao
Broad Institute of MIT and Harvard, Cambridge, MA, USA
Tobias C. Walther & Robert V. Farese Jr
Howard Hughes Medical Institute, Boston, MA, USA
Tobias C. Walther

Authors

Kun Wang
View author publications
You can also search for this author in PubMed Google Scholar
Chia-Wei Lee
View author publications
You can also search for this author in PubMed Google Scholar
Xuewu Sui
View author publications
You can also search for this author in PubMed Google Scholar
Siyoung Kim
View author publications
You can also search for this author in PubMed Google Scholar
Shuhui Wang
View author publications
You can also search for this author in PubMed Google Scholar
Aidan B. Higgs
View author publications
You can also search for this author in PubMed Google Scholar
Aaron J. Baublis
View author publications
You can also search for this author in PubMed Google Scholar
Gregory A. Voth
View author publications
You can also search for this author in PubMed Google Scholar
Maofu Liao
View author publications
You can also search for this author in PubMed Google Scholar
Tobias C. Walther
View author publications
You can also search for this author in PubMed Google Scholar
Robert V. Farese Jr
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

K.W., M.L., T.C.W. and R.V.F. conceived the project. K.W. performed protein expression, purification, cryo-EM grid preparation, data processing, model building, molecular docking and VirtualFlow drug screen studies. M.L., S.W. and X.S. advised on cryo-EM data processing. K.W., C.-W.L., and A.B.H. performed the mutagenesis and activity studies. C.W.L. performed the fluorescence protease protection assay. S.K. and G.A.V. performed molecular dynamics simulations. A.J.B. performed the mass spectrometry analysis. K.W., M.L., T.C.W. and R.V.F. wrote the manuscript. All authors analyzed and discussed the results and contributed to the manuscript.

Corresponding authors

Correspondence to Maofu Liao, Tobias C. Walther or Robert V. Farese Jr.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Sarah Rouse and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Inforamtion File

Peer Review File

Description of Additional Supplementary Files

Supplementary Dataset 1

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Wang, K., Lee, CW., Sui, X. et al. The structure of phosphatidylinositol remodeling MBOAT7 reveals its catalytic mechanism and enables inhibitor identification. Nat Commun 14, 3533 (2023). https://doi.org/10.1038/s41467-023-38932-5

Download citation

Received: 23 September 2022
Accepted: 22 May 2023
Published: 14 June 2023
DOI: https://doi.org/10.1038/s41467-023-38932-5

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.