Molecular Insights into Human Phosphatidylserine Synthase 1 Reveal Its Inhibition Promotes LDL Uptake

Summary:

In mammalian cells, two phosphatidylserine (PS) synthases drive PS synthesis. Gain-of-function mutations in the Ptdss1 gene lead to heightened PS production, causing Lenz-Majewski syndrome (LMS). Recently, pharmacological inhibition of PSS1 has been shown to suppress tumorigenesis. Here, we report the cryo-EM structures of wild-type human PSS1 (PSS1^WT), the LMS-causing Pro269Ser mutant (PSS1^P269S) and PSS1^WT in complex with its inhibitor DS55980254. PSS1 contains 10 transmembrane-helices (TMs), with TMs 4–8 forming a catalytic core in the luminal leaflet. These structures revealed a working mechanism of PSS1 akin to the postulated mechanisms of the membrane-bound O-acyltransferase family. Additionally, we showed that both PS and DS55980254 can allosterically inhibit PSS1, and that inhibition by DS55980254 activates the SREBP pathways, thus enhancing the expression of LDL receptors and increasing cellular LDL uptake. This work uncovers a mechanism of mammalian PS synthesis and suggests that selective PSS1 inhibitors have the potential to lower blood cholesterol levels.

Graphical Abstract

In Brief:

PSS1 employs its catalytic histidine and a calcium ion in the transmembrane core to replace the polar head of PC or PE for PS synthesis and specific PSS1 inhibitors can block this synthesis to activate the SREBP pathway, increasing LDL receptor expression, and promoting LDL uptake in cells.

CELL-D-24-00876R3

Structural Insights into Human Phosphatidylserine Synthase 1 and How Its Inhibition Triggers SREBP Activation

Introduction

Phosphatidylserine (PS), the most abundant negatively charged glycerophospholipid in eukaryotic membranes, functions as a component of the cell membrane and plays vital roles in intracellular signaling, cell death, blood coagulation, cholesterol trafficking, and viral infection^1–5. It has also been shown to protect the health of nerve cell membranes and myelin^6,7. PS is preferentially distributed in the cytosolic leaflet of the plasma membrane and is also found in late endocytic compartments. The movement of PS from the inner to the outer leaflet of the plasma membrane takes place during apoptosis, which is thought to be pivotal in triggering cell death^8,9. PS is also exploited by many pathogens for their cell entry and propagation, for example, HIV and Ebola viruses can attack host cells by binding to exposed PS^10–12. On the surface of activated platelets, the exposed PS engages coagulation factors, leading to an increased production of thrombin, indicating its crucial role in significantly promoting blood clotting^13,14. Moreover, PS has been shown to modulate the activity of many proteins, such as protein kinase C, Raf-1, Na⁺/K⁺ ATPase, and the AMPA glutamate receptor^15–17.

The mechanisms of phospholipid biosynthesis from yeast to mammalian cells are typically conserved. In general, cells make Phosphatidylcholine (PC) or Phosphatidylethanolamine (PE) via the CDP-choline pathway or the CDP-ethanolamine pathway, but the mechanism of PS synthesis is distinct from that of the other phospholipids¹⁸. In yeast, PS is generated by the replacement of cytidine monophosphate of CDP-diacylglycerol with l-serine, whereas in mammalian cells, the de novo synthesis of PS occurs at the endoplasmic reticulum (ER) in a calcium-dependent manner¹⁹. Two membrane enzymes, named phosphatidylserine synthases 1 (PSS1) and phosphatidylserine synthases 2 (PSS2), are responsible for exchanging the polar head group of PC or PE by l-serine^19,20 (Figure 1A). PSS2 shows a preference for PE as its substrate, whereas PSS1 can utilize either PE or PC¹⁹. Animal studies found that mice with either Ptdss1 or Ptdss2 gene deficiency are viable, however, the knockout of both genes is lethal²¹. While PSS1 and PSS2 are well conserved across different species (Figure S1A) with an amino acid sequence similarity of over 50%, their expression profiles vary depending on the tissue type^22,23.

Figure 1. Functional characterization and overall structure of human PSS1.

(A) Schematic of PS synthesis. The structures of serine, choline and ethanolamine are shown.

(B) The activity of PSS1^WT and PSS1^P269S. Data are mean ± S.D. (n=3). P values (two-sided) were calculated by Student’s t tests using GraphPad Prism v.9; **, P<0.01; ****, P<0.0001.

(C) DS55980254 (termed as “DS”) inhibits the activity of PSS1^WT and PSS1^P269S in vitro. Data are mean ± S.D. (n=3). The 2D chemical structure of DS is shown. The IC₅₀ of DS to PSS1^WT is 1.65 μM (the deviation range is 1.22–2.26 μM). The IC₅₀ of DS to PSS1^P269S is 1.27 μM (the deviation range is 0.87–1.87 μM).

(D) Overall structure showing PSS1^WT dimer viewed from the side of the membrane. The cryo-EM map of the putative PE is shown.

(E) View of the dimer from the cytosol. The cryo-EM map of the putative PI is shown.

(F) View of the dimer from the lumen. TMs1-3 and TM9 are located at the interface and are involved in dimer assembly. TM5 and TM6 contribute to engage the calcium (green sphere).

(G) Structural details of the calcium binding site.

(H) Structural details of PI-mediated dimeric interface. The putative PC (gray sticks), PS (yellow sticks), PI (red sticks) and PE (cyan sticks) are shown. The interactions between calcium and residues are indicated by dashed lines. R349 from another protomer is underlined.

See also Figures S1 and S2.

Systematic alanine scanning revealed that the catalytic site of PS synthesis is in the luminal leaflet of PSS1^24,25. In addition, SERINC proteins, which incorporate serine into membranes, were identified to facilitate the synthesis of PS²⁶ by providing ample amounts of accessible l-serine. However, higher concentrations of PS in the cell membrane can inhibit the activity of PSS1, indicating a feedback regulation of PS production¹⁹. Despite these previous studies, the details of the catalytic and regulatory mechanisms of PS synthesis remain unclear.

Membrane concentration of PS is highly correlated to cholesterol metabolism. Low-density lipoprotein (LDL)-derived cholesterol first moves from the lysosome to the plasma membrane and then to the ER. The movement from the plasma membrane to the ER has been shown to be highly dependent on the presence of PSS1^27,28. PS also serves as a negative regulator of acyl-CoA cholesterol acyltransferases (ACATs), which convert cholesterol to cholesteryl ester for cholesterol storage²⁹. These findings indicate an important role of PSS1 in the intracellular cholesterol transport and homeostasis. Moreover, gain-of-function mutations on the Ptdss1 gene cause Lenz-Majewski syndrome (LMS), which is characterized by a combination of intellectual disability, sclerosing bone dysplasia, and distinct appearance anomalies³⁰. These mutants have been shown to disrupt the pattern formation of the actin cytoskeleton in osteoclasts³¹.

Recently, two studies demonstrated that pharmacological inhibition of PSS1 can suppress the growth of certain tumors and contribute to current therapeutic approaches for treating Ptdss2-deficient cancer and B cell lymphomas^32,33. Thus, structural information of PSS1 will help the design of its potent inhibitors. Here, we report cryo-electron microscopy (cryo-EM) structures of wild-type human PSS1 (PSS1^WT), its P269S mutant (PSS1^P269S) and inhibitor-bound PSS1^WT, which along with functional assays and molecular dynamics (MD) simulations uncover the working mechanism of PSS1. Moreover, our functional studies demonstrate that pharmacological inhibition of PSS1 may contribute to stimulate LDL uptake in cells.

Results

Overall Structure of PSS1^WT

We overexpressed Flag-tagged human PSS1^WT and PSS1^P269S in HEK293 cells and purified the recombinant proteins by an anti-Flag M2 resin with CaCl₂ at a final concentration of 5 mM. Gel filtration reveals that the resulting proteins migrate as a single peak (Figure S1B). To further validate the activity of recombinant PSS1, we used liquid chromatography–mass spectrometry (LC-MS/MS) to examine the synthesized PS. We supplemented l-serine into liposomes containing 2 mM 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC) at different concentrations and incubated with PSS1^WT or PSS1^P269S at 37 °C for 90 minutes. The lipids were extracted and analyzed by LC-MS/MS. Our results showed that PSS1^P269S exhibits more robust activity than PSS1^WT, confirming P269S as a gain-of-function mutation (Figure 1B). The IC₅₀ of DS55980254 to PSS1^WT and PSS1^P269S is 1.65 and 1.27 μM respectively in the presence of 200 μM serine, showing proper enzymatic activity of recombinant PSS1^WT and PSS1^P269S in vitro (Figure 1C).

We determined the structure of PSS1^WT by cryo-EM at 2.7-Å resolution (Figure 1D–F, Figure S2, and Table S1). Two PSS1 molecules form a homodimer, and the overall dimensions of the dimer are ~100 Å × 55 Å × 50 Å. Each monomer contains four peripheral α-helices (PH) and 10 transmembrane helices (TMs) with the dimensions of ~60 Å × 55 Å × 50 Å, consistent with the previous prediction of its topology²⁵. C2 symmetry was applied during data processing. The local resolution of the transmembrane helix core is about ~2.5-Å, which allows us to assign the side chains of most residues and most of associated endogenous lipids (Figure S2E). A homology search using Dali³⁴ shows that the overall structure of PSS1 does not share any similar fold with membrane proteins whose structures have been determined before. TMs 4–8 form the catalytic cavity in the luminal leaflet, which opens to both the lipid bilayer and lumen (Figure 1F). A phospholipid density was found between the luminal halves of TM4 and TM8 in the cryo-EM map of PSS1^WT (Figure 1D). Based on the morphology of this density and the chemical environment of the binding site, we tentatively assigned this lipid as a PE. We calculated the binding free energy of either POPC or POPE to PSS1^WT using the molecular mechanics-Poisson-Boltzmann surface area (MM-PBSA) method with the implicit membrane model³⁵. The results showed that POPC has a slightly stronger binding affinity compared to POPE (Figure S3A). Previous findings indicated that PSS1 may associate with calcium¹⁹. Consistent with this result, a calcium ion is coordinated by residues Glu197, Glu200, Gln217 and Asp221 in TM5 and TM6 (Figure 1G).

The dimer interface between the two PSS1 molecules comprises about 2,500 Ų. TMs 1–3 and TM9 are involved in the assembly of the dimer (Figure 1F). Interestingly, two phospholipids were found in the dimeric interface. Owing to the electron density of these lipids in the cryo-EM map, we tentatively assigned them as phosphoinositides (PIs) (Figure 1E). TMs 1–2 and TM9 of the other protomer generate the PI binding site. The main-chain carbonyls of Leu86 and Phe88 and the guanidino group of Arg102 along with the guanidino group of Arg349 in the other protomer engage the polar head of PI (Figure 1H). The structural analysis shows that PS association and the engagement of phospholipid substrates in PSS1 do not depend on its dimeric state. However, it is possible that the dimeric assembly plays an essential role in stabilizing the enzyme, thus indirectly facilitating PS production. The rationale behind the dimeric assembly for PSS1 function remains unclear and warrants further investigation.

PS-binding sites in the cytosolic leaflet

Two phospholipid-binding pockets are found around PH4 in the cytosolic leaflet of each PSS1 protomer (Figure 2A). A phospholipid density was observed in each positively charged cavity (Figure 2B). Based on the morphology of the density and the chemical environment of the binding site, we assigned each lipid as PS. Owing to the flexible nature, the lipid tails of PS1 were not completely observed in the cryo-EM map. The first PS binding pocket (termed “Site 1”) consists of the linker between PH1 and TM1, the linker between TM2 and TM3, and PH4, while the second pocket (termed “Site 2”) is formed by TMs 3, 4, 8 and PH4. The binding free energy of PS to PSS1^WT is −70.2 kcal/mol for the PS in Site 1 (PS1) and −83.4 kcal/mol for the PS in Site 2 (PS2) from the MD simulations (Figure S3B).

Figure 2. The cytosolic PS-binding sites.

(A) Luminal view of two PS-binding sites. One PSS1 monomer is colored in blue, and the other is colored in cyan. The PS molecules are shown by yellow sticks. The residues that contribute to the rigidity of the sites are indicated. Pro269 is highlighted in red. PSs are only labeled in one subunit, but present in both subunits with C2 symmetry.

(B) Electrostatic surface representation of the cytosolic PS-binding sites.

(C) Interaction details of the PS1-binding site. The cryo-EM map of PS1 is shown.

(D) Interaction details of the PS2-binding site. Residues are represented as sticks; dashed lines represent hydrophilic interactions with the distance. The cryo-EM map of PS2 is shown.

See also Figure S2.

Two π-cation interactions formed by Arg95/Trp247 and Arg336/Trp272 stabilize both cavities (Figure 2A). In Site 1, residues Arg95, Arg262, and Gln266 bind to PS1 via hydrophilic interactions, while residues Phe37, Phe250, Ala263, and Phe267 are responsible for accommodating the fatty acid chains of PS1 (Figure 2C). In Site 2, there are more extensive interactions between PS and PSS1 than that in Site 1, which is consistent with the stronger binding in Site 2 from the MD simulations. Residues Thr273 and Arg336 and the main-chain carbonyl of Arg95 and Phe267 interact with the polar head of PS2, while residues Phe93, Leu100, Trp101, Met178, Phe267, and Leu325 generate a hydrophobic environment to host the fatty acid chains of PS2 (Figure 2D). Alanine scanning demonstrated that mutating Arg95, Arg262, Gln266, or Arg336 to alanine increased the production of PS²⁴, probably owing to the disruption of these two sites. Moreover, two LMS-causing gain-of-function mutations, Leu265Pro and Pro269Ser, are also located in this area (Figure 2A). A previous functional analysis using patient’s samples revealed higher activity of PSS1^L265P and PSS1^P269S compared to that of PSS1^{WT 30}. Therefore, it is most likely that PSS1 may sense cytoplasmic levels of PS via these two PS-binding sites to regulate its activity. Since PS can inhibit the activity of PSS1^19,30 and two PS molecules bind to the cytosolic pockets of PSS1^WT, we speculate the structure of PSS1^WT represents the inactive state.

Overall Structure of PSS1^P269S

The structure of PSS1^P269S was determined by cryo-EM at 3.3-Å resolution (Figure 3A, Figure S4 and Table S1). In this structure there are also some lipid-like densities between the luminal halves of TM4 and TM8, between TM7 and TM9 and near cytosolic leaflet in the cryo-EM map of PSS1^P269S (Figure S4F), however owing to the resolution limitation of the map, we could not unambiguously model them. Although the overall structure of PSS1^P269S shares a similar fold to that of PSS1^WT, there are several marked conformational changes (Figure 3B). PH1, as well as the linkers that form the PS-binding sites (PH1 to TM1 and TM6 to TM7), are invisible in the cryo-EM map of PSS1^P269S. However, these structural elements are still notable when the cryo-EM map of PSS1^WT has been filtered to 3.3-Å (Figure S4G). Our MD simulations showed that PH4 (Figure 3C, lower panel) and the linker between PH1 and TM1 (Figure 3C, upper panel) exhibit higher root mean square fluctuations (RMSF) in PSS1^WT without PS on a 200 ns time scale compared to in the presence of PS. This indicates that these regions, which are key structural features of the PS-binding sites, are more flexible without PS in the cytosolic sites.

Figure 3. Overall structure of PSS1^P269S.

(A) Overall structure showing PSS1^P269S dimer viewed from the side of the membrane.

(B) The comparison of TMs in PSS1^P269S (yellow, dark green ball) and PSS1^WT (blue, light green ball).

(C) R.M.S.F. of linker between PH1 and TM1 (upper) and PH4 (lower) in PSS1^WT with (blue) and without PS binding (green). Data of each residue is an average of two monomers in three parallel groups: $\sqrt{\frac{{\sum }_{\text{traj}=1}^3{\sum }_{\text{monomer}=1}^2{(RMS{F}_{\text{monomer},\text{traj}})}^2}{6}}$

(D) Luminal view of PSS1^P269S compared to PSS1^WT. The conformational changes are indicated by arrows.

(E) Structural comparison of the entrances of the phospholipid substrate.

(F) Structural comparison of the linker between TM5 and TM6.

(G) Changes in the distance between representative amino acid residues (C_α) and the calcium ion with and without PS binding (from 200 ns MD). ΔDistance=Distance in PSS^WT without PS binding – Distance in PSS^WT with PS binding. Data is an average of three parallel groups. Each group contains two PSS1 molecules in one homodimer.

See also Figures S1, S2 and S4.

TMs 2, 3 and 9 that contribute to the dimeric interface in both structures are well aligned; however, TMs 4–6, 10, and PH3 undergo a notable shift toward to the edge of PSS1 (Figure 3B and D). Structural analysis shows that Phe168, Gly171, Gly175, and Met178 in TM4 and Trp321, Leu325, and Gly328 in TM8 create the gate for phospholipid access (Figure 3E). Although the conformations of TM8 in both structures appear similar, there is a ~5-Å movement of Phe168 at the luminal end of TM4 in the structure of PSS1^P269S, causing an opening of the phospholipid-binding site in the luminal leaflet that is absent in the structure of PSS1^WT (Figure 3E).

In the structure of PSS1^P269S, Glu212 in the linker between TM5 and TM6 as well as Glu197, Glu200, and Asp221 directly bind to the calcium ion (Figure 3F), which is consistent with the AlphaFold predicted model of PSS1^{WT 36} (Figure S4H), but different from the PSS1^WT(Figure 1G). These conformational changes lead to the catalytic core shifting towards to the edge of the enzyme. The MD simulations showed the backbones of the linker between TM5 and TM6 including residues 207–213 move closer to the calcium ion in the absence of PS-binding on a 200 ns time scale; in contrast, residues 214–219 move away from the calcium ion (Figure 3G). Notably, Glu197, Glu200, and Asp221 remain at similar positions regardless of the presence of PS (Figure 3G and Figure S3C). This finding is consistent with our structural observations on the conformational changes in both structures (Figure 3F). A previous study showed that Asn209 plays a crucial role in the l-serine exchange during PS production²⁴ and our structural analysis reveals Asn209 in PSS1^P269S is 4-Å closer to calcium than that in PSS1^WT (Figure 3F). Thus, it is tempting to speculate that the structure of PSS1^P269S represents an active state of PSS1^WT without PS binding. In the absence of PS binding, the arrangement of the calcium binding site may facilitate the access of serine to the catalytic core, which would promote the PS synthesis reaction. Furthermore, the binding free energy of POPC or POPE to PSS1^WT without PS in the cytosolic sites is lower than it is with PS, indicating a favorable recruitment of phospholipid substrates in the active state (Figure S3A).

Catalytic Mechanism

TMs 4–8 form the catalytic cavity in the luminal leaflet, which opens to both the lipid bilayer and lumen (Figure 4A and B). The cavities in both structures are negatively charged, which is favorable for engaging the positively charged phospholipid substrates (PC and PE), while the calcium ion would facilitate the recruitment of serine into the cavity. Structural analysis reveals that sufficient space is available around the luminal halves of TMs 4–6, particularly in proximity to calcium and His172 within the catalytic center, and a serine can be docked into the calcium site (Figure 4C). Our MD simulations showed that the free energy of binding of serine to PSS1^P269S monomer is −8.1 kcal/mol (Figure S3D). His172 may function as a catalytic base to subtract the hydroxyl proton of the serine, which remains trapped by the calcium ion, during the nucleophilic attack of the hydroxyl group on the ester bond between the phosphate group and the polar head of PC or PE to form PS (Figure 4D). After the reaction, the negatively charged PS would be released due to the repulsive electrostatic forces between PS and the cavity (Figure 4A and B).

Figure 4. The catalytic mechanism of PSS1.

(A) and (B) Electrostatic surface representation of catalytic cavity of PSS1^WT (A in blue) and PSS1^P269S (B in yellow).

(C) The docking model of serine (gray sticks) in the calcium binding site of PSS1^P269S.

(D) A working model of PSS1-meidated PS synthesis.

(E) Functional validation of the catalytic residue and lipid entrance. Data are mean ± S.D. (n=3).

(F) Salt bridge between Lys179 and Glu301 induces a shift of Phe168. Residues are represented as sticks; dashed lines represent hydrophilic interactions.

See also Figures S1, S2 and S7.

To validate our model, our functional assay showed that the PSS1^H172A presents no activity compared to that of PSS1^WT, although the purification results show that PSS1^H172A is well behaved in the detergent solution (Figure 4E and Figure S1B). When we mutated Phe168, Gly171 and Gly175 individually to tryptophan to introduce steric hindrance in the lipid gate, PSS1^F168W, PSS1^G171W and PSS1^G175W showed minimal enzymatic activity (Figure 4E and Figure S1B). It has also been shown that Glu197Ala, Glu200Ala, Glu212Ala, and Asp221Ala mutations almost completely abolished PSS1 activity²⁴, supporting their critical roles in coordinating calcium, which in turn is essential for PS synthesis.

In the structure of PSS1^P269S, a salt bridge forms between Lys179 and Glu301, causing the cytosolic leaflet of TM4 to be drawn towards TM7. Consequently, this salt bridge pushes the luminal leaflet of TM4 away from TM8 (Figure 4F). Further MD analysis indicates when PS is not bound, the distance between the Cα atoms of Phe168 and Trp321 increases, leading to the opening of the catalytic core facilitating the access of the phospholipid substrates (Figure S3E). Our functional analysis showed that PSS1^F168A has almost no activity (Figure 4E and Figure S1B), suggesting that Phe168 plays an important role in engaging the phospholipid substrates, as observed from the structure (Figure 3E). The conformational changes of Phe168 and Asn209 in both structures along with the MD simulations suggest that the catalytic core may be easily accessed by phospholipids and serine in the absence of PS. This sheds light on the mechanism underlying how PS in the cytosolic pockets modulates the activity of PSS1.

Inhibitor-bound State

Most recently, several PSS1 specific inhibitors, all of which share a similar backbone, have been identified^32,33. To elucidate the binding mode of these inhibitors, we expressed PSS1^WT in the presence of DS55980254 at a final concentration of 1 μM and purified the complex with DS55980254 at 10 μM, determining its structure at 2.87-Å (Figure 5A, Figure S5, and Table S1). DS55980254 binds to the cavity that is generated by TM3, PH2, TM7 and TM8 in the luminal leaflet (Figure 5B). Hydrophilic interactions with DS55980254 are provided by residues Asp134, Ser320, Arg323 and the main chain of Phe313. Phe313 forms a π-π interaction with the inhibitor while His317 forms a π-cation interaction. Importantly, residues Phe313 and Ser320 are not conserved in PSS2, underscoring the specificity of DS55980254 (Figure S1). Our functional analysis also showed that DS55980254 at a final concentration of 100 μM does not inhibit the base-exchange activity of PSS1^F313M/S320V (Figure 5C). Comparison with the structure of apo PSS1^WT reveals that DS55980254 triggers a conformational change of TM8 and fixes the conformation of TM7 with the loop between TM7 and TM8 via a hydrophilic bond between Lys308 and Phe313 (Figure 5D). To further dissect the inhibitory mechanism, we performed MD simulations to analyze the dynamics of PSS1 with or without DS55980254 binding in the absence of PS (Figure 5E). The results show the dynamics of the cytosolic PS-binding sites (residues 245–275, magenta region in Figure 5A) are reduced when DS55980254 binds to PSS1, which is akin to our structure in the presence of PS. Our finding concludes that PS and DS55980254 both act allosterically, although they bind to different sites (Figures 2A and 5A).

Figure 5. The inhibition of PSS1 by its specific inhibitor DS55980254.

(A) Overall structure of DS55980254 (labeled as “DS”) bound PSS1^WT dimer viewed from the side of the membrane. The cryo-EM map of DS (sticks) is shown. The residues 245–275 were colored in magenta.

(B) Interaction details between DS and PSS1. Dashed lines represent hydrophilic interactions.

(C) Functional validation of the DS55980254-binding residues. Data are mean ± S.D. (n=3 independent experiments). P value (two-sided) was calculated by Student’s t tests using GraphPad Prism v.9; ns, P > 0.05.

(D) The comparison of TM7 and TM8 between the inhibitor-bound PSS1^WT (light blue) and apo PSS1^WT (blue). The conformational changes are indicated by red arrows.

(E) R.M.S.F. of residues 165–300 in PSS1^WT with DS (yellow), PS (blue) and without DS and PS (green) during 200 ns MD simulations. Data of each residue is an average of two monomers in three parallel groups: $\sqrt{\frac{{\sum }_{\text{traj}=1}^3{\sum }_{\text{monomer}=1}^2{(RMS{F}_{\text{monomer},\text{traj}})}^2}{6}}$

See also Figures S1, S2 and S5.

PSS1 inhibitor activates SREBP pathway and stimulates LDL uptake

Two studies have revealed that the movement of accessible cholesterol from the PM to the ER depends on GRAMD1/Aster proteins^37,38. PS binds to the Aster proteins, which has been shown to be indispensable for Aster-mediated cholesterol transport³⁷. Additionally, PS is also required for an Aster-independent pathway for cholesterol transport²⁸. These studies demonstrate the necessity of PS for cholesterol transport to the ER. Moreover, depletion of Ptdss1 in CHO-K1 cells led to a deficiency of intracellular PS and triggered the cleavage of sterol regulatory element binding protein-2 (SREBP-2) to turn on the SREBP pathway²⁸.

To test the potential link between PSS1 enzymatic activity and the activation of the SREBP pathway, we conducted a SREBP-2 cleavage assay in Ptdss1^−/− CHO-K1 cells²⁷ by observing the ratio between the nuclear fragment and the membrane-bound precursor of SREBP-2 by SDS-PAGE. A previous study showed that in wild-type CHO-K1 cells, LDL can inhibit the cleavage of SREBP-2 by ~80%²⁸. Our results show that the LDL-induced inhibition of SREBP-2 cleavage has been abolished in Ptdss1^−/− cells (Figure 6A, lanes 1 and 2), while inhibition was restored to 79% when Flag-tagged PSS1^WT was overexpressed in Ptdss1^−/− cells (Figure 6A, lanes 3 and 4). This is consistent with the previous finding²⁸. Notably, adding LDL into the medium did not prevent the proteolytic process of SREBP-2 in the cells that overexpressed Flag-tagged PSS1^WT with DS55980254 at a final concentration of 1 μM (Figure 6A, lanes 5 and 6). When we overexpressed Flag-tagged PSS1^H172A, a catalytic mutant (Figure 4E), LDL inhibited the proteolytic processing of SREBP-2 only by 20% (Figure 6A, lanes 7 and 8). These findings indicate a direct correlation between the activity of PSS1 and the cleavage of SREBP-2.

Figure 6. PSS1 deficiency triggers the SREBP-2 cleavage and reduces the ER cholesterol concentration.

(A) The SREBP-2 processing with and without DS55980254 (labeled as “DS”) in Ptdss1^−/− CHO-K1 cells. On day 0, cells were set up in 6-well plates at a density of 25 × 10⁴ cells/well. On day 1, cells were switched to 1 ml medium A containing 10% FCS and 100 μl baculovirus expressed hPSS1^WT or hPSS1^H172A were supplemented into the medium. The inhibitor (DS) was supplemented into the medium at a final concentration of 1 μM at the time of infection and was kept at a constant concentration in each buffer throughout the experiment. On day 2, cells were switched to cholesterol-depletion medium A containing 1% hydroxypropyl-β-cyclodextrin (HPCD), which can deplete the free cholesterol in cells. After incubation for 1 hour, cells received cholesterol-depletion medium A in the absence or presence of 100 μg protein/ml LDL. After 6 hours, cells were harvested for immunoblotting of SREBP-2, Flag and Histone H3. P, precursor. N, nuclear. SREBP cleavage was quantified using ImageJ. For each lane, the ratio of nuclear to total SREBP-2 (nuclear + precursor) was calculated.

(B) Diagram of ER membrane fractionation scheme.

(C) and (D) denote major fractions recovered and analyzed by immunoblot analysis. CHO-K1 cells (C) and Ptdss1^−/− CHO-K1 cells (D) were treated according to the fractionation scheme as described in METHOD DETAILS. Aliquots representing equal volumes of each fraction (A–F) were subjected to immunoblot analysis for the indicated organelle markers.

(E) Cholesterol content of the purified ER membranes in CHO-K1 cells and Ptdss1^−/− CHO-K1 cells. Lipids were extracted from the purified ER, and the amounts of cholesterol and phospholipids were quantified as described in METHOD DETAILS. Mole % of cholesterol in the ER from different cells was calculated as the ratio of cholesterol to phospholipids plus cholesterol. Data are mean ± S.D. (n=4). P value (two-sided) was calculated by Student’s t tests using GraphPad Prism v.9; **, P<0.01.

(F) Working model of PSS1 inhibitors in simulating LDL uptake.

To investigate the direct relationship between PSS1 and cholesterol concentration in the ER, we isolated the ER fractions from regular CHO-K1 cells and Ptdss1^−/− CHO-K1 cells according to a previously established protocol³⁹ (Figure 6B–D). Lipidomics analysis showed that the mole ratio of cholesterol to phospholipids plus cholesterol in the ER of Ptdss1^−/− cells is ~35% lower than that in the regular cells (Figure 6E). This result supports our findings from the SREBP-2 cleavage assay, which implies that there is a relatively low cholesterol concentration in the ER of Ptdss1^−/− cells (Figure 6A). Since PSS1 deficiency triggers the activation of SREBP, we hypothesize that PSS1 inhibitors are promising compounds to increase the uptake of LDL-derived cholesterol (Figure 6F).

To validate this hypothesis, we conducted the SREBP-2 cleavage assay in CHO-K1 cells treated with DS55980254 from 0.01 μM to 1 μM. The proteolytic processing of SREBP-2 increased as did the mRNA level of LDLRs (Figure 7A–B). When SV589j cells, a line of SV40-immortalized human fibroblasts²⁷, were treated with DS55980254 at a final concentration of 1 μM, the cleavage of SREBP-2 increased (Figure 7C). More importantly, the surface expression of LDLRs in treated cells was three times higher than in untreated cells (Figure 7D and Figure S6A), and after a two-hour treatment with fluorescence labeled LDL, the uptake of LDL doubled in the presence of DS55980254 compared to untreated cells (Figure 7E and Figure S6B). The cholesterol staining assay showed that inhibiting PSS1 activity does not lead to cholesterol accumulation in lysosomes, in contrast, inhibiting NPC1 by its inhibitor U18666A⁴⁰ causes the substantial accumulation of cholesterol in lysosomes (Figure 7F).

Figure 7. DS55980254 stimulates LDLR expression and LDL uptake without causing cholesterol accumulation in lysosomes or overexpression of HMGCR in the ER.

(A) The SREBP-2 processing and (B) LDLR mRNA level with and without DS55980254 (labeled as “DS”) in CHO-K1 cells. On day 0, cells were set up in 6-well plates at a density of 15 × 10⁴ cells/well. On day 1, DS was supplemented into the medium at different concentration as indicated and was kept at a constant concentration in each buffer throughout the experiment. On day 2, cells were switched to cholesterol-depletion medium A containing 1% hydroxypropyl-β-cyclodextrin (HPCD). After incubation for 1 hour, the cells received cholesterol-depletion medium A in presence of 100 μg protein/ml LDL. After 18 hours, cells were harvested for immunoblotting of SREBP-2, Flag and Histone H3 and measuring LDLR mRNA level. P, precursor. N, nuclear. SREBP cleavage was quantified using ImageJ. For each lane, the ratio of nuclear to total SREBP-2 (nuclear + precursor) was calculated. Data of LDL mRNA level is presented as mean ± S.D. (n=3 biological replicates, with each point representing the average of three technical replicates).

(C) The SREBP-2 processing, (D) surface LDLR expression level and (E) BODIPY FL-LDL uptake with and without DS in SV589j cells. On day 0, cells were set up in 6-well plates at a density of 25 × 10⁴ cells/well. After plating for 6 hours, cells were supplemented with DMSO (solvent control) or 1μM DS, which was kept at a constant concentration in each buffer throughout the experiment. On day 1, cells were switched to cholesterol-depletion medium B. After incubation for 16 hours, the cells received cholesterol-depletion medium B in presence of 50 μg protein/ml LDL. After 24 hours, cells were harvested for immunoblotting of SREBP-2, Flag and Histone H3 or by incubation with EDTA, washed, incubated with PE-anti-LDLR, and subjected to flow cytometry. The rest of cells were supplemented with 5 μg protein/ml BODIPY FL-LDL. After 2 hours, cells were harvested for flow cytometry. Data are mean ± S.D. (n=3). P value (two-sided) was calculated by Student’s t tests using GraphPad Prism v.9; **, P<0.01.

(F) Cholesterol distribution in CHO-K1 cells with different compounds treatment. On day 0, CHO-K1 cells were plated on 12-mm glass coverslips in medium A with 5% FCS. On day 1, cells were incubated with DMSO (solvent control), 1μM DS or 1μM U18666A. Both compounds were kept at a constant concentration in each buffer throughout the experiment. On day 2, cells were switched to cholesterol-depletion medium A. After a 12-hour incubation, cells were switched to medium B containing 5% LPDS, 10 μM compactin, and 200 μM mevalonate. After 6 h, cells were imaged using filipin (Scale bar: 20 μm.).

(G) Measurement of sterol synthesis. On day 0, CHO-K1 cells were set up in 6 cm plates at a density of 4 × 10⁵ cells/plate. On day 1, cells were incubated with DMSO (solvent control), 1 μM DS or 1 μM compactin (CPN). Compounds were kept at a constant concentration in each buffer throughout the experiment. On day 2, cells were switched to medium A supplemented with 5% FCS and 3 μCi/ml [¹⁴C] acetate as well as cold acetate at a final concentration of 0.1 mM. After 4 hours, cells were harvested for the metabolic labeling study and immunoblotting of HMG-CoA reductase, and Histone H3. Data are mean ± S.D. (n=3). P values (two-sided) were calculated by Student’s t tests using GraphPad Prism v.9; *, P<0.05.

See also Figure S6.

To explore the effect of DS55980254 on cholesterol synthesis, we measured the de novo cellular sterol amount after a 4-hour incubation with radiolabeled acetate. Our results showed that DS55980254 increases sterol levels by 10% compared to the solvent control; in contrast, compactin, which inhibits the activity of HMG-CoA reductase (HMGCR, the rate-limiting enzyme in cholesterol synthesis), reduces sterol levels by 10% compared to the solvent control (Figure 7G). Moreover, compactin induces a considerable overexpression of HMGCR, whereas DS55980254 does not (Figure 7G).

Our results suggest that abolishing the enzymatic activity of PSS1 leads to lower cholesterol levels in the ER. Furthermore, DS55980254 can trigger SREBP-2 cleavage and promote LDL uptake in cells while only slightly increasing cholesterol synthesis, but does not induce a notable overexpression of HMGCR. Since selective inhibition of PSS1 does not induce ER stress in the presence of PSS2³², it is worthwhile to test whether PSS1-specific inhibitors can reduce blood cholesterol levels.

Discussion

The generation of PS in mammalian cells exclusively occurs through PS synthases. The phospholipid substrate gains entry to the enzyme through a gate formed between the luminal halves Notably, three glycine residues (Gly171, Gly175, and Gly328) that face the gate reduce steric hindrance, facilitating the unimpeded access of the phospholipid (Figure 3E). l-serine, the other substrate, can be incorporated into the membrane from the cytosol by SERINC proteins. A previous study demonstrated that the overexpression of mammalian SERINC1, SERINC2, or SERINC5 in COS cells increases the enzymatic activity of PSS1²⁶. Prior to the reaction, l-serine can be sequestered by the calcium in the core. This interaction ensures that l-serine attains an optimal proximity to the catalytic histidine, facilitating the deprotonation of its hydroxyl group. The activated serine is poised to initiate an attack on the ester bond linking the phosphate group and the polar head of PC or PE. Since PS appears to be more abundant in the cytoplasmic leaflet than in the luminal leaflet in the ER⁴¹, the newly synthesized PS must be flipped to accumulate on the cytosolic leaflet. It is highly possible that without PS in the cytosolic sites, PSS1 represents an active state. In turn, excessive PS molecules in the cytosolic leaflet become negative regulators to reduce the enzymatic activity and preventing the overproduction of PS.

If the structure of PSS1^WT represents the inactive state, why could we detect its activity through the LC-MS/MS analysis? It is conceivable that POPC-rich liposomes serve as mimics for PS-deficient membranes, prompting PSS1 to release its endogenously bound PS into the lipid environment. This release converts PSS1 from an inactive to an active state. Since the orientation of PSS1 reconstituted into the liposome can be random, as PS is synthesized, the PS-binding site on both leaflets becomes occupied, leading to a decrease in the activity of PSS1^WT. In contrast, the PS-binding sites were disrupted in PSS1^P269S, preventing its activity from being regulated by the newly synthesized PS, resulting in a higher activity compared to PSS1^WT (Figure 1B).

Intriguingly, the mechanism of PSS1 bears resemblance to the previously proposed mechanisms of membrane-bound O-acyl transferases (MBOATs)^42–48, although the architecture of PSS1 does not align with any known MBOATs and there is no evolutionary correlation between them (Figure S7A). As ER-membrane enzymes, MBOATs catalyze the transfer of an acyl group from an acyl-CoA to protein or lipid substrates⁴⁹ (Figure S7B). For instance, Porcupine employs a catalytic histidine to trigger the activation of a serine residue on Wnt for its lipidation⁴⁷. l-serine and Wnt can access these enzymes via their luminal side and a catalytic histidine triggers serine activation by deprotonating its hydroxyl group to facilitate nucleophilic attack on the ester bond of their substrates (thioester bond of acyl-CoA and phosphoester bond of PC or PE). In contrast, the fatty acid donors traverse distinct pathways to reach the catalytic core. Acyl-CoA approaches the catalytic histidine of MBOATs from the cytosol; conversely, PC or PE gains access to the catalytic histidine of PSS1 from the luminal leaflet. Following product formation, most lipidated Wnts are transported to their trafficking carrier, Wntless, through a transit complex⁵⁰. It remains unclear whether the newly synthesized PS diffuses directly into the luminal leaflet or if PSS1 forms a transient complex with a flippase, which can then shuttle the PS to the cytosolic leaflet.

Given that PS plays a crucial role as a signaling molecule, the synthesis of PS must be tightly regulated by cells. Our structural findings, combined with previous discoveries²⁴, suggest that cells employ three strategies to mediate the appropriate production of PS within the ER membrane. First, the activity of SERINC proteins may control the abundance and therefore the accessibility of l-serine to the reaction center. Second, the newly synthesized PS must be flipped from the luminal leaflet to cytosolic leaflet. PS flippases couple the energy from the hydrolysis of ATP to the transport of PS from the outer to the inner leaflet thereby maintaining an asymmetric distribution of PS in the ER membrane. A previous study showed that ATPase activity of phospholipid-transporting ATPase IA (ATP8A1) is stimulated by PS in vitro⁵¹. Third, excessive PS can bind to the cytosolic pockets of PSS1 to regulate its enzymatic activity, inhibiting the production of PS. The concentration of PS in the cytosolic leaflet will be reduced after PS is distributed to other intracellular organelles. Then, the reduced concentration of PS in the membrane will cause the PS in the binding pockets to diffuse into the lipid bilayer, thereby re-activating its PS synthesis.

Limitation of the study

We determined our structures using detergent micelles, and it is conceivable that PSS1 might exhibit distinct conformations within different types of nanodiscs or the native cell membrane. Another interesting question is how PS is flipped after production and whether PSS1 can form a complex with one of the PS flippases. Future in vitro biochemical and cell biological work may uncover this puzzle. Moreover, in vivo studies are needed to determine whether targeting PSS1 could either serve as a statin alternative or be used in conjunction with statins to lower blood cholesterol levels.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Xiaochun Li (xiaochun.li@utsouthwestern.edu).

Materials availability

Plasmids generated in this study and the raw cryo-EM dataset are available upon request from the lead contact.

Data and code availability

• The 3D cryo-EM density maps have been deposited in the Electron Microscopy Data Bank (EMDB) under the accession numbers EMD-44178, EMD-44179 and EMD-44180. Atomic coordinates for the atomic model have been deposited in the Protein Data Bank (PDB) under the accession numbers 9B4E, 9B4F and 9B4G.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cell lines

Spodoptera frugiperda Sf9 cells (ATCC) were used to generate baculovirus and maintained in Sf-900 III SFM medium (Gibco) at 27 °C with shaking. HEK 293S GnTI⁻ cells (ATCC) were used for protein expression and maintained in Freestyle 293 expression medium (Gibco) with 2% FCS at 37 °C and 8% CO₂ with shaking. Medium A is a 1:1 mixture of Ham’s F-12 medium and DMEM containing 2 mM L-glutamine. Medium B is Dulbecco’s modified Eagle’s medium (DMEM)-low glucose (1000 mg/L). Cholesterol-depletion medium A consists of medium A supplemented with 5% newborn calf lipoprotein-deficient serum (LPDS)⁵², 10 μM compactin, and 200 μM mevalonate. Cholesterol-depletion medium B consists of medium A supplemented with 5% LPDS, 30 μM compactin, and 200 μM mevalonate. All media except Sf-900 III SFM medium are supplemented with 100 U/ml of penicillin and 100 μg/ml of streptomycin sulfate. CHO-K1 cells (Chinese hamster) were maintained in medium A supplemented with 5% FCS. CHO-K1 cells lacking PSS1 (designated as Ptdss1^−/− cells) were generated using CRISPR-Cas9 by the Brown/Goldstein lab²⁷ and grown in medium A containing 10% FCS. SV589j cells, a clonal line of SV40-immortalized human fibroblasts derived from SV589 cells by the Brown/Goldstein lab²⁷, were grown in medium A containing 5% FCS. Buffer B is Hank’s Balanced Salt Solution (HBSS) containing 2% FCS and 5 mM sodium EDTA, pH 8.0. All cells were routinely monitored for mycoplasma contamination.

METHOD DETAILS

Protein expression and purification

The complementary DNA (cDNA) of human PSS1 (Protein ID NP_055569.1) was cloned into pEG-BacMam with a C-terminal Flag-tag. Point mutations were introduced into the coding region of PSS1 by site-directed mutagenesis using the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies). The coding region of each plasmid was sequenced to ensure the integrity of the construct. The protein was expressed in HEK-293S GnTI⁻ cells using the baculovirus system. To generate baculovirus, 5 μg bacmid was transfected into 1 × 10⁷ sf9 cells. After total 72 h, the P1 virus was harvested. For the P2 virus, 1 ml P1 was added to 50 ml 3 × 10⁶ sf9 cells, and then 50 ml medium was added the next day. After total 72 h, the P2 virus was harvested for protein expression.

Protein purification was performed as previously described²⁹. Briefly, 800 ml of HEK-293S GnTI⁻ cells at a density of 3 × 10⁶ were infected with 50 ml of P2 virus. After 8 hours, 10 mM sodium butyrate was added, and cells were then incubated at 37 °C to express proteins. For the PSS1^WT-inhibitor complex, DS55980254 (ProbeChem) was added to the culture at a final concentration of 1 μM upon infection and maintained at a final concentration of 10 μM in all subsequent purification steps. The cells were collected after 48 h of infection and were disrupted by sonication in buffer A (20 mM HEPES, pH 7.5, 150 mM NaCl, 5mM CaCl₂) supplemented with 1 mM phenylmethylsulfonyl fluoride and 5 μg/ml leupeptin. After low-speed centrifugation (4,000 r.p.m, rotor: A-4–62, Eppendorf, 4 °C, 5 min), the resulting supernatant was incubated in buffer A with 1% (w/v) lauryl maltose neopentyl glycol (LMNG) (Anatrace) for 1 h at 4 °C. The lysate was then centrifuged at high-speed (18,000 r.p.m, rotor: F21–8x50y, Thermo scientific, 4 °C, 30 min) to remove insoluble components, and the supernatant was loaded onto the Flag M2 affinity resin (Sigma-Aldrich). After two tandem washes, the protein was eluted in buffer A supplemented with 100 μg/ml 3×Flag peptide, 0.01% (w/v) LMNG and concentrated by Amicon Centrifugal Filters (Millipore). The concentrated protein was purified by a Superose 6 Increase 10/300 size-exclusion chromatography column (GE Healthcare) in buffer A and 0.06% (w/v) Digitonin. Finally, the peak fractions were pooled, concentrated, then directly flash frozen in liquid nitrogen and stored at −80 °C. Typically, 800 ml cells can produce 400 μg protein.

EM sample preparation and imaging

3 μl of protein sample at ~8 mg ml⁻¹ (~150 μM) was added to Quantifoil R1.2/1.3 400 mesh Au holey carbon grids (Quantifoil), blotted using a Vitrobot Mark IV (FEI) (15 force, 5 s blotting time and 100% humidity at 22 °C), and frozen in liquid ethane. For the PSS1^WT-inhibitor complex, DS55980254 was added into the sample at the final concentration of 150 μM before grid freezing. The grids were imaged in a 300 kV Titan Krios (FEI). For PSS1^WT and PSS1^WT-inhibitor complex, raw movie stacks were collected with a Gatan K3 Summit direct electron detector (Gatan). Data were collected using SerialEM at 0.834 Å per pixel and a nominal defocus range of −0.8 to −1.8 μm. Images were recorded for 5 s exposures in 50 subframes to give a total dose of roughly 60 electrons per Ų. For PSS1^P269S, raw movie stacks were collected with a Falcon 4i camera at 0.738 Å per pixel and a nominal defocus range of −0.8 to −1.8 μm. The exposure time for each micrograph was 4 s with a total dose of ~60 electrons per Ų.

Imaging processing and 3D reconstruction

For PSS1^WT and DS55980254-bound PSS1^WT, dark subtracted images were first normalized by gain reference (0.834 Å per pixel). Drift correction was performed using MotionCor2⁵³ and performed in RELION3.1⁵⁴. The contrast transfer function (CTF) was estimated using CTFFIND4⁵⁵. For PSS1^P269S, dark subtracted images were first normalized by gain reference (0.738 Å per pixel). Drift correction and CTF were performed and estimated using CryoSPARC v4.4.1⁵⁶. Auto picking for all datasets was performed with crYOLO-v1.7.6 or 1.9.3⁵⁷ using the general model with the particle threshold of 0.1. Particle extraction was performed in RELION3.1 or CryoSPARC v4.4.1 with a box size of 300 pixels for PSS1^WT and DS55980254-bound PSS1^WT, and 340 pixels for PSS1^P269S. Subsequent two-dimensional classification, multi-class ab initio modeling, heterogenous three-dimensional (3D) refinement, nonuniform refinement, CTF refinement and local refinement of the best class were performed in CryoSPARC v.3.3.0 or v4.4.1 (Figures. S2, S4 and S5). All the masks for the refinement were automatically generated by CryoSPARC during the nonuniform refinement and local refinement.

The initial models were built de novo based on AlphaFold2 predicted structures and then manually adjusted and refined using COOT⁵⁸. Due to the limited resolution, residues 1–11, 143–151 and 410–473 of PSS1^WT, residues 1–11, 142–151 and 410–473 of DS55980254-bound PSS1^WT, and residues 1–39, 249–283 and 410–473 of PSS1^P269S were not built. Residues 154–162 of PSS1^WT, and residues 154, 156–158, 210 and 212 of DS55980254-bound PSS1^WT were built as alanine. The models were refined in real space using PHENIX⁵⁹. For cross-validations, MolProbity⁶⁰ was used to validate the final model. Local resolutions were estimated using CryoSPARC. Structure figures were generated using COOT⁵⁸, PyMOL (http://www.pymol.org) and ChimeraX⁶¹.

ER isolation

On day 0, both CHO-K1cells and Ptdss1^−/− CHO-K1 cells were set up in medium A with 10% FCS at 7 × 10⁵ cells/10 cm dish. On day 3, the cells were washed with phosphate-buffered saline, and then scraped into 15 ml tubes. All further operations were carried out at 4 °C. The cell pellets were collected after centrifugation at 1000 g for 10 min and resuspended in 1 ml of ice-cold Buffer C (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, and protease inhibitors: 50 μg/ml leupeptin, 25 μg/ml pepstatin A, 10 μg/ml aprotinin, 25 μg/ml phenylmethyl sulfonylchloride) with 15% (w/v) sucrose. Cells were then disrupted by passaging them 10 times through a ball-bearing homogenizer with a 10 μm clearance. The homogenized cells (Fraction A, see fractionation scheme in Figure 6B) were centrifuged at 7000 g for 15 min to yield a pellet and a supernatant. The supernatant was diluted to a total volume of 3 ml using Buffer C containing 15% sucrose. A discontinuous sucrose gradient was generated in an ULTRA CLEAR tube (Beckman Instruments) by overlaying the following sucrose solutions all in Buffer C: 2 ml 45%, 4 ml 30%, 3 ml of the diluted supernatant in 15% sucrose, and 1 ml 7.5%. The gradient was centrifuged at 100,000 g in a TH641 rotor (Thermo Scientific) for 1 h and allowed to slow down without braking, after which two bands of membranes were clearly visible and collected using a Pasteur pipet. The sucrose concentration in the light membrane ranged from 14%–18%, and 34%–38% in the heavy membrane fraction (Fraction C, about 800 μl). A discontinuous iodixanol gradient was generated by underlaying, in succession, 2.25 ml of 19%, 21%, 23%, and 25% (v/v) iodixanol, all in Buffer C. This discontinuous gradient was allowed to stand for 1–2 h, allowing for diffusion across the interfaces to form a continuous gradient. The heavy membrane fraction from above was loaded at the bottom of this gradient, which was then centrifuged for 2 h at 110,000 g in a TH641 rotor and allowed to slow down without braking. Fractions (0.8 ml each) were collected from the bottom. Lipids were extracted from ER fraction (Fraction E, in Figure 6B) by adding an equal volume of chloroform/methanol (1:1, v/v). Bottom organic layer was collected in a glass tube after vortex, spun at 2000 rpm (rotor: A-4–62, Eppendorf) for 10 minutes and evaporated under vacuum for overnight.

Cholesterol was further extracted using a modified MTBE method (2:1:1 MTBE:MeOH:PBS v/v) mixture in a 16 × 100 mm screw cap test tube with PTFE-lined caps⁶². The samples were vortexed and centrifuged to enhance phase separation. The upper layer (organic) was transferred to a separate screw cap vial with a Pasteur Pipette. The extraction process was repeated by first adding an additional 1-ml aliquot of MTBE to the sample and following the procedure described above. Extracts were dried under a gentle stream of nitrogen at 40 °C and reconstituted in methanol. Cholesterol levels were measured with isotope dilution mass spectrometry using a deuterated analog of cholesterol (d7) (Avanti Polar Lipids) added to the sample prior to extraction as a surrogate standard. Cholesterol was measured using high performance liquid chromatography (HPLC, Shimadzu LC30 coupled to a triple quadrupole mass spectrometer (MS, SCIEX API 5000) through an atmospheric pressure chemical ionization source⁶³. Total phospholipid levels were determined by a colorimetric assay⁶⁴ that measures inorganic phosphate released by Malachite Green Phosphate Assay Kit after acidic digestion. Cholesterol content in the ER is expressed as the mole percentage of total lipids (phospholipids + cholesterol).

Assaying of base-exchange activities by LC-MS/MS

POPC liposomes were prepared by sonication in a buffer containing 50 mM HEPES, pH 7.5 and 5 mM CaCl₂. Serine-exchange activity was measured by supplementation of 500 nM PSS1 or PSS1 variants into liposomes and indicated concentration of serine. Assays were performed in a 100 μl volume for 90 min at 37 °C and were terminated by adding 10 μl of 0.5 M EDTA. For the inhibition assay, DS55980254 was preincubated with PSS1 or PSS1 variants at indicated concentration for 10 min at room temperature.

Lipids were extracted from the liposomes using a modified methyl-tert-butyl ether (mTBE) method⁶². Briefly, 1 ml Milli-Q water, 1 ml of methanol and 2 ml of mTBE were added to glass tubes containing the sample, vortexed, and centrifuged. The organic phase was collected, spiked with 20 μL of a 1:5 diluted Splash Lipidomix standard solution (Avanti lipids, Alabaster, USA), dried under N₂ and resuspended in hexane. Lipids were analyzed by LC-MS/MS using a SCIEX QTRAP 6500⁺ (SCIEX, Framingham, USA) equipped with a Shimadzu LC-30AD (Shimadzu, Columbia, USA) HPLC system and a 150x2.1 mm, 5 μm silica column (Supelco, Bellefonte, USA). Samples were injected at a flow rate of 0.3 ml/min starting at 97.5% solvent A (hexane) a 2.5% solvent B (mTBE). Solvent B was increased to 5% over 3 min and then to 60% over 6 min.

Following this, solvent B was decreased to 0% during 30 s while Solvent C (90:10 (v/v) isopropanol-water) was set at 20% and increased to 40% over the next 11 min. Solvent C is increased to 44% over 6 min and then to 60% over 50 s. The system was held at 60% solvent C for 1 min prior to re-equilibration at 2.5% of solvent B for 5 min at a 1.2 ml/min flow rate. Solvent D (95:5 (v/v) acetonitrile-water with 10 mM Ammonium acetate) was infused post-column at 0.03 ml/min. The column oven temperature was maintained at 25 °C. Data was acquired in negative ionization mode using multiple reaction monitoring (MRM) and analyzed using MultiQuant software (SCIEX). The identified lipid species were normalized to its corresponding internal standard.

SREBP-2 processing and western blot

The details of cell plating and treatment are described in figure legends. Cells were treated with 25 μg/ml of N-acetyl-Leu-Leu-norLeucinal (ALLN) 1 h prior to harvesting for western blot of SREBP-2 and Histone H3 as described below.

Cells were lysed by the addition of PBS containing 0.3% (w/v) SDS and 150 U/ml of Benzonase Nuclease. Aliquots were mixed with 4x loading dye, after which 20 μl protein were applied to 4–12% or 4–20% gradient gels. After electrophoresis, the proteins were transferred to nitrocellulose filters, which were then incubated overnight at 4 °C with one of the following antibodies dissolved in 5% (w/v) nonfat milk: 5 μg/ml anti-SREBP-2, anti-Flag (1:1500), anti-Histone H3 (1:5000), anti-TOM20 (1:1000), anti-CREB (1:1000), anti-EEA1 (1:1000), anti-GM130 (1:1000), anti-Na⁺/K⁺ ATPase (1:1000), anti-Calnexin (1:1000), anti-NPC1 (1:1000), anti-ABCD3 (1:1000) and 5 μg/ml anti-HMGCR. Bound antibodies were visualized by chemiluminescence using Supersignal West Pico Chemiluminescent Substrate (Thermo Scientific) after incubation for 0.5 h at room temperature with goat anti-rabbit IgG conjugated to horseradish peroxidase (1:2500) or horse anti-mouse IgG conjugated to horseradish peroxidase (1:2500). The images were scanned using an Odyssey FC Imager (Dual-Mode Imaging System; 2 min integration time) and analyzed using Image Studio lite v5.2.

Quantitative real-time PCR

Total RNA was isolated from cells using the RNeasy Plus Universal Mini Kit (QIAGEN) according to the manufacturer’s instructions and then subjected to reverse transcription. Three replicate samples were subjected to real-time PCR analysis as described previously⁶⁵. The real time PCR contained, in a final volume of 20 μl, 20 ng of reverse transcribed total RNA, 167 nM forward and reverse primers, and 10 μl of 2× SYBR Green PCR Master Mix (PerkinElmer Life Sciences). PCR reactions were carried out in 384-well plates using the ViiA7 Real-Time PCR System (Applied Biosystems). All reactions were done in triplicate. mRNA for acidic ribosomal phosphoprotein 36B4 served as an invariant control. Relative amounts of mRNAs were calculated using the comparative C_T method. The primer sequences used for PCR are listed here:

36B4_forward: TCTGCACTCACGCTTCCTAGAG;

36B4_reverse: GGCAACAGTCGGGTAACCAA;

LDLR_forward: AGACACATGCGACAGGAATGAG;

LDLR_reverse: GACCCACTTGCTGGCGATA.

Surface LDLR expression

Anti-LDLR, a mouse monoclonal antibody that binds to human LDLR⁶⁶, is produced by the Brown/Goldstein lab. Phycoerythrin-conjugated anti-LDLR (PE-anti-LDLR) was generated using PE/R-Phycoerythrin Conjugation Kit (Abcam) according to the manufacturer’s instructions. After plating and incubation as described in the Figure legends, cells were released from monolayers by incubation with buffer B, after which ~1 × 10⁶ cells were suspended in buffer B and centrifuged for 5 min at 1000 rpm (rotor: TX-750, Thermo scientific). Each cell pellet was resuspended in 0.1 ml buffer B containing 2 μg/ml PE-anti-LDLR. After incubation for 15 min at room temperature, cells were washed once with buffer B and then resuspended in buffer B to a concentration of ~1 × 10⁶ cells/ml for flow cytometry.

BODIPY FL-LDL uptake

After plating and 2-hour incubation with BODIPY FL-LDL as described in the Figure legends, cells were released from monolayers by treatment with buffer B and then then centrifuged at 4 °C for 5 min at 1000 rpm (rotor: TX-750, Thermo scientific). The pelleted cells were washed once by 1 ml buffer B, resuspended in buffer B, and subjected to flow cytometry.

Metabolic labeling study

After plating and incubation as described in the figure legends, cells were lysed by 0.1 N NaOH and then supplemented with ethanol and KOH at final concentrations of 40% and 15%, respectively. The samples were saponified at 75 °C for 60 min. After saponification, the total sterols were extracted by petroleum ether. The organic phase was evaporated until completely dry and then solubilized in Complete Counting Cocktail for the determination of de novo cholesterol synthesis via liquid scintillation counting. To establish background radioactivity, the cells were chilled to 4 °C, re-fed with radiolabeled medium and immediately washed, saponified and extracted. An aliquot of each sample was taken for protein concentration determination using the BCA protein assay reagent (Thermo Scientific).

Flow cytometry

Resuspended cells were adjusted to a concentration of ~1 × 10⁶ cells/ml with buffer B in a single U-bottom tube and loaded into a MACSQuant Analyzer 16 flow cytometer. Dead cells and cell doublets were excluded by gating so that >20,000 cells were available for analysis in each condition. The data were analyzed using MACSQuantify v2.13.3 software.

Visualization of cholesterol distribution by fluorescence microscopy

After plating and incubation as described in the Figure legends, cells were fixed in 3.7% (w/v) formaldehyde in PBS for 15 min at room temperature. After washing with PBS, cells were labelled with 25 μg/ml filipin in PBS for 30 min at room temperature, then washed 3 times in PBS, and mounted in Shandon Immu-mount (Thermo Scientific). Fluorescence images were acquired using a Plan-Apochromat ×63/1.4 oil DIC objective (Zeiss, Oberkochen, Germany), a Zeiss LSM 800 microscope (Zeiss), and ZEN imaging software (Zeiss). Images of different conditions were captured with the same parameters.

Construction of protein models for MD simulations

The structure of PSS1^WT from cryo-EM was modelled in a lipid bilayer. The structure contains the homodimer of PSS1. Within each homodimer, two product PS molecules are in the allosteric binding site. Each monomer also binds two 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) molecules and each active site binds a Ca²⁺ ion. Two dipalmitoyl phosphatidylinositol (PI) molecules are built at the contact interface of the two monomers. The structural model of PSS1 without PS-binding was constructed by removing the coordinate of PS at the allosteric site. Hydrogen atoms were added using PROPKA3⁶⁷. A disulfide bond was constructed between Cys153 and Cys213. The transmembrane domain was then placed into an intact lipid bilayer by replacing overlapping lipid molecules using the CHARMM-GUI⁶⁸. The lipid bilayer consists of upper (cytosol) and lower (extracellular space) leaflets, both of which contain 100% POPC. The protein-membrane system was next solvated in 22.5 Å thickness of water with 0.15 M KCl, resulting in 127 K⁺ ions, 125 Cl⁻ ions, and 48282 water molecules for PSS1 with PS and 130 K⁺ ions, 132 Cl⁻ ions, and 49889 water molecules for PSS1 without PS. The PSS1 model with PS-binding involves 233990 atoms in a tetragonal box of dimension 150.1 Å × 150.1 Å × 111.6 Å. The PSS1 without PS-binding involves 239487 atoms in a tetragonal box of dimension 150.1 Å × 150.1 Å × 114.0 Å.

Molecular dynamics simulations

MD simulations were conducted using AMBER22 software suit⁶⁹. We used CHARMM-GUI to generate required input files^70,71. The AMBER ff14SB⁷² force field was used for the protein part, and the generalized AMBER force field^73,74 was used for the bound molecules PS, PC, and PI with AM1-BCC model to determine the atomic charges⁷⁵. The lipid bilayer and solvent water were modeled with Lipid21⁷⁶ and TIP3P force fields, respectively. The SHAKE algorithm⁷⁷ was used to constrain all hydrogen-containing bonds. The solvated protein-membrane complex was first optimized using the steepest descent method for 2500 steps followed by the conjugate gradient method for another 2500 steps. A constraint of 250 kcal·mol⁻¹·rad⁻² was applied to the dihedral angle defined by C1-C3-C2-O2 and the double bond of each POPC molecule during the geometry optimization. Positional restraints relative to the initial structure were defined for the protein, bound molecules (except Ca²⁺), and all phosphorus atoms of the POPC molecule. The force constants were 10.0 kcal·mol⁻¹·Å⁻² for proteins and bound molecules and 2.5 kcal·mol⁻¹· Å⁻² for phosphorus. Next, the system was equilibrated with molecular dynamics at 300 K. The equilibration followed the default of CHARMM-GUI. The constraints gradually decreased to none during the equilibration. The system was heated from 0 to 300 K and equilibrated for 250 ps under constant volume at 300 K, and further equilibrated at 300 K and 1 atm for 1625 ns. The whole equilibration cost 1875 ps. After equilibration, the production runs for 200 ns and output the trajectories every 100 ps, resulting in a total of 2000 snapshots for each production run. The production was performed with a time step of 2 fs. The Langevin thermostat⁷⁸ and Berendsen barostat⁷⁹ were used to control the temperature and pressure, respectively. Three parallel simulations were conducted for both PSS1 with and without PS. The input topology and structure file, structural restraint files, and parameter files for each MD step were included in the supporting data. All MD trajectories with membrane and water stripped were also included. All MD-related files can be found at https://doi.org/10.5281/zenodo.11404258.

Trajectory analyses

All atomic distance and conformational fluctuation analyses were conducted using the cpptraj utility of AMBER⁸⁰. To quantify the binding conformation at the active site, we calculated the distance between key residues and the Ca²⁺. For Glu197, Glu200, Glu212, and Asp221, we calculated the distance between their carboxylate group (represented by the middle point of the two sidechain carboxylic oxygen atoms) and the Ca²⁺. For residues 197 to 221, we also calculated the distance between each Cα to Ca²⁺. We calculated the distance between the Cα atoms of Phe168 and Trp321 to evaluate the opening of the catalytic core. To investigate the conformational dynamics of PSS1, we calculated the root-mean-square fluctuation (RMSF) for each residue. We included the backbone heavy atoms in the calculation. Conformations of each monomer in each trajectory were first overlayed to generate an average structure. This average structure was then used in the corresponding trajectory to calculate the RMSF for each residue.

The binding free energy analyses were based on the MM-PBSA method enabled by the MMPBSA.py utility of AMBER⁸¹. We used a uniform membrane dielectric constant in a slab-like implicit membrane to model the membrane environment³⁵. The PSS1 dimer with Ca²⁺ ions in the active sites were treated as the receptor. Ligands except the ones of interest, if any, were treated as solvent together with the double-layer membrane, K⁺, Cl⁻, and water molecules. To reduce the computational cost, we evenly sampled 50 out of 2000 snapshots in each trajectory. Other specifics were included in an example input file at https://doi.org/10.5281/zenodo.11404258.

Reproducibility

All enzymatic activity, metabolic and imaging experiments were repeated at least two times on different days. Similar results were obtained.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses of data were performed using GraphPad Prism 9. Quantification methods and tools used are described in each relevant section of the METHOD DETAILS or figure legends.

Supplementary Material

1. Figure S1 Sequence alignment and purification of PSS1, related to Figures 1, 3, 4 and 5.

(A) Sequence alignment of human PSS1 and PSS2 with mouse, chicken, xenopus, and zebrafish PSS1. The transmembrane helices and the residue numbers of human PSS1 are indicated above the protein sequence. The specific residues necessary for calcium binding are indicated by circles, the catalytic histidine is indicated by an asterisk and the cysteines for disulfate bond are indicated by purple squares. The residues necessary for inhibitor binding are indicated by dark blue squares.

(B) Purification of human PSS1^WT and its variants for cryo-EM study and activity assays. Representative Superose 6 increase 10/30 gel-filtration chromatogram of PSS1^WT, PSS1^P269S, PSS1H172A, PSS1F168A, PSS1F168W, PSS1G171W, PSS1G175W and PSS1F313M/S320V in buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 5 mM CaCl₂ and 0.06% Digitonin. The peak fractions used in this study are indicated by red arrows.

2. Figure S2 Data processing of PSS1^WT, related to Figures 1 and 2.

(A) Representative cryo-EM image of PSS1^WT. Scale bar, 40 nm.

(B) Selected 2D class averages of PSS1^WT.

(C) Summary of cryo-EM data processing procedures. FSC curves between two half maps are shown. The mask, which was generated by CryoSPARC for the final refinement, is shown in gray.

(D) Angular distribution of PSS1^WT particles in the final round of 3D refinement using CryoSPARC.

(E) Local resolution of cryo-EM map. Maps are colored according to local resolution, estimated using CryoSPARC.

(F) Cryo-EM map of the transmembrane helices.

3. Figure S3 200 ns MD simulations of PSS1^WT with and without PS, related to Figures 3, 4 and 5.

(A) The binding free energy of PE and PC to PSS1^WT monomer with and without PS from three MD simulations. This is also the case for (B) and (D). Unit: kcal/mol.

(B) The binding free energy of PS to PSS1^WT monomer from three MD simulations. Unit: kcal/mol.

(C) The representative curves showing the distance between the side chain center of the corresponding residue and calcium ions in the structure of PSS1 with (blue) or without PS (green).

(D) The binding free energy of serine to PSS1^P269S monomer from three MD simulations. Unit: kcal/mol.

(E) Distance between C_α atoms of Phe168 and Trp321 in the PSS1^WT two monomers with (blue) or without PS (green).

4. Figure S4 Data processing of PSS1^P269S, related to Figure 3.

(A) Summary of cryo-EM data processing procedures. The mask, which was generated by CryoSPARC for the final refinement, is shown in gray.

(B) Angular distribution of PSS1^P269S particles in the final round of 3D refinement using CryoSPARC. FSC curves between two half maps are shown.

(C) Local resolution of cryo-EM map. Maps are colored according to local resolution, estimated using CryoSPARC.

(D) Cryo-EM map of the transmembrane helices.

(E) The extra density at the lipid entrance in the cryo-EM map of PSS1^P269S.

(F) The extra lipid-like density in the cryo-EM map of PSS1^P269S.

(G) Comparison of the cryo-EM maps of PSS1^P269S and PSS1^WT indicates the flexibility of PH4 in PSS1^P269S. The cryo-EM map of PSS1^WT has been filtered to 3.3-Å.

(H) Comparison of calcium binding sites of PSS1^P269S and AF model (gray). The calcium in the structure of PSS1^P269S is shown as a green sphere. Dashed line represents hydrophilic interactions.

5. Figure S5 Data processing of DS55980254-bound PSS1^WT, related to Figure 5.

(A) Summary of cryo-EM data processing procedures. FSC curves between two half maps are shown. The mask, which was generated by CryoSPARC for the final refinement, is shown in gray.

(B) Angular distribution of DS55980254 bound PSS1^WT particles in the final round of 3D refinement using CryoSPARC.

(C) Local resolution of cryo-EM map. Maps are colored according to local resolution, estimated using CryoSPARC.

(D) Cryo-EM map of the transmembrane helices.

6. Figure S6 Measurement of LDLR expression and LDL uptake by flow cytometry, related to Figure 7.

(A) Measurement of LDLRs by flow cytometry in SV589j cells incubated with LDL. On day 0, the indicated cells were set up in medium A with 5% FCS. After plating for 6 hours, cells were supplemented with DMSO (solvent control) or 1 μM DS, which was kept at a constant concentration in each buffer throughout the experiment. On day 1, cells were switched to cholesterol-depletion medium B. After 16 h, cells then received the above cholesterol-depletion medium containing 50 μg protein/ml of LDL. After 24 h, the cells were harvested by incubation with EDTA, washed, incubated with PE-anti-LDLR, and subjected to flow cytometry (METHOD DETAILS).

(B) Measurement of BODIPY FL-LDL uptake by flow cytometry in SV589j cells. On day 0, the indicated cells were set up in medium A with 5% FCS. After plating for 6 hours, cells were supplemented with DMSO (solvent control) or 1 μM DS, which was kept at a constant concentration in each buffer throughout the experiment. On day 1, cells were switched to cholesterol-depletion medium B. After 16 h, cells then received the above cholesterol-depletion medium containing 50 μg protein/ml of LDL. After 24 h, the cells were treated by 5 μg protein/ml of BODIPY FL-LDL for 2 hours, then the cells were harvested by incubation with EDTA, washed and subjected to flow cytometry (METHOD DETAILS).

7. Figure S7 Phylogenetic tree and working mechanism of MBOATs, related to Figure 4.

(A) Phylogenetic tree of PSSs and MBOATs. Human, mouse, frog, zebrafish, Arabidopsis, yeast and bacteria proteins were used for the comparison. Numbers at the nodes indicates the bootstrap values on neighbor joining analysis. The calculation was performed in MEGA X.

(B) Working mechanisms of PORCN and ACAT1. The catalytic histidine activates the hydroxyl group of the serine of Wnt (PORCN) or cholesterol (ACAT1), which would attack the thioester bond of acyl-CoA substrate, to form the product.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal Anti-Flag antibody	Sigma-Aldrich	Cat# F7425
Rabbit monoclonal Anti-SREBP-2	Sigma-Aldrich	Cat# MABS1988
HRP Rabbit polyclonal to Histone H3 antibody	Abcam	Cat# ab21054
Rabbit Anti-TOM20 antibody	Protein Tech	Cat# 11802-1-AP
Mouse Anti-CREB antibody	Thermo Fisher	Cat# 35-0900
Mouse Anti-EEA1 antibody	Sigma-Aldrich	Cat# E7659
Mouse Anti-GM130 antibody	BD Biosciences	Cat# 610822
Rabbit Anti-Na+/K+ ATPase antibody	Abcam	Cat# ab76020
Rabbit Anti-Calnexin antibody	Bio-techne	Cat# NB100-1965
Mouse Anti-NPC1	Abcam	Cat# 134113
Rabbit Anti-ABCD3	Affinity Biosciences	Cat# DF8315
Mouse Anti-HMGCR	Sigma-Aldrich	Cat# MABS1233
Anti-Mouse IgG, HRP-linked antibody	Cell Signaling Technology	Cat# 7076S
Anti-Rabbit IgG, HRP-linked antibody	Cell Signaling Technology	Cat# 7074S
Bacterial and virus strains
E. coli DH5α Competent Cells	GoldBio	Cat# CC-101-TR
E. coli DH10Bac Competent Cells	Thermo Fisher	Cat# 10361012
Chemicals, peptides, and recombinant proteins
LB broth	Fisher Scientific	Cat# BP1426-500
Sf-900 III SFM medium	Gibco	Cat# 12658027
Freestyle 293 expression medium	Gibco	Cat# 12338018
DMEM/F12 50/50 Mix medium	Corning	Cat# 10-090-CV
DMEM-low glucose (1000 mg/l)	Sigma-Aldrich	Cat# D6046
Fetal Bovine Serum (FBS)	GeminiBio	Cat# 100-500
Filipin III from Streptomyces filipinensis	Sigma-Aldrich	Cat# F4767-1 MG
Newborn calf lipoprotein-deficient serum (LPDS, d<1.215 g/ml)	Goldstein et al.52	N/A
Penicillin-Streptomycin solution (100x)	Corning	Cat# 30-002-CI
HEPES	Gibco	Cat# 11344041
NaCl	Fisher Scientific	Cat# S271
Sodium butyrate	Millipore-Sigma	Cat# 303410
Phenylmethylsulfonyl fluoride (PMSF)	GoldBio	Cat# P-470-25
Leupeptin	Biosynth	Cat# ILP-4041
Lauryl maltose neopentyl glycol (LMNG)	Anatrace	Cat# NG310
Digitonin	Thermo Scientific	Cat# 407560050
Anti-Flag M2 resin	Millipore-Sigma	Cat# A2220
3x Flag peptide	ApexBio	Cat# A6001
14C-Acetate	American Radiolabeled Chemicals	Cat# ARC 0173A
Petroleum Ether	Fisher Chemical	Cat# E139-1
DS55980254	ProbeChem	Cat# PC-49548
Cellfectin II reagent	Gibco	Cat# 10362-100
Phosphate buffered saline with Tween 20 (PBST)	Millipore-Sigma	Cat# 08057-100TAB-F
Benzonase nuclease	Millipore-Sigma	Cat# 70746-3
Mevalonate	Sigma-Aldrich	Cat# M4667
Compactin	Sigma-Aldrich	Cat# M2147
Dimethyl Sulfoxide (DMSO)	Sigma-Aldrich	Cat# D8418
EDTA, sodium, pH 8.0	Thermo Fisher	Cat# 15575020
Hydroxypropyl-β-cyclodextrin	Cyclodextrin Technologies Development, Inc.	Cat# THPB-P
Hank’s Balanced Salt Solution (HBSS)	Gibco	Cat# 14025-092
LDL, human (d1.019-1.063 g/ml)	Goldstein et al.52	N/A
Lipoprotein, low density from human plasma	Sigma-Aldrich	Cat# SAE0053-10MG
BODIPY FL-LDL	Invitrogen	Cat# L3483
PE/R-Phycoerythrin Conjugation Kit	Abcam	Cat# ab102918
N-Acetyl-Leu-Leu-norLeucinal (ALLN)	Calbiochem	Cat# 208719
1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC)	Avanti Polar Lipids	Cat# 850457C-25mg
Phosphate Buffered Saline (PBS)	Corning	Cat# 21-031-CV
QuikChange II XL Site-Directed Mutagenesis Kit	Agilent	Cat# 200522
Sodium Dodecyl Sulfate (SDS)	Sigma-Aldrich	Cat# 71736
Bolt™ Bis-Tris Plus Mini Protein Gels, 4–12%, 1.0 mm, WedgeWell™ format	Invitrogen	Cat# NW04127B0X
Novex™ Tris-Glycine Mini Protein Gels, 4–20%, 1.0 mm, WedgeWell™ format	Invitrogen	Cat# XP04205B0X
U18666A	Sigma-Aldrich	Cat# U3633
Hexane	Sigma-Aldrich	Cat# 293253
Acetonitrile	Sigma-Aldrich	Cat# 34998
Ammonium Acetate	Sigma-Aldrich	Cat# 421311
Splash Lipidomix standard solution	Avanti Polar Lipids	Cat# 330707
Critical commercial assays
Malachite Green Phosphate Assay Kit	Millipore-Sigma	Cat# MAK307-1KT
Deposited data
Cryo-EM structure of PSS1WT	This paper	PDB: 9B4E; EMDB: EMD-44178
Cryo-EM structure of PSS1P269S	This paper	PDB: 9B4F; EMDB: EMD-44179
Cryo-EM structure of DS55980254-bound PSS1WT	This paper	PDB: 9B4G; EMDB: EMD-44180
Experimental models: Cell lines
Sf9 insect cell	ATCC	Cat# CRL-1711; RRID: CVCL_0549
HEK-293S GnTI− cell	ATCC	Cat# CRL-3022; RRID: CVCL_A785
CHO-K1 cell	ATCC	Cat# CCL-61; RRID: CVCL_0214
Ptdss1−/− CHO-K1 cell	Trinh et al.27	N/A
SV589j	Trinh et al.27	N/A
Recombinant DNA
pEG-BacMam-PSS1	This paper	N/A
Software and algorithms
Image Studio Lite v5.2	LI-COR	N/A
MACSQuantify Software 2.13	Meltenyi Biotec	N/A
MotionCor2	Zheng et al.53	https://emcore.ucsf.edu/ucsf-motioncor2
RELION 3.1	Zivanov et al.54	https://www3.mrc-lmb.cam.ac.uk/relion/
CTFFIND4	Rohou and Grigorieff55	http://grigoriefflab.janelia.org/ctffind4
CryoSPARC v3.3.0 or v4.4.1	Punjani et al.56	https://cryosparc.com/
SPHIRE-crYOLO v1.7.6 or 1.9.3	Wagner et al.57	http://sphire.mpg.de/
Coot 0.8.8	Emsley et al.58	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Phenix 1.17	Adams et al.59	http://www.phenix-online.org/
UCSF ChimeraX 1.5	Pettersen et al61	https://www.cgl.ucsf.edu/chimerax/
PyMOL 2.3.4	Schrodinger	https://pymol.org/2/
Prism 9	GraphPad	https://www.graphpad.com/
PROPKA3	Olsson et al.67	https://github.com/jensengroup/propka
CHARMM	Jo et al.68	https://www.charmm.org/
AMBER22	The Amber Project	https://ambermd.org/AmberMD.php
Other
R1.2/1.3 400 mesh Au holey carbon grids	Quantifoil	Cat# 1210627
Superose 6 Increase 10/300 GL column	Cytiva	Cat# 29091596
Milli-Q IQ7000 Ultrapure Water System	Sigma-Aldrich	N/A

Highlights:

• The catalytic mechanism of PSS1 is akin to the working principle of MBOATs.

• Phosphatidylserine and DS55980254 are allosteric negative modulators of PSS1.

• PSS1 deficiency lowers ER cholesterol levels, activating the SREBP pathway.

• Specific PSS1 inhibitors may promote LDLR expression and LDL uptake in cells.

Inclusion and diversity

We support inclusive, diverse, and equitable conduct of research.