P4-ATPase control over phosphoinositide membrane asymmetry and neomycin resistance

Abstract

Neomycin, an aminoglycoside antibiotic, has robust antibacterial properties, yet its clinical utility is curtailed by its nephrotoxicity and ototoxicity. The mechanism by which the polycationic neomycin enters specific eukaryotic cell types remains poorly understood. In budding yeast, NEO1 is required for neomycin resistance and encodes a phospholipid flippase that establishes membrane asymmetry. Here, we show that mutations altering Neo1 substrate recognition cause neomycin hypersensitivity by exposing phosphatidylinositol-4-phosphate (PI4P) in the plasma membrane extracellular leaflet. Human cells also expose extracellular PI4P upon knockdown of ATP9A, a Neo1 ortholog and ATP9A expression level correlates to neomycin sensitivity. In yeast, the extracellular PI4P is initially produced in the cytosolic leaflet of the plasma membrane and then delivered by Osh6-dependent nonvesicular transport to the endoplasmic reticulum (ER). Here, a portion of PI4P escapes degradation by the Sac1 phosphatase by entering the ER lumenal leaflet. COPII vesicles transport lumenal PI4P to the Golgi where Neo1 flips this substrate back to the cytosolic leaflet. Cryo-EM reveals that PI4P binds Neo1 within the substrate translocation pathway. Loss of Neo1 activity in the Golgi allows secretion of extracellular PI4P, which serves as a neomycin receptor and facilitates its endocytic uptake. These findings unveil novel mechanisms of aminoglycoside sensitivity and phosphoinositide homeostasis, with important implications for signaling by extracellular phosphoinositides.

Main

Neomycin was first discovered in the 1940s and has been widely used for decades to treat bacterial infections^1,2. It is derived from Streptomyces fradiae, a species of soil bacteria, and is known for its broad-spectrum antibacterial activity, particularly against gram-negative bacteria¹. Neomycin exerts its antibacterial effects by binding to bacterial ribosomes and inhibiting protein synthesis, and it is also capable of disrupting translation in eukaryotic cells^3,4. However, most mammalian cell types are resistant to neomycin because the polycationic antibiotic cannot penetrate into the cell^2,5. Despite this general resistance, aminoglycosides like neomycin can cause various toxic effects on mammalian tissues. These include convulsions following intracisternal injection, inhibition of neuromuscular and ganglionic transmission, as well as nephrotoxicity and ototoxicity affecting the cochlea and vestibular system⁶. Toxicity of neomycin is attributed to its ability to enter particular cell types, such as the hair cells of the inner ear^7–9. The precise biochemical mechanisms behind the neomycin sensitivity of specific eukaryotic cells remain unclear.

Neomycin specifically binds to PI4P and phosphatidylinositol 4,5-bisphosphate (PI(4,5)P₂)^8,10–12, lipids that play crucial roles in signal transduction, membrane trafficking, and cytoskeletal dynamics through their ability to recruit and activate various effectors¹³. PI kinases synthesize phosphoinositides (PIPs) in the cytosolic leaflet of the Golgi, plasma membrane, and endolysosomal system¹⁴. Most other phospholipids in the cell are produced in the cytosolic leaflet of the ER. An ER scramblase allows rapid and energy-independent flip-flop, so newly synthesized phospholipids will populate both membrane leaflets^15–18. As the membrane flows through the Golgi to the plasma membrane, energy-dependent phospholipid flippases in the P4-ATPase family transport phosphatidylserine (PS) and phosphatidylethanolamine (PE) from the lumenal or extracellular leaflet to the cytosolic leaflet, while leaving phosphatidylcholine and most sphingolipids behind¹⁹. Thus, the plasma membrane is asymmetric with PS/PE and PIPs enriched in the cytosolic leaflet, and phosphatidylcholine and sphingolipids populating the extracellular leaflet.

P4-ATPases are highly conserved and budding yeast S. cerevisiae expresses five P4-ATPases Neo1, Drs2, Dnf1, Dnf2, and Dnf3. These lipid flippases localize to different cellular compartments and transport different lipid substrates^20–23. Neo1 and Drs2 are Golgi-localized PS/PE flippases, orthologous to human ATP9A/B and ATP8A1/2, respectively, required to establish plasma membrane PS/PE asymmetry^21,24–26. PIPs were not considered P4-ATPase substrates because they are not synthesized at the ER and presumably would not have access to the lumenal leaflet of the secretory pathway. However, PI4P can be delivered to the ER by lipid transfer proteins in the oxysterol-binding protein family, where it is exchanged for PS or cholesterol/ergosterol^27–30, raising the possibility that PI4P can flip-flop to the ER lumen and enter the secretory pathway.

This study reveals that PI4P enters the secretory pathway’s lumen, is transported to the Golgi by COPII-dependent vesicular transport and is flipped by Neo1 back to the cytosolic leaflet. Abrogation of Neo1’s ability to bind PI4P allows transport of this lipid to the extracellular leaflet of the plasma membrane, where it binds neomycin. We provide structural and biochemical evidence that PI4P is a transport substrate of Neo1, and evidence that this pathway is conserved in human cells.

Results

Neo1 substrate translocation pathway mutations cause neomycin sensitivity.

NEO1 was originally discovered in a screen for overexpressed genes that conferred neomycin resistance to wild-type (WT) Saccharomyces cerevisiae³¹. Disruption of NEO1 is lethal but temperature-conditional mutants have been generated (neo1–1), and these cells are extremely sensitive to neomycin at the permissive growth temperature³². In addition, we previously isolated separation-of-function mutations in the Neo1 substrate translocation pathway (Fig. 1a boxed region) that cause aberrant exposure of only PS, only PE, or both substrates in the extracellular leaflet of the plasma membrane (Fig. 1a, b)²⁴. To ascertain if neomycin sensitivity is due to loss of PS or PE asymmetry, we examined the growth of these neo1 mutants on solid media with or without neomycin (Extended Data Fig. 1). A subset of the mutants including S221L, R247L, S452Q, and P456A were sensitive to neomycin and displayed an intermediate growth phenotype relative to wild-type (WT) cells and the neo1–1 mutant (Fig. 1c), which exposes PS and PE¹⁹. We also observed growth inhibition of these neomycin-sensitive (neo1^s) mutants in liquid media containing as little as 50 – 100 ug/ml neomycin, whereas WT cells are resistant to more than 1000 ug/ml neomycin (Fig. 1d). Curiously, the pattern of neomycin sensitivity did not correlate well to loss of PS or PE asymmetry. For example, the Q209G and S452Q Neo1 mutants both expose PS, but Q209G is neomycin resistant, while S452Q is neomycin sensitive (Fig. 1b-d and²⁴). Similarly, Q193A exposes PE but is resistant to neomycin, while the PE-exposing S221L and R247L mutants are neomycin-sensitive. To further address the relationship between PS/PE asymmetry and neomycin sensitivity, we examined other flippase mutants (drs2Δ, dnf1,2Δ, and dnf3Δ) that exhibit a loss of PS/PE membrane asymmetry and found that all are neomycin resistant (Fig. 1e). Dop1 and Mon2 are large cytosolic proteins that interact with Neo1 and are required for Neo1 activity^33,34. Consistently, dop1–1 and mon2Δ mutants exhibit the same degree of neomycin sensitivity as neo1–1 (Fig. 1e,f). Consequently, neomycin sensitivity does not appear to result from loss of PS/PE asymmetry and is a unique phenotype of neo1 relative to the other P4-ATPase mutants. Therefore, we hypothesized that Neo1 transports an unknown lipid substrate linked to neomycin sensitivity.

Figure 1: Separation-of-function mutations in Neo1 that cause neomycin sensitivity do not correlate with loss of PS or PE asymmetry.

a, The structure of S. cerevisiae Neo1 with the major domains displayed in different colors (PDB ID 9BS1, this study). The dashed box is the substrate translocation pathway enlarged in (b). b, Transmembrane (TM) segments TM1, TM2, and TM4 form the substrate transport pathway of Neo1, and mutation of residues shown causes exposure of PS and/or PE on the extracellular leaflet of the plasma membrane. c, Neomycin sensitivity of Neo1 substrate transport pathway mutants at 26 °C on YPD and YPD Neomycin (500 μg/ml) plates, relative to WT and neo1–1 (a temperature-sensitive strain displaying a loss of PS and PE membrane asymmetry at this permissive growth temperature). d, Neomycin sensitivity in liquid culture of strains expressing the indicated Neo1 variants. The data represent growth relative to WT cells in the absence of the drug. Two-way ANOVA was performed to test the variance and comparisons with WT Neo1 were made with Tukey’s post hoc analysis (n = 3, ±standard deviation (SD). ****p < 0.000,ns represents non-significant .e, f, Neomycin sensitivity of S. cerevisiae flippase mutants (neo1–1, drs2Δ, dnf1,2Δ, dnf3Δ) and strains deficient for Neo1 regulators (dop1–1, mon2Δ). The data represent growth relative to WT cells in absence of the drug. Two-way ANOVA was performed to test the variance and comparisons with WT Neo1 were made with Tukey’s post hoc analysis (n = 3, ±standard deviation (SD). ns represents non-significant.

Neomycin sensitive mutants expose PI4P in the plasma membrane extracellular leaflet

Neomycin binds tightly to phosphoinositides (PIPs), but the biological significance of this interaction is unclear^8,10–12. Budding yeast express a phosphatidylinositol 4-kinase (PI 4-kinase, Stt4) and a phosphatidylinositol 4-phosphate 5-kinase (PI4P-5-kinase, Mss4) to produce PI4P and (PI(4,5)P₂), respectively, in the cytosolic leaflet of the plasma membrane (Fig. 2a)^35–37. To test whether neo1 mutants lose asymmetry of PI4P and/or PI(4,5)P₂ and expose PIPs in the extracellular leaflet, we probed cells with recombinantly purified GFP-tagged P4C domain (PI4P binding domain from SidC), specifically binding to PI4P, and the PLC_PH, pleckstrin homology (PH) domain of phospholipase C delta, which binds to PI(4,5)P₂^38–40 (Extended Data Fig. 2g). Intriguingly, at the non-permissive growth temperature (38 °C), only GFP-SidC_P4C bound to neo1–1 spheroplasts (cells lacking the cell wall), with GFP-PLC_PH failing to bind. Notably, the GFP-SidC_P4C fluorescence predominantly localized to the cell periphery, exhibiting no discernible intracellular fluorescence (Fig. 2b). This distinct pattern of GFP-SidC_P4C binding to the cell surface was observed in all other neomycin-sensitive Neo1 mutants as well. In contrast, neither probe bound to the neomycin-resistant, PS-exposing Q209G or PE-exposing Q193A mutants (Fig. 2c, Extended Data Fig. 2e).

Figure 2: neo1 mutants cells expose PI4P in the plasma membrane extracellular leaflet.

a, Phosphoinositide synthesis in the cytoplasmic leaflet of the plasma membrane and the use of recombinantly purified GFP-SidC_P4C and GFP-PH_PLC to probe for external PI4P and PI(4,5)P₂, respectively. b, Inactivation of neo1–1 at the nonpermissive temperature (38 °C) causes PI4P exposure. Fluorescent images were inverted to black signal on white background for clarity. The right panel is the fluorescence signal intensity of the GFP-probe and left panel shows the DIC panel to display yeast cells. Scale bar = 2 μm c, Quantification of GFP-SidC_P4C and GFP-PLC_PH binding to the cell surface of cells expressing Neo1 variants. One-way ANOVA was performed to test the variance and comparisons with WT Neo1 were made with Tukey’s post hoc analysis (n = 20, ±standard deviation (SD). P<0.0001 is ****. d, The GFP-SidC_P4C(R652Q) PI4P binding defective mutant fails to bind neo1–1 cells. One-way ANOVA was performed to test the variance and comparisons with GFP-SidC_P4C were made with Dunnett’s multiple comparisons test (n = 20, ±standard deviation (SD). P<0.0001 is **** . e, stt4–4 suppresses the neomycin sensitivity of neo1S221L or neo1S452Q at 33°C. g, pik1–83 fails to suppress the neomycin sensitivity of neo1S221L or neo1S452Q at 34°C. The data represents growth relative to cells in the absence of the drug. Two-way ANOVA was performed to test the variance and comparisons were made with Tukey’s post hoc analysis (n = 3, ±standard deviation (SD). **p < 0.01, ns represents non-significant.

To validate the specificity of GFP-SidC_P4C for PI4P, we used a GFP-SidC_P4C(R652Q) variant with impaired PI4P recognition³⁸, which failed to bind to neo1–1 cells (Fig. 2d, Extended Data Fig. 2f). We co-stained Neo1-S221L cells with GFP-SidC_P4C and propidium iodide to distinguish between live and dead cells. Most of the cells in the population were stained with GFP-SidC_P4C, with very few instances of GFP-SidC_P4C-positive cells co-staining with propidium iodide (Extended Data Fig. 2c,d). Thus, cells exposing PI4P were alive and capable of maintaining the permeability barrier to propidium iodide. Importantly, dop1–1 and mon2Δ cells, but none of the other flippase mutants, exposed PI4P (Extended Data Fig. 2a,b). Thus, a strong correlation exists between mutants that expose PI4P in the outer leaflet and neomycin sensitivity. These findings also suggest that Neo1 is a PI4P flippase that normally prevents exposure of PI4P in the extracellular leaflet.

Extracellular PI4P is initially produced in the cytosolic leaflet of the plasma membrane.

PI4P is synthesized by the PI 4-kinase Stt4 at the plasma membrane, and Pik1 at the late Golgi^36,41. To explore the source of extracellular PI4P in neo1 mutants, we created neo1 stt4–4 and neo1 pik1–83 double mutants. We assayed these double mutants for neomycin sensitivity at a semipermissive growth temperature where the stt4–4 (33 °C) and pik1–83 (34 °C) alleles would be partially defective (Extended Data Fig. 3a)^36,37,42,43. STT4 and PIK1 are essential genes, so complete inactivation of these temperature-sensitive alleles would prevent growth. Attenuating PI4P synthesis with stt4–4 suppressed neomycin sensitivity caused by neo1-S221L or neo1-S452Q, as indicated by better growth of the neo1 stt4–4 double mutants relative to the neo1 single mutant in the presence of neomycin (Fig. 2e). In contrast, the pik1 mutation had no discernible effect on neomycin sensitivity of neo1 mutants (Fig. 2f). Moreover, a PI4P 5-kinase mutation (mss4–102³⁷) did not significantly influence neomycin sensitivity in the neo1 background (Extended Data Fig. 3b,c). Collectively, these results indicate that neo1 sensitivity to neomycin is primarily attributed to PI4P synthesized by Stt4 at the plasma membrane. No loss of PI(4,5)P₂ asymmetry was detected in the neo1 mutants, nor did we detect a significant contribution of this lipid to neomycin sensitivity of yeast (Extended Data Fig. 3d).

ATP9A depletion causes PI4P exposure in the plasma membrane extracellular leaflet

ATP9A and ATP9B are the mammalian orthologs of Neo1⁴⁴, and sequence comparison indicates that most of the Neo1 residues crucial for establishing PI4P asymmetry are conserved in these mammalian flippases (Extended Data Fig. 4a). Neomycin is poorly absorbed by most human cells and is toxic to only a few tissues, such as kidney cells and inner ear hair cells^7,9. ATP9A is typically more highly expressed than ATP9B in human tissues, and both are weakly expressed in the kidney (Extended Data Fig. 4c,d). To explore a potential link between ATP9A expression and neomycin sensitivity, we examined ATP9A protein expression levels in HeLa and HEK293 (kidney) cells.

Interestingly, ATP9A is highly expressed in HeLa cells but is expressed at low levels in HEK293 cells (Extended Data Fig. 4b,e,f). HeLa cells were resistant to 5000 μg/ml of neomycin, while HEK293 cells displayed moderate sensitivity (Fig. 3a,d). siRNA-mediated knockdown of ATP9A in HeLa cells rendered them sensitive to neomycin. Similarly, ATP9A depletion in HEK293 cells further increased their neomycin sensitivity (Fig. 3a,d).

Figure 3: ATP9A knockdown cells are neomycin sensitive and expose PI4P in the plasma membrane extracellular leaflet.

a, Neomycin sensitivity of wild type HeLa cells and ATP9A knockdown cells. The data represent % viable cells relative to the viable cells in absence of the neomycin. One-way ANOVA was performed to test the variance and comparisons with control cells were made with Dunnett’s multiple comparisons test (n = 3, ±standard deviation (SD). **p < 0.01. b, ATP9A knockdown HeLa cells expose PI4P in extracellular leaflets. Hoechst dye was used as nuclear stain. Scale bar = 10 μm c, Quantification of GFP-SidC_P4C and GFP-PLC_PH binding to the cell surface of ATP9A knockdown cells. One-way ANOVA was performed to test the variance and comparisons with control siRNA treated cells were made with Tukey’s post hoc analysis (n = 20, ±standard deviation (SD). P<0.0001 is ****. d, Neomycin sensitivity of wild type HEK293 cells and ATP9A knockdown cells. The data represent % viable cells relative to the viable cells in absence of the neomycin. One-way ANOVA was performed to test the variance and comparisons with WT cells were made with Dunnett’s multiple comparisons test (n = 3, ±standard deviation (SD). **p < 0.01. e, ATP9A knockdown HEK293 cells expose PI4P in extracellular leaflets. Scale bar = 10 μm f, Quantification of GFP-SidC_P4C and GFP-PLC_PH binding to the cell surface of ATP9A knockdown cells. One-way ANOVA was performed to test the variance and comparisons with control siRNA treated cells were made with Tukey’s post hoc analysis (n = 20, ±standard deviation (SD). P<0.0001 is ****.

Next, we tested whether siRNA-mediated knockdown of ATP9A in HeLa and HEK293 cells would cause exposure of extracellular PI4P. As expected, GFP-SidC_P4C bound to the extracellular leaflet of ATP9A knockdown cells, and no GFP-PLC_PH binding was detected (Fig. 3b,c,e,f). In contrast, HeLa cells that were fixed and permeabilized stained efficiently with both probes (Extended Data Fig. 2g). Thus, ATP9A knockdown in human cells leads to the exposure of PI4P on the extracellular leaflet. These results indicate that ATP9A plays a crucial role in maintaining PI4P asymmetry and neomycin resistance and suggest that the PI4P flippase activity of Neo1 and ATP9A is conserved from yeast to mammals.

Cryo-EM captures PI4P at a Neo1 substrate exit site

ATP hydrolysis by P-type ATPases is typically stimulated by the transported substrate^19,45. Therefore, we examined whether PI4P can stimulate ATP hydrolysis by Neo1. Purified Neo1 was incubated with PE (a known substrate)^24,46 or PI4P, revealing a significant stimulation of ATPase activity by both lipids (Fig. 4a). Another yeast P4-ATPase, Drs2, is allosterically activated by PI4P binding interactions near the C-terminus, outside of the substrate translocation pathway. For Drs2, the addition of substrate (PS) or PI4P alone minimally activates ATPase activity, while the combination of both lipids is required for full activation^47–49. In contrast, the combination of PE and PI4P provided no additional stimulation of Neo1 ATPase activity compared to PE alone (Fig. 4a). This result suggests that PI4P is a Neo1 transport substrate rather than an allosteric activator. We also tested PI4P stimulation of the Neo1 mutants. The S221L and S452Q entry gate mutants surprisingly enhanced the PI4P-stimulated ATPase activity, and the exit gate mutants R247L and P456A reduced this activity (Fig. 4b). This is an unexpected result because these four mutations all cause a similar hypersensitivity to neomycin and loss of PI4P asymmetry (Fig. 1c,2c).

Figure 4: Cryo-EM structure of Neo1-PI4P and comparison of PI4P binding in Neo1 and Drs2.

a, ATP hydrolysis by purified WT Neo1 with or without 0.02 mM PI4P or 0.1 mM PE (substrate) or both PI4P+PE. One-way ANOVA was performed to test the variance and comparisons with no substrate or PE were made with Dunnett’s multiple comparisons test (n = 3, ±standard deviation (SD****p < 0.0001, ns-not significant). b, ATP hydrolysis activity of WT Neo1 and Neo1 mutants at increasing PI4P concentration. Data points represent the mean ± SD in triplicate. c, Global views of Neo1-PI4P in E2P state with the 4 key residues in substrate transport pathway shown in sticks and PI4P shown in teal spheres. d, Middle: the structure of Neo1-PI4P is superimposed with Drs2-PI4P (gray, PDB ID 6ROJ) and shown in cartoon (RMSD = 4.0 Å). Note the different binding site of PI4P as an activator in Drs2 (gray spheres) and as a substrate in Neo1 (teal spheres). The PI4P binding site residues in Drs2 are highlighted in red box. Left: enlarged view of the PI4P binding site in Drs2. Right: enlarged view of PI4P binding site in Neo1. PI4P and interacting residues are shown in sticks. Hydrogen bonds between the PI4P headgroup and flippase residues are labeled by dashed black lines. The EM density of PI4P in Neo1 is superimposed on the atomic model and shown as a transparent surface. e, Mutation in PI4P binding residue causes neomycin sensitivity. HE: H472 E475. The data represents growth relative to WT cells in absence of the drug. Two-way ANOVA was performed to test the variance and comparisons with WT Neo1 were made with Tukey’s post hoc analysis (n = 3, ±standard deviation (SD). ****p < 0.0001.

Major lipid substrates have been observed structurally either at the flippase substrate entry site (near the extracellular or lumenal leaflet, pre-flipping) or at the substrate exit site (near the cytosolic leaflet, post-flipping) in the lipid transport pathway^45,50–52. To investigate how PI4P binds in Neo1, we reconstituted the purified Neo1 in peptidiscs with added PI4P^53,54, trapped the flippase by BeF₃⁻ in the substrate binding E2P state⁵⁰, and determined a cryo-EM structure of the Neo1-PI4P complex at an overall resolution of 3.7 Å (Extended Data Fig. 5 and Extended Data Table 1). We observed a lipid-like EM density that fits well with a modeled PI4P (Fig. 4c). PI4P is at the substrate exit site, representing a post-flipping PI4P in Neo1. Interestingly, the lipid entry site is closed, similar to the closed entry site observed in the Neo1 E2P structure previously determined in detergents (Extended Data Fig. 5). The headgroup of PI4P resides in a large cavity next to TM4, with the terminal phosphate forming hydrogen bonds with a pair of histidine residues (His472 and His476) (Fig. 4d, Neo1-PI4P). Glu475 forms two hydrogen bonds with the PI4P headgroup to stabilize the substrate further. We substituted Glu475, His472, and His476 with alanine, and none of the single mutants showed neomycin sensitivity (Fig. 4e). However, the H472A E475A and H472G E475G double mutants (EH-AA and EH-GG) displayed neomycin sensitivity similar to the Neo1S221L mutants. Thus, the PI4P binding site observed by cryo-EM is important for neomycin resistance.

PI4P binds in the same exit site position and orientation observed for substrate lipid bound in the Dnf1-Lem3 flippase (Extended Data Fig. 6a,b). Thus, the PI4P binding site in Neo1 is within the substrate translocation pathway and is on the opposite side of the membrane domain relative to the PI4P binding site in Drs2 (Fig. 4c). Remarkably, despite the different functions (substrate vs regulatory) and binding sites, the PI4P headgroup is similarly coordinated in Drs2 and Neo1. In Drs2, the binding involves hydrogen bonding with a histidine, a lysine, and a tyrosine (Fig. 4c), but with additional H-bonds between the linking phosphate and a duet of tryptophan and arginine residues from TM10. In contrast, the PI4P binding is weaker in Neo1, perhaps evolved for facile substrate release into the membrane (Fig. 4d). The sequence comparison of PI4P binding site suggests conserved positively charged residues or aromatic residues in other flippases (Extended Data Fig. 7a,b,c). In sum, these data provide biochemical and structural evidence that PI4P is a translocation substrate of Neo1 rather than an allosteric regulator.

Non-vesicular transport of PI4P to the ER and vesicular transport through the Golgi is required for extracellular PI4P exposure.

Neo1 localizes to the Golgi complex, raising the question of how it controls the asymmetry of PI4P synthesized in the cytosolic leaflet of the plasma membrane. To define the mechanism by which PI4P synthesized in the cytosolic leaflet is exposed on the extracellular leaflet of the plasma membrane, we tested a hypothesis based on the unique metabolism of this phosphoinositide proposed to occur at ER-plasma membrane contact sites. After synthesis at the plasma membrane, PI4P can be transported to the endoplasmic reticulum (ER) cytosolic leaflet via the lipid transfer protein Osh6, which in turn acquires phosphatidylserine (PS) from the ER and transports it back to the plasma membrane (Fig. 5a, Wild type)^27–30. PI4P delivered to the ER was presumed to be completely degraded to phosphatidylinositol (PI) by the ER-localized Sac1 phosphatase^27,55,56. However, any lipids present at the ER can undergo rapid flip-flop through an ER scramblase and equilibrate between the lumenal and cytosolic leaflets^17,18,57,58. Indeed, lumenal PI4P has been detected by freeze-fracture electron microscopy studies of the yeast ER, particularly in sac1Δ cells⁵⁹. Therefore, we hypothesized that PI4P could reach the ER lumenal leaflet and travel by vesicular transport to the Golgi, where the Golgi-localized Neo1 would flip PI4P from the lumenal leaflet to the cytosolic leaflet in wild-type cells. The inability of Neo1 mutants to recognize PI4P would then allow the incorporation of PI4P into exocytic vesicles budding from the Golgi, finally resulting in PI4P exposure on the extracellular leaflet of the plasma membrane where it would be accessible to neomycin (Fig. 5a, neo1 mutant).

Figure 5: Non-vesicular and vesicular transport is required for PI4P exposure.

a, A model for the trafficking of PI4P and the role of Neo1 in maintaining PI4P membrane asymmetry. PM: Plasma membrane, ER: Endoplasmic Reticulum, PS: Phosphatidylserine b, sac1Δ cells are neomycin sensitive. c, Overexpression of wild-type NEO1 (2μ NEO1) suppresses sac1Δ neomycin sensitivity. d, sac1Δ cells expose extracellular PI4P and overexpression of NEO1 suppresses PI4P exposure. One-way ANOVA was performed to test the variance, and comparisons with were made with Tukey’s multiple comparisons test (n = 20, ±standard deviation (SD). P<0.001 is ***, P<0.01 is **. e, Deletion of OSH6 (osh6Δ) reduces neomycin sensitivity of neo1 mutants. f, Reducing the rate of vesicular transport with sec23–1, sec12–4, sec18–1, or sec14–1 suppresses neomycin sensitivity of neo1–1 at 26 °C. All the comparisons were calculated via a One-Way ANOVA followed by Tukey’s multiple comparisons test. Colors represent comparison to WT, pairwise comparisons shown with brackets and asterisk. The data represent growth relative to WT cells in absence of the drug (n = 3, ±standard deviation (SD). **p < 0.01, ***p < 0.001, ****p < 0.0001. For b, c, e, and f, cells were grown at 26°C.

To test this hypothesis, we first examined sac1Δ cells, which exhibit elevated levels of PI4P at the ER, including the lumenal leaflet^55,59. Consistently, sac1Δ cells displayed neomycin sensitivity similar to neo1 mutant cells and exposed PI4P in the extracellular leaflet (Fig. 5b-d). However, the extreme sensitivity of sac1Δ to neomycin was unexpected because these cells are expressing wild-type Neo1. This result implies that there is insufficient Neo1 to handle the higher flux of PI4P flowing from the ER to the Golgi lumen in sac1Δ cells. To explore this further, we overexpressed wild-type NEO1 in sac1Δ cells (sac1Δ/2μ NEO1), which restored neomycin resistance and reduced PI4P exposure to near wild-type levels (Fig. 5c,d, Extended Data Fig. 8a). Thus, the combined activities of Sac1 and Neo1 are required to prevent extracellular exposure of PI4P. We attempted to produce sac1Δ neo1 double mutants to assess their combined deficit on neomycin sensitivity but the double mutants were either inviable or grew very poorly (Extended Data Fig. 8b). This synthetic lethal genetic interaction supports the importance of Neo1 to PI4P metabolism.

To probe the role of non-vesicular transport by Osh lipid transfer proteins in PI4P exposure and neomycin sensitivity, we created double mutants of neo1–1 with osh1Δ - osh7Δ. Interestingly, only the osh6Δ mutation suppressed neomycin sensitivity of neo1–1, neo1-S221L, and neo1-S452Q (Fig. 5e, Extended Data Fig. 8c,). Osh6 exchanges PI4P from the plasma membrane to the ER cytosol, swapping it with PS, which is then returned to the plasma membrane. Osh4 is the major PI4P/ergosterol transport protein acting at the Golgi where Neo1 and Pik1 are localized^28,56,60, and surprisingly osh4Δ did not influence the neomycin sensitivity of neo1–1. These results indicate that PI4P synthesized by Stt4 and transported from the plasma membrane to the ER via the lipid transfer protein Osh6 is crucial for PI4P exposure.

To investigate the involvement of vesicular transport in PI4P exposure and neomycin sensitivity, we generated double mutants of neo1–1 with temperature-sensitive mutations that perturb COPII vesicle budding from the ER (sec23 and sec12)^61,62, fusion with the Golgi (sec18–1)^63,64, or budding of exocytic vesicles from the Golgi (sec14–1)^65,66. At a permissive growth temperature where protein transport is partially perturbed, the sec mutations either completely (sec23–1 and sec14–1) or partially (sec12–4, sec18–1) suppressed neomycin sensitivity of neo1–1, indicating that vesicular transport from the ER through the Golgi to the plasma membrane is required for PI4P exposure (Fig. 5f).

Phosphatidylinositol 4-phosphate (PI4P) serves as a receptor for neomycin cellular entry.

To address how PI4P exposure in the extracellular leaflet confers neomycin sensitivity, we covalently conjugated neomycin with Texas Red (Neo-TR) to visualize how neomycin is taken up (Fig. 6a). Strikingly, there was no significant entry or binding of Neo-TR observed in WT Neo1 cells or the neomycin-resistant Neo1-Q193A and Neo1-Q209G mutants. Conversely, Neo1 mutants that expose PI4P in the extracellular leaflet—namely, S221L, R247L, S452Q, and P456A cells—exhibited substantial binding of Neo-TR to the plasma membrane and uptake of the probe to intracellular compartments (Fig. 6b,c). This Neo-TR uptake pattern aligns seamlessly with the exposure of PI4P. Neo-TR localized to the plasma membrane, intracellular punctate structures marked by the late Golgi protein Sec7 and/or endosomal protein Tlg1, and the vacuolar limiting membrane (Fig. 6b,c, Extended Data Fig. 9a,b). This localization pattern implies that extracellular PI4P serves as a binding site for neomycin, and once bound to the plasma membrane, neomycin enters the cells through endocytosis, ultimately reaching the vacuole.

Figure 6: Neomycin binds to exposed PI4P and is endocytosed.

a, Use of neomycin-Texas Red (Neo-TR) to probe PI4P exposed on the extracellular leaflet of the membrane. b, Neo-TR binds neomycin-sensitive neo1 mutants. Scale bar = 2 μm. c, Quantification of mean fluorescence intensity of cells stained with Neo-TR. One-way ANOVA was performed to test the variance and comparisons with WT Neo1 were made with Dunnett’s multiple comparisons test (n = 20, ±standard deviation (SD) ****p < 0.0001). d, LatA blocks Neo-TR internalization. e, Quantification of Neo-TR fluorescence intensity at the PM. For all quantification, data from ~20 cells from three independent experiments were obtained and analyzed. One-way ANOVA was performed to test the variance and comparisons with different time points were made with Tukey’s multiple comparisons test (n = 20, ±standard deviation (SD) ****p < 0.0001) Scale bar = 2 μm. f, Neo-TR binds to the ATP9A knockdown HeLa cells. Hoechst dye was used as a nuclear stain. g, Quantification of total Neo-TR fluorescence intensity of cells (n=20) Scale Bar = 10 μm. h, Neomycin-sensitive HEK293 and ATP9A knockdown cells bind Neo-TR. Scale Bar = 10 μm. i, Quantification of total Neo-TR fluorescence intensity (n=20). One-way ANOVA was performed to test the variance and comparisons with control cells were made with Dunnett’s multiple comparisons test (n = 20, ±standard deviation (SD) ****p < 0.0001).

To further investigate the role of endocytosis in Neo-TR entry, we used the endocytosis inhibitor latrunculin A (LatA), known for impeding F-actin formation⁶⁷. Cells were treated with LatA for varying durations, and Neo-TR uptake was measured (Fig. 6d,e). As described earlier, WT Neo1 cells did not exhibit binding to Neo1-TR, showing no fluorescence. In the case of Neo1S221L mutant cells that had not been treated with LatA (t=0), Neo-TR was primarily localized to the plasma membrane with a few intracellular punctate structures. As the duration of LatA treatment progressed to 30 mins and 60 mins, there was an increase in Neo-TR localization at the plasma membrane compared to t=0, and the intracellular puncta and vacuolar staining diminished (Fig. 6d,e). Upon removal of LatA, Neo-TR was internalized and found primarily in intracellular puncta. We also knocked out the endocytic adaptor gene SLA2⁶⁸ in the neo1S221L background and tested Neo-TR uptake. As expected, we observed increased plasma membrane Neo1-TR levels in the neo1S221L sla2Δ strain and reduced internal staining (Extended Data Fig. 9c,d). Surprisingly, the neo1S221L sla2Δ strain still displayed sensitivity to neomycin, suggesting that blocking neomycin endocytosis does not protect cells from its toxicity (Extended Data Fig. 9e).

Next, we examined whether Neo-TR could bind to neomycin-sensitive and PI4P-exposing human cells depleted for ATP9A. Control HeLa cells showed no binding of Neo-TR, but ATP9A knockdown cells showed significant Neo-TR at the plasma membrane, and intracellular punctate structures that were primarily excluded from the nucleus (Fig. 6f,g). Interestingly, Neo1-TR significantly stained control HEK293 cells, and ATP9A knockdown cells showed increased Neo-TR at the plasma membrane, intracellular punctate structures, and the nucleus (Fig. 6f,g). In contrast to Hela ATP9A siRNA cells, Neo1-TR predominantly localized to the nucleus of control and ATP9A knockdown HEK293 cells. These results support the hypothesis that PI4P provides a cell surface receptor for neomycin cellular uptake.

Discussion

Here, we describe a mechanism by which the P4-ATPase Neo1 in budding yeast and the orthologous ATP9A in human cells confer neomycin resistance by preventing phosphatidylinositol-4-phosphate (PI4P) secretion to the extracellular leaflet of the plasma membrane. In cells deficient for Neo1/ATP9A, the extracellular PI4P is a receptor that concentrates neomycin on the cell surface and facilitates cellular entry of this aminoglycoside antibiotic. In addition to the previously described PS and PE flippase activity^21,24,46,69, Neo1/ATP9A also functions as a PI4P flippase, transporting PI4P from the lumenal leaflet of the Golgi to the cytosolic leaflet. These discoveries provide a mechanistic explanation for the cell-type-specific toxicity of neomycin and have broader implications for phosphoinositide homeostasis in antibiotic resistance, cellular physiology, and human pathology^13,70.

In humans, ATP9A homozygous recessive mutations cause neurodevelopmental defects leading to postnatal microcephaly, intellectual disability, attention deficit hyperactivity disorder, and hypotonia^71–73. Neo1 and ATP9A play essential roles in protein trafficking between the Golgi and endosomes, and perturbation of these pathways may cause these neurodevelopmental defects in patients^44,71,74. It is also possible that the loss of plasma membrane asymmetry and perturbation of phosphoinositide metabolism contribute to neuronal dysfunction⁷⁵. The findings presented here have significant implications for understanding the toxicity of aminoglycosides like neomycin. The cell-type-specific expression of ATP9A and ATP9B may explain why certain tissues, such as kidney cells and inner ear hair cells, are particularly susceptible to neomycin-induced damage (Extended Data Fig. 4c-f). The expression level and activity of human Sac1 (SACM1L) may also contribute to neomycin sensitivity. Additionally, identifying PI4P as a neomycin receptor opens new avenues for developing strategies to mitigate aminoglycoside toxicity, such as targeting PI4P exposure or modulating flippase activity.

Neomycin binds to extracellular PI4P^9,10 and is subsequently endocytosed into the cell, as demonstrated by the uptake of Texas Red-conjugated neomycin (Neo-TR) in Neo1 and ATP9A deficient cells. This uptake is mediated by endocytosis in yeast, as inhibition of endocytic pathway reduces Neo-TR internalization. These observations suggest that PI4P serves as a cell surface receptor for neomycin, facilitating its entry into cells and subsequent accumulation in endocytic and vacuolar compartments. Surprisingly, inhibition of endocytosis reduces the Neo-TR internalization but does not affect yeast cytotoxicity, suggesting that neomycin can directly cross the plasma membrane to intoxicate the cell. The difference in the localization pattern of Neo-TR between HeLa and HEK293 cells also suggests different cytotoxic mechanisms for these two cell types. Although the precise mechanism for how the polycationic neomycin crosses a lipid bilayer and inhibits growth remains to be determined, binding to exposed PI4P is a critical determinant of cytotoxicity.

We uncovered a pathway for exposure of PI4P in the extracellular leaflet of the plasma membrane of neo1 mutants that entails both nonvesicular and vesicular transport of this lipid. This pathway starts with the synthesis of PI4P in the cytosolic leaflet of the plasma membrane by Stt4³⁷ and its Osh6-dependent, non-vesicular transport to the ER. Of the seven Osh proteins in budding yeast, only Osh6 appears to contribute significantly to the exposure of PI4P. Osh7 is thought to act redundantly with Osh6 for phosphatidylserine transport from the ER to the plasma membrane, but we did not detect a role for Osh7 in this pathway. However, it is possible that Osh7 overexpression in the osh6Δ background would reveal a contribution to the PI4P flux. Similarly, PI4P produced at the Golgi by Pik1 and potentially transported to the ER by Osh4 did not contribute to its extracellular exposure. The biosynthetic ER scramblase likely mediates the penetration of PI4P into the ER lumen, where it is protected from degradation by Sac1. Finally, COPII-dependent ER to Golgi transport and Sec14-dependent Golgi to plasma membrane vesicular transport deliver PI4P to the extracellular leaflet of the plasma membrane. The conservation of this pathway in human cells, where knockdown of the Neo1 ortholog ATP9A leads to extracellular PI4P exposure and neomycin sensitivity, underscores the evolutionary significance of this mechanism.

PI4P exposure in neo1 mutants and the ATP9A knockdown cells suggest that a substantial flux of PI4P escapes degradation by Sac1 and enters the ER lumen⁵⁹. This raises the possibility that PI4P could regulate proteins of the ER, Golgi and plasma membrane by interacting with lumenal/extracellular domains. In mammalian cells, extracellular PI4P could also mediate cell-cell signaling events^13,70,76. In addition, the loss of Sac1 greatly increases the undercurrent of PI4P in the secretory pathway, overwhelming the capacity of Neo1 to pump this lipid to the cytosolic leaflet. Sac1 is regulated by nutrient availability, and glucose deprivation causes Sac1 redistribution to the Golgi complex^77,78. Reduction of Sac1 at the ER could increase the flux of PI4P entering the ER lumen and traveling through the secretory pathway. Determining if lumenal/extracellular phosphoinositides contribute to nutrient acquisition or energy metabolism will be interesting.

The cryo-EM structure of Neo1 in the E2P state with bound PI4P provides further insight into substrate selection and transport mechanisms by P4-ATPases. From a collection of mutants targeting the substrate translocation pathway, we initially identified four mutations that caused neomycin sensitivity and PI4P exposure. Two of these mutations are in the entry gate substrate-binding site (S221L, S452Q) as defined in related P4-ATPases, one is near the “hydrophobic gate” separating the entry and exit sites (P456A), and one is near the exit gate substrate binding site (R247L). The cryo-EM structure shows PI4P is in the exit gate substrate binding site and identified residues critical for headgroup coordination (H472, H476, and E475). Consistent with a flippase-substrate interaction, the neo1-H472A E475A double mutant causes a loss of PI4P asymmetry and hypersensitivity to neomycin. As previously observed, the entry site is closed in the E2P structure and is presumably opened in vivo through interactions with Dop1 and Mon2. P4-ATPases do not require lipid substrate for autophosphorylation (E1 to E1P to E2P); instead, substrates induce dephosphorylation (E2P to E2) to stimulate overall ATPase activity. PI4P stimulated purified Neo1 ATPase activity as expected for a transport substrate. Interestingly, the P456A and exit gate R247L mutations diminish, but the entry gate S221L and S452Q mutations enhance, the ability of PI4P to stimulate ATPase activity. This unexpected behavior of the entry gate mutations suggests they are opening the entry gate binding site to enhance substrate-stimulated ATPase activity. S221L and S452Q may disrupt PI4P transport by allowing PS or PE to outcompete PI4P at the entry gate binding site in vivo. Further studies are needed to fully understand the biochemical consequences of these mutations.

In summary, this study sheds light on the broader role of extracellular phosphoinositides in cellular signaling and membrane dynamics. The exposure of PI4P on the cell surface, as observed in Neo1 and ATP9A deficient cells, may have implications for other physiological processes, including cell signaling, membrane trafficking, and pathogen interactions^13,70,71. Further investigation into the regulation and function of extracellular phosphoinositides will undoubtedly yield new insights into how these signaling lipids regulate cell physiology and impact disease mechanisms.

Methods

Yeast strains and plasmid construction:

The strains and plasmids used in the study are listed in Extended Data Table 2. All yeast culture reagents were purchased from Sigma-Aldrich and BD Scientific, and strains were grown in YPD or minimal selective media. Yeast transformation was performed using the LiAC-PEG method⁷⁹. Single and double deletion strains were generated using the homologous recombination method⁸⁰. Open reading frames were replaced with the kanMX6, CloNAT, or His3MX6 cassettes and confirmed by PCR. Neomycin was purchased from Gibco. Colony growth assays were performed by spotting 1 OD and 10-fold serial dilutions on YPD and YPD+Neomycin (500 μg/ml) plates. Plates were incubated for 2–3 days at 26 °C or indicated temperature before capturing the image. The images of yeast colonies are representative of three biological replicates, each derived from independently isolated strains with the same genotype. DNA constructs and mutations were generated through Gibson assembly (New England BioLabs) and quick-change mutagenesis (Agilent).

Cell Culture and SiRNA knockdown:

HeLa cell lines were obtained as gifts from Hye-Won Shin Lab (Kyoto University) and HEK293 cells were obtained as gifts from Lars Plate (Vanderbilt University). Cells were cultured at 37 °C with 5% CO₂ in DMEM (Gibco) with 10% FBS. Cell lines were routinely monitored for mycoplasma contamination. Predesigned siRNA duplexes against human ATP9A were purchased from Thermo Fisher Scientific with following sequences: ATP9A SiRNA#1: GAUUCCUCAUCAUUGGGUAtt (s19607) and ATP9A SiRNA#2: CGUGGGAACUUUUACCCGAtt (s19609). Control SiRNA Silencer^™ Select Negative Control No. 1 siRNA was also purchased from Thermo Fisher Scientific (4390844). Hela cells were seeded on six-well plates at a density of 1.5×10⁵ cells/well. Cells were transfected with 30 nM siRNA using Lipofectamine RNAiMAX Reagent (Invitrogen-13778150). The knockdown efficiency was validated by western blots after siRNAs were transfected for 48 hr using Anti-ATP9A antibody.

Protein extraction and Immunoblotting:

Cells were washed with PBS and scraped from the plate in RIPA buffer (Sigma Aldrich R0287) plus protease and phosphatase inhibitors. Cells were transferred into a clean microcentrifuge tube and vortexed for 20 s at high speed. Cells were centrifuged at 14,000g for 15 min, and supernatant was collected in a microcentrifuge tube. Protein concentration was measured using Pierce BCA protein assay Kit (Thermo Scientific-23225). 20 μg of protein samples were loaded on a 4–20% gradient SDS–PAGE gel (Bio-Rad Cat. #4561095) and transferred to PVDF membranes. ATP9A was probed using primary antibody Rabbit Anti-ATP9A antibody (ab234873, 1:1000) and secondary antibody anti-Rabbit HRP (Promega W4011, 1:10000). The loading control β-actin was probed using primary Mouse Anti-β-Actin antibody (CST#3700, 1:10000) and secondary anti-Mouse HRP (Promega W4021,1:10000) antibody.

Neomycin sensitivity assay for yeast cells:

Cells grown to the mid-log phase were diluted to 0.1 OD600 in fresh YPD medium, and 100 μL of the resulting cells were dispensed into each well of a 96-well plate, with or without neomycin, in an additional 100 μL of YPD. Neomycin dilutions were calculated to achieve final concentrations in a total well volume of 200 μL. The plates were then incubated at 26 °C for 20 hr. Cell concentrations were measured in OD600/mL using a Multimode Plate Reader Synergy HT (Biotek). Growth, relative to the vehicle control (no drug), was considered as 100% growth. All values represent averages from at least three biological replicates ± standard deviation. Statistical analysis were determined using a mixed-effects analysis model followed by a Dunnett’s or Tukey’s multiple comparison analysis.

Neomycin sensitivity assay for mammalian cells:

HeLa cells or HEK293 cells were treated with respective siRNA for 48 hrs in 6 well plates. Cells were harvested and approximately 5,000 cells were seeded on 96 well plates in 100 μL of complete media (DMEM+FBS+ P/S) and allowed to attach dish for 24 hr. After 24 hrs 100 μL of complete media was replaced with the 100 μL of Neomycin containing media (DMEM+FBS+ Neomycin) and incubated for 48 hrs at 37 °C. Cell viability was measured using WST-1 Assay Kit (ab65473). Briefly 10 μL of WST-1 reagent was added to each well and incubated for 4 hrs at 37 °C for formazan formation by viable cells. The amount of formazan produced by viable cells was quantified using Multimode Plate Reader Synergy HT (Biotek) at 440 nm. Viable cells at the no drug condition were considered as 100% growth. All values represent averages from at least three biological replicates ± standard deviation.

GFP-SidC_P4C protein expression and purification:

Recombinant GFP-SidC_P4C and GFP-SidC_P4C(R562Q) was expressed and purified from Escherichia coli as previously described³⁸. Briefly, E. coli Rosetta2 pLysS strains containing the expression plasmids were cultured in LB media supplemented with 50 μg/ml kanamycin and 25 μg/ml chloramphenicol until reaching mid-log phase. Protein expression was induced overnight at 18 °C by adding 0.1 mM isopropyl-β-D-thiogalactopyranoside (IPTG). Harvested cells were resuspended in a buffer composed of 20 mM Tris-HCl (pH 7.5) and 150 mM NaCl and lysed using a cell disrupter (Constant Systems). The soluble fractions were obtained through centrifugation at 18,000 rpm for 30 min at 4 °C and were then incubated with TALON metal affinity resins (Takara-635502) for 1 hr at 4 °C. The protein-bound resins were thoroughly washed with the lysis buffer. Protein was eluted by stepwise gradient (15, 30, 40, 50, 60, 180, 180, and 240 mM imidazole), and fractions containing the protein were analyzed by SDS-PAGE and peak fractions were pooled together. An Amicon Ultra-15 centrifugal filter unit (Millipore, UFC903008) was used to dialyze into an imidazole free buffer containing 20 mM Tris, pH 7.5, and 50 mM NaCl.

GFP-PLC_PH expression and purification:

Recombinant GFP-PLC_PH was expressed and purified from E. coli BL21 (DE3) as previously described³⁹. E. coli strain expressing GFP-PLC_PH was grown in Terrific Broth containing 4 ml/liter glycerol, 50 ug/ml kanamycin and 0.8% (wt/vol) D-glucose. Cells were induced with 0.1 mM IPTG at 0.6 OD for 20 h at 23 °C. Cells were harvested and resuspended in lysis buffer containing 25 ml of 300 mM NaCl, 40 mM Tris, 15 mM imidazole, 5% (vol/vol) glycerol, pH 7.4 (10 °C), 5 mM β-mercaptoethanol, protease inhibitors and DNase. Cells were lysed using a cell disrupter (Constant Systems) and the lysate was centrifuged at 15,000 rpm for 20 min. The supernatant was incubated overnight with TALON metal affinity resin (Takara-635502) at 4 °C. The resin was washed twice with 10–15 mM imidazole in lysis buffer and eluted using a stepwise imidazole gradient. Peak fractions analyzed by SDS-PAGE were pooled and dialyzed (Thermo Fisher, 66380; Slide-A-Lyzer Dialysis Cassettes) into imidazole-free buffer (150 mM NaCl, 20 mM Tris, 5% (vol/vol) glycerol and 0.5 mM β-mercaptoethanol, pH 7.4).

GFP-SidC_P4C and GFP PH_PLC binding assay for yeast cells:

Yeast strains were grown in YPD media and cells were harvested while in log phase (0.5 OD600/mL). Cells were washed once in 50 mM Tris–Cl, pH 7.4, and resuspended at 50 OD600/mL in 50 mM Tris–Cl, pH 7.4, containing 1.4 M sorbitol. Cells were then converted to spheroplasts during a 45-min incubation with 1 μl of Zymolyase-100T (10 mg/mL in water; Seikagaku Corp. Tokyo, Japan) per 10 OD600units of cells. Spheroplasts were resuspended at 1 OD600cells/mL in buffer (50 mM Tris, pH 7.4, 50 mM NaCl with 1.4 M sorbitol). 2 μl of GFP-SidC_P4C (12 μg/ml) was added to 200 μl spheroplast suspension, and the mixture was incubated for 20 mins on ice. Cells were washed two times with buffer (50 mM Tris, pH 7.4, 50 mM NaCl with 1.4 M sorbitol) and cells were then mounted on glass slides and observed immediately at room temperature. Images were acquired using a DeltaVision Elite Imaging System equipped with a 100×, 1.4 NA oil immersion objective lens followed by deconvolution using softWoRx software (GE Healthcare Cytiva, Pittsburgh, PA). Images were analyzed and fluorescence intensity was quantified using ImageJ (1.54g). For GFP-PLC_PH binding 1 μl (20 μg/ml) of purified protein was incubated with 200 μl spheroplast suspension, and the mixture was incubated for 20 min on ice. Cells were washed two times with buffer (50 mM Tris, pH 7.4, 50 mM NaCl with 1.4 M sorbitol) and mounted on the glass slide. Images were acquired at room temperature and GFP-PLC_PH binding was quantified using ImageJ.

GFP-SidC_P4C and GFP PH_PLC binding assay for mammalian cells:

Hela cells were treated with respective siRNA for 48 hrs in 6 well plates. Cells were harvested and approximately 50,000 cells were seeded on 35 mm Glass bottom MatTek Dishes (P35G-1.5–14-C) and allowed to attach dish for 24 hr. After 24 hrs cells were washed with 1 ml of 1X PBS and Hoechst stain (33342, 1:1000) in PBS was added to the cells and incubated for 15 min at 37 °C. Cells were washed two times with PBS and 10 μl of GFP-SidC_P4C (12 μg/ml) or 5 μl (20 μg/ml) of GFP-PLC_PH in 1 ml PBS was added to 1 ml ice cold PBS, and the mixture was incubated for 40 min on ice. Cells were washed two times with 1X PBS. Images were acquired using Inverted LSM880-Airyscan (Zeiss) microscope equipped with a 63×, 1.4 NA oil immersion objective lens followed by deconvolution using Zen Desk software. Images were analyzed and GFP-SidC_P4C or GFP-PLC_PH binding was quantified using ImageJ (1.54g). HeLa cells were permeabilized and stained as previously described⁴⁰. HeLa cells were cultured in a 6-well uncoated glass dish (10 mm diameter) to 80% confluence in complete media. After removing the media, cells were washed with 1 ml of 1X PBS. Hoechst stain (33342, 1:1000) in PBS was then added and incubated for 15 min at 37 °C. Following two PBS washes, cells were fixed with 4% PFA and 0.2% glutaraldehyde at room temperature for 20 min. The cells were then washed twice with 50 mM ammonium chloride and incubated for 20 min. The six-well plates were placed on ice and washed twice with ice-cold PIPES buffer. Fresh staining solution (5% BSA, 0.5% (vol/wt) saponin, GFP-SidC_P4C or GFP-PLC_PH) was prepared, and 200 μl of this solution was added to each sample and incubated on ice for 45 min. After staining, the cells were washed three times with ice-cold PIPES buffer. Post-fixation was performed with 2% PFA in PBS on ice for 10 min, followed by 10 min at room temperature. Cells were then washed with 50 mM NH₄Cl in PBS, rinsed with water, mounted, and sealed using ProLong Diamond Antifade.

Neomycin Texas red labeling:

Neomycin was conjugated to Texas red as previously described⁸¹. Neomycin sulfate hydrate (Sigma-Aldrich) was used at a final concentration of 115.6 mg/ml. 115.6 mg neomycin sulfate hydrate was initially resuspended in 500 μl of deionized water, to which was added 176 μl of 0.5 M K₂CO₃ at pH 9.0, and 120 μl of Texas Red-X-succinimidyl ester (Life Technologies) dissolved in dimethylformamide at 2.5 mM. The reaction volume was adjusted to 1 ml with deionized water and the solution was incubated overnight at 4 °C to complete the conjugation reaction.

Neomycin-TR uptake assay for yeast cells:

Cells were grown up to mid log phase in rich media YPD. Cells were harvested and washed three times in SD media. 2 μl of Neo-TR from stock solution was added to 50 μl of cells and incubated on ice for 20 min. Cells were washed two times with fresh media and resuspended in fresh media. Images were collected using DeltaVision Elite Imaging System equipped with a 100×, 1.4 NA oil immersion objective lens followed by deconvolution using softWoRx software (GE Healthcare Cytiva, Pittsburgh, PA). The fluorescence intensity was measured using ImageJ. For LatA experiments cells were treated with Neo-TR for 20 min and treated with LatA for 30 and 60 mins. Cells were washed and resuspended in fresh media. Cells were kept on ice until mounted on the coverslip.

Neomycin-TR uptake assay for mammalian cells:

HeLa cells and HEK 293 cells were treated with respective siRNA for 48 hrs in 6 well plates. Cells were harvested and approximately 50,000 cells were seeded on 35 mm Glass bottom MatTek Dishes (P35G-1.5–14-C) and allowed to attach dish for 24 hr. After 24 hrs cells were washed with 1ml of 1X PBS and Hoechst stain (33342, 1:1000) in PBS was added to the cells and incubated for 15 min at 37 °C. Cells were washed two times with PBS and 20 μl of Neo-TR from stock solution was added to 1 ml ice cold PBS, and the mixture was incubated for 30 min on ice. Cells were washed two times with 1X PBS. Images were acquired using Inverted LSM880-Airyscan (Zeiss) microscope equipped with a 63×, 1.4 NA oil immersion objective lens followed by deconvolution using Zen Desk software. Images were analyzed and Neo-TR fluorescence intensity was quantified using ImageJ (1.54g).

Molecular cloning:

The gene sequence encoding Neo1 from S. cerevisiae was PCR-amplified from genomic DNA of S. cerevisiae BY4741 obtained from Horizon Discovery (Waterbeach, UK) with a pair of primers 5′-GATTACAAGGATGACGACGATAAGCCTAACCCTCCTTCATTTAAGTCACATAAGC-3′/5′-ATTAAGAGCTCATGGAGTGGCAAACTCTTGCAC-3′. The gene was inserted using ligation-independent cloning into a vector modified from pJK369 (Plasmid #73594) acquired from Addgene (Watertown, MA) for the expression of Neo1. After DNA sequence verification (Plasmidsaurus, Eugene, OR), the S. cerevisiae strain BY4741 Δpep4::kanMX6 (Horizon Discovery, Waterbeach, UK) was transformed with the constructed vector using Quick & Easy Yeast Transformation Mix (Takara Bio, San Jose, CA). The expressed Neo1 had an N-terminal FLAG-tag.

Protein expression and purification:

Cells were grown in 1.2 liters of SD -Ura (Synthetic Drop-out without uracil) medium (20 g/L glucose, 6.7 g/L yeast nitrogen base without amino acids, and 1.92 g/L yeast synthetic drop-out medium supplements without uracil) at 30 °C, shaking at 250 rpm, until exhaustion of glucose in the culture monitored using a Contour Next One blood glucose meter (Ascensia Diabetes Care, Parsippany, NJ). Gene expression was induced with a final concentration of 2% galactose in 4.8 liters of YP medium (10 g/L yeast extract and 20 g/L peptone) and cells were grown for an additional 16 hours at 30 °C, 220 rpm. Cells were harvested by centrifugation at 10,000 × g at 4 °C for 5 min, cell pellet was washed and resuspended in protein buffer (20 mM HEPES, pH 7.4, 150 mM NaCl and 1 mM MgCl₂), then dripped into liquid nitrogen and finally stored at −80 °C before cell lysis. Cells were lysed using a SPEX Freezer/Mill 6875 (SPEX SamplePrep, Metuchen, NJ) with the following settings: pre-cool 1 min, run time 2 min, cool time 1 min, cycles 5, and rate 15 CPS (cycle per second). The milling was repeated one time, then the lysed cells were resuspended in protein buffer also containing 1x ProBlock Gold Yeast/Fungi Protease Inhibitor Cocktail (GoldBio, St Louis, MO) and 10 μL Benzonase Nuclease HC (EMD Millipore, Burlington, MA). The resuspension was further incubated at 4 °C for 30 min with gentle stirring. After centrifugation at 6,000 × g for 15 min at 4 °C, the supernatant was filtered through an Econo-Column chromatography column with a 20 μm pore size bed support (Bio-Rad, Hercules, CA). The resulting protein solution was ultracentrifuged at 137,900 × g for 2 hours at 4°C, and the membrane pellet was resuspended in resuspension buffer (20 mM Tris, pH 7.4, 500 mM NaCl, 10% (w/v) glycerol and 1 mM MgCl₂) and homogenized with a Dounce homogenizer. The resulting membrane homogenate was treated with final concentrations of 1% DDM (n-Dodecyl-β-D-Maltopyranoside), 0.1% CHS (Cholesteryl Hemisuccinate Tris Salt) and 1 mM AEBSF and incubated at 4 °C for 1 hour with constant stirring. Extracted membrane proteins were collected after ultracentrifuge at 137,900 × g for 1 hour at 4 °C, and then mixed with 2 mL bed volume of pre-equilibrated anti-DYKDDDDK G1 affinity resin (GenScript, Piscataway, NJ) and incubated at 4 °C for 3 hrs with stirring. The mixture was then transferred to an Econo-Column chromatography column (Bio-Rad, Hercules, CA) at 4 °C. After flow-through collection, the column was washed with protein buffer with detergents (0.01% LMNG and 0.001% CHS) containing 0 mM, 0.25 mM and 0 mM ATP in sequence in 10 bed volumes. Two strategies have been reported in the literature for removing endogenously bound lipids in a lipid flippase, either via excessive column wash⁵¹ or by wash with ATP during purification⁸². We used the later approach by adding an ATP wash step. Finally, the column was developed with 3 bed volumes of protein buffer with detergents (0.01% LMNG and 0.001% CHS) and 0.5 mg/mL DYKDDDDK peptide. The eluate was collected and concentrated to 0.5 mL with a 50 kDa cutoff centrifugal filter (EMD Millipore, Burlington, MA) at 4,000 × g at 4 °C. Finally, the protein concentrate was further purified with a Superose 6 Increase 10/300 GL column (Cytiva, Marlborough, MA) at a flow rate of 0.5 mL/min in final protein buffer with detergents (0.0025% LMNG and 0.00025% CHS). The peak fractions of the target protein were pooled and concentrated before prompt use or flash-freezing in liquid nitrogen and storage at −80 °C.

Peptidisc reconstitution of Neo1:

Concentrated Neo1 in 5.325 mg/mL was first incubated with 0.5 mM PI4P dissolved in 4 mM sodium cholate at 4 °C for 3 hours, and then added with 300 μL pH-adjusted peptidisc solution containing 0.5 mg peptidisc peptide (Peptidisc Biotech, Vancouver, BC, Canada) and incubated further on ice for 30 min. Neo1 in peptidiscs was further purified with a Superose 6 Increase 10/300 GL column (Cytiva, Marlborough, MA) at a flow rate of 0.5 mL/min in final protein buffer without detergents (20 mM HEPES, pH 7.4, 150 mM NaCl and 1 mM MgCl₂). The peak fractions of the target protein were pooled and concentrated to 1.511 mg/mL.

Cryo-EM grid preparation and data collection:

For the E2P state of peptidisc-reconstituted Neo1 bound with PI4P, Neo1 in a final protein concentration of 1.058 mg/mL was incubated with 10 mM NaF, 2 mM BeSO₄, 5 mM MgCl₂ and 0.5 mM PI4P dissolved in DMSO (dimethyl sulfoxide)/ethanol (v/v=1:1) at 4 °C overnight. After glow discharge cleaning for 15 s in a Gatan Solarus Model 950 plasma cleaner (Gatan, Pleasanton, CA), the 400-mesh gold C-Flat R1.2/1.3 holey carbon grids (Electron Microscopy Sciences, Hatfield, PA) were applied with 3 μL protein sample. Sample vitrification was carried out in a Vitrobot Mark IV (Thermo Fisher Scientific, Waltham, MA) at 5 °C and 100% relative humidity with the following settings: blot time 4 s, blot force 2, wait time 0 s. The EM grids were plunge-frozen in liquid ethane, cooled by liquid nitrogen. Automated cryo-EM data collection was performed on a 300 kV Titian Krios electron microscope controlled by SerialEM v4.0.12⁸³ in multi-hole mode. Movie stacks were recorded at 105,000× magnification with the objective lens defocus values varied from −1.3 to −1.6 μm, in a K3 direct electron detector (Gatan, Pleasanton, CA) in super-resolution mode. The calibrated image pixel size was 0.828 Å at specimen level. A GIF Quantum energy filter was also used to remove inelastically scattered electrons with the slit width set to 20 eV. For each movie micrograph, a total of 50 frames were captured with 1 s exposure time and a total electron dose of 58 e⁻/Ų.

Cryo-EM image processing and 3D reconstruction:

Image processing was carried out in cryoSPARC v4.2. Movie stacks were aligned by patch motion correction for beam-induced movement, then the contrast transfer function (CTF) parameter of each micrograph was determined by patch CTF estimation, and the CTF effect was subsequently corrected. After manual inspection to remove poor-quality images, the remaining micrographs were frame aligned and dose weighted. Raw particle images were first picked using templates generated from a preliminary 3D map. The picked raw particles were subjected to a 2D classification. After removing obvious “junk” classes and also duplicates with a minimum separation distance of 50 Å, the resulting particles were used for ab initio 3D reconstructions. After three rounds of heterogeneous refinement, the good class particles were subjected to a 2D classification to select rare-view classes for Topaz⁸⁴. The Topaz-picked raw particles were subjected to two rounds of heterogeneous refinement followed by a 2D classification, from which the good class particles were combined with previously template-picked good class particles. The combined and duplicate-removed particles were further processed by non-uniform refinement followed by local refinement to obtain the final map.

Model building, refinement and validation:

The initial atomic model was built by fitting the Neo1 AlphaFold model⁸⁵ into the EM map. The atomic model was then rebuilt manually and refined with Coot v0.9.8.3⁸⁶ and ISOLDE v1.6⁸⁷ followed by iterative real-space refinement in Phenix v1.20.1⁸⁸. The final model was validated by MolProbity v4.5.1⁸⁹. Structural figures were prepared in UCSF ChimeraX v1.7⁹⁰.

ATP hydrolysis assay:

ATPase activity assays were conducted using BIOMOL Green Reagent (Enzo Life Sciences, Farmingdale, NY) to quantify released inorganic phosphate. Lipids PE and PI4P were solubilized with 20 mM and 4 mM sodium cholate, respectively, in 20 mM HEPES, pH 7.4, and 150 mM NaCl. Each reaction mixture contained 0.025 mg/mL protein, 0.0025% LMNG, 0.00025% CHS, 20 mM HEPES, pH 7.4, 150 mM NaCl, 10 mM MgCl₂, and 0.25 mM ATP. Reactions were carried out at 37 °C for 15 min and immediately terminated by the addition of the reagent. After incubation of the mixture for 20 min at room temperature, the absorbance at 620 nm was measured using a microplate reader SpectraMax M2e (Molecular Devices, San Jose, CA). The phosphate concentration was determined by calibration using the phosphate standard (BML-KI102).

Data availability:

The cryo-EM 3D map of the S. cerevisiae Neo1 in E2P state and bound with PI4P has been deposited in the Electron Microscopy Data Bank under accession code EMD-44850. The corresponding atomic model has been deposited in the Protein Data Bank under accession code 9BS1. The EM data are available from the authors upon reasonable request. Source data are provided with this paper.