P4-ATPases control phosphoinositide membrane asymmetry and neomycin resistance

Abstract

The aminoglycoside antibiotic neomycin 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. Cryogenic electron microscopy reveals PI4P binding to Neo1 within the substrate translocation pathway. PI4P enters the lumen of the endoplasmic reticulum and is flipped by Neo1 at the Golgi to prevent PI4P secretion to the cell surface. Deficiency of the orthologous ATP9A in human cells also causes exposure of PI4P and neomycin sensitivity. These findings unveil conserved mechanisms of aminoglycoside sensitivity and phosphoinositide homoeostasis, with important implications for signalling by extracellular phosphoinositides.

Fig. 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 colours (PDB ID 9BS1, this study). The dashed box is the substrate translocation pathway enlarged in b. b, The transmembrane (TM) segments TM1, TM2 and TM4 form the substrate transport pathway of Neo1, and the 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⁻¹) 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. The mixed model analysis was performed to test the variance and comparisons with WT Neo1 were made with Tukey’s multiple comparison analysis (n = 3 biologically independent experiments; ****P < 0.0001). The data are the mean ± s.d. e,f, Neomycin sensitivity of S. cerevisiae flippase mutants (neo1-1, drs2∆, dnf1,2∆ and dnf3∆) and strains deficient for Neo1 regulators (dop1-1 and mon2∆) on YPD + neomycin plates (e) and liquid culture (f). The data represent the growth relative to WT cells in absence of the drug. A mixed model analysis was performed to test the variance and comparisons with WT Neo1 were made with Tukey’s multiple comparison analysis (n = 3 biologically independent experiments; ****P < 0.0001). n.s., not significant. The data are the mean ± s.d. Source data

Fig. 2. neo1 mutant cells expose PI4P in the plasma membrane extracellular leaflet.

a, PIP 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. PM, plasma membrane; PI, phosphatidylinositol. b, Inactivation of neo1-1 at the non-permissive temperature (38 °C) causes PI4P exposure. The fluorescent images were inverted to black signal on white background for clarity. The fluorescence signal intensity of the GFP probe (left) and the differential interference contrast (DIC) panel to display yeast cells (right) are shown. Scale bar, 2 μm. c, A quantification of GFP–SidC_P4C and GFP–PLC_PH binding to the cell surface of cells expressing Neo1 variants. Two-way ANOVA was performed to test the variance and comparisons with WT Neo1 were made with Dunnett’s multiple analysis (n = 20 cells from three biologically independent experiments; P < 0.0001). The data are the mean ± s.d. d, The GFP–SidC_P4C(R652Q) PI4P binding defective mutant fails to bind neo1-1 cells. A one-way ANOVA was performed to test the variance and comparisons with WT Neo1 were made with Dunnett’s multiple analysis (n = 20 cells from three biologically independent experiments; ****P < 0.0001). The data are the mean ± s.d. e, stt4-4 suppresses the neomycin sensitivity of neo1S221L or neo1S452Q at 33 °C. f, pik1-83 fails to suppress the neomycin sensitivity of neo1S221L or neo1S452Q at 34 °C. The data represent growth relative to cells in the absence of the drug. A mixed effect model was performed to test the variance and comparisons with WT Neo1 were made with Tukey’s multiple comparison test (n = 3 biologically independent experiments; **P = 0.0069, 0.0068). n.s., not significant. The data are the mean ± s.d. Source data

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

a, Neomycin sensitivity of WT HeLa cells and ATP9A-knockdown cells. The data represent the per cent of viable cells relative to the viable cells in absence of the neomycin. NS, non-silencing (control) siRNA. A mixed effect model was performed to test the variance and comparisons with control cells were made with Dunnett’s multiple comparisons test (n = 3 biologically independent experiments; **P = 0.0054, 0.0025). The data are the mean ± s.d. b, ATP9A-knockdown HeLa cells expose PI4P in extracellular leaflets. Hoechst dye was used as nuclear stain. Scale bar, 10 μm. c, A quantification of GFP–SidC_P4C and GFP–PLC_PH binding to the cell surface of ATP9A-knockdown cells. A two-way ANOVA was performed to test the variance and comparisons with control siRNA treated cells were made with Dunnett’s multiple comparisons test (n = 20 cells from three biologically independent experiments; ****P < 0.0001). The data are the mean ± s.d. d, Neomycin sensitivity of WT HEK293 cells and ATP9A-knockdown cells. The data represent the per cent of viable cells relative to the viable cells in absence of the neomycin. A mixed effect model was performed to test the variance and comparisons with control cells were made with Dunnett’s multiple comparisons test (n = 3 biologically independent experiments; **P = 0.0024, 0.0014). The data are the mean ± s.d. e, ATP9A-knockdown HEK293 cells expose PI4P in extracellular leaflets. Scale bar, 10 μm. f, A quantification of GFP–SidC_P4C and GFP–PLC_PH binding to the cell surface of ATP9A-knockdown cells. A two-way ANOVA was performed to test the variance and comparisons with control siRNA treated cells were made with Dunnett’s multiple comparisons test (n = 20 cells from three biologically independent experiments; ****P < 0.0001). The data are the mean ± s.d. Source data

Fig. 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. A 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 biologically independent experiments; ****P < 0.0001). n.s., not significant. The data are the mean ± s.d. b, ATP hydrolysis activity of WT Neo1 and Neo1 mutants at increasing PI4P concentration. The data points represent the mean ± s.d. from three biologically independent experiments. c, Global views of Neo1-PI4P in E2P state with the four 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 (grey, PDB ID 6ROJ) and shown in cartoon (root mean square deviation of 4.0 Å). Note the different binding site of PI4P as an activator in Drs2 (grey spheres) and as a substrate in Neo1 (teal spheres). The PI4P binding site residues in Drs2 are highlighted in the red box. Left: an enlarged view of the PI4P binding site in Drs2. Right: an enlarged view of PI4P binding site in Neo1. PI4P and interacting residues are shown in sticks. The hydrogen bonds between the PI4P headgroup and flippase residues are labelled by black dashed 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 represent growth relative to WT cells in absence of the drug. A mixed effect model was performed to test the variance and comparisons with WT Neo1 were made with Tukey’s multiple comparison test (n = 3 biologically independent experiments; ****P < 0.0001). The data are the mean ± s.d. Source data

Fig. 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; P_i, inorganic phosphate; PI, phosphatidylinositol; PS, phosphatidylserine. b, sac1∆ cells are neomycin sensitive. c, Overexpression of WT NEO1 (2 µ NEO1) suppresses sac1∆ neomycin sensitivity. A mixed effect model was performed to test the variance and comparisons with WT Neo1 were made with Tukey’s multiple comparison test (n = 3 biologically independent experiments; **P = 0.0031). The data are the mean ± s.d. d, sac1∆ cells expose extracellular PI4P and overexpression of NEO1 suppresses PI4P exposure. A two-way ANOVA was performed to test the variance, and comparisons with were made with Tukey’s multiple comparisons test (n = 20 cells from three biologically independent experiments; ****P < 0.0001). The data are the mean ± s.d. e, Deletion of OSH6 (osh6∆) 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 mixed effect model to test the variance and comparisons with WT Neo1 were made with Tukey’s multiple comparison test. The colours represent a comparison with WT, and the pairwise comparisons are shown with brackets and asterisk. The data represent growth relative to WT cells in absence of the drug (n = 3 biologically independent experiments; **P = 0.0029; ***P = 0.0005, 0.0009; ****P < 0.0001). The data are the mean ± s.d. For b, c, e and f, the cells were grown at 26 °C. Source data

Fig. 6. Neomycin binds to exposed PI4P and is endocytosed.

a, Use of 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, A quantification of mean fluorescence intensity of cells stained with Neo–TR. A one-way ANOVA was performed to test the variance and comparisons with WT Neo1 were made with Dunnett’s multiple comparisons test (n = 20 cells from three biologically independent experiments; ****P < 0.0001). The data are the mean ± s.d. d, LatA blocks Neo–TR internalization. Scale bar, 2 μm. e, A quantification of Neo–TR fluorescence intensity at the plasma membrane (PM). For all quantification, the data from ~20 cells from three independent experiments were obtained and analysed. A one-way ANOVA was performed to test the variance and comparisons with different timepoints were made with Tukey’s multiple comparisons test (n = 20 cells from three biologically independent experiments; **P = 0.0014; ****P < 0.0001). The data are the mean ± s.d. f, Neo–TR binds to the ATP9A-knockdown HeLa cells. Hoechst dye was used as a nuclear stain. g, A quantification of total Neo–TR fluorescence intensity of cells (n = 20 cells from three biologically independent experiments). Scale bar, 10 µm. h, Neomycin-sensitive HEK293 and ATP9A-knockdown cells bind Neo–TR. Scale bar, 10 µm. i, A quantification of total Neo–TR fluorescence intensity (n = 20 cells from three biologically independent experiments). For g and i, a two-tailed unpaired t-test was performed to test the variance and comparisons with control cells (n = 20, mean ± s.d. ****P < 0.0001). Source data

Extended Data Fig. 1. Neo1 substrate transport pathway mutants are sensitive to neomycin.

Neomycin sensitivity assay of neo1 substrate transport pathway mutants at 26 °C on YPD and YPD Neomycin (500 µg/ml) plates.

Extended Data Fig. 2. Neomycin sensitive neo1 substrate transport pathway mutants expose PI4P in extracellular leaflets.

a, neo1-1, dop1-1 and mon2∆ mutant exposes PI4P on extracellular leaflets probed using recombinantly purified GFP-SidC_P4C. 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. b, Quantification of GFP-SidC_P4C binding to the cell surface of the WT cells and flippase mutant strains. Two-way ANOVA was performed to test the variance and comparisons with WT cells were made with Tukey’s multiple comparison test. (n = 20 cells from three biologically independent experiments, Data are mean ± (SD). **** P < 0.001). c, Intact live cells expose PI4P on the extracellular leaflet. Scale bar = 2 μm. d, Quantification of the percentage of the cells exposing PI4P stained with PI. Two-way ANOVA was performed to test the variance and comparisons with WT Neo1 cells were made with Šídák’s multiple comparisons test. (n = 3 biologically independent experiments, Data are mean ±(SD).**** P < 0.001). e, Perturbation of plasma membrane phosphoinositide asymmetry was tested in neo1 mutant strains. (n = 20 cells from three biologically independent experiments). Scale bar = 2 μm. f, Only GFP-SidC_P4C binds to neomycin-sensitive mutant cells at the rim of cells, suggesting exposure of PI4P on extracellular leaflet of the plasma membrane. (n = 20 cells from three biologically independent experiments). Scale bar = 2 μm. g, HeLa cells were fixed, permeabilized and stained with recombinant biosensors GFP-SidC_P4C or GFP-PLC_PH. (n = 20 cells from three biologically independent experiments). Scale bar = 10 μm. Source data

Extended Data Fig. 3. Confirmation of pik1-83, stt4-4 and mss4-102 mutants.

a, Live cell fluorescence microscopy of intracellular (cytosolic) GFP-SidC_P4C in PI kinase mutant cells pik1-83 and stt4-4, at permissive (28 °C) and non-permissive temperature (37 °C). (n = 20 cells from three biologically independent experiments). b, Neo1 neomycin sensitive mutants do not have any effect on localization of intracellular PI(4,5)P₂ probed with GFP-PLC_PH expressed in the yeast cytosol. Scale Bar =2 µm. n = 3 biological replicates.(n = 20 cells from three biologically independent experiments). c, A PI 5-kinase Mss4 mutant allele fails to suppress the neomycin sensitivity of neo1 mutants. The data represent growth relative to WT cells without the drug. Mixed model analysis was performed to test the variance and comparisons with neo1S221L or neo1S452Q were made with Tukey’s multiple comparisons test (n = 3 biologically independent experiments, Data are mean ± SD). ns represents not significant. d, localization of GFP-PLC_PH expressed in the cytosol of a mss4-102 mutant at permissive (28 °C) and non-permissive temperature (37 °C) Scale Bar = 2 µm.(n = 20 cells from three biologically independent experiments). Source data

Extended Data Fig. 4. Expression of ATP9A and ATP9B.

a, Sequence alignment of Neo1 with human and mouse orthologs ATP9A, ATP9B and C. elegans Tat5. b, Expression of ATP9A in HeLa cells and HEK293 cell lysates by immunoblotting. (n = 3 biologically independent experiments). c,d RNA-seq data of ATP9A and ATP9B in tissues. e,f RNA- seq data of ATP9A and ATP9B in commonly used cell lines. Arrow highlights the expression of ATP9A in Kidney tissues, HeLa cells and HEK293 cells. The expression plot images were adopted from Genotype-Tissue Expression (GTEx) Portal and Human Protein Atlas (ATP9A: https://www.proteinatlas.org/ENSG00000054793-ATP9A, ATP9B: https://www.proteinatlas.org/ENSG00000166377-ATP9B). The data used for the analyses in c,d,e,f described in this manuscript were obtained from the GTEx Portal on 01/20/2025. Source data

Extended Data Fig. 5. Cryo-EM structural determination of peptidisc-reconstituted Neo1 bound with PI4P in the E2P state.

a, Data processing workflow. b, Representative raw micrograph. A total of 6,166 such micrographs were recorded. c, Representative 2D classes. d, Gold-standard Fourier shell correlation (GSFSC) curve for the 3D reconstruction. e, Angular distribution heat map for the 3D reconstruction.

Extended Data Fig. 6. Comparison of substrate binding sites in Neo1 and Dnf1.

Substrate entry site is open in the E2P state of Dnf1 in detergents (PDB ID 7KYC) (a), but is closed in the peptidisc-reconstituted Neo1 in E2P state (b, this study), and in the E2P state of Neo1 in detergents (PDB ID 7RD6) (c). Residues surrounding the substrate entry site are shown in sticks. The entry site is more tightly closed in detergent-reconstituted Neo1, involving additional residues T979, V224 and P225.

Extended Data Fig. 7. Mutation in the Neo1 PI4P binding region does not affect cell growth.

a, Sequence alignment of Neo1 PI4P binding region with human orthologs ATP9A, ATP9B and Drs2. Sc: Saccharomyces cerevisiae. Hs: Homo sapiens. b, All of the PI4P binding region Neo1 mutants support viability of neo1∆ cells. We transformed a neo1∆pURA3-NEO1 strain with HIS3-marked plasmids harboring the indicated Neo1 variant. Cells were spotted on complete media SD to select both plasmids and on SD-5-FOA plates to pop out pURA3-NEO1 plasmid which will allow the expression on 5-FOA plates. NEO1 is an essential gene, and all Neo1 mutants were able to complement the growth defect of neo1∆ strain. c, Sequence alignment of Drs2 PI4P binding region with Neo1, ATP9A and ATP9B. The Drs2 PI4P binding site residues are not well conserved in Neo1/ATP9A/ATP9B.

Extended Data Fig. 8. sac1∆ cells expose PI4P in extracellular leaflets and neomycin sensitivity assay of neo1-1 osh mutants.

a, sac1∆ cells expose PI4P in extracellular leaflets and overexpression of NEO1 from a 2 µ plasmid suppresses the PI4P exposure. 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. n = 3 independent biological replicates. b, neo1 temperature sensitive mutants (ts) neo1-1 sac1∆ or neo1-2 sac1∆ mutants are inviable or grow poorly on 5-FOA plates. We transformed a sac1∆ neo1∆ pURA3-NEO1 strain with LEU-marked plasmids harboring the indicated Neo1 temperature sensitive variant. Cells were spotted on complete media SD to select both plasmids and on SD-5-FOA plates to pop out pURA3-NEO1 plasmid which will allow the expression of Neo1 variant on 5-FOA plates. c, Deletion of only osh6∆ suppresses the neomycin sensitivity of neo1 mutants. The data represent growth relative to WT cells without the drug. Mixed model analysis was performed to test the variance and comparisons with neo1-1 were made with Tukey’s multiple comparison test (n = 3 biologically independent experiments, Data are mean ± SD). ns represents not significant. *** represents p value = 0.007). Source data

Extended Data Fig. 9. Neomycin binds to exposed PI4P and is endocytosed.

a, Neo-TR binds neomycin-sensitive neo1 mutants and endocytosed to the trans-Golgi Network (Sec7) and endosomal compartment (Tlg1). Scale bar = 2 μm. b, Colocalization of Neo-TR with TGN marker Sec7 and endosomal marker Tlg1 (n = 20 cells from three biologically independent experiments). Data are mean ± SD. c, knockout of sla2∆ in neomycin sensitive mutant neo1 S221L blocks Neo-TR internalization. Scale bar = 2 μm d, Quantification of Neo-TR fluorescence intensity at the PM. For all quantification, (n = 20 cells from three biologically independent experiments). The two-tailed unpaired t-test was used for comparison. Data are mean ±(SD). **** represents p < 0.0001. e, Deletion of sla2∆ fails to suppress the neomycin sensitivity of neo1 mutants neo1S221L. The data represent growth relative to WT cells without the drug. Mixed model analysis was performed to test the variance and comparisons with neo1S221L were made with Tukey’s multiple comparison test.(n = 3 three biologically independent experiments, Data are mean ± SD, ns represents not significant). Source data