Characterization of a PIP Binding Site in the N-Terminal Domain of V-ATPase a4 and Its Role in Plasma Membrane Association

Vacuolar ATPases (V-ATPases) are multi-subunit ATP-dependent proton pumps necessary for cellular functions, including pH regulation and membrane fusion. The evidence suggests that the V-ATPase a-subunit’s interaction with the membrane signaling lipid phosphatidylinositol (PIPs) regulates the recruitment of V-ATPase complexes to specific membranes. We generated a homology model of the N-terminal domain of the human a4 isoform (a4NT) using Phyre2.0 and propose a lipid binding domain within the distal lobe of the a4NT. We identified a basic motif, K234IKK237, critical for interaction with phosphoinositides (PIP), and found similar basic residue motifs in all four mammalian and both yeast a-isoforms. We tested PIP binding of wildtype and mutant a4NT in vitro. In protein lipid overlay assays, the double mutation K234A/K237A and the autosomal recessive distal renal tubular-causing mutation K237del reduced both PIP binding and association with liposomes enriched with PI(4,5)P2, a PIP enriched within plasma membranes. Circular dichroism spectra of the mutant protein were comparable to wildtype, indicating that mutations affected lipid binding, not protein structure. When expressed in HEK293, wildtype a4NT localized to the plasma membrane in fluorescence microscopy and co-purified with the microsomal membrane fraction in cellular fractionation experiments. a4NT mutants showed reduced membrane association and decreased plasma membrane localization. Depletion of PI(4,5)P2 by ionomycin caused reduced membrane association of the WT a4NT protein. Our data suggest that information contained within the soluble a4NT is sufficient for membrane association and that PI(4,5)P2 binding capacity is involved in a4 V-ATPase plasma membrane retention.


Introduction
Vacuolar ATPases (V-ATPases) are ATP-dependent proton pumps that maintain the luminal pH of intracellular organelles [1][2][3][4]. V-ATPases in specialized cells, such as epididymis, kidney and bone cells, are targeted to the plasma membrane for extracellular pH regulation [5][6][7][8], or in the case of metastasizing cells, for extracellular acidification that enables invasiveness [9][10][11]. Eukaryotic V-ATPases consist of 14 subunits distributed into two sectors (V 1 and V o ), with many subunits having multiple isoforms [1,2,12,13]. V 1 contains nucleotide binding sites and is responsible for ATP hydrolysis. ATP hydrolysis creates a driving force, resulting in the rotation of a membrane-bound barrel, composed 32% identity with the a4NT isoform. Using the structure alignment, we identified a putative PIP binding site in the distal lobe of a4NT ( Figure 1A, inset), which shares a similar secondary structure arrangement with the basic pocket of the PH domain in Osh3, a defined PI(4)P-binding protein (PDB: 4ic4). Compared to the Osh3 PH domain [46], the putative binding site in a4 adopts a more compact incomplete barrel structure, consisting of three parallel β-strands and two α-helices. Considering the importance of basic or aromatic amino acids in PIP interaction in non-canonical PIP binding domains, we identified the basic motif K 234 IKK 237 within the putative binding domain ( Figure 1B). This basic motif, which may be critical for the interaction with the PIP acidic head groups, is accessible to the plasma membrane lipid and is conserved in all four mammalian NTs, as well as yeast Vph1p and Stv1p. There are basic/aromatic residues variation within this motif which we hypothesize confers PIP specificity to different a isoforms (Figure 2A).

Homology Model of a4NT Revealed a Putative Lipid Binding Motif K 234 IKK 237 Conserved in Yeast and All Four Mammalian Isoforms
We used Phyre2.0 to generate a three-dimensional structure model of the cytoplasmic N-terminal domain of the human a4 isoform (a4NT) with the available cryo-electron-microscopy-generated structure of yeast Vph1p-NT (PDB: 3j9t.b) ( Figure 1A), which shares 32% identity with the a4NT isoform. Using the structure alignment, we identified a putative PIP binding site in the distal lobe of a4NT ( Figure 1A, inset), which shares a similar secondary structure arrangement with the basic pocket of the PH domain in Osh3, a defined PI(4)P-binding protein (PDB: 4ic4). Compared to the Osh3 PH domain [46], the putative binding site in a4 adopts a more compact incomplete barrel structure, consisting of three parallel β-strands and two α-helices. Considering the importance of basic or aromatic amino acids in PIP interaction in non-canonical PIP binding domains, we identified the basic motif K 234 IKK 237 within the putative binding domain ( Figure 1B). This basic motif, which may be critical for the interaction with the PIP acidic head groups, is accessible to the plasma membrane lipid and is conserved in all four mammalian NTs, as well as yeast Vph1p and Stv1p. There are basic/aromatic residues variation within this motif which we hypothesize confers PIP specificity to different a isoforms (Figure 2A).  . Inset: the putative phosphoinositide binding domain consists of three parallel β-strands and two α-helices. Highlighted in green is the conserved binding motif K 234 IKK 237 . (B) Fitting of the a4NT model (red) in the place of Vph1 N-terminal domain in the available cryo-EM generated structure of S. cerevisae V-ATPase (PDB: 3j9t) (cyan), with the PIP binding motif highlighted in green, and the side chain of two residues, K234 and K237, is shown to implicate the plasma membrane access (inset: enlarged view) (red: a4NT, cyan: S. cerevisae V-ATPase, green: key residues). motif K 234 IKK 237 . (B) Fitting of the a4NT model (red) in the place of Vph1 N-terminal domain in the available cryo-EM generated structure of S.cerevisae V-ATPase (PDB: 3j9t) (cyan), with the PIP binding motif highlighted in green, and the side chain of two residues, K234 and K237, is shown to implicate the plasma membrane access (inset: enlarged view) (red: a4NT, cyan: S.cerevisae V-ATPase, green: key residues).

-Enriched Liposomes
To assess our putative PIP binding motif, we generated two mutations in a4NT: K234A/K237A and K237del. Both mutations are in the putative binding motif and K237del is a human mutation causing distal renal acidosis [44].
We tested in vitro PIP interactions of wildtype and mutant a4NTs with a protein-lipid overlay assay and a PIP-enriched liposome pull-down assay. The PIP overlay assay demonstrated significant decreases in the PIP binding of both mutants compared to wildtype ( Figure 2B,C). The results do not show any statistically significant PIP specificity for the a4 isoform. This may reflect the dimensional constraints of the PIPs being blotted onto the nitrocellulose membrane in the overlay assay; membrane curvature is often required for lipid specificity in PIP interaction [47,48] and is absent in overlay blot assays.
As a4-containing V-ATPases are targeted to the plasma membrane of renal intercalated cells, we were particularly interested in the interaction of the a4 isoform with plasma membrane PIPs. To mimic the plasma membrane environment and structure, we used PolyPIPosomes liposomes enriched with PI(3,4)P 2 , PI(4,5)P 2 and PIP3, which are mostly found in the plasma membrane. Using liposome pull-down assays, wildtype and mutant K234A/K237A were incubated with these different PolyPIPosomes. While the K234A/K237A mutants reduced association with all three liposomes, only the associations with the PI(4,5)P 2 -enriched liposomes were statistically significant ( Figure 3A,B). PI(4,5)P 2 is concentrated in the plasma membrane and plays a major role in both the regulation and recruitment of plasma membrane proteins. We also assessed the binding of the human mutation K237del with PI(4,5)P 2 -enriched liposomes; we observed a significant decrease in the liposome binding of the deletion mutant compared to wildtype ( Figure 3B). Interestingly, when testing the binding of a4 proteins to PI(4)P, a PIP not enriched in the plasma membrane but rather enriched in the Golgi, there was no significant difference in the binding between wildtype and K237del ( Figure 3C). do not affect protein secondary structure. Circular dichroism (CD) spectra of a4NT wildtype (black) and mutant K234A/K237A (blue) in 50 mM Tris pH8.0, and in addition of 10 mM SDS (wildtype: green; mutant: red).

Mutation within the Putative a4-PIP Binding Motif Reduced In Vitro PIP Binding and Interactions with PI(4,5)P2-Enriched Liposomes
To assess our putative PIP binding motif, we generated two mutations in a4NT: K234A/K237A and K237del. Both mutations are in the putative binding motif and K237del is a human mutation causing distal renal acidosis [44].
We tested in vitro PIP interactions of wildtype and mutant a4NTs with a proteinlipid overlay assay and a PIP-enriched liposome pull-down assay. The PIP overlay assay demonstrated significant decreases in the PIP binding of both mutants compared to wildtype ( Figure 2B,C). The results do not show any statistically significant PIP specificity for the a4 isoform. This may reflect the dimensional constraints of the PIPs being blotted onto the nitrocellulose membrane in the overlay assay; membrane curvature is often required for lipid specificity in PIP interaction [47,48] and is absent in overlay blot assays.
As a4-containing V-ATPases are targeted to the plasma membrane of renal intercalated cells, we were particularly interested in the interaction of the a4 isoform with plasma membrane PIPs. To mimic the plasma membrane environment and structure, we used PolyPIPosomes liposomes enriched with PI(3,4)P2, PI(4,5)P2 and PIP3, which are mostly found in the plasma membrane. Using liposome pull-down assays, wildtype and mutant K234A/K237A were incubated with these different PolyPIPosomes. While the K234A/K237A mutants reduced association with all three liposomes, only the associations with the PI(4,5)P2-enriched liposomes were statistically significant ( Figure 3A,B). PI(4,5)P2 is concentrated in the plasma membrane and plays a major role in both the regulation and recruitment of plasma membrane proteins. We also assessed the binding of the human mutation K237del with PI(4,5)P2-enriched liposomes; we observed a significant decrease in the liposome binding of the deletion mutant compared to wildtype ( Figure 3B). Interestingly, when testing the binding   To ask whether the mutations affected PIP binding or overall protein structure, we assessed protein folding using circular dichroism (CD) spectroscopy of both WT and mutant protein in an aqueous environment, in the presence and absence of SDS ( Figure 2D). WT a4NT contains largely a helical structure, exhibited by two negative minima at 222 nm and 208 nm, consistent with the structural model of a4NT. The CD spectra of the mutants were comparable to the wildtype with overlapping peaks at those specific wavelengths, indicating that the mutations disrupt PIP binding without altering protein structure (Figure 2D). In CD spectroscopy, the presence of micelles enhance helicity characteristics of a membrane-bound protein by creating a membrane-like environment. In this study, we added 10 mM SDS (SDS to peptide ratio was 370:1) to determine whether the presence of SDS micelles could increase the helicity of the proteins. The CD spectra of both WT and mutant a4 showed deeper minima at 208 nm and 222 nm, suggesting that the cytosolic WT and mutant proteins have membrane-bound protein characteristics ( Figure 2D). The non-overlapping positive spectra at 190-200nm could be accounted for by differences in the trace amount of salt or imidazole in the protein buffer. As the secondary protein structure appears unaffected by the mutations, our results suggest that the K 234 IKK 237 motif is, in part, involved in the interaction with PIPs, and mutations within this motif reduce interaction with PI(4,5)P2.

K234A/K237A and K237del Mutations Reduce a4NT-Membrane Association In Vivo
To assess the effect of a4NT mutations on membrane association in vivo, the FLAGtagged wildtype and mutants K234A/K237A and K237del a4NT were expressed in HEK293 cells and subcellular fractionation experiments were performed to determine whether the mutations affected protein retention in microsomal fractions. There was a statistically significant decrease in the amount of K237del and K234A/K237A mutants in the microsomal fraction compared to wildtype ( Figure 4A,B). To ask whether the mutations affected PIP binding or overall protein structure, we assessed protein folding using circular dichroism (CD) spectroscopy of both WT and mutant protein in an aqueous environment, in the presence and absence of SDS ( Figure 2D). WT a4NT contains largely a helical structure, exhibited by two negative minima at 222 nm and 208 nm, consistent with the structural model of a4NT. The CD spectra of the mutants were comparable to the wildtype with overlapping peaks at those specific wavelengths, indicating that the mutations disrupt PIP binding without altering protein structure ( Figure 2D). In CD spectroscopy, the presence of micelles enhance helicity characteristics of a membranebound protein by creating a membrane-like environment. In this study, we added 10 mM SDS (SDS to peptide ratio was 370:1) to determine whether the presence of SDS micelles could increase the helicity of the proteins. The CD spectra of both WT and mutant a4 showed deeper minima at 208 nm and 222 nm, suggesting that the cytosolic WT and mutant proteins have membrane-bound protein characteristics ( Figure 2D). The non-overlapping positive spectra at 190-200 nm could be accounted for by differences in the trace amount of salt or imidazole in the protein buffer. As the secondary protein structure appears unaffected by the mutations, our results suggest that the K 234 IKK 237 motif is, in part, involved in the interaction with PIPs, and mutations within this motif reduce interaction with PI(4,5)P2.

K234A/K237A and K237del Mutations Reduce a4NT-Membrane Association In Vivo
To assess the effect of a4NT mutations on membrane association in vivo, the FLAGtagged wildtype and mutants K234A/K237A and K237del a4NT were expressed in HEK293 cells and subcellular fractionation experiments were performed to determine whether the mutations affected protein retention in microsomal fractions. There was a statistically significant decrease in the amount of K237del and K234A/K237A mutants in the microsomal fraction compared to wildtype ( Figure 4A,B).  Immunofluorescence microscopy was used to assess whether the mutations affect membrane localization of the proteins. Wildtype FLAG-tagged a4NT (green) are enriched in the vicinity of the plasma membrane, labelled with wheat germ agglutinin (WGA, red) ( Figure 5A, top panel). This indicates that even though a4NT lacks a transmembrane domain, cytosolic a4NT contains sufficient information for plasma membrane retention. By contrast, both mutant proteins were mainly in the cytosolic fraction and did not appear in the vicinity of the plasma membrane ( Figure 5A, middle horizontal panels). Quantification by assessing protein signal intensity at the vicinity of the plasma membrane marker revealed a 75% decrease in signal detected at the plasma membrane for both mutants (Figure 5B), consistent with the in vitro data above. Immunofluorescence microscopy was used to assess whether the mutations affect membrane localization of the proteins. Wildtype FLAG-tagged a4NT (green) are enriched in the vicinity of the plasma membrane, labelled with wheat germ agglutinin (WGA, red) ( Figure 5A, top panel). This indicates that even though a4NT lacks a transmembrane domain, cytosolic a4NT contains sufficient information for plasma membrane retention. By contrast, both mutant proteins were mainly in the cytosolic fraction and did not appear in the vicinity of the plasma membrane ( Figure 5A, middle horizontal panels). Quantification by assessing protein signal intensity at the vicinity of the plasma membrane marker revealed a 75% decrease in signal detected at the plasma membrane for both mutants ( Figure 5B), consistent with the in vitro data above.

Ionomycin Reduces a4NT-Membrane Association In Vivo
We have observed that mutations within the putative lipid binding motif disrupt membrane association both in vitro and in vivo. As further proof of PIP involvement, we then asked whether modification of the plasma membrane phosphoinositides could affect a4NT membrane localization. Treatment of cells with ionomycin, a Ca 2+ ionophore, leads to the activation of PLCs, resulting in the rapid breakdown of PI(4,5)P 2 [49]. To analyze the correlation between PI(4,5)P2 changes and the redistribution of the a4NT at the plasma membrane, HEK293 cells transfected with FLAG-tagged wildtype a4NT were treated with 5 µM of ionomycin at room temperature for 15 min. Wildtype a4NT (red) appear in some regions of the plasma membrane, labelled with anti-Na + /K + -ATPase antibodies (green); ionomycin treatment reduced wildtype a4NT signal at the plasma membrane ( Figure 6A). Quantification was performed by dividing the fluorescence of the membrane-associated WT signal with total protein fluorescence. There was an approximately 50% reduction in a4NT associated with membrane in cells treated with ionomycin ( Figure 6B) suggesting that depletion of the plasma membrane PI(4,5)P 2 reduces a4NT membrane association. As a control, we showed that plasma membrane localization of the PH domain of PLC, an established PI(4,5)P 2 -binding protein, was similarly reduced in the presence of ionomycin ( Figure 6C). Plasmids containing FLAG-tagged a4NT wildtype (WT), mutants K234A/K237A and K237del were transfected in HEK293 cells. Cells were fixed and stained for membrane marker Alexa-647 conjugated wheat germ agglutinin (WGA) (red), followed by permeabilization and staining for FLAGtagged proteins (green). White arrows indicate regions of plasma membrane localization. The scale bar represents 10 μm (inset at the top panel is enlarged area of colocalization). (B) Quantification of the green signal intensity at the vicinity of the plasma membrane (red). Data represent mean value of at least 20 cells assessed SEM from three independent experiments. Statistical significance was analyzed by one-way ANOVA with Dunnett's multiple comparisons test comparing mutants to WT. ** indicates p < 0.01, *** indicates p < 0.001.

Ionomycin Reduces a4NT-Membrane Association In Vivo
We have observed that mutations within the putative lipid binding motif disrupt membrane association both in vitro and in vivo. As further proof of PIP involvement, we then asked whether modification of the plasma membrane phosphoinositides could affect a4NT membrane localization. Treatment of cells with ionomycin, a Ca 2+ ionophore, leads to the activation of PLCs, resulting in the rapid breakdown of PI(4,5)P2 [49]. To analyze the correlation between PI(4,5)P2 changes and the redistribution of the a4NT at the plasma membrane, HEK293 cells transfected with FLAG-tagged wildtype a4NT were treated with 5 μM of ionomycin at room temperature for 15 min. Wildtype a4NT (red) appear in some regions of the plasma membrane, labelled with anti-Na + /K + -ATPase antibodies (green); ionomycin treatment reduced wildtype a4NT signal at the plasma membrane ( Figure 6A). Quantification was performed by dividing the fluorescence of the membrane-associated WT signal with total protein fluorescence. There was an approximately 50% reduction in a4NT associated with membrane in cells treated with ionomycin ( Figure 6B) suggesting that depletion of the plasma membrane PI(4,5)P2 reduces a4NT membrane association. As a control, we showed that plasma membrane localization of the PH domain of PLC, an established PI(4,5)P2-binding protein, was similarly reduced in the presence of ionomycin ( Figure 6C).  A pair t-test was run to analyze the significance in mean difference. ** indicates p < 0.001. (C) Treatment with ionomycin causes depletion of membrane PI(4,5)P2, resulting in mislocalization of PH-PLC. Plasmids containing GFPtagged PH-PLC were transfected in HEK293 cells. Cells were treated with 5 μM of ionomycin at room temperature for 15 min before harvest. Cells were fixed and permeabilized, then stained for membrane marker anti-Na + /K + -ATPase with Alexa Fluor 568 (red). Quantification of the resulting immunofluorescent images were performed by measuring the relative intensity of the red signal overlapping with the green signal to the total red signal. Data represent mean value of at least 20 cells assessed ±SEM from 3 independent experiments. A pair t-test was run to analyze the significance in mean difference. * indicates p < 0.05.

Discussion
Here, we show that the V-ATPase a4 isoform, predominantly localized in the plasma membrane of the renal intercalated cells in vivo, interacts with PI(4,5)P2, a PIP concentrated on the plasma membrane. To support the physiological significance of this interaction, mutations within a putative PIP binding site reduced the a4-PI(4,5)P2 interaction. Further, reduction of PI(4,5)P2 levels decreased a4NT plasma membrane localization. Studies in yeast have shown that V-ATPases containing the vacuolar isoform Vph1p can be recruited to membranes in a PI (3,5)P2-dependent manner [23]. These data all suggest that there is a PIP interaction site within the cytosolic N-terminal domain of the a-subunits that recognizes phosphoinositides in specific membranes and plays a role in the spatial regulation of V-ATPases.
Pleckstrin homology (PH) domains are a major type of membrane binding domains, and have been well characterized as binding modules to PIPs [50]. They consist of two sheets curving to form a barrel with basic key residues within the flexible loops connecting the -sheet, and the barrel structure is enclosed by a C-terminal -helix. However, many membrane proteins bind to PIPs without a canonical domain but rather use positively charged patches. Here, we proposed a putative PIP binding domain in the distal lobe of the N-terminal domain of the a-subunit, which has a similar structural arrangement to the basic pocket for PI headgroup binding in the Osh3 protein [46]. Screening for basic or aromatic patches that facilitate the interaction with PIP headgroup, we identified the basic motif, (K/R)X(K/R)(K/R), within the putative binding domain. This motif is conserved in all four human a-subunit isoforms and yeast Vph1p and Stv1p, with some variations in A pair t-test was run to analyze the significance in mean difference. ** indicates p < 0.001. (C) Treatment with ionomycin causes depletion of membrane PI(4,5)P 2 , resulting in mislocalization of PH-PLC. Plasmids containing GFP-tagged PH-PLC were transfected in HEK293 cells. Cells were treated with 5 µM of ionomycin at room temperature for 15 min before harvest. Cells were fixed and permeabilized, then stained for membrane marker anti-Na + /K + -ATPase with Alexa Fluor 568 (red). Quantification of the resulting immunofluorescent images were performed by measuring the relative intensity of the red signal overlapping with the green signal to the total red signal. Data represent mean value of at least 20 cells assessed ±SEM from 3 independent experiments. A pair t-test was run to analyze the significance in mean difference. * indicates p < 0.05.

Discussion
Here, we show that the V-ATPase a4 isoform, predominantly localized in the plasma membrane of the renal intercalated cells in vivo, interacts with PI(4,5)P 2 , a PIP concentrated on the plasma membrane. To support the physiological significance of this interaction, mutations within a putative PIP binding site reduced the a4-PI(4,5)P 2 interaction. Further, reduction of PI(4,5)P 2 levels decreased a4NT plasma membrane localization. Studies in yeast have shown that V-ATPases containing the vacuolar isoform Vph1p can be recruited to membranes in a PI(3,5)P 2 -dependent manner [23]. These data all suggest that there is a PIP interaction site within the cytosolic N-terminal domain of the a-subunits that recognizes phosphoinositides in specific membranes and plays a role in the spatial regulation of V-ATPases.
Pleckstrin homology (PH) domains are a major type of membrane binding domains, and have been well characterized as binding modules to PIPs [50]. They consist of two βsheets curving to form a barrel with basic key residues within the flexible loops connecting the β-sheet, and the barrel structure is enclosed by a C-terminal α-helix. However, many membrane proteins bind to PIPs without a canonical domain but rather use positively charged patches. Here, we proposed a putative PIP binding domain in the distal lobe of the N-terminal domain of the a-subunit, which has a similar structural arrangement to the basic pocket for PI headgroup binding in the Osh3 protein [46]. Screening for basic or aromatic patches that facilitate the interaction with PIP headgroup, we identified the basic motif, (K/R)X(K/R)(K/R), within the putative binding domain. This motif is conserved in all four human a-subunit isoforms and yeast Vph1p and Stv1p, with some variations in the amino acid composition, which, we hypothesize, confers the specificity of different isoforms ( Figure 2D). Interestingly, fitting of the a4NT homology model in the completed structure of V-ATPases showed that the side chain of the basic residues oriented outward, allowing access to membrane PIP headgroups ( Figure 1B). Furthermore, mutations within these motifs have been identified in V-ATPase-related diseases, including a4.K237del found in patients with distal renal tubular acidosis [45], and a2. K237_V238del associated with cutis laxa [15]. Taken together, these data indicate a basic motif involved in V-ATPase a-subunit-PIP interactions and V-ATPases regulation. In a yeast study, Banerjee et al. suggested that residue K84 within Stv1p is critical for PI(4)P interaction [22]. The K84 residue resides on a flexible loop exposed to the membrane and is within the Stv1p Golgi localization signal-W 83 KY [21]; however, this residue is not conserved in humans and Vph1p. Studies of PIP binding domains indicate that highly conserved basic motifs are necessary, but not exclusive, for interaction with PIPs [51]. Further, protein folding plays an additional key role in bringing distant residues in the sequence to closer proximity in 3D structure, which altogether strengthens protein-lipid interactions. We hypothesize that in Stv1p, the putative binding motif and the K84 residues work cooperatively to strengthen and/or define the specificity of the protein and lipid interaction.
We present evidence that the cytosolic N-terminal domain of the a4 is sufficient for membrane association and that mutations within the motif K 234 IKK 237 disrupt binding to the plasma membrane enriched PIP, PI(4,5)P 2 . This is consistent with previous work which showed that the targeting signal for V-ATPases lies within the N-terminal domain of the a-subunit [21] and V-ATPases containing different isoforms of the a-subunit are targeted at different locations within the cells. We hypothesize that the variation in the basic motifs may, in part, account for the different PIP specificity of the four isoforms, conferring differential regulation of V-ATPases at specific membranes.
Each of the seven PIP species has a unique cellular distribution that renders them a lipid address for different organelles. Multiple studies have shown that PIPs are key regulators in membrane trafficking, responsible for recruitment of signaling effectors. Here, we show that reducing plasma membrane PI(4,5)P 2 concentrations using ionomycin decreased a4NT plasma membrane localization of the a subunit. Studies using the metastatic breast cancer cell lines MBA-231 showed that enhanced a3 and a4 expression and upregulation of PI(4,5)P 2 were both critical for formation of invadopodia and cell invasion [52]; this infers that PIP-a-subunit interactions are required for V-ATPase involvement in metastasizing cells and suggests the PIP binding site as a novel target for cancer therapeutics. In summary, we propose a lipid binding domain within the N-terminal domain of the V-ATPase a-subunit and present in vitro and in vivo evidence that a4NT can bind to PIPs through this domain. Further studies are now required to elucidate whether PIP-V-ATPase interactions confer either spatial or enzymatic regulation or a combination of both.

Protein Structure Prediction
A homology model of the N-terminus of the a4 isoform (amino acid 1-420) was constructed using Phyre2.0 with reference coordinates of S. cerevisiae V-ATPase state 1 Vph1p-NT (PDB: 3j9t.b). Confidence interval of the model was 100%. Sequence alignment of all 4 human a-isoforms and yeast Vph1p, Stv1p were obtained by Chimera.

Protein-Lipid Overlay Assay
PIP Array membranes (Echelon, Santa Clara, CA, USA) were blocked overnight at 4 • C in blocking buffer (10 mM Tris pH 8.0, 150 mM NaCl, 0.1% Tween-20, 3% fatty acidfree BSA). A total of 12 µg of purified proteins was diluted in 10 mL blocking buffer and incubated with the membrane for 2 h at 4 • C. The membranes were washed 3 times for 10 min with cold TBS-T (10 mM Tris pH 8.0, 150 mM NaCl, 0.1% (v/v) Tween-20). PIP membranes were then incubated with mouse anti-His antibody (1:2000 dilution) in blocking buffer for 1 h, washed 3 times in cold TBS-T, and incubated with horseradish peroxidaseconjugated anti-mouse immunoglobulin G in blocking buffer for 1h. After 3 washes with cold TBS-T, membranes were developed by enhanced chemiluminescence immunoblotting detection reagent (GE Healthcare, Chicago, IL, USA) and exposed for 30 s.

PolyPIPosome Pull-Down Assay
A total of 10 µg of purified proteins was incubated with 20 µL 1 mM PI(4,5)P 2 -PolyPIPosome (Echelon) and 200 µL of binding buffer (50 mM Tris pH 8.0, 150 mM NaCl and 0.05% Nonidet P-40) and incubate rotating at 4 • C for 2 h. The protein-bound liposomes were pelleted by centrifugation at 13,000 rpm for 10 min, followed by 3 washes with 200 µL of binding buffer each. Pellets were resuspended in 20 µL of binding buffer and 20 µL of 2× SDS sample buffer, separated by SDS-PAGE and transferred to nitrocellulose membrane. Membranes were then incubated with mouse anti-His antibody (1:5000 dilution) in TBS-T for 1 h, washed another 3 times with TBS-T and incubated with horseradish-peroxidaseconjugated anti-mouse immunoglobulin G in TBS-T for 1 h. After 3 washes with TBS-T, membranes were developed by enhanced chemiluminescence immunoblotting detection reagent (GE Healthcare) and exposed for 5-10 s.

HEK293 Transfection and Cellular Fractionation
HEK293 cells (ATCC) were cultured on 10 cm culture dishes in Dulbecco's modified Eagle's medium (DMEM) (Gibco) containing 10% fetal bovine serum (FBS) and 0.5% antibiotics, and grown in a 95% air, 5% CO 2 humidified environment at 37 • C. pcDNA3 plasmids of human a4 N-terminal domain (amino acid 1-421) wildtype and mutant K234A/K237A, 5 µg of plasmid/dish, were transfected into HEK293 cells using PolyJet Reagent (SignaGen, Frederick, MD, USA) in accordance with the procedure recommended by the manufacturer. Cells were harvested 30 h post transfection and pelleted by centrifugation (100× g). Cells were resuspended in homogenized buffer (250 mM sucrose, 1 mM EDTA, 10 mM HEPES, protease inhibitor cocktails) and lysed in a Dounce homogenizer. Lysates were separated by low-speed centrifugation (1000× g) for 10 min. Supernatants were subjected to ultracentrifugation (80,000× g) for 1 h to collect microsomal fractions and cytosolic fractions. Microsomal pellets were resuspended in 100 µL of homogenized buffer. The fractions were analyzed by immunoblots as described above.

Immunofluorescence
HEK293 cells were cultured as described above and seeded on coverslips in 6-well plates. Cells were co-transfected with plasmids of human a4 N-terminal domain (WT and mutants) and fluorescent proteins using PolyJet Reagent (SignaGen) in accordance with the procedure recommended by the manufacturer. Cells were fixed with 4% paraformaldehyde (PFA), permeabilized with 0.2% Triton-X and stained with anti-Flag antibody (Abcam, Cambridge, UK). For ionomycin treatment, cells were treated with 5 µM of ionomycin at room temperature for 15 min right before fixing. Images were acquired with a confocal microscope (Leica Confocal SP8, Wetzlar, Germany) using the 63× oil objective. Colocalization were measured with Mander's coefficient M1, which calculates the percentage of total signal from the green channel which overlaps with the signal from the red channel [53].

Statistical Analysis
GraphPad Prism 9.4.1 software was used for statistical analysis and statistical graph production. One-way ANOVA followed by Dunnett's multiple comparison test or Student's t-test were used as indicated in figure legends. In figure, asterisks were used as follows: * indicates p < 0.05, ** indicates p < 0.01, *** indicates p < 0.001. The experimental results are expressed as the mean ± SEM.
Author Contributions: A.C. performed all experiments except those represented in Figure 2D, which were performed by M.G. and Figure 3C