Human V-ATPase a-subunit isoforms bind specifically to distinct phosphoinositide phospholipids

Abstract

Vacuolar H⁺-ATPases (V-ATPases) are highly conserved multisubunit enzymes that maintain the distinct pH of eukaryotic organelles. The integral membrane a-subunit is encoded by tissue- and organelle-specific isoforms, and its cytosolic N-terminal domain (aNT) modulates organelle-specific regulation and targeting of V-ATPases. Organelle membranes have specific phosphatidylinositol phosphate (PIP) lipid enrichment linked to maintenance of organelle pH. In yeast, the aNT domains of the two a-subunit isoforms bind PIP lipids enriched in the organelle membranes where they reside; these interactions affect activity and regulatory properties of the V-ATPases containing each isoform. Humans have four a-subunit isoforms, and we hypothesize that the aNT domains of these isoforms will also bind to specific PIP lipids. The a1 and a2 isoforms of human V-ATPase a-subunits are localized to endolysosomes and Golgi, respectively. We determined that bacterially expressed Hua1NT and Hua2NT bind specifically to endolysosomal PIP lipids PI(3)P and PI(3,5)P₂ and Golgi enriched PI(4)P, respectively. Despite the lack of canonical PIP-binding sites, we identified potential binding sites in the HuaNT domains by sequence comparisons and existing subunit structures and models. We found that mutations at a similar location in the distal loops of both HuaNT isoforms compromise binding to their cognate PIP lipids, suggesting that these loops encode PIP specificity of the a-subunit isoforms. These data suggest a mechanism through which PIP lipid binding could stabilize and activate V-ATPases in distinct organelles.

The membrane bound organelles in eukaryotic cells each have a distinct, tightly regulated luminal pH important for organelle function and cell growth. This specific pH supports organelle identity and is established principally by vacuolar H⁺-ATPases (V-ATPases) (1, 2). V-ATPases are highly conserved multisubunit enzymes, functioning as ATP-driven proton pumps (3). They are comprised of 14 subunits organized in two operational subcomplexes, V₁ and V_o (4). The catalytic V₁ subcomplex is a peripheral membrane complex on the cytosolic face of the membrane and the V_o subcomplex is embedded in the membrane (5). The V₁ subcomplex executes ATP hydrolysis and the V_o subcomplex contains the proton pore (4, 6, 7). V-ATPases are ubiquitously expressed and localized to multiple organelles, including the Golgi, endosomes, secretory vesicles, lysosomes, and the lysosome-like yeast and plant vacuoles (2, 7). Their principal function is organelle acidification and through this function they impact processes such as receptor mediated endocytosis, protein processing, trafficking and degradation, neurotransmitter loading, and modulation of signaling pathways such as Notch and mechanistic target of rapamycin (2, 3, 8). In mammals, V-ATPases are also present in the plasma membrane of specific cells, including kidney intercalated cells, epididymal clear cells, and osteoclasts. In these settings, they acidify the extracellular space and function in pH homeostasis, bone resorption, urine acidification, and sperm maturation (9, 10, 11).

The a-subunit of the V_o domain of V-ATPases is a 100 kDa subunit, which consists of subunit amino terminal (aNT) and subunit carboxy-terminal (aCT) domains (Fig. 1A). The aNT is on the cytosolic face of the membrane and hydrophilic and the aCT is integral to the membrane and hydrophobic (4, 12). The aCT domain is involved in proton translocation (13, 14, 15). Cryo-EM studies establish that the aNT domain resembles a folded hairpin, comprised of globular proximal and distal ends connected by a coiled-coil linker, giving it a dumbbell shape (12). The aNT domain functions as a regulatory core for several reasons. It is positioned at the V₁-V_o interface and is capable of interacting with the three peripheral stator stalks in the assembled enzyme. In the intact, active V-ATPase, the aNT proximal and distal domains are pulled away from the central stalks because of their interactions with stator stalks (4, 16), while in autoinhibited V_o that is not bound to V₁, the aNT domain collapses toward the central stalk (12). These conformational changes affect both the proximal and distal ends of aNT and may be involved in V-ATPase regulation by reversible disassembly (17). In addition, aNT domains interact with cellular factors essential for V-ATPase regulation. These factors include RAVE, PIP lipids, and regulatory proteins like glycolytic enzymes (3, 18, 19). Significantly, the information for compartment specific targeting, localization and signaling of V-ATPase to specific cellular membranes also is encoded in the aNT domain (1, 15).

Figure 1.

Human V-ATPase structure and liposome pelleting assay.A, cryo-EM structure of V-ATPase from human embryonic kidney cell line HEK293F (PDB 6WM2). The cytosolic and transmembrane sectors are denoted as V₁ and V_o. The V_o aNT domain is shown in red and aCT domain in blue. The structure of the a1NT is shown on the left, with the proximal and distal ends indicated. B, overview of the liposome pelleting assay. Liposomes of defined lipid content are combined with varied protein concentrations, incubated, and subjected to ultracentrifugation as describe in Experimental procedures. Liposomes bound to protein are in pellet and unbound protein are in supernatant. The image is created with BioRender.com. aCT, subunit carboxy terminal; PI, phosphatidylinositol; saNT, subunit amino terminal; V-ATPase, vacuolar H⁺-ATPase.

Importantly, the V-ATPase a-subunits are expressed as multiple isoforms that are critical for targeting of V-ATPases to specific compartments and administer their manifold organelle-specific functions to V-ATPase (20, 21, 22). Yeast has two a-subunit isoforms, Vph1 and Stv1, which are localized at steady state to vacuoles and Golgi (5, 23). Higher organisms, including humans, have four isoforms designated a1-a4 (24). The a1, a2, and a3 isoforms are expressed ubiquitously but can also be enriched in certain organelles or cell types. The a1 isoform is present in lysosomes, but is also enriched in synaptic vesicles (25). The a2 isoform is predominantly localized to the Golgi apparatus but has also been found in endosomes (26). The a3 isoform is present in lysosomes, particularly secretory lysosomes, phagosomes, and the plasma membrane of osteoclasts (27, 28). The a4 isoform is specific to renal intercalated cells and epididymal clear cells (9). Genetic defects in V-ATPase activity due to loss of specific subunit isoforms are implicated in several diseases, including distal renal tubule acidosis (a4 mutations) (9), cutis laxa type II (a2 mutations) (29), and autosomal recessive osteopetrosis (a3 mutations) (30). Also, aberrant V-ATPase activity is implicated in cancer and viral infection (11, 31). This makes V-ATPase an attractive target for drug development.

Another important feature of eukaryotic organelle membranes is the enrichment of phosphatidylinositol phosphate (PIP) lipids. PIPs are transiently generated from phosphatidylinositol (PI) lipids in outer leaflets of organelle membranes by lipid kinases and phosphatases (32). Their levels are scarce, but they can modulate vesicular trafficking, activate or inhibit ion channels, pumps and transporters, and maintain organelle physiology (33, 34, 35). In eukaryotic cells, PIP lipids are enriched in specific intracellular membranes where they help to specify organelle identity (33, 36, 37, 38). PI(3)P is enriched in early endosomes, but is also present in late endosomes and lysosomes (36, 37, 39, 40). PI(4)P is the chief PIP lipid in the Golgi apparatus but is found in plasma membranes as well (25, 37). PI(3,5)P₂ is predominantly localized in yeast vacuoles and mammalian late endosomes, lysosomes, and multivesicular bodies (32, 37, 38, 41, 42). PI(4,5)P₂ is enriched in the plasma membrane (43).

V-ATPase activity is regulated by PIP lipids in yeast. V-ATPases containing the Vph1 a-subunit isoform are activated and stabilized by salt stress and high extracellular pH in a PI(3,5)P₂-dependent manner (44). Addition of short chain PI(3,5)P₂ to isolated vacuoles increases V-ATPase and proton pumping activity (45). Mutations in the distal domain of Vph1NT compromise this in vitro activation, and when present in intact Vph1, also impact the cellular response to osmotic stress (45). Elevating PI(3,5)P₂ levels by constitutive activation of Fab1 kinase or salt shock reversibly recruits Vph1NT to membranes in vivo in the absence of other subunits (44). Taken together, these data suggest that Vph1NT can interact with PI(3,5)P₂. PI(4)P is also essential for Stv1 localization and activity. In vitro, Stv1NT binds PI(4)P and this association is impaired by a mutation in the proximal domain of Stv1NT (1). When this mutation is introduced into full-length Stv1, or Golgi PI(4)P levels are lowered, Stv1-containing V-ATPases are mislocalized to the vacuole (1). Addition of PI(4)P was shown to enhance activity of Stv1-containing V-ATPases (46). These data highlight the important of PIP interactions with V-ATPase a-subunit isoforms.

In this paper, we aim to determine whether PIP binding is conserved in mammalian a-subunit isoforms, and to elucidate any PIP specificity of the human a1 and a2 isoforms. We have tested the specificity of N-terminal domain of human V_o a-subunit a1 (Hua1NT) and a2 (Hua2NT) for intracellular PIP lipids in a liposome pelleting assay and found that Hua1NT binds preferentially to liposomes containing PI(3)P and PI(3,5)P₂, lipids typically enriched in endosomes and lysosomes. In contrast, Hua2NT shows a preference for Golgi-enriched lipid PI(4)P. We have identified potential PIP-specific binding sites for both Hua1NT and Hua2NT in their distal domains. The data demonstrate the association between specific V-ATPase subunit isoforms and the PIP lipids enriched in the organelles where they reside. Defining PIP-binding codes on V-ATPase will improve our understanding of organelle specific pH control and could provide new avenues for controlling V-ATPase subpopulations.

Results

Hua1NT binds the endolysosomal PIP lipids PI(3)P and PI(3,5)P2 in vitro

The human a1 and a2 isoforms reside in endolysosomal compartments and Golgi, respectively, in most mammalian cells. Based on the results in yeast indicating that vacuolar Vph1NT binds to PI(3,5)P₂ and Stv1NT binds to PI(4)P, we hypothesized that the human a1 and a2 subunits might also bind to the predominant intracellular PIP lipids in their organelles of residence, including PI(3)P, PI(3,5)P₂, and PI(4)P. In addition, based on current structures and results with the yeast a-subunits, we hypothesized that the cytosolic N-terminal domains would be most likely to contain PIP lipid binding sites.

To investigate whether the Hua1NT and Hua2NT isoforms exhibit specific binding to PIP lipids, we expressed the NT domain only of each isoform (Fig. 1A) and tested their binding to liposomes in a quantitative liposome pelleting assay (Fig. 1B) (47). The Hua1NT domain (aa 1–356) was fused to maltose-binding protein (MBP) at its N terminus and a FLAG tag at its C terminus to generate MBP-Hua1NT-FLAG. The tagged protein was expressed in Escherichia coli and purified by sequential affinity chromatography on amylose and anti-FLAG columns, followed by size-exclusion chromatography. Purity was analyzed by SDS-PAGE (Fig. 2A, left). The Hua1NT construct eluted at the molecular mass of a monomer (M) from the gel filtration column (Fig. 2A, right). This fraction was used to characterize the in vitro binding specificity of Hua1NT and different PIP lipids. In this assay, different concentrations of the expressed protein were mixed with liposomes containing 5 mol% of the indicated PIP lipid, then incubated at room temperature to allow interaction, followed by ultracentrifugation (described in Experimental procedures). The same concentration of protein was centrifuged without any liposomes in order to determine whether there is any precipitation; this is the “protein only” sample. As an additional control, we incubate the proteins with liposomes without PIP lipids, but with an equal amount of phosphatidyl serine (PS) replacing the PIPs. This sample serves as a control for liposome binding that is not PIP-specific. For all samples, supernatant and pellet fractions were obtained as described in Experimental procedures and analyzed by SDS-PAGE and immunoblotting.

Figure 2.

Hua1NT binds the endolysosomal PIP lipids PI(3)P and PI(3,5)P₂in vitro.A, left, Coomassie-stained SDS-PAGE showing purified MBP-Hua1NT-FLAG after amylose, FLAG column, and gel filtration purification, with the molecular mass markers indicated. Right, elution profile of MBP-Hua1NT-FLAG from gel filtration column. The peak corresponding to the monomeric molecular mass (M) was collected and used for the liposome pelleting assay. B, anti-FLAG immunoblots of MBP-Hua1NT-FLAG in supernatant (S) and pellet (P) fractions collected after ultracentrifugation. Protein only samples contained no liposomes. Control samples used liposomes containing no PIP lipid. Experimental samples used liposomes containing 5% of the indicated PIP lipids. The experiments were performed using different protein concentrations as indicated to the right of the blots. C, stacked plots showing average proportion (±SEM) of protein bound to control and experimental samples. Control binding is indicated by gray area on the bottom of each stack, PIP-specific binding is indicated by the colored portion at the top. D, PIP-specific binding of MBP-Hua1NT-FLAG at different protein concentrations. The proportion of protein bound to control in C was subtracted from the total to give the PIP-specific binding at each protein concentration. At least three assays were performed for each protein concentration and PIP lipid, and each dot represents a distinct assay (biological replicate). Bars represent mean ± SEM. Statistical significance was determined by ordinary one-way ANOVA with Tukey’s multiple comparison test. ∗∗∗∗p < 0.0001, ∗∗∗p < 0.0005, ∗∗p < 0.005, ∗p < 0.05, and ns p > 0.05. HuaNT, N-terminal domain of human V_o a-subunit; MBP, maltose-binding protein; PIP, phosphatidylinositol phosphate.

Hua1NT interacted with both PI(3)P and PI(3,5)P₂ as indicated by the presence of bands in pellet (P) fraction at all concentrations (Fig. 2B). In contrast, Hua1NT pelleted much less with PI(4)P liposomes, suggesting weaker binding (Fig. 2B). The relative amount of protein in the supernatant and pellet fractions were determined by quantification of band intensity by ImageJ for each experiment and dividing the signal from the pellet over the total (supernatant + pellet) signal. The level of precipitation in the protein only sample was negligible. There was some variability in individual samples, but each experiment was repeated at least three times with freshly prepared protein and liposomes. The average proportion of total protein that bound to the PIP-containing and control liposomes over several experiments are plotted in Figure 2C. The control, PIP-independent binding (bottom of stacked bars for each protein concentration) is relatively low. In Figure 2D, PIP-independent binding is subtracted in order to compare the level of PIP-specific binding at each protein concentration. These data indicate that Hua1NT binds specifically PI(3)P and PI(3,5)P2 liposomes, but shows little binding to PI(4)P-containing liposomes. In addition, binding is nearly complete at the lowest concentration of protein measured (0.5 μM), indicating submicromolar affinity of the protein for these lipids. These data suggest that the Hua1 isoform, which resides in the endosomes and lysosomes of many cells, shows preferential binding to PI(3)P and PI(3,5)P₂, the lipids enriched in these organelles.

Hua2NT binds the Golgi PIP lipid PI(4)P in vitro

We next tested the specificity of Hua2NT for the same three PIP lipids. MBP-Hua2NT-FLAG (aa 1–364) was expressed in E. coli and purified similarly to Hua1NT. A Coomassie stained SDS-PAGE of purified MBP-Hua2NT-FLAG after affinity purification and gel filtration and its elution profile are depicted in Figure 3A. The liposome pelleting assay was performed to test the interaction of Hua2NT with liposomes of distinct PIP content. We found that at different concentrations, Hua2NT interacted with PI(4)P most strongly, as indicated by the bands in pellet (P) fractions (Fig. 3B). There is again negligible pelleting in the protein only sample. The pelleting of Hua2NT with PI(3)P and PI(3,5)P₂ was less than that with PI(4)P (Fig. 3B). The proportion of protein at different concentrations bound to the PIP lipids and control was determined as described above and is plotted in Figure 3C. The binding to control liposomes is relatively low (Fig. 3C) and was subtracted in Figure 3D. The comparison indicates Hua2NT binds specifically and with relatively high affinity to PI(4)P (Fig. 3D). These data suggest that the Hua2 isoform, which functions in Golgi, shows a preference for the Golgi-enriched PI(4)P.

Figure 3.

Hua2NT binds the Golgi PIP lipid PI(4)P in vitro.A, left, Coomassie-stained SDS-PAGE showing purified MBP-Hua2NT-FLAG after amylose, FLAG column, and gel filtration purification, with molecular mass markers. Right, elution profile of MBP-Hua2NT-FLAG from gel filtration column. The peak corresponding to the monomeric molecular mass (M) was collected and used for liposome pelleting assay. B, anti-FLAG immunoblots of MBP-Hua2NT-FLAG of supernatant and pellet fractions from protein only, control, and experimental liposome pelleting assays performed as described in Figure 2. C, stacked plots of the mean proportion of protein bound to control liposomes (gray, bottom) versus that specific to PIP-containing liposomes (colored, top). Error bars indicate SEM. D, proportion of PIP-specific binding of MBP-Hua2NT-FLAG to liposomes containing the indicated lipids at different protein concentrations. At least three assays were performed for each protein concentration and PIP lipid, and each dot represents a distinct biological replicate. Band intensities were determined by ImageJ. Statistical significance was determined by ordinary one-way ANOVA with Tukey’s multiple comparison tests. ∗∗∗∗p < 0.0001, ∗∗∗p < 0.0005, ∗∗p < 0.005, ∗p < 0.05, ns p > 0.05. HuaNT, N-terminal domain of human V_o a-subunit; MBP, maltose-binding protein; PIP, phosphatidylinositol phosphate.

Identification of candidate PIP lipid binding sites on human a1NT and a2NT

Once the binding specificities of Hua1NT and Hua2NT isoforms for different PIP lipids were established, we wanted to identify candidate PIP lipid binding sites in these isoforms. Identifying these binding sites will aid in understanding the functions of PIP lipids in vivo and recognizing PIP binding codes on V-ATPase subunit isoforms might aid in understanding regulation and targeting of V-ATPases. Many cytosolic or peripheral membrane proteins bind PIP lipids with an orthodox binding site like PX, PHD, or FYVE (32, 36). None of these canonical binding sites are present in yeast or human V-ATPase a-subunit isoforms. However, some membrane proteins can bind PIPs through noncanonical binding sites. To identify candidate PIP lipid binding sites, we used structure and homology modeling to identify aNT mutations for testing and then tested the mutant proteins for altered PIP lipid binding. The cryo-EM structure of human V-ATPase containing the a1 isoform from HEK293F cells (PDB 6WM2) has been reported (48). There is no structure for human V-ATPases containing the a2 isoform, so we used the molecular threading software PHYRE2 (sbg.bio.ic.ac.uk/phyre2.html) (49) to model Hua2NT on the Hua1NT structure. An overlay of the a1NT structure and a2NT model is depicted in Figure 4A. The two subunit isoforms are moderately conserved (48.2% identity) (Fig. S1). To identify candidate PIP lipid binding sites, we applied three major criteria. These include: (a) minimal homology between isoforms that show distinct PIP binding specificities, (b) presence of basic amino acid residues, capable of ionic interactions with negatively charged PIP head groups (possibly flanked by aromatic amino acids), and (c) predicted orientation of the binding sites toward the cytoplasmic leaflet of membrane, where the PIP lipids reside, in the assembled V-ATPase complex. In Hua1NT, Y²¹⁴VH and G²³⁹FR were identified as candidate binding sequences (Fig. 4B, left) that satisfied the criteria and in Hua2NT K²²¹WY and G¹⁸³KV were identified (Fig. 4B, right). All of these residues are present in the distal domain in the loops in Hua1NT and Hua2NT. Although yeast Stv1NT has a binding site in a proximal loop, the corresponding proximal loops Hua1NT and Hua2NT are shorter and have no basic residues.

Figure 4.

Identification of candidate PIP lipid binding sites on human a1NT and a2NT.A, overlay of ribbon diagrams for the Hua1NT structure and Hua2NT model. The gold ribbon represents Hua1NT, and the light blue ribbon represents Hua2NT. B, left, Hua1NT candidate PIP lipid binding sites Y²¹⁴VH and G²³⁹FR with side chains shown in red and orange. Right, candidate PIP lipid binding sites of Hua2NT, K²²¹WY, and G¹⁸³KV with side chains shown in blue and cyan. HuaNT, N-terminal domain of human V_o a-subunit; PIP, phosphatidylinositol phosphate.

Hua1NT requires Y²¹⁴VH sequence to bind PI(3)P and PI(3,5)P₂

The candidate binding sites identified in Hua1NT, Y²¹⁴VH, and G²³⁹FR were mutated using the primers listed in Table 1, to yield MBP-Hua1NT(Y²¹⁴VH-AVA)-FLAG and MBP-Hua1NT(G²³⁹FR-AAA)-FLAG fusion proteins. The purified mutant proteins had similar purification profiles to the WT Hua1NT (Fig. 5A, left), and the monomer fractions (M) from gel filtration (Fig. 5A, right) were again used for liposome pelleting assay. Since WT Hua1NT showed binding specificity to liposomes containing PI(3)P and PI(3,5)P₂, we tested the mutant proteins for binding to liposomes containing these PIP lipids. The immunoblots show that Hua1NT with Y²¹⁴VH—AVA mutation has reduced binding to liposomes containing PI(3)P and PI(3,5)P₂ as depicted by the more intense bands in supernatant (S) (Fig. 5B, left) and the lower proportion of bound protein (Fig. 5B, right). However, Hua1NT with G²³⁹FR—AAA mutation still binds to the liposomes containing PI(3)P and PI(3,5)P₂ as indicated by bands in the pellets (P) (Fig. 5C, left) and the relatively high proportion of protein bound to liposomes containing both lipids (Fig. 5C, right). The PIP-specific binding to both PI(3)P and PI(3,5)P₂ of the WT and mutant versions of Hua1NT are compared in Figure 5D. The graph shows that Hua1NT with mutation at Y²¹⁴VH site has almost no detectable binding for PI(3)P or PI(3,5)P₂ and Hua1NT with mutation at G²³⁹FR site has slightly reduced binding to both lipids at low concentrations, suggesting a somewhat lower affinity, but binds similarly to WT at higher concentrations (Fig. 5D). These data indicate that Hua1NT requires the Y²¹⁴VH sequence to bind PI(3)P and PI(3,5)P2 in vitro.

Table 1.

Oligonucleotides used in this study

Primer name	Sequence (5’ → 3”)
Hua1YVH-AVA F	CGACGCCGTGGCCAAGTCTGTGTTTATCATTTTCTTCC
Hua1YVH-AVA R	TTGGCCACGGCGTCGCCAGTCACAGGATC
Hua1NT(GFR-AAA)F	TGAAGCCGCCGCCGCCTCACTCTATCCCTGTCC
Hua1NT(GFR-AAA)R	GCGGCGGCGGCTTCACAGATTTTCTTGACTCTGT
a2K221Afor	TTGAAGACCCTGAAACAGGGGAAGTCATAGCATGGTATGTCTTTTTAATATCCTTTTGGGGA
a2K221Arev	TCCCCAAAAGGATATTAAAAAGACATACCATGCTATGACTTCCCCTGTTTCAGGGTCTTCA
Hua2NT(GKV-AAV)F	CCAAGCCGCCGTGGAAGCATTTGAAAAAATGTTGTGGA
Hua2NT(GKV-AAV)R	TCCACGGCGGCTTGGTTAATTAGGCCAGACACA

Abbreviation: HuaNT, N-terminal domain of human V_o a-subunit.

Figure 5.

Hua1NT requires Y²¹⁴VH sequence to bind PI(3)P and PI(3,5)P₂.A, left, Coomassie-stained SDS-PAGE showing purified MBP-Hua1(Y²¹⁴VH-AVA)NT-FLAG and MBP-Hua1(G²³⁹FR-AAA)NT-FLAG after gel filtration. Right, elution profiles of the two mutant proteins. The peaks corresponding to the monomeric molecular mass (M) were collected and used in the liposome pelleting assay. B, left, anti-FLAG immunoblot of MBP-Hua1(Y²¹⁴VH-AVA)NT-FLAG at different concentrations with supernatant (S) and pellet (P) fractions collected from protein only and control samples as described in Figure 2 or from liposomes 5% of PI(3)P or PI(3,5)P2 as indicated. Right, stacked plots showing the proportion of protein in the pellet in control samples (gray) or from samples containing the indicated PIP lipid (color) as described in Figure 2. C, left, anti-FLAG immunoblot of MBP-Hua1(G²³⁹FR-AAA)NT-FLAG at different concentrations in supernatant (S) and pellet (P) fractions collected from pelleting with liposomes containing 5% of the PIP lipids. Right, stacked plot indicating the proportion of total protein bound to control (gray) versus PIP-specific binding (colored portion) at each protein concentration. Error bars represent SEM. D, comparison of the proportion bound to PI(3)P (left) and PI(3,5)P2 (right) for each mutant to the WT protein shown in Figure 2. At least three assays were performed for each concentration and PIP lipid and each dot represents a distinct biological replicate. Comparison was done by ordinary one-way ANOVA with Tukey’s multiple comparison tests. ∗∗∗∗p < 0.0001, ∗∗∗p < 0.0005, ∗∗p < 0.005. HuaNT, N-terminal domain of human V_o a-subunit; MBP, maltose-binding protein; PI, phosphatidylinositol; PIP, phosphatidylinositol phosphate.

Hua2NT requires K²²¹WY sequence to bind PI(4)P

Interestingly, yeast Stv1, which resides in the Golgi, contains a high-affinity binding site for PI(4)P in a W⁸³KY sequence at its proximal end; mutation of only the lysine in this sequence abolished PI(4)P binding (1). Given the similarity to the K²²¹WY sequence in the distal end of Hua2NT, we mutated the lysine of the K²²¹WY sequence to alanine to yield MBP-Hua2NT(K²²¹A)-FLAG using the primers listed in Table 1. In addition, the G¹⁸³KV sequence, was mutated to generate an MBP-Hua2NT(G¹⁸³KV-AAV)-FLAG fusion protein. The mutated proteins had similar purification profiles to WT Hua2NT (Fig. 6A). The monomer fractions (M) were used in liposome pelleting assays. Since WT Hua2NT showed binding specificity for PI(4)P, we tested the binding of the mutants to PI(4)P-containing liposomes. The immunoblots show that Hua2NT with K²²¹A mutation binds poorly to PI(4)P-containing liposomes, and the low proportion of MBP-Hua2NT(K²²¹A)-FLAG in pellets (P) over several experiments (Fig. 6B). Comparison of the PIP-specific binding of Hua2NT mutant (K²²¹A) with binding of the WT Hua2NT to PI(4)P-containing liposomes (Fig. 6C) indicates significantly reduced binding of the mutant protein. For MBP-Hua2NT(G¹⁸³KV-AAV)-FLAG, the immunoblot and graph depicting proportion of protein bound (Fig. 6D) show that, curiously, Hua2NT with mutation at G¹⁸³KV sequence binds to PI(4)P lipids at low protein concentrations (0.5 μM), but seems to have reduced binding to PI(4)P at higher concentration of protein. This behavior is very different than that of Hua2NT with the K²²¹A mutation or the WT and mutant forms of Stv1NT and Hua1NT and will require further analysis to understand. However, it is clear that mutation of Hua2NT K²²¹ to A greatly decreases binding to PI(4)P.

Figure 6.

Hua2NT requires K²²¹to bind PI(4)P.A, left, Coomassie-stained SDS-PAGE showing purified MBP-Hua2(K²²¹WY-AWY)NT-FLAG and MBP-Hua2(G¹⁸³KV-AAV)NT-FLAG proteins with molecular mass marker. Right, elution profiles of MBP-Hua2(K²²¹WY-AWY)NT-FLAG and MBP-Hua2(G¹⁸³KV-AAV)NT-FLAG from gel filtration column. The peaks corresponding to the monomeric molecular mass (M) were collected and used for liposome pelleting assays. B, left, anti-FLAG immunoblot of MBP-Hua2(K²²¹WY-AWY)NT-FLAG in supernatant (S) and pellet (P) fractions collected from protein only, control, and liposomes containing 5% PI(4)P. Total protein concentrations are indicated to the right of the blots. The lines indicate nonadjacent lanes from the same blot. Right, stacked plots showing the proportion of total protein in the control (gray) and the PI(4)P-specific proportion (orange). C, comparison of PI(4)P-specific binding of WT Hua2NT (from Fig. 3) to that of the mutant K²²¹WY. Each dot represents a distinct biological replicate. Statistical significance was determined by ordinary one-way ANOVA with Tukey’s multiple comparison tests. ∗∗∗∗p < 0.0001, ∗∗∗p < 0.0005, ∗∗p < 0.005, ∗p < 0.05, and ns p > 0.05. D, left, anti-FLAG immunoblot of MBP-Hua2(G¹⁸³KV-AAV)NT-FLAG at different concentrations in supernatant (S) and pellet (P) fractions collected from controls or liposomes containing 5% PI(4)P. The lines indicate nonadjacent lanes from the same blot. Right, stacked plots of average proportion of total protein bound to control liposomes (gray) versus PI(4)P-specific binding (gold). HuaNT, N-terminal domain of human V_o a-subunit; MBP, maltose-binding protein; PI, phosphatidylinositol.

Different PIP binding specificities are encoded at a similar location in aNT isoform structures

These data highlight the importance of the Y²¹⁴VH sequence in Hua1NT and K²²¹WY in the Hua2NT sequence for binding to their respective PIP lipids. The sequence alignment in Figure 7A indicates that these two sequences are at very similar positions in the aligned sequences. A comparison of the three-dimensional structures indicates that both of the sequences lie in a poorly conserved loop at the distal end of the aNT isoforms. Remarkably, the sequence implicated in binding of yeast Vph1NT to PI(3,5)P₂ is in a very similar position as shown in Figure 7B. This distal domain loop is poorly resolved in most V-ATPase cryo-EM structures, suggesting that it may be mobile, but is resolved one of the yeast (50) and some of the human V-ATPase structures (48, 51). Overall structure of V-ATPases is highly conserved between organisms and between complexes containing different a-subunit isoforms. These data suggest that there is a variable loop in the distal domain of multiple aNT isoforms that has the capacity for binding to PIP lipids. Intriguingly, different sequences at this location appear to support distinct PIP specificity.

Figure 7.

Different PIP-binding specificities are encoded at a similar location in aNT isoform structures.A, sequence alignment of a region in the distal end of Vph1NT, Hua1NT, and Hua2NT. Sequences that compromised binding to PIP lipids when mutated are highlighted. B, ribbon diagram of distal end binding domains of Hua1NT, Hua2NT, and Vph1NT showing the respective binding sequences, Y²¹⁴VH (red), K²²¹WY (blue), and K²³¹TREYKHK (green). HuaNT, N-terminal domain of human V_o a-subunit; PIP, phosphatidylinositol phosphate.

The a1NT and a2NT domains recruit to compartments enriched in their cognate PIP lipids

All full-length V_o a-subunit isoforms initially enter the endoplasmic reticulum and undergo initial assembly steps before proceeding to their organelles of residence. However, in yeast, the cytosolic aNT domains, which never enter the endoplasmic reticulum, are recruited to organelle membranes in a PIP-specific manner. Significantly, this recruitment does not require binding to other V-ATPase subunits, since loss of membrane-integrated V_o a-subunits destabilizes other V_o subunits and isolated V₁ is unable to bind to V_o or Vph1NT (52, 53). Stv1NT-GFP is recruited to Golgi membranes that contain PI(4)P and Vph1NT-GFP is recruited to vacuolar/endosomal membranes that contain PI(3,5)P2. In order to determine whether mammalian aNT domains behave similarly, we fused the a1NT domain to GFP (a1NT-GFP) and transiently transfected the construct into 4T1 mouse breast cancer cells. We then compared its localization to LAMP2, a resident of the endolysosomal compartments enriched in PI(3)P and PI(3,5)P2, by indirect immunofluorescence microscopy. As shown in Figure 8A, the a1NT-GFP signal is present in multiple puncta that partially overlap with the LAMP2 signal. We also expressed the a1NT(A²¹⁴VA) mutant that had shown compromised binding to PI(3)P and PI(3,5)P2 in vitro (Fig. 8B) in this cell line and compared its localization with LAMP2. As shown in Figure 8B, the a1NT-GFP mutant also recruited to puncta. However, linescans from both WT and mutant a1NT-GFP compared to LAMP2 (Fig. 8, C and D), suggest that the a1NT WT has extensive overlap with LAMP2 signal, while the mutant a1NT exhibited less overlap. Although, we did not fully characterize the nature of the puncta, these data suggest that the a1NT domain can be recruited to its organelle of residence even in the absence of the the full V_o targeting machinery, as in yeast, and introduction of a mutation that weakens PI(3)P and PI(3,5)P2 binding compromises this recruitment.

Figure 8.

Localization of expressed a1NT-GFP in mouse 4T1 breast cancer cells.A, micrograph of a single 4T1 cell expressing a1NT-GFP probed with anti-GFP antibody (green) and anti-LAMP2 antibody (red) for visualization by indirect immunofluorescence microscopy. B, micrograph of a cells expressing a1NT-GFP containing the Y²¹⁴VH-AVA mutation visualized as in A. Both A and B are single images from near the center of a Z-stack, selected as having a readily visible nucleus. C, small section of the same images shown in A, illustrating where lines were drawn across the images using FIJI-ImageJ. Fluorescence intensities at positions along each line are plotted below the images, with intensities from the LAMP2 image plotted in blue and intensities from the a1NT image plotted in orange. D, same as C, except that the lines were drawn across portions of the images in B. Intensities from the LAMP2 image are again plotted in blue and those from the mutant a1NT image are plotted in orange.

Based on the results with Golgi Stv1NT in yeast, we hypothesized that in mammalian cells a2NT might be recruited to the Golgi apparatus, which is enriched in PI(4)P. As described above, we expressed a2NT-GFP in 4T1 cells. We then compared its localization with the Golgi marker TGN46 by indirect immunofluorescence microscopy. As shown in Figure 9A, both a2NT-GFP and TGN46 are present in a very similar perinuclear distribution that is clearly different than the distribution of a1NT-GFP (Fig. 8A). In addition, a2NT-GFP was also present at several ruffled regions of the plasma membrane where there is little or no staining by the TGN46 antibody. Interestingly, PI(4)P is present in the plasma membrane as well as the Golgi apparatus, but Stv1-GFP preferentially recognizes the Golgi pool. However, a2NT-GFP may be recognizing a PI(4)P-enriched plasma membrane domain that would not be occupied by TGN46. Further studies will be necessary to identify this domain, but it was consistently observed with a2NT-GFP. We also expressed the mutant a2NT that compromised binding to PI(4)P-containing liposomes in 4T1 cells. Figure 9B shows that the mutant a2NT has a distribution very similar to the WT, with extensive overlap with TGN46 as well as some plasma membrane staining. An overlay showing the cell body (Fig. S2) confirms that this staining is at or near the plasma membrane. It is worth noting that while yeast Stv1NT mutants that failed to bind to PI(4)P in vitro also failed to recruit to membranes in vivo (1), the yeast NT constructs were integrated into the genome and expressed at endogenous levels. In contrast, the transiently transfected human a1NT-GFP and a2NT-GFP constructs are likely overexpressed. The reduced affinity of the a2NT-GFP mutant observed in vitro may be sufficient to support recruitment of this protein to membranes in cells when it is present at high concentrations.

Figure 9.

Localization of expressed a2NT-GFP in mouse 4T1 breast cancer cells.A, micrographs of 4T1 cell expressing WT a2NT-GFP probed with anti-GFP antibody (green) and anti-TGN46 antibody (red) in indirect immunofluorescence microscopy. B, micrograph of 4T1 cell expressing the a2NT-GFP containing the K²²¹A mutation, probed as described in A. Both A and B are single images from near the center of a Z-stack, selected because as having a readily visible nucleus.

Discussion

PIP binding is a conserved feature of V-ATPase aNT domains

These data suggest that the capacity for PIP-specific binding is present in a-subunit isoforms from mammals as well as yeast, despite the lack of previously characterized PIP-binding domains in any of the a-subunit isoforms. However, there are both similarities and differences in the binding of yeast and human isoforms to PIP lipids.

Human a2NT and yeast Stv1 are both primarily in Golgi-localized V-ATPases. Both bind tightly and specifically to liposomes containing PI(4)P, the PIP lipid enriched in the Golgi apparatus. In addition, both a2NT-GFP and Stv1NT-GFP are recruited to Golgi membranes when expressed in cells. The mutations that compromised PI(4)P binding, K⁸⁴A in the W⁸³KY of yeast Stv1NT and mutation of K²²¹A in a K²²¹WY sequence in human a2NT, appear to occur in a similar sequence context but are present at opposite ends of the aNT domain. Stv1 is longer than most a-subunit isoforms across species, and much of this additional length is in insertions in the NT domain, including a longer and more basic membrane-adjacent loop at the proximal end that contains the W⁸³KY sequence. In the a2NT model, there are no basic amino acids in the shorter proximal loop. However, it is also notable that although mutation of Stv1 K⁸⁴ abolishes binding to PI(4)P liposomes in vitro and transfers PI(4)P binding to Vph1NT when added at a similar position, experiments with chimeras between regions of Stv1NT and Vph1NT suggest that the distal end of Stv1NT may also help support high-affinity binding to PI(4)P (54). In yeast, PI(4)P binding is linked to both activation of V-ATPase activity in vitro and retention or retrieval of Stv1-V-ATPases to the Golgi apparatus in vivo. We do not yet know the functional effects of PI(4)P binding to the mammalian a2 isoform. Transient transfection of full-length a-subunit isoforms, with or without PIP-binding mutations, is very likely to disrupt overall a-subunit isoform distribution and function. Introduction of the K²²¹A mutation into the genomic locus of human a2 (ATP6V0a2) will be necessary to determine whether the functional effects of PI(4)P binding are similar to those observed in yeast.

Comparison of the human a1NT and yeast Vph1NT reveals sequence conservation between the regions that compromise PI(3,5)P₂ binding in vitro when mutated (Fig. 7). In yeast, mutations in either of two adjacent sequences K²³¹TREY and K²³⁶HK block PI(3,5)P₂ activation of V-ATPases in isolated vacuoles and cause defects in PI(3,5)P₂-dependent osmotic stress responses (45). Interestingly, the human Y²¹⁴VH sequence implicated in binding of Hua1NT to PI(3,5)P₂ aligns with a Y²³⁵KH sequence in the middle of the two sequences implicated in yeast Vph1NT binding to PI(3,5)P₂ and both are followed by an additional lysine that was not mutated in this study, but could potentially contribute to binding (Fig. 7). In addition, both Vph1NT-GFP and a1NT-GFP recruit to endolysosomal compartments when expressed in cells. However, there are important differences as well. Although human a1NT binds tightly and specifically to liposomes containing either PI(3)P or PI(3,5)P₂, Vph1NT binds poorly to liposomes in vitro (54), suggesting a lower affinity for PIP lipids than Hua1NT. In addition, V-ATPase activity in yeast could be activated by PI(3,5)P₂ but not by PI(3)P, suggesting the two lipids have distinct effects on activity, and potential differences in binding (45). It is difficult to justify differences in specificity and affinity purely from the sequences in this region, as Vph1NT has additional basic amino acids including a lysine in the middle of the homologous sequence that could potentially contribute to binding to either or both of these PIP lipids (Fig. 7). For reasons described above for the a2 isoform, we have not yet determined the effects of PIP binding to the mammalian a1 isoform on V-ATPase activity. However, the effects of the a1 mutation on cellular distribution of a1NT-GFP are promising, and suggest that future genome editing to incorporate this mutation could yield interesting insights about effects of PIP binding on a1-containing V-ATPases. Functionally, the V-ATPase in the yeast vacuole plays a critical role in the osmotic stress response, which is characterized by dramatic (20-fold) changes in PI(3,5)P₂ levels in yeast (44, 55). In this context, the rather weak affinity of Vph1NT for PI(3,5)P₂ may be important for the robust regulatory response to the rapid rise of PI(3,5)P₂ with osmotic stress. Although PI(3,5)P₂ is also extremely important in the endolysosomal system of higher eukaryotes, the changes in lipid levels with environmental stress are less dramatic (56), so higher affinity binding could accommodate a basal level of interaction between V-ATPases and the PIP lipids and support the response to stress. In addition, it is also important to recognize that the affinities of aNT domains for liposomes in solution may not fully reflect the binding of aNT domains tethered to the membrane by attachment to aCT and thus confined in proximity to the lipid bilayer.

Notably, both Vph1NT and Hua1NT are likely to transit through the Golgi en route to endosomes and lysosomes, but neither binds significantly to PI(4)P. This is potentially consistent with PIP binding occurring when they reach their compartments of residence, possibly accompanied by activation or other functional effects.

How could PIP-binding sites in the distal domain of aNTs affect V-ATPase structure and activity?

As described above, it is remarkable that the human a1NT and a2NT show distinct PIP-specific binding that can be disrupted by mutations in sequences mapping to such a similar structural location (Fig. 7A). We have found that binding to PI(3,5)P₂ activates V-ATPases containing the Vph1 isoform and also stabilizes them against disassembly of the V₁ subcomplex from V_o (44). The evidence that Vph1NT is regulated by sequences at this same location suggests that PIP binding to the mammalian a-subunit isoform to this region could also be functionally important. Structures are available for both intact yeast V-ATPases and isolated V_o subcomplexes containing Vph1, and we compared the position of the loop containing the PI(3,5)P₂-binding sequence in both (Fig. 10). Both the proximal and distal domains of Vph1NT collapse toward the center of the V_o subcomplex in the absence of V₁ (12, 17). In this conformation, the loop implicated in PI(3,5)P2 binding is above the c-ring and does not appear to be readily accessible to membrane lipids. In contrast, Vph1NT is in a more extended conformation in the assembled Vph1-containing V-ATPases where it now would have access to membrane lipids (50, 57). In this extended conformation, Vph1NT interacts with V₁ subunits in two of the three peripheral stalks, as required for stable V-ATPase assembly. The distal end of Vph1NT, specifically, interacts with one of the stalks comprised of E and G subunits (4, 16). A similar extended conformation of Hua1NT is seen in structures of the intact V-ATPases containing this subunit (48, 51, 58). We hypothesize that PIP binding to the extended conformation of aNT helps to stabilize the assembled conformation. Reversible disassembly of V₁ from V_o, possibly initiated by departure of V₁ subunit C, is a characteristic feature of V-ATPases from multiple species (4). It is possible that the V-ATPase is poised for disassembly even in its assembled state, and stabilizing factors, such as PIP binding are able to favor or stabilize the assembled conformation.

Figure 10.

Position of the PI(3,5)P₂-binding sequence in the assembled yeast V-ATPase and disassembled V_osector. Comparison of position of PI(3,5)P₂-binding sequence (circled red) in assembled V-ATPase and autoinhibited V_o sector, both containing the Vph1 isoform. The figures have been adapted from PDB 7FDA for the fully assembled complex (50) and PDB 6C6L for the isolated V_o complex (12). PI, phosphatidylinositol; V-ATPase, vacuolar H⁺-ATPase.

There is still much to learn about the interaction between PIP lipids and V-ATPases, although this work indicates that the capacity for these interactions is conserved and focuses attention on the distal end of aNT isoforms as sites of PIP binding and regulation. Mutation of the endogenous isoform genes will be important in the future, given the presence of multiple a-subunit isoforms and the potential for overexpression to perturb isoform distribution or introduce functional compensation. In addition, although cryo-EM structures of V-ATPases have identified lipids within the V_o subcomplex, there is as yet no structure that shows direct interaction of the aNT distal end with lipid headgroups. Achieving such a structure may require restoration of PIP lipids after solubilization of V-ATPases, with the goal of preserving PIP interactions through reconstitution.

Experimental procedures

Cloning, expression, and purification of HuaNT isoforms from E. coli

MBP-Hua1NT(1–356)-FLAG fusion protein was generated by Anne Smardon (SUNY Upstate Medical University). PCR fragments containing the first 356 amino acids from human cDNA ATP6Voa1 were cloned into pGEM and excised with EcoRI and SalI, then cloned into pMAL:PPase cut with the same enzymes. Construction of MBP-Hua2NT-FLAG domain was described (1). The mutant constructs, MBP-Hua1(Y²¹⁴VH-AVA)NT-FLAG, MBP-Hua1(G²³⁹FR-AAA)NT-FLAG, MBP-Hua2(K²²¹WY-AWY)NT-FLAG and MBP-Hua2(G¹⁸³KV-AAV)NT-FLAG were constructed by introducing point mutations using In-Fusion HD Cloning (Takara Bio, Inc) with the primers listed in Table 1. The constructs were transformed into Rosetta [genotype: F− ompT hsdSB(rB− mB−) gal dcm (DE3) pRARE (CamR)] (Novogen) competent E. coli cells and grown to a density of A₆₀₀ of 0.6 in 2.5% Luria-Bertani broth (Fisher Bioreagents) with 125 μg/ml ampicillin (Sigma-Aldrich) and 34 μg/ml chloramphenicol(Sigma-Aldrich). Expression of protein was then induced by addition of 0.3 mM IPTG, followed by incubation at 19 °C for 16 h. The cells were pelleted at 3000 rpm at 4 °C for 20 min and resuspended in 25 ml amylose column buffer (20 mM Tris–HCl, pH 7.4, 0.2 M NaCl, and 1 mM EDTA). 2 mM MgCl₂, 1 mg/ml lysozyme, 2 μg/ml DNase, 5 mM DTT, 1 mM of PMSF were added, and the mixture was sonicated on ice for cell lysis, followed by slow rocking at 4 °C for 30 min. The lysate was centrifuged at 15,000 rpm at 4 °C for 20 min. The supernatant was diluted 1:4 with the same buffer and purified through 5 ml amylose and 1 ml FLAG affinity purification columns sequentially. Amylose resin was purchased from New England BioLabs and anti-FLAG M2 resin from Sigma. Purification in the amylose column involved running the lysate twice through the column, followed by a wash with ten column volumes of the same buffer. Proteins were eluted from amylose column with 10 mM maltose monohydrate. One milliliter fractions were collected and 5 mM DTT was added. Peak fractions with protein concentrations giving an A₂₈₀ >0.5 were pooled and dialyzed in dialysis tubing with a molecular weight cutoff of 3.5 kD (Spectrum Labs) in FLAG buffer (50 mM Tris–HCl, pH 7.4, and 150 mM NaCl) overnight at 4 °C. The dialyzed mixture was added to an anti-FLAG M2 column (Sigma), for the second round of purification. The column was washed with ten column volumes of FLAG buffer, and proteins were eluted as 1 ml fractions with 100 μg/ml FLAG peptide (Sigma). 5 mM DTT was added to each fraction and peak fractions as determined from Coomassie blue gel electrophoresis were pooled and concentrated in Vivaspin 15-max speed filter (molecular weight cutoff of 30 kD) (Sartorius Stedim Lab Ltd) to a final concentration of at least 1 mg/ml. The concentrated protein was run on a Superdex Increase 200 (Cytiva) gel-filtration FPLC column on Biologic DuoFlow System (Bio-Rad). The fraction corresponding to the peak corresponding to the molecular mass of the monomer protein (∼80 kDa) was collected and was used for liposome pelleting assay.

Liposome preparation and liposome pelleting assay

Lipids were purchased from Avanti Polar Lipids as lyophilized powder and dissolved in CHCl₃: CH₃OH: H₂O in 20:9:1 ratio. Liposomes with 50 mol% 16:0 phosphatidylcholine, 25 mol% 16:0 PS, 18 mol% 16:0 phosphatidylethanolamine, 2 mol% 16:0 nitrobenzoxadiazole-PE (for visualizing liposomes), and 5 mol% 18:1 phosphatidylinositol phospholipids PI(3)P, PI(4)P, or PI(3,5)P2, were prepared in liposome making buffer (50 mM Tris, 25 mM NaCl, pH 7.4) by extrusion as described by Banerjee and Kane, 2017 (1). Control liposomes, without PIP lipids were prepared with an extra 5 mol% 16:0 PS.

Liposome pelleting assays were performed by incubating the prepared liposomes with the expressed protein of interest at room temperature for 30 min to allow protein–lipid interaction. The salt concentration was adjusted to 60 mM with NaCl. The final volume of the mixture was 100 μl and samples contained protein concentrations of 0.5, 0.75, 1, and 1.5 μM with a final lipid concentration of 0.33 mM. Following incubation, the mixture was centrifuged at 95,000 rpm at 4 °C for 30 min in a TLA-100 fixed angle rotor using Beckman Coulter mini ultracentrifuge. The supernatant was collected, and the pellet was resuspended in 100 μl buffer with 50 mM Tris–HCl and 60 mM NaCl, pH 7.4. Both supernatant and pellet samples were precipitated with 10% trichloroacetic acid, and the resulting pellets were washed with cold acetone and resuspended in 30 μl cracking buffer (50 mM Tris–HCl, pH 6.8, 8 M urea, 5% SDS and 1 mM EDTA). Equal volumes of each sample were separated by SDS-PAGE and transferred into nitrocellulose membrane. The blots were probed with mouse monoclonal anti-FLAG M2 antibody (Sigma) and anti-mouse immunoglobulin G (IgG) alkaline phosphatase–conjugated antibody (Promega). Band intensities were quantified by Image J. To calculate proportion of total protein that bound to the liposomes with PIP or without PIP (control), the band intensities of the pellet fractions were divided by the total band intensity of the supernatant and pellet. The PIP-specific binding was calculated by subtracting the band intensity of PIP-independent binding (control) from band intensity of PIP-specific binding. Freshly prepared liposomes and purified proteins were used in the experiments.

Protein structure prediction

The homology model of Hua2NT (1–364) was generated by PHYRE2 (Protein Fold Recognition server, One-to-One threading software) (49) from the structure of a1-containg V-ATPase from human embryonic kidney cell line HEK293F (PDB 6WM2) (48). NCBI-BLASTp program was used to align the protein sequences of the two HuaNT isoforms and identify regions that are not conserved between isoforms and have positively charged basic amino acids with aromatic (and hydrophobic) amino acids. Structures were visualized on UCSF Chimera Version 1.16 (59).

Statistical analysis

All experiments were performed at least three times for each protein concentration and PIP lipid. ImageJ was used to quantify the bands in the immunoblots. Data is represented as mean ± SEM. Significance was determined using ordinary one-way ANOVA with multiple comparisons in GraphPad Prism 9. p values: ∗p < 0.05, ∗∗p < 0.005, ∗∗∗p < 0.0005, and ∗∗∗∗p < 0.00005.

Expression and localization of human a1NT and a2NT in 4T1 mouse breast cancer cells

The pCMV-GFP plasmid (60) was purchased from Addgene (Plasmid #11153). The WT and mutant NT domains of human a1 and a2 were PCR amplified and cloned between nucleotides 593 and 1719 of pCMV-GFP polylinker using the Takara InFusion cloning kit according to the manufacturer’s instructions. The sequences of resulting plasmids were confirmed by DNA sequencing. 4T1 mouse breast cancer cells were purchased from American Type Culture Collection.

In preparation for microscopy, round coverslips were washed with water and ethanol then coated with human collagen 4 at a concentration of 50 μg/ml for 30 min, followed by UV sterilization. The coverslips were placed in the wells of a 24-well plate and approximately 5000 4T1 cells were seeded per well and allowed to grow overnight, then transfected with the desired plasmid in Lipofectamine 3000 (Thermo Fisher Scientific) according to the manufacturer’s instructions. Transfected cells were grown for 2 days, then fixed with 4% formaldehyde in PBS for 10 min, and permeabilized with 0.1% Triton X-100 for 10 min. Fixed and permeabilized cells were incubated in blocking buffer (1% bovine serum albumin, 22.5 mg/ml glycine, 0.1% Tween in PBS) for 1 h.

The a1NT-GFP and a2NT-GFP constructs were both detected with mouse anti-GFP antibody (Proteintech 66002) diluted 1:200 in blocking buffer. Localization of a1NT-GFP was detected using mouse anti-GFP as a primary antibody and goat anti-mouse AlexaFluor 488 (Thermo Fisher Scientific A32723) at a 1:1000 dilution in blocking buffer as a secondary antibody. Rat monoclonal anti-LAMP2 antibody (ABL-93) was obtained from the University of Iowa Developmental Studies Hybridoma Bank (originally deposited by J.T. August) and diluted 1:100 in blocking buffer before use. Anti-LAMP2 binding was detected with goat anti-rat AlexFluor 594 (Thermo Fisher Scientific A-11007) diluted 1:1000 in blocking buffer. For visualization of a2NT, the anti-GFP antibody was labeled with the FlexAble CoraLite Plus 488 kit for mouse IgG1 (Proteintech). The anti-TGN46 primary antibody (Proteintech 66477) was labeled with FlexAble CoraLite Plus 647 for mouse IgG1 and used at a 1:500 dilution. Fixed and permeabilized cells were incubated with antibody for 1 h, and then washed three times, for 5 min each, in blocking buffer. Cells containing a1NT-GFP were then incubated in secondary antibody for 1 h. After the final incubation all samples were subjected to a final three washes of 5 min each. Coverslips were removed from the wells and mounted face down on glass microscope slides with ProLong Glass antifade mounting medium.

Labeled cells were visualized on a Nikon Ti2-E SoRa spinning disk confocal microscope equipped with 488 and 561 nm wavelength lasers. Cells were visualized with a 60x 1.4NA oil immersion objective with the microscope operating in 2.8x SoRa mode. Z-stacks of 25 to 40 slices were collected for each cell type. Laser intensities and exposure times were the same across samples for visualization of each primary antibody. Images were initially acquired using Nikon Elements (microscope.healthcare.nikon/products/software/nis-elements) software, then further analyzed using Fiji-ImageJ version 2.0.0 (Fiji.sc). Figures 8 and 9 represent single slices from the Z-stack, selected as having a well-defined nucleus. Line scans were performed and quantitated in Fiji and the data were plotted using Microsoft Excel (www.microsoft.com).

Data availability

Any data and materials will be shared upon request by contacting Patricia Kane (kanepm@upstate.edu). In addition, supporting data and micrographs will be available at https.upstate.figshare.com.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by George M. Carman