Cvm1 and its paralogue Cvm2 function as a complex at vacuolar membrane contact sites

Abstract

Membrane contact sites are regions where organelle membranes come together, and serve as platforms for metabolite exchange, process organization, and regulation of organelle dynamics. The yeast vacuole, equivalent to lysosomes in higher eukaryotes, functions as a degradative organelle, storage compartment, and signaling hub, establishing contacts with multiple organelles. We previously identified the protein Cvm1 as a component of vacuole contact sites with mitochondria, the nuclear endoplasmic reticulum (ER), and peroxisomes. Here, we investigate Cvm1-mediated contacts and show that the contacts with mitochondria require the porins Por1 and Por2. Additionally, Cvm1 forms a protein complex with its paralogue Yml020w, which we designate as Cvm2. Bioinformatic analysis predicts that both proteins contain an α/β-hydrolase fold. Notably, the predicted catalytic triad of Cvm2 is essential for its in vivo function, while Cvm1 lacks an active site. Complex formation is necessary for the function of the proteins, and Cvm1 targets the complex to the vacuole by binding phosphatidylinositol-3-phosphate on this membrane. Overexpression of this complex generates extended contacts between the vacuole and the peripheral ER. Collectively, our work describes the novel Cvm1–Cvm2 complex and molecular interactions important for its function as part of vacuolar contact sites.

• Membrane contact sites are appositions between organelle membranes, allowing material exchange, organization of metabolic processes, and regulation of organelle dynamics. Cvm1 is part of contacts of the yeast vacuole with mitochondria, peroxisomes, and the ER.

• The authors find that Cvm1 requires the yeast VDAC homologues to establish contacts with the mitochondria and that it binds PI(3)P on the vacuole. Furthermore, Cvm1 functions in complex with its paralogue Yml020w, now named Cvm2, which is likely a hydrolase.

• This work identifies the molecular interactions that establish vacuolar contacts and provides insights into their function and regulation.

INTRODUCTION

Eukaryotic cells contain membrane-enclosed organelles that allow the coexistence of biochemical environments specifically suited for different reactions. The necessity of communication among these compartments is partially fulfilled by membrane contact sites (MCS) between them. These structures are formed by protein complexes, called tethers, which interact with both membranes, establishing an apposition that does not result in fusion (reviewed in Eisenberg-Bord et al., 2016 and Prinz et al., 2020). Every organelle in the cell forms such contact sites (Valm et al., 2017; Kakimoto et al., 2018; Shai et al., 2018). At the molecular level, one of the main functions of these structures is the exchange of metabolites. These metabolites can be membrane lipids, a process catalyzed by lipid transfer proteins, or soluble luminal metabolites, which require the presence of channels at the contact site structure. The functions of the contacts can also be mediated at the molecular level by the physical attachment of the organelles, influencing their positioning in the cell, or by the generation of a subcompartment within organelle membranes with specific characteristics, for example, through the presence of enzymes (reviewed in Prinz et al, 2020).

Through these molecular functions, MCS play key roles in diverse cellular processes. Metabolite transport at contact sites enables metabolic pathways with steps in different compartments. For example, the transport of ceramide from the endoplasmic reticulum (ER) to the trans-Golgi network at a contact site between them, allows the production of sphingomyelin in this compartment (reviewed in Kumagai and Hanada, 2019). Lipid transport at contact sites, mediated by the Vps13 family of bulk lipid transporters, can also provide material for the biogenesis of rapidly growing membranes (reviewed in Leonzino et al., 2021). These structures constitute subcompartments within organelle membranes, and the conditional localization of proteins to them can serve as a regulatory mechanism. For instance, in yeast, under starvation conditions, the relocalization of the Hmg-CoA reductases to a contact site drives increased flux through the mevalonate pathway (Rogers et al., 2021). Contact sites are also key players in intracellular signaling, a paramount example being calcium signaling: the ER, as the main calcium storage organelle in metazoan cells, can control calcium entry through the plasma membrane, through a contact site established between the two membranes by the ER calcium sensor STIM1 and the plasma membrane calcium channel ORAI1 (Liou et al., 2005; Zhang et al., 2005; Luik et al., 2006; Prakriya et al., 2006; Vig et al., 2006; Zhang et al., 2006). Finally, the dynamics of organelles relies heavily on their contacts with other organelles. For example, mitochondrial division, a process important for the genomic and metabolic health of the organelle, happens primarily at sites where the mitochondrial network interacts with the ER (Friedman et al., 2011). In yeast, contact sites of the mitochondrial network with the cell cortex mediate both the retention of mitochondria in the mother cell and their inclusion in the daughter cell (Swayne et al., 2011; Pernice et al., 2016). These examples highlight that MCS are major determinants of the metabolic functions of organelles, the communication among them, their regulation by the metabolic state of the cell, and their life cycle, including biogenesis, dynamics, and inheritance.

The yeast vacuole is the equivalent organelle to lysosomes in higher eukaryotes, and it functions as the main degradative organelle of the cell. Vacuoles are the endpoint of the endocytic and autophagic pathways that deliver cargo for degradation and recycling (Cowles et al., 1997; Thumm, 2000; reviewed in Bowers and Stevens, 2005; reviewed in Huotari and Helenius, 2011). The vacuole is also a storage compartment for various metabolites, such as phosphate and amino acids, and can store metal ions for detoxification (reviewed in Li and Kane, 2009). Additionally, these organelles are crucial for cellular signaling as the place of activity of the target of rapamycin complex 1 (TORC1), a key regulator of cell growth and proliferation (Binda et al., 2009). Transporters in the membrane of the vacuole regulate the transport of metabolites and ions, and therefore play an important role in cellular homeostasis (Thumm, 2000; reviewed in Li and Kane, 2009). The importance of lysosomes in higher eukaryotes is demonstrated by the association of lysosomal dysfunction with many diseases such as lysosomal storage disorders, Alzheimer's and Parkinson diseases (reviewed in Colacurcio and Nixon, 2016).

Vacuoles form MCS with several organelles, including the ER, mitochondria, peroxisomes, lipid droplets, and the plasma membrane (Kakimoto et al., 2018; Shai et al., 2018). The contact sites of the vacuole with the nuclear ER, known as the nuclear–vacuolar junctions (NVJ) are prominent in yeast. One of its tethers is formed by the interaction of Nvj1, an integral membrane protein of the outer nuclear membrane, with the vacuolar protein Vac8 (Pan et al., 2000). NVJs are dynamic contact sites that respond to nutrient changes, by expanding upon starvation conditions. Under these conditions, the NVJs function as a platform to generate lipid droplets regulated by the tether Mdm1 (Hariri et al., 2018). This contact site serves as a hub for lipid metabolism, as the lipid transport proteins Osh1, Nvj2 and Lam6, as well as the enzymes Tsc13 and Hmg1, involved in lipid production, localize there under specific conditions (Levine and Munro, 2001; Kvam et al., 2005; Toulmay and Prinz, 2012; Elbaz-Alon et al., 2015; Rogers et al., 2021).

The MCSs of the vacuole with peroxisomes were identified via a split-Venus reporter in Saccharomyces cerevisiae (Shai et al., 2018). These MCSs remain uncharacterized, however, the peroxisomal protein Pex3 has been suggested to function as a tether of this contact site in Hansenula polymorpha (Wu et al., 2019).

We described previously the protein Cvm1 (contacts of the vacuolar membrane) as a novel component of three MCSs of the vacuole in S. cerevisiae: with the mitochondria, with the nuclear ER, and with peroxisomes. At the vacuolar-mitochondrial contact site, Cvm1 functions as a tether, independently of the previously described tether Vps39. The main localization of endogenous Cvm1 is at the NVJs (Bisinski et al., 2022). We showed that Cvm1 is related to sphingolipid homeostasis as changes in the expression level of the protein leads to altered sensitivity to myriocin, an inhibitor of the serine palmitoyltransferase, the enzyme that catalyzes the first and rate-limiting step of sphingolipid biosynthesis (reviewed in Hanada, 2003; Wadsworth et al., 2013; Bisinski et al., 2022). Additionally, the overexpression of Cvm1 leads to decreased content of complex sphingolipids. Interestingly, sphingolipid levels regulate Cvm1 localization, as decreased levels result in induction of Cvm1-mediated contact sites (Bisinski et al., 2022).

Given that Cvm1 is a component of three different contact sites of the vacuole, it seems to have a pivotal role in the integration of this organelle in the cell. Thus, we set out to describe Cvm1-mediated contact sites, by interrogating the molecular environment of the protein via pull-down experiments. We find that the mitochondrial porin Por1 copurifies with Cvm1, and Cvm1 requires Por1 or its paralogue Por2 to establish contact sites with the mitochondria. Additionally, Cvm1 interacts and is functionally related to its paralogue, the protein of unknown function Yml020w, which we name Cvm2. Bioinformatic modeling predicts that both proteins have an α/β-hydrolase fold, but while Cvm2 has a conserved active site, Cvm1 does not. The residues that form the predicted active site in Cvm2 are necessary for its in vivo function. Cvm2 localizes to the vacuolar membrane and accumulates in the NVJs through its interaction with Cvm1, which binds to phosphatidylinositol-3-phosphate (PI(3)P) at the vacuolar membrane. Cvm1 and Cvm2 form a complex, which is essential for their function in vivo. Finally, we show that overexpression of the Cvm1–Cvm2 complex induces extended vacuolar-peripheral ER contact sites.

RESULTS

Por1 copurifies with Cvm1, and Cvm1 requires porins at the mitochondria to establish vCLAMPs

We previously identified the protein Cvm1 as a novel component of different contact sites of the vacuole (Bisinski et al., 2022). To understand the molecular environment of Cvm1-mediated contact sites, we performed SILAC-based GFP-Trap pull-down experiments with N-terminally green fluorescent protein (GFP)-tagged Cvm1, which we previously showed to be functional (Bisinski et al., 2022; Figure 1A; Supplemental File S1).

FIGURE 1:

Cvm1 interacts with Por1 and requires porins at the mitochondria to establish vCLAMPs. (A) SILAC-based GFP-Trap pull-down of Cvm1 N-terminally tagged with GFP (GFP-Cvm1). Control cells were labeled with light isotopes and cells expressing GFP-Cvm1 were labeled with heavy isotopes. The log₁₀ of protein intensity is plotted against the log₂ of the normalized heavy/light SILAC ratio. Dots representing significantly enriched proteins (p < 0.001) are colored in pink and labeled, other identified proteins are represented by turquoise dots. (B and C) Analysis of vCLAMP formation by overexpressed GFP-Cvm1 in presence or absence of Por1 and Por2. The subcellular localization of GFP-Cvm1 overexpressed under the control of the NOP1 promoter was analyzed relative to the mitochondrial marker Cit1-2xmKate2 and the vacuole lumen stained with CMAC. The analysis was performed in a control strain or a strain lacking Por1 and expressing Por2 under the inducible GALL promoter, which is repressed in the presence of glucose. Cells were either incubated in media containing galactose or shifted to media containing glucose for 12 h. Panel B shows representative images, the scale bar represents 5 µm. Panel C shows the quantification of the percentage of cells with Cvm1-vCLAMPs. The bars represent the average of three independent experiments, shown as squares. Statistical comparison was performed using a one-way ANOVA and a Tukey post hoc test (*P > 0.05; **P <0.01; ***P <0.001).

Among the proteins most strongly coenriched with Cvm1 was the protein Por1, a porin located in the outer mitochondrial membrane. Por1 is one of two yeast homologues of the mammalian VDAC channels (Blachly-Dyson et al., 1997). We previously described that Cvm1 is a tether of the vacuole-mitochondria contact site, also known as vCLAMP (vacuolar and mitochondrial patch) (Bisinski et al., 2022). We thus wondered whether Cvm1 localization to this contact site depends on its interaction with Por1. To analyze this, we used a strain in which Cvm1 is N-terminally tagged with GFP and overexpressed under the control of the NOP1 promoter, conditions under which Cvm1 forms extended vCLAMPs, which can be visualized by fluorescence microscopy as patches of enrichment of Cvm1 along the vacuole membrane in the regions proximal to the mitochondria. To analyze the requirement of porins in the mitochondrial outer membrane, we used a strain in which Por1 is deleted and its paralogue Por2 is expressed under the control of the inducible GALL promoter, a shorter version of the GAL1 promoter that induces lower expression (Mumberg et al., 1994). This promoter is repressed in the presence of glucose, so under these conditions neither paralogue should be expressed, preventing that Por2 compensates for the lack of Por1 (Gligorovski et al., 2023). We used this strategy because cells lacking both paralogues are very sick (Kim et al., 1986).

For microscopy analysis, the cells were either incubated in media containing galactose or shifted to media containing glucose for 12 h. We then quantified the percentage of cells containing Cvm1 vCLAMPs, which were defined as accumulations of GFP-Cvm1 on the surface of the vacuole in the region that colocalized with the mitochondrial network (Cit1-2xmKate2; Figure 1B). In the control strain, in which both Por1 and Por2 are endogenously expressed, ∼80% (glucose) and ∼100% (galactose) of cells showed Cvm1 vCLAMPs. In the strain lacking Por1, when grown in galactose ∼70% of cells showed Cvm1 vCLAMPs, indicating that lack of Por1 alone affects vCLAMP formation. In addition, when the cells were shifted to glucose media for 12 h to inhibit the expression of Por2, the presence of vCLAMPs was further decreased to ∼30% of the cells (Figure 1, B and C).

These experiments indicate that the extended vCLAMPs promoted by the overexpression of Cvm1 require the presence of the VDAC homologues Por1 and Por2. Given the physical interaction indicated by the copurification, this suggests that the tethering of the vacuole to the mitochondria could be caused by the interaction of Cvm1 with Por1/Por2.

Cvm1 physically interacts with its functionally related paralogue Yml020w (Cvm2)

Another protein strongly coenriched with Cvm1 was the protein of unknown function Yml020w. (Figure 1A). Based on the data of the present article, showing a strong relation between Cvm1 and Yml020w, we have decided to name this protein Cvm2.

Cvm2 is the protein most similar to Cvm1 in the yeast genome, as evidenced by our JackHMMER searches (Johnson et al., 2010). An alignment of both proteins with orthologues from related species can be observed in Supplemental File S2. Analysis of the sequence of Cvm1 and Cvm2 using DISOPRED3 (Jones and Cozzetto, 2015) and HHPred (Söding et al., 2005) predicted a similar domain architecture for them, with a disordered N-terminal domain and homology to α/β-hydrolase fold proteins in the central region (Figure 2A). We searched for structurally similar proteins using VAST (Madej et al., 2020), and aligned the structure of the yeast α/β-hydrolase Bna7 (PDB accession number 1VKH, Arndt et al., 2005) with the AlphaFold-predicted structural models of Cvm1 and Cvm2. This analysis suggests the presence of a canonical aspartate–histidine–serine catalytic triad in Cvm2, but not in Cvm1 (Figure 2B). Based on this, Cvm2 would be an enzyme, while Cvm1 would be a protein with an enzyme fold lacking a catalytic site, that is, a pseudoenzyme. Two α/β-hydrolases identified as structurally similar by VAST are also included in the alignment in Supplemental File S2 (Arndt et al, 2005; Qian et al., 2009). Our phylogenetic analysis shows that the orthologues of Cvm2 are present in more fungi species than Cvm1 orthologues, the origin of which could be explained by gene duplication in a common ancestor of a subset of the Saccharomycetaceae family (Figure 2; Supplemental Figure S1A). For all proteins included in the different bioinformatic analysis, Supplemental Table S1 lists their accession numbers and the organisms that they belong to, including their taxonomy.

FIGURE 2:

Cvm1 interacts with its functionally related paralogue Yml020w (Cvm2). (A) Domain architecture of Cvm1 and Cvm2. Diagrams displaying the predicted N-terminal disordered domain and the predicted α/β-hydrolase fold in the proteins. (B) Structural alignment of the AlphaFold predicted structures of Cvm1 and Cvm2 with the active hydrolase Bna7, which was predicted as a similar structure by VAST. The depicted regions show the active site of Bna7 with an aspartate–histidine–serine catalytic triad. The AlphaFold predictions include a similar catalytic triad in Cvm2, but not in Cvm1. (C) SILAC-based GFP-Trap pull-down of Cvm2 N-terminally tagged with GFP expressed under the control of the strong and constitutive TEF1 promoter (↑GFP-Cvm2). Light isotope-labeled control cells and heavy isotope-labeled cells expressing ↑GFP-Cvm2 were used. The log₁₀ of protein intensity is plotted against the log₂ of the normalized heavy/light SILAC ratio. Dots representing significantly enriched proteins (P < 0.001) are colored in pink and labeled, other identified proteins are represented by turquoise dots. (D) Analysis of the myriocin sensitivity of cells lacking Cvm2. A wt strain, a strain lacking Cvm2, and two clones lacking Cvm2 carrying a plasmid with the CVM2 gene were serially diluted and spotted on YPAD media and YPAD media containing 1 µM myriocin. (E) Analysis of the myriocin sensitivity of cells overexpressing Cvm2. A wt strain and a strain overexpressing Cvm2 under the control of the strong constitutive TEF1 promoter were serially diluted and spotted on YPAD media and YPAD media containing 1 µM myriocin. (F) Analysis of the myriocin sensitivity of mutants of Cvm2 in the putative catalytic site. A wt strain, a strain in which a stop codon has been introduced after 25 amino acids in Cvm2 (Cvm2 H25Stop), or strains carrying mutations in two amino acids that would form the catalytic site of Cvm2 (Cvm2 D496A and Cvm2 S384A) were serially diluted and spotted on YPAD media and YPAD media containing 1.25 µM myriocin. (G) Analysis of the myriocin sensitivity of a mutant of Cvm1 in the only possibly catalytically active amino acid in the region where similar hydrolases have the active site. A wt strain, a strain in which a stop codon has been introduced after 20 amino acids in Cvm1 (Cvm1 Q25Stop), or a strain carrying a mutation in the only amino acid that could be possibly catalytically active in the region where similar hydrolases have an active site (Cvm1 H335A) were serially diluted and spotted on YPAD media and YPAD media containing 1.25 µM myriocin. (H) Analysis of the myriocin sensitivity of a strain overexpressing Cvm1 in combination of a mutation of Cvm2 in the putative catalytic site. A wt strain, a strain overexpressing Cvm1 from the strong constitutive TEF1 promoter (↑ Cvm1), a strain carrying a mutation in the putative catalytic site of Cvm2 (Cvm2 S384A), and a strain carrying both mutations combined (↑Cvm1 Cvm2 S384A) were serially diluted and spotted on YPAD media and YPAD media containing 1.7 µM myriocin.

To confirm the physical interaction of Cvm1 with Cvm2, we performed pull-down experiments with N-terminally GFP-tagged Cvm2. In this case, Cvm1 was also among the most strongly coenriched proteins, confirming the physical interaction (Figure 2C; Supplemental File S3). Interestingly, Por1 was also among the most highly coenriched proteins with Cvm2, further confirming that the proteins are part of the same interaction network.

Given the phylogenetic connection and physical interaction, we wondered whether Cvm2 would participate in the same cellular processes as Cvm1. Because we previously showed that changes in expression levels of Cvm1 lead to altered sensitivity to myriocin (Bisinski et al., 2022), we tested the myriocin sensitivity of Cvm2 deletion and overexpression strains. Indeed, cells lacking Cvm2 are hypersensitive to myriocin, phenocopying the deletion of CVM1 (Figure 2D; Bisinski et al., 2022). This phenotype can be suppressed by reintroducing the gene in a plasmid, indicating that it is specific for the lack of the protein and not a secondary effect of the genomic modification (Figure 2D). The overexpression of Cvm2 from the strong constitutive TEF1 promoter, on the other hand, leads to a hyperresistance of the cells to myriocin (Figure 2E).

To analyze whether Cvm2 is indeed a hydrolase and whether this activity is related to the myriocin-sensitivity phenotype, we compared the myriocin sensitivity of strains carrying mutations in two amino acids that would form the active site according to our structural predictions: aspartate 496 and serine 384. As a control lacking Cvm2 function, we used a strain in which the codon encoding for Histidine at position 25 of Cvm2 has been mutated to a stop codon by CRISPR/Cas9 (Cvm2 H25Stop), instead of a deletion of the whole open reading frame to minimize the genomic modifications. Figure 2F shows that both point mutations in the predicted active site of Cvm2 confer hypersensitivity to myriocin, comparable with the phenotype observed in the strain lacking Cvm2 because of the introduction of a premature stop codon. This indicates that these residues are essential for the function of the protein related to this phenotype. As previously noted, structural predictions suggest that Cvm1 lacks a catalytic site. Nonetheless, we mutated Histidine 335 in this protein, the only residue near the predicted active site region with potential catalytic activity. Figure 2G shows that the Cvm1-H335A mutant exhibits wild-type (wt) levels of myriocin sensitivity, indicating that this residue is dispensable for Cvm1 function in this context. Together, these results support the interpretation that Cvm2 functions as a hydrolase, while Cvm1 likely acts as a catalytically inactive pseudoenzyme. Moreover, they suggest that the enzymatic activity of Cvm2 underlies the observed phenotype.

We have shown before that overexpression of Cvm1 alone can cause hyperresistance to myriocin (Bisinski et al, 2022). If Cvm1 is a pseudoenzyme that binds to an active enzyme, we wondered whether Cvm1 would require Cvm2 to produce this phenotype. Indeed in cells lacking Cvm2, the overexpression of Cvm1 fails to produce resistance to myriocin (Figure 2; Supplemental Figure S1B). In particular, it requires the active site of Cvm2, as it also fails to produce resistance to myriocin in a background where the active site of Cvm2 is mutated (Figure 2H).

Taken together, these results indicate that Cvm1 physically interacts with and is functionally related to its paralogue Cvm2, both of which have a predicted α/β-hydrolase fold. Cvm2 likely contains an active site, while Cvm1 would be a pseudoenzyme. The active site of Cvm2 is related to the myrocin sensitivity phenotype of the gene, and required by Cvm1 to produce resistance to myriocin when overexpressed.

Cvm2 localizes to the vacuolar membrane and the NVJ dependent on Cvm1

To address the subcellular localization of Cvm2, we tagged it with the fluorescent protein mNeonGreen (Shaner et al., 2013). We used the myriocin hypersensitivity phenotype of cells that lack Cvm2 to address the functionality of C-terminal and N-terminal–tagged versions, expressed under the control of the endogenous promoter. Growth tests on solid media containing myriocin showed that C-terminal–tagged Cvm2 with mNeonGreen (Cvm2-mNG) is not functional. Complementation of the Cvm2 H25Stop strain with plasmids expressing N-terminal–tagged Cvm2 under the control of its endogenous promoter using two different linkers (3xGGSG and PAPAP) showed that they are both functional, with the rigid PAPAP linker resulting in full complementation (Figure 3; Supplemental Figure S1A). For this reason, Cvm2 tagged N-terminally with mNeonGreen with a rigid PAPAP linker is used for all subcellular localization experiments in this work and is noted from now on and in the figures as mNG-Cvm2.

FIGURE 3:

Cvm2 localizes at the vacuolar membrane and is enriched in the NVJ, dependent on Cvm1. (A) Fluorescence microscopy analysis of the subcellular localization of Cvm2 tagged N-terminally with mNeonGreen (mNG-Cvm2) relative to the general ER marker Sec63 C-terminally tagged with 2xmKate2 (Sec63-2xmK2) and the vacuolar lumen stained with CMAC. The scale bar represents 5 µm. (B) Fluorescence microscopy analysis of mNG-Cvm2 relative to the NVJ marker Nvj1 tagged C-terminally with 2xmKate2 (Nvj1-2xmK2) and the vacuolar lumen stained with CMAC. The line profile to the right of the merged image shows the normalized fluorescence intensity of mNG-Cvm2 and Nvj1-2xmK2 along the yellow line drawn over the vacuolar surface, as shown in the merged image. The scale bar represents 5 µm. (C) Quantification of the enrichment of mNG-Cvm2 in the NVJ relative to the rest of the vacuole membrane. The enrichment factor was calculated as the mean intensity value of mNG-Cvm2 along a line prolife within the NVJ and then divided by the mean intensity value in a line profile along the rest of the vacuole membrane for each cell. Each dot represents one cell, and the bigger dots represent the mean for each of three independent experiments (colored in green, blue, and red). For each experiment, at least 20 cells were analyzed. (D) Fluorescence microscopy analysis of mNG-Cvm2 in a control strain and in a strain lacking Cvm1. The vacuolar lumen is stained with CMAC. The line profiles show the normalized fluorescence intensity of mNG-Cvm2 and the CMAC signal along the yellow line in the merged images. The scale bar represents 5 µm. (E) Fluorescence microscopy analysis of N-terminally GFP tagged Cvm1 (GFP-Cvm1) in a control strain and in a strain lacking Cvm2. The vacuolar lumen is stained with CMAC. The line profiles show the normalized fluorescence intensity of GFP-Cvm1 and the CMAC signal along the yellow line in the merged images. The scale bar represents 5 µm.

Fluorescence microscopy analysis of mNG-Cvm2 shows that it localizes to the vacuole membrane, as evidenced by the correlation of the signal with the staining of the vacuole lumen with CMAC (Figure 3A). In addition to the homogeneous signal of mNG-Cvm2 along the vacuole membrane, dot-like accumulations are observed. These accumulations colocalize with the ER marker Sec63-2xmKate2 (Brodsky and Schekman, 1993), in the region that corresponds to the nuclear ER (Figure 3A). This region could represent the NVJ, a contact site of which Cvm1 is a component (Bisinski et al., 2022). Therefore, we confirmed that these accumulations represented the NVJ, by colocalization with its most classical marker, the protein Nvj1 (Pan et al., 2000). Indeed, the accumulations of mNG-Cvm2 at the vacuolar membrane colocalize with Nvj1-2xmKate2, as clearly observed in the line profile (Figure 3B). There is a cell-to-cell variability of the extent of Cvm2 enrichment in the NVJs, as we described previously for Cvm1. To characterize this, we quantified this enrichment compared with the rest of the vacuole membrane in the form of a ratio, and observed an average enrichment of 1.8, with cells ranging from no enrichment (ratio close to 1) to cells with very strong enrichments of up to 4 to 5 (Figure 3C).

Because we established that Cvm1 and Cvm2 physically interact with each other, and they share the localization to the vacuole membrane and the NVJs, we wondered whether the proteins depend on each other to achieve their subcellular localization. Indeed, this is the case, as the absence of Cvm1 leads to Cvm2 losing its localization to the vacuole membrane, which can be observed in the line profiles below the microscopy images (Figure 3D). Cvm1, however, does not require the presence of Cvm2 for its localization to the vacuole membrane (Figure 3E). The steady-state protein levels of Cvm1 and Cvm2 are not significantly altered in the absence of their respective binding partners (Figure 3; Supplemental Figure S1, B and C), indicating that the observed localization changes are not due to differences in protein abundance.

We conclude that Cvm2 localizes to the vacuolar membrane and is enriched at the NVJs. Cvm2 depends on the presence of Cvm1 to achieve its subcellular localization.

Cvm1 localizes to the vacuolar membrane by binding to PI(3)P via a positively charged patch

Because Cvm1 does not require Cvm2 for its localization to the vacuole, we wondered how it achieves its localization at this organelle. Analysis of the sequence of Cvm1 with either DeepTMHMM (Hallgren et al., 2022) or TOPCONS (Tsirigos et al., 2015) predicted no transmembrane domains in the protein. Indeed, biochemical analysis indicated that Cvm1 is peripherally associated with membranes, as it sediments with membranes but can be extracted by treatment with 100 mM sodium carbonate or 1 M sodium chloride, suggesting that the localization to membranes is mediated by electrostatic interactions (Figure 4A).

FIGURE 4:

Cvm1 localizes to the vacuolar membrane by binding PI(3)P via a positively charged patch. (A) Analysis of the membrane association of Cvm1. A subcellular fractionation was performed using cells expressing GFP-tagged Cvm1. The fraction that sedimented at 20,000 × g (P) was resuspended in a buffer containing either 100 mM Na₂CO₃, 1% TX-100 or 1M NaCl to analyze whether GFP-Cvm1 continued to sediment or was found in the supernatant after these treatments. For the immunoblot, a primary antibody detecting GFP was used. (B) Fluorescence microscopy analysis of N-terminally GFP tagged Cvm1 (GFP-Cvm1) in a control strain and in strains lacking either Vps34 or Fab1. The vacuolar lumen is stained with CMAC. The line profiles to the right of the microscopy images show the normalized fluorescence intensity of GFP-Cvm1 and the CMAC signal along the yellow line in the merged images. The scale bar represents 5 µm. (C) Liposome sedimentation assay with recombinant purified Cvm1 from K. marxianus (KmCvm1). The liposomes contained no charged lipids (neutral), or PS and different PIPs adjusted to the same overall charge. The graphs represent the mean value of three independent experiments with SEM. Significances were calculated using a one-way ANOVA with a Holm-Sidak test (*P < 0.05). (D) Surface model representation of the AlphaFold prediction of Cvm1 from S. cerevisiae based on the coulombic electrostatic potential (Varadi et al., 2022). Negatively charged areas are colored red and positively charged areas in blue. (E) Enlargement of the basic patch marked by the red square in panel D. The residues that form the surface are indicated. (F) Fluorescence microscopy analysis of control GFP-Cvm1 and the GFP-Cvm1 (K221E R764E) mutant. The vacuolar lumen is stained with CMAC. The line profiles to the right of the microscopy images show the normalized fluorescence intensity of GFP-Cvm1 or GFP-Cvm1 (K221E R764E) and the CMAC signal along the yellow line in the merged images. The scale bar represents 5 µm. (G) Analysis of the myriocin sensitivity of a strain carrying the Cvm1 (K221E R764E). Cultures of the indicated strains were serially diluted and spotted on YPD plates or YPD plates containing 1.7 µM myriocin. (H) GFP-Trap pull-down of GFP-Cvm1 or GFP-Cvm1 (K221E R764E). Label-free pull-downs of a strain expressing GFP-Cvm1, a strain expressing GFP-Cvm1 (K221E R764E), and a control strain expressing no GFP were performed in triplicates. The graph shows the log₂ of the protein intensity ratio between the GFP-Cvm1 strain and the control strain, plotted against the log₂ of the protein intensity ration between the GFP-Cvm1 (K221E R764E) and the control strain. Each dot represents one protein, Cvm1 and Cvm2 are marked in green.

In the pull-down experiments using Cvm1 as bait protein (Figure 1A), no candidate proteins were identified that could recruit Cvm1 to the vacuolar membrane. Because phosphoinositides are one of the major identity determinants of organelle membranes, used for many proteins and in particular many MCS to recognize them, we hypothesized that they might be required for Cvm1 to localize to the vacuole. The main phosphoinositides in the endocytic pathway are (PI(3)P), produced by the PI(3)P kinase Vps34 at endosomes and phosphatidylinositol-3,5-bisphosphate (PI(3,5)P2), produced from PI(3)P by the PI(3,5)P₂-kinase Fab1 at late endosomes and the vacuole (Herman and Emr, 1990; Stack et al., 1993; Yamamoto et al., 1995). In cells lacking Vps34, Cvm1 completely loses the localization to the vacuole, which is not the case in the absence of Fab1. This can be observed in the line profiles to the right of the microscopy images (Figure 4B), suggesting that Cvm1 binds to PI(3)P at the vacuolar membrane.

To confirm that Cvm1 indeed binds to PI(3)P-positive membranes directly and that the loss of subcellular localization is not an indirect effect, we produced recombinant protein in Escherichia coli. Expression and purification trials identified Cvm1 from Kluveromyces marxianus (KmCvm1) as the most suitable protein, and we were able to purify a truncated version of KmCvm1 lacking the N-terminal region predicted to be disordered to high purity (KmCvm1[64-714]) (Figure 4; Supplemental Figure S1, A and B). This purified protein bound efficiently to liposomes containing PI(3)P in liposome sedimentation assays, but not to liposomes containing other phosphoinositides or charged lipids (Figure 4C). This indicates that Cvm1 binds directly to PI(3)P-containing membranes.

To understand how Cvm1 binds to PI(3)P at the vacuolar membrane, we checked for predicted membrane-binding domains in the protein. However, algorithms in domain databases such as InterPro or Pfam (Blum et al., 2021; Mistry et al., 2021) did not identify any. Because we used a truncated version of Cvm1 missing the N-terminal disordered domain for the liposome sedimentation assays, the protein binds to PI(3)P via its folded region. Comparison of the structures predicted by AlphaFold (Jumper et al., 2021) for Cvm1 from S. cerevisiae and K. marxianus identified a surface patch containing several basic amino acids, which could thus mediate binding to PI(3)P. This surface is predicted to be formed by histidines 208 and 231, arginines 213 and 764, and lysines 221 and 730 (Figure 4, D and E). Interestingly, these amino acids are conserved in Cvm1 orthologues from other yeast species, but are not present in Cvm2 (Figure 4; Supplemental Figure S1C). Mutating lysine 221 and arginine 764 to glutamate by CRISPR/Cas9, thus inverting the charge, results in Cvm1 losing the localization to the vacuole (Figure 4F). This indicates that these residues are important for Cvm1 to achieve its vacuolar localization.

To further characterize this mutant, we tested its myriocin sensitivity and its ability to interact with Cvm2. The mutant exhibited hypersensitivity to myriocin, comparable with that observed in cells lacking Cvm1, suggesting that membrane binding is essential for Cvm1 function (Figure 4G). In addition, the mutant failed to interact with Cvm2, implying that membrane association may be a prerequisite for the Cvm1–Cvm2 interaction (Figure 4H; Supplemental File S4).

Collectively, our results indicate that Cvm1 binds to PI(3)P at the vacuolar membrane via a positively charged patch and targets Cvm2 to this organelle.

Cvm1 and Cvm2 form a protein complex necessary for their function

Based on the interaction of Cvm1 and Cvm2, their functional relationship and the dependence of Cvm2 on Cvm1 for subcellular localization, we wondered whether the two proteins form a complex. To test this, we built a strain in which both proteins are expressed under the control of the strong inducible GAL1 promoter and Cvm1 is N-terminally tagged with the 3xFLAG tag. 3xFLAG-Cvm1 is functional, as the cells containing this modification are not hypersensitive to myriocin as cells containing Cvm1 in which a stop codon has been introduced at position 20 of Cvm1 (Figure 5; Supplemental Figure S1B). Purification of Cvm1 from this strain resulted in a strong copurification of Cvm2 (Figure 5A), with the identity of the proteins confirmed by mass spectrometry, and both proteins run together in size exclusion chromatography (Figure 5; Supplemental Figure S1A). These experiments show that the two proteins form a complex.

FIGURE 5:

Cvm1 and Cvm2 form a protein complex necessary for their function. (A) Affinity purification of N-terminally 3xFLAG tagged Cvm1 from a strain expressing Cvm1 and Cvm2 under the control of the strong inducible GAL1 promoter. The identity of the purified proteins, confirmed by mass spectrometry, is indicated to the right of the SDS–PAGE. (B) AlphaFold multimer prediction of the structure of a Cvm1-Cvm2 dimer. The predicted complex includes three key interaction sites: site 1: Cvm1-D721 and Cvm2-R618, site 2: Cvm1-D804 and Cvm2-K602, and site 3: Cvm1-H720, -W717 and -Y713 with Cvm2-Y611 and -W615. (C and D) Affinity purifications of N-terminally 3xFLAG tagged Cvm1 from a strain expressing Cvm1 and Cvm2 wt or the indicated mutants under the control of the strong inducible GAL1 promoter. Purified proteins are indicated to the right of the SDS–PAGE. (E and F) Wt strains or strains expressing the indicated Cvm1 or Cvm2 mutants were serial diluted and spotted on YPAD media or YPAD media containing 1.25 µM myriocin. (G) Fluorescence microscopy analysis of mNG-Cvm2 in a control strain and in strain expressing the mutant Cvm1(D721R). The vacuolar membrane is stained with FM4-64. The line profiles to the right of the microscopy images show the normalized fluorescence intensity of mNG-Cvm2 and the FM4-64 signal along the yellow line in the merged images. The scale bar represents 5 µm.

We used AlphaFold multimer (Evans et al., 2022) to predict the structure of a putative Cvm1/Cvm2 dimer, which resulted in a high-confidence model (Figure 5; Supplemental Figure S1, C and D). This model indicates an extensive interface with three key interaction sites. Interaction site 1 involves a salt bridge between Cvm1-D721 and Cvm2-R663 or Cvm2-R618, site 2 is formed by a salt bridge between Cvm1-D804 and Cvm2-K602, and interaction site 3 is formed by hydrophobic interactions between Cvm1-H720, W717 and Y713 with Cvm2-Y611 and W615 (Figure 5B). We mutated these residues by either inverting the charges of polar amino acids (aspartates to arginines, arginines, and lysines to glutamate) or introducing charged residues (glutamate) instead of hydrophobic ones in either Cvm1 or Cvm2 and performed the purifications using these strains. Interestingly, mutations of the amino acids at interaction sites 1 and 3 either in Cvm1 or Cvm2 resulted in loss of the copurification of Cvm2, suggesting that these interfaces are necessary for complex formation. This was not the case when the putative interaction site 2 was disrupted (Figure 5, C and D).

Given that mutations Cvm1 D721R, Cvm1 Y713E W717E, Cvm2 R618E, and Cvm2 Y611E W615E disrupt complex formation, we used them to address whether complex formation is necessary for the functionality of these proteins. Cells expressing any of these variants showed hypersensitivity to myriocin, comparable with variants in which a stop codon was introduced after a few amino acids by CRISPR/Cas9 (Cvm1 Q20Stop or Cvm2 H25Stop). In contrast, strains containing the mutations that do not result in complex disassembly (Cvm1 D804R and Cvm2 K602E) are not hypersensitive to myriocin (Figure 5, E and F). This shows that complex formation is necessary for the in vivo role of Cvm1 and Cvm2, responsible for their effect on the cellular myriocin sensitivity.

We then wondered whether the deletion or the overexpression of both proteins of the complex would have an additional effect on cell growth. A double mutant lacking expression of both proteins (Cvm1 Q20Stop or Cvm2 H25Stop) shows a growth defect on media containing myriocin that is similar when either of the proteins are not expressed individually. This is the kind of genetic interaction expected for proteins that function as part of a complex, as lack of one of the subunits already results in lack of the complex (Figure 5; Supplemental Figure S1E). In contrast, cooverexpression of Cvm1 and Cvm2 leads to a stronger resistance to myriocin compared with when just one protein is overexpressed, even at high concentrations of the drug (Figure 5; Supplemental Figure S1F).

Because we described previously that the localization of Cvm2 to the vacuole depends on the presence of Cvm1, we wondered whether the disruption of complex formation has the same effect. Indeed, when the complex formation is disrupted by introducing the mutation Cvm1 D721R, Cvm2 loses the localization to the vacuole, as can be appreciated in the line profile to the right of the microscopy images (Figure 5G). Suggesting that the presence of Cvm1 is not sufficient for Cvm2 localization to the vacuolar membrane, but the complex formation is required for Cvm2 subcellular localization.

In our purifications, both Cvm1 and Cvm2 run as double bands in SDS–PAGE, which normally hints at posttranslational modifications of the proteins. In the case of Cvm2, this corresponds to phosphorylation, as the second band disappears upon incubation of the sample with the lambda protein phosphatase (Figure 5; Supplemental Figure S2A). This treatment did not change the double-band pattern for Cvm1. Given the substantial size shift in the band, we speculated that this was due to ubiquitination of the protein. To analyze this, we used a strain in which the UBI4 gene expressing ubiquitin is replaced by a 6xHIS tagged ubiquitin separated by an 8 amino acid linker in between and expressed under the control of the strong constitutive TEF1 promoter for purification experiments (HIS-ALINQERA-Ub; MacDonald et al, 2017). Additionally, either Cvm1 and Cvm2 or just Cvm1 were expressed under the control of the strong and inducible GAL1 promoter and Cvm1 was N-terminally tagged with the 3xFLAG tag. Immunostaining analysis of purified 3xFLAG-Cvm1 with anti-His and anti-Flag antibodies showed that Cvm1 is indeed modified with the HIS-ALINQERA-Ubiquitin, corresponding to the upper band (Figure 5; Supplemental Figure S2B). For confirmation, we purified HIS-tagged proteins from a HIS-ALINQERA-Ub strain, in which Cvm1 and Cvm2 were expressed under the control of the GAL1 promoter and Cvm1 was N-terminally tagged with the 3xFLAG tag, using Ni-NTA resin to enrich proteins modified with the HIS-ALINQERA-Ubiquitin protein. Immunostaining analysis with an anti-Flag antibody showed that Cvm1 was coenriched together with the HIS-tagged ubiquitin (Figure 5; Supplemental Figure S2C). We used mass spectrometry as an additional method to confirm the ubiquitination of Cvm1. Ubiquitin is attached to lysine residues via its C-terminus, which ends with the sequence arginine–glycine–glycine (Arg–Gly–Gly). When ubiquitinated proteins are treated with trypsin, which digests after lysine or arginine, the modification inhibits digestion after the modified lysine. Additionally, trypsin cuts after the arginine in the C-terminal sequence of ubiquitin, leaving a Gly–Gly dipeptide as a modification in the substrate protein, which can be detected by mass spectrometry (Fulzele and Bennet, 2018; Udeshi et al., 2020). Analysis of purified Cvm1 digested with trypsin identified several Gly–Gly modified peptides, confirming the ubiquitination of Cvm1 (Figure 5; Supplemental Figure S2D).

From these data, we conclude that Cvm1 and Cvm2 form a protein complex, which is necessary for their function and the localization of Cvm2 to the vacuolar membrane. In vivo, both proteins undergo posttranslational modifications, Cvm1 is ubiquitinated and Cvm2 is phosphorylated.

Overexpression of the Cvm1–Cvm2 complex results in extended vacuole-peripheral ER contact sites

We have previously shown that endogenously expressed Cvm1 is mainly localized at the vacuolar membrane with accumulations in the NVJs. However, when overexpressed, it shows an increased localization at the vCLAMPs, as these are considerably expanded (Bisinski et al., 2022). In this study, we found that the main localization of endogenously expressed Cvm2 is at the vacuolar membrane and with accumulations in the NVJs as for Cvm1. We therefore wondered whether the localization of Cvm1 to the vCLAMPS under overexpressed conditions is due to the fact that only one protein of the complex is overexpressed. We decided to assess the subcellular localization of the proteins when both subunits of the complex are overexpressed. Of note, this is the situation that produces the strongest resistance to myriocin. To do this, we introduced a TEF1 promoter upstream of both genes and fused either Cvm1 or Cvm2 to a fluorescent protein.

In the double overexpression strain, both GFP-Cvm1 and mNG-Cvm2 localized to the vacuole membrane and showed strong accumulations at the NVJ (Figure 6, A and B). However, mNG-Cvm2 shows additional signal along the ER, as seen by colocalization with the general ER marker Sec63 and in dots located away from the vacuole (Figure 6B).

FIGURE 6:

Overexpression of the Cvm1–Cvm2 complex results in extended vacuole-peripheral ER contact sites. (A and B) Fluorescence microscopy analysis of GFP-Cvm1 or mNG-Cvm2 in a strain expressing Cvm1 and Cvm2 under the control of the strong and constitutive TEF1 promoter. The proteins were imaged relative to the general ER marker Sec63, tagged either with 2xmKate2 (Sec63-2xmK2, A) or the HaloTag and stained with JFX646 (Sec62-HT, B) and the vacuolar lumen stained with CMAC. The scale bar represents 2 µm. (C and D) Representative fluorescence microscopy images depicting the small vacuoles (CMAC) surrounded by the peripheral ER (Sec63-2xmK2, C or Sec63-HT, D) observed in strains overexpressing Cvm1 and Cvm2. Cvm1 is tagged with GFP and Cvm2 with mNeonGreen. The scale bar represents 2 µm. (E–G) SBF-SEM analysis of control cells or cells expressing Cvm1 and Cvm2 under the control of the strong constitutive TEF1 promoter. Three independent replicates were performed and cells were quantified for the presence of peripheral ER-vacuole contact sites. Panel E shows representative examples of what was considered NVJs, peripheral ER-vacuole contact sites or no contact sites (V, vacuole; nER, nuclear ER; pER, peripheral ER). Panel F shows the result of this quantification, significances were calculated using a one-way ANOVA and Tukey post hoc test (***P < 0.005). Panel G shows a tridimensional reconstruction of one peripheral ER-vacuole contact site based on segmentation of the SBF-SEM data. The nuclear ER is colored in purple, the peripheral ER in magenta, and the vacuole in green. (H and I) Analysis of the signal intensity of GFP-Cvm1 or mNG-Cvm2 on peroxisomes in a strain where they are both overexpressed, and the relationship with the proximity of peroxisomes to the vacuole surface. Panel H shows the quantification of the fluorescence intensity of GFP-Cvm1 and mNG-Cvm2 in peroxisomes relative to the cellular mean, and separated according to the proximity of the peroxisomes to the vacuole. The distribution of intensities between peroxisomes close or away from the vacuole was compared using a Kolmogorov–Smirnoff test (**P < 0.01; n.s. P > 0.05). Panel I shows representative microscopy images of peroxisomes and their GFP-Cvm1 or mNG-Cvm2 signal. The ratio between this signal and the cell mean is indicated, as well as the distance to the vacuole. One image contains two peroxisomal structures marked as left (L) and right (R). Peroxisomes are marked with Pex14-HaloTag labeled with JFX656 (Pex14-HT) and the vacuolar lumen is stained with CMAC. The scale bar represents 2 µm. (J) Fluorescence microscopy analysis of mNG-Cvm2 colocalization with the Golgi marker Mnn9 tagged with the HaloTag and labeled with JFX650 (Mnn9-HT) in a strain expressing Cvm1 and Cvm2 under the control of the strong constitutive TEF1 promoter. The vacuole lumen is stained with CMAC. To the right, zoomed in cropped microscopy images are shown with the Fire look-up table for the channels of Mnn9-HT and mNG-Cvm2, white arrows are added as reference. The scale bars represent 2 µm.

A striking phenotype observed for both proteins in the double overexpression strain was that they were strongly enriched in small vacuoles, which appear to be surrounded by ER-tubules (Figure 6, A, C, and D). This suggested that the overexpression of the Cvm1–Cvm2 complex resulted in the generation of extended peripheral ER-vacuole contact sites. To confirm that these were indeed contact sites and quantify their abundance, we used serial block-face scanning electron microscopy (SBF-SEM) to quantify the existence of peripheral ER-vacuole contact sites. As suggested by the fluorescence microscopy images, contact sites between the peripheral ER and the vacuole were very abundant in the double overexpression strain, present in ∼75% of cells, and rare in the control strain (Figure 6F). Examples of what was considered peripheral ER-vacuole contact sites, NVJs or not contact sites can be seen in Figure 6E. These peripheral-ER vacuole contact sites of the double overexpression strain mostly corresponded to small vacuoles covered in a good portion of their surface by the ER, as was previously suggested by the fluorescence microscopy experiments. A reconstruction of one such contact can be observed in Figure 6G, with the nuclear envelope reconstructed as a purple surface, the peripheral ER as a magenta surface and the small vacuole in green. To better illustrate the three-dimensional (3D) architecture of the structure, Supplemental Video S1 shows an animation of the SBF-SEM slices of this cell and the generated model and Figure 6 and Supplemental Figure S1A show different slices of the structure.

Movie S1.

Download video file ^{(13.6MB, mp4)}

Animation of the different slices of the SBF-SEM experiment and the 3D model generated of the vacuole-peripheral ER membrane contact site.

Another interesting observation is that Cvm2 localizes in dots that are away from the vacuole. To characterize the cytosolic dots formed by mNG-Cvm2 in the strain overexpressing Cvm1 and Cvm2, we colocalized them with different organelle markers. We observed that mNG-Cvm2 had local maxima in peroxisomes marked by the peroxisomal marker Pex14, tagged with a Halo Tag (Pex14-HT) (Albertini et al., 1997). mNG-Cvm2 was enriched above the mean cellular level in these structures, independent of whether they were close or away from the vacuole, as can be appreciated in the quantification in Figure 6H and the examples in Figure 6I. GFP-Cvm1, on the contrary, only showed enrichment in peroxisomes that were found on the surface of the vacuole (Figure 6, H and I), as previously reported (Bisinski et al., 2022).

In addition, Golgi complex structures marked by the Golgi marker Mnn9 (Jungmann and Munro, 1998) also represented local maxima for the mNG-Cvm2 signal in the strain overexpressing Cvm1 and Cvm2 (Figure 6J), independently of their distance to the vacuole surface (Figure 6; Supplemental Figure S2B). This was not the case for GFP-Cvm1 (Figure 6; Supplemental Figure S2A). Imaging of Cvm1 and Cvm2 at endogenous levels in comparison with Mnn9 or Pex14 showed no enrichment in the structures (Figure 6; Supplemental Figure S2, C–E).

We conclude that overexpression of Cvm1 and Cvm2 complex leads to extended contact sites between the vacuole and the peripheral ER, which are enriched in the complex, and to Cvm2 localizing to peroxisomes and Golgi structures distant from the vacuole, which are negative for Cvm1.

DISCUSSION

Accumulating evidence has established MCS as crucial determinants of organelle function. We have recently identified Cvm1 as a component of contact sites of the vacuole with three different organelles. Furthermore, the protein affects cellular lipid homeostasis and its localization to contact sites is subject to regulation. These facts hinted at Cvm1 as an important player in vacuolar function, and therefore we set out to gain insight into Cvm1-mediated contact sites by identifying additional components. Our approach to this question was through the identification of proteins that copurify with Cvm1.

An important finding from our copurification experiments was that Cvm1 interacts with its paralogue, Cvm2, a previously uncharacterized protein. Furthermore, we show that the two proteins form a complex and similarly affect cellular physiology (see model in Figure 7). Inspection of the sequence of these paralogues shows homology to proteins containing an α/β-hydrolase fold, a widespread structural fold of hydrolytic enzymes, which can have a wide variety of substrates (Ollis et al., 1992; reviewed in Silman and Sussman, 2008; Bauer et al., 2020). We compared the predicted AlphaFold structures of Cvm1 and Cvm2 with active α/β-hydrolases predicted to have a similar structure, and saw that Cvm2 contains a potential catalytic triad, while Cvm1 does not. The amino acids that form this triad are required for the in vivo function of Cvm2. This suggests that the Cvm1–Cvm2 complex would consist of an active enzyme coupled to a pseudoenzyme. In many cases, pseudoenzymes have functions related to those of their active counterparts, for instance, some pseudokinases act as allosteric activators or inhibitors of active kinases (reviewed in Goldberg and Sreelatha, 2023). Cvm1 and Cvm2 acting within a complex suggests a similar situation, which is supported by the genetic interactions that we observed: the overexpression phenotype of Cvm1 in myriocin sensitivity requires the presence of the catalytic site of Cvm2. What the substrate of Cvm2 might be, remains to be investigated in further studies. The altered sensitivity to myriocin of cells lacking or overexpressing Cvm1 and Cvm2 suggests a role of these proteins in sphingolipid metabolism. It would thus be interesting to investigate whether the effect of the complex on these lipids is direct or indirect, especially since the main localization of the complex is in the NVJ, a contact site known as a hub of lipid metabolism and transport (Levine and Munro, 2001; Kvam et al., 2005; Toulmay and Prinz, 2012; Elbaz-Alon et al., 2015).

FIGURE 7:

Cvm1 and Cvm2 form a complex at different vacuolar MCS. Diagram summarizing the findings of the present study. Cvm1 forms a complex with its paralogue Cvm2. Cvm2 is recruited to the vacuole membrane by Cvm1, which binds PI(3)P-positive membranes. The complex likely bind Por1/2 at the mitochondrial outer membrane to establish vCLAMPs. Overexpression of Cvm1 causes a massive expansion of the vCLAMPs, while cooverexpression of Cvm1 and Cvm2 enlarges the vacuole-peripheral ER contact site. Structural predictions suggest that Cvm2 is a hydrolase and that Cvm1 is a pseudoenzyme.

Phylogenetically, Cvm2 orthologues are more widespread, while a group of species from the Saccharomycetaceae family contains additionally a copy of Cvm1, possibly due to a gene duplication event. As such, if the original function of the shared common ancestor between Cvm1 and Cvm2 was that of a hydrolytic enzyme acting at contact sites, by working as part of a heterotypic complex, Cvm1 and Cvm2 could have specialized in different molecular roles required for this function. Cvm2 would act as the active enzyme and Cvm1 as the targeting factor required for contact site localization and tethering of membranes. This hypothesis is consistent with our findings that Cvm2 localizes to the vacuole membrane based on its interaction with Cvm1, which recognizes PI(3)P-positive membranes. This is also in line with our previous work, which showed that overexpression of Cvm1 alone massively expands the contact site between vacuole and mitochondria, indicating that Cvm2 was not required to establish it. Interestingly, when Cvm2 is also overexpressed, we observe the expansion of the vacuole-ER contact site, indicating that the presence of Cvm2 affects the partition of Cvm1 among different contact sites (see model in Figure 7).

A second result of our copurification approach was identifying the porin Por1 as a binding partner of Cvm1. In addition, we show that Cvm1-mediated vCLAMPs decrease significantly when Por1 is absent and decrease even further when its homologue Por2 is additionally not expressed. In our copurification experiments we did not identify Por2. The reason for this could be the abundance of the porin proteins. Por1 is one of the most abundant proteins in yeast and comparison of the absolute protein abundance of Por1 and Por2 reported in mass spectrometry experiments shows that Por1 is ∼200 times more abundant than Por2 (Ho et al., 2018). Because the depletion of the two yeast porins results in a drastic decrease of Cvm1-mediated vacuole-mitochondria contact sites, we hypothesize that these contact sites could be established by Cvm1 binding to Por1/Por2 on the mitochondrial membrane (Figure 7). Definitive clarification of how the tethering takes place would require additional experiments, like the detection of crosslinks between the two proteins. In this work, we have established that Cvm1 binds to the vacuole membrane via its interaction with PI(3)P. The interaction partners of Cvm1, or the Cvm1–Cvm2 complex at the ER and the peroxisomes to be part of the NVJs and the peroxisome-vacuole contact site, remain unknown (Figure 7).

If the VDAC channels are indeed the binding partner that allows Cvm1 to tether vacuoles to mitochondria, the high abundance of Por1 would explain the fact that overexpression of Cvm1 could cause massive contact sites with the mitochondria. The formed contacts were so extensive that they caused the mitochondria to wrap around the vacuole, indicating that the binding partner was not a limiting factor (Bisinski et al, 2022). Interestingly, the VDAC channel, the orthologue of Por1 and Por2 in mammalian cells, has been previously reported to be part of MCSs (Szabadkai et al., 2006). This is now another example of abundant proteins on the surface of organelles with unrelated functions, which moonlight as contact site tethers. Another prominent case of this is the translocase of the outer membrane complex at the mitochondria, which is involved in many contact sites of this organelle, such as with the ER and with the vacuole (Lahiri et al., 2014; Murley et al., 2015; González Montoro et al., 2018). It is interesting to hypothesize the evolutionary advantage of using these already abundant complexes as a way to interact with the organelles, or that they may regulate cross-talk between contact site function and other processes.

An interesting point is that we previously observed that the localization of Cvm1 to contact sites was subject to regulation. We now additionally uncover that while we can isolate Cvm1–Cvm2 as a biochemically stable complex, when we overexpress both proteins, we also observe pools of localization of Cvm2 where Cvm1 is not located, in peroxisomes and Golgi structures distant from the vacuole. One possibility is that these subcellular localizations are an artifact of overexpression, since we could not detect them at endogenous levels. Overexpression could saturate binding partners, causing the excess protein molecules to expose normally hidden regions, potentially enabling interactions that do not occur under physiological expression levels. Another possibility is that the complex is dynamic, with its subunits having roles for which they function independently and roles for which they function as a complex. There are other examples of this, like the HOPS complex, which is a vesicular tethering complex important for many vesicular transport steps, while its subunit Vps39 has an independent role at MCS (Balderhaar and Ungermann, 2013; Hönscher et al., 2014; González Montoro et al., 2018; González Montoro et al., 2021). Thus there are at least two aspects of the Cvm1–Cvm2 complex that appear to undergo regulation in vivo: complex assembly and recruitment to contact sites. We identified that both proteins undergo posttranslational modifications, which could be possible molecular mechanisms of regulation. There are precedents of tethering at contact sites being regulated by phosphorylation, like the phosphorylation of the VAP proteins that regulates the binding to their different interaction partners at different MCS (Kumagai et al., 2014; Di Mattia et al., 2020; Kors et al., 2022). Ubiquitination regulates the mammalian ER-mitochondria tether Mfn2, with different outcomes depending on which domain is modified (Sugiura et al., 2013; Basso et al., 2018; McLelland et al., 2018). The fact that PI(3)P is the binding partner of Cvm1 on the vacuole membrane provides another possible regulatory mechanism for the complex, as the availability of phosphoinositides is variable and responds to many stimuli (reviewed in Posor et al., 2022). In particular, it is a known mechanism for regulation of contact site tethering, as tethers can be recruited by PIPs at membranes, such as the mammalian ER resident extended synaptotagmins, which bind PI(4,5)P₂ at the plasma membrane and facilitate tethering of the ER and the plasma membrane (Giordano et al., 2013).

Taken together, the novel molecular interactions that we identify as part of vacuolar contact sites provide insights into how these contact sites are established, their possible molecular function, and how they could be regulated.

MATERIALS AND METHODS

Request a protocol through Bio-protocol

Plasmids, strains, and molecular biology

S. cerevisiae strains were based on either BY4741 or BY4742. Genetic manipulation was performed by homologous recombination of PCR-amplified cassettes as it was described in (Janke et al., 2004) or by CRISPR Cas9 as described in Generoso et al. (2016). The yeast strains used in this study are listed in Supplemental Table S2.

The plasmid complementing CVM2 was created by PCR amplification of a cassette containing the CVM2 open reading frame plus 335 base pairs upstream and 297 base pairs downstream. The following primers were used: oAL158, 5′-AAAGAGCTCGTATGTGAACGGCTTGTGTAC-3′, and oAL159 5′-AAAACTAGTCAGTTCGCGGCTCTCATCTAC-3′. The PCR product was digested with the restriction enzymes SpeI and SacI and ligated into a pRS315-based plasmid containing the KanMX resistance cassette (Sikorski and Hieter, 1989).

To generate the Cvm1(D721R) mutant in the genome by CRISPR Cas9, 500 ng of the plasmid pAGM 146 and the oligonucleotide oAGM597 were transformed in the strain BY4741 Cvm2 H25Stop pRS403 Cvm2pr-mNeonGreen-PAPAP-Cvm2. The plasmid pAGM 146 was generated by PCR amplification of the plasmid pCU5003 using the primers oAGM 595 and oAGM 596. The selected S. cerevisiae transformant was confirmed by sequencing.

To generate the GFP-Cvm1(K221E R764E) mutant by CRISPR Cas9, first K221 was mutated to. For this, 500 ng of the plasmid pAGM 229 and the oligonucleotide oAGM 926 were transformed in the strain GFP-Cvm1. Plasmid pAGM 229 was generated by PCR amplification of the plasmid pCU5003 using the primers oAGM 924 and oAGM 925. Upon confirmation of a selected S. cerevisiae transformant by sequencing 500 ng of the plasmid pAGM 237 and the oligonucleotide oAGM 1010 were transformed in the strain GFP-Cvm1(K221E) for mutation of R764 to glutamate. Plasmid pAGM 237 was generated by PCR amplification of the plasmid pCU5003 using the primers oAGM 1008 and oAGM 1009. The selected S. cerevisiae transformant was confirmed by sequencing it.

The mutations Cvm2(H25Stop) and Cvm1(Q20Stop) were generated by transformation of 500 ng of the plasmids pAGM 109 (Cvm1 Q20Stop) or pAGM 110 (Cvm2 H25Stop) and the oligonucleotides oAGM 369 (Cvm1 Q20Stop) or oAGM 372 (Cvm2 H25Stop) in the strain BY4741. The plasmid pAGM 109 was generated by PRC amplification of the plasmid pCU5003 using the primers oAGM 367 and oAGM 368. The plasmid pAGM 110 was generated the same way using the primers oAGM 370 and oAGM 371. The selective S. cerevisiae transformants were confirmed by sequencing.

The plasmids and oligonucleotides used in this study are listed in Supplemental Tables S3.

Sequence alignment and phylogenetic tree inference

Cvm1 and Cvm2 protein sequences from S. cerevisiae were used as inputs to JackHMMER (Johnson et al., 2010), which was used to iteratively search the Uniprot database for Cvm1 and Cvm2 orthologues. From the two resulting alignments, we selected a total of 18 ascomycota organisms, which displayed a total of 29 Cmv1 and Cvm2 orthologue protein sequences. Muscle (Edgar, 2004) was used to perform a profile–profile alignment of the Cvm1 and Cvm2 orthologue alignments to produce a single alignment file. Tree inference was performed using IQ-TREE 1.6.12 (Nguyen et al., 2015). The best fit model using the BIC criterion according to Modelfinder was the LG+F+G4 substitution model. To assess branch support, IQ-TREE was used to perform ultrafast bootstrap approximation (UFboot) with 1000 replicates (Minh et al., 2013) and SH-like approximate likelihood ratio test (SH-aLRT) also with 1000 bootstrap replicates. Figures were generated using Figtree (http://tree.bio.ed.ac.uk/software/figtree/), which was also used to midpoint-root the tree.

For the alignment shown in Supplemental File S2, we kept the sequences belonging to 18 organisms from the Basidiomycota, Mucoromycota and Ascomycota phylums, Candida glabrata, Kluyveromyces lactis, Lachancea thermotolerans, Tetrapisispora blattae, Tetrapisispora phaffii, Vanderwaltozyma polyspora, S. cerevisiae, Zygosaccharomyces mellis, Candida albicans, Debaryomyces hansenii, Yarrowia lipolytica, Schizosaccharomyces japonicus, Aspergillus fumigatus, Neurospora crassa, Galerina marginata, Ustilago hordei, Glomus cerebriforme, and Rhizopus microspores. We used Clustal Omega and MUSCLE to align these sequences with the sequences corresponding to the PDB structures for 3ICW (Candida antarctica Lipase B) and 1VKH (S. cerevisiae serine hydrolase), which were found as structurally similar to Cvm1 and Cvm2, respectively, by VAST (Madej et al., 2020) (see below). The final alignment was manually curated using Aliview (Larsson, 2014).

Structural alignment

We retrieved the predicted structures for Cvm1 and Cvm2 from the AlphaFold database (Varadi et al., 2022) as a query to search for similar protein structures present in the NCBI's molecular modeling database using VAST-Search (Madej et al., 2020). We obtained multiple hits to alpha–beta hydrolases such as lipases and serine hydrolases. Among these, we chose the PDB structure of 3ICW (Candida antarctica Lipase B) to perform a structural alignment to the AlphaFold predictions of Cvm1 and Cvm2 protein structures. UCSF Chimera (Pettersen et al., 2004) was used to produce structural alignment as well as the molecular graphics.

Fluorescence microscopy, image quantification, and statistical analysis

Yeast cells were grown to logarithmic phase in rich medium (YPAD: yeast extract, peptone, adenine, and dextrose) or in synthetic complete medium (SDC), containing essential amino acids. To stain the acidic lumen of the vacuole, yeast cells were incubated with 20 µM 7-Amino-4-Chlormethylcumarin (CMAC, Invitrogen) for 15 min, followed by one wash with SDC medium. Proteins containing a HaloTag were labeled with the Janelia Fluor 646 ligand (Lavis Laboratory). The yeast cells (0.5 OD units) were incubated with 2.5 µM of Janelia Fluor 646 for 30 min, followed by 10 washing steps with filtered SDC medium (Day et al., 2018; Los et al., 2008).

Cells were imaged live in SDC medium at 23°C on an Olympus IX-71 inverted microscope equipped with 100 × NA 1.49 and 60 × NA 1.40 objectives; a sCMOS camera (PCO); an InsightSSI illumination system; 4′,6-diamidino-2-phenylindole, GFP, mCherry, and Cy5 filters; and SoftWoRx software (Applied Precision). We used z-stacks with 200-, 250-, or 350-nm spacing for constrained-iterative deconvolution (SoftWoRx). Microscopy image processing and quantification was performed using ImageJ (National Institutes of Health, Bethesda, MD).

Quantification of the mNeonGreen-PAPAP-Cvm2 enrichment in the NVJ was performed by calculation of the enrichment factor. For this, a line profile was drawn along the region positive for Nvj1-mKate2 and a second one along the vacuolar membrane. The enrichment factor was calculated by dividing the mean intensity value of mNeonGreen-PAPAP-Cvm2 of the first-line profile by the mean intensity value of the second. The experiment was performed in independent biological triplicates, and ≥20 cells were analyzed for each experiment. The single-cell data, and the means for each experiment are shown in a SuperPlot, according to Lord et al. (2020).

For the quantification of the fluorescence intensity of GFP-Cvm1 or mNG-Cvm2 in peroxisomes or Golgi compartments, classifying them according to their distance to the vacuole (Figure 5, H and I; Supplemental Figure S1B), the following procedure was used. The fluorescence intensity value in the cell was measured for the GFP/mNG channel in one equatorial z-plane using a manually defined regions of interest (ROI). The peroxisome and vacuole channels were thresholded in ImageJ using the MaxEntropy, the Golgi channel was thresholded using the Otsu algorithm. The thresholded images were used to generate 3D ROIs with the 3D ROI Manager plugin from the ImageJ 3D Suite (Ollion et al., 2013). For each peroxisomal/Golgi ROI, the distance to the vacuole and the mean fluorescence intensity were measured using the 3D ROI Manager. The intensity for each compartment in the GFP/mNG channel was normalized by the mean fluorescence in the cell, and the compartments were classified as close (≤0.1 µm) or distant (>0.1 µm) to the vacuole. Three independent experiments were performed, represented by different colors in the graph, and for each experiment at least 25 peroxisomal/Golgi structures were analyzed. The distribution of intensities between compartments close and distant to the vacuole was compared using a two-sample Kolmogorov–Smirnov test.

SBF-SEM

For SBF-SEM, yeast cells were fixed in 2.5% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.4. The subsequent processing followed a modified version of the NCMIR rOTO-postfixation protocol (Deerinck et al., 2010) ensuring pronounced contrast and electron dose resistance for consecutive imaging in hard Epon resin.

Following fixation, samples underwent a centrifugation step for 5 min at 800 × g. They were then postfixed in 1% potassium permanganate, followed by treatment with 2% osmium tetroxide (Science Services) in 0.1 M cacodylate buffer, pH 7.4 for 1 h each. afterward samples were incubated in 2.5% potassium ferrocyanide (Riedel de Haën) in cacodylate buffer for an additional 30 min. All steps were performed on ice with gentle agitation, excluding a washing step between osmication and ferrocyanide treatment (Hua et al., 2015). Subsequently, cells were treated with 1% (wt/vol) thiocarbohydrazide (Riedel de Haën) in dH20 for 20 min, followed by an additional 2% osmication step in water for 30 min at room temperature. Further treatment involved incubation in 1% tannic acid (Riedel de Haën) in dH2O. Following dH20 washes, samples were immersed in 1% aqueous uranyl acetate overnight at 4°C, brought up to 50°C, and washed with prewarmed water. The final step involved incubation in freshly prepared Walton's lead aspartate (Pb (NO3)2 (Carl-Roth), L-Aspartate (Serva), KOH (Merck)) for 30 min at 60°C. Following this, cells were dehydrated through a graded ethanol series (30, 50, 70, and 90%) on ice for 7 min each, rinsed in anhydrous ethanol twice for 7 min, and twice in anhydrous acetone (Carl-Roth) for 10 min at room temperature. Subsequently, cells were infiltrated in an ascending Epon:acetone mixture (1:3, 1:1, and 3:1) for 2 h each, followed by an additional incubation in a hard mixture of 100% Epon 812 (Sigma) overnight. Final curing took place in hard Epon mixture in Eppendorf tubes at 60°C for 48 h.

Once polymerized, the bottom of the tubes was trimmed off from the resin block, including the pelleted cells. Samples were then trimmed into small blocks suitable for SBF-SEM, glued to aluminum rivets using two-component conductive silver epoxy adhesive, and additionally coated with a 20-nm-thick gold layer. The rivet containing the mounted resin block was inserted into the 3View2XP Gatan stage (Gatan, Pleasanton, CA), fitted in a Jeol JSM 7200F (JEOL, Tokyo, Japan), and aligned parallel to the diamond knife-edge. The cells proved to be stable under imaging conditions of 3.1 kV accelerating voltage, high vacuum mode of 10 Pa, utilizing a 30 nm condenser aperture, and a positive stage bias of 600 V. Imaging parameters were set to 3 nm pixel size, 1.8 µs dwell time, with an ablation interval of 30 nm, and an image size of 10240 × 10240 pixels. In total, an approximate volume of 30 × 30 × 7.5 µm (comprising 250 slices) was acquired. Image acquisition was controlled via Gatan Digital Micrograph software (Version 3.32.2403.0). Subsequent postprocessing, including alignment, filtering, and segmentations, was performed manually in Microscopy Image Browser (Version 2.7) (Belevich et al., 2016). To maintain computational resources within manageable limits, the final stacks were binned before exporting the files for volumetric visualization in Amira 3D (Version 2021.1, Thermo Fisher Scientific).

Growth tests

For growth assays in solid medium, the cells were grown overnight in liquid YPAD medium, diluted the next morning and grown to exponential phase. Cells were diluted to an OD600 of 0.25 followed by serial 1:10 dilutions, which were then spotted onto YPAD agar plates with or without myriocin from Mycelia sterilia (Sigma-Aldrich) with the indicated concentrations. Plates were monitored for 1 to 4 d and imaged with the Bio-Rad ChemiDoc MP imaging System. The growth assay were performed in triplicates with one representative picture shown for each condition.

Whole-cell lysates, mNG pull-downs, SDS–PAGE, and Western blot

To test the expression levels of GFP-Cvm1 in the presence and absence of Cvm1, yeast cells were grown to logarithmic phase in liquid YPAD medium. A total of 30 OD₆₀₀ units were harvested and resuspended in 300 µl of buffer [PBS, 1% Triton X-100, 1 mM phenylmethylsulfonylfluoride (PMSF), 0.1 mg/ml leupeptin, 0.5 mg/ml pepstatin A, and 0.1 mM Pefabloc]. A total of 200 µl of acid-washed glass beads were added and the cells were mechanically lysed twice for 4 min with a Disruptor Genie device (Scientific Industries), with a 5-min break on ice in between. The supernatant was transferred to a new tube and centrifuged or 20 min at 20,000 × g at 4°C. A 75 µl sample was taken for SDS analysis, sample buffer was added and heated up at 80°C for 10 min.

To test the expression levels of mNG-Cvm1 in the presence and absence of Cvm1, yeast cells were grown to logarithmic phase in liquid YPAD medium. 100 OD₆₀₀ units were harvested and resuspended in 350 µl of buffer (50 mM Tris pH = 8, 1% Triton X-100, 1 mM PMSF, 0.1 mg/ml leupeptin, 0.5 mg/ml pepstatin A, and 0.1 mM Pefabloc). 200 µl of acid-washed glass beads were added and the cells were mechanically lysed twice in a FastPrep device (6 m/s for 40 s; MP Biomedicals), with a 5-min break on ice in between. The supernatant was transferred to a new tube and 200 µl of buffer was added to the glass beads and pooled with the supernatant. The lysate was centrifuged for 20 min at 20,000 × g at 4°C. A sample was taken for SDS–PAGE analysis, sample buffer was added and the sample was heated up at 80°C for 10 min (Input). The supernatant was added to 12.5 µl mNeon-Trap agarose beads (Chromotek), which were previously equilibrated with the buffer. The samples were incubated for 30 min at 4°C, afterwards centrifuged at 300 × g for 1 min at 4°C and a sample was taken for SDS–PAGE analysis, sample buffer was added and the sample was heated up as before (Flow Through). The beads were washed twice with 1 ml buffer (300 × g, 1 min, 4°C). For elution sample buffer was added to the beads after the last washing step, the samples were heated up as before and the supernatant was taken as the pull-down sample.

Proteins were separated by SDS–PAGE in 7.5% Tris-HCL acrylamide/bisacrylamide gels and transferred to a nitrocellulose membrane (GE Healthcare) by established protocols. Membranes were blocked for 30 min in PBS with 5% milk (wt/vol) after transfer. Subsequently, the primary antibody was incubated either for 1.5 h at room temperature or overnight at 4°C with light shaking. After the incubation, the Membranes were washed four times with PBS for 5 min, followed by incubation with a 1:20,000 dilution of a fluorescent dye-coupled secondary antibody (Thermo Fisher Scientific/Invitrogen). Fluorescent detection of the signal was performed in a Li-Cor Odyssey scanner. All antibodies used in this study are listed in Supplemental Table S4.

Pull-down of proteins ubiquitinated with 6xHis-ALINQUERA-ubiquitin for Western blot analysis

Yeast cells expressing 3xFLAG-Cvm1 and Cvm2 under the control of the inducible GAL1pr and Ubiquitin N-terminally tagged with 6xHIS and a linker in between (ALINQERA) (MacDonald et al., 2017) were grown until log-phase in YPA medium containing 2% galactose. Yeast cells expressing 3xFLAG-Cvm1 endogenously and Ubiquitin N-terminally tagged with 6xHIS and a linker in between were grown until log-phase in YPAD medium. A total of 400 to 500 OD-Units were harvested by centrifugation (4000 × g, 10 min, 4°C), washed with water and the pellets frozen in liquid nitrogen. The cells were then resuspended in Lysis buffer (50 mM HEPES/NaOH pH = 7.4, 300 mM NaCL,1.5 mM MgCl₂, 5% glycerol) supplemented with 1 mM PMSF, 1 × FY protease inhibitor mix (Serva) and lysed by mechanical disruption with 0.1 mm glass beads (Sartorius) using a FastPrep Homogenizer (2 × 6m/sec for 40 s; MP Biomedicals). The membranes were pelleted by centrifugation at 130,000 × g for 25 min at 4°C and subsequently resuspended in Extraction buffer (50 mM HEPES/NaOH pH = 7.4, 300 mM NaCl,1.5 mM MgCl₂, 5% glycerol, 20 mM Imidazole, 1% Triton X-100). Membranes were then solubilized for 30 min at 4°C in a turning wheel. Remaining debris was removed by centrifugation at 21,000 × g for 20 min at 4°C. An aliquot was taken from the supernatant for the Input control. The supernatant was then combined with pre-equilibrated Ni-NTA agarose (Qiagen) and incubated for 1 h at 4°C in a turning wheel. The Ni-NTA agarose was pelleted by centrifugation (300 × g, 1 min, 4°C), and washed four times with the extraction buffer. Competitive elution with Imidazole was performed twice by addition of the Elution buffer (50 mM HEPES/NaOH pH = 7.4, 300 mM NaCl,1.5 mM MgCl₂, 5% glycerol, 350 mM Imidazole, 1% Triton X-100) for 5 min at 4°C in a turning wheel. For endogenously expressed 3xFLAG-Cvm1, the whole eluate was precipitated by chloroform/methanol precipitation and the pellet resuspended in 1 × SDS sample buffer (4% SDS, 0.05% bromophenol blue, 0.0625 M Tris, pH 7.4, 2.5% β-mercaptoethanol, and 10% glycerol) by agitating at 80°C and sonication. The eluate from the overexpressed complex was combined with 4x SDS sample buffer. Proteins were separated by SDS–PAGE in 7.5% Tris-HCl acrylamide/bisacrylamide gels and transferred to a nitrocellulose membrane. Mouse monoclonal anti-FLAG (Invitrogen) antibodies were used to detect the 3xFLAG tagged Cvm1. All antibodies used in this study are listed in Supplemental Table S4.

GFP-Trap pull-downs and mass spectrometry analysis

The pull-downs shown in Figures 1A and 2C were performed with SILAC labeling of the cells. For this, yeast strains auxotrophic for lysine were grown in synthetic medium containing either 30 mg/ml l-lysine or 30 mg/ml 13C6;15N2-l-Lysine (Cambridge Isotope Laboratories). The pull-down shown in Figure 4H was label-free, for this, three independent cultures of each strain were grown in the YPAD medium.

The cells were grown to logarithmic phase and 350 OD-Units were harvested and washed once with water. The pellet was resuspended in lysis buffer (20 mM HEPES pH 7.4, 150 mM KoAC, 1 mM MgCl2, 5% Glycerol, 1% GDN) supplemented with 1 mM PMSF and 1 × FY protease inhibitor mix (Serva) and lysed by mechanical disruption with 0.1 mm glass beads (Sartorius) using a FastPrep Homogenizer (2 × 6m/sec for 40 s; MP Biomedicals). Remaining cell debris was removed by centrifugation (21,000 × g, 20 min, 4°C), followed by the protein concentration determination of the lysate with the Protein Assay Dye Reagent Concentrate (Bio-Rad). Each sample was normalized to the same protein concentration. The normalized lysates were then added to 12.5 µl slurry GFP Trap beads (Chromotek), respectively, which were preincubated with lysis buffer. Samples were incubated at 4°C for 15 min in a turning wheel. The GFP trap beads were centrifuged for 1 min at 300 × g and washed twice with lysis buffer and four times with the lysis buffer without detergent. The samples were further processed and digested with LysC for mass spectrometry with the iST 96x Kit (Preomics) according to the manufacturer's instructions. For SILAC-based experiments, the corresponding SILAC pairs were mixed 1:1 and analyzed together in the mass spectrometer, for label-free experiments, the samples were analyzed independently.

Experiments in Figures 1A and 2C were analyzed by mass spectrometry using reversed-phase chromatography on a Thermo Ultimate 3000 RSLCnano system connected to a Q ExactivePlus mass spectrometer (Thermo Fisher Scientific) through a nano-electrospray ion source. Of note, 50-cm PepMap C18 easy spray columns (Thermo Fisher Scientific) with an inner diameter of 75 µm were used for peptide separation with a constant temperature of 40°C. For elution of the peptides from the column a linear gradient of acetonitrile from 10 to 35% in 0.1% formic acid for 118 min at a constant flow rate of 300 nl/min was used, which was followed by direct electrospraying into the mass spectrometer. Acquiring of mass spectra on the Q-Exactive Plus was done in a data-dependent mode to automatically switch between full scan MS and up to 10 data-dependent MS/MS scans. A 50 ms was the maximum injection time for full scans, with a target value of 3,000,000 at a resolution of 70,000 at m/z = 200. The 10 most intense multiply charged ions (z = 2) from the survey scan were selected with an isolation width of 1.6 Th and fragment with higher energy collision dissociation (Olsen et al., 2007) with normalized collision energies of 27. Target values for MS/MS were set at 100,000 with a maximum injection time of 80 ms at a resolution of 17,500 at m/z = 200. To avoid repetitive sequencing, the dynamic exclusion of sequenced peptides was set at 30 s.

The experiment shown in Figure 4H was analyzed using reversed-phase chromatography through a Captive Spray Ion source on a Thermo UltiMate 3000 RSLCnano system that was connected to a TimsTOF HT mass spectrometer (Bruker Corporation, Bremen). The peptides were separated on an Aurora Gen3 C18 column (25 cm × 75 × 1.6 µm) with CSI emitter (Ionoptics, Australia) at 40°C. The peptides were eluted using a linear gradient of acetonitrile from 10 to 35% in 0.1% formic acid for 44 min, while the flow rate was constant 300 nL/min, followed by a 7-min increase to 50% and followed by 4 min to reach 85% buffer B. After elution, the peptides were electro sprayed into the mass spectrometer at an electrospray voltage of 1.5 kV and 3 L/min Dry Gas. The TimsTOF MS settings were adjusted to positive ion polarity with an MS range from 100 to 1700 m/z.

For analysis of resulting MS and MS/MS spectra, we used MaxQuant software (version 1.6.0.13, https://www.maxquant.org/) applying the integrated ANDROMEDA search algorithms (Cox and Mann, 2008; Cox et al., 2011) and Perseur (V2.0.7.0, www.maxquant.org/perseus). Resulting peak lists were compared against local databases for S. cerevisiae (obtained from the Saccharomyces Genome database, Stanford University), with common contaminants added.

Purification of the 3xFLAG-tagged Cvm1–Cvm2 complex from yeast

Yeast cells were diluted from an overnight precultured in 4 l of YPA medium (yeast extract, peptone, adenine) containing 2% galactose. Cells were grown for 24 h to an OD₆₀₀ of 3 to 4 and harvested by centrifugation (4000 × g, 10 min, 4°C). The pellet was once washed with ice-cold H₂O and a second time with the ice-cold purification buffer (PB: 300 mM NaCl, 50 mM HEPES/NaOH pH 7.4, 1.5 mM MgCl2, 10% Glycerol). The washed pellet was subsequently resuspended in a 1:2 ratio (wt/vol) in PB supplemented with 1 mM dithiothreitol (AppliChem GmbH) and a 1:50 dilution of a 50x Protease inhibitor mix containing 500 µM E64 (Carl Roth), 0,025 mg/ml Pepstatin A (Carl Roth) and 100 mM o-Phenantroline (Sigma-Aldrich). The cell suspension was then dropwise frozen in liquid nitrogen and lysed with 15  ×  2 min cycles at 12 CPS in a 6875D Freezer/Mill Dual-Chamber Cryogenic Grinder constantly cooled with liquid nitrogen (SPEX SamplePrep LLC). The powder was thawed in PB containing 1 mM DTT and the Protease inhibitor Mix. The resulting lysate was centrifuged twice (1000 × g, 10 min, 4°C) and the membranes pelleted at 100,000 × g for 45 min at 4°C. The pelleted membranes were then solubilized in 10 ml PB containing protease inhibitors and 1% glyco-diosgenin (GDN) (SERVA) in a turning-wheel for 60 min at 4°C. Subsequently, nonsolubilized membranes were pelleted at 100,000 × g for 30 min at 4°C. The supernatant was combined with 1.5 ml of anti-FLAG M2 affinity gel (Sigma-Aldrich) and incubated for 45 min at 4°C on a nutator. The beads were centrifuged at 1800 × g for 30 s and washed once with 20 ml of PB + 0.1% GDN. After the transfer to a 2.5 ml MoBiCol column (MoBiTec), the beads were washed stepwise with 10 ml PB + 0.01% GDN, 5 ml PB + 0.01% GDN + 1 mM ATP and 10 ml PB + 0.01% GDN. Elution of bound proteins was performed twice in PB + 0.01% GDN together with 3xFLAG-peptide for 30 min at 4°C. The eluates were collected by centrifugation (300 × g, 4°C, 30 s) and concentrated in a 100 kDa Amicon Ultra centrifugal filter (Merck Millipore), preequilibrated with PB + 1% GDN.

Analysis of the membrane association of Cvm1

Cells expressing GFP-Cvm1 were diluted from an overnight preculture in YDP and grown to mid-log phase at 30°C with shaking. The whole experiment was performed at 4°C or in an ice-bath. 200 OD units were harvested by centrifugation, resuspended in 500 µl lysis buffer (50 mM TrisHcl pH = 7.4, 150 mM NaCl) and lysed in a FastPrep machine (MP biomedicals) twice for 40 s. After lysis, the total volume was completed to 2 ml using lysis buffer, and the lysate was centrifuged for 3 min at 300 × g, to remove nuclei and unbroken cells. The supernatant was split into 200 µl aliquots, which were centrifuged for 30 min at 20,000 × g at 4°C. After centrifugation, one supernatant was removed and processed for SDS–PAGE. One pellet sample was resuspended in 200 µl lysis buffer and processed for SDS–PAGE. The other pellet fractions were resuspended in 200 µl of different extraction buffers: one pellet was resuspended in lysis buffer plus 1% Triton X-100, one pellet was resuspended in lysis buffer with 1M NaCl, and one pellet was resuspended in Na₂CO₃ buffer (20 mM HEPES pH 7.4, 200 mM Na2CO3). All samples were centrifuged for 30 min at 20,000 × g at 4°C, the supernatants were removed and processed for SDS–PAGE. Processing for SDS–PAGE included addition of 4x SDS sample buffer and heating to 95°C for 10 min. Samples were analyzed by SDS–PAGE on a 10% Bis-Tris Gel and Western blot on a nitrocellulose membrane using antibodies directed against GFP. All antibodies used in this study are listed in Supplemental Table S4.

Purification of KmCvm1 64-718 from E. coli

KmCvm1 64-718 was expressed recombinantly in E. coli BL21 and purified via 6xHis-affinity tag. Briefly, the cell pellet was resuspended in buffer P (50 mM NaH2PO4, 300 mM NaCl, 5% glycerol, 10 mM imidazole) supplemented with 2.5 µg/ml DNaseI, 100 µg/ml lysozyme, and 1:1000 protease inhibitor mix HP (Serva), mechanically lysed (Microfluidizer) centrifuged (45 min, 39,000 × g, 4°C) and the cleared supernatant was loaded three times on a Ni-NTA resin (SERVA) gravity flow column equilibrated in buffer P. Contaminants were removed within four washing steps (twice 20 ml buffer P with 20 mM imidaziole and twice 20 ml buffer P with 50 mM imidazole). The bound protein was eluted with buffer P containing 250 mM imidazole and the tag was cleaved off with SUMO protease. The protein was concentrated and further purified by size exclusion chromatography (Superdex 200 pg, 16/600 HiLoad, GE Healthcare). Protein was used freshly or mixed with Glycerol (10% f.c.), snap-frozen in LN2 and stored at −70°C.

Liposome sedimentation assay

Liposomes were prepared as described in Fitzian et al. (2021). Lipid mixes of different compositions (neutral: 18 mol% DOPE/72 mol% DOPC; PS: 18 mol% DOPE/64 mol%DOPC/8 mol% PS; PIP: 18 mol% DOPE/68 mol%DOPC/2 mol% PS/2 mol% PIP) were dried and resuspended in Lipo-Assay buffer (LAB: 20 mM HEPES pH 7.4, 150 mM NaCl) supplemented with 5% sucrose (wt/vol). After ten freeze-thaw cycles, liposomes were stored at −70°C. For sedimentation assays, thawed liposomes were freshly extruded to 400 nm average diameter using polycarbonate membranes.

KmCvm1 64-718 was purified by size exclusion in lipo-assay-buffer (LAB) supplemented with 2 mM TCEP. A total of 2 µM protein in 150 µl LAB with 2 mM TCEP and 50 µl extruded liposomes (2 mM) were mixed. The reaction was incubated at 25°C for 20 min and centrifuged at 21,000 × g, 4°C for 30 min. The supernatant was removed and precipitated with TCA. The pellet and the precipitated supernatant fractions were resuspended in 1x SDS-loading dye and analyzed by SDS–PAGE and Coomassie staining. Band intensities were measured with ImageJ 1.50i and mean values were calculated with Origin Pro 2023 10.0.0.154 (Academic). Three individual biological repeats performed in technical quadruplicates are plotted as mean with SE. Significances were calculated using one-way ANOVA with the Holm-Sidak test (*P < 0.05; **P < 0.001; ***P < 0.0001).

Abbreviations used:

ER

endoplasmic reticulum

GFP

green fluorescent protein

mNG

mNeonGreen

NVJ

nucleus-vacuolar junction

SBF-SEM

serial block-face scanning electron microscopy

Ub

Ubiquitin

vCLAMPs

vacuole and mitochondrial patch.