The I-BAR protein Ivy1 is an effector of the Rab7 GTPase Ypt7 involved in vacuole membrane homeostasis

The yeast vacuole is equivalent to the mammalian lysosome and is required to degrade macromolecules, such as proteins and lipids, in its hydrolytic environment and make the resulting metabolites available for the cell.

Several trafficking pathways, such as the endocytic and the AP-3 pathway, as well as autophagy, direct cargo to the vacuole and require the Rab7 GTPase Ypt7 for the final fusion event (Haas et al., 1995; Wichmann et al., 1992). Ypt7 is – like all Rab proteins – a switch-like protein that is converted to its active GTP-bound form by the dimeric Mon1–Ccz1 complex, an endosomal guanine nucleotide exchange factor (GEF) (Gerondopoulos et al., 2012; Nordmann et al., 2010; Cui et al., 2014; Singh et al., 2014). The active form of Ypt7 can bind to several effectors that have been implicated in membrane trafficking. On endosomes, Ypt7-GTP interacts with the retromer complex and, thus, supports recycling of transport receptors back to the Golgi complex (Balderhaar et al., 2010; Liu et al., 2012). For fusion of endosomes with vacuoles, Ypt7 binds to the hexameric HOPS tethering complex through its subunits Vps41 and Vps39 (Bröcker et al., 2012; Seals et al., 2000; Wurmser et al., 2000), which is followed by SNARE-mediated fusion (Stroupe et al., 2009).

The role of Rab GTPases in events not directly connected to membrane trafficking is an emerging theme. For instance, Ypt7 acts in the formation of a contact zone between vacuoles and mitochondria (Elbaz-Alon et al., 2014; Hönscher et al., 2014). Metazoan Rab7 has also been linked to cellular signaling, and interacts with the PI-3-kinase Vps34 that is similar to Rab5 (Shin et al., 2005; Stein et al., 2003). Likewise, Rab5 and its yeast homolog Vps21 have functions beyond membrane fusion (Durán and Hall, 2012; Numrich and Ungermann, 2014). Rab5 binds the transcription factor APPL and, thus, regulates cell proliferation (Miaczynska et al., 2004), whereas Vps21 affects cellular Ca²⁺ signaling (Nickerson et al., 2012).

Over the last years, it has become clear that the lysosome is a central hub for cellular signaling (Bar-Peled and Sabatini, 2012; Jewell et al., 2013; Wullschleger et al., 2006). The surface of yeast vacuoles harbors the EGO complex, which contains the Ego1 membrane protein Ego3, and the two Rag GTPases Gtr1 and Gtr2 (Binda et al., 2009; Bonfils et al., 2012; Dubouloz et al., 2005; Zhang et al., 2012). A similar heptameric complex, called LAMTOR, has been identified in metazoan cells (Bar-Peled et al., 2012; Sancak et al., 2010). The EGO–LAMTOR complex binds to the target of rapamycin complex 1 (TORC1), which is also found on yeast vacuoles and mammalian lysosomes (Berchtold and Walther, 2009; Sturgill et al., 2008).

Loss of TORC1 activity results in induction of autophagy that can be divided into macro- and microautophagy (Chen and Klionsky, 2011; Mizushima et al., 2011). Microautophagy is known to sequester specific cargos or cytosolic portions into inward protrusions at the vacuolar or lysosomal surface, and has been described as a mechanism to selectively degrade mitochondria, peroxisomes, lipid droplets, parts of the ER, and parts of the nucleus (Bernales et al., 2006; Böckler and Westermann, 2014; Roberts et al., 2003; Schuck et al., 2014; van Zutphen et al., 2014; Wang et al., 2014). All these processes are largely dependent on the core machinery of autophagy consisting of 18 autophagy-related genes (ATG) genes, but the mechanism how invaginations are formed is poorly understood.

Interestingly, TORC1 activity has been linked to the subunits of the HOPS complex and Rab7 (Flinn et al., 2010; Zurita-Martinez et al., 2007), leading to the question about how trafficking and amino-acid sensing are coordinated. In this study, we identify Ivy1, an inverse (I)-BAR protein, as an effector of Ypt7 on vacuoles, that we link to membrane dynamics and TORC1 activity. Our results show that Ypt7, in complex with Ivy1, could coordinate cellular signaling and/or the regulation of metabolism, and homeostasis of the vacuole membrane.

In a previous study, we used an overexpression screen to search for proteins affecting vacuolar morphology (Arlt et al., 2011). Strong overexpression of S. cerevisiae IVY1 from the 2-µm plasmid resulted in massive vacuole fragmentation (Fig. 1A), which is in agreement with previous findings (Lazar et al., 2002). High cellular levels of this protein also caused defects in protein sorting along the endocytic pathway (Fig. 1B) (Arlt et al., 2011). Although vacuole delivery of cargo of the AP-3 pathway, the artificial fusion between GFP-tagged Snc1 and the transmembrane domain of the vacuolar v-SNARE Nyv1 (GNS), and Vps10 recycling from endosomes were not affected in cells overexpressing IVY1, the endocytosis of the Ste3 receptor was as defective as in vps21Δ cells, which lack the Rab5-like GTPase of endosomes (Fig. 1B). By using immuno-electron microscopy (IEM), ultrastructural analysis of the strain overexpressing IVY1 revealed strong accumulation of multivesicular bodies proximal to the vacuole (Fig. 1C,D), and localized Ivy1 both on the vacuolar surface and in cytosolic filamentous protein aggregates between the MVBs (Fig. 1D–F). These data suggested that Ivy1, as an interactor of Ypt7 and/or the HOPS subunit Vps33 (Lazar et al., 2002), can control fusion on vacuoles and multivesicular bodies. To test this potential role of Ivy1 in controlling fusion, we purified recombinant Ivy1 and added it to the established vacuole fusion assay (Haas et al., 1994). This assay requires both Ypt7 and HOPS, and measures the luminal mixing of isolated vacuoles (see Materials and Methods). Addition of Ivy1 inhibited homotypic vacuole fusion efficiently (Fig. 1G,H). To find out whether Ivy1 acts on Ypt7, we performed the same titration experiment with Ivy1 in the absence of ATP but the presence of the SNARE Vam7 (Thorngren et al., 2004). Under these conditions, Ypt7 is tightly bound to HOPS (Brett and Merz, 2008; Cabrera et al., 2009; LaGrassa and Ungermann, 2005), and is relieved by phosphorylation of HOPS in the presence of ATP (LaGrassa and Ungermann, 2005). Ivy1 only inhibited fusion in the presence of ATP (Fig. 1H), suggesting that this protein directly interacts with Ypt7.

To determine the relation of Ivy1 with Ypt7 and HOPS, we traced the GFP-tagged protein in wild-type and various deletion strains. In wild-type cells, endogenous Ivy1 was present in distinct puncta on the vacuolar surface (Fig. 2A). This localization was not affected by deletion of its putative interaction partner, the HOPS subunit Vps33, or Vps41 – another HOPS subunit (Fig. 2A). In the absence of Ypt7, however, Ivy1 was almost entirely cytosolic, indicating that Ivy1 requires Ypt7 to localize to vacuoles (Fig. 2A). In agreement, overexpression of IVY1 in cells that express GFP-tagged Ypt7 resulted in a strong relocalization of Ypt7 to dot-like structures that colocalized with the lipophilic dye FM4-64 (Fig. 2B). This suggests that the excess amount of Ivy1 traps Ypt7 at endosomes, which is in agreement with the observed accumulation of MVBs following ultrastructural analyses (Fig. 1C,D). In contrast, the subcellular distribution of Vps33 remained unaffected by overexpression of Ivy1 (Fig. 2B), indicating that the reported Vps33 interaction with Ivy1 (Lazar et al., 2002) is not important for intracellular function of Ivy1.

To test for direct interaction between Ivy1 and Ypt7, we incubated recombinant Ivy1 with purified GST-tagged Ypt7 or Vps21 that had been preloaded with GDP or GTPγS. Ivy1 was efficiently and specifically isolated together with Ypt7-GTP (Fig. 2C). The interaction is comparable to the interaction between Ypt7 and HOPS, which was analyzed in parallel to confirm that Ypt7 was loaded with GTP (Fig. 2C, bottom panel). To further verify the close proximity between Ivy1 and Ypt7, we used bimolecular fluorescence complementation, also known and hereafter referred to as split-YFP approach), where each protein is tagged either with the N-terminal (VN) or C-terminal (VC) segment of the Venus variant of GFP (Sung and Huh, 2007). The YFP-signal was, indeed, observed on vacuoles of wild-type cells and strongly diminished in those of vps11Δ cells, which have defective vacuoles (Fig. 2D). We, thus, conclude that Ivy1 is a so-far-unknown vacuolar effector of Ypt7.

If the Ypt7-dependent localization of Ivy1 were required to control fusion at the vacuole, one would expect colocalization with selected marker proteins of organelles involved in these events. Apart from the colocalization with Ypt7, we did not detect overlapping localization with the mCherry-tagged ESCRT-III subunit Snf7, the vacuolar Vac8 protein found at nuclear vacuolar junction (NVJs) (Pan et al., 2000) or with mitochondria, which form a recently identified contact site with vacuoles termed vCLAMP (Elbaz-Alon et al., 2014; Hönscher et al., 2014) (Fig. 3A). Similarly, the Atg8-positive pre-autophagosomal structure (PAS) (seen in Fig. 3A, bottom images, in the cell to the left) did not overlap with Ivy1 on vacuoles (shown in enlargement).

As our data did not agree with a role for Ivy1 in endosomal fusion, we searched for other proteins that colocalized on the vacuole. We previously have characterized the palmitoylated Ego1/Meh1 protein on vacuoles (Hou et al., 2005), which is part of the EGO complex comprising Ego1, Ego3, and the two Rag GTPases Gtr1 and Gtr2 (Dubouloz et al., 2005; Levine et al., 2013; Zhang et al., 2012). The yeast EGO complex and mammalian LAMTOR both activate TORC1 in the presence of amino acids (Bar-Peled et al., 2012; Binda et al., 2009; Bonfils et al., 2012; Dubouloz et al., 2005). Interestingly, the HOPS subunit Vps39, which requires Ypt7 for its vacuolar localization (Bröcker et al., 2012), colocalizes with the EGO complex and has been implicated in EGO function (Binda et al., 2009). We, therefore, localized Ivy1 relative to EGO components, and observed both Ivy1 and EGO components in the same domains on vacuoles (Fig. 3B) that also localized with Ypt7 and Vps39 (Fig. 3C) (Binda et al., 2009).

To test for close proximity to the EGO complex, we utilized again the split-YFP approach. Within this analysis, we observed a clear and specific signal when Ivy1-VC was expressed with Gtr2-VN or Ego3-VN, and a weak signal when it was expressed with Iml1-VN (Fig. 3D), the identified GTPase activating protein (GAP) of Gtr1 (Bar-Peled et al., 2013; Panchaud et al., 2013). VN-Gtr2 did not yield any signal, suggesting that the position of the tag could bias the read-out of this assay. Together, these data suggest that Ivy1, indeed, localized to the vicinity of the EGO complex. To further substantiate these findings, we used the membrane-based split-ubiquitin two-hybrid system (Nikko and Andre, 2007) and, again, observed interactions of Ivy1 with components of the EGO complex, such as Gtr1 (Fig. 3E). Gtr2 was not detected here, presumably because of the same tag-related problem as in the split-YFP assay. Additionally, Ivy1 strongly interacted with itself (Fig. 3E), indicating that the protein forms dimers or oligomers as suggested by the IEM analysis (Fig. 1C,D). We conclude that Ivy1 is, indeed, proximal to the EGO complex on the vacuole, which might provide a functional link.

We next wondered whether the proximity of Ivy1 to the EGO complex has functional consequences on TORC1 signaling. Deletion of IVY1 alone did not affect the localization of EGO to dot-like structures on vacuoles as revealed by Ego3-GFP distribution (Fig. 3F), indicating that Ivy1 is not a structural component that is responsible for the localization of the EGO complex. Vice versa, deletion of EGO subunits such as Ego3 did not change the localization of Ivy1 (Fig. 3G). Thus, Ivy1 and EGO are not necessary for localization of each other but may have a functional relationship. We, thus, used growth assays in the presence of the well-established TORC1 inhibitor rapamycin. Growth on rapamycin-containing plates identifies mutants that cause defects in TORC1 activation. Consequently, deletion of EGO1 or EGO3 results in a growth defect on rapamycin. However, ivy1Δ cells did not show any growth impairment on rapamycin (Fig. 4A). We speculated that Ivy1 might affect TORC1 activity more indirectly or even act as a negative regulator. Under these conditions, a growth defect on rapamycin-containing plates would not be expected for ivy1Δ cells.

We, therefore, searched for conditions under which Ivy1 is important or essential for growth and, thus, crossed ivy1Δ cells against the entire deletion library of viable yeast-knockout strains and analyzed the growth phenotype of the respective double mutants (Tong et al., 2001). Among the respective double mutants that were most seriously compromised for growth were those, in which ivy1Δ was combined with mutations within specific subunits of the V-ATPase (Fig. 4B). This was an interesting observation because the V-ATPase has been identified as a possible sensor of amino acid availability within the vacuole (Zoncu et al., 2011). In addition, the V-ATPase is involved in acidification of the vacuole, which can be bypassed by growing cells in low pH medium – as shown for the vma2 mutant – but results in sensitivity to high pH (Fig. 4C). Since ivy1Δ cells did not show sensitivity, we excluded a function regarding the regulation of vacuolar pH (Fig. 4C).

We next analyzed the effects of loss of Ivy1 in combination with loss of individual V-ATPase subunits in our assays, and focused primarily on the ivy1Δ vma16Δ mutant. Because of the connection to TORC1, we tested this double mutant for its sensitivity to rapamycin. Under the conditions applied, the ivy1Δ vma16Δ double knockout was hypersensitive to rapamycin, as known for ego1Δ cells (Fig. 4D). We then tested whether the TORC1 activity in these strains was reduced as expected, and thus analyzed the phosphorylation status of the TORC1 target Sch9 to this end (Urban et al., 2007). Although each single deletion incurred only a minor defect in TORC1 activity, the ivy1Δ vma16Δ double mutant was strongly impaired (Fig. 4E,F). This defect is not due to reduced proteolytic activity of the mutant vacuoles because processing of the vacuolar alkaline phosphatase was still functional in ivy1Δ vma16Δ cells, but lost when the general peptidase Pep4 is absent (Fig. 4G). This suggests that the loss of TORC1 activity analyzed under these conditions is probably not owing to a problem in the acidification of the vacuole lumen. Furthermore, localization of the EGO complex is not influenced in the ivy1Δ vma16Δ mutant (Fig. 4H).

To unravel the reason for the growth defect on rapamycin and the reduced TORC1 activity, we analyzed the morphology of ivy1Δ vma16Δ vacuoles. Surprisingly, we noticed strong accumulation of membranous material in the vacuole lumen, which was stained by the lipophilic dye FM4-64 (Fig. 5A). Similar observations were made when a vma6Δ deletion with a defect in a V₀-ATPase subunit was combined with ivy1Δ, indicating that the general loss of V-ATPase activity causes in the same phenotype. To understand the origin of these membranous structures, we examined the double mutant at ultrastructural level by electron microscopy and observed numerous large vesicles whose content was indistinguishable from the cytoplasm within the vacuole lumen (Fig. 5B). We hypothesize that these structures are autophagosomes and, therefore, analyzed the efficiency of autophagy by monitoring the processing of GFP-Atg8 to free GFP – a standard method to assess the progression of this pathway (Klionsky et al., 2012) – when cells were starved by nitrogen (Fig. 5C). Whereas wild-type and ivy1Δ cells showed similar profiles of GFP-Atg8 processing, autophagy was already induced in vma16Δ cells during vegetative growth (as indicated by free GFP at the first time point) and increased over time. However, additional deletion of IVY1 did not enhance this process. We, therefore, asked whether the vesicular structures in the vacuole of ivy1Δ vma16Δ cells are the results of the fusion of autophagosomes. When we monitored these cells by fluorescence microscopy, we noticed that, when stained with the vacuole-specific dye CMAC, the vacuole lumen but not the vesicular structures was positive for GFP-Atg8, a marker protein for autophagosomes (Fig. 5D). This suggests that these vesicles are connected to the vacuolar surface. To examine this in more detail, we recorded Z-stacks of vacuoles, and confirmed that the apparent vesicular structures are, indeed, continuous with the vacuole limiting membrane (Fig. 5E, supplementary material Movie S1). Thus, ivy1Δ vma16Δ vacuoles do not accumulate vesicles inside the lumen but have an enlarged and invaginated vacuolar surface.

As the deletion of VMA16 already leads to increase of basal autophagy, we asked whether the same phenotype also occurs when we combine EGO mutants (which have lesser TORC1 activity) with the deletion of ivy1. Neither loss of EGO activity in an ivy1Δ mutant nor the increase of EGO activity (due to deletion of IML1) in a ivy1Δ vma16Δ double mutant affected the expanded membrane phenotype (Fig. 5F). This suggests that Ivy1 is more directly involved in maintaining the vacuolar membrane homeostasis and, thus, affects TORC1 activity.

To understand the role of Ivy1 in membrane dynamics, we searched for putative homologues by using HHpred Software (toolkit.tuebingen.mpg.de/hhpred), and readily identified the mammalian proteins IRSp53 and missing-in-metastasis (MIM) (Mattila et al., 2007). These proteins share with Ivy1 a central I-BAR/IMD domain, but the homology does not extend to regions outside this motif (Fig. 6A). IRSp53 and MIM are both involved in the formation of filopodia of mammalian cells, and are able to induce inward protrusions on liposomes (Becalska et al., 2013; Mattila et al., 2007). To analyze the function of Ivy1 on membranes, we purified full-length Ivy1 with a N-terminal mGFP tag that – unlike the isolated I-BAR domain – worked well during the biochemical characterization (our unpublished observations). Gel filtration of the purified protein resulted in a substantial high-molecular-mass peak indicative of multimerization, and a clearly distinct fraction of dimeric protein. As BAR domain proteins form dimers in solution (Frost et al., 2009; Mattila et al., 2007), we used this fraction for further experiments (Fig. 6B).

To analyze Ivy1, we prepared giant unilamellar vesicles (GUVs) by using a vacuolar lipid composition that supports efficient fusion of liposomes that carry vacuolar SNAREs (Mima and Wickner, 2009; Mima et al., 2008; Stroupe et al., 2009; Zinser et al., 1991). We then added Ivy1 to those GUVs. Strikingly, Ivy1 localized to specific domains that often contained invaginated or deformed membranes (Fig. 6C). Localization of Ivy1 to the GUV surface did not require Ypt7, presumably because its ability to recognize the membrane surface itself, even though Ypt7 has a profound effect on Ivy1 behavior (see below). Occasionally, Ivy1 was also found within such invaginations, which agrees with an active role of Ivy1 in inward membrane remodeling (Fig. 6C, bottom). As these domains were rather stable, we were, however, unable to determine whether Ivy1 senses or induces membrane perturbation. A close-up view indicated that Ivy1 polymerized into scaffold-like structures along the GUV membrane, in agreement with its ability to self-interact (Fig. 3E), form domains in vivo (Fig. 2A) and self-associate into multimer in vitro (Fig. 6B).

As Ivy1 binds directly to Ypt7-GTP (Fig. 2C) and interacts with phosphoinositides, such as phosphatidylinositol (3)-phosphate (PI3P) (Fig. 6E), we tested the influence of either factor on the behavior of Ivy1 on GUVs. When we incubated Ivy1 with GUVs that lacked phosphoinositides, Ivy1 was homogenously distributed over the entire membrane (Fig. 6F). Again, domain localization of Ivy1 was observed in the presence of PI3P but also in that of phosphatidylinositol (3,5)-bisphosphate (PI-3,5-P₂), which are both present on the yeast vacuoles (Fig. 6G,H, top rows). We then tagged Ypt7 at its C terminus with hexahistidine and added this fusion protein to GUVs containing DOGS-NTA lipids (1,2-dioleoyl-sn-glycero-3-[(N-(5-amino-1-carboxypentyl)iminodiacetic acid)succinyl]), and Ypt7 was correctly recruited to the GUV membrane. Interestingly, the addition of Ypt7-GTP prevented the domain formation of Ivy1 on membranes (Fig. 6G,H, bottom rows). This suggests that the amount of Ypt7-GTP and of PI3P on vacuole membranes affects the degree of Ivy1-dependent domain formation and, thus, function in vivo.

Owing to its behavior in vitro, we postulated that Ivy1 has a similar role in vivo. We, thus, analyzed the effect of the mutation of eight positively charged residues within the I-BAR domain had on Ivy1 function; the residues were identified on the basis of homology modeling, by using the IRSp53 structure (Lee et al., 2007) as a template. This mutated version of Ivy1 was unable to complement the defects in an ivy1Δ vma16Δ mutant regarding vacuolar morphology and rapamycin hypersensitivity (Fig. 7A,B), suggesting that these residues within the predicted I-BAR domain are important for Ivy1 function.

Furthermore, we hypothesized that Ivy1 acts during microautophagy that had previously been observed in response to nitrogen starvation (Muller et al., 2000) or in situations of stress, such as starvation (Toulmay and Prinz, 2012; Wang et al., 2014). During starvation, we – indeed – found Ivy1-GFP in microautophagic invaginations in 18% of the vacuoles (Fig. 7C). Recent work indicated that the vacuole membrane forms distinct domains upon prolonged starvation, and Ivy1 was identified as one of the marker proteins for such domains (Toulmay and Prinz, 2013; Wang et al., 2014). The localization to these domains is strongly affected by the alteration of PI3P levels (Fig. 7D). If Ivy1 were to accumulate on such domains and induce a negative curvature, it might be able to regulate membrane content, stabilize negative curvature, or support membrane protein degradation. We reasoned that prolonged heat would induce sufficient stress to vacuoles in order to distinguish between these options and, thus, followed the fate of Ivy1-GFP before and after prolonged heat shock (Fig. 7E-G). We, indeed, observed a dramatic relocalization of Ivy1-GFP from a dot-like location on the vacuole membrane to domains with high negative curvature (Fig. 7E), which is in agreement with the predicted membrane preference of the I-BAR domain within Ivy1. Ivy1 was almost exclusively in domains of negative curvature (Fig. 7F, enlarged area, 7G), which became stable over time and are probably not microautophagy intermediates but, rather, are required to stabilize the vacuole during heat shock. Their appearance was similar to the sterol-rich domains recently described for vacuoles of cells in stationary phase (Toulmay and Prinz, 2013; Wang et al., 2014). However, Ivy1 is not necessary for the induction or maintenance of these membrane invaginations and domain formation (marked here with Ego1-GFP), because the ivy1Δ mutant shows the same vacuolar morphology under conditions of heat shock (Fig. 7H). Importantly, this assay shows that Ivy1 has, indeed, a preference for negative curvature as predicted from its domain structure.

Our in vitro experiments showed that Ivy1 interacts with several phosphoinositides (Fig. 6E), and that the presence of PI3P or PI-3,5-P₂ might affect the behavior of Ivy1 on membranes (Fig. 6F–H). We, thus, analyzed mutants affecting PIP levels on vacuoles and noticed that the deletion of Fab1 – the PI-3,5-kinase of yeast – which results in a very large vacuole (Gary et al., 1998), led to a redistribution of Ivy1-GFP from a dot-like to a more homogenous localization along the vacuole surface (see below). The same large vacuole phenotype has been observed in several other mutants involved in PI-3,5-P2 biogenesis, such as atg18Δ, vac7Δ and fig4Δ, suggesting that membrane turnover is be defective (Duex et al., 2006; Gary et al., 2002). We, thus, asked whether Ivy1 is a marker that can be used to observe membrane remodeling upon Fab1 induction. For this, we placed FAB1 under control of the inducible GAL1 promotor, which represses Fab1 production in the presence of glucose and mimics the deletion phenotype (Fig. 8A) (Gary et al., 1998). Like in the deletion phenotype, Ivy1-GFP was distributed over the entire vacuole surface (Fig. 8A). When we induced Fab1 by adding galactose, we frequently observed Ivy1 in dots within the vacuole lumen (Fig. 8A,B). As these structures did not stain well with FM4-64, we are not yet sure whether Ivy1 just localizes to or induces these structures. However, the location of Ivy1, again, implies a role in microautophagy events. Ivy1 seems to be induced during starvation, which also results in accumulation of lipid droplets proximal to the vacuole (Wang et al., 2014). Therefore, we asked whether Ivy1 localization correlates with alterations of the vacuolar surface. We grew cells for several days and observed Ivy1 localization (Fig. 8C). Whereas Ivy1 remained in dots until the diauxic shift, it accumulated in domains on the vacuole during stationary phase. Some Ivy1 protein became visible within the vacuole lumen after 4 days, suggesting that a small fraction of Ivy1 is degraded during microautophagy, which is known to occur in cells in stationary phase. However, the main fraction of Ivy1 protein remained stable within the cell, and we did not observe massive degradation of Ivy1 under these circumstances. Together, our data provide evidence that the I-BAR protein Ivy1 is involved in vacuole membrane homeostasis. Moreover, it is probably the first protein that can be specifically used as a marker protein in order to visualize microautophagy processes at the vacuole.

Our data demonstrate that Ivy1 is a novel Ypt7 effector, which controls membrane homeostasis of the yeast vacuole. Ivy1 requires Ypt7 for its recruitment to vacuoles, and its function is controlled by the PI3P content on the surface of this organelle. Ivy1 strongly colocalizes and interacts with the EGO complex, suggesting that it is involved in regulating TORC1 activity. Its main function seems to be the regulation of membrane homeostasis of the vacuole, probably together with other proteins. In agreement with this notion, we observed that the concomitant loss of Ivy1 and the V-ATPase subunits results in a strong expansion of the vacuolar surface with multiple invaginations, which indicates that Ivy1 is a factor required to reduce vacuolar membrane. Taking GUV membranes as a model system, we further show that Ivy1 can self-organize into distinct membrane microcompartments that require PI3P and are modulated by the Ypt7 content (Fig. 6). This analysis also revealed that Ivy1 has a preference for negative curvature. To understand its organization in vivo, we searched for conditions, where vacuoles would require such proteins and identified Ivy1 in vacuolar domains of negative curvature that formed during microautophagy, starvation and response to heat shock. The close connection to Ypt7 and the EGO complex suggest that Ivy1 connects trafficking and signaling processes in the context of vacuole membrane biogenesis.

Ivy1 was initially thought to be implicated in endocytic trafficking because its overproduction resulted in vacuole fragmentation and the appearance of MVBs (Lazar et al., 2002). However, we consider this unlikely because – even though we also observed a similar impact on vacuole biogenesis and endocytic trafficking (Fig. 1) – this defect required massive overexpression of Ivy1, which leads to an accumulation of Ivy1 deposits proximal to MVBs and vacuoles (Fig. 1C-F). Similarly, the inhibition of the vacuole fusion assay is the simple consequence of the ability of Ivy1 to bind Ypt7-GTP on vacuoles. In the same study, the HOPS subunit Vps33 was suggested to be a main binding partner of Ivy1 (Lazar et al., 2002) but, so far, we did not find evidence for this. Ivy1 overexpression did not affect relocalization of Vps33 to vacuolar membranes (whereas Ypt7 was clustered on vacuoles) and Vps33 deletion did not affect Ivy1 localization (Fig. 2). The previously observed rescue of Ivy1-induced vacuole fragmentation in response to Vps33 overexpression could be a consequence of better stability of the HOPS complex on vacuoles. Whether Vps33 is involved in the localization of HOPS with the EGO complex and Ivy1 on vacuolar membranes will require further analyses.

Our data provide strong arguments in favour of a so-far-unknown Ypt7-dependent function of Ivy1 in the organization of the vacuolar membrane (Fig. 8D). Athough Ivy1 is not required for the formation of vacuolar domains (Toulmay and Prinz, 2013) or the localization of the EGO complex (shown here), it strongly accumulates in distinct membrane subdomanins on the vacuole together with the EGO complex, Ypt7 and Vps39 as one of the tested HOPS subunits (Fig. 2) (Binda et al., 2009). These domains are proximal to the previously characterized contact site for vacuoles and mitochondria or endosomes, suggesting that this zone provides a template for the assembly of multiple membrane proteins. Strikingly, Ivy1 participates to a distinct sterol-rich domain on vacuoles that is enriched under conditions of starvation (Toulmay and Prinz, 2013). Recent data further show that the same vacuolar domains are the site of starvation-induced lipophagy (Wang et al., 2014), which is a long-term adaptation of cells to nutrient deprivation. This process depends on a number of autophagy proteins, including the autophagy-specific PI3-kinase complex (van Zutphen et al., 2014; Wang et al., 2014). Ivy1 strongly localizes to these zones together with the PI3-kinase complex during stationary phase (Wang et al., 2014). We believe that the massively enlarged vacuolar membrane in mutants lacking Ivy1 and V-ATPase subunits could be the result of a defect in the organization of this subdomain that, consequently, affects microautophagy and signaling through the EGO complex. Potentially, Ypt7 binding by Ivy1 also limits the available Ypt7 pool and, thus, favors lipophagy and other types of microautophagy under conditions of starvation, thus providing a convenient shift from trafficking to (micro)autophagy programs.

Recently, a new I-BAR protein that affects phagocytosis and cytokinesis was characterized in Dictyostelium discoideum (Linkner et al., 2014). Like Ivy1, IBARa localizes to the vacuole in a PI3P-dependent manner and does not seem to affect the formation of filopodia. However, unlike Ivy1, IBARa is basically reduced to its I-BAR domain, which – by itself – induces liposome invaginations. Even though the cellular phenotypes upon IBARa deletion suggest a direct role in phagocytosis, the precise function of the protein in vivo remains elusive. Our data also did not reveal whether Ivy1 itself drives protrusions into the vacuole, although we located the protein exactly at these sites. Furthermore, the function of Ivy1 might be regulated post-translationally. Because the isolated I-BAR domain of Ivy1 was insoluble following overexpression, we were unable to test its ability to deform membranes. However, neither Ivy1 overexpression nor the addition of full-length Ivy1 to GUVs (Fig. 6) seemed to result in the induction of protrusions. We, thus, consider it more likely that Ivy1 follows and, possibly, promotes negative membrane curvature, rather than inducing it. Future analyses of Ivy1 mutants will be necessary to dissect this issue in more detail.

In summary, our data provide a molecular link of an I-BAR protein to the Rab7-like Ypt7 GTPase and, possibly, to the PI3P content on vacuoles. Moreover, they suggests a crosstalk between one of the main nutrient-sensing signal cascades and microautophagy (Fig. 8D), resulting in an attractive model outlining the regulation of vacuolar membrane homeostasis as an adaptation to nutrient availability.

All yeast strains used in this study are listed in supplementary material Table S1. For cloning, IVY1 coding sequence was amplified from S. cerevisiae genomic DNA with Phusion polymerase (Thermo Scientific) and cloned into E. coli expression vectors. Plasmids were transformed into E. coli BL21 (DE3) Rosetta cells.

To overproduce Ivy1 from E. coli with an N-terminal His-tag, the coding region of IVY1 was cloned into the BamHI/XhoI sites of a pET32c vector. To generate mGFP-tagged Ivy1, IVY1 was cloned into the BamHI/XhoI sites of a pCOLAHS vector. Afterwards, mGFP was cloned in front of IVY1 into the BamHI/BglII sites of the pCOLAHS vector, that had been digested with BamHI. The pCOLAHS vector was derived from the pCOLADuet1 vector by cloning a SUMO-tag in frame with the His-tag. All plasmids are listed in supplementary material Table S2.

The fusion assay employs two sets of isolated vacuoles. One lacks the alkaline phosphatase Pho8 but contains a full set of proteases, whereas the other contains the alkaline phosphatase as the inactive pro-form owing to the absence of the Pep4-protease. During fusion, lumenal mixing results in cleavage and activation of Pho8, which can be assayed spectrophotometrically (Cabrera and Ungermann, 2008). Vacuoles were purified from strains BJ3505 (pep4Δ) and DKY6281 (pho8Δ). Fusion reactions containing 3 µg of each vacuole type were performed in fusion reaction buffer (10 mM PIPES/KOH pH 6.8, 5 mM MgCl₂, 125 mM KCl, 0.2 M sorbitol), containing an ATP-regenerating system. Reactions were incubated for 90 min at 26°C, and then developed as described (LaGrassa and Ungermann, 2005).

For microscopy, yeast cells were grown in yeast peptone dextrose (YPD) medium, yeast peptone galactose (YPG) or in selective medium to an OD₆₀₀ of about 1, collected by centrifugation (3 min at 4000 g, 20°C), washed with synthetic medium containing either glucose or galactose, and immediately analyzed by using fluorescence microscopy. Staining of the cells with FM4-64 was performed as described (LaGrassa and Ungermann, 2005). In brief, cells were incubated with 30 μM FM4-64 for 30 min, washed with the corresponding medium and further incubated for 1.5 h in fresh medium. Images were acquired using a Leica DM5500 B microscope equipped with a SPOT Pursuit camera with GFP, RFP, FM4-64 and DIC (differential interference contrast) filters or using a DeltaVision Elite fluorescence microscope (Applied Precision, Issaquah, WA) equipped with a CoolSNAP HQ Camera using FITC, YFP, TRITC, DAPI and mCherry filters. Images acquired with the Leica microscope were deconvolved using the Metamorph software. Images acquired with DeltaVision Elite were deconvolved using SoftWorx. Pictures were processed using Adobe Photoshop CS4 or ImageJ.

The tagging of genes of interest for the bimolecular fluorescence complementation assay (BiFC, also known as split-YFP approach) was carried out as described (Sung and Huh, 2007). In brief, proteins of interest are tagged either with the N-terminus or the C-terminus of the Venus fluorescent protein. Complementation of Venus fluorescence was examined by using the YFP fluorescence channel.

Examination of yeast cells by electron microscopy and IEM was carried out as described (Griffith et al., 2008). For the IEM, cells were fixed, embedded in gelatin and cryo-sectioned. This was followed by immuno-labeling with an anti-GFP antibody (Abcam) and by protein-A–gold incubation to localize Ivy1. Sections were analysed by using an electron microscope (1200 EX; JEOL).

TORC1 activity was measured as described (Urban et al., 2007). Phosphorylation of the C-terminal part of HA-tagged Sch9 was used as readout for TORC1 activity. In brief, whole-protein extracts were prepared, treated with 2-nitro-5-thiocyanatobenzoic acid (NTCB), and further analyzed by SDS-PAGE and western blotting. Sch9 was visualized by decoration with anti-HA. Quantification of TORC1 activity was carried out as previously reported (Binda et al., 2009). Wild-type TORC1 activity was set to 100%.

The glutathione–Rab pull-down experiment was done as described before (Markgraf et al., 2009). recombinant GST-tagged Rab proteins were loaded with 1 mM GDP or GTPS in 20 mM HEPES/NaOH pH 7.4. 150 µg Rab proteins were coupled to GSH beads. Rab proteins were incubated with recombinant His-tagged Ivy1 (or control proteins) for 1 h at 4°C on a nutator mixer. Beads were washed three times with 20 mM HEPES-NaOH, 100 mM NaCl, 1 mM MgCl, 0.1% (w/v) Triton X-100, and proteins were eluted with 20 mM HEPES-NaOH, 100 mM NaCl, 1 mM MgCl, 0.1% (w/v) TritonX-100+20 mM EDTA. The eluates were then precipitated with trichloracetic acid (TCA), and analyzed by SDS-PAGE and western blotting. As a loading control, the bound GST–Rab GTPase was eluted from the beads by boiling in sample buffer, and analyzed by SDS-PAGE and Coomassie staining.

E. coli BL21 (DE3) Rosetta cells containing the IVY1-plasmids or YPT7/VPS21-plasmids were grown until OD₆₀₀ of 0.8; expression was induced with 0.5 mM IPTG overnight at 16°C. Cells were harvested and lysed in 50 mM Tris/HCl pH 7.5, 150 mM NaCl, 1 mM PMSF, 1× protease inhibitor cocktail; 1×=0.1 mg/ml of leupeptin, 1 mM o-phenanthroline, 0.5 mg/ml of pepstatin A, 0.1 mM Pefabloc. Lysates were centrifuged 15 min at 30,000 g and the cleared supernatant was added to Ni-NTA beads for His-tagged protein or to GSH-beads for GST-tagged protein, followed by an incubation for 1 h at 4°C on a nutator mixer. Ni-NTA-beads were washed with 25 ml buffer containing 20 mM imidazole. His-tagged Ivy1 was eluted from beads with buffer containing 0.3 M imidazole. His-SUMO-mGFP-Ivy1 was eluted using the SUMO protease. GST-Ivy1 was eluted using buffer containing 15 mM reduced glutathione. The buffer was finally exchanged by dialyzing the eluted proteins against 10 mM PIPES/KOH pH 6.8, 200 mM sorbitol, 150 mM KCl, 5 mM MgCl₂ containing 10% glycerol.

Giant unilamellar vesicles (GUVs) were prepared by electroformation as described before (Romanov et al., 2012). Different lipid compositions were used as indicated in the figure legends. Lipids were from Avanti polar lipids and Echelon Biosciences.

Tandem affinity purification (TAP) of HOPS complex was carried out as described (Ostrowicz et al., 2010; Puig et al., 2001). In brief, logarithmically growing yeast cells were lysed in 50 mM HEPES/NaOH pH 7.4, 150 mM NaCl, 1.5 mM MgCl₂, centrifuged for 10 min at 20,000 g and 1 h at 100,000 g. The cleared lysate was incubated for 1 h with IgG Sepharose beads. Bound protein was eluted using TEV protease. Eluates were analyzed using SDS-PAGE and western blotting.

We thank Markus Babst for suggestions and members of the Ungermann lab for discussions. This work was supported by the DFG (UN111/7-1), the SFB 944 (project P11) and by the Hans-Mühlenhoff foundation (to C.U.), and the Swiss National Foundation (to C.D.V.). F.R. is supported by ECHO (700.59.003), ALW Open Program (821.02.017 and 822.02.014), DFG-NWO cooperation (DN82-303) and ZonMW VICI (016.130.606) grants.