The vacuolar kinase Yck3 maintains organelle fragmentation by regulating the HOPS tethering complex

Homeostasis of eukaryotic cells is largely dependent on dynamic compartmentalization of the endomembrane system. Compartments are designed so that they can exchange materials and undergo dramatic morphological changes in order to meet the demands of metabolism, growth, and environment. Organelle architecture seems to be governed by the same processes that facilitate inter-compartmental exchange, namely, membrane fission and fusion (for reviews see Mellman and Warren, 2000; Bonifacino and Glick, 2004). Although the basic machineries of this so-called vesicular transport are well characterized, we understand less about the coordinated mechanisms that keep them under spatiotemporal control. This regulation is essential for normal and pathological pathways of organelle assembly and disassembly and, in fact, provides membrane transport with the context that results in a functional cell. Thus, the understanding of transport regulation is a primary focus for cell biology.

The lysosome-like vacuole of budding yeast Saccharomyces cervisiae is a robust model for studying the cell biological aspects of regulated membrane flux. Several principles of vesicle targeting and membrane fusion have been established through genetic and cell biological studies of vacuole biogenesis and biochemical analysis of isolated vacuoles (Burd et al., 1998; Mullins and Bonifacino, 2001;Wickner, 2002). Vacuoles are particularly suitable for studying organelle architecture, because they are generally large, low copy, and regulate their morphology in response to many of the same signals that control morphogenesis of other organelles (Conibear and Stevens, 2002; Weisman, 2003). For example, vacuole inheritance is coordinated by the cell cycle. Early in G1, vesicular-tubular “segregation” structures bud from the vacuole and migrate from the mother cell into the emerging daughter, where they fuse to reform the characteristic low copy vacuole (for review see Weisman, 2003). Additionally, vacuoles are sensors for environmental stress. When yeast are placed into hypertonic medium, vacuoles undergo a rapid decrease in volume via a process involving phospholipid synthesis, in order to restore osmotic balance to the cell (Bone et al., 1998; Nass and Rao, 1999; Bonangelino et al., 2002b). Here, we will refer to this volume decrease as vacuole “fragmentation,” but it may actually be the result of a combination of fragmentation, tubulation, ruffling (crenellation), deflation, and retrograde transport. Conversely, vacuole fusion represents an adaptation for hypotonically-stressed cells, allowing cells to accommodate the influx of water by increasing the vacuole volume. Cell cycle–dependent inheritance and fission/fusion during osmotic stress are among several examples of situations in which vacuoles undergo regulated responses to changes in cell physiology (Weisman, 2003).

A number of components involved in vacuole fusion, fission, and inheritance have been identified (Wickner, 2002; Weisman, 2003). Despite these advances, we still do not understand how these antagonistic processes of organelle growth and disassembly are regulated. What signals induce vacuole segregation structures, fragmentation during salt stress, or vacuole growth after inheritance has been completed? To address these questions, we sought to characterize mutants with defects in vacuole morphology, beginning with those having enlarged (class D) vacuoles (Bonangelino et al., 2002a; Seeley et al., 2002). We screened these mutants for their ability to undergo regulated in vivo fragmentation during hypertonic stress, in an attempt to pinpoint the molecular cause for their lost morphological flexibility. Here, we found that failure to undergo fission is only one explanation for the class D vacuole phenotype. We report that negative regulation of fusion by the vacuolar casein kinase I (CKI) Yck3 is an additional mechanism for vacuole size control.

We used the phenomenon of vacuole volume shrinkage during high salt stress as the basis for a limited screen of the haploid Euroscarf deletion library, focusing on mutants that have partial class D vacuole morphology (for stains see Tables I and II). As depicted in Fig. 1 A, cultures were grown to log phase, and vacuoles were labeled by pulse-chase with the lipophilic dye FM4-64, which follows the endocytic pathway to the vacuole membrane (Vida and Emr, 1995). Cells were then transferred to fresh medium containing 0.4 M NaCl, which is a salt concentration sufficient to drive fragmentation of wild-type (wt) vacuoles (Bonangelino et al., 2002b). After 10 or 60 min of growth under these conditions, vacuole morphology (Fig. 1 B) was visualized by fluorescence microscopy. The results of the screen are summarized in Table II (see Table S1 for a detailed analysis).

As expected, wt vacuoles became fragmented within a 10-min salt exposure, whereas fab1Δ vacuoles stayed large and round (Fig. S1, A and B). Fab1 is a lipid kinase whose increased activity—the generation of phosphatidylinositol 3,5-bisphosphate (PtdIns[3,5]P₂)—is necessary for vacuole fragmentation during hypertonic stress (for review see Weisman, 2003). Similarly, vps3Δ vacuoles failed to fragment (Fig. S1 C). The VPS3 gene product plays a role in vacuole protein transport, and is required for vacuole inheritance (Raymond et al., 1990). The class D phenotypes of vps3Δ and fab1Δ might therefore be due to a fission defect.

This, however, is not a satisfying explanation for the phenotype seen for other class D mutants; in most mutants, vacuoles became fragmented. Fragmentation was not dependent on the Hog1 MAPK pathway, the major mechanism for long-term adaptation to high osmolarity (Fig. S1 D; for review see Mager and Siderius, 2002). This is consistent with the finding that PtdIns(3,5)P₂ synthesis during hyper-osmotic stress is also independent of this pathway (Dove et al., 1997). Interestingly, we observed that vacuole fragmentation was maintained in the wt strain and most mutants, including the hog1 deletion, during prolonged salt exposure (Table II; Fig. S1). This was unexpected because PtdIns(3,5)P₂ increases only transiently in the presence of 0.4 M NaCl (Dove et al., 1997). Prolonged vacuole fragmentation may thus require an apparatus beyond PtdIns(3,5)P₂-directed fission, perhaps involving PtdIns(3,5)P₂ effectors and/or the active inhibition of vacuole fusion.

The yck3Δ mutant exhibited a unique stress response, which was consistent with a failure to down-regulate vacuole fusion (Table II; Fig. S1 E). After 10 min of salt stress, yck3Δ vacuoles achieved a level of vacuole fragmentation similar to wt, demonstrating that Yck3 is not an upstream regulator of the PtdIns(3,5)P₂-related fission cascade. However, although wt vacuoles stayed small, yck3Δ's became large again (Fig. 1 C; Fig. S1, A and E). We analyzed the same deletion in a strain that typically has a single, enlarged vacuole (BJ3505, one of the vacuole fusion reporter strains), and confirmed these observations (Fig. 2, A and B). If deletion of yck3 results in failure to down-regulate vacuole fusion during hypertonic stress, introduction of an additional mutation that blocks vacuole fusion should revert this phenotype. Nyv1 is a vacuolar R-SNARE that seems to function exclusively in homotypic vacuole fusion (Fischer von Mollard and Stevens, 1999). Its deletion, however, does not disturb vacuole morphology, leading to the suggestion that Nyv1 can be replaced by other SNAREs in vivo (Thorngren et al., 2004). To address these issues, we constructed an nyv1Δ yck3Δ double deletion. Nyv1Δ vacuoles, as well as nyv1Δ yck3Δ vacuoles, became dramatically fragmented during hypertonic stress and stayed that way (Fig. 2, C and D). This indicates that Nyv1 is required for fusion of yck3Δ vacuoles in vivo.

YCK3/CKI3 encodes one of four budding yeast CKI isoforms (X. Wang et al., 1996). CKI family kinases are abundant, conserved, and multifunctional. Several studies have implicated them in vesicular transport in metazoans and yeast (X. Wang et al., 1996; Panek et al., 1997; Murakami et al., 1999; for review see Gross and Anderson, 1998), but a general mechanism for their action has been elusive. Yck3 has no known in vivo substrates and, apart from its localization to vacuoles via the adaptor protein (AP)-3–dependent route (Sun et al., 2004), is a mystery.

These results (and Yck3's vacuolar enrichment; Fig. 3 A, 6D; Sun et al., 2004) led us to hypothesize that Yck3 is a negative regulator of in vivo vacuole fusion during hypertonic stress. To address this possibility, we isolated yck3Δ vacuoles, and used a quantitative content mixing assay (Haas, 1995) to analyze their fusion in vitro. Yck3Δ vacuoles fused, on average, 40% better than the cognate wt vacuoles (Fig. 3 C). This result did not appear to be due to an increase in vacuole fusion machinery or assay reporter protein (Fig. 3, E and F; Fig. 7), and a yck3Δ vacuole size increase was observed with an independent visual fusion assay (Fig. S2, A and B).

If Yck3 is a negative regulator of vacuole fusion, then vacuoles containing excess Yck3 should fuse poorly. Indeed, overproduction of YCK3 (GAL-YCK3) caused a 75% reduction in in vitro vacuole fusion, even when fusion was stimulated with recombinant Sec18 (Fig. 3 D). These vacuoles have near wt levels of Pho8 reporter protein and vacuole markers (Fig. 3, E–G; Fig. 6 D), but do appear to have other defects in vacuole biogenesis: whereas growth in galactose causes wt vacuoles to become fragmented in vivo, GAL-YCK3 vacuoles are large and FM4-64 localization is dispersed, indicating a delay in endosome-to-vacuole transport (Fig. S2 D).

In vitro vacuole fusion is divisible into several assayable stages (priming, docking, and content mixing), making it possible to determine the step in which a particular mutant is impaired (for review see Wickner, 2002). During priming, SNAREs and tethering proteins in a “cis-complex” are disassembled in an ATP- and Sec18/17 (yeast NSF/α-SNAP)-dependent manner. Functionality of this stage is signaled by Sec17 release from the vacuole (Mayer et al., 1996). Sec17 release from both GAL-YCK3 and yck3Δ vacuoles was comparable to wt, indicating that Yck3 does not affect the priming stage of vacuole fusion (Fig. 3 G).

Docking, which leads to the stable association of membranes, can be subdivided into two stages: tethering, which requires the Rab GTPase Ypt7 and the multi-subunit homotypic fusion and protein sorting (HOPS vacuole)/class C-Vps complex (Price et al., 2000; Sato et al., 2000; Seals et al., 2000; Wurmser et al., 2000), is thought to facilitate the initial contacts between membranes. This is followed by trans-SNARE complex formation, which irreversibly connects membranes before lipid mixing (for review see Jahn et al., 2003). Docking can be assayed in vitro by adding FM4-64 to isolated vacuoles and observing vacuole clustering by microscopy (Mayer and Wickner, 1997). Even though wt vacuoles from galactose-grown cells are small (Fig. 3 H; Fig. S2 D), they form clusters in an ATP-dependent manner. GAL-YCK3 vacuoles, in contrast, do not cluster, indicating that these vacuoles cannot dock (Fig. 3, H and I).

If vacuoles with extra Yck3 cannot dock, is better docking the reason why yck3Δ vacuoles fuse better? Inhibitors that block priming, HOPS function, SNARE complex formation, and downstream events had similar effects on wt and yck3Δ fusion (Fig. 4 A). Remarkably, though, two Ypt7 inhibitors, when added as recombinant proteins to the fusion reaction, were unable to block fusion between yck3Δ vacuoles (Fig. 4 B). Neither the guanyl-nucleotide dissociation inhibitor Gdi1 (Haas et al., 1995), nor Gyp7-47, the normally highly potent active site of the Rab GTPase activating protein (GAP) Gyp7 (Eitzen et al., 2000), were able to block yck3Δ fusion, suggesting that the tethering stage is indeed altered in this mutant. Deletion of yck3 did not appear to overcome the requirement for Ypt7/HOPS in tethering, because antibodies against Ypt7 and the HOPS subunit Vps41 (Fig. 4 A) both inhibited fusion, and both proteins are needed for yck3Δ vacuole biogenesis (Fig. S2 E). In confirmation of previous observations that Ypt7 (Haas et al., 1995; Mayer and Wickner, 1997) is required on both partner membranes, when Gdi1 or Gyp7-47 were added to fusion reactions between wt (susceptible) and yck3Δ (resistant) vacuoles, fusion was inhibited (Fig. 4 C). Thus, Yck3 affects the tethering stage of vacuole fusion.

We analyzed the in vivo distribution of GFP-tagged proteins involved in tethering and later stages of vacuole fusion. GFP-Vps41 was strikingly enriched in punctate structures and vacuole interfaces in the yck3Δ strain, in contrast to wt cells (Fig. 5 A, arrows; Fig. S3). This suggests that, in yck3Δ, GFP-Vps41 is present in vacuole docking sites. GFP-Pho8 (a transmembrane protein not involved in fusion) is not enriched in these sites in either strain (Fig. 5 B), and other proteins—Ypt7 (Fig. 5 C), the Q-SNARE Vam3 (Fig. 5 D), and the late-acting fusion factor Vac8 (Fig. 5 E), each of which is known to be enriched at docking sites (Wada et al., 1997; Pan and Goldfarb, 1998; Wang et al., 2002)—are similarly distributed in both strains.

The above observations suggested to us a direct role for Yck3 in tethering complex modulation. Does Yck3 phosphorylate Vps41? Because we could not purify enough kinase or soluble Vps41 for in vitro phosphorylation studies, we took advantage of the fact that, when isolated vacuole fractions are incubated with ATP, Vps41 becomes retarded in its SDS-PAGE mobility (this study; Price et al., 2000). Is this due to Yck3 activity? We collected vacuole-containing fractions from wt and yck3Δ cells and found that the ATP-induced Vps41 up-shift is indeed dependent on Yck3 (Fig. 6 A). Of all class D or kinase deletion mutants (Seeley et al., 2002) tested, including the other CKI isoforms (unpublished data), yck3Δ was the only strain completely abolished for the Vps41 up-shift, indicating a specific role for Yck3 in posttranslational modification of Vps41.

We purified modified Vps41 from vacuoles, and upon phosphatase treatment, it returned to its original gel mobility (Fig. 6 B), showing that the up-shift was due to phosphorylation. Interestingly, although Vps41 has 49 potential serine/threonine phosphorylation sites (NetPhos 2.0), its gel shift is completely dependent on Yck3 (Fig. 6). To demonstrate that Vps41 is an in vivo substrate for Yck3, we analyzed Vps41 phosphorylation in the GAL-YCK3 strain. Indeed, in the presence of excess Yck3 (Fig. 6 C, bottom), Vps41 was constitutively phosphorylated (Fig. 6 C, top, and D). Further substantiating a direct connection between Yck3 and Vps41, we found that they are associated on isolated vacuoles (Fig. 6 E). We also observed that Vps41 on yck3Δ vacuoles is a competent Yck3 substrate: when Yck3⁺ vacuoles containing tagged Vps41 were incubated together with yck3Δ vacuoles containing untagged Vps41, untagged Vps41 became phosphorylated in trans (Fig. 6 F).

Gdi1 and Gyp7-47 are unable to inhibit yck3Δ vacuole fusion (Fig. 4 B). Previous analysis of vacuole tethering has shown that, after priming, the HOPS complex becomes associated with Ypt7 (Price et al., 2000). Thus, when vacuoles are incubated with ATP, Gdi1 and Gyp7-47 can extract the HOPS subunits Vps41 and Vam6, presumably because they lose their receptor on the membrane. This extraction requires two ATP-dependent steps: priming, and a second, unknown step (Price et al., 2000). In addition, a recent study found that an excess of the Q-SNARE Vam7 can overcome the ATP-requirement (priming) for vacuole fusion. This ATP-independent fusion reaction is resistant to inhibition by Gdi1 (Thorngren et al., 2004). We hypothesized that this step might be Yck3 phosphorylation of Vps41. To test this, we incubated isolated vacuoles with ATP and either Gdi1 or Gyp7-47 and analyzed Vps41 binding. As shown in Fig. 7 (A and B), in the presence of ATP, Vps41 becomes partly released from wt vacuoles (in a reaction enhanced by Gyp7-47/Gdi1), but not from yck3Δ vacuoles. The release of another HOPS subunit, the Ypt7 guanine nucleotide exchange factor (GEF) Vam6 (Wurmser et al., 2000), is also dependent on Yck3 (Fig. 7 C). This is not, however, the case for Gdi1-dependent extraction of Ypt7: it is as efficiently extracted from yck3Δ vacuoles as from wt membranes (Fig. 7 B). Given that yck3Δ vacuole fusion is not inhibited by Gdi1 or Gyp7-47, and the HOPS subunits Vps41 and Vam6 are more stably associated with these vacuoles, we suggest that HOPS-associated Ypt7 on yck3Δ vacuoles is more active than on wt vacuoles.

Based on the above observations, we hypothesized that Yck3 might phosphorylate Vps41 in order to block tethering during hypertonic stress. As shown in Fig. 8 (A and B), 30–40% of total cellular Vps41 became phosphorylated upon moving cells from YPD to 0.9 M NaCl, whereas Vps41 in yck3Δ cells was not phosphorylated; yck3Δ vacuoles do become large again in vivo even in 1 M NaCl (Fig. S4). Interestingly, when wt cells were relieved of salt stress, phosphorylation was immediately reversed (Fig. 8 B). It is plausible that rapid dephosphorylation of Vps41 permits the reestablishment of docking sites and resumption of membrane fusion. In support of this, GFP-Vps41 is dramatically enriched at interfaces between fragmented yck3Δ vacuoles when cells are hypertonically shocked (Fig. 8 C).

What do these results imply for vacuole behavior in unstressed, normally growing cells? We observed that salt-induced vacuole fragmentation resembles the formation of vacuole inheritance structures (Fig. 2; Fig. S1). Moreover, several vacuole mutants that cannot fragment during hypertonic stress also fail to form segregation structures and are thus defective in vacuole inheritance (this study; Raymond et al., 1990; Y.X. Wang et al., 1996; Gomes de Mesquita et al., 1996; Bonangelino et al., 1997; Weisman, 2003). Therefore, we asked whether Yck3-mediated vacuole fusion inhibition is also needed for the maintenance of segregation structures, and found that yck3Δ cells had a pronounced inheritance defect: ∼50% of yck3Δ cells, compared with ∼80% of wt cells, had labeled vacuoles in their buds after a long chase period (Table III; Fig. 9) . Furthermore, the yck3Δ vacuoles that were inherited were often reduced in size compared with wt, and tubulo-vesicular segregation structures were not observed; Fig. 9, arrows in wt). Strikingly, many yck3Δ cells lost their dye label over time, which is consistent with a defect in vacuole inheritance (Fig. 9; Table III, compare 45-min and 4–5-h chase). These defects in yck3Δ vacuole inheritance are less severe than in mutants deleted for inheritance machinery (vac8Δ, vac17Δ;Wang et al., 1998; Tang et al., 2003), and, indeed, inheritance in the yck3Δ vac8Δ mutant was nearly abolished (Table III, not depicted). These data exclude Yck3 as vacuole inheritance machinery per se, but strongly indicate that it is needed to ensure the fidelity of this trafficking event.

In this study, we analyzed yeast vacuole fragmentation during hypertonic stress in order to identify genes that regulate organelle morphology (Fig. 1; Table II). We confirmed the requirement for PtdIns(3,5)P₂ synthesis in the initial vacuole volume decrease, and found VPS3, a gene required for vacuole biogenesis and inheritance (Raymond et al., 1990), to also be involved in this process. Furthermore, we showed that budding yeast vacuole fragmentation in response to hypertonic stress is, in contrast to the fission yeast Schizosaccharomyces pombe (Bone et al., 1998), independent of the Hog1 MAPK pathway. Because vacuoles in these two organisms differ dramatically in morphology (fission yeast vacuoles are typically small and numerous), it is likely that they use different mechanisms for membrane remodeling.

As a result of this screen, we identified a previously uncharacterized vacuolar kinase, the CKI isoform Yck3. Deletion of yck3 had an interesting effect on vacuole morphology during salt stress: while vacuoles were able to undergo the initial decrease in volume, they became large again in a SNARE-dependent manner (Fig. 1 C; Fig. 2; Fig. 8 C), leading us to hypothesize that Yck3 prevents vacuoles from refusing during hypertonic stress. It makes sense that such a mechanism would be required. When cells are exposed to an increase in environmental salt, they need to maintain the osmotic balance of the cytosol. Therefore, the large vacuole quickly releases water by membrane fission through a process that involves transient phospholipid remodeling (Dove et al., 1997; Bonangelino et al., 2002b). Because a subsequent increase in vacuole volume would disturb the cytosolic homeostasis just achieved, a second response might be that vacuole fusion is also actively prevented. This could keep the cells alive until either the environment returns to normal or the cytosolic glycerol content increases via activation of the Hog1 pathway (Mager and Siderius, 2002; Weisman, 2003).

To confirm the role of Yck3 as a potential negative regulator of vacuole fusion, we showed that in vitro vacuole fusion is increased when Yck3 is absent, and blocked at the docking stage when overproduced (Fig. 3). This, together with our striking observation that vacuole fusion becomes resistant to Rab GTPase inhibitors when yck3 is deleted (Fig. 4 B), suggests that Yck3 is a regulator of vacuole tethering. In support of this possibility, we found that Yck3 is specifically needed for the phosphorylation of the HOPS tethering subunit Vps41 (Fig. 6). When Vps41 cannot be phosphorylated, its membrane association is altered: it accumulates at vacuole docking sites in vivo (Fig. 5 A; Fig. 8 C) and, along with the HOPS complex subunit/Rab-GEF Vam6, cannot be released from membranes in vitro (Fig. 7). We propose that, through Vps41 phosphorylation, Yck3 controls the activation state of the Rab Ypt7: Vps41 phosphorylation decreases the Rab/HOPS interaction on vacuoles, leading to reduced HOPS-associated GEF activity and increased access of the Rab-GAP and Gdi1. Thus, under conditions when Yck3 is more active, more Vps41 becomes phosphorylated, and causing tethering complex inactivation and fusion inhibition; by this, fragmented vacuoles stay fragmented. If Yck3′s activity is reduced, Vps41 and Vam6 are more stably associated with Ypt7, tethering is efficient, and fusion proceeds.

In support of our model, we found that Vps41 phosphorylation and vacuole fragmentation during hypertonic stress are indeed correlated in vivo (Fig. 8). Despite the strength of this correlation and its attractiveness as a model, we acknowledge that Yck3 may regulate tethering through multiple substrates, as this stage of vacuole fusion involves many factors in several, inter-connected complexes (Müller et al., 2002; Wickner, 2002; Wang et al., 2003). We have not yet been able to determine the exact relationship between Vps41 and Yck3 in regulating vacuole fusion and architecture, because Vps41 is essential for several stages of vacuole biogenesis (see below), including transport of Yck3 itself (unpublished data).

It is fitting that the HOPS tethering complex would be an important locus for integration of environmental/developmental signals with vacuole fusion regulation. Vesicle tethering complexes, as effectors of small GTPases of the Rab/Ypt family, have been hypothesized to regulate the specificity and timing of membrane fusion (for reviews see Zerial and McBride, 2001; Whyte and Munro, 2002), although we still lack a satisfying picture of how tethering is accomplished, let alone regulated. Further adding to their potential (and complexity), putative tethering complexes seem to have diverse roles in membrane trafficking. Each of the six subunits of the HOPS complex, for example, is required for several pathways of vacuole biogenesis, and deletion of individual genes causes cells to become temperature sensitive for growth and defective in other membrane transport events (Wada et al., 1992; Rieder and Emr, 1997; Srivastava et al., 2000; Wurmser et al., 2000; Peterson and Emr, 2001). Vps41 is especially optimal as a control center module, as it seems to be involved in fusion and fission. As part of the HOPS complex, it regulates vacuole fusion (Price et al., 2000; Wurmser et al., 2000). Through its interaction with the AP-3 complex, Vps41 is needed to form Golgi-derived vesicles that follow the alkaline phosphatase (ALP) pathway to the vacuole (Piper et al., 1997; Rehling et al., 1999; Darsow et al., 2001). This pathway transports ALP (the PHO8 gene product) and a subset of vacuolar SNAREs (for review see Mullins and Bonifacino, 2001), although the functional significance of this seemingly minor, nonendosomal route to the vacuole is still not clear. Might this pathway also be important for vacuole budding/fragmentation? Perhaps Yck3 flicks a fusion-to-fission switch. Phosphorylated Vps41 might change its contacts on the vacuole membrane, causing it to participate in budding while distracting it from its fusion role. In this context, it is noteworthy that all four budding yeast CKI isoforms can regulate small GTPases involved in vesicle formation. In one report, Yck1/2/3 overexpression compensated for an ARF-GAP deletion (X. Wang et al., 1996); in another study, inactivation of Hrr25 rescued a SAR-GEF defect in ER-Golgi transport (Murakami et al., 1999). Based on all this, we propose that organelle-specific CKI isoforms promote remodeling of the tethering/coat complexes in which these GTPases are embedded. CKI isoforms might therefore play a critical role in fine-tuning small GTPase function, perhaps even coordinating fusion/fission regulation.

Yck3 is needed to maintain vacuole fragmentation in response to high environmental salt and is also required for efficient vacuole inheritance (Fig. 9), suggesting that it plays a similar role in both pathways.

We propose a general model for regulated changes in vacuole morphology, using the framework of cell cycle–controlled vacuole inheritance (Fig. 10): early in G1, vacuoles extend tubules via lipid remodeling (step 1). This depends on (a) generation of PtdIns(3,5)P₂ from PI(3)P and (b) Vps3. Because PtdIns(3,5)P₂ levels can fluctuate (Dove et al., 1997), this step is reversible. As tubules extend, they pinch off vesicles, which must be prevented from fusing back (step 2). Phosphorylation by Yck3 of Vps41 (and possibly other proteins) is required at this stage, in order to block HOPS-mediated tethering and prevent homotypic vacuole fusion (as well as transport vesicle-to-vacuole fusion). We speculate that Yck3 phosphorylation of Vps41 also transforms it into a budding factor. In the absence of yck3, this stage is reversible: vesicles might pinch off, but they fuse back, and inheritance becomes a challenge. In the presence of active Yck3, however, these vesicles accumulate, and attach to actin filaments through the interaction of Vac8, Vac17, and the type V myosin, Myo2 (step 3; Tang et al., 2003). Thus anchored to the cytoskeleton, vacuole fragments migrate vectorially into the bud, where they fuse, reassembling the large vacuole (step 4).

The above model of vacuole inheritance has clear parallels to mitotic Golgi division in mammalian cells (for reviews see Shorter and Warren, 2002; Colanzi et al., 2003): before mitosis, the normally stacked cisternae detach and form tubules, which then fragment into vesicles. These vesicles are prevented from refusing by the same mechanism that unstacks the cisternae: phosphorylation of matrix proteins, including the vesicle tethering factor GM130, by cell cycle–controlled kinases (Lowe et al., 1998). The Golgi fragments are then distributed between nascent daughter cells, and, late in mitosis, fusion is resumed and a stacked Golgi is reestablished. Thus, fragmentation of both Golgi and yeast vacuoles is accompanied by tethering complex phosphorylation and inactivation, and the consequent inhibition of membrane fusion. By this, vesicles accumulate, which can be distributed between nascent cells.

In conclusion, we found that the CKI Yck3 acts on the HOPS tethering complex to negatively regulate vacuole fusion. In doing so, it exerts influence over vacuole assembly and disassembly, and thus controls vacuole architecture. Our findings, together with our current knowledge of Golgi inheritance, lead us to propose that tethering complex phosphorylation is a general mechanism for reversible fusion inhibition during organelle remodeling.

See figure legends for methods not described here. For yeast strains (listed in Table I) and plasmid construction see Online supplemental materials.

FM4-64 (Molecular Probes) pulse-chase analysis of whole cells was as previously described. Vacuole inheritance experiments were according to Catlett and Weisman (1998), and other labeling experiments were done as in Vida and Emr (1995). Vacuoles were visualized by phase contrast and fluorescence microscopy as described below, using the filter set 23.

For GFP microscopy, cells were grown to mid-log phase in filter-sterilized SD medium supplemented with all amino acids (except Vac8-GFP [Fig. 5 E] and GFP-Vps41 in Fig. 8 C, in which SD-Leu was used). Cells were washed and resuspended fresh medium, an aliquot was placed onto a glass slide, covered with a coverslip, and immediately analyzed at RT. Images were acquired with a Zeiss Axiovert 35 microscope equipped with AxioCam, with filter set 10 or phase contrast, with a 100× objective. Color images were acquired with Zeiss AxioVision 3.1 software, and processed using Adobe Photoshop 7.0. Figures show representative fields from multiple experiments.

All biochemical reagents were purchased from Sigma-Aldrich or Roth, unless indicated. All reagents added to vacuoles were prepared in, or dialyzed into, 10 mM Pipes-KOH, pH 6.8, 200 mM sorbitol (PS) buffer.

All antibodies used in this study were rabbit polyclonal as described previously (Veit et al., 2001). Antibodies against His₆-tagged Vps41, Vam6 (Price et al., 2000), and Pho8 were provided by W. Wickner (Dartmouth Medical School, Hanover, NH). IgGs, Fab fragments, and affinity purified antibodies were prepared as described previously (Price et al., 2000); anti-Yck3 was raised in rabbits against the kinase domain of Yck3 (aa 1–333), fused to a GST tag and purified from E. coli (using the pGEX4T3-Yck3NTD vector), and affinity purified with an Affi-Gel 10 (Bio-Rad Laboratories) column containing GST-Yck3(1-333).

To analyze protein content of yeast cells, cells were lysed in 0.25N NaOH, 140 mM β-ME, 3mM PMSF, followed by 50% TCA precipitation, and acetone wash; equal amounts (0.25–0.5 OD₆₀₀ equivalents of cells) were analyzed by Western blot. For extractions in which Vps41 phosphorylation was analyzed, 0.25 OD₆₀₀ U of cells in mid-log phase were lysed as above, but with the addition of Complete Mini protease inhibitor cocktail (PIC; Roche; 1 tablet/12 ml of buffer), and phosphatase inhibitors (50 mM NaF, 0.1 mM sodium orthovanadate, 5 mM β-glycerophosphate).

Cultures (20 ml) were grown to OD₆₀₀ = 0.8–1.0 in YPD. Until lysis, cells were treated as for the vacuole purification, adjusted relative to the smaller culture size. Spheroplasts were resuspended in 1 ml of lysis buffer (200 mM sorbitol, 50 mM KOAc, 2 mM EDTA, 20 mM Hepes-KOH pH 6.8, 1× PIC [Haas, 1995], 1 mM PMSF, 1 mM DTT), and cells were osmotically lysed in DEAE-dextran/HCl. The total extract was centrifuged at 400 g, 5 min, 4°C, and the supernatant (S4) was then centrifuged for 15 min at 13,000 g. The “P13” pellets were collected and diluted to 0.3–0.4 mg/ml in PS buffer containing 0.1× PIC.

Vacuoles (purified according to Haas et al., 1995) or P13 fractions were incubated at 26°C in PS buffer supplemented with salts (Mayer et al., 1996; 0.5 mM MgCl₂, 0.5 mM MnCl₂, 125 mM KCl), CoA (10 μM), with or without an ATP-regenerating system (0.5 mM ATP, 40 mM creatine phosphate, 0.1 mg/ml creatine kinase). Where indicated, reactions also contained recombinant Gdi1, His₆-Sec18, GST-Vam7 PX domain (aa 1–122; Boeddinghaus et al., 2002), or Gyp7-47-His₆ (Eitzen et al., 2000).

Vacuole fusion using BY4741 pep4Δ and pho8Δ vacuoles was measured by a biochemical complementation assay as described previously (Haas, 1995). Fusion was determined by measuring the absorbance of at least duplicate samples at 400 nm. Except where indicated, the results are expressed by subtracting the signal from an incubation lacking ATP. 1 U of fusion activity is defined as 1 μmol p-nitrophenol developed per min and μg pep4Δ vacuoles.

Reactions (200 μl) contained 5 μg vacuoles, 1× fusion reaction salts (as above), 0.1% Triton X-100, and 30 μg/ml proteinase K, in PS buffer. After 15 min at 30°C, enzymatic activity was quenched with the addition of 1 mM PMSF, and fusion reaction developer solution was added, followed by quenching with glycine, pH 11.5 (Haas, 1995). Pho8 activity was measured spectrophotometrically, as for the vacuole fusion assay.

Online supplemental materials, including plasmid and yeast strain construction, the Vps41 purification in Fig. 6 B, Figs. S1–S4, and Table S1, are available.

We thank all members of the Ungermann lab for support, especially Lars Dietrich for many thoughtful ideas and discussions, Christine Boeddinghaus, Haitong Hou, and Jan Rohde for sharing observations and reagents, and Gabriela Müller and Nadine Decker for expert technical assistance. David Goldfarb (University of Rochester, Rochester, NY), Ed Hurt (University of Heidelberg, Heidelberg, Germany), Michael Knop (EMBL, Heidelberg, Heidelberg, Germany), Sean Munro (MRC, Cambridge, UK), Aki Nakano (Molecular Membrane Biology Laboratory, RIKEN, Saitama, Japan), Janet Shaw (University of Utah, Salt Lake City, UT), Tom Stevens (University of Oregon, Eugene, OR), and Bill Wickner kindly shared plasmids, yeasts, and serum.

This work was supported by a grant of the DFG, the SFB638, the EMBO Young Investigator Programme, the Fonds der Chemischen Industrie (C. Ungermann), and a pre-doctoral fellowship from the National Science Foundation NSFGRFP (to T.J. LaGrassa).