Enolase Activates Homotypic Vacuole Fusion and Protein Transport to the Vacuole in Yeast*

Membrane fusion and protein trafficking to the vacuole are complex processes involving many proteins and lipids. Cytosol from Saccharomyces cerevisiae contains a high Mr activity, which stimulates the in vitro homotypic fusion of isolated yeast vacuoles. Here we purify this activity and identify it as enolase (Eno1p and Eno2p). Enolase is a cytosolic glycolytic enzyme, but a small portion of enolase is bound to vacuoles. Recombinant Eno1p or Eno2p stimulates in vitro vacuole fusion, as does a catalytically inactive mutant enolase, suggesting a role for enolase in fusion that is separate from its glycolytic function. Either deletion of the non-essential ENO1 gene or diminished expression of the essential ENO2 gene causes vacuole fragmentation in vivo, reflecting reduced fusion. Combining an ENO1 deletion with ENO2-deficient expression causes a more severe fragmentation phenotype. Vacuoles from enolase 1 and 2-deficient cells are unable to fuse in vitro. Immunoblots of vacuoles from wild type and mutant strains reveal that enolase deficiency also prevents normal protein sorting to the vacuole, exacerbating the fusion defect. Band 3 has been shown to bind glycolytic enzymes to membranes of mammalian erythrocytes. Bor1p, the yeast band 3 homolog, localizes to the vacuole. Its loss results in the mislocalization of enolase and other vacuole fusion proteins. These studies show that enolase stimulates vacuole fusion and that enolase and Bor1p regulate selective protein trafficking to the vacuole.

Membrane fusion and protein trafficking to the vacuole are complex processes involving many proteins and lipids. Cytosol from Saccharomyces cerevisiae contains a high M r activity, which stimulates the in vitro homotypic fusion of isolated yeast vacuoles. Here we purify this activity and identify it as enolase (Eno1p and Eno2p). Enolase is a cytosolic glycolytic enzyme, but a small portion of enolase is bound to vacuoles. Recombinant Eno1p or Eno2p stimulates in vitro vacuole fusion, as does a catalytically inactive mutant enolase, suggesting a role for enolase in fusion that is separate from its glycolytic function. Either deletion of the non-essential ENO1 gene or diminished expression of the essential ENO2 gene causes vacuole fragmentation in vivo, reflecting reduced fusion. Combining an ENO1 deletion with ENO2-deficient expression causes a more severe fragmentation phenotype. Vacuoles from enolase 1 and 2-deficient cells are unable to fuse in vitro. Immunoblots of vacuoles from wild type and mutant strains reveal that enolase deficiency also prevents normal protein sorting to the vacuole, exacerbating the fusion defect. Band 3 has been shown to bind glycolytic enzymes to membranes of mammalian erythrocytes. Bor1p, the yeast band 3 homolog, localizes to the vacuole. Its loss results in the mislocalization of enolase and other vacuole fusion proteins. These studies show that enolase stimulates vacuole fusion and that enolase and Bor1p regulate selective protein trafficking to the vacuole.
Proteins and lipids are delivered to subcellular compartments through selective budding from a donor organelle, traffic through the cytosol, and docking and fusion to the target compartment. The mechanism and components involved in membrane fusion are conserved from yeast to human. Yeast homotypic vacuole fusion is a particularly convenient model for dissecting this process, given the facile isolation of pure vacuoles and the well developed genetics of this organism. Vacuoles fission and fuse in the cell in response to osmotic stimuli (1). Defects in fusion allow unbalanced fission and, hence, a phenotype of vacuole fragmentation (2). In vitro, the fusion of purified vacuoles can be monitored by assays measuring the resulting mixing of compartments (3). Using this in vitro vacuole fusion assay, we have previously reported that yeast cytosol contains two fractions that stimulate fusion, a low M r activity and a high M r activity, termed HMA 2 (4 -6). Here we purify HMA and identify it as enolase 1 (Eno1p) and enolase 2 (Eno2p). Eno1p and Eno2p are 95% identical. Although they both function as glycolytic enzymes, ENO2 is essential, whereas ENO1 is non-essential (7,8).
Because band 3, an integral membrane protein, binds glycolytic proteins to the membrane of human erythrocytes (9), we looked for a similar protein in yeast that might be responsible for the localization of cytosolic enolase to the vacuole. Bor1p has sequence homology to band 3, but little is known of its function (10). We show here that Bor1 is localized to vacuoles and that its loss results in defects in the localization of proteins, including enolase, to the vacuole.
Vacuole Isolation and in Vitro Fusion Assay-Unless otherwise indicated, vacuoles were isolated from yeast strains BJ3505 and DKY6281 for in vitro fusion (12). Where indicated, vacuole suspensions were mixed with 0.2 volume of 50% glycerol and 10 mM Pipes-KOH, pH 6.8, 200 mM sorbitol to final concentration of 3 g/l and frozen dropwise in liquid N 2 . Fusion reactions (40 l) contained 3 g of vacuoles lacking the protease Pep4p, 3 g of vacuoles from cells without Pho8p, 5 mM MgCl 2 , 10 mM Pipes-KOH, pH 6.8, 1 mM ATP, 40 mM creatine phosphate, 0.5 mg/ml creatine kinase, and 10 M coenzyme A. Fusion reactions with frozen vacuoles also contained 83 mM NH 4 Cl and 200 mM sorbitol; reactions with fresh BJ3505 and DKY6281 vacuoles contained 98 mM KCl and 455 mM sorbitol, whereas those with fresh BDY3 and BDY4 vacuoles had 98 mM KCl and 200 mM sorbitol. Fusion assays were incubated for 90 min at 27°C, and fusion was measured by assaying alkaline phosphatase (6).
Tetracycline-repressable Strains-Unless otherwise noted, all tetracycline-repressible strains were obtained from OpenBiosystems. These strains were grown in 2% yeast peptone dextrose at 30°C, and proteins were repressed with the addition of doxycycline (Sigma) dissolved at 10 mg/ml in EtOH. All experiments were performed at least three times with consistent results.

RESULTS
To identify factors involved in yeast vacuole fusion, we fractionated yeast cytosol and monitored the ability of fractions to stimulate in vitro vacuole fusion. Although freshly isolated vacuoles do not require cytosol or added components for fusion, the harsher procedure used to make frozen vacuoles (4) yields vacuoles, which are more responsive to cytosol and added purified components. Freezing the vacuoles may remove or denature fusion factors or may partially permeabilize the vacuoles; cytosolic factors could rescue such a defect.
Gel filtration of cytosol yields two peaks of activity, of high and low M r , which stimulate in vitro fusion of frozen vacuoles (4). We have previously identified the low M r activities as thioredoxin and IB2 (4,9). To identify the HMA, proteins from K-91 yeast cytosol were resolved (Table 1) by HR400 gel filtration. Fractions were tested for their ability to stimulate in vitro vacuole fusion. HMA emerged after the bulk of cytosolic proteins near 60 kDa (data not shown). This fraction was concentrated 10-fold by lyophilization and further separated by Q-Sepharose Fast Flow. The HMA activity was in the unbound fraction. HMA was further resolved by butyl-Sepharose 4 Fast Flow and by Green 19 chromatography. A silver-stained gel showed a dominant band in the active fractions (Fig. 1). This band was excised and identified by reversephase HPLC nano-electrospray tandem mass spectrometry as a mixture of enolase 1 and enolase 2.
It is surprising that enolase appears to be involved in membrane fusion, as its recognized cellular role is in glycolysis and its localization is cytosolic. To determine whether enolase is also on the vacuole, we performed immunoblot analysis of isolated vacuoles using anti-Eno1-His 6 antibody (which recognizes both Eno1p and Eno2p), anti-Vam7 antibody, and anti-Cdc48 antibody. As expected, Vam7p is predominantly on the vacuole, whereas Cdc48p is largely absent from vacuoles. Using scanning densitometry, we found enolase to be present on vacuoles (Fig.  2) at ϳ1 ng/100 ng of total vacuolar protein.
To directly test whether enolase is HMA or a contaminant that had co-purified with HMA, we tested recombinant enolase for its capacity to stimulate in vitro vacuole fusion (Fig. 3A). Eno1p-His 6 , Eno1p-GST, Eno2p-His 6 , and Eno2p-GST stimulated in vitro vacuole fusion in a dose-dependent manner. H159A Eno2p-GST, a point mutant deficient in glycolytic enzymatic activity (15), has comparable activity to the wild type Eno2p-GST (Fig. 3B). As controls we note that adding similar amounts of bovine serum albumin gave no stimulation and adding Vam7p, a known vacuolar fusion protein, stimulated in vitro fusion 5-fold (data not shown). As further confirmation that enolase is not

TABLE 1
Purification of HMA Concentrated suspensions of K-91 yeast cells were frozen drop-wise in liquid nitrogen, lysed by blender (Waring Commercial) in liquid N 2 , and stored at Ϫ80°C. The lysates were then thawed and centrifuged (8,000 ϫ g, 10 min, 4°C), and 30 ml of supernatant was applied to a 1L HR400 Gel Filtration (Amersham Biosciences) column (equilibrated in 10 mM Pipes, pH 6.8, 9.5-ml fractions). Each fraction was assayed for stimulation of in vitro vacuole fusion, and the active fractions were pooled. These fractions were lyophilized, resuspended in 0.1 volume of water, and dialyzed into 10 mM Pipes-KOH, pH 6.8, using regenerated cellulose dialysis tubing with a nominal M r cut off of 3,500 (Fisherbrand). The resulting fraction was adjusted to pH 8 by the addition of 10 N KOH and applied to a 2.5-ml Q Sepharose Fast Flow column. The fusion-stimulatory activity was in the unbound fractions. The active fraction was then adjusted to 4 M NaCl, 200 mM sorbitol, 10 mM Pipes, pH 9, and applied to a butyl-Sepharose 4 Fast Flow column equilibrated in 4 M NaCl, 200 mM sorbitol, 10 mM Pipes-KOH, pH 9. Fractions were dialyzed in to 200 mM sorbitol, 10 mM Pipes, pH 6.8, using regenerated cellulose dialysis tubing and tested for activity in the in vitro fusion assay. The active fraction (unbound) was then applied to a 3.8-ml Reactive Green 19 (Sigma) column and eluted with 0.2 M NH 4 Cl, 200 mM sorbitol, 10 mM Pipes-KOH, pH 6.8 (750-l fractions). Active fractions were pooled. Protein concentration was determined by the Bradford method (Bio-Rad). One unit of fusion stimulatory activity increases the in vitro fusion assay signal by 0.1 A 400 nm . Step

Enolase, Bor1p, and Vacuole Fusion
acting through glycolysis, we noted that fusion was not affected by 2-phosphoglycerate (an enolase substrate), other glycolytic enzymes (glyceraldehyde-3-phosphate dehydrogenase, phosphoglycerate kinase, or pyruvate kinase), or inhibitors of glycolysis (2-deoxy-D-glucose or quercetin) (data not shown). Thus enolase is the active component in the purified HMA fraction but does not act through glycolysis.
To test the role of enolase in vacuole fusion in vivo, we assessed the ability of cells lacking enolase to maintain normal vacuole structure. Because Eno2p is essential for cell viability, we obtained a strain with a tetracycline-repressible promoter substituted for the native ENO2 promoter to assess how the loss of Eno2p affects vacuole fusion. When grown in the presence of varied concentrations of tetracycline, these cells showed a dose-dependent decrease in the amount of enolase protein, as monitored by immunoblot (Fig. 4A). This decrease in enolase was accompanied by vacuole fragmentation (Fig. 4B). Cells in which the tetracycline-repressible promoter was inserted into the promoters of known fusion factors, Erg25p (an ergosterol biosynthetic enzyme) and Sec17p, showed increased vacuole fragmentation during tetracycline repression (Fig. 5). Strains in which ENO1 is deleted are deficient in vacuole fusion, resulting in a fragmented vacuole phenotype in 41% of    cells (Fig. 6B). Without added inhibitor, 26% of cells from the eno2-tet R strain have fragmented vacuoles, reflecting the fact that the tetracycline-repressible promoter is considerably weaker than the normal ENO2 promoter. When the loss of Eno1p is combined with a deficiency in Eno2p (eno1⌬, eno2-tet R ), 97% of the cells show highly fragmented vacuoles (Fig. 6, A and B). This clearly indicates an in vivo role for enolase in maintaining normal vacuole structure. Vacuoles from pep4⌬ and pho8⌬ derivatives (BDY3 and BDY4, respectively) of the eno1⌬, FIGURE 6. Vacuole fusion in an eno1⌬, eno2-tet R strain. A, the vacuoles in parental and eno1⌬, eno2-tet R strains were visualized using FM4-64 after 12 h in 2%YPD medium at 30°C without tetracycline. B, after FM4-64 staining, the percent of cells with fragmented vacuoles was quantified in a blind fashion. C, vacuoles from either wild type or eno1⌬, eno2-tet R strains were assayed for fusion in standard assays, on ice or at 27°C, either without further addition or with Eno2-GST (4 g), Vam7p (0.4 g) or anti-body to Vam3p (4 g).
eno2-tet R strain were unable to fuse in vitro and fusion was not restored by the addition of either Eno2p-GST or Vam7p (Fig. 6C). The lack of stimulation by enolase suggests additional fusion defects in these vacuoles in addition to the loss of enolase. The resistance to Vam7p, a strong activator of fusion, supports the concept of multiple defects resulting from lowered levels of enolase. To determine whether these fusion defects are because of a vacuolar deficiency in other fusion factors, lysates and vacuoles were assayed by immunoblot (Fig. 7) for relevant vacuolar proteins. Using scanning densitometry, we found 76% less enolase in the cell lysate and at least 80% less enolase on the vacuoles of enolase-deficient cells than in the cell lysate or on the vacuoles of the wild type cells. Many proteins (CPY, Pep4p, Sec17p, and Ypt7p) were present at the same levels on wild type and enolasedeficient vacuoles, indicating that vacuoles were formed and isolated intact. However, other proteins involved in vacuole fusion (Sec18p, Vps33p, Vam6p, Vti1p, Vam3p, and Nyv1p) were present at normal levels in enolase-deficient cells but were only found on the vacuoles at reduced levels.
Enolase appears to bind to vacuoles through a peripheral membrane association, because the need for Eno2p addition for in vitro fusion is increased by exposure to salt, which removes peripherally associated proteins (data not shown). What is the receptor for enolase association with vacuoles? Band 3 has been shown to localize glycolytic enzymes to the membranes of human erythrocytes (9). The yeast protein Bor1p has sequence homology to band 3 (10). To test for Bor1p on the vacuole, total cell extracts and vacuoles from strains containing a Bor1-GFP fusion protein were immunoblotted for GFP. Bor1-GFP is enriched on vacuoles as compared with the total cell lysate (Fig. 8A). The importance of vacuolar Bor1p is supported by the vacuolar phenotype of a bor1⌬ strain. Approximately 90% of cells lacking Bor1p have fragmented vacuoles (Fig. 8C ). To determine whether Bor1p is needed for enolase localization to the vacuole, immunoblots were performed on vacuoles from the bor1⌬ strain (Fig. 8B). Vacuolar markers ALP and CPY were present in amounts comparable to their abundance on vacuoles from the wild type strain, whereas Eno1, 2p, Sec17p, Sec18p, Ypt7p, and Vam3p were severely lacking on vacuoles from the bor1⌬ strain. These results establish that yeast Bor1p is localized to vacuoles and show a novel role for Bor1p in the trafficking of proteins to the vacuole.

DISCUSSION
We have purified a HMA that stimulates vacuole fusion and identified that activity as enolase 1 and enolase 2. Enolase is an abundant cytosolic protein and a glycolytic enzyme. To test whether enolase was a contaminant in HMA and not the active protein per se, recombinant proteins were tested for stimulation of the in vitro fusion assay. Either recombinant Eno1p or Eno2p stimulate in vitro vacuole fusion, confirming that the activity in the fractions purified from yeast are not due to a minor component. The ability of both Eno1p and Eno2p to stimulate fusion is not surprising because they are 95% identical (7,8).
Enolase too has a role that is separate from its function in glycolysis. Vacuole fusion is not stimulated by the addition of the substrate or product of enolase nor by the addition of other glycolytic enzymes and is not inhibited by the addition of glycolytic inhibitors. A catalytically inactive enolase mutant stimulates in vitro fusion as well as wild type enolase ((15), Fig. 2B). Thus enolase participates in glycolysis and in vacuole fusion by distinct mechanisms.
Enolase-deficient cells (i.e. those lacking Eno1p, those deficient in Eno2p, and those lacking Eno1p and deficient in Eno2p) exhibit vacuole fragmentation (Fig. 6B), indicating a role for enolase in vacuole fusion, in trafficking fusion factors to the vacuole, or both. Indeed, enolase deficiency causes a decrease in vacuolar levels of other proteins required for fusion (Fig. 7) demonstrating a role for enolase in trafficking of proteins to the vacuole in addition to its role in fusion.
Why does recombinant enolase stimulate the in vitro fusion of wild type vacuoles but not enolase-deficient vacuoles? One would expect that adding recombinant enolase to enolase-deficient vacuoles would rescue fusion if there were no secondary defects in vacuolar protein composition when enolase is deficient. However, we have found that enolase-deficient vacuoles have the secondary defect of lacking proteins that are required for fusion. This explains the inability of enolase to rescue the fusion of vacuoles from enolase-deficient cells.
Enolase has been reported to bind to a subunit of the adaptor protein complex 3 (22), a complex that provides vesicle structure and cargo specificity for vesicles moving between the Golgi and vacuole (22,23). Both ALP and Vam3p, which traffic using the adaptor protein 3 complex-containing vesicles, (23), are deficient on vacuoles when enolase is limited (Fig. 7).
Sequence comparisons of yeast proteins to human band 3 protein, which localizes glycolytic enzyme complexes to membranes, identified Bor1p as the yeast homologue of band 3 (9,10). Little is known about Bor1p in yeast other than the enhanced efflux of boron when Bor1p is overexpressed (24). To determine whether Bor1p is involved in enolase localization to the vacuole, we examined vacuoles from bor1⌬ cells.