The Vesicle Transport Protein Vps33p Is an ATP-binding Protein That Localizes to the Cytosol in an Energy-dependent Manner*

Molecular mechanisms of vesicle transport between the prevacuolar compartment and the vacuole in yeast or the lysosome in mammalian cells are poorly under-stood. To learn more about the specificity of this intercompartmental step, we have examined the subcellular localization of a SEC1 homologue, Vps33p, a protein im-plicated to function in transport between the prevacuolar compartment and the vacuole. Following short pulses, 80–90% of newly synthesized Vps33p cofractionated with a cytosolic enzyme marker after making permeabilized yeast cells. However, during a chase, 20–40% of Vps33p fractionated with permeabilized cell membranes in a time-dependent fashion with a half-time of ; 40 min. Depletion of cellular ATP increased the association rate to a half-time of ; 4 min and caused 80–90% of newly synthesized Vps33p to be associated with permeabilized cell membranes. The association of Vps33p with permeabilized cell membranes was reversible after restoring cells with glucose before permeabilization. The N -ethylmaleimide-sensitive fusion protein homo-logue, Sec18p, a protein with known ATP binding and -GFP GFP pRS426 pBG423–2GFP. Kex2-GFP 4-kilobase pair Bam HI- Bam HI pYcP-KX2 ligated the Bam HI site of pRS416, pBG416-KX. pBG416-KX polymerase chain reaction Spe I/ Sac I ends, with Spe I/ ligated pBG426- an equal volume of 4% SDS and 4 m M dithiothreitol. The samples were transferred to a 1.5-ml tube, and the wells were rinsed with 2% SDS and 2 m M dithiothreitol and pooled with the first portion. The samples were heated (65–100 °C) for 2–5 min. The reactions were then immunoprecipitated for Vps33p, subjected to SDS-PAGE, and subjected to autoradiography or PhosphorImager analysis. Ten percent of every reaction was not immunoprecipitated and was analyzed for the total proteins that were photoaffinity-labeled. Important control reactions included using creatine phosphokinase as a positive control, irradiating the 8-azido-[ a - 32 P]ATP before the addition of the yeast extract, and omitting the irradiation step. Photoaffinity labeling of creatine kinase with 8-azido-[ a - 32 P]ATP was consistently competed greater than 80% with 2 m M ATP. We never observed photoaffinity labeling of Vps33p when the 8-azido-[ a - 32 P]ATP was photolyzed before the addition of extracts, indicating that secondary reactants were not produced. Sim- ilarly, we never observed labeling of Vps33p when we did not photoac-tivate the 8-azido-[ a - 32 P]ATP, indicating that the azido group on carbon 8 of the adenine ring was the only reactive moiety. For competition experiments, unlabeled nucleotides were generally added 2–10 min before the addition of the 8-azido-[ a - 32 P]ATP. When augmenting the reactions with unlabeled 8-azido-ATP, a mixture of the cold and radio- active nucleotides was made, and the two were added together.

P]ATP, could bind directly to Vps33p with low affinity. Interestingly, excess unlabeled ATP could enhance photoaffinity labeling of 8-azido-[␣-32 P]ATP to Vps33p, suggesting cooperative binding, which was not observed with excess GTP. Importantly, we did not detect significant photolabeling after deleting amino acid regions in Vps33p that show similarity to ATP interaction motifs. We visualized these events in living yeast cells after fusing the jellyfish green fluorescent protein (GFP) to the C terminus of full-length Vps33p. In metabolically active cells, the fully functional Vps33p-GFP fusion protein appeared to stain throughout the cytoplasm with one or two very bright fluorescent spots near the vacuole. After depleting cellular ATP, Vps33p-GFP appeared to localize with a punctate morphology, which was also reversible upon restoring cells with glucose. Overall, these data support a model where Vps33p cycles between soluble and particulate forms in an ATP-dependent manner, which may facilitate the specificity of transport vesicle docking or targeting to the yeast lysosome/vacuole.
The compartmentalized nature of eukaryotic cells demands high fidelity mechanisms to ensure maintenance of each organelle's unique content. Organelles of the secretory and endocytic system rely on vesicle-mediated transport pathways not only to move protein cargo throughout the cell but also to preserve their specific protein and lipid compositions.
The SNARE hypothesis describes fundamental molecular mechanisms that give rise to the specificity of vesicle-mediated protein transport in eukaryotic cells (1). The basic tenet of this notion is that interactions between integral membrane proteins in transport vesicles (v-SNAREs) and in their cognate target organelles (t-SNAREs) direct the specificity of all intercompartmental events (2). Although substantial evidence exists to support the existence of many v-and t-SNARE proteins, precisely how they function in vesicle transport specificity is continually being revised as more details are uncovered for different systems (3,4). However, a fact remaining clear and central to the SNARE hypothesis is that multimeric interactions of v-and t-SNAREs with cytosolic proteins play an essential role in the precision of vesicle-mediated transport (5,6).
The robust genetics of the yeast, Saccharomyces cerevisiae, has identified four genes whose cytosolic protein products are believed to help specify vesicle docking and/or fusion. Fittingly, one of these, SEC1, not only was the first secretory gene identified through mutant isolation (7) but was also the namesake of this family of proteins that are ubiquitously expressed in eukaryotic organisms (for reviews see Refs. 8 -10). The Sec1 group of proteins in yeast consists of Sec1p (11), Vps45p (12,13), Sly1p (14), and Vps33p (15,16).
The VPS33 gene was identified in a selection for yeast mutants with defects in vacuolar protein sorting (17). Loss of VPS33 function results in three prominent phenotypes; 1) temperature-sensitive growth (restrictive temperature is 38°C), 2) severe missorting of both soluble and membrane vacuolar proteins, and 3) abnormal vacuole morphology. Although the vacuole in vps33⌬ mutants is very fragmented, it is not as severe as the other class C vps mutants, end1/vps11, vps18, and vps16 (18). The VPS33 gene product, Vps33p, is 691 amino acids long, mostly hydrophilic, and cytoplasmically localized. The primary sequence of Vps33p contains two regions that are similar to the type A and B (also called Walker A and B) nucleotide consensus sequences (15), suggesting that it may bind or hydrolyze ATP, which has never been established.
In this study, we have examined ATP binding in the possible function of Vps33p. Our data indicate that Vps33p can indeed bind ATP and that deletion of a region that includes the type B consensus pattern results in both the loss of ATP binding and a nonfunctional protein. Further, Vps33p reversibly changed its localization from the cytosol to a particulate fraction when ATP was depleted in yeast cells, which was also observed for the NSF 1 homologue, Sec18p.

Plasmid Construction
A series of green fluorescent protein (GFP) fusion vectors were constructed by inserting a BamHI-BamHI fragment coding for a "super bright" GFP variant, M2 GFP (20) from pCSK-100 (21), kindly provided by Xiaolan Ma and William Margolin (University of Texas Medical School, Houston, TX). into the unique BamHI site of both pRS416 and pRS426 (22,23). This resulted in four vectors with two orientations of GFP having several multiple cloning sites. These vectors are named pBG416-GFP-SK, pBG416-GFP-KS (cen/ARS), pBG426-GFP-SK, and pBG426-GFP-KS (2 origin). Two plasmids, both containing a 3.7kilobase pair SalI-XbaI DNA fragment, were used for expression of the wild-type VPS33 gene. The first plasmid uses pRS416 (cen/ARS) as vector, pJWY33-39; the second uses pRS426 (2 origin) as vector, pBG33-126. The plasmid, pPRP33-100, is a pBluescript KS vector (Stratagene, La Jolla, CA) with the 3.7-kilobase pair SalI-XbaI VPS33 DNA fragment. To translate the VPS33 gene in vitro, 1600 base pairs were removed between the SalI site and the ATG codon in pPRP33-100, generating pPRP33-101. This placed the VPS33 gene under control of the T3 RNA polymerase promoter.
For the remaining constructs, site-directed mutagenesis (24) was used to add restriction sites or delete nucleotides. An AatII site was put at Ϫ9 base pairs from the VPS33 ATG codon in pJWY33-39 using the oligonucleotide 5Ј-CTATTCATCGTGACGTCATTTGATAAAGTTGG-3Ј to generate pBG33-116. To delete 57 nucleotides encoding amino acids 661-679 in VPS33 (the type B-like nucleotide interaction region), an oligonucleotide (5Ј-TATAGAGTTCATGATCCTTGTGCCATTGATCA-ACAAATGCTTCATTATAGCTATTTCACCC-3Ј) was used to loop out the corresponding DNA segment in pJWY33-39. The subsequent VPS33 deletion mutant was subcloned (SalI-XbaI) into pRS426 to generate pCT33-202. Oligonucleotides were used to place BamHI sites in VPS33, in frame to GFP, at amino acids 423 (5Ј-CAACTCGAGTAT-AGGATCCTCAATGTCGTT-3Ј) and 691 (5Ј-CTATCATATAATGGATC-CGATATAGAGTTC-3Ј) in pJWY33-39. The resulting plasmids, pPRP33-302 (BamHI at position 423), and pPRP33-303 (BamHI at position 691 were used to construct pPRP33-304, which has both BamHI sites by replacing a 1-kilobase pair XhoI-XbaI fragment in pPRP33-302 with the corresponding XhoI-XbaI fragment from pPRP33-303. To put the M2 GFP mutant under control of the VPS33 promoter, pBG33-116 was cut with AatII, blunted, and then cut with SalI and ligated into pBG426-GFP-KS cut with SalI and SmaI to generate pBG33pr-2GFP. To make a GFP fusion at the C terminus of VPS33, a BamHI-BamHI M2 GFP fragment was ligated into pPRP33-303 at the engineered BamHI site (pBG33691-GFP). To express this from a 2 plasmid, the SphI and SacI fragment in pBG33-126 was replaced with the SphI and SacI fragment from pBG33691-GFP, making pBG33691-2GFP. A truncated VPS33-GFP fusion was made by cutting pPRP33-304 with BamHI and ligating a BamHI-BamHI M2 GFP fragment, creating pBG33423-GFP. For expression of this truncated GFP fusion, pBG423-GFP was cut with SalI and SacI and inserted into pRS426 to make pBG423-2GFP. To make a Kex2-GFP fusion, a 4-kilobase pair BamHI-BamHI fragment from pYcP-KX2 (25) was ligated into the BamHI site of pRS416, creating pBG416-KX. The KEX2 gene was amplified from pBG416-KX via polymerase chain reaction with SpeI/SacI ends, cut with SpeI/SacI, and ligated into pBG426-GFP-SK, making pBGKex2-2GFP. This contains GFP at the full-length C-terminal tail of Kex2p. A plasmid encoding an alkaline phosphatase-GFP fusion (Pho8p-GFP) was kindly provided by Greg Odorizzi and Scott Emr (University of California, San Diego, La Jolla, CA).

Radiolabeling, Immunoprecipitations, and Permeabilized Cell Fractionation
The preparation and radiolabeling of yeast spheroplasts with Tran 35 S-label (ICN Radiochemicals, Costa Mesa, CA) used standard methods previously described (26). The preparation of permeabilized yeast spheroplasts was also via a standardized protocol previously described (27). Immunoprecipitations for Vps33p were all treated with IgGsorb (The Enzyme Center, Waltham, MA) for Ն 20 min before the addition of antiserum (15). Extracts were rocked overnight at 4°C and then treated with protein A-Sepharose (Amersham Pharmacia Biotech). The protein A-Sepharose pellets were washed sequentially with two 1-ml portions of 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, and 0.5% Tween 20 (Tween 20 buffer), two 1-ml portions of Tween 20 buffer containing 2 M urea, and two 1-ml portions of 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 1 mM EDTA. After drying, the immunoprecipitates were boiled for 4 min in standard Laemmli sample buffer (2ϫ concentration), and SDS-PAGE was performed as described (28). Most gels were dried and subjected to autoradiography using Kodak BioMax film. In some cases, the gels were subjected to phosphor imaging analysis using a STORM PhosphorImager (Molecular Dynamics, Sunnyvale, CA).

Fluorescence Microscopy
Cells expressing various GFP fusion constructs were grown in minimal proline medium with 0.5% yeast extract to middle or late log phase. They were concentrated to ϳ20 A 600 units/ml and examined in the same medium. For energy depletion experiments, the cells were washed two or three times in minimal proline medium without glucose and then resuspended in the same medium containing 10 mM NaN 3 and 10 mM NaF for periods up to 2 h. To reverse the effects of energy depletion, the cells were washed two or three times in minimal proline medium containing glucose and then incubated for 60 -90 min at 30°C. All micrographs were digitally captured on a Zeiss Axioskop epifluorescence microscope using an Optronics (Goleta, CA) DEI-750 CCD color camera with Adobe Premiere software (Adobe Systems, Mountain View, CA), a TARGA 2000 video board (Truevision, Inc., Santa Clara, CA), and a PowerPC Macintosh 9500 (Apple Computer, Cupertino, CA). All images were edited with Adobe Photoshop software (Adobe Systems).
Photoaffinity Reactions-Reactions were typically 12.5-25 l in total volume and consisted of ϳ125 g of membrane extract, 20 -40 M 8-azido-[␣-32 P]ATP (ICN Radiochemicals, Costa Mesa, CA, or RPI Corp., Mt. Prospect, IL). All reactions were performed in a 96-well microtiter dish. In general, the reactions were set up and preincubated for 2-10 min prior to UV irradiation. The microtiter dish was placed ϳ2 cm below a hand-held 254-nm UV lamp (UVP, Upland, CA, model UVGL-25), and the reactions were irradiated once for 1 min, followed with a 1-min interval of no irradiation and then a second irradiation for 1 min. In some experiments, the filter was removed from the UV light source to increase the radiation. The reactions were stopped by adding G6PDH, glucose-6-phosphate dehydrogenase; CPY, carboxypeptidase Y; NSF, N-ethylmaleimide-sensitive fusion protein; PAGE, polyacrylamide gel electrophoresis. an equal volume of 4% SDS and 4 mM dithiothreitol. The samples were transferred to a 1.5-ml tube, and the wells were rinsed with 2% SDS and 2 mM dithiothreitol and pooled with the first portion. The samples were heated (65-100°C) for 2-5 min. The reactions were then immunoprecipitated for Vps33p, subjected to SDS-PAGE, and subjected to autoradiography or PhosphorImager analysis. Ten percent of every reaction was not immunoprecipitated and was analyzed for the total proteins that were photoaffinity-labeled. Important control reactions included using creatine phosphokinase as a positive control, irradiating the 8-azido-[␣-32 P]ATP before the addition of the yeast extract, and omitting the irradiation step. Photoaffinity labeling of creatine kinase with 8-azido-[␣-32 P]ATP was consistently competed greater than 80% with 2 mM ATP. We never observed photoaffinity labeling of Vps33p when the 8-azido-[␣-32 P]ATP was photolyzed before the addition of extracts, indicating that secondary reactants were not produced. Similarly, we never observed labeling of Vps33p when we did not photoactivate the 8-azido-[␣-32 P]ATP, indicating that the azido group on carbon 8 of the adenine ring was the only reactive moiety. For competition experiments, unlabeled nucleotides were generally added 2-10 min before the addition of the 8-azido-[␣-32 P]ATP. When augmenting the reactions with unlabeled 8-azido-ATP, a mixture of the cold and radioactive nucleotides was made, and the two were added together.

Vps33p
Fractionates with Permeabilized Yeast Cell Membranes-When the VPS33 gene was first cloned, its protein product, Vps33p, was demonstrated to be a soluble cytoplasmic protein through subcellular fractionation (15). When examining the biochemical function of Vps33p, we used an in vitro reconstitution assay that measures intercompartmental transport of procarboxypeptidase Y to the yeast vacuole in permeabilized cells (27). The vps33-4 allele has a single L646P change and retains partial function for vacuolar protein sorting (15,29,30). Spheroplasts from a vps33-4 mutant strain were able to mature 35-40% of carboxypeptidase Y (CPY) in vivo, while the majority was secreted as the Golgi-modified p2 precursor (data not shown). However, even with this capacity for CPY maturation, vps33-4-permeabilized cell membranes were unable to transport p2CPY in vitro when incubated with ATP and a cytosolic extract made from a VPS33 yeast strain (data not shown). In contrast, 40% maturation of p2CPY was achieved in VPS33-permeabilized cell membranes under these same conditions (data not shown). Significant maturation of p2CPY (32%) was still observed if VPS33-permeabilized cell membranes were incubated with a cytosolic extract made from a vps33⌬ strain (data not shown). From these results, measuring protein transport to the yeast vacuole, we concluded that the functional fraction of Vps33p was associated with permeabilized cell membranes and not the cytosol. Thus, we examined the fractionation of Vps33p more carefully in wild-type permeabilized cells to determine if any of the protein was associated with a particulate fraction.
After a 5-min pulse with Tran 35 S-label, ϳ80% of Vps33p cofractionated with the cytosolic marker protein, glucose-6phosphate dehydrogenase (G6PDH, Fig. 1A lanes 1 and 2). However, during a 60-min chase, Vps33p fractionated away from the cytosolic supernatant fraction and was progressively recovered more in the permeabilized cell membrane pellet fraction (Fig. 1A, lanes 2-14). Although this separation into a supernatant and pellet was subjected to just 13,000 ϫ g, the Vps33p found in the supernatant continued to remain soluble after sedimentation at Ͼ125,000 ϫ g, suggesting that it was indeed soluble. Thus, at steady state, as much as 40% of Vps33p was associated with permeabilized cell membranes in excess of the cytosolic marker G6PDH in this experiment. Subsequent analyses indicated that a more consistent value for the particulate fraction of Vps33p at steady state was 15-20% (see Figs. 2 and 6). The apparent half-time for the association of Vps33p with permeabilized cell membranes was 25-30 min (Fig. 1A).

Loss of Cellular Energy Increases the Rate and Amount of
Vps33p Insolublility-We suspected that cellular energy played a role in the association/disassociation of Vps33p with FIG. 1. Newly synthesized Vps33p becomes insoluble in a timeand energy-dependent manner. A, spheroplasts from a wild-type yeast strain (SEY6210) were pulse-labeled with Tran 35 S-label for 5 min and chased with unlabeled Met and Cys for the indicated times. At each time point, the cells were treated under conditions that permeabilize the plasma membrane (freeze/thaw and four hypotonic washes (26) to generate a cytosolic fraction (S) and permeabilized cell pellet (P). Each fraction was immunoprecipitated for Vps33p and the cytosolic marker G6PDH. The immunoprecipitates were subjected to SDS-PAGE and autoradiography. B, spheroplasts from a wild-type yeast strain (SEY6210) were pulse-labeled with Tran 35 S-label for 15 min and chased with unlabeled Met and Cys plus 10 mM NaN 3 and 10 mM NaF for the indicated times. At each time point, the cells were treated as described for A.
membranes. To examine this possibility, cells were treated with agents that deplete cellular ATP. Sodium azide and sodium fluoride inhibit mitochondrial respiration (31) and glycolysis (32), respectively, and thus act to completely stop the cell from maintaining homeostatic ATP concentrations. In yeast, NaN 3 and NaF are the most potent compounds for total and rapid depletion of intracellular ATP (33).
Depletion of intracellular ATP with NaN 3 and NaF had a pronounced affect on the fractionation behavior of Vps33p. After a 15-min pulse with Tran 35 S-label, ϳ85% of the newly synthesized Vps33p cofractionated with the cytosol (Fig. 1B, lanes 1 and 2) as observed previously (Fig. 1A, lanes 1 and 2). However, if NaN 3 and NaF (10 mM each) were added during the chase, Vps33p rapidly and completely fractionated with permeabilized cell membranes (Fig. 1B, lanes 3-12). The apparent half-time for this association was 3-4 min, which was 8 -9-fold faster than the rate observed in the absence of energy poisons in vivo (Fig. 1B). The effect of energy depletion was most likely specific to properties associated with Vps33p, because the cytosolic enzyme, G6PDH, remained soluble throughout the course of NaN 3 /NaF treatment. If energy poisons were withheld from the chase, nearly 90% of the newly synthesized Vps33p remained soluble and cofractionated with G6PDH ( Fig.  1B, lanes 13 and 14).
We tested whether the action of NaN 3 and NaF was reversible and extended this analysis to two other proteins, Sec18p and Vps45p. The NSF homolog, Sec18p, was examined because it exists in soluble and particulate forms (34) with putative ATP binding activity like mammalian NSF (35,36). The Sec1 homolog, Vps45p, also exists in soluble and particulate forms (12) and is 22% identical to Vps33p (8). However, unlike NSF/ Sec18p, Vps45p shows no sequence similarity to proteins that bind or hydrolyze ATP (12,13). If spheroplasts were fractionated after pulse/chase radiolabeling in glucose medium, Ͼ80% of Vps33p and Sec18p and ϳ55% of Vps45p cofractionated with the cytosolic marker G6PDH after 30 min of chase ( Fig. 2A,  lanes 1 and 2). In contrast, treatment with NaN 3 and NaF for 30 min caused Ͼ85% of both Vps33p and Sec18p to fractionate with permeabilized cell membranes, but the amount of Vps45p that was membrane-associated did not increase significantly ( Fig. 2A, lanes 3 and 4). After the cells were washed several times in medium to remove the NaN 3 and NaF, replenished with fresh glucose, and incubated for another 30 min, Ͼ90% of Vps33p and Ͼ70% of Sec18p fractionated again with the cytosol ( Fig. 2A, lanes 5 and 6). This suggested that the effects of energy depletion were reversible. Although not shown, the fractionation behavior of Vps33p was the same when cycloheximide (100 g/ml) was added to the cells during energy regeneration (after NaN 3 and NaF treatment), indicating that new protein synthesis was not required to displace the membraneassociated pool of Vps33p.
We also examined the ability of cells to reverse intercompartmental protein transport to the vacuole after energy depletion. After a 15-min pulse with [ 35 S]Met/Cys, ϳ18% of CPY was present as the mature vacuolar form, while the remaining ϳ85% represented the p1 and p2 precursor forms (Fig. 2B). Nearly complete maturation of CPY occurred in the presence of glucose but was completely blocked when NaN 3 and NaF were added during a 30-min chase period (Fig. 2B). If the energy poisons were washed free of the cells, Ͼ95% maturation of CPY took place during a further 30-min chase in glucose (Fig. 2B). Overall, these results suggest that yeast spheroplasts can withstand the deleterious effects of extreme energy depletion, which argued that the insolubility of Vps33p and Sec18p were not due to irreversible toxicity.
Vps33p Aggregates after Energy Depletion but Can Be Dissociated with ATP and Cytosol-To examine in more detail the biochemical nature of the Vps33p particulate association, various treatments were used to solubilize the protein. After 30 min in NaN 3 and NaF, permeabilized cell membranes were obtained from yeast spheroplasts and treated with KCl (1 M), Triton X-100 (1%), urea (2 M), and sodium carbonate, pH 11 (0.1 M). Only sodium carbonate was effective at releasing a signifi-FIG. 2. The effects of energy depletion are reversible and extend to the NSF homologue, Sec18p, but not to the SEC1 homologue, Vps45p. A, spheroplasts from a wild-type yeast strain (SEY6210) were pulse-labeled with Tran 35 S-label for 15 min and chased for 30 min. They were then chased with and without 10 mM NaN 3 and 10 mM NaF for 30 min, as indicated. After 30 min of energy depletion, an aliquot of cells was washed several times to remove the energy poisons and incubated for 30 min in fresh glucose medium (lanes 7 and 8). At each time point, the cells were treated under conditions that permeabilize the plasma membrane (freeze/thaw and four hypotonic washes) to generate a cytosolic fraction (S) and permeabilized cell pellet (P). Each fraction was immunoprecipitated for Vps33p and the cytosolic marker G6PDH and then sequentially for Sec18p and Vps45p. Then fractions were subjected to SDS-PAGE and autoradiography. B, spheroplasts from a wild-type yeast strain (SEY6210) were radiolabeled and fractionated as in A but immunoprecipitated for CPY. After SDS-PAGE and autoradiography, each sample was quantitated for the amount of mature CPY (mCPY), which indicated the capability for energy-dependent transport through the early secretory pathway. cant amount (33%) of Vps33p from the membranes (Fig. 3A,  lanes 9 and 10), while the other reagents released less than 10% (Fig. 3A, lanes 3-8). This result suggested the interaction between Vps33p and permeabilized cell membranes after energy depletion was strong enough to resist disruption with high salt, nonionic detergent, or denaturing reagents. One possibility to explain this behavior is protein aggregation after energy depletion.
Since restoration of cellular energy led to solubilization of Vps33p in vivo (Fig. 1A), we studied the role of ATP in dissociation in vitro. After energy depletion, permeabilized cells were incubated in physiological buffer for 60 min at 25°C, and only 2% of Vps33p was detected in the supernatant after sedimentation of the membranes (Fig. 3B, lanes 1 and 2). When ATP (1 mM plus a regeneration mix) was included in the incubation, ϳ7% of Vps33p was detected in the supernatant after sedimentation of the membranes (Fig. 3B, lanes 3 and 4). Incubation with a cytosolic protein extract (5 mg/ml) released only ϳ3% of the particulate Vps33p (Fig. 3B, lanes 5 and 6). However, adding back both ATP and a cytosolic protein extract resulted in the release of 22% of the bound Vps33p (Fig. 3B,  lanes 7 and 8). This was a 3-fold increase in dissociation over incubation with ATP alone and a 7-fold increase over incubation with cytosol alone. Moreover, solubilization of Vps33p with ATP and cytosol was much more efficient than treatment with high salt, a nonionic detergent, or a denaturing agent (Fig. 3A). This suggests that activity of a cytosolic protein factor(s) was required to solubilize Vps33p from permeabilized cell membranes in an ATP-dependent fashion.
We further examined the characteristics of the insoluble Vps33p fraction with differential and sucrose gradient centrifugation before and after energy depletion. Permeabilized cell membranes were extracted under iso-osmotic conditions in a low ionic strength buffer (0.8 M sorbitol, 10 mM triethanolamine-acetic acid, pH 7.8, and 1 mM EDTA), which dissociates the membranes into crude organelles (26). The relative distribution of Vps33p during differential centrifugation after extraction of permeabilized cell membranes was very similar between energy-rich and energy-depleted cells (Fig. 4A). Both energy conditions showed ϳ30% of Vps33p in the low speed P2 pellet and ϳ10% in the high speed P3 pellet (Fig. 4A, lanes 4  and 6 and lanes 10 and 12). However, a significant difference was observed when these pellet fractions were applied to the top of a sucrose gradient and subjected to equilibrium centrifugation. In pellets from cells labeled in medium free of energy poisons, Vps33p settled to a major peak at fraction 8 (ϳ33% sucrose) and a minor peak at fraction 15 (ϳ46% sucrose, Fig.  4B). In contrast, Vps33p shifted predominantly to the bottom of the gradient (ϳ47% sucrose) in pellets from cells labeled in medium with NaN 3 and NaF (Fig. 4B). Along with the inability to solubilize Vps33p with detergent (Fig. 3A), these result suggests that energy deprivation caused aggregation of the protein to an insoluble form.
Vps33p Can Be Labeled with Photoactive ATP-Several possibilities could explain the effect of energy starvation and rapid loss of Vps33p localization from the cytosol to a particulate fraction. One prospect could be that Vps33p is a reversible substrate for a modifying enzyme like a protein kinase or phosphatase. The lack of ATP in the cell may trap the phosphorylation status of Vps33p and this could result in its aggregation. Indeed, Vps33p is a phosphoprotein in vivo, but we have not observed a difference in its fractionation properties dependent on phosphorylation (data not shown). More importantly, treatment of cells with NaN 3 and NaF did not change the phosphorylation status of Vps33p (data not shown). A simpler explanation for the energy-dependent solubility might be that Vps33p directly binds ATP, and this binding is necessary to keep the protein cytosolic.
Covalent cross-linking to a photosensitive ATP analog, 8-azido-ATP was examined to test the simplest model for localization of Vps33p. A high speed membrane fraction was made from a yeast strain deleted for the chromosomal copy of VPS33 but expressing the wild-type gene from a multicopy plasmid (pBG33-126). To demonstrate specificity of the 8-azido-ATP ligand, we examined all of the proteins in the extract. Over a dozen abundant, well resolved polypeptides existed in the high speed pellet fraction (Fig. 5A). However, only three bands were detected after irradiation with 8-azido-[␣-32 P]ATP, indicating that this ligand reacted with a subset of the many proteins present (Fig. 5A). After immunoprecipitation of extracts for FIG. 3. Vps33p is highly insoluble after energy depletion. A, spheroplasts from a wild-type yeast strain (SEY6210) were pulse-labeled with Tran 35 S-label for 15 min. They were then chased with and without 10 mM NaN 3 and 10 mM NaF for 30 min, as indicated. The cells were treated under conditions that permeabilize the plasma membrane (freeze/thaw and four hypotonic washes) to generate a permeabilized cell membrane pellet. The membrane pellet was then divided into five equal portions and treated on ice for 15 min with buffer, 1 M KCl, 1% Triton X-100 (TX-100), 2 M urea, or 0.1 M NaCO 3 , as indicated. The membranes were collected via centrifugation, and the supernatant (S) and pellet (P) were immunoprecipitated for Vps33p and subjected to SDS-PAGE and autoradiography. B, the same conditions as described for A were used to generate radiolabeled permeabilized cell membranes. However, four equal portions were incubated at 25°C for 60 min with buffer alone, 1 mM ATP (plus regeneration components), a cytosolic protein extract (5 mg/ml), or both ATP and cytosol (5 mg/ml), as indicated. After incubation, the membranes were collected via centrifugation and the supernatant and pellet were immunoprecipitated for Vps33p and subjected to SDS-PAGE and autoradiography.
Vps33p, a radiolabeled band was detectable on SDS-PAGE (Fig. 5C, lane 2). This band was not observed if the vps33⌬ strain was transformed with a vector control (pRS426), confirming its identity as Vps33p (Fig. 5C, lane 1). Although not shown here, this band was also not detected if UV irradiation was withheld from the wild-type extracts, demonstrating that photoactivation was required for affinity labeling of Vps33p. As a specific positive control in our experiments, we used creatine kinase, since it is a known ATP-binding protein and has been successfully photolableled with 8-azido-ATP (37). As a specific negative control, we used CPY, the vacuolar peptidase that neither requires ATP for activity nor has sequence similarity to ATP-binding proteins (38). As expected, creatine kinase, but not CPY, was photoaffinity-labeled with 8-azido-ATP (Fig. 5B).
Two important criteria in affinity labeling with photoactive nucleotides are to demonstrate competitive binding and saturable binding. Surprisingly, ATP did not appear to compete the binding, but rather it enhanced the ability to photoaffinitylabel Vps33p with 8-azido-ATP. An approximately 7-and 20fold increase was observed with 1 and 5 mM unlabeled ATP, respectively (Fig. 5C, lanes 3 and 5). When the same reactions were performed with creatine kinase, Ͼ90% inhibition of pho-toaffinity labeleling was observed (data not shown). This apparent enhancement of photoaffinity labeling was specific to ATP, since 1 or 5 mM GTP had no effect on radiolabeling of Vps33p (Fig. 5C, lanes 4 and 6).
The stimulation of photoaffinity labeling of Vps33p with 8-azido-[␣-32 P]ATP by unlabeled ATP most likely reflected either an improvement in cross-linking efficiency or cooperative binding. One approach to improve photocross-linking efficiency is to increase the concentration of the azido nucleotide. For example, supplementing photoaffinity reactions with unlabeled 8-azido-ATP enhances cross-linking to oligoadenylate synthetase (39). A proportional increase in photoaffinity labeling of Vps33p with 8-azido-[␣-32 P]ATP was observed in the presence of unlabeled 8-azido-ATP (Fig. 5D). This increase in binding was nearly linear between 0.1 and 1.0 mM 8-azido-ATP and appeared to show half-maximal saturation at ϳ0.5 mM (Fig. 5D). This suggested that Vps33p exhibited a relatively low affinity for ATP binding under these conditions. Importantly, if the extract was first illuminated with UV light in the presence of 2.0 mM 8-azido-ATP followed by a second photoreaction with 8-azido-[␣-32 P]ATP, the labeling of Vps33p was decreased nearly 4-fold (Fig. 5D, lane 6). This suggested that low affinity ATP binding of Vps33p could be saturated. Creatine kinase did not show this behavior and was significantly inhibited with 1 mM unlabeled 8-azido-ATP (Fig. 5B). Creatine kinase displays half-maximal saturation of 8-azido ATP binding at 12 M (37), Ͼ40-fold higher affinity than what we observed in Vps33p (Fig.  5D). With this apparent affinity difference in mind, photolabeling of Vps33p was competed Ͼ90% with ATP (1 mM) when we performed the experiment in the presence of 1 mM 8-azido-ATP (Fig. 5E, lane 2). Using ADP as a competitor was ineffective until the concentration reached 10 mM (Fig. 5E, lane 9). This suggests that Vps33p has a greater affinity for ATP than ADP.
Deleting the Putative Nucleotide Hydrolysis Region from Vps33p Abolishes ATP Binding, Energy-dependent Cytosol Distribution, and Vacuolar Protein Sorting-We examined if the regions of Vps33p that show similarity to proteins that bind or hydrolyze ATP were involved in photoaffinity labeling with 8-azido-ATP. Using site-directed mutagenesis to change conserved amino acid residues within these motifs was avoided because Vps33p does not strictly conform to the known consensus sequence motifs. Moreover, changing aspartate 678 to asparagine (D678N) did not show defective phenotypes (data not shown). Therefore, we deleted 19 amino acids, residues 661-679, because this encompasses all of the region that shows strong similarity to proteins with demonstrated ATP binding or hydrolysis activity like NSF/Sec18p (36), Vps4p (40), Escherichia coli SecA (41), adenylate kinase (42), Hsp70 (43), and Hsp90 (44). A Vps33p mutant lacking these 19 amino acids showed an 18-fold decrease in binding to 8-azido-ATP, which was unchanged in the presence of excess ATP (Fig. 6A, lanes 3  and 4). This suggested that these 19 amino acids were directly or indirectly critical for forming an ATP binding site or were important in maintaining the stability of Vps33p (see below). Removing the 19-amino acid putative ATP hydrolysis region in Vps33p also abolished the ability of the protein to function in vacuolar protein sorting. Maturation of carboxypeptidase Y was not detectable even after a 90-min chase (Fig. 6B).
We further examined the characteristics of Vps33p deleted for amino acids 661-679 in more detail to determine its stability and behavior during energy depletion. After a 15-min pulse with Tran 35 S-label, greater than 75% of the mutant protein was recovered in the cytosolic supernatant fraction of permeabilized cells, which was very similar to wild-type Vps33p (Fig.  7A, lanes 1 and 2). However, after a 30-and 90-min chase, the majority of the deletion mutant protein was recovered in the FIG. 4. Subcellular fractionation of Vps33p. A, differential extraction and centrifugation. Spheroplasts from a wild-type yeast strain (SEY6210) were pulse-labeled and fractionated using the same conditions as in Fig. 3A to generate the S1 and P1. The P1 pellet was then further extracted three times with 0.8 M sorbitol, 10 mM triethanolamine-acetic acid, pH 7.8, and 1 mM EDTA and subjected to centrifugation at 16,000 ϫ g each time to generate the S2 and P2. The S2 supernatant was then subjected to centrifugation at 125,000 ϫ g to generate the S3 and P3. Each fraction was immunoprecipitated for Vps33p and the cytosolic marker G6PDH and subjected to SDS-PAGE and autoradiography. B, Equilibrium sucrose gradient centrifugation. The P2 and P3 pellets (from A) were resuspended, mixed together, and applied to the top of a discontinuous sucrose gradient (26). After centrifugation at 170,000 ϫ g for 15 h, fractions were taken from the top, immunoprecipitated for Vps33p, and subjected to SDS-PAGE and autoradiography.

Energy-dependent Localization of Vps33p
membrane pellet fraction of permeabilized cells, whereas wildtype Vps33p maintained its cytosolic distribution (Fig. 7, C and  D, lanes 3 and 4 and lanes 7 and 8). Moreover, the amount of mutant Vps33p in the pellet after 90 min of chase in glucose (80%) was very similar to the amount of wild-type Vps33p in the pellet after 30 min of energy depletion (83%; compare Fig.  7A, lanes 7 and 8 to Fig. 7B, lanes 5 and 6). This observation indicates that the 19-amino acid region of Vps33p, that shows strong similarity to ATP binding proteins and ATPases, was required for maintaining its predominantly cytosolic distribution. Additionally, this region may play a role in efficient local-ization of Vps33p to a membrane during rapid loss of cellular ATP. After a 30-min treatment with NaN 3 and NaF, twice as much of the deletion mutant protein was recovered in the cytosolic supernatant fraction of permeabilized cells compared with wild-type Vps33p (Fig. 7, A and B, lanes 5 and 6). Another observation from this analysis was that overexpression of Vps33p did not significantly change its distribution in the cytosol or membrane fractions. Using a multicopy plasmid with the endogenous VPS33 promoter gave 10 -15-fold greater levels of Vps33p in cells. Even after energy depletion, the ratio of cytosolic to membrane-associated Vps33p with increased pro-FIG. 5. Vps33p can be labeled with an 8-azido ATP. A, a crude membrane pellet (125,000 ϫ g) was made from a wild-type yeast strain with a multicopy plasmid encoding the VPS33 gene. Approximately 25 g of protein was subjected to SDS-PAGE before and after incubation on ice with ϳ20 M 8-azido-[␣-32 P]ATP (1-min irradiation). The unlabeled protein sample was stained with silver nitrate, and the labeled sample was subjected to PhosphorImager analysis, as indicated. The position of molecular weight standards is indicated with the mass in kDa. B, approximately 250 g of protein from the 125,000 ϫ g pellet or 5 g of purified creatine kinase was incubated on ice with ϳ20 M 8-azido-[␣-32 P]ATP with or without 1 mM unlabeled 8-azido-ATP, as indicated. After irradiation, both samples were subjected to SDS-PAGE and PhosphorImager analysis (the yeast extract was immunoprecipitated for Vps33p prior to the gel). As a negative control, ϳ65 g of a 125,000 ϫ g supernatant extract was reacted with ϳ20 M 8-azido-[␣-32 P]ATP (1-min irradiation) and immunoprecipitated for carboxypeptidase Y or probed on an immunoblot, as indicated. C, a yeast strain deleted for the chromosomal copy of VPS33 (vps33⌬) was transformed with a multicopy plasmid encoding the complete VPS33 gene or a vector control, as indicated. The cells were lysed by vigorous agitation with glass beads, and a 125,000 ϫ g pellet was obtained. Approximately 125 g of protein was incubated on ice with ϳ16 M 8-azido-[␣- 32  tein levels was the same as when the protein was examined at chromosomal levels, which is estimated to be as low as 0.005% of total cellular protein (15). Overall, these analyses suggested that deletion of the putative ATP hydrolysis region in Vps33p abolished ATP binding and intercompartmental transport to the vacuole and impaired cytosolic localization.
VPS33p Is Localized to Punctate Structures after Energy Depletion in Vivo-To examine the intracellular location of Vps33p, fusion proteins were constructed with the jellyfish green fluorescent protein. This approach was chosen over immunofluorescence because it allowed for a dynamic assessment of intracellular localization in living cells. Three GFP fusion constructs were made to the VPS33 gene all under control of the VPS33 promoter. One was made to the promoter region alone with no coding sequence (VPS33p r -GFP), a second was made to the first 423 amino acids (Vps33p 423 -GFP), and a third was made to the full length protein (Vps33p 691 -GFP). Both hybrid proteins had GFP at their C termini. Importantly, the Vps33p 691 -GFP construct complemented all phenotypes associated with loss of VPS33 function, including growth at 38°C, proper sorting of vacuolar proteins, and normal vacuole morphology. The truncated fusion, Vps33p 423 -GFP, did not complement any of these phenotypes when expressed in a yeast strain deleted for the entire VPS33 coding region. This suggested that a functionally important polypeptide tract of Vps33p was in the C-terminal third of the protein between amino acids 423 and 691.
The Vps33p 691 -GFP fusion protein was localized as a haze throughout the cytoplasm of cells in glucose medium (Fig. 8,  Vps33p 691 -GFP panel). This pattern was consistent with the fractionation experiments described earlier (cf. Figs. 1 and 2). In addition to the cytoplasmic haze, the Vps33p 691 -GFP fusion

Energy-dependent Localization of Vps33p
protein was also localized to one or two bright spots per cell in glucose, which were usually near the vacuole (Fig. 8,  Vps33p 691 -GFP panel). Significantly, these perivacuolar spots were not observed in cells expressing the truncated Vps33p 423 -GFP fusion protein or GFP alone (Fig. 8, Vps33p 423 -GFP and VPS33p r -GFP panels). The exclusive cytoplasmic localization of these two control constructs argued that the C-terminal third of Vps33p might contain information that caused the protein to associate with a specific subcellular organelle or structural complex. Furthermore, quantitation of the fluorescence in several cells with image analysis software (NIH Image) revealed that the cytoplasmic haze accounted for ϳ80 -FIG. 8. Full-length Vps33p fused to GFP displays reversible punctate morphology after energy depletion. The coding region for the jellyfish GFP was fused in-frame to the C terminus of the full-length coding region of Vps33p (Vps33p 691 -GFP) under expression from its endogenous promoter. As controls, GFP was fused to the first 423 amino acids (Vps33p 423 -GFP) and just to the VPS33 promoter VPS33 pr -GFP, as indicated. All gene constructs were subcloned into multicopy yeast expression vectors. The plasmid encoding Vps33p 691 -GFP was transformed into a yeast strain (BGY3300) with a complete chromosomal deletion of the VPS33 open reading frame. The plasmids encoding Vps33p 423 -GFP and VPS33 pr -GFP were transformed into an isogenic wild-type strain (SEY6210). Fusions of GFP to Kex2 (Kex2-GFP) and alkaline phosphatase (Pho8p-GFP) were also examined but in yeast strain TVY1 (SEY6210 ⌬pep4::LEU2). Cells were taken from exponentially growing cultures in selective medium (Glucose panels). They were then treated with NaN 3 and NaF (10 mM each) for 90 min (Glucose 3 NaN 3 /NaF panels). The energy poisons were then washed away, and the cells were treated with fresh glucose for 90 min (NaN 3 /NaF 3 Glucose panels). At each condition, 4 -6 fields of cells were digitally captured with a fluorescence microscope using an fluorescein isothiocyanate filter set. The images were then composed so as to show at least 10 cells per condition. The arrows in the Vps33p 691 -GFP panel point to punctate structures. Bar, 5 m. 85% of the total signal. The remaining 15-20% of the signal comprised the one or two fluorescent spots. These values correlate well with the fractionation data of Vps33p, where 80 -85% of the protein was cytosolic and 15-20% was particulate under steady state conditions. Energy depletion of cells with NaN 3 and NaF treatment changed the localization pattern of Vps33p 691 -GFP. The fluorescence was no longer a predominant cytoplasmic haze but instead consisted of 10 -20 punctate dots scattered throughout the cytoplasm (Fig. 8). This pattern was again compatible with the fractionation data where Vps33p was predominantly particulate after energy depletion (Figs. 1 and 2). Cells expressing the control constructs, Vps33p 423 -GFP and VPS33p r -GFP, did not show these numerous fluorescent spots during energy depletion and maintained their fluorescent haze in the cytoplasm (Fig. 8). Importantly, this behavior was completely reversible after washing away the NaN 3 /NaF and supplying the cells with fresh glucose. The punctate dots were no longer apparent, and the fluorescence was again predominantly cytoplasmic with the one or two perivacuolar spots per cell (Fig. 8).
We further examined the nature of the many tiny spots observed after energy depletion with the Vps33p 691 -GFP fusion protein. One possibility for their identity was the Golgi complex, since this organelle often has punctate morphology when viewed with fluorescence microscopy in yeast (45). We constructed a GFP fusion protein to the cytoplasmic C terminus of Kex2p, a marker for the late Golgi complex in yeast. In glucose medium, the Kex2p-GFP protein presented a pattern very reminiscent of immunofluorescence studies (46) with 5-10 punctate structures throughout the cytoplasm (Fig. 8). After energy depletion, the morphology of the Kex2p-GFP protein was distinct from that observed for Vps33p 691 -GFP. Rather than dispersing into numerous spots, the Kex2p-GFP appeared to cluster in one or two larger structures per cell (Fig. 8). This suggests that the identity of the small spots seen with the Vps33p 691 -GFP protein may not be the Golgi complex. Furthermore, the localization of an alkaline phosphatase-GFP (Pho8p-GFP) fusion protein (kindly provided by Greg Odorizzi and Scott Emr), a vacuole marker, suggested that the identities of energy-depleted Vps33p 691 -GFP spots were most likely not related to the vacuole. After energy depletion, the vacuoles in general did not fragment into smaller structures but instead enlarged and took on a swollen appearance, a common morphology when cells are starved for nutrients (33). DISCUSSION In these studies, we have presented data to provide evidence for a new biochemical function of a previously identified gene product. Our work establishes Vps33p as a low affinity ATP binding protein that shows energy-dependent cytosolic localization. This conclusion rests on several different types of experimental data. Depleting cellular ATP results in a rapid change of Vps33p from a cytosolic protein to one that fractionates as a particulate protein. Restoring cellular energy with glucose reverses this effect. In glucose medium, a full-length Vps33p-GFP fusion was distributed throughout the cytoplasm. After energy depletion, the GFP fluorescence was localized to punctate structures, which returned to the cytoplasm with fresh glucose. The particulate form of Vps33p after energy depletion required both ATP and a cytosolic extract for efficient release to a soluble form in vitro. The ATP analog, 8-azido-ATP, could bind directly to Vps33p in a saturable manner. Deletion of a 19-amino acid region in Vps33p, similar to many ATP binding proteins and ATPases, prevented photoaffinity labeling with 8-azido-ATP. The deletion mutant protein was also particulate even under energy rich conditions and did not have the ability to support vacuolar delivery of CPY. Overall, the data suggest a model where Vps33p may function in vesicle traffic by cycling between soluble and particulate forms dependent on ATP binding or hydrolysis.
The first clue that Vps33p may interact with ATP was its sequence similarity to proteins that bind or hydrolyze this nucleotide (15). The most common and easily identified sequences are the type A (putative triphosphate-binding), and type B (adenine-binding), consensus regions (47,48). Frequent renditions of the type A and B sequences are (A/G)XXXXGK-(T/S) and (H/R/K)X (5)(6)(7)(8) -HyXHyHy(D/E) (where Hy is hydrophobic), respectively (49). Although Vps33p does not precisely match the type A sequence (15), the putative type B sequence in Vps33p is 666 KKGINKRFIIIAD 678 and matches the consensus pattern more closely except for an alanine residue. The 19-amino acid deletion (residues 661-679) encompassing the type B-like sequence, prevented Vps33p from efficiently binding to 8-azido-ATP and resulted in a nonfunctional protein.
This suggests that ATP binding may be necessary for Vps33p function. Furthermore, the deletion of the type B region causes Vps33p to aggregate in ATP-rich conditions, arguing that ATP binding or hydrolysis is required for solubility. Reconstituting the release of Vps33p from membranes with ATP and cytosol in vitro supports this conclusion and suggests that other protein factors contribute to cytosolic localization.
The binding characteristics of Vps33p indicate that it has a low affinity for ATP under our experimental conditions. From the data, we estimate that half-maximal saturation occurs at ϳ0.5 mM, which is well within the physiological range of 2-5 mM (50). With this low affinity, the binding site would most likely not be constitutively occupied, and this affinity suggests that binding to ATP may serve a regulatory role in Vps33p function. Multimeric interactions with other proteins or itself may give rise to a conformation of Vps33p with higher affinity for ATP. In support of this possibility, photoaffinity labeling of Vps33p was significantly more efficient in particulate extracts compared with cytosolic extracts. These binding properties of Vps33p imply that it does not interact with ATP in the same manner as proteins like the AAA-type ATPases. For instance, Cdc48p (51), NSF/Sec18p (34,52), and Vps4p (40) all have 50 -60 amino acids between the conserved lysine and aspartate residues of their type A and B consensus sequences. The 183residue stretch separating lysine 494 and aspartate 678 in Vps33p is considerably longer, suggesting a different structure for the functional nucleotide binding fold. Although deletion of 19 residues encompassing the type A-like sequence also reduced binding of 8-azido-ATP to Vps33p, the resulting protein had nearly wild-type function. 2 The ATP-specific enhancement in photoaffinity labeling of Vps33p with 8-azido-ATP insinuates a component of positive cooperativity in its nucleotide binding. This is best illustrated with the data demonstrating a 90% decrease in 8-azido-ATP with 1 mM ATP, a slight increase in 8-azido-ATP binding with 2.5 mM and 5.0 mM ATP, and an 80% decrease with 10 mM ATP (Fig. 5E). Although one possibility for this property would be differences in binding affinity between ATP and 8-azido-ATP, another explanation is positive cooperativity. For example, the activity of cAMP phosphodiesterase is stimulated over 10-fold with 1-10 M cGMP but inhibited nearly 10-fold with 100 -1000 M cGMP (53). Since we are measuring only nucleotide binding and not enzymatic activity in our assays, a cooperative interaction is difficult to establish. At present, we do not know if Vps33p is an ATPase, and this issue is best reconciled with purified protein. The ATP binding characteristics of Vps33p are very atypical, especially if viewed from the standpoint of a single protein or nucleotide binding site. The extracts that we used for ATP binding were very crude and most likely contain other proteins that could interact with Vps33p. Positive cooperativity in ATP binding may happen if a second site on another protein needed to be saturated or occupied causing a conformational change in Vps33p to form its ATP binding site. This type of cooperative binding occurs in the rubisco activase complex (54,55), CTP synthetase (56), and E. coli GroEL-GroES chaperone complex (57,58). This idea gives rise to an intriguing possibility that the type B-like sequence in Vps33p interacts with another protein in such a way as to form an ATP binding site. The heteromultimeric formation of a specific nucleotide binding pocket would provide unique specificity to a vesicle docking or transport reaction.
Although the precise molecular details of ATP binding to Vps33p are unknown, the reliance on sequence specificity conventions or "consensus" sequences should not be overestimated in establishing these features. The case of Hsp90 is an excellent example where it was concluded not to bind ATP after using ATP-agarose chromatography, 8-azido-ATP photoaffinity labeling, fluorescence nucleotide analogs, and sequence analyses (49). However, the recent structure of an Hsp90 N-terminal domain co-crystallized with ATP and/or ADP refutes these biochemical results (44). Moreover, the first 220 amino acids of Hsp90 used for the crystallization do not contain a type A or B consensus sequence. Likewise, synapsins I and II were recently shown to bind ATP, and both of these proteins also lack the typical consensus sequences (59). This indicates that many different ATP binding folds exist and that primary amino acid sequence data are often over-or misinterpreted (60). The interaction of Vps33p with ATP does not conform to characteristics of a conventional nucleotide binding site. Molecular details of the interaction may be resolved when putative protein partners of Vps33p are identified.
The energy-dependent localization of Vps33p most likely reflects an ATP-driven conformational change in the protein. Such alterations in protein structure are well documented in oligomeric proteins that bind or hydrolyze ATP such as heat shock chaperones (61) and the N-ethylmaleimide sensitive fusion protein, NSF (36,62). The oligomeric nature of Vps33p is unknown. The fractionation behavior of both Vps33p and the NSF homolog, Sec18p, but not Vps45p, most likely reflects an energy-driven cycle that continually localizes these proteins between soluble and particulate forms with homeostatic ATP concentrations.
The behavior of Vps33p in living yeast cells under ATP-rich and ATP-depleted conditions provides new observations for interpretation of its cellular location. As predicted from permeabilized cell preparations, 75-85% of Vps33p is cytosolic, with the remaining portion localizing to fluorescent spots, which appeared perivacuolar. This localization pattern is reminiscent of the prevacuolar compartment in yeast, which is aberrantly exaggerated in class E VPS mutants (63)(64)(65). However, our data with equilibrium sucrose density gradients suggest that the 15-20% of Vps33p that is particulate does not cofractionate with the prevacuolar compartment as defined by the presence of pro-CPY (26). Therefore, the perivacuolar fluorescent spots may be a novel organelle or subcellular structure. Recently, a comprehensive study has shown genetic and physical interactions among the four yeast VPS class C genes and gene products (30). The VPS18 and VPS11 gene products form a major hetero-oligomer with minor amounts of associated Vps16p and Vps33p. This complex fractionates as both a dense insoluble protein complex and as a peripheral component of the vacuole (30). The sedimentation characteristics of the Vps11p-Vps18p insoluble fraction are very similar to what we have observed with Vps33p after energy depletion. This suggests that ATP is required to dissociate Vps33p from the Vps11p-Vps18p complex.
The many SEC1 family members suggest that these proteins function at every vesicle-mediated step in eukaryotic cells (8,10). However, the precise molecular mechanism by which Sec1like proteins execute their role in vesicle docking/fusion is unknown. The most recent findings suggest that Sec1-like proteins play a negative regulatory role in SNARE complex interactions. The Drosophila Sec1 homologue, Rop1, can negatively regulate neurotransmitter release (66), and yeast Sly1p can prevent v-SNARE/t-SNARE interactions in endoplasmic reticulum to Golgi transport (67). On the other hand, a rat Sly1 homologue, rSly1p, can positively influence transport between the endoplasmic reticulum and Golgi (68), so it seems likely that Sec1-like proteins do not perform a common function at different vesicle-mediated steps along the secretory and endocytic pathways. From this perspective, Vps33p may display a unique activity not provided by any of the other SEC1 family members, since they are not predicted to bind or hydrolyze ATP.
Future studies on Vps33p will focus on establishing the details of ATP binding in the molecular mechanism of transport between the prevacuolar compartment and lysosome/vacuole. In particular, we will determine if Vps33p can hydrolyze ATP after purification of a recombinant protein. The possible multimeric nature of Vps33p will be resolved with gel filtration analysis and sedimentation through glycerol gradients. The recently described genetic interaction between VPS33 and the vacuolar t-SNARE, VAM3, suggests that their protein products may directly interact (29). We will test this with protein crosslinking to determine if Vam3p is associated with Vps33p after energy depletion. Examination of the full-length Vps33p-GFP fusion protein in a class E VPS mutant background stained with the vital endocytosis marker, FM 4 -64, should further clarify whether or not Vps33p localizes to the prevacuolar compartment. Reconstitution of intercompartmental transport to the vacuole will be the definitive system to dissect the molecular details of ATP binding and Vps33p function. We are poised to exploit this system with new temperature-sensitive VPS33 alleles that show a rapid block in vacuolar protein transport at elevated temperatures. 2 Biochemical complementation of vps33 mutants will aid in proving or disproving if ATP binding or hydrolysis functions to cycle Vps33p from the cytosol to a membrane or membrane complex, allowing vesicular transfer to the yeast vacuole.