Nonclassical Protein Sorting to the Yeast Vacuole*

Fungal cells spend most of their time in stationary phase (1, 2), yet most research has focused on cells in logarithmic conditions. In the wild, these cells frequently encounter periods of limiting nutrient conditions. Accordingly, recycling, as opposed to simply degradation, of macromolecules is a routine and critical feature of cell biology. It is important to explore the ways in which cells adapt to environmental changes, including starvation, to understand cellular physiology. The basic processes of intracellular transport, as well as many of the specific components, are conserved among yeast and higher eukaryotes, making studies in this experimentally tractable organism relevant to other systems.

specific process of lysosomal uptake has been described by Dice and colleagues (22), who have identified a lysosomal surface receptor that is involved in recognition of proteins bearing a pentapeptide (KFERQ) motif. This import mechanism requires members of the hsp70 family including an intralysosomal hsc73 (23). These data suggest the presence of protein translocation machinery in the lysosomal membrane, a property not generally ascribed to this organelle.

The Resident Hydrolase Aminopeptidase I Is Localized to the Vacuole Independent of the Secretory Pathway
In addition to the pathways mentioned above, other routes are used for protein delivery to the vacuole. One of the recently characterized mechanisms is the cytoplasm to vacuole targeting pathway used to deliver the resident hydrolase aminopeptidase I (API) to the vacuole (24). An examination of the biosynthesis of API revealed differences from the standard pattern seen with vacuolar proteins that transit through the secretory pathway. All of the characterized vacuolar hydrolases that transit through the ER and Golgi complex undergo proteolytic and/or glycosyl modifications during transit. Following the removal of cleavable signal sequences, these proteins receive a core glycosylation. Subsequent carbohydrate additions, upon delivery to and passage through the Golgi complex, are detected as an increase in molecular mass. Delivery to the vacuole is often accompanied by the removal of a propeptide segment. The half-time for vacuolar delivery through this route is 5-10 min for most of these proteins.
In contrast to the secretory pathway proteins, API is not glycosylated even though it has N-linked glycosylation sites (24). It lacks a signal sequence or consensus signal sequence cleavage site (25,26). In accordance with the lack of a means to enter the secretory pathway, the precursor form of API is not found within the endoplasmic reticulum or Golgi complex but rather within the cytosol (24). API contains a propeptide at the N terminus that is removed in the vacuole by a proteinase B-dependent reaction. The half-time for propeptide processing, and presumably vacuolar delivery, is 30 -40 min, substantially longer than that for proteins that utilize the secretory pathway. Similarly, the import of API is relatively insensitive to sec mutants, which were isolated based on defects in transit through the secretory pathway. In particular, Sec gene products that are specific to the early steps of the secretory pathway, ER to Golgi complex transport, are not required for API delivery to the vacuole (24). Many of the vps mutants that were isolated based on defects in vacuolar targeting of carboxypeptidase Y show essentially normal processing kinetics for API. Finally, unlike vacuolar proteins that traverse the secretory pathway, overexpression of API does not lead to secretion from the cell. These results indicate that API does not enter the vacuole through the secretory pathway but rather uses an alternate mechanism to attain its correct subcellular localization.

The Aminopeptidase I Propeptide Is Required
for Vacuolar Targeting Many vacuolar hydrolases are synthesized as zymogens containing a propeptide segment that keeps the enzyme inactive. The maintenance of a latent state may serve to protect the cell from the hydrolytic activity prior to the arrival of the enzyme in the vacuole. The propeptides of vacuolar hydrolases may also be involved in folding and/or targeting of the hydrolase (4). The propeptides of proteinase A and carboxypeptidase Y have been shown to contain vacuolar targeting determinants; however, no consensus sequences have been defined for yeast vacuolar proteins. The API propeptide is composed of two ␣-helices (27). The first of these forms an amphipathic structure; unlike mitochondrial targeting signals, the charged half is composed of both basic and acidic residues. Sitespecific mutagenesis was used to analyze the location of targeting information in the API propeptide (28,29). Small deletions anywhere within the first helix of the propeptide resulted in a complete block in targeting. Similar deletions within the second helix had no affect on targeting; in this case, the altered proteins were delivered to the vacuole with wild type kinetics. Deletion of the entire second helix, however, blocks API targeting suggesting some role for this part of the precursor protein in the import process.
A series of random mutants was also generated within the API propeptide (28). All mutations that caused precursor accumulation mapped within the first helix. The precursor API was shown to reside in the cytosol in a protease-sensitive form. This result confirmed that the propeptide mutations caused a defect in targeting and not simply processing of API. Hence, the first helix of the API propeptide contains information that is necessary for delivery to the vacuole. The targeting information in the first and second helices is also sufficient for vacuolar localization; this segment of API can target a passenger protein to the vacuole. Hybrid proteins containing the N-terminal helices of API fused to the green fluorescent protein, are localized to the vacuole. 2

Biochemical and Genetic Identification of Targeting Components
The API propeptide presumably interacts with subcellular sorting components such as a binding protein and/or receptor. To identify these sorting components, an in vitro system was established to reconstitute API import (30). The import reaction is temperature-dependent, having an optimum at approximately 30°C. In addition, import is inhibited both in vitro and in vivo at 14°C. Inhibition at 14°C suggests that import does not occur through a proteinaceous channel as in the ER, mitochondria, or chloroplasts; translocation in these organelles can occur at temperatures as low as 0°C, as long as protein unfolding is not limiting. A block in transport at 14°C suggests a vesicle-mediated event.
The in vitro import reaction is also energy-dependent (30). When import was carried out in the presence of non-hydrolyzable ATP or GTP analogs, or when ATP was depleted, processing of the API precursor was blocked. Inhibition of vacuolar ATPase activity with specific inhibitors or using vma mutants resulted in reduced processing of API. In these cases, however, the block may have been indirect and reflect decreased access to the precursor protein (see below).
The vacuolar localization of API was also analyzed through a classical genetic approach (31,32). Because the mature form of API is stable in the vacuole lumen, wild type cells accumulate this form of the protein. To identify mutants in the import pathway, mutagenized yeast cells were screened for precursor accumulation. Mutants were obtained that are termed cvt, for cytoplasm to vacuole targeting defective. These mutants accumulate the precursor form of API but (for most) do not affect vacuolar delivery of proteins that transit through the secretory pathway. In all but two cvt mutants, precursor API is located in the cytosol, indicating that these mutants are blocked in targeting and do not deliver the precursor protein to the vacuole.
To determine if these mutants are unique, complementation studies were carried out between cvt mutants and other mutant strains known to affect vacuolar protein delivery. The complementation studies revealed that some of the cvt mutants are probably allelic to certain vps mutants. This is expected because some of the vps mutants display major defects in vacuole morphology, in some cases lacking a detectable vacuole; these mutants do not present a proper target for API delivery. In addition, some of the components needed for API delivery may reach the vacuole through the secretory pathway.
A more substantial overlap is detected between the cvt mutants and two groups of mutants that were isolated based on defects in macroautophagy, the apg and aut mutants (32)(33)(34)(35). This overlap is surprising because there are substantial differences between the Cvt pathway and macroautophagy. For example, API import is kinetically slower than secretory pathway transit but is much faster than macroautophagy. Along these lines, macroautophagy also shows a lower yield of protein uptake. The Cvt pathway is biosynthetic, delivering a resident hydrolase to the vacuole. Accordingly, API import occurs during vegetative growth. In contrast, macroautophagy is a degradative pathway and, while occurring at a basal level in rich media, is induced under starvation conditions. Finally, in agreement with their respective roles in cellular physiology, Cvt import is specific for API and perhaps other hydrolases, whereas macroautophagy is nonspecific and is used to deliver bulk cytosol to the vacuole.

Aminopeptidase I Transits to the Vacuole and Imports as a Dodecamer
To understand the molecular basis for the overlap between the Cvt and macroautophagic pathways, the native state of API was examined during the import process. The mature hydrolase was reported to be a dodecamer in the vacuole (36 -38). To determine when oligomerization occurs, and specifically if the precursor is a dodecamer, a kinetic analysis was carried out (39). The results indicated that oligomerization into a dodecamer occurred rapidly, with a half-time of approximately 3 min. This is much shorter than the half-time of processing and indicates that oligomerization is not rate-limiting in the import process and occurs prior to vacuolar delivery. Oligomerization studies combined with subcellular fractionation showed that the oligomer forms in the cytosol; monomeric API can only be found in a soluble fraction whereas a low speed pellet fraction contains exclusively the dodecameric form. The transition from a soluble to sedimentable form may be indicative of binding to a target membrane and/or the formation of a pelletable complex.
A temperature-sensitive API propeptide mutant was utilized to follow the import of API into the vacuole following the initial oligomerization event. An alteration of lysine to arginine at position 12 of the propeptide, K12R, results in a temperature-sensitive targeting phenotype (28); at the nonpermissive temperature, precursor K12R API accumulates in the pelletable oligomeric form. Because the K12R phenotype is thermally reversible, it can be used to follow the stages of import subsequent to the initial oligomerization reaction and binding event. Analysis of the K12R mutant revealed that precursor API transits to and imports into the vacuole in the dodecameric form (39). This makes translocation through a proteinaceous channel unlikely because the oligomeric precursor is approximately 732 kDa in mass. This result, coupled with the in vitro and genetic studies, suggests that precursor API enters the vacuole through a vesicle-mediated process that is similar to macroautophagy.

Macroautophagic Protein Import Involves a Double Membrane Vesicle
The process of macroautophagy had been characterized in yeast primarily through morphological studies and more recently by molecular and classical genetic techniques (14,15,33). Macroautophagy begins with the formation of an enwrapping membrane that sequesters cytosol. Upon completion, this membrane forms a double membrane vesicle that is termed an autophagosome (AP). The origin of the autophagosomal membrane is not known. The fact that APs appear to be fairly uniform in size suggests that they may originate from a pre-existing organelle such as a specialized region of the ER. However, it is not clear that the necessary targeting components, needed for subsequent delivery to the vacuole, would reside in the membrane of this organelle. In addition, inhibition by cycloheximide (14) suggests that at least some aspects of macroautophagy require de novo protein synthesis.
Following targeting to the vacuole, the outer membrane of the AP fuses with the vacuole membrane. The machinery needed for targeting, docking, and fusion have not been well characterized but are likely to be similar to those involved in similar processes in other parts of the cell (40 -43). For example, the vacuolar t-SNARE Vam3p, which is required for homotypic vacuole fusion (44), is also needed for API delivery to the vacuole (45). Along these lines, preliminary evidence indicates a role for the rab protein Ypt7p and the SEC19 gene product, GDP dissociation inhibitor, in API import. 3 Fusion of the APs with the vacuole allows the release of the interior single membrane vesicle into the vacuole lumen. This vesicle, termed an autophagic body (AB), is eventually broken down by vacuolar hydrolases, allowing access to the lumenal contents.
Yeast strains deficient in vacuolar hydrolase activity accumulate ABs within the vacuole lumen (14). Breakdown of the AB is pHdependent (46). The elevated vacuolar pH in vma mutants stabilizes ABs. This may explain the apparent block in API processing in vma mutants.

Biochemical Studies Support a Macroautophagy-like
Mechanism for API Delivery Applying the morphological data on macroautophagy to API import allows specific predictions about the Cvt pathway. In particular, if API is imported by a process similar to macroautophagy, precursor API should reside in the cytosol in a double membrane vesicle analogous to an autophagosome. Similarly, the precursor protein should accumulate within single membrane vesicles, equivalent to ABs, in the vacuole lumen in mutants that are defective in vesicle breakdown. This model for API import was tested using mutants defective in various stages of the localization process.
The Vps18 protein is part of a complex that is required for transport to, or fusion with, the vacuole membrane (47). A temperature-sensitive vps18 mutant accumulates precursor API (48). Pulse-chase studies showed that precursor API in the vps18 temperature-sensitive mutant grown in rich medium went from a protease-sensitive to a protease-insensitive form, consistent with its enclosure within a membrane-bound compartment. Subcellular fractionation studies demonstrated that the precursor was not located within the vacuole, consistent with its sequestration in cytosolic vesicles.
The cvt17 mutant accumulates precursor API within the vacuole. Light microscopy indicates that the cvt17 mutant is defective in vacuolar vesicle breakdown. Precursor API cofractionates with vacuoles in the cvt17 mutant. Differential osmotic lysis conditions, allowing lysis of the vacuole membrane but not subvacuolar vesicles, revealed that the precursor API could be fractionated away from vacuolar lumenal proteins (48). Protease protection studies confirmed that the subvacuolar precursor API is contained within a membrane-enclosed compartment. Analysis of marker proteins demonstrated that these subvacuolar compartments, termed Cvt bodies, were distinct from vacuole membrane vesicles.
These results suggested that precursor API resides within both cytosolic and vacuolar vesicles. The only way for a subvacuolar vesicle to be derived from the fusion of a cytosolic vesicle with the vacuole is for the cytosolic vesicle to be double membraned.

API Delivery Morphologically Resembles Macroautophagy
The topology of API during import and the morphology of the transit vesicles were directly examined using immunoelectron microscopy (48,49). Precursor API is strikingly absent from the majority of the cytoplasm. Instead of a random distribution, it clusters into complexes. What are presumed to be initial complexes are apparently devoid of membrane or at least of a unit membrane structure. The nature of these complexes, termed Cvt complexes, is not known. The Cvt complexes are subsequently surrounded by membrane, resulting in the formation of double membrane Cvt vesicles. Precursor API was also detected in vesicles in the vacuole lumen in the cvt17 mutant. These morphological data provide direct confirmation of the biochemical and genetic data suggesting that API import occurs through a vesicle-mediated process.

Macroautophagy and the Cvt Pathway Display
Different Kinetics Although the data cited above indicated a substantial mechanistic overlap between the Cvt and macroautophagic pathways, there are still differences. To directly compare the two pathways simultaneously, a macroautophagic marker was analyzed. ALP is a resident vacuolar membrane protein that transits to the vacuole through a portion of the secretory pathway. ALP is a type II integral membrane protein with a C-terminal propeptide that faces the vacuole lumen and that is removed upon vacuolar delivery (50). The N terminus of ALP contains a transmembrane domain that serves as an internal uncleaved signal sequence that is needed to gain entry into the ER. Removal of a portion of the N terminus of ALP including the transmembrane domain generated a construct, Pho8⌬60p, that is no longer able to enter the secretory pathway (51). The only way for Pho8⌬60p to be delivered to the vacuole is by macroautophagy. Vacuolar delivery can be monitored by following removal of the C-terminal propeptide.
Pho8⌬60p was used as a marker for macroautophagy and API as a marker for the Cvt pathway. Examining vacuolar uptake of these two proteins under vegetative and starvation conditions revealed differences between the two pathways (35). API import is rapid and complete under both conditions. The kinetics of import are essentially identical in rich or nitrogen-deficient media. In contrast, uptake of Pho8⌬60p does not occur in rich media. Pho8⌬60p is taken into the vacuole upon induction of macroautophagy by nitrogen starvation. The kinetics of uptake, however, are much slower than seen for API, and the yield of the import process is incomplete; uptake of Pho8⌬60p plateaus at approximately 30% import. By three criteria, the conditions of import, the rate, and the yield, the Cvt pathway and macroautophagy are distinguishable.
API import occurs under both vegetative and starvation conditions. Macroautophagic import cannot account for API import under vegetative conditions. To examine the nature of the import process under starvation conditions, a morphological analysis was employed (49). API import was followed during a shift from vegetative to starvation conditions. During vegetative growth, precursor API in the form of Cvt complexes was detected in Cvt vesicles as described above. These vesicles are approximately 150 nm in diameter and contain an electron-dense core that appears to be distinct from cytosol. Upon starvation, the Cvt complexes are seen inside autophagosomes. The ratio of Cvt vesicles to APs decreased as starvation conditions progressed. The APs are substantially larger than Cvt vesicles, having a diameter of 400 to 900 nm. In addition, the APs contain cytosol along with the Cvt complex.

FIG. 1. Aminopeptidase I is imported into the vacuole by a vesicle-mediated process.
The precursor protein oligomerizes in the cytosol and forms complexes that are surrounded by a sequestering membrane. The environmental conditions that determine the specific type of vesicle used for import are described in the text. These double membrane vesicles fuse with the vacuole delivering a subvacuolar vesicle that is broken down allowing API maturation.

The Machinery Used for API Import Adapts to the Environmental Conditions
These data suggest that API is imported into the vacuole by two processes that substantially overlap and that share many components (Fig. 1). Under vegetative conditions, import occurs through the Cvt pathway. Starvation signals the need for the turnover of cytosolic proteins. By some mechanism, Cvt vesicles are no longer synthesized or are modified to become APs. The signaling mechanism is not known but may involve kinases such as Tor2p (52). Because API uptake remains rapid and complete under starvation conditions, specific machinery must still operate to ensure efficient uptake during starvation. It is interesting to note that API levels increase substantially during nitrogen limitation (35). The increase in API synthesis is accommodated by an increase in the import capacity via the use of APs. If API is a critical hydrolase during starvation, the cell ensures an adequate supply of this enzyme by connecting its biosynthesis to macroautophagic uptake.

Conclusions
The synthesis and degradation of organelles and macromolecules are controlled in part by the availability of metabolic substrates. If non-fermentable carbon sources, such as glycerol or ethanol, become available, mitochondrial synthesis increases. Similarly, peroxisomes proliferate in the presence of oleic acid or methanol. These organelles are selectively delivered to the vacuole for degradation if glucose or ethanol become available for fermentative or respiratory metabolism, respectively. When yeast cells experience conditions of nitrogen or carbon starvation following growth in rich medium, macroautophagic uptake of bulk cytosol is induced. The turnover of macromolecules and organelles within the vacuole lumen provides the cell with critical building blocks. Tremendous amounts of membrane synthesis, flow, and recycling are involved in these varied biogenesis and degradation events, yet they are not well understood.
There are several possible explanations for the use of so many delivery mechanisms to the yeast vacuole and to the overlap between the Cvt and macroautophagy pathways. First, the vacuole contains numerous hydrolases that appear to have overlapping specificities. Proteolytic capacity is critical for survival under certain environmental conditions. By utilizing multiple targeting pathways for resident hydrolases, the cell ensures at least a partial complement of vacuolar enzymes. Second, the yeast vacuole has functional similarities to the mammalian lysosome, the plant vacuole, and the contractile vacuole of slime mold and other organisms. In higher eukaryotes, the targeting pathways have diverged allowing specific delivery of proteins to each of these compartments. Because yeast maintain these diverse functions within the vacuole, various pathways are needed for transport of the corresponding proteins. Third, the dodecameric nature of API may be critical for its function and/or stability. The oligomeric protein is incapable of entering the vacuole through the secretory pathway and must use an alternate mechanism. The most efficient use of cellular resources is to adapt existing machinery, in this case the macroautophagic pathway, to allow efficient transport of API to the vacuole.
Many questions remain to be answered concerning the various targeting pathways highlighted in this review. The mechanistic functions of the many gene products implicated in the VPS-dependent secretory pathway targeting route have not been established. Relatively few proteins have been identified that are specific for the alternate pathway used by alkaline phosphatase. The degradative pathways for protein and organelle turnover are, in some cases, quite specific. The means by which this specificity is achieved are not known. All of these processes involve membrane flow to the vacuole. The origin of the membranes, the ways in which they form and/or package their cargo, how they are targeted to the vacuole, and whether the phospholipid constituents are recycled are topics that need to be further addressed. Because of these many questions, nonclassical protein targeting to the vacuole remains an exciting field of cell biology research.