Vid28 Protein Is Required for the Association of Vacuole Import and Degradation (Vid) Vesicles with Actin Patches and the Retention of Vid Vesicle Proteins in the Intracellular Fraction*

Background: The Vid pathway is linked to the nonclassical secretory and internalizing pathways. Results: Vid vesicle proteins are in the extracellular fraction when VID28 is deleted or mutated. Conclusion: VID28 is required for the retention of Vid vesicle proteins in the intracellular fraction. Significance: Learning how signal-less proteins are secreted is critical for understanding the nonclassical secretory pathway. Gluconeogenic enzymes are induced when Saccharomyces cerevisiae are starved of glucose. However, when glucose is added to prolonged starved cells, these enzymes are degraded in the vacuole via the vacuole import and degradation (Vid) pathway. The Vid pathway is linked to the nonclassical secretory and internalizing pathways. In prolonged starved cells, substantial amounts of the key gluconeogenic enzyme fructose-1,6-bisphosphatase (FBPase) are in the extracellular fraction (periplasm). However, when glucose is added to glucose-starved cells, levels of extracellular FBPase decrease rapidly. Ultrastructural studies indicate that FBPase is in Vid/endosomes following glucose addition, suggesting that FBPase is internalized in response to glucose refeeding. Under the same conditions, the majority of Vid vesicle proteins are in the intracellular fraction. In yeast, actin polymerization is involved in endocytosis. Vid vesicles associate with actin patches initially, and they dissociate later. Here, we show that VID28 plays a critical role in the association of Vid vesicles with actin patches and the retention of Vid vesicle proteins in the intracellular fraction. Vid28p was distributed to Vid vesicles and interacted with other Vid vesicle proteins. Vid28p contains an Armadillo (ARM) domain required for FBPase degradation. When VID28 was deleted or when the ARM domain was mutated, Vid vesicles failed to co-localize with actin patches, and Vid vesicle proteins appeared in the extracellular fraction. We suggest that the ARM domain is required for the association of Vid vesicles with actin patches and the retention of Vid vesicle proteins in the intracellular fraction.

Steady-state protein levels are determined by protein synthesis and protein degradation. Two major protein degradation pathways degrade intracellular proteins: the proteasome-de-pendent and the lysosome-dependent pathways. The yeast vacuole is homologous to mammalian lysosomes and contains a variety of hydrolytic enzymes that degrade proteins or macromolecules that are targeted to the vacuole (1)(2)(3)(4)(5). The vacuolar protein sorting pathway is a biosynthetic pathway that targets carboxykinase Y from the endoplasmic reticulum to the Golgi and to the vacuole for processing (2)(3)(4)(5). The cytoplasm-to-vacuole targeting pathway delivers aminopeptidase I from the cytoplasm to the vacuole for maturation (6,7). The nonselective autophagic pathway degrades cytosolic proteins and organelles under starvation condition (8 -11). Organelles such as peroxisomes can be degraded by pexophagy (12,13). Extracellular molecules are internalized and delivered to the vacuole via multiple endocytic pathways, such as the receptor-mediated endocytosis and fluid-phase endocytosis (14 -18).
For the Vid pathway, multiple FBPase-containing compartments have been purified from wild-type cells (35). Vid vesicles are 30 -50 nm in diameter (43). The UBC1 gene is critical for the biogenesis of Vid vesicles (44). The import of FBPase into Vid vesicles requires the heat shock protein Ssa1p/Ssa2p, cyclophilin A, and Vid22p (34,38,39). Vid24p is a peripheral protein on Vid vesicles (45). Coat protein I (COPI) coatomer proteins are also localized to Vid vesicles and are required to recruit Vid24p to Vid vesicles (40). Vid vesicles can aggregate to form clusters. Vid/endosomes contained clusters of Vid vesicles and have been purified from cells that were starved of glucose and then replenished with glucose for 20 min (35). It is not known how Vid vesicles cluster and how they associate with actin patches. In yeast, actin polymerization is involved in the early steps of endocytosis (14 -17). Vid vesicle proteins Sec28p, Vid24p, and Vid30p associate with actin patches initially and dissociate later (35,46). Cargo proteins FBPase and MDH2 also transiently associate with actin patches and dissociate later (35).
The Vid pathway is linked to the nonclassical secretory and internalizing pathways. When cells are starved for a prolonged time, substantial amounts of FBPase are in the extracellular fraction/periplasm (47). Levels of FBPase in the extracellular fraction/periplasm decreased when glucose was added to glucose-starved cells. Following glucose addition for 15 min, FBPase is in Vid/endosomes, suggesting that FBPase is internalized in response to glucose refeeding. Because FBPase does not contain the classical endoplasmic reticulum-Golgi signal sequence and the internalizing sequence, this protein is likely to be secreted and internalized via the nonclassical secretory and internalizing pathways. Under the same experimental conditions, Vid vesicle proteins Vid24p, Sec28p, and Vid30p are in the intracellular fraction (47).
A number of VID genes have been identified. However, only a handful of these genes have been characterized (39,41,46). Several of the VID genes such as VID28 and VID30 are involved in both proteasome and vacuole pathways (28,42,49). Here, we examined the roles that VID28 plays in the vacuole pathway. Vid28p was distributed to Vid vesicles and co-localized with actin patches. Vid28p interacted with other Vid vesicle proteins including Vid24p, Sec28p, and Vid30p. In the absence of the VID28 gene, co-localization of Vid vesicles with actin patches was inhibited, and Vid vesicle proteins appeared in the extracellular fraction. Vid28p contains an ARM domain. When the ARM domain was deleted, co-localization of Vid vesicles with actin patches was inhibited, and Vid vesicle proteins were in the extracellular fraction. We suggest that the ARM domain is required for the association of Vid vesicles with actin patches and the retention of Vid vesicle proteins in the intracellular fraction.

EXPERIMENTAL PROCEDURES
Strains, Media, and Antibodies-The yeast strains used in the study are listed in Table 1. The primers used in this study are shown in Table 2. Cells were grown in medium containing low glucose in YPKG (1% yeast extracts, 2% peptone, 1% potassium acetate and 0.5% glucose) for 3 days as the starvation condition. Cells were aliquoted and transferred to YPD (1% yeast extracts, 2% peptone and 2% glucose) medium for the indicated time points. Anti-HA antibodies were purchased from Roche Applied Science, and anti-V5 antibodies were from Invitrogen. Anti-Pil1p antibodies were from Dr. Dickson (University of Kentucky). Anti-actin antibodies were from Abcam. Rhodamine-conjugated phalloidin was purchased from Invitrogen.
FBPase Degradation, Differential Centrifugation, and His Pulldown Experiments-To study FBPase degradation, cells were grown in YPKG for 3 days and aliquoted. Cells were transferred to YPD for 0, 2, and 3 h. Total lysates were obtained and examined for the degradation of FBPase by Western blotting using anti-FBPase antibodies. For differential centrifugation, wild-type cells were glucose-starved for 3 days and re-fed with glucose for 20 min. Cell lysates were sequentially centrifuged at 13,000 ϫ g for 10 min, 100,000 ϫ g for 2 h, and 200,000 ϫ g for 2 h. Vid vesicles are enriched in the 200,000 ϫ g (P200) pellet fraction, whereas soluble proteins are enriched in the 200,000 ϫ g supernatant (S200) fraction. The P200 and S200 fractions were resolved by SDS-PAGE and examined for the distribution of FBPase, Vid24p-HA, Vid24p-V5-His, Vid28p-HA, Vid30p-HA, and Vid28p-V5-His by Western blotting. For His pulldown experiments, cells co-expressing various tags were glucose-starved for 3 days and transferred to medium containing high glucose for 20 min. Cells were harvested and solubilized in buffer containing 2% Triton X-100. His pulldowns were performed according to the manufacturer's instructions (Qiagen). Samples were separated into the total, unbound, and bound fractions and then immunoblotted with FBPase and Sec28p antibodies. Vid24p-HA, Vid28p-HA, and Vid30p-HA were detected with HA antibodies, whereas Vid24p-V5-His, Vid28p-V5-His, and Vid30p-V5-His were detected with anti-V5 antibodies.
Fluorescence Microscopy-For actin staining, yeast cells expressing GFP-tagged proteins were grown under starvation conditions in 2 ml of YPKG for 3 days. Cells were processed as described (35,46). Actin was stained with 1 l of rhodamineconjugated phalloidin at 0.2 units/l in methanol and incubated for 30 min in the dark at room temperature. GFP and actin were visualized at 26°C with FLUAR 100ϫ objective lens (1.30 NA) using FITC and rhodamine filters. Images were taken using a Zeiss Axiovert S100 inverted microscope with an AxioCam MRm charge-coupled device camera and AxioVision version 4.5 software.
Extraction of Extracellular Proteins-Cells were grown in YPKG for 3 days and aliquoted. The same amounts of cells (OD 600 ϭ 10) were transferred to medium containing glucose for 0, 15, and 30 min. Cells were pelleted and extracted using the protocol as described (50). Cells were incubated with 100 l of extraction buffer containing 0.1 M Tris, pH 9.4, and 10 mM ␤-mercaptoethanol in a 37°C shaker for 15 min. Following incubation, cells were pelleted, and the supernatant fraction was centrifuged at 16,000 ϫ g for 30 s at room temperature. Proteins from the supernatant (called the extracellular fraction in this study) were precipitated using 15% trichloroacetic acid,

Vid28p, Nonclassical Secretion, and the Vid Pathway
washed, and resuspended in SDS-PAGE buffer. Cell pellets (called the intracellular fraction in this study) were also lysed and resuspended in SDS-PAGE buffer. Both intracellular and extracellular fractions were examined by Western blotting with anti-FBPase, anti-Sec28p, anti-HA, and anti-Pil1p antibodies.
Plasmid Preparations-The VID28 gene was amplified by PCR reactions using P157F and P158R as the forward and reverse primers ( Table 2). In some experiments, P40F and P40R were used as the forward and reverse primers ( Table 2). PCR products were purified and cloned into a TOPO plasmid (Invitrogen). Vid28p-V5-His was digested with NruI and SwaI to remove the 2 sequence and religated. The resulting plasmid was linearized with HindIII and integrated into the VID28 locus. Following the integration, Vid28p-V5-His was expressed under its own promoter. The second copy of untagged Vid28p was not expressed because of the absence of galactose in the medium. Vid24p-V5-His was digested with EcoRI and integrated, whereas Vid30p-V5-His was digested with HindIII and integrated. Vid28p-GFP-HIS3 was produced by digesting a plasmid containing Vid28p-V5-His-URA3 with PacI and SacII and ligated with a fragment produced by PacI and SacII digestion of the FBPase-GFP-HIS3 plasmid. The ARM domain mutation was produced by digesting the Vid28p-V5-His plasmid or the Vid28p-GFP-HIS3 plasmid with MscI and Pml1 and religated. The resulting deletion was confirmed by DNA sequencing at the Core Facility of the Penn State University College of Medicine.

VID28 Is Required for the Degradation of FBPase in the
Vacuole-The VID28 gene was originally identified via a transposon library screening for mutants that showed defective FBPase degradation in the vacuole-dependent pathway. FBPase is degraded in the vacuole when glucose is added to cells that are starved of glucose for 3 days (42). To confirm that the VID28 gene was involved in the vacuole-dependent pathway, wildtype and ⌬vid28 cells were grown in low glucose media for 3 days, re-fed with glucose, and examined for FBPase degradation (Fig. 1A). In 3-day-starved wild-type cells, FBPase was degraded in response to glucose. In contrast, FBPase degradation was impaired in the 3-day-starved ⌬vid28 mutant, suggesting that Vid28p is involved in the degradation of FBPase in the vacuole.
FBPase Associates with Vid Vesicles in the ⌬vid28 Mutant-For the Vid pathway, FBPase is associated with Vid vesicles that are enriched in the 200,000 ϫ g pellet fraction. We determined whether the VID28 gene has a role in this process. If the VID28 gene is required for the association of FBPase with Vid vesicles, FBPase should not be detected in the Vid vesicle-enriched fraction in the ⌬vid28 strain. In contrast, if the VID28 gene is not involved in this process, FBPase should be present in the Vid vesicle-enriched fraction in cells lacking the VID28 gene. Wildtype cells and the ⌬vid28 mutant cells were starved of glucose for 3 days and transferred to a medium containing glucose for 20 min. Cell lysates were subjected to differential centrifugation and examined for the distribution of FBPase in the Vid vesicleand cytosol-enriched fractions. In wild-type cells, FBPase was distributed in both the Vid vesicle-enriched and the cytosolenriched fractions (Fig. 1B). In the ⌬vid28 mutant, most of the FBPase was in the Vid vesicle-enriched fraction (Fig. 1B). Small amounts of FBPase were in the cytosol-enriched fraction.
The presence of FBPase in the Vid vesicle-enriched fraction in the ⌬vid28 mutant suggests that Vid vesicles are formed in the ⌬vid28 strain. Therefore, the Vid vesicle protein Vid24p should also be detected in the Vid vesicle-enriched fraction in the ⌬vid28 strain. Levels of Vid24p were low during glucose starvation and increased following the addition of glucose for 20 -30 min (45). The ⌬vid28 strain expressing Vid24p-HA was glucose-starved for 3 days, transferred to a medium containing fresh glucose for 20 min, and examined for the distribution of Vid24p in the Vid vesicle-and cytosol-enriched fractions. In wild-type cells expressing Vid24p-HA, this protein was in both the Vid vesicle-enriched and the cytosol-enriched fractions (Fig. 1C). In the ⌬vid28 mutant, most of the Vid24p-HA was in the Vid vesicle fraction (Fig. 1C). It is not clear why FBPase and Vid24p-HA were present at low levels in the cytosol-enriched fraction in the ⌬vid28 strain. However, high amounts of Vid24p in the Vid vesicle fraction suggest that Vid vesicles are formed.
The ⌬vid28 Mutant Affects the Association of FBPase and Vid24p with Actin Patches-We determined whether VID28 has a role in the association of FBPase and Vid24p with actin patches. Wild-type strains and ⌬vid28 strains expressing either FBPase-GFP or GFP-Vid24p were starved of glucose for 3 days and re-fed with glucose for 0, 30, and 60 min. The distribution of FBPase-GFP and GFP-Vid24p with actin patches was examined using a protocol that gives consistent and reliable results regarding the distribution of GFP-tagged proteins with actin patches stained with phalloidin conjugated with rhodamine (35,46). In wild-type cells, the association of FBPase-GFP with actin patches increased in response to glucose addition for up to 30 min ( Fig. 2A). Less co-localization of FBPase with actin patches was seen at the 60-min time point. In the ⌬vid28 strain, FIGURE 1. VID28 is required for FBPase degradation in prolonged starved cells. A, wild-type cells and ⌬vid28 cells were grown in media containing low glucose for 3 days, transferred to media containing high glucose for 0, 2, and 3 h, and examined for the degradation of FBPase. B, wild-type strains and ⌬vid28 strains were glucose-starved and re-fed with glucose for 20 min. Cells were harvested, and lysates were subjected to differential centrifugation. The distribution of FBPase in the Vid vesicle-enriched fraction (V) and the cytosolenriched fraction (C) was examined. C, wild-type strains and ⌬vid28 strains expressing Vid24p-HA were glucose-starved and re-fed with glucose for 20 min. The distribution of Vid24p-HA in the Vid vesicle-enriched fraction and cytosol-enriched fraction was determined. APRIL 26, 2013 • VOLUME 288 • NUMBER 17 little co-localization of FBPase-GFP with actin patches was observed before and after the addition of glucose (Fig. 2B).

Vid28p, Nonclassical Secretion, and the Vid Pathway
We determined whether cells lacking the VID28 gene also reduced co-localization of the Vid vesicle protein Vid24p with actin patches. In wild-type cells expressing GFP-Vid24p, co-localization of this protein with actin patches was observed at the t ϭ 0-min and the t ϭ 30-min time points (Fig. 2C). Less co-localization was seen at the t ϭ 60-min time point (Fig. 2C). In the ⌬vid28 cells, little co-localization of GFP-Vid24p with actin patches was observed during glucose starvation and following glucose addition (Fig. 2D). Thus, the association of FBPase and Vid24p with actin patches is inhibited in cells lacking the VID28 gene.
Vid28p Is Distributed to the Vid Vesicle-enriched Fraction-We next examined the expression and the distribution of Vid28p. Vid28p was tagged with HA and expressed in wild-type cells that were glucose-starved and transferred to glucose for the indicated time points. Levels of Vid28p-HA did not increase in response to glucose (Fig. 3A). FBPase was degraded in Vid28p-HA-tagged cells, suggesting that HA tagging does not affect FBPase degradation. We also tagged Vid28p with V5-His and expressed this protein in wild-type cells that were starved of glucose and then replenished with glucose (Fig. 3B). Levels of Vid28p-V5-His did not increase following the addition of glucose (Fig. 3B). Moreover, FBPase was degraded in cells expressing Vid28p-V5-His, suggesting that V5-His tagging does not interfere with the degradation of FBPase.
Next, we investigated whether Vid28p was distributed to the Vid vesicle-enriched fraction. Wild-type cells expressing Vid28p-HA or Vid28p-V5-His were starved of glucose, transferred to a medium containing fresh glucose for 20 min, and examined for the distribution of Vid28p in the Vid vesicle-and cytosol-enriched fractions. In strains expressing Vid28p-HA, most of the Vid28p-HA was in the Vid vesicle-enriched fraction (Fig. 3C). FBPase was in both the cytosolic and the Vid vesicle fractions. Likewise, in strains expressing Vid28p-V5-His, the majority of this protein was in the Vid vesicle-enriched fraction, and low amounts were in the cytosolic fraction (Fig. 3D). FBPase was in both the Vid vesicle-enriched and the cytosolenriched fractions. Thus, Vid28p is localized to the Vid vesicleenriched fraction.
Vid28p Co-localizes with Actin Patches-We determined whether Vid28p co-localized with actin patches. Vid28p-GFP was expressed in wild-type cells that were starved of glucose and then transferred to media containing high levels of glucose for the indicated time points. FBPase was degraded in wild-type cells expressing Vid28p-GFP (Fig. 3E), suggesting that GFP tagging does not inhibit Vid28p function in FBPase degradation.
In wild-type cells expressing Vid28p-GFP, co-localization of Vid28p-GFP with actin patches was observed following the addition of glucose for up to 30 min (Fig. 3F). Less co-localization was seen at the 60-min time point. Thus, Vid28p and actin patches associate initially and dissociate later in response to glucose replenishment.
Vid28p Interacts with Vid24p-Because Vid28p was in the Vid vesicle-enriched fraction, we examined whether Vid28p interacts with other Vid vesicle proteins. We first determined the interaction of Vid28p with Vid24p. Wild-type cells co-ex- expressing FBPase-GFP were glucose-starved and re-fed with glucose for 0, 30, and 60 min. The distribution of FBPase and actin patches was examined using fluorescence microscopy. C and D, wild-type strains (C) and ⌬vid28 strains (D) expressing GFP-Vid24p were glucose-starved and transferred to media containing high glucose for 0, 30, and 60 min. The distribution of GFP-Vid24p with actin patches was determined by fluorescence microscopy.

Vid28p, Nonclassical Secretion, and the Vid Pathway
pressing Vid28p-V5-His and Vid24p-HA were glucose-starved and re-fed with glucose for 20 min. Vid28p-V5-His was pulled down from total lysates, and proteins were separated into the total, unbound, and bound fractions. The presence of Vid24p-HA in these fractions was then examined by immunoblotting (Fig. 4A). Most of the Vid28p-V5-His was in the bound fraction. The majority of Vid24p-HA was also in the bound fraction. We have shown previously that COPI coatomer proteins interact with Vid24p (40). As expected, high levels of Sec28p were detected in the bound fraction that contained Vid24p-HA. Coatomer proteins are involved in other protein trafficking pathways. As such, a portion of Sec28p was detected in the unbound fraction. Under the same conditions, most of the FBPase was in the unbound fraction. We also determined whether or not actin was in the bound fraction. The majority of actin was in the unbound fraction.
We performed similar experiments using wild-type cells that co-expressed Vid24p-V5-His and Vid28p-HA. This strain was and Vid28p-V5-His (D) were starved of glucose and transferred to media containing high glucose for 20 min. Total lysates were subjected to differential centrifugation and examined for the distribution of Vid28p-HA, Vid28p-V5-His, and FBPase in the Vid vesicle-and cytosol-enriched fractions. E, wild-type cells expressing Vid28p-GFP were glucose-starved and examined for the degradation of FBPase in response to glucose. F, the association of Vid28p-GFP and actin patches was determined using fluorescence microscopy. Vid24p-V5-His and Vid28p-HA were co-expressed in wild-type cells that were starved and re-fed with glucose for 20 min. Vid24p-V5-His was pulled down from total lysates and examined for the presence of Vid28p-HA, Sec28p, FBPase, and actin in the total, unbound, and bound fractions. C, Vid28p-V5-His and Vid24p-HA were co-expressed in the ⌬sec28 mutant that was glucose-starved and replenished with glucose for 20 min. Vid28p-V5-His was pulled down from total lysates. Levels of Vid28p-V5-His, Vid24p-HA, FBPase, and actin in the total, unbound, and bound fractions were determined. D, Vid28p-V5-His and Vid24p-HA were expressed in the ⌬vid30 strain that was starved and re-fed with glucose for 20 min. Vid28p-V5-His was pulled down from total lysates, and levels of Vid28p-V5-His, Vid24p-HA, Sec28p, FBPase, and actin in the total, unbound, and bound fractions were determined. Relative distribution of proteins in the unbound and bound fractions was summarized in Table 3. APRIL 26, 2013 • VOLUME 288 • NUMBER 17 glucose-starved and re-fed with glucose for 20 min (Fig. 4B). Most of the Vid24p-V5-His was in the bound fraction. The majority of Vid28p-HA was also in the bound fraction. Likewise, high levels of Sec28p were in the bound fraction. FBPase and actin were mainly in the unbound fraction. Therefore, Vid24p, Vid28p, and Sec28p co-precipitated in the bound fraction. However, FBPase and actin were not in the materials that contain Vid24p, Vid28p, and Sec28p.

Vid28p, Nonclassical Secretion, and the Vid Pathway
The Absence of SEC28 Interferes with Vid28p Interaction with Vid24p-We next determined whether the absence of SEC28 affected the interaction of Vid28p with Vid24p. We produced a ⌬sec28 strain that co-expressed Vid28p-V5-His and Vid24p-HA. Cells were glucose-starved and re-fed with glucose for 20 min. Vid28p-V5-His was pulled down from total lysates, and the distribution of Vid28p-V5-His and Vid24p-HA was examined (Fig. 4C). In the ⌬sec28 strain, most of the Vid28p-V5-His was precipitated in the bound fraction. The majority of Vid24p-HA was in the unbound fraction. FBPase and actin were mainly in the unbound fraction. Therefore, the absence of SEC28 decreases levels of Vid24p precipitated by Vid28p.
We examined whether or not the absence of VID30 affected the interaction of Vid28p with Vid24p. The ⌬vid30 strain was transformed to express Vid28p-V5-His and Vid24p-HA. Cells were glucose-starved for 3 days and re-fed with glucose for 20 min (Fig. 4D). Vid28p-V5-His was pulled down in the bound fraction. Most of the Vid24p-HA was also in the bound fraction. Likewise, high levels of Sec28p were in the bound fraction. FBPase and actin were mainly in the unbound fraction. Therefore, the absence of VID30 does not affect the levels of Vid24p and Vid28p in the bound fraction.
Vid28p Interacts with Vid30p-We determined whether Vid28p interacted with Vid30p. Wild-type cells that co-expressed Vid28p-V5-His and Vid30p-HA were glucose-starved and re-fed with glucose. Vid28p-V5-His was pulled down from total lysates, and the distribution of Vid30p-HA in the unbound and bound fractions was examined (Fig. 5A). Most of the Vid28p-V5-His and Vid30p-HA was in the bound fraction. High levels of Sec28p were also in the bound fraction. FBPase and actin were in the unbound fraction.
Similar experiments were performed using wild-type cells that co-expressed Vid30p-V5-His and Vid28p-HA that were glucose-starved and transferred to media containing high glucose (Fig. 5B). Vid30p-V5-His was in the bound fraction. High levels of Vid28p-HA and Sec28p were also in the bound fraction. FBPase and actin, on the other hand, were mostly in the unbound fraction. These experiments suggest that Vid28p also interacts with Vid30p.
The Absence of SEC28 Inhibits Vid28p Interaction with Vid30p-We determined whether the absence of SEC28 affected Vid28p interaction with Vid30p. Vid28p-V5-His and Vid30p-HA were co-expressed in cells lacking the SEC28 gene. Cells were glucose-starved for 3 days and re-fed with glucose. Vid28p-V5-His was pulled down from total lysates and examined for the presence of Vid30p-HA in the unbound and bound fractions (Fig. 5C). In cells lacking SEC28, the majority of Vid28p-V5-His was in the bound fraction. Most of the Vid30p-HA was in the unbound fraction. FBPase and actin were in the unbound fraction. Thus, the absence of Sec28p reduces Vid28p interaction with Vid30p.
We examined whether the absence of VID24 affected the interaction of Vid28p with Vid30p. The ⌬vid24 mutant strain was transformed to express Vid28p-V5-His and Vid30p-HA. Cells were starved of glucose for 3 days and re-fed with glucose (Fig. 5D). Vid28p-V5-His was pulled down from total lysates, and the presence of Vid30p-HA in the unbound and bound fractions was then examined (Fig. 5D). In the ⌬vid24 strain, both Vid28p-V5-His and Vid30p-HA were in the bound frac-  Table 3.

Vid28p, Nonclassical Secretion, and the Vid Pathway
tion. Lower levels of Sec28p were in the bound fraction. FBPase and actin were in the unbound fraction. Therefore, the absence of VID24 does not affect the levels of Vid28p precipitated by Vid30p in the bound fraction.
VID28 Is Required for Vid24p Interaction with Vid30p-We have shown previously that Vid24p interacts with Vid30p (46). We determined whether the absence of VID28 interfered with the interaction of Vid24p with Vid30p. Wild-type cells and the ⌬vid28 cells that co-expressed Vid24p-V5-His and Vid30p-HA were starved of glucose and re-fed with glucose. Vid24p-V5-His was pulled down from total lysates and examined for the presence of Vid30p-HA in the unbound and bound fractions (Fig.  6A). In wild-type cells, Vid24p-V5-His and Vid30p-HA were in the bound fraction. High levels of Sec28p were also in the bound fraction, whereas most of the FBPase and actin was in the unbound fraction.
In cells lacking VID28, most of the Vid24p-V5-His was in the bound fraction (Fig. 6B). However, Vid30p-HA was mainly distributed in the unbound fraction. Sec28p levels also decreased in the bound fraction. Most of the FBPase and actin was in the unbound fraction. Therefore, the absence of VID28 decreases interaction of Vid24p with Vid30p.
Sec28p, Vid24p, and Vid30p Are in the Extracellular Fraction in the ⌬vid28 Mutant-Using a cell extraction protocol that extracts extracellular proteins from whole cells, we have demonstrated that substantial amounts of FBPase were in the extracellular fraction in prolonged starved wild-type cells (47). Levels of FBPase in the extracellular fraction decreased following glucose addition to glucose-starved wild-type cells. However, the majority of Sec28p, Vid24p, and Vid30p were in the intracellular fraction in wild-type cells before and after glucose addition (47). We next examined whether VID28 has a role in the distribution of FBPase and Vid vesicle proteins in the intracellular and extracellular fractions. Wild-type strains and the ⌬vid28 strains expressing either Vid24p-HA or Vid30p-HA were glucose-starved for 3 days and transferred to media containing high levels of glucose for 0, 15, and 30 min. The distribution of FBPase, Sec28p, Vid24p, and Vid30p in the intracellular and extracellular fractions was then examined.
In wild-type cells, FBPase was detected in the extracellular fraction during glucose starvation (Fig. 7A). Levels of extracellular FBPase decreased in response to glucose (Fig. 7A). The  Relative distribution of these proteins in the intracellular and extracellular fractions in glucose-starved wild-type strain and the ⌬vid28 strain was determined using the National Institutes of Health ImageJ software and listed in Table 5. APRIL 26, 2013 • VOLUME 288 • NUMBER 17 majority of Sec28p, Vid24p, and Vid30p were in the intracellular fraction in wild-type cells before and after glucose addition (Fig. 7A). Pil1p is the major component of eisosomes. This protein was detectable in both intracellular and extracellular fractions in wild-type cells during glucose starvation and following glucose addition (Fig. 7A).

Vid28p, Nonclassical Secretion, and the Vid Pathway
In the ⌬vid28 mutant, FBPase was also in the extracellular fraction during glucose starvation, and levels of extracellular FBPase decreased in response to glucose (Fig. 7B). In this strain, Sec28p, Vid24p, and Vid30p were detected in the extracellular fraction during glucose starvation (Fig. 7B). Moreover, levels of these proteins decreased in response to glucose replenishment. Pil1p was in both intracellular and extracellular fractions before and after the addition of glucose to the ⌬vid28 cells. Therefore, the absence of VID28 results in the appearance of Sec28p, Vid24p, and Vid30p in the extracellular fraction during glucose starvation.
Vid28p Lacking the ARM Domain Inhibits FBPase Degradation-Vid28p contains an ARM domain located in amino acids 624 -873. We determined whether the ARM domain has a role in the Vid pathway. We produced the deletion of the ARM domain that was fused with the V5-His tag. The ⌬ARM-Vid28p-V5-His protein was expressed during glucose starvation, and levels decreased following the glucose addition for 60 min. In the ⌬ARM strain, FBPase degradation was inhibited (Fig. 8A), suggesting that this domain plays a role in the Vid pathway.
We determined whether ⌬ARM-Vid28p-V5-His was localized to the Vid vesicle-enriched fraction. Cells expressing ⌬ARM-Vid28p-V5-His were glucose-starved and re-fed with glucose for 20 min. The distribution of ⌬ARM-Vid28p-V5-His in the Vid vesicle-and cytosol-enriched fractions was examined (Fig. 8B). Most of the ⌬ARM-Vid28p-V5-His was in the Vid vesicle-enriched fraction. FBPase was in both the Vid vesicle-enriched and the cytosol-enriched fractions (Fig. 8B). To examine whether or not the deletion of the ARM domain affected the distribution of other Vid vesicle proteins to the Vid vesicle-enriched fraction, we produced a strain that co-expressed ⌬ARM-Vid28p-V5-His and Vid24p-HA that was glucose-starved and re-fed with glucose. Most of the Vid24p-HA was in the Vid vesicle-enriched fraction (Fig. 8B). Likewise, in a strain that co-expressed ⌬ARM-Vid28p-V5-His and Vid30p-HA, most of the Vid30p-HA was in the Vid vesicle-enriched fraction (Fig.  8B). Thus, Vid28p lacking the ARM domain is distributed to the Vid vesicle fraction. Furthermore, the localization of Vid24p, Vid30p, and FBPase to the Vid vesicle-enriched fraction is not affected by the deletion of the ARM domain. We attempted to determine whether Vid28p lacking the ARM domain co-localized with actin patches. Unfortunately, the ⌬ARM-Vid28p-GFP was expressed at low levels and was not used for further studies.
We next examined whether the absence of the ARM domain affected the interaction of Vid28p with Vid24p. Cells co-expressing ⌬ARM-Vid28p-V5-His and Vid24p-HA were starved and re-fed with glucose for 20 min. Total lysates were subjected to the pulldown experiments and examined for the presence of these proteins in the total, unbound, and bound fractions (Fig.   8C). ⌬ARM-Vid28p-V5-His was precipitated in the bound fraction. Levels of Vid24p-HA and Sec28p in the bound fraction decreased. FBPase and actin were in the unbound fraction. These results suggest that the absence of the ARM domain interferes with Vid28p interaction with Vid24p.
We determined whether the absence of the ARM domain affected Vid28p interaction with Vid30p. Cells that co-expressed ⌬ARM-Vid28p-V5-His and Vid30p-HA were glucosestarved, re-fed with glucose, and examined for the presence of these proteins in the total, unbound, and bound fractions (Fig.  8D). The majority of ⌬ARM-Vid28p-V5-His was in the bound fraction. Relative levels of Vid30p-HA in the bound fraction decreased. Most of the Sec28p, FBPase, and actin was in the unbound fraction.
FBPase and Vid24p Fail to Co-localize with Actin Patches in the ⌬ARM Mutant-We determined whether the deletion of the ARM domain affected the association of FBPase and Vid24p with actin patches. FBPase-GFP and GFP-Vid24p were expressed in cells in which the ARM domain was deleted. Cells were glucose-starved for 3 days and transferred to media con-

Vid28p, Nonclassical Secretion, and the Vid Pathway
taining fresh glucose for the indicated time points. Little co-localization of FBPase-GFP with actin patches was observed during glucose starvation or following the addition of glucose for up to 60 min (Fig. 9A). Likewise, GFP-Vid24p did not show obvious co-localization with actin patches in cells lacking the ARM domain before and after the addition of glucose (Fig. 9B). Therefore, the deletion of the ARM domain reduces the association of FBPase and Vid24p with actin patches.
The Absence of the ARM Domain Results in the Appearance of Sec28p, Vid24p, and Vid30p in the Extracellular Fraction-Finally, we determined whether the deletion of the ARM domain affected the distribution of FBPase, Sec28p, Vid24p, and Vid30p in the intracellular and extracellular fractions. In both the wild-type cells and the ⌬ARM cells, FBPase was in the extracellular fraction, and levels of extracellular FBPase decreased in response to glucose (Fig. 10, A and B). In wild-type cells, most of the Sec28p, Vid24p, and Vid30p was in the intracellular fraction. In the ⌬ARM strain, these proteins were detected in the extracellular fraction during glucose starvation (Fig. 10B). Levels of these proteins in the extracellular fraction decreased when glucose was added to glucose-starved ⌬ARM strain (Fig.  10B). Pil1p was in the intracellular and extracellular fractions before and after glucose addition in both the wild-type strains and the ⌬ARM strains.

DISCUSSION
In this study, we described the function of the VID28 gene in the Vid pathway. When the VID28 gene was deleted, co-local-ization of FBPase and Vid24p with actin patches was inhibited, and the interaction of Vid24p with Vid30p was interrupted. Furthermore, Sec28p, Vid24p, and Vid30p were in the extracellular fraction in the ⌬vid28 mutant. We suggest that VID28 is involved in the association of Vid vesicles with actin patches, the interaction of Vid24p with Vid30p, and the retention of Sec28p, Vid24p, and Vid30p in the intracellular fraction. However, FBPase and Vid24p were still detected in the Vid vesicleenriched fraction in the ⌬vid28 strain. Therefore, VID28 is not   Table 4. APRIL 26, 2013 • VOLUME 288 • NUMBER 17 involved in the formation of Vid vesicles and the association of FBPase with Vid vesicles.

Vid28p, Nonclassical Secretion, and the Vid Pathway
Vid28p interacts with other Vid vesicle proteins including Vid24p, Sec28p, and Vid30p. Based on these interaction studies (summarized in Table 3 and Table 4), we propose a model in which these proteins are organized in the following order: Vid24p-Sec28p-Vid28p-Sec28p-Vid30p. This model is consistent with our interaction data showing that: 1) the interaction of Vid28p and Vid24p was inhibited in cells lacking SEC28; 2) the interaction of Vid28p and Vid24p was preserved in cells lacking VID30; 3) the interaction of Vid28p and Vid30p was affected in cells lacking SEC28; 4) the interaction of Vid28p with Vid30p was observed in cells lacking VID24; and 5) the interaction of Vid24p and Vid30p was disrupted in the absence of VID28. This complex may be required for Vid vesicles to cluster to form large aggregates. When the interaction is disrupted, vesicles may not aggregate, and hence the association with actin patches is inhibited. This may result in the release of Vid vesicles into the extracellular fraction. Consistent with this idea, Sec28p, Vid24p, and Vid30p were in the extracellular fraction in the ⌬vid28 cell that was glucose-starved ( Table 5 ). The presence of these proteins in the extracellular fraction during glucose starvation is unlikely to result from cell lysis. In our experiments, cells were aliquoted prior to the addition of glucose. As such, the same amounts of cells were used to extract extracellular proteins from whole cells. If cell lysis had occurred, levels of these proteins should have been similar before and after the addition of glucose. In fact, levels of these proteins in the extracellular fraction decreased rapidly following the addition of glucose to the ⌬vid28 cells for 15 min. Therefore, cell lysis cannot explain the rapid decline of these proteins in the extracellular fraction in these strains following glucose replenishment. Under these same conditions, Pil1p was detectable in the extracellular fraction following glucose addition.
Vid vesicles are 30 -50 nm in diameter and may exist in two forms. Individual Vid vesicles are below the detection limit for fluorescence microscopy. If these vesicles are scattered in the cytoplasm, they may appear as diffused staining in the cytoplasm. However, when these vesicles are clustered together, they may appear as punctate structures as observed by fluorescence microscopy. Therefore, the Vid vesicles that associate with actin patches as seen by fluorescence microscopy are likely the clustered form of Vid vesicles. Individual Vid vesicles may not be detected by transmission electron microscopy using thin sections of embedded cells either, because the membranes may not be preserved well using this technique.
At present, it is not known how Vid vesicles associate with actin. Vid vesicles may aggregate to form clusters, and actin then associates with clusters of Vid vesicles. Co-localization of Vid vesicles with actin patches is observed in glucose-starved cells, suggesting that Vid vesicles and actin patches already associate prior to the addition of glucose. In the absence of VID28, FBPase and Vid24p were in the Vid vesicle-enriched fraction, suggesting that small vesicles are formed. However, they failed to co-localize with actin patches. Actin patches stained with phalloidin were still observed in the ⌬vid28 mutant. These actin patches may be partially assembled in this strain. Alternatively, they may be fully assembled but not functional. If actin patches are preassembled, they may be added directly to the surface of clusters of Vid vesicles to form Vid clusters/actin. The ARM domain is required for the association of FBPase and Vid24p with actin patches and for the retention of these

Vid28p, Nonclassical Secretion, and the Vid Pathway
Vid proteins in the intracellular fraction. When the ARM domain was deleted, FBPase and Vid24p failed to co-localize with actin patches. Furthermore, Sec28p, Vid24p, and Vid30p were also detected in the extracellular fraction in glucosestarved cells (Table 5). Because levels of FBPase and Vid vesicle proteins in the extracellular fraction decreased in response to glucose in the ⌬ARM mutant, this domain is unlikely to be involved in the decline of these proteins in the extracellular fraction.
We propose that Vid28p is required for the aggregation of Vid vesicles to form clusters (Fig. 11). When the VID28 gene is deleted or mutated, Vid vesicles cannot aggregate to form clusters. Hence the association with actin patches is blocked. This may lead to the appearance of Vid vesicle proteins Sec28p, Vid24p, and Vid30p in the extracellular fraction. FBPase was in the extracellular fraction in the wild-type strain and the ⌬vid28 and ⌬ARM strains, suggesting that this cargo protein is secreted into the extracellular fraction independent of the VID28 gene. Interestingly, levels of these proteins in the extracellular fraction all decreased in response to glucose in the ⌬vid28 and the ⌬ARM strains, suggesting that the decline of these proteins in the extracellular fraction is independent of the VID28 gene. The decrease in levels of these proteins in the extracellular fraction may result from internalization into the cytoplasm, degradation in the extracellular space, or release into the media.
Small vesicles called exosomes are released from a variety of mammalian cells and are important for cell-cell communication (51)(52)(53). Exosomes are 40 -100 nm in diameter and have a density of 1.1-1.2 g/ml, whereas Vid vesicles are 30 -50 nm in diameter and have a density of 1.2 g/ml. Interestingly, gluconeogenic enzymes such as FBPase have been identified in exosomes purified from NIT-1 insulinoma cells (48). Multivesicular bodies have been observed in cells that secrete exosomes. However, the mechanisms responsible for the secretion of small vesicles have not been elucidated. The Vid pathway may be a valuable model system to study the trafficking of small vesicles in response to changes in the environment.  11. The proposed roles of VID28 in the Vid pathway. Vid28p is distributed in the Vid vesicle-enriched fraction and interacts with other vesicle proteins including Sec28p, Vid24p, and Vid30p. Vid vesicles may exist in two forms. When the VID28 gene was deleted or mutated, FBPase and Vid24p were in the Vid vesicle fraction, but they failed to co-localize with actin patches. These results suggest that small vesicles were formed. However, they failed to aggregate. We propose that VID28 is required for Vid vesicles to aggregate to form clusters. When clusters cannot form, Sec28p, Vid24p, and Vid30p appear in the extracellular fraction. FBPase was in the extracellular fraction in both the wild-type cells and the ⌬vid28 cells, suggesting that the presence of FBPase in the extracellular fraction is independent of the VID28 gene. Levels of FBPase, Sec28p, Vid24p, and Vid30p in the extracellular fraction decreased rapidly in response to glucose. Therefore, VID28 is unlikely to be involved in the decline of these proteins in the extracellular fraction.