Intravacuolar membrane lysis in Saccharomyces cerevisiae. Does vacuolar targeting of Cvt17/Aut5p affect its function?

The integral membrane protein Cvt17/Aut5p is a putative lipase essential for intravacuolar lysis of autophagic bodies. It is localized at the endoplasmic reticulum, from which it is targeted via the multivesicular body (MVB) pathway to intravacuolar MVB vesicles. Proteinase protection experiments now demonstrate that the Aut5 amino terminus is located in the cytosol, and the carboxyl terminus is located inside the ER lumen. In contrast to procarboxypeptidase S, targeting of Cvt17/Aut5p to MVB vesicles is not blocked in cells lacking the ubiquitin ligase Tul1p or the deubiquitinating enzyme Doa4p. Also, truncation of the amino-terminal cytosolic Cvt17/Aut5p domain does not inhibit its targeting to MVB vesicles. These findings suggest that similar to Sna3p sorting of Cvt17/Aut5p to MVB vesicles is independent of ubiquitination. By fusing the ER retention/retrieval signal HDEL to the carboxyl terminus of Cvt17/Aut5p, we generated a construct that is held back at the ER. Detailed analysis of this construct suggests an essential role of vacuolar targeting of Cvt17/Aut5p for its function. Consistently, aut5Delta cells are found impaired in vacuolar degradation of autophagocytosed peroxisomes. Importantly, biochemical and morphological data further suggest involvement of Cvt17/Aut5p in disintegration of intravacuolar MVB vesicles. This points to a general function of Cvt17/Aut5p in intravacuolar membrane breakdown.

The integral membrane protein Cvt17/Aut5p is a putative lipase essential for intravacuolar lysis of autophagic bodies. It is localized at the endoplasmic reticulum, from which it is targeted via the multivesicular body (MVB) pathway to intravacuolar MVB vesicles. Proteinase protection experiments now demonstrate that the Aut5 amino terminus is located in the cytosol, and the carboxyl terminus is located inside the ER lumen. In contrast to procarboxypeptidase S, targeting of Cvt17/ Aut5p to MVB vesicles is not blocked in cells lacking the ubiquitin ligase Tul1p or the deubiquitinating enzyme Doa4p. Also, truncation of the amino-terminal cytosolic Cvt17/Aut5p domain does not inhibit its targeting to MVB vesicles. These findings suggest that similar to Sna3p sorting of Cvt17/Aut5p to MVB vesicles is independent of ubiquitination. By fusing the ER retention/ retrieval signal HDEL to the carboxyl terminus of Cvt17/ Aut5p, we generated a construct that is held back at the ER. Detailed analysis of this construct suggests an essential role of vacuolar targeting of Cvt17/Aut5p for its function. Consistently, aut5⌬ cells are found impaired in vacuolar degradation of autophagocytosed peroxisomes. Importantly, biochemical and morphological data further suggest involvement of Cvt17/Aut5p in disintegration of intravacuolar MVB vesicles. This points to a general function of Cvt17/Aut5p in intravacuolar membrane breakdown.
Autophagy is a starvation-induced transport pathway delivering intracellular material for degradation to the lysosome (vacuole) (for review see Refs. 1 and 2). Three independent approaches in the model eukaryote Saccharomyces cerevisiae identify numerous autophagic proteins termed Apg (3), Aut (4), and Cvt proteins (5).
In starving cells proaminopeptidase I is specifically targeted to the vacuole via otherwise unspecific autophagy. Proaminopeptidase I is proteolytically matured in the vacuole; this opens a convenient way to monitor autophagy. In non-starved cells proaminopeptidase I transport is taken over by the Cvt pathway. The Cvt pathway and autophagy are morphologically very similar and use many common components (5,6); however, the Cvt pathway does not transport cytosolic material (7).
Autophagy starts at the preautophagosomal (perivacuolar) structure (8 -10) with the formation of transport vesicles (au-tophagosomes), which nonspecifically enclose parts of the cytosol. Autophagy differs from other protein transport pathways by using transport intermediates (autophagosomes) surrounded by two membrane layers. Consequently, after fusion with the vacuolar membrane, their cytosolic content is not released into the vacuole lumen but is instead released as membrane-enclosed autophagic bodies. Therefore, before vacuolar breakdown of the autophagocytosed material the membrane of autophagic bodies has to be lysed. Clearly, this lysis of membranes must be strictly limited to the membranes of autophagic bodies and must not affect the integrity of the vacuolar limiting membrane. Specific intracellular membrane lysis is a fascinating feature of eukaryotic cells, which is also of medical interest, since it is involved in the pathogenesis of some microorganisms (11).
In yeast, vacuolar proteinases A (encoded by the PEP4 gene) and B (PRB1 gene) are required for lysis of autophagic bodies (12), but their molecular function in disintegrating lipid membranes remains enigmatic. Further components of the lysis machinery, Aut4p (13) and Cvt17/Aut5p (14,15), were recently uncovered. Importantly, Cvt17/Aut5p contains a lipase (or esterase) active site motif, which by site-directed mutagenesis of the active site serine was shown to be essential for its activity (14,15). Our previous work demonstrated that the integral membrane protein Cvt17/Aut5p is targeted from the ER, 1 where a significant steady state pool is detectable via the multivesicular body (MVB) pathway to ϳ50-nm intravacuolar MVB vesicles, which in wild-type cells are degraded dependent on vacuolar proteinase A (15). The MVB pathway starts at the prevacuolar compartment (late endosome) (16,17). Here, dependent on several Vps class E proteins, some membrane proteins are sorted to membrane regions of the prevacuolar compartment, which afterward invaginate and bud off as ϳ50-nm MVB vesicles into the interior of the prevacuolar compartment. This process results in formation of a prevacuolar compartment filled with vesicles, a structure termed the multivesicular body. After its fusion with the vacuole the MVB vesicles are released into the vacuolar lumen and degraded. Two different modes have been described for sorting of membrane proteins to MVB vesicles. Procarboxypeptidase S sorting requires ubiquitin conjugation at its lysine residue at position 8 by the ubiquitin ligase Tul1p and the presence of Doa4p, which releases ubiquitin from ubiquitin-protein conjugates (16,18). In contrast, sorting of Sna3p to MVB vesicles is independent of ubiquitination (19).
We here show that Cvt17/Aut5p has a membrane topology 1 The abbreviations used are: ER, endoplasmic reticulum; MVB, multivesicular body; MES, 2-(N-morpholino)ethanesulfonic acid; MOPS, 3-(N-morpholino)propanesulfonic acid; SC, medium, synthetic complete medium; GFP, green fluorescent protein; proCPS, procarboxypeptidase S. similar to procarboxypeptidase S, with its amino terminus located in the cytosol and the carboxyl terminus in the ER lumen. However, in contrast to procarboxypeptidase S, the sorting of Cvt17/Aut5p to MVB vesicles takes place in tul1⌬ and in doa4⌬ cells, and no obvious sorting signal is found in its aminoterminal cytosolic domain. These findings suggest that Cvt17/ Aut5p is sorted to MVB vesicles independent of ubiquitination, similar to Sna3p.
Furthermore a Aut5-HA-HDEL protein carrying the ER retention/retrieval signal HDEL at its carboxyl terminus proposes an essential function of the vacuolar targeting of Aut5p for lysis of autophagic bodies. This points to a function of Aut5p at or after the prevacuolar compartment. After this finding, we further tested Aut5p for a function in lysing other intravacuolar vesicles. Indeed, aut5⌬ cells during starvation target peroxisomes to their vacuoles but are impaired in their disintegration. We additionally demonstrate the involvement of Aut5p in the lysis of intravacuolar MVB vesicles.

EXPERIMENTAL PROCEDURES
Strains and Growth Media-Media were prepared according to Ausubel et al. (20). If not otherwise mentioned cells were grown in synthetic complete (SC) medium containing 2% glucose. For induction of the GAL1 promotor cells were grown overnight in SC medium containing 2% galactose. Starvation was done in 1% potassium acetate.
Strains are listed in Table I. A PCR fragment consisting of the kanamycin resistance gene flanked by up-and downstream sequences of PRB1 was generated using oligonucleotides Prbkan1 (GGAGTTCT-TCCCATACAAACTTAAGAGTCCAATTAGCTTCCAGCTGAAGCTTC-GTACGC) and Prbkan2 (ATTAAATAATATTCAATT TATCAAGAATA-TCTCTCACTTGCATAGGCCACTAGTGGATCTG) and plasmid pUG6 (21). Chromosomal replacement of PRB1 in WCG4a with this fragment yielded YUE59. Strains YUE92 and YUE94 were generated analogously. Correct gene replacement was confirmed by Southern blotting. Crossing of YMS30 (22) and YIS4 followed by tetrad dissection yielded YUE37. YUE40 is an ascospore from a cross of YMS30 and YUE14 (15). YUE74 and YUE77 were generated by crossing YMS30 with YUE59 and Y05789 with Y12763, respectively, and subsequent tetrad dissection. YUE87 is an ascospore from a cross of YUE37 and YMTA. YUE90 was generated using strain Y04883 and the pep4knockout plasmid KS-PRA1⌬-HIS3 (4).
Construction of Plasmids-The mutations in the cytosolic domain of AUT5::HA 3 (K4R, K9R, ⌬2-12; pUE26, pUE27, pUE38, respectively) were generated by PCR. For the first PCR the primers MUTup (15) and AUT5K4 (tctttcttgaagggcttCtatgcaacattcaatag), AUT5K9 (ggagaagcaa-atctcCttcttgaagggcttttatg), del2-12 (catcctagatgcaaaggagacattcaatagaatatttccc), respectively, were used with pUE7 as template. The PCR products were used as megaprimers in a second PCR together with MUTdown (15) and pUE7 as template. The PCR products were bluntligated into pRS426 cut with SmaI. The constructs were confirmed by sequencing. To introduce HDEL at the carboxyl terminus of AUT5-HA 3 a PCR using the oligonucleotides MUTup and HDEL (CTATTACAAT-TCATCATGGCCGGCGTAATCCGGCAC) and pUE7 as template was performed. The PCR product was subcloned into pRS426 to yield pUE29 -1, into pRS316 to yield pUE30, into pRS315 to yield pUE36, and into pRS425 to yield pUE37. For the construction of pUE41 a PCR fragment was generated using primers AUT5-R3 (GATGCAAAGGAG-AAGCAAATCTCTTTCTTGAAGGGCTTTTATGCAAGCACTGAGCGC-GTAATCTG) and MCS-F4 (TGTAATACGACTCACTATAGGGCGAAT-TGGAGCTCCACCGCGGTGGAATTCGAGCTCGTTTAAAC) and plasmid pFA6a-His3MX6-PGAL1-3HA (23) as template. This PCR product was inserted by homologous recombination within YIS4 cells into plasmid pUE5 and cut with NotI, AatII, and HpaI. For this the PCR fragment together with the cut vector were transformed into YIS4, and transformants were selected on SC medium lacking uracil and then replica-plated on SC medium lacking histidine (SC-his). The recombinant plasmid was rescued from transformants able to grow on SC-his, and the correct recombination was confirmed. The plasmids are listed in Table II.
Immunoblotting-Cells were grown as indicated, and 3 A 600 units of cells were harvested, lysed, and prepared for Western blotting. The samples were resuspended in Laemmli buffer and incubated at 37°C for 30 min with vigorous agitation. Equal amounts of protein were loaded on each lane of standard 7.5 or 10% acrylamide gels, subjected to SDS-PAGE, and electroblotted on polyvinylidene difluoride membranes (Amersham Biosciences). Proteins on immunoblots were visualized by ECL detection (Amersham Biosciences) according to the manufacturer's instructions.
Deglycosylation-Thirty A 600 units of stationary phase cells were harvested and treated as described (15). Immunoprecipitation was done with anti-HA antibody for 2 h at 4°C followed by 1 h of incubation with protein A-Sepharose (Amersham Biosciences). Deglycosylation was achieved by treating samples with endoglycosidase H (Roche Molecular Biochemicals) at 37°C for 1 h.
Protease Protection Experiment-Fifty A 600 units of cells were spheroplasted in SB buffer (1.4 M sorbitol, 50 mM K 2 HPO 4 , 10 mM NaN 3 , 40 mM ␤-mercaptoethanol, pH 7.5) containing 0.3 mg of zymolyase-100T for 1 h at 30°C. The spheroplasted cells were gently lysed in lysis buffer (0.8 M sorbitol, 10 mM MOPS, 1 mM EDTA, pH 7.2) using a tissue grinder. The lysate was cleared of the remaining cells and debris by repeated centrifugation for 5 min at 2000 ϫ g. The cleared lysate was split into aliquots corresponding to 12.5 A 600 . From the time of lysis, all material was kept on ice. Membranes were separated by 30 min of centrifugation at 20,000 ϫ g at 4°C. For protease treatment of the pellet, trypsin was added to a final concentration of 0.5 mg/ml after resuspension of the pellet in lysis buffer, and the samples were incubated for 30 min on ice. If noted, Triton X-100 was present at 1%. Digestions were stopped by adding trichloroacetic acid to a concentration of 10%. The trichloroacetic acid pellets were resuspended in Laemmli buffer and analyzed by SDS-PAGE and immunoblotting.
Indirect Immunofluorescence and Fluorescence Microscopy-Immunofluorescence was performed as described (15). Cells were labeled with mouse anti-HA antibody; as secondary antibody Cy3-conjugated goatanti-mouse immunoglobulin G (Dianova) was used. For visualization of GFP fusions cells were grown to log phase, if not otherwise mentioned, stained with FM4 -64 (Molecular Probes) (25), and viewed with a Zeiss Axioscope 2 Plus microscope equipped with an Axiocam digital image system.
Measurement of Pexophagy-The induction of peroxisomes was done according to Hutchins et al. (26). Briefly, logarithmically growing cells were shifted to synthetic glycerol medium (0.67% yeast nitrogen base without amino acids, 50 mM MES, 50 mM MOPS, 3% glycerol, 0.1% glucose, pH 5.5) for 12 h at 30°C. After the addition of a 10ϫ yeast extract/peptone solution to a final concentration of 1% yeast extract and 2% peptone, the cells were incubated for additional 4 h. For peroxisome induction cells were then washed and transferred to YTO (0.67% yeast nitrogen base without amino acids, 0.1% Tween 40, 0.1% oleic acid) for 19 h. To induce peroxisome degradation cells were shifted to SD-N (0.17% yeast nitrogen base without amino acids and ammonium sulfate, 2% glucose). Aliquots were taken at the indicated times and either prepared for immunoblot analysis using antibodies against Fox3p or directly analyzed in fluorescence microscopy with GFP-SKL. Electron Microscopy-Electron microscopy after permanganate fixation and Epon embedding was done as described (27).

RESULTS
Aut5p is targeted from the ER via the Golgi and the prevacuolar compartment (late endosome) to the vacuolar lumen at 50-nm MVB vesicles (15). In wild-type vacuoles the MVB vesicles carrying Aut5p are broken down, resulting in a half-life of Aut5p of 50 -70 min (14,15). Our previous indirect immunofluorescence microscopy indicated a significant steady state level of Aut5p at the ER (15). We therefore speculated that Aut5p might act at the ER probably by modifying specific lipids, which after transport to autophagosomes render them competent for intravacuolar lysis. If this is true, vacuolar transport would only reflect the turnover of the protein and would be dispensable for its function. Alternatively, Aut5p might function at the prevacuolar compartment or inside the vacuole; in this scenario its vacuolar targeting would be essential. To distinguish between these possibilities, we wanted here to block the vacuolar targeting of Aut5p. One idea was to look for vacuolar-targeting sequences within Aut5p. Site-directed mutagenesis of such targeting sequences should then prevent its vacuolar entry. We started our search for vacuolar-targeting sequences with an evaluation of the Aut5p membrane topology.
At the ER the Amino Terminus of Aut5p Is Exposed to the Cytosol and Its Carboxyl Terminus Points Inside the Lumen-The molecular function of vacuolar proteinase A and B in lysing the lipidous membranes of autophagic bodies is enigmatic. It is tempting to speculate that they might proteolytically activate Aut5p. Our previous analysis of biologically active carboxylterminally HA-tagged Aut5p (Aut5-HAp) did not suggest an amino-terminal processing of Aut5p (15). We generated here an amino-terminally HA-tagged Aut5p (HA-Aut5p) and expressed this under control of the inducible GAL1 promotor to check for carboxyl-terminal processing. Complementation of the proaminopeptidase I maturation defect of aut5⌬ cells indicated biological activity of HA-Aut5p (Fig. 1A, lane 3). Interestingly, in immunoblots HA antibodies detected bands with lower molecular mass than Aut5-HAp (Fig. 1A, lanes 3, 4, and 6), which were absent in glucose-grown cells (Fig. 1A, lane 7). Identical bands were detected with a polyclonal antibody against Aut5p (not shown), confirming their identity with Aut5p species. After immunoprecipitation with HA antibodies and subsequent deglycosylation with endoglycosidase H in aut5⌬ cells and in cells either deficient in vacuolar proteinase A (pep4⌬) or B (prb1⌬), HA-Aut5p did not show unambiguously different mobilities in Western blots (Fig. 1B, lanes 2-5). In pep4⌬ cells the HA-Aut5p band appeared broader, however, suggesting the presence of higher molecular mass species (Fig. 1B, lane 4). If this might indicate proteolytic processing at the carboxyl terminus is unclear at the moment and must be the subject of further detailed studies.
HA-Aut5p was pelletable in lysed spheroplasts (Fig. 1C), indicating it was membrane-associated. We further confirmed in indirect immunofluorescence the localization of HA-Aut5p to the ER in aut5⌬ cells (Fig. 1D). As expected the typical ring- 2-m URA3 P GAL1 Invitrogen like staining around the nucleus and staining near the plasma membrane was seen. Accordingly, in cells lacking vacuolar proteinase A (pep4⌬) HA-Aut5p was detectable inside the vacuole (Fig. 1E, left), whose position is easily visible in Nomarski optics (Fig. 1E, middle). To determine the topology of Aut5p we made proteinase protection experiments using the amino-and carboxyl-terminally HA-tagged Aut5p. In aut5⌬ cells, where Aut5p is located at the ER, HA-Aut5p was proteinase-accessible even in the absence of the detergent Triton X-100 (Fig. 1F,  lanes 8 -10), whereas Aut5-HAp was proteinase-protected ( Fig.   FIG. 1. The amino terminus of Aut5p is located in the cytosol, and its carboxyl terminus is located in the ER lumen. A, Aut5p carrying an HA epitope at its amino terminus (HA-Aut5p) is biologically active and heterogeneous in molecular mass. aut5⌬ or pep4⌬ (lacking vacuolar proteinase A) cells expressing Aut5-HAp from plasmid pUE7 (Aut5-HAp, expressed with its native promotor; lanes 1, 2, 5) or HA-Aut5p from pUE41 (HA-Aut5p, expressed with a GAL1-promotor; lanes 3, 4, 6, 7) were grown in glucose (Glc) or galactose (Gal) medium as indicated and analyzed in immunoblots with antibodies against HA and proaminopeptidase I. pAPI, proaminopeptidase I; mAPI, mature aminopeptidase I. As loading control phosphoglycerate kinase (PGK) is shown. B, the heterogeneity of HA-Aut5p is mostly due to glycosylation. Lysates of galactose grown aut5⌬, pep4⌬ (lacking vacuolar proteinase A), and prb1⌬ (lacking vacuolar proteinase B) cells expressing HA-Aut5p (pUE41) were immunoprecipitated with antibodies against HA and after deglycosylation with endoglycosidase H, probed in immunoblots with antibodies against HA. As controls glucose-grown aut5⌬ cells expressing Aut5-HA (pUE7) or carrying the empty pYES2 vector were included. Further details are given under "Results." C, HA-Aut5p is pelletable. aut5⌬ cells transformed with pUE7 (Aut5-HA) or pUE41 (GAL1::HA-AUT5) were grown in glucose or galactose medium, respectively, spheroplasted, and lysed using a tissue grinder. After removing cell debris and nonlysed spheroplasts, the total lysate (T) was centrifuged at 13,000 rpm for 30 min to generate a pellet (P 13 ) and supernatant (S). Immunoblotting with HA antibodies identified the Aut5p species. As the control the blot was reprobed with antibodies against soluble, cytosolic phosphoglycerate kinase (PGK). D and E, indirect immunofluorescence microscopy of HA-Aut5p expressed in aut5⌬ (D) or pep4⌬ (lacking vacuolar proteinase A) cells (E). Cells were grown in galactose medium to stationary growth phase and then fixed with formaldehyde followed by spheroplasting with zymolyase. The spheroplasted cells were then incubated with primary antibody against HA, and afterward, with secondary Cy3-coupled antibody. Nuclear DNA was stained with 4,6-diamidino-2-phenylindole (DAPI). From left to right, immunofluorescence (HA-Aut5p), Nomarski optics (NOM), showing the vacuole, and nuclear staining with 4,6-diamidino-2-phenylindole. Note the typical ER staining pattern of HA-Aut5p in D, i.e. a ring-like localization around the nucleus and at regions near the plasma membrane, and of the vacuole lumen in cells lacking vacuolar proteinase A (pep4⌬) (E). Bar, 10 m. F, proteinase protection experiment. Cells were lysed as described under "Experimental Procedures." The cleared lysate (Total) was divided into aliquots and centrifuged at 20,000 ϫ g for 30 min. The supernatant (Sup) was removed, and the pellets were resuspended in buffer and either treated with 0.5 mg/ml Trypsin (Tryp) or 0.5 mg/ml Trypsin and 1% Triton X-100 (TrypϩTx). After 30 min on ice the samples were trichloroacetic acid-precipitated and prepared for Western blot analysis. Aut5p was detected using antibodies to HA. As the control, the soluble ER lumenal protein Kar2p is shown. 1F, lanes [3][4][5]. This suggests that the Aut5 amino terminus is located in the cytosol, and the carboxyl terminus is located in the ER lumen.
Sorting of Aut5p via the MVB Pathway Does Not Depend on Ubiquitination-Sorting of procarboxypeptidase S (proCPS) via the MVB pathway requires the ubiquitin ligase Tul1p. Tul1p ubiquitinates the lysine residue 8 of proCPS, which is located in a 19-amino acid amino-terminal domain in the cytosol just preceding the proCPS transmembrane domain (16,18). A lack of ubiquitination in tul1⌬ cells leads to missorting of proCPS to the vacuolar-limiting membrane (18) (Fig. 2B). Our proteinase protection experiments now demonstrate a similar topology for Aut5p, which exposes a 14-amino acid aminoterminal stretch to the cytosol followed by a transmembrane domain (Fig. 3A). As in the case of proCPS the amino-terminal cytosolic domain of Aut5p contains two lysine residues at positions 4 and 9 (Fig. 3A). We therefore checked in indirect im-munofluorescence whether sorting of Aut5-HAp to the vacuolar lumen via the MVB pathway depends on Tul1p. Because Aut5-HAp is rapidly degraded in the vacuole, we used tul1⌬ pep4⌬ cells lacking vacuolar proteinase A for this experiment. Interestingly, in TUL1-deficient cells a significant vacuolar pool of Aut5-HAp was detectable ( Fig. 2A). As a control we confirmed mislocalization of GFP-CPS to the vacuolar-limiting membrane in tul1⌬ pep4⌬ cells (Fig. 2B). Sna3p is another cargo of the MVB pathway, but its sorting does not require ubiquitin conjugation (18,19). As a further control, we checked the localization of Sna3-GFP in tul1⌬ pep4⌬ cells in direct fluorescence microscopy. As expected, Sna3-GFP was correctly localized in the vacuole lumen (Fig. 2C). Taken together, these results suggest that the targeting of Aut5-HAp to the vacuole lumen does not require Tul1p.
Doa4p is a protease that specifically releases ubiquitin from ubiquitin-protein conjugates and, thus, replenishes the free ubiquitin pool of the cells. A lack of Doa4p therefore affects all ubiquitin-dependent processes (28). Consistently, missorting of GFP-CPS to the vacuolar-limiting membrane occurred in doa4⌬ cells (Fig. 2G) (16,19). To prevent vacuolar degradation of Aut5-HAp we used cells deficient in the vacuolar proteinases A, B, and Y (pep4⌬ prb1⌬ prc1⌬). Indirect immunofluorescence microscopy indicated that Aut5-HAp is targeted to the vacuole lumen irrespective of presence of Doa4p (Fig. 2, D and E).
In proCPS the first lysine (position 8) of the cytosolic aminoterminal stretch is the target site for ubiquitin conjugation (16) and, thus, essential for proCPS targeting to MVB vesicles. Using site-directed mutagenesis we replaced lysines 4 and 9 of Aut5-HAp with arginine. To further evaluate, if there is any sorting signal within the 14 amino acids of the amino-terminal cytosolic domain, we generated a truncated Aut5-HA(del2-12)p lacking amino acids 2-12. We expressed these constructs in aut5⌬ cells. Indirect immunofluorescence microscopy confirmed the normal ER localization of all these Aut5-HAp species (not shown). In aut5⌬ pep4⌬ cells significant amounts of these mutant proteins were detectable in the vacuole in addition to the ER localization (Fig. 3, B-D). These findings further argue against a ubiquitination-dependent sorting of Aut5p as well as against the presence of sorting determinants in the amino-terminal cytosolic domain. Interestingly, these constructs complemented the proaminopeptidase I maturation defect in aut5⌬ cells both under non-starvation conditions, where the Cvt pathway is active (Fig. 3E), and after starvation induction of autophagy (Fig. 3F). The constructs also complemented the defect in lysis of autophagic bodies in aut5⌬ cells (Fig. 3G). Some aut5⌬ cells expressing Aut5(K9R)-HAp exhibited few autophagic bodies in their vacuoles (Fig. 3G), indicating slightly retarded degradation. This might indicate a slightly reduced activity of the mutated protein, since accumulation of autophagic bodies is more sensitive in monitoring autophagy than proaminopeptidase I maturation. Taken together our findings suggest that the sorting of Aut5p is independent of ubiquitination and that the Aut5p amino-terminal cytosolic domain contains no sorting information nor is it essential for activity.
An  essential, we wanted to block vacuolar targeting of Aut5-HA and check whether this interferes with lysis of autophagic bodies. We therefore next analyzed the biological activity and localization of Aut5-HAp in vps23⌬ and vps28⌬ cells. In these Vps class E mutants protein sorting via the MVB pathway is disturbed, leading to accumulation of its cargoes at the prevacuolar compartment and their mislocalization to the vacuolar membrane (16). To allow detection of a vacuolar Aut5-HAp pool in these Vps class E mutants, we used cells defective in the vacuolar proteinases A, B, and Y (pep4⌬ prb1⌬ prc1⌬). Indirect immunofluorescence confirmed localization of Aut5-HAp to the prevacuolar compartment and the vacuolar-limiting membrane in vps23⌬ and vps28⌬ cells. No significant amounts of intravacuolar Aut5-HAp were detected ( Fig. 4C and data not shown). In immunoblots no significant accumulation of proaminopeptidase I was detectable in these Vps class E mutants, neither when the Cvt pathway is active (non-starved cells) nor after starvation induction of autophagy (Fig. 4D). As men-tioned, light microscopic evaluation of autophagic body lysis is more sensitive than proaminopeptidase I maturation. Because in our wild-type strain vacuoles of starved cells are more readily visible in Nomarski optics, we deleted VPS23 and VPS28 in WCG4a. Consistent with proaminopeptidase I maturation vps23⌬ cells showed no vacuolar accumulation of autophagic bodies after starvation (Fig. 4E). Also, most of the vps28⌬ cells accumulated no autophagic bodies. Some cells accumulated few autophagic bodies in each cell less than observed in aut5⌬ cells (Fig. 4E). Also in vps28⌬ pep4⌬ cells indirect immunofluorescence microscopy confirmed that the bulk transport of Aut5-HAp to the vacuolar lumen is inhibited (not shown). However, this does not exclude that the Aut5-HAp mislocalized to the vacuolar-limiting membrane or the small amounts that might still reach the vacuole lumen are sufficient for lysis of almost all autophagic bodies.
For a more detailed analysis, we therefore generated an Aut5-HAp species carrying a HDEL motif at its carboxyl terminus (see "Experimental Procedures"). The HDEL motif in yeast functions as an ER retention/retrieval signal, i.e. HDELproteins leaking to the early Golgi are continuously retrieved back to the ER (29). Indirect immunofluorescence microscopy indeed confirmed ER localization of Aut5-HA-HDEL in aut5⌬ pep4⌬ cells and, if any, detected only minor amounts in the vacuolar lumen (Fig. 5B). In yeast after linkage of core Nglycan within the ER, further glycosylation in the Golgi includes the addition of ␣-1,6-mannose residues (30). Accordingly, Aut5-HA-HDEL exhibits compared with Aut5-HAp a significantly enhanced ␣-1,6-mannose glycosylation pattern due to its repeated retrieval from the Golgi (Fig. 5C). As a control the samples were further treated with N-glycosidase F and endoglycosidase H (Fig. 5C); both enzymes are widely used to release N-glycans from proteins. Taken together the findings confirm ER retention/retrieval of Aut5-HA-HDEL. Interestingly, Aut5-HA-HDEL expressed from a centromeric plasmid with its native promotor only partly complemented the proaminopeptidase I maturation defect in aut5⌬ cells (Fig. 5D, lane 3). Overexpression of Aut5-HA-HDEL from a two micron plasmid however lead to a more complete proaminopeptidase I maturation (Fig. 5D, lane 4). This dosage dependent complementation is in agreement with the idea, that a small amount of Aut5-HA-HDEL, which might still reach the vacuole is sufficient for lysis of autophagic bodies. The MVB-sorting defect in vps28⌬ cells should further reduce the amount of Aut5-HA-HDEL reaching the vacuolar lumen. Indeed, vps28⌬ aut5⌬ cells expressing Aut5-HA-HDEL from a centromeric plasmid contained almost exclusively proaminopeptidase I (Fig. 5D, lane  6), and overexpression from a 2-m plasmid led to only partial proaminopeptidase I maturation (Fig. 5D, lane 7). The enhancement of the proaminopeptidase I maturation defect in vps28⌬ cells further supports the idea that small amounts of Aut5-HA-HDEL are sufficient to lyse autophagic bodies. Taken together our results do not suggest a function of Aut5p at the ER, but at later stages of its sorting pathway, namely at the prevacuolar compartment or the vacuole.
Aut5p Is Essential for Pexophagy-Because our findings suggest a function of Aut5p at the prevacuolar compartment or the vacuole, we next analyzed whether breakdown of other vesicular intermediates in the vacuole also depends on Aut5p. Growth of S. cerevisiae cells in medium containing oleic acid as the sole carbon source induces proliferation of peroxisomes. When these cells are shifted to nitrogen starvation, peroxisomes are specifically targeted to and degraded in the vacuole in a process called pexophagy (26). For morphological analysis of pexophagy we used a plasmid encoded GFP-SKL fusion protein (31). The carboxyl-terminal peroxisomal targeting sig- nal 1 (SKL) targets the GFP to peroxisomes, which thus become visible in fluorescence microscopy as cytosolic green dots. In wild-type cells vacuolar degradation of GFP-SKL-containing peroxisomes liberates a quite proteolysis-resistant GFP into the vacuole lumen, yielding a homogeneously fluorescent vacuole (Fig. 6B). In contrast, a defective vacuolar breakdown of peroxisomes in pep4⌬ cells lacking vacuolar proteinase A results in vacuolar accumulation of green dots (Fig. 6B). After induction of pexophagy, aut5⌬ cells expressing GFP-SKL clearly accumulated distinct green dots in their vacuoles (Fig.  6B). This indicates vacuolar uptake of peroxisomes in aut5⌬ cells but a defective vacuolar breakdown. This is in agreement with the defect of aut5⌬ cells in lysing autophagic bodies. We further confirmed the peroxisomal degradation defect in aut5⌬ cells in immunoblots using the peroxisomal matrix protein Fox3p (3-ketoacyl-CoA thiolase) (26) as a marker. In wild-type cells Fox3p levels were reduced during starvation, but this reduction was not observed in aut5⌬ or pep4⌬ cells (Fig. 6C).
Aut5p Affects Intravacuolar Lysis of MVB Vesicles-Vacuolar lysis of autophagic bodies requires Aut5p (15), (14), Aut4p (13), and vacuolar proteinase B (12). In addition to autophagic bodies, MVB vesicles are also disintegrated in the vacuole. The components needed for lysing MVB vesicles have not been studied in detail so far. We were especially interested in determining if Aut5p, located on the MVB vesicles, is also involved in their disintegration. To biochemically monitor the integrity of intravacuolar MVB vesicles we used two marker proteins (i) GFP-CPS, a MVB cargo whose sorting requires ubiquitination (16) and (ii) Sna3-GFP, which is sorted independent of ubiquitin conjugation (19). As illustrated in Fig. 7, both fusion proteins expose their GFP moiety into the interior of intravacuolar MVB vesicles. Undigested fusion proteins therefore indicate intact MVB vesicles. GFP-CPS can be proteolytically cleaved closely after its transmembrane domain even if the MVB vesicles are intact (Fig. 7, A and B). In this case an intermediate sized GFP species (GFP*) is formed. When the MVB vesicles are lysed from both fusion proteins, proteolysis resistant-free GFP is released. We analyzed logarithmically growing and nitrogen-starved cells. Our analysis yielded similar results using GFP-CPS or Sna3-GFP as markers. In grow-  Fig. 1E and analyzed with antibodies against HA. Note the localization of Aut5-HA-HDEL to the ER (ring-like staining around the nucleus and near the plasma membrane), whereas no significant labeling is seen in the vacuole. Bar, 10 m. DAPI, 4,6-diamidino-2-phenylindole; NOM, Nomarski optics. C, ␣-1,6-mannose linkages were monitored in immunoblots with specific antibodies. Crude extracts of stationary cells expressing Aut5-HA (pUE35), Aut5-HA-HDEL (pUE37), or an empty vector (pRS425) were immunoblotted and probed with antibodies against ␣-1,6-mannose linkages (upper panel) and HA (lower panel). Samples were immunoprecipitated with HA antibodies and either deglycosylated with endoglycosidase H or N-glycosidase F or mock-treated. An asterisk marks cross-reacting material. Aut5p* corresponds to deglycosylated Aut5p species. D, Aut5-HA-HDEL shows a dosage-dependent complementation of the proaminopeptidase I maturation defect of aut5⌬ cells. vps28⌬ aut5⌬ cells and aut5⌬ cells expressing Aut5-HA or Aut5-HA-HDEL from a centromeric (CEN) or 2-m plasmid (2) were grown to stationary phase and analyzed in immunoblots with antibodies against proaminopeptidase I (upper panel), against HA (middle), and cytosolic phosphoglycerate kinase (PGK) (lower panel). An asterisk marks cross-reacting material. After quantification using ImageQuant, the amount of mature aminopeptidase I was expressed as the percentage of the total amount of mature and proaminopeptidase I present in the sample. pAPI, proaminopeptidase I; mAPI, mature aminopeptidase I. ing wild-type and aut4⌬ cells large amounts of free GFP indicated lysis of MVB vesicles (Fig. 7, A and B, lanes 1 and 3). As expected, pep4⌬ cells showed no disintegration of MVB vesicles (Fig. 7, A and B, lane 4). Interestingly, growing aut5⌬ cells exhibited a significantly reduced amount of free GFP (Fig. 7, A  and B, lane 2), indicating a reduced lysis of MVB vesicles. Starved wild-type, aut4⌬, and pep4⌬ cells showed results similar to growing cells (Fig. 7). Starved aut5⌬ cells contained a significant level of undigested Sna3-GFP (Fig. 7A, lane 6) and GFP* (Fig. 7B, lane 6). These findings suggest a function of Aut5p in MVB vesicle lysis both in growing and starved cells. To exclude the occurrence of proteolysis during cell lysis, we confirmed the presence of proaminopeptidase I in the aut5⌬ extracts analyzed in Fig. 7, A and B by immunoblotting with anti-aminopeptidase I antibodies (not shown). To further exclude that the defects in MVB vesicle lysis in aut5⌬ cells are caused by missorting of the marker proteins, we checked their localization by fluorescence microscopy. Both GFP-CPS and Sna3-GFP localized to the vacuole lumen in aut5⌬ cells (Fig. 7,  C and D).
To further confirm that the GFP-CPS and Sna3-GFP degradation defects are due to defects in MVB vesicle lysis, we performed electron microscopy. Because in starved cells the large number of accumulating autophagic bodies interferes with detection of the 50-nm MVB vesicles, we generated mutant strains also lacking the autophagy protein Apg1/ Aut3p. A lack of this serine/threonine protein kinase selectively abolishes formation of Cvt vesicles and autophagic bodies (22,32). aut3⌬ aut5⌬ pep4⌬ mutant cells accumulated 50-nm vesicles in their vacuoles (Fig. 8C). This shows that Aut5p is not essential for formation of 50-nm vesicles. The electron microscopic analysis corroborated our biochemical study by showing 50-nm vesicles in aut3⌬ aut5⌬ cells (Fig.  8A). Compared with aut3⌬ pep4⌬ cells (Fig. 8B) aut3⌬ aut5⌬ cells (Fig. 8A) showed fewer but clearly visible intravacuolar 50-nm vesicles. DISCUSSION To learn more about the site of action of Aut5p, we wanted to block its vacuolar targeting. We started with proteinase protection experiments (Fig. 1) using amino-and carboxyl-terminally HA-tagged Aut5p to evaluate the Aut5p topology. The experiments demonstrated a cytosolic localization of the Aut5p amino terminus, with the carboxyl terminus trapped in the ER lumen. This suggests a localization of the lipase active site inside the ER and is consistent with the previously observed glycosylation of this part of Aut5p (14,15). Interestingly, our findings now demonstrate the existence of an amino-terminal 14-amino acid-long cytosolic domain of Aut5p just before the transmembrane domain. This is reminiscent to the topology of proCPS. proCPS is targeted to MVB vesicles after Tul1p-dependent ubiquitination of lysine 8. Lysine 8 is located in a short cytosolic amino-terminal stretch preceding the transmembrane domain of proCPS (16,18). Because of the topological similarities, we analyzed the two lysines found in the cytosolic aminoterminal Aut5p domain for a vacuolar-targeting function. However, in contrast to proCPS replacement of lysines 4 and 9 of Aut5-HAp with arginine did not prevent its vacuolar localization in indirect immunofluorescence microscopy (Fig. 3, B and  C). In addition, the complete deletion of the Aut5 amino-terminal cytosolic domain (amino acids 2-12) did not abolish its vacuolar localization nor its activity (Fig. 3, D-G). Together with vacuolar targeting in tul1⌬ and in doa4⌬ cells, this argues against a ubiquitin-dependent targeting of Aut5p to MVB ves- Bar, 10 m. C, cells were treated as described in A, and after a shift to nitrogenfree medium, aliquots were taken at the indicated times and processed for immunoblotting using antibodies against the peroxisomal marker protein Fox3p. The amounts of Fox3p were quantified using ImageQuant and expressed as the percentage of the amount present at time point 0.
icles. Aut5p in this respect resembles Sna3-GFP (19). This finding is highly interesting; however, the lack of a specific sorting signal in the amino-terminal cytosolic domain of Aut5p prevented us from using the mutated Aut5p species to block its vacuolar targeting.
We therefore next analyzed Vps class E mutants, where Aut5-HAp is retained at the prevacuolar compartment and partly mislocalized to the vacuolar membrane (Fig. 4C). Under starvation and non-starvation conditions the tested Vps class E mutants showed in immunoblots mature aminopeptidase I, suggesting the occurrence of the Cvt and autophagic pathway. Light microscopic examination of autophagic body lysis is more sensitive to detect autophagy defects than the maturation of proaminopeptidase I. vps23⌬ cells showed wild-type like lysis of autophagic bodies. Some vps28⌬ cells, however, accumulated a few autophagic bodies in their vacuoles, but fewer than did aut5⌬ cells (Fig. 4E). We followed this first hint for a non-ER function by generating an Aut5-HA-HDEL protein. We confirmed its retention/retrieval at the ER in indirect immunofluorescence microscopy (Fig. 5B) and by analyzing its glycosylation pattern (Fig. 5C). Most interestingly, in aut5⌬ cells expression of Aut5-HA-HDEL from a centromeric plasmid only partially complemented the proaminopeptidase I maturation defect (Fig. 5D, lane 3), but overexpression from a 2-m plasmid complemented almost completely (Fig. 5D, lane 4). Although indirect immunofluorescence microscopy did not detect a vacuolar pool of Aut5-HA-HDEL, we hypothesized that small amounts of Aut5-HA-HDEL, which might still leave the ER, might be responsible for lysis of autophagic bodies. If this is true, combining the MVB-sorting defect of Vps class E mutants with the ER retention/retrieval of Aut5-HA-HDEL should further enhance the proaminopeptidase I maturation defect. Indeed, centromeric expression of Aut5-HA-HDEL in aut5⌬ vps28⌬ cells resulted in almost no proaminopeptidase I maturation (Fig. 5, lane 6), and overexpression of Aut5-HA-HDEL in these cells only led to partial maturation. This suggests that a small amount of Aut5p is sufficient for vesicle lysis, consistent with the idea of an enzymatic function. Our findings further suggest that Vps28-dependent sorting of Aut5p to MVB vesicles is essential for its biological function.
Within the vacuolar lumen not only autophagic bodies but also numerous MVB vesicles are lysed. We wanted to know whether Aut5p is also involved in the lysis of these MVB vesicles. We used GFP-CPS (16) and Sna3-GFP (18) with similar results as marker proteins to monitor the integrity of MVB vesicles in immunoblots (Fig. 7, A and B). Electron microscopy further confirmed that the degradation defects observed with the GFP fusion proteins correspond to defects in lysing the MVB vesicles (Fig. 8). Most interestingly, in growing and starved cells lacking Aut5p a significantly reduced breakdown of MVB vesicles was detected. This points to an additional function of Aut5p in lysing these membranes. As a control we checked in aut3⌬ aut5⌬ pep4⌬ cells that Aut5p is not obviously needed for biogenesis of the MVB vesicles (Fig. 8C). In agreement with the observed overlapping function of several proteins between pexophagy and autophagy (26,33), we could further demonstrate (Fig. 6) that aut5⌬ cells are able to take up peroxisomes in their vacuoles but are defective in their breakdown. Taken together our findings suggest a function of Aut5p at the prevacuolar compartment (late endosome) or at the vacuole and point to a more general role of Aut5p in lysis of intravacuolar vesicles.
Several ways that Aut5p could mediate vesicle breakdown seem conceivable. Based on the essential role of the active site motif characteristic for lipases and esterases, Aut5p might act as an unspecific hydrolase directly attacking membranes inside the vacuole. In this case it would be crucial for the cells to prevent untimely activation of Aut5p during its transit to the vacuole. Selective activation within the vacuolar lumen might be achieved by proteolytic maturation. So far, our analysis of amino-terminally (Fig. 1) and carboxyl-terminally (15) HAtagged Aut5p did not clearly detect a matured Aut5p species; however, the observed broader band of HA-Aut5p in cells lacking vacuolar proteinase A (pep4⌬) needs further detailed studies. Alternatively, Aut5p might be activated by the acidic vacuolar pH. However, neither of these activation strategies explains how lysis of the vacuolar limiting membrane is prevented. Another possibility that explains the specificity would be an activation of Aut5p by interaction with another protein such as a colipase. To identify such a putative interacting protein, we made a high copy suppressor screen using aut5⌬ cells, but under the conditions used, this did not detect any suppressors. Also, in our hands a two-hybrid screen using Aut5p as bait did not result in detection of a valuable interaction partner. Interestingly, a large scale two-hybrid approach (34) pointed to the inositol phosphosphingolipid phospholipase C Isc1p (35, 36) as a putative Aut5p-interacting protein. We chromosomally deleted ISC1, but under the conditions tested light microscopic examination did not show vacuolar accumulation of autophagic bodies during starvation (not shown). This does not support a direct involvement of Isc1p in lysis of autophagic bodies. In general, the idea of Aut5p as a hydrolase with low substrate specificity seems unlikely, since this demands sophisticated mechanisms for controlling its activity. If Aut5p acts as a hydrolase, a high specificity for molecules present only at its target membranes would significantly limit its risk for the integrity of the cell.
In an alternate scenario in the cytosol a multivesicular body, i.e. a late endosome (prevacuolar compartment) filled with MVB vesicles, might fuse with an autophagosome. Within the resulting organelle the MVB vesicles then could fuse with the inner membrane layer of autophagosomes. This would deliver FIG. 8. Electron microscopy demonstrates accumulation of 50-nm vesicles in the vacuoles of aut5⌬ cells. Because accumulation of Cvt or autophagic bodies would interfere with the detection of MVB vesicles, we here use aut3⌬ cells, which are deficient in biogenesis of Cvt vesicles and autophagosomes. aut3⌬ aut5⌬ (A), aut3⌬ pep4⌬ (B), and aut3⌬ pep4⌬ aut5⌬ (C) cells were starved for 4 h in 1% potassium acetate, fixed with permanganate, and processed for epon embedding and electron microscopy. On the right magnified cutouts are shown. Arrowheads point to vesicles within the vacuole. Bars, 500 nm.
Aut5p to the inner membrane of autophagosomes and, thus, to autophagic bodies. In this scenario Aut5p would attack in the vacuolar lumen those membranes where it is located. As discussed, a high substrate selectivity of Aut5p would also be expected in this scenario. In mammalian cells indeed fusions between endosomes and autophagosomes, resulting in the formation of amphisomes, have been reported (37). One should also take into account the possibility that Aut5p might function already at or inside the multivesicular body. Because in all scenarios a high substrate selectivity of Aut5p seems likely, it is a challenging task for future work to identify such a putative Aut5p substrate.