Biochemical Journal

Research article

The phosphoinositide 3-kinase Vps34p is required for pexophagy in Saccharomyces cerevisiae

Silke Grunau , Dorothee Lay , Sabrina Mindthoff , Harald W. Platta , Wolfgang Girzalsky , Wilhelm W. Just , Ralf Erdmann


PIds (phosphoinositides) are phosphorylated derivatives of the membrane phospholipid PtdIns that have emerged as key regulators of many aspects of cellular physiology. We have discovered a PtdIns3P-synthesizing activity in peroxisomes of Saccharomyces cerevisiae and have demonstrated that the lipid kinase Vps34p is already associated with peroxisomes during biogenesis. However, although Vps34 is required, it is not essential for optimal peroxisome biogenesis. The function of Vps34p-containing complex I as well as a subset of PtdIns3P-binding proteins proved to be mandatory for the regulated degradation of peroxisomes. This demonstrates that PtdIns3P-mediated signalling is required for pexophagy.

  • peroxisome
  • pexophagy
  • phosphoinositide
  • Saccharomyces cerevisiae
  • Vps34p


Peroxisomes are subcellular organelles bound by a single membrane that are ubiquitously present in eukaryotic cells. They have common metabolic functions such as the oxidation of fatty acids and the generation and removal of hydrogen peroxide. Peroxisomes have also been implicated in a series of other metabolic activities such as the synthesis of plasmalogens, isoprenoids, penicillin and lysine, and the catabolism of polyamines, D-amino acids and methanol [1]. A remarkable property of peroxisomes is their unique variability in enzyme content and metabolic function depending on the cell or tissue type, or the needs of both the cell or the whole organism [2]. The mechanism regulating the abundance of peroxisomes can be subdivided into different stages. When fermentable sugars are present, yeast β-oxidation genes are turned off by a mechanism termed glucose repression [3]. When non-fermentable carbon sources are present, these genes become up-regulated resulting in an increased gene expression including those encoding peroxisomal proteins [4]. Degradation of fatty acids in fungi and plants is exclusively localized in peroxisomes [5]. In Saccharomyces cerevisiae growth on oleic acid as the sole carbon source results in a striking proliferation of peroxisomes and is accompanied by a massive induction of genes encoding enzymes of the peroxisomal β-oxidation system [4]. When the peroxisome proliferation stimulus is removed, and/or peroxisomal metabolism is no longer required, peroxisomal abundance is decreased rapidly by a selective autophagic process termed pexophagy [6]. Interestingly, different lipids are involved in the regulation of this process. The peroxisomal induction is under the control of nuclear transcription factors regulated by lipid-derived peroxisome proliferators [4]. In S. cerevisiae, oleic acid serves as a specific signal for peroxisomal proliferation and the transcriptional regulation of genes encoding peroxisomal function is based on the ORE (oleate response element) [4].

Specific degradation of peroxisomes via pexophagy has been linked to the PtdIns3P and PtdIns4P pathways with purported roles as factors of the machinery facilitating engulfment and subsequent degradation of peroxisomes in the vacuolar lumen contain PId (phosphoinositide)-binding domains [7]. However, PtdIns3P and PtdIns4P as well as the two identified pexophagy-specific binding proteins Atg24p and Atg26p have been shown to localize to autophagic membranes during pexophagy and not to the peroxisomes themselves [8]. In the present paper we report our discovery that yeast peroxisomes are capable of synthesizing PtdIns3P and demonstrate that Vps34p, the only PId 3-kinase in S. cerevisiae, is localized at peroxisomes under inducing conditions. Cells lacking Vps34p are impaired in selective peroxisomal degradation, indicating that PtdIns3P signalling is required for the regulation of pexophagy.


Strains and culture conditions

The S. cerevisiae strains used in the present study are listed in Table 1. Genomic tagging of VPS34 was performed as described previously [9] using primers RE1946 and RE1947 for the PCR. The vps34Δ strain was constructed by deletion of VPS34 of the wild-type UTL7-A as described previously [9]. Transformants were selected for the appropriate marker and proper integration was confirmed by PCR. Yeast complete medium [YPD: 1% (w/v) yeast extract/2% (w/v) peptone/2% (w/v) glucose] and minimal medium [YNB: 0.67% yeast nitrogen base without amino acids/0.5% glucose] have been described previously [10]. Media contained 0.1% oleic acid and 0.05% Tween 40 (YNO; induction) or 2% glucose (repression) or 2% ethanol (up-regulation) plus 0.1% yeast extract and 0.67% yeast nitrogen base without amino acids, adjusted to pH 6.0. For pexophagy studies, strains were grown on YNO medium over night. Cells were harvested, washed with sterile water and shifted to the pexophagy medium containing 2% glucose/0.17% yeast nitrogen base without amino acids, adjusted to pH 6.0. When necessary, auxotrophic requirements were added according to [11].

View this table:
Table 1 Strains used in the present study

Euroscarf, Institute of Molecular Biosciences, Johann Wolfgang Goethe-University Frankfurt, Frankfurt, Germany.


The plasmid pAR2 encoding GFP (green fluorescent protein)–PTS (peroxisomal targeting signal) 1 [12] was used as a peroxisomal marker. The localization of GFP fusion proteins was also analysed in the yHPR251 strain, which expresses PTS2–DsRed as peroxisomal marker from an integrated plasmid in strain UTL-7A. For EGFP (enhanced GFP)-tagging of Vps34p, the VPS34 open reading frame was amplified by PCR with genomic S. cerevisiae DNA as the template and RE1988 and RE1989 as primers (Table 2). The VPS34 PCR product was digested with SmaI/SalI and cloned into pUG36. pGEX-6P-3 plasmids encoding the double FYVE (conserved in Fab1, YOTB, Vacl and EEAI)-finger of mouse Hrs (amino acids 147–223) or its mutated form carrying a C215S mutation in both FYVE domains [13] were cut with EcoRI and SalI. The isolated fragments were ligated into the EcoRI/SalI digested pUG36, resulting in EGFP–2×FYVE and EGFP–2×FYVE(C215S).

View this table:
Table 2 Oligonucleotides used in the present study

Antibodies and immunoblotting

Immunoblots were incubated with polyclonal rabbit antibodies raised against Pcs60p [14], Pex11p [15], ProtA (Protein A; Sigma), Kar2p [16], Pep12p (Molecular Probes), ALP (alkaline phosphatase; Molecular Probes), Gas1p [17], Fbp1p [18], Fox1p [19], Fox3p [20] and Pex13p [15]. HRP (horseradish peroxidase)-coupled anti-rabbit antibody (Sigma–Aldrich) was used as a secondary antibody, and the blots were developed using the ECL (enhanced chemiluminescence) system (Amersham Buchler).

Isolation of peroxisomes

Preparation of the yeast spheroplasts, cell homogenization and determination of the suborganellar localization of proteins was performed as described previously [10]. For density gradient centrifugation, postnuclear supernatants (10 mg of protein) was prepared and loaded on to preformed 2.25–24% (w/v) Optiprep (Iodixanol) gradients. Peroxisomes were separated from other organelles in a vertical rotor (Sorvall TV 860, 1.5 h, 19500 rev./min at 4 °C). Fractions were collected from the bottom and were subjected to enzyme and refractive index measurements as well as immunoblot analysis.

Radioactive labelling of cell organelles and chromatographic separations

The organelles obtained from the gradient fractions were diluted in a sucrose buffer and sedimented for 1 h at 38000 rev./min (Beckman 70.1 rotor). Labelling, TLC and the preparation of [32P]-labelled PId standards were performed as described previously [21,22]. HPLC analysis of phosphoinositols was carried out as described previously [22] on a 4.6 mm × 250 mm Partisil SAX column (Whatman) using a Merck-Hitachi LaChrom Elite HPLC system equipped with an Ramona Star solid scintillator radioflow detector (Raytest). The gradient used for elution of the HPLC column was made of buffer A which contained only distilled water and buffer B which contained 1 M (NH4)4HPO4 (pH 3.8). The gradient was run at 0% buffer B for 10 min, increased to 25% buffer B over 60 min and then increased to 100% buffer B over 20 min. The flow rate was 1 ml/min.


Whole-cell yeast extracts were prepared from oleic acid-induced cells as previously described [15]. GFP- and DsRed-tagged proteins were monitored by live cell imaging with a Zeiss Axioplan fluorescence microscope and AxioVision 4.1 software. The activity of catalase (EC was determined as described previously [23].


S. cerevisiae peroxisomes contain PIds

Peroxisomes that were isolated from rat liver harbour the PId kinase activities necessary for the formation of different PIds such as PtdIns4P, PtdIns(4,5)P2 and PtdIns(3,5)P2 [21]. These findings prompted us to analyse whether peroxisomes from S. cerevisiae also possess the enzymatic capability to synthesize PIds. As other endomembranes such as endosomes, vacuoles, multivesicular bodies and the Golgi apparatus maintain this enzyme activity [24], separation of peroxisomes from these organelles was a prerequisite for the present study. To this end, organelles were separated from a cell-free homogenate of oleic acid-induced wild-type cells by density gradient centrifugation, using a mixture of sucrose and Optiprep as the gradient mediums. After centrifugation, 23 fractions were collected and subjected to the detection of organellar markers by enzyme measurement and immunoblot analysis. The localization of the peroxisomal markers showed that catalase, thiolase (Fox3p), Pex11p and Pex13p peroxisomes peaked at the bottom of the gradient at a high density of 1.21 g/cm3, separated well from mitochondria, vacuoles and endosomes (Figure 1A). The presence of a smaller amount of thiolase in fractions with a lower density is probably due to its release from ruptured organelles. The minor amount of the peroxisomal membrane markers Pex11p and Pex13p that is detected in lighter fractions is supposed to reflect the position of immature peroxisomes. The mitochondrial porin migrated to fractions of medium density. Vacuolar ALP [25] and endosomal Pep12p [26] were both found in the upper fractions of the gradient. The ER (endoplasmic reticulum) was distributed over a range of fractions as indicated by the detection of Kar2p [16], but separated well from the peroxisomal fractions.

Figure 1 Peroxisomes have a PtdIns3P-synthesizing activity

Organelles from a cell-free homogenate of oleic acid-induced wild-type yeast cells were separated by density gradient centrifugation (2.25%–24% Optiprep, 18% sucrose) and the fractions were then subjected to (A) measurement of the activity of catalase used as peroxisomal marker or immunoblot analysis with antibodies against the following marker proteins: Fox3p, Pex11p and Pex13p (peroxisomes); Porin (mitochondria); Kar2p (ER); ALP (vacuole); Pep12p (late endosomes) and; Fbp1p (cytosol). (B) Organelles from selected fractions were analysed for PId-synthesizing activities by in vitro labelling with [γ-32P]ATP followed by TLC of the extracted PIds. (C) PIds synthesized in vitro by the peroxisomal fractions (fractions 3 and 4) were analysed further by HPLC (top panel). For the identification of the synthesized PIds, PtdIns3P (middle panel) or PtdIns4P (bottom panel) were added to the samples as external standards and analysed as above. In comparison with standard PIds, PtdIns3P was identified to be synthesized by peroxisomes.

To test their capability to synthesize PIds, the organelles of the gradient fractions were collected by centrifugation and subjected to [γ-32P]ATP labelling. Labelled phospholipids were extracted from the samples and analysed by TLC (Figure 1B). Most of the synthesized PIds appeared in the gradient fractions 19–21, those predominantly containing endosomal membranes (Figure 1B). Significant amounts of newly synthesized PIds were also observed in the fractions containing mitochondria and ER. Most interestingly, in the high-density peroxisomal fractions, PId synthesis was also detected (Figure 1B, lane 3). To identify the nature of these peroxisomal PIds, the labelled phospholipids were extracted, deacylated and separated by HPLC (Figure 1C, top panel). For clear identification, the samples were supplemented with PtdIns3P or PtdIns4P standards which were added prior to the HPLC (Figure 1C, middle and bottom panels). Comparison of the elution profiles of peroxisomal PIds with those of the external PId standards revealed that PtdIns3P is the major product generated by peroxisome labelling.

A portion of the lipid kinase Vps34p localizes to peroxisomes

Our results demonstrate that yeast peroxisomes possess the ability to synthesize PtdIns3P and thus are associated with a PId 3-kinase. The only PId 3-kinase known in S. cerevisiae is Vps34p, which has been shown to synthesize PtdIns3P at both the vacuolar and the endosomal membranes [27,28]. To investigate whether Vps34p is also responsible for the peroxisomal synthesis of PtdIns3P, the subcellular localization of the protein was analysed under peroxisome-inducing conditions. First, we tested for the co-localization of GFP-tagged Vps34p with peroxisomes by fluorescence microscopy (Figure 2A). GFP alone, or GFP fused to the N-terminal part of Vps34p, was co-expressed in wild-type cells together with PTS2–DsRed which served as a peroxisomal marker. GFP alone led to an overall cytosolic staining, whereas GFP–Vps34p exhibited a weak cytosolic staining but a clear punctuated pattern that co-localized with the PTS2–DsRed staining, demonstrating that the fusion protein localizes at least partially to peroxisomes.

Figure 2 Vps34p is associated with peroxisomes

Localization studies were performed with a GFP- or TEV–ProtA-tagged Vps34p. (A) Either plasmid-encoded GFP or GFP fused to Vps34p were co-expressed with the peroxisomal marker PTS2–DsRed in wild-type cells. Transformants were grown on oleic acid plates for 48 h and analysed by fluorescence microscopy. Non-tagged GFP exhibited an overall cytosolic labelling, whereas GFP–Vps34p led to a punctated staining pattern which to a significant extent co-localized with the peroxisomal marker PTS2–DsRed. Scale bar, 10 μm. (B) A cell-free homogenate of oleic acid-induced wild-type cells expressing plasmid encoded Vps34p–ProtA was loaded on to a linear a 2.25%–24% Optiprep gradient containing 18% sucrose. The fractions were subjected to immunoblot analysis with antibodies against the following marker proteins: Pcs60p (peroxisomes); Porin (mitochondria); Kar2p (ER); ALP (vacuole); Pep12p (late endosomes) and; Gas1p (plasma membrane). The result indicated that Vps34p was detected predominantly in the vacuolar and endosomal fractions; however, a significant portion was also localized to peroxisomes.

We then corroborated these observations by conducting subcellular fractionation studies. A cell-free homogenate of oleic acid-induced wild-type cells expressing a genomically tagged VPS34 gene coding for a Vps34p–ProtA fusion protein (Vps34p–TEV–ProtA) was prepared and organelles were separated by density gradient centrifugation (Figure 2B). As judged by immunoblot analysis of the obtained fractions, peroxisomes predominantly migrated to the bottom fractions as indicated by the presence of the peroxisomal marker protein Pcs60p. In contrast, a mitochondrial marker protein (porin) and an ER marker protein (Kar2p) were both found in the middle of the gradient, clearly separated from the peak peroxisomal fraction. Vacuolar ALP, endosomal Pep12p and Gas1p, a marker protein of the plasma membrane [17], remained in the top fractions of the gradient. Consistent with its well known cytosolic, vacuolar and endosomal localizations, the majority of Vps34p remained in the top gradient fractions. However, a significant portion of Vps34p was detected at a higher density, co-localizing with peroxisomal Pcs60p (Figure 2B, lane 3). Considered together, the localization studies demonstrate that a significant portion of Vps34 is associated with peroxisomes.

Vps34p is not required for peroxisome biogenesis

Vps34p controls several pathways in membrane trafficking and organelle formation. Thus a requirement for Vps34p in peroxisome formation was investigated. Cell homogenates were prepared from the vps34Δ strain and the organelles were separated by density gradient centrifugation (Figure 3A). The majority of the peroxisomal marker Pcs60p was detected in the high-density fractions, as it is typical for mature peroxisomes, indicating that the formation of peroxisomes seems not to be disturbed. Next, the vps34Δ-strain was analysed for the peroxisomal targeting of GFP fused to PTS1 (GFP–PTS1). Fluorescence microscopy analysis of oleic acid-induced live cells revealed a punctated staining pattern in the wild-type strain that is typical for peroxisomal labelling (Figure 3B). Expression of GFP–SKL in pex5Δ-cells, which have impaired peroxisomal protein import, exhibited a cytosolic fluorescence pattern typical for these cells. The fluorescence microscopy pattern observed for the vps34Δ cells was similar to that in the wild-type strain, suggesting that vps34Δ cells are still able to import peroxisomal matrix proteins.

Figure 3 vps34Δ cells are not affected in peroxisome biogenesis, but exhibit a defect in peroxisome function

(A) Peroxisomes of oleic acid-induced vps34Δ cells were isolated by density gradient fractionation as described for wild-type cells. The gradient fractions obtained were subjected to immunoblot analysis. Migration of the peroxisomal marker protein Pcs60p to high-density fractions demonstrated the presence of mature peroxisomes and indicated that deficiency in Vps34p did not severely affect peroxisome biogenesis. (B) The PTS1 marker protein GFP–PTS1 was transformed in wild-type, pex5Δ and vps34Δ-cells. The transformed strains were grown on oleic acid plates for 2 days and then examined by fluorescence microscopy. In contrast with pex5Δ cells that were impaired in PTS1-dependent matrix protein import and mislocalized the marker protein to the cytosol, both wild-type and vps34Δ exhibited a punctated GFP staining. These results demonstrated that the marker protein localized to peroxisomes, excluding defects in peroxisomal protein import. Scale bar 2.5 μm. (C) The indicated strains were spotted as a series of 10-fold dilutions on medium containing either glucose or oleic acid as the sole carbon source and incubated for 5 days at 26 or 30 °C as indicated. All strains were capable of utilizing glucose as the sole carbon source at 26 and 30 °C. On the oleic acid medium, although slower than the wild-type, vps34Δ still grew at 26 °C. At 30 °C vps34Δ exhibited a severe growth defect on the oleic acid medium, suggesting that the cells are affected in peroxisome function. (D) The indicated strains were grown on glucose and oleate and the growth time was adjusted to the different generation times. Whole-cell extracts were prepared and oleate induction was monitored by immunoblot analysis of Fox3p. Porin served as a loading control.

Peroxisomes are required for growth of S. cerevisiae under the condition of fatty acids being the single carbon source, but they are dispensable for growth on other fermentable and non-fermentable carbon sources. To test whether Vps34p is required for peroxisome function, the growth abilities of vps34Δ cells on either glucose or oleic acid as a carbon source were analysed. As the vps34-mutant is temperature-sensitive [29], growth tests were performed at different temperatures. All strains were found to be capable of utilizing glucose as the sole carbon source at both 26 °C and 30 °C (Figure 3C). In accordance with its sensitivity to temperature, vps34Δ exhibited significantly reduced growth at 37 °C (results not shown). When oleic acid was the sole carbon source, the wild-type cells grew at all temperatures. In contrast, the pex3Δ-strain, defective in peroxisome biogenesis, was unable to grow on this medium as expected for these cells [10]. At 26 °C, vps34Δ still grew on oleic acid medium although slower than the wild-type; however, at 30 °C, the cells exhibited a severe growth reduction suggesting that they are affected in peroxisome function. As peroxisomes were present, indicating that the biogenesis of the organelles seemed not to be affected, the peroxisomal enzyme composition was analysed in more detail. We monitored the level of the peroxisomal marker enzyme Fox3p by immunoblot analysis of lysates derived from wild-type vps34Δ and vps15Δ cultures. Fox3p was not detectable in cells grown in 2% glucose medium, which is in agreement with the fact that the FOX3 locus is glucose-repressed (Figure 3D). When tested on oleate medium, the samples derived from the vps34Δ- and vps15Δ-strains contained only a small amount of Fox3p compared with the wild-type strain (results not shown). However, we also noticed that the two deletion strains grew significantly slower than the wild-type. After adjustment of the generation time (wild-type, 14 h compared with deletion strains, 28 h), we found that the Fox3p level in vps34Δ and vps15Δ cells was comparable with the wild-type strain (Figure 3D). Immunoblotting of mitochondrial porin was also conducted to serve as a loading control. Taken together, these results indicate that Vps34p function is not essential for peroxisome biogenesis, but that it does have an effect on the growth rate of the cells.

Vps34p complex I is involved in degradation of peroxisomes

In S. cerevisiae, the level of peroxisomes and peroxisomal enzymes is adjusted in response to different nutritional sources. When cells are grown on oleic acid as the sole carbon source, proliferation of peroxisomes and synthesis of peroxisomal enzymes is induced. Conversely, a subsequent shift to glucose-rich or nitrogen-limiting conditions results in regulated peroxisome degradation (pexophagy) requiring the activity of vacuolar hydrolases [6]. Deletion of the putative Vps34p orthologue Pdd1p from Hansenula polymorpha affects peroxisome degradation and the sorting of the vacuolar proteinase HpCPY [30]. On the basis of the assumption that Vps34p may carry out a similar function, we analysed the selective degradation of peroxisomes in the presence and absence of Vps34p. Cells were grown on an oleic acid medium and subsequently shifted to glucose-rich medium to induce pexophagy. Samples were then taken at different time points, whole-cell lysates were prepared and analysed by immunoblot analysis. The degradation of peroxisomal thiolase within 24 h of the change of medium served as a marker for functional pexophagy. Under these conditions, the steady-state concentration of mitochondrial porin was not affected, demonstrating that the experimental design is suitable to study specifically the degradation of peroxisomes (Figure 4A). Pep4p is a vacuolar aspartyl protease and required for the post-translational precursor maturation of vacuolar proteinases [31]. Deletion of this protease severely affects the vacuolar degradation of Fox3p that is still visible after 24 h. Interestingly, in cells lacking Vps34p, a significant amount of thiolase was still present after 24 h on the glucose medium. These results demonstrate that peroxisomal degradation was slowed down and that Vps34p plays a role in pexophagy in S. cerevisiae. By contrast, the deletion of the gene encoding PtdIns3P 5-kinase Fab1p had no effect on pexophagy. This finding indicates that PtdIns3P is not required as an intermediate for the generation of PtdIns(3,5)P2, but is itself the crucial signalling lipid required for pexophagy.

Figure 4 Vps34p- and PtdIns3P-mediated signalling is required for pexophagy

(A) Strains were grown on an oleic acid medium for 16 h and subsequently shifted to a glucose-containing medium that induces pexophagy. After distinct periods samples were taken, whole-cell lysates were prepared and then subjected to SDS/PAGE and immunoblot analysis. Mitochondrial porin served as an internal control. Wild-type cells degraded Fox3p (thiolase) and thus peroxisomes were considered to be degraded via pexophagy within 24 h. vps34Δ cells showed a significantly reduced degradation rate, indicative of defective pexophagy. pep4Δ, known to have impaired pexophagy, served as control. (B) In order to analyse whether Vps34p-complex I (Vps15p/Vps30p/Atg14p) or complex II (Vps15p/Vps30p/Vps38p) is involved in pexophagy the corresponding deletion strains were tested for pexophagy as described in (A). (C) Plasmid-encoded EGFP–2xFYVE or EGFP–2xFYVE(C215S) were co-expressed with the peroxisomal marker PTS2–DsRed in wild-type cells. Transformants were grown on oleic acid plates for 48 h and analysed by fluorescence microscopy. The EGFP–2xFYVE strain displayed a punctated pattern which co-localized to a significant extent with the peroxisomal marker PTS2–DsRed, whereas the mutant form of the FYVE domain did not co-localize with the peroxisomal marker. Scale bar, 10 μm. (D) Pexophagy was monitored in cells constitutively expressing either EGFP–2xFYVE or EGFP–2xFYVE(C215S) as described previously. The expression of the FYVE domain interfered with the degradation of the peroxisomal marker protein Fox3p, whereas the mutated form still allowed pexophagy to occur.

In yeast, Vps34p forms part of two separate protein complexes [8]. Both complexes contain Vps15p and Vps30p(Atg6) associated with Vps34p. Complex I is connected functionally to autophagy via its specific component Atg14p. Complex II contains Vps38p instead of Atg14p and is required for the sorting of vacuolar proteins. Although the single deletions of Vps15p,Vps30p and Atg14p resulted in a block of pexophagic degradation, the deletion of Vps38p did not influence this process (Figure 4B). The finding that Atg14p, but not Vps38p, is required for the autophagic removal of peroxisomes indicates that the PtdIns3P generated by complex I is a requisite for this process.

We next analysed by which major mechanism Vps34p may regulate pexophagy via its product PtdIns3P. In principle, accumulation of PtdIns3P itself could be required to define a certain membrane topology (discussed in [32]). The second possibility is that it recruits a subset of PtdIns3P-binding proteins. Several of such PtdIns3P-effector proteins have been described in different species. Most of these effector proteins interact with PtdIns3P via the FYVE zinc finger domain [33] or via the PX (Phox-homology) domain [34,35]. We transformed cells expressing PTS2–DsRed with either EGFP–2×FYVE or EGFP–2×FYVE(C215S) and monitored their localization in oleate-grown cells by fluorescence microscopy (Figure 4C). The EGFP–2×FYVE expression generated a punctated pattern. These signals co-localized in most cases with the peroxisomal marker PTS2–DsRed. In marked contrast, the EGFP–2×FYVE(C215S) mutant, which has a reduced capacity to bind PtdIns3P [13,33], displayed no co-localization with the peroxisomal marker, demonstrating the specificity of the previously observed association. This provides further evidence, using an alternative method, for the specific localization of PtdIns3P to peroxisomes in oleate-induced cells and thus adds weight to the data presented in Figure 1. In addition, the specific localization of the FYVE probe to PtdIns3P-containing peroxisomal structures indicates that the peroxisomal PtdIns3P pool is accessible to PtdIns3P-binding-module-containing proteins. It has been described that overexpression of the FYVE domain is capable of disrupting PtdIns3P-mediated functions, most likely by competing with PtdIns3P-binding proteins [36]. In order to investigate whether PtdIns3P itself or if PtdIns3P-effector-mediated signalling is required for pexophagy, we overexpressed the double FYVE domain or alternatively the corresponding C215S mutant (Figure 4D). Similar to the assay described above, the oleate-grown cells were shifted to glucose-medium to induce pexophagy. The overexpression of the FYVE domain inhibited the degradation of Fox3p, whereas the FYVE(C215S) mutant had no decrease in Fox3p level. These data indicate that Vps34p regulates pexophagy via PtdIns3P effectors that are probably directed to the peroxisomal pool of PtdIns3P.

Identification of downstream effectors of PtdIns3P

The proteome of S. cerevisiae contains five proteins harbouring a FYVE domain which binds exclusively to PtdIns3P. Furthermore, yeast has 15 proteins containing a PX domain which enables the recognition of PtdIns3P as well as other types of PtdInsP. Eight other proteins have been reported to interact with PtdIns3P via an non-canonical interaction motive [37]. To identify the downstream effectors of PtdIns3P, we screened the corresponding 28 deletion strains for a defect in pexophagy (Figure 5). We found that 11 of the deletion strains displayed a delayed degradation of Fox3p or a near total block of Fox3p breakdown.

Figure 5 A screen for PtdIns3P-binding proteins involved in pexophagy

The indicated strains were grown on oleic acid-containing medium for 16 h and subsequently shifted to a glucose-containing medium to induce pexophagy. After different periods of time, samples were taken, whole-cell lysates were prepared and subjected to SDS/PAGE and immunoblot analysis. Mitochondrial porin served as an internal control. The degradation of Fox3p (thiolase) served as a marker for pexophagy, which in wild-type cells led to the disposal of peroxisomes within 24 h. The deletion strains were grouped into proteins with (A) a FYVE domain, (B) a PX domain or (C) proteins with a non-canonical-binding motif.

Results reported previously have established a role for the S. cerevisiae Atg proteins Atg18p, Atg20p, Atg24p and Atg27p in pexophagy [8], whereas we show in the present paper for the first time that the homologue of the H. polymorpha Vam7p [38], a vacuolar SNARE (soluble N-ethylmaleimide-sensitive fusion protein-attachment protein receptor) complex protein, is essential for this process in S. cerevisiae. An equally novel finding is the principal connection of six other proteins to the autophagic degradation of peroxisomes. The deletion strains bem3Δ, mvp1Δ and vps27Δ had a slightly delayed degradation of Fox3p, whereas the deletion strains arc1Δ, bem1Δ and pep7Δ showed a significantly reduced breakdown of Fox3p (Figure 5). Pep7p, Vps27p and Mvp1 are connected to membrane traffic and protein sorting between different cellular compartments, whereas Arc1p is a tRNA-binding protein. Bem1p, as well as Bem3p, is connected to Cdc42p-mediated signalling.

The functionalities of these proteins may give the first indications of a connection, either directly or indirectly, in the involvement of these processes with the autophagic degradation of peroxisomes.


In the present paper we have reported the detection of a PtdIns3P-synthesizing activity associated with peroxisomes from S. cerevisiae and the identification of Vps34p as the responsible kinase. We show that Vps34p is associated with peroxisomes under inducing conditions and provide evidence that the protein is vital for the controlled degradation of peroxisomes by pexophagy.

Previous findings have demonstrated that peroxisomal membranes contain PIds [21,39]. In rat liver peroxisomes, PtdIns is used as a substrate and converted into PtdIns4P, PtdIns(3,5)P2 and PtdIns(4,5)P2 [21]. However, although some PtdIns3P was detected in vivo in peroxisomes, it was barely detected in vitro, raising the question of whether peroxisomes themselves are able to perform the enzyme activity required for its synthesis or not. In the present paper, we identified the synthesis of PtdIns3P at the peroxisomal membrane of S. cerevisiae (Figure 1). In yeast, Vps34p is the only known kinase which phosphorylates PtdIns at the 3′ hydroxy position [40]. Vps34p and its orthologues from different species have been implicated in the control of important pathways, including the endosome to lysosome transport, multivesicular body formation and autophagy. Accordingly, the enzymes and their products have been localized to endosomal and vacuolar membranes and were detected on PAS (pre-autophagosomal structures) [8]. Although it has been known that Vps34p is associated with autophagic and endosomal membrane systems [41], we have now demonstrated that a significant portion of Vps34p is also associated with peroxisomes (Figure 2). The reason why peroxisomal localization has not been observed previously is probably owing to the fact that previous localization studies were performed with glucose-grown cells possessing peroxisomes in their non-induced state. This property is not unique for Vps34p. For example, the small G-protein Rho1p is predominantly tethered to the membranes of the secretory pathway, but is specifically recruited to peroxisomes upon their induction [42]. Interestingly, besides to endosomes, vacuoles and PAS, Vps34p was also localized to dot-like structures considered to be representing a novel undefined compartment [43] that, in the light of our data, might represent peroxisomes.

Interestingly, we found Vps34p to localize to peroxisomes during the biogenesis of peroxisomes, raising the question of whether Vps34p is involved in this process. However, the GFP–PTS1 cargo could still be imported efficiently into the peroxisomal matrix in the vps34Δ background. Although the growth of the vps34Δ strain was significantly slower on oleate plates compared with the wild-type strain, the deletion of VPS34 proved not to be essential for peroxisome biogenesis. One explanation for the slower growth on oleate could be that Vps34p is involved in the derepression of oleate-inducible genes [44]. This result was obtained in a screen for mutant strains that were affected in glucose repression, glycerol derepression or oleate induction of the POT1 locus. Using FACS to monitor the expression level of Pot1p–GFP, [44] found that the deletion of VPS34 had a significant effect on the transition from the glucose-repressed state to either a glycerol-derepressed or a oleate-induced state as judged by a reduced signal intensity of the Pot1p–GFP probe. We obtained similar results when we monitored the level of Fox3p by immunoblotting. Although Fox3p was strongly induced in wild-type cells after a shift from a glucose to a oleate medium, the vps34Δ strain contained only little Fox3p. However, we also noted that the cells themselves grew with an approx. 50% slower rate than the wild-type cells. As the amount of Fox3p was similar in all strains after adjustment of the generation time, we concluded that Vps34p may not be directly involved in the derepression of oleate-inducible genes, but that it seems to have a more pleiotropic effect on cell growth and the rate of organellar biogenesis. In addition, the fact that the vps34Δ strain also grows slower on glucose medium points to a more pleiotropic effect.

Originally, VPS34 was discovered as a gene required for the sorting of CPY (carboxypeptidase Y) to the vacuole [29]. In accordance with this function, a number of phenotypes were described that were related to vacuolar malfunctions. These include temperature-sensitive growth, disturbed vacuolar segregation and the missorting of a number of soluble vacuolar proteinases resulting in impaired vacuolar degradation. Experiments designed to isolate mutants affected specifically in peroxisome degradation in H. polymorpha identified Pdd1p, a putative orthologue of Vps34p [30]. The pdd1 mutant is blocked in the initial sequestration of peroxisome degradation during either glucose or ethanol adaptation. We show that in S. cerevisiae Vps34p is required for the regulated degradation of peroxisomes by pexophagy, as demonstrated by the reduced disappearance of the peroxisomal marker protein Fox3p (thiolase) in vps34Δ cells after shifting from oleate induction to a glucose medium. (Figure 4A). In addition, we found that the Vps34p complex I, which contains the autophagy-specific component Atg14p, is essential for pexophagy, whereas the Vps38p-containing complex II is dispensable. This finding is novel, although it should be noted that a recent publication in Pichia pastoris showed that the strain lacking the Vps38-orthologue Uvrag is not affected in pexophagy [45].

In the present study, we performed a screen for downstream effectors of Vps34p function (Figure 5). The PtdIns3P-binding Atg proteins Atg18p, Atg20p, Atg24p and Atg27p of S. cerevisiae have been reported previously to be linked with pexophagy [8]. Considering the species-specific differences of the pexophagy machinery, we confirmed that Atg21p, which is essential for pexophagy in H. polymorpha [46], is dispensable for this process in S. cerevisiae, as previously reported [47]. In this context, it is interesting to note that the H. polymorpha homologue of Vam7p, which is part of the vacuolar SNARE complex [38], is essential for pexophagy in S. cerevisiae. In addition, we showed that some deletion strains are affected in pexophagy via proteins that have not been linked before to this process or even autophagy-related pathways in general.

The deletion strains lacking Vps27p or Mvp1p exhibit a mild defect in pexophagy by displaying a slightly delayed degradation of Fox3p. This might indicate that their association with pexophagy is the modulation of this process or the association is indirect by affecting more general sorting pathways. Vps27p is part of the endosomal ESCRT (endosomal sorting complex required for transport) machinery that is required for the biogenesis of multivesicular bodies and the sorting of ubiquitinated proteins to the vacuole, but is also involved in the mainte-nance of the vacuolar morphology. It has been shown that the delivery of certain hydrolases to the vacuole is perturbed in yeast ESCRT mutants [48]. Mvp1p is an SH3 domain-containing protein that acts in concert with the dynamin-like GTPase Vps1p to promote membrane traffic to the vacuole [49]. Vps1p has also been shown to regulate peroxisome abundance, inheritance and also fission [50,51]. The dynamin-like GTPase Dnm1p is part of the peroxisomal, as well as mitochondrial, fission machinery [51,52] and it is interesting to note that dnm1Δ cells are affected in mitophagy [53].

The mutant strains lacking Pep7p, Arc1p, Bem1p or Bem3p exhibit a severe defect in pexophagy, indicating that these proteins might play an important role in this process. Pep7p is a multivalent adaptor protein involved in vesicle-mediated vacuolar protein sorting and vacuolar inheritance [54]. Another possible connection to autophagic processes is its interaction to the t-SNARE Pep12p [55], which is required for selective autophagy of mitochondria [53].

Arc1p has been found to bind to tRNA and is supposed to function as a cofactor of methyl- and glutamyl-tRNA synthetases [56]. It is interesting to note that Arc1p has been suggested to be the target of the serine/threonine kinase Atg1p [57], which is required for the early steps of autophagic processes including pexophagy [58].

Work on Bem1p has focused on its role as scaffold protein for complexes of the essential GTPase Cdc42p with its GEF (guanine-nucleotide-exchange factor) Cdc24p and their functional role in cell polarity, assembly of the bud emergence site and the reorganization of the actin cytoskeleton [59]. Additional data indicate a role of Bem1p as a positive regulator of homotypic fusion of yeast vacuoles [60]. Other possible links to autophagic degradation of peroxisomes might be indicated by association of Bem1p with actin, Vac8p and Vam7p [60], because these proteins are also required for pexophagy [6]. Bem3p is another interaction partner of Bem1p, and bem3Δ cells are also affected in pexophagy. Bem3p is a GAP (GTPase-activating protein) for Cdc42p, hinting to a second possible link between Cdc42p complexes and the autophagic degradation of peroxisomes.

Clearly, further work is required to fully analyse the functional connection of the tested proteins to selective autophagy of peroxisomes. It will be important to elucidate which autophagy-related pathways and membrane trafficking pathways are interconnected by these proteins and thus at which step, be it specific or indirect, they influence pexophagy.

The mechanisms by which Vps34p and its product PtdIns3P regulate pexophagy are still incompletely understood. In line with the finding that Vps34p function is required at an early stage in autophagy, yeast cells affected in VPS34 seem to block pexophagy at very early stage as judged by electron microscopy [30]. It has been reported previously that PtdIns3P is present at the PAS as well as on the vertex and boundary of the pexophagosome/vacuole fusion site [61] during pexophagy in P. pastoris. The finding that Vps34p and PtdIns3P do not only localize to autophagic membranes, but also to the potential target of selective autophagy raises the possibility that formation of the initiation may start at the target itself.

Our data are clear in that yeast peroxisomes functionally recruit Vps34p leading to the formation of PtdIns3P, which is required for pexophagy. However, the fact that rat liver peroxisomes besides PtdIns3P synthesize other distinct PIds raises the question of whether PtdIns3P is the only PId synthesized in the yeast peroxisomal membrane. The discovery that Fab1p, a PtdIns3P 5-kinase, is required for growth on oleic acid [62] opens up the possibility that peroxisomal PtdIns3P might be processed further. However, our data show that Fab1p is not required for pexophagy (Figure 4A), indicating that PIds might play a more complex role in the regulation of peroxisome function and biogenesis.


Silke Grunau carried out organellar purification, localization studies of Vp34p and the growth test. TLC as well as HPLC was performed by Dorothee Lay, and pexophagy experiments were performed by Sabrina Mindthoff and Harald Platta. Wilhelm Just and Ralf Erdmann contributed to the experimental designs and data analysis. Wolfgang Girzalsky, Harald Platta and Ralf Erdmann wrote the manuscript.


H. W. P. was supported by the EMBO Longterm Fellowship [grant number ALTF11902008]. This work was supported by the Deutsche Forschungsgemeinschaft (SFB642) (ATP- and GTP-dependent steps in the biogenasis of peroxisomes) and the Fonds of the Ruhr-Universität Bochum.


We thank Ulrike Freimann for technical assistance.

Abbreviations: ALP, alkaline phosphatase; EGFP, enhanced GFP; ER, endoplasmic reticulum; ESCRT, endosomal sorting complex required for transport; GFP, green fluorscent protein; PAS, pre-autophagosomal structures; PId, phosphoinositide; ProtA, Protein A; PTS, peroxisomal targeting signal; PX, Phox-homology; SNARE, soluble N-ethylmaleimide-sensitive fusion protein-attachment protein receptor


View Abstract