Pex9p is a new yeast peroxisomal import receptor for PTS1-containing proteins

Peroxisomes are ubiquitous cell organelles with high metabolic versatility. The most-widespread functions include decomposition of oxygen species and of other toxic metabolites, as well as breakdown of lipid molecules. The organellar enzyme composition differs between species and between tissue-specific cells, as well as within unicellular populations, as a response to a change of environmental or developmental conditions (Kunau, 2006).

The diversity and change of enzymatic content is facilitated by conserved and efficient import machinery, enabling directional targeting of folded and oligomerized proteins into the lumen of peroxisomes (reviewed in Kim and Hettema, 2015; Liu et al., 2012; Schliebs et al., 2010). Important determinants of the import cascade are (i) peroxisomal targeting signals (PTSs) exposed by peroxisomal matrix proteins, (ii) cycling import receptors and (iii) multimeric import machinery at the peroxisomal membrane that comprises numerous peroxins (Distel et al., 1996), which import folded proteins.

The best-characterized PTSs comprise a tripeptide – SKL – or a conserved variant at the C-terminus of peroxisomal proteins [type I PTS (PTS1)] (Gould et al., 1989, 1987), or a nonapeptide with the consensus sequence (R/K-L/I/V-x₅-H/Q-L/A/F, where x represents any amino acid), which is located near the N-terminus [type 2 PTS (PTS2)] (de Hoop and Ab, 1992; Swinkels et al., 1991). PTS1 and PTS2 are recognized in the cytosol by the cognate cytosolic receptors Pex5p and Pex7p, respectively (Marzioch et al., 1994; Van der Leij et al., 1993).

The majority of peroxisomal enzymes use the Pex5p-dependent pathway. Most of these contain the canonical PTS1, but a few ‘hook’ onto this pathway through non-PTS signals (Schäfer et al., 2004), or they bind to other PTS1 cargo proteins, resulting in so-called ‘piggy-back transport’ (Yang et al., 2001). The receptor–cargo complex docks at the peroxisomal membrane through interaction with either Pex14p or Pex13p (Albertini et al., 1997; Bottger et al., 2000). Subsequently, Pex5p together with Pex14p forms a transient pore through the membrane, which allows traversal of oligomeric cargo complexes up to a size of 9 nm (Meinecke et al., 2010). After release of cargo–protein into the matrix, Pex5p is released from the membrane into the cytosol, where it is available for further rounds of import. Receptor recycling involves monoubiquitylation of Pex5p, the concerted action of two AAA (ATPases associated with various cellular activities) peroxins – Pex1p and Pex6p – as well as deubiquitylation of the receptor (Okumoto et al., 2014; Platta et al., 2008, 2007).

In contrast to Pex5p, the PTS2-receptor Pex7p is unable to target cargo proteins to the peroxisomal membrane autonomously (Schliebs and Kunau, 2006). In mammalian cells, PTS1 and PTS2 pathways converge in the cytosol, where Pex7p binds to a splice variant of Pex5p (Braverman et al., 1998; Mukai and Fujiki, 2006). In yeast cells, both pathways are clearly separated using distinct translocation pores (Montilla-Martinez et al., 2015). In yeast, cargo-loaded Pex7p forms a heteromeric complex with one of two cytosolic co-receptors, either Pex18p or Pex21p (Grunau et al., 2009; Purdue et al., 1998). These are paralogs with redundant function with respect to peroxisomal import of the β-oxidation enzyme Fox3p, the predominant PTS2-containing protein in cells that have been grown on fatty-acid-containing medium. Recently, it has been shown that Pex21p, but not Pex18p, is required for the peroxisomal import of Gpd1p, a PTS2-containing protein that is induced under osmotic stress (Effelsberg et al., 2015).

Here, we report the identification of a second PTS1 receptor in Saccharomyces cerevisiae. In contrast to Pex5p, which serves as a general import receptor for PTS1-containing proteins, the paralog with the systematic name Ymr018wp, now designated Pex9p, provides specificity for a subset of peroxisomal enzymes under strictly defined growth conditions. The differentially regulated import pathways underscore the role of peroxisomes as multi-purpose organelles that contribute to the adaptation of baker's yeast to different environmental conditions.

The hypothetical open reading frame YMR018W of Saccharomyces cerevisiae codes for a protein that shares high sequence similarity with the PTS1 receptor Pex5p (Amery et al., 2001) (Fig. 1). Here, we report on a role for the gene product of YMR018W in peroxisome biogenesis. In a parallel and independent approach, Ymr018wp was found to have the same function (Yifrach et al., 2016); hence, we jointly named this new peroxin Pex9p.

Phylogeny analysis of yeast tetratricopeptide repeat (TPR)-containing proteins revealed that Pex9p is the closest relative of Pex5p (Fig. S1). The overall sequence similarity is 27% but mostly comprises the C-terminal part, which contains at least six conserved TPR-domains in both proteins (Fig. 1A). Common sequence patterns between Pex5p and Pex9p are also present within their N-terminal halves. Strikingly, both proteins contain di-aromatic pentapeptide motifs, which are known to mediate binding to Pex14p and to Pex13p, membrane-bound peroxins of the peroxisomal docking complex (Fig. 1B) (Douangamath et al., 2002; Kerssen et al., 2006; Pires et al., 2003). In addition, the functionally relevant and highly conserved cysteine residue at position 6 is also present in Pex9p. In Pex5p, this cysteine (Cys6) is monoubiquitylated during the import cycle, which is a crucial step for recycling of the receptor (Williams et al., 2007).

In order to investigate whether Pex9p and Pex5p share not only structural but also functional similarities, we tested the interaction between Pex9p and potential peroxisomal binding partners by performing yeast two-hybrid analyses (Fig. 2). Unlike Pex5p, full-length Pex9p did not exhibit auto-activation when fused to the DNA-binding domain (Fig. 2A). Pex9p did not exhibit interactions with the peroxins Pex1p, Pex3p through to Pex8p, Pex10p, Pex11p, Pex12p, Pex15p, Pex17p, Pex18p, Pex19p and Pex21p (data not shown). Also, Pex13p, which strongly interacts with Pex5p through one of its WxxxF motifs (where x is any amino acid) (Douangamath et al., 2002; Pires et al., 2003), failed to bind to Pex9p (Fig. 2A). However, interaction of Pex9p with the second PTS1-receptor docking protein Pex14p was clearly observed (Fig. 2A). To map the Pex14p-binding sites, we used N- and C-terminal truncated versions of Pex9p. An N-terminal fragment comprising 111 amino acids still facilitated strong binding to Pex14p. A fragment containing the N-terminal 90 amino acids did not interact. The most striking structural feature of the 21-amino-acid sequence that distinguished the two fragments is the presence of one WxxxF motif.

Further two-hybrid interactions of Pex9p were detected with the two malate synthases of the yeast – Mls1p and Mls2p (Fig. 2B). The isoenzymes share an overall sequence identity of 81%, and both contain at their C-termini the putative PTS1 sequence SKL. The interaction also occurs with an N-terminally truncated version of Pex9p (Fig. 2B; amino acids 175–514). This construct still contains the TPR domains, which, in Pex5p, are known to form a PTS1-binding domain (Gatto et al., 2000). Interestingly, another peroxisomal protein, Pcs60p, which also harbors the C-terminal targeting sequence SKL and which is known to interact with Pex5p (Hagen et al., 2015), did not interact with Pex9p. In addition, no interaction between Fox1p, a Pex5p-dependent non-PTS cargo protein, and Pex9p could be detected.

To analyze the expression of Pex9p under control of its own promoter, the protein was genomically tagged with Protein A. Immunoblot analysis revealed that the steady-state level of Pex9p significantly increased during growth on oleate-containing medium (Fig. 3). This is in accordance with results obtained by using a high-throughput approach (Smith et al., 2002), which suggests oleate-inducible transcription of PEX9. The tight regulation by oleate is also indicated by the fact that Pex9p was not detected in significant amounts in oleate-free media (Fig. 3A). This was seen during growth on glucose as the sole carbon source as well as during growth on the non-fermentable carbon sources like ethanol (Fig. 3A). Under these conditions, Pex9p as well as Fox3p were repressed, whereas Pex5p and both the malate synthases Mls1p and Mls2p were abundant (Fig. 3A). In addition, expression of PEX18 seems to be determined by the presence of predicted oleate-response elements (OREs) upstream of the gene start codon (Fig. S2). This sequence pattern is absent in the PEX21 gene.

Our analysis also shows that both Pex5p as well as the potential PTS1-receptor Pex9p are expressed in cells growing on oleate medium. However, Pex9p cannot compensate for the lack of Pex5p function under these growth conditions (Fig. 3B). The data indicate that Pex9p could function as an alternative import receptor for a subset of PTS1 enzymes, not including oleate-converting enzymes like Fox1p.

Subcellular fractionation studies of yeast cells expressing Protein-A-fused Pex9p revealed a partial peroxisomal localization of the putative receptor. In OptiPrep™–sucrose density gradients of homogenates from oleate-induced cells, the majority of Pex9p co-migrated with the peroxisomal marker enzyme Fox3p and the membrane-bound Pex5p at fractions of high density (fraction 3), which is characteristic for peroxisomes (Fig. 4A). Small amounts of Pex9p were found in fractions 15–17 of the gradient, which typically contain mitochondria and endoplasmic reticulum (ER), and in less-dense fractions at the top of the gradient, which is typical for cytosolic proteins such as Pgk1p. A similar sedimentation distribution was observed for Pex5p (Fig. 4A; Bottger et al., 2000; Schäfer et al., 2004). The peroxisome association of Pex9p was corroborated by performing fluorescence microscopy analysis, showing colocalization of GFP-tagged Pex9p and the peroxisomal marker mCherry–Ant1p (Fig. 4B).

In order to determine the subperoxisomal localization of Pex9p, organellar fractions that had been isolated from spheroplasts of cells expressing Pex9p–Protein-A were subjected to chemical extraction and protease protection assays. Pex9p was resistant to both low-salt and high-salt extraction but, in contrast to the typical integral membrane protein Pex10p, was released from the membranes under alkaline conditions (Fig. 4C). However, as for Pex5p, a fraction of Pex9p was retained in the sediment after carbonate-mediated extraction (Fig. 4C), suggesting integration into the peroxisomal membrane. The embedding of Pex9p into the peroxisomal membrane was further supported by its partial resistance against externally added protease (Fig. 4D). Therefore, the majority of peroxisomal Pex9p is peripherally attached to the peroxisomal membrane, which reflects the situation for Pex5p (Kerssen et al., 2006).

Both the malate synthases Mls1p and Mls2p contain a classic PTS1 sequence, SKL, at their C-termini, and for Mls1p, peroxisomal localization has been described. Interestingly, Mls1p is localized in the cytosol under glucose-mediated repression of gene expression and associates with peroxisomes during growth in the presence of oleate (Kunze et al., 2002). Subcellular fractionation studies using isopycnic density gradient centrifugation as well as fluorescence microscopy localization showed that Mls1p and Mls2p were localized in peroxisomes in the presence of oleate (Fig. 5A). Both proteins were predominantly located in mature peroxisomes (fractions 3 and 4), with small amounts located in cytosolic fractions (fractions 25–27) and in particulate fractions (fractions 11–19), which also contained ER- and mitochondrial marker proteins. These fractions also contained small amounts of other peroxisomal proteins, including Fox1p, Fox3p, Mdh3p, Pex5p and Pex14p, which identifies the associated organelles as a putative subpopulation of peroxisomes that have lower densities.

Subcellular fractionation of Δpex9 cells revealed a similar distribution of marker enzymes as for those in wild-type cells. In accordance with the ability of the Δpex9 cells to use oleate as the sole carbon source, all peroxisomal marker enzymes – including both malate synthases – localized to the high-density peroxisomal fractions (Fig. 5A). This result indicates that Pex5p has the ability to target both malate synthases to peroxisomes.

Δpex5 yeast cells are characterized by a defect in the import of peroxisomal PTS1-containing proteins. This leads to the absence of mature high-density peroxisomes, but the mutant cells still contain peroxisomal membrane remnants that are also referred to as ‘ghosts’, which are still import-competent with respect to PTS2-containing proteins (Hettema et al., 2000). Analysis of the Δpex5 mutant revealed the expected import defects for the PTS1-containing proteins Mdh3p and Pcs60p, and for the non-PTS protein Fox1p, but Mls1p and Mls2p showed co-segregation with the peroxisomal marker protein Pex14p and the PTS2-containing protein Fox3p (fractions 11–19). This suggests that import of these proteins into peroxisomal remnants can still take place in the absence of Pex5p. This finding is corroborated by fluorescence microscopy analysis, which showed a punctate pattern of the GFP-fused Mls1p that was largely congruent with the pattern of the peroxisomal membrane marker mCherry–Ant1p.

Remarkably, Δpex5Δpex9 cells showed a clear mislocalization of both Mls1p and Mls2p to the cytosol when compared with single-mutant Δpex5 cells. This indicates that Pex9p can mediate the peroxisomal import of both malate synthases in oleate-grown cells.

In order to validate the contribution of Pex9p to the import efficiency of both malate synthases in the presence of Pex5p, differential centrifugation analyses of oleate-induced wild-type, Δpex9, Δpex5 and Δpex5Δpex9 cells were compared (Fig. 5C). The presence of the proteins in the organellar sediment was estimated by performing immunoblot analysis as a measure for their import efficiency. The highest import rates were observed for cells containing both receptors, whereas no import could be detected in the double-deletion strain. The negligible amounts of both malate synthases detected in the organellar fractions of Δpex5Δpex9 cells as well as in fractions 13–17 of the density gradient analysis (Fig. 5A) are probably due to aggregation or association with other cellular organelles. This is also suggested by the peroxisome-free Pex3p-deficient cells, which also showed small amounts of particulate malate synthases. However, cells that lacked either Pex9p or Pex5p exhibited reduced import levels when compared with those of wild-type cells.

Pex9p is a new peroxin that is involved in peroxisomal import of a subset of matrix proteins. The new peroxin has been designated Pex9p. This gene symbol had previously been incorrectly assigned and was now orphan (Distel et al., 1996; Kiel et al., 2006). Sequence and structural similarities between Pex5p and Pex9p suggest that the genes represent paralogs that evolved through gene duplication, probably leading to the sharing of similar functions and interaction partners but to different expression profiles (Guan et al., 2007).

Pex5p recognizes its cargo proteins in the cytosol, and the receptor–cargo complex docks at the peroxisomal membrane through interaction with either Pex14p or Pex13p (Hasan et al., 2013). Subsequently, Pex5p together with Pex14p forms a transient import pore, which allows traversal of folded proteins across the membrane (Meinecke et al., 2010). Pex9p interacts with the receptor docking protein Pex14p through a region in the N-terminal domain that contains a WxxxF motif – a conserved motif that is known to act as the binding interface for Pex5p (Saidowsky et al., 2001). This finding supports the view that Pex5p and Pex9p bind to the same domain of Pex14p, most likely to the N-terminal WxxxF-binding domain. As reported for Pex5p (Kerssen et al., 2006), the majority of Pex9p is localized in the cytosol, but a fraction of the protein is associated with the peroxisomal membrane. The partial protection against proteases and limited resistance against alkaline treatment suggests that a fraction of Pex9p, despite the lack of predictable transmembrane-spanning regions, becomes integrated into the peroxisomal membrane. This might indicate a dynamic topology throughout the peroxisomal protein import cycle, which has also been reported for the PTS1-receptor Pex5p (Kerssen et al., 2006). Receptor recycling involves monoubiquitylation of Pex5p at a conserved cysteine residue, which contributes to the ATP-dependent release of the receptor from the membrane (Platta et al., 2014). This cysteine is conserved in Pex9p (Fig. 1), suggesting that Pex9p might also be ubiquitylated during the import process. Taken together, the current evidence suggests that also Pex9p performs a similar import cycle to that of Pex5p, including cargo recognition in the cytosol, Pex14p-mediated docking of the cargo-loaded receptor to the membrane and contribution to cargo translocation. Then, Pex9p is thought to be ubiquitylated at the conserved cysteine residue and released from the membrane in an ATP-dependent manner by the two AAA-peroxins Pex1p and Pex6p. This, however, needs to be experimentally tested. This also accounts for the question of whether Pex9p contributes to pore formation, as Pex5p does, and is deubiquitylated, like Pex5p, so that it becomes available in the cytosol for another round of import.

Pex9p interacts with the cargos Mls1p and Mls2p but not with the well-known Pex5p cargo Pcs60p. Non-binding of a PTS1 peptide to the TPR-domains of Pex9p has been reported previously (Amery et al., 2001). Therefore, it is possible that additional contact sites, besides the PTS1 sequence, provide the high specificity for Mls1p and Mls2p. By contrast, the non-PTS cargo protein Fox1p interacts with amino acids 239–300 within the N-terminal half of the PTS1-receptor Pex5p (Klein et al., 2002). Within these regions, the sequence similarities of Pex9p and Pex5p are very low (Fig. 1A). The deletion of the PEX9 gene does not lead to a defect in β-oxidation because some of the enzymes involved in this pathway, such as Fox1p, are imported only through Pex5p.

Unlike Pex5p, Pex9p is not expressed in glucose- or ethanol-grown cells, but it is strongly induced by oleate. Under these conditions, Pex9p acts as a cytosolic and membrane-bound peroxisome import receptor for both malate synthase isoenzymes, Mls1p and Mls2p. Oleate-mediated induction and catabolite repression is governed by well-conserved oleate- and glucose-responsive promoter elements that are typically found in promoter regions of genes encoding peroxisomal enzymes (Einerhand et al., 1995, 1993; Kos et al., 1995). Indeed, OREs, as well as elements involved in glucose-mediated repression, like cis-acting sites for binding of ABF1 and RPA (URS1) (Einerhand et al., 1995), are present in the promoter regions of PEX9 (Fig. S2). In accordance with its abundance in cells grown on glucose-containing medium, elements for glucose-mediated repression were not found in the promoters of PEX5 (Fig. S2). This difference in expression of Pex5p and Pex9p is similar to the expression profiles of the two co-receptors of the PTS2-dependent import pathways. Whereas Pex21p expression seems to be almost constitutive, and thereby resembles the expression profile of Pex5p, Pex18p as well as Pex9p are induced by the presence of oleate (Effelsberg et al., 2015; Purdue et al., 1998).

Pex9p can bind and import both the malate synthases Mls1p and Mls2p into peroxisomes. Although Pex5p can also import both enzymes in oleate-grown cells, the single deletion of the PEX9 gene causes a reduction in the import of Mls1p and Mls2p, even in the presence of Pex5p. This suggests that the fast rate of malate synthase import achieved in the presence of oleate is the sum of the activities of both PTS1 receptors.

It has been shown previously that in cells grown on ethanol as the sole carbon source, Mls1p is primarily enzymatically active in the cytosol, whereas in cells shifted to oleate-containing medium, the same enzyme localizes in peroxisomes (Kunze et al., 2002). This is now explained by the oleate-induced expression of Pex9p, which increases the efficiency of peroxisomal targeting of malate synthases (Fig. 6). In fact, on ethanol, acetyl-CoA is produced in the cytosol, explaining the presence of the acetyl-CoA-consuming malate synthase activities in the cytosol. On oleate, acetyl-CoA is predominantly produced in peroxisomes. Accordingly, under these conditions, Pex9p triggers the efficient transport of the malate synthases from the cytosol to the site of acetyl-CoA production in peroxisomes. Similar control mechanisms of protein import into peroxisomes have been reported for PTS2-containing proteins. Pex21p, the constitutively expressed co-receptor of the PTS2 import pathway, is able to target Fox3p. However, it is unable to import all of the newly expressed Fox3p in the presence of oleate (Purdue et al., 1998), which requires the alternative co-receptor Pex18p (Fig. S3) (Effelsberg et al., 2015; Purdue et al., 1998).

In summary, four pathways for import of peroxisomal matrix enzymes exist in baker's yeast: two general pathways that depend on the constitutively expressed PTS1- and PTS2-receptors Pex5p and Pex21p, respectively, and two oleate-inducible pathways, requiring the PTS1- and PTS2-receptors Pex9p and Pex18p, respectively (Fig. 6; Fig. S3). This regulation of peroxisomal import contributes to the adaptation of peroxisomal metabolism to environmental changes and underlines the role of peroxisomes as ‘multi-purpose organelles’, which adapt their metabolic functions according to cellular needs. The adaptation of protein import of a cellular organelle in general, and peroxisomes in particular, to different environmental conditions by expressing distinct receptors is a novel addition to our understanding of metabolic control and to the concept of regulated protein targeting.

S. cerevisiae strains used in this study are listed in Table S1. All strains were grown at 30°C. YPD medium contained 2% glucose, 2% peptone and 1% yeast extract. Selective media contained 0.17% yeast nitrogen base without amino acids, 0.1% yeast extract, 5% ammonium sulfate, adjusted to pH 6.0, supplemented with 0.3% glucose (YNBG) or with a mixture of 0.1% glucose, 0.1% oleic acid, 0.05% Tween-40 (YNBGO) and nutritional supplements according to the selection markers.

Gene deletions and genomic integrations of TEV-Protein A as C-terminal fusions were performed as described previously (Güldener et al., 1996; Knop et al., 1999).

For generation of yeast and Escherichia coli expression plasmids, the corresponding coding DNA regions were amplified by performing PCR using the indicated primers (Table S2) and genomic DNA of a wild-type strain as a template. For two-hybrid analysis and localization studies, DNA fragments were cloned into pPC86 and pPC97, and pUG35 and pUG36, respectively.

Cells were cultivated on YNBGO medium. Post-nuclear supernatants derived from wild-type, Δpex9, Δpex5 and Δpex5Δpex9 strains were loaded onto linear OptiPrep™-sucrose [2.25–24% (w/v) iodixanol containing 18% (w/v) sucrose] density gradients. For differential centrifugation studies, wild-type, Δpex9, Δpex5 and Δpex5Δpex9 cells were inoculated in YNBG at an initial optical density at 600nm (OD₆₀₀) of 0.1 and were grown for 10 h (OD₆₀₀ ∼2). Oleate was added to a final concentration of 0.1%, and the cultures were incubated for an additional 10 h. Post-nuclear supernatants were loaded onto a cushion formed with 1 ml of 60% sucrose and 750 µl of 7% sucrose. The samples were centrifuged at 20.000 g for 30 min in SW41 Ti rotor (Beckmann Coulter). The sediments were collected and diluted to the initial volume. Equal volumes were precipitated with trichloroacetic acid and analyzed by immunoblotting.

Post-nuclear supernatants and isopycnic density gradient centrifugations were prepared as described previously (Cramer et al., 2015). For chemical extractions of membrane proteins, organellar sediments of cells expressing Pex9p-TPA (Pex9p fused to Protein A containing a tobacco etch virus protease cleavage site) resulting from centrifugation at 25,000 g were obtained as described previously (Erdmann and Blobel, 1996). Organellar membranes were incubated with 10 mM Tris-HCl pH 8.0 to release soluble matrix proteins, 10 mM Tris-HCl pH 8.0+500 mM KCl to extract proteins loosely associated with the membrane, and with 100 mM Na₂CO₃ pH 11.5 to separate peripheral from putative integral membrane proteins. After a 45-min incubation for each step, membranes were sedimented at 200,000 g for 60 min through an 18% sucrose cushion. Equal volumes of supernatants and sediments were precipitated with trichloracetic acid (12.5%) and subjected to immunoblot analyses. Protease protection assays were performed with equal fractions of post-nuclear supernatants containing 0.67 mg/ml of protein. One of the fractions was supplemented with 0.2% Triton-X100 before Proteinase K (400 µg protease/mg protein) was added. Digestion was performed first for 15 min on ice, then for an additional 45 min at room temperature, and were stopped by adding 10 µl PMSF.

Wide-field fluorescence imaging was performed on a Zeiss Axioskop50 fluorescence microscope (Zeiss). Images were taken with a Princeton Instruments 1300Y digital camera. GFP signal was visualized with a 450–490 nm band pass excitation filter, a 510-nm dichromatic mirror and a 515–565-nm band pass emission filter. mCherry fluorescence was visualized with a 546/12 nm band pass excitation filter, a 560-nm dichromatic mirror and a 575–640-nm band pass emission filter (Nagotu et al., 2012).

For generation of antibodies against Mls2p, the plasmid pGST-MLS2 (Table S1) was transformed into E. coli BL21-DE3 Star, resulting in isopropyl-β-D-thiogalactopyranosid-inducible expression of GST-tagged Mls2p, which was purified by using affinity chromatography with glutathione Sepharose 4B (GE Healthcare, Munich, Germany), cleaved by thrombin and further purified by using size-exclusion chromatography through a Superdex 200 column (GE Healthcare, Munich, Germany). Polyclonal antibodies were raised against the purified recombinant Mls2p (Pineda, Berlin, Germany).

Immunoblots were incubated with polyclonal rabbit antibodies raised against Mls1p (a kind gift from Andreas Hartig, University of Vienna, Vienna, Austria; 1:2.000), Mls2p (this study; 1:10.000), Pex5p (1:10.000) (Albertini et al., 1997), Pex10p (1:10.000) (Albertini et al., 2001), Pex14p (1:10.000) (Albertini et al., 1997), Fox3p (1:10.000) (Erdmann and Kunau, 1994), Pcs60p (1:10.000) (Blobel and Erdmann, 1996), Fox1p (1:10.000) (Schäfer et al., 2004), Mdh3p (1:2.500) (Steffan and McAlister-Henn, 1992), Por1p (1:10.000) (Kerssen et al., 2006), Kar2p (1:10.000) (Rose et al., 1989), Pgk1p (1:7.000) (Invitrogen, Karlsruhe, Germany), Pxa1p (1:10.000) (Rucktaschel et al., 2009), Protein A (1:10.000) (Sigma-Aldrich, Munich, Germany) and GFP–GST (1:5.000) (Birschmann et al., 2003). The antibodies against Mls1p and Mls2p did not cross-react. Primary antibody was detected with a IRDye 800CW goat anti-rabbit IgG secondary antibody (LI-COR Bioscience, Bad Homburg, Germany) followed by detection using the Infrared Imaging System (LI-COR Bioscience, Bad Homburg, Germany). Semi-quantitative analyses of immunoblot signals were obtained by using the ‘Infrared Imaging System Application Software Version 3.0’ (LI-COR Bioscience, Bad Homburg, Germany).

Double transformants of wild-type PCY2 cells expressing the GAL4-fusion proteins were selected, and β-galactosidase activity was determined by performing a filter assay using X-Gal as the substrate, as previously described (Rehling et al., 1996).

We thank Thomas Schröter (Ruhr-Universität Bochum, Germany) for providing plasmid pTSC13-mCherry-Ant1p. The work presented in Figs.1B, 2–4, 5A and 6 has been adapted from Daniel Effelsberg's doctoral thesis (Analysis of alternative import routes into peroxisomes of the yeast Saccharomyces cerevisiae, 2015, Ruhr-Universität Bochum, Germany). We thank Maya Schuldiner and Einat Zalckvar for their collegiality.