The budding yeast Pex5p receptor directs Fox2p and Cta1p into peroxisomes via its N-terminal region near the FxxxW domain

Peroxisomes are small vesicular organelles that are present in almost all eukaryotic cells. They are surrounded by a single membrane and contain diverse sets of enzymes that are involved in several metabolic pathways, including fatty acid β-oxidation and lipid biosynthesis. Peroxisomes are characterized by the presence of a number of oxidases that produce hydrogen peroxide, which is decomposed by catalase (Tolbert, 1981).

The comparison of peroxisomal protein contents in different tissues and species demonstrates that some tissues and species lack peroxisomal proteins that are present in others, which clearly reflects the ability of peroxisomes to adjust to specific metabolic functions. Even within evolutionarily related groups, such as mammals, peroxisome diversity is observed (Islinger et al., 2010). Moreover, the peroxisomal proteome may be modified to adapt to changes in environmental conditions and the metabolic state of the cell. A comprehensive account of peroxisome biogenesis and functions can be found in the recent special issue of Biochimica et Biophysica Acta (2016, vol. 1863, pp. 787-1069). When compared with the proteomes of other organelles, the peroxisomal proteome is relatively small. Currently, the number of known peroxisomal proteins in humans is 101, while that in budding yeast (Saccharomyces cerevisiae Meyen ex E.C. Hansen) is 75 (http://www.peroxisomedb.org/, accessed 17 July, 2018) (Schlüter et al., 2010). Of those proteins, only a subset is localized in the peroxisomal lumen; however, the mechanisms for the import of these few proteins into peroxisomes are surprisingly diverse. Peroxisomal matrix proteins are imported post-translationally, and the specificity of transport depends on the peroxisome-specific targeting signal encoded in the protein sequence. Thus far, two such signals are known and have been well-characterized: peroxisomal targeting signal 1 (PTS1) and peroxisomal targeting signal 2 (PTS2) (Girzalsky et al., 2010).

The PTS1 targeting signal was established as a tripeptide located at the extreme C-terminus of proteins (Gould et al., 1989). While an SKL sequence appears to be the most typical, permutation analysis led to the formulation of a more flexible rule for the PTS1 signal: (S/A/C)-(K/H/R)-(L) (Swinkels et al., 1992). The comparison of many PTS1 sequences in mammals, plants, protozoa and yeast allowed the generalization of the PTS1 composition as follows: (small)-(basic)-(large non-polar) amino acids. In silico analysis generalized the consensus even further (Nötzel et al., 2016). The tripeptide PTS1-like sequence is present in the majority of peroxisomal matrix proteins and is necessary and usually sufficient to direct reporter proteins to peroxisomes. Interestingly, however, some PTS1 variants are not active in a heterologous context, although they function in the proteins in which they naturally occur. This behavior implies that some additional information may be needed for the import of these proteins. Therefore, the complete PTS1 signal ensuring the proper recognition of peroxisomal proteins has been proposed to include up to nine amino acid residues preceding the C-terminal tripeptide (Brocard and Hartig, 2006).

A much smaller group of peroxisomal matrix proteins uses a PTS2 signal consisting of nine amino acids with the consensus sequence (R/K)-(L/I/V)-X₅-(H/Q)-(L/A/F) located within the first 40 N-terminal amino acids of the protein (Lazarow, 2006). To date, only two proteins with functional PTS2 sequences have been identified in S. cerevisiae, 3-ketoacyl-CoA thiolase (Fox3p/Pot1p) (Erdmann, 1994) and glycerol-3-phosphate dehydrogenase (Gpd1p) (Jung et al., 2010). Curiously, the import of the nicotinamidase Pnc1p into peroxisomes also depends on the PTS2 signal, but it relies on the signal in Gpd1p via a piggy-back import mechanism (Effelsberg et al., 2015; Kumar et al., 2016; Saryi et al., 2017), another feature that is characteristic of peroxisome biogenesis.

Two well-known peroxisomal proteins in S. cerevisiae function as import receptors: Pex5p with tetratricopeptide repeat (TPR) domains and Pex7p with WD-40 domains, which recognize PTS1 and PTS2 signals, respectively. While Pex5p can function as an autonomous receptor, Pex7p requires the co-receptors Pex18p and Pex21p (Purdue et al., 1998), which stabilize and target the peroxisomal membrane cargo-bound Pex7p (Stein et al., 2002). Recently, another yeast protein, which for two decades was a known paralog of Pex5p with no assigned function, was shown to participate in the import of a subset of PTS1-dependent peroxisomal matrix proteins and was named Pex9p (Effelsberg et al., 2016; Yifrach et al., 2016). However, its role in the peroxisomal import machinery is still not fully understood.

The receptor-assisted translocation of proteins into the peroxisomal lumen occurs via a transient pore that is formed in the peroxisomal membrane by Pex5p and Pex14p peroxins and, depending on the cargo size, can be as wide as 9 nm (Meinecke et al., 2010). This mechanism distinguishes peroxisomes from other cellular organelles, and the temporary nature of the import pores explains past contradictory findings demonstrating that the peroxisomal membrane is able to maintain a pH gradient (Waterham et al., 1990) but is permeable to bulky particles (Walton et al., 1995). This unique mechanism allows the peroxisome to internalize proteins in their native folded conformation and oligomeric protein complexes. As a consequence, proteins that lack their own PTS signal can be imported into peroxisomes in association with PTS-containing proteins via the piggy-back mechanism mentioned above. Curiously, two distinct types of membrane pores are created in the peroxisomal membrane for the separate transport of Pex5p- or Pex7p-dependent cargos (Montilla-Martinez et al., 2015).

Interestingly, this situation in S. cerevisiae is not universal. Peroxisome import mechanisms appear to be subject to evolutionary diversification. Sometimes, the orthologous proteins from various systematic groups are imported via different routes. For example, while the import of thiolase is dependent on the PTS2 signal and Pex7p receptor in most organisms, in Caenorhabditis elegans, thiolase contains the PTS1 signal (Bun-Ya et al., 1997); a PTS2-dependent import route and its components appear to be absent from this organism (Motley et al., 2000).

Another example is AOx, the key enzyme in peroxisomal β-oxidation. These proteins in Yarrowia lipolytica, S. cerevisiae and many other yeast species have no known PTS signals (Wang et al., 1999; https://portals.broadinstitute.org/cgi-bin/regev/orthogroups/show_orthogroup.cgi?orf=YGL205W, accessed 17 July, 2018) but are efficiently imported into peroxisomes, whereas in mammals and Pichia pastoris, they are imported via the PTS1 route (Koller et al., 1999). Thus, the import mechanisms of peroxisomal matrix proteins appear to be as complex and diverse as the functions of this organelle and are definitely not fully understood.

After the PTS1- and PTS2-dependent import routes were basically determined, the import of the AOx enzyme, which in several yeast species possesses neither of these signals, remained especially puzzling. However, studies trying to solve this puzzle have been scarce. Almost two decades ago, the import of AOx into peroxisomes in S. cerevisiae was shown to depend on Pex5p, but it was also revealed that AOx interacts with the region between residues 136 and 292 in this receptor, which is outside the C-terminal half containing the TPR domain that is responsible for the recognition of the PTS1 signal. Moreover, the same 136-292 aa fragment was found to bind the carnitine acetyltransferase Cat2p (Skoneczny and Lazarow, 1998). Curiously, the latter protein was shown to be imported into peroxisomes regardless of the presence of the PTS1 signal on its C-terminus, although the signal itself was proven to be fully functional in a heterologous context (Elgersma et al., 1995). These results were later corroborated by narrowing down the region in the Pex5p polypeptide that is important for the import of these two proteins to the 250-270 aa fragment encompassing the diaromatic motif FxxxW. Individual residues indispensable for AOx and/or Cat2p recognition were identified within this motif and at neighboring positions (Klein et al., 2002). Therefore, the two budding yeast proteins AOx and Cat2p appeared to utilize an alternative peroxisomal import pathway, which was tentatively called the PTS3 route, even though its components remain unknown to date. In particular, we do not know the PTS3 signal itself – i.e. the consensus sequence – if one exists, of the amino acids that direct AOx and Cat2p into peroxisomes.

Another question that remains is the number of proteins that actually utilize this route. Are AOx and Cat2p the only budding yeast proteins that are imported into peroxisomes via the hypothetical PTS3 route? Most peroxisomal matrix proteins contain a PTS1 or PTS1-like tripeptide at their C-terminus. Notably, however, many of them were assigned to the PTS1-dependent import route solely because of the presence of the PTS1 signal. The dependence of their import on this signal had been implied but not proven directly. Therefore, a reasonable question is whether the PTS1 signal is necessary and sufficient to drive these proteins into peroxisomes in all cases or if it is sometimes dispensable, as in the case of Cat2p. Finally, there are proteins that are functionally associated with peroxisomes but not shown to be localized in this compartment and do not contain either PTS1 or PTS2. Perhaps some of them do reside in the peroxisomal matrix and are imported via the PTS3 route.

Although ambiguities in peroxisomal import signaling have been recognized since exceptions from the strict rules regarding PTS1 (and also PTS2) were reported, no attempt has been made to answer the above questions. This paper aims to address them. By employing several complementary approaches, we were able to demonstrate that in addition to AOx and Cat2p, at least two other proteins, Fox2p and Cta1p, are partially dependent on interactions with the N-terminal region of Pex5p for import. The Ile264 amino acid residue in Pex5p, which is adjacent to the FxxxW diaromatic domain that was previously shown to be crucial for recognizing AOx and Cat2p as cargos, is also important for the efficient import of Fox2p and Cta1p. Moreover, Pex9p appears to be able to substitute, albeit weakly, for the absent Pex5p in the import of these two proteins. That substitution most likely relies on the N-terminal part of Pex9p. Although Pex9p does not contain the FxxxW diaromatic motif, it displays strong homology to the region of Pex5p that contains this motif. In light of our data, the peroxisomal import mechanisms appear to be even more complex than previously believed.

We aimed to identify potential peroxisomal matrix PTS3-type proteins whose transport is dependent on the Pex5p receptor but not on its TPR domain, which is involved in the transport of proteins containing the classical PTS1 signal at the C-terminus. To assign a peroxisomal protein to this novel group of proteins, we established the following initial criteria: (1) a protein that is transported into peroxisomes even though it does not contain the PTS1 signal, or the signal can be removed without affecting protein peroxisomal localization; (2) a protein that is not transported into peroxisomes in cells lacking the Pex5p receptor; and (3) a protein that is transported into peroxisomes in cells lacking the Pex7p receptor, thus showing its independence of the putative PTS2 signal recognition. Two approaches were undertaken to discover the novel target proteins transported via the PTS3 pathway and thus bearing this as yet unidentified PTS3 signal: an experimental approach and an ‘educated guess’ approach.

As an experimental approach to identify further peroxisomal proteins whose import may depend on the PTS3 signal, we used metal affinity chromatography to identify proteins that physically interact with the N-terminal region of Pex5p encompassing the putative PTS3 recognition domain. Since it was previously demonstrated in a two-hybrid test and by metal affinity chromatography that AOx and Cat2p physically interact with a small fragment of Pex5p comprising 136-292 amino acid residues (Skoneczny and Lazarow, 1998), we used the same 6×His-tagged polypeptide (His-Pex5p^136-292) as bait to affinity capture other proteins that might be imported into peroxisomes via the same route as AOx and Cat2p. Protein samples were extracted from S. cerevisiae cells grown in oleate-containing medium to induce peroxisome proliferation. Moreover, we used a pox1Δ,cat2Δ double-deletion strain to increase the purification yield of the prospective protein since we had previously noticed signs of competition between AOx and Cat2p for binding to the His-Pex5p^136-292 bait. The resulting specifically bound protein was identified by mass spectrometry as Fox2p, a bifunctional enzyme with 3-hydroxyacyl-CoA dehydrogenase and enoyl-CoA hydratase activity in the peroxisome fatty acid β-oxidation pathway that converts trans-2-enoyl-CoA esters to 3-ketoacyl-CoA esters (Hiltunen et al., 1992). The authenticity of the Fox2p retained on the affinity column was confirmed with cell extracts from the pox1Δ,cat2Δ,fox2Δ triple-deletion strain. No corresponding protein band was seen in the eluate from the affinity column loaded with the protein extract from the cells of this strain (data not shown). Thus, Fox2p could be another peroxisomal protein transported via the novel PTS3 pathway even though, like Cat2p, it possesses a PTS1 signal (-SKL). For that reason, it was commonly assumed to be targeted to the peroxisomal matrix via the PTS1 pathway; although, to the best of our knowledge, it was never demonstrated experimentally. To verify the functionality of the Fox2p C-terminal SKL sequence and the adjacent amino acid residues, hybrid genes were constructed encoding proteins in which the PTS1 sequence of Fox2p was removed and its C-terminus blocked by the green fluorescent protein (GFP) tag. Cells from the wild-type (WT) and mutant yeast strains were transformed with this plasmid and with the plasmid encoding the peroxisome localization marker mRFP-SKL. The localization of the fusion proteins was checked by fluorescence microscopy in cells grown in oleic acid-containing medium to induce peroxisome proliferation (see Materials and Methods for details of plasmid construction and yeast cell growth conditions). As shown in Fig. 1A, Fox2p with a non-functional PTS1 was present in yeast cells in the form of particles that colocalized with the peroxisomal fluorescence marker.

An additional approach to find further peroxisomal matrix proteins imported by the PTS3-dependent route involved analysis of the available literature and amino acid sequence data. Candidate proteins were selected based on the following principles: (1) there are reports indicating or suggesting the peroxisomal localization of the protein, and (2) the protein lacks both PTS1 and PTS2 targeting signals, or it does have the PTS1 signal, but no literature data or conflicting data exist regarding its indispensability for import. We also included S. cerevisiae proteins that are homologues of peroxisomal proteins from other organisms that also lack known peroxisomal targeting sequences. Finally, the list of candidate S. cerevisiae proteins that could be transported via the PTS3 route included Pcd1p, Faa2p, Mdh2p and Cta1p; however, the preliminary localization experiments excluded the first three from further analysis. Only the import of Cta1p, the peroxisomal catalase A, was analyzed in more detail. Earlier attempts to characterize the mechanism of yeast Cta1p recognition and targeting to the peroxisomes suggested that in addition to its C-terminal SKF tripeptide, some other internal sequence(s) could be involved in its import (Kragler et al., 1993). This possibility was not subjected to any further studies, but to us, the import of Cta1p appeared likely to depend at least in part on the same mechanisms as that of AOx and Cat2p.

To test this assumption, we constructed a plasmid containing the hybrid gene encoding Cta1p protein from which the SKF sequence was removed and with the C-terminus blocked by GFP tagging. Cells from the WT and mutant yeast strains were transformed with this plasmid and with the plasmid encoding the peroxisome localization marker mRFP-SKL. The localization of the fusion proteins was checked by fluorescence microscopy in cells grown in oleic acid-containing medium to induce peroxisome proliferation. As shown in Fig. 1B, the Cta1p-GFP fusion protein was found in peroxisomes, which indicated that its import could be independent of the PTS1 signal.

The images shown in Fig. 1A-D revealed that the PTS1-independent import of Cta1p and Fox2p might not be as efficient as the import of AOx and Cat2p, the two PTS3-dependent proteins discovered earlier. Both AOx and Cat2p were tagged with GFP at the C-terminus and were observed almost exclusively in particles colocalizing with the peroxisomal marker, whereas Cta1p-GFP and Fox2p-GFP, in addition to peroxisomes, were also observed in the cytosol. This difference could mean that these enzymes are natively present in the cytosol in addition to peroxisomes or that their C-terminal regions have some contribution to their peroxisomal import. To resolve this, we tested the efficiency of Fox2p import when it was tagged with GFP at the N-terminus (GFP-Fox2p). The tagged protein is virtually exclusively present in peroxisomes (Fig. 2A, top row), whereas N-terminally tagged Fox2p lacking its C-terminal SKL tripeptide (GFP-Fox2p-ΔSKL) is imported less efficiently (Fig. 2B, top row), similarly to the C-terminally GFP-tagged construct.

By exploiting the fact that enzymatic activity and, hence, its levels in subcellular fractions can be measured easily, we quantified the relative contributions of PTS1-dependent and PTS1-independent Pex5p-Cta1p interactions to the import of Cta1p into peroxisomes by testing the native protein and the protein with its C-terminal SKF signal replaced with YIS. Both proteins were expressed from centromeric, single-copy YCp50-derivative vectors (YCp-Cta1p and YCp-Cta1p-YIS, respectively). The GC1-8B yeast strain transformed with these plasmids was devoid of endogenous Cta1p and Ctt1p catalases. Transformant cells grown in oleate-containing medium were fractionated into an organellar fraction containing peroxisomes and a cytosol fraction, and the catalase activity was measured in both fractions as described in the Materials and Methods. The peroxisomal portion of catalase A without PTS1 was ∼20% smaller than wild-type catalase A (Fig. 2C), so removal of the PTS1 signal had only a minor effect on its import competence.

One of the features of peroxisome biogenesis is the ability of this organelle to import mature oligomeric protein complexes. As a consequence, polypeptides with no peroxisomal targeting signal can enter peroxisomes while bound or ‘piggy-backed’ to polypeptides that contain one. This workaround mechanism was first demonstrated for thiolase (Glover et al., 1994), and evidence has since accumulated that this mechanism is widespread in S. cerevisiae (Yang et al., 2001; Effelsberg et al., 2015; Kumar et al., 2016; Saryi et al., 2017) and other species (Titorenko et al., 2002; Lee et al., 1997). Therefore, we tested whether the observed import of Fox2-GFP and Cta1p-GFP, which had their C-termini blocked with a GFP tag, might be due to the formation of complexes with the respective endogenous native proteins. This could occur for Cta1p-GFP, as native catalase A is known to be a tetramer (Seah et al., 1973); however, to the best of our knowledge, no data are available on the oligomeric state of S. cerevisiae Fox2p protein. However, the fruit fly peroxisomal multifunctional enzyme was shown crystallographically to be a dimer (Haataja et al., 2011). If the same is true for its budding yeast orthologue, the piggy-back import of PTS1-less Fox2p-GFP bound to native Fox2p is conceivable. However, as shown in Fig. 1A, the efficiency of Fox2p-GFP import in fox2Δ cells lacking the native Fox2p appeared to be similar to that of WT cells. The same was true for Cta1p-GFP (Fig. 1B), which displayed a comparable efficiency of import in WT and cta1Δ cells. The piggy-back mode of import for plasmid-encoded Cta1p-YIS (Fig. 2C) was also excluded because of the lack of genome-encoded native Cta1p in the GC1-8B strain. Incidentally, as shown in Fig. 1C,D, the import of neither the C-terminally GFP-tagged AOx, known to be an octamer in its native state, nor that of Cat2p-GFP showed signs of dependence on the piggy-back mechanism, as both were equally efficient in WT cells and in the respective deletion strains pox1Δ and cat2Δ.

Two known peroxisomal import routes relying on PTS1 and PTS2 targeting signals are handled by the Pex5p and Pex7p cargo receptors, respectively. None of the studied proteins have ever been associated with the PTS2-dependent peroxisomal import route, nor do they contain any recognizable PTS2 signals. However, for the sake of completeness, we tested the influence of the absence of either Pex5p or Pex7p on the import of GFP-tagged Fox2p and Cta1p into peroxisomes. As shown in Fig. 1A-D, the absence of Pex7p (pex7Δ strain) had no influence on the import of these proteins, just as in the case of AOx and Cat2p. In contrast, as shown in Fig. 2A,B,D,E, their import depended on the Pex5p receptor (compare rows labeled WT and pex5Δ in panels A, B, D and E in Fig. 2). This dependence classifies them into the same group as AOx and Cat2p. In addition to the images, we collected quantitative data by counting cells expressing GFP-tagged Fox2p and Cta1p to determine the fraction that showed at least partial particulate fluorescence (Fig. 2F). The data obtained for constructs encoding Fox2p and Cta1p with their C-termini blocked by GFP (Fig. 2D,E), together with the data presented in Fig. 1A,B, clearly indicate that the import of these proteins into peroxisomes can be driven by Pex5p without recognition of the PTS1 signal by its TPR domain. These findings are consistent with the results from our affinity purification experiment for Fox2p and fully support the previously postulated dual-domain nature of Pex5p. These results also suggest that, in addition to AOx and Cat2p, the import of Fox2p and Cta1p may depend, at least in part, on the hypothetical PTS3 route.

Interestingly, although the importance of Pex5p for the import of Fox2p and Cta1p was beyond doubt, we could occasionally see residual peroxisomal localization for both these proteins in the pex5Δ strain (see rows labeled pex5Δ in Fig. 2A,B,D,E). A paralog of Pex5p, Pex9p, has been recently characterized in S. cerevisiae (Effelsberg et al., 2016; Yifrach et al., 2016), and its participation in the import of the subset of PTS1-dependent peroxisomal matrix proteins was demonstrated. Therefore, we tested whether this protein contributes to Fox2p and Cta1p import and is therefore responsible for their visible residual peroxisomal localization. As shown in Fig. 2, absence of the PEX9 gene had no noticeable effect on the import of the studied proteins into peroxisomes, but the absence of both the PEX5 and PEX9 genes had a stronger detrimental effect on the import of GFP-tagged Fox2p and Cta1p than the absence of the PEX5 gene alone; in the pex5Δ,pex9Δ double-deletion strain, both proteins were observed exclusively in the cytosol. Note that while the images in the rows labeled WT, pex9Δ and pex5Δ,pex9Δ are representative of whole-cell populations, the images in the rows labeled pex5Δ display the rare instances of cells with particulate GFP-tagged Fox2p or Cta1p. Most of the cells from the pex5Δ strain looked exactly like those shown in the rows labeled pex5Δ,pex9Δ. Notably, the colocalization of Fox2p or Cta1p with the PTS2-DsRed peroxisomal marker was observed only in cells that displayed weak fluorescence in the green channel, i.e. contained low levels of GFP-tagged proteins. The quantitative data (Fig. 2F) clearly show that 5-10% of pex5Δ cells expressing every GFP fusion construct used in this experiment displayed weak but visible particulate fluorescence colocalizing with the peroxisomal marker, whereas in pex5Δ,pex9Δ cells, the fluorescence was purely cytosolic. These data suggest that although Pex9p appears not to significantly contribute to the import of either protein in the presence of Pex5p, it can to some extent act as a substitute receptor for them when Pex5p is missing. Notably, since this phenomenon is observed for C-terminally GFP-tagged Fox2p and Cta1p (Fig. 2D,E), Pex9p, similarly to Pex5p, appears to be able to drive the residual import of these proteins through a mechanism that does not involve the recognition of the PTS1 signal by its TPR domain.

As previously postulated, Pex5p appears to have two domains that interact with cargo proteins: the well-known TPR domain recognizing the C-terminal PTS1 and the less well-defined internal region in the N-terminal half of its polypeptide. Amino acid residues that are important for the Pex5p interaction with AOx and Cat2p were identified within this region, and one of them, Ile264, was found to be important for the interaction with both proteins but not with the elements in the translocon complex consisting of Pex13p and Pex14p (Klein et al., 2002). Therefore, we tested whether the substitution of this residue affects Fox2p and Cta1p import. Fig. 3 shows the effect of replacing isoleucine 264 with lysine at the corresponding position in the Pex5p polypeptide on the import of these proteins. Indeed, the substitution of Ile264 did affect the import of Fox2p-GFP and Cta1p-GFP, both with their C-termini blocked by the GFP tag, diminishing the number of cells with visible green fluorescence particles that colocalized with the peroxisomal marker by 20% and 35%, respectively (see Fig. 3A). Interestingly, however, in yeast strains expressing N-terminally tagged GFP-Fox2p with its PTS1 signal exposed, the difference in the numbers of cells with GFP particles observed in the presence of WT Pex5p and in the presence of its I264→K-mutated version were less pronounced. Ile264 residue is located near the FxxxW motif known to be important for the interaction of Pex5p receptor with Pex14p, a subunit of the peroxisomal translocon (Williams et al., 2005). Therefore, its substitution could potentially influence the interaction between these peroxins and in consequence affect the Pex5p-dependent import as a whole, so we tested if substitution of the Ile264 residue in Pex5p affects import that relies on the PTS1 targeting signal. As shown in Fig. 3B, substitution of the Ile264 residue in Pex5p had no effect on the proportion of cells displaying particulate localization of mRFP-SKL, a marker protein imported into peroxisomes exclusively via a PTS1-dependent route. Notably, in the absence of Pex5p, we did not observe any residual peroxisomal localization that was similar to that seen for GFP fusions of Fox2p and Cta1p.

Analogous results were obtained in the experiment in which we tested the influence of the I264→K mutation in Pex5p on the distribution of Fox2p-GFP (Fig. 3C) and native catalase A (Fig. 3D) between the organellar and cytosolic fractions. Fox2p-GFP was quantified by western blotting with anti-GFP antibodies in the same strain as used previously (Fig. 3A). Catalase A was quantified by its enzymatic activity in the pex5Δ derivative of the GC1-8B strain, which is devoid of both yeast catalases, and in which we expressed the native Cta1p protein and the WT or mutated I264→K-variant of Pex5p. In both cases, yeast cells were grown in YPEO medium and fractionated into an organellar fraction containing peroxisomes and a cytosol fraction. The distribution of both proteins between the organellar and cytosolic fractions revealed that the substitution of the Ile264 residue in Pex5p disturbed the import of both native Cta1p and Fox2p-GFP by ∼35%.

The intriguing case of the budding yeast peroxisomal acyl-CoA oxidase that somehow enters the lumen of its destined organelle despite possessing neither of the identified targeting signals has been known for several decades. While the PTS1- and PTS2-dependent peroxisomal import routes are well-characterized and are evolutionarily conserved from simple unicellular eukaryotes to mammals, the import of the budding yeast AOx is a curiosity because its orthologues from many other organisms do have PTS1 signals on their C-termini. In addition to S. cerevisiae, some other yeast species contain AOx proteins with no recognizable PTS1 or PTS2 (Wang et al., 1999) that probably follow the same route into peroxisomes as the budding yeast AOx.

No less intriguing were the cases of proteins such as S. cerevisiae Cat2p (Elgersma et al., 1995) and Hansenula polymorpha alcohol oxidase (Gunkel et al., 2004), which do have a PTS1 signal able to direct reporter proteins to peroxisomes, yet they are seemingly dispensable within these proteins. These observations challenge the common belief, which has evidently been oversimplified, regarding the import mechanisms of peroxisomal matrix proteins. Once PTS1 and PTS2 were discovered in the 1990s (Gould et al., 1989; Swinkels et al., 1991), most peroxisomal proteins were attributed to the PTS1- or PTS2-dependent route based on the mere presence of the respective signals in their amino acid sequences. The fact that in budding yeast, PTS1-containing Cat2p and AOx, which lacks PTS1, both share a mechanism for peroxisomal internalization that involves interaction with the N-terminal region of Pex5p, especially with its diaromatic FxxxW motif, indicates the need to re-evaluate Pex5p- and PTS1-dependent import. This new mechanism itself is unknown, as are the other elements of this hypothetical import route, such as the signal sequences within AOx and Cat2p polypeptides, which, although currently unknown, are sometimes called PTS3 signals. Is this mechanism distinct enough to qualify it as a separate peroxisomal import route? An attempt to answer these questions is the subject of a separate study. In this study, we set another goal, to identify other peroxisomal proteins, in addition to AOx and Cat2p, whose import into peroxisomes in budding yeast depends on the PTS3 route. Those could be the only two proteins belonging to this group. Only three proteins in S. cerevisiae are known to be imported via the PTS2 pathway (Erdmann, 1994; Effelsberg et al., 2015), yet the entire machinery consisting of three peroxins is preserved for them (Purdue et al., 1998).

Nonetheless, we had reason to assume that the list of PTS3-dependent peroxisomal proteins may indeed be longer. Other proteins in addition to AOx are associated with peroxisomes but contain neither PTS1 nor PTS2 signals. These proteins could be imported into peroxisomes but not previously categorized as such because of the absence of any recognizable PTS. Moreover, the case of Cat2p indicated that the presence of PTS1 is not the ultimate proof that the protein utilizes the PTS1-dependent import route. That appeared to be true for both proteins described in this study. Fox2p contains the canonical C-terminal PTS1 tripeptide SKL, yet in metal affinity chromatography, it bound the same 136-292 aa fragment of Pex5p as AOx and Cat2p did. This interaction also occurs in vivo, as Fox2p GFP-tagging at its C-terminus only partially affected its import (see Fig. 1A). Cta1p, the candidate resulting from the analysis of the literature data and protein sequences, contains a working variant of PTS1, SKF, although its masking with the GFP tag or its removal had limited impact on the import of Cta1p into peroxisomes. C-terminally GFP-tagged Fox2p and Cta1p proteins entered peroxisomes equally efficiently in the presence and absence of their respective native, genome-encoded proteins, which excludes their piggy-back-assisted import. The fully efficient import observed in the absence of the Pex7p receptor also excludes their dependence on PTS2 (Fig. 1A,B). Comparing these images with the ones in Fig. 1C,D revealed that the efficiency of the import of Fox2p and Cta1p was lower than that of AOx and lower than that of Cat2p when the latter is not dependent on the PTS1 signal. Similarly to AOx and Cat2p, the observed import of Fox2p and Cta1p occurs without interaction between their C-termini and the Pex5p TPR region, but it does depend on the presence of the Pex5p receptor (Fig. 2). This observation is compatible with the hypothesis of the dual functionality of Pex5p peroxin: (1) it contains a PTS1 signal receptor within its TPR region located in the C-terminal half of the polypeptide and (2) it also contains a receptor for the hypothetical PTS3 signal in AOx and Cat2p located in the region closer to its N-terminus, between residues 250 and 270.

By employing a mutated variant of Pex5p bearing the I264→K substitution, which blocks the interaction of Pex5p with AOx and Cat2p, the participation of the PTS3 pathway in the import of Fox2p and Cta1p can be estimated as between 20 and 35%, depending on the assessment method used (Fig. 3).

As shown in Fig. 2C and Fig. 3D, substitution of the C-terminal PTS1 of Cta1p-SKF with YIS lowered its import efficiency by ∼20%, whereas substitution of Ile264 in Pex5p lowered the import efficiency of native Cta1p by ∼35%. Superficially, one could expect these two numbers to sum up to 100%, but the modifications introduced into Cta1p and Pex5p most likely only partially affect the PTS1- and PTS3-dependent import of Cta1p. The region in Cta1p upstream from the SKF tripeptide probably also contributes to Cta1p interactions with the Pex5p TPR domain. Likewise, the substitution of Ile264 in Pex5p may not completely abolish the interaction of Cta1p with the N-terminal part of Pex5p, as other amino acid residues in this region were also shown to be important for the interaction of Pex5p with AOx and Cat2p (Klein et al., 2002).

It is noteworthy that testing the importance of Ile264 for the import efficiency of Fox2p and Cta1p with the two methods gave similar results (compare Fig 3A,C and D). The percentage of cells displaying the peroxisome-located GFP fusion proteins is considerably higher than the percentage of Fox2p-GFP or native catalase A present in the organellar fraction, but this can be explained by the presence of Fox2p-GFP and Cta1p-GFP in most cells, both in particles and in cytosol (see Fig. 2D,E).

Our data are also in agreement with previous reports on the import of Cta1p that revealed the dispensability of its C-terminus for its import into peroxisomes and revealed instead the importance of certain internal region(s) (Kragler et al., 1993). We demonstrated the importance of the N-terminal region of Pex5p for the import of Cta1p, which entitles us to assign Cta1p, as well as Fox2p, together with AOx and Cat2p, to the common group of peroxisomal proteins whose import into this organelle is dependent, at least partially, on their interaction with the specific region residing within the N-terminal region of the Pex5p polypeptide that surrounds the FxxxW diaromatic motif.

Analysis of our data suggests that the importance of the interaction with this region for cargo recognition and/or import is somewhat different for Cta1p and Fox2p. While the substitution of Ile264 in Pex5p manifests in a similar way in the presence of the SKF sequence at the Cta1p C-terminus and when its PTS1 is blocked by the GFP tag (compare Fig. 3A and D), peroxisomal localization of Fox2p is affected more strongly by Ile264 substitution when its C-terminus is blocked (see Fig. 3A).

Our results revealed yet another interesting peculiarity regarding the import of Fox2p and Cta1p into peroxisomes. The examination of many pex5Δ cells expressing either of these proteins allowed us to notice an occasional particulate distribution of GFP fluorescence that colocalized with the PTS2-DsRed marker imported by the PTS2 route, in addition to the uniform cytosolic fluorescence (see rows labeled pex5Δ in Fig. 2A,B,D,E). The origin of this initially disturbing observation was clarified by the use of the pex5Δ,pex9Δ double-deletion strain, in which no Fox2p or Cta1p particles could be observed. For many years, S. cerevisiae Pex9p had been known only under the systematic name Ymr018wp and had no assigned function. Recently, however, the Mls1p and Mls2p proteins have been shown to be imported into peroxisomes in oleate-induced cells in a Pex9p-dependent manner (Effelsberg et al., 2016; Yifrach et al., 2016). Our results demonstrated that in the absence of Pex5p, Pex9p can also exhibit some receptor activity towards Fox2p and Cta1p (Fig. 2). Moreover, it must bind these proteins with the polypeptide region outside the TPR domain because the C-termini of these cargo proteins are completely blocked by the GFP tag. No firm conclusions regarding the functional assignments of particular regions of the Pex9p protein can be drawn yet. Nevertheless, the alignment of the N-terminal parts of Pex5p and Pex9p, as shown in Fig. 4, reveals regions of quite strong homology between them. These regions not only surround the first two WxxxF diaromatic motifs and the conserved Cys6 amino acid residue (marked with green boxes) but also the AOx- and Cat2p-interacting 250-270 aa region of Pex5p containing the third FxxxW diaromatic motif (marked with red box) that, in light of our present results, also appear to be important for the interaction with Fox2p and Cta1p. Therefore, we might speculate that the respective homologous region in the Pex9p polypeptide is also involved in interactions with Fox2p and Cta1p.

Another line of speculation may be based on the fact that the region surrounding the FxxxW diaromatic motif in Pex5p was previously shown to be important for the interaction of this peroxin with Pex14p, a subunit of the peroxisomal translocon (Williams et al., 2005). One could argue that when Ile264 is missing from Pex5p, the import of AOx, Cat2p, Fox2p and Cta1p is compromised due to an impaired interaction of Pex5p with Pex14p and not with its cargo proteins. However, the interaction of three of these cargo proteins with the discussed Pex5p region had been demonstrated directly. Moreover, the substitution of Ile264 did not affect the Pex5p- and PTS1-dependent import of mRFP-SKL. Nevertheless, it is conceivable that the competition between these cargo proteins and Pex14p for binding to the same region in Pex5p does exist. Moreover, it is not unthinkable that the displacement of cargo by Pex14p binding is a necessary step involving Pex14p, Pex5p and its cargo protein en route to the peroxisomal lumen. Notably, this competition has been demonstrated for mammalian proteins. It was shown that interactions between human catalase and human Pex5p is disrupted by the binding of the Pex14p N-terminal fragment to the Pex5p N-terminal domain (Freitas et al., 2011). That would also suggest that the binding of peroxisomal cargo proteins to Pex5p N-terminal region is more common not only among various Pex5p-dependent S. cerevisiae proteins but also among proteins from other systematic groups. If that were the case, then the question arises of why only two PTS1-containing proteins, interacting with both TPR and FxxxW regions, were identified in our study. Was our search not thorough enough? While we cannot answer this question yet, it should be mentioned that only Fox2p, not Cta1p, interacted with the Pex5p N-terminal domain in the metal affinity chromatography approach. Therefore, the interaction of other PTS1-containing proteins with the Pex5p N-terminal region and the partial dependence of their import on this region cannot be ruled out.

In summary, our results clearly show that the Pex5p receptor-dependent import of peroxisomal matrix proteins is more convoluted than that described in current models of peroxisome biogenesis, perhaps not only in S. cerevisiae but also in other yeast species. The results also show that the newly discovered Pex9p receptor has a more universal role in peroxisomal matrix protein import than recently reported. The data presented here, together with previous findings about AOx and Cat2p, expand our knowledge on Pex5p-dependent peroxisomal import routes, which justifies opening a new chapter in the studies of peroxisomal protein import mechanisms. They also contribute to better understanding the still unresolved issue of the PTS1- and PTS2-independent import of proteins into the peroxisomal matrix.

The wild-type BY 4741 (MATa leu2Δ0 ura3Δ0 his3Δ1 met15Δ0) strain (Brachmann et al., 1998) and its derivatives of the MATa or MATα mating type, as necessary, from the Saccharomyces Genome Deletion Project (http://www-sequence.stanford.edu/group/yeast_deletion_project/) carrying the pox1Δ, fox2Δ, cat2Δ, pex5Δ, pex7Δ or pex9Δ deletion alleles (obtained from Open Biosystems, Huntsville, USA) were used in most experiments described in this paper. To study the intracellular distribution of plasmid-encoded catalase A with or without the PTS1 signal in the presence of Pex5p protein variants, the GC1-8B (MATa leu2-3 leu2-112 ura3 trp1 ctt1-1 cta1-2) strain devoid of both catalase activities (Cohen et al., 1985) and its pex5Δ derivative were used as hosts. Double-deletion strains with the BY genetic background were constructed by crossing appropriate single-deletion parental strains, followed by sporulation and tetrad analysis using standard yeast genetics procedures (Sherman, 2002). The pex5Δ derivative of the GC1-8B strain was made by transformation of the parental strain with the kanMX deletion cassette, which was PCR-amplified from the pex5Δ knockout collection strain with the PEX5PRO and PEX5UTR primer pair (see Table S1 for a full list of the primers).

Yeast cells were grown on standard YPD medium (1% yeast extract, 2% peptone and 2% dextrose) at 28°C. To induce peroxisome proliferation, oleate-containing YPEO medium was used (1% yeast extract, 2% bactopeptone, 2% ethanol, 0.25% Tween 80 and 0.1% oleic acid). Liquid (0.67% yeast nitrogen base and 2% glucose) or solid (+2% agar) minimal SC medium supplemented with appropriate amino acids was used for yeast transformant selection and for growing yeast cells whenever plasmid maintenance was required.

To construct the single-copy plasmid containing the native CTA1 gene, a 2779 bp EcoRI DNA fragment encompassing the CTA1 open reading frame together with upstream 818 nucleotides and downstream 413 nucleotides was cut out from the Yep13-CTA1-III plasmid (Cohen et al., 1985) and cloned into the EcoRI site in the YCp50 vector (Rose et al., 1987), thus creating the YCp-CTA1 plasmid. The YCp-CTA1-YIS plasmid containing the mutated CTA1 gene encoding catalase A with the C-terminal SKF (Ser-Lys-Phe) sequence replaced by YIS (Tyr-Ile-Ser) was constructed by site-directed mutagenesis (see Table S1 for primer sequences). The 1147 bp fragment of the CTA1 coding region down to the STOP codon was PCR-amplified with the CTAupAge and CTAs_rev primer pair. A 500 bp CTA1 gene 3′-untranslated region fragment, including the very end of its open reading frame, was amplified with the CTAs_for and CTAdwHin primer pair. The two PCR products were combined and used as a template in the final PCR reaction with the CTAupAge and CTAdwHin primers. The resulting 1647 bp fragment was digested with AgeI and HindIII and cloned into the YCp-CTA1 plasmid digested with the same enzymes. The resulting YCp-CTA-YIS plasmid-encoded Cta1p with its C-terminal tripeptide replaced by the amino acid residues YIS.

To obtain plasmids encoding fusion proteins with green fluorescent protein (GFP) located at the N- or C-terminus, either possessing or lacking the C-terminal PTS1 signal, the respective sequences encoding the open reading frame or the open reading frame and 300-500 base pairs from the promoter sequence (with or without the last three codons) were PCR-amplified from yeast genomic DNA with the appropriate primer pairs containing restriction sites (see Table S1 for primer sequences). Amplification products were digested with the appropriate restriction enzymes and cloned into the pUG36 (for N-terminal fusions) or pUG35 or pUG23 (for C-terminal fusions) plasmids (Güldener and Hegemann, http://mips.helmholtz-muenchen.de/proj/yeast/info/tools/hegemann/gfp.html) linearized with the same enzymes. By this strategy, the GFP-Fox2p, GFP-Fox2p-ΔSKL, Fox2p-GFP and Cta1p-GFP plasmids were constructed. To create Cat2p-GFP encoding GFP-tagged Cat2p with the N-terminal mitochondrial leader removed, two primer pairs were used (see Table S1) to amplify the CAT2 gene promoter and the CAT2 open reading frame (ORF) starting at the second downstream ATG codon. Both fragments were inserted sequentially into the pUG35 vector linearized with the appropriate restriction enzymes. To obtain the AOx-GFP plasmid, the entire POX1 ORF together with 454 bp of its promoter (up to the BglII restriction site) was seamlessly fused to the GFP ORF ending with the SacI restriction site. The hybrid gene with a blunt-ended BglII site was inserted at the EcoRV and SacI sites in the pRS316 vector. To construct the pRS-Pex5pI264K plasmid encoding Pex5p with the I264K amino acid substitution within the putative PTS3-recognizing domain, the PEX5 ORF was subjected to site-directed mutagenesis by PCR amplification with the I264K_up and I264K_lo primer pair. The mutagenized PEX5 ORF was digested with Bsu36I and BstAPI and cloned into the pRS-PEX5WT plasmid (Kerssen et al., 2006) linearized with the same restriction enzymes. The PTS2-DsRed plasmid expressing the peroxisomal marker was constructed by the insertion of the XbaI-PstI fragment cut out from the pHPR131 plasmid (Stein et al., 2002), which encodes the DsRed protein with the attached PTS2 signal from S. cerevisiae thiolase, into the same sites in the pRS415 vector. The mRFP-SKL plasmid expressing the peroxisomal marker was constructed by the insertion of the XhoI-XbaI fragment cut out from pRS316 mRFP-SKL (Fagarasanu et al., 2009), which encodes the mRFP protein ending with the PTS1 signal, into the same sites in the pRS415 vector.

The iProof™ High-Fidelity DNA Polymerase (Bio-Rad, Hercules, CA, USA) was used to amplify the PCR fragments, and the accuracy of all constructs was verified by sequencing. The XL-1 Blue [MRF’Δ (mcrA)183Δ (mcrCB-hsdSMR-mrr)173 endA1 supE44] E. coli strain from Stratagene was used for DNA cloning and plasmid propagation. Plasmids were introduced into yeast cells using the high-efficiency lithium acetate transformation method (Gietz et al., 1995).

Yeast cells were grown overnight in liquid YPEO medium to induce the expression of genes encoding peroxisomal proteins. Cells from 200 ml of culture were spun down, washed with demineralized water, suspended in TD buffer (100 mM Tris-HCl pH 8.0, 5 mM EDTA and 5 mM dithiothreitol; 3 ml volume per 1 g of cell wet weight) and incubated for 30 min at 28°C with gentle shaking. The cells were then spun down, washed as before and suspended in Zymolyase digestion buffer (1.3 M sorbitol, 5 mM MOPS pH 7.2, 1 mM EDTA, and 0.05% Zymolyase 100 T; 2 ml of buffer per 1 g of yeast wet weight) and incubated for 45 min at 34°C. The resulting spheroplasts were spun down at 800 g, resuspended in homogenization buffer (0.65 M sorbitol, 25 mM MES-NaOH pH 6.0, 0.5 mM EDTA and Pierce™ Protease Inhibitors (1 tablet per 50 ml of buffer); 2 ml per 1 g of starting cell wet weight) and gently disrupted by using a Potter homogenizer (approximately 10 strokes of a pestle rotating at 1000 RPM). The cell homogenate was separated by centrifugation at 3000 g for 5 min into a fraction containing unbroken cells and nuclear membranes and a supernatant containing other organelles. The supernatant was centrifuged at 25,000 g for 30 min to yield a pellet containing peroxisomes and mitochondria and a soluble fraction containing the cytoplasm. The organellar pellet was resuspended in the same volume of homogenization buffer as that used in the initial homogenization step.

Equal volumes of organelle suspensions and supernatants were separated by SDS-PAGE on 10% gel and the separated proteins were transferred onto Immobilon^®-P nitrocellulose membrane (Millipore, Burlington, MA, USA) using the semi-dry method. GFP was detected using mouse anti-GFP primary antibody (11814460001, Roche Applied Science, Mannheim, Germany; 1:5000) and goat anti-mouse HRP-conjugated secondary antibody (P 0447 from Dako Denmark A/S, Glostrup, Denmark). Blots were incubated with the SuperSignal^® West Femto (Thermo Scientific, Rockford, IL, USA) and the chemiluminescence signal was digitized with Fluorchem SP CCD camera (Alpha Innotech, San Leandro, CA, USA). The protein bands were quantified with ImageQuant 5.0 software (Molecular Dynamics Inc., Sunnyvale, CA, USA).

The catalase activity was followed spectrophotometrically at 240 nm per the decomposition of H₂O₂. The activity was expressed as µmol of H₂O₂ decomposed×min⁻¹×mg of protein⁻¹ (Beers and Sizer, 1952).

To visualize the protein localization, yeast strains were transformed with plasmids expressing GFP-tagged proteins and co-transformed with plasmids expressing the mRFP-SKL or PTS2-DsRed protein, which is a peroxisome marker imported independently of the Pex5p receptor. Transformants were grown on YPEO medium to induce peroxisome proliferation. The cells were imaged with a Zeiss Axio Imager.M2 fluorescence microscope (Carl Zeiss AG, Oberkochen, Germany) with an EC Plan-NEOFLUAR 100× objective and 38HE-GFP or 20HE-rhodamine filter sets and with Nomarski optics for bright-field imaging.

The iProof™ High-Fidelity DNA Polymerase was purchased from Bio-Rad (Hercules, CA, USA). All other reagents were of analytical grade and purity.