An ancestral role in peroxisome assembly is retained by the divisional peroxin Pex11 in the yeast Yarrowia lipolytica

Peroxisomes are membrane-bounded organelles found in most eukaryotic cells. They are dynamic and highly responsive to changing environmental cues. The hallmark metabolic pathways of peroxisomes are the β-oxidation of fatty acids coupled to the controlled decomposition of hydrogen peroxide by catalase, but peroxisomes also exhibit specialized biochemical functions depending on cell type. The peroxisomes of some eukaryotes have such highly specialized functions that they are known by distinctive names (Gabaldón, 2010). Peroxisomes in trypanosomes are known as glycosomes because they compartmentalize glycolysis enzymes (Michels et al., 2006), and some plant peroxisomes are called glyoxysomes because they contain glyoxylate cycle enzymes (Hayashi et al., 2000). Filamentous fungi have modified peroxisomes called Woronin bodies involved in the maintenance of cellular integrity (Liu et al., 2008). The necessity for functional peroxisomes in human development and health is evident from the peroxisome biogenesis disorders (PBDs), a spectrum of fatal diseases in which peroxisomes fail to assemble correctly (Waterham and Ebberink, 2012; Smith and Aitchison, 2013).

Peroxisomes form through two pathways – growth and division of existing peroxisomes and de novo biogenesis from the endoplasmic reticulum (ER). The extent to which one or the other pathway contributes to peroxisome formation is different for different cells. In actively growing cells of the yeast Saccharomyces cerevisiae, peroxisome formation is predominantly by growth and division, whereas de novo peroxisome formation can occur in the event of a catastrophic loss of peroxisomes from cells (Motley and Hettema, 2007). The situation in mammalian cells is less clear, and studies have concluded that either de novo peroxisome formation or fission of existing peroxisomes can dominate (Veenhuis and van der Klei, 2014).

Peroxisome division is achieved by both general organelle divisional proteins, including the dynamin-related GTPases, and peroxisome-specific divisional proteins, notably those of the Pex11 family of peroxins. In S. cerevisiae, the Pex11 family is composed of Pex11p – the founding member of the family – Pex25p and Pex27p, whereas in humans, the family is made up of PEX11α, PEX11β and PEX11γ forms (Smith and Aitchison, 2013).

The role of Pex11p in peroxisome division is well established. Deletion of the PEX11 gene in S. cerevisiae results in cells with fewer, larger peroxisomes (Erdmann and Blobel, 1995), whereas overexpression of PEX11 results in cells with an increased number of smaller peroxisomes or elongated structures that are thought to be peroxisomes in the process of dividing (Marshall et al., 1995). Pex11p acts to elongate peroxisomes prior to their scission and subsequent separation and is proposed to assemble on the peroxisomal membrane at specific sites, stimulating the accumulation of phospholipids (Koch et al., 2010). Matrix proteins can then translocate across the growing tubules, and Pex11p recruits dynamin-related proteins for membrane scission (Koch et al., 2010). Pex11p has also been shown to exhibit membrane remodeling activity in vitro (Opaliński et al., 2011). In human cells, PEX11γ is thought to recruit PEX11α and PEX11β to the peroxisomal membrane to form PEX11-enriched patches, leading to peroxisome elongation (Koch et al., 2010). Overall, these findings support a role for Pex11p in peroxisome division through membrane remodeling.

Pex11p, Pex25p and Pex27p in S. cerevisiae share sequence similarity. Likewise, all three proteins share partially redundant function in peroxisome division; overexpression of PEX11, PEX25 or PEX27 leads to cells with increased numbers of small peroxisomes, whereas deletion of any of these genes leads to cells with decreased numbers of enlarged peroxisomes (Smith et al., 2002; Rottensteiner et al., 2003; Tam et al., 2003). Bioinformatic analysis identified Pex11Bp, Pex11Cp and Pex11/25p as putative homologs of Pex11p in fungi (Kiel et al., 2006). This study raised the interesting possibility that additional members of the Pex11p family have yet to be identified. We have now completed the most comprehensive comparative genomics survey of the Pex11 protein family to date in order to understand the basis and significance of the conservation and evolution of the Pex11 protein family in yeast and other eukaryotes. We show that Pex11p itself is highly conserved and ancestral, and has undergone numerous lineage-specific duplications, whereas other Pex11 protein family members are fungal-specific innovations. We functionally characterized the bioinformatically predicted Pex11 protein family members Pex11p, Pex11Cp and Pex11/25p of the yeast Yarrowia lipolytica and showed a conserved role in the regulation of peroxisome size and number for Pex11Cp and Pex11/25p. We also showed that deletion of PEX11 in Y. lipolytica unexpectedly led to cells that lack morphologically identifiable peroxisomes, are defective in peroxisomal matrix protein import and show enhanced degradation of peroxisomal membrane proteins, i.e. they exhibit the classical pex phenotype, which has not been observed to date in cells deleted for PEX11. Our results are consistent with an ancestral role for Pex11p in de novo peroxisome assembly.

Two studies of PEX gene evolution found that most PEX genes are of eukaryotic origin and that there are patterns of peroxin conservation and loss across the lineages of the eukaryotic tree of life (Gabaldón et al., 2006; Schlüter et al., 2006). Since then, new PEX genes have been identified (Managadze et al., 2010; Tower et al., 2011), and the number of eukaryotic genome sequencing projects has increased dramatically, allowing for a more complete picture of peroxisomal protein evolution. We therefore revisited the evolution and distribution of the Pex11 protein family by completing a comparative genomics survey using 125 genomes (supplementary material Table S3) spanning the six eukaryotic supergroups (Walker et al., 2011).

Our analysis showed that Pex11p is conserved throughout the diversity of eukaryotes, and was likely present at the time of the last eukaryotic common ancestor (LECA) (Fig. 1). Numerous genomes contain multiple paralogs of Pex11p; three copies are present in both Homo sapiens and Trypanosoma brucei – i.e. Pex11p and the Pex11p-related proteins GIM5A and GIM5B (Maier et al., 2001) – and five copies are present in the plants Arabidopsis thaliana (Lingard and Trelease, 2006) and Oryza sativa (Nayidu et al., 2008). Three copies of Pex11p are present in the fly Drosophila melanogaster in addition to single paralogs identified previously (Mast et al., 2011; Faust et al., 2012). Additional genomes encode multiple copies of Pex11p; there are two or three copies in Archaeplastida genomes (Chlamydomonas reinhardtii, Micromonas sp., Ostreococcus tauri, Populus trichocarpa and Volvox carteri), and five copies in the rhizarian alga Bigelowiella natans and the ciliate Tetrahymena thermophila. Peroxisomes have not been reported in many parasites (de Souza et al., 2004; Gabaldón et al., 2006; Gabaldón, 2010), and it was therefore not unexpected that Pex11p was not identified in the parasite genomes analyzed in this study – Encephalitozoon cuniculi, Entamoeba histolytica, Giardia sp., Trichomonas vaginalis, Theileria parva and Plasmodium falciparum. Pex11p was identified in the parasite Toxoplasma gondii, although peroxisomes have not been observed in this organism (Ding et al., 2000). Failure to identify Pex11p homologs in the remaining eukaryotic genomes could be because an organism lacks peroxisomes but has never been experimentally characterized, or the genomes encode highly divergent sequences that were not detected by our search algorithms, or because of poor genome assembly and/or coverage.

Unlike for most parasites, a divergent peroxisome called the glycosome has been reported for the Kinetoplastida (Gualdrón-López et al., 2012). The trypanosome genomes encode multiple Pex11 proteins, with Pex11p itself or GIM5 as the top reciprocal pHMMer hits in the T. brucei genome (Fig. 1). An evolutionary relationship was proposed to exist between GIM5 and trypanosome Pex11p on the basis of sequence similarity and common function (Voncken et al., 2003). T. brucei GIM5A and GIM5B share 13% and 14% sequence identity with T. brucei Pex11p (Voncken et al., 2003). Knockdown of GIM5 expression resulted in decreased numbers of larger glycosomes, whereas overexpression of GIM5 resulted in increased numbers of smaller glycosomes (Maier et al., 2001; Voncken et al., 2003). However, we did not find strong support for kinetoplastid Pex11p and GIM5 proteins being homologous, as reciprocal pHMMer searches into kinetoplastid genomes retrieved only the query sequence. Although GIM5 proteins could still be true Pex11p homologs, they are highly divergent in sequence and frequently failed to retrieve Pex11p homologs in non-kinetoplastid genomes (data not shown). A phylogenetic analysis of the Pex11 family proteins in Excavata resolved Pex11p and GIM5 into distinct clades, predating the split of kinetoplastids from their relative, Bodo saltans (supplementary material Fig. S1).

With few exceptions, the fungal genomes in this study all encoded one Pex11p ortholog. Similarly, Pex11Cp was identified across all lineages of Fungi studied, but less frequently in other supergroups. However, paralogs of the remaining Pex11 family proteins were observed in Fungi. Pex27p, initially reported in S. cerevisiae, was identified only in the Saccharomyces species S. paradoxes and S. bayanus (Fig. 1). Pex25p homologs were restricted to the Saccharomycotina, a lineage of ascomycete fungi (James et al., 2006). Pex11Bp, Pex11Cp and Pex11/25p are present in many more fungal genomes than just the filamentous Fungi, as initially reported (Kiel et al., 2006). Pex11Bp homologs were identified in members of the Pezizomycotina, which is a collective of fungal lineages more basal to the Saccharomycotina (James et al., 2006). Pex11/25p homologs, with the exception of a Pex11/25p homolog identified in Y. lipolytica, were restricted to more basal fungal lineages outside the Ascomycota. These findings indicate that all Pex11 family proteins, with the exception of Pex11p itself, appear to be present only in the Fungi, and their distribution might have been sculpted by fungal evolution.

The Pex11 protein family has undergone multiple expansions in diverse eukaryotic lineages that, by and large, are not observed for the other peroxins (Smith and Aitchison, 2013). Our comparative genomics survey confirmed that S. cerevisiae Pex25p and Pex27p are homologs, as a pHMMer search into the S. cerevisiae genome with Pex25p as the query retrieved Pex25p but also Pex27p as the next-best hit (E-value = 5.3×10⁻¹⁰), whereas a pHMMer search with Pex27p as query retrieved Pex27p but also Pex25p as the next-best hit (E-value = 5.6×10⁻¹²). We also investigated the specific evolutionary relationship between Pex25p and Pex27p, and Pex11p. A Pex11p hidden Markov model (HMM) retrieved S. cerevisiae Pex25p and Pex27p, in addition to Pex11p. Similarly, a Pex25p HMM retrieved Pex11p in some Fungi (data not shown). A Pex27p HMM did not retrieve Pex11p, probably because Pex27p is restricted to the genus Saccharomyces.

Are the proposed Pex11 family proteins Pex11Bp, Pex11Cp and Pex11/25p genuine Pex11p homologs? HMMer searches with Pex11Bp or Pex11Cp as queries retrieved Pex11p homologs in yeast, in addition to Pex11Bp or Pex11Cp homologs (data not shown). Likewise, HMMer searches with Pex11p queries retrieved some yeast Pex11Bp and Pex11Cp homologs, as well as Pex11p homologs (data not shown). The identities of some proteins identified in our comparative genomics survey were sometimes ambiguous. For example, Chlamydomonas reinhardtii, Micromonas sp., Volvox carteri, Bigelowiella natans, Tetrahymena thermophila and Naegleria gruberi proteins were retrieved by both Pex11p and Pex11Bp HMMs (supplementary material Table S1). A Pex11Cp HMM retrieved proteins not only in Nematostella vectensis, Bigelowiella natans, Ectocarpus siliculosus, Phytophthora ramorum and Guillardia theta, but also retrieved Pex11γ in H. sapiens, Mus musculus, Canis familiaris and Danio rerio. Pex11/25p was originally identified in Y. lipolytica as a protein with weak similarity to both Pex11p and Pex25p (Kiel et al., 2006); however, our alignments do not show evidence of a clear fusion protein (data not shown). Searches with a Pex11p HMM in the Y. lipolytica genome retrieved the original Y. lipolytica Pex11/25p query in addition to Pex11p. pHMMer searches with Pex11p and Pex11/25p queries occasionally retrieved the same yeast proteins (data not shown).

Clearly, the evolution of the Pex11 protein family is quite complex. We therefore undertook a phylogenetic analysis to elucidate further the evolutionary histories of the Pex11 family proteins in the Opisthokonta supergroup and to clarify the classification of some homologs whose identities were ambiguous based on reciprocal pHMMer searches alone (Fig. 2). A strongly supported clade comprising mammalian Pex11γ and fungal Pex11Cp homologs was observed (Fig. 2). Strikingly, this suggests that these proteins are specific orthologs of the same gene and the products of a gene duplication event occurring in at least the ancestor of opisthokonts. This analysis resolved each of the fungal Pex11 family proteins into well supported clades. Pex11Bp, identified in Pezizomycotina fungi, arose within the Pex11p clade and appears to be a PEX11 gene duplication in these fungal lineages. Pex27p arose in the Pex25p clade; PEX25 and PEX27 have been identified as ohnologs arising from a Saccharomyces whole genome duplication (Byrne and Wolfe, 2005). Based on the comparative genomic and phylogenetic analyses, the conservation and broad distribution of Pex11p identifies it as the ancestral member of the Pex11 protein family and shows that other members of this family, aside from Pex11Cp, are fungal-specific innovations.

This and earlier (Kiel et al., 2006; Schlüter et al., 2006; Gabaldón et al., 2006) comparative genomics surveys and phylogenetic analyses have elucidated the evolution of the Pex11 protein family in Fungi and other eukaryotes. In Y. lipolytica, all members of the Pex11 protein family have been predicted only in silico, and nothing is known regarding their localization in the cell and whether they have a role in peroxisome biogenesis. We therefore chose to functionally characterize the Pex11 protein family, i.e. Pex11p, Pex11Cp and Pex11/25p, in Y. lipolytica. Y. lipolytica shares characteristics with both the pezizomycete and the saccharomycete yeasts, and its genome has been shown previously to be taxonomically informative for understanding the evolution of Rab GTPases in yeasts (Pereira-Leal, 2008).

We first determined whether the Y. lipolytica Pex11 protein family members are peroxisomal (Fig. 3A). We tagged Pex11p, Pex11Cp and Pex11/25p at their C-termini with mCherry (mC) and visualized their localization by confocal microscopy. Pex11p–mC, Pex11Cp–mC and Pex11/25p–mC colocalized with the GFP-tagged peroxisomal matrix protein thiolase (Pot1p–GFP) to punctate structures characteristic of peroxisomes. Subcellular fractionation showed that Pex11p–mC, Pex11Cp–mC and Pex11/25p–mC localized preferentially to a 20,000 g pellet (20KgP) fraction enriched for peroxisomes and not to a supernatant fraction (20KgS) (Fig. 3B). Density gradient centrifugation of the 20KgP fraction showed that these three proteins cofractionate with peroxisomal Pot1p but not with mitochondrial Sdh2p (Fig. 3C). These data show that Pex11p, Pex11Cp and Pex11/25p are peroxisomal in Y. lipolytica.

We next determined whether Pex11p, Pex11Cp and Pex11/25p associate with the peroxisomal membrane. Pex11p is predicted to have two transmembrane helices (amino acids 103–122 and 137–156), but no transmembrane domains were predicted for Pex11Cp or Pex11/25p (http://www.cbs.dtu.dk/services/TMHMM-2.0/). The 20KgP fraction was subjected to hypotonic lysis in dilute alkali Tris buffer followed by centrifugation (Fig. 3D). Pex11p–mC, Pex11Cp–mC and Pex11/25p–mC localized preferentially to the pellet fraction (Ti8P) enriched for membrane proteins like the peroxisomal integral membrane protein Pex2p (Eitzen et al., 1996) and not to the supernatant fraction (Ti8S) enriched for matrix proteins like Pot1p. Ti8P fractions were next subjected to extraction with alkali sodium bicarbonate followed by centrifugation. Pex11p–mC, Pex11Cp–mC and Pex11/25p–mC localized preferentially to the pellet fraction (CO₃P) enriched for integral membrane proteins, as did Pex2p, in contrast to Pot1p (Fig. 3D). Taken together, these observations indicate that Pex11p, Pex11Cp and Pex11/25p are peroxisomal integral membrane proteins.

Transcript and protein levels of PEX11 gene family members PEX11 and PEX25 increase, whereas those of family member PEX27 remain constant, in S. cerevisiae during incubation in oleic acid medium (Erdmann and Blobel, 1995; Smith et al., 2002; Tam et al., 2003). Semi-quantitative RT-PCR was performed to determine transcript levels of the PEX11, PEX11C and PEX11/25 genes in Y. lipolytica during a switch from glucose medium to oleic acid medium (Fig. 3E). Quantification showed that the transcript levels of PEX11, PEX11C and PEX11/25 increased with time of incubation in oleic acid medium, whereas transcript levels for the cytosolic enzyme G6PDH remained constant and served as an internal control (Fig. 3E).

Deletion of genes like PEX3 encoding a peroxin essential for the formation of the peroxisomal membrane leads to abnormal peroxisome assembly and failure of yeast strains to grow on medium containing oleic acid, a carbon source whose metabolism requires functional peroxisomes (Fig. 4A). Growth of pex11CΔ and pex11/25Δ strains on oleic acid medium was similar to that of the wild-type strain. In contrast, the pex11Δ strain, like the pex3Δ strain, displayed no growth on oleic acid medium, and thus Pex11p does not share redundant function with Pex11Cp or Pex11/25p. Our findings also suggest that Y. lipolytica pex11Δ cells lack functional peroxisomes, in contrast to what has been observed for other eukaryotic cells lacking functional Pex11p, which contain fewer and enlarged peroxisomes, prompting further investigation into a possible role for Pex11p in peroxisome assembly in Y. lipolytica.

We next examined peroxisome morphology in the pex11Δ strain, as the inability to grow on oleic acid medium is often indicative of defective peroxisome assembly. Deletion of PEX11 in cells expressing genomically encoded Pot1p–GFP and incubated in oleic acid medium showed fluorescently labeled tendrillar structures (supplementary material Fig. S2A), However, the presence of elongated peroxisomes in these pex11Δ cells was not supported by electron microscopy (Fig. 5A), biochemical analysis (Fig. 6A) or immunofluorescence analysis (supplementary material Fig. S2B) of the pex11Δ strain. Moreover, pex3Δ cells expressing genomically encoded Pot1p–GFP also showed a similar fluorescent elongated and tendrillar pattern in oleic acid medium (supplementary material Fig. S2A); however, pex3Δ cells of Y. lipolytica failed to assemble functional peroxisomes and showed a general mislocalization of peroxisomal matrix proteins to the cytosol (Bascom et al., 2003). We speculate that the elongated fluorescent structures observed in pex11Δ and pex3Δ cells expressing genomically integrated POT1-GFP are an artifact of oligomerization of Pot1p–GFP that is mislocalized to the cytosol. Confocal microscopy showed that monomeric Pot1p–mRFP expressed from plasmid in pex11Δ cells showed a generalized fluorescence pattern characteristic of the cytosol and lacked the readily identifiable punctate peroxisomes still observed in wild-type cells (Fig. 4B).

We also examined whether deletion of PEX11C or PEX11/25 affected peroxisome dynamics, particularly peroxisome division, which is well documented for deletions of genes encoding other Pex11 protein family members. We used confocal microscopy to observe wild-type, pex11CΔ and pex11/25Δ cells expressing genomically encoded Pot1p–GFP in both glucose medium and oleic acid medium. Cells of the pex11CΔ and pex11/25Δ strains in glucose-containing YEPD showed no significant difference in peroxisome number compared to wild-type cells by confocal microscopy (Fig. 4C). In contrast, pex11CΔ and pex11/25Δ cells in oleic-acid-containing YPBO contained fewer peroxisomes compared with wild-type cells (Fig. 4C).

Electron microscopy showed canonical spherical peroxisomes in wild-type cells (Fig. 5B) but no evidence of wild-type peroxisomes in pex11Δ cells (Fig. 5A). However, clusters of small vesicles were observed in some pex11Δ cells (Fig. 5A, boxed regions and insets). These findings, together with evidence from confocal microscopy (Fig. 4B; supplementary material Fig. S2B), show that pex11Δ cells fail to assemble morphologically identifiable, functional peroxisomes. Morphometric analysis revealed that pex11CΔ and pex11/25Δ cells contained fewer and larger peroxisomes compared with wild-type cells (Fig. 5B–D).

Unlike other eukaryotes investigated so far, Y. lipolytica cells deleted for PEX11 do not contain morphologically identifiable peroxisomes. These unexpected findings prompted us to examine the localization of peroxisomal matrix and membrane proteins in this strain. Cells of the wild-type strain and of the deletion strains pex11Δ, pex11CΔ and pex11/25Δ were subjected to differential centrifugation to yield a 20KgS fraction and a 20KgP fraction enriched for ‘mature’ peroxisomes (Titorenko and Rachubinski, 2000; Titorenko et al., 2000). The 20KgS fraction was subjected to ultracentrifugation at 200,000 g to yield a pellet (200KgP) fraction enriched for small vesicles, including small peroxisomal vesicles (Titorenko and Rachubinski, 2000; Titorenko et al., 2000) and a supernatant (200KgS) fraction enriched for cytosol. Immunoblotting was performed with antibodies against matrix proteins with different peroxisomal targeting signals (PTSs): the PTS1-containing protein isocitrate lyase (ICL), a 62-kDa protein recognized by antibodies against the tripeptide Ser-Lys-Leu (SKL) PTS1 consensus sequence, the PTS2-containing protein Pot1p and five acyl-CoA oxidase subunits (Aox1–5) that contain neither a PTS1 nor a PTS2 sequence. These matrix proteins localized preferentially to the 20KgP fraction from wild-type, pex11CΔ and pex11/25Δ cells but, in contrast, were detected predominately in the 20KgS and 200KgS fractions from pex11Δ cells (Fig. 6A). These findings indicate that pex11Δ cells are defective in the import of peroxisomal matrix proteins, regardless of their PTS type.

Like matrix proteins, the peroxisomal membrane proteins Pex2p, Pex3Bp and Pex24p were enriched in the 20KgP fraction from wild-type, pex11CΔ and pex11/25Δ cells (Fig. 6A). Surprisingly, Pex2p and Pex24p were not detected in subcellular fractions from pex11Δ cells, whereas a small amount of Pex3Bp was found in the 200KgP fraction containing small vesicles. These data are consistent with a preferential degradation of peroxisomal membrane proteins in the absence of ‘mature’ peroxisomes in pex11Δ cells and are further supportive of pex11Δ cells being defective in peroxisome assembly.

In wild-type cells, Pot1p is imported into the peroxisomal matrix as a 45-kDa precursor form (pPot1p), where it is proteolytically processed to its mature 43-kDa form (mPot1p) (Titorenko and Rachubinski, 1998). Only mPot1p was detected in the lysates of wild-type, pex11CΔ and pex11/25Δ cells by immunoblot; however, only pPot1p was detected in the pex11Δ lysate (Fig. 6B), further supporting that pex11Δ cells are defective in matrix protein import.

We next examined the effect of overexpressing PEX11, PEX11C or PEX11/25 on peroxisome size and number. Electron microscopy analysis showed that cells overexpressing PEX11C or PEX11/25, but not PEX11, contained increased numbers of smaller peroxisomes in oleic acid medium (Fig. 7A–C). In glucose medium, cells overexpressing PEX11, PEX11C or PEX11/25 all contained increased numbers of peroxisomes (supplementary material Fig. S3). Taken together, the data from our gene deletion and overexpression studies demonstrate a classical Pex11 protein family member role in maintenance of peroxisome size and number for Pex11Cp and Pex11/25p, but also unexpectedly show that Pex11p in Y. lipolytica functions differentially in controlling peroxisome size and number under different carbon conditions and, more notably, functions as a bona fide peroxisome biogenic protein, akin to Pex1p or Pex14p for example (Smith and Aitchison, 2013).

Deletion of the PEX11 gene led to the inability of Y. lipolytica to grow on oleic-acid-containing YPBO agar (Fig. 4A). Transformation of the Y. lipolytica pex11Δ strain with plasmid expressing Y. lipolytica PEX11 re-established the growth of the strain on YPBO (Fig. 8), showing that the failure of pex11Δ cells to grow on oleic acid medium was due to the absence of the Y. lipolytica PEX11 gene. We next tested whether Hansenula polymorpha Pex11p, Pichia pastoris Pex11p, S. cerevisiae Pex11p and Y. lipolytica Pex11Cp or Pex11/25p could correct the aberrant peroxisome assembly in Y. lipoytica pex11Δ cells and restore their ability to grow on oleic acid medium. Expression of plasmid-borne H. polymorpha and P. pastoris PEX11, but not S. cerevisiae PEX11 or Y. lipolytica PEX11C or PEX11/25, restored growth of the Y. lipolytica pex11Δ strain on oleic acid medium (Fig. 8), suggesting that H. polymorpha and P. pastoris Pex11p have retained a peroxisome biogenic function present in Y. lipolytica Pex11p that has been lost in S. cerevisiae Pex11p and is absent in Y. lipolytica Pex11Cp or Pex11/25p (Fig. 2).

Our results show that Y. lipolytica pex11Δ cells do not assemble functional peroxisomes, do not contain morphologically identifiable peroxisomes, mislocalize peroxisomal matrix proteins and more readily degrade peroxisomal membrane proteins, all of which support an unprecedented role for the divisional protein Pex11 in de novo peroxisome assembly. Peroxisomes are linked to the cellular endomembrane trafficking system by the ER. Peroxisomal membrane proteins sample the ER en route to peroxisomes, and peroxisome growth and division are facilitated by proteins and lipids coming from the ER. The functions of the peroxins Pex3p and Pex19p in these processes are well documented in different eukaryotes (Smith and Aitchison, 2013; Tabak et al., 2013). In Y. lipolytica, pex3Δ cells lack any peroxisomal structures (Bascom et al., 2003) and pex19Δ cells, although they do contain morphologically identifiable peroxisomes, mislocalize peroxisomal matrix proteins to the cytosol (Lambkin and Rachubinski, 2001). We assessed the localization of Pex11p–mC, Pex11Cp–mC and Pex11/25p–mC in pex3Δ and pex19Δ cells by confocal microscopy. Distinctive ‘rosette’ structures decorated by Pex11p–mC were observed in pex3Δ and pex19Δ cells grown in either YEPD or YPBO (supplementary material Fig. S4A, boxed regions and insets). Pex11Cp–mC in both pex3Δ and pex19Δ cells exhibited a distinctive cortical and perinuclear localization reminiscent of the ER (supplementary material Fig. S4A) and similar in appearance to the pattern exhibited by the ER-localized marker protein, GFP–HDEL (supplementary material Fig. S4B). In contrast, Pex11/25p–mC exhibited a generalized cortical localization in pex3Δ and pex19Δ cells (supplementary material Fig. S4A). Our observations suggest that in pex3Δ and pex19Δ cells, Pex11p, Pex11Cp and Pex11/25p are localized to membrane structures and, in the case of Pex11Cp, to membranes of the secretory system that do not assemble into functional peroxisomes.

Limited functional analysis has been performed on the novel in-silico-predicted Pex11p-related proteins in Fungi (Kiel et al., 2006). The Pex11 protein family in the fungus Penicillium chrysogenum was recently characterized, and Pex11p and Pex11Cp were found to be peroxisomal, but Pex11Bp was localized to the ER (Opaliński et al., 2012). Deletion of PEX11B or PEX11C had no significant effect on peroxisome number, whereas overexpression of PEX11B resulted in clusters of smaller peroxisomes. The authors concluded that Pex11Bp might play some role in peroxisome biogenesis, although not in de novo peroxisome assembly. The Pex11 protein family in the yeast H. polymorpha was also recently investigated, and Pex11p, Pex11Cp and Pex25p were shown to be peroxisomal (Saraya et al., 2011). Intriguingly, H. polymorpha Pex25p, like S. cerevisiae Pex25p, was shown to be required for de novo peroxisome biogenesis, as the presence of Pex25p, but not Pex11p, was needed to reintroduce peroxisomes into a peroxisome-deficient strain.

We have now demonstrated that Pex11Cp and Pex11/25p regulate peroxisome size and number. In this way, Pex11Cp and Pex11/25p function similarly to Pex11 family proteins in other eukaryotes. Y. lipolytica pex11CΔ or pex11/25Δ cells grown in glucose medium do not contain significantly different numbers of peroxisomes from wild-type cells, whereas pex11CΔ and pex11/25Δ strains grown in oleic acid medium contain fewer peroxisomes than wild-type cells. These observations suggest that Pex11Cp and Pex11/25p in Y. lipolytica might act to regulate peroxisome size and number in a proliferating rather than constitutively dividing peroxisome population.

Pex11 family proteins are associated with the peroxisomal membrane, but their specific membrane topology varies with the organism (Schrader et al., 2012). S. cerevisiae Pex11p, Pex25p and Pex27p appear to be peripherally associated with the peroxisomal membrane, although some controversy still exists regarding the strength of the association between Pex11p and the membrane. Mammalian Pex11 proteins are integral membrane proteins with one or two transmembrane-spanning helices. We showed that Y. lipolytica Pex11p is a peroxisomal integral membrane protein (Fig. 3D), with two membrane-spanning domains predicted. The membrane topology of Y. lipolytica Pex11p therefore more closely resembles its human homologs than S. cerevisiae Pex11p. Organelle extraction also demonstrated that Pex11Cp and Pex11/25p are integral to the peroxisomal membrane (Fig. 3D). Although transmembrane domains were not predicted for Pex11Cp or Pex11/25p, these proteins contain regions of hydrophobicity that might promote integral association with the peroxisomal membrane.

pex3Δ and pex19Δ cells of Y. lipolytica are defective in peroxisome biogenesis and lack wild-type peroxisomes (Lambkin and Rachubinski, 2001; Bascom et al., 2003). Here, we showed that Pex11p is also implicated in peroxisome biogenesis in Y. lipolytica and demonstrated a putative localization of Pex11 family proteins to compartments of the secretory system in pex3Δ and pex19Δ strains (supplementary material Fig. S4A,B). Pex11p–mC was found in distinctive rosette structures in pex3Δ and pex19Δ cells (supplementary material Fig. S4A). Pex11p could be modifying the ER membrane in these strains; however, we did not observe evidence of modified ER in electron micrographs of pex11Δ cells. The specific nature of these structures awaits further investigation.

Why does deletion of Y. lipolytica PEX11 alone produce a defect in peroxisome assembly resulting in the absence of peroxisomes? One explanation could be the nature of peroxisome dynamics in Y. lipolytica, particularly in comparison to those in S. cerevisiae. Yeast peroxisomes increase in size and number when grown in medium containing a non-fermentable carbon course. One key reason why Y. lipolytica is such a tractable model to study peroxisome biogenesis is because of its ability to efficiently utilize hydrophobic substrates such as oleic acid, which is accompanied by extensive proliferation of peroxisomes. However, growth of S. cerevisiae in oleic acid medium has been shown to result primarily in increased peroxisomal size rather than increased peroxisomal number (Tower et al., 2011). In contrast, Y. lipolytica cells contain greatly enlarged and much more numerous peroxisomes when grown in oleic acid medium as compared to growth in glucose medium (Smith and Rachubinski, 2001; Chang et al., 2009; 2013). The increased number of peroxisomes in Y. lipolytica cells, coupled with the responsiveness of this yeast to the presence of fatty acid with regards to its increased capacity to proliferate peroxisomes, could explain why a defect in peroxisome assembly was not heretofore observed in other yeasts deleted for PEX11. The loss of Pex11p in Y. lipolytica could result in the uncoupling of the process of de novo peroxisome biogenesis from the process of peroxisome growth and division. This uncoupling could, in turn, lead to dysregulation of peroxisome biogenesis and loss of the organelle, evidence for which is observed in our inability to detect several peroxisomal membrane proteins by immunoblotting of subcellular fractions of pex11Δ cells (Fig. 6A).

Another reason why a role in peroxisome assembly per se had not been before attributed to the Pex11 protein is that peroxisome assembly, rather than peroxisome division, was an ancestral function of the Pex11 protein and this function has been retained by other members of the Pex11 protein family. For instance, although humans have three Pex11 proteins, only mutation of PEX11β leads to mislocalization of a matrix protein and pathology associated with the PBDs (Ebberink et al., 2012). Similarly, a role in de novo peroxisome biogenesis has been attributed to Pex25p in S. cerevisiae and H. polymorpha (Smith et al., 2002; Saraya et al., 2011). Our phylogenetic analysis demonstrated that Pex25p did not arise by duplication of the gene encoding Pex11p at the Y. lipolytica branch point and resolved the Pex11 family of proteins in Fungi into at least three well-supported clades (Pex11p/Pex11Bp, Pex11Cp, Pex11/25p/Pex25p/Pex27p) (Fig. 2). These data point to multiple expansions of the Pex11 family, giving rise to the various Pex11 proteins in Fungi. More specifically, given that both Pex11p and Pex25p have been implicated in peroxisome biogenesis, our analysis suggests that there were multiple ancestral Pex11 sequences present in the ancestor of Fungi. One of these sequences had a role in peroxisome biogenesis, and different paralogs have retained this ancestral peroxisome biogenic function. The ability of H. polymorpha and P. pastoris PEX11, but not S. cerevisiae PEX11, to correct the growth defect of the Y. lipolytica pex11Δ strain on oleic acid medium (Fig. 8), the metabolism of which requires functional peroxisomes, suggests that H. polymorpha and P. pastoris Pex11 proteins have retained a peroxisome biogenic function that is present in Y. lipolytica Pex11p and that has been lost in S. cerevisiae Pex11p.

Pex11 family proteins can form homo-oligomeric and hetero-oligomeric protein complexes (Schrader et al., 2012), and constructive neutral evolution supports the retention of gene duplication events within such complexes. Thus, while there is one set of evolutionary pressures supporting the expansion of the PEX11 family, which is possibly independent of the function of the Pex11 family protein complex, taxon-specific functional specialization could selectively retain the separate functions of the complex within distinct paralogs or divide the task among all members in an idiosyncratic manner. This hypothesis is supported by different lines of functional evidence in Fungi, as no single Pex11 family protein is essential for peroxisome biogenesis in P. chrysogenum (Opaliński et al., 2012), whereas the ancestral role of PEX11 in peroxisome biogenesis has been retained by PEX11 in Y. lipolytica, and perhaps by H. polymorpha and P. pastoris PEX11, as well as by PEX25 in other yeasts (Smith et al., 2002; Saraya et al., 2011).

Unlike most Pex proteins, the Pex11 protein family has undergone multiple expansions. Our analysis resolved mammalian Pex11α/Pex11β and Pex11γ into separate clades, pre-dating the Opisthokonta, or earlier (Fig. 2). The distribution observed for the Pex11 protein family and its roles in peroxisomal dynamics are indicative of a protein family that has undergone multiple independent gene duplications and losses repeatedly through diverse eukaryotic lineages.

In closing, we have shown a role for the classical divisional peroxin Pex11p in peroxisome assembly in the yeast Y. lipolytica. We have also provided insight into novel members of the Pex11 protein family in Fungi by demonstrating that Pex11Cp and Pex11/25p play a conserved role in the regulation of peroxisome size and number. By undertaking an evolutionary cell biological study on the history of a protein family and studying peroxisomes in diverse eukaryotes, we were able to uncover a new and unexpected role for Pex11p in de novo peroxisome assembly, with implications for human health through increased understanding of a new class of PBD arising from mutation of a gene encoding a Pex11 protein isoform.

Genomes surveyed in this study are listed in supplementary material Table S3. To identify putative Pex11p, Pex11/25p, Pex25p and Pex27p homologs, H. sapiens, S. cerevisiae and/or Y. lipolytica protein sequences (supplementary material Table S1) were used as queries for pHMMer (Eddy, 2009) searches against locally hosted genomes. Candidate homologs with an expect (E-) value less than or equal to 0.05 were subjected to reciprocal pHMMer searches against the query genome and the non-redundant genome. Retrieval of the original query as the top reciprocal pHMMer hit with an E-value less than or equal to 0.05 and retrieval of a named homolog as the top reciprocal pHMMer hit in the non-redundant database were the criteria for the identification of homologs. Verified homologs were aligned using MUSCLE v.3.6 (Edgar, 2004). Hidden Markov models (HMM) (Eddy, 1998) were built from the alignments and used by HMMer 3.0 (Eddy, 2009) for searches against locally hosted genomes. Candidate homologs with E-values less than or equal to 0.05 were verified as described above. Validated homologs from the same eukaryotic supergroup and/or experimentally characterized homologs from the literature were used as queries for pHMMer searches. Newly validated homologs were added to the HMM iteratively, and HMMer searches were repeated until no additional putative homologs were identified. For genomes without protein sequences, nHMMer (Wheeler and Eddy, 2013) was used to search nucleotide databases following the procedure described above.

To identify putative Pex11Bp and Pex11Cp homologs, protein sequences for known Pex11Bp and Pex11Cp homologs (Kiel et al., 2006) (supplementary material Table S1) were aligned using MUSCLE 3.6. HMMs were built from the alignments and used by HMMer 3.0 to search locally hosted genomes. Candidate homologs were verified as described above. Newly validated homologs were added to the HMM iteratively, and HMMer searches were repeated until no additional putative homologs were identified. For genomes without protein sequences, HMMs were used by nHMMer to search nucleotide databases following the procedure described above.

Evolutionary relationships between taxa (Fig. 1) were determined using http://www.ncbi.nlm.nih.gov/taxonomy (Aime et al., 2006; Hibbett, 2006; James et al., 2006; Suh et al., 2006; Wapinski et al., 2007; Boehm et al., 2009; Schoch et al., 2009; Medina et al., 2011; Walker et al., 2011; Ebersberger et al., 2012; Mast et al., 2012).

MUSCLE v.3.6 was used to align validated homologs. ZORRO (Wu et al., 2012) was used to calculate HMM scores for each position in the alignment. Alignments (available upon request) were masked and trimmed in Mesquite v.2.75 (www.mesquiteproject.org). ProtTest v.1.3 (Abascal et al., 2005) was used to determine the optimal model of sequence evolution. MrBayes v.3.2 (Ronquist and Huelsenbeck, 2003) was used for Bayesian analysis to produce posterior probability values. Analyses were run for 1,000,000 Markov chain Monte Carlo generations. Two independent runs were performed, with convergence of the results confirmed by ensuring a Splits Frequency of <0.1. Burn-in values were obtained by removing all trees prior to a graphically determined plateau. Additionally, PhyML v.2.44 (Guindon and Gascuel, 2003) and RAxML v.8.0.0 (Stamatakis, 2006) were used for maximum likelihood analyses, with bootstrap values based on 100 pseudoreplicates of each dataset. RAxML trees were run using CIPRES (http://www.phylo.org/index.php).

Y. lipolytica strains are listed in supplementary material Table S2. Strains were cultured at 30°C unless otherwise indicated. Plasmid-containing strains were cultured in synthetic minimal medium or in YEPD medium or YPBO medium supplemented with hygromycin B (125 µg/ml) or nourseothricin (100 µg/ml). Synthetic minimal medium consisted of 0.67% yeast nitrogen base without amino acids, 2% glucose, 1× complete supplement mixture without leucine or uracil. YEPD, YPBO, YEPA and YNO media have been described previously (Chang et al., 2013).

Y. lipolytica PEX11 (XP_501425.1), PEX11C (XP_501447.1) and PEX11/25 (XP_503276.1) were disrupted using a fusion PCR-based integrative procedure (Davidson et al., 2002).

pTC3 (Lin et al., 1999), pUB4 (Kerscher et al., 2001), pINA443 (Nuttley et al., 1993) and pINA445 (Nuttley et al., 1993) have been described previously. pTC3-NAT was constructed from pTC3 by replacing URA3 with the Streptomyces nouresi NAT gene conferring resistance to nourseothricin. The chimeric genes PEX11-mCherry and PEX11/25-mCherry flanked by promoter and terminator regions of the POT1 gene were inserted into pUB4. PEX11C-mCherry flanked by promoter and terminator regions of POT1 was cloned into pTC3-NAT. H. polymorpha PEX11, P. pastoris PEX11, S. cerevisiae PEX11 and Y. lipolytica PEX11, PEX11C and PEX11/25 flanked by promoter and terminator regions of POT1 were inserted into pUB4. The gene encoding the ER chimeric marker protein GFP–HDEL was inserted into pINA445. pUB4-mRFP-SKL and pUB4-POT1-mRFP have been described previously (Chang et al., 2009; 2013).

Confocal fluorescence microscopy (Tower et al., 2011), electron microscopy and morphometric analysis (Eitzen et al., 1997; Tam et al., 2003), and immunofluorescence microscopy (Tam and Rachubinski, 2002) were performed as described previously.

Images were deconvolved using Huygens Professional Software (Scientific Volume Imaging BV). Three-dimensional (3D) data sets were deconvolved using an iterative Classic Maximum Likelihood Estimation confocal algorithm with an experimentally derived point spread function. Imaris software (Bitplane) was used to create maximum intensity projections of the deconvolved 3D datasets. Transmission images of yeast cells were processed by applying a Gaussian filter in Huygens, and blue color was applied to images using Imaris software. Internal structures in images were removed in Photoshop to better visualize fluorescence data.

Wild-type cells expressing Pex11p–mC, Pex11Cp–mC or Pex11/25p–mC were grown in glucose medium and transferred to oleic acid medium for 8 h. Subcellular fractionation to yield a postnuclear supernatant (PNS) fraction, a 20KgP pellet fraction enriched for peroxisomes, a 20KgS supernatant fraction enriched for cytosol, a 200KgP pellet fraction enriched for small vesicles and a 200KgS cytosolic supernatant fraction was performed as described previously (Tam and Rachubinski, 2002; Chang et al., 2009).

Peroxisomes were purified by isopycnic centrifugation on a discontinuous Nycodenz gradient as described previously (Chang et al., 2009). Extraction of peroxisomes with dilute alkali Tris to yield a pellet (Ti8P) fraction enriched for membrane proteins and a supernatant (Ti8S) fraction enriched for matrix proteins was performed as described previously (Vizeacoumar et al., 2003). Extraction of the Ti8P fraction with alkali Na₂CO₃ and centrifugation to yield a pellet (CO₃P) fraction enriched for integral membrane proteins and a supernatant (CO₃S) fraction enriched for peripheral membrane proteins were performed as described previously (Chang et al., 2009).

Total RNA was isolated using the RiboPure Yeast Kit (Life Technologies). First-strand cDNA was synthesized using SuperScript VILO MasterMix (Life Technologies).

Antibodies have been described previously (Tam and Rachubinski, 2002; Chang et al., 2009). Rabbit anti-DsRed polyclonal antibody was from Takara Bio. Alexa-Fluor-488-conjugated anti-guinea-pig-IgG secondary antibody was from Life Technologies. Antigen–antibody complexes were detected by enhanced chemiluminescence.

We thank Emily Herman for writing scripts used in the comparative genomics survey and for helpful discussion; Hanna Kroliczak, Richard Poirier, Elena Savidov and Nasser Tahbaz for expert technical assistance; and members of the Rachubinski and Dacks laboratories for helpful discussion. We dedicate this manuscript to the memory of our good friend and colleague, Richard Poirier.