PEX11 proteins attract Mff and human Fis1 to coordinate peroxisomal fission

Most eukaryotic cells contain peroxisomes, small organelles essential for several cellular functions mainly associated with the metabolism of lipids. These round organelles harbor a crucial detoxifying function in response to stress assaults. Their function has also been associated with the process of ageing and in antiviral innate immunity (Angermüller et al., 2009; Dixit et al., 2010; Koepke et al., 2007; Wanders and Waterham, 2006). Accordingly, the peroxisomal compartment adapts to changes in cellular microenvironments through proliferation and specific degradation (Fagarasanu et al., 2007; Oku and Sakai, 2010). Studies on peroxisome turnover in mammalian cells revealed a half-life of about two days (de Duve et al., 1974; Huybrechts et al., 2009). To guarantee the maintenance of peroxisomes under varying conditions, mechanisms exist that insure their steady-state either by growth and division of pre-existing organelles or via de novo biogenesis from the endoplasmic reticulum (Geuze et al., 2003; Motley and Hettema, 2007; Toro et al., 2007; Toro et al., 2009). Both pathways seem to continually replenish the pool of peroxisomes in human cells. In vivo analyses using an engineered photoactivatable peroxisomal membrane protein suggested that peroxisomes prevalently arose de novo (Kim et al., 2006). However, the underlying mechanisms and regulations are only poorly understood. Proteins have been characterized that participate in the assembly and function of peroxisomes. These proteins, called peroxins, are encoded by PEX genes (Distel et al., 1996). Mutations in several PEX genes have been associated with the development of lethal genetic diseases characterized as peroxisomal biogenesis disorders or PBDs, e.g. Zellweger syndrome (Steinberg et al., 2006).

The peroxisome growth and division model can be divided into several steps including: (1) peroxisome polarization; (2) membrane protrusion; (3) membrane elongation, (4) import of membrane and matrix proteins; and (5) membrane scission. This process gives rise to the formation a new daughter organelle (Delille et al., 2010; Fagarasanu et al., 2007; Koch and Brocard, 2011; Koch et al., 2010). Within the factors involved in this process, proteins of the PEX11 family and fission factors shared with mitochondria play essential roles. The mammalian genome codes for three different PEX11 proteins, PEX11α, PEX11β and PEX11γ, which exhibit different expression patterns (Abe and Fujiki, 1998; Abe et al., 1998; Li et al., 2002a; Li and Gould, 2002; Schrader et al., 1998; Tanaka et al., 2003). Knock-out mice models have been studied that lack PEX11α and PEX11β (Li et al., 2002a; Li et al., 2002b; Li and Gould, 2002). Taken together, the expression of PEX11α is inducible through variations of the environment and this protein is not essential for the formation of functional peroxisomes. In contrast, absence of the otherwise constitutively expressed PEX11β leads to the development of Zellweger-like symptoms in mice.

Interestingly, proteins of the mitochondrial fission machinery, the tail-anchored protein human Fis1 (hFis1) as well as the dynamin-related protein DRP1, also execute their function at the peroxisomal membrane (Gandre-Babbe and van der Bliek, 2008; Koch et al., 2003; Koch et al., 2005). Protein interaction analyses revealed the association of PEX11 proteins and hFis1 suggesting that PEX11 participates in the recruitment of the membrane fission machinery onto peroxisomes (Kobayashi et al., 2007; Koch et al., 2010). This interpretation was challenged by the finding that the mitochondrial fission factor, Mff acted as efficient DRP1 recruitment factor, assigning hFis1 a rather regulatory function on DRP1 (Otera et al., 2010). Indeed, the fission machinery recruited to peroxisomes must act in tight coordination with membrane elongation factors to facilitate membrane constriction and division. We previously showed that proteins of the PEX11 family represent such a class of proteins (Koch et al., 2010). Their overexpression affected the morphology of peroxisomes in cultured human cells causing their elongation. Ultimately, excess of PEX11 proteins in the cells leads to the formation of structures composed of juxtaposed elongated peroxisomes (JEPs) and to the disappearance of distinct round peroxisomes. Thus, following the formation of JEPs in cells represents a useful tool to analyze the details of the molecular mechanism involved in peroxisome proliferation.

Mechanistic insights into peroxisomal membrane elongation were recently provided through the finding that yeast PEX11 proteins and mammalian PEX11α and PEX11β harbor an amphipathic α-helix. In vitro assays on liposomes with peptides from Saccharomyces cerevisiae (ScPex11p), Hansenula polymorpha (HpPex11p) and Homo sapiens (HsPEX11α) showed their ability to elongate membranes (Opaliński et al., 2011). In the absence of its amphipathic helix, HpPex11p was unable to protrude the peroxisomal membrane in the yeast H. polymorpha.

Most organisms contain several PEX11 proteins. This implies that either each protein plays a different role in the proliferation process or that these represent redundant factors. Recent studies on PEX11 proteins from different yeast species demonstrated that each member of the PEX11 family is involved in a different pathway that leads to the formation of peroxisomes. While Pex11p is involved in growth and division of pre-existing peroxisomes, Pex25p regulates de novo biogenesis from the ER, as pex3Δpex25Δ mutant yeast cells expressing plasmid-born Pex3p were unable to generate peroxisomes (Huber et al., 2012; Saraya et al., 2011). In contrast to the yeast proteins Pex11p and Pex25p, the sequences of the three mammalian PEX11 proteins are closely related. Indeed, they share high amino acid sequence homology (83%) and altogether they are more closely related to Pex11p than to Pex25p. This raises the question how three independent but similar proteins act in concert or influence each other during peroxisome proliferation in mammalian cells.

Using a biochemical approach, we determined the detailed topology of PEX11γ at the peroxisomal membrane and identified functional motifs in its sequence including an amphipathic domain. We demonstrate the requirement for this region for membrane elongation and show that modulating the activity of the amphipathic domain directly influences elongation of the peroxisomal membrane in human cells. Based on peroxisomal targeting of PEX11γ mutated versions, we further elucidate how this protein interacts with the other PEX11 proteins and with members of the fission machinery. Especially, we establish the interaction of Mff with the PEX11 proteins and its localization at the peroxisomal membrane. We propose a mechanistic model for peroxisome proliferation in mammalian cells.

PEX11 proteins have been identified in most eukaryotic organisms and all proteins studied were shown to localize at the peroxisomal membrane (Abe and Fujiki, 1998; Abe et al., 1998; Schrader et al., 1998; Tanaka et al., 2003). However, depending on the organism various topologies were proposed for PEX11 proteins. While ScPex11p was suggested to localize in the inner side of the peroxisomal membrane (Marshall et al., 1996), differential permeabilization experiments showed that the mammalian PEX11 proteins exposed their both termini to the cytosol (Abe et al., 1998; Schrader et al., 1998; Tanaka et al., 2003). Their exact topology in the peroxisomal membrane, however, remains to be elucidated. To tackle this issue, we first performed in silico analysis of PEX11γ revealing two hydrophobic stretches [amino acids (aa) 127–149 and aa 209–227; Fig. 1A]. Then, to determine whether it integrated into the peroxisomal membrane or only attached to it through protein–protein interactions, we carried out a carbonate extraction on human cell lysates expressing PEX11γ–FLAG. Similar to the well-studied peroxisomal membrane protein PEX14 (Will et al., 1999), PEX11γ was not extractable with sodium carbonate and was exclusively detected in the membrane pellet, indicating that it is a true integral membrane protein (Fig. 1B). This raises the question, whether the two hydrophobic stretches identified represent domains that entirely cross the membrane or whether they stand as anchors for PEX11γ, which are only buried into the phospholipid bilayer.

To analyze the membrane topology of PEX11γ, we employed a biochemical approach based on the property of PEG-maleimide (mPEG) to selectively react with reduced cysteines, thereby leading to a mobility shift of proteins in SDS-PAGE. As hydrophilic substance, mPEG cannot cross intact lipid membranes. Notably, wild-type PEX11γ contains six cysteines, all present in the N-terminal moiety (Fig. 2A), allowing the analysis of the membrane topology of this region and the contribution of these cysteines to the function of PEX11γ. To establish the method, an engineered peroxisomal matrix marker containing four accessible cysteines (EGFP–C4-Px) was expressed in human cells and tested for pegylation (Fig. 2B). EGFP–C4-Px was entirely imported into the peroxisomal matrix, as confirmed by microscopic analysis and colocalization with the peroxisomal protein PEX14 (supplementary material Fig. S1). Hence, it was only pegylated when Triton X-100 was added to the protein extracts (Fig. 2B). The small portion of modified EGFP–C4-Px in the absence of detergent might indicate the minor peroxisomal leakage that occurs during the preparation of the cell lysates or small amounts of cytosolic proteins en route to peroxisomes. Further, to insure the reliability of our assay with regard to the study of peroxisomal membrane proteins, we tested the genuine integral membrane protein PEX14 and ectopically expressed PEX3–EGFP. Accessibility experiments showed that the C-terminus of PEX14 faces the cytosol (Oliveira et al., 2002; Will et al., 1999). Accordingly, in our assay its single cysteine (C362) was almost fully pegylated without addition of detergent demonstrating the accessibility of this part of PEX14 from the cytosol (Fig. 2B). The peroxisomal membrane protein PEX3 contains seven cysteine residues. In our pegylation assay two cysteines could be modified without detergent whereas seven cysteines were accessible when the Triton X-100 was added. This is in agreement with structural analyses showing that only one cysteine of PEX3 faces the peroxisomal matrix. Among the six cytosolic cysteines two form a disulfide bridge, two are constrained between helical domains and two are freely accessible from the cytosol (Sato et al., 2010; Schmidt et al., 2010).

With PEX11γ, despite the presence of six cysteines along the N-terminal half, we observed only one clear band-shift after pegylation, indicating that when peroxisomes were kept intact a single cysteine was substrate for mPEG (Fig. 2C). To identify the accessible cysteine, we altered the sequence of the PEX11γ protein by individually replacing all cysteines to alanines and performed new pegylation assays. We reasoned that, if the modifiable cysteine was mutated, no pegylation should be visualized. While mutations at positions 27/28, 59, 91 and 106 did not influence the pegylation state of PEX11γ, mutation C39A hardly showed pegylation (Fig. 2C). This could be due either to steric hindrance or to involvement of these cysteine residues in reversible disulfide bridges. Note that a small fraction of the protein was modified, suggesting that although not fully accessible to mPEG all cysteines might face the cytosol. Although mutation C39A led to a significant increase in non-pegylated PEX11γ species, obviously another cysteine was at least partially accessible for mPEG.

To rule out compensatory effects, we engineered a version of PEX11γ devoid of cysteine residues, PEX11γ^C0 as well as several PEX11γ forms each containing a single cysteine. Expectedly, PEX11γ^C0 was not pegylated, and each mutant with a single cysteine was modified in the presence of detergent (Fig. 2D). Yet again, PEX11γ^C39 was the only version that was fully pegylated in the absence of Triton X-100 as confirmed through the absence of the non-modified PEX11γ band in this lane (Fig. 2D). These observations validate that cysteine 39 is freely accessible from the cytosol. Although PEX11γ cysteines seem to be only partially accessible for mPEG modification due to their close vicinity to the phospholipid bilayer, the presence of disulfide bridges or steric hindrance our results demonstrate that the N-terminus of PEX11γ lies on the cytosolic face of the peroxisomal membrane.

An explanation for the differential pegylation pattern in the N-terminal half of PEX11γ could be that the cysteines influence the stability or localization of PEX11γ. We previously showed that human PEX11 proteins act as membrane elongation factors and their overexpression were always associated with excessive elongation and clustering of peroxisomes eventually leading to the formation of juxtaposed elongated peroxisomes (JEPs) (Koch et al., 2010). Consequently, we reasoned that functional PEX11γ should be able to localize to peroxisomes, induce strong elongation of the peroxisomal membrane and form JEPs. We assessed this property for EGFP–PEX11γ^C39A, EGFP–PEX11γ^C59A and EGFP–PEX11γ^C0, in comparison to wild-type EGFP–PEX11γ using immunofluorescence staining for PEX14. All versions of PEX11γ clearly localized to peroxisomes, and influenced the shape of the peroxisomal membrane inducing elongation and JEP formation as confirmed by statistical evaluation (Fig. 2E). Interestingly, in contrast to cells overexpressing wild-type PEX11γ, cells expressing the C0 mutated version also presented many small peroxisomes suggesting a role for PEX11γ N-terminal region in peroxisome proliferation. However, all tested mutations behaved like wild-type with regard to the formation of JEP structures (see statistics Fig. 2E). We conclude that PEX11γ N-terminus resides in the cytosol and that none of the cysteines influence the membrane-elongation properties of PEX11γ. Consequently, membrane elongation and JEP formation must be allied to the function of PEX11γ C-terminal half. Alternatively, PEX11γ could trigger membrane elongation through activation of another protein. We therefore, sought to analyze the C-terminus of PEX11γ in more details.

The C-terminal half of PEX11γ is highly hydrophobic and contains three predicted helical regions, two of which fulfill the requirements for membrane spanning domains (Fig. 3A). To determine the exact membrane topology of this region, we introduced cysteine residues in or between the hydrophobic helices and performed pegylation assays (Fig. 3A,B).

As shown in Fig. 3B, cysteines introduced at positions 134 (A134C) and 219 (T219C) were inaccessible for pegylation, which demonstrates that these parts of the protein were protected. These are likely inserted into the membrane as suggested from the in silico prediction (Fig. 1A). Because the C-terminus of PEX11γ presents highly ordered secondary structures two of which being hydrophobic, absence of pegylation due to steric hindrance can be ruled out (Fig. 1). Interestingly, for all other PEX11γ mutant versions, namely, A160C, A206C and A238C, an additional band of reduced electrophoretic mobility was observed as compared with wild-type showing that these were pegylated, and accessible, in the absence of detergent (Fig. 3B). These results imply that the two hydrophobic α-helices do indeed represent membrane-buried regions, suggesting anchoring of the protein in the cytosolic face of the peroxisomal membrane. To confirm the activity of mutated PEX11γ proteins we tested for their ability to influence the peroxisomal morphology and form JEPs (Fig. 3C). The effects of expressing either mutated PEX11γ–FLAG were undistinguishable from wild-type as supported by the statistical evaluation of peroxisomal number and shape (Fig. 3C). Fig. 3D shows a summary of PEX11γ cysteine accessibility and the deduced topology of the protein in the peroxisomal membrane.

Our prediction using different in silico approaches (see Materials and Methods) identified another α-helical domain between the two hydrophobic regions (aa 176–193; Fig. 4A). Close inspection of this domain predicts the presence of a strong amphipathic α-helix. Such amphipathic regions have already been identified in some PEX11 proteins including, ScPex11p, HpPex11p and HsPEX11α and HsPEX11β (Opalinski et al., 2011). Those were shown to play a role in membrane elongation in vitro and deletion of this region in HpPex11p led to the absence of peroxisomal elongation in the yeast H. polymorpha; however, such effect has never been shown for the mammalian PEX11 proteins. In vitro studies using PEX11α peptides only showed the ability of the peptides to elongate neutral small unilamellar vesicles (Opaliński et al., 2011</figref>). In contrast to the regions described for PEX11α, the predicted amphipathic α-helix of PEX11γ features a large distinct hydrophobic face (Fig. 4B), and its polar face contains several charged residues. Visualization through 3D modeling clearly shows that the negatively charged residues (D191, E181) are concentrated in the middle of the helix.

To study the potential function of this amphipathic domain on the regulation of PEX11γ, we introduced a proline at position 182, which breaks the helical structure (PEX11γ^A182P), and analyzed the effect of this mutation on peroxisome morphology (Fig. 4B, right). Upon ectopic expression in HEK293T cells, EGFP– PEX11γ^A182P localized to peroxisomes as shown using immunofluorescence staining for PEX14 (Fig. 5A). Remarkably, cells expressing this latter mutation did not display significant peroxisome elongation and did not form JEPs. Instead, peroxisomes presented only faint elongations. In contrast, as we reported previously (Koch et al., 2010), expression of EGFP–PEX11γ typically induced strong membrane elongation and formation of JEPs. These results outline the mechanistic importance of the amphipathic region for PEX11γ function and, for the first time, strongly point to the idea that PEX11γ is capable of protruding the peroxisomal membrane.

Interestingly, we identified a region containing four prolines out of 10 amino acid residues a few residues upstream of the α-helical domain (Fig. 4A). The existence of such proline-rich motif might indicate a regulatory function. Indeed, prolines exist in two isoforms, cis or trans, the latter being the more prominent isoform in natural proteins. Proline isomerization has been suggested to play a role in protein folding and in determining the tertiary structure of proteins as for instance in caveolin-1 (Aoki et al., 2010). Especially, such isomerization can constitute a switch to change the overall protein structure to either activate/inactivate the protein itself or modulate its interactions with other factors (Feng et al., 2011; Sarkar et al., 2011). The isomerization is prevalently achieved through the action of peptidyl-prolyl-cis/trans-isomerases (PPIs), enzymes that catalyze this isomerization step (Lu et al., 2007). Overall, this proline-rich motif might play a pivotal role to regulate the amphipathic region and we sought to modify this motif by mutating the proline at position 158 to an alanine and test the resulting version, PEX11γ^P158A, for its ability to affect peroxisome morphology.

Similarly to EGFP–PEX11γ^A182P, EGFP–PEX11γ^P158A localized to peroxisomes (Fig. 5B). In this case, however, peroxisome elongation was not abolished but only postponed. Up to 72 hours after transfection individual round peroxisomes were present in the cells, but the number of peroxisomes was highly increased as compared with overexpression of EGFP–PEX11γ. At later time points, the cells presented elongated peroxisomes and JEPs formed which correlated with a decrease in the number of individual peroxisomes. This confirms that PEX11γ^P158A can trigger elongation of the peroxisomal membrane. Yet, the kinetics seem to be altered suggesting that the proline-rich motif might indeed be involved in the regulation of the amphipathic region. Fig. 5C shows the statistical evaluation of the effect of these PEX11γ mutations on peroxisomes 2 and 9 days after transfection. Cells expressing EGFP–PEX11γ^P152A,A182P did not form JEPs. This demonstrates the dominant effect of mutation A182P (helix break) and confirms the key function of the amphipathic region of PEX11γ in the molecular mechanism leading to elongation of the peroxisomal membrane (Fig. 5B,C).

Two prolines in the identified proline-rich motif, P165/167, resemble a motif, ‘PLP’, recently identified as binding sequence for hFis1 (Serasinghe et al., 2010), a factor of the peroxisomal fission machinery already shown to interact with the PEX11 proteins (Kobayashi et al., 2007; Koch et al., 2010). To decipher whether this tripeptide is involved in the regulation of PEX11γ, we chose to study the effect of its deletion from the PEX11γ sequence in vivo (Fig. 4A).

EGFP–PEX11γ lacking aa 165–167, EGFP–PEX11^noPLP, localized to peroxisomes and led to effects on the peroxisomal membrane that were similar to those of EGFP–PEX11γ (Fig. 5D). In fact, more peroxisomes were elongated and JEP formation was possibly enhanced as the structures usually appeared bigger. We previously reported that the dramatic elongation of peroxisomes was due to the out-titration of the fission machinery (Koch et al., 2010). Accordingly, JEPs could be dissolved by overexpression of hFis1 and compensated for the high amounts of PEX11 proteins in overexpressing cells. If the PLP motif of PEX11γ is involved in hFis1 recruitment then, overexpression of PEX11γ^noPLP protein should immediately lead to extensive JEP formation. However, such observation is prone to subjective interpretation and thus we sought to perform a reverse experiment.

Rather than following the activity of the mutated PEX11γ version through analysis of peroxisome elongation, we focused on the dissolution of the JEP structures already present in the cell. We co-expressed myc–hFis1 and either wild-type EGFP–PEX11γ or EGFP–PEX11γ^noPLP and evaluated the peroxisome morphology 24 and 48 hours after transfection (Fig. 5E,F). While cells co-expressing EGFP–PEX11γ and myc–hFis1 presented many small, round peroxisomes as expected, the simultaneous expression of myc–hFis1 and EGFP–PEX11γ^noPLP led to the appearance of significantly elongated peroxisomes (Fig. 5E,F). Obviously, in this case the concomitant expression of hFis1 had a mediocre effect on peroxisome morphology, which could be due to weakened interaction between the two proteins. However, we could not detect a significant change in hFis1 pull down using PEX11γ^noPLP as bait (results not shown). Direct interaction between PEX11β and hFis1 was demonstrated in vitro (Kobayashi et al., 2007) and we showed previously that PEX11β and PEX11γ interacted in vivo (Koch et al., 2010). Hence, formation of a ternary complex between PEX11β, PEX11γ and hFis1 might be essential for peroxisome division. Our results suggest that the PLP motif in PEX11γ rather than being directly involved in hFis1 binding influences the activity of hFis1.

The factors involved in peroxisome proliferation in human cells, PEX11 proteins and fission factors, act together to ensure that the number of peroxisomes adapts to the metabolic requirement of the cells. In vitro experiments showed that PEX11β directly interacts with hFis1 (Kobayashi et al., 2007). We reported previously that beside its homodimerization, PEX11γ interacts with both PEX11α and PEX11β. However, the latter two proteins did not co-precipitate. Moreover, in affinity purification all three PEX11 proteins co-purified with hFis1 (Koch et al., 2010). This strongly favors the idea that PEX11 proteins act in concert to execute their functions.

To analyze whether the introduced mutations in PEX11γ, namely A182P, P158A and C0 affected the ability of the protein to interact with hFis1, PEX11β or wild-type PEX11γ, we performed affinity purifications and tested for co-precipitating proteins (Fig. 6A,B). HEK293T cells expressing the appropriate plasmid pairs were lysed in buffer containing 0.2% digitonin. The protein hFis1 co-precipitated with all tested PEX11γ mutations. Yet, the amounts of EGFP–myc–hFis1 proteins obtained using PEX11γ^A182P–FLAG as bait, were significantly higher than those yielded with wild-type PEX11γ (Fig. 6A). Interestingly, the amounts of hFis1 obtained with PEX11γ^P158A were undistinguishable from those obtained with non-mutated PEX11γ indicating that this proline is not required for hFis1 binding per se. Alternatively, another PEX11 protein could influence hFis1 binding. To differentiate between these possibilities, we analyzed the binding of PEX11γ mutants with wild-type versions of either PEX11β or PEX11γ. We identified EGFP–PEX11β and EGFP–PEX11γ in affinity purifications with all mutated versions of PEX11γ (Fig. 6B). Although EGFP–PEX11γ co-precipitated with PEX11γ^A182P–FLAG to a degree similar to the wild-type PEX11γ–FLAG, the amounts of EGFP–PEX11β identified were much lower as visualized by western blotting. This finding suggests that co-precipitation of hFis1 and PEX11γ is not due to association via PEX11β but rather that PEX11γ directly interacts with hFis1.

We re-evaluated this finding through immunoprecipitations from lysates originating from cells co-expressing three proteins, namely, a mutated or non-mutated version of PEX11γ–FLAG, EGFP–myc–hFis1 and either EGFP–PEX11β or EGFP–PEX11γ (supplementary material Fig. S2). Because they have a different molecular weight, the EGFP-tagged proteins can be visualized as two distinct bands through western blot analysis with anti-GFP antibodies. Our results show that upon overexpression of EGFP–PEX11β, little amounts of hFis1 co-precipitated with wild-type PEX11γ but these were increased upon expression of PEX11γ^A182P. In contrast, when EGFP–PEX11γ was expressed more hFis1 was visualized in the affinity purified fractions. However, in the presence of EGFP–PEX11γ, PEX11γ^A182P–FLAG co-precipitated hFis1 in amounts similar to the non-mutated PEX11γ–FLAG (supplementary material Fig. S2). These data demonstrate that hFis1 indeed directly interacts with PEX11γ and that this interaction strongly depends on the interaction of PEX11γ with other PEX11 proteins.

Previous studies on yeast Pex11p have suggested that its homodimerization depends on the presence of a disulfide bridge through cysteine at position 3 (Marshall et al., 1996). In contrast, a version of human PEX11γ lacking cysteine residues (C0) co-precipitated hFis1, PEX11β and wild-type PEX11γ showing that these interactions did not require the formation of disulfide bridges (Fig. 6A,B). Hence, a mechanism similar to that proposed for ScPex11p is unlikely for the human PEX11γ.

In vivo as well as in vitro analyses established that the mitochondrial fission factor, Mff, recruits DRP1 to the mitochondrial outer membrane (Otera et al., 2010). Knock-down experiments also showed that peroxisomes elongated in the absence of Mff (Gandre-Babbe and van der Bliek, 2008). However, whether Mff localizes to peroxisomes and cooperates with PEX11 proteins has not been studied. Hence, we tested whether the PEX11 proteins also interacted with Mff to coordinate peroxisomal fission. Our pull-down experiments show that Mff co-precipitated with both PEX11β and PEX11γ, however, the co-purified amounts were higher with PEX11β (Fig. 6C). When PEX11γ^A182P was used as bait, equal amounts of Mff as compared with PEX11γ could be visualized in the affinity purifications (Fig. 6C). This is in opposite to the results obtained with hFis1 (Fig. 6A). To unambiguously determine whether a portion of Mff resides at the peroxisomal membrane as expected from the co-immunoprecipitation experiments, we co-expressed EGFP–Mff and a peroxisomal matrix marker, mCherry–Px, and stained the mitochondria with Mitotracker (Fig. 6D). Most EGFP signal localized to mitochondria that appeared fragmented as previously reported for overexpression of Mff (Gandre-Babbe and van der Bliek, 2008; Otera et al., 2010). In addition, a significant portion was present exclusively on peroxisomes. This finding was confirmed through 3D object analyses. To avoid unclear co-localization due to an overlap between the peroxisomal and mitochondrial signals, we first removed all Mff signals that also contained mitochondrial staining. Then, intersections between Mff and peroxisomes were made visible (Fig. 6D). Although we cannot exclude that the peroxisomes present near mitochondria also contain Mff, for our co-localization study we only considered isolated peroxisomes. Interestingly, similar to the effect of hFis1 expression, when Mff was overexpressed PEX11γ-expressing cells did not contain JEP (Fig. 6E). Moreover, in contrast to hFis1, upon co-expression of Mff no elongated peroxisomes were invoked by the expression of PEX11γ^noPLP (Fig. 5E, Fig. 6F). Overall, these findings confirm the involvement of Mff in PEX11-driven peroxisomal fission and suggest a slightly different role for PEX11β and PEX11γ.

Because concomitant expression of PEX11β and PEX11γ seemed to differentially influence the binding capabilities of PEX11γ to the fission machinery, we analyzed the effects of their co-expression on peroxisomes. Overproduction of PEX11β alone first leads to an increased number of small and round peroxisomes and, at a later time point, to the formation of peroxisome clusters similar to PEX11γ (Fig. 7A) (Delille et al., 2010; Koch et al., 2010). Here, upon overexpression of mRFP–PEX11β and EGFP–PEX11γ, elongated peroxisomes and JEPs were observed even 24 h after transfection (Fig. 7A). Remarkably, in addition to JEPs the cells contained many small peroxisomes, indicating that the activities of PEX11β and PEX11γ on peroxisome proliferations are different and that these both proteins play an important role in this process. In contrast, concomitant expression of mRFP–PEX11β and EGFP–PEX11γ^A182P did neither lead to strong peroxisome elongation nor to the formation of JEPs (Fig. 7A). Rather, we observed a dramatic increase in the number of small peroxisomes (see statistics Fig. 7A). The manifestation of a high number of peroxisomes was also observed 72 h after transfection a time point at which overexpression of PEX11β alone led to the formation of JEPs. Although unable to protrude the peroxisomal membrane, PEX11γ lacking its proline at position 182 (PEX11γ^A182P) interacted with hFis1 suggesting the requirement for a subtle interplay between members of the peroxisome proliferation machinery for proper maintenance of the organelle. A regulated cascade of molecular interactions between members of the PEX11 protein family and the fission machinery seems to coordinate the sequence of events leading to peroxisome proliferation.

The number of peroxisomes per cell is rigorously maintained through the coordination of proliferation via de novo biogenesis from the ER (Kim et al., 2006; Toro et al., 2009), growth and division from pre-existing peroxisomes (Purdue and Lazarow, 2001) and degradation of the organelles by pexophagy (Oku and Sakai, 2010). It has been shown that the various members of the PEX11 protein family are involved in pathways that lead to the formation of peroxisomes in yeast, plant and human cells (Koch et al., 2010; Lingard and Trelease, 2006; Huber et al., 2012). Besides their interaction with the fission machinery, in some PEX11 proteins, an amphipathic α-helix was suggested to provide a mechanism for their mode of action on the peroxisomal membrane (Opalinski et al., 2011).

Here, we present a detailed analysis of PEX11γ topology at the peroxisomal membrane, confirming that both N- and C-termini face the cytosol. Using mPEG, a cysteine-selective reagent, we mapped the accessibility of all six cysteines of PEX11γ along its N-terminal half (Fig. 2). Interestingly, only cysteine at position 39 seemed to be freely accessible for pegylation, the others might be involved in a tertiary structure protecting them from pegylation. As control, we engineered PEX11γ^C0 devoid of cysteines. This mutated version still localized to peroxisomes and acted on the peroxisomal membranes similar to PEX11γ (Fig. 2E). Thus, our finding implies that the region required for membrane elongation resides in the C-terminal half of the protein. Nevertheless, cysteines in the N-terminal moiety could be involved in maintaining the structure of the protein similar to ScPex11p which was proposed to homodimerize via a disulfide bridge at cysteine 3 (Marshall et al., 1996). However, our immunoprecipitation experiments confirmed that PEX11β, wild-type PEX11γ and hFis1 co-purified with PEX11γ^C0 (Fig. 6A,B). Ultimately, this shows that no covalent cysteine bond is required for protein–protein interactions which had already been implied by the necessity to use the mild, membrane-preserving detergent digitonin during immunoprecipitations. Indeed, in previous experiments we showed that the hydrophobic regions of PEX11β and also PEX11α were required for interaction with PEX11γ and hFis1 (Koch et al., 2010). If at all, the cysteines of PEX11γ might rather stabilize the overall structure of the cytosolic N-terminal part of the protein or strengthen protein interactions through transient disulfide bridges.

Close inspection of the C-terminal half of PEX11γ revealed two membrane-inserted helices that anchor and span an amphipathic helical region in the peroxisomal membrane (Figs 3, 4, Fig. 5A). Herein, the amphipathic helix is inserted from the cytosol to generate positive membrane curvature necessary for protrusion and subsequent fission of the peroxisomal membrane. Noteworthy, in contrast to other PEX11 proteins the amphipathic region of PEX11γ is located at its C-terminus. Besides, the membrane-anchored PEX11γ (Fig. 1) stands out within members of the PEX11 protein family because it features two membrane-buried segments, whereas in PEX11α and PEX11β a single hydrophobic region was predicted, and shown to be required for their proper localization (Koch et al., 2010). The two membrane-bound regions of PEX11γ might enable the protein to properly position its amphipathic region and maintain protein–protein interaction spatially controlled in the membrane. Here we show that the amphipathic domain of PEX11γ is indeed necessary for both membrane elongation and interaction with PEX11β and hFis1 (Fig. 5A–C, Fig. 6A,B). It might be regulated by a proline-rich motif, where cis/trans isomerization could represent a plausible regulatory mechanism. In fact, mutation of the proline at position 158 of PEX11γ led to delayed peroxisome elongation and correlated with an increase in peroxisome number at early time points (Fig. 5B,C). Moreover, deletion of the PLP motif in PEX11γ strongly affected the properties of JEPs with regard to their susceptibility to fission factors (Fig. 5D–F). Our findings that co-production of PEX11γ and PEX11β led to a dual phenotype suggest that these two proteins act at different levels in the process of peroxisome proliferation (Fig. 7A). While PEX11β increases the number of peroxisomes, PEX11γ promotes elongation of the peroxisomal membrane. However, these two mechanisms are not mutually exclusive and most likely influence each other to ensure adequate regulation.

Obviously, two cases can be distinguished: (1) PEX11γ homodimerizes, and cannot interact with hFis1, or (2) PEX11γ heterodimerizes, and efficiently binds hFis1. In the first case, PEX11γ would be involved in membrane elongation, in the second case PEX11γ would rather act on the fission machinery. Such explanation is plausible since the events of membrane elongation and fission, must be coordinated and are unlikely to occur simultaneously. Differential di- or even oligomerization and interaction with hFis1 synchronized with steric adjustment of membrane-bending motifs might represent a mechanism to control the progress of peroxisomal proliferation.

Additionally, a third player of the fission machinery, Mff, should now be included in the scheme of PEX11-controlled peroxisomal fission. Obviously, PEX11 proteins do not only interact with hFis1, but also with Mff which we clearly found to be present on peroxisomes (Fig. 6C,D). Furthermore, Mff counteracted JEPs formation by wild-type PEX11γ and mutated PEX11γ^noPLP (Fig. 6E,F), suggesting a role in peroxisome proliferation similar to that in mitochondrial proliferation as recruitment factor for DRP1 (Gandre-Babbe and van der Bliek, 2008; Otera et al., 2010). In contrast to hFis1, Mff is the first protein of the fission machinery that showed different interaction properties for the various PEX11 proteins suggesting a stronger interaction with PEX11β. Thus, it is conceivable that PEX11β recruits or positions the fission machinery whose action would then be triggered by PEX11γ once the peroxisomal membrane is properly remodeled and protruded for fission (Fig. 7B).

In this work, we have established new tools, namely PEX11 proteins that target to peroxisomes but lose their ability to perform their task, that help decipher the cascade of molecular interactions leading to polarized elongation of the peroxisomal membrane and fission of the organelle. Our studies on mutated versions of PEX11γ show that this protein has an uncharacteristic topology at the peroxisomal membrane (Fig. 3D).

The exact mechanism by which DRP1 severs organellar membranes in vivo and the function of hFis1 and Mff in this process are poorly understood. Resolving these issues will be crucial for a deeper understanding of both peroxisomal and mitochondrial proliferation. With regard to the proliferation of peroxisomes several questions remain especially on the mode of lipid recruitment during proliferation of the organelles. Presumably, growth/division and de novo biogenesis of peroxisomes integrate at the stage of lipid uptake. Whether PEX11 proteins are involved in this process remains to be elucidated.

Interestingly, a correlation has been noted in some mammalian systems between the high number of peroxisomes in hippocampal neurons and protection against neurodegeneration (Santos et al., 2005). Peroxisomes may indeed play a primordial role in the protection against accumulation of β-amyloid peptides and their function might reduce the pathological development of Alzheimer′s disease (Kou et al., 2011). For instance, similar to increased mitochondrial fission during apoptosis, peroxisome elongation might represent a good assessment for the pathogenesis of neurological disorders. Knowing the factors involved in this process and the effects caused by their malfunction could lead to the development of new diagnostic targets to differentiate between peroxisomal and mitochondrial disorders.

Plasmids coding for EGFP–HsPEX11γ, HsPEX11γ–FLAG, EGFP–myc–hFis1, myc–hFis1 and mCherry–Px were described before (Koch et al., 2010). EGFP–Mff was purchased from GeneCopoeia. EGPF–C4–SKL was engineered by insertion of a linker (annealed oligonucleotides CB396/397) in the EGFP–C1 (Clontech, BglII/EcoRI). EGFP–Px was described before (Stanley et al., 2006). For mRFP–HsPEX11β and –PEX11γ, the coding sequence of PEX11β and PEX11γ were amplified and inserted into pcDNA3.1-mRFP (EcoRI/XhoI) obtained from Jeffrey Gerst (Weizmann Institute of Science, Israel). Mutations were introduced in the original EGFP–HsPEX11γ-encoding plasmid through site-directed mutagenesis via PCR with oligonucleotide pairs harboring the respective mutation (see supplementary material Table S1), followed by DpnI digestion. All mutations were verified by sequencing of the obtained plasmids.

TMPred was used for predicting the hydrophobic segments (http://www.ch.embnet.org/software/TMPRED_form.html). The α-helical content was predicted using different algorithms, JPred (Cole et al., 2008), NetSurfP (Petersen et al., 2009), JUFO (Meiler and Baker, 2003) and PSIPred (Jones, 1999).

Human embryonic kidney cells (HEK293T) were cultured in DMEM (+10% FCS, +1% penicillin/streptomycin; PAA Laboratories, Pasching, Austria) at 37°C (5% CO₂). Cells were transfected using FuGene6 (Roche) or nucleofected (Amaxa). For microscopic analysis, cells were fixed with 3.7% formaldehyde in PBS (15 min) and embedded in Mowiol supplemented with 25 mg/ml DABCO (Carl Roth, Karlsruhe, Germany). Immunofluorescences were carried out as described previously (Koch et al., 2010). Mitochondria were stained using Mitotracker Deep Red FM (Molecular Probes). Nuclei were counterstained with Hoechst 33342 (Molecular Probes). For immunoprecipitation, cells were lysed 48 hours after transfection in lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA) containing 0.2% digitonin. After centrifugation (800 g, 5 min) cells were transferred onto columns containing anti-FLAG M2 affinity gel (Sigma-Aldrich), incubated (2 hours, 4°C) and washed extensively (0.5 M Tris-HCl pH 7.4, 1.5 M NaCl). Immunoprecipitates were eluted from the columns using 3× FLAG peptides (150 ng/µl; Sigma-Aldrich) for 30 minutes at 4°C. Aliquots from inputs (I) and eluates (E) were analyzed by western blot.

HEK293T cells expressing the appropriate proteins were harvested 48 h after transfection, washed in PBS, and resuspended in pegylation buffer (Antonenkov et al., 2004, without DTT). Cells were lysed with a Potter-Elvehjem (1500 rpm, 10 strokes) and cell debris were pelleted by centrifugation (800 g, 10 min). Equal fractions of the supernatant were incubated with 4 mM mPEG [O-2(-maleimidoethyl)-O′-methylpolyethylene glycol 5.000; Sigma-Aldrich] or 0.1% v/v Triton X-100 or both for 1 hour at 4°C. In control experiments, DTT (1 mg/ml) was added (not shown). Pegylation was stopped by addition of 1 µl β-mercaptoethanol to the samples. Equal fractions were loaded onto an SDS gel and analyzed by western blotting.

HEK293T cells expressing the appropriate protein were harvested 48 hours after transfection, washed in PBS and resuspended in lysis buffer (10 mM Tris-HCl pH 7.4, 1 mM EDTA) with complete protease inhibitors (Roche). An aliquot of these crude extracts (IP) was removed and stored. Cells were homogenized using a Potter-Elvehjem (1500 rpm, 10 strokes) and centrifuged (100,000 g, 60 min). The pellet was resuspended in high salt buffer (10 mM Tris pH 7.4, 0.5 M KCl) and mixed for 30 min. After centrifugation (100,000 g, 60 min), the supernatant (S1) was stored and the pellet was resuspended in carbonate buffer (100 mM Na₂CO₃). After mixing on a rotating wheel for 30 min and centrifugation (100,000 g, 60 min), the supernatant (S2) and final membrane pellet (MP) were stored. Equal fraction of IP, S1, S2 and MP were loaded onto an SDS gel for western blot analysis.

Rabbit anti-HsPEX14 antibodies were a kind gift from Ralf Erdmann (Ruhr University, Bochum, Germany). Alexa Fluor 594 donkey anti-rabbit and Alexa Fluor 488 goat anti-mouse antibodies were purchased from Molecular Probes (Invitrogen). Rabbit anti-GFP antibodies were a kind gift from Michael Rout (The Rockefeller University, New York, USA). Deep Red donkey anti-mouse antibodies were purchased from Jackson Laboratories. Anti-FLAG M2 monoclonal antibodies (HRP conjugated) was purchased from Sigma-Aldrich. HRP-conjugated sheep anti-mouse and donkey anti-rabbit antibodies were purchased from GE Healthcare.

Confocal images were acquired on a LSM510META, Zeiss (Neofluar 100×1.45, pixel size 45×45 nm, z-stacks 200 nm, 1.6 µs pixel dwell time, 12-bit) using a 405 nm laser (BP420–480) for Hoechst staining, 488 nm laser (BP500–550) for GFP, 561 nm laser (LP585 or BP 575–615) for mCherry/mRFP and 633 nm laser (LP650) for deep red dyes. Cells were randomly chosen, and detector gain and amplifier offset were adjusted to avoid clipping. All images were deconvolved using the QMLE algorithm of Huygens Professional (SVI, The Netherlands), projected (maximum intensity) and adjusted in ImageJ. Object colocalization analysis was performed in Huygens Professional. Briefly, objects were created for peroxisomes (mCherry–Px), mitochondria (Mitotracker Deep Red FM) and Mff (EGFP–Mff). Then each Mff-object that also contained mitochondrial staining was removed and excluded from further analysis. Intersections of the residual Mff objects with peroxisomal objects were calculated. Quantifications of peroxisome number were performed in ImageJ. A maximum intensity projection was segmented and thresholded using k-means clustering (ij-Plugins.sourceforge.net/plugins/segmentation/index.html). Peroxisomes were classified as either normal (area <1 µm²; Feret's diameter <0.75 µm; circularity >0.8), elongated (area 0.1–1 µm²; Feret's diameter >0.75 µm; circularity <0.8) or JEPs based on the area of the peroxisomes (>1 µm²). At least twenty randomly chosen cells from three independent experiments were analyzed. The data are presented as mean ± standard error of the mean (s.e.m.). The statistical significance was evaluated as indicated using the Wilcoxon rank sum test (http://faculty.vassar.edu/lowry/VassarStats.html).

We thank Christine David for the plasmid pEGFP-N1-PEX3.