Pex11pβ-mediated growth and division of mammalian peroxisomes follows a maturation pathway

Peroxisomes are ubiquitous subcellular organelles, which multiply by growth and division but can also form de novo via the endoplasmic reticulum. Growth and division of peroxisomes in mammalian cells involves elongation, membrane constriction and final fission. Dynamin-like protein (DLP1/Drp1) and its membrane adaptor Fis1 function in the later stages of peroxisome division, whereas the membrane peroxin Pex11pβ appears to act early in the process. We have discovered that a Pex11pβ-YFPm fusion protein can be used as a specific tool to further dissect peroxisomal growth and division. Pex11pβ-YFPm inhibited peroxisomal segmentation and division, but resulted in the formation of pre-peroxisomal membrane structures composed of globular domains and tubular extensions. Peroxisomal matrix and membrane proteins were targeted to distinct regions of the peroxisomal structures. Pex11pβ-mediated membrane formation was initiated at pre-existing peroxisomes, indicating that growth and division follows a multistep maturation pathway and that formation of mammalian peroxisomes is more complex than simple division of a pre-existing organelle. The implications of these findings on the mechanisms of peroxisome formation and membrane deformation are discussed.


Introduction
Peroxisomes are ubiquitous subcellular organelles that compartmentalize enzymes responsible for several crucial metabolic processes such as fatty acid oxidation, biosynthesis of ether phospholipids and metabolism of reactive oxygen species (Wanders and Waterham, 2006). Peroxisomes are highly flexible and dynamic organelles that rapidly assemble, multiply and degrade in response to metabolic needs.
The underlying mechanisms leading to the formation and multiplication of peroxisomes are still a matter of debate. Exciting advances in peroxisome research have recently challenged the general view that peroxisomes are autonomous organelles. Compelling evidence has been presented that peroxisomes can arise de novo from the endoplasmic reticulum (ER) via a maturation process (Hettema and Motley, 2009;Hoepfner et al., 2005;Titorenko and Rachubinski, 1998;Titorenko and Rachubinski, 2001). The peroxins Pex3p, Pex19p and Pex16p are required to elaborate and maintain the peroxisomal membrane, and are implicated in the import of peroxisomal membrane proteins (PMPs) (Fujiki et al., 2006;Heiland and Erdmann, 2005). Their loss of function leads to peroxisome absence, whereas re-introduction of the missing genes restores peroxisome formation in yeast and mammals, which has lead to the proposal that there can be de novo synthesis of peroxisomes (Faber et al., 2002;Haan et al., 2006;Hoepfner et al., 2005;Kragt et al., 2005;Matsuzono et al., 1999;Muntau et al., 2000;South and Gould, 1999;Titorenko and Rachubinski, 2001). Peroxisomes can also multiply by growth and division (reviewed by Fagarasanu et al., 2007;Schrader and Fahimi, 2006). To what extend these alternative pathways of multiplication contribute to peroxisome formation might vary among organisms.
In elegant studies in Saccharomyces cerevisiae it has recently been shown that in wild-type cells, peroxisomes multiply by growth and division and do not form de novo (Motley et al., 2008), whereas in mammalian cells both mechanisms may operate simultaneously (Kim et al., 2006).
Peroxisomes in mammalian cells often exhibit tubular morphology (Schrader et al., 1996;Yamamoto and Fahimi, 1987). These elongated peroxisomes have been observed to divide into spherical organelles, thus contributing to peroxisome multiplication (Schrader and Fahimi, 2006;Yan et al., 2005). Peroxisome elongation and division is promoted by extra-and intracellular stimuli such as growth factors, fatty acids and reactive oxygen species (Schrader and Fahimi, 2006;Sinclair et al., 2009;Yamamoto and Fahimi, 1987;Yan et al., 2005), or during the cell cycle in plants (Lingard et al., 2008). In the last few years great progress has been made in the identification of components of the growth and division machinery of peroxisomes in yeast, mammals and plants (Fagarasanu et al., 2007;Lingard et al., 2008;Lingard and Trelease, 2006;Orth et al., 2007;Schrader and Fahimi, 2006;Thoms and Erdmann, 2005). Growth and division of peroxisomes in mammalian cells proceeds through at least three distinct steps: elongation of peroxisomes, constriction of the peroxisomal membrane, and final division. The mammalian Pex11p has three isoforms, and one of them, Pex11p, has been demonstrated to promote peroxisome elongation and division , whereas the dynamin-like GTPase DLP1 and its membrane adaptor Fis1 appear to mediate final fission of the peroxisomal membrane (Kobayashi et al., 2007;Koch et al., 2003;Koch et al., 2005;Schrader, 2006). Interestingly, DLP1, Fis1 and the recently identified Mff (Gandre-Babbe and van der Bliek, 2008) are present on both peroxisomes and mitochondria, thus indicating organelle crosstalk and coordinated biogenesis (Camoes et al., 2009;Muntau et al., 2000;Schrader and Yoon, 2007). DLPs and Fis1p are also involved in peroxisome division in yeast and in plants (Hoepfner et al., 2001;Kuravi et al., 2006;Lingard et al., 2008;Mano et al., 2004;Motley et al., 2008;Vizeacoumar et al., 2006), whereas de novo formation of peroxisomes appears to be independent of DLPs (Jourdain et al., 2007;Motley and Hettema, 2007;Nagotu et al., 2008).
Pex11p and Pex11p-related proteins are unique components of the peroxisomal division and proliferation machinery (Thoms and Erdmann, 2005;Yan et al., 2005). A loss of Pex11p is usually accompanied by a reduced number of peroxisomes. Mammalian Pex11p, Pex11p and Pex11p are thought to be integral membrane proteins with their N-and C-termini exposed to the cytosol. All Pex11 proteins interact with themselves, and are likely to form oligomers (Li and Gould, 2003;Lingard et al., 2008;Marshall et al., 1996;Marshall et al., 1995;Rottensteiner et al., 2003;Tam et al., 2003). A Pex11 knockout mouse is viable, with no obvious effect on peroxisome number or metabolism (Li et al., 2002a). By contrast, knockout of Pex11p causes neonatal lethality and is accompanied by several defects reminiscent of Zellweger syndrome (Li et al., 2002b). Pex11p is not directly involved in final fission of peroxisomes, and most probably functions upstream of DLP1 by promoting peroxisome elongation (Kobayashi et al., 2007;Koch et al., 2004;Koch et al., 2003). Recently, an interaction of Pex11p with Fis1 has been proposed (Kobayashi et al., 2007), and interactive roles of mammalian (and plant) Pex11p, Fis1p and DLP are beginning to emerge Koch et al., 2003;Koch et al., 2005;Li and Gould, 2003;Lingard et al., 2008). However, the biochemical properties of Pex11p are still a matter of debate (Thoms and Erdmann, 2005). Efforts to dissect the growth and division process of peroxisomes have so far been focussed on the manipulation of DLP1 and Fis1 function, which are supposed to act late in the growth and division process. In the present study, we have further dissected the process of peroxisomal growth and division by manipulation of Pex11p, which is thought to act early on in this process. We provide the first evidence that Pex11p-mediated growth and division of peroxisomes follows a multistep maturation pathway that is initiated at the pre-existing peroxisome.

Pex11p-YFP induces tubular peroxisomal accumulations (TPAs) and inhibits the formation of spherical peroxisomes
In order to manipulate and dissect peroxisomal growth and division, we generated truncated and tagged versions of Pex11p. These constructs were expressed in COS-7 cells (Fig. 1A), which were then screened for alterations of peroxisomal growth and division. Expression of Myc-tagged (N-or C-terminally) or untagged Pex11p has been shown to induce a pronounced elongation of peroxisomes (Fig. 1B), which is followed by peroxisome division and the formation of small spherical peroxisomes ( Fig. 1C) (Koch et al., 2003;Lingard and Trelease, 2006;Schrader et al., 1998). When COS-7 cells were transfected with Pex11p fused to monomeric YFP m at the C-terminus (Pex11p-YFP), striking morphological changes of the peroxisomal compartment were detected ( Fig. 1D-H). A minority of cells (1-5%) contained spherical peroxisomes (Fig. 1D,G), whereas in the majority the peroxisomes were tubular forming 2751 Pex11p-mediated peroxisome maturation accumulations. These structures were found to be distributed uniformly within the cytoplasm (Fig. 1E), or to concentrate at one side of the nucleus ( Fig. 1F; supplementary material Fig. S1J-L). Usually, several small tubular accumulations were observed, which appeared to be thicker than the tubular peroxisomes found in controls (Fig. 1B). This suggests that the tubular morphology is stabilized by expression of Pex11p-YFP. Cells with only a few tubular structures were also detected and increased with time in culture (Fig. 1G,H). Many of the structures, which we termed TPAs (tubular peroxisomal accumulations), had a twisted, curly or cork-screw-like appearance. Interestingly, the formation of spherical peroxisomes by constriction and fission of tubular peroxisomes, which is promoted by the expression of Myc-tagged (N-or C-terminally) or untagged Pex11p (Koch et al., 2003;Lingard et al., 2008;Schrader et al., 1998), was completely inhibited. TPAs were not observed to constrict or to divide and to give rise to spherical organelles (Fig. 1G). Thus, other peroxisomal structures than TPAs appeared to be absent from Pex11p-YFP-expressing cells (Figs 4, 5; see supplementary material Fig. S2). Similar morphological changes of peroxisomes after expression of Pex11p-YFP were obtained with a variety of other cell lines of human (HepG2, HeLa) or rodent (Fao, AR42J) origin (unpublished data), with COS-7 cells exhibiting very prominent TPAs. As pre-peroxisomal membrane compartments have been associated with the formation of peroxisomes, we investigated the TPAs in more detail.

TPA formation depends on Pex11p-YFP
In order to demonstrate that TPA formation was not just due to the overexpression of Pex11p or a YFP-or GFP-tagged peroxisomal membrane protein, several control experiments were performed.
Transfection of COS-7 cells with a Pex11p construct bearing the YFP-tag at the N-terminus (YFP-Pex11p), or with Pex11p-YFP, a Pex11p isoform with similar membrane topology, did not result in TPA formation ( Fig. 2A-D). Pex11p-YFP had only a slight effect on peroxisome elongation, as described for Myc-tagged Pex11p (Delille and Schrader, 2008;Schrader et al., 1998), whereas YFP-Pex11p was able to induce peroxisome elongation and division as demonstrated for Myc-tagged Pex11p ( Fig. 2M; supplementary material Fig. S3). Similarly, expression of Pex3p-GFP either alone or in combination with Pex11p-Myc did not promote TPA formation (Fig. 2E,F). Replacing the YFP tag with a HaloTag, a different monomeric protein of similar size (Los et al., 2008), also resulted in the formation of peroxisomal aggregates, whereas they were not found after expression of a Pex16p-HaloTag fusion protein (Fig. 2G,H; supplementary material Fig. S4). In addition, we expressed Pex11p-YFP in Pex19p-deficient fibroblasts. In these cells peroxisomes are completely absent (Muntau et al., 2003) and Pex11p-YFP is exclusively targeted to mitochondria. No clustering of mitochondria was observed under these conditions. Interestingly, a fragmentation of mitochondria occurred (supplementary material Fig. S4). These data demonstrate that TPA formation is not related to the expression of a peroxisomal membrane protein carrying a fluorescent tag, or to the expression of Pex11p. Even the expression of N-terminally-tagged YFP-Pex11p, or Pex11p-YFP, a homologue of Pex11p with similar membrane topology, was unable to induce TPAs. Apparently, TPA formation depends on the localization of Pex11p (and not e.g. Pex11p) at the peroxisomal (and not the mitochondrial) membrane and requires C-terminal localization of the YFP tag.

C-terminal truncations of Pex11p inhibit peroxisome elongation
The above observations suggested that the YFP tag might block the C-terminal cytoplasmic tail of Pex11p, making it inaccessible and thus, inhibiting putative protein interactions, for example with components of the peroxisomal division machinery. Removal of different portions of the C-terminus could therefore lead to TPA formation as well. We generated C-terminally truncated versions of Pex11p with a Myc-tag at the N-terminus and examined their influence on peroxisomal targeting and TPA formation. As demonstrated, after expression of Myc-Pex11p in COS-7 cells, a prominent elongation of peroxisomes but no TPA formation was observed (Fig. 1B). Next, we expressed a Pex11p lacking the putative five-amino-acid cytoplasmic tail (Myc-Pex11p5). The truncated protein was efficiently targeted to peroxisomes as  S3). The Cterminal tail appeared to be dispensable for peroxisome elongation and division. Next we expressed a Pex11p construct lacking the putative transmembrane domain (Myc-Pex11p30; Fig. 2K,L). The truncated protein was still targeted to peroxisomes indicating that the targeting information within the N-terminal part is sufficient for peroxisomal localization. More interestingly, cells expressing Myc-Pex11p30 exhibited only spherical peroxisomes, and peroxisome elongation was drastically reduced (Fig. 2M). Similar observations were made with a Myc-Pex11p60 construct (not shown). These results indicate that the last C-terminal 30 amino acids of Pex11p are required for the elongation of peroxisomes. The very last five amino acids are dispensable for peroxisome elongation and division, but their loss results in a delay in peroxisome elongation. As the addition of a C-terminal YFP does not interfere with the ability of Pex11p to elongate peroxisomes, we assume that Pex11p-YFP acts as a 'dominant-negative' mutant,

2753
Pex11p-mediated peroxisome maturation which may inhibit other downstream events in the formation of spherical peroxisomes, such as the proper assembly of the constriction and division machineries.

TPAs represent a pre-peroxisomal membrane compartment composed of tubular membrane extensions and mature globular peroxisomes
To examine TPA morphology at the ultrastructural level, we performed electron microscopy of Pex11p-YFP transfected cells. As shown in Fig. 3, consistent with light microscopy, COS-7 cells expressing Pex11p-YFP contained accumulations of elongated membranes. In larger accumulations, the tubular membranes were observed to form ordered stack-like structures (Fig. 3B, open arrow). Interestingly, the membrane tubules appeared thinner than regular elongated peroxisomes (approx. 35-60 nm vs. 70-150 nm). Immunoelectron microscopy revealed that the membrane tubules could be decorated with antibodies directed to GFP or YFP, indicating that they contained Pex11p-YFP (Fig. 3D). More interestingly, bulbous or spherical membrane structures were observed at one end of the membrane tubules ( Fig. 3A,C, arrows). To achieve a specific labeling of peroxisomes, we used the alkaline DAB (diaminobenzidine) reaction for catalase (Angermuller and Fahimi, 1981). As this method does not work well with COS-7 cells , we used HepG2 cells, which also showed TPA formation after expression of Pex11p-YFP, and have been used for DAB cytochemistry before (Schrader et al., 1994). Interestingly, DAB staining was predominantly seen in the globular structures at the endings of the tubules (Fig. 3C). These findings confirm that the TPAs formed after expression of Pex11p-YFP are composed of ordered, stack-like tubular membranes, which appear to contain no or low amounts of catalase. These membrane tubules are in direct luminal and membrane continuity with globular or bulbous structures, which appear to be mature, catalasecontaining peroxisomes. At this point we hypothesize that these are peroxisomes caught during the growth and division process and propose that the TPAs represent a pre-peroxisomal membrane compartment, which is composed of tubular membrane extensions and mature (spherical) peroxisomes attached to them.

The distribution of matrix proteins in TPAs
To investigate the distribution of matrix proteins within TPAs, we co-transfected COS-7 cells with Pex11p-YFP and a construct coding for a DsRed fusion protein carrying a peroxisomal targeting signal 1 at the C-terminus (DsRed-PTS1). Like other fluorescent fusion proteins carrying a PTS1 for peroxisomal matrix protein import (for example, EGFP-PTS1, mRuby-PTS1), DsRed-PTS1 was properly and exclusively targeted to peroxisomes in control cells (unpublished data). When coexpressed with Pex11p-YFP, DsRed-PTS1 was almost exclusively found in the globular structures of the TPAs, and not in the tubular membrane extensions ( Fig. 4A-C). By contrast, Pex11p-YFP mainly localized to the tubular membrane extensions and not to the spherical peroxisomes attached to them. Deconvolution microscopy revealed structures resembling a peroxisomal 'apple tree', with the globular, DsRed-PTS1-positive structures representing the apples, and the tubular, stem-like Pex11p-YFP-positive membranes representing the tree.
2754 Journal of Cell Science 123 (16) Similar observations were made when COS-7 cells expressing Pex11p-YFP were stained with antibodies directed to peroxisomal catalase or acyl-CoA oxidase, a key matrix enzyme of peroxisomal -oxidation (unpublished data). These findings further confirm that the globular parts of the TPAs represent mature, importcompetent peroxisomes.

The distribution of PMPs in TPAs
Next, we addressed the distribution of different PMPs (peroxisomal membrane proteins) within TPAs. First, COS-7 cells expressing Pex11p-YFP were stained with antibodies directed to PMP70, a peroxisomal ABC transporter (Fig. 4D-F). PMP70 was found to localize predominantly to the spherical parts of the TPAs, and not to the tubular membrane extensions labeled by Pex11pYFP. To investigate the localization of newly synthesized PMPs to the TPAs, Similar observations were made for PMP22/Pxmp2-HA, which has recently been proposed to be a channel-forming peroxisomal membrane protein involved in metabolite transfer (Rokka et al., 2009). In addition, ACBD5.1-Myc, an acyl-CoA-binding protein with a potential transmembrane domain  was predominantly targeted to the globular structures (supplementary material Fig. S5). Coexpression of Myc-Pex11p resulted in a prominent labeling of the globular structures as well with a weaker staining of the tubular extensions (supplementary material Fig. S5).
Pex11p has been reported to promote peroxisomal division and multiplication (Kobayashi et al., 2007;Koch et al., 2003;Koch et al., 2005;Li and Gould, 2002;Schrader et al., 1998). Thus, we investigated whether expression of Pex11p-Myc would be sufficient to overcome the Pex11p-YFP-mediated block in peroxisomal division. Pex11p-Myc was targeted to the tubular membrane structures of the TPAs, which also contained Pex11p-YFP, and not to the globular domains ( Fig. 5A-C). Furthermore, its expression was not sufficient to release the block in peroxisomal division and did not result in a division of the TPAs or in the formation of spherical peroxisomes.
We next investigated the targeting of the peroxins Pex3p, Pex19p and Pex16p, which have been implicated in membrane biogenesis of peroxisomes and are supposed to act early in peroxisome biogenesis. Pex19p is suggested to function as a cycling receptor and/or chaperone for PMPs, and to be recruited to the peroxisome by the membrane receptor Pex3p (Fujiki et al., 2006). Coexpression of Pex11p-YFP and HA-Pex19p revealed targeting of HA-Pex19p to the whole membrane of the TPAs, including the spherical compartments and the tubular membrane extensions, where it colocalized with Pex11p-YFP (Fig. 5D-F). In addition, a cytoplasmic localization was observed, which relates to the role of Pex19p as a cycling receptor and/or chaperone. When the experiment was performed with Pex3p-Myc or Myc-Pex16p, similar results were obtained, with both proteins labeling the spherical and the tubular membrane structures (Fig. 5G-L). In contrast to Pex19p and Pex3p, Myc-Pex16p sometimes labeled the tubules less prominently. We also coexpressed a Myc-tagged version of Pex12p, a zinc RING finger protein thought to be involved in the recycling of the Pex5p receptor for matrix proteins, which was found on all membranes of the TPAs. Furthermore, Pex14p, an essential component of the docking and/or translocation machinery of matrix proteins, was detected on globular and (less prominently) on tubular membranes (Fig. 6C). Our findings clearly demonstrate that different PMPs are targeted to distinct regions within the TPAs and maintain specific membrane localization, although the globular and tubular membrane compartments form a continuum (Fig. 6). PMPs with a putative metabolic function (e.g. PMP70, PMP22/Pxmp2, ACBD5.1) are predominantly localized to the spherical domains of the TPAs, which in addition are importcompetent for peroxisomal matrix proteins. Pex11p, which promotes peroxisome elongation and division, is exclusively localized to the tubular membrane extensions. By contrast, peroxins involved in membrane biogenesis and PMP import such as Pex19p, Pex3p and Pex16p are found on both the spherical and the tubular membrane compartments. Of course, membrane mobility, topology and complex formation have to be considered as well. Our data further support the idea that the globular domains of the TPAs represent a mature, import-competent peroxisomal structure, whereas the tubular extensions represent a pre-peroxisomal membrane compartment that contains major PMPs required for membrane elaboration, but has not (yet) acquired import competence for matrix proteins.

Tubular membrane extensions are formed by pre-existing peroxisomes
The import-competent, globular domains of the TPAs can either be interpreted as new peroxisomes, which are about to form at the ends of the tubular membrane structures (analogous to ripening apples at the ends of the trees) or alternatively, they might be pre-existing peroxisomes that will give rise to tubular membrane extensions. To distinguish between these two alternatives we transfected COS-7 cells with Pex11p-YFP using a well-established electroporation protocol  that allows detection of the expressed protein soon after transfection. A time-course experiment was performed, and cells were fixed 1-6 hours after electroporation and processed for immunofluorescence using antibodies to PMP70 (Fig. 7A-G). Interestingly, at early time points Pex11p-YFP was targeted to pre-existing, spherical peroxisomes positive for PMP70. A complete colocalization of Pex11p-YFP and PMP70 staining was observed in the majority of the cells expressing Pex11p-YFP (Fig. 7A-D). Shortly after, small nose-like tubular extensions were observed to form at one side of the spherical peroxisomes (Fig. 7E,F). These extensions were positive for Pex11p-YFP, but did not show labeling for PMP70. After longer periods of time, the Pex11p-YFP-positive membrane extensions became  more prominent and elongated (Fig. 7G). These membrane extensions were highly dynamic (supplementary material Movie 1). More importantly, Pex11p-YFP appeared to concentrate in the growing tubular extensions. Later, the elongated peroxisomes started to form TPAs (Fig. 1D-H). Within these TPAs, Pex11p-YFP was predominantly localized in the tubular membrane extensions and no longer in the globular (PMP70-positive) peroxisomes attached to them. Peroxisomal constriction and division was completely abolished, and single, spherical peroxisomes were no longer observed. Expression of Pex11p-Myc has been shown to induce peroxisome elongation, and to promote constriction and division of the elongated peroxisomes into small spherical organelles . We repeated the time-course experiment with Pex11p-Myc (Fig. 7H-M). Similar to Pex11p-YFP, Pex11p-Myc first localized to spherical PMP70-positive peroxisomes. After about 60 minutes, small, nose-like Pex11p-Myc-positive (but PMP70-negative) extensions arose from the spherical peroxisomes. In contrast to Pex11p-YFP the tubular extensions were observed to segment and constrict showing a 'beads on a string'-like appearance (Fig.  7K-M). To investigate the import of matrix proteins into the preexisting and newly formed peroxisomal structures upon overexpression of Pex11p-Myc, we applied the HaloTag technology Los et al., 2008). COS-7 cells expressing HaloTag-catalase were first incubated with the cell-permeable HaloTag TMR ligand (red), then transfected with Pex11p-Myc, and immediately cultured in the presence of the cell-permeable HaloTag R110 Direct ligand (green; Fig. 8). Interestingly, the Pex11p-Myc-positive tubules were initially negative for catalase (Fig. 8A). Occasionally, newly imported catalase was seen at the tubule tips. The constricted tubules and beads contained only newly imported HaloTag-catalase, whereas the attached globular peroxisomes contained both pre-imported and newly imported HaloTag-catalase (Fig. 8B). As initially reported , Pex11p-Myc was absent from the bead-like structures, but was found inbetween, at the constriction sites ( Fig. 7K-M). Later (>12-24 hours), fission into small spherical peroxisomes was observed (Fig. 1). We would like to emphasize that the latter is inhibited by a lack of DLP1 (supplementary material Fig. S6) and results in the accumulation of elongated but constricted membranes . Interestingly, the 'beads' within these membranes are positive for matrix and membrane proteins (Fig.  7N,O; supplementary material Fig. S6)  indicating that loss of DLP1 function causes a block at a later stage than Pex11p-YFP. In contrast to Pex11p-YFP, the formation of TPAs was not observed after expression of Pex11p-Myc or loss of DLP1.
These findings demonstrate that Pex11p initially localizes to the membranes of pre-existing peroxisomes where it initiates the formation of a tubular membrane extension. This membrane extension is enriched in Pex11p, but is not yet import competent for matrix proteins (or PMP70). In contrast to Pex11p-Myc, expression of Pex11p-YFP appears to act as a 'dominant-negative' mutant, which abolishes division of the tubular membranes as well as further PMP insertion (or segregation) and matrix protein import into the forming, bead-like peroxisomes. Furthermore, the formation of TPAs occurs at later time points. We would like to emphasize that the TPAs were not labeled by ER marker proteins and represent a membrane compartment distinct from the ER (supplementary material Fig. S1).

TPA formation can also be induced by manipulation of Fis1
Fis1, a tail-anchored membrane protein of both peroxisomes and mitochondria, is thought to act as a membrane receptor for DLP1. Similar to Pex11p, expression of Fis1 has also been reported to promote peroxisome division and multiplication (Kobayashi et al., 2007;Koch et al., 2005). Furthermore, Fis1 has been suggested to interact with Pex11p or/and to be recruit to peroxisomes via Pex11 proteins in plants (Kobayashi et al., 2007;Lingard et al., 2008). After coexpression of Pex11p-YFP and Myc-Fis1, the latter was predominantly targeted to the tubular membrane extensions of the TPAs, but not to the globular organelle structures (Fig. 9A-F). Similar to Pex11p-Myc (Fig. 5D-F), Myc-Fis1 was not sufficient to overcome the block in peroxisomal division.
We recently reported that coexpression of Fis1 and Pex11p changed the uniform intracellular distribution of peroxisomes . Coexpression of a GFP-tagged Fis1 (GFP-Fis1) and Pex11p-Myc (but not expression of GFP-Fis1 or Pex11p-Myc alone) resulted in the formation of peroxisomal accumulations with a tubulo-reticular appearance and a block in peroxisomal division. Furthermore, GFP-Fis1 and Pex11p-Myc were observed to colocalize on these structures. We re-evaluated these recent findings and discovered that those accumulations were similar to TPAs formed by Pex11p-YFP (Fig. 9G-I). This was confirmed by ultrastructural studies of GFP-Fis1-and Pex11p-Myc-transfected cells (Fig. 9J). Labeling for PMP70 (or matrix proteins, not shown) revealed that the spherical organelle structures at the ends of the tubules were mature peroxisomes, whereas GFP-Fis1 and Pex11p-Myc colocalized on the tubular membrane extensions (Fig. 9G-I). These observations show that TPAs can also be generated independently of Pex11p-YFP. Our results further support the suggested interactive roles of Fis1 and Pex11p. We propose that specific manipulations of either Fis1 or Pex11p disturb the assembly of a functional constriction-fission complex, inhibiting peroxisomal division, and thus promoting TPA formation.

Pex11p-YFP -a novel tool to study peroxisomal growth and division
Pex11p plays important roles in peroxisome morphogenesis. Its expression promotes peroxisome elongation and division in mammalian cells (Li and Gould, 2002;Schrader et al., 1998), and its absence, as in PEX11 null mice results in reduced numbers of peroxisomes and severe developmental defects (Li et al., 2002b). Interestingly, the addition of a monomeric YFP to the C-terminus of Pex11p did not interfere with peroxisome elongation, but 2758 Journal of Cell Science 123 (16) resulted in the formation of tubular peroxisomal membrane compartments (named TPAs), which had lost their ability to divide. These alterations were specific for Pex11p-YFP. The only exception identified so far was GFP-Fis1 promoting TPA formation when coexpressed with Pex11p-Myc (see below). Deletions of the C-terminus of Pex11p either led to a delay (Myc-Pex11p5) or abrogated peroxisome elongation completely (Myc-Pex11p30), indicating that the morphological alterations observed were not merely the result of an inaccessibility of the Cterminus itself. We assume that the generated Pex11p-YFP fusion protein acts as a 'dominant-negative mutant' on the proper assembly of a functional constriction-division complex -presumably because of sterical hindrance. Interactive roles of Fis1 and Pex11 proteins have recently been described (Kobayashi et al., 2007;Lingard et al., 2008). The N-terminal region of Pex11p has been demonstrated to be required for homo-oligomerization of Pex11p and appeared indispensable for its peroxisome-multiplying activity, whereas the C-terminal region is supposed to mediate binding to Fis1 (Kobayashi et al., 2007). It is thus possible that the interaction of Myc-Pex11p5C with endogenous Pex11p explains the observed delay in peroxisome elongation. Remarkably, independent of Pex11p-YFP, coexpression of GFP-Fis1 and Pex11p-Myc also resulted in a block in peroxisome division and TPA formation. Although we initially failed to demonstrate an interaction of Fis1 and Pex11p , our findings are in line with a coordinated and interactive role of both proteins in peroxisome proliferation. This is also supported by the fact that both Pex11p and Fis1 colocalize predominantly at the tubular membrane extensions of the TPAs. It should be noted that in addition to Pex11p, Fis1 also has been suggested to form homo-oligomers (Serasinghe and Yoon, 2008). In plant cells homo-and heterooligomerizations among all five Arabidopsis PEX11 isoforms, and interactions with FIS1b have been described (Lingard et al., 2008). Moreover, the mitochondrial fission factor Mff is thought to play a, yet unknown, role in peroxisome division (Gandre-Babbe and van der Bliek, 2008). We propose that Pex11p-YFP initially elongates peroxisomal membranes, but inhibits peroxisomal segmentation and division, which subsequently results in the formation of stacked membranes. In addition to sterical hindrance during the formation of a functional constriction-division complex, the Pex11p-YFP fusion protein as a whole may have some intrinsic 'stacking' activity, and tight trans-membrane interactions of Pex11p-YFP may explain the block in peroxisome division. Stacking may also be due to interactions between cytoplasmic domains of other proteins on apposing tubular peroxisomal membranes, which might accumulate, based on the block in peroxisome division. These putative binding interactions might be of physiological significance, for example during peroxisome morphogenesis, organelle-peroxisome interactions or lipid transfer via the ER. Although the mechanistic reasons for the inhibitory function of Pex11p-YFP are not completely clear, our generated Pex11p-YFP fusion protein represents a specific and useful novel tool to further dissect peroxisome proliferation and to investigate early events in peroxisomal growth and division.

Pex11p-mediated peroxisome maturation
An important finding of this study, exploiting the generated Pex11p-YFP fusion protein, is the observation that Pex11pmediated growth (elongation) and division of pre-existing peroxisomes follows a multistep maturation pathway (Fig. 10) including the following steps: (1) Pex11p-mediated formation of a peroxisomal subdomain at one side of a pre-existing peroxisome; (2) growth (extension) of this domain and formation of a peroxisomal membrane compartment; this compartment contains some but not all PMPs (for example, the 'early peroxins' Pex3p, Pex19p, Pex16p, which are required for peroxisome membrane biogenesis and PMP import, but not PMP70 or PMP22) and is thus not yet fully import-competent for peroxisomal matrix proteins; (3) segmentation (constriction) of the membrane compartment (presumably mediated by Pex11p, Fis1 and other, yet unknown, components; DLP1 does not appear to be required for membrane constriction ; (4) assembly and/or activation of the import machinery and subsequent import of PMPs and matrix proteins; (5) final division into spherical peroxisomes mediated by DLP1. Similar maturation pathways, which are commonly initiated by the formation of an early peroxisomal membrane compartment and its stepwise conversion into a mature, metabolically active peroxisome compartment, have been proposed for peroxisomal growth and division in yeast (Veenhuis et al., 2000) and in some aspects also resemble ER-dependent peroxisome maturation (Titorenko and Rachubinski, 2009;van der Zand et al., 2006). In those models, maturation is achieved by selective and stepwise import of certain PMPs, membrane lipids and matrix proteins. Our observations clearly demonstrate that growth and division of mammalian peroxisomes (at least the process mediated by Pex11p overexpression) is more complex than simple division of a preexisting organelle and per se represents a process of biogenesis.
It is probable that membrane growth and extension also involve the transfer of lipids from the ER, their main source, to the 2759 Pex11p-mediated peroxisome maturation Fig. 10. Peroxisome proliferation by growth and division follows a multistep maturation pathway. Pex11p mediates the formation of a tubular membrane extension emerging from one side of a spherical, mature peroxisome. The membrane extension grows and acquires a distinct set of PMPs, but contains no, or only a few, matrix proteins. Subsequently, the membrane extension segments and constricts. This step can be inhibited by expression of Pex11p-YFP which keeps the peroxisomes in an elongated state and results in TPA formation. The constricted peroxisomal tubule acquires other PMPs and matrix proteins and is finally divided by hFis1 and DLP1 into several spherical peroxisomes. Loss of DLP1 inhibits fission resulting in the accumulation of constricted peroxisomal tubules positive for PMPs and matrix proteins  peroxisomal membrane. Although the mechanism(s) of phospholipid transfer from the ER (or other organelles) to peroxisomes is unknown, there are three potential mechanisms.
(1) ER-derived vesicular transport; however, in a recent report the vesicle-based ER-to-peroxisome transfer of phospholipids has been questioned (Raychaudhuri and Prinz, 2008). (2) Membrane-membrane interaction with ER subdomains (contact sites) through a protein-based, but yet unknown, phospholipid transfer; such a mechanism, which would resemble lipid transfer to mitochondria, is supported by the close association observed between peroxisomes and the smooth ER (Grabenbauer et al., 2000;Novikoff and Shin, 1964;Yamamoto and Fahimi, 1987;Zaar et al., 1987). (3) Direct luminal connection with ER subdomains; such a direct connection has been observed in mouse dendritic cells by three-dimensional image reconstruction Tabak et al., 2003;van der Zand et al., 2006). Although the TPAs are not accessible for proteins which localize to the ER (supplementary material Fig. S1), we cannot exclude the possibility that direct luminal connections between the TPAs and the ER may exist. However, we would assume that they are of a transient nature.
As the globular and extended tubular membrane domains show membrane continuity, specific mechanisms must exist that restrict the mobility of the PMPs and keep them within the globular or tubular membrane domains. This might be mediated by protein oligomerization, but might also involve a specific lipid environment. In this respect, a role for lipid microdomains in the maturation of peroxisomes in the yeast Yarrowia lipolytica has been proposed (Boukh-Viner et al., 2005;Titorenko and Rachubinski, 2009). The expression of Pex11p-YFP might thus be a valuable tool to specifically isolate tubular peroxisomal membranes and analyze their lipid composition.
How is the specific targeting of the PMPs to the globular and extended tubular membrane domains within the TPAs achieved? It is probable that Pex11p and Fis1 are directly targeted to the tubular membrane extensions. An interaction of Pex11 proteins and Fis1 with Pex19p, the peroxisomal import receptor for PMPs, has been demonstrated (Delille and Schrader, 2008;Jones et al., 2004), and no evidence for localization to the ER has been obtained (Lingard et al., 2008) (this study). Other PMPs are thought to be routed indirectly to peroxisomes through the ER (e.g. Pex3p) by an, as yet unknown, mechanism (Haan et al., 2006;Hoepfner et al., 2005;Kragt et al., 2005;Tam et al., 2005). It should be noted that our findings do not exclude an indirect route for PMPs to TPAs. If this transport involves the formation of pre-peroxisomal carriers from the ER, and if different carriers with a distinct PMP composition exist, they must have the ability to distinguish between the globular and tubular membrane domains prior to fusion. Alternatively, all carriers might fuse over the whole of the globular and tubular domains or at a specific site (e.g. at the neck region). This exciting scenario would require specific, yet undefined sorting and/or retention mechanisms within the peroxisomal membrane, and requires further investigation.
Surprisingly, the formation of new spherical peroxisomes is inhibited in COS-7 cells containing TPAs. We have demonstrated that this is due to a block in the division of pre-existing peroxisomes. However, recent studies in yeast and mammalian cells have shown that peroxisomes can multiply either by division or by de novo formation (Hoepfner et al., 2005;Kim et al., 2006;Motley and Hettema, 2007;Nagotu et al., 2008). Whereas in yeast peroxisomes only form de novo in the absence of preexisting peroxisomes, and multiply by division in wild-type cells (Motley et al., 2008;Nagotu et al., 2008), in mammalian cells de novo formation and multiplication by division has been proposed to occur simultaneously (Kim et al., 2006). As we did not observe the formation of new spherical peroxisomes in the cell types that we studied, we might face a situation similar to that in yeast, where only the complete loss of peroxisomes triggers de novo formation. Alternatively, Pex11p might be required for de novo formation as well suggesting an overlap in the components involved. It should also be noted that recent data supporting simultaneous peroxisome formation by both de novo synthesis and division are based on the assumption that daughter peroxisomes, formed by division of pre-existing peroxisomes, should all contain peroxisomal components from their mother peroxisomes (Kim et al., 2006). As demonstrated this is not the case for Pex11p-mediated peroxisomal growth and division. It is, however, possible that other, condition-specific characteristics may result in a more equal, symmetric division of peroxisomes (peroxisome replication). Evidence for non-symmetrical fission of peroxisomes, obtained using HaloTag technology, has been reported recently ). An uneven distribution of matrix proteins has also been observed under peroxisomeproliferating conditions in regenerating rat liver (Yamamoto and Fahimi, 1987) or as a general mechanism upon fission in the yeast Hansenula polymorpha (Nagotu et al., 2008).

Pex11p-a 'morphogenic' peroxin
Different models of how Pex11p might affect peroxisome number have been proposed (Lingard and Trelease, 2006;Orth et al., 2007;Thoms and Erdmann, 2005), but in many cases it is still unclear how this effect is exerted. Studies are also complicated by the existence of several Pex11 isoforms and Pex11-related proteins, which differ in their tissue specificity and might have multiple or more specialized, non-overlapping functions during peroxisome biogenesis. In the case of Pex11p our data strongly support that it is a 'morphogenic' protein that is responsible for the generation of a tubular peroxisomal subcompartment, which is a pre-requisite for subsequent peroxisome division and proliferation. We provide new evidence that Pex11p initiates the formation of a membrane extension at one side of a pre-existing peroxisome, promotes and/or induces membrane elongation, and, more importantly, also partitions into peroxisome membrane tubules. Pex11p also concentrates at constriction sites prior to membrane fission. Thus, it might mediate a narrowing of the tubular membranes to support exclusion of PMPs and matrix proteins from these regions and to prepare the ground for the assembly of the division machinery. Interestingly, the Pex11p-YFP-mediated membrane extensions have a diameter similar to peroxisomal constriction sites. Analogous to Pex11p, concentration of H. polymorpha Pex11p during fission at the base of a tubular peroxisome extension has been observed in a DNM1 deletion strain (Nagotu et al., 2008). Our data strongly support that mammalian Pex11p functions as a peroxisomal scaffold and/or structural protein that forms and stabilizes highly curved peroxisome membrane tubules. These properties are likely to depend on the abundance of Pex11p, and its oligomerization. In this respect, Pex11p resembles the reticulons and Dp1/Yop1p, which shape the tubular ER (Shibata et al., 2006). Future work to unravel the biological roles of Pex11 proteins promises to yield more fascinating insights into the mechanisms of peroxisome formation and regulation of peroxisome shape and dynamics. the G418-enriched cells were (i) incubated for 48 hours with the cell-permeable HaloTag TMR ligand (final concentration: 250 nM; Promega), (ii) extensively washed with PBS and incubated in standard growth medium for 24 hours, (iii) processed for electroporation (Neon Transfection System, Invitrogen) with the plasmid encoding Pex11p-Myc, and (iv) immediately cultured in the presence of the cell-permeable HaloTag R110 Direct ligand (final concentration: 10 nM; Promega). Cells were fixed after 24 hours at the onset of peroxisome division and processed for immunofluorescence microscopy. Fluorescence was evaluated on a CellM imaging station (Olympus) equipped with U-MNUA2, U-MNIBA3 and U-MWIY2 fluorescence mirror units.

Electron microscopy
For routine electron microscopy, cultured cells were fixed with Ito's fixative for 30 minutes at room temperature. After postfixation with 1% K 4 Fe(CN) 6 -reduced osmium tetroxide (1 hour at 4°C), samples were stained with 0.3% uranyl acetate. For cytochemical localization of catalase (Angermuller and Fahimi, 1981), HepG2 cells were incubated in alkaline DAB medium for 1 hour at 37°C followed by postfixation in 1% aqueous osmium tetroxide and 2% uranyl acetate (Schrader et al., 1994). The samples were dehydrated in a graded series of alcohol and embedded in Epon 812 (Polysciences Ltd, Eppenheim, Germany). For immunostaining, samples were fixed in 0.1 M cacodylate buffer, pH 7.35 containing 2% PFA, 0.1% glutaraldehyde (Serva, Heidelberg, Germany). The dehydrated samples were embedded in Lowicryl K4M (Polysciences Ltd) and polymerized at -20°C with UV light (360 nm) for 48-72 hours. Thin sections (70 nm) were incubated with polyclonal antibodies directed to G(Y)FP (1:200-500) and visualized using 10 nm protein A-gold solution (J. Slot, University of Utrecht, The Netherlands) at a dilution of 1:60 or 1:70, both in 0.5% BSA in PBS. Sections were stained with uranyl acetate and lead citrate, and analyzed using a Zeiss EM 109 electron microscope.

Quantification and statistical analysis of data
For quantitative evaluation of peroxisome morphology, 100-200 cells per coverslip were examined and categorized as cells with elongated, tubular (>2 m in length) or spherical peroxisomes (0.3-1 m; including rod-shaped peroxisomes) or TPAs. Usually, three to five coverslips per preparation were analyzed, and three to five independent experiments were performed. Data are presented as means ± s.d.