Peroxisomes control mitochondrial dynamics and the mitochondrion-dependent apoptosis pathway

Peroxisomes are organelles bound by a single membrane that play essential roles in metabolic functions such as oxidation of fatty acid chains, catabolism of reactive oxygen species (ROS), and synthesis of ether phospholipids in all eukaryotic cells. The peroxin (Pex) family of proteins is required for the assembly and function of peroxisomes (Waterham and Ebberink, 2012; Fujiki et al., 2012). Deficiency of Pex3, a peroxisomal membrane protein necessary for membrane assembly, thus results in the complete loss of peroxisomes (Muntau et al., 2000), whereas deficiency of Pex5, a peroxisomal transporter, results in the loss of peroxisomal matrix proteins (Otera et al., 1998). Peroxisomal dysfunction due to Pex gene mutations is detrimental to human development, as evidenced by human autosomal recessive genetic diseases (known as peroxisome biogenesis disorders), such as Zellweger syndrome, which results in death within the first year of life (Goldfischer et al., 1973). Peroxisome-deficient mice also die during the neonatal period (Maxwell et al., 2003; Baes et al., 1997). Both patients with peroxisome biogenesis disorders and peroxisome-deficient mice manifest a variety of characteristics including neurological dysfunction, hypotonia and craniofacial abnormalities (Trompier et al., 2014; Waterham and Ebberink, 2012; Muntau et al., 2000).

Peroxisomes collaborate with other organelles in various physiological and pathological contexts. In particular, peroxisomes engage in a functional interplay with mitochondria with regard to the degradation of fatty acids and ROS detoxification as well as to antiviral immunity (Lismont et al., 2015; Schrader and Yoon, 2007; Dixit et al., 2010). The interplay between peroxisomes and mitochondria is highlighted by the observation that the loss of Pex genes gives rise to abnormalities in mitochondrial structure and metabolic function. For instance, deletion of Pex5 in mouse hepatocytes affects the structure of mitochondrial inner and outer membranes as well as giving rise to abnormal (swollen) cristae (Goldfischer et al., 1973; Peeters et al., 2015; Baumgart et al., 2001), reduced activity of oxidative phosphorylation (OXPHOS) complexes, loss of the mitochondrial membrane potential and increased ROS levels (Peeters et al., 2015). Deletion of Pex5 has also been shown to increase the number of mitochondria as well as the level of glycolytic activity, possibly as a compensatory response to the impairment of OXPHOS (Peeters et al., 2015). Deletion of Pex13 in mouse brain or of Pex19 in fly larvae resulted in similar dysfunction of OXPHOS, elevated ROS levels, and an increased abundance of mitochondria (Rahim et al., 2016; Bülow et al., 2018). Furthermore, human patients harboring PEX16 mutations manifest myopathy accompanied by mitochondrial abnormalities (Salpietro et al., 2015). It remains unclear, however, which of these various phenotypes in mice, flies and humans reflect primary effects of peroxisome deficiency or are secondary to primary effects such as ROS accumulation.

In addition to their roles in OXPHOS and redox regulation, mitochondria are key players in the regulation of apoptosis. Various proteins that are normally localized to the intermembrane space of mitochondria, including cytochrome c, are released into the cytosol on apoptosis induction. In the cytosol, cytochrome c interacts with Apaf-1 and pro-caspase-9 to form a large protein complex known as the apoptosome. The resulting increase in the autocatalytic activity of pro-caspase-9 leads to the cleavage and activation of pro-caspase-3 and pro-caspase-7, and the active forms of these latter two enzymes then execute apoptosis by cleaving numerous substrates (Wang and Youle, 2009).

Mitochondria are highly dynamic organelles that continually change their morphology by fission and fusion processes, which contributes to mitochondrial quality control and the induction of apoptosis (Detmer and Chan, 2007; Suen et al., 2008). Mitochondrial fission and fusion are mediated by evolutionarily conserved members of the dynamin family of proteins. Fission is thus mediated by cytosolic dynamins, such as dynamin-related protein 1 (Drp1; also known as DNM1L) and Dyn2 (also known as DNM2), whereas fusion is mediated by the membrane-anchored dynamins Mfn1 and Mfn2, and Opa1 in mammals (Detmer and Chan, 2007; Lee et al., 2016). The fission process mediated by Drp1 appears to play a central role in the induction of cytochrome c release and subsequent apoptosis in various physiological and pathological contexts (Westermann, 2010).

Recent studies have shown that interactions with other organelles contribute to the regulation of mitochondrial dynamics. Sites of contact between mitochondria with endoplasmic reticulum (ER) (known as mitochondrion-associated membranes, MAMs) play an important role in the regulation of mitochondrial fission (Friedman et al., 2011) and have been implicated in that of apoptosis (Prudent et al., 2015; Yang et al., 2018; Hoppins and Nunnari, 2012). Lysosomes also participate in the regulation of mitochondrial dynamics (Wong et al., 2018). Mitochondria and peroxisomes share key regulators for their fission including Drp1, and as well as its receptors Fis1 and Mff (Kobayashi et al., 2007; Camões et al., 2009; Delille et al., 2009; Schrader, 2006). However, whether peroxisomes also regulate mitochondrial dynamics and mitochondrion-mediated apoptosis has remained unclear.

Here, we investigated the role of peroxisomes in the regulation of mitochondrial dynamics, caspase activation and apoptosis by deleting Pex3 or Pex5 in mouse embryonic fibroblasts (MEFs) under conditions in which the cytosolic ROS levels do not increase substantially. We found that deletion of either Pex3 or Pex5 resulted in fragmentation of mitochondria, the appearance of cytochrome c in the cytosol, and an increase in the amounts of cleaved caspase-9 and caspase-3. Importantly, restoration of Pex3 or Pex5 expression in the corresponding knockout (KO) MEFs attenuated these effects. Furthermore, we found that ablation of Pex3 greatly enhanced the induction of apoptosis by the DNA-damaging agent etoposide. Our results thus suggest that peroxisomes regulate mitochondrial dynamics, caspase activity and cell death so as to reduce cellular sensitivity to damaging insults.

To examine an acute effect of peroxisome deficiency, we took advantage of MEFs derived from Pex3^fl/fl;Rosa-Cre-ER^T2 mice, which are homozygous for a floxed allele of Pex3 and harbor a tamoxifen-inducible transgene for Cre recombinase (Fig. S1A). We immortalized these cells by introducing SV40 large T antigen and deleted Pex3 by adding 4-hydroxytamoxifen. Immunoblot analysis detected Pex3 protein in control MEFs (not exposed to 4-hydroxytamoxifen) but not in Pex3 KO MEFs (Fig. S1B). Immunofluorescence analysis also detected almost no punctate signals for Pex14 or for EGFP tagged with peroxisome-targeting signal 1 (PTS1), the signaling peptide present in peroxisome matrix proteins, in the Pex3 KO MEFs (Fig. S1C), indicating the successful depletion of peroxisomes in these cells.

With the use of our Pex3 KO MEFs, we then set out to identify mitochondrial phenotypes of peroxisome deficiency that could be rescued by reintroduction of peroxisomes. We examined mitochondrial morphology by observing the intracellular distribution of Tom20, a mitochondrial outer membrane protein, and ATP synthase β, a mitochondrial inner membrane protein, and found that the extent of mitochondrial fragmentation was increased in Pex3 KO MEFs (Fig. 1A,B). Huygens-based quantification indicated that the size of mitochondria was significantly smaller in Pex3 KO MEFs than that in control MEFs (Fig. 1C,D). The length of mitochondria also became shorter in Pex3 KO MEFs compared to that in control MEFs (Fig. S2B–E). We then asked whether restoration of Pex3 expression in these Pex3 KO cells would rescue this mitochondrial phenotype. Infection of Pex3 KO MEFs with a retrovirus encoding Pex3 increased the abundance of Pex3 protein and induced the formation of peroxisomes (Fig. S1D,E). Importantly, this reintroduction of Pex3 resulted in elongation of mitochondria in the Pex3-deficient MEFs (Fig. 1E–H), indicating that Pex3 suppresses mitochondrial fragmentation in a reversible manner.

Given that the effect of Pex3 deletion on mitochondrial morphology might have been the result of a Pex3-specific function unrelated to peroxisome formation, we examined whether deletion of a different Pex gene, Pex5, confered a similar phenotype. Deletion of Pex5 would be expected to result in loss of peroxisomal matrix proteins but retention of the peroxisomal membrane, whereas that of Pex3 results in the complete loss of peroxisomes. To disrupt Pex5 in MEFs, we adopted the CRISPR-Cas9 system with a guide (g)RNA targeted to Pex5 (Fig. S3A). We confirmed disruption of Pex5 gene (Fig. S3B) as well as the loss of Pex5 protein (Fig. 2A) in the targeted cells. Examination of mitochondrial morphology revealed that the extent of mitochondrial fragmentation was increased in the Pex5 KO MEFs compared with control (wild-type, WT) MEFs (Fig. 2B,C). Furthermore, this phenotype of Pex5 deficiency was rescued by retrovirus-mediated restoration of Pex5 expression (Fig. 2D–F). These results thus support the notion that peroxisomal functions are involved in suppressing mitochondrial fragmentation.

We next examined whether an increase (rather than a decrease) in the number of peroxisomes might also affect mitochondrial morphology. We thus exposed MEFs to 4-phenylbutyrate (4-PBA), an inducer of peroxisome proliferation. We confirmed that 4-PBA increased the abundance of peroxisomes, as detected by immunostaining of Pex14, in control MEFs (Fig. 3A; Fig. S4A,B). Furthermore, we found that 4-PBA induced mitochondrial elongation in these cells (Fig. 3A,C,D; Fig. S4C,D). We also quantified the total mitochondrial number and volume, and found that 4-PBA treatment reduced the number of mitochondria without inducing significant changes in their total volume (Fig. S4G,H). These results further support the idea that 4-PBA treatment accelerated the mitochondrial fusion process or suppressed the fission process. Importantly, however, 4-PBA did not induce mitochondrial elongation or suppress mitochondrial fragmentation in Pex3 KO MEFs (Fig. 3B,C,E; Fig. S4E,F), suggesting that the induction of mitochondrial elongation by 4-PBA requires peroxisomes. Taken together, these results indicate that peroxisome abundance correlates well with mitochondrial volume and length.

Given the mitochondrial fragmentation apparent in peroxisome-deficient cells, we next examined mitochondrial structure in more detail by electron microscopy (EM). Mitochondria were indeed smaller and shorter in Pex3 KO MEFs compared with control MEFs (Fig. 4A–C), consistent with the results of confocal fluorescence microscopy (Fig. 1A,B). The structure of cristae also appeared to have collapsed, with the presence of an indistinct and irregular inner membrane, in Pex3 KO MEFs (Fig. 4D), similar to the morphology previously observed in Pex5-deficient hepatocytes (Peeters et al., 2015; Baumgart et al., 2001).

Given that mitochondrial fragmentation and collapsed cristae are associated with the release of cytochrome c from these organelles and the intrinsic (mitochondrion-dependent) pathway of apoptosis (Suen et al., 2008; Otera et al., 2016), we examined whether Pex3 deletion affected the distribution of cytochrome c. Whereas cytochrome c immunoreactivity was detected almost exclusively in mitochondria in control MEFs, as shown by its overlap with that of Tom20, cytochrome c signals were diffusely distributed in the cytosol in addition to their punctate mitochondrial distribution in Pex3 KO MEFs (Fig. 5A,B). A high cell density appeared to further increase the amount of cytochrome c in the cytosol of Pex3 KO cells (Fig. S5A). Quantitative analysis revealed that the colocalization of cytochrome c with Tom20, as reflected by Pearson's correlation coefficient (r), was significantly reduced in Pex3 KO MEFs compared with control MEFs (Fig. 5C), indicating that Pex3 deletion indeed induced the diffusion of cytochrome c. Furthermore, subcellular fractionation analysis showed an increased ratio of cytosolic to mitochondrial cytochrome c in Pex3 KO MEFs as compared with control MEFs (Fig. 5D,E). This result further confirms the diffusion of cytochrome c to the cytoplasm in Pex3 KO MEFs. Importantly, forced expression of Pex3 was sufficient to restore the normal (mitochondrial) distribution of cytochrome c in Pex3 KO MEFs (Fig. 5F–H). Together, these results thus show that Pex3 suppresses the diffusion of cytochrome c.

We then examined whether deletion of Pex5 results in a similar redistribution of cytochrome c. Cytochrome c signals were indeed found to be diffusely distributed in the cytosol of Pex5 KO MEFs (Fig. 6A,B). We also confirmed that forced expression of Pex5 restored the mitochondrial localization of cytochrome c in the Pex5 KO cells (Fig. 6C,D). Together, our results thus indicate that functional peroxisomes, which require both Pex3 and Pex5, are necessary for suppression of cytochrome c diffusion.

We next addressed the mechanism by which the diffusion of cytochrome c is increased in peroxisome-deficient MEFs. Although previous studies have shown that long-term deletion of Pex genes results in ROS accumulation (Bülow et al., 2018; Rahim et al., 2016), the cytosolic ROS level of our Pex3 KO MEFs as measured with a CellROX assay kit did not appear to differ from that of control MEFs (Fig. S6A,B). Under the same conditions, the treatment of these MEFs with tert-butylhydroperoxide (TBHP), a ROS inducer, increased CellROX signals (Fig. S6A,B). Moreover, control and Pex3 KO MEFs did not show a detectable difference in mitochondrial ROS levels as monitored by using MitoSOX (Fig. S6C,D). These results suggest that (relatively acute) Pex3 deletion did not overtly increase ROS levels in mitochondria or the cytosol of our cultured MEFs. We also examined the rate of oxygen consumption in these cells, given that the abundance and activity of OXPHOS components are reduced after the deletion of Pex genes (Peeters et al., 2015). We again found, however, that control and Pex3 KO MEFs did not differ significantly in their basal or ATP-linked rates of oxygen consumption (Fig. S6E,F), suggesting that the mitochondrial OXPHOS system remains intact after Pex3 deletion in MEFs. Considering that we did not observe overt ROS accumulation or altered oxygen consumption in these cells, it was unlikely that an increase in ROS levels was responsible for the induction of cytochrome c diffusion. Indeed, treatment of Pex3 KO MEFs with the ROS scavenger N-acetylcysteine (NAC) did not suppress mitochondrial fragmentation or cytochrome c diffusion, whereas such treatment did attenuate the TBHP-induced increase in CellROX signal intensity (Fig. S6G–J).

Given that Drp1 plays a pivotal role in mitochondrial fragmentation (fission) and cytochrome c release (Estaquier and Arnoult, 2007; Otera et al., 2016), and that Drp1 localizes not only to mitochondria but also to peroxisomes (Tanaka et al., 2006; Waterham et al., 2007), we examined whether Pex3 deletion affects the abundance or subcellular localization of Drp1. Immunofluorescence analysis showed that the amounts of Drp1 both in the cytosol and associated with mitochondria appeared to increase in Pex3 KO MEFs compared with control MEFs (Fig. 7A). A western blot analysis also showed that the total amount of Drp1 proteins was increased upon Pex3 KO (Fig. S7A,B). Importantly, the extent of colocalization of Drp1, as well as that of its receptor Mff, with mitochondria was significantly higher in Pex3 KO MEFs than in control cells both in immunofluorescence and subcellular fractionation analyses (Fig. 7B–D; Fig. S7E–I), indicating that Pex3 ablation results in an increased localization of Drp1 to mitochondria, possibly through increasing the amount of Mff at mitochondria.

To examine whether Drp1 is responsible for the mitochondrial fragmentation and cytochrome c diffusion observed in Pex3 KO MEFs, we suppressed the function of Drp1 by introducing a catalytically inactive mutant (K38A) of the protein that has been shown to act in a dominant-negative manner (Frank et al., 2001). Expression of Drp1(K38A) indeed both restored the elongated morphology of mitochondria and attenuated cytochrome c diffusion in Pex3 KO MEFs (Fig. 7E–G). In addition, siRNA-mediated knockdown of Drp1 rescued the morphological alterations of mitochondria in Pex3 KO MEFs (Fig. S7E,F). Together, these results thus suggested that Pex3 deletion induces mitochondrial fragmentation and cytochrome c diffusion by promoting the localization of Drp1 to mitochondria.

The release of cytochrome c from mitochondria triggers activation of the Apaf-1–caspase-9 complex (apoptosome) and caspase-3 and thereby induces apoptosis (Hyman and Yuan, 2012). We therefore examined the effect of Pex3 deletion on caspase activity and found that the levels of the cleaved forms of caspase-9 and caspase-3 were increased in Pex3 KO MEFs compared with control MEFs (Fig. 8A,B). These results thus suggested that Pex3 suppresses the activation of caspase-9 and caspase-3 under basal conditions in MEFs. In contrast, we did not detect any significant difference in the fraction of annexin V–positive (apoptotic) cells between control and Pex3 KO MEFs (Fig. 8C,D), suggesting that caspase activation induced by Pex3 deletion is not sufficient to trigger apoptosis.

We hypothesized that the elevated basal activity of caspase-9 and caspase-3 in Pex3 KO MEFs might increase the vulnerability of these cells to cellular stressors that can trigger mitochondria-dependent apoptosis. To test this, we treated peroxisome-deficient cells with the DNA-damaging agent etoposide, a well-established inducer of mitochondria-dependent apoptosis (Yang et al., 1997; Wei et al., 2001; Karpinich et al., 2002). Both Pex3 KO MEFs and Pex5 KO MEFs manifested increased levels of caspase-3 activation in response to etoposide treatment compared with the corresponding control MEFs (Fig. 8E,F). Furthermore, annexin V staining revealed that Pex3 KO MEFs underwent apoptosis to a significantly greater extent than did control MEFs in response to etoposide treatment (Fig. 8G,H). Re-expression of Pex3 in Pex3 KO MEFs significantly suppressed the cellular apoptosis induced in response to etoposide (Fig. 8I), demonstrating that the phenotype was ascribable to Pex3 deletion. These results thus suggest that peroxisomes prevent excessive caspase activity and the induction of apoptosis, and that they thereby increase the resistance of cells to cellular stress such as that associated with DNA damage.

Although the cooperation between peroxisomes and mitochondria with regard to cellular metabolism has been extensively studied, the possible contribution of peroxisomes to mitochondrial fission–fusion dynamics has remained largely unknown. Our results now show that peroxisomes play a key role in determination of the balance between mitochondrial fission and fusion, with this balance being essential for a wide range of biological processes including cellular responsiveness to stressors (Khacho et al., 2016; Weir et al., 2017; Detmer and Chan, 2007). Indeed, our data also show that peroxisomes are important for protection of cells from mitochondrion-dependent apoptosis in response to DNA damage. Our study therefore provides a new basis for understanding the function of peroxisomes.

We found that the loss of peroxisomes induces the fragmentation of mitochondria in MEFs (Figs 1A–D, 2B,C, 4A–C; Fig. S2), whereas some studies have shown that the loss of peroxisomes in mouse serotogenic neurons and Drosophila Malpighian tubules instead resulted in the enlargement of mitochondria (Rahim et al., 2016; Bülow et al., 2018). Furthermore, a recent study has reported that human patient-derived fibroblasts lacking Pex3 did not exhibit changes in mitochondrial morphology (Sugiura et al., 2017). In addition to at the level of cells, tissues and species, these differences might be attributable to the long-term ablation of peroxins in these studies, which likely resulted in secondary effects due to the accumulation of ROS and subsequent cellular damages. In this study, we took advantage of a MEF culture in which the intracellular ROS level was not significantly increased after peroxin ablation and found that the mitochondrial fragmentation and cytochrome c diffusion induced by peroxin gene deletion were rescued by restoration of peroxin expression, indicating that these phenomena are the primary effects of peroxisomal loss (Figs 1E–H, 2E,F, 5F–H, 6C,D).

The mechanism responsible for peroxisome-mediated regulation of mitochondrial fission–fusion dynamics remains unknown. Given that the fission machineries of both organelles share components such as Drp1, Fis1 and Mff in mammalian cells (Waterham et al., 2007; Tanaka et al., 2006; Schrader et al., 2012), peroxisomes and mitochondria may compete for these components. Indeed, we found that Pex3 deletion increased localization of Drp1 and its receptor Mff to mitochondria (Fig. 7A–E, Fig. S7E–I), suggesting that the absence of peroxisomes promotes the recruitment of Drp1 to mitochondria by increasing the amount of its receptor at mitochondria. Furthermore, the inhibition of Drp1 rescued mitochondrial fragmentation in Pex3 KO MEFs (Fig. 7E–G; Fig. S7I,J). These results thus imply that Drp1 function is involved in the peroxisome-mediated regulation of mitochondrial dynamics. Several studies have also shown that peroxisomes are located adjacent to MAMs, which are described as mitochondrial constriction sites (Cohen et al., 2014; Mattiazzi Ušaj et al., 2015; Friedman et al., 2011; Horner et al., 2011). Given that peroxisomes also make physical contact with mitochondria (Fransen et al., 2017), these observations raise the possibility that peroxisomes compete with the ER for mitochondrial contact sites. Whether peroxisomes actually regulate MAM formation warrants future study. Furthermore, a recent study has shown that lysosomes also make contacts with mitochondria and regulate mitochondrial fission (Wong et al., 2018). Indeed, >80% of mitochondrial fission sites were found to contact lysosomes, whereas <20% of such sites contacted peroxisomes. Peroxisome–mitochondrion contacts may thus hamper or promote the interaction between lysosomes and mitochondria, resulting in modulation of the mitochondrial fission process. Together, these previous and present observations reveal multiple types of interorganellar communication that coordinately regulate mitochondrial fission–fusion dynamics.

Intriguingly, we noticed that the deletion of Pex5 affected mitochondrial morphology less dramatically than the deletion of Pex3. Furthermore, 4-PBA treatment slightly induced mitochondrial elongation in Pex5 KO MEFs as well as causing a slight increase in the number of peroxisomes (Fig. S8A,B). These results suggest that peroxisomal membranes or membrane proteins, which are retained in Pex5 KO MEFs but not in Pex3 KO MEFs, play a role in regulating mitochondrial morphology in addition to peroxisomal matrix proteins.

Given the findings in this study, we propose that Drp1 mediates mitochondrial fragmentation and subsequent cytochrome c diffusion in peroxisome-deficient cells. However, it remains unknown what molecular mechanism underlies cytochrome c diffusion after Drp1-mediated mitochondrial fragmentation in peroxisome-deficient cells. One possibility is that peroxisomes compete with mitochondria for some components necessary for cytochrome c diffusion. Cytochrome c is known to be released through the pore composed of the Bcl-2 family members Bax and Bak1 at the outer membrane of mitochondria (Tait and Green, 2010). In line with this, we checked the mitochondrial localization of Bax in Pex3 KO MEFs and found that the amounts of mitochondrial and cytosolic Bax in Pex3 KO MEFs were comparable to those in control MEFs (Fig. S5B,C). This result indicates that mitochondrial localization of Bax is not substantially affected by Pex3 deletion. Intriguingly, Fujiki and colleagues have reported that a fraction of Bak1 also localizes to peroxisomes (Hosoi et al., 2017). Elimination of peroxisomes may thus alter localization of Bak1 from peroxisomes to mitochondria, and the increased mitochondrial Bak1 may thereby facilitate cytochrome c diffusion in peroxisome-deficient cells. It would be important to test the notion that molecules shared by mitochondria and peroxisomes mediate their interorganellar communications.

Our results revealed not only the fragmentation of mitochondria in response to the loss of peroxisomes, but also the elongation of mitochondria in response to treatment of cells with an inducer of peroxisome proliferation, 4-PBA (Fig. 3; Fig. S4). These findings suggest that peroxisomal abundance is an important determinant of mitochondrial dynamics. Cellular conditions that affect the abundance of peroxisomes might thus also influence mitochondrial dynamics through peroxisomes. In this regard, cellular stressors such as UV light exposure and elevated ROS levels increase the number of peroxisomes in both plant and mammalian cells (Schrader and Fahimi, 2006). This increase in peroxisomal number or abundance may thus contribute to a protective response to allow cells to cope with stress via suppression of mitochondrial fragmentation and caspase activation. Such a notion is consistent with our present results showing that peroxisomes reduce cellular sensitivity to toxic insults.

Fatty acids, such as oleic acid, and a high-fat diet are also thought to increase the abundance of peroxisomes (Reddy and Mannaerts, 1994; Ishii et al., 1980; Diano et al., 2011; Lock et al., 1989; Veenhuis et al., 1987). It would thus be of interest to determine whether the high level of fatty acid synthesis apparent in adult neural stem-progenitor cells (Knobloch et al., 2013) confers resistance to cellular stress through an increase in the number of peroxisomes. Indeed, the abundance of peroxisomes is known to be high in radial glia cells preserved for a long period and to be reduced by aging (Ahlemeyer et al., 2007), with such changes possibly having consequences for mitochondrial regulation in these cells.

Mitochondrion-dependent activation of caspases contributes not only to removal of unnecessary cells during development or damaged cells exposed to stress stimuli but also to regulation of tissue stem cell differentiation and terminal differentiation of myoblasts, erythroblasts and keratinocytes (Hollville and Deshmukh, 2017). Furthermore, non-apoptotic caspase activation plays a key regulatory role in the pruning of neurites and the formation and maturation of neural circuits in the nervous system (Unsain and Barker, 2015). For example, caspase-9 is necessary for axon pruning in dorsal root ganglion neurons and cervical sympathetic neurons (Simon et al., 2012; Cusack et al., 2013). Non-apoptotic caspase activation is also implicated in regulation of the internalization of AMPA-sensitive glutamate receptors, which contributes to long-term depression in hippocampal neurons (Li et al., 2010). Non-apoptotic activation of caspases is thus essential for the control of various cellular processes. The activation of caspases at a sublethal level in peroxisome-deficient cells observed in the present study suggests that peroxisomes limit caspase activation under low-stress conditions (Fig. 8). It will be of interest to examine the possible role of peroxisomes in various biological processes that require non-apoptotic caspase activation.

Individuals with Zellweger syndrome and peroxin-deficient mice manifest severe defects in various organs including the brain, bone, muscle, kidney and liver. The mechanisms underlying this broad range of abnormalities, however, remain unknown. Dysfunction of the mitochondrial fusion machinery also gives rise to neurodegenerative diseases, muscle atrophy and osteogenic abnormalities (Detmer and Chan, 2007; Chen et al., 2010; Romanello et al., 2010; Touvier et al., 2015). Degeneration of Purkinje cells, one of the most prominent features of patients with Zellweger syndrome (Barry and O'Keeffe, 2013; Trompier et al., 2014), is thus also observed in mice with Purkinje cell-specific deficiency of Mfn2 (Chen et al., 2007). The peroxisome-dependent regulation of mitochondria uncovered in the present study therefore raises the possibility that excessive mitochondrial fragmentation plays a causal role in the pathogenesis of Zellweger syndrome. If so, our findings may provide a basis for the development of new therapies for this lethal disease.

Immunoblot analysis was performed as described previously (Okazaki et al., 2013). Immune complexes were detected with a chemiluminescence reagent (100 mM Tris-HCl pH 8.5, 1.25 mM luminol, 0.2 mM coumaric acid, 0.009% H₂O₂) and an Image Quant LAS4000 instrument (GE Healthcare). Blot intensities were measured with Image J software. In Figs 5D, 7C and Fig. S5B, subcellular fractionation analysis was performed by using a mitochondria/cytosol fractionation kit (K256-25; Biovision, Milpitas, CA,) according to the manufacturer's instructions. See ‘Antibodies’ section below for details of antibodies used.

Cells were fixed with 4% formaldehyde for 10 min at 37°C, permeabilized with 0.2% Triton X-100 in PBS for 5 min, and incubated for 30 min in PBS containing 2% fetal bovine serum (FBS) and 2% bovine serum albumin (BSA) (blocking buffer). They were then exposed first for 24 h at 4°C to primary antibodies in blocking buffer and then for 1 h at room temperature to Alexa Fluor-conjugated secondary antibodies (Thermo Fisher Scientific) and Hoechst 33342 in blocking buffer. Moviol were used as mounting medium. Images were acquired with a TCS SP5 confocal microscope (Leica) and were processed with Photoshop CS software (Adobe). For cytochrome c staining, only secondary antibodies were utilized for immunofluorescence in order to check the specificity of these signals. The Pearson's correlation coefficient (r) for the colocalization of Tom20 and cytochrome c as well as the Manders' M1 coefficient for the colocalization of Tom20 and Drp1 were calculated with Coloc 2 of Fiji. See ‘Antibodies’ section below for details of antibodies used.

In Fig. 1A,B,E,F, Fig. 2B,C,E,F, Fig. 3A–E, Fig. 7E,F, Fig. S6H,I, mitochondria were labeled by antibodies against Tom20 or ATP synthase β. Mitochondrial morphology of individual cells was evaluated in a double-blinded analysis and classified into ‘fragmented mitochondria’, ‘intermediate mitochondria’ and ‘elongated mitochondria’ (see Fig. S2A).

In Fig. 1C,D,G,H and Fig. 4A–H, samples were prepared in the same way as in the immunofluorescence experiments, except that ProLong Diamond (Thermo Fisher Scientific) were used as mounting medium. We took the 3D images in order to match the Nyquist condition (pixel size of x and y, 50 nm; z, 100 nm). These images were deconvoluted in Huygens software (Scientific Volume Imaging). After the deconvolution process, individual mitochondria were defined with the Huygens object analyzer. The values of seed and threshold were described in each figure legends. After the definition of mitochondria, their voxels and length of were calculated with object analyzer. Cellular average of mitochondrial volume and/or length were calculated in each condition.

Cells were fixed in 0.1 M phosphate buffer (pH 7.2) containing 2% glutaraldehyde and 2% paraformaldehyde, exposed to 1% OsO₄, dehydrated, and embedded in Epon 812. Ultrathin sections (60 nm) were cut with an ultramicrotome (UC6, Leica Microsystems), stained with uranyl acetate and lead citrate, and examined with a Hitachi HT7700 electron microscope. The area and major axis of mitochondria in images were measured with the use of Fiji software.

Cells were incubated with 5 μM MitoSOX or 500 nM CellROX for 30 min at 37°C, isolated by exposure to trypsin, and resuspended in PBS containing 3% FBS for analysis with a FACSAria flow cytometer (BD Biosciences).

The oxygen consumption rate (OCR) of cells was measured with the use of a Seahorse XF24 Extracellular Flux Analyzer (Seahorse Biosciences). Cells were plated in 24-well Seahorse plates and cultured overnight, after which the medium was replaced with Seahorse XF Base medium supplemented with 10 mM glucose, 1 mM pyruvate and GlutaMAX (2 ml/liter, Thermo Fisher Scientific). The cells were placed in a 37°C incubator without CO₂ before loading into the analyzer. After measurement of basal respiration, the cells were exposed to 1 μM oligomycin to measure the proton leak, to 1 μM carbonylcyanide m-chlorophenylhydrazone (CCCP) to measure the maximal OCR, and to 0.5 μM rotenone and 0.5 μM antimycin A to measure the non-mitochondrial OCR. The ATP-linked OCR was calculated by subtracting the proton leak from basal OCR. Cells plated simultaneously in 96-well plates were counted to normalize OCR values.

Cells were stained with Cy5-coupled annexin V (Promokine) according to the manufacturer's instructions. Flow cytometric analysis of the stained cells was performed with a FACSAria flow cytometer (BD Biosciences).

MEFs and Plat-E cells (Morita et al., 2000) were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS and 1% penicillin-streptomycin. Plat-E cells were transfected with the use of the GeneJuice Transfection Reagent (Merck Millipore), whereas transfection of MEFs was performed with Lipofectamine 2000 or with Lipofectamine and PLUS Reagents (Thermo Fisher Scientific).

C57BL/6 mice harboring the Pex3^{tm3a(EUCOMM)Wtsi} allele obtained from the EUCOMM (European Conditional Mouse Mutagenesis Program) consortium were crossed with Act-FLP transgenic mice (Kono et al., 2017) to remove the FRT-flanked region and subsequently with Rosa-CreER^T2 transgenic mice (obtained from the U.S. National Cancer Institute) (Fig. S1). Mice heterozygous for the floxed allele of Pex3 were mated, and the resulting homozygous embryos were isolated for preparation of MEFs. All animal experiments were performed according to approved guidelines. The MEFs were immortalized by the introduction of SV40 large T antigen as described previously (Ando et al., 2000), and they were then treated with 1 nM 4-hydroxytamoxifen to remove the loxP-flanked region. Immortalized MEFs treated with ethanol vehicle instead of 4-hydroxytamoxifen were prepared as control cells.

3T3 MEFs (kindly provided by Hidenori Ichijo, Graduate School of Pharmaceutical Sciences, The University of Tokyo, Japan) were transfected with the KO vector (see ‘Plasmids’ section below) and were then sorted with a FACSAria flow cytometer (BD Biosciences) to obtain GFP-positive cells, which were seeded as single cells in a 96-well plate.

For confirmation of Pex5 deletion in MEFs, the cells were collected and lysed with genotyping buffer [50 mM KCl, 10 mM Tris-HCl pH 8.3, 1.5 mM MgCl₂, 0.1 mg/ml gelatin, 0.45% NP-40, 0.45% Tween 20, proteinase K (500 μg/ml, Kanto Chemical)] or lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl pH 8.1) and were then incubated consecutively at 55°C for 3 h and 98°C for 10 min. The Pex5 locus was amplified by PCR with the use of KOD FX Neo (Toyobo) and the forward and reverse primers 5′-TCCCTTCCCCCAGCCCACTCCGGGTGCCTC-3′ and 5′-TCGGCGATGAATTCTTGGGACCAGTCGGTCTCATT-3′. The PCR products were ligated into PCR-Blunt (Thermo Fisher Scientific) for sequencing by Eurofin Genomics.

Plat-E cells were transfected with pMXs-IG, or either pMXs-Pex3-IG vector encoding human Pex3 or pMXs-Pex5S-IG vectors encoding Chinese hamster Pex5S, or pMXs-Drp1(K38A)-IG vector encoding rat Drp1(K38A) (Morita et al., 2000). After 3 days, the culture supernatants were harvested for isolation of retroviruses. Pex3 KO or Pex5 KO MEFs were infected with the corresponding peroxin retrovirus or the control virus, after which the cells were sorted with a FACSAria flow cytometer (BD Biosciences) to obtain GFP-positive cells. For preparing Drp1(K38A)-infected cells, Pex3 KO MEFs were infected with the Drp1(K38A) retrovirus or the control virus and infected cells were sorted in the same way as above.

The plasmid pUcD2Hyg/EGFP-PTS1 was described previously (Tamura et al., 1998). Full-length cDNAs for human Pex3 or Chinese hamster Pex5S (His-ClPex5S-HA) (Ghaedi et al., 2000; Matsumura et al., 2000) were subcloned into the BamHI and XhoI sites of the pMXs-IG vector (kindly provided by Toshio Kitamura, Division of Cellular Therapy/Division of Stem Cell Signaling, The Institute of Medical Sciences, The University of Tokyo, Japan). The p3xFLAG-ratDrp1K38A plasmid encoding rat Drp1(K38A) was kindly provided by Naotada Ishihara, Department of Biological Sciences, Graduate School of Science, Osaka University, Japan. Full-length cDNAs for rat Drp1(K38A) were subcloned into the EcoRI and XhoI sites of the pMXs-IG vector. For generation of the CRISPR vector for Pex5 deletion, a pair of oligonucleotides encoding the gRNA (forward, 5′-CACCGCTGGTCACCATGGCAATGC-3′; reverse, 5′-AAACGCATTGCCATGGTGACCAGC-3′) was annealed and ligated into the px458 vector (Ran et al., 2013).

Silencer Select siRNAs (Thermo Fisher Scientific) were used for Drp1-knockdown experiments. Cells were transfected with siRNA oligonucleotides with the Lipofectamine RNAiMAX Reagent (Thermo Fisher Scientific). The cells were then used for subsequent assays after incubation for 72 h. The siRNA sequences were 5′-CGAUUGAAGGAACCGCAAATT-3′ and 5′-GCAAUUGAGCUAGCGUAUATT-3′. The negative control #1 and #2 siRNA [Thermo Fisher Scientific, catalog nos 4390843 (#1) and 4390847 (#2)] were used as controls.

NAC was obtained from Sigma-Aldrich. Etoposide, Hoechst 33342, a CellROX Green Flow Cytometry Assay Kit (including TBHP, NAC), ProLong Diamond, and MitoSOX Red Reagents were obtained from Thermo Fisher Scientific. The Seahorse XF Cell Mito Stress Test Kit was obtained from Primetech. 4-PBA was obtained from Tocris.

Polyclonal and monoclonal antibodies to cleaved caspase-3 (#9661, #9664; 1:500) and to Mfn2 (#9482S; 1:500) were obtained from Cell Signaling; antibodies to p38 MAPK (sc-535; 1:1000), to Tom20 (sc-11415; 1:400) and to Bax (sc-493; 1:1000) were from Santa Cruz Biotechnology; those to α-tubulin (T6199; 1:1000) were from Sigma; those to cytochrome c [556432; 1:400 for immunofluorescence (IF), 556433; 1:500 for western blotting (WB)] and to Drp1 (611112; 1:400 for IF, 1:1000 for WB) were from BD Pharmingen; those to Pex14 (10594-1-AP; 1:400) and to Mff (17090-1-AP; 1:100 for IF, 1:500 for WB) were from Proteintech; those to Pex3 (HPA042830; 1:500) were from Atlas Antibodies; those to ATP synthase β (A21351; 1:400) were from Thermo Fisher Scientific; those to caspase-9 (M054-3; 1:500) were from MBL Life Science; those to Mfn1 (ab57602; 1:500) and to VDAC2 (ab47104; 1:500) were from Abcam; and those to Pex5 were described previously (Okumoto et al., 2014).

Quantitative data are presented as means±s.e.m. and were compared with Scheffe's test or the two-tailed unpaired or paired Student's t-test. A P value of <0.05 was considered statistically significant.

We thank M. Okajima (Graduate School of Pharmaceutical Sciences, The University of Tokyo) for technical assistance, N. Ishihara (Institute of Life Science, Kurume University) and T. Kitamura (The Institute of Medical Sciences, The University of Tokyo) for providing the plasmids, H. Ichijo (Graduate School of Pharmaceutical Sciences, The University of Tokyo) for providing the cells. We also appreciate our laboratory members for discussions.