Mitochondria and peroxisomes cooperate in versatile metabolic processes, exchange many molecules, and also share proteins required for fission and quality control (Schrader et al., 2015; Wanders et al., 2016). This peroxisome–mitochondria connection has been excellently discussed before (Fransen et al., 2017; Costello et al., 2018), and we only focus on several examples. Both organelles can be cellular sites for β-oxidation of fatty acids—while in several fungi and plants, β-oxidation occurs exclusively inside of peroxisomes, other fungi and animals harbor full sets of β-oxidation enzymes inside of both organelles (Maggio-Hall and Keller, 2004; Poirier et al., 2006; Goepfert and Poirier, 2007; Kretschmer et al., 2012b; Camões et al., 2015). An important difference between the compartments is the mechanisms they use for the import of proteins: While mitochondrial proteins often rely on N-terminal targeting signals and can be imported in an unfolded or flexible state (Backes and Herrmann, 2017), import into the peroxisomal matrix is different.

To exchange intermediates of the β-oxidation pathway, substitution of a coenzyme A moiety with carnitine is one possibility to shuttle acetyl and acyl groups as acetyl- or acylcarnitine units (Antonenkov and Hiltunen, 2012). Acetylcarnitine transferases are dually localized enzymes, and isoforms with different targeting signals have been reported, which can be generated via alternative transcriptional and translational start sites or differential splicing (Corti et al., 1994; Elgersma et al., 1995; Ueda et al., 1998; Houten et al., 2020). Within mitochondria, carnitine is replaced by coenzyme A to enable further metabolization (Houten et al., 2020).

A second cellular pathway enables the exchange of C2 units between peroxisomes and mitochondria, albeit more indirectly. Citrate can be generated from oxaloacetate via citrate synthases, which are not only part of the Krebs cycle but also part of the glyoxylate cycle, partially located inside of peroxisomes (Kim et al., 1986; Kunze et al., 2006). The citrate synthase Cit2 has overlapping functions with acetylcarnitine transferase Cat2p in S. cerevisiae and is contained in peroxisomes (Van Roermund et al., 1995; van Roermund et al., 1999; Swiegers et al., 2001; Shai et al., 2018). Of interest, Cit2p can be targeted to mitochondria, as well as to peroxisomes, and can compensate for the absence of mitochondrial citrate synthase Cit1p (Kim et al., 1986; Hoppins et al., 2011; Lee et al., 2011; Morgenstern et al., 2017). Citrate can be shuttled via dedicated transport proteins linking the mitochondrial matrix to the cytosol (Kaplan et al., 1995; Palmieri, 2004).

How are proteins distributed that do not come in different isoforms but contain a mitochondrial targeting signal at the N-terminus and a PTS1 at the C-terminus? Previously, it was discussed that these usually end up in mitochondria, because the N-terminal signal will exit the ribosome first, can directly interact with the mitochondrial import machinery, and thus will dominate a PTS1—a hierarchy of targeting signals is likely to exist (Kunze and Berger, 2015). Indeed, the mitochondrial ribosomal protein Mrp7p contains a functional PTS1 in S. cerevisiae but does not localize in peroxisomes presumably because the N-terminal signal is recognized first (Neuberger et al., 2004). However, several bona fide peroxisomal proteins such as the thiolase Tes1p or the catalase Cta1p have been identified in highly purified mitochondria of S. cerevisiae (Morgenstern et al., 2017).

Another PTS1-containing protein—the protein phosphatase Ptc5p—has been recently shown to first target mitochondria where its N-terminus is proteolytically removed. Subsequently, the protein is translocated into the peroxisomal lumen via interaction with Pex5p (Stehlik et al., 2020) (Figure 2). Further proteins were also shown to be retargeted from mitochondria to the cytosol, e.g., S. cerevisiae fumarase (Stein et al., 1994; Yogev et al., 2010). The sorting of Ptc5p is different as the presence of Pex5p is critical for export from mitochondria, indicating an interaction between Ptc5p and Pex5p prior to full mitochondrial import. One target of Ptc5p in peroxisomes is Gpd1p (Stehlik et al., 2020), and dephosphorylation of Gpd1p increases its activity (Lee et al., 2012). In line with these data, a synthetic growth defect was observed for strains deleted for PTC5 and MDH3 (Costanzo et al., 2016; see the previous chapter on the redundancy of Gpd1p and Mdh3p). Together, these data suggest a network consisting of Ptc5p, Gpd1p, and Mdh3p to regulate peroxisomal NADH metabolism presumably in a manner adapting the different compartments to each other. In A. thaliana, the sorting of NAD(P)H dehydrogenases may resemble Ptc5p from S. cerevisiae as competing N- and C-terminal targeting signals for mitochondria and peroxisomes are involved (Carrie et al., 2008).

Targeting of proteins with competing targeting signals may represent a more generic approach to regulate the interaction and communication of compartments—the overexpression of several of these proteins increases the fraction of peroxisomes associated with mitochondria (Stehlik et al., 2020). In mammalian cells, the overexpression of the dual affinity protein ACBD2/ECI2 also enhances the proximity of peroxisomes and mitochondria (Fan et al., 2016). The protein Cnm1p with competing targeting signals at its termini can increase the attachment of mitochondria to the nuclear envelope in S. cerevisiae, suggesting a similar mechanism occurring between another pair of organelles (Eisenberg-Bord et al., 2021).

In addition to luminal proteins, peroxisomes and mitochondria share a number of membrane proteins or membrane-associated proteins with various tasks (also see Fransen et al., 2017; Costello et al., 2018). A key process controlled by overlapping factors is fission. Both in S. cerevisiae and in mammalian cells, the C-terminally tail-anchored protein Fis1p/FIS1, dynamin-related GTPases, and additional shared proteins such as the WD40 repeat-containing protein Caf4p belong to the factors involved in the fission process (Koch et al., 2003; Praefcke and McMahon, 2004; Koch et al., 2005; Kuravi et al., 2006; Motley et al., 2008; Motley et al., 2015; Castro et al., 2018). C-terminally tail-anchored membrane proteins such as FIS1 tend to be dually localized, and it was shown that the hydrophobicity of the transmembrane domain together with changes in the charge of the luminal tail determines targeting efficiency and hence the subcellular distribution (Costello et al., 2017a). Remarkably, the peroxisomal targeting factor Pex19p is required for the correct sorting of Fis1p to mitochondria and to peroxisomes in S. cerevisiae (Cichocki et al., 2018) (Figure 2). How the distribution of proteins such as Fis1p is adapted to the needs of both organelles is not known yet.

Further proteins operating at mitochondria and peroxisomes are ATPases termed Msp1p in S. cerevisiae or ATAD in mammals, which can extract superfluous or mistargeted tail-anchored proteins (Chen Y.-C. et al., 2014; Okreglak and Walter, 2014; Weir et al., 2017). Of interest, a variety of PMPs can be removed from mitochondria via Msp1p/ATAD, suggesting a broader range of substrates (Nuebel et al., 2021). Furthermore, these authors observed that many peroxisomal proteins are not degraded or downregulated in the absence of preexisting peroxisomes, but functional PMP-containing protein complexes assemble on mitochondria, which causes unfavorable consequences such as the import of PTS1 cargo into mitochondria (Nuebel et al., 2021). Mitochondrial dysfunction in the absence of functional peroxisomes is emerging as one cause of symptoms in patients lacking peroxisomes (Wanders, 2014; Lismont et al., 2019; Schrader et al., 2020; Nuebel et al., 2021). The ubiquitin ligase MARCH5—another protein for quality control—was shown to localize to peroxisomes in addition to its known location at mitochondria in mammals. At peroxisomes, MARCH5 is involved in controlling selective autophagy (Zheng et al., 2021).

A multi-subunit protein structure associated with both organelles in fungi is the ERMES (ER–mitochondria encounter structure) complex, which has a role in tethering of mitochondria to the ER and to peroxisomes to form putative three-way junctions (Kornmann et al., 2009; Cohen et al., 2014; Ušaj et al., 2015; Kundu and Pasrija, 2020). Moreover, the ERMES regulating GTPase Gem1p can be found in peroxisomes and mitochondria of S. cerevisiae (Kornmann et al., 2011; Cichocki et al., 2018). In mammalian cells, the Gem1p ortholog MIRO1 is dually localized as well and has a role in recruiting the lipid transfer protein VPS13D probably involved in sustaining organellar growth and lipid exchange (Castro et al., 2018; Baldwin et al., 2021; Guillén-Samander et al., 2021). Thus, mitochondria and peroxisomes use an overlapping set of proteins to connect to the ER.

A different mode of transport to accomplish the dual localization of selected proteins is vesicular trafficking from one organelle to the other. Dynamin-independent carriers were proposed to transport the ubiquitin ligase MAPL from mitochondria to peroxisomes (Neuspiel et al., 2008). More recently, it was suggested that de novo peroxisome formation involves mitochondria-derived vesicles (Sugiura et al., 2017).

All these data point to a close relationship of mitochondria and peroxisomes not only in terms of metabolism but also in terms of biogenesis, quality control, and turnover. This remarkable overlap might reflect peroxisome evolution as partially mitochondria-derived organelles (Bolte et al., 2015; Speijer, 2017). Are peroxisomes indeed outposts of mitochondria, and how can this be in line with the key role of the endoplasmic reticulum for peroxisome biogenesis (Joshi et al., 2017; see also in the next chapter)?

We speculate that peroxisomes may have emerged from an ancient ER-derived quality control compartment involved in clearing specific mitochondrial proteins, especially under conditions of oxidative stress. In this perspective, the ERAD-related peroxisomal import machinery (Schliebs et al., 2010) could be regarded as an extraction machine for mitochondrial proteins working in trans. Particularly, the oxidative enzymes of the β-oxidation pathway may be a major burden for mitochondrial metabolism, which may explain why this pathway has been completely transferred to peroxisomes of several species, e.g., S. cerevisiae (Hiltunen et al., 2003; Poirier et al., 2006; Speijer, 2017). The indirect targeting of the phosphatase Ptc5p via mitochondrial transit (Stehlik et al., 2020) could be a snapshot or remnant of this evolutionary scenario. In this effect, peroxisomes may represent ancient molecular vacuum cleaners of the mitochondrial surface (Figure 3), which became autonomous over time by hitchhiking the mitochondrial division and quality control machinery. Further shared proteins might still be obscure—one study indicated a role for the dually localized mitofusin ortholog Fzo1 from S. cerevisiae for tethering peroxisomes to mitochondria (Shai et al., 2018).

The endoplasmic reticulum (ER) is the major source of cellular lipids and has a key role in supporting the growth of the entire cell (Ferro-Novick et al., 2013). Early work already revealed intimate connections between this large-supply organelle and peroxisomes (Novikoff and Shin, 1964). Metabolites are exchanged between both compartments, e.g., during the biosynthesis of ether lipids (Wanders et al., 2016). In S. cerevisiae, the sorting of the multifunctional PMP Pex3—a key factor in regulating PMP import into peroxisomes—from the ER to peroxisomes was described (Hoepfner et al., 2005; Jansen and van der Klei, 2019). ER-derived vesicular carriers were reported to contain peroxisomal proteins, and machinery was uncovered that is relevant for the emergence of peroxisomal vesicles from the ER in yeasts (Titorenko et al., 2000; Lam et al., 2010; Agrawal et al., 2011; Van der Zand et al., 2012; Agrawal et al., 2016; Mast et al., 2018). Hence, vesicular trafficking from the ER represents one road for lipids and proteins to the peroxisome and also operates in mammalian cells (Kim et al., 2006; Aranovich et al., 2014; Sugiura et al., 2017). In S. cerevisiae and other yeasts, ER-resident microdomain-forming reticulon proteins play a critical role in tethering of preexisting peroxisomes (David et al., 2013; Mast et al., 2016; Wu et al., 2020; Ferreira and Carvalho, 2021). Interestingly, the same domains may be required for the budding of peroxisome precursors (Joshi et al., 2016; Joshi et al., 2018; Wang et al., 2018). As an alternative to vesicle formation, direct lipid transfer promotes the growth of peroxisomes and recruitment of lipid transfer proteins such as VPS13 is involved in peroxisome biogenesis both in yeasts and in mammals (Raychaudhuri and Prinz, 2008; Baldwin et al., 2021; Guillén-Samander et al., 2021; Yuan et al., 2022). How these mechanisms for lipid supply each contribute to peroxisome formation is an exciting question for future work. An intimate physical connection between peroxisomes and the ER is probably required for the correct activity of both mechanisms. This can, e.g., be supported through an interaction between the peroxisomal acyl-CoA binding protein ACBD5 and the ER membrane-associated protein VAPB in mammalian cells (Costello et al., 2017b; Hua et al., 2017).

Many PMPs are synthesized in the vicinity of the peroxisome of S. cerevisiae (Zipor et al., 2009; Dahan et al., 2022), and direct insertion into the peroxisomal membrane was demonstrated for several of them (Sacksteder et al., 2000; Fang et al., 2004; Jones et al., 2004; Matsuzaki and Fujiki, 2008; Yagita et al., 2013; Chen Y. et al., 2014). Interestingly, other PMPs are synthesized proximal to the ER and may reach the peroxisome via ER transit (Jan et al., 2014). For Pex3, dual localization to peroxisomes and the ER of S. cerevisiae was shown by fractionation experiments (Mast et al., 2016). Hence, the ER is not only an intermediate but also a relevant steady-state location for certain PMPs, indicating functions beyond ER transit. Indeed, in mammalian cells ER-localized PEX3 is involved in the sorting of a protein designated for lipid droplets (Schrul and Kopito, 2016). In addition to PEX3, the soluble targeting factor PEX19 is associated with the ER and involved in targeting reticulon-homology proteins to this organelle besides its prominent function in peroxisome biogenesis (Yamamoto and Sakisaka, 2018; Zimmermann et al., 2021).

The coat protein I (COPI) complex localizes primarily along the organelles of the secretory pathway where it is involved in retrograde transport (Barlowe and Miller, 2013). In addition, this protein assembly can be associated with peroxisomes and also with mitochondria, but the molecular function of the localization at these organelles is not fully understood (Passreiter et al., 1998; Barlowe and Miller, 2013; David et al., 2013; Zabezhinsky et al., 2016). Remarkably, COPI-dependent sorting of a viral protein from peroxisomes to the ER was reported (McCartney et al., 2005). The small GTPase Arf1p—a major regulator of COPI—shows multiple subcellular localizations including peroxisomes and mitochondria in S. cerevisiae (Ackema et al., 2014; Yofe et al., 2017). So far, the role of COPI-dependent vesicular trafficking for peroxisome biogenesis is elusive although in S. cerevisiae, COPI components can be purified together with the subdomain-forming protein Pex30p, which is involved in peroxisome formation (David et al., 2013). Pex30p is also found at sites of lipid droplet formation in S. cerevisiae, and COPI was shown to be involved in formation of this organelle in Drosophila cells (Wilfling et al., 2014; Joshi et al., 2018; Wang et al., 2018; Ferreira and Carvalho, 2021). Further research is required to understand how key factors of the secretory pathway are dynamically distributed between the different organelles and how this coordinates organelle biogenesis at different sites. Nevertheless, the discussed examples emphasize that the ER and peroxisomes are possible alternative destinations of several factors required for the proper maintenance of each.

Another interesting example of dual targeting was described in A. thaliana—a purple acid phosphatase-containing competing targeting signals is localized to the ER and to peroxisomes (Kataya et al., 2016). Various other phosphatases with noncanonical PTS1 motifs were identified in this study that may reside inside of peroxisomes and in additional compartments including the nucleus (Kataya et al., 2016).

Different types of pathogenic attack trigger the nuclear localization of plant catalases (Inaba et al., 2011; Zhang M. et al., 2015). A recent study in A. thaliana uncovered that catalase is localized both in peroxisomes and in the nucleus even in the absence of any infection (Al-Hajaya et al., 2022). Regulation of dual targeting via metabolites such as H₂O₂ is likely to operate in plants similar to what was observed in mammals (Okumoto et al., 2020). Peroxisomes together with mitochondria have a key role in controlling ROS production, which is important for cellular signaling and needs to be maintained on a level acceptable for the integrity of cells, especially of the nucleus (Fransen et al., 2017; Sies, 2017; Li et al., 2021; Paradis et al., 2021).

Interestingly, the hypoxia-inducible transcription factor Hif1 and hydroxylases regulating its activity are found in peroxisomes and mitochondria (Khan et al., 2006; Li et al., 2019). Sequestration upon reoxygenation was suggested as a possible biological function of peroxisomal targeting (Khan et al., 2006).

Lipid droplets are sites of cellular fat storage and mobilization. Their biosynthesis and function are both intricately linked to peroxisomes. Both organelles require overlapping proteins for biogenesis, e.g., the lipodystrophy protein seipin and other factors including Pex30p (Szymanski et al., 2007; Salo et al., 2016; Joshi et al., 2017; Joshi et al., 2018; Wang et al., 2018). The targeting factor PEX19 and ER-localized PEX3 facilitate the insertion of the membrane protein UBXD8 into a subdomain of the ER membrane, which turned out as a prerequisite for subsequent transfer to lipid droplets in mammalian cells (Schrul and Kopito, 2016). Upon depletion of PEX19, UBXD8 appeared predominantly in mitochondria. Of interest, a proper targeting of UBXD8 requires C-terminal farnesylation of PEX19 (Schrul and Kopito, 2016).

Peroxisomes and lipid droplets are physically connected, and membrane protrusions emerging from peroxisomes have been shown to attach to lipid droplets in S. cerevisiae (Binns et al., 2006). In A. thaliana, a dually localized lipase can be sorted from peroxisomes to lipid droplets via similar protrusions in the course of fat mobilization during seed germination (Thazar-Poulot et al., 2015). In a similar way, in fasting animals triglyceride lipases are translocated from peroxisomes to lipid droplets depending on the targeting factor PEX5. This process was proposed to involve sites of organelle contact (Kong et al., 2020) (Figure 4).

Since peroxisomes are enclosed by a lipid bilayer and lipid droplets only by a monolayer, the dual localization of membrane proteins is likely to depend on dual topology. This was demonstrated for mammalian fatty acyl-CoA reductase 1, which is targeted to peroxisomes and lipid droplets and transits via the ER (Exner et al., 2019) (Figure 4).

Knowledge about proteins destined for peroxisomes and the plasma membrane or peroxisomes and endosomes is limited. It has been described that fungal peroxisomes hitchhike on motile endosomes to move through fungal hyphae (Guimaraes et al., 2015; Salogiannis et al., 2016). In A. nidulans, the association between endosomes and peroxisomes is mediated by the linker protein PxdA (Salogiannis et al., 2016). Proteins required for the maturation of endosomes such as members of the ESCRT (endosomal sorting complex required for transport) pathway (Henne et al., 2011) are involved in peroxisome biogenesis (Mast et al., 2018). Of interest, recent work revealed intraluminal vesicles inside of plant peroxisomes, which might also require the function of the ESCRT pathway, hinting at some parallels to multi-vesicular bodies (Wright and Bartel, 2020).

Insights into the molecular mechanisms that regulate the distribution of proteins among their different destinations are of interest to understand communication and homeostasis of cellular compartments. Studies on the dual localization of catalase in mammalian cells could represent a blueprint—it was shown that an elevated cytosolic concentration of H₂O₂ provokes cytosolic retention of this detoxifying enzyme (Walton et al., 2017). Recent studies showed phosphorylation of the peroxisomal membrane Pex14 in response to oxidative stress and during mitosis, which reduces the peroxisomal import of catalase to a greater extent than the import of other tested matrix proteins (Okumoto et al., 2020; Yamashita et al., 2020). Modification of the peroxisomal import machinery, hence, has an important role to adapt the peroxisomal and cytosolic protein composition to cellular demands. In fungi, the peroxisomal import of catalase is exceptional, and it relies on a noncanonical PTS1 and a second unusual peroxisomal targeting signal closer to the N-terminus (Kragler et al., 1993; Williams et al., 2012). Substitution of the C-terminus with a canonical targeting signal lowers catalase activity and leads to the aggregation of the enzyme (Williams et al., 2012). In S. cerevisiae, peroxisomal localization of the glyoxylate cycle enzyme citrate synthase 2 is reduced upon expression of a version of Pex14 mimicking a phosphorylated state (Schummer et al., 2020). Phosphorylation of the targeting factor Pex5 has been implicated in the regulation of pexophagy upon stress exposure in mammalian cells and occurs in different species (Zhang J. et al., 2015; Oeljeklaus et al., 2016). How phosphorylation of Pex5 modulates the import of specific cargo besides regulating peroxisome breakdown remains to be established. Redox regulation at an N-terminal cysteine residue is another way to control Pex5 activity (Ma et al., 2013).

Phosphorylation of dually localized cargo proteins is also involved in regulating their intracellular distribution (Figure 1). S. cerevisiae glycerol-3-phosphate dehydrogenase Gpd1 is phosphorylated in the vicinity of the PTS2 (Jung et al., 2010). While phosphomimetic variants of Gpd1 are efficiently targeted to the peroxisome, variants resembling non-phosphorylated Gpd1 are retained in the cytosol.

Many fungal species encode at least two versions of the PTS1 receptor Pex5 (Kiel et al., 2006). The expression of the Pex5 paralog Pex9 is induced upon incubation of S. cerevisiae cells in oleic acid medium and regulates the import of a subset of peroxisomal matrix proteins with specific targeting signals (Effelsberg et al., 2016; Yifrach et al., 2016; Yifrach et al., 2022). In the corn smut fungus Ustilago maydis, two Pex5 paralogs with different cargo selectivity are required for peroxisome function (Ast et al., 2022). The concentration or availability of different targeting factors at peroxisomes and at other organelles may emerge as an additional regulatory device to control the subcellular distribution of proteins.