Dually targeted proteins regulate proximity between peroxisomes and partner organelles

Peroxisomes play a central role in fatty acid metabolism. To correctly target to peroxisomes, proteins require specialized targeting signals. One mystery in the field is sorting of proteins that carry both a targeting signal for peroxisomes as well as for other organelles such as mitochondria or the endoplasmic reticulum (ER). Exploring several of these dually localized proteins in Saccharomyces cerevisiae, we observed that they can act as dynamic tethers bridging organelles together through an affinity for organelle-destined targeting factors. We show that this mode of tethering involves the peroxisome import machinery, the ER– mitochondria encounter structure (ERMES) in the case of mitochondria and the GET complex in the case of the ER. Depletion of each of the targeting factors resulted in the accumulation of smaller peroxisomes. We propose that dual targeting of proteins occurs at contact sites and that protein import per se contributes to the maintenance of these membrane proximities. This introduces a previously unexplored concept of how targeting of dual affinity proteins can support organelle attachment, growth and communication.


Introduction
Metabolic pathways in eukaryotic cells are often compartmentalized inside membraneenclosed organelles such as mitochondria and peroxisomes [1]. Localized proteins synthesized in the cytosol must target and translocate into their organelles of residence correctly, efficiently and in a regulated manner [2][3][4]. The majority of mitochondrial proteins are imported from the cytosol in a conformationally flexible state via evolutionary conserved protein complexes called translocase of the outer membrane (TOM) and translocase of the inner mitochondrial membrane (TIM). Many proteins destined for the mitochondrial matrix or inner membrane contain Nterminal targeting signals that are cleaved by proteases upon import (Wiedemann & Pfanner, 2017).
Protein translocation into the matrix of peroxisomes requires distinct targeting signals termed peroxisome targeting signal type 1 and 2 (PTS1 and PTS2, respectively). PTS1 motifs are recognized by the cytosolic targeting factor Pex5 and translocated via interaction of Pex5 with Pex13/14 on the peroxisomal membrane allowing the translocation of folded and even oligomerized protein complexes in a yet unresolved mechanism [4,6,7]. Peroxisomes can multiply by growth and division, but unlike mitochondria can also form de novo in the absence of mature peroxisomes from the endoplasmic reticulum (ER) and possibly also involving mitochondria [4,8,9].
For many years proteins were studied that either have a C-terminal PTS1 or an N-terminal mitochondrial targeting signal (MTS). Recently, it was shown that the protein phosphatase Ptc5 of the yeast S. cerevisiae (from hereon called simply yeast) contains both targeting signals and exhibits a dual localization to peroxisomes and mitochondria [10]. Ptc5 is processed by the inner membrane peptidase (IMP) complex inside mitochondria and reaches the peroxisomal We then utilized overexpressed internally Myc-tagged versions of two candidate proteins integrated into the LEU2 locus: Cat2 and Pxp2. Even without the RFP tag, Pxp2-Myc-PTS and Cat2-Myc-PTS both increased the association of peroxisomes and mitochondria ( Fig. 1G) with Pxp2 giving a stronger effect. A difference between both proteins was also reflected by their subcellular distribution. Whereas Cat2-RFP-PTS was found to localize distributed over mitochondria and in peroxisomes, Pxp2-RFP-PTS occurred in foci at mitochondria frequently proximal to a peroxisome but also inside peroxisomes with no contact to mitochondria ( Fig. 1B and 1D). Pxp2-RFP-PTS became more evenly distributed in mitochondria upon deletion of PEX5 (Fig. 1D). Thus, overexpression of proteins that have a mitochondrial and a peroxisomal targeting signal can enhance the formation of contacts between both organelles.

Loss of the matrix protein targeting machinery reduces the number of contacts between peroxisomes and mitochondria
We reasoned that if tethering is caused by dual targeting then we should be able to affect it by eliminating the targeting factor Pex5. We therefore quantified the number of peroxisomes associated with mitochondria in control and ∆pex5 cells. A significant reduction of peroxisomes proximal to mitochondria was observed in the absence of Pex5 ( Fig. 2A and 2B) indicating a role of the peroxisomal matrix targeting machinery for the generation of contacts -something unexpected if only membrane proteins were required for tethering as is usually assumed.
Indeed, we could block enhanced tethering triggered by overexpression of Pxp2-RFP-PTS1 by deleting PEX5 (Fig. 2C). Furthermore, we quantified peroxisome number and detected an increase of smaller appearing peroxisomes upon deletion of PEX5 (Fig. 2D). Increased movement of peroxisomes was detected in ∆pex5 cells (Movie 1), a phenotype previously described for other mutants with reduced organelle tethering [e.g. ref 24]. These data show that overexpression of dual affinity PTS1 proteins and depletion of Pex5 have a reciprocal effect on peroxisome number and appearance.
To confirm our findings and reduce the possibility that the observed phenotype resulted from secondary, downstream effects we took an approach for rapid depletion of the Pex5 receptor Pex13 [7], using an auxin-inducible degron (AID) tag [27,28] (Fig. 2E). Depletion of Pex13 resulted in reduced import of RFP-PTS1 (Fig. 2F). Association of peroxisomes and mitochondria was quantified in a strain co-expressing Cat2-RFP-PTS as an additional tether relative to a control strain ( Fig. 2G and 2H, Fig. S3). (E) Scheme of the genetic modifications used for auxin-dependent depletion of Pex13 in (F), (G) and (H). The endogenous PEX13 locus was genetically engineered to encode a translational fusion of Pex13 with a C-terminal auxin-inducible degron (AID) and six hemagglutinin (HA) tags. Pex13 degradation is mediated by the F-box protein AFB2 from A. thaliana, which was expressed from the ADH1 promotor (F) Fluorescence microscopic images of indicated strains expressing the peroxisomal membrane protein Ant1-YFP (green), the mitochondrial inner membrane protein Tim50-CFP (blue) and RFP-PTS (red) after a 4 h incubation period in the absence (-Auxin) or presence (+Auxin) of 2 mM indole-3-acetic acid (left). Auxin-dependent depletion of Pex13-AID-HA at indicated time points was analyzed by SDS-PAGE and immunoblot (right). Por1 served as a loading control. (G) Subcellular localization of Ant1-YFP and Tim50-CFP of indicated strains was analyzed in the presence of 2 mM indole-3-acetic acid at indicated time points (left). Arrows indicate peroxisomes in close proximity to mitochondria. The fraction of peroxisomes in contact with mitochondria (PxM) relative to the total peroxisome count (PxT) of the indicated strain was quantified at the indicated time points and growth conditions (right). (H) Identical to (G), except that the cells also expressed Cat2-RFP-PTS1 to increase PerMit contacts. (I) Schematic illustration of a galactose-inducible conditional pex5 mutant overexpressing Ptc5 ΔTM -RFP-PTS to generate additional PerMit contacts. (J) Quantification of the fraction of peroxisomes in contact with mitochondria (PxM) relative to the total peroxisome count (PxT) of the strain described in (H) at the indicated time points and growth conditions from images derived from time lapse microscopy. Scale bars represent 5 µm. Quantifications are based on n = 3 experiments. Each color represents one experiment. Error bars represent SEM. P-values were calculated using a two-sided unpaired Student's t-test.
Both strains started with a different degree of organelle association but depletion of Pex13-AID-HA by auxin addition reduced the number of peroxisomes proximal to mitochondria to a similar basal level supporting a role of matrix protein targeting in driving organelle association.
To corroborate these results we created a strain with conditional PEX5 expression by putting it under control of the GAL promoter (Fig. 2I). Immunoblot and fluorescence microscopy of RFP-PTS1 expressing cells confirmed that Pex5 levels correlated with PTS1 import (Fig. S3). We then overexpressed a Ptc5∆TMD-RFP-PTS, which is a dual targeted derivative of Ptc5 lacking the transmembrane domain [10]. Automated time lapse imaging revealed that induction of Pex5 elevated the number of peroxisomes close to mitochondria, whereas its reduction had the opposite effect (Fig. 2J). Since single planes were imaged, lower basal overlap of mitochondria and peroxisomes were observed in the absence of Pex5 (compare to Fig. 2B). Together, these data suggest a function of peroxisomal matrix protein import for formation of PerMit contacts.

Identification of factors that regulate dual targeting of Ptc5
To identify additional factors involved in dual targeting and therefore potentially also tethering, we conducted a high content genetic screen focused on Ptc5 sorting from mitochondria to peroxisomes (Fig. 3A). Peroxisomal import of Ptc5-RFP-PTS depends on mitochondrial processing and subsequent Pex5 dependent targeting [10]. This transit mechanism may require direct contact between the organelles. Crossings of a wildtype strain expressing Ptc5-RFP-PTS and Pex3-GFP as a peroxisomal marker with arrayed libraries containing yeast deletion strains and hypomorphic mutants of essential genes were performed and sporulated to select for haploid cells containing both fluorescent proteins together with a genetic perturbation (Fig. 3A) [29,30]. These include many expected genes essential for peroxisome biogenesis but also distinct genes e.g. SOM1. Som1 was shown previously to be part of the IMP complex [31], which we could confirm (Fig. 3D). Moreover, we identified Pex34 as a factor for correct sorting of Ptc5-RFP-PTS to peroxisomes (Fig. 3C). Pex34 was identified as a potential PerMit tether [16]. This result strengthen our hypothesis that the proximity of mitochondria and peroxisomes is crucial for peroxisomal targeting of Ptc5-RFP-PTS via mitochondrial transit and that our assay is likely to report on tethering.

The ERMES complex regulates dual targeting between mitochondria and peroxisomes
An interesting gene target from the screen was Mdm10 (a critical factor for connecting mitochondria to the ER as part of the ER-mitochondria encounter structure (ERMES)) (Figs. 3 and 4A). The ERMES complex consists of the ER-anchored protein Mmm1 and three mitochondrial proteins (Mdm10, Mdm12 and Mdm34) [32]. The integral membrane protein Mdm10 has an additional function in assisting import of mitochondrial ß-barrel proteins [33].
It was therefore unclear whether the effect of Mdm10 was through the ERMES or by affecting mitochondrial protein import. To address whether the observed phenotype is specific to Mdm10 or shared by other ERMES complex members, we repeated our assay in freshly made strains. In all ERMES mutants targeting of Ptc5-RFP-PTS to peroxisomes was significantly reduced (Figs. 4B and S4A) with ∆mdm10 displaying the strongest phenotype.
To test the stage at which ERMES mutants affected the dually targeted proteins, we first analyzed the initial steps of mitochondrial protein import. All mutants influenced mitochondrial import and preprotein processing as determined by testing a truncated variant of Ptc5-RFP (Fig.   4C). Moreover, we observed an import defect for several bona fide mitochondrial proteins in ∆mdm10 cells (Fig. S4B) demonstrating that loss of ERMES complex alters translocation of mitochondrial proteins in a more general way.
To follow up on these results, we determined the localization of other proteins that contain competing targeting signals in ∆mdm10 cells. Cat2-RFP-PTS behaved similar to Ptc5-RFP-PTS and was retained in mitochondria while targeting to peroxisomes was reduced (Fig. 4D).
In contrast, the localization of Pxp2-RFP-PTS, Tes1-RFP-PTS and Mss2-RFP-PTS displayed the opposite pattern of localization in Mdm10 depleted cells with reduced mitochondrial targeting but normal peroxisome targeting (Fig. 4E). Proteolytic processing defects could not be observed for any of these proteins suggesting a noncanonical targeting signal for mitochondria ( Fig. 4F). Quantification showed a reduction of mitochondrial foci containing Pxp2-RFP-PTS (Fig. 4G). Proteins containing a classical mitochondrial presequence such as Ptc5 may therefore require ERMES-mediated proximity to translocate to peroxisomes, whereas proteins without a canonical signal do not efficiently target mitochondria in the absence of ERMES. Given these results, we suggest that previous literature showing that overexpression of Mdm10 and Gem1 enhanced PerMit contacts [16] exposed their ability to provide excess of dual affinity proteins to mitochondria. ERMES components could be directly involved in mitochondrial targeting as we found that Pxp2-RFP-PTS foci regularly (38% +/-5%) overlapped with Mmm1-mNeonGreen foci ( Fig. 4H; Supplementary table 3). An indirect function of ERMES is likewise possible and a potential saturation of the import machinery on the mitochondrial surface provokes the altered localization of dual affinity proteins such as Pxp2, Mss2 and Tes1.

ERMES controls proximity of peroxisomes and mitochondria
Quantification of peroxisomes in all ERMES mutants showed that they contained a higher number of small peroxisomes in a manner similar to a ∆pex5 strain (Fig. 2) and that import of RFP-PTS1 was reduced (Fig. 5A). Moreover, the fraction of peroxisomes in contact to mitochondria declined in the ERMES mutants ( Fig. 5B and 5C). These phenotypes often reverted and, for example, a ∆mdm12 deletion strain with wildtype like growth and regular mitochondrial morphology retained a higher number of peroxisomes (Fig. S5). Indeed, phenotypic suppression of ERMES mutants by second site mutations is regularly observed [34]. However, the effect of ERMES loss on peroxisome number, proximity to mitochondria and matrix targeting speaks for a specific function of ERMES proteins on peroxisome formation independent of mitochondrial dysfunction as was also suggested elsewhere [17,35].
The above led us to consider that ERMES may affect the proximity of mitochondria and peroxisomes -directly or indirectly. Since we had noticed that overexpression of dual-affinity proteins resulted in increased proximity and a reduced number of peroxisomes ( Fig. 1), we tested whether a synthetic tether -designed to connect peroxisomes and mitochondria (Fig. 5D) also affected peroxisomes in a manner that rescued loss of ERMES components. We found that in both ∆mdm10 and ∆mdm34 cells expression of a synthetic tether decreased peroxisome number (Figs. 5E and S6). In addition, expression of this synthetic tether activated substantial expansion of peroxisomes ( Fig. 5E and 5F).
Upon deletion of MDM34 (and less so in ∆mdm10) we were able to visualize Pxp2-RFP-PTS, but not the soluble peroxisomal marker protein RFP-PTS (Figs. 5F and S6) accumulated at junctions between mitochondria and the enlarged peroxisomes (Figs. 5E and S6). This observation supports the function of Pxp2-RFP-PTS as a dual affinity tether distributed at sites of organelle contact. Furthermore, expression of the synthetic tether partially restored peroxisomal targeting of Ptc5-RFP-PTS in ∆mdm10 cells (Fig. 5G). These findings strengthen a proposed role of ERMES as a regulator of the interaction of mitochondria and peroxisomes.
Deletion of PEX5 and deletion of MDM10 both resulted in a higher number of small peroxisomes, with ∆mdm10 having the stronger effect ( Fig. 5H and 5I). The phenotype of the double mutant was more reminiscent to ∆mdm10 single mutants in line with a potential connection of both factors (Fig. 5H and 5I). The stronger phenotype of the ∆mdm10 deletion strain points to additional functions of ERMES for peroxisome formation beside its role in regulating trafficking of dual affinity PTS1 proteins. These may include stabilization of contacts via interaction with membrane proteins such as Pex11 as previously suggested [17] or enhanced activatity of fission proteins such as the dually localized tail-anchored (TA) fission protein Fis1 on peroxisomes [36] in cells lacking ERMES. Indeed, additional deletion of FIS1 partially suppressed peroxisome accumulation in ∆mdm10 cells (Fig. 5J and K).

The TA-protein Pex15 contributes to ER-peroxisome proximities
In analogy to dual targeting of Cat2-RFP-PTS, Ptc5-RFP-PTS and Pxp2-RFP-PTS1, sorting of dually targeted membrane proteins could occur at organelle contact sites and may contribute to their formation or stabilization. Many peroxisomal membrane proteins localize to the ER or mitochondria in peroxisome deficient strains lacking the chaperone Pex19 or the membrane protein Pex3 [37,38]. One such protein is the TA protein Pex15 (Pex26 in mammals) which can be sorted via the ER as well as directly imported into peroxisomes via Pex19 [39][40][41][42][43][44]. Pex15 can be targeted and translocated into the ER via the GET complex [39]. We therefore expressed To test whether it is the ER transit of Pex15 that is required for maintaining regular peroxisomes, we overexpressed GFP-Pex15g via a galactose inducible promoter and observed suppression of the ∆get2 and ∆get3 phenotypes ( Fig. 6E; Fig. S7) suggesting that mistargeting of Pex15 contributes to the generation of aberrant peroxisomes in mutants lacking a functional GET complex. Prolonged overexpression of PA-GFP-Pex15g resulted in peroxisomes attached to the ER (Fig. 6F) resembling peroxisomes attached to mitochondria by overexpression of dual affinity PTS1 proteins (Fig. 1). Thus, trafficking of a distinct dual affinity protein may happen at an organelle contact site and promote this connection.

Synthetic genetics interactions reveal redundant modules for tethering via dual targeting at mitochondria and the ER
If tethering of peroxisomes to the ER can be supported by proteins sorted via the ER then this is a parallel or redundant way to create ER-peroxisome contacts. Pex30 and orthologous proteins from other fungi are known to be involved in ER-peroxisome tethering, regulate membrane flux to peroxisomes, and act as a hub for peroxisome biogenesis at the ER [18,19 ,45-49]. It was previously shown that there is a negative genetic interaction between ∆get2 and ∆pex30 [50,51]. Similar to ∆get2 and ∆mdm10, ∆pex30 cells are characterized by an elevated number of smaller, highly mobile peroxisomes (Movie2) [45]. We observed substantial cytosolic mistargeting of mCherry-PTS1, reduced cell growth and altered mitochondrial morphology in ∆get2∆pex30 mutants (Figs. 7A and S8). Cells with intact mitochondria showed the lowest number of peroxisomes, however peroxisomes that remained frequently in proximity to mitochondria (Figs 7A and S8). We conclude that the formation of functional peroxisomes requires tethering facilitated by parallel mechanisms -Pex30 on the one hand and protein import via the GET complex on the other. Mitochondrial dysfunction in the double mutant may be caused by mistargeting of proteins designated for other organelles including peroxisomes.
Further synthetic genetic interactions were detected in comprehensive interaction maps [50,51] e.g. ∆mdm10∆pex30 (positive genetic or buffering) and ∆mdm12∆get2 (negative genetic or synthetic lethal) (Fig. 7B). The positive genetic interaction between mdm10 and pex30 hints to some level of epistasis. Indeed, in a ∆mdm10∆pex30 strain both the mitochondrial morphology phenotype and the growth defect of ∆mdm10 single mutants were partially suppressed showing that absence of Pex30 counteracts absence of Mdm10 ( Fig. 7C and 7D). Glucose medium improved the growth of the ∆mdm10∆get2 double mutant compared to the ∆mdm10 single mutant which speaks for separate functions of Mdm10 and Mdm12 beyond their common role in ERMES complex formation (Fig. 7B -7D). Recently, entry of mitochondrial cargo into the ER via the GET complex was described [52]. Elevated ER targeting of mitochondrial proteins could contribute to the growth defect of ∆mdm10 single mutants on glucose-containing medium (Fig. 7D). In addition to enhanced cytosolic localization of mCherry-PTS1, ∆get2∆mdm10 cells contained an increased number of small peroxisomes compared to each of the single mutants ( Fig. 7E and 7F). These data are consistent with the hypothesis that ERMES and GET complex exhibit a similar role -both are involved in regulating peroxisome growth and division by sorting of dual affinity cargo.

Discussion
We have found that sorting of proteins with targeting signals for two distinct membranes involves sites of interorganellar contact and can concomitantly support their stabilization. This conclusion derives from an analysis of the phenotypic similarity of ∆get2, ∆mdm10, ∆pex5 and ∆pex30 strains with respect to peroxisome number and appearance (Fig. 8A), the phenotypes of several double mutants and from the characterization of different types of dual affinity proteins (Fig. 8B).  [53,54], they seem to be ideally suited to support organelle interactions by binding to different targeting factors and membrane bound translocation machineries or by trafficking from one organelle to the other. Dual affinity proteins may even define regions of reduced identity, which are likely to coincide with organelle contacts.
Localization of peroxisomal membrane proteins to mitochondria has been observed in mutants deficient in Pex19, components of the GET complex or the AAA-ATPase Msp1 [38,39,55].
Mitochondrial targeting of overexpressed Pex15 is observed upon deletion of components of the GET complex but not in wildtype cells [39], thus sorting via the ER probably represents a major avenue of transit. Other previous observations are consistent with our concept. Increased levels of Ant1, Fis1 and Pex11 enhanced PerMit contacts [16]. Ant1 and Pex11 accumulate in mitochondria of mutants lacking components required for peroxisomal membrane protein import [15,17]. Another intriguing observation is that while many peroxisomal membrane proteins can target to peroxisomes without transitioning through the ER [56], several peroxisomal membrane proteins have evolved to be synthesized in vicinity to the ER and translocate from it [57]. Hence a potential reason for such rerouting may be for tethering purposes. One example is the integral membrane protein Pex3 that can either transit through the ER or directly insert into peroxisomes [58,59]. Artificial targeting of Pex3 to mitochondria is compatible with peroxisome biogenesis in S. cerevisiae and provokes tethering of peroxisomes to this organelle [60]. Protein trafficking via a vesicular carrier [40,[61][62][63][64][65][66] or direct membrane flow [24,25,[67][68][69][70] accompanied by protein transit at contact sites may be parallel processes and may both depend on dual affinity proteins.
The decision to which organelle a peroxisome will preferentially attach can be influenced by the metabolic state of the cell [16]. In oleic acid medium proteins such as Tes1 and Cat2 are highly induced which may account for the observation of more frequent connections to mitochondria [16,71].
Is this mode of targeting and tethering only true for peroxisomal proteins? The recently identified tether for mitochondria and the nuclear envelope Cnm1 shows hallmarks of a dual affinity protein comprising a targeting signal for the ER and for mitochondria at the N-and Ctermini, respectively [72]. The discovery of an ER surface retrieval pathway for mitochondrial membrane proteins [73] exemplifies a related mechanism. The chaperone Djp1 critical in this pathway could function similar to Pex5 or Pex19 -the latter has unexpected role in sorting of mitochondrial TA proteins, as well [74]. Of interest, we detected ∆djp1 among the mutants that decrease import of Ptc5-RFP-PTS to peroxisomes ( Fig. 3B and 3C). It was shown previously that overexpression of Djp1 increases PerMit contacts [16] and that loss of Djp1 affects peroxisomal protein import [75]. The GET complex mediates import of peroxisomal and mitochondrial proteins into the ER [39,52]. Several studies uncovered an overlap of factors for peroxisome and mitochondrial biogenesis in mammalian cells [76][77][78]. Accordingly, shared machinery controls protein targeting, biogenesis and division of both organelles and may contribute to tethering as well.
Upon discovery of ERMES as a tether structure, a link between protein transport, membrane flux and tethering mediated by the mitochondrial import protein Mdm10 was suggested [32].
Coupling of protein and membrane trafficking is a common principle in the secretory pathway [79] and it might also occur for peroxisomes at different contact sites. Simple predictions are difficult to make for redundant pathways based not on a single targeting factor but on the sum of interactions created by dual targeting of many proteins each with a unique affinity to the different targeting factors. In respect to the peroxisome, the functional redundancy of potential membrane sources and contact sites is an obstacle in classifying mutants defective in peroxisome biogenesis. A unifying hypothesis is complicated by the fact that several proteins can be directly imported into peroxisomes or traffic via a different organelle. We conclude that the dynamic interplay of membranes at interorganellar contact sites is influenced by targeting of proteins with dual localization signals.

Acknowledgements
We thank Marisa Piscator for technical assistance in the Bölker laboratory. We recognize Bob Lesch for organizing data and Claudia Morales for support in the Schekman laboratory. We

Yeast strains, plasmids, and oligonucleotides
All S. cerevisiae strains used in this study have the genetic background of BY4741 or BY4742 [80] and are listed in supplementary table 4. Yeast cells were transformed as previously described using Li-acetate, PEG and ssDNA [81]. Gene replacement or endogenous tagging was carried out using PCR-amplification of cassettes based on an established toolbox system.

Yeast media and growth conditions
Strains were grown in either YPD (1% yeast extract, 2% peptone, 2% glucose) medium or synthetic medium containing 2% glucose, 0.67% yeast nitrogen base without amino acid, 0.5% ammonium sulfate and complementary supplement mixtures (CSM) or drop-out mixtures lacking appropriate amino acids when experiments required selection. For galactose induction of RFP-Pex15g, cells were grown overnight in synthetic medium containing 2% raffinose, 0.1% glucose, 0.67% yeast nitrogen base without amino acids, 0.5% ammonium sulfate and CSM-His or CSM-His-Leu overnight. Cells were washed once with H2O and diluted in 2% raffinose, 0.67% yeast nitrogen base without amino acids and CSM-His or CSM-His-Leu. After 2h 2% galactose was added and cells were incubated for 3 h. For regulation of Pex5 expression, cells were grown overnight in YP (1% yeast extract, 2% peptone) medium containing 2% raffinose and 0.1% glucose. Cells were washed once with H2O and diluted to an OD600 of 1 in YP containing 2% raffinose. After 2 h, cells were harvested and resuspended in YP medium containing 2% galactose. After 45 min, cells were again harvested and resuspended in YPD medium. Samples were taken at indicated time points and analyzed either by fluorescence microscopy or SDS-PAGE and immunoblot. For depletion of Pex13, cells were precultured in either YPD or synthetic medium containing 2% glucose, 0.67% yeast nitrogen base without amino acids, 0.5% ammonium sulfate and dropout-mix lacking histidine. Cells were diluted in the appropriate medium to an OD600 of 0.5 and incubated at 30°C for 2 h. The culture was then split into two aliquots of equal volume and one culture was supplemented with 2 mM indole-3-acetic acid (auxin). Both cultures were incubated at 30°C and at indicated time points samples were taken and analyzed by either fluorescence microscopy or SDS-PAGE and immunoblot.
For growth assays on plates 5 µl of serial tenfold dilutions of logarithmically growing cells (starting OD600 = 1) were spotted onto solid synthetic media either containing 2% glucose or 2% glycerol and were incubated for 2 -5 days at 30°C. Cells expressing PA-GFP-Pex15g under control of a methionine regulatable promotor (Lam et al., 2010) were grown in the presence of 200 µM methionine or 50 µM methionine to achieve substantial overexpression.

High content screening
To create collections of haploid strains, we constructed a Synthetic Genetic Array (SGA) compatible query strain co-expressing the peroxisomal marker Pex3-GFP (endogenous) together with Ptc5-RFP-PTS1 (overexpression). Using automated mating approaches the query strain was crossed with deletion and DAmP arrayed collections [29,30]. Yeast manipulations in high-density format were performed on a RoToR bench top colony arrayer (Singer Instruments). In short: mating was performed on rich medium plates, selection for diploid cells was performed on SD monosodium glutamate (MSG) plates containing geneticin (Formedium, 500 µg ml −1 ), nourseothricin (WERNER BioAgents "ClonNat", 200 µg ml −1 ) and hygromycin B (Formedium, 500 µg ml −1 ). Sporulation was induced by transferring cells to nitrogen starvation media plates for 7 d. Haploid cells containing the desired mutations were selected by transferring cells to SD MSG plates containing geneticin, neurseothricin and hygromycin B, alongside the toxic amino-acid derivatives canavanine and thialysine (Sigma-Aldrich) to select against remaining diploids, and lacking leucine to select for spores with an alpha mating type.
The collections were visualized using an automated microscopy setup. In brief, cells were Nuclei were removed from homogenates by centrifugation two times at 600 x g. For differential centrifugation, 1 mg PNS fraction was used. PNS (100 µl) fractions were centrifuged at 13k x g for 5 min. The resulting supernatant was centrifuged at 100k x g for 20 min. All fractions were analyzed by SDS-PAGE and immunoblot in amounts representing the initial volume.

Density gradient centrifugation
Post-nuclear supernatants (PNS) for density gradient centrifugation were prepared using a modified protocol from Cramer et al., 2015 (steps 1-16). Cells were precultured in 500 ml YP medium supplemented with 0.1% glucose at 30 °C overnight. Cells were harvested at 6000 × g for 6 min, washed once with 30 ml sterile water, and resuspended in 500 ml YNBO medium. For protein localization analysis, z-projections of deconvolved image stacks of the fluorescent channels were used. Deconvolution was performed on the z-stacks by the ImageJ plugin DeconvolutionLab with 25 iterations of the Richardson-Lucy algorithm [93]. Pearson's correlation coefficients were determined with Volocity 5.3.

Automated time lapse microscopy
Exponentially growing liquid cultures were spotted on 1.5 % agarose pads containing synthetic media. Images were taken with a Zeiss Axio Observer. Z1 microscope equipped with a ×100/1.46 Oil DIC objective and a pco.edge 4.2 sCMOS camera (PCO).
An X-Cite 120PC metal halide light source (EXFO, Canada) and ET-YFP or ET-TexasRed filter cubes (Chroma, USA) were used for fluorescence detection. Images were recorded with VisiView 3.3.0.6 (Visitron Systems) and processed with Metamorph 7.7.5 (Molecular Devices) and Adobe Illustrator CS6 (Adobe Systems). Images were taken every hour and cells were incubated at 30°C.

EndoHF treatment
Protein extract samples (5µl) were subjected to EndoHf (New England Biolabs) treatment according to manufactures instructions. EndoHf (4µl, 4000 units) was used per reaction. To resolve different mobility of glycosylated and non-glycosylated proteins, we used 7.5% Trisglycine midi-gels (Bio-Rad).

Statistics and reproducibility
Microscopic data was collected from 3 independent S. cerevisiae cultures. Five images per culture were quantified. Pearson's correlation coefficient was calculated with Volocity 5.5.2.
Superplots [95] and Student's t-tests were computed using RStudio 1.2.1335 with R 3.6.0. Blots are structured as follows: center line, mean; error bars, standard error of the mean; big circles, mean of experiments; and small circles, data points of experiments. P-values were calculated using an unpaired, two-sided Student's t-test. For calculation of Pearson's correlation coefficients all analyzed images contained ten or more cells and one image represents one data point. For quantification of contacts between mitochondria and peroxisomes five images per strain with at least five cells from 3 independent experiments were analyzed. Quantification was performed by manual inspection of cells. Peroxisomes were counted as mitochondria associated if no black pixels were detected between fluorescent signals of respective marker proteins. Inspection was carried out without knowledge of the genotypes. All experiments were at least repeated three times with similar results.