Pex24 and Pex32 are required to tether peroxisomes to the ER for organelle biogenesis, positioning and segregation in yeast

Peroxins are defined as proteins that play a role in peroxisome biogenesis, including peroxisomal matrix protein import, membrane biogenesis and organelle proliferation (Distel et al., 1996). Most peroxins are peroxisomal or cytosolic proteins that are transiently recruited to the organelle. Recent studies in Saccharomyces cerevisiae showed that a family of peroxins, called the Pex23 protein family (Kiel et al., 2006), localize to the endoplasmic reticulum (ER) (David et al., 2013; Joshi et al., 2016; Mast et al., 2016). The function of these peroxins is still poorly understood and is the subject of this study.

Proteins of the Pex23 family are characterized by a DysF domain. The DysF domain was first identified in human dysferlin. Dysferlin is important for fusion of vesicles with the sarcolemma at the site of muscle injury (Bansal and Campbell, 2004; Lek et al., 2011; Bansal et al., 2003). Human dysferlin contains multiple C2 domains, which play a direct role in the above membrane repair process; however, the function of the DysF domain in dysferlin is still obscure.

Yarrowia lipolytica Pex23 was the first DysF-domain-containing peroxin that was identified (Brown et al., 2000). The number of Pex23 family members varies in different yeast species and their nomenclature is confusing (Fig. 1A). Hansenula polymorpha and Pichia pastoris contain four Pex23 family proteins, but S. cerevisiae has five and Y. lipolytica has only three. Mutants lacking one of these peroxins show diverse peroxisomal phenotypes ranging from a partial peroxisomal matrix protein import defect to enhanced or decreased peroxisome numbers (Brown et al., 2000; Tam and Rachubinski, 2002; Vizeacoumar et al., 2003, 2004; Yan et al., 2008).

Initially, Pex23 family proteins were thought to localize to peroxisomes (Brown et al., 2000; Tam and Rachubinski, 2002; Vizeacoumar et al., 2003, 2004; Yan et al., 2008). However, later studies indicated that S. cerevisiae Pex23 family proteins are ER proteins that form complexes with the ER resident reticulons Rtn1, Rtn2 and Yop1 (David et al., 2013; Mast et al., 2016). S. cerevisiae Pex30 and its paralog Pex31 have been implicated in the formation of peroxisome–ER contact sites, where they regulate de novo peroxisome formation from the ER (David et al., 2013; Joshi et al., 2016; Mast et al., 2016; Wang et al., 2018). S. cerevisiae Inp1 has also been implicated in the formation of peroxisome–ER contacts, but serves a different function, namely in peroxisome retention during yeast budding (Knoblach et al., 2013).

S. cerevisiae Pex30 and Pex31 (ScPex30 and ScPex31) contain a reticulon-like domain and have membrane-shaping properties (Joshi et al., 2016). ER regions where Pex30 accumulates are important for the regulation of pre-peroxisomal vesicle (PPV) formation, but also play a role in lipid droplet biogenesis (Joshi et al., 2018; Wang et al., 2018; Lv et al., 2019).

So far, S. cerevisiae Pex29, Pex30 and Pex31 have been extensively studied. However, our knowledge on other members of the S. cerevisiae Pex23 protein family, as well as on these proteins from other yeast species, is still relatively scarce.

Here, we studied all four Pex23 family members of the yeast H. polymorpha. Our results indicate that these proteins localize to the ER and accumulate at membrane contact sites, including peroxisome–ER contacts and nucleus–vacuole junctions (NVJs). Pex24 and Pex32, but not Pex23 and Pex29, predominantly localize to peroxisome–ER contact sites. Moreover, deletion of PEX24 or PEX32, but not of PEX23 or PEX29, results in major aberrations in peroxisome biology, including defects in peroxisomal matrix protein import and membrane growth, organelle proliferation, positioning and segregation. These defects are accompanied by the disruption of close associations between the peroxisomal and ER membranes, indicating that these proteins are crucial for peroxisome–ER contact site formation. Introduction of an artificial peroxisome–ER tether suppresses the peroxisomal phenotypes, indicating that Pex24 and Pex32 contribute to the tethering of peroxisomes to the ER.

Further studies on the function of Pex32 indicate that the accumulation of this protein at peroxisome–ER contacts is lost in cells lacking the peroxisomal membrane protein Pex11. Moreover, in pex11 cells peroxisome–ER contacts are defective like in pex32 cells. These results are consistent with the view that Pex11 is also important for peroxisome–ER associations. Deletion of PEX32 in pex3 atg1 cells did not result in a change in the abundance or morphology of PPVs, suggesting that Pex32 is not involved in the regulation of PPV formation.

Construction of a phylogenetic tree of Pex23 family members from four different yeast species indicated that two subfamilies (the Pex23 and Pex24 subfamilies) can be distinguished (Fig. 1A). All H. polymorpha members contain a DysF domain at the C terminus. H. polymorpha Pex32 (HpPex32) is much shorter than the other family members, which is mostly due to the lack of an unstructured fragment at the amino terminus of this protein (Fig. 1B).

HpPex23 ends with a Lys-Lys-Lys-Glu stretch of residues, similar to the Lys-Lys-Xaa-Xaa (where Xaa indicates any amino acid) found in S. cerevisiae Pex30 (David et al., 2013), whereas HpPex24 ends with Lys-Lys-Arg. These C termini might represent di-lysine motifs, which are recognized by coatomer subunits and important for retrograde transport to the ER (Ma and Goldberg, 2013). The C termini of HpPex29 and HpPex32 do not contain di-lysine motifs.

Secondary structure prediction indicated that all four sequences contain between two to four transmembrane helices and a C-terminal domain dominated by β-sheets (Fig. 1B). It has been previously argued that a reticulon-like domain was observed in this family of proteins, particularly in ScPex30 and ScPex31 (Joshi et al., 2016). Indeed, a similar domain prediction can be found for HpPex23 using HHpred on the Pfam-A database. This detection extends from residue 100 to 233 of HpPex23. However, this detection has an Expect value (E-value) of 2 and a probability of 92.38, making it a borderline detection. Similar borderline domain predictions are detected for HpPex24, HpPex29 and HpPex32. A Trp residue is also present at the N terminus of this potential domain and aligns with the classically conserved Trp residue of other Pex reticulon-like domains.

The localization of the four H. polymorpha Pex23 family proteins was determined using fluorescence microscopy (FM) using strains producing C-terminally GFP-tagged proteins under control of their endogenous promoters (Fig. 2A,B). Functional peroxisomes are essential for growth of H. polymorpha on methanol. All four strains grew similarly to the wild-type (WT) control on medium containing methanol, indicating that tagging with GFP at the extreme C terminus does not affect the function of Pex23 family proteins in peroxisome biology (Table S1).

Protein localization was assessed using cells that were grown in medium containing glucose (peroxisome-repressing conditions). In these conditions the cells generally contain a single small peroxisome that is associated with the ER (Wu et al., 2018). As shown in Fig. 2B, all four proteins colocalized with the ER marker BiP–mCherry–HDEL, predominantly at the cortical ER. Frequently, a patch of Pex23–GFP was observed at the nuclear envelope as well (Fig. 2A,B). In Pex24–GFP- or Pex32–GFP-producing strains generally one fluorescent spot was detected per cell, which invariably localized close to the single peroxisome marked with Pex14–mKate2. More fluorescent spots were present in cells of Pex23–GFP- and Pex29–GFP-producing strains, and one of them invariably was present in the vicinity of the Pex14–mKate2 spot (Fig. 2A).

Upon overproduction, all four H. polymorpha Pex23 family proteins showed a typical cortical ER and nuclear envelope pattern of localization, supporting that they represent genuine ER proteins. FM analysis revealed that the overproduced proteins were not evenly distributed over the ER, but were present in spots and patches. In all strains, one cortical patch localized in the vicinity of the peroxisome (here marked with DsRed–SKL) (Fig. 2C). Relatively large patches of GFP fluorescence were frequently observed at the nuclear envelope in cells overproducing Pex24–GFP. Colocalization studies with the nucleus–vacuole junction (NVJ) protein Vac8 indicated that these patches represent NVJs (Fig. 2D). Pex23–GFP also accumulated at NVJs when produced under control of its own promoter.

Western blot analysis showed that the levels of all four GFP fusion proteins were very low when produced under control of their endogenous promoters. In fact, Pex24–GFP and Pex32–GFP were below the limit of detection, whereas faint bands were detected on blots of lysates from Pex23–GFP- and Pex29–GFP-producing cells. Upon overproduction, all four GFP-fusion proteins were readily detected (Fig. 2E).

Our data support observations in S. cerevisiae, where Pex23 family proteins localize to the ER, including at regions where peroxisomes and the ER are in close vicinity (David et al., 2013; Mast et al., 2016). The presence of a proportion of HpPex23 and overproduced HpPex24 at NVJs suggests that Pex23 family proteins are also components of other membrane contacts.

To study the role of the Pex23 family proteins we constructed four H. polymorpha deletion strains, pex23, pex24, pex29 and pex32. First, we analyzed whether Pex23 family proteins are important for peroxisomal matrix protein import in glucose-grown cells producing the matrix marker GFP–SKL using widefield fluorescence microscopy (FM). GFP–SKL mislocalized to the cytosol in a proportion of the pex32 cells (Fig. 3A), whereas cytosolic fluorescence was occasionally observed in pex23 and pex24 cells, but never in pex29 cells. In pex32 cultures, typically three types of cells could be discriminated, namely: (1) cells with a GFP spot without cytosolic fluorescence, (2) cells with a GFP spot in conjunction with cytosolic GFP and (3) cells with only cytosolic GFP. This indicates that although matrix protein import is strongly compromised in some of the cells, Pex32 is not essential for the assembly of a functional importomer.

Next, we quantified the number of GFP containing spots using confocal laser scanning microscopy (CLSM) and a custom-made plugin for ImageJ (Thomas et al., 2015). In these images, cytosolic fluorescence was not detected in any of the strains due to the lower sensitivity of CLSM relative to widefield FM. The average number of GFP spots per cell was similar in WT and pex29 cells but reduced in the other three deletion strains. The strongest reduction was observed in pex24 and pex32 cells (Fig. 3B). Frequency distributions show that these reductions are accompanied by an increase in the percentage of cells lacking a GFP spot (Fig. 3B). These results indicate that Pex24 and Pex32 are important for normal peroxisome abundance.

Finally, we analyzed the strains in peroxisome-inducing growth conditions (methanol). Mislocalization of peroxisomal matrix enzymes affects methylotrophic growth (van der Klei et al., 2006). We therefore routinely grow peroxisome-deficient mutants on a mixture of glycerol and methanol (Knoops et al., 2014). At these conditions, cells grow on glycerol (which does not require peroxisome function), while methanol is used as additional carbon and energy source, depending on the severity of the peroxisome function defect. Growth experiments using glycerol-methanol medium revealed that the strongest growth defects were in the pex32 and pex24 strains. Cells of the pex23 strain showed only a minor reduction in growth, whereas pex29 cells grew similar to WT controls (Fig. 3D). These data indicate that the function of peroxisomes is strongly compromised in the absence of Pex24 and Pex32, but not in cells lacking Pex29.

Quantification of structures marked with the H. polymorpha peroxisomal membrane marker Pmp47–GFP (Cepińska et al., 2011) indicated that in these growth conditions, relative to WT controls, peroxisome abundance was reduced, especially in pex24 and pex32 cells (Fig. 3C). Moreover, CLSM revealed that pex23, pex24 and pex32 cells frequently contained a peroxisome of enhanced size (Fig. 3C).

In conclusion, pex24 and pex32 cells showed the most severe peroxisomal phenotypes, whereas pex29 cells were similar to WT and pex23 cells had minor peroxisomal defects.

In S. cerevisiae, Pex23 family proteins and Inp1 play a role in the formation of peroxisome–ER contacts (David et al., 2013; Knoblach et al., 2013; Mast et al., 2016). By measuring the distance between peroxisomal and ER membranes in electron micrographs, we analyzed the role of H. polymorpha Pex23 proteins in the formation of peroxisome–ER contacts (Fig. 4). At membrane contact sites, the two membranes are usually separated by a distance smaller than 30 nm. As shown in Fig. 4A, in WT controls the distance between both membranes was generally less than 10 nm for ∼80% of the peroxisomal profiles analyzed, whereas the distance was larger than 30 nm in less than 10% of the profiles analyzed (Fig. 4A). In cells of the pex23 and pex29 strains, only 40% of the organellar profiles had a measured distance of less than 10 nm, and this percentage further dropped for pex24 and pex32 cells to 10–20% (Fig. 4A,B). These changes were not related to a decrease in total cortical ER, which instead slightly increased (Fig. 4D). Deletion of INP1 had no effect on the distance at peroxisome–ER contact sites (Fig. 4C), in line with our recent observation that H. polymorpha Inp1 associates peroxisomes to the plasma membrane (Wu, 2020). Based on these observations, we conclude that Pex24 and Pex32 – and to a lesser extent Pex29 and Pex23, but not Inp1 – play crucial roles in the formation of tight membrane contacts between peroxisome and ER membranes.

In glucose-grown WT cells, the single peroxisome is invariably localized at the cell cortex. FM analysis of the position of peroxisomes demonstrated that peroxisomes remained close to the cell cortex upon deletion of either PEX32 or INP1. However, in a pex32 inp1 double mutant, peroxisomes were more frequently observed in the central part of the cells, indicating that Pex32 and Inp1 together contribute to the cortical association of peroxisomes (Fig. 4E).

In budding WT cells, at least one peroxisome is retained in the mother cells, whereas another one is transported to the nascent bud. Quantification of peroxisomes in mother cells and buds indicated that the organelles normally segregated between mother cells and buds of the pex29 strain, similar to segregation observed in WT controls. Cultures of pex23, pex24 and pex32 cells, however, showed aberrant peroxisome segregation patterns. In pex24 cultures, a large fraction of the budding cells contained peroxisomes solely in the buds, indicative of a defect in retention of peroxisomes in mother cells (Fig. 4F). A similar, but stronger retention defect was observed in inp1 control cells, known to be defective in peroxisome retention (Fig. 4F).

Our data show that close associations between peroxisomes and the ER require Pex23 family proteins, of which Pex24 and Pex32 are paramount. Inp1 is not crucial for the formation of these associations. Our data furthermore show that these associations contribute to peroxisome positioning at the cell cortex and proper peroxisome segregation in budding cells.

To study whether the effect of the absence of Pex24 and Pex32 on peroxisome biology is due to the loss of peroxisome–ER contacts, we introduced an artificial tether in an attempt to reassociate both organelles. This approach is based on studies in S. cerevisiae, in which the absence of proteins of the ER–mitochondria encounter structure (ERMES) is partially complemented by artificially anchoring mitochondria to the ER (Kornmann et al., 2009). To this end we constructed an artificial tether protein consisting of full-length Pex14 and the tail anchor of the ER protein Ubc6, separated by two heme-agglutinin tags (HA). This construct (P_ADH1Pex14–HA–HA–Ubc6^TA), termed ERPER, was introduced in WT and the four deletion strains (Fig. 5A). Electron microscopy (EM) showed that introduction of ERPER resulted in regions of close apposition (<10 nm) between the ER/nuclear envelope and the peroxisomal membranes (Fig. 5B,C). Immuno-EM using anti-HA antibodies confirmed the presence of ERPER tether protein at these regions (Fig. 5B). EM also showed that, upon growth on a mixture of methanol and glycerol, multiple peroxisomes were present in all mutant strains producing ERPER, as was also the case in WT controls producing ERPER (Fig. 5C), an observation that was confirmed by FM (Fig. 5D). Furthermore, in pex32 cells producing both ERPER and GFP–SKL, cytosolic fluorescence was not detectable (Fig. 5D), indicating that the matrix protein import defect was suppressed by ERPER. Peroxisome quantification showed that peroxisome numbers in pex24 and pex32 cells containing ERPER were similar to those in WT control cells producing ERPER (Fig. 5E; compare with Fig. 3C).

Introduction of ERPER did not affect the growth of WT, whereas it partially suppressed the growth defects that were observed for the pex24 and pex32 deletion strains on glycerol-methanol medium (Fig. 5F), confirming that the tether restored peroxisome matrix protein import and peroxisome function. As expected, the tether did not affect growth of pex29 cells on methanol (Fig. S1B). Furthermore, the minor growth defect of pex23 was not suppressed by ERPER, suggesting that this defect is not caused by altered peroxisome–ER contacts (Fig. S1C).

Introduction of a control construct containing P_ADH1Pex14, which does not cause tethering of peroxisomes to the ER, did not alter peroxisome biogenesis or function in WT cells (Fig. S1A). Only introduction of ERPER (P_ADH1Pex14–HA–HA–Ubc6^TA), and not overexpression of Pex14 under the same promoter (P_ADH1Pex14; Pex14++), suppressed the growth defect of pex32 cells on glycerol-methanol, confirming that artificial tethering and not solely the enhanced Pex14 levels are responsible for suppression of the phenotype (Fig. S1D).

From this we conclude that the severe peroxisome defects in pex24 and pex32 cells are related to a loss in tight peroxisome–ER contacts.

Membrane contacts of cell organelles with the ER have been implicated in lipid transfer. To test whether the Pex24- and Pex32-dependent peroxisome–ER contacts are important for expansion of peroxisomal membranes, we compared the average peroxisomal membrane surface per cell in the four deletion strains relative to the WT control. A plug-in for the analysis of CLSM images allows quantification of the average diameter of peroxisomes by fitting spheres in data obtained from the green channel of combined z-slices of glycerol-methanol grown, Pmp47–GFP-producing cells (Thomas et al., 2015). From these data we estimated the average peroxisomal membrane surface per cell. As shown in Fig. 5G, these values were reduced in pex24 and pex32 cells relative to values for pex23, pex29 and WT cells. Because we are aware of the drawbacks of analyzing organelle sizes by FM (the limited resolution of FM may cause an overestimation of the diameter of very small organelles that are more abundant in WT cells), we also quantified the average length of peroxisomal membranes in cell sections using EM (Fig. 5H). This analysis confirmed that in pex32 cells especially, but also in pex24 cells, the peroxisomal membrane surface is reduced.

Similar analyses of the pex24 and pex32 strains containing ERPER showed that the average peroxisome membrane surface area per cell was increased (Fig. 5G,H), suggesting that Pex24- and Pex32-dependent contacts might contribute to lipid supply and hence peroxisomal membrane expansion.

Because pex32 cells showed the strongest peroxisome phenotype, we confined our further studies to Pex32. First, we analyzed whether the phenotype of pex32 cells could be suppressed by overproduction of any of the other members of the Pex23 protein family. To this purpose, the corresponding genes were placed under control of the strong amine-inducible amine oxidase promoter (P_AMO). Quantitative analysis of FM images of glucose-methylamine-grown cells indicated that upon overexpression of PEX32 in pex32 cells peroxisome numbers increased. This was not the case upon overexpression of PEX23, PEX24 or PEX29 (Fig. 6A). Similarly, overexpression of PEX32, but not of PEX23, PEX24 or PEX29, almost completely restored the growth defect of pex32 cells on methanol (Fig. 6B). These data show that PEX23, PEX24 and PEX29 are not functionally redundant with PEX32.

Next, we performed correlative light and electron microscopy (CLEM) to analyze Pex32–GFP localization at high resolution. In order to obtain sufficient fluorescence signal, Pex32–GFP was slightly overexpressed by placing the gene under control of the P_AMO promoter and inducing expression for a short period. Under these conditions, generally only a single fluorescent spot was detected per cell. EM analysis revealed that the fluorescent spot characteristically localized at the region where the ER and peroxisomal membrane were closely associated (Fig. 6C). In total, four tomograms were analysed, and in all of them the Pex32–GFP-dependent fluorescent spot was present at the peroxisome–ER contact site.

Next, we examined whether the peroxisome–ER association is required for concentrating Pex32. To address this, we localized Pex32–GFP in a pex3 atg1 double deletion strain, which lacks normal peroxisomes but contains PPVs (Knoops et al., 2014). In these cells Pex32–GFP accumulation in a spot was lost. Instead, multiple fainter Pex32–GFP spots were observed (Fig. 6D). In a pex5 atg1 control strain generally one or a few Pex32–GFP spots were present, as observed in WT cells. In pex5 atg1 cells, small peroxisomes occur that are defective in PTS1 protein import but that harbor the complete set of peroxisomal membrane proteins (PMPs). Because PPVs in pex3 atg1 cells and peroxisomes in pex5 atg1 cells differ in PMP composition, we argued that those PMPs that are absent in PPVs might contribute to the accumulation of Pex32–GFP in spots. One of these PMPs is Pex11 (Knoops et al., 2014). We therefore also investigated the formation of Pex32–GFP spots in cells lacking Pex11. FM indicated that in pex11 cells, but not in pex25 controls, the bright Pex32–GFP spots were lost (Fig. 6D). Pex25 is also a PMP and belongs to the same protein family as Pex11. Western blot analysis showed that Pex32–GFP levels in these mutants are similar to those in WT controls, indicating that the absence of the clear Pex32–GFP spots was not due to reduced protein levels (Fig. S2). These data suggest that Pex11, but not Pex25, is specifically required for the accumulation of Pex32–GFP at peroxisome–ER contact sites.

H. polymorpha pex11 cells have several features in common with pex32 cells. These cells show reduced growth on methanol, contain fewer, but larger, peroxisomes and show a peroxisome segregation defect (Krikken et al., 2009). This led us to examine whether pex11 cells are also defective in peroxisome–ER contacts. Indeed, EM analysis showed that the distance between ER and peroxisomal membranes increased in pex11 cells, as was observed for pex32 cells (Fig. 6E). These data indicate that ER-localized Pex32 together with peroxisomal Pex11 contribute to the formation of peroxisome–ER contacts.

S. cerevisiae Pex30 and Pex31 are involved in the regulation of PPV formation. The absence of these proteins was reported to either stimulate (David et al., 2013; Mast et al., 2016) or delay (Joshi et al., 2016; Wang et al., 2018) PPV formation. Possibly, this relates to differences between the assays that were used to monitor PPV formation. Using Pex14–GFP as a marker for PPVs, Joshi and colleagues showed that deletion of PEX30 or PEX31 resulted in a significant decrease in the number of Pex14–GFP spots in S. cerevisiae pex3 atg1 cells (Joshi et al., 2016). A similar analysis in H. polymorpha revealed that deletion of PEX32 in pex3 atg1 cells did not alter the abundance of Pex14–GFP spots (Fig. 7A,B). CLEM analysis revealed that the Pex14–GFP spots in pex3 atg1 pex32 cells represent clusters of small vesicles (Fig. 7C). As shown in Fig. 7D, pex32 pex3 atg1 and pex3 atg1 control cells contain morphologically very similar clusters of vesicles. These data indicate that H. polymorpha Pex32 does not play an important role in the regulation of PPV formation.

Here, we show that all four members of the H. polymorpha Pex23 protein family (Pex23, Pex24, Pex29 and Pex32) localize to the ER. Of these, Pex32 and Pex24 predominantly accumulated at peroxisome–ER contacts and appeared to be very important for multiple peroxisome features. Pex23 is less important for peroxisomes, and we could not detect a peroxisomal phenotype in cells lacking Pex29. Possibly, Pex23 and Pex29 play redundant roles in peroxisome biology or are involved in other functions and hence do not represent true peroxins. Pex23 also accumulated at NVJs, suggesting that Pex23 family proteins might be intrinsic contact site proteins. Initial studies revealed that in H. polymorpha pex23 and pex29 cells, but not in pex24 and pex32 cells, mitochondrial morphology and lipid body abundance is altered, suggesting that these proteins might contribute to the formation of other organelles (F.W., unpublished). Indeed, in S. cerevisiae, ER domains enriched in Pex30 are the sites where most nascent lipid droplets form (Joshi et al., 2018).

Analysis of an evolutionary tree revealed that HpPex23 proteins can be partitioned into two major subgroups, one containing HpPex23 and HpPex32 and the other HpPex24 and HpPex29. There is no clear correlation between subgroup and molecular function, because the strongest peroxisomal phenotypes occurred in the absence of HpPex24 and HpPex32.

The absence of H. polymorpha Pex24 and Pex32 resulted in the loss of peroxisome–ER contacts, accompanied by several peroxisome defects. These phenotypes could be suppressed by an artificial peroxisome ER tether protein, indicating that Pex24 and Pex32 function as contact site tethers. The peroxisomal membrane protein Pex11 also contributes to the formation of these contacts; however, we do not know whether Pex11 contributes directly or indirectly to peroxisome–ER contact formation. Interestingly, previous P. pastoris Pex11 pulldown experiments resulted in the identification of Pex31, a member of the P. pastoris Pex23 protein family (Yan et al., 2008). Moreover, David and colleagues (David et al., 2013) identified ScPex11 as a specific binding partner in ScPex29 complexes, supporting the presence of Pex11 in protein complexes at peroxisome–ER contacts. S. cerevisiae Pex11 is also a component of peroxisome–mitochondrion contact sites, indicating that Pex11 contributes to the formation of different membrane contacts (Mattiazzi Ušaj et al., 2015).

Our data suggest that Pex24 and Pex32 are components of tether complexes that bridge peroxisomes to the ER. However, they do not meet all three criteria suggested for bona fide tethers by Eisenberg-Bord and colleagues (Eisenberg-Bord et al., 2016). These authors proposed that tethers (1) localize or accumulate at the contact site, (2) have the structural capacity to mediate binding to two opposing membranes and (3) exert a tethering force, the existence of which may be established by demonstrating rescue using artificial tethers, among other means. Here we show that H. polymorpha Pex24 and Pex32 accumulate at peroxisome–ER contact sites (criterion 1) and that an artificial tether can rescue phenotypes caused by the absence of these proteins (criterion 3). Further studies are required to determine whether Pex24 and Pex32 also meet criterion 2.

The loss of peroxisome–ER contacts causes multiple phenotypes. It is not unprecedented that a contact-site-resident protein is involved in various processes. For instance the vacuolar membrane protein Vac8 functions in NVJs, vacuole fusion and inheritance in S. cerevisiae (Pan and Goldfarb, 1998). Moreover, the mitochondrial outer membrane protein Mdm10 is a component of ERMES and required for membrane protein insertion (Kornmann et al., 2009; Meisinger et al., 2004; Wiedemann and Pfanner, 2017).

A possible function of the Pex24-, Pex32- and Pex11-dependent peroxisome–ER contacts includes transfer of lipids from the ER to peroxisomes. Indeed, we observed reduced peroxisomal membrane surfaces in cells lacking Pex24 or Pex32. Yeast peroxisomes lack lipid biosynthetic enzymes; hence, expansion of the peroxisomal membrane relies on the supply of lipids from other sources. In S. cerevisiae, peroxisomal membrane lipids may originate from multiple sources, including the mitochondrion, the Golgi apparatus, the vacuole and the ER (Flis et al., 2015; Rosenberger et al., 2009). Indeed, evidence for non-vesicular lipid transport between the ER and peroxisomes in yeast has been reported previously (Raychaudhuri and Prinz, 2008).

In glucose-grown H. polymorpha cells the single peroxisome invariably associates with the edge of cortical ER sheets, where the ER is highly curved (Wu et al., 2018). Using CLEM, we showed that Pex32 specifically localizes to these regions. This is consistent with studies in S. cerevisiae that revealed that members of the Pex23 family occur in complexes with the ER-shaping reticulons, Rtn1, Rtn2, and Yop1 (David et al., 2013; Joshi et al., 2016; Mast et al., 2016). ER-shaping proteins have been implicated in lipid exchange between the ER and mitochondria in S. cerevisiae (Voss et al., 2012). Therefore, it is tempting to speculate that highly curved ER regions where H. polymorpha Pex24 and Pex32 localize function in lipid transport. Also, like S. cerevisiae Pex30 and Pex31, HpPex23 family proteins have a reticulon-like domain and thus might have membrane shaping properties (Joshi et al., 2016). Peroxisome–ER contact sites that contribute to phospholipid transport have also recently been identified in mammals. At these sites, the ER proteins VAPA and VAPB interact with the peroxisome membrane proteins ACBD4 and ACBD5 (Costello et al., 2017a,b; Hua et al., 2017).

Another role of peroxisome–ER contacts may be in peroxisome fission. Mitochondrion–ER contacts are important in the selection of fission sites (Friedman et al., 2011). A comparable mechanism might occur for peroxisomes. This is suggested by the presence of enlarged peroxisomes in pex24 and pex32 cells, similar to those observed in pex11 cells, which are known to be defective in peroxisome fission (Williams et al., 2015). A possible alternative explanation for the enlarged peroxisomes in H. polymorpha pex23, pex24 and pex32 cells is a change in membrane lipid composition, which might interfere with peroxisome fission. Although the absence of S. cerevisiae Pex30 changes the ER phospholipid composition (Wang et al., 2018), it is unknown whether this peroxin influences the phospholipid content of the peroxisomal membrane.

The peroxisome–ER contacts described in this study also contribute to peroxisome positioning at the cell cortex and proper segregation of the organelles between mother cells and buds. So far, only yeast Inp1 was implicated in peroxisome retention (Fagarasanu et al., 2005; Krikken et al., 2009). Here we show that HpPex24 contributes to peroxisome retention in mother cells as well. We previously reported that H. polymorpha pex11 cells show a peroxisome retention defect, underscoring the role of Pex11 in the formation of peroxisome–ER contacts (Krikken et al., 2009).

Proteins of the Pex23 family are implicated in the regulation of PPV formation, but are not required for their formation. Using different experimental approaches, the absence of S. cerevisiae Pex30 or Pex31 was shown to stimulate (David et al., 2013; Mast et al., 2016) or delay (Joshi et al., 2016; Wang et al., 2018) PPV formation. We show that in H. polymorpha, deletion of PEX32 in pex3 atg1 cells has no major effect on the abundance or morphology of PPVs, suggesting that H. polymorpha Pex32 does not play an important role in the regulation of PPV formation.

In conclusion, our data indicate that Pex24 and Pex32 contribute to the tethering of peroxisomes to the ER at membrane contact sites. These contacts play multiple functions, including in peroxisome biogenesis, membrane growth, organelle proliferation and segregation.

The H. polymorpha strains used in this study are listed in Table S2. Yeast cells were grown in batch cultures at 37°C on mineral medium (MM) (Van Dijken et al., 1976) supplemented with 0.5% glucose, 0.5% methanol or a mixture of 0.5% methanol and 0.05% glycerol as carbon sources, and 0.25% ammonium sulfate or 0.25% methylamine as nitrogen sources. When required, amino acids were added to the medium to a final concentration of 30 µg/ml. Transformants were selected on YND plates [0.67% yeast nitrogen base without amino acids (YNB; Difco, BD) and 0.5% glucose] or on YPD plates (1% yeast extract, 1% peptone and 1% glucose) containing 2% agar supplemented with 100 µg/ml zeocin (Invitrogen), 300 µg/ml hygromycin B (Invitrogen) or 100 µg/ml nourseothricin (WERNER BioAgents).

The plasmids and primers used in this study are listed in Tables S3 and S4. All plasmid integrations were performed as described previously (Faber et al., 1994). All integrations were confirmed by PCR and all deletions were confirmed by PCR and Southern blotting.

A plasmid encoding Pex23–mGFP was constructed as follows: a PCR fragment encoding the C terminus of PEX23 was obtained using primers Pex23 GFP-fw and Pex23 GFP-rev with H. polymorpha NCYC495 genomic DNA as a template. The obtained PCR fragment was digested with BglII and HindIII, and inserted between the BglII and HindIII sites of plasmid pHIPZ-mGFP fusinator. BsmBI-linearized pHIPZ PEX23-mGFP was transformed into yku80 cells, producing the strain Pex23–mGFP.

The same methods were used to construct Pex24–mGFP, Pex29–mGFP and Pex32–mGFP strains. PCR was performed on WT genomic DNA with primers Pex24 fw and Pex24 rev to amplify the C terminus of PEX24, primers Pex29 fw and Pex29 rev to amplify the C terminus of PEX29, and primers Pex32 fw and Pex32 rev to amplify the C terminus of PEX32. The obtained PCR fragment of PEX24 was digested with BglII and HindIII, the PCR fragment of PEX29 and the PCR fragment of PEX32 were restricted by BamHI and HindIII. These three digested fragments were inserted between the BglII and HindIII sites of the pHIPZ-mGFP fusinator plasmid. BclI-linearized pHIPZ PEX24-mGFP, NruI-linearized pHIPZ PEX29-mGFP and MfeI-linearized pHIPZ PEX32-mGFP were transformed into yku80 cells separately, producing strains Pex24–mGFP, Pex29–mGFP and Pex32–mGFP. MunI-linearized pHIPH PEX14–mKate2 was transformed into Pex23–mGFP, Pex24–mGFP, Pex29–mGFP and Pex32–mGFP cells for colocalization studies.

For the colocalization of Pex23 family proteins with the ER, DraI-linearized pHIPX7 BiP_N30–mCherry–HDEL was integrated into Pex24–mGFP and Pex29–mGFP cells, and StuI-linearized pHIPX7 BiP_N30–mCherry–HDEL was transformed into Pex23–mGFP cells and Pex32–mGFP cells. Plasmid pHIPX7 BiP_N30–mCherry–HDEL was constructed as follows: first, a PCR fragment containing BiP was obtained with primers KN18 and KN19 using WT genomic DNA as templates. The obtained fragment was digested with BamHI and HindIII, inserted between the BamHI and HindIII sites of pBlueScript II, resulting in plasmid pBS-BiP. Then a PCR fragment containing GFP–HDEL was obtained with primers KN14 and KN17 using pANL29 as template, the resulting fragment was digested with SalI and BglII, and then inserted between the SalI and BglII sites of pBS-BiP, resulting in pBS-BiP_N30–GFP–HDEL. Subsequently, pBS-BiP_N30–GFP–HDEL was digested with BamHI/SalI and inserted between the BamHI/SalI sites of pHIPX7 to obtain pHIPX7 BiP_N30–GFP–HDEL. Plasmid pHIPX7 BiP_N30–GFP–HDEL was digested with BamHI/EcoRI and inserted between the BamHI/EcoRI sites of pHIPX4, resulting in pHIPX4 BiP_N30–GFP–HDEL. NotI and SalI were used to digest pHIPX4 BiP_N30–GFP–HDEL and inserted between the NotI and SalI sites of pHIPZ4 DsRed–SKL to obtain plasmid pRSA017. Later, a PCR fragment was obtained using primers BIPmCh1_fw and BIPmCh1_rev on plasmid pMCE02, the resulting fragment was inserted between BglII and SalI sites of pRSA017 to obtain pHIPZ4 BiP_N30–mCherry–HDEL. Finally, a PCR fragment was obtained by primers BIPmCh2_fw and BIPmCh1_rev using plasmid pHIPZ4 BiP_N30–mCherry–HDEL as a template, the resulting fragment was inserted between BglII and SalI sites of pHIPX7 BiP_N30–GFP–HDEL, resulting in pHIPX7 BiP_N30–mCherry–HDEL.

A plasmid encoding Pex24–mGFP behind the inducible promoter amine oxidase (P_AMO) was constructed as follows: a PCR fragment containing PEX24–mGFP was obtained using primers Pex24GFP fw and Pex24GFP rev with Pex24–mGFP genomic DNA as template. This PCR product and pHIPH5 were restricted by SbfI and BamHI and ligated, which resulted in pHIPH5 PEX24–mGFP. PmlI-linearized pHIPH5 PEX24–mGFP was transformed into yku80 or pex32::DsRed–SKL cells resulting in strain P_AMOPex24–mGFP or strain pex32::DsRed–SKL::P_AMOPex24–mGFP. Plasmid pHIPH5 was constructed using NotI- and SphI-digested pHIPZ5, inserted into the NotI and SphI sites of pHIPH4.

The plasmid pHIPH5 PEX29–mGFP and plasmid pHIPH5 PEX32–mGFP were constructed in the same way. Primers Pex29ov-fw and Pex29ov-rev were used to amplify a PCR fragment containing PEX29–mGFP using Pex29–mGFP genomic DNA as the template. Primers Pex32ov-fw and Pex32ov-rev were used to obtain a PCR fragment containing PEX32–mGFP with Pex32–mGFP genomic DNA as the template. PCR products of PEX29–mGFP and PEX32–mGFP were restricted using SbfI and BclI, and inserted between the SbfI and BclI sites of pHIPH5 PEX24–mGFP, respectively, to make plasmid pHIPH5 PEX29–mGFP and pHIPH5 PEX32–mGFP. NarI-linearized pHIPH5 PEX29–mGFP and pHIPH5 PEX32–mGFP were integrated into yku80 or pex32::DsRed–SKL cells separately to overproduce Pex29–mGFP and Pex32–mGFP.

The plasmid of pHIPH5 PEX23–mGFP was constructed in two steps. First, a PCR fragment containing partial (no start codon) PEX23–mGFP was obtained using primers Pex23ov-fw and Pex23ov-rev with Pex23–mGFP genomic DNA as a template. The PCR product and pHIPH5 PEX24–mGFP were restricted using SbfI and BamHI and ligated to produce pHIPH5 PEX23p–mGFP. Next, a PCR using primers Pex23ov2-fw and Pex23ov2-rev was performed to obtain the left partial (with start codon) PEX23–mGFP fragment using plasmid pHIPH5 PEX24–mGFP as template. The PCR product and pHIPH5 PEX23p–mGFP were restricted using NotI and BamHI, then ligated to produce pHIPH5 PEX23–mGFP. NarI-linearized pHIPH5 PEX23–mGFP was transformed into yku80 or pex32::DsRed–SKL cells to overproduce Pex23–mGFP.

EcoRI-linearized pHIPN18 DsRed–SKL was integrated into yku80, P_AMOPex23–mGFP, P_AMOPex24–mGFP, P_AMOPex29–mGFP and P_AMOPex32–mGFP cells. A plasmid encoding pHIPN18 DsRed–SKL was constructed as follows: a vector fragment was obtained by HindIII and SalI digestion of pHIPN18 GFP–SKL, whereas the DsRed–SKL insertion fragment was obtained by HindIII and SalI digestion of pHIPZ4 DsRed–SKL; ligation resulted in the plasmid pHIPN18 DsRed–SKL. Plasmid pHIPN18 GFP–SKL was constructed by inserting NotI- and XbaI-digested pAMK94 into the NotI and XbaI sites of pHIPN4. Plasmid pAMK94 was constructed as follows: a PCR fragment containing ADH1 was amplified using primers ADH1 fw and ADH1 rev with WT genomic DNA as template. NotI- and HindIII-digested PCR product was then inserted into NotI and HindIII sites of pHIPZ4 eGFP–SKL.

MunI-linearized pHIPN VAC8–mKate2 was integrated into Pex23–mGFP and P_AMOPex24–GFP cells to produce Vac8–mKate2. Plasmid pHIPN VAC8–mKate2 was constructed by fragment ligation from HindIII/SalI digested plasmid pHIPZ VAC8–mKate2 and HindIII/SalI digested plasmid pHIPN PEX14–mCherry. Plasmid pHIPZ VAC8–GFP and plasmid pHIPZ PEX14–mKate2 were digested with HindIII and BglII and ligated to obtain plasmid pHIPZ VAC8–mKate2. Plasmid pHIPZ VAC8–GFP was constructed by amplification of the VAC8 gene, lacking the stop codon, using primers Vac8_BglII R and Vac8_F and genomic DNA as template. The resulting PCR product was digested with HindIII and BglII, and ligated between the HindIII and BglII sites of the pHIPZ–mGFP fusinator plasmid.

The pex23 deletion strain was constructed by replacing the PEX23 region with the zeocin resistance gene as follows: first, a PCR fragment containing the zeocin resistance gene and 50 bp of the PEX23 flanking regions was amplified with primers PEX23-Fw and PEX23-Rev using plasmid pENTR221-zeocin as template. The resulting PEX23 deletion cassette was transformed into yku80 cells to obtain strain pex23. PEX24, PEX29 and PEX32 were also replaced by the zeocin resistance gene in the same way. Primers for the PEX24 deletion cassette were PEX24-Fw and PEX24-Rev, primers for the PEX29 deletion cassette were dPEX29-F and dPEX29-R, and primers for the PEX32 deletion cassette were dPEX32-F and dPEX32-R. These three deletion cassettes were transformed into yku80 cells, producing pex24, pex29 and pex32, respectively.

For expression of GFP–SKL in WT (yku80) and deletion mutant strains, StuI-linearized pHIPN7 GFP–SKL was transformed into pex23 and pex24 mutant cells, and AhdI-linearized pFEM35 was transformed into yku80, pex29 and pex32 mutant cells.

The MunI-linearized pHIPN PMP47–mGFP plasmid was transformed into pex23, pex24, pex29 and pex32 cells. Plasmid pHIPN PMP47–mGFP was constructed as follows: a PCR fragment encoding the nourseothricin resistance gene was obtained with primers Nat-fwd and Nat-rev using plasmid pHIPN4 as a template. The obtained PCR fragment was digested with NotI and XhoI and inserted between the NotI and XhoI sites of pMCE7, resulting in plasmid pHIPN PMP47–mGFP.

The DraI-linearized pAMK15 was transformed into pex32 cells to obtain a strain producing DsRed–SKL.

To introduce an artificial peroxisome–ER tether, two plasmids, pARM115 (pHIPH18 PEX14) and pARM118 (pHIPH18 PEX14–2HA–UBC6), were constructed as follows. A PCR fragment containing PEX14 was amplified with primers Pex14-HindIII-fw and Pex14-PspXI-rev using WT genomic DNA as a template. The PCR fragment was digested with HindIII and PspXI, then inserted between the HindIII and SalI sites of pAMK94 to get plasmid pHIPZ18 PEX14. A NotI/BpiI digested fragment from plasmid pHIPZ18 PEX14 and a NotI/BpiI digested fragment from plasmid pHIPH4 were ligated, resulting in plasmid pARM115. The AgeI-linearized pARM115 was transformed into yku80::GFP–SKL and pex32::GFP–SKL cells to produce P_ADH1Pex14 (Pex14++). A PCR fragment containing PEX14–2×HA was amplified by primers HindIII-Pex14 and Pex14-HA-HA. A fragment containing 2×HA–UBC6 was amplified with primers HAHA-Ubc6 and Ubc6-PspXI using WT genomic DNA as template. The obtained PCR fragments were purified and used as templates together with primers HindIII-Pex14 and Ubc6-PspXI in a second PCR reaction. The obtained overlap PCR fragment was digested with HindIII and PspXI, and inserted between the HindIII and SalI sites of pAMK94, resulting in plasmid pARM053 (pHIPZ18 PEX14–2HA–UBC6). A NotI/BpiI digested fragment from plasmid pAMK053 and a NotI/BpiI digested fragment from plasmid pHIPH4 were ligated, resulting in plasmid pARM118. Then the AgeI-linearized pARM118 was transformed into yku80::GFP–SKL, yku80::Pmp47–GFP, pex23::GFP–SKL, pex24::GFP–SKL, pex24::Pmp47–GFP, pex29::GFP–SKL, pex32::GFP–SKL and pex32::Pmp47–GFP cells, to produce P_ADH1Pex14–2HA–Ubc6 (ERPER)-expressing strains.

The BglII-linearized pHIPZ PEX32–mGFP was transformed into pex3 atg1::Pex14–mCherry, pex5 atg1::Pex14–mCherry, pex11 and pex25 cells, to produce Pex32–mGFP-expressing stains. BlpI-linearized pARM014 (pHIPX7 PEX14–mCherry) was transformed into pex5 atg1 cells, which resulted in pex5 atg1::Pex14–mCherry. Plasmid pARM014 was constructed through the following steps: first, a PCR fragment containing Pex14–mCherry was amplified with primers PRARM001 and PRARM002 using pSEM01 as a template. The obtained PCR fragment was digested with NotI and HindIII, and inserted between the NotI and HindIII sites of plasmid pHIPX7, resulting in plasmid pARM014. An ATG1 deletion cassette was amplified by PCR with primers pDEL-ATG1-fwd and pDEL-ATG1-rev using plasmid pARM011 as template. Then the PCR product was integrated into pex5 cells to make the pex5 atg1 mutant.

Two plasmids allowing disruption of H. polymorpha PEX25 were constructed using Multisite Gateway technology, as follows. First, the 5′ and 3′ flanking regions of the PEX25 gene were amplified by PCR with primers RSAPex25-1 and RSAPex25-2, and RSAPex25-3 and RSAPex25-4, respectively, using H. polymorpha NCYC495 genomic DNA as a template. The resulting fragments were then recombined in donor vectors pDONR P4-P1R and pDONR P2R-P3, resulting in plasmids pENTR-PEX25 5′ and pENTR-PEX25 3′, respectively. Then, PCR amplification was performed using primers attB1-Ptef1-forward and attB2-Ttef1-reverse using pHIPN4 as the template. The resulting PCR fragment was recombined into vector pDONR-221, yielding entry vector pENTR-221-NAT. Recombination of the entry vectors pENTR-PEX25 5′, pENTR-221-NAT, and pENTR-PEX25 3′, and the destination vector pDEST-R4-R3, resulted in pRSA018. Then, a PEX25 disruption cassette containing nourseothricin resistance gene was amplified with primers RSAPex25-5 and RSAPex25-6 using pRSA018 as a template. To create the pex25 mutant, the PEX25 disruption cassette was transformed into yku80 cells. BlpI-linearized pHIPH PEX14–mCherry was integrated into pex11::Pex32–mGFP or pex25::Pex32–mGFP to produce Pex14–mCherry.

To construct the pex32 inp1 mutant, a PCR fragment containing an INP1 deletion cassette was amplified with primers dInp1FW-F and dInp1-REV using plasmid pHIPH5 as a template. The resulting INP1 deletion cassette was transformed into pex32 cells for double deletion of pex32 inp1. The AhdI-linearized pFEM35 was transformed into pex32 inp1 to produce GFP–SKL-expressing cells.

To construct the pex3 atg1 pex32 strain, a PCR fragment containing the PEX32 deletion cassette was amplified with primers dPex32-F and dPex32-R using pex32 genomic DNA as a template. The resulting PEX32 deletion cassette was transformed into pex3 atg1 cells to get a triple mutant of pex3 atg1 pex32. XhoI-linearized pHIPN PEX14–mGFP plasmid was integrated into pex3 atg1 pex32 cells.

The plasmid encoding pHIPN PEX14–mGFP was constructed as follows: a PCR fragment containing the nourseothricin resistance gene was obtained using primers Nat fw and Nat rev with plasmid pHIPN4 as a template. The PCR product and pSNA12 were digested with NsiI and NotI, then ligated to produce pHIPN PEX14–mGFP.

DNA restriction enzymes were used as recommended by the suppliers (Thermo Fisher Scientific or New England Biolabs). PCR for cloning was carried out using Phusion High-Fidelity DNA Polymerase (Thermo Fisher Scientific). An initial selection of positive transformants by colony PCR was carried out using Phire polymerase (Thermo Fisher Scientific). For DNA and amino acid sequence analysis, the Clone Manager 5 program (Scientific and Educational Software, Durham, NC) was used.

For western blot analysis, total cell extracts were prepared as described previously (Baerends et al., 2000). Samples in Fig. 2E were denatured in urea loading buffer (25mM Tris-HCl pH 6.8, 0.8% SDS, 3.5% glycerol, 4 M urea, 2% β-mercaptoethanol and 0.008% Bromophenol Blue). Blots were decorated using anti-GFP antibodies (sc-996, Santa Cruz Biotech; 1:2000 dilution), anti-pyruvate carboxylase-1 (Pyc1) antibodies (Ozimek et al., 2007, 1:10,000 dilution) or anti-Pex14 antibodies (Komori et al., 1997, 1:10,000 dilution). Secondary goat anti-rabbit (31460) or goat anti-mouse (31430) antibodies conjugated to horseradish peroxidase (HRP) (Thermo Scientific; 1:5000 dilution) were used for detection. Blots were scanned by using a densitometer (GS-710; Bio-Rad Laboratories).

Widefield FM images of living cells and of cryosections for CLEM were captured at room temperature using a 100×1.30 NA objective (Carl Zeiss, Oberkochen, Germany). Images were obtained from the cells in growth medium using a fluorescence microscope (Axioscope A1; Carl Zeiss), Micro-Manager 1.4 software and a digital camera (Coolsnap HQ²; Photometrics). GFP fluorescence was visualized using a 470/40 nm band-pass excitation filter, a 495 nm dichromatic mirror, and a 525/50 nm band-pass emission filter. DsRed fluorescence was visualized using a 546/12 nm band-pass excitation filter, a 560 nm dichromatic mirror, and a 575–640 nm band-pass emission filter. mCherry and mKate2 fluorescence were visualized using a 587/25 nm band-pass excitation filter, a 605 nm dichromatic mirror, and a 670/70 nm band-pass emission filter.

Confocal images were captured with an LSM800 Airyscan confocal microscope (Carl Zeiss) using Zen 2.3 software (Carl Zeiss) and a 100×/1.40 plan apochromat objective and GaAsP detectors. For quantitative analysis of peroxisomes or Pex14–mGFP fluorescent spots, z-stacks were made of randomly chosen fields.

Image analysis was performed using ImageJ, all brightfield images have been adjusted to only show cell outlines. Figures were prepared using Adobe Illustrator software.

For morphological analysis, cells were fixed in 1.5% potassium permanganate, post-stained with 0.5% uranyl acetate and embedded in Epon [a mixture of Glycid ether (51.5% w/v; Serva, 151414), Methylnadic anhydride (47.3% w/v; Serva, 140573) and 2,4,6-Tris (dimethylaminomethyl)phenol (1.2% w/v; Santa Cruz, F0112)]. Image analysis and distance measurements are performed using ImageJ. For the quantification of the ER, the total length of the plasma membrane and the peripheral ER was measured from cell sections, and from this the percentage of the cortex covered by the ER was calculated. Correlative light and electron microscopy (CLEM) was performed using cryosections, as described previously (Knoops et al., 2015). After fluorescence imaging, the grid was post-stained and embedded in a mixture of 0.5% uranyl acetate and 0.5% methylcellulose. Acquisition of the double-tilt tomography series was performed manually in a CM12 TEM (Philips) running at 100 kV, and included a tilt range of 40° to −40° with 2.5° increments. To construct the CLEM images, pictures taken with FM and EM were aligned using the eC-CLEM plugin in Icy (Paul-Gilloteaux et al., 2017) (http://icy.bioimageanalysis.org). Reconstruction of the tomograms was performed using the IMOD software package (https://bio3d.colorado.edu/imod/).

Immuno-EM was performed as described previously (Thomas et al., 2018). Labeling of HA was performed using monoclonal antibodies (Sigma-Aldrich H9658; 1:100 dilution) followed by goat anti-mouse antibodies conjugated to 6 nm gold (Aurion, The Netherlands; 1:20 dilution).

Homologous sequences were detected using BLASTP with an E-value threshold of 1e-5 (Altschul et al., 1990). Linear and secondary structure predictions were realized using Foundation (Bordin et al., 2018).

The multiple sequence alignment used as input was created using ClustalOmega (Sievers et al., 2011), using default parameters, and manually curated in Jalview (Waterhouse et al., 2009). The tree was generated using PhyML 3.1 (Guindon et al., 2010) using the LG matrix, 100 bootstraps, tree and leaves refinement, SPR moves, and amino acids substitution rates determined empirically.

For peroxisome membrane surface area calculation: the average peroxisome volume (V) and average peroxisome number per cell (N) were determined using a plugin for ImageJ (Thomas et al., 2015) from two independent experiments (2×300 cells were counted). The formula V=(4/3)πr³ was used to calculate peroxisome radius (r), and formula S=4πr² was used to calculate the average peroxisome surface area (S). The average peroxisome number per cell N multiplied with S is the peroxisome membrane surface area per cell.

For the calculation of the distance between GFP–SKL spots and the cell cortex, cells containing GFP spots were selected and processed using ImageJ. Subsequently, the distance between the middle of the GFP spot and the cell outline was measured. For cells containing two or more GFP spots, only the spot closest to the cell outline was used.

Peroxisome inheritance quantification was performed using a method published previously (Krikken et al., 2009).

We thank Benjan Karnebeek, Malgorzata Krygowska and Kim van Maldegem for assistance in strain construction and Tim Levine (University College London, UK) for advice in protein sequence analysis.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.246983.reviewer-comments.pdf