Dnm1p-dependent peroxisome fission requires Caf4p, Mdv1p and Fis1p

Eukaryotic cells contain a set of functionally distinct membrane-bound compartments called organelles. The maintenance of a full set of organelles is fundamental to eukaryotic life. With every cell division, organelles are duplicated and segregated between daughter cells. Organelles multiply either by fission of existing organelles or by de novo formation.

Peroxisomes are organelles that are found in almost all eukaryotic cells. The number, size and shape of peroxisomes varies between different cell types and different environmental conditions, from small spherical organelles to large elaborate reticular structures. For instance, in Saccharomyces cerevisiae, peroxisomes proliferate in response to growth on oleic acid, a long-chain fatty acid. Recently, we have shown that, in wild-type cells, peroxisomes multiply by fission of pre-existing peroxisomes, but they can also form de novo in cells temporarily devoid of these organelles (Motley and Hettema, 2007).

The family of dynamin-related proteins (DRPs) has been implicated in multiple membrane-remodelling events in eukaryotic cells. Whereas some DRPs are specific to a particular organelle, others seem to be more promiscuous. In yeast, Vps1p was identified as being required for the transport of vacuolar hydrolases from Golgi to endosomes (Rothman et al., 1990; Vater et al., 1992). Later, it was also shown to be involved in vacuole fusion (Luo and Chang, 2000; Peters et al., 2004) and in secretion of a subset of secretory proteins (Harsay and Schekman, 2002), and to be required for normal peroxisome abundance (Hoepfner et al., 2001). Vps1p requires Pex19 for its recruitment to peroxisomes, although the exact function of Pex19p is unclear (Vizeacoumar et al., 2006). A second yeast DRP, Dnm1p, was shown to be required for mitochondrial fission (Bleazard et al., 1999) and to regulate peroxisome abundance to a minor extent (Kuravi et al., 2006).

Genetic and biochemical approaches have identified three additional proteins (Fis1p, Mdv1p and Caf4p) that are required for mitochondrial fission and that cooperate with Dnm1p (for a review, see Hoppins et al., 2007). Fis1p recruits Dnm1p to mitochondrial membranes in concert with Mdv1p and Caf4p (Cerveny and Jensen, 2003; Griffin et al., 2005; Mozdy et al., 2000; Tieu and Nunnari, 2000; Tieu et al., 2002). Both Caf4p and Mdv1p bind Dnm1p, and, in an mdv1Δ/caf4Δ double mutant, much of the Dnm1p is dissociated from mitochondria (Schauss et al., 2006). Mdv1p seems to play a more important role in fission than Caf4p, because mdv1Δ cells show a clear fission defect whereas caf4Δ cells do not (Griffin et al., 2005). Num1p also has a role in recruiting Dnm1p to mitochondria: cells lacking Num1p display a decrease in mitochondrial fission (Cerveny et al., 2007; Schauss and McBride, 2007).

In mammalian cells, a single DRP (dynamin-like protein 1, DLP1) is thought to be required for peroxisome fission (Koch et al., 2003). DLP1 resembles Dnm1p and is also required for mitochondrial fission (Smirnova et al., 2001). A patient with a mutation in DLP1 has recently been described, and this mutation results in a lethal disorder in which fission both of peroxisomes and of mitochondria is impaired (Waterham et al., 2007).

Human FIS1 is required for normal mitochondrial and peroxisomal morphology. Both DLP1 and human FIS1 partially localise to peroxisomes (Koch et al., 2005; Li and Gould, 2003). Recently, it was also shown in yeast cells that Fis1p and Dnm1p partially localise to peroxisomes (Kuravi et al., 2006). Mutants lacking Dnm1p or Fis1p are not affected in peroxisome abundance when grown on a fermentable carbon source. However, when shifted to oleic acid as the carbon source, the number of peroxisomes failed to increase as much as was observed in wild-type cells (Kuravi et al., 2006). Additionally, a vps1Δ/dnm1Δ double mutant exhibited a more severe decrease in peroxisome number than the vps1Δ single mutant. These results suggest that Vps1p and Dnm1p are partially redundant in the regulation of peroxisome abundance. Using an assay designed to study peroxisome fission in vivo, we showed that the decreased abundance in a vps1Δ/dnm1Δ mutant results from a strong reduction in peroxisome fission (Motley and Hettema, 2007).

Here, we describe our analysis of the role of Dnm1p in controlling peroxisome abundance in cells grown on a fermentable carbon source. Under these conditions, a constant peroxisomal number was maintained with no requirement for peroxisome proliferation. We show that Dnm1p-dependent fission of peroxisomes depends on Fis1p and that, similarly to mitochondrial fission, Caf4p and Mdv1p are required for Dnm1p-dependent peroxisome fission, although, unlike for mitochondria, their roles in peroxisome fission are redundant. Furthermore, we show that Vps1p-dependent fission is not dependent on the presence of Fis1p, Caf4p or Mdv1p. We are able to suppress the fission defect in vps1Δ cells by (1) overexpression of Dnm1p and (2) redirecting Fis1p from mitochondria to peroxisomes. Dnm1p contributes little to peroxisome fission during fermentative growth, but its contribution appears far greater under conditions of mitochondrial dysfunction. We show that, although Dnm1p and Vps1p are partially redundant for peroxisome fission, they also exhibit specific and non-overlapping functions: overexpression of Vps1p does not rescue the mitochondrial fission defect in a dnm1Δ mutant, and overexpression of Dnm1p does not rescue the vacuolar fusion defect in vps1Δ cells.

We conclude that fission of peroxisomes in yeast is controlled by two different dynamin-related proteins that are recruited to peroxisomes independently of each other by distinct proteins. Although these DRPs display redundancy for their role in peroxisome fission, at least some of their other roles within the cell are non-overlapping.

With every cell division, peroxisomes duplicate by fission. We have shown previously that the dynamin-related protein Vps1p is required for this process (Motley and Hettema, 2007). Recently, it was shown that peroxisome abundance in a vps1Δ/dnm1Δ double mutant is even lower than in a vps1Δ single mutant, suggesting that Dnm1p also plays a role in peroxisome fission (Kuravi et al., 2006). This was further supported by the observation that some colocalisation of Dnm1p and Fis1p was found with peroxisomes (Kuravi et al., 2006).

We analysed yeast strains lacking one of the genes encoding mitochondrial fission components. We transformed these strains with a well-established fluorescent marker appended with a type I peroxisomal targeting signal (GFP-PTS1), grew them to log phase on glucose medium as described in the Materials and Methods, and counted peroxisomes in dnm1Δ, fis1Δ, caf4Δ, mdv1Δ and num1Δ cells (Fig. 1). No clear differences in peroxisome number were observed between these mutants and wild-type cells, in agreement with previous findings (Kuravi et al., 2006). Because the contribution of Dnm1p to peroxisomal fission only becomes clear when Vps1-dependent fission is blocked, we introduced a vps1Δ mutation into the mitochondrial fission mutants. vps1Δ/fis1Δ double mutants contained a single peroxisomal structure in the majority of dividing cells (Fig. 1). Neither vps1Δ/caf4Δ nor vps1Δ/mdv1Δ cells displayed a peroxisome-abundance phenotype that was more severe than that of vps1Δ alone. However, the triple mutant vps1Δ/caf4Δ/mdv1Δ showed a phenotype that was indistinguishable from that of vps1Δ/dnm1Δ or vps1Δ/fis1Δ cells.

We conclude that the machinery used for mitochondrial fission is used also for peroxisomal fission. Our results suggest that Dnm1p is recruited to peroxisomes via Fis1p, Mdv1p and Caf4p, whereby Caf4p and Mdv1p are redundant. Furthermore, our results show that Vps1p operates independently of Fis1p, Caf4p, Mdv1p and Dnm1p.

vps1Δ/num1Δ cells did not show a reduction of peroxisome abundance compared to vps1Δ cells, suggesting that Num1p is not required for Dnm1p-dependent peroxisome function. Num1p was not further studied.

Our results suggest that two independent systems operate in peroxisome fission, one relying on Dnm1p and a second relying on Vps1p. Vps1p plays the major role in peroxisome fission: vps1Δ cells have a clear phenotype whereas dnm1Δ cells do not. However, when Dnm1p is overexpressed in a vps1Δ mutant, peroxisome number normalises (Fig. 2). This shows that: (1) Dnm1p can substitute for Vps1p in peroxisome fission and (2) that Dnm1p is limiting in Dnm1p-dependent (peroxisomal) membrane fission. By contrast, Dnm1p overexpression does not restore peroxisome abundance to normal levels in vps1Δ/fis1Δ or vps1Δ/caf4Δ/mdv1Δ mutants. We conclude that peroxisome abundance depends on two redundant machineries, of which the Vps1p-containing machinery is the major contributor.

In the experiments described above, we analysed peroxisome abundance only. We subsequently tested whether Dnm1p-mediated peroxisome fission is affected in cells lacking Fis1p, or Caf4p and Mdv1p by using an assay we developed recently (Motley and Hettema, 2007). vps1Δ/dnm1Δ cells were pulse-labelled with GFP-PTS1 (see Materials and Methods), and mated with vps1Δ/dnm1Δ cells pulse-labelled with HcRed-PTS1 and overexpressing Dnm1p. In the latter strain, the red peroxisomes were small and abundant, whereas, in the former strain, the GFP-labelled peroxisome was a single elongated structure. After mating and subsequent cytoplasmic mixing, Dnm1p diffused from one mating partner to the other, and the pre-labelled green peroxisome was divided almost instantly into small peroxisomal structures (Fig. 3A×B mating). In these mating cells, red peroxisomes coexisted with green peroxisomes. However, when vps1Δ/dnm1Δ cells overexpressing Dnm1p were mated with vps1Δ/fis1Δ or vps1Δ/caf4Δ/mdv1Δ mutants, there was a delay before fission of the pre-labelled peroxisome began to occur: initially, the red peroxisomes from the vps1Δ/dnm1Δ mating partner coexisted with green tubulated peroxisome from the vps1Δ/fis1Δ (A×D) or vps1Δ/caf4Δ/mdv1Δ (A×E) mating partner. That this fission of peroxisomes is carried out by Dnm1p is clear from the mating of vps1Δ/dnm1Δ with vps1Δ/dnm1Δ cells in the absence (B×C) or presence (A×B) of overexpressed Dnm1p. We conclude that Fis1p, Caf4p and Mdv1p are required for Dnm1p-dependent peroxisome fission.

A large fraction of Dnm1p is associated with mitochondria in a Fis1p-dependent manner (Cerveny and Jensen, 2003; Mozdy et al., 2000; Tieu et al., 2002). Rescue of the vps1Δ phenotype by overexpression of Dnm1p suggests that Dnm1p is limiting for peroxisome fission. We argued that, if the Dnm1p that was usually associated with mitochondria could be redirected to peroxisomes, then endogenous levels of Dnm1p might be sufficient to overcome the peroxisomal fission defect in vps1Δ cells. To test this hypothesis, we made use of the observation that exchanging the C-terminal membrane-anchor sequence of Fis1p with that of the peroxisomal membrane protein Pex15p results in an exclusive localisation of the resulting fusion protein to peroxisomes (Halbach et al., 2006). We expressed this fusion protein under control of the FIS1 promoter in vps1Δ/fis1Δ cells and analysed peroxisome and mitochondrial morphology. Whereas expression of Fis1p restored peroxisome abundance to that observed in vps1Δ cells, when expressing the Fis1p-Pex15p fusion protein, peroxisome abundance was restored to wild-type levels. This indicates an increased level of Dnm1-dependent peroxisome fission. By contrast, expression of the fusion protein did not restore the mitochondrial fission defect, whereas expression of Fis1p restored mitochondrial fission to wild-type levels (Fig. 4). These observations show that peroxisomes and mitochondria compete for Dnm1p, and that Fis1p plays a pivotal role in distributing Dnm1p between peroxisomes and mitochondria.

Similarly to the recruitment of Mdv1p and Caf4p to mitochondria via Fis1p, our results suggest that Fis1p also recruits these proteins to peroxisomal membranes. To test this, we analysed GFP fusions of both Caf4p and Mdv1p for their localisation in vps1Δ/caf4Δ/mdv1Δ cells (Fig. 5). Peroxisome number was restored in the presence of these fusion proteins, indicating that the GFP fusions were functional. These cells gave clearer localisations of GFP-Mdv1p and GFP-Caf4p than wild-type cells, presumably because of lack of endogenous protein competing for localisation. Double labelling of GFP-Mdv1p or GFP-Caf4p with HcRed-PTS1 in vps1Δ/mdv1Δ/caf4Δ cells showed a limited amount of colocalisation. As expected, most GFP punctae were distinct from peroxisomes and most likely represent the mitochondrion-associated pool of these proteins. Peroxisomal labelling was hard to detect in flattened z-stacks but was easier to detect in single focal planes (Fig. 5A,D). GFP-Mdv1p or GFP-Caf4p were not detected on 100% of peroxisomes.

We have shown above that expression of Fis1p-Pex15p increases Dnm1p-dependent peroxisome fission in an Mdv1p- and Caf4p-dependent manner. Therefore, we expected GFP-Caf4p and -Mdv1p levels on peroxisomes to be increased in cells expressing Fis1p-Pex15p. To test this, we performed a mating assay: one mating partner comprised vps1Δ/fis1Δ cells expressing Fis1p-Pex15p in combination with either GFP-Caf4p or GFP-Mdv1p; the other mating partner was a mutant lacking peroxisomes (pex3Δ) and expressing HcRed-PTS1. Upon cell fusion and cytoplasmic mixing, HcRed-PTS1 was imported into the Fis1p-Pex15p-containing peroxisomes of the mating partner (Fig. 5B,E). Now, a more pronounced colocalisation of GFP-Mdv1p and GFP-Caf4p with peroxisomes was observed (compare Fig. 5B with A and E with D). However, both GFP fusion proteins were, for the most part, cytosolic when expressed in fis1Δ cells, showing their dependence on Fis1p for membrane association (Fig. 5C,F). We conclude that Mdv1p and Caf4p associate with peroxisomes, and that this association depends on Fis1p.

Peroxisomes have been shown to multiply in response to mitochondrial dysfunction (Butow and Avadhani, 2004). In a genome-wide screen for peroxisome-morphology mutants, we observed many mutants in mitochondrial genes that display increased peroxisomal abundance, including mutants in mitochondrial ribosomal genes, i.e. MRP5, MRPL8, MRPL49 (G.P.W. and E.E.H., unpublished data) and the nuclear-encoded F1F0-ATP synthase subunits (Fig. 6). We tested whether the increased peroxisome abundance in the atp7Δ and atp17Δ cells is dependent on DRPs. First, we disrupted the ATP7 or ATP17 gene in vps1Δ cells, which normally have a reduced number of peroxisomes. Strikingly, vps1Δ/atp7Δ cells and vps1Δ/atp17Δ cells displayed an increased peroxisome abundance compared with both vps1Δ cells and wild-type cells, implying that Vps1p is not necessary for this proliferation. By contrast, in vps1Δ/dnm1Δ/atp7Δ cells and vps1Δ/dnm1Δ/atp17Δ cells, a single peroxisomal structure was observed. These results indicate that DRPs are required for peroxisome proliferation during mitochondrial dysfunction, with Dnm1p having a major contribution. To test whether Dnm1p is solely responsible for this proliferation or whether, in its absence, Vps1p can take over, we constructed a dnm1Δ/atp7Δ strain. This strain also displays an increased peroxisome abundance. We conclude that both Dnm1p and Vps1p are involved in peroxisome proliferation in response to mitochondrial dysfunction. Whereas Dnm1p only contributes to a minor extent under non-proliferative conditions, under peroxisome-proliferation conditions, the contribution of Dnm1p is much greater.

We analysed a third DRP, Mgm1p, for its function in peroxisome multiplication. Mgm1p has been shown to be required for mitochondrial fusion, for the proper assembly of F₁F₀-ATPase and for cristae formation (Amutha et al., 2004; Meeusen et al., 2006). We found that mgm1Δ cells show increased peroxisome abundance (Fig. 6), although the abundance of peroxisomes in these cells was very sensitive to growth conditions (see Materials and Methods). vps1Δ/mgm1Δ mutants displayed an increased peroxisome abundance similar to that seen in mgm1Δ cells and atpΔ cells (Fig. 6). vps1Δ/dnm1Δ/mgm1Δ triple mutants, however, showed a huge decrease in peroxisome number: most cells contained a single peroxisome. Because mature peroxisomes do not fuse (Motley and Hettema, 2007), the phenotype of vps1Δ/dnm1Δ/mgm1Δ cells implies that the increased peroxisome number in mgm1Δ and mgm1Δ/vps1Δ cells is a result of excessive fission by DRPs.

We tested whether Dnm1p and Vps1p are redundant in other functions besides peroxisomal fission. vps1Δ cells have a vacuolar fusion defect that can be visualised by allowing the cells to take up FM4-64 to steady-state levels (Vida and Emr, 1995). In wild-type cells, FM4-64 accumulates in the vacuolar membrane, whereas, in vps1Δ cells, staining is most intense in the smaller, fragmented vacuolar structures (Fig. 7). Overexpression of Dnm1p did not rescue the vacuolar fusion defect observed in these cells (Fig. 7). Similarly, we found that the mitochondrial fission defect in dnm1Δ cells was not rescued by overexpression of Vps1p (Fig. 7). We conclude that Vps1p and Dnm1p are partially redundant for their role in peroxisome fission, with their functions on other organelles being non-overlapping.

In this paper, we establish that the accessory proteins required for Dnm1p-dependent peroxisome fission are the same as those required for Dnm1-dependent mitochondrial fission, whereas Vps1p-mediated peroxisome fission is independent of these proteins. We show that effective regulation of peroxisome fission seems to require both Vps1p and Dnm1p, and that their relative contributions vary with growth conditions.

The role of the dynamin-related protein Dnm1p in mitochondrial fission is well established (for a review, see Hoppins et al., 2007). Dnm1p-mediated mitochondrial fission requires the mitochondrial outer-membrane protein Fis1p, and the adaptors Mdv1p and Caf4p. The two paralogous DRPs Vps1p and Dnm1p have been shown to be important for a normal abundance of peroxisomes, and both have been found to be associated with peroxisomes (Hoepfner et al., 2001; Vizeacoumar et al., 2006; Kuravi et al., 2006). We have recently shown that Vps1p and Dnm1p control peroxisome abundance by stimulating fission (Motley and Hettema, 2007).

Whereas the association of Vps1p with peroxisomes requires interaction with Pex19p, it has been shown that Fis1p is required for Dnm1p recruitment to peroxisomes (Vizeacoumar et al., 2006; Kuravi et al., 2006). In this paper, we show that not only Dnm1p but also Caf4p and Mdv1p are recruited to peroxisomes, and that recruitment is dependent on Fis1p. Only a small amount of GFP-Mdv1p and GFP-Caf4p was found to colocalise with some peroxisomes, suggesting that the association is only temporary.

Using our recently developed mating assay, we showed a requirement for Fis1p, Mdv1p and Caf4p in Dnm1p-dependent peroxisome fission. When vps1Δ/dnm1Δ cells overexpressing Dnm1p were mated with vps1Δ/dnm1Δ cells, fusion of the peroxisomal structure in the Dnm1p-deficient mating partner occurred soon after cytoplasmic mixing. However, when vps1Δ/dnm1Δ cells overexpressing Dnm1p were mated with either vps1Δ/fis1Δ or vps1Δ/caf4Δ/mdv1Δ cells, the peroxisomal structure lacking Fis1p or Caf4p and Mdv1p initially failed to divide in spite of the presence of Dnm1p in the (now mixed) cytoplasm of the mating cells. Only after several hours, when the zygote was being formed, did the vps1Δ/fis1Δ peroxisome begin to divide. The delay in fission suggests that Fis1p, Caf4p and Mdv1p do not associate with the peroxisomal structures directly after cell fusion, and that equilibration of the existing pool is slow or that synthesis of these factors is required. This is not surprising because Fis1p is an integral membrane protein, and Caf4p and Mdv1p are peripheral membrane proteins.

We found that Mdv1p and Caf4p can substitute functionally for each other on peroxisomes. This is not the case for mitochondria, because mdv1Δ cells have a mitochondrial fission defect that is much more pronounced than that of caf4Δ cells (Griffin et al., 2005). Indeed, Caf4p has been shown to be required for a more peripheral distribution of Dnm1p. This pool of Dnm1p is not immediately involved in the fission process (Schauss et al., 2006). Another factor that acts in concert with Dnm1p is Num1p. Num1p is required for normal mitochondrial morphology and seems to couple mitochondrial inheritance to fission (Cerveny et al., 2007). Our data do not support a role for this protein in peroxisome fission.

Because Dnm1p- and Vps1p-dependent peroxisome fission appear to be partially redundant, we tested whether Dnm1p could compensate for a Vps1p deficiency. We found that overexpression of Dnm1p can rescue the partial peroxisome fission defect observed in cells lacking Vps1p. This suppression depends on the presence of Fis1p and on that of Caf4p and Mdv1p. Our results imply that peroxisomes normally have to compete with mitochondria for Dnm1p. Cells lacking either Fis1p or Caf4p display an increase in cytosolic Dnm1p (Schauss et al., 2006). The same is observed in cells lacking Num1p (Cerveny et al., 2007). However, vps1Δ/num1Δ double mutants do not restore peroxisome number to wild-type levels, nor do caf4Δ/vps1Δ double mutants. This implies that simply increasing the level of Dnm1p in the cytosol is not sufficient to restore peroxisome fission in vps1Δ cells. Because Fis1p is essential for Dnm1-dependent peroxisomal and mitochondrial fission, we redirected Fis1p to an exclusively peroxisomal location. A Fis1p-Pex15p fusion expressed in vps1Δ/fis1Δ cells failed to rescue the mitochondrial fission defect, whereas peroxisome abundance was restored to normal. This shows the importance of Fis1p in localising Dnm1p to mitochondria versus peroxisomes.

We analysed the third DRP present in yeast, Mgm1p. We show that cells lacking Mgm1p have increased peroxisome abundance and that this increase is a result of excessive DRP-dependent fission. Is this reflecting a direct role of Mgm1 on peroxisome dynamics? We favour the interpretation that this increase is not a direct effect of Mgm1p on peroxisomes but rather a response to mitochondrial dysfunction. First, we were unable to colocalise Mgm1p with peroxisomes (not shown). Furthermore, peroxisome proliferation has been reported in response to mitochondrial dysfunction (Butow and Avadhani, 2004) and mgm1Δ cells display a variety of mitochondrial defects, including a failure to assemble their F₁F₀-ATPase properly and increased loss of mitochondrial DNA (mtDNA). Indeed, all ATP-synthase mutants we tested showed an increase in peroxisome number and this increase was dependent on DRPs. Our data are compatible with the interpretation that the increase in peroxisome abundance in mgm1Δ cells is a result of the mitochondrial dysfunction. However, we cannot exclude a more direct role of Mgm1p in peroxisome dynamics.

Whereas the fission of peroxisomes in humans and Hansenula polymorpha (Nagotu et al., 2007) depends on a single DRP, the presence of two homologous DRPs on peroxisomes in S. cerevisiae is intriguing (Schrader and Yoon, 2007). Despite their extensive amino acid sequence identity, Dnm1p and Vps1p are recruited independently to peroxisomes. Furthermore, they cannot substitute for each other in their other functions (for instance, vacuole fusion and mitochondrial fission). This illustrates that these DRPs are functionally distinct and suggests a functional significance for the presence of these two DRPs on peroxisomes in yeast. We can only guess what the significance is.

Because mitochondria and peroxisomes are metabolically linked, a coordinated regulation of their biogenesis is not surprising, and has been described in both humans and yeast (for a review, see Schrader and Yoon, 2007). Peroxisomes proliferate in cells with dysfunctional mitochondria (Butow and Avadhani, 2004). We show that this proliferation is dependent on DRPs. Whereas Dnm1p makes only a minor contribution to peroxisome fission under standard growth conditions, in mutants with mitochondrial dysfunction, the contribution of Dnm1p appears to be much greater (compare vps1Δ with vps1Δ/atp7Δ). This suggests that yeast is able to coordinate the division of peroxisomes with the functional state of mitochondria via Dnm1p. We are currently investigating the molecular basis for this.

Yeast strains were derivatives of BY4741 (MATA his3-Δ1 leu2-Δ0 met15-Δ0 ura3-Δ0) or BY4742 (MATa his3-Δ1 leu2-Δ0 lys2-Δ0 ura3-Δ0) obtained from the EUROSCARF consortium. Double or triple gene deletions were made by replacing the entire coding sequence of the mutated genes with a marker (Schizosaccharomyces pombe HIS5 or the Klebsiella pneumoniae hygromycin B phosphotransferase gene cassette that confers resistance to Hygromycin B) (Goldstein and McCusker, 1999). dnm1Δ/vps1Δ, mdv1Δ/vps1Δ, caf4Δ/vps1Δ, mdv1Δ/caf4Δ/vps1Δ, num1Δ/vps1Δ and Δfis1Δ/vps1 were generated by replacing the VPS1 reading frame with the HIS5 cassette in dnm1Δ, mdv1Δ, caf4Δ, mdv1Δ/caf4Δ, num1Δ and fis1Δ, respectively. The MGM1, ATP7 and ATP17 open reading frames were replaced by that of the Hygromycin cassette to generate dnm1Δ/vps1Δ/mgm1Δ, dnm1Δ/vps1Δ/atp7Δ, dnm1Δ/vps1Δ/atp17Δ, vps1Δ/atp7Δ, vps1Δ/atp17Δ and dnm1Δ/atp7Δ. The mdv1Δ/caf4Δ double mutant was constructed by replacing the CAF4 open reading frame with that of the Hygromycin cassette in the mdv1Δ mutant.

URA3 and LEU2 centromere plasmids were derived from Ycplac33 and Ycplac111 (Gietz and Sugino, 1988). GFP-PTS1 is a peroxisomal luminal GFP marker protein appended with the well-characterised peroxisomal targeting signal type 1 (PTS1) (Gould et al., 1988). A far-red peroxisomal luminal marker was made by appending a variant of the Heteractis crispa chromoprotein (HcRed) with the PTS1. As source of HcRed, we used HcRed-Tandem with optimised yeast codon usage (Evrogen, Moscow, Russia).

Constitutive expression of GFP-PTS1 and HcRed-PTS1 was under control of the TPI1 promoter and the HIS3 promoter, respectively. Dnm1p overexpression was achieved using the TPI1 promoter. All constitutive expression constructs contained the PGK1 terminator. The Fis1p-Pex15p fusion protein was expressed from a construct containing the FIS1 promoter, and the fusion protein contained the cytoplasmic domain of Fis1p and the C-terminal tail anchor of Pex15p. Caf4p and Mdv1p were N-terminally tagged with GFP. Conditional-expression constructs contained the GAL1 promoter. In order to reduce the half-life of the transcript, we replaced the PGK1 terminator with the MFA2 terminator (Duttagupta et al., 2003; LaGrandeur and Parker, 1999).

For all experiments, cells were grown overnight in selective glucose medium. For analysis of phenotypes by microscopy, cells were subsequently diluted to OD 0.1 in 2% glucose medium + casamino acids and grown for two to three cell divisions (4-6 hours), so that phenotypes were analysed under conditions whereby cells are actively maintaining their peroxisome number. In the case of the mgm1Δ and ATP-synthase mutants, peroxisome number varied significantly depending on growth media, but reproducible results were obtained using 2% glucose + casamino acid medium. Cells were fixed (see below) for 5 minutes before imaging. For the experiment described in Fig. 3, an overnight culture was used to inoculate selective galactose medium at an OD₆₀₀ of 0.1 to allow induction of reporter proteins for 3 hours. Cells were then switched to selective glucose medium for 2 hours, to shut down expression of the gal-inducible reporter protein, before mating. For mating, cells were collected by filtration onto a 0.22-μm Millipore nitrocellulose filter (type GS, 25-mm diameter) and this filter was incubated, cells side up, on a pre-warmed YPD plate at 30°C. 1×10⁷ cells of each strain were collected per 25-mm filter.

After 2 hours, cells were harvested by vortexing the filter in selective glucose medium and fixed for 5 minutes by adding formaldehyde to 3.6%. Free formaldehyde groups were quenched in 0.1 M ammonium chloride/1×PBS. Cells were imaged within 1 hour of fixing because loss of fluorescence intensity and increased autofluorescence was seen in fixed cells left for extended periods. For each experiment, >100 cells were examined and images are representative of the findings. For Fig. 1, budding cells were counted as single cells.

Cells were grown in 2% glucose selective medium to log phase (OD 0.5). 1 ml of log-phase cells were pelleted, resuspended in YPD containing 20 μM FM4-64 and incubated at 30°C for 15 minutes. They were then pelleted, resuspended in 1 ml of YPD and incubated at 30°C for 30 minutes. Subsequently, the cells were washed in 1 ml of water before being resuspended in 2% glucose medium, ready for imaging.

Live and fixed cells were analysed at with an Axiovert 200M (Zeiss) equipped with Exfo X-cite 120 excitation light source, band-pass filters (Zeiss and Chroma), and alpha Plan-Fluar 100×/1.45 NA or A-Plan 40×/0.65 NA Ph2 objective lens (Zeiss) and Hamamatsu Orca ER digital camera. Image acquisition was performed using Openlab software (Improvision) at 21°C. Fluorescence images were collected as 0.2-μm z-stacks and merged into one plane after contrast enhancing in Openlab, and processed further in Photoshop except when stated differently in the text or figure legends. Bright-field images were collected in one plane. Blue colour was applied to the bright-field image using Adobe Photoshop. The level of the bright-field images was modified, and the image was blurred, sharpened and blurred again before one more round of level adjustment so that only the circumference of the cell was visible.

The authors thank Peter Piper and Stephan Mills for help with the robotics, and Nina Rajala and Naomi Saggers for help with strain construction. This research was supported by a Wellcome Trust Career Development Fellowship in Basic Biomedical Sciences awarded to E.H.H. and a Royal Society Research Grant.