Yeast cells contain a heterogeneous population of peroxisomes that segregate asymmetrically during cell division

Eukaryotic cells contain membrane-bound compartments called organelles that carry out specific functions. Among them, peroxisomes are organelles bound by a single membrane that are present in almost all eukaryotic cells. These physiologically important organelles harbor a large variety of metabolic and non-metabolic functions, depending on the organism, its developmental stage and the tissue. A common peroxisome function is hydrogen peroxide metabolism (Kohlwein et al., 2013; Wanders and Waterham, 2006). Along with mitochondria, peroxisomes contribute to reactive oxygen species homeostasis and are thought to play an important role in cellular ageing (Lefevre et al., 2015; Manivannan et al., 2012; Titorenko and Terlecky, 2011).

Two modes of peroxisome proliferation have been proposed, namely multiplication by fission of pre-existing organelles and de novo formation from the endoplasmic reticulum. Although controversy remains about the extent to which each of these two mechanisms contributes to the total cellular peroxisome population in different organisms, data obtained in the yeast species Hansenula polymorpha (Nagotu et al., 2008) and Saccharomyces cerevisiae (Motley and Hettema, 2007) indicate that peroxisome fission is the major pathway in these organisms.

According to the so-called growth and division model of peroxisome proliferation (Lazarow and Fujiki, 1985), a pre-existing peroxisome grows by incorporating newly synthesized membrane and matrix constituents until it reaches a mature size. Subsequently, this organelle divides, a process that involves the three consecutive steps of elongation, constriction and membrane scission (Lazarow and Fujiki, 1985; Motley and Hettema, 2007). Potentially, peroxisomes can divide symmetrically or asymmetrically. Symmetric division of peroxisomes leads to an equal distribution of the components of the mother organelle in the two daughter organelles. Upon growth, these daughter organelles have similar fractions of proteins originating from the mother organelle and of newly incorporated proteins. In contrast, asymmetric division of peroxisomes produces dissimilar organelles. Upon further growth of the smaller organelles, the result is one organelle containing predominantly proteins originating from the mother organelle and a second organelle that is mainly composed of newly synthesized components.

It has been suggested that peroxisomes in mammalian cells divide asymmetrically (Schrader et al., 2012), resulting in larger, mature mother organelles in conjunction with smaller daughter peroxisomes. Fluorescence microscopy data from yeast indicate that peroxisomes are heterogeneous with respect to the distribution of certain peroxisomal membrane proteins (Cepińska et al., 2011; Yofe et al., 2016). However, it is unknown whether the organelle population is also heterogeneous with respect to age.

Using a tandem fluorescent protein timer (Khmelinskii et al., 2012), we studied the heterogeneity of peroxisomes in S. cerevisiae. This approach can distinguish between relatively young and old peroxisomes. Our data indicate that, within one cell, peroxisomes differ with respect to the age of their matrix protein content. In addition, we observed that the relatively old organelles are retained in the mother cells during yeast budding, whereas the younger ones are transported to the buds. This careful segregation of older and younger organelles is maintained during multiple budding events. Interestingly, the replicative lifespan (RLS) of mother cells of an inp2 deletion strain was longer than that of a wild-type control. In this strain, both the older and newer organelles are retained in the mother cells. Our data suggest that in wild-type yeast cells the older, possibly deteriorated, organelles are retained in the mother cell, whereas the younger more vital organelles are preferentially transported to the buds.

To determine the relative age of the peroxisomal matrix content, we used a fusion protein consisting of DsRed1 and superfolder GFP (sfGFP) (Khmelinskii et al., 2012) with the peroxisomal targeting signal –SKL at the extreme C-terminus (Fig. 1A), placed under control of the constitutive promoter of the glyceraldehyde 3-phosphate dehydrogenase gene (pTHD3). DsRed1 has a very long maturation time (half-life ∼11 h) (Campbell et al., 2002), whereas sfGFP maturation takes only a few minutes (Pédelacq et al., 2006). Upon import of this fusion protein into peroxisomes, sfGFP immediately shows green fluorescence, whereas the red fluorescence gradually increases with time. Thus, the ratio of the fluorescence intensities of DsRed1 to sfGFP can serve as a measure of the age of peroxisomal matrix protein content. If peroxisomes divide symmetrically, resulting in daughter organelles that have a similar capacity to import matrix proteins, all organelles within one cell are expected to show a similar ratio of green to red fluorescence (Fig. 1B). However, asymmetric peroxisome fission probably results in a heterogeneous population of organelles, with a range of DsRed1/sfGFP fluorescence intensity ratios (Fig. 1B).

As shown in Fig. 1C, fluorescence microscopy of glucose-grown cells revealed multiple green fluorescent spots per cell, whereas only a subset of organelles also showed bright red fluorescence (Fig. 1C). We measured fluorescence intensities of DsRed1 and sfGFP on individual peroxisomes, which were normalized to the total fluorescence intensity of the respective fluorophore in the cell. Under our experimental conditions, all peroxisomes showed both green and red fluorescence (Fig. S1). As expected, correlation between fluorescence intensities of these fluorophores on individual peroxisomes showed no linear relationship. Instead, peroxisomes with relatively low sfGFP fluorescence intensities showed much lower DsRed1 intensities than expected for a linear relationship (Fig. S1). This result shows that indeed peroxisomes occur in cells, which contain newly synthesized DsRed1–sfGFP fusion protein, of which the DsRed1 portion of this fusion is not yet fully mature. To capture the whole range of heterogeneity, we did not use any fluorescence intensity cutoffs in our quantification except for removal of background fluorescence. As expected, quantification of the DsRed1/sfGFP fluorescence intensity ratios on individual peroxisomes revealed a large heterogeneity in glucose-grown cells (Fig. 1D). A similar heterogeneity was observed when cells were grown on oleic acid, a condition that induces peroxisome proliferation (Fig. 1E).

The heterogeneity was not caused by proteolytic degradation of the fusion protein, because western blot analysis indicated that almost all protein was present in the full-length form (Fig. S2). The observed heterogeneity confirms that peroxisomes divide asymmetrically in both glucose- and oleic acid-grown yeast cells,.

Next, we investigated the fate of nascent and mature organelles during growth of cells on glucose. Fluorescence microscopy of a strain constitutively producing DsRed1–sfGFP–SKL revealed that peroxisomes with high red fluorescence were generally localized in the mother cells, whereas organelles that showed low or no red fluorescence were more abundant in the buds (Fig. 2A). This was confirmed by a significantly higher ratio of DsRed1 to sfGFP fluorescence intensity on individual peroxisomes in mother cells relative to those present in developing buds (Fig. 2B). Furthermore, the fluorescence intensity ratio of total cellular DsRed1 to sfGFP was also higher in the mother cells relative to the buds, confirming that organelles with older peroxisomal matrix proteins are preferentially retained in the mother cells (Fig. S3A). Similar results were obtained when cells were grown on oleic acid (Fig. 2C; Fig. S3B).

S. cerevisiae mother cells typically produce approximately 25 buds before they die. Using a microfluidics device, we checked whether the observed asymmetry in peroxisome inheritance was maintained during multiple consecutive budding events of an individual mother cell. In this experimental set-up, individual mother cells trapped in a microfluidics device are imaged by fluorescence microscopy during multiple budding events, while the daughter cells are continuously removed by a flow of fresh medium (Lee et al., 2012). A representative example is shown in Fig. 2D, where the peroxisome inheritance of one individual yeast mother cell is monitored during 21 consecutive budding events (total time of imaging approximately 30 h). This experiment shows that the peroxisome with the highest DsRed1/sfGFP fluorescence intensity ratio was always retained in the mother cell, whereas the other organelles were segregated between mother and daughter cells (Fig. 2D).

Our results indicate that retention of older peroxisomes in mother cells and segregation of nascent peroxisomes to buds is carefully controlled during the entire replicative lifespan of the mother cells. The retention of the older peroxisomes in the mother cell is most probably independent of their larger size because peroxisomes measure up to 0.2 µm in diameter in glucose-grown S. cerevisiae cells (Veenhuis et al., 1987) whereas the diameter of the bud neck is about 1 µm (Bertin et al., 2012). Hence, even the largest organelles can easily pass through the bud neck. Moreover, if the largest peroxisomes became stuck at the bud neck because of their size, they would accumulate close to the bud neck, which was never observed.

Inheritance of peroxisomes protein 1 (Inp1) is a peroxisomal membrane protein that plays a role in peroxisome retention in the mother cell, whereas Inp2 is required to transport peroxisomes to the yeast bud via actin filaments (Fagarasanu et al., 2005, 2006). Localization of both proteins in cells of strains producing DsRed1–SKL revealed that Inp1–GFP was mainly present in mother cells on peroxisomes with high DsRed1 intensities (Fig. 3A), whereas Inp2–GFP was enriched in daughter cells on peroxisomes with low DsRed1 fluorescence intensity (Fig. 3B).

Quantification of GFP fluorescent spots in the dividing cells showed that Inp1–GFP spots were abundant in the mother cells whereas Inp2–GFP spots were prevalent in buds (Fig. 3C). Moreover, total GFP fluorescence intensity quantification in the mothers and buds further confirmed that Inp1–GFP fluorescence intensity was higher in mother cells, whereas Inp2–GFP intensity was higher in buds (Fig. 3D).

These observations are in line with earlier observations in S. cerevisiae and H. polymorpha, which indicated that Inp1-containing peroxisomes are retained in mother cells, whereas organelles that move to the bud harbor Inp2 (Cepińska et al., 2011; Knoblach et al., 2013). We now show for the first time that Inp1 and Inp2, which both are produced in a cell cycle dependent way, associate with mature and nascent organelles, respectively. Inp1 is recruited to peroxisomes by the membrane protein Pex3, which is present on all organelles (Knoblach et al., 2013). Possibly, the levels of Pex3 increase in time on the older organelles, thus increasing the capacity of the organelles to bind Inp1. Inp2 is an integral peroxisomal membrane protein. How this protein preferably inserts into the membranes of the younger organelles is unknown. A possible mechanism is that higher curvature of the membranes in the smaller, younger organelles plays a role.

The enhanced level of Inp1 on older peroxisomes enables effective retention of these organelles in the mother cells by not allowing their transfer to the daughter cell. By contrast, the presence of Inp2 on the nascent organelles preferentially transfers these peroxisomes to the bud. This asymmetry could prevent transfer of damaged, mature organelles to newly formed buds.

To test the physiological relevance of the careful partitioning of mature and nascent organelles between mother and daughter cells, we analyzed whether interfering in this process affects the RLS of mother cells.

First, our studies revealed that intact peroxisomes are important for yeast RLS because pex3 cells, which lack functional peroxisomes, show a reduced RLS relative to wild-type controls (Fig. 4A; mean RLS of pex3 cells is 20.4, compared with 25.7 for the wild-type control).

Next, we analyzed inp1 and inp2 mutant strains. In cells lacking INP1, all peroxisomes are transported to the developing buds, leaving mother cells devoid of peroxisomes. Conversely, deletion of INP2 results in the retention of all organelles in the mother cell and the formation of buds without peroxisomes. In inp1 cells, peroxisome numbers are decreased (Fagarasanu et al., 2005), whereas inp2 cells show a heterogeneity of peroxisome numbers, with some cells containing more peroxisomes and others less than in wild-type cells (Fagarasanu et al., 2006). As shown in Fig. 4B, the RLS of inp1 cells is reduced relative to the wild-type control, whereas that of inp2 cells is enhanced. Obviously, the loss of all peroxisomes in inp1 mother cells decreases the lifespan of these cells. Conversely, our data suggest that the retention of all organelles in the mother cells positively affects the RLS.

It has been suggested that, in S. cerevisiae, peroxisomes are not important when cells are grown on glucose, because peroxisome-deficient mutants normally grow on medium containing glucose. However, our data show that glucose-grown pex3 cells, which do not contain functional peroxisomes, have a shorter RLS than wild-type controls. Similarly, mother cells of an INP1-deficient strain, which are unable to retain peroxisomes, show a reduced RLS in glucose-containing medium. These observations indicate that functional peroxisomes are important for the RLS of glucose-grown yeast cells. The finding that retention of all peroxisomes in mother cells of the inp2 strain has a positive effect on yeast RLS further supports this conclusion. Possibly, enhanced levels of peroxisomal catalase in mother cells of the inp2 mutant strain contribute to reduced levels of reactive oxygen species, which may enhance the RLS. However, other as-yet-unknown peroxisome-bound metabolic or non-metabolic functions might also contribute to extension of the RLS (Fransen et al., 2012).

The S. cerevisiae strains used in this study are listed in Table S1. Cells were grown at 30°C on mineral medium (MM) (Van Dijken et al., 1976) containing 0.25% ammonium sulfate and 2% glucose or 0.1% oleic acid. MM was supplemented with the required amino acids or uracil to a final concentration of 20 μg/ml (histidine and methionine) or 30 μg/ml (leucine, lysine and uracil). For growth on agar plates, YPD medium (1% yeast extract, 1% peptone, 1% glucose) was supplemented with 2% agar. For selection of zeocin- or nourseothricin-resistant transformant, YPD plates containing 200 µg/ml zeocin (Invitrogen) or 100 µg/ml nourseothricin (Werner BioAgents) were used. To select transformants based on amino acid prototrophy, yeast nitrogen base (YNB) plates without amino acids (Difco; BD), containing 1% glucose and 2% agar, was used with the appropriate amino acids. For cloning purposes, Escherichia coli DH5α was used, which was cultured at 30°C on Luria–Bertani medium supplemented with the appropriate antibiotics.

Yeast strains, plasmids and primers used in this study are listed in Tables S1, S2 and S3, respectively.

To construct pSAN02, DsRED1 was amplified from pYM35 (Janke et al., 2004) using primers DsRED1-Fw and DsRed1-Rev, digested with BamHI/HindIII and cloned into pSL34 (Lefevre et al., 2013), resulting in plasmid pP_MET25-DsRed1-SKL. Subsequently, DsRed1–SKL was amplified from pP_MET25-DsRed1-SKL using primers DsRED1-1 and DsRED1-2, digested with HindIII/SalI and cloned into pHIPN4 (Saraya et al., 2012), resulting in plasmid pHIPN4-DsRed1-SKL. The TDH3 promoter was amplified from S. cerevisiae genomic DNA using primers TDH3-4 and TDH3-5, digested with NotI/HindIII and cloned into pHIPN4-DsRed1-SKL, which resulted in plasmid pP_TDH3-DsRed1-SKL (pSAN02). pSAN02 was linearized using MfeI and integrated in the THD3 promoter region of the S. cerevisiae genome.

For construction of a plasmid encoding a fusion protein of DsRed1 and sfGFP containing the peroxisomal targeting signal 1 (PTS1) sequence –SKL at the extreme C-terminus, an mCherry–sfGFP fragment was first amplified from pMaM17 (Khmelinskii et al., 2012) using primer pair MC.sfGFP-Fw/MC.sfGFP-Rev. This PCR fragment was digested with AatII/XhoI and cloned into pSL33, resulting in plasmid pP_MET25-mCherry-sfGFP-SKL. From this plasmid, an mCherry–sfGFP–SKL fragment was amplified using primer pair MC.sfGFP-1/MC.sfGFP-2 and a HindIII/SalI fragment was cloned into pHIPZ4 (Salomons et al., 2000), resulting in pHIPZ-mCherry-sfGFP-SKL. The DsRed1 was amplified from pYM35 using DsRed1.HindIII_F and DsRED2.OL; sfGFP–SKL was amplified from pHIPZ-mCherry-sfGFP-SKL using primers sfGFP.DsRED1.OL and sfGFP_R. Both fragments were joined together by overlap PCR and the HindIII/SalI fragment cloned into pSAN02, resulting in plasmid pP_TDH3-DsRed1-sfGFP-SKL (pSAN03). Plasmid pSAN03 was linearized by restriction enzyme MfeI and integrated into the S. cerevisiae genome.

To construct pSL33, the P_MET25-DsRed-SKL-tcyc1 fragment was amplified from pUG34-DsRed-SKL (Kuravi et al., 2006) using primer pair DsRed-1/DsRed-2. The obtained PCR product was digested with KpnI/XbaI and cloned into pBSII KS⁺, resulting in pSL32. The nourseothricin resistance gene was amplified from pAG25 (Goldstein and McCusker, 1999) using primer pair Nat1.1/Nat1.2, digested with SacII/KpnI and cloned into pSL32, resulting in pSL33.

Wide-field fluorescence images (Fig 1C; Fig 2A) were acquired using a 100×1.30 NA Plan-Neofluar objective (Carl Zeiss). These images were captured by an inverted microscope (Observer Z1; Carl Zeiss) using AxioVision software (Carl Zeiss) and a digital camera (CoolSNAP HQ²; Photometrics). The GFP signal was visualized with 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 with a 546⁄12-nm bandpass excitation filter, a 560-nm dichromatic mirror and a 575–640-nm bandpass emission filter. For both DsRed1 and sfGFP image acquisition, an exposure time of 500 ms was used. Under these conditions, the samples were not overexposed and the fluorescence intensities of the fluorophores not saturated, as established by the maximum pixel intensity value.

Confocal images (Figs 2D, 3A,B) were acquired with a confocal microscope (LSM510; Carl Zeiss) equipped with photomultiplier tubes (Hamamatsu Photonics) and Zen 2009 software (Carl Zeiss). GFP fluorescence was visualized by excitation with a 488-nm argon ion laser (Lasos), and emission detected using a 500–550-nm band-pass emission filter. The DsRed signal was visualized by excitation with a 543-nm helium neon laser (Lasos) and emission was detected using a 565–615-nm bandpass emission filter.

Image analysis was performed using ImageJ software. For quantification, peroxisomes were first selected based on green fluorescence and then the fluorescence intensities of sfGFP and DsRed1 on each peroxisome were measured. The fluorescence intensity on each peroxisome was corrected by subtracting the background intensity from a region in the cell without peroxisomes. Moreover, fluorescence intensities of sfGFP and DsRed1 on individual peroxisomes in a cell were normalized by dividing by the total intensity (sum of all peroxisomes) of the respective fluorophore in the cell. The normalized fluorescence intensities on each peroxisome were used to calculate the DsRed1/sfGFP fluorescence intensity ratio.

To study peroxisome inheritance, a single cell microfluidic dissection device was used with a constant supply of mineral medium. Time-lapse images were acquired using the confocal microscope LSM510. The objective and microfluidic dissection device were kept at 30°C.

The RLS of yeast strains was measured using a microfluidics dissection device (Lee et al., 2012) at 30°C. Mineral medium supplemented with 2% glucose and appropriate amino acids (without yeast extract) was supplied at a flow rate of 5–7 µl/min throughout the experiment. Time-lapse bright field images were acquired every 30 min using a wide field inverted microscope (Observer Z1; Carl Zeiss). The number of buds produced by each cell was counted and the RLS curve obtained using Kaplan–Meier analysis.

We would like to thank Kevin Knoops and Arjen M. Krikken for helpful discussions.