The very-long-chain fatty acid elongase Elo2 rescues lethal defects associated with loss of the nuclear barrier function in fission yeast cells

Fatty acids are components of lipids, which constitute the intracellular membranes of eukaryotic cells, and are classified by their carbon chain length. Long-chain fatty acids have a chain length of 11-20 carbon atoms (C11-C20), with C16 and C18 being the most abundant in mammalian cells. Fatty acids with a chain length >20 carbon atoms are called very-long-chain fatty acids; they are less abundant, accounting for only a few percent of the total fatty acids in the cell. However, they play crucial roles that cannot be substituted by long-chain fatty acids: most very-long-chain fatty acids constitute sphingolipids, which play important roles as skin barriers and in neural functions (Mizutani et al., 2009; Imgrund et al., 2009; Kihara, 2016). Fatty acids are elongated by a group of enzymes embedded in the endoplasmic reticulum (ER) membrane (reviewed in Kihara, 2012); the most important ones are fatty acid elongases because they catalyze the rate-limiting step of the fatty acid elongation cycle. Cells cannot utilize extracellular very-long-chain fatty acids since they are – unlike long-chain fatty acids – rarely transported through the plasma membrane (Sassa et al., 2014). Therefore, very-long-chain fatty acids must be synthesized in the cell. Mammalian cells have seven fatty acid elongase genes (ELVOL1-ELOVL7) (Leonard et al., 2004; Jakobsson et al., 2006). Several diseases related to fatty acid metabolism, such as Stargardt muscular dystrophy, ichthyosis and neurological disorders, have been found in humans (reviewed in Kihara, 2012; Logan and Anderson, 2014; Agbaga, 2016; Kihara, 2016). In S. cerevisiae, there are three genes encoding fatty acid elongases: ELO1, ELO2 and ELO3 (Toke and Martin, 1996; Oh et al., 1997). Elo1 catalyzes elongation of long-chain fatty acids, whereas Elo2 and Elo3 catalyze that of very-long-chain fatty acids. The double deletion of S. cerevisiae Elo2 and Elo3 is lethal (Revardel et al., 1995; Silve et al., 1996), indicating that very-long-chain fatty acids are essential for cell growth. However, the roles of very-long-chain fatty acids in the maintenance of intracellular membranes are not well understood.

The eukaryotic genome is compartmentalized by the nuclear envelope, which contains lipid membranes. Functioning of the nuclear envelope is crucial for genome integrity and cell viability in eukaryotes. The nuclear envelope is composed of the inner and outer nuclear membranes, which are continuous with the ER membrane. Maintaining the homeostasis of these membranes is an important cellular activity performed by a group of lipid metabolic enzymes. However, it is unknown how the lipid metabolic enzymes regulate functions of the nuclear and ER membranes.

Lem2 and Man1 belong to the conserved nuclear membrane proteins of the LAP2-Emerin- Man1 (LEM)-domain protein family. The LEM domain itself was initially identified as a chain of ∼40 amino acids common to the N-terminal nucleoplasmic region of LAP2, emerin, and Man1 in metazoans (Lin et al., 2000; Laguri et al., 2001; Brachner et al., 2005). The LEM-domain protein family was later expanded to include Lem2 (LEMD2 in mammals), Lem3 (ANKLE1 in mammals), Lem4 (ANKLE2 in mammals), and Lem5 of metazoans (reviewed in Lee and Wilson, 2004; Wagner and Krohne, 2007; Barton et al., 2015). Heh1 (officially called Src1) and Heh2 – identified in the budding yeast Saccharomyces cerevisiae (King et al., 2006; Grund et al., 2008) – display a LEM-like helix-extension-helix (HEH) domain in their N-terminal region. They also have an MAN1–Src1–C-terminal (MSC) domain in their C-terminal region, similar to the domain present in metazoan Lem2 and Man1 proteins (Mans et al., 2004; Brachner and Foisner, 2011). Lem2 and Man1 are also conserved in the fission yeast Schizosaccharomyces pombe, show domains similar to the LEM and MSC domains, as well as two transmembrane domains (Hiraoka et al., 2011; Gonzalez et al., 2012; Steglich et al., 2012; Barrales et al., 2016; Tange et al., 2016; Yang et al., 2017).

In S. pombe, the loss of Lem2 causes pleiotropic phenotypes associated with chromosome instability, such as the defective formation of pericentromeric heterochromatin and the frequent loss of a mini-chromosome (Barrales et al., 2016; Tange et al., 2016). Lem2 interacts with Bqt4, another nuclear membrane protein that plays a direct role in anchoring telomeres to the nuclear membrane (Chikashige et al., 2009). Cells lacking both Lem2 and Bqt4 show synthetic growth defects, whereas the loss of either one does not affect cell growth (Tange et al., 2016), suggesting that there is a functional interaction between the two proteins. In this study, we show that the conserved fatty acid elongase Elo2 rescues synthetic defects caused by the absence of Lem2 and Bqt4.

Initially, we searched for multicopy suppressors of the synthetic growth defect phenotype in cells lacking Lem2 and Bqt4. Since the lem2Δbqt4Δ double mutant is lethal during vegetative growth, we constructed a strain with a deletion in the bqt4⁺ gene (bqt4Δ) that also bears a conditional lem2 shut-off construct, in which the lem2⁺ gene is expressed under the nmt81 promoter (for the strain construct, see Table S1). In this strain, lem2⁺ expression is shut off in the presence of thiamine, as confirmed by western blotting (Fig. S1). As shown in Fig. 1A, cells of the bqt4Δ lem2 shut-off strain were viable in the absence of thiamine (lem2 ON), but showed growth defects in the presence of thiamine (lem2 shut-off). As a control, multicopy plasmids bearing the lem2⁺ or bqt4⁺ gene rescued the growth defect of the bqt4Δ lem2 shut-off strain in the presence of thiamine (Fig. 1A). In the screening experiments, we transformed the bqt4Δ lem2 shut-off strain with a plasmid library of the S. pombe genome, and obtained 610 suppressor colonies among ∼80,000 transformants in two independent experiments (see Materials and Methods). Colony PCR and/or DNA sequencing revealed that 314 suppressor colonies contained the lem2⁺ gene, whereas 230 contained the bqt4⁺ gene. We purified plasmids from 66 unidentified suppressors and determined their DNA sequences. All these plasmids contained the same DNA sequence (see ‘suppressor 1’ in Fig. 1A), which included the SPAC1B2.03c gene (elo2⁺). By transforming the bqt4Δ lem2 shut-off strain with this single gene, we identified the elo2⁺ gene as an effective suppressor of the lem2Δbqt4Δ double-deletion phenotype (elo2⁺ in Fig. 1A). No other genes were found in the two rounds of suppressor screening. The S. pombe genome has a gene homologous to elo2⁺ (SPAC1639.01c; elo1⁺), but this gene did not rescue the bqt4Δ lem2 shut-off strain phenotype (elo1⁺ in Fig. 1A). Furthermore, the elo2⁺ gene rescued the bqt4Δ lem2 shut-off strain phenotype in the elo1Δ background (Fig. S2). These results indicate that Elo2 alone can suppress the synthetic lethal phenotype caused by the loss of Lem2 and Bqt4.

We examined the intracellular localization of Elo1 and Elo2. In wild-type cells, Elo1-GFP and Elo2-GFP were located at intracellular membranes, including the nuclear and ER membranes (Fig. 1B). This localization was not affected in mutants depleted of Lem2, Bqt4 or both proteins (lem2Δ, bqt4Δ or bqt4Δ lem2 shut-off) (Fig. 1B). Hence, the intracellular localization of Elo1 and Elo2 at the nuclear and ER membranes is independent of Lem2 or Bqt4.

Humans have seven elo2⁺ homologs (ELOVL1 to ELOVL7) whereas S. cerevisiae has three elo2⁺ homologs (ELO1, ELO2 and ELO3); the amino acid sequence alignment of these homologs is shown in Fig. S3. As summarized in Fig. 2A, fatty acids with >18 carbon atoms are synthesized by a series of carbon chain elongations in the ER membrane. Fatty acid elongases (ELOVL1 to ELOVL7 in mammals, Elo1 to Elo3 in S. cerevisiae) exhibit specificities for fatty acyl-CoA substrates (Ohno et al., 2010; Sassa et al., 2018; Oh et al., 1997; Rössler et al., 2003).

For genetic analysis, we first tried to construct an S. pombe strain with a deletion in elo2⁺ (elo2Δ) as this gene had been annotated as non-essential in a public database PomBase (https://www.pombase.org). However, tetrad analysis showed that elo2Δ is lethal during vegetative growth (Fig. 2B). Thus, we constructed a conditional elo2 shut-off strain, in which elo2⁺ is expressed under the nmt41 promoter. This strain displayed a growth defect when elo2⁺ was shut off in the presence of thiamine; this growth defect was rescued by the expression of elo2⁺, but not by elo1⁺ or lem2⁺ (Fig. 2C). By using the conditional elo2 shut-off strain, we examined whether human and S. cerevisiae homologs of elo2⁺ can rescue the growth defect. In our experiments, three human homologs (ELOVL1, ELOVL3 and ELOVL7) and two S. cerevisiae homologs (ELO2 and ELO3) rescued the growth defect of the S. pombe elo2 shut-off strain (Fig. 2C). Furthermore, the S. pombe elo2⁺ gene, but not the S. pombe elo1⁺ gene, was able to rescue the growth defects of the S. cerevisiae elo3Δ ELO2 shut-off strain (Fig. 2D), whereas the double deletion of ELO2 and ELO3 was lethal in S. cerevisiae. These results indicate that S. pombe Elo2 is an ortholog of human ELOVL1, ELOVL3 and ELOVL7, and S. cerevisiae Elo2 and Elo3.

To understand how Elo2 affects lipid metabolism in the absence of Lem2 and Bqt4, we measured the lipid composition of S. pombe cells using liquid chromatography coupled with tandem mass spectrometry (LC–MS/MS). Most saturated and mono-unsaturated very-long-chain fatty acids are incorporated into ceramides, which are composed of a fatty acid and a long-chain base (Fig. 3A) (Kihara, 2016). The fatty acid moiety of ceramides in S. cerevisiae is almost exclusively C26:0 (Ejsing et al., 2009). Given that amount of fatty acids is kinetically affected by elongation and ceramide synthesis, we measured the amounts of ceramides with the respective number of carbon atoms as the fatty acid metabolites. Our LC-MS/MS measurements showed that C20 phytosphingosine [t20:0; (2S,3S,4R)-2-aminoicosane-1,3,4-triol] is a main long-chain base of ceramides in S. pombe (as previously reported in Garton et al., 2003), and that t20:0/24:0 phytoceramide (a conjugate of C20:0 phytosphingosine and C24:0 fatty acid) is the most abundant species of ceramide in S. pombe (Fig. 3B,C). t20:0/24:0 phytoceramide greatly decreased in the bqt4Δ lem2 shut-off strain, but partially recovered with increased expression of elo2⁺ (Fig. 3B,C). This result suggests that loss of Lem2 and Bqt4 results in C24:0 fatty acid reduction, and that this reduction is partially restored by increased amounts of Elo2.

To determine whether the enzyme activity of Elo2 is required to suppress the loss of Lem2 and Bqt4, we mutated the conserved ¹⁶⁵HxxHH¹⁶⁹ (where x represents any amino acid) Elo2 motif that is required for enzyme activity (Denic and Weissman, 2007) to HxxAH (Elo2-H168A). Expression of Elo2-H168A failed to rescue growth defects in the elo2 shut-off strain and, accordingly, failed to rescue growth defects in the bqt4Δ lem2 shut-off strain, suggesting that the suppression of the lem2Δ bqt4Δ phenotype is directly related to Elo2 function (Fig. 3D,E). Thus, the enzyme activity of Elo2 is likely to be required to suppress the loss of Lem2 and Bqt4.

Various pleiotropic phenotypes of the lem2Δ mutant have been reported, such as slow growth, hydroxyurea (HU) sensitivity, a high loss rate mini-chromosome and defective pericentromeric gene silencing (Xu et al., 2016; Barrales et al., 2016; Tange et al., 2016). Surprisingly, all these phenotypes were rescued by increased expression of elo2⁺ (Fig. 4A-E). In the experiments shown in Fig. 4, elo2⁺ was expressed under the adh11 promoter on the chromosome (see Fig. S4). Cells of lem2Δ displayed slow growth in a YES-rich medium (lem2Δ in Fig. 4A left), but not in EMMG (lem2Δ in Fig. 4A right). The slow growth of lem2Δ was rescued by the increased expression of elo2⁺ (lem2Δ+elo2⁺ in Fig. 4A right). Interestingly, this rescue required Bqt4, as the increased expression of elo2⁺ in the bqt4Δ lem2Δ double mutant resulted in similarly slow growth as in lem2Δ (compare bqt4Δ lem2Δ+elo2⁺ with lem2Δ in Fig. 4A right). Cells of lem2Δ also showed sensitivity to HU when grown in EMMG or YES (Fig. 4B) compared to the rad3Δ control cells. The increased expression of elo2⁺ suppressed HU sensitivity of lem2Δ but not of rad3Δ (Fig. 4B), suggesting that the suppression is not a direct result of the DNA damage repair pathway. Again, this suppression required Bqt4, as the increased expression of elo2⁺ in the bqt4Δ lem2Δ double mutant resulted in the same level of HU sensitivity as in lem2Δ (compare bqt4Δ lem2Δ+elo2⁺ with lem2Δ in Fig. 4B). The high loss rate of a mini-chromosome observed in lem2Δ was abrogated by the increased expression of elo2⁺ when grown in YES or EMMG (Fig. 4C,D). The silencing defect at pericentromeric heterochromatin in lem2Δ was partially rescued by the increased expression of elo2⁺ (compare lem2Δ with lem2Δ+elo2⁺ in Fig. 4E left). Cells of bqt4Δ also showed a slight silencing defect when grown in YES, which was eliminated by the increased expression of elo2 (compare bqt4Δ with bqt4Δ+elo2⁺ in Fig. 4E left). At the same time, the bqt4Δ mutant exhibited a telomere detachment phenotype (Chikashige et al., 2009). However, it was not rescued by elo2⁺ expression (Fig. 4F,G). The results that the increased expression of elo2⁺ rescues various mutant phenotypes suggest that Elo2 is involved in maintaining nuclear integrity.

Because the fatty acid elongase Elo2 suppressed the synthetic lethality of lem2Δ bqt4Δ, we first examined its functions in maintaining the nuclear membrane integrity. To this end, we observed the behaviors of nuclear proteins. In the wild type S. pombe cells, nuclear proteins remained in the nucleus throughout the cell cycle (Fig. 5A). In contrast, transient leakage of nuclear proteins occasionally occurred in lem2Δ (see ‘2 min’ in Fig. 5B) and bqt4Δ (see ‘2-4 min’ in Fig. 5C) and, more strikingly, in the bqt4Δ lem2 shut-off strain (Fig. 5D). In those cells, nuclear proteins often remained in the cytoplasm for a long time (see ‘−12 to −6 min’ and ‘12-20 min’ in Fig. 5D). In the bqt4Δ lem2 shut-off strain, Rna1 (S. pombe RanGAP), which is normally located in the cytoplasm, entered the nucleus at the same time that the nuclear proteins leaked out to the cytoplasm (see ‘20-55 s’ in Fig. S5). Leakage of nuclear proteins was not observed in the elo2 shut-off strain (Fig. S6). These results suggested that there was a loss of the nuclear barrier function in the absence of Lem2 and Bqt4.

To examine the nuclear membrane structure in cells with or without leakage of nuclear proteins, we used correlative light-electron microscopy (CLEM). In normal cells with an intact nuclear membrane, nuclear proteins remained in the nucleus (Fig. 6A,D). In cells exhibiting a moderate level of nuclear protein leakage, invaginations and/or aberrant vacuole-like membranes inside the nucleus were frequently observed (Fig. 6B,D). In cells showing a high level of nuclear protein leakage, ruptures were often present in the nuclear envelope (Fig. 6C,D). Thus, the leakage of nuclear proteins associated with the nuclear envelope ruptures are a possible cause of the synthetic lethal defects of bqt4Δ lem2Δ mutants.

To investigate this possibility, we continuously followed cell divisions of living cells (for live-cell imaging, see Movie 1). Even with transient leakage of nuclear proteins, the cells continued to grow as long as nuclear proteins re-entered the nucleus. However, cells with persistent loss of nuclear proteins often failed to divide. We measured the ratio of the fluorescence signal in the cytoplasm relative to the nucleus as an indication of nuclear protein leakage and plotted it versus the generation time (Fig. 7). In the early stages of shut off (0-3 h after shut off; Fig. 7A), the cytoplasm-to-nucleus fluorescence ratio was low and the generation time was about 3 h (median 3.1 h), indicating the retention of nuclear proteins in the nucleus and normal cell division. As time passed after shut off (measurements at 3-6 h in Fig. 7B and 6-9 h in Fig. 7C), the cytoplasm-to-nucleus ratio and the generation time gradually increased (median of the generation time was 4.5 h and 9.8 h, respectively), indicating that the cell division becomes defective as the leakage of nuclear proteins progresses. A generation time longer than 5 h can be regarded as a cell growth defect. At the later stage (9-12 h after shut off; Fig. 7D), the cytoplasm/nucleus ratio increased, indicating the persistent loss of nuclear proteins and, accordingly, the cell division was defective. The plots in Fig. 7A-D were superimposed on a single plot (Fig. 7E). These results indicate significant positive correlation between the cytoplasm-to-nucleus ratio and the generation time (see legend to Fig. 7E). Taken together, our results suggest that a cause of lethality of the bqt4Δ lem2Δ double-deletion mutant is the leakage of nuclear proteins caused by the loss of the nuclear barrier function.

We then examined whether Elo2 prevents the leakage of nuclear proteins in the bqt4Δ lem2Δ double-deletion mutant. We measured the frequency of cells leaking nuclear proteins by observing GFP-GST-NLS (Fig. 8A). Measurements showed that the frequency of the leakage increased to 35% in lem2Δ cells, and to 10% in bqt4Δ cells compared to 0% in wild type cells cultured in YES; the frequency of nuclear protein leakage in these cells significantly decreased with the addition of elo2⁺ (Fig. 8B, left). Similar effects of elo2⁺ on preventing nuclear protein leakage were also observed in cells cultured in EMMG (Fig. 8B, right). These results suggested that Elo2 suppressed the leakage of nuclear proteins.

Because nuclear protein leakage occurred more frequently and during longer periods in the bqt4Δ lem2 shut-off strain as shown in Fig. 5D, we measured the frequency of nuclear protein leakage in this strain (Fig. 8C,D). As time passed after shut off, the frequency of nuclear protein leakage gradually increased (Fig. 8D, upper left). This frequency decreased with the expression of elo2⁺ (Fig. 8D, upper right), suggesting that the loss of the nuclear barrier function can be recovered by the increased expression of elo2⁺.

The conserved fatty acid elongase Elo2 suppresses the pleiotropic phenotypes of the lem2Δ strain. Some of these phenotypes are nutrition-dependent (observed in rich medium but not in minimum medium), while others are nutrition-independent (observed in rich and minimum media). Nutrition-dependent phenotypes include defective pericentromeric heterochromatin, mini-chromosome loss and slow growth (Barrales et al., 2016; Tange et al., 2016); nutrition-independent phenotypes include HU sensitivity (Fig. 4B in this study; Xu et al., 2016) and bqt4Δ synthetic lethality (Tange et al., 2016). Duplication of the lnp1⁺ gene has been shown to suppress the nutrition-dependent phenotypes of the lem2Δ strain but not the nutrition-independent lethal phenotype of the lem2Δ bqt4Δ strain (Tange et al., 2016). This result is consistent with the fact that lnp1⁺ was not obtained as a lem2Δ bqt4Δ suppressor in this study. Therefore, Elo2 affects Lem2 and Bqt4 function via different pathways to Lnp1.

The N-terminal region of Lem2 plays a role in its positioning at the centromere, whereas its C-terminal region plays a role in silencing at the pericentromeres and telomeres (Barrales et al., 2016). Although The N-terminal region of Lem2 binds to Bqt4, the C-terminal region is sufficient to rescue the synthetic lethality of lem2Δ bqt4Δ (Tange et al., 2016; Hirano et al., 2018). Therefore, the interaction of Lem2-Bqt4 is not necessary for cell viability. Bqt4 plays a role in anchoring telomeres to the nuclear envelope (Chikashige et al., 2009); however, a non-telomeric function of Bqt4 is likely to be involved in the synthetic lethality of lem2Δ bqt4Δ, as discussed previously (Tange et al., 2016; Hirano et al., 2018). Our study also demonstrated that telomere anchoring was not rescued by elo2⁺ expression; therefore, Elo2 is likely to compensate for crucial, currently unknown, functions that overlap between Lem2 and the non-telomeric function of Bqt4.

The nuclear envelope is generally believed to be a stable structure. However, recent studies have revealed that transient ruptures of the nuclear envelope occur in interphase under various circumstances. For example, in cancer cells these transient nuclear membrane ruptures occur particularly during migration (Vargas et al., 2012; Denais et al., 2016), which causes metastasis and, more frequently, in laminopathy cells that have lost the integrity of the nuclear lamina (De Vos et al., 2011; reviewed in Hatch and Hetzer, 2014). It has also been established that nuclear membrane ruptures drive genome instability as chromosomes become exposed to the cytoplasm (Vietri et al., 2015; reviewed in Hatch and Hetzer, 2014; Lim et al., 2016). This is consistent with the genome instability phenotypes observed in S. pombe lem2Δ cells (Tange et al., 2016), and this instability can be suppressed by the increased expression of Elo2, as observed in this study.

Our study also demonstrated that the nuclear membrane ruptures occur frequently in the absence of the nuclear membrane proteins Lem2 and Bqt4 in S. pombe. These transient nuclear ruptures are restored without apparent defects through a pathway mediated by the endosomal sorting complex required for transport (ESCRT) III (Raab et al., 2016). The recruitment of ESCRT-III components, such as CHMP7, to the reforming nucleus in early telophase depends on Lem2 in humans (Olmos et al., 2016; Gu et al., 2017). The S. pombe homolog of human CHMP7, cmp7⁺, genetically interacts with lem2⁺ associated with the ESCRT-III function (Gu et al., 2017). These reports suggest that Lem2 plays a role of resealing the nuclear membrane through the function of ESCRT-III. However, even in the absence of Lem2, Elo2 prevents nuclear protein leakage. Thus, Elo2 probably plays a role independently of ESCRT-III. This is supported by the fact that elo2⁺ is the only one extragenic suppressor of lem2Δ bqt4Δ cells found so far, and cmp7⁺ has not been identified in the suppressor screening.

Our study shows that the defects caused by the loss of Lem2 and Bqt4 are rescued by increased expression of the very-long-fatty acid elongase Elo2. Because Elo2 is essential for cell viability, continuously supplying the very-long-chain fatty acids or the phytoceramides is probably important for maintaining membrane integrity and genome integrity, especially when frequent and persistent ruptures of the nuclear membrane occur – as has been observed in cancer cells. Our finding demonstrates the importance of very-long-chain fatty acids or phytoceramides in nuclear membrane integrity and also in genome stability. Understanding lipid metabolism involving very long-chain fatty acids will provide new insights into the mechanisms protecting the nuclear membrane integrity and the genome stability in many biological events, such as proliferation, differentiation, diseases and cancer development. As very-long-chain fatty acid elongases are widely conserved in eukaryotes, our results in S. pombe is likely to be applicable to a wide range of eukaryotic organisms.

All S. pombe and S. cerevisiae strains used in this study are listed in Table S1. S. pombe strains were maintained in the synthetic Edinburgh minimal medium, containing 5 mg/l sodium glutamate (EMMG) (Tange et al., 2016) or EMMG5S (EMMG supplemented with 225 μg/ml of adenine, uracil, lysine, leucine and histidine), to avoid the undesired genome rearrangements that sometimes occur in lem2Δ cells cultured in the glucose- rich yeast extract with supplements (YES) medium (Tange et al., 2016). For lem2 shut-off experiments, 20 μM thiamine (Nacalai Tesque, Kyoto, Japan) was added to media. S. cerevisiae strains were cultured in a synthetic complete (SC) medium (0.67% yeast nitrogen base) containing 2% D-glucose (SD) or 2% galactose (SG), supplemented with an amino acid mix of Dropout Supplement −His/−Ura (Clontech catalog# 630422).

All plasmids were constructed using the NEBuilder (New England BioLabs, MA) systems, and enzymatic digestion and ligation (TaKaRa, Shiga, Japan) systems, according to the manufacturer's instructions. Gene disruptions in lem2⁺, bqt4⁺, elo2⁺, rad3⁺ and clr4⁺, as well as chromosomal fusion of GFP or mCherry to elo1⁺, elo2⁺ and ish1⁺, were performed using a PCR-based gene targeting method (Wach, 1996; Bähler et al., 1998). The correct disruptions and integrations were confirmed using genomic PCR at the 5′ and 3′ ends. Details related to the construction of the plasmids and yeast strains are described in Tables S1 and S2.

Multicopy suppressors were screened by using two S. pombe genomic DNA libraries, i.e. pTN-F2 (Nakase et al., 2001) and pTN-L1 (Nakamura et al., 2001). In the first round of screening, the S. pombe cYK186-1D (bqt4Δ lem2 shut-off) strain was transformed with the pTN-F2 library. Approximately 2000 transformants were screened on YES plates and 13 suppressor colonies were obtained. We purified plasmids from all the suppressors by using a yeast miniprep method (Nasmyth and Reed, 1980), and determined their DNA sequences: the lem2⁺ sequence was present in five suppressors, the bqt4⁺ sequence in four suppressors, and an identical genomic region (Chr1:2808169-2817505 containing the elo2⁺ sequence) in four suppressors. In the second round of screening, the S. pombe cYK341-5A (bqt4Δ lem2 shut-off) strain with a more-severe growth defect in the lem2 shut-off condition than the cYK186-1D strain was transformed with the pTN-L1 library. Approximately 80,000 transformants were screened and 597 suppressors were obtained. After removing the suppressors containing the lem2⁺ and bqt4⁺ genomic loci by colony PCR, we purified plasmids from the remaining 62 suppressors and identified the elo2⁺ sequence in the plasmids by PCR and DNA sequencing. From the two rounds of screening, we obtained a total of 610 suppressor colonies, comprising 314 lem2⁺-containing colonies, 230 bqt4⁺-containing colonies and 66 elo2⁺-containing colonies.

Tetrad analysis was performed as described previously (Moreno et al., 1991). Briefly, YK455 and cYK278-1A were mixed and cultured on plated malt extract medium at 26°C for 3 days for sporulation. Spores were dissected on plated YES medium using a dissection microscope (MSM300, Singer Instruments), and grown at 30°C for 4 days. Microcolonies formed from each spore were replicated on YES supplemented with 100 μg/ml geneticin (G418 disulfate, Nacalai Tesque, Kyoto, Japan) and EMMG plates, and then grown at 30°C for 2 days.

A mini-chromosome loss assay was performed as described previously (Allshire et al., 1995; Tange et al., 2016). Briefly, cells harboring a mini-chromosome Ch16 (Matsumoto et al., 1987) were precultured in liquid EMMG without supplements, and then cultured at 30°C for 2-4 days on YE or EMMG plates supplemented with 10 μg/ml adenine. For quantification, we counted colonies that were half red (half-sectored colonies) and white colonies. The rate of mini-chromosome loss was expressed as the number of half-sectored colonies per total number of white and half-sectored colonies.

A heterochromatin silencing assay was performed as described previously (Allshire et al., 1995; Barrales et al., 2016). Briefly, the strains harboring the ura4⁺ gene at imr1L (the left innermost repeat region of centromere 1) were precultured in EMMG5S, then plated and grown at 30°C for 2-4 days either on YE supplemented with uracil and adenine, YE with adenine and 1 mg/ml-5′-fluoroorotic acid (5-FOA) (Wako, Osaka, Japan), EMMG with uracil and adenine, or on EMMG with low uracil (50 μg/ml), adenine and 5-FOA.

Intracellular GFP and mCherry fusion proteins in living cells were observed using a DeltaVision microscope system (GE Healthcare Inc.) equipped with a cooled CCD camera (CoolSNAP HQ2, Photometrics, AZ) and a 60× Plan-ApoN SC oil immersion objective lens (NA=1.40, Olympus, Tokyo, Japan), or a DeltaVision OMX system (GE Healthcare Inc.) equipped with an EM-CCD camera (Cascade II, Photometrics, AZ) and a 100× UPlanSApo silicone-immersion objective lens (NA=1.35, Olympus, Tokyo, Japan). The collected data were computationally processed by the ND-SAFIR de-noising software (Boulanger et al., 2010), followed by iterative 3D deconvolution in Priism (Chen et al., 1996). The brightness of the images was adjusted using the Fiji software (Schindelin et al., 2012) without changing the gamma settings. For quantification, original unprocessed images were used.

Time-lapse observation was performed as described previously (Asakawa and Hiraoka, 2009). Briefly, cells were cultured overnight in liquid EMMG or EMMG5S at 30°C to reach logarithmic growth phase. Living cells were immobilized on a coverslip coated with lectin in a 35 mm glass-bottomed culture dish (MatTek, MA) and observed at 30°C in liquid EMMG supplemented with thiamine by using a DeltaVision microscope. For short-term observation in Fig. 5, images were collected at ∼15 focal sections with 0.3 μm interval at each time point. To follow cell divisions of living cells in Fig. 7, images of only GFP-GST-NLS were collected every 5 min at three focal sections to minimize the excitation light during long-term observation.

The nucleus was distinguished by the Ish1-mCherry signal when available, and total GFP fluorescence intensities in the nucleus and cytoplasm were quantified on a single focal plane image of the original unprocessed images by using image processing software ImageJ (http://imagej.nih.gov/ij/). The total fluorescence intensity in the cytoplasm was obtained by subtracting the nuclear fluorescence intensity from that of the whole cell. Following the subtraction of the background fluorescence measured outside the cell, the mean fluorescence intensity per pixel, the nucleus-to-cytoplasm ratio of mean fluorescence intensity, and the total fluorescence intensities in the nucleus, cytoplasm and whole cell were calculated. For quantification of the nuclear protein leakage rate shown in Fig. 8B,D, snapshot images of mitotic cells were selected and used. The leakage rate was represented as the cytoplasm-to-nucleus ratio calculated by using mean fluorescence intensities in the nucleus and cytoplasm. Microsoft Excel was used for calculations.

We defined the time from one cell division to the next as the cell generation time. We used the time-lapse images of the first 12 h to measure the generation time and the next 12 h to observe the cell fate, given that the normal doubling time of S. pombe lies well below 12 h (2.5-3 h in EMMG) (Moreno et al., 1991). If the next cell division did not occur until the end of the observation period, the generation time was counted as longer than 12 h.

Correlative light-electron microscopy (CLEM) imaging was performed as described previously (Asakawa et al., 2010). Briefly, the bqt4Δ lem2 shut-off strain was incubated in liquid EMMG supplemented with thiamine at 30°C for 16 h. Cells were cultured on a dish with a gridded coverslip (ibidi, Martinsried, Germany) and fixed by replacing the medium with a fixative (2% glutaraldehyde in 0.1 M sodium phosphate buffer pH 7.2). After 1 min, the cells were observed under a fluorescent microscope (DeltaVision system), and then further fixed with fresh fixative at 4°C for 2 h. After washing with 0.1 M sodium phosphate buffer, the cells were fixed again in 1.2% KMnO₄ overnight, dehydrated and embedded in epoxy resin. Serial sections of 80 nm were stained with uranyl acetate and lead citrate, and analyzed by using a transmission electron microscope (JEM1400Plus, JEOL, Japan) at 80 kV.

Cells of the bqt4Δ lem2 shut-off strain were precultured overnight in liquid EMMG at 30°C. The cells were transferred into liquid EMMG with or without thiamine, and cultured at 30°C for a further 48 h, then harvested by centrifugation. The cells (1.0×10⁷) were suspended in 100 µl of H₂O and mixed with 375 µl of chloroform:methanol:12 N formic acid (100:200:1, vol:vol:vol). As an internal standard, 5 pmol of t18:0-d₉C16:0 [N-palmitoyl(d₉) D-ribo-phytosphingosine, Avanti Polar Lipids, Alabaster, AL] was added. After vigorous mixing at room temperature for 1 min, lipids were extracted by successive adding and mixing of 125 µl of chloroform and 125 µl of H₂O. Phases were separated by centrifugation (20,000 g at room temperature for 3 min), and the organic (lower) phase was recovered, then dried and dissolved in 50 µl of chloroform/methanol (1:2, vol:vol). Lipids were subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis as described previously (Ohno et al., 2017). Ceramide species composed of C20 phytosphingosine (t20:0) and a non-hydroxy or a hydroxy fatty acid were detected by multiple reaction monitoring, selecting m/z of specific ceramide species [M+H]⁺ at first MS (Q1) and m/z 328.3, which corresponded to the C20 phytosphingosine fragment ion at second MS (Q3) (listed in Table S3).

We thank Dr Kazuya Kabayama (Osaka University) for his insightful discussion, Dr Takeshi Sakuno (The University of Tokyo) for providing plasmids, Ms Chizuru Ohtsuki (Osaka University) for her technical assistance. We obtained the S. pombe genomic libraries from the National Bio-Resource Project, Japan.