Sphingolipids regulate telomere clustering by affecting the transcription of genes involved in telomere homeostasis

Telomeres are organized in a unique structure to protect them from DNA damage signaling and chromosomal end-to-end fusion, and to maintain telomere length. In some organisms, including mammals, flies, nematodes, and yeasts, telomeres are clustered at the nuclear periphery (Chikashige et al., 2006; Crabbe et al., 2012; Ferreira et al., 2013; Gotta et al., 1996; Wesolowska et al., 2013). Telomere clustering is expected to be functionally important, as the precise subnuclear organization of chromosomes impacts upon essential aspects of gene regulation and genome stability. However, the mechanisms and roles of telomere clustering remain poorly understood.

In the budding yeast Saccharomyces cerevisiae, the 32 telomeres of haploid cells cluster into three to eight foci, which are usually enriched at the nuclear periphery under wild-type conditions (Kueng et al., 2013; Wellinger and Zakian, 2012). The telomere anchoring to the periphery is mediated by two redundant pathways; interactions between Sir4, a telomere-associated protein, and Esc1 and Mps3, inner nuclear membrane proteins, or between Yku80 and telomerase and Mps3 (Taddei et al., 2010). In addition to anchoring, telomere clustering requires Mps3 (Antoniacci et al., 2007), which belongs to the conserved SUN (Sad1-UNC84 homology) protein family (Jaspersen et al., 2006), although telomere clustering and anchoring seem to be functionally distinct phenomena (Horigome et al., 2011; Ruault et al., 2011). Sir3 and Mms21 are also required for telomere clustering in budding yeast (Gotta et al., 1996; Hoze et al., 2013; Kueng et al., 2013; Ruault et al., 2011; Zhao and Blobel, 2005). In fission yeast, Bqt1 and Bqt2, which associate with Sad1 have been implicated in meiosis-specific telomere clustering to the spindle pole body (Chikashige et al., 2006). It has also been reported that the ribosome biogenesis factors Ebp2 and Rrs1 mediate telomere clustering through interaction with Mps3, independently of anchoring (Horigome et al., 2011). Interestingly, Mps3 plays a role in regulating lipid homeostasis (Friederichs et al., 2011). Indeed, cells lacking Mps3 exhibit an abnormal composition of sphingolipids and neutral lipids. More importantly, changes in lipid composition suppress the phenotypes of the MPS3-G186K dominant mutant (Friederichs et al., 2011). These studies suggest that telomere homeostasis is linked to lipid metabolism, although the impact of lipid metabolism on telomere organization is still unclear.

Sphingolipids are an important class of lipids of eukaryotic cell membranes, serving not only as building blocks of membranes and specific lipid domains, but also as signaling molecules in a wide variety of biological processes (Breslow and Weissman, 2010; Dickson, 2008; Gault et al., 2010; Hla and Dannenberg, 2012; Montefusco et al., 2014). All sphingolipids contain a backbone of sphingoid bases that are synthesized de novo from serine and palmitoyl-CoA (Breslow and Weissman, 2010; Dickson, 2008; Funato et al., 2002). Sphingoid bases are acylated to form ceramides in the endoplasmic reticulum (ER). Then, ceramides are transported to the Golgi, where ceramides are converted into more complex sphingolipids (Breslow and Weissman, 2010; Dickson, 2008; Funato et al., 2002). In budding yeast, transport of ceramides to the Golgi can be achieved by both vesicular and non-vesicular mechanisms (Funato and Riezman, 2001), and complex sphingolipids such as inositol phosphorylceramide (IPC) and mannosyl IPC (MIPC) are delivered from the Golgi to their final destinations via the vesicular pathway (Funato et al., 2002; Malathi et al., 2004; Schnabl et al., 2005).

Biosynthesis and transport must be tightly regulated to ensure sphingolipid homeostasis. If sphingolipid homeostasis is challenged, the change in lipid composition will be sensed and signal to downstream targets, leading to adjustment and/or adaptation of lipid composition to preserve cellular functions. In yeast, Orm1 and Orm2 proteins, ER transmembrane proteins, have been shown to control sphingoid base and complex sphingolipid synthesis (Aguilera-Romero et al., 2014; Breslow et al., 2010; Shimobayashi et al., 2013). Recent findings indicate that the regulation by Orm proteins is achieved by feedback loops through mechanisms dependent on target of rapamycin complexes (TORC1 and TORC2) (Aguilera-Romero et al., 2014; Berchtold et al., 2012; Shimobayashi et al., 2013). Sch9, an effector of TORC1 is also a regulator of sphingolipid homeostasis (Swinnen et al., 2014). A further level of regulation involves transport steps. Our previous studies have shown that Arv1 is a positive regulator of ceramide transport (Kajiwara et al., 2008). A screen for high-copy suppressors of the arv1Δ mutation identified GPI15, which encodes an enzyme involved in the biosynthesis of the glycosylphosphatidylinositol (GPI) anchor, and we demonstrated that GPI anchor synthesis regulates IPC production (Kajiwara et al., 2008). Recently, it has also been shown that Osh proteins are required to transport ceramides from the ER to the Golgi (Kajiwara et al., 2014). Ncr1, which is the yeast ortholog of mammalian Niemann Pick C1 protein (NPC1), might regulate sorting of complex sphingolipids at post-Golgi compartments (Malathi et al., 2004).

In this study, we identified three genes, DEP1, EBS1 and UPF1 (also known as NAM7) as high-copy suppressors of arv1Δ mutation. DEP1 encodes a component of the histone deacetylase (HDAC) complex Rpd3L (Keogh et al., 2005). EBS1 and UPF1 encode proteins involved in nonsense-mediated mRNA decay (NMD) (Luke et al., 2007). Because deletion of these suppressor genes leads to short telomeres (Askree et al., 2004; Zhou et al., 2000), we tested a possible role for Arv1 in telomere structure and organization. We show that arv1Δ mutation compromises telomere clustering, but does not affect telomere length or the telomere position effect (TPE), the silencing of telomere-proximal genes. Mutations of other genes affecting GPI anchor and sphingolipid synthesis also result in defects in telomere clustering, suggesting that sphingolipids regulate telomere clustering. Moreover, our data indicate that low levels of sphingolipids lead to reduced expression levels of genes involved in telomere homeostasis such as SIR3 and EST3, and that EST3 overexpression rescues the telomere clustering defect in arv1Δ mutant. As loss of Arv1 activates the DNA damage checkpoint, we propose that telomere clustering is crucial for telomere protection.

We have previously found that Arv1 is required for efficient delivery of an early GPI intermediate, glucosaminyl(acyl)phosphatidylinositol (GlcN-acylPI), to the first mannosyltransferase in the ER lumen during GPI anchor synthesis in S. cerevisiae (Kajiwara et al., 2008). To investigate further the function of Arv1 and the physiological roles of GPI anchor synthesis, we screened for multicopy suppressor genes that rescued the growth defect of arv1Δ cells under heat stress. A YEp13 based yeast genomic library was used for screening as described previously (Kajiwara et al., 2008), yielding three independent plasmids. DNA sequences of the plasmids, subcloning of DNA fragments and testing for their ability to suppress the temperature-sensitive phenotype, revealed that overexpression of DEP1, EBS1 and UPF1 could rescue the arv1Δ growth defect at 37°C (Fig. 1A,B).

Deletion of EBS1 or UPF1, as well as the deletion of UPF2 (also known as NMD2) or UPF3, which triggers an NMD defect, confers resistance to rapamycin, the inhibitor of the TOR kinase (Luke et al., 2007). To identify functional relationships between ARV1 and suppressor genes, we measured growth phenotypes of deletion mutants on plates containing rapamycin (Fig. 1C). As reported previously, ebs1Δ and upf1Δ were resistant to rapamycin, whereas arv1Δ, like dep1Δ, did not show rapamycin resistance. These results indicate that the temperature-sensitive growth phenotype of arv1Δ does not result from a defect in NMD pathway.

Deletions of the suppressor genes DEP1, EBS1 and UPF1 lead to short telomeres (Askree et al., 2004; Zhou et al., 2000). Genome-wide analysis of telomere length has also revealed that ARV1 deletion mutant cells have short telomeres (Askree et al., 2004). These results raised the possibility that arv1Δ cells are defective in maintaining telomere structure, and thus the arv1Δ growth phenotype can be rescued by overexpressing genes related to telomere structure or function. To explore this possibility, we first re-examined whether arv1Δ cells exhibit shorter telomeres than wild-type cells. We analyzed the effect of ARV1 deletion on telomere length, with the different genetic backgrounds, BY4742, RH6082 and BY4741 (Fig. 2A), and W303-1B (Fig. 2C), in order to exclude the possibility that the observed result was specific to the genetic background used. As previously reported (Askree et al., 2004), telomeres in tel1Δ or upf1Δ cells derived from BY4742 were shorter than those of wild-type cells (Fig. 2A), whereas telomeres in rif1Δ cells were longer (Fig. 2C). However, no detectable shortening of telomeres in arv1Δ cells compared with the wild-type strains was observed in any genetic background (Fig. 2A,C). Although the reason for the discrepancy in results with arv1Δ cells between the two studies is unclear, it might be explained by different experimental conditions [e.g. differences in cultivation, which was in solid medium (Askree et al., 2004) versus liquid medium (this study), or temperature]. Regardless, our data suggest that the suppression of the arv1Δ phenotype by DEP1, EBS1 and UPF1 is not directly related to telomere length.

In budding yeast, the 32 telomeres of haploid cells have been shown to cluster, forming three to eight foci at the nuclear periphery (Antoniacci et al., 2007; Horigome et al., 2011; Kueng et al., 2013; Ruault et al., 2011; Wellinger and Zakian, 2012). We next examined whether deletion of ARV1 affects the clustering of telomeres. We tagged the telomere repeat binding protein Rap1 at its N-terminus with CFP to visualize the foci (as in Fig. 2B, left) and scored the numbers of telomere-associated Rap1 foci by 3D live-cell imaging, as reported previously (Horigome et al., 2011). In BY-derived background strains, wild-type haploid cells contained 1–5 telomere clusters, whereas arv1Δ cells exhibited increased numbers (Fig. 2B, right), indicating that deletion of ARV1 has an impact on telomeric clustering. This phenotype is not strain-background-specific, because similar results were also obtained in the W303 strains (Fig. 2D).

Given that Arv1 is involved in GPI anchor synthesis (Kajiwara et al., 2008), we tested whether mutations in other genes affecting GPI anchor synthesis have effects on telomere structure. As shown in Fig. 2C,D, neither gpi2-7 nor gpi3-10 affected telomere length, whereas both mutants displayed an increased number of Rap1 foci, suggesting that GPI anchor synthesis is required for maintaining telomere clustering.

GPI anchor synthesis controls not only transport of GPI-anchored proteins but also synthesis of sphingolipids (Kajiwara et al., 2008). Indeed, mutants defective in GPI anchor synthesis exhibit a defect in incorporation of ceramides into IPC, one of the complex sphingolipids of yeast. To test whether reduced formation of IPC affects telomere clustering, we treated wild-type cells with aureobasidin A (AbA) and examined telomere clustering. AbA is a potent inhibitor of IPC synthase, causing decreased levels of complex sphingolipids and increased levels of ceramides (Kajiwara et al., 2012). We found that the number of Rap1 foci was increased by treatment with AbA (Fig. 3A). We next tested whether complex sphingolipid levels contribute to maintenance of telomere clustering, by quantifying the number of telomere clusters in the lcb1-100 strain. This mutant is defective in serine palmitoyltransferase catalyzing the first step of sphingolipid synthesis and results in reduced levels of all sphingolipids (Zanolari et al., 2000). As shown in Fig 3B, the number of Rap1 foci was remarkably increased in the lcb1-100 mutant cells compared with the wild-type cells, implying that the reduced sphingolipid levels contribute to the dispersion of telomere clusters. Thus, we conclude that complex sphingolipids are required to maintain telomere clustering.

Searching for the mechanism by which the level of sphingolipids affects telomeric clustering we identified genes that are associated with sphingolipid-mediated telomere dispersion. We analyzed the genome-wide gene expression profile of arv1Δ mutant and compared it to the profile of wild-type cells. Expression of 438 known genes was misregulated by more than twofold upon deletion of ARV1 (265 upregulated, 173 downregulated) (supplementary material Fig. S1A,B; Tables S2–S4). A total of 62 genes were upregulated by more than threefold (supplementary material Tables S4 and S5), whereas 45 genes were downregulated by more than threefold (supplementary material Tables S4, S6). The functional categorization of genes misregulated more than twofold revealed that genes involved in protein synthesis (28%), metabolism and lipid (13%), glycosylation and cell wall (12%), and transport (11%) were upregulated in the arv1Δ mutant (supplementary material Fig. S1A), whereas genes involved in metabolism and lipid (32%), transport (16%) were markedly downregulated (supplementary material Fig. S1B).

Although the experiments were carried out under different culture conditions (SD versus YPD medium) and with different genetic backgrounds (RH6082 versus W303), we compared our data with the previously reported microarray data from the arv1Δ mutant (Shechtman et al., 2011). As shown in supplementary material Fig. S1C,D, there were 33 genes whose changes in expression levels (greater than twofold) overlapped in the two studies. The upregulated and downregulated groups included 24 and nine genes, respectively. The overlapping genes, with more than a threefold increase were GSC2, MCD4, RTA1, SIL1 and YLR104W (also known as LCL2) (supplementary material Fig. S1C, underlined). The proteins encoded by GSC2, MCD4 and YLR104W are involved in cell wall integrity, and MCD4 encodes a protein essential for GPI anchor synthesis. RTA1 encodes a protein involved in 7-aminocholesterol resistance, and SIL1 encodes a nucleotide exchange factor for the ER luminal chaperone Kar2, which has a protective role in response to ER stress. The increased expression levels of these genes are in good agreement with the functions of Arv1p in GPI anchor synthesis, sterol and ER homeostasis (Kajiwara et al., 2012, 2008; Shechtman et al., 2011; Tinkelenberg et al., 2000), and the increased MCD4 levels in the arv1Δ mutant suggest that GPI anchor synthesis might signal in a feedback loop to Mcd4. By contrast, the overlapping genes with more than a threefold decrease in transcription included RPI1 and SST2. RPI1 encodes a transcription factor, which is important for ethanol tolerance. SST2 encodes a negative regulator of the pheromone response. MFA2, encoding the yeast mating pheromone a-factor, which induces pheromone response pathway, was also reduced upon deletion of ARV1 (supplementary material Fig. S1D). It is notable that CLN1 and CLN2, which encode G1 cyclins involved in regulation of the cell cycle, were strongly upregulated in the arv1Δ mutant (supplementary material Table S5). Because G1 cyclins negatively regulate the pheromone response pathway (Bloom and Cross, 2007; Doncic et al., 2011), the increased expression of CLN1 and CLN2, and decreased expression of MFA2 might contribute to the defects in pheromone-induced G1 cell cycle arrest and mating that have been observed in the arv1Δ mutant (Villasmil et al., 2011). A particularly intriguing finding is the downregulation of EST3 (supplementary material Table S6), which encodes a non-catalytic component of telomerase. Another non-catalytic subunit of telomerase, Est1, and the catalytic subunit Est2 have been shown to anchor telomeres to the nuclear envelope (Schober et al., 2009). Est1 is known to interact with the SUN domain protein Mps3, which localizes to the peripheral nuclear envelope and is required for the formation of telomere clusters (Antoniacci et al., 2007). These findings might support the idea that telomerase is required for telomere clustering. However, it is unclear whether reduced expression of EST3 is responsible for sphingolipid-mediated telomere dispersion in arv1Δ mutant cells.

We wondered whether transcription of EST3 was downregulated in a sphingolipid-dependent manner. To address this question, we analyzed the expression profile of the arv1Δ sur2Δ double mutant, because the two mutations of ARV1 and SUR2 synergistically affect the incorporation of ceramide into IPC (Fig. 4A) and the sensitivity to AbA (Fig. 4B), which seems to reflect the steady-state levels of complex sphingolipids (Kajiwara et al., 2014). SUR2 is a gene responsible for hydroxylation of sphingoid bases, and as reported previously (Guan et al., 2009), deletion of SUR2 led to a significant reduction in incorporation into IPC (Fig. 4A). Microarray analysis showed that three genes were upregulated (two genes>twofold; one gene>threefold) and 13 genes were downregulated (12 genes>twofold; 1 gene>threefold) in sur2Δ mutant (supplementary material Tables S4, S7 and S8). A total of 296 genes were upregulated more than twofold in the arv1Δ sur2Δ double mutant and 118 genes were upregulated more than threefold (supplementary material Tables S4 and S9), whereas 600 genes and 243 genes were downregulated more than twofold and threefold, respectively (supplementary material Tables S4 and S10). Fig. 4C,D shows the genes whose expression levels changed more than threefold in arv1Δ cells (compared to the wild-type cells), and then further changed more than 1.5-fold in arv1Δ sur2Δ cells (compared to the arv1Δ cells). The increased group included five genes, ALD6, CYC1, MCD4, MET14 and SEC72 (Fig. 4C). By contrast, the decreased group included 23 genes, and ∼30% of the genes encoded proteins with undefined functions (Fig. 4D). EST3 was also included in this decreased group, suggesting that transcription of EST3 is downregulated by the reduced levels of complex sphingolipids.

Other genes involved in telomere homeostasis such as TEN1, YKU80, SIR3 and TEL1 were also repressed in both arv1Δ and arv1Δ sur2Δ cells (Fig. 4E–H). Mutations in YKU80 or SIR3 have been shown to affect telomere clustering (Gotta et al., 1996; Laroche et al., 1998; Ruault et al., 2011). Although not statistically significant (P=0.11, arv1Δ versus arv1Δ sur2Δ), it is noteworthy that like for EST3, SIR3 downregulation in the arv1Δ mutant was further exacerbated in the absence of Sur2. For the rest, statistically significant repression (P<0.05) in arv1Δ and arv1Δ sur2Δ cells or sphingolipid-dependent repression of transcription was not observed. Interestingly, POL32 was upregulated in the sur2Δ or arv1Δ mutant cells. Although the increases were not statistically significant, POL32 upregulation might be required for adaptation of cell to low sphingolipid levels. Collectively, these results suggest that sphingolipids regulate expression of telomere genes, specifically EST3, TEN1, YKU80, SIR3 and TEL1 at the transcriptional level.

As EST3 mutations cause telomere shortening and senescence (Askree et al., 2004; Lendvay et al., 1996), and EST3 was downregulated in arv1Δ mutant cells (Fig. 4D), we next addressed the functional consequences of such expression change on the phenotypes of the arv1Δ mutant. As shown in Fig. 5A, when overexpressed constitutively under the control of the GPD promoter, Est3 rescued the arv1Δ growth defect at 37°C. Furthermore, Est3 overexpression suppressed the telomere clustering defect in the arv1Δ mutant (Fig. 5B). These results suggest that the two phenotypes in the arv1Δ mutant might be due to a decrease in sphingolipid-dependent expression level of Est3. However, because est3Δ cells can grow normally at 37°C (Fig. 5C), it is likely that the regulation of EST3 expression by sphingolipid signaling is not essential for maintaining cell growth upon heat stress. Importantly, deletion of EST3 in combination with arv1Δ resulted in a synthetic growth defect (Fig. 5C), suggesting that Est3 and Arv1 function in parallel pathways to allow cells to grow at high temperatures (Fig. 5D). This model is consistent with the data showing that est3Δ has a high co-fitness value with the arv1Δ strain (supplementary material Fig. S2A). Interestingly, est3Δ cells showed high co-fitness values to strains defective in IPC synthesis such as aur1Δ and kei1Δ, which also have high values with arv1Δ (supplementary material Fig. S2B), supporting the hypothesis that there is a functional interaction between Est3 and the IPC level.

Because overexpression of Est3 rescued the telomere clustering defect of the arv1Δ mutant, we examined whether this effect was due to an increase in telomerase action. To test this, we overexpressed Est1 in the arv1Δ mutant and quantified the number of telomere clusters. Indeed, we observed that overexpression of Est1 suppressed the telomere clustering defect (Fig. 5B). Moreover, overexpression of Est1 resulted in a suppression of the temperature sensitivity of arv1Δ mutant strain (Fig. 5A). Although the temperature sensitivity of the arv1Δ mutant strain was not suppressed by overexpression of Est2, the restoration of telomere clustering was observed upon Est2 overexpression (data not shown). Collectively, these data indicate that the suppression can be achieved by a telomerase-dependent action.

Deletion of ARV1 resulted in telomere dispersion, but did not impact upon telomere length. In addition, the deletion of ARV1 did not compromise silencing of telomere-proximal genes (called telomere position effect, TPE), because there were no significant changes in expression of subtelomeric genes such as ADH4, COS8 and YAR073W in arv1Δ mutant cells (supplementary material Fig. S3). Dysfunctional (uncapped or short) telomeres have been shown to elicit the DNA damage response pathway, leading to accumulation of single-stranded DNA (ssDNA), activation of the DNA damage checkpoint (Dewar and Lydall, 2010; Longhese, 2008; Maringele and Lydall, 2002; Teo and Jackson, 2001; Zubko et al., 2004) and cell cycle arrest or delay (IJpma and Greider, 2003; Pang et al., 2003; Qi et al., 2008, 2003). Therefore, we wanted to know whether the slow growth of arv1Δ mutant was due to activation of a telomeric DNA damage checkpoint. Loss of Exo1, which is a DNA exonuclease and contributes to the generation of ssDNA at uncapped telomeres, leads to suppression of the DNA damage checkpoint activation and growth defects in strains with impaired telomere integrity (cdc13-1, yku70Δ, ten1-101) (Dewar and Lydall, 2010; Maringele and Lydall, 2002; Xu et al., 2009; Zubko et al., 2004). We generated diploid strains heterozygous for exo1Δ and arv1Δ mutations, which we sporulated and dissected tetrads from, and then determined the colony size and genotypes after spores were allowed to form colonies for several days. As shown in Fig. 6A, arv1Δ exo1Δ double mutants grew better than arv1Δ mutants on the YPD dissection plate. It is known that Mre11 functions redundantly with Exo1 during DNA resection (Garcia et al., 2011; Moreau et al., 2001). Thus, we also investigated whether the absence of Mre11 restores the slow growth of arv1Δ cells. arv1Δ mre11Δ double mutant cells formed larger colonies than arv1Δ cells (Fig. 6B). Because, unlike mre11Δ (Garcia et al., 2011), arv1Δ cells were not sensitive to methyl methanesulphonate (MMS), a DNA damaging agent (Fig. 6C), the slow growth phenotype of arv1Δ cells is not due to cell growth arrest or death by a defect in DNA double-strand break (DSB) repair. We therefore conclude that slow growth of arv1Δ cells results from DNA damage checkpoint activation, which is most likely caused by unprotected telomeres that occur upon impaired telomere clustering. Consistent with this, we observed that a DNA damage checkpoint protein fused to GFP, Ddc2–GFP, showed a diffuse nuclear localization in wild-type cells but that it formed subnuclear foci in arv1Δ cells (Fig. 6D,E), with an average of 1.3 foci per nucleus, which is a similar number to the number of DNA damage sites (Melo et al., 2001). The formation of Ddc2 foci in arv1Δ cells was suppressed by overexpression of Est3, suggesting that the phenotype of arv1Δ is due to a telomere defect.

We have previously found that inhibition of complex sphingolipid synthesis and mutants defective in GPI anchor synthesis stimulate ER-stress-mediated and caspase-dependent apoptosis (Kajiwara et al., 2012). In addition, mutations in gpi genes reduce chronological lifespan (Kajiwara et al., 2012). Dysfunctional telomeres are known to be associated with senescence and chronological aging (Lendvay et al., 1996; Ponnusamy et al., 2008; Qi et al., 2008; Schober et al., 2009). To determine whether telomeres play a protective role in the cell death induced by imbalanced sphingolipid metabolism, we tested the effect of overexpression of telomerase subunits on sensitivity of arv1Δ cells to AbA. We found that overexpression of EST1 or EST3 suppressed the sensitivity of arv1Δ cells (Fig. 7A). An increased loss of cell viability by ARV1 deletion upon AbA treatment was also rescued by overexpression of EST1 or EST3 (data not shown). These results suggest that telomerase antagonizes the apoptotic signaling. This does not seem to be due to the action of telomeres on the GPI anchor or sphingolipid synthesis because defects in GPI-anchored protein maturation (Fig. 7B) and IPC synthesis (data not shown), or the hypersensitivity arv1Δ cells to Calcofluor White (data not shown) were not altered by the overexpression of EST1 or EST3. Furthermore, we previously demonstrated that deletion of YCA1, which encodes a yeast metacaspase (Madeo et al., 2002), confers resistance to AbA, and suppresses the hypersensitivity of arv1Δ cells to AbA and the increased apoptotic cell death in arv1Δ cells upon AbA treatment (Kajiwara et al., 2012). However, the telomere clustering defect in arv1Δ cells was not rescued by YCA1 deletion (Fig. 7C), indicating that telomere clustering defect is not caused by the cell death signals from Yca1 induced by impaired sphingolipid metabolism. This is consistent with a model that telomere clustering is epistatic to Yca1-dependent cell death (Fig. 7D).

In this study, we have shown in two different genetic backgrounds, BY and W303, that arv1Δ mutation compromises telomere clustering. Because mutations in other GPI anchor synthesis genes, and lcb1-100 mutation and AbA treatment, which affect sphingolipid metabolism, led to defects in telomere clustering, we conclude that appropriate complex sphingolipid levels are required for telomere clustering. Furthermore, we provide evidence that mutations leading to defective complex sphingolipid synthesis downregulate the expression of genes involved in telomere homeostasis, such as SIR3 and EST3, and that overexpression of EST3 rescues the telomere clustering defect in the arv1Δ mutant. These results suggest that sphingolipids regulate telomere clustering at the transcriptional level. Loss of Arv1 does not affect telomere length and the TPE, but leads to activation of DNA damage checkpoint. In addition, EST3 overexpression counteracts apoptotic signals triggered by impaired sphingolipid metabolism. Thus, we propose that telomere clustering plays a crucial role in the protection of chromosome ends.

The screen for multi-copy suppressor genes that rescue the temperature-sensitive growth phenotype of arv1Δ mutant identified three genes (DEP1, EBS1 and UPF1) and telomerase genes. Because DEP1 expression is significantly reduced in the arv1Δ mutant (supplementary material Fig. S1E), the reason for the suppression of growth phenotype appears to be simply the dosage compensation by overexpression of DEP1. Dep1 is a component of the Rpd3L histone deacetylase (HDAC) complex, and like mutations in other genes (e.g. RPD3, SIN3, UME1, SPA30 and PHO23) of the Rpd3 complex, mutation of DEP1 displays a thermosensitive growth phenotype (Ruiz-Roig et al., 2010). The Rad3 complex is known to be involved not only in cell adaptation during the heat stress response but also in regulating heterochromatic boundary that restricts the spread of gene silencing (Ehrentraut et al., 2010; Zhou et al., 2009). In fact, deletion of DEP1 reduces the transcription of subtelomeric genes. It is conceivable that the functions of Rad3 complex might be linked to the temperature-sensitive growth phenotype of arv1Δ mutant. However, deletion of ARV1 did not impact upon the TPE (supplementary material Fig. S3). Interestingly, rpd3 and sin3 were identified as suppressor mutations of Ca²⁺-sensitive csg2Δ mutant, which is defective in MIPC synthesis (Beeler et al., 1998), suggesting a connection between sphingolipids and the Rpd3 complex. Thus, the suppression by DEP1 overexpression might be associated with a Rad3-complex-mediated regulation of sphingolipid synthesis.

Deletion of ARV1 did not significantly affect the levels of EBS1 or UPF1 (supplementary material Fig. S1F), suggesting that the suppression of the temperature-sensitive growth defect in arv1Δ mutant caused by overexpression of EBS1 or UPF1 is not due to restoration of their functional impairments. Because expression of genes encoding telomerase subunits and regulators, such as EST1, EST2, EST3, TEN1 and STN1, has been shown to be upregulated by loss of the NMD pathway (such as in the upf1Δ mutant) (Addinall et al., 2011; Dahlseid et al., 2003; He et al., 2003), and because Stn1 is a negative regulator of telomerase action (Addinall et al., 2011; Dahlseid et al., 2003; Puglisi et al., 2008), the suppression of phenotypes in arv1Δ mutant by overexpression of EBS1 or UPF1 can be explained by a decreased expression of STN1 through NMD. This is consistent with our observation that EBS1 overexpression rescued not only the temperature-sensitive growth defect (Fig. 1B and Fig. 5A) but also telomere clustering defect in arv1Δ mutant (Fig. 5B). However, our northern blot analysis showed that steady-state levels of STN1 mRNA were not decreased in arv1Δ strains when EBS1 or UPF1 was overexpressed (data not shown). Thus, alteration of Stn1 protein levels might not be responsible for the restoration of phenotypes of arv1Δ cells. Given that NMD might control other protein levels, there is another possible interpretation. Recently, it has been reported that failure of quality control, in order to remove misfolded proteins from the plasma membranes during heat stress, is toxic to the cell (Zhao et al., 2013). Here, we found that deletion of ARV1 upregulated two genes, MUP1 (2.65-fold±1.73, P=0.174) and HXT3 (1.35-fold±0.05, P<0.001, mean±s.d., n=3), which encode integral membrane proteins known to be rapidly degraded in response to heat stress to protect cells from their toxic accumulation at plasma membrane (Zhao et al., 2013). Given that it is also expected that loss of sphingolipids exacerbates the proteotoxic stress because of aberrant plasma membrane lipid composition, it is possible that an increased proteotoxic stress contributes to the phenotypes of arv1Δ mutant cells and might be alleviated by reduction of integral membrane protein levels by NMD.

The mechanism by which impaired sphingolipid metabolism affects expression of genes involved in telomere homeostasis remains unknown. EST3 expression is upregulated upon deletion of SFP1 (Cipollina et al., 2008). Because Sfp1 mediates TORC1 signals to promote ribosome biogenesis (RiBi) and ribosomal protein gene expression (Loewith and Hall, 2011), it is possible that sphingolipid-dependent repression of EST3 expression might result from an increased Sfp1-dependent regulation of gene transcription mediated by TORC1. This is consistent with the observation that ARV1 deletion upregulates RiBi and ribosomal protein expression (supplementary material Fig. S4A,E). Moreover, inhibition of TORC1 by rapamycin has shown to upregulate expression of genes involved in regulation of telomere structure, such as SIR3, CDC13 and ESC1 (Hardwick et al., 1999), suggesting that TORC1 and Sfp1 might regulate telomere homeostasis at the transcriptional level. However, because arv1- and sur2-null mutations have no additive effect on the expression of RiBi and ribosomal proteins (supplementary material Fig. S4D,H), the increased RiBi and ribosomal protein expression in the arv1Δ mutant would occur through sphingolipid-independent mechanisms. We found that ARV1 deletion downregulates DEP1 expression in a sphingolipid-independent manner (supplementary material Fig. S1E). As the Rpd3L HDAC complex is implicated in the repression of RiBi and ribosomal protein genes (Huber et al., 2011), the increased RiBi and ribosomal protein expression levels observed in arv1Δ mutant might be due to a decreased expression level of DEP1.

TORC1 might play positive roles in regulating telomere homeostasis at the transcriptional or translational level. Intriguingly, although YKU70 or YKU80 expression is not affected by rapamycin treatment (Hardwick et al., 1999), the protein levels of Yku70 and Yku80 are reduced in the presence of rapamycin (Ungar et al., 2011). In mammals, it has been reported that rapamycin causes a decrease in the mRNA level of the human TERT gene, which encodes the catalytic subunit of telomerase (Zhou et al., 2003) and is downregulated by ceramides (Patwardhan and Liu, 2011). Recent studies have suggested that impaired sphingolipid metabolism attenuates TORC1 activity (Mousley et al., 2012). Here, arv1Δ cells showed decreased expression level of EGO3, which encodes a component of EGO complex that interacts and activates TORC1 (Loewith and Hall, 2011), and arv1- and sur2-null mutations had an additive effect on the EGO3 transcription (supplementary material Fig. S4I), suggesting a sphingolipid-dependent regulation of TORC1 activity. Consistent with this, we found that ARV1 deletion conferred significant hypersensitivity to rapamycin, compared with wild-type (data not shown). Functional links between sphingolipid metabolism, TORC activity and the transcription of genes involved in telomere homeostasis remain to be determined.

Although detailed mechanisms on how sphingolipids regulate the expression levels of genes involved in telomere homeostasis currently remain unknown, this relationship might be important with regard to telomere clustering. Like EST3, SIR3 expression is reduced in arv1Δ mutant cells and this appears to occur through a sphingolipid-dependent mechanism (Fig. 4E). Mutations in SIR3 do affect telomere clustering (Gotta et al., 1996), and Sir3 has been proposed to stimulate trans-interactions between telomeres, independently of silencing or anchoring to the nuclear periphery (Ruault et al., 2011). Because telomere clustering is determined by the cellular amount of Sir3 (Hoze et al., 2013; Ruault et al., 2011), the reduced expression of SIR3 might be primarily responsible for the telomere clustering defect in arv1Δ mutant cells.

In addition to Sir3, the integral nuclear membrane SUN domain protein Mps3 contributes to formation of telomere cluster (Antoniacci et al., 2007). Mps3 has been shown to interact with Est1, raising the possibility that telomerase might regulate telomere clustering by anchoring telomeres at the nuclear periphery (Schober et al., 2009). This model is consistent with our results, which indicate that an increased action of telomerase rescues the telomere clustering defect. Interestingly, besides its role in telomere anchoring, Mps3 plays a role in regulating the lipid composition in the cell. Indeed, cells lacking Mps3 function have aberrant lipid profiles, including a decrease in IPC, and an increase in MIPC and triacylglycerols (Friederichs et al., 2011). As changes in lipid composition suppress the abnormal nuclear morphology of the MPS3-G186K dominant mutant (Friederichs et al., 2011), it is possible that nuclear lipid composition influences the function of Mps3. Recent studies have also shown that inhibition of IPC synthesis by AbA treatment led to an abnormal nuclear morphology (Kajiwara et al., 2012). Thus, the impaired Mps3 function due to changes in nuclear lipid composition could also account for the telomere clustering defect seen in arv1Δ cells. Further studies on the functional relationships between nuclear lipid composition and telomere anchoring might elucidate the roles of sphingolipids in telomere clustering.

Telomere clustering is defective in arv1Δ mutant cells, but TPE is normal, suggesting that telomere clustering is dispensable for TPE. This is consistent with previous studies indicating that telomere clustering and TPE are separable phenotypes (Ruault et al., 2011). The relationship between telomere clustering and anchoring, as well as telomere clustering and length, appear to be also functionally distinct phenomena. It has been shown that the ribosome biogenesis factors Ebp2 and Rrs1 are required for telomere clustering but not for telomere anchoring (Horigome et al., 2011), and that telomere length is not affected by overexpression of a soluble N-terminal domain of Mps3 that disrupts telomere anchoring (Schober et al., 2009). Consistently, we observed no significant alteration in telomere length in arv1Δ mutant cells. Importantly, our results suggest that the arv1Δ mutant activates an Exo1- and MRX (Mre11–Rad50–Xrs1)-dependent DNA damage checkpoint, and that overexpression of telomerase subunits counteracts apoptotic signals triggered by impaired sphingolipid metabolism. As dysfunctional telomeres have been shown to result in activation of the DNA damage checkpoint (Dewar and Lydall, 2010; Longhese, 2008; Maringele and Lydall, 2002; Teo and Jackson, 2001; Zubko et al., 2004), cell cycle arrest in G2/M (IJpma and Greider, 2003; Pang et al., 2003; Qi et al., 2008) and apoptotic cell death (Qi et al., 2008, 2003), it is tempting to speculate that the telomeric regions in the arv1Δ mutant are not protected from DNA damage checkpoint signaling. Previous observations have shown that a reduction in the amount of complex sphingolipids caused by AbA treatment or acc1 mutation, but not lcb1-100 mutation, induced cell cycle G2/M arrest (Al-Feel et al., 2003; Endo et al., 1997; Jenkins and Hannun, 2001) and that the arv1Δ mutant exhibited a synthetic growth defect with the cdc13-1 mutant leading to uncapped telomeres (Addinall et al., 2008) are consistent with the hypothesis. Therefore, although the mechanisms are largely unknown, telomere clustering may play a crucial role in telomere protection. It is also possible that telomere clustering might be required for efficient repair of DNA double strand breaks in subtelomeric region (Therizols et al., 2006). However, we consider this unlikely because the arv1Δ mutant showed a sensitivity to MMS that was similar to that of the wild-type cells.

All strains used in this study are listed in supplementary material Table S1. Yeast cultivations, genetic manipulations and strain construction were carried out as described previously (Kajiwara et al., 2008). S. cerevisiae wild-type strains BY4741 and BY4742, and related single deletion strains were purchased from Open Biosystems (Huntsville, AL). FKY3177-3180 strains were generated by crossing the BY4742-derived est3 deletion and FKY671, followed by sporulation and subsequent dissection of the spores. FKY1751 and FKY1752 were generated from W303-1A strain by the PCR-based one-step gene replacement method (Baudin et al., 1993). FKY1784-1787 and FKY2303 strains were generated by crossing FKY245 and FKY1752. The deletion strains were confirmed by PCR. FKY3389 was generated by PCR-mediated tagging of DDC2 with a GFP–HIS fragment in W303-1B, and FKY3423 and FKY3424 were created by crossing FKY3389 and FKY2303. Strains FKY2353 and FKY2362 were generated by crossing DL2828 with W303-1A, and DL2829 with W303-1A, respectively. RH1800, RH3809, RH6082, W303-1A, W303-1B, DL2828 and DL2829 yeast strains were kind gifts from Dr H. Riezman (University of Geneva, Geneva , Switzerland), Dr Y. Jigami (National Institute of Advanced Industrial Science and Technology, Tsukuba city, Ibaraki, Japan) and Dr P. Orlean (University of Illinois, Urbana, IL).

The genetic screen for multicopy suppressors rescuing the temperature-sensitive growth phenotype of arv1Δ cells was carried out as described previously (Kajiwara et al., 2008). Genomic DNA fragments containing DEP1, EBS1 and UPF1 were isolated from the multicopy suppressor YEp13 plasmids. Because multiple genes were present in the fragment containing DEP1 or EBS1 in the original library, the DEP1 and EBS1 genes were subcloned into pRS425 and YEp13 plasmids, respectively, and the resulting plasmids, pRS425/SNY8ΔC-DEP1 and YEp13/EBS1 were used to determine their abilities to suppress the temperature-sensitive growth defect of arv1Δ cells. To construct plasmids overexpressing EBS1, EST1 and EST3, the DNA fragments containing their open reading frames were amplified by PCR and cloned into pRS426GPD (2 μ, URA3) and pRS424GPD (2 μ, LEU2), which contain the GPD promoter and the CYC terminator, to obtain pRS426GPD/EBS1 (FKP434), pRS426GPD/EST1 (FKP435), pRS426GPD/EST3 (FKP437) and pRS424GPD/EBS1 (FKP605), pRS424GPD/EST1 (FKP606), pRS424GPD/EST3 (FKP608), respectively.

Temperature and drug sensitivity assays were performed as described previously (Kajiwara et al., 2012). Briefly, cells were grown to log phase at 25°C in synthetic dextrose (SD) minimal medium supplemented with the appropriate nutrients to select for plasmids (Kajiwara et al., 2008), and were serially 1:5 diluted in sterile water. Then, 8 µl of each fivefold serial dilution were spotted onto SD plates and incubated at the indicated temperatures. Drug sensitivity tests were carried out by spotting diluted yeast cultures on SD plates containing aureobasidin A (AbA) (Takara, Japan) or on YPD plates containing rapamycin (Santa Cruz Biotechnology, USA) followed by incubation at 25°C for 3–5 days.

Telomere length was measured as described previously (Craven and Petes, 1999). Yeast genomic DNA was prepared according to standard protocols, and digested with XhoI. The DNA fragments were separated by electrophoresis and transferred to Hybond N+ membranes. Then, the membranes were probed with a digoxigenin-labeled Y′ probe (Craven and Petes, 1999). This probe was prepared by PCR amplification using the template yeast genomic DNA and the primers Y′-F (5′-ACACACTCTCTCACATCTACC-3′) and Y′-R (5′-TTGCGTTCCATGACGAGCGC-3′), and visualized using anti-digoxigenin antibody conjugated to alkaline phosphatase and CDP-Star substrate.

For the telomere clustering assay, cells were transformed with plasmid pRS306-CFP-RAP1 after digestion with SpeI, as described previously (Horigome et al., 2011). The transformants were grown overnight at 25°C in SD lacking uracil and observed with a fluorescence microscope (Olympus BX51, Japan). The number of CFP–Rap1 foci was counted in deconvolved 3D stacks, typically 24 stacks of 0.2 μm, and the percentage of nuclei was determined by counting at least 100 nuclei for each sample.

To assess Ddc2–GFP foci, cells were grown overnight to log phase at 30°C and viewed under a fluorescence microscope. Images were acquired with identical exposure time.

In vivo [³H]myo-inositol labeling was performed as described previously (Kajiwara et al., 2008). Cells grown overnight in semi-synthetic medium, SDYE at 25°C were resuspended in SD medium without inositol, and labeled with 10 μCi of [³H]myo-inositol (PerkinElmer Life Sciences, Japan) at 25°C for 2 h. The reaction was stopped by placing the mixture on ice and by adding 10 mM NaF and 10 mM NaN₃. The cells were washed with cold water, and lipids were extracted. Half of the sample was subjected to mild alkaline hydrolysis to deacylate glycophospholipids as described previously (Kajiwara et al., 2008). The pooled organic phases were desalted by partitioning with n-butanol, and lipids were dried under nitrogen. The lipids were analyzed by thin-layer chromatography (TLC) using solvent system I, chloroform, methanol and 0.25% KCl (55:45:10, vol/vol/vol). Radiolabeled lipids were visualized and quantified on an FLA-7000 system (Fujifilm, Japan).

Total RNA was prepared from cells grown in SD medium, and poly(A)⁺ RNA was purified using the Oligotex-dt30 mRNA purification kit (Takara, Japan). Poly(A)⁺ RNA was amplified, biotin-labeled and hybridized to oligonucleotide arrays (GeneChip YG-S98, Affymetrix), and hybridization intensities were analyzed as described previously (Tamaki et al., 2005). The experiments were independently carried out three times. Genes whose expression level was increased or decreased in comparison with the wild-type strain are listed in supplementary material Tables S2, S3 and S5–S10, based on the following criteria: the mean change of three experiments was more than twofold (supplementary material Tables S2, S3, S7–S10) or threefold (supplementary material Tables S5, S6). Mean value and standard deviation (s.d.) were obtained from three independent experiments. P-values for the t-test are *P<0.05; **P<0.01; ***P<0.001. For genes listed in supplementary material Tables S2, S3 S5 and S6, the reliability of the binding ratio was also judged based on the change in P-value (increase; P<0.003, decrease; P>0.996).

Cell lysate was prepared as described previously (Urban et al., 2007). Proteins were separated on SDS-PAGE and analyzed by western blotting. Blots were probed with rabbit polyclonal antibodies against Gas1 and CPY (gifts from Dr Riezman, University of Geneva, Geneva, Switzerland) and a peroxidase-conjugated affinity-purified anti-rabbit-IgG antibody (Sigma, St Louis, MO).

We thank Howard Riezman (University of Geneva) for the RH1800, RH3809 and RH6082 strains and antibodies against Gas1 and CPY; Yoshifumi Jigami (National Institute of Advanced Industrial Science and Technology) for the W303-1A and W303-1B strains; Peter Orlean (University of Illinois) for the DL2828 and DL2829 strains. We also thank Keiko Nakagawa for technical assistance in gene-chip analysis.