Caspase-dependent and -independent lipotoxic cell-death pathways in fission yeast

Lipotoxicity is postulated as an etiology underpinning non-insulin-dependent diabetes mellitus, cardiomyopathy and other obesity-linked metabolic disorders (Unger and Orci, 2002). Relevant studies have highlighted the close relationship between perturbed lipid homeostasis and endogenous cell-death programs. In this model, chronic surplus-energy intake that exceeds the lipid-storage capacity of the adipose tissue causes a spillover of free fatty acids to non-adipose tissues. This consequently initiates ectopic build-up of toxic lipid intermediates, which are causally related to apoptosis and organ dysfunctions (Schaffer, 2003).

The signaling processes involved are complex and there appears to be cell-type-specific metabolic routing of the excess fatty acids to various secondary lipid metabolites. Insights gained under varied physiological settings have implicated different lipids, most notably ceramide and diacylglycerol (DAG). In pancreatic islet cells and numerous other cell types, superfluous palmitate is channeled into the de novo synthesis of ceramide, which triggers apoptosis (Dyntar et al., 2001; Shimabukuro et al., 1998; Turpin et al., 2006). However, the causative role of ceramide is controversial because palmitate-induced apoptosis can also occur through ceramide-independent means, such as the generation of reactive oxygen species (ROS) (Listenberger et al., 2001; Lupi et al., 2002; Sparagna et al., 2000). In other cell types, such as rat neonatal cardiomyocytes, palmitate-stimulated apoptosis is mediated by neither ceramide nor ROS, but by the diminution of mitochondrial cardiolipin and by cytochrome c release (Hardy et al., 2003; Ostrander et al., 2001). In addition to these processes, upstream DAG accumulation, activation of protein kinase C (PKC) and mitogen-activated protein-kinase and ROS production are the key events in the lipoapoptosis of vascular-smooth-muscle cells (Inoguchi et al., 2000; Inoguchi et al., 1994; Yu et al., 2001). Although the observed heterogeneity can to a large extent be attributed to the diverse genetic and metabolic backgrounds in various models, it remains a challenge to elucidate the precise molecular determinants that define the prevalence or relative influence of individual signaling pathways under particular physiological contexts.

Multiple forms of programmed cell death (PCD) have been described in mammals on the basis of discrete morphological criteria (Kroemer et al., 2005). Historically, apoptosis is the most intensively studied form, but increasing evidence has pinpointed the existence of other forms of PCD, for example autophagic cell death, mitotic catastrophe and senescence (Kroemer et al., 2005). Advances in the molecular understanding of mammalian PCD have unraveled the immense diversity of signal-transduction pathways and the enormous multitude of cell-death machineries (Danial and Korsmeyer, 2004; Golstein and Kroemer, 2007). As with many other biological studies, the overwhelming complexity of mammalian systems calls for comparative studies in simpler systems, such as yeasts.

Since the last decade, a growing community of biologists has turned to yeasts as model organisms in which to study PCD. PCD in the budding yeast Saccharomyces cerevisiae exhibits numerous functional and mechanistic similarities to its counterpart in mammalian cells at the cellular, organelle and even macromolecular levels. Morphological features of apoptosis are detected as a result of ectopic gene expression, intracellular defects or unfavorable growth circumstances (Buttner et al., 2006). In many cases of yeast PCD, mitochondrial events are proven to be pivotal (Eisenberg et al., 2007; Hardwick and Cheng, 2004). Additionally, yeasts possess several orthologs of mammalian cell-death mediators (Frohlich et al., 2007). Metacaspase Yca1 in S. cerevisiae was shown to undergo caspase-like proteolytic processing and mediate apoptosis under a wide range of stimuli, including in conditions involving hydrogen peroxide, acetic acid, chronological aging, hyper-osmotic shock, viral killer toxins, valproic acid, deleterious mutations in mRNA decapping, DNA replication and protein deubiquitylation, and heterologous expression of α-synuclein (Bettiga et al., 2004; Flower et al., 2005; Madeo et al., 2002; Mazzoni et al., 2005; Mitsui et al., 2005; Reiter et al., 2005; Silva et al., 2005; Weinberger et al., 2005). Overexpression of Trypanosoma brucei metacaspase promotes mitochondrion-associated cell death in S. cerevisiae, and Leishmania major metacaspase can functionally replace S. cerevisiae Yca1, alluding to the functional conservation of metacaspases in lower eukaryotes (Gonzalez et al., 2007; Szallies et al., 2002). Evidence of PCD has also been presented in the fission yeast Schizosaccharomyces pombe (Low and Yang, 2008; Marchetti et al., 2006; Mutoh and Kitajima, 2007; Roux et al., 2006), but its core machineries, such as metacaspase, are not documented as yet.

Recently, we reported in the fission yeast a novel form of endogenically triggered lipoapoptosis arising from a genetic obstruction in triacylglycerol (TAG) synthesis (Zhang et al., 2003). Mutants with double deletion (i.e. knockout) of the two major DAG acyltransferase genes, plh1⁺ and dga1⁺ (DKO mutants), were found to exhibit elevated levels of DAG and to undergo apoptosis upon entry into stationary phase as well as when challenged with exogenous fatty acids and DAG. Ceramide, however, induces cell death that is not associated with apoptotic markers. It is of particular interest to investigate the mechanisms underlying lipid-induced cell death in DKO mutants, which may serve as an excellent model in which to examine the links between lipid signaling molecules and internal cell-death pathways.

In this report, we further characterize lipotoxic cell death in fission yeast, i.e. cell death resulting from abnormal metabolism of intracellular lipids. We delineate the serial molecular events at the onset of lipoapoptosis. Global lipid profiling using a versatile high-performance-liquid-chromatography–mass-spectrometry (HPLC-MS) method further sheds light on the distinct lipotoxic roles of DAG and ceramide in S. pombe. We examined the energetic status of mitochondria and ROS production, and assessed the involvement of the putative cell-death effectors metacaspase Pca1, the Bcl2-homology domain 3 (BH3)-only protein Rad9, and DAG-binding proteins Pck1 and Bzz1. We conclude that fission yeast displays plasticity of metacaspase-dependent and -independent modes of cell death with distinct morphology under different conditions and stimuli.

To better understand the cell-death process in TAG-deficient mutants, we sought simple yet reliable markers with which to monitor cell death. To this end, we observed progressive pyknosis (condensation of chromatin), and karyorhexis (breakdown of the nuclear envelope) in the DKO cells as they entered stationary phase in rich medium (Fig. 1) (Zhang et al., 2003). In order to visualize the nuclear envelope, the mutant loci plh1::his3⁺ and dga1::kanMX4 were crossed to a genetic background with a genome integration of the green fluorescence protein (GFP) at the C-terminus of the nuclear-pore protein Nup124 (Balasundaram et al., 1999). Nup124-GFP was expressed at endogenous levels and its fluorescence could be monitored under the microscope. During log-phase growth in rich medium (cell density of 0.4×10⁷-2.0×10⁷ cells/ml), DKO cells exhibited high viability, as shown by the high colony-forming efficiency on solid rich medium (Fig. 1A). The chromatin and nuclear morphology appeared normal and were completely indistinguishable from that of wild-type cells (Fig. 1C; the micrograph for wild-type cells at log phase is shown in supplementary material Fig. S1). Unlike other higher eukaryotes, S. pombe undergoes `closed' mitotic nuclear division, during which nuclear-envelope dissolution does not take place (Demeter et al., 1995).

However, as DKO cells transited into post-log phase (3×10⁷-5×10⁷ cells/ml), their viability decreased dramatically, denoting a rapid, irreversible commitment to cell death, because these cells failed to be revived when plated onto fresh solid medium (Fig. 1A). This was accompanied by chromatin condensation in the vast majority of the population, with some cells also showing nuclear-envelope breakdown (Fig. 1D). The chromatin condensation that was detected in DKO cells was readily distinguishable from that of the metaphasic chromosomes by its higher intensity and multi-focal nature (Fig. 1H), which is probably similar to that observed in another case, in which it is referred to as hypercondensation (Demeter et al., 1995).

Upon reaching early stationary phase (6×10⁷-8×10⁷ cells/ml), extensive chromatin condensation and fragmentation ensued with the concomitant breakdown of the nuclear envelope into membrane-enclosed compartments (Fig. 1E). Advanced breakdown and rupture of the nuclear envelope and dispersion of the hypercondensed chromatin into the cytoplasm were generally observed when DKO cells had reached the saturation density (8×10⁷-10×10⁷ cells/ml, compare to the wild-type saturation density of >12×10⁷ cells/ml) for 24 hours or more (Fig. 1F). The nucleus of the wild-type cells remained intact throughout the above time frame (Fig. 1G).

Surprisingly, however, lipotoxic cell death in minimal medium was quantitatively and qualitatively different from that in rich medium: the onset of cell death was much delayed in minimal medium (Fig. 1B) and not coupled to characteristic pyknotic nuclear alterations as early events (Fig. 1I-L). This denoted that an entirely different cell-death pathway was invoked. In minimal medium, the general morphology of DKO cells was similar to that of wild type, except that the percentage of cells that displayed clumped chromatin (Fig. 1H) and lost cellular ultrastructures (characterized by diminished or delocalized Nup124-GFP fluorescence) was always higher in DKO cells (Fig. 1K,L).

To determine whether a functional link exists between malfunctional intracellular-lipid regulation and the mechanisms of cell death, the global lipid profiles at various growth phases in rich medium were obtained using HPLC-MS. The disruption of the DAG-acyltransferase genes plh1⁺ and dga1⁺ in DKO cells led to a remarkable reduction in intracellular TAG levels, corroborating published results using radioisotope labeling (Fig. 2A) (Zhang et al., 2003). Conversely, levels of DAG, as the direct metabolic precursor of TAG, were elevated by 84% (for DAG C36:2, relative to wild type) at log phase and more markedly, i.e. 205% elevated (for DAG C36:2, relative to wild type), at early stationary phase (Fig. 2B).

In fact, after the block in DAG esterification, DKO cells were much more sensitive to high concentrations of exogenous fatty acids and DAG than were wild-type cells (Fig. 2D,E). We demonstrated previously that the removal of cellular DAG, but not the inhibition of ceramide biosynthesis, specifically attenuated fatty-acid-induced apoptosis; hence, fatty-acid toxicity is primarily mediated by DAG (Zhang et al., 2003). When exposed to DAG and oleic acid (similar was observed for palmitic acid), log-phase DKO cells revealed phenotypes that highly resembled the case at stationary phase in rich medium, i.e. chromatin hypercondensation and nuclear-envelope breakdown (Fig. 2F), presumably due to the activation of a common death pathway.

By contrast, the levels of sphingolipids (only the result for phytoceramide analysis is shown) were also affected in DKO cells, i.e. increased by 52% and 35% at log and early-stationary phases, respectively (Fig. 2C). We also noticed a poor intracellular buffering capacity for sphingolipids in DKO cells when compared with wild-type cells (Fig. 2E) (Zhang et al., 2003). All the sphingolipids tested, such as C₂-ceramide, dihydrosphingosine and phytosphingosine, however, induced cell death that was similarly not accompanied by stereotypical nuclear lesions (Fig. 2F; micrographs for dihydrosphingosine and phytosphingosine treatments not shown). This was consistent with the notion that, in S. pombe, DAG is the key lipid species that induces apoptotic cell death, whereas ceramide evokes yet another cell-death program that apparently does not involve early nuclear events. Taken together, all the above results indicate that multiple lipotoxic cell-death pathways or mechanisms exist in the fission yeast and the engagement in certain cell-death pathways is highly context-specific depending on the conditions or stimuli.

In order to investigate whether the various forms of cell death described above are caused by simple cell lysis, vital stainings with propidium iodide (PI) and Phloxin B were performed (Fig. 3A,B). These dyes serve as indicators for plasma-membrane integrity and general metabolic activities because they are actively pumped out from live cells but accumulate in dead cells or those with compromised metabolic activities (Moreno et al., 1991; Roux et al., 2006). It was observed that, at early stationary phase in rich medium, although the majority of the DKO population had lost their viability (Fig. 1A), most of them still displayed intact plasma membrane (Fig. 3A,B) and active, albeit diminished, respiratory activities (Fig. 5E, elaborated later). Substantial loss of plasma-membrane integrity commenced only from day 1 of stationary phase, whereas a remnant of respiration could still be detected up until day 2 of stationary phase (Fig. 3A,B and Fig. 5E).

These observations suggested an active dying process, with the loss of replicative capacity and pyknotic nuclear changes preceding the actual death as an endpoint, i.e. the general cessation of metabolic activities and the ultimate dismantling of cellular ultrastructures (Kroemer et al., 2005). In other words, this form of cell death could not be explained by spontaneous cell lysis because the permeabilization of plasma membrane occurred much later than the loss of viability and the presence of the osmotic-stabilizing agent sorbitol did not modify the death pattern (under the condition discussed below) (Wysocki and Kron, 2004). The precedence of viability loss before cell lysis was similarly observed in the cell death that was induced by exogenous lipids or stationary-phase entry in minimal medium (data not shown).

The nuclear-envelope disintegration that was observed during the death process in DKO cells upon entry into stationary phase was highly reminiscent of that observed following pim1 mutation (Demeter et al., 1995). pim1⁺ encodes a chromatin-associated guanine nucleotide exchange factor of the RCC1 protein family. A temperature-sensitive mutation in pim1 led to the accumulation of cells arrested with a medial septum, condensed chromatin and fragmented nuclear envelope, which eventually lost viability (Demeter et al., 1995). Interestingly, upon the induction of cell death in rich medium, most DKO cells were double nucleated (although the nuclei were largely deformed or fragmented in the advanced stage; Fig. 1D-F) and septated (Fig. 3C,D), implying a post-mitotic arrest. By contrast, wild-type cells at stationary phase were typically rounded, contained a single nucleus (Fig. 1G) and did not form a septum (Fig. 3C,D). Similar to pim1 mutation, the development of the terminal phenotype in DKO cells was probably dependent on the passage through mitosis and might be caused by crucial cellular defects from post-mitosis to cytokinesis.

In agreement with our conjecture, exogenous fatty acids and DAG also led to increased percentages of septated cells (Fig. 2F; numerical data not shown). By contrast, the cell death in minimal medium was not linked to high septation indices (Fig. 1L; numerical data not shown), further illustrating the mechanistic divergence of these discrete pathways.

Having carefully characterized the cell-death phenotypes under various conditions, we next sought to examine the role of well-established regulators of apoptosis (e.g. mitochondria) in lipotoxic cell death. However, owing to the lengthy transition period and stochastic onset of cell death, we found it infeasible to precisely follow the kinetics of molecular events taking place as the cultures entered stationary phase. Thus, we developed an assay involving the use of cell-conditioned media (see Materials and Methods for the preparation of conditioned media). We hypothesized that the medium of a stationary-phase culture provided the essential extracellular environment or signals that induced cells to enter stationary phase. Indeed, conditioned rich medium almost exactly mimicked stationary-phase conditions: it recapitulated not only identical morphological changes (Fig. 4A) and viability patterns (Fig. 4B), but also the intracellular-lipid profiles (Fig. 4C-E) in wild-type and DKO cells.

Acute transfer of a log-phase wild-type culture into the conditioned rich medium induced the characteristic stationary-phase morphology with high survival rates within 1 hour (Fig. 4A,B). After switching to the conditioned rich medium, a fraction of the wild-type cells were still able to complete one to two rounds of cell cycles before finally becoming arrested with a short, rounded shape. DKO cells, by contrast, failed to adapt to the acute switch to the conditioned rich medium, and immediately arrested growth and developed the terminal-cell-death phenotype in approximately 30 minutes. This might be causally linked to the dramatic surge of intracellular DAG in DKO cells (Fig. 4D). In the conditioned rich medium, DAG (C36:2) and phytoceramide were elevated by 192% and 32%, respectively, in DKO cells (relative to wild type; Fig. 4D,E), with profiles virtually identical to those in rich medium at stationary phase (Fig. 2A-C).

From the proportions of the population that were double-nucleated or post-mitotic, that formed a medial septum, that were viable, that displayed condensed chromatin or disintegrated nuclear envelope, respectively, we deduced that, upon entry into stationary phase in rich medium, the elevated levels of DAG caused a post-mitotic arrest and rapid commitment to cell death (Fig. 4B,D,F). This was soon followed by chromatin condensation, which in turn preceded nuclear disintegration (Fig. 4G). The proportions of cells stained positive with PI remained minority in the course of 6 hours (data not shown) and supplementation of the osmosis-stabilizing agent sorbitol yielded no quantifiable changes (Fig. 5B; micrograph not shown), again showing that non-specific lysis was not the predominant mode of cell death.

Based on the above, the chief merit of the conditioned medium rested in its virtue of stationary-phase induction in an acute manner. This validated its use in combination with pharmacological agents such as mitochondrial inhibitors in order to examine the effects of these agents on cell death at stationary phase. These pharmacological agents strongly forestalled the stationary-phase entry if directly added into log-phase cultures (data not shown).

Both in mammals and the budding yeast, mitochondria are central to apoptosis not only as targets of destruction but also as active cell-death perpetrators (Hardwick and Cheng, 2004). To elucidate the mechanism of lipoapoptosis in the DKO cells, we addressed the role of mitochondria by using conditioned media. As depicted in Fig. 5, complex-IV inhibitors cyanide and azide were potent attenuators of lipotoxic cell death in the conditioned rich medium, and completely abolished the apoptotic nuclear changes (Fig. 5A,B; azide-treated DKO cells appeared indistinguishable from cyanide-treated DKO cells). Uncoupler of mitochondrial membrane potential carbonyl cyanide p-trifluoromethoxy phenyl hydrazone (FCCP) retarded the development of morphological changes, but its effect on viability was insignificant (Fig. 5A,B). These results suggested that normal mitochondrial functions were needed for both the induction as well as the progression of lipotoxic cell death, analogous to acetic acid and pheromone- and/or amiodarone-induced PCD (Ludovico et al., 2002; Pozniakovsky et al., 2005). The inhibitor of F₀F₁-ATPase, oligomycin, had no effect on viability nor morphology (Fig. 5A,B) (Ludovico et al., 2002). This form of lipotoxic cell death was probably not dependent on de novo protein synthesis because cycloheximide exerted no effect at all concentrations tested (Fig. 5A,B). It should be noted that, at the concentrations tested, cyanide and azide were deleterious to log-phase DKO cells (Fig. 5C shows the results for cyanide treatment in rich medium), indicating the vital and lethal functions of mitochondria under different conditions.

In accordance with our hypothesis that DAG was the lipid molecule that promoted mitochondrion-dependent apoptosis in rich medium, cyanide effectively abrogated the toxicity of exogenous DAG (Fig. 5C). Respiring mitochondria also appeared to be necessary for ceramide-induced cell death (Fig. 5C). In conjunction with this, we observed that in conditioned minimal medium DKO cells DKO cells died to a greater extent than wild type cells, although the induction time required was significantly longer than in the conditioned rich medium (supplementary material Fig. S2A). In contrast to the case in the conditioned rich medium, mitochondrial inhibitor cyanide exacerbated lipotoxic cell death in the conditioned minimal medium (supplementary material Fig. S2B). This suggested that mitochondrial activities were not required for lipotoxic cell death in minimal medium.

Similar to mitochondria, generation of ROS is also a centerpiece of PCD in mammalian cells and budding yeast. In some cases of yeast PCD, an initial rise in the mitochondrial transmembrane potential (ΔΨ_m) (hyperpolarization) or increase in the energy coupling promotes a downstream oxidative burst, which then initiates mitochondrial de-energization or functional impairment (Eisenberg et al., 2007). We similarly noticed a transient hyperpolarization in DKO cells upon the induction of cell death in rich medium (Fig. 5D). This transient hyperpolarization was immediately followed by a depolarization with a concomitant reduction in respiratory activities (Fig. 5D-F). We also noted a build-up of ROS that coincided with the diminution of mitochondrial activities (Fig. 5D-F and Fig. 6A). The percentage of dying DKO cells with high levels of ROS [i.e. 2′,7′-dichlorofluorescein diacetate (DCFH-DA)-positive and PI-negative) decreased with time after entry into stationary phase (Fig. 6B). The upper panel of Fig. 6C shows the control experiments for the specificity of the double stainings with DCFH-DA and PI. Dead cells did not take up DCFH-DA non-specifically but were stained with PI. Addition of hydrogen peroxide, a condition known to induce oxidative stress in S. cerevisiae (Madeo et al., 1999), similarly induced massive generation of ROS in wild-type S. pombe cells (Fig. 6C, middle panel).

The generation of ROS was likewise observed in the lipoapoptotic cell death that was triggered by the conditioned rich medium, but was completely annulled by cyanide (Fig. 6C, lower panel), positioning the oxidative burst downstream of mitochondrial activities. Furthermore, free-radical spin trap 3,3,5,5-tetramethylpyrroline N-oxide and hypoxic conditions alleviated the lipoapoptotic cell death of DKO mutants (data not shown) (Zhang et al., 2003). In summary, these observations denoted that mitochondria and ROS play an active part in the lipoapoptosis of S. pombe.

Next, we dissected the genetic requirements of the various forms of lipotoxic cell death described above. Pca1 (for pombe caspase 1, GeneDB systematic name SPCC1840.04) is a metacaspase in fission yeast that displays over 50% identity to S. cerevisiae metacaspase Yca1 at the amino acid level. Pca1 was predicted to have a molecular mass of 46.6 kDa, to consist of 425 amino acid and have a conserved cysteine-270–histidine-271 catalytic diad in the putative caspase domain. When GFP was tagged to the C-terminus of Pca1, a C-terminal subunit of approximately 10 kDa was consistently detected by western blotting, in addition to the full-length and intermediate truncated bands (Fig. 7). The cleavage was totally abolished when cysteine 270 was specifically mutated to alanine. This indicated that, similar to Yca1 (Madeo et al., 2002), Leishmania major (Gonzalez et al., 2007) and Arabidopsis (Watanabe and Lam, 2005) metacaspases, Pca1 underwent auto-proteolytic processing that was dependent on its catalytic cysteine residue.

The deletion of pca1⁺ from DKO (Δpca1 triple knockout or Δpca1 TKO) did not inhibit lipotoxic cell death in rich medium (Fig. 8A), conditioned rich medium or following treatments with fatty acids, DAG or ceramide (data not shown). Interestingly however, Δpca1 TKO cells demonstrated significantly delayed cell death in minimal medium as compared with DKO cells (Fig. 8C).

We also investigated the roles of several putative cell-death mediators in the various forms of lipotoxic cell death by disrupting these candidate genes in a number of DKO genetic backgrounds. We hypothesized that the lipoapoptotic or lipotoxic cell-death pathways in fission yeast can be conceptualized into three consecutive stages: initiation, execution and destruction (Fig. 9) (Low et al., 2005). In the initiation stage, the interaction between upstream lipid molecules, such as DAG, and their molecular targets results in the activation of the core death machineries (the execution stage), leading to cell-death enactment. This culminates lastly in the dismantling of cellular structures, such as the nucleus, in the destruction stage.

We reasoned that, in the initiation stage, lipid molecules first bind to effectors with conserved lipid-binding motifs. In fission yeast, there are three gene products that contain conserved DAG-binding domain 1 (C1), a cysteine-rich domain that mediates the recruitment of proteins to the membrane (Spitaler and Cantrell, 2004). The C1 domain was first identified in PKC and later in other non-kinase proteins, such as Munc13, and participates in DAG- and/or phorbol-ester-induced apoptosis (Colon-Gonzalez and Kazanietz, 2006). Pck1 and Pck2 are two C1-containing PKC homologs in fission yeast. Their known biological roles so far have been ascribed to the maintenance of cell-wall integrity and morphogenesis (Arellano et al., 1999). Bzz1 (SPBC12C2.05c) is another C1-containing non-kinase protein of unannotated physiological function in fission yeast, but studies on its homolog in budding yeast point to its role in actin polymerization and endocytosis (Soulard et al., 2005).

For the execution stage, we analyzed S. pombe rad9⁺. Rad9 is a component of the complex Rad9-Hus1-Rad1; this complex functions as a sensor in the DNA-damage and replication checkpoint (Caspari et al., 2000). Mutations in rad9 led to a failure to halt the cell cycle in G2 phase in the presence of DNA damage; hence, mutants were susceptible to a range of radiation- and drug-induced genotoxicity (Komatsu et al., 2000a). Paradoxically, overproduction of S. pombe Rad9 or mammalian Rad9 induced apoptosis in human cells, which apoptosis could be blocked by co-expression of Bcl2 (Komatsu et al., 2000b). Similar to Bad, Bid and Bim, S. pombe Rad9 contains a single BH3 domain, which was essential for its physical interaction with Bcl2 and the pro-apoptotic effect.

Another candidate gene examined was apg6⁺ (systematic name SPAC20G8.10c). Apg6 belongs to the Beclin family of autophagy proteins, but its role in autophagy is yet to be confirmed in fission yeast. Its homolog in S. cerevisiae, Apg6 (also known as Vps30) is crucial for the formation of autophagic vesicles (Abeliovich and Klionsky, 2001). In some cell-death models, increases in autophagic vesicles and lysosomal activity occur as early events in the cell-death process. The autophagic pathway per se constitutes type-II PCD, which is caspase-independent and non-apoptotic in nature (Gozuacik and Kimchi, 2007).

We observed that, in minimal medium, single deletions of pck1⁺, bzz1⁺ or rad9⁺, but not pck2⁺ or apg6⁺, reverted the viability of the respective TKO cells to that of the wild type (Fig. 8C,D, compare with Fig. 1B). Similar to the case of pca1⁺, single disruptions of all these genes did not confer resistance to all the other forms of lipotoxic cell death (Fig. 8A,B, data not shown).

In this study, we first delineate distinct apoptotic and non-apoptotic forms of cell death that are not products of simple cell lysis in S. pombe. Subsequently, we show that the central roles of mitochondria and ROS in PCD are conserved in the lipoapoptosis of S. pombe. Finally, we identify genetic components that are crucial to lipotoxic cell death.

Despite the similitude of yeast and mammalian PCD, using yeasts as models for apoptosis research is in its infancy and has been controversial (Wysocki and Kron, 2004). Here, we provide evidence that, as a result of perturbed intracellular-lipid homeostasis, fission yeast not only undergoes PCD, but does so via multiple pathways that are distinct with respect to morphological changes, genetic criteria and cellular mechanisms. Our work represents the first in-depth analysis of cell-death programs in fission yeast.

Lipids are subject to stringent spatiotemporal regulation, the failure of which is often associated with pathophysiological conditions (Schaffer, 2003). Lipotoxicity has also been reported in other organisms, such as Drosophila, that are informative systems employed to study the mechanisms of lipid-induced cell death (Buszczak et al., 2002; Phan et al., 2007; Xu et al., 2003). In the current study, we employ a TAG-deficient S. pombe mutant as a model sensitized to intracellular changes in lipids to investigate the involvement of lipid molecules in cell-death mechanisms. We showed that, in rich medium, a block in TAG synthesis elicits a post-mitotic arrest upon entry into stationary phase, followed by a mitochondrion-dependent pathway that entails prominent chromatin condensation and nuclear-envelope fragmentation. By means of cell-conditioned media, we demonstrated that the nutritional or quorum cue that prompts cells to enter stationary phase resides in the extracellular milieu. As cells arrest their mitotic cycles and assume the stationary-phase states, an adaptive lipid remodeling is likely to take place as the system diverts the synthesis of phospholipids to the synthesis of storage-neutral lipids (TAGs and sterol esters) (Hosaka and Yamashita, 1984). Because the esterification of DAG is prevented in DKO cells, DAG, instead of taking its normal metabolic course, is transformed into an active inducer of apoptotic cell death (Fig. 9) (Zhang et al., 2003). How DAG arrests cells post-mitotically and activates the downstream cell-death events is unknown, but this intriguing finding has opened up new avenues of investigation.

The genes involved in this pathway remain unclear because single disruptions of genes encoding putative cell-death mediators did not reverse the cell death. This could be due to the limitation of this hypothesis-driven approach. There are a number of other candidate genes, such as homologs of apoptosis-inducing factor (AIF) and pbh1⁺, which encodes a baculoviral inhibitor-of-apoptosis repeat (BIR) protein (Rajagopalan and Balasubramanian, 2002; Walter et al., 2006). Several homologs of AIF are present in S. pombe, yet the deletion of one of these, namely SPAC29A4.01C (designated aif1⁺) did not inhibit lipoapoptotic cell death (data not shown). Another possibility is the redundancy of the candidate genes under this condition, in which the cell-death signal is possibly transduced at its upper limit, although this hypothesis is yet to be verified.

Since the morphological description in the early days, pyknosis and karyorhexis are the close-to-definitive hallmarks of apoptosis (Zamzami and Kroemer, 1999). In S. pombe, DAG-induced cell death in rich medium is associated with pronounced early nuclear events, whereas the paucity of nuclear events in lipotoxic cell death in minimal medium and ceramide-induced cell death hints at the existence of autonomous death programs operating in parallel with apoptosis but subject to other controls (Table 1) (Low and Yang, 2008). The plasticity of the modes of cell death under varying conditions and stimuli observed here is highly suggestive of the diversification of cell-death pathways even in unicellular eukaryotes. Teleologically speaking, the sophistication of cell-death programs allows for a tighter control of cell death and its execution solely under strictly necessary circumstances.

There are other examples of the existence of multiple cell-death pathways in lower eukaryotes. In S. cerevisiae, an apoptotic pathway mediated by the endonuclease-G homolog Nuc1 is independent of Yca1 and Aif1, and prevails only under conditions in which mitochondrial functions are potentiated (Buttner et al., 2007). Multiple upstream signaling pathways govern waves of cell death with different kinetics in response to mating pheromones (Zhang et al., 2006). The developmental cell death of the protist Dictyostelium is typically vacuolar or autophagic and is characterized by early commitment to cell death, extensive vacuolization, focal chromatin condensation and delayed plasma-membrane permeabilization (Tresse et al., 2007). The deletion of autophagy gene atg1 suppresses the morphological changes, but not the cell death itself, resulting in a necrotic cell death (Kosta et al., 2004). It is postulated that this necrotic pathway might be primordial but has become vestigial in the course of evolution and might serve to complement or backup the more-developed or refined vacuolar- and/or autophagic pathway.

Albeit devoid of noticeable morphological changes indicative of any currently known pathway, the delayed cell death in minimal medium might represent a non-apoptotic, non-vacuolar death pathway in S. pombe that is genetically dependent on pca1⁺, rad9⁺, pck1⁺ and bzz1⁺ (Fig. 9; Table 1). Potential differences in lipid profiles and stress signaling in rich and minimal media might account for the discrepancy in the cell-death patterns observed. In fact, medium-dependent cellular responses have been presented in other studies that underline fundamental differences in the signaling mechanisms in various medium backgrounds, such as the cAMP pathway (Takeda et al., 1995). This further suggests that interplay between upstream signaling modules, such as the stress-activated protein-kinase Sty1 pathway or the target-of-rapamycin (TOR) pathway, might play a decisive role in the initiation stage of lipotoxic cell death (Low et al., 2005). In line with this, our preliminary results revealed that, at log phase, DKO cells display a higher basal level of Sty1 phosphorylation and are much more sensitive to rapamycin-induced cell death with the concomitant generation of ROS (C.P.L, L.P.L. and H.Y., unpublished).

We also showed that S. pombe metacaspase Pca1 exhibits similar biochemical properties to its S. cerevisiae counterpart Yca1. However, we failed to detect any consistent cytoprotection in pca1-deletion strains in conditions of hydrogen peroxide, acetic acid or chronological aging (data not shown). In some genetic backgrounds, we even noticed that Δpca1 displays lower viability than either the isogenic (in targeted gene disruptions) or parental (in genetic crossings) wild-type strains (data not shown).

The disparity seen in these two organisms is nonetheless not surprising, because they are vastly divergent from each other in evolution (Sipiczki, 2000). In the mould Aspergillus fumigatus, simultaneous deletion of the two metacaspase genes CasA and CasB does not enhance the survival under the stimuli stated above, but on the contrary, is detrimental to growth under endoplasmic-reticulum stress (Richie et al., 2007). Paracaspase is not required for the developmental cell death of Dictyostelium and the overall growth, but unknown aspartate-directed cysteine proteases might be crucial in the pre-death developmental stages (Olie et al., 1998; Roisin-Bouffay et al., 2004). pca1⁺ is upregulated at the transcript level under hydrogen peroxide, cadmium, heat, sorbitol and methyl methanesulfonate treatments (Chen et al., 2003; Lim et al., 2007), hinting at its stress-adaptive (pro-survival) and/or suicidal (pro-death) role. It is of particular interest that, in our hands, pca1 deletion alleviates lipotoxic cell death only in minimal medium. This, however, adds S. pombe to the expanding list of eukaryotes in which metacaspases are implicated in the regulation of cell death, including Arabidopsis thaliana (He et al., 2008), budding yeast (Madeo et al., 2002), and the protozoa Leishmania donovani (Lee et al., 2007) and Plasmodium falciparum (Meslin et al., 2007).

Functional dualism has been described for a number of death regulators in yeast, including nuclease Nuc1, which is involved in mitochondrial-DNA recombination, and mitochondrial-fission protein Dnm1, suggesting the evolutionary sophistication of gene functions (Buttner et al., 2007; Cheng et al., 2006). Recently, Rad9 has also been shown to be cytoprotective under oxidative stress (Kang et al., 2007). It remains to be ascertained whether Pca1 and Rad9 fit into the category of proteins with dual opposing functions.

Our data also indicate a pro-death role of Pck1, but not Pck2, in lipotoxic cell death of S. pombe. Given that the genome of S. pombe is to-date the smallest among all sequenced eukaryotic genomes and displays a minimal gene redundancy, Pck1 and Pck2 are likely to have distinct physiological roles or be subject to differential regulations (Perez and Calonge, 2002). By contrast, there is only one PKC homolog, Pck1, in S. cerevisiae. A recent genomic screen for sensitivity to isoprenoid farnesol unveiled the pro-survival role of Pck1 (Fairn et al., 2007). ROS generation by the electron-transport chain was shown to be the primary mechanism of farnesol toxicity. In the presence of farnesol, Pck1 translocates into mitochondria where it might phosphorylate its substrates to confer cytoprotection against ROS.

In summary, three distinct forms of lipotoxic cell death in fission yeast have been identified in this study (Table 1). Our results also highlight the importance of mitochondria and ROS in the lipoapoptosis of S. pombe. The biochemical mechanisms of the various genetic components for the lipotoxic cell death identified in this study might be further investigated to provide insights into lipid-mediated cell-death signaling. Finally, our work establishes S. pombe as a novel alternative system for PCD research. Powered by the rich repertoire of genetic tools available in the fission yeast and the robust analytical techniques, continued work will probably be informative and might unravel important governing principles underlying lipid signaling and cell demise.

Rich medium (YES) and minimal medium (EMM) were prepared as previously described (Moreno et al., 1991; Zhang et al., 2003). All S. pombe strains used in this study are listed in supplementary material Table S1. All growth and treatment incubation was performed at 30°C with shaking, in culture tubes (less aeration) or flasks (good aeration), which produced results with similar trends. The cold-sensitive mutant nda3 KM311 was grown to log phase in YES at 30°C and was arrested at metaphase by incubating at 20°C for 6 hours (Toda et al., 1983).

All gene disruptions were performed by replacing the entire coding regions with the respective markers using the long-flanking homologous-recombination method (Wach et al., 1994; Zhang et al., 2003). The primer sequences for the respective gene-disruption cassettes are listed in supplementary material Table S2. To disrupt dga1⁺ using kan^R, an ∼1.5 kb restriction fragment containing the kanMX4 gene from pFA6akanMX4 was used to replace the ura4⁺ marker in the gene-replacement cassette for dga1⁺ (Zhang et al., 2003). Genotypes of mutants were unambiguously determined by positive and negative diagnostic PCR procedures. Positive PCR tested for the incorporation of the markers at the specific loci, whereas negative PCR served as a negative control for clones tested positive and detected the presence of the coding regions. The primer sequences for diagnostic PCR are listed in supplementary material Table S3.

Colony-forming or clonogenic assays were performed as previously described (Zhang et al., 2003). We did not observe significant discrepancy when using diluents other than sterile water, such as culture media or 1.2 M sorbitol. The cell densities of the suspension cultures were accurately determined by hemacytometric counting. Viability is expressed as percentages of cells that formed visible colonies on solid rich medium. A viability of 100% corresponds to all plated cells (approximately 500 cells per plate) forming visible colonies.

Visualization of DNA with 4′,6-diamidino-2-phenylindole (DAPI) was performed as described previously (Zhang et al., 2003). For septum staining, formaldehyde-fixed cells were processed as for DAPI staining, except that they were finally resuspended in 0.01% (w/v) aniline blue. For vital staining with Phloxin B, cells were incubated in 0.4 mg/ml Phloxin B in phosphate-buffered saline (PBS) and incubated at 30°C for 5 minutes, then washed with an equal volume of PBS. All samples were viewed under a Leica DM2500 transmitted light and fluorescence microscope under 1000× magnification using acquisition software FW4000. At least 1000 cells were examined for each sample. For vital staining with PI, cells were resuspended in PBS containing 50 μg/ml of PI just before flow-cytometric analyses (see below).

ΔΨ_m was assessed by staining with rhodamine 123 (Rh123) as described previously, with modifications (Ludovico et al., 2001). Cells were incubated in sterile water containing 50 nM Rh123 at 2.5×10⁶ cells/ml at room temperature in the dark for 10 minutes. To examine the effects of mitochondrial inhibitors on ΔΨ_m, cells were pre-incubated with the inhibitors in sterile water at room temperature for 10 minutes prior to Rh123 staining.

To assess intracellular ROS levels, cells were incubated in YES containing 60 μM DCFH-DA at 0.5×10⁷ cells/ml at 30°C in the dark for 20 minutes. Just before analysis, cells were co-stained with PI, as described above for analysis of plasma-membrane integrity. For a positive ROS control, wild-type cells were grown to log phase in YES and exposed to 2.5 mM hydrogen peroxide for 2 hours. For a negative ROS control, heat-killed cells (presumably died by necrosis) were prepared by incubating wild-type cells at 80°C for 5 minutes.

Flow-cytometric analyses were done using an EPICS Altra (Beckman Coulter, USA) or a CyAn Lx (DakoCytomation, USA) flow cytometer. Data-acquisition software were EXPO 32 and Summit, respectively. At least 10,000 cells were analyzed per sample. Post-run data analyses were performed using software WinMDI v2.8 and Simmit v4.3, accordingly.

1.5×10⁸ cells were accurately resuspended in 1 ml of 20 mM glucose in PBS and the oxygen consumption was monitored using Oxygen Liquid-Phase Oxygen Measurement System supported by Oxygraph Plus software (Hansatech Instruments, UK) for 2 minutes at 30°C. Subsequently, potassium cyanide (KCN) was added at 7.5 mM to ensure that the oxygen consumption was specific to mitochondrial respiration.

The masses of dry cell pellets were recorded before lysis by lyticase digestion and glass beads. Total-lipid extraction was then carried out using a modified Bligh and Dyer method with extended extraction time and multiple extractions (Bligh and Dyer, 1959). Quantitative determination of lipids was performed using an Agilent 1200 HPLC system and 3200 Q-Trap mass spectrometer (Applied Biosystems, USA). Results were expressed in arbitrary units, which represent counts per second normalized to internal standard counts and average total phospholipid counts at log phase per mg of dry mass.

Exogenous-lipid treatments on cells grown to log phase in rich medium were performed as described previously, with some modifications (Zhang et al., 2003). For fatty-acid treatments, oleic and palmitic acids were first dissolved in 10% Brij58 and then completely mixed with culture media at a 1:9 volume ratio (the final concentration of Brij58 being 1%). Cells were centrifuged at 2000 g and resuspended in fatty-acid-containing media. DAG (1,2-dioctanoyl-sn-glycerol), C₂ ceramide (N-acetyl-D-sphingosine), dihydrosphingosine (DL-1,3-dihydroxy-2-aminooctadecane) and phytosphingosine (4-hydrosphinganine from S. cerevisiae) were dissolved in DMSO and added directly into cultures. All lipid treatments were carried out for 2 hours. To examine the effects of mitochondrial inhibitors on exogenous-lipid-induced cell death, drugs were added after the exogenous lipids were added into the cultures.

Strict adherence to the stated conditions was necessary for reproducible results. Cell-conditioned rich medium was prepared as follows: wild-type cells were first refreshed to log phase, and later diluted with fresh YES to 1×10⁶ cells/ml in a culture flask at the vessel:culture volume ratio of approximately 3:1 and incubated for approximately 18 hours at 30°C with shaking until the culture just reached the saturation density of 1.2×10⁸-1.4×10⁸ cells/ml. The culture medium was filter-sterilized and kept at 4°C. Separate batches of medium conditioned with different strains in the same way yielded indistinguishable results. Conditioned minimal medium was prepared in the similar way except that the starting cell density was 2×10⁶ cell/ml and the culture was allowed to grow to day 3 of stationary phase before harvesting. To induce acute entry into stationary phase, cells were grown to log phase, centrifuged and washed with an equal volume of sterile deionized water before being resuspended in the conditioned media. Mitochondrial inhibitors and other pharmacological agents were added at the indicated concentrations.

Full-length pca1⁺ cDNA was subcloned into pREP41 and pREP41-C′GFP (both on nmt promoter). Overexpression was de-repressed by growing yeast cells in EMM-leucine in the absence of thiamine. Cysteine-270 residues of both plasmids were specifically mutated to alanine using the QuikChange Site-Directed Mutagenesis kit (Stratagene, USA). The N-terminal ∼440 bp of the full-length pca1⁺ cDNA was amplified and subcloned into vector pGEX-4T-1. The corresponding recombinant protein from Escherichia coli was used to immunize rabbits (animal work done by Biogenes GmbH, Germany).

All experiments were performed in multiplicates (n⩾3) and repeated in at least two genetic backgrounds. The effects of pca1 deletion were also examined in strains with scrambled genetic backgrounds obtained from genetic crossings, in addition to mutants generated from defined wild-type strains. Statistical analyses were performed using unpaired Student's t-test, and differences with P<0.05 are considered as statistically significant.

This work is supported by research grants from the National Medical Research Council and the Ministry of Education, Singapore, and by a start-up grant from the University of New South Wales, Australia. We are grateful to Mohan Balasubramaniam and Ge Wanzhong for generously supplying strains (MBY series) and relevant techniques. We thank David Balasundaram for providing the Nup124p-GFP strain, Sepp Kohlwein and Julia Petschnigg for technical assistance on fatty-acid treatments and useful discussion, and Jeffrey S. Armstrong for help on measuring oxygen consumption. We acknowledge Kelly Chua, Ai Fang Ong and Chang Xing for their participation in this work. F.M. and S.B. are grateful to the grant `LIPOTOX' from the FWF.