Yeast Cells Lacking All Known Ceramide Synthases Continue to Make Complex Sphingolipids and to Incorporate Ceramides into Glycosylphosphatidylinositol (GPI) Anchors*

In yeast, the inositolphosphorylceramides mostly contain C26:0 fatty acids. Inositolphosphorylceramides were considered to be important for viability because the inositolphosphorylceramide synthase AUR1 is essential. However, lcb1Δ cells, unable to make sphingoid bases and inositolphosphorylceramides, are viable if they harbor SLC1-1, a gain of function mutation in the 1-acyl-glycerol-3-phosphate acyltransferase SLC1. SLC1-1 allows the incorporation of C26:0 fatty acids into phosphatidylinositol (PI), thus generating PI″, an abnormal, C26-containing PI, presumably acting as surrogate for inositolphosphorylceramide. Here we show that the lethality of the simultaneous deletion of the known ceramide synthases LAG1/LAC1/LIP1 and YPC1/YDC1 can be rescued by the expression of SLC1-1 or the overexpression of AUR1. Moreover, lag1Δ lac1Δ ypc1Δ ydc1Δ (4Δ) quadruple mutants have been reported to be viable in certain genetic backgrounds but to still make some abnormal uncharacterized inositol-containing sphingolipids. Indeed, we find that 4Δ quadruple mutants make substantial amounts of unphysiological inositolphosphorylphytosphingosines but that they also still make small amounts of normal inositolphosphorylceramides. Moreover, 4Δ strains incorporate exogenously added sphingoid bases into inositolphosphorylceramides, indicating that these cells still possess an unknown pathway allowing the synthesis of ceramides. 4Δ cells also still add quite normal amounts of ceramides to glycosylphosphatidylinositol anchors. Synthesis of inositolphosphorylceramides and inositolphosphorylphytosphingosines is operated by Aur1p and is essential for growth of all 4Δ cells unless they contain SLC1-1. PI″, however, is made without the help of Aur1p. Furthermore, mannosylation of PI″ is required for the survival of sphingolipid-deficient strains, which depend on SLC1-1. In contrast to lcb1Δ SLC1-1, 4Δ SLC1-1 cells grow at 37 °C but remain thermosensitive at 44 °C.

The acyl-CoA dependent biosynthesis of ceramide is operated by Lag1p and Lac1p (see Fig. 1A), two highly homologous and functionally redundant endoplasmic reticulum proteins that are only active when forming a complex with Lip1p (7)(8)(9). Concomitant deletion of LAG1 and LAC1 causes a significant growth defect in the W303 genetic background, and the same double deletion is lethal in the YPK9 background (10,11). Ypc1p and Ydc1p are two highly homologous ceramidases also residing in the endoplasmic reticulum (see Fig.  1A). When the acyl-CoA-dependent ceramide synthases are inhibited or deleted, Ypc1p and Ydc1p can catalyze the reverse reaction, i.e. the condensation of free fatty acids with accumulating long chain bases (LCBs) (12,13).
The question of whether sphingolipids are required for yeast cell survival has received slightly conflicting answers over the last two decades. A series of seminal studies showed that lcb1⌬ cells (see Fig. 1) are auxotrophic for PHS but that a suppressor allele in the lysophosphatidate acyltransferase SLC1 can make them independent of exogenous PHS (14). The suppressor allele SLC1-1 allows lcb1⌬ cells to synthesize phosphatidylinositol (PI) with a C26:0 fatty acid instead of C18:1 at the sn-2 position of the glycerol moiety. This lipid, herein dubbed PIЉ, can be mannosylated in the same way as IPCs, unlike normal PI (15). Nevertheless, many sphingolipiddependent processes such as endocytosis of Lucifer yellow or raft integration, surface transport, and surface stabilization at 37°C of the essential plasma membrane H ϩ ATPase Pma1p are fully operational in lcb1⌬ SLC1-1 cells (16,17). These latter studies led to the concept that all essential functions of sphingolipids including the stabilization of Pma1p depend on C26-containing lipids but not necessarily on sphingolipids.
Although various studies demonstrated that LCB phosphates, M(IP) 2 Cs, and MIPCs are dispensable (see Fig. 1, A and B) (18 -22), IPCs were suspected to be essential because the IPC synthase AUR1 was found to be essential (see Fig.  1B) and because Aureobasidin A (AbA), a highly specific inhibitor of Aur1p, rapidly killed growing yeast cells (23,24). However, the essentiality of IPCs was questioned when it became clear that the concomitant deletion of all ceramide synthases (LAG1, LAC1, YPC1, YDC1) was not lethal in the W303 background and that W303 lag1⌬ lac1⌬ ypc1⌬ ydc1⌬ (W303.4⌬) cells could grow in the presence of AbA (25). It was, however, suspected that these cells grew because they made some abnormal, uncharacterized lipids, which were resistant to mild alkaline hydrolysis and could be labeled metabolically with [ (25,26). A more recent study has demonstrated, however, that W303.4⌬ cells stop growing on high concentrations of AbA or on AbA at 37°C and also stop making these abnormal inositolphosphorylsphingolipids (13). These data suggested that the essential functions of IPCs could be taken over not only by PIЉ but also by these abnormal, uncharacterized inositolphosphorylsphingolipids.
In the present study, the abnormal inositolphosphorylsphingolipids are characterized and shown to be dispensable in 4⌬ cells only in the presence of SLC1-1. The study documents that through an unknown pathway, W303.4⌬ cells, lacking all known ceramide synthases, still make ceramides, which they add to GPI anchors and use for the synthesis of small amounts of normal IPCs.

EXPERIMENTAL PROCEDURES
Strains, Growth Conditions, and Materials-S. cerevisiae strains are listed in supplemental Table SI, and plasmids are  listed in supplemental Table SII. Cells were grown on rich medium (YPD, YPG, or YPR, containing 2% glucose (D), raffinose (R), or galactose (G) as a carbon source) or synthetic complete medium (SD, SR, or SG (27) or nitrogen base (YNB, United States Biological)) containing also amino acids (aa), uracil (U), and adenine (A), or high amounts (50 mg/liter) of inositol (I) instead of only 2 mg/liter. We also used "Lester medium" (LM), a complete synthetic medium buffered at pH 6.0, containing 4% glucose or 4% galactose, 50 mg/liter inositol but without LCBs and Tergitol (15). LM was designed for cells unable to make LCBs. Methionine was often omitted to increase the transcription from the MET25 promoter (28).
Construction of Strains and Plasmids-W4⌬ strains were constructed from 4⌬.LAG1 strains described before (29). After transfection of plasmids containing SLC1, SLC1-1, or AUR1, the covering plasmids containing LAG1 and/or LAC1 were shuffled out using 5Ј-fluoroorotic acid (FOA). (Unfortunately, the FOA selection does not work on LM.) To construct plasmids containing SLC1 or SLC1-1, the open reading frames were amplified by PCR using oligonucleotides SLC1.F1 and SLC1.R1 (supplemental Table SIII) and genomic DNA from SJ21R or lcb1⌬ SLC1-1 as templates. The PCR products were doubly digested with SpeI and XhoI and ligated into the similarly digested p423-Met25 or p413-Met25 vectors to yield pBF26, pBF27, and pBF212. 4⌬ cells grew best when SLC1-1 was present on a centromeric vector behind the MET25 promoter and in the absence of methionine. For the overexpression of AUR1, we generated the plasmid pYC272-AUR1-LEU2. For this, the LEU2 promoter and coding sequence were amplified by PCR using primers LEU2Not1.F1 and LEU2Not1.R1, and the product was cut by NotI and ligated into pYC272 (30), a 2 vector containing the AbA-resistant AUR1-C allele as selective marker. All inserts were verified by DNA sequencing. The kanMX marker was changed to URA3 or ADE2 as described (31). The strain FBY952 (pmi40 lcb1⌬ SLC1-1) was generated by crossing lcb1⌬ SLC1-1 (4R3-17C) with strain C4 (pmi40). Further methods described before and used with minor modifications are described in the supplemental data.

SLC1-1 Significantly Increases the Fitness of lag1⌬ lac1⌬
ypc1⌬ ydc1⌬ Mutants-The lac1⌬ lag1⌬ mutant is dead in YPK9 but viable in the W303 background (10,11). In both backgrounds, we had previously generated quadruple lac1⌬ lag1⌬ ypc1⌬ ydc1 mutants containing pBM150-LAG1, an URA3 vector harboring LAG1, here named W4⌬.LAG1 and Y4⌬.LAG1 (29). It was not possible to force out the LAG1bearing plasmid from 4⌬.LAG1 in the presence of FOA on FIGURE 1. Sphingolipid biosynthesis and utilization in yeast. A, biosynthesis of ceramides through two different pathways. Gene names are in italics, and enzyme inhibitors are in bold italics. Lag1p or Lac1p are functional only in a complex with Lip1p (9). Cer-2 to Cer-5 designate ceramides with between two and five hydroxyl groups. Only genes relevant to this study are shown. B, biosynthesis of complex sphingolipids. GDP-M ϭ GDPmannose. C, integration of ceramides into GPI anchors. ordinary complete synthetic media even in the presence of 10% glycerol or the water-soluble synthetic ceramide C6-DHS (not shown; supplemental Fig. S1). However, W4⌬.LAG1 could spontaneously loose the LAG1-bearing plasmid on LM, a synthetic, pH-buffered medium containing 4% glucose and 50 mg/liter inositol (13). As SLC1-1, a gain of function allele of SLC1, can rescue lcb1⌬ cells (see the Introduction and Fig.  1), we tested whether it also would improve the fitness of 4⌬ cells. This indeed was the case; 4⌬ cells harboring SLC1-1 were able to loose pBM150-LAG1 and grow on ordinary complete synthetic media, whereas cells harboring the wild type (WT) allele of SLC1 grew only very slowly or not at all ( Fig.  2A; supplemental Fig. S2). 4⌬ cells harboring a centromeric SLC1-1 vector in the W303 and YPK9 backgrounds were called W4⌬.SLC1-1 and Y4⌬.SLC1-1; the ones containing a SLC1-1 multicopy vector were called W4⌬.mSLC1-1 and Y4⌬.mSLC1-1. Their genotypes were confirmed by PCR (supplemental Fig. S3). Overall, W4⌬.SLC1-1 cells were more robust than Y4⌬.SLC1-1 because they grew to higher densities in liquid media and grew at 37°C (not shown; Fig. 2B). Thus, even after the introduction of SLC1-1, YPK9 cells tolerate the deletion of ceramide synthases less well than W303 cells.
4⌬.SLC1-1 Cells Still Add Normal Ceramides to the GPI Anchors-GPI anchoring is essential for yeast, survival and most GPI anchors of yeast possess an inositolphosphorylceramide, which is introduced into the anchor by lipid remodeling (32). Cwh43⌬ and other mutants, which do not incorporate ceramides into GPI anchors, are viable (33)(34)(35), indicating that ceramide is not an essential component of GPI anchors. The incorporation of ceramides into GPI anchors supposedly is analogous to the IPC synthase reaction carried out by Aur1p, but it seems to be independent of Aur1p because it cannot be inhibited by AbA (36). As 4⌬.SLC1-1 cells should not make ceramides, we wanted to see what lipids they utilize for GPI anchoring. Anchor lipids can be analyzed by metabolically labeling cells with [ 3 H]inositol and then purifying the labeled GPI proteins, digesting them with protease, further purifying the labeled anchor peptides by octyl-Sepharose chromatography, releasing the [ 3 H]inositol-labeled lipid moieties using nitrous acid, and analyzing them by TLC/fluorography (37). As shown in Fig. 3, W4⌬.SLC1-1 cells make significant amounts of the typical anchor lipids IPC-3 and IPC-4, containing PHS-C26:0 and PHS-C26:0-OH as a ceramide moiety, respectively (38). IPC-3 and IPC-4 anchors are resistant to mild base treatment, which hydrolyzes diacylglycerol (DAG)-based anchor lipids such as pG1 but not ceramide-based ones (supplemental Fig. S4B, lanes 5 and 6). Controls indicate that no lipids are seen when peptides are not treated with nitrous acid (Fig. 3, lane 9), indicating that the lipids in lanes 1-8 truly are anchor lipids and are not contaminating free lipids. Radioscanning shows that  W4⌬.SLC1-1, lcb1⌬ SLC1-1, and its corresponding parental WT line SJ21R were precultured in LM, and W4⌬.LAC1 was grown in LM containing raffinose as carbon source. 20 A 600 units of exponentially growing cells were preincubated for 10 min and labeled with 80 Ci of [ 3 H]inositol for 120 min at 30°C in SDaaUA or SRaaUA in the presence of the indicated inhibitors. Proteins were delipidated, and GPI anchor peptides were prepared and treated with nitrous acid (lanes 1-8) or control-incubated (lane 9). The following are indicated below each lane: the radioactivity in lipid extracts containing free lipids (Ci F), the radioactivity in lipids released from GPI anchor peptides (cpmϫ10 Ϫ3 A), as well as the radioactivity in ceramide-containing anchors appearing as IPC-3 and IPC-4 (having three or four hydroxyls in their ceramide moiety), expressed as a percentage of total radioactivity per lane (% IPC). The total anchor lipids were spotted onto the TLC when less than 10 5 cpm had been obtained; otherwise 10 5 cpm/lane were spotted. Lane 1 contains an aliquot of the lipid extract of SJ21R WT cells showing that its main free IPC is IPC-4, whereas GPI anchors also contain a large proportion of IPC-3. The TLC was developed in solvent 55:45:5, and spots were visualized by fluorography. Myr, myriocin. FEBRUARY 25, 2011 • VOLUME 286 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 6771
W4⌬.SLC1-1 cells incorporate significant amounts of [ 3 H]inositol into GPI anchors, amounts that are comparable with the ones found in W4⌬.LAC1 or S21JR WT cells (Fig. 3, lanes 2, 6, and 8) or in BY4742-derived mutants plc1⌬ or pgc1⌬, which were found to not have any GPI remodeling defect (supplemental Fig. S4C, lanes 5-7). As may be expected, lcb1⌬ SLC1-1 cells, although incorporating [ 3 H]inositol into lipids less efficiently than the corresponding SJ21R WT cells, made only pG1 type anchors (Fig. 1C), i.e. anchors with a C26:0-containing DAG but no anchors containing IPC-3 or IPC-4 ( Fig. 3, lanes 7 and 8). In W4⌬.SLC1-1 cells, the incorporation of [ 3 H]inositol into free lipids and GPI anchors could be reduced by myriocin (Fig. 1A), an inhibitor of the serine palmitoyl transferase, and to a lesser extent by AbA (Fig. 1B), but the ratio of ((IPC-3 ϩ IPC-4) total anchor lipids) did not change (% IPC) (Fig. 3, lanes 2-5). These two findings can be explained in the sense that (a) in WT cells, only a very small percentage of LCBs is utilized for GPI anchors, and such small amounts are made even in the presence of myriocin and (b) Aur1p is not involved in the addition of ceramides to GPI anchors (36). Only normal ceramide anchors were also found in a fraction of more polar anchor lipids and in other types of 4⌬ cell lines (supplemental Fig. S4).
For better characterization, lipid extracts from W303.4⌬ and W4⌬.SLC1-1 cells were analyzed by LC-ESI-MS. Searching for ions potentially corresponding to lipid a (Fig. 4, asterisk), we found that W303.4⌬ cells contained good amounts of ions corresponding to the expected mass of inositol-phospho-PHS18 (m/z ϭ 558.6) and inositol-phospho-PHS20 (m/z ϭ 586.6) (  5B). Substantial amounts of the same compound were also found in the other 4⌬ strains tested (supplemental Fig. S6A). Inositol-phospho-PHS was recently found in the fatty acid elongation mutants elo2⌬ and elo3⌬ and was named lyso-IPC (39). (The elo3⌬ mutant has difficulty making ceramides because C26-CoA and C24-CoA,  the optimal substrates for Lac1p/Lag1p, are lacking (40)). As shown in supplemental Fig. S6B, a direct comparison showed that the relative percentage of lyso-IPCs among the total of negative ions was 35-fold higher in W303.4⌬ cells than WT cells and 5-fold higher than in elo3⌬ cells. Lyso-IPCs are expected to be mild base-resistant and to migrate in TLC significantly less than IPCs, as is the case for lipid a (Fig. 4). Lyso-IPCs therefore most likely correspond to the major mild base-resistant [ 3
The LC-ESI-MS analysis also showed that 4⌬ cells still contain small amounts of IPC44 and IPC46 with three, four, or five hydroxyl groups in their ceramide moiety, the ratio of intensities of IPC/PI in 4⌬ mutants being 1.3-1.4% of the ratio observed in WT cells (Fig. 6, A and B). Fragmentation of IPCs in 4⌬ cells resulted in characteristic neutral losses of 162 Da (inositol-H 2 O) and 180 Da (inositol), confirming their identity as IPCs (not shown). The drastic reduction in the IPC/PI ratio is the result of a 40-fold reduction in IPC intensities and a concomitant ϳ2-fold increase in PI intensities. This can be explained by the fact that in WT cells, 40% of PI is used to make complex sphingolipids ( Fig. 1) (41). IPC-3 is usually very minor in WT cells but was accounting for more than a third of IPCs in 4⌬ cells (Fig. 6B). Moreover, 4⌬ cells contained 15.5 times more PIs with a total of 42 and 44 carbon atoms in the two fatty acids than WT (Fig. 6C). This is expected because only few very long chain fatty acids can be utilized for making sphingolipids in 4⌬ cells, and they thus spill over into PI. Interestingly, the PI44 species were of similar abundance whether or not SLC1-1 was present, indicating that the increased fitness of SLC1-1-bearing strains is not due to higher amounts of C26-containing PI but rather to the presence of some PIЉ (having C26:0 in sn-2) together with PIЈ (i.e. PI with C26:0 in sn-1), the latter being the only species present in cells not bearing SLC1-1.
4⌬ Cells Make Ceramides from Exogenously Added Long Chain Bases-The presence of small amounts of normal IPCs and ceramides in 4⌬ cells is puzzling because cells lack all of the known ceramide synthases. We have struggled with the question of whether the cells themselves make these lipids or whether they are taken up from the media. To resolve this, we tried to force 4⌬ cells to make IPCs from exogenously added precursors by specifically blocking either fatty acid or LCB biosynthesis. Although cerulenin, an inhibitor of fatty acid synthase, killed the cells in the presence of exogenous fatty acids, cells were quite resistant to myriocin (see below, Fig.  8A). As shown in Fig. 7A, the IPCs made by 4⌬ cells in the presence of myriocin and LCBs containing either 18 or 20 carbon atoms were quite different. In the presence of DHS18, the cells mainly made IPCs with 44 carbon atoms and three, four, or five hydroxyls in the ceramide; in the presence of DHS20, they made mainly IPCs with 46 carbon atoms. IPC44 and IPC46 are produced by the condensation of a C26:0 fatty acid with an LCB18 or LCB20, respectively. The predominance of IPC46s in cells fed with DHS-C20 also shows that myriocin efficiently blocked the endogenous synthesis of LCB18. Incidentally, the data argue that cells do not make IPC-5 species from exogenous LCBs. Myriocin also strongly blocked the biosynthesis of lyso-IPCs in 4⌬ cells (Fig. 7B), as may be expected from supplemental Fig. S5. Exogenous DHS18 and DHS20 both were replacing endogenous LCBs efficiently for IPC biosynthesis (Fig. 7A), but exogenous DHS20 was incorporated into lyso-IPCs more efficiently than DHS18, or else lyso-IPC20s were more stable than lyso-IPC18s (Fig. 7B). Also, endogenous LCB18 was more prevalent in IPCs than endogenous LCB20, whereas the inverse was true for lyso-IPCs (Figs. 5A and 7; supplemental Fig. S6A), although both IPCs as well as lyso-IPCs most likely are made by Aur1p (Fig. 4B, lanes 13 and 14) (13). Moreover, lyso-IPCs always contained PHS, not DHS, even if derived from exogenous DHS (not shown).
Although the data in Fig. 7 clearly demonstrate that 4⌬ cells can make ceramides, they do not exclude that part of the IPCs of 4⌬ cells are derived from sphingolipids contaminating the synthetic complete LM. Note that IPCs could be taken up, hydrolyzed by Isc1p, and then utilized by Aur1p during a metabolic labeling experiment with [ 3 H]inositol. However, further arguments against an external source of ceramides in 4⌬ cells are the following. 1) lcb1⌬ SLC1-1 cells grown in the same medium cells do not make those lipids (supplemental Fig. S7), nor do they incorporate ceramides into GPI anchors (Fig. 3, lane 7). 2) Previous attempts to rescue the growth phenotype of lcb1⌬ cells by adding ceramides (DHS-C6, DHS-C24, PHS-C26-OH) or M(IP) 2 C to the growth medium were unsuccessful (42). 3) Growth media extracted with ethyl ether to remove apolar lipid contaminants supported normal cell growth of 4⌬ cells, and the deliberate supplementation of growth media with crude, sphingolipid-containing lipid extracts from WT cells did not improve the growth of 4⌬ cells (not shown). 4) In some experiments, myriocin treatment abrogated biosynthesis of all mild base-resistant IPCs in 4⌬ cells (supplemental Fig. S5, lanes 1-4). In summary, all our findings argue in favor of an unknown further pathway in 4⌬ cells allowing them to synthesize small amounts of ceramides.
To see whether ceramides could be generated non-enzymatically, we incubated [ 3 H]C16:0 or [ 3 H]C16:0-CoA with LCBs during 4 h at physiological pH with either yeast lipid extract or boiled yeast microsomes as a support. No generation of ceramides was observed (not shown).
We found that we could not clone W303.4⌬ cells by micromanipulation, possibly because they die upon physical separation of daughter from mother cells. We therefore wanted to be sure that the strains we were looking at were not contaminated by some other yeast species. Low stringency PCR followed by DNA sequencing showed, however, that all DNA sequences we obtained are present in the genome of S. cerevisiae, making it highly unlikely that another species is contaminating our 4⌬ strains (supplemental Fig. S8).
Overexpression of AUR1 Can Rescue Viability of YPK9.4⌬ Cells-The presence of high amounts of lyso-IPCs in 4⌬ cells raised the possibility that the ability to make high amounts of these lipids may decide whether a given genetic background can tolerate the deletion of all ceramide synthases, as is the case for W303 but not for YPK9. DNA sequencing of the open reading frame of the genomic AUR1 in W303.4⌬ (FBY958-Lnew) as well as in Y2⌬.LAC1, a YPK9 lag1⌬ lac1⌬ strain unable to loose the covering pBM150-LAC1 plasmid, showed that both strains had the identical AUR1 coding sequence. This did not exclude that W303 contained higher amounts of the IPC synthase Aur1p or of Kei1p, the second essential subunit of the IPC synthase (43). As overexpression of AUR1 can suppress the growth defect of kei1-1 mutants, we tried to see whether the introduction of extra copies of AUR1 into Y4⌬.LAG1 cells would allow them to loose the pBM150-LAG1 plasmid on FOA. This indeed was the case. The thus generated Y4⌬.AUR1 strain also made significant amounts of lyso-IPCs (supplemental Fig. S6A). Assuming that the affinity of Aur1p for free LCBs is much lower than that for ceramides, one would predict that lyso-IPCs are made efficiently only by cells having high concentrations of LCBs and low levels of ceramides, conditions that are met in all 4⌬ cells. The fact that Y4⌬.AUR1 cells are viable, whereas YPK9.4⌬ cells are inviable, also raises the possibility that lyso-IPCs, like PIЉ, can act as a substitute for normal IPCs, but we cannot exclude that the overexpression of Aur1p allows for more efficient synthesis of essential IPCs rather than lyso-IPCs.

4⌬.SLC1-1 Cells Are Resistant to High Concentrations of Aureobasidin A-Various single-point mutations in AUR1
were reported to make WT cells resistant to very high concentrations (Ͼ20 or 25 g/ml) of AbA, i.e. 100 -500-fold more resistant than WT cells, and this without affecting their sensitivity to other drugs (44,45). This argues that AbA is a specific inhibitor of IPC synthase. AbA suppresses the biosynthesis of lyso-IPCs in W303.4⌬ (13) as well as W4⌬.SLC1-1 cells (Fig. 4B) and blocks growth of W303.4⌬ cells (13). As shown in Fig. 8A, 4⌬.SLC1-1 cells are partially resistant to AbA, but AbA clearly reduces their cloning efficiency. However, W4⌬.SLC1-1 cells could be further propagated on high concentrations of AbA (Fig. 8B), much in contrast to W303.4⌬ cells lacking SLC1-1 (13). These data suggest that lyso-IPCs or else the small amounts of IPCs/MIPCs contribute to the robustness of W4⌬.SLC1-1 cells but that the introduction of SLC1-1 has rendered these sphingolipids dispensable for cell survival. 4⌬ cells were less sensitive to myriocin than to AbA (Fig. 8A), in keeping with the finding that myriocin could not suppress the addition of ceramides to GPI anchors (Fig. 3, lane 3) nor the synthesis of small amounts of IPC-4 (Fig. 4B, lane 12), probably because myriocin cannot completely block LCB biosynthesis.
AUR1 Is Not Required for the Incorporation of C26:0 Fatty Acids into Phosphatidylinositol-In the original study identifying SLC1 as an sn-2-specific acyl-glycerol-phosphate acyltransferase, the authors found that C26:0 fatty acids were incorporated by SLC1-1p only into PI but not into other glycerophospholipids (14). The simplest hypothesis to explain this finding is that SLC1-1p makes some phosphatidic acid with C26:0 in sn-2 (PAЉ), that this PAЉ is transformed into CDP-DAGЉ, and that only PI synthase (PIS1) is able to utilize CDP-DAGЉ but phosphatidylserine synthase (CHO1) is not. Based on the partial AbA sensitivity of 4⌬.SLC1-1 cells, we considered the alternative hypothesis that the PAЉ would be degraded by Pah1p to DAGЉ containing C26:0 in sn-2 and that in the absence of ceramides, DAGЉ may be mistaken by IPC synthase Aur1p as a ceramide so that Aur1p would transfer inositolphosphate from a PI onto this DAGЉ, thereby generating PIЉ. To test for this, we exploited the fact that the presence of MPIЉ is evidence for the synthesis of PIЉ. Indeed, although PIЈ (i.e. a PI with C26:0 in sn-1) does not serve as an acceptor for mannoses transferred by Csg1p or Chs1p, PIЉ (with C26:0 in sn-2) does (15,46 (Fig. 9A, lane 2, and 9C, lane 16). As expected, adding myriocin or AbA to the labeling reaction greatly diminishes the labeling of these lipids (Fig. 9A, lanes 2, 4, and 6). (The origin of the band migrating as the upper of the two MIPC bands after mild base hydrolysis is not known (Fig. 9A, lanes 3, 5, and 7)). In pmi40 lcb1⌬ SLC1-1 cells, no labeling of MIPC or M(IP) 2 C is observed, but one can see three mild base-sensitive bands migrating at and slightly above the position of MIPC; these bands most likely correspond to MPIЉ (Fig. 9B, boxed, lanes 12 and 13, and 9C,  lane 17). This MPIЉ is sensitive to phospholipase A 2 (not shown) and is labeled equally strongly also in the presence of myriocin or AbA (Fig. 9B, lanes 8, 10, and 12). Also, several other mild base-sensitive and one mild base-resistant band are present in pmi40 lcb1⌬ SLC1-1 but not pmi40 cells; some may correspond to lyso-MPIЉs. pmi40 lcb1⌬ SLC1-1 precultured and/or labeled in the presence of PHS generated both MIPCs and MPIЉ (Fig. 9D). In this case also, AbA blocked the biosynthesis of the major species of MIPC but not of MPIЉ (Fig. 9D, lanes 23 and 25). Thus, Aur1p is not required to make mannosylated PIЉ, and hence to make PIЉ, and this suggests that the exclusive incorporation of C26 into PI seen in SLC1-1 cells reflects the fact that Pis1p can use CDP-DAGЉ with a C26 in sn-2, whereas Cho1p cannot. Alternatively, C26 may be introduced into PI through a SLC1-1-dependent lipid deacylation-reacylation cycle as Slc1p was shown to acylate lyso-PI and to prefer this substrate to lysophosphatidylcholine or lysophosphatidylethanolamine (47). At any event, the partial AbA sensitivity of Y4⌬.SLC1-1 and W4⌬.SLC1-1 (Fig. 8A) cannot be explained by an effect of AbA on the biosynthesis of PIЉ.
Mannosylation of C26:0 Containing Phosphatidylinositol Helps the Survival of Sphingolipid-deficient Cells-Individual deletions of either CSG1 or CSH1, the functionally redundant IPC-specific mannosyltransferases, have little effect on MIPC biosynthesis, whereas simultaneous deletion of both genes totally abolishes mannosylation of IPC (Fig. 1B). Also, csg1⌬ csh1⌬ cells grow normally, showing that mannosylation of Starting with a cell suspension with an A 600 of 10, serial 10-fold dilutions were prepared and plated on LM plates containing AbA, myriocin, or methanol (MeOH). Plates were incubated at 30°C and photographed after 3 and 6 days. B, cells from the colonies growing on AbA (next to the white asterisk in A) were streaked out onto new plates containing or not 2.5 g/ml AbA, together with W303 WT cells. Plates were incubated at 24°C and photographed after 3 days (left and middle); growing W303.4⌬ cells were again taken from the middle plate, streaked onto a fresh plate, and once more cultured for 3 days (right).
IPCs and formation of MIPC are not required for cell survival (20,21). We assumed that PIЉ is mannosylated by Csg1 or Csh1p. To see whether the MPIЉ forms generated by SLC1-1 cells lacking sphingolipids are important for cell survival, we treated csh1⌬, csg1⌬, csh1⌬ csg1⌬, and WT cells expressing SLC1 or SLC1-1 with myriocin. As shown in Fig. 10, all mutants not expressing the suppressor allele SLC1-1 were highly sensitive to myriocin. SLC1-1 was able to rescue cells expressing CSG1, whereas csg1⌬ SLC1-1 and csg1⌬ csh1⌬ SLC1-1 cells were completely unable to grow. This result suggests that mannosylation of PIЉ, i.e. the biosynthesis of MPIЉ, is essential for the viability of cells lacking sphingolipids, and the presence of the minor mannosyltransferase Csh1p is not sufficient for cell survival on myriocin, either because it is catalytically less active than Csg1p or because it does not utilize PIЉ as a substrate. Lipid extract of 4⌬ cells contained negative ions of m/z 1137, which corresponds to the expected m/z of mannosyl-PI44:1 with oleic acid in sn-1 and C26:0 in sn-2, albeit in quantities too low to allow fragmentation (not shown).
W4⌬.SLC1-1 Cells Are Heat Shock-intolerant-Free LCBs, LCB phosphates, as well as ceramides get elevated when yeast cells are heat-shocked by a shift from 24 to 37°C (18,48,49). As summarized in Table 1, lcb1⌬ SLC1-1 cells cannot grow at 37°C and other stress conditions (50). Similarly, lcb1-100 mutant cells show no increase in the levels of PHS and DHS during heat stress, cannot grow at 37°C, and show reduced survival during stronger heat stresses (51)(52)(53). Moreover, lcb1-100 mutant cells cannot initiate translation of Hsp70 mRNAs after a heat-induced global translation arrest (54). W303.4⌬ cells and W303 lag1 ts lac1⌬ cells carrying a thermolabile LAG1 allele grow much less well at 37°C than at 24°C (13,25). This suggested that ceramide and/or derivatives thereof are also required for survival and growth at elevated temperatures. Here we found that W4⌬.SLC1-1 cells grow quite well at 37°C (Fig. 2B), indicating that the need for ceramide and/or derivatives thereof can be overcome by some  SLC1-1-dependent lipid. To explore whether SLC1-1-dependent lipids also confer resistance to a stronger heat stress, we tested the plating efficiency of W4⌬.SLC1-1 cells having been cultured at 24°C and then having been shifted for various times to 44°C following the protocol previously used for lcb1-100 cells by Friant et al. (53) (Fig. 11A). W303 WT cells resisted this heat treatment despite being notorious for harboring a truncated SSD1 allele, Ssd1p being a cochaperone for Hsp104p, a mutation that renders W303 more heat-sensitive than other WT strains (55) (Fig. 11). In contrast, W4⌬.SLC1-1 cells were found to be sensitive to a temperature shift to 44°C. The time required to kill cells varied from 12 to 30 min. Curiously, cells survived better when plated on PHS then without (not shown). Moreover, the overexpression of UBI4, encoding for ubiquitin, could not rescue the heat sensitivity of W4⌬.SLC1-1 cells (Fig. 11B), whereas the same plasmid had been reported to rescue lcb1-100 cells under the same heat stress conditions, presumably by accelerating the ubiquitinylation and proteasomal degradation of aggregated proteins (53). Inspection of the overview in Table 1 shows that only W4⌬.SLC1-1 cells grow at 37°C. This suggests that for cells lacking ceramide and IPCs, both an SLC1-1-dependent lipid and significant PHS levels need to be present to allow cells to grow at 37°C. Our data further suggest that different kinds of damage are caused by severe heat stress in W4⌬.SLC1-1 and lcb1-100 cells. W4⌬.SLC1-1 cells having very high levels of LCBs may not have any problem inducing translation of mRNAs for heat shock proteins and therefore are not helped by UBI4 overexpression, but they must be unable to survive 44°C because of the almost complete lack of ceramides and complex sphingolipids. In contrast, membranes of lcb1-100 SLC1-1 cells, which still have considerable amounts of complex sphingolipids, may better resist the high temperatures but have difficulty reinitiating heat shock proteins because they are unable to raise LCB levels during heat stress (52). Altogether the data illustrate that even the concomitant presence of large amounts of PHS, of small amounts of IPCs, of lyso-IPCs, and of SLC1-1-dependent lipids cannot compensate for the absence of the normal sphingolipids during a strong heat stress at 44°C.

DISCUSSION
The present report identifies the most abundant abnormal sphingolipids of lag1⌬ lac1⌬ ypc1⌬ ydc1⌬ cells as lyso-IPCs. Lyso-IPCs indeed are expected to be metabolically labeled with radioactive inositol, DHS, and phosphate, as reported for lipids a and b, X1, and X2 before (25,26). The synthesis of lyso-IPCs is blocked by AbA ( Fig. 4B; supplemental Fig. S5) (13), and lyso-IPCs therefore seem to be made by Aur1p. Aur1p has previously been shown to use not only the physiological C26-and C24-containing ceramides but also ceramides with shorter fatty acids of 2, 6, or 16 carbon atoms (13,56), and this report clearly shows that in the absence of other substrates, Aur1p can even use free LCBs, albeit with a preference for LCB-C20 over LCB-C18. However, we could not find ions corresponding to mannosylated lyso-IPCs, suggesting that lyso-IPCs are not a substrate for Csg1p or Csh1p.
This study was triggered by the stunning observation that 4⌬ cells continue to add ceramides to GPI anchors (Fig. 4). Comparison of the relative amounts of [ 3 H]inositol in GPI proteins having ceramide anchors (Fig. 3, lanes 2 and 6) and in free IPCs (supplemental Table SV) indicates that the loss of ceramide synthases brings IPCs down 15-25-fold, whereas ceramide-containing GPI anchors are down only 2-fold (see supplemental calculation). It therefore appears that cells lacking known ceramide synthases continue to make small amounts of ceramides and incorporate them preferentially into GPI anchors, whereas they cease by and large to make IPCs. Two alternative hypotheses can explain the preferential  incorporation of ceramides into GPI anchors. Either the Cwh43p ceramide remodelase has a higher affinity for ceramides than Aur1p, or the ceramides of 4⌬ cells are generated by a process, which channels them preferentially into the ceramide remodelase. At any event, the TLC mobility of IPCs present in GPI anchors of 4⌬ cells and WT cells is the same, suggesting that the unknown enzyme generating ceramides in 4⌬ cells makes ceramides containing very long chain fatty acids, as are found in WT GPI anchors (57). It may be that Cwh43p, the ceramide remodelase, has a high specificity for C42-and C44-ceramides because it accepts neither ceramides with shorter fatty acids that are generated in 4⌬.Lass5 cells (supplemental Fig. S4C, lane 3) (29) nor non-acylated LCBs.
On the other hand, as stated before, Aur1p accepts ceramides with C20:0 down to C2:0 fatty acids as substrates (13,29,56). Thus, if the unknown ceramide synthase of 4⌬ cells would make ceramides with shorter fatty acids, they should be transformed into IPCs. As this is not the case, we can assume that the unknown ceramide synthase of 4⌬ cells is specific for very long chain fatty acids. Aureobasidin A at 2.5 g/ml stops the growth of W303.4⌬ cells (13), whereas it slows, but does not stop, the growth of W4⌬.SLC1-1 cells (Fig. 8B). The growth inhibitory effect of AbA cannot be due to an influence on GPI anchoring as it was shown that AbA does not block ceramide incorporation into GPI anchors even at concentrations up to 10 g/ml (36). It is also becoming clear from our studies that Aur1p is not required for making PIЉ (Fig. 9), and so the inhibitory effect of AbA on the growth of 4⌬.SLC1-1 cells must be achieved through blocking the synthesis of IPCs and/or lyso-IPCs, not of PIЉ and MPIЉ. Although the immediate toxic effect of AbA in WT cells is most likely due to the toxicity of accumulating ceramides (13,25), this kind of toxic effect of AbA can safely be excluded in 4⌬ cells. The data therefore suggest that lyso-IPCs and/or IPCs make an essential contribution to the cell viability of W4⌬.SLC1-1 cells, a contribution that becomes dispensable only if cells harbor SLC1-1 (Fig. 8B).
PIЉ only represents 1-2% of PI in SLC1-1 lcb1⌬ cells (15). Although it seemed reasonable to assume that lcb1⌬ SLC1-1 cells need PIЉ, MPIЉ, and/or inositol-phospho-MPIЉ for cell survival, it was not formally excluded that that they only require PAЉ or DAGЉ or just a pathway that reduces toxic levels of C26:0 (15). However, the fact that mannosylation of PIЉ is important for survival of SLC1-1 cells lacking LCBs (Fig. 10) strongly suggests that MPIЉ and/or inositol-phospho-MPIЉ are the life-saving lipids in cells lacking complex sphingolipids. In WT cells, about 1% of PI carries a C26 fatty acid in sn-1 (46), and the percentage of this PIЈ is massively increased in lac1⌬ lag1⌬ cells (26,29) (Fig. 6C). However, it seems that PIЈ is not able to rescue lcb1⌬ or YPK9.4⌬ cells, but only PIЉ, carrying a C26:0 in sn-2. It may be speculated that the PIЈ fails to rescue lcb1⌬ or YPK9.4⌬ cells not because its biophysical properties are very different from PIЉ but because it is not a substrate for Csg1p and therefore cannot be mannosylated to yield the essential mannosylated forms of PI. Although SLC1-1 expression had the same life-saving effect on YPK9.4⌬ cells as on lcb1⌬ cells, we could not directly confirm that mannosylation of PIЉ is also crucial for the survival of 4⌬ cells because 4⌬ cells are not very robust and the introduction of additional mutations is difficult. Further efforts will be needed to identify the enzyme producing the ceramides in 4⌬ cells.