Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
In the yeast Saccharomyces cerevisiae, glycosylphosphatidylinositol (GPI)-anchored proteins play important roles in cell wall biogenesis/assembly and the formation of lipid microdomains. The lipid moieties of mature GPI-anchored proteins in yeast typically contain either ceramide moieties or diacylglycerol. Recent studies have identified that the GPI phospholipase A2 Per1p and O-acyltransferase Gup1p play essential roles in diacylglycerol-type lipid remodelling of GPI-anchored proteins, while Cwh43p is involved in the remodelling of lipid moieties to ceramide. It has been generally proposed that phosphatidylinositol with diacylglycerol containing a C26 saturated fatty acid, which is generated by the sequential activity of Per1p and Gup1p, is converted to inositolphosphorylceramide by Cwh43p. In this report, we constructed double-mutant strains defective in lipid remodelling and investigated their growth phenotypes and the lipid moieties of GPI-anchored proteins. Based on our analyses of single- and double-mutants of proteins involved in lipid remodelling, we demonstrate that an alternative pathway, in which lyso-phosphatidylinositol generated by Per1p is used as a substrate for Cwh43p, is involved in the remodelling of GPI lipid moieties to ceramide when the normal sequential pathway is inhibited. In addition, mass spectrometric analysis of lipid species of Flag-tagged Gas1p revealed that Gas1p contains ceramide moieties in its GPI anchor.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Glycosylphosphatidylinositol (GPI) modification of proteins is conserved among mammals, yeast and protozoa (Ferguson, 1999; Kinoshita and Inoue, 2000; Orlean and Menon, 2007; Pittet and Conzelmann, 2007). In the budding yeast Saccharomyces cerevisiae, approximately 60 proteins are predicted to be cell surface anchored by GPI moieties (Caro et al., 1997). A number of these proteins, including Gas1p, are anchored to the plasma membrane, whereas others appear to be mainly cell-wall-linked (Caro et al., 1997; Hamada et al., 1999; Kapteyn et al., 1999). Yeast GPI-anchored proteins (GPI-APs) are predominantly involved in cell-wall integrity and assembly (De Groot et al., 2005; Klis et al., 2006).
The biosynthesis and attachment of GPI to target proteins occur on the endoplasmic reticulum (ER) membrane. GPIs, which have conserved core structures, are preassembled in the ER in a multi-step pathway prior to being transferred to target proteins. Following preassembly, which involves the sequential addition of glucosamine, an acyl chain, mannose residues and ethanolamine phosphates to phosphatidylinositol (PI), GPI precursors are then attached to ER-translocated proproteins bearing a C-terminal signal sequence for GPI attachment. After the GPI modification of proteins, the acyl group on inositol is typically removed from GPI in the ER (Tanaka et al., 2004; Fujita et al., 2006a).
Mature GPI-APs contain saturated fatty acyl (or alkyl) chains in the lipid moiety of GPI (McConville et al., 1993; Brewis et al., 1995; Benting et al., 1999), even though GPI is synthesized from conventional PI, which usually contain unsaturated fatty acyl chains. Thus, the fatty acyl chain of GPI lipid moieties undergoes a lipid remodelling process, in which the acyl chains are converted from an unsaturated to a saturated form (Fujita and Jigami, 2008; Fujita and Kinoshita, 2010; 2012). In mammalian cells, nearly all mature GPI-APs contain PI with two saturated fatty acyl or alkyl chains after replacement of the unsaturated fatty acid at the sn-2 position of lipid moieties with a saturated chain (Maeda et al., 2007). In contrast, mature GPI-APs in yeast contain two different types of lipid moieties: one is a diacylglycerol with a C26 fatty acid at the sn-2 position, and the other is a ceramide lipid moiety consisting mainly of phytosphingosine (PHS; t18:0) and a C26:0 fatty acid (Conzelmann et al., 1992; Fankhauser et al., 1993; Sipos et al., 1994; 1997; Reggiori et al., 1997). In both types of lipid moieties, the C26 fatty acid may be hydroxylated (Sipos et al., 1997); however, it remains unclear which GPI-APs contain a ceramide-type lipid in the GPI anchor, although ceramide lipids are most commonly found in yeast GPI. An exception is the plasma membrane-localized GPI-AP Gas1p, which is reported to contain the diacylglycerol-type lipid moiety in its GPI anchor (Fankhauser et al., 1993).
In yeast, lipid remodelling is initiated in the ER after the deacylation of inositol; removal of fatty acid from inositol by Bst1p is a prerequisite for GPI lipid remodelling (Ghugtyal et al., 2007). Next, an unsaturated acyl chain at the sn-2 position of diacylglycerol is removed through the action of Per1p, which possesses phospholipase A2 activity, to form lyso-GPI (Fujita et al., 2006b). A C26 saturated acid chain is then transferred to the sn-2 position of diacylglycerol by the O-acyltransferase Gup1p (Bosson et al., 2006). These lipid-remodelling events are essential for the efficient transport of GPI-APs from the ER to the Golgi apparatus and for their association with lipid microdomains, termed lipid rafts (Bosson et al., 2006; Fujita et al., 2006b; Maeda et al., 2007). The diacylglycerol of GPI lipids is often replaced by a ceramide moiety consisting of PHS with a C26 fatty acid, a process that occurs in the ER (Sipos et al., 1997). CWH43 is reported to be involved in the remodelling of GPI lipids to ceramides (Ghugtyal et al., 2007; Umemura et al., 2007). Evidence also suggests that the ethanolamine phosphate groups of GPI glycan moieties play an important role for lipid remodelling to ceramides (Benachour et al., 1999; Zhu et al., 2006). Correctly remodelled GPI-APs are recognized by the p24 complex, incorporated to COPII vesicles at the ER exit site, and then transported from the ER to the Golgi apparatus (Castillon et al., 2009; 2011; Fujita et al., 2011). In the Golgi, the ceramide PHS moiety appears to be further modified with a hydroxy-C26 fatty acid (Reggiori et al., 1997; Sipos et al., 1997).
Lipid microdomains are thought to be sphingolipid- and sterol-rich regions of the plasma membrane, and can be biochemically isolated as a detergent-resistant membrane (DRM) fraction (Brown and Rose, 1992; Simons and Ikonen, 1997). GPI-APs are also one of the major components of lipid microdomains (Brown and Rose, 1992). We previously demonstrated that GPI-APs are rarely found in DRMs in S. cerevisiae per1Δ and gup1Δ cells (Fujita et al., 2006b), strongly suggesting that lipid remodelling is required for the association of GPI-APs with lipid microdomains. GPI-APs appear necessary for the recruitment of several transmembrane proteins, including the high-affinity tryptophan permease Tat2p, to the plasma membrane via lipid microdomains. Notably, Tat2p does not associate with DRMs in either gwt1-10 or gaa1-1 cells, which contain mutated GPI inositol acyltransferase and transamidase proteins respectively (Okamoto et al., 2006). Despite these findings, the effects of GPI lipid remodelling on the localization of transmembrane proteins that associate with lipid microdomains remain to be elucidated.
Here, we report the identification of an alternative pathway for the remodelling of GPI lipids to ceramides based on phenotypic analyses of single- and double-mutant cells of gup1 and cwh43 in S. cerevisiae. Mass spectrometry analysis of Flag-tagged Gas1p revealed that Gas1p, which was previously reported not to be modified by ceramide, contains ceramide moieties in its GPI anchor. We also found that the lipid remodelling of GPI-APs has an effect on the association of transmembrane protein Tat2p with lipid microdomains, as well as GPI-APs.
gup1Δ cwh43Δ double-mutant cells show more severe phenotypes than gup1Δ or cwh43Δ single-mutant cells
Recently, our laboratory and the Conzelmann group reported that PER1, GUP1 and CWH43, which encode GPI phospholipase A2, O-acyltransferase and a putative ceramide remodelase, respectively, are involved in the lipid remodelling of GPI-AP in yeast (Bosson et al., 2006; Fujita et al., 2006b; Ghugtyal et al., 2007; Umemura et al., 2007). Based on these studies, two possible pathways were proposed for the conversion of diacylglycerol to ceramides in GPI lipid moieties (Umemura et al., 2007). The first is a sequential pathway in which a diacylglycerol generated by sequential reactions involving Per1p and Gup1p, and containing a C26 saturated fatty acid (pG1) is converted to a ceramide moiety (inositolphosphorylceramide; IPC) by Cwh43p (Fig. 1A, left panel). The second is a divergent pathway that uses the lyso-PI generated by Per1p as a substrate for two reactions catalysed by Gup1p and Cwh43p (Fig. 1A, right panel). To determine the functional position of Cwh43p in the yeast GPI lipid remodelling pathway and investigate the molecular mechanism of ceramide conversion, we applied a genetic approach by constructing double-deletion mutants of PER1, GUP and CWH43.
As per1Δ, gup1Δ and cwh43Δ mutants in the BY4741 genetic background (Brachmann et al., 1998) exhibit only minimal growth defects, we constructed deletion mutant strains in the W303 genetic background (Thomas and Rothstein, 1989). The per1Δ, gup1Δ and cwh43Δ mutants in the W303 genetic background showed more severe phenotypes, characterized by increased sensitivity to high temperature and calcofluor white, a fluorescent stain that binds to cell wall containing chitin and impairs yeast cell growth, than the corresponding mutants in the BY4741 genetic background (Fig. S1). Therefore, we also constructed double-deletion mutants in the W303 genetic background and investigated their phenotypes. In the W303 genetic background, gup1Δ cwh43Δ double-mutant cells displayed a slow-growth phenotype when cultured on YPAD medium at 30°C (Fig. 1B). In addition, the gup1Δ cwh43Δ double-mutant cells had higher sensitivity to high temperature (35°C), calcofluor white and SDS, a detergent that disrupts the cell wall and/or plasma membrane, than gup1Δ or cwh43Δ single-mutant cells (Fig. 1B). The gup1Δ cwh43Δ double-mutant cells were also hypersensitive to cycloheximide (CHX), which disturbs protein synthesis, whereas the gup1Δ and cwh43Δ single-mutant cells were not (Fig. S2). Notably, the per1Δ gup1Δ double-mutant cells grew better in the presence of SDS than the per1Δ single-mutant cells (Fig. 1B). The phenotype of the per1Δ cwh43Δ double-mutant cells was comparable to that of the per1Δ single-mutant cells. If the lipid-remodelling pathway operated exclusively sequentially, both gup1Δ and gup1Δ cwh43Δ cells should exhibit the same phenotypes. However, these results suggest that the pathway is branched or that a bypass pathway exists, in which lyso-PI is changed to a ceramide by the action of Cwh43p.
Analysis of catalytically inactive Gup1p and Cwh43p mutants
As it is possible that the deletion of GUP1 and CWH43 might disrupt other lipid-remodelling pathways or even other unrelated pathways, we constructed gup1 and cwh43 mutants in which catalytically dead versions of Gup1p and Cwh43p were produced. It was reported that Gup1pH447A and Cwh43pR882A point-mutant proteins are stable, but catalytically inactive, and that the phenotypes of gup1-H447A and cwh43-R882A mutant cells are similar to those of gup1Δ and cwh43Δ deletion mutant cells respectively (Bosson et al., 2006; Umemura et al., 2007). Here, the phenotypes of gup1Δ cwh43Δ double-mutant cells simultaneously expressing the Gup1pH447A and Cwh43pR882A point-mutant proteins were investigated and found to be nearly the same as those of gup1Δ cwh43Δ cells (Fig. S3A). In particular, the point-mutant cells displayed a slow-growth phenotype when cultured on YPAD medium at 30°C and also had increased sensitivity to high temperature (35°C), calcofluor white and SDS than gup1Δ cwh43Δ cells in which either the genes for gup1-H447A and CWH43 or GUP1 and cwh43-R882A were simultaneously expressed (Fig. S3A). In all strains, it was confirmed that the wild-type and mutant Gup1p and Cwh43p proteins were produced, although the amount of these proteins was different between the strains; the results of immunoblot analysis indicated that the Gup1pH447A protein was less stable than wild-type Gup1p and that the production level of mutant Cwh43pR882A was higher than that of wild-type Cwh43p in the examined strains (Fig. S3B). These results confirmed that the phenotypes observed in the gup1Δ cwh43Δ double-mutant cells were due to loss of the catalytic activity of the corresponding proteins and not due to other defects, such as the destabilization of protein complexes.
Analysis of GPI lipid species obtained from GPI-anchored proteins
We next examined the lipid species of GPI-APs purified from the single- and double-mutant cells. Cells were first labelled with tritiated inositol, and lipid moieties of GPI-APs were then prepared and analysed by thin-layer chromatography (TLC). In wild-type cells, the lipid species were mainly identified as IPC/B, consisting of PHS and a C26:0 fatty acid, and pG1, consisting of PI with a C26:0 fatty acid at the sn-2 position (Fig. 2, lane 1). The per1Δ cells accumulated PI with a short unsaturated fatty acid (pG2) due to the defect in GPI-PLA2, together with a much smaller amount of IPC/C, consisting of PHS and a hydroxylated C26:0 fatty acid (Fig. 2, lane 2), as was reported previously for this mutant (Fujita et al., 2006b). The gup1Δ cells accumulated lyso-PI and low amounts of IPC/B due to the defect in the acylation of lyso-PI at the sn-2 position (Fig. 2, lane 3), a finding that was consistent with previous reports (Bosson et al., 2006; Fujita et al., 2006b). In contrast, the cwh43Δ cells predominantly accumulated pG1 and lyso-PI, but neither IPC/B nor IPC/C were detected, suggesting a complete loss of ceramide conversion in these cells (Fig. 2, lane 4). This speculation is also consistent with previous findings (Ghugtyal et al., 2007; Umemura et al., 2007), suggesting that the profile of GPI lipid species from strains in the W303 genetic background is the same as those in the BY4741 genetic background. It is noteworthy that the gup1Δ cwh43Δ double-mutant cells completely lost the ceramide species that was detected in gup1Δ single-mutant cells (Fig. 2, lanes 3 and 7). This result supports the existence of an alternative pathway for lipid remodelling that involves Cwh43p in S. cerevisiae.
Release of Gas1p in gup1Δ, cwh43Δ and gup1Δ cwh43Δ cells
The β-1,3-glucanosyltransferase Gas1p is one of the most abundant and well-characterized GPI-APs in S. cerevisiae (Conzelmann et al., 1988; Popolo and Vai, 1999). N-linked ER-type Man8 oligosaccharides, O-linked Man1 residues, and a GPI anchor are transferred to the primary 60 kDa translation product of Gas1p in the ER, yielding an immature 105 kDa polypeptide (ER form). Further modification of the ER form oligosaccharide chains occurs in the Golgi, generating the mature 125 kDa form of Gas1p (Golgi form). Gas1p is predominantly located in the plasma membrane, but it is also linked to cell wall β-1,3-glucans (de Sampaïo et al., 1999).
We examined the maturation and secretion of Gas1p by immunoblotting with an anti-Gas1p polyclonal antibody. Compared with wild-type cells, intracellular Gas1p was clearly decreased in gup1Δ cells, but was detected at high levels in the medium (Fig. 3, lanes 1, 2, 5 and 6), as reported previously (Fujita et al., 2006b). The amount of intracellular Gas1p purified from cwh43Δ cells was comparable to that obtained from wild-type cells, consistent with the findings of Umemura et al. (2007), although the ER form accumulated at higher levels in the mutant cells (Fig. 3, lane 3). In gup1Δ cwh43Δ double-mutant cells, intracellular levels of the Golgi form of Gas1p were hardly detectable, and a significant amount of Gas1p had been released into the medium (Fig. 3, lanes 4 and 8). The effect of cwh43 deletion on the release of Gas1p in gup1Δ cells is again consistent with the existence of an alternative lipid-remodelling pathway.
Analysis of GPI lipid species of Gas1p
The lipid moiety of the GPI anchor in Gas1p is reported to be a diacylglycerol-type lipid containing a C26 fatty acid, termed the pG1 form (Fankhauser et al., 1993). Therefore, we anticipated that the deletion of CWH43 would not affect the localization of Gas1p, even in gup1Δ mutant cells. However, CWH43 deletion in gup1Δ mutant cells decreased the retention of Gas1p (Fig. 3). Two possibilities may account for this phenomenon. The first is that CWH43, in addition to GUP1, might be involved in the remodelling of GPI lipid moieties in Gas1p to the pG1-type, while the second is that the lipid moiety of Gas1p might not be a diacylglycerol-type, but in fact is a ceramide-type, and that both GUP1 and CWH43 are required for the complete maturation of the Gas1p lipid moieties. To elucidate this issue, we used a triple quadrupole mass spectrometer (MS) with a chip-based ionization source to analyse the GPI lipid species of Gas1p. In this experiment, we used mutant strains constructed in the BY4741 genetic background, as we were unable to obtain sufficient signal intensity in the MS analysis of GPI lipid species of Gas1p obtained from strains in the W303 genetic background.
Flag-tagged Gas1p was expressed under control of its own promoter in wild-type cells and was then extracted from the plasma membrane, purified using anti-Flag-beads, applied to SDS-PAGE, transferred to polyvinylidene difluoride (PVDF) membrane, and then subjected to sodium nitrite treatment to release PI from GPI by deamination of glucosamine residues (Fontaine et al., 2003). The released lipids were selectively analysed by neutral-loss scanning of 277 Da (phosphoryl inositol + NH4) in positive-ion mode, and a precursor-ion peak (m/z 955) was detected (Fig. 4A). MS2 and MS3 analyses of the peak revealed that the GPI lipid moiety of Gas1p was not PI, but was IPC, consisting of PHS and a saturated C26 fatty acid (t18:0-C26:0 phytoceramide) (Fig. 4A).
We also purified Flag-Gas1p from cwh43Δ cells and analysed the lipid released from the tagged protein by precursor-ion scanning of 241 Da (phosphoryl inositol − H2O). The peak patterns clearly differed from those of the wild-type cells, suggesting that the lipid moieties were not IPC, but rather PI (Fig. 4B). MS2 analyses of the three precursor-ion peaks (m/z 893, 921 and 949) from cwh43Δ cells revealed that the peaks corresponded to PIs with fatty acyl moieties of C14:0-24:0, C14:0-26:0 and C16:0-26:0 respectively (Fig. 4B). This result indicates that Cwh43p remodels the PI moiety of Gas1p to ceramide.
gup1Δ cwh43Δ cells show growth defects on low-tryptophan medium
We previously reported that gwt1-10 cells exhibit decreased uptake of tryptophan due to mislocalization of Tat2p, a high-affinity tryptophan permease that associates with lipid microdomains, and are sensitive to tryptophan concentrations in the growth medium (Okamoto et al., 2006). We have also reported that Gas1p does not associate with lipid microdomains in per1Δ or gup1Δ cells (Fujita et al., 2006b). Together, these observations led us to speculate that the poor growth of gup1Δ cwh43Δ double-mutant cells might be caused by the inefficient uptake of tryptophan.
To investigate the effects of tryptophan concentration on the lipid-remodelling mutants, the per1Δ, gup1Δ, cwh43Δ and the double-mutant cells were cultured on media supplemented with low (4 mg l−1), medium (20 mg l−1) and high concentrations (200 mg l−1) of tryptophan (Fig. 5A). The per1Δ, gup1Δ, per1Δ cwh43Δ and gup1Δ cwh43Δ cells displayed a slow-growth phenotype, even on medium containing a normal concentration of tryptophan (20 mg l−1), although the growth defect was more obvious on the low-tryptophan medium (4 mg l−1). The growth defect could not be recovered by the addition of a high concentration of tryptophan (200 mg l−1) to the medium, suggesting that the growth defect was not caused by tryptophan deficiency. The growth rate of cwh43Δ cells was comparable to wild-type cells on all conditions. Interestingly, the growth of gup1Δ cwh43Δ double-mutant cells was more severely affected by a low tryptophan concentration than that of gup1Δ single-mutant cells, again showing the additive effects of deleting both GUP1 and CWH43. These results also suggest the existence of an alternative pathway in which lyso-PI is changed to ceramide by the activity of Cwh43p.
Localization of Tat2p in per1Δ, gup1Δ and cwh43Δ cells
To further examine the relationship between the lipid remodelling of GPI and lipid microdomains in yeast, we attempted to determine the cellular localization of fluorescent protein-tagged Tat2p under tryptophan-free conditions. At low medium concentrations of tryptophan, Tat2p is sorted to the plasma membrane, where it actively transports tryptophan into cells, whereas at high tryptophan concentrations, Tat2p is sorted to the vacuole for degradation (Umebayashi and Nakano, 2003). Monomeric red fluorescent protein (mRFP) (Campbell et al., 2002) was attached to the carboxyl terminus of Tat2p, and a single copy of the tagged gene was expressed in the control and mutant cells under control of the TAT2 promoter. In these cells, the chromosomal TAT2 gene was disrupted to maintain the native expression levels of Tat2p. Tat2p-mRFP fluorescence was detected at the plasma membrane of wild-type and cwh43Δ cells grown in tryptophan-free medium, but was mainly detected in the vacuoles of per1Δ and gup1Δ cells (Fig. 5B, upper panels). These results indicated that the conversion of GPI lipid moieties to the pG1 form was required at a minimum for the proper localization of Tat2p, but that the conversion of the moieties to ceramide was not essential.
Next, we attempted to ascertain whether the mislocalization of Tat2p in per1Δ and gup1Δ cells is specific to lipid microdomain-associated transmembrane proteins. For the analysis, we focused on the hexose transporter, Hxt1p, which is not associated with lipid microdomains (Malinska et al., 2003), and constructed cells carrying chromosomal Hxt1p with a C-terminal GFP fusion to examine the cellular localization of Hxt1p. In all examined strains, Hxt1p-GFP was specifically localized at the cell surface (Fig. 5B, lower panels), suggesting that defective GPI lipid remodelling only affects lipid microdomain-associated transmembrane proteins.
Tat2p in the DRM fraction is decreased in per1Δ and gup1Δ cells
We previously reported that Tat2p is unable to associate with DRMs in gwt1-10 and gaa1-1 cells (Okamoto et al., 2006). To determine if the mislocalization of Tat2p-mRFP is related to DRM association, we investigated whether Tat2p associates with DRMs in mutant cells with impaired GPI lipid remodelling.
Wild-type, per1Δ, gup1Δ and cwh43Δ cells were mechanically disrupted and membrane fractions were extracted with 20 mM CHAPS at 4°C and then fractionated by centrifugation on an Optiprep density gradient. Pma1p, a control DRM-associated protein, was located primarily in fractions 2 and 3 of the examined strains. In wild-type cells, Tat2p was also found predominantly in fractions 2 and 3, indicating it associated with DRMs (Fig. 6). Tat2p was also detected at low levels in fractions 5–8, which did not contain DRMs, indicating that forms of Tat2p exist that do not associate with DRMs. In per1Δ and gup1Δ cells, the amount of Tat2p in fractions 5–8 was higher than that observed in fractions 2 and 3, indicating that the form of Tat2p that does not associate with DRMs is increased in these mutant strains. The distribution of Tat2p in cwh43Δ cells was comparable to that in wild-type cells. The fractionation results were consistent with the localization patterns of mRFP-tagged Tat2p (Fig. 5B).
Possible model for the GPI lipid remodelling pathway
Based on the results of our group and others (Sipos et al., 1997; Bosson et al., 2006; Fujita et al., 2006b), we had previously proposed that two main possible pathways, namely a sequential and a divergent pathway, could explain the remodelling of GPI lipids to ceramides (Umemura et al., 2007). In the sequential pathway, pG1-type lipid, which is generated by the sequential activity of Per1p and Gup1p, is converted to IPC by Cwh43p (Fig. 1A, left panel). Alternatively, the remodelling pathway may diverge, with the lyso-PI generated by Per1p being used as a substrate for two reactions catalysed by Gup1p and Cwh43p (Fig. 1A, right panel). Moreover, we also speculated the existence of a third pathway that bypasses both the sequential and divergent pathways (Fig. 1A, dotted arrows), and involves the direct conversion of pG2-type lipid to IPC. This speculation is consistent with the findings that GPI-APs from per1Δ gup1Δ double-mutant cells accumulate IPC/C in addition to pG2 in vivo (Fujita et al., 2006b). To confirm the existence of these possible pathways and determine their importance, detailed analysis of the molecular mechanisms of GPI lipid remodelling to ceramides was warranted.
In the present study, we first report that gup1Δ cwh43Δ double-mutant cells, which accumulate lyso-PI-type GPI anchors, but completely lack ceramide-type GPI anchors (Fig. 2), have a higher sensitivity to high temperature, calcofluor white and SDS than gup1Δ and cwh43Δ cells (Fig. 1B). These results are consistent with the existence of both an alternative pathway, as well as a sequential pathway, because the increased sensitivity of gup1Δ cells towards high temperature and reagents by cwh43 mutation cannot be explained if the lipid-remodelling pathway functioned strictly sequentially. The levels of the Golgi form of Gas1p, a major GPI-AP, are reduced in gup1Δ cells and mostly unaffected in cwh43Δ cells, but are hardly detectable in gup1Δ cwh43Δ cells (Fig. 3). This finding again indicates that cwh43 mutation causes adverse phenotypic effects and that Gas1p containing lyso-PI-type GPI anchors does not associate with lipid microdomains.
Although the observed phenotype of the gup1Δ cwh43Δ double mutant suggests that the alternative pathway functions in S. cerevisiae cells, it appears that the sequential pathway is the main route. If only the alternative pathway were active, mutation of GUP1 would not cause any effect on the lipid remodelling of Gas1p, as the GPI lipid moiety would be changed from lyso-PI to IPC by Cwh43p, irrespective of the functionality of Gup1p. However, in gup1Δ mutant cells, a large amount of Gas1p was indeed released into the culture medium (Fig. 3), and Gas1p was previously found to not associate with lipid microdomains in gup1Δ mutant cells (Fujita et al., 2006b), indicating that both GUP1 and CWH43 genes are required for correct remodelling of Gas1p. A reasonable explanation for the observed findings is that a large portion of the GPI lipid moieties of Gas1p are changed to the pG1 form by Gup1p before further lipid remodelling to IPC by Cwh43p, which is consistent with the sequential pathway. The aggravation of the growth deficiency of gup1Δ cells by cwh43 deletion can be explained if it is assumed that the Gas1p remains linked to the surface via a ceramide-based anchor owing to the function of Cwh43p in the alternative pathway in gup1Δ cells, and that the deletion of CWH43 removes the residual part of Gas1p (and other GPI-APs) on the cell surface. Another possible explanation for the phenotype of gup1Δ cells would be that Gup1p is required for the full function of Cwh43p. Interestingly, the microsomes of gup1Δ cells still incorporate tritiated dihydrosphingosine into proteins, albeit with lower efficiency than wild-type cells (Bosson et al., 2009). Although this result appears to be consistent with the existence of the alternative pathway, it should be noted that the ceramide-based anchor lipids synthesized in vitro by gup1Δ-derived microsomes are abnormal (Bosson et al., 2009), suggesting that Gup1p is required for the full function of Cwh43p. It is possible that Gup1p and Cwh43p form a complex under physiological conditions, although we could not detect protein–protein interactions between Gup1p and Cwh43p under our experimental conditions (data not shown).
Based on our present results and those of previous reports, we propose a model for GPI lipid remodelling in yeast (Fig. 7). In this model, Cwh43p preferentially uses pG1 as a substrate and converts pG1 to IPC, as was reported previously in support of the sequential pathway model (Sipos et al., 1997). When the lyso-PI or pG2 form accumulates as a result of gup1 or per1 mutation, Cwh43p is able to convert lyso-PI and pG2 to IPC. The C-terminal domain of Cwh43p contains a region characteristic of DNase I-like superfamily proteins and is also observed in Isc1p, Inp51p, Inp52p, Inp53p and Inp54p. Isc1p is an inositol phosphosphingolipid phospholipase C (Sawai et al., 2000), and Inp51/52/53/54 possess polyphosphatidylinositol 5-phosphatase activity (Stolz et al., 1998a,b; Guo et al., 1999; Wiradjaja et al., 2001), suggesting that these proteins commonly recognize the structure of inositol phosphate. Therefore, it is conceivable that Cwh43p generally recognizes the inositol phosphate in GPI anchors and changes the lipid moieties of GPI to ceramide. Cwh43p may have GPI-PLC-like activity for hydrolysing the linkage between inositol phosphate and glycerol, thereby simultaneously replacing the glycerol moiety with ceramide in a similar fashion to the reversible sphingolipid ceramide N-deacylase reaction (Mitsutake et al., 1997), which is highly efficient for the incorporation of fatty acyl chains in place of water. In lipid microdomains, Cwh43p may easily accomplish the incorporation of ceramide instead of water during the hydrolysis of inositol-phosphate linkages.
Phenotypes of double-mutant cells other than gup1Δ cwh43Δ
We also investigated the phenotypes of double-mutant cells other than gup1Δ cwh43Δ and found that per1Δ gup1Δ cells grow more rapidly than per1Δ single-mutant cells on medium containing SDS (Fig. 1B). This result was unexpected, considering that Per1p, a phospholipase A2, and Gup1p, an O-acyltransferase, are thought to act sequentially during the remodelling of GPI lipids. According to the lipid analysis of total GPI-APs, more IPC/C lipid was detected in GPI-APs from per1Δ gup1Δ double-mutant cells than in those from per1Δ cells (Fig. 2; lanes 2 and 5), as reported previously (Fujita et al., 2006b). Therefore, it is possible that the double mutation may induce a bypass pathway that facilitates production of the ceramide-type GPI anchor lipid, which may act to reinforce the plasma membrane and confer resistance against detergents.
In contrast with gup1Δ cells, cwh43 deletion had only minimal effect on the aggravation of the growth deficiency of per1Δ cells, and per1Δ cwh43Δ double-mutant cells displayed a similar phenotype as per1Δ single-mutant cells (Fig. 1B). In both per1Δ and per1Δ cwh43Δ cells, most anchor lipids are the pG2 form (Fig. 2; lanes 2 and 6); therefore, it is possible that the pathway from the pG2 to IPC forms (Fig. 7; dotted grey arrow), which was described in two previous reports (Fujita et al., 2006b; Umemura et al., 2007), may be quite limited in the specific GPI-APs or conditions.
Transport and association of Tat2p with lipid microdomains
We initially speculated that the slow-growth phenotype of gup1Δ cwh43Δ double-mutant cells might result from a tryptophan deficiency, because we have previously found that several GPI synthesis mutants (e.g. gwt1-10, gpi7Δ and gaa1-1) mislocalize Tat2p (Okamoto et al., 2006). Although the cold-sensitive phenotype of gwt1-10 and gpi7Δ cells is suppressed by high tryptophan concentrations (Okamoto et al., 2006), the slow-growth phenotype of the gup1Δ cwh43Δ double-mutant cells is not suppressed, indicating that factors other than tryptophan are critical for the growth of gup1Δ cwh43Δ double-mutant cells. GUP1 was originally reported to be involved in glycerol uptake (Holst et al., 2000) and a wide range of crucial cellular processes, including those mediating cell preservation and function (Ferreira et al., 2006). Therefore, although it cannot be ruled out that the slow growth phenotype of gup1Δ cwh43Δ double-mutant cells might be due to the impairment of other cellular processes involving GUP1 other than the lipid remodelling of GPI, it is most likely that the deficiency of GPI-APs on the plasma membrane and cell wall results in the severe growth defect of the double mutant.
Although the slow-growth phenotype of gup1Δ cwh43Δ cells does not appear to be related with tryptophan, we found that per1Δ and gup1Δ mutant cells contain a considerable amount of cellular Tat2p that does not associate with lipid microdomains and is missorted to the vacuole, whereas the localization of Tat2p in cwh43Δ mutant cells was comparable to wild-type cells. Based on these findings, it is conceivable that GPI-APs possessing lipid moieties of the pG1 form are required for the stable association of Tat2p with lipid microdomains, but that GPI-APs with the ceramide form are not essential for this association. Notably, the phenotype of per1Δ and gup1Δ mutant cells differs from that observed for gwt1-10, gpi7Δ and gaa1-1 cells, in which Tat2p does not exit from the ER (Okamoto et al., 2006). To elucidate the underlying factors for this phenotypic difference, it is important to consider the exit of GPI-APs and Tat2p from the ER. GPI-APs are transported from the ER to Golgi via vesicles that are distinct from those containing other proteins (Muñiz et al., 2001). In addition, lipid remodelling is required for the concentration of GPI-APs into ER exit sites (Castillon et al., 2009). Taken together, these data support a model in which Tat2p is co-transported to Golgi with GPI-APs. In gwt1-10, gpi7Δ and gaa1-1 mutant cells, the amount of GPI-APs in the ER would be markedly decreased due to the defective synthesis of GPI. It is possible that a certain critical amount of GPI-APs is required to direct Tat2p into ER exit sites and/or incorporate Tat2p into ER microdomains. In contrast, a normal amount of GPI-AP is expected to be synthesized in per1Δ and gup1Δ cells, although the structures of the GPI lipid moieties are aberrant. In these cells, Gas1p cannot associate with microdomains and is therefore secreted (Bosson et al., 2006; Fujita et al., 2006b). These findings imply that GPI-APs are transported to the plasma membrane via an unknown pathway in per1Δ and gup1Δ mutant cells. It is likely that Tat2p is co-transported with such GPI-APs to the plasma membrane, but are not stably retained in lipid microdomains, and are then transported to vacuoles via endosomes for degradation. In S. cerevisiae, p24 complex regulates the transport and quality control of GPI-AP by monitoring anchor remodelling (Castillon et al., 2011). It is proposed that a portion of unremodelled GPI-APs escape from the quality control system at the ER exit site, are transported to Golgi, and are then retrotransported from Golgi to the ER via the p24 complex (Castillon et al., 2011). In per1Δ and gup1Δ cells, Tat2p might exit from the ER together with unremodelled GPI-APs that escape.
Gas1p does not associate with DRMs in per1Δ and gup1Δ mutant cells (Fujita et al., 2006b) and a considerable amount of Tat2p also does not interact with DRMs in these mutant cells, as shown in the present study, although the remaining Tat2p proteins still associate with DRMs (Fig. 6). These findings appear inconsistent with our previous report that GPI-APs are required for the association of Tat2p with microdomains (Okamoto et al., 2006). We speculate that unremodelled GPI-APs that escape from the quality control system in the ER and are transported to the cell surface may help Tat2p to associate with lipid microdomains, but are not able themselves to stably associate with lipid microdomains. Another important factor that should be taken into consideration is that DRMs, which are biochemically defined, are conceptually not equivalent to lipid microdomains (Jacobson and Dietrich, 1999; Munro, 2003; Morris et al., 2011); thus, the situation in lipid microdomains might differ from that in DRMs.
GPI anchor of Flag-Gas1p from wild-type cells possesses ceramide moieties
Our data indicate that the GPI lipid moieties of Flag-Gas1p in wild-type cells are predominantly ceramides (Fig. 4). As this determination was made from the analysis of MS3, which gives highly reliable data, we are confident of this assignment. To our knowledge, this is the first report to describe that a specific GPI-AP (Gas1p) of S. cerevisiae predominantly possesses a specific GPI lipid species (t18:0-C26:0 ceramide). However, it was reported that the GPI lipid moiety of Gas1p is a diacylglycerol type containing a C26 fatty acid, and does not contain a ceramide moiety (Fankhauser et al., 1993). One possible reason for this apparent discrepancy is that the material used for analyses in these two studies differed; here, recombinant Flag-tagged Gas1p was used, whereas purified untagged Gas1p was analysed by Fankhauser et al. (1993). However, it is unlikely that the addition of a Flag tag would influence the main lipid species of the GPI anchor, as the Flag tag was inserted at the N-terminal region of Gas1p and the GPI anchor is attached to the C terminus. It is more likely that the difference in the GPI lipid species was due to differences in the culture or purification conditions used for the preparation of Gas1p. It is noteworthy that the GPI lipid moiety of Gel1p, a homologue of Gas1p in Aspergillus fumigatus, is also a ceramide type (Fontaine et al., 2003).
The GPI lipid moieties of Flag-Gas1p in cwh43Δ cells are predominantly diacylglycerol (Fig. 4), indicating that Cwh43p remodels the GPI lipid species of Gas1p to ceramide. Our preliminary experiments using liquid chromatography/MS/MS by neutral-loss scanning of 277 Da (phosphoryl inositol + NH4) in positive-ion mode suggested that the GPI lipid species of Gas1p in wild-type and cwh43Δ cells are ceramide and diacylglycerol respectively (data not shown). Therefore, we further analysed the GPI lipid species of Gas1p in wild-type and cwh43Δ cells by neutral-loss scanning in positive-ion mode, which is suitable for the analysis of ceramide species, and by precursor-ion scanning in negative-ion mode, which is suitable for the analysis of diacylglycerol species.
Although Gas1p possesses ceramide in its GPI lipid moiety and Cwh43p is involved in the GPI lipid remodelling of Gas1p to ceramide, deletion of CWH43 has little effect on the localization of Gas1p, as Gas1p remains associated with lipid microdomains even in cwh43Δ cells (Umemura et al., 2007). An important issue to resolve is the physiological role of lipid remodelling to ceramides by Cwh43p, as a number of unanswered questions remain. For example, does the conversion of GPI lipid moieties to ceramides strictly depend on the species of GPI-AP? In other words, do GPI-APs that are exclusively modified with diacylglycerol-type lipid moieties exist among the approximately 60 species of GPI-AP of S. cerevisiae? Further studies are necessary to elucidate the molecular mechanisms by which the glycerol moiety of GPI anchors is replaced with ceramides by Cwh43p and to determine the biological function of such remodelling.
Media, strains and gene disruption
The composition of the YPAD and synthetic complete (SC) media used for the culture of yeast cells was described previously (Sherman, 1991). The S. cerevisiae strains used in this study are listed in Table 1. A one-step PCR-mediated technique was used for the gene disruption of PER1, GUP1 and CWH43 using plasmids pFA6a-kanMX6, pFA6a-TRP1 and pFA6a-His3MX6 as PCR templates (Longtine et al., 1998) and the primer pairs per1u-F1′ (5′-TGTGAAACCATACCCTTCGGGAGAAAAGAAACAGAAGTGTGGCAAGAAATCGGATCCCCGGGTTAATTAA-3′) and per1d-R1-2 (5′- AATACATATATAACTATTATTCATGATAGTAAACAGGTGAATTAAGATATGAATTCGAGCTCGTTTAAAC-3′), Dgup1-F1 (5′-AGGCAAAAACAAAGCGTGAATCAGCATTTGTTAACAGACAATTGCAAGAACGGATCCCCGGGTTAATTAA-3′) and Dgup1-R1 (5′-ATACATACATGATAGCAGTGTTATACAATTGATATTCGTAAATTTGGCATGAATTCGAGCTCGTTTAAAC-3′), and cwh43u-F1 (5′-TTTCTCGAGGAATAAGTAACCAGGAATACAGAAGGTATCCACCGCCAGTTCGGATCCCCGGGTTAATTAA-3′) and cwh43d-R1 (5′-CAGTACACACAATGTGATTACACTGATTTATAAAACCACCTTACGGCCTCGAATTCGAGCTCGTTTAAAC-3′) respectively.
The construction of plasmids pMO10 (TAT2-mRFP, CEN, LEU2), pMO12 (TAT2-3× HA, CEN, LEU2) and pMO14 (GFP-HDEL, CEN, TRP1), which were used for investigation of the localization of Tat2p, was previously described (Okamoto et al., 2006). The construction of plasmid pMF924 (3× FLAG-GAS1, CEN, LEU2), which was used for MS analysis of the GPI lipid moieties of Gas1p, was also previously reported (Fujita et al., 2006b).
The plasmid pRS314-GUP1-FLAG was constructed as follows. The promoter region and open reading frame (ORF) of GUP1 were amplified by PCR from genomic DNA using the primers GUP1F (5′-AAAAAGAATTCGGGACGACCGGTGTAAATGCTCCA-3′) and GUP1R-3FLAG (5′-AAAAAGTCGACTCAGCTAGTCTTGTCATCGTCATCCTTGTAATCGATGTCATGATCTTTATAATCACCGTCATGGTCTTTGTAGTCACTAGCGCATTTTAGGTAAATTCCGTGCCT-3′). The amplified fragment was digested with EcoRI and SalI, and inserted into pRS316T, which contains the GPI7 terminator region inserted into the XhoI/KpnI-digested pRS316 (Sikorski and Hieter, 1989; Fujita et al., 2006b) to generate pRS316T-GUP1-FLAG (GUP1-3× FLAG, CEN, URA3). pRS316T-GUP1-FLAG was digested with SacI and KpnI, and the excised 2.9-kb fragment, which contained the GUP1 promoter and ORF, 3× FLAG, and GPI7 terminator, was inserted into SacI/KpnI-digested pRS314 (Sikorski and Hieter, 1989) to construct pRS314-GUP1-FLAG (GUP1-3× FLAG, CEN, TRP1).
The plasmid pRS314-gup1-H447A-FLAG was constructed as follows. The 447th histidine of Gup1p was substituted with alanine using the QuikChange Site-directed Mutagenesis kit (Stratagene, La Jolla, CA) and pRS316T-GUP1-FLAG as a PCR template, generating the plasmid pRS316T-gup1-H447A-FLAG. This plasmid was digested with SacI and KpnI, and the excised 2.9 kb fragment was inserted into SacI/KpnI-digested pRS314 to construct pRS314-gup1-H447A-FLAG.
The construction of plasmid pRS316T-CWH43-HA (CWH43-3× HA, CEN, URA3) was previously described (Umemura et al., 2007). The 882nd arginine of Cwh43p was substituted with alanine using the QuikChange Site-directed mutagenesis kit and pRS316T-CWH43-HA as a PCR template to construct the plasmid pRS316T-cwh43-R882A-HA.
Analysis of isotope-labelled PI moieties of GPI-anchored proteins
[3H]inositol-labelled PI moieties were prepared from GPI-APs as described previously (Sipos et al., 1997; Guillas et al., 2000), and analysed by TLC on a Silica gel 60 plate (Merck, Darmstadt, Germany) using solvent system 1 (55:45:10, vol/vol CHCl3 : CH3OH : 0.25% KCl). Reaction products separated on TLC plates were detected by autoradiography and analysed using a Molecular Imager FX system (Bio-Rad, Hercules, CA).
Preparation of cell lysates and immunoblotting analysis
Cell lysates were prepared as described previously (Umemura et al., 2007). Briefly, cells were grown in YPAD medium until reaching the exponential phase and were then collected by centrifugation. The cells were washed and broken using glass beads in TNPI buffer [50 mM Tris-HCl, pH 7.5, 150 mM NaCl, protease inhibitor cocktail (Roche, Basel, Switzerland) and 1 mM phenylmethylsulfonyl fluoride] at 4°C. After removing cell debris by centrifugation (400 g, 5 min), the cell extracts were centrifuged at 13 000 g for 20 min to sediment the ER-rich fraction. The resulting ER-rich pellet was dissolved in SDS sample buffer, separated by SDS-PAGE, and analysed by immunoblotting using anti-HA monoclonal antibody (mAb) 16B12 (1:2000; Covance, Princeton, NJ), anti-FLAG mAb M2 (1:10000; Sigma-Aldrich, St Louis, MO), anti-Gas1p peptide polyclonal antibody (1:2000; kindly provided by Dr Katsura Hata, Eisai, Tokyo, Japan) and anti-Pma1p mAb 40B7 (1:10000; EnCor Biotechnology, Gainesville, FL).
Isolation of DRMs
Cells were grown at 25°C in SC until reaching the exponential phase, and 5 × 108 cells were then collected by centrifugation. DRMs for the analysis of Gas1p were isolated as described previously (Okamoto et al., 2006). Briefly, after incubation with 20 mM CHAPS for 30 min on ice, cell lysates were subjected to buoyant density gradient centrifugation using Optiprep solution (Axis-Shield PoC, Oslo, Norway) for 5.5 h at 37 000 r.p.m. (130 000 g) in a SW55Ti rotor (Beckman, Fullerton, CA). After centrifugation, nine fractions of equal volume were collected starting from the top of the gradient. Each fraction was mixed with SDS sample buffer and subjected to SDS-PAGE and immunoblotting.
For the imaging of Tat2p-mRFP and Hxt1p-GFP fusion proteins, cells were grown to the early exponential phase (OD660 of ∼ 0.2) at 25°C in SC medium. The cellular localization of Tat2p-mRFP and Hxt1p-GFP was visualized using an Olympus BX50 epifluorescence microscope system (Olympus, Tokyo, Japan) equipped with a MicroMAX cooled CCD camera (Roper Scientific/Princeton Instruments, Trenton, NJ). Captured images were processed using IPLab software (Scanalytics, Fairfax, VA).
Isolation of PIs from Flag-tagged Gas1p
Wild-type (BY4741) and cwh43Δ (BY-cwh43) strains were transformed with the plasmid pMF924 (Fujita et al., 2006b) for the expression of Flag-tagged Gas1p under control of its own promoter. Transformed cells were grown in 4 l YPAD medium until reaching an OD660 of ∼ 2, harvested, and Flag-Gas1p was then purified as described previously (Fujita et al., 2006b). Flag-Gas1p was subjected to SDS-PAGE and transferred onto PVDF membrane. PI moieties were released from Flag-Gas1p by sodium nitrite treatment, extracted with 1-butanol, and then purified using a silica column, as previously described (Fontaine et al., 2003; Maeda et al., 2007).
Mass spectrometric analysis of IPC and PIs from Flag-Gas1p
Electrospray ionization mass spectrometry (ESI-MS) analyses were performed using a 4000Q TRAP triple quadrupole MS (AB SCIEX, Foster City, CA) with a chip-based ionization source (TriVersa NanoMate; Advion BioSystems, Ithaca, NY). Released IPC and PI from GPI were directly subjected to flow injection analysis by neutral-loss (positive-ion mode) or precursor-ion scanning (negative-ion mode) of the phosphoryl inositol (Taguchi et al., 2005). The parameter settings used for ESI-MS were 1.2 kV (positive-ion mode) or −1.2 kV (negative-ion mode) for ion spray voltage, and 25–60 V (positive-ion mode) or −40 to −60 V (negative-ion mode) for collision energy. The mobile phase composition was chloroform : methanol (1:2) containing 5 mM ammonium formate and was applied at a flow rate of 200 nl min−1, as previously reported (Ikeda et al., 2011).
We are grateful to Dr Andreas Conzelmann (University of Fribourg) for helpful discussions. We thank Dr Katsura Hata (Eisai) for providing the anti-Gas1p peptide polyclonal antibody and also thank Mr Toru Watanabe (AIST) for critical reading of the manuscript. Part of this study was supported by a grant from the Japan Science and Technology Agency (Research for Promoting Technological Seeds) and JSPS KAKENHI (Grant number 23570185).