Glycosylphosphatidylinositol (GPI)‐anchored membrane proteins of Saccharomyces cerevisiae exist with two types of lipid moiety—diacylglycerol or ceramide—both of which contain 26:0 fatty acids. To understand at which stage of biosynthesis these long‐chain fatty acids become incorporated into diacylglycerol anchors, we compared the phosphatidylinositol moieties isolated from myo‐[2‐3H]inositol‐labelled protein anchors and from GPI intermediates. There is no evidence for the presence of long‐chain fatty acids in any intermediate of GPI biosynthesis. However, GPI‐anchored proteins contain either the phosphatidylinositol moiety characteristic of the precursor lipids or a version with a long‐chain fatty acid in the sn‐2 position of glycerol. The introduction of long‐chain fatty acids into sn‐2 occurs in the endoplasmic reticulum (ER) and is independent of the sn‐2‐specific acyltransferase SLC1. Analysis of ceramide anchors revealed the presence of two types of ceramide, one added in the ER and another more polar molecule which is found only on proteins which have reached the mid Golgi. In summary, the lipid of GPI‐anchored proteins can be exchanged by at least three different remodelling pathways: (i) remodelling from diacylglycerol to ceramide in the ER as proposed previously; (ii) remodelling from diacylglycerol to a more hydrophobic diacylglycerol with a long‐chain fatty acid in sn‐2 in the ER; and (iii) remodelling to a more polar ceramide in the Golgi.
Numerous glycoproteins of Saccharomyces cerevisiae become attached to a glycosylphosphatidylinositol (GPI) anchor and many ultimately appear in the cell wall or at the plasma membrane (Hardwick et al., 1992; Stratford, 1994; Cawley et al., 1995; Komano and Fuller, 1995). The biosynthesis of GPI‐anchored proteins follows the same basic rules in all eukaryotes, including yeast (Conzelmann et al., 1990; Englund, 1993; Fankhauser et al., 1993; Nuoffer et al., 1993). Nevertheless, different organisms contain widely differing kinds of lipid moieties in their GPI anchors which, in most organisms, do not at all reflect the lipid moieties present in phosphatidylinositol (PI) which would appear as a natural starting point for the GPI biosynthesis in the endoplasmic reticulum (ER) (Roberts et al., 1988b; Luhrs and Slomiany, 1989; Bütikofer et al., 1990, 1992; Walter et al., 1990; McConville and Ferguson, 1993; Patnaik et al., 1993; Doering et al., 1994; Redman et al., 1994; Brewis et al., 1995; Serrano et al., 1995; Treumann et al., 1995; Kapteyn et al., 1996). In principle, this difference could come about by the specificity of the GlcNAc‐transferase which initiates GPI biosynthesis by transferring GlcNAc onto PI or by the prevalence of a rare kind of PI in the environment of this enzyme. Alternatively, lipid moieties might become exchanged either before or after transfer of the GPI onto proteins (Englund, 1993). This latter model has been found to apply to the blood stage form of Trypanosoma brucei where longer fatty acids are replaced by myristic acid through sequential deacylation/reacylation reactions on sn‐2, then sn‐1 shortly before the GPI is attached to the protein (Masterson et al., 1990). A remodelling mechanism has also been proposed to explain the prevalence of sn‐1‐alkyl, 2‐acylglycerol in mammalian cells (Singh et al., 1994). The situation in yeast is peculiar since two very different types of lipid moieties can be found: ceramides and another lipid which is sensitive to mild base treatment and can be removed by PI‐specific phospholipase C (PI‐PLC), probably diacylglycerol. The ceramides are found on the majority of yeast anchors; they mainly consist of C18:0 phytosphingosine and a C26:0 fatty acid and are different from the main ceramide found in the abundant inositolphosphoceramides (IPCs) (Smith and Lester, 1974; Lester and Dickson, 1993). Alternatively, Gas1p, a well‐characterized GPI protein of yeast, is made with a C26:0 fatty acid (FA)‐containing, base‐sensitive lipid (Fankhauser et al., 1993). In both types of lipid moieties the C26:0 may be hydroxylated on C2. Complete GPIs (CPs) ready to be transferred to proteins are present only in very low amounts in wild‐type cells but can be detected in lipid extracts of [2‐3H]mannose‐labelled pmi40, a strain with a conditional defect in mannose biosynthesis incorporating elevated amounts of exogenously added [3H]mannose at 37°C (Sipos et al., 1994). CPs also become detectable in extracts from [3H]inositol‐labelled gaa1 and gpi8 cells which both have a conditional mutation affecting the addition of CPs to proteins (Benghezal et al., 1995; Hamburger et al., 1995). In all these cases, the CPs are completely susceptible to mild base treatment, supporting the notion that the ceramide moieties may be introduced at a later stage by lipid remodelling of protein anchors. The work presented here addresses the question of whether the C26:0 long‐chain FAs of mild base‐sensitive anchors are present already at the beginning of GPI biosynthesis, or whether they are introduced only at a later stage.
It appeared impossible to obtain GPI intermediates in sufficient quantity and purity as required for GC–MS. We therefore considered radiolabelling with lipid precursors such as acetate or palmitate in order to increase the sensitivity of detection. Acetate, although uniformly labelling all possible lipid moieties to comparable specific activity, was considered inappropriate because GPI intermediates would represent only a very scanty fraction among all labelled lipids and their purification seemed out of reach. As to the use of palmitate, we anticipated the same problem and in addition that it would introduce a heavy bias towards palmitate‐containing lipid moieties. To avoid these inconveniences we resorted to metabolic labelling with [3H]inositol. In contrast to the other possible labels, inositol labels only a subset of phospholipids, namely the various PIs, IPCs and GPI intermediates. Inositol labelling also was expected uniformly to label all possible GPIs irrespective of their lipid moieties and thus, we could be confident not to bias the result of our analysis by the label. The disadvantage of [3H]inositol labelling consisted of the fact that direct isolation of the isolated lipid moiety was not feasible, so that only more complex fragments containing the lipid and the [3H]inositol, i.e. the PI or IPC moieties could be analysed. Nevertheless, this approach clearly identified two novel types of lipid remodelling in yeast GPI anchors.
Early GPI intermediates contain a more polar PI moiety than mature proteins
Previously established methods (Conzelmann et al., 1992) were optimized to ascertain complete delipidation of GPI‐anchored proteins without major losses of proteins and quantitative cleavage of GPIs or GPI‐containing peptides with nitrous acid (see Materials and methods). The standard procedure used in all further preparations yielded pronase‐treated inositol‐labelled anchor peptides which were free of any contaminating lipids (Figure 1, lanes 1 and 3) since all labelled material remained close to the origin of the TLC. By contrast, all known GPI intermediates except for the complete GPI precursors CP1 and CP2 move off the origin in the solvent system used here (not shown). The protocol used routinely for nitrous acid treatment quantitatively liberated the [3H]inositol‐labelled lipid moieties of these anchor peptides (Figure 1, lanes 2 and 4). After a long pulse, three lipid species were detectable in this TLC system whereby only the fastest‐migrating species (pG1) was base‐sensitive, the other lipid moieties (pC1, pC2) being ceramides (see below). This type of analysis also revealed that, within the limits of the resolution of our TLC analysis, cells use the same types of lipids for GPI anchoring, whether labelled at 24°C or 37°C (Figure 1). To compare the PI moieties obtained from anchor peptides with those present in early GPI intermediates we had to label mutant cells, since no GPI intermediates accumulate to detectable levels in wild‐type cells (Puoti et al., 1991; Sipos et al., 1994). sec53 cells have a ts mutation affecting GDP‐Man biosynthesis and, at 37°C, accumulate a very hydrophobic lipid which also is seen in other mutants such as dpm1, sec59, gpi4, gpi5, gpi6 and gpi9 (Benghezal et al., 1995). Its partial characterization strongly suggested that it consists of GlcN–acylPI (Orlean, 1990; Conzelmann et al., 1992). To isolate the PI moiety from this GlcN–acylPI precursor we had to remove the GlcN and the FA which is linked to the inositol ring. [All GPI intermediates of S.cerevisiae contain an acyl chain on the inositol, although this acyl has not been found on protein‐bound GPI anchors (Fankhauser et al., 1993)]. As shown in Figure 2A, the removal of the GlcN by nitrous acid reduces rather than increases the mobility of GlcN–acylPI (Figure 2A), lane 4 versus lane 2). This apparent increase in hydrophilicity may be observed because the free amine of GlcN neutralizes the negative charge of the phosphodiester. This hypothesis is supported also by the finding that N‐acetylation of the GlcN–acylPI significantly decreases rather than increases its mobility (not shown). The removal of GlcN by nitrous acid is quantitative, since the residual acylPI moiety is resistant towards GPI‐specific phospholipase D (GPI–PLD), an enzyme which only acts on GlcN‐containing intermediates (Davitz et al., 1989) (Figure 2A, lanes 3 and 4). In contrast, the purified GPI‐intermediate (Figure 2A, lane 2) was quantitatively hydrolysed by GPI–PLD (Figure 2A, lane 1). Additional proof for the proposed GlcN–acylPI structure of the sec53 intermediate was obtained by demonstrating that the [3H]inositol‐labelled lipid after purification still is resistant to PI–PLC (not shown), that it reduces its mobility upon N‐acetylation, and that its GPI–PLD product (Figure 2A, lane 1) increases its mobility (i.e. its hydrophobicity) in TLC upon N‐acetylation or nitrous acid treatment (not shown). Partial deacylation by methanolic ammonia has been reported to preferentially cleave the acyl chain on the inositol (Roberts et al., 1988a) and generated fragments co‐migrating with PI, lyso‐PI and inositolphosphoglycerol (Figure 2A, lane 5). These assignments were confirmed by treating these fragments with PI–PLC, an enzyme which is expected to hydrolyse both PI and lyso‐PI, but is unable to cleave inositol‐acylated forms of this intermediate (Figure 2B, lanes 2–4). As expected, the species migrating with PI and lyso‐PI were reduced to inositol‐phosphate by PI–PLC (not shown). When the partially deacylated lipid co‐migrating with PI in lane 3 was once more treated with methanolic ammonia, it was transformed into the lower band of lane 3 which co‐migrates with lyso‐PI (not shown). These data clearly demonstrate that the partial deacylation products of lane 5 co‐migrating with PI and lyso‐PI carry no acyl chains on the inositol.
Importantly, the PI moiety generated from GlcN–acylPI of sec53 had a lower Rf, i.e. was apparently less hydrophobic than the PI moiety obtained from the GPI anchors of mature proteins (Figure 2A, lanes 5 and 6). Mild base treatment of anchor peptides generated inositolphosphoglycerol from the PI moiety, whereas anchor‐derived IPCs (pC1 and pC2) were not hydrolysed (Figure 2A, lane 7). The difference between the major and a minor, slightly faster migrating, deacylation product could not be identified.
To exclude strain differences in the utilization of fatty acids for GPI biosynthesis, we compared the mobilities of the GlcN–acylPI intermediate of various mutants. As shown in Figure 3A, although coming from different genetic backgrounds, all mutants analysed produced a GlcN–acylPI of identical mobility. The products of methanolic ammonia also migrated similarly (Figure 3B) and were all PI–PLC‐sensitive (Figure 3C)
Fatty acids in sn‐2 of the GlcN–acylPI intermediate and the mature GPI anchor are different
To obtain more information on their structure, the PI moieties from the early sec53 intermediate and mature anchors were treated with phospholipase A2 (PLA2). Working with the deamination product of the early sec53 intermediate was difficult because, during storage of the purified lipid at−20°C or at 4°C, this early intermediate became transformed into apparently more hydrophobic and less hydrophobic species. When the major species was repurified, the other species reappeared upon storage (not shown). This leads us to believe that the various forms of acylPI may be generated by transesterification, resulting in the spontaneous migration of the acyl group on the inositol ring. Transesterification of acetyl groups on sialic acids and FA migration on lyso‐glycerophospholipids at neutral or mildly basic pH are commonly observed (Plückthun and Dennis, 1982; Kamerling et al., 1987; Butor et al., 1993) and the appearance of multiple forms of the GlcN–acylPI has been observed previously by others (Costello and Orlean, 1992). However, the more and less hydrophobic species may also arise through oxidative degradation of unsaturated FAs. Although the acylPI moiety of the early sec53 GPI intermediate thus migrated as one major and three minor species (Figure 4, lane 2), all these bands yielded mainly PI, lyso‐PI and glycerophospho‐inositol upon mild deacylation with methanolic NH3 (Figure 4, lane 4). The identity of PI was further confirmed by PLA2 treatment whereby PI was degraded to lyso‐PI, whereas all forms of acylPI were resistant to PLA2 (lanes 1 and 5). PLA2 treatment of the anchor lipid moieties obtained from mature anchor peptides yielded a lyso‐PI which migrated to the same position in our TLC system (Figure 4, lane 6). This indicates that the FA in the sn‐1 position of mature diacylglycerol‐containing anchors is the same as in the early precursor of sec53. This leaves us with the expectation that it is the length of the FA in sn‐2 which is responsible for the higher mobility of the anchor‐derived PI as compared with the PI from the early sec53 intermediate (Figure 2A, lane 6 versus lane 5; Figure 4, lane 7 versus lane 4). Indeed, by treating anchor lipid moieties with methanolic NH3, a treatment which hydrolyses the acyls in sn‐1 and sn‐2 at comparable rates, we generated two forms of lyso‐PI, one co‐migrating with the PLA2 product of the early sec53 intermediate and a second one which has a higher Rf (Figure 4, lanes 8 and 9). This second, more hydrophobic lyso‐PI is completely resistant to PLA2 (Figure 4, lane 10). It is noteworthy, that this second more hydrophobic lyso‐PI is not generated from the early sec53 intermediate (Figure 4, lanes 4 and 5). A band with somewhat higher mobility (Figure 4, lanes 3–5, lozenge) must be a different fragment since it is susceptible to PLA2. These results indicate that the early GPI precursor of sec53 and mature mild base‐sensitive anchors contain diacylglycerol moieties which differ by the length of the FA in sn‐2. This finding implies the existence of some form of remodelling step in between these two biosynthetic stages.
Lipid extracts of wild‐type cells contain large amounts of PI–PLC‐resistant GPI structures
The mannosylation defect of sec53 is known to have pleiotropic effects on protein glycosylation, protein translocation into the ER and vesicular traffic from the ER to the Golgi (Ferro‐Novick et al., 1984a,b; Feldman et al., 1987; Kepes and Schekman, 1988). Therefore, we were concerned that the early precursors accumulating in sec53, sec59, dpm1, gpi4 and gpi9, all mutants with defects in mannosylation reactions, may contain different FAs than the true intermediate of wild‐type cells. Although until now we had been unable to detect any GPI intermediate in the extracts of [3H]inositol‐labelled wild‐type cells, we decided to investigate if any GPI intermediate may go undetected because it is hidden by co‐migrating PI, IPCs or mannosylated forms of IPC. In developing a suitable strategy we assumed that potential GPI intermediates of wild‐type cells, like the previously described abnormal GPIs accumulating in GPI biosynthesis mutants, would resist PI–PLC treatment (Orlean, 1990; Conzelmann et al., 1992; Benghezal et al., 1995). Thus, starting with [3H]inositol‐labelled lipid extracts, PI–PLC was used to remove the multiple species of PI which represent the most abundant inositol‐labelled lipids. The subsequent nitrous acid treatment was expected to liberate the acylPI moiety from all potential GPI intermediates. Indeed, as shown in Figure 5A, PI–PLC treatment removed the bulk of PI so that only 5–11% of c.p.m. of the initial lipids were left after PI–PLC treatment (Figure 5, lanes 2 and 5). The subsequent nitrous acid treatment led to the appearance of at least six very hydrophobic bands migrating to the high Rf region expected for acylPIs (Figure 5A, lanes 3 and 6). These bands did not appear if the PI–PLC‐treated lipid extract was treated with GPI–PLD before nitrous acid (not shown). Importantly, these bands accounted for a significant fraction of counts not only in gpi8 (Figure 5, lane 6), a mutant which accumulates late GPI intermediates because of its deficiency in the addition of GPIs to proteins (Benghezal et al., 1995), but also in W303 and X2180 wild‐type cells (Figure 5, lanes 3, 7 and 9). The material shown in lanes 3 and 6 was further compared with a similarly treated extract from sec53 (Figure 5B, lane 10) in a solvent system which yielded fewer but more widely separated bands, which could easily be purified for further analysis (see below). Whereas all strains contained bands A, B′/C′ and B″/C″, sec53 contained an additional two species labelled D′ and D″ (Figure 5B). It should be noted that nitrous acid attacks the free amine of GlcN, a sugar component which is highly characteristic of GPIs. Bands A–D were therefore considered to be derived from hitherto unrecognized, inositol‐acylated GPIs since only this kind of inositol‐containing lipids can be expected to be PI–PLC resistant and to significantly increase its mobility on TLC upon nitrous acid deamination. Until now, we have not analysed the glycan moieties of these putative GPI intermediates of wild‐type cells. Since lipids B′/C′ and B″/C″ represent potential GPI anchor precursor lipids of wild‐type cells, we decided to purify all these novel GPI‐derived acylPIs and to compare their deacylation products with the ones from mature anchors.
Fatty acids in sn‐2 differentiate GPI precursors and mature GPI anchors in wild‐type cells
The putative acylPIs were purified by preparative TLC whereby the upper and lower half of the two major species of wild‐type cells were scraped separately so as to yield B′, C′, B″ and C″ as separate fractions. Upon rechromatography of the purified lipids we observed that B′ rechromatographed as B′ plus a significant amount of B″ and a trace of A; and that B″ rechromatographed as B″ plus a significant amount of B′ and a trace of A. Thus, we had to accept the fact that these lipids were not stable, but able to interconvert. The same interconversion was observed for the pair C′/C″ and the pair D′/D″. This phenomenon is analogous to the instability of the deamination products of the early GPI intermediate of sec53 (Figure 4, lane 2). Interconverting lipids were pooled and pools were further analysed after deacylation with methanolic NH3 and PLA2 treatment (Figure 6A). The former treatment produced relatively homogeneous populations of PI and lyso‐PI from each pool (Figure 6A, lanes 5 and 7) which co‐migrated with similarly treated, regular PI from the same cells (Figure 6B. The fact, that a homogeneous PI and a homogeneous lyso‐PI pool was obtained from each pool of interconverting lipids supports the idea that interconversion among members of these pools arises through transesterification of the FA on the inositol which is changing its site of attachment. The PI populations derived from B′/B″ and C′/C″ have slightly different Rfs (Figure 6A, lanes 5 and 7), indicating that the difference between these two GPI precursor lipids resides in the diacylglycerol moieties, although additional differences in the inositol‐bound fatty acyl chain cannot be ruled out. Since the partial deacylation with methanolic NH3 is not specific for sn‐1‐ or sn‐2‐acyls, the relative homogeneity of the lyso‐PI population generated from each pool suggests that any given GPI precursor has two FAs of similar length in sn‐1 and sn‐2. To compare the PI moieties derived from pools B′/B″ and C′/C″ of wild‐type cells with the PI moieties derived from the GPI protein anchors of the same cells, the lipids were hydrolysed by PLA2 treatment and methanolic NH3. This analysis led to the same conclusion as the comparison with the GlcN–acylPI intermediate of sec53: (i) protein anchors contain a more hydrophobic PI moiety (pG1) than GPI intermediates (Figure 6A), lanes 2, 5 and 7); (ii) partial deacylation with methanolic NH3 produces two distinct lyso‐glycerophospholipids from anchor PIs (Figure 6A, lane 3, lipids a and b) whereas a single homogeneous population is obtained from GPI intermediates (Figure 6A, lanes 5 and 7, lipid b); and (iii) the more polar lyso‐PI from anchor peptides (lipid b) co‐migrates with the lyso‐PI from GPI intermediates. This indicates that GPI intermediates and diacylglycerol‐containing GPI anchors of wild‐type cells differ with respect to the length of the FA in sn‐2 whereas, by the resolution power of our TLC system, the FA in sn‐1 does not appear to be different.
We also prepared [3H]inositol‐labelled late GPI intermediates from gpi7 and gpi8 which accumulate both the close‐to‐complete and complete GPI precursor ready for attachment to cells (Benghezal et al., 1995). These late GPI intermediates also have a PI moiety which migrates similarly as the PI from the earlier intermediates (Figure 6C, lanes 19–22). The PI moiety of CP2 from gpi8 is less hydrophobic than pG1, and its only lyso‐form co‐migrates with the lyso‐PI of GPI intermediates (Figure 6D. This suggests that GPI intermediates up to the completed GPI do not have a long‐chain FA at position sn‐2.
The remodelling of GPI anchors in sn‐2 does not require Slc1p
IPCs contain the largest lipid moiety among the phospholipids of S.cerevisiae as their ceramide comprises a long‐chain base and a C26 fatty acid (Smith and Lester, 1974). Strains unable to make long‐chain bases like lcb1::URA3 require some kind of long‐chain base (dihydrosphingosine or phytosphingosine) in their growth medium. Yet, a mutation in SLC1 (slc1‐1) suppresses this auxotrophy (Dickson et al., 1990). Slc1p can function as a 1‐acyl‐sn‐glycerol‐3‐phosphate acyltransferase in Escherichia coli but Slc1p is not essential for growth of S.cerevisiae (Nagiec et al., 1993). The lcb::URA3 slc1‐1 strain makes anomalous mannosylated forms of PI with a C26 in the sn‐2 position which mimic the lacking mannosylated forms of IPC (Lester et al., 1993). Therefore, the slc1‐1 suppressor allele is thought to encode a variant enzyme with an altered substrate specificity that enables it to use C26 instead of C18 or C16 FAs when acylating the sn‐2 position of inositol‐containing glycerolipids (Nagiec et al., 1993). We decided to test the hypothesis that SLC1 is a component of an enzyme complex involved in the remodelling of GPI anchors at sn‐2. For this, the PI moieties from the GPI anchor peptides of the SLC1 deletion strain were compared with the same material from the corresponding wild‐type. As shown in Figure 7, the deletion strain still made normal amounts of diacylglycerol‐type GPI anchors, the Rfs of which were identical to those of wild‐type cells. Thus, it appears that SLC1 is either not involved in the sn‐2‐remodelling of diacylglycerol‐containing GPI anchors or that a redundant gene can compensate for its deletion.
Long‐chain FAs are introduced at sn‐2 rapidly after addition of GPIs to proteins
To study the kinetics of the appearance of long‐chain FAs in sn‐2 of GPI anchors, a pulse–chase experiment was performed. As shown in Figure 8A, after 5 min of pulse, two different PI moieties, pG1 and pG2, were detected. pG1 represents the mature anchor lipid with a long‐chain FA in sn‐2, whereas the less polar pG2 migrates close to the PI from pool C′/C″ (see below) and co‐migrates with the free PI of these cells (not shown). pG2 was easily detectable after short pulses and when labelling with a high ratio of [3H]inositol/OD. For chase, only cycloheximide is effective in blocking the further synthesis of labelled anchors, as the radiolabelled PI remains available for GPI synthesis for some time even if further incorporation of [3H]inositol into PI is blocked by the addition of an excess of non‐radioactive inositol. In the presence of cycloheximide, pG1 and pG2 were disappearing whereas anchor‐derived IPC moieties pC1 and pC2 increased (Figure 8). Also, a lysoform of PI was discovered which disappeared with similar kinetics as pG1 and pG2. The quantitative evaluation of this experiment is shown in Figure 8B. It can be seen that the total counts are still strongly increasing during the first 5–10 min of chase in the presence of cycloheximide. This can be rationalized by the fact that part of proteins is translocated only post‐translationally (Schekman, 1994) and that GPI anchor addition may not be instantaneous after translocation is accomplished. Moreover, for the preparation of anchor peptides, efficient delipidation of GPI proteins required affinity chromatography on concanavalin A–Sepharose. Thus, the rise in total counts during the first minute of chase may also be due to a delay in the transfer of N‐glycans onto newly made proteins. From the quantitation it appears that different anchor lipids peak at different times: pG1 is increasing during the first 5 min of chase, while lyso‐PI and pG2 are already decreasing. Similarly, pC1 peaks at 5–10 min of chase while pG2 and pG1 are decreasing. Finally, pC2 is peaking at 15 min when all other lipid moieties are decreasing. With regard to the introduction of long‐chain FAs into sn‐2, the kinetic data are compatible with the idea that lyso‐PI and pG2 may be precursors of pG1, although the kinetics by themselves cannot prove any such relationship. PLA2 treatment of a mixture of anchor lipids containing pG1 and pG2 and lyso‐PI (Figure 9B, lane 6) results in a single lyso‐PI which co‐migrates with the lyso‐PI detectable among anchor lipids after short pulses (Figure 9B, lane 7). This indicates that pG1, pG2 and lyso‐PI from anchor peptides have a similar FA in sn‐1, but differ at sn‐2 (Figure 9B). These data are compatible with the idea that pG2‐type anchors are remodelled to pG1‐type anchors through acyl exchange in sn‐2 and that the lyso‐PI observed after short pulses is an intermediate in this reaction. Side‐by‐side comparison of pG2 with PI moieties from pools B′/B″ and C′/C″ (Figure 9A showed that pG2 did not exactly co‐migrate with either one, but rather migrated slightly less than pool C′/C″. Similarly, the lyso‐PI appearing during short pulses on GPI anchors (Figure 9A, lane 3) was found to co‐migrate with the lower half of the lyso‐PI forms derived from the partial deacylation of pools B′/B″ and C′/C″ (Figure 9A, lanes 1 and 2). This however might derive from the fact that the latter contain a mixture of two lyso‐PI forms having the monoacyl substituent in either sn‐1 or sn‐2, whereas the natural lyso‐PI appearing in short pulses might contain only one of such substituents. It should also be borne in mind that the partial deacylation may release the FAs from numerous unlabelled phospholipids and that the presence of unlabelled lipids may slightly influence the mobility of the labelled lipids. Thus, we feel that we reach the limits of resolution of our TLC system and cannot at present decide whether pG2 still carries the same lipid moieties as the GPI intermediates, or if pG2 is already the product of a remodelling reaction.
Subcellular localization of lipid remodelling events can be studied using secretion mutants
To determine the subcellular localization of the introduction of long‐chain FAs in the sn‐2 position of diacylglycerol‐containing GPI anchors and of other remodelling events, we labelled secretion mutants in which the secretion is conditionally blocked at various stages of the secretory pathway (Novick et al., 1980; Esmon et al., 1981). Figure 10A indicates that, in wild‐type cells, the majority of pG1‐ and pG2‐type anchors disappear during a chase of 30 min (lanes 2 and 3). Also, pC2 is appearing only during chase and this is in agreement with data in Figure 8. In sec12 and sec18—two mutants in which secretory vesicles either cannot bud off the ER or cannot fuse with the subsequent cis Golgi compartment respectively—pG1 seems to be stable during chase. Thus, the data suggest that pG1‐ and pG2‐ and pC1‐type anchors can be made in the ER, whereas the pC2‐type anchors are made in the Golgi. In an analogous experiment with a longer pulse and a higher amount of radioactivity, data could be quantitated by radioscanning of TLC plates, as shown in Figure 10B. This experiment confirmed the findings of the previous experiment and showed that pG1 is stable also in sec7 and remains partially stabilized in sec14 cells which are blocked after the cis or the mid Golgi, respectively (Novick et al., 1980; Franzusoff and Schekman, 1989). pC2‐type anchors seem to be made efficiently in sec14 but not sec7, thus suggesting that pC2‐type anchors may be generated only in the mid Golgi. When passing the pronase digest of the delipidated protein extract over octyl–Sepharose, we always observe a run‐through fraction of labelled peptides which does not adhere to octyl–Sepharose. Some soluble forms of GPI peptides might be expected based on the postulated cleavage of certain GPI‐anchored proteins at the cell surface as they become integrated into the cell wall (Muller and Bandlow, 1993; Lu et al., 1994, 1995; Muller et al., 1996), but here we also find some soluble inositol‐labelled material when labelled proteins are retained in the ER. The soluble fraction has not yet been structurally analysed and the data have been included only to allow for a complete account of labelled material. The interpretation of pulse–chase experiments in secretion mutants thus allows us to conclude that pG1 and pG2 anchors are made in the ER and that at least part of them seem not to be further remodelled in this organelle. However, the data do not indicate why pG1‐ and pG2‐type anchors are labile in the mid Golgi and later compartments. This lability may be due to exchange of lipid moieties during a further remodelling step, to rapid turnover of proteins with this type of anchor, or to loss of the inositolphospholipid moiety during integration into the cell wall.
The spectrum of FAs in glycerophospholipids of S.cerevisiae is relatively narrow and comprises mainly C16:1, C18:1, accompanied by minor amounts of C14:0, C16:0 and C18:0 (Hunter and Rose, 1972; Wagner and Paltauf, 1994). Yet, analysis of FAs shows very significant differences in the relative frequency of different FAs among the major phospholipid classes and also between the sn‐1 and sn‐2 positions (Wagner and Paltauf, 1994). On the other hand, very long C26:0 FAs have been found in sphingolipids and in the diacylglycerol‐containing GPI anchor of Gas1p (Smith and Lester, 1974; Fankhauser et al., 1993). In this report we used metabolic labelling with [3H]inositol in order to compare the lipid moieties of different forms of GPI anchors and GPI intermediates. Although this method does not allow for a direct analysis of the lipid component, it was chosen because it was not expected to influence the cell in its choice of lipid components for GPIs. Nevertheless, we have to consider that, in inositol‐free medium, inositol production is the rate‐limiting step in the formation of PI and that the addition of a bolus of [3H]inositol might transiently increase PI synthesis. Therefore, we cannot be sure that the FA profile of PIs produced during labelling is exactly identical with that of the PI produced during exponential growth. To minimize this problem, we used a standard protocol in which no more than 10 μCi were used to label 1 OD600 unit of cells. Our calculation indicates that, in SDC medium, 1 OD600 unit of exponentially growing cell generates ∼2×10−11 mol PI per min. When adding a bolus of 10 μCi of [3H]inositol/OD of cells equivalent to 3.3× 10−10 mol of inositol, 80% of the counts are incorporated into PI within 10 min. Assuming that endogenous inositol synthesis is not down‐regulated by the exogenously added [3H]inositol, this means that during 10 min, the rate of PI synthesis might increase transiently by 130% at most. Thus, we cannot exclude that our labelling procedure introduces a small distortion of the lipid profile of GPIs but that this effect—at least under standard conditions—is probably only minor. We also took care to avoid differences stemming from variations in the genetic background, since all mutant strains were compared with the corresponding wild‐type strains. At the level of the resolution of our TLC systems, our analysis failed to unravel any difference in PI moieties of GPI anchors between W303 and X2180 or between cells grown at 24 and 37°C, respectively. Temperature has been reported to have virtually no influence on the type of FAs used for phospholipid biosynthesis by S.cerevisiae (Hunter and Rose, 1972).
Our studies revealed the existence of previously unrecognized GPIs in the lipid extract of [3H]inositol‐labelled cells. Further analysis showed that these GPIs are heterogeneous and migrate in TLC (solvent system 2) at various positions between PI and M(IP)2C (unpublished results). It would appear that the amount of this material is not less abundant in wild‐type cells than in gpi8 or gpi7 (not shown), i.e. in mutants which accumulate some form of GPI intermediate due to a defect in GPI biosynthesis. At present, we do not have an explanation for the appearance of pool D′/D″ in sec53. We believe that PIs from pool D′/D″ have very short FAs but this material has not yet been fully characterized. The appearance of pool D′/D″ seems to be specific for mannosylation mutants and not to be a secondary effect of the secretion block of sec53 since it was also found in pmi40 but not in sec12 (not shown).
Our studies establish the existence of two previously unrecognized types of lipid remodelling in yeast. One type is postulated in order to explain the presence of pG1‐type PI moieties on GPI anchors, a species which is clearly absent from all free GPIs. This finding raises a number of questions: (i) what kind of primary anchor lipid is acted upon to give rise to pG1‐type lipids?; (ii) what is the nature of the FA that is introduced in sn‐2?; (iii) what part of the GPI does get exchanged during remodelling?; and (iv) what enzymes are mediating this remodelling? In analogy to Gas1p, the mild base‐sensitive GPI anchor of which has been characterized, it seems reasonable to assume that pG1‐type anchors contain a C26:0 FA in sn‐2 (Fankhauser et al., 1993). Studies on the rate of incorporation of radiolabelled FAs demonstrated a very rapid non‐random incorporation of [3H]palmitic and [3H]oleic acid into either the sn‐1 or the sn‐2 position of different classes of glycerophospholipids of S.cerevisiae and led to the suggestion that yeast sustains an sn‐2‐specific deacylation/reacylation cycle similar to that of mammalian cells (Wagner and Paltauf, 1994). An analogous reaction might be responsible for the presence of long‐chain FAs in the sn‐2 position of pG1‐type anchors. However, at present we cannot exclude that remodelling is achieved through the replacement of diacylglycerol, phosphatidic acid or an even larger moiety. Exchange of a PI moiety for a different PI moiety however seems unlikely since we could not find any hydrophobic PI of the same TLC mobility as pG1 in the lipid extracts of [3H]inositol‐labelled cells (not shown). It must also be stressed that remodelling in sn‐2 is detected by our experiments because the remodelling exchanges a short FA against a long one. However, our data cannot exclude concomitant remodelling events resulting in more conservative substitutions. Indeed, the exchange of myristate for myristate (myristate exchange) on the mature, GPI‐anchored VSG of Trypanosoma brucei has recently been described (Buxbaum et al., 1996).
A further type of remodelling was established through the use of secretion mutants. Previous studies had already established that the ceramide moieties of mature GPI anchors of X2180 mainly contain C18:0 phytosphingosine and C26:0 FAs with minor amounts of C20:0 phytosphingosine and monohydroxylated C26:0 (Fankhauser et al., 1993). It is clear from the present and previous studies (Conzelmann et al., 1992) that pC1, the main IPC derived from GPI anchors of mature proteins is different, i.e. is more hydrophobic than the main cellular IPC (IPC/C, IPC‐II), the latter consisting of phytosphingosine with a C2‐hydroxylated C26:0 (Smith and Lester, 1974; Hechtberger et al., 1994) and is presumably synthesized in the ER (Puoti et al., 1991). Our study also confirms that some remodelling reaction which introduces ceramides can occur in the ER (Conzelmann et al., 1992). This study now demonstrates that only pC1, i.e. the more hydrophobic type of IPC, is introduced in the ER whereas pC2—the anchor‐derived IPC co‐migrating with IPC/C—only appears on GPI proteins reaching the mid Golgi. This suggests that GPI anchors and IPCs made in the ER ultimately contain different ceramides. The exact nature of the ceramide added onto GPI anchors in the ER is currently under study. Based on the results of this study we have to postulate a further remodelling step to explain the appearance of pC2 in anchors reaching the mid Golgi apparatus. This postulate raises the same type of questions as referred to earlier. In particular, it is not clear what type of GPI anchor lipids become substituted by pC2. Our present working hypothesis is that pC1 contains either C18 phytosphingosine with a C26:0 non‐hydroxylated FA or dihydrosphingosine with a hydroxylated C26:0 FA and that the Golgi remodelling step is introducing a further hydroxyl group into the lipid moiety. Studies to establish the exact nature of pC2 are under way.
In several experiments analysing material from cells labelled by only a short pulse, nitrous acid treatment not only released pG1, pG2 and pC1 but also traces of acylPI from anchor peptides. This suggests that inositol‐acylated GPIs can be transferred to proteins in yeast. On the other hand, the low amount of this species and the absence of non‐acylated CPs in yeast (Sipos et al., 1994) suggest that acyls are removed from inositols either shortly before, during, or immediately after transfer of CPs to proteins. A similar situation seems to prevail in trypanosomes (Güther and Ferguson, 1995).
At present, the role of lipid remodelling on yeast GPI anchors is not clear. One might be led to think that lipid remodelling of anchors is of minor importance, since mutants which can grow in the absence of ceramides have been described (Nagiec et al., 1993). We do not subscribe to this view since first, these mutants grow very poorly in the absence of exogenous long‐chain bases (unpublished results). Second, they only can grow when harbouring the slc1‐1 allele which allows them to introduce C26 FAs into sn‐2 of ordinary PIs (Lester et al., 1993). It may be more than a coincidence that all remodelling pathways introduce long‐chain FAs into the lipid moiety of GPI‐anchored proteins either by introducing C26‐containing ceramides or by introducing the long‐chain FA in the sn‐2 position of glycerophospholipid anchors. At present the data indicate that, with the possible exception of pG2‐type anchors, all GPI‐anchored proteins are becoming remodelled in a way as to receive a long‐chain FA. It may be speculated that the introduction of long‐chain FAs is required for the sorting of GPI proteins into specialized membrane domains or transport vesicles. We hope that mutants with defects in the remodelling enzymes will shed more light on the functional aspects of this phenomenon.
Materials and methods
Strains, growth conditions and materials
S.cerevisiae strains were W303‐1A (MATa ade2‐1 can1‐100 ura3‐1 leu2‐3,112 trp1‐1 his3‐11,15), X2180‐1A (MATa SUC2 mal gal2 CUP1), SJ21R (MATa ura3‐52 leu2‐3,112 ade1 MEL1), YNN5 (MATa ura3‐52 leu2‐3,112 ade1 MEL1 slc1Δ2::LEU2), SF 294‐2B (MATa sec7‐1), SF 226‐1C (MATa sec12‐4), HMSF169 (MATa sec14‐1), HMSF176 (MATa sec18‐1), HMSF331 (MATa sec53‐6), SF402‐4D (MATa sec59‐1), FBY12 (MATa ade2‐1 ura3‐1 leu2‐3,112 trp1‐1 his3‐11,15 gpi4‐1), FBY15 (MATα ade2‐1 leu2‐3,112 trp1‐1 his3‐11,15 gpi7‐1), FBY11 (MATα ade2‐1 leu2‐3,112 gpi8‐1), FBY16 (MATα ade2‐1 ura3‐1 leu2‐3,112 trp1‐1 his3‐11,15 lys2 gpi9‐1), XD6‐2B (MATα dpm1::LEU2 ura3‐52 leu2‐3,112 trp1‐Δ1 lys2‐801) harbouring DPM1 on plasmid pDM6, and XD6‐2B harbouring the thermosensitive dpm1‐6 allele on plasmid pDM8‐6 instead of pDM6. Cells were kept on YPD plates and were cultured in SD medium containing salts, vitamins (but no inositol), trace elements and 2% glucose as a carbon source; SDC medium is the same but contains 1% casein hydrolysate in addition. Uracil and adenine, each at 40 μg/ml, were added where required. The OD of dilute cell suspensions was measured in a 1 cm cuvette at 600 nm; ‘1 OD’ refers to cells contained in 1 ml of a cell suspension having an OD600 of 1 and is equivalent to ∼107 cells. myo[2‐3H]inositol (20 Ci/mmol) was from ARC Inc., St Louis, MO, USA. 1000‐fold purified GPI–PLD from bovine serum was a kind gift of Dr Urs Brodbeck, Institut für Biochemie und Molekularbiologie, University of Bern, Switzerland. Phospholipase A2 isolated from bee venom was from Sigma. 7N NH3 in methanol (methanolic ammonia) was bought from Janssen Chimica (Geel, Belgium). Concanavalin A–Sepharose and octyl–Sepharose 4B were from Pharmacia (Uppsala, Sweden). PI–PLC from Bacillus cereus and Pronase from Streptomyces griseus were from Boehringer‐Mannheim, Germany. TLC silica gel 60 plates were from Merck ABS, Zurich.
Metabolic labelling with [3H]inositol
Exponentially growing cells were pulse labelled for 5–30 min with [3H]inositol in SDC medium at 10 OD/ml with 5–10 μCi/OD of cells or, to visualize better pG2, at 15 μCi/OD. For chase, cells were diluted 4‐fold with SDC.
Preparation of anchor peptides
Radiolabelled cells were broken with glass beads and extensively delipidated with chloroform:methanol:water (10:10:3). The lysate was then resuspended in 200 μl of ethanol:water:diethylether:pyridine:25% NH3 (15:15:5:1:0.018) by sonication for 15 s and the suspension was left for 15 min at room temperature (Hanson and Lester, 1980). Samples were dried in a Speed‐Vac evaporator and glycoproteins isolated over concanavalin A–Sepharose after resolubilizing proteins in sample buffer as described (Conzelmann et al., 1990). Concanavalin A–Sepharose beads were washed until the effluent contained <500 c.p.m./ml. GPI peptides were eluted by extensive Pronase treatment and subsequent boiling of the concanavalin A beads. Anchor peptides were further purified over octyl–Sepharose as described (Sipos et al., 1994).
Purification of GPI intermediates
Total lipid extracts were digested with 0.02 U of PI–PLC in 20% propanol as described (Puoti and Conzelmann, 1993) for 5 h or overnight when extracts from more than 20 OD were to be treated. After drying the samples, the total lipid extracts were treated by nitrous acid as described (Güther et al., 1994); controls contained 0.5 M NaCl instead of 0.5 M NaNO2. Lipids were then extracted twice with 150 μl of water‐saturated butanol and back‐extracted by 50 μl of butanol‐saturated water. The HNO2‐generated lipids (more hydrophobic than any other [3H]inositol‐labelled lipid) were purified twice by preparative TLC using solvent system 3 (see below).
Specific release of inositol‐linked acyl groups was performed using methanolic NH3 in 200 μl for 2 h at 30°C in an air‐heated incubator; controls were incubated only with methanol. Mild base treatment for complete deacylation of glycerophospholipids was done using 8 N NH3 in 300 μl methanol:water (1:1) for 6 h at 37°C or using NaOH as described (Becker and Lester, 1980). Nitrous acid treatments were done as described (Güther et al., 1994). GPI–PLD and PI–PLC treatments were performed as described (Davitz et al., 1989; Puoti and Conzelmann, 1993). For PLA2 treatment, lipids were resolubilized in 30 μl buffer (25 mM Tris–HCl, pH 7.5, 2 mM CaCl2, 0.1% sodium deoxycholate) and treated with 2 U of enzyme for 5 h at 37°C. For TLC, lipids were desalted by butanol extraction if significant salt concentrations had to be removed, lipids were dried and resolubilized in chloroform:methanol (1:1) for application. Pre‐equilibrated tanks with the following solvent systems were used for ascending TLC: solvent system 1 [chloroform:methanol:0.25% KCl (55:45:10, v/v)]; solvent system 2 [chloroform:methanol:water (10:10:3, v/v)]; solvent system 3 [chloroform:methanol:0.25% KCl (55:45:5, v/v)]. In most cases, radioactivity on TLC plates was monitored and quantitated by one‐ and two‐dimensional radioscanning using a Berthold radioscanner.
We would like to thank Dr Urs Brodbeck for GPI–PLD, Dr M.A.J. Ferguson for radiolabelled lipid standards from T.brucei, Dr Philippe Renaud for helpful discussions, and Mohammed Benghezal and Urs Meyer for critically reading the manuscript and developing autoradiographs. This work was supported by grant 3100‐032515 of the Swiss National Foundation to A.C.
- Copyright © 1997 European Molecular Biology Organization