Csg1p and Newly Identified Csh1p Function in Mannosylinositol Phosphorylceramide Synthesis by Interacting with Csg2p*

Csg1p and Csg2p have been shown to be involved in the synthesis of mannosylinositol phosphorylceramide (MIPC) from inositol phosphorylceramide. YBR161w , termed CSH1 here, encodes a protein that exhibits a strong similarity to Csg1p. To examine whether Csh1p also functions in MIPC synthesis, we performed a [ 3 H]di-hydrosphingosine labeling experiment. (cid:1) csg1 cells exhibited only a reduction in the synthesis of mannosy-lated sphingolipids compared with wild-type cells, whereas the (cid:1) csg1 (cid:1) csh1 double deletion mutant exhibited a total loss. These results indicated that Csg1p and Csh1p have redundant functions in MIPC synthesis. Analyses using (cid:1) csg1 and (cid:1) csh1 cells in the (cid:1) ipt1 , (cid:1) sur2 , or (cid:1) scs7 genetic background demonstrated that Csh1p has a different substrate specificity from Csg1p. We also revealed that Csg2p interacts with both Csg1p and Csh1p. Deletion of the CSG2 gene reduced the Csg1p activity and abolished the Csh1p activity. These results suggested that two distinct inositol phosphorylceramide

Sphingolipids are ubiquitous and abundant components of eukaryotic plasma membranes. In the yeast Saccharomyces cerevisiae, there are three classes of sphingolipids that contain myo-inositol, namely inositol phosphorylceramide (IPC), 1 mannosylinositol phosphorylceramide (MIPC), and mannosyldiinositol phosphorylceramide (M(IP) 2 C). These sphingolipids contain one of five ceramide backbones that differ in hydroxylation state (Fig. 1). Dihydroceramide (ceramide A in Fig. 1) contains a dihydrosphingosine (DHS)-type long chain base and a C26 fatty acid. Sur2p can convert DHS to phytosphingosine or dihydroceramide to phytoceramide (ceramide B in Fig. 1) (1). Hydroxylation of the C26 fatty acid of ceramide A and ceramide B at the C-2 position by Scs7p yields ceramide BЈ and ceramide C, respectively ( Fig. 1) (1, 2). Ceramide D is generated by further hydroxylation, at an unknown position, of the fatty acid moiety of ceramide C, a reaction requiring Ccc2p, a possible Cu 2ϩ transporter (3). IPCs are synthesized from ceramides by Aur1p in the lumen of the Golgi apparatus (4,5). Csg1p/Sur1p and Csg2p/Cls2p are involved in the conversion of the IPCs to MIPCs (3,6), which, finally, are modified by Ipt1p to form the corresponding M(IP) 2 Cs (7).
Although neither Csg1p nor Csg2p is required for vegetative growth, both seem to be important for Ca 2ϩ signaling, because csg1 and csg2 mutants were originally isolated as Ca 2ϩ -sensitive mutants (8). Certain suppressor genes for a Ca 2ϩ -sensitive csg2 mutant were found to be genes involved in sphingolipid synthesis (1-3, 6, 9). This finding led to a later study that revealed the involvement of Csg1p and Csg2p in MIPC synthesis (3,6). Their precise functions and their relationship in MIPC synthesis remains unclear, although Csg1p is predicted to have a catalytic function based on the presence of a stretch of 93 amino acids with homology to those found in two yeast ␣-1,6-mannosyltranferases, Och1p and Hoc1p (3). Structurally, Csg2p is an integral membrane protein with up to 10 transmembrane segments that, when overexpressed, localizes to the endoplasmic reticulum (8,10). Csg2p is also known to have an EF-Ca 2ϩ -binding domain and has been implicated in the regulation of a nonexchangeable, intracellular Ca 2ϩ pool (11).
The predicted product of the open reading frame YBR161w (termed here CSH1 for CSG1/SUR1 homolog) is highly homologous to Csg1p. In this study we examined the possibility that this open reading frame is actually involved in the MIPC synthesis. The MIPC synthesis activity exhibited by a ⌬csg1 mutant was found to be completely lacking in a ⌬csg1 ⌬csh1 double mutant, indicating that both Csg1p and Csh1p are involved in the MIPC synthesis, redundantly. Moreover, we found that Csg2p associates with both Csg1p and Csh1p. Thus, Csg1p-Csg2p and Csh1p-Csg2p complexes may function as two distinct IPC mannosyltransferases.

EXPERIMENTAL PROCEDURES
Yeast Strains and Media-S. cerevisiae strains used are listed in Table I. The ⌬csg1::HIS3 and ⌬csg2::URA3 cells were constructed by replacing the 0.44-kb ClaI-EcoRI region in the CSG1 gene and the 0.41-kb EcoRI-HincII region in the CSG2 gene with HIS3 and URA3 markers, respectively. The ⌬csh1::LEU2 and ⌬csh1::KanMX4 cells were constructed by replacing the 0.59-kb EcoRV-XbaI region in the CSH1 gene and the entire open reading frame of the CSH1 gene with the LEU2 marker and the KanMX4 marker, respectively. The ⌬ipt1::KanMX4, ⌬sur2::KanMX4, ⌬scs7::KanMX4, and ⌬pep4::KanMX4 cells were constructed by replacing their entire open reading frames with the KanMX4 marker. The cells were grown either in YPD medium (1% yeast extract, 2% peptone, and 2% glucose) or in synthetic complete medium (0.67% yeast nitrogen base and 2% glucose) containing nutritional supplements.
Immunoblotting-Cell lysates for immunoblotting were prepared as described previously (13). Lysates equivalent to 0.2 A 600 cells were subject to SDS-PAGE and transferred to Immobilon TM polyvinylidene difluoride membrane (Millipore). The resulting membrane was incubated with a 1:1000 dilution of the anti-HA antibody HA7 (Sigma) or the anti-Myc antibody PL14 (Medical & Biological Laboratories) for 1 h and then with a 1:5000 dilution of horseradish peroxidase-conjugated donkey anti-mouse IgG F(abЈ) 2 fragment (Amersham Biosciences) for 1 h. Labeling was detected by the ECL detection method (Amersham Biosciences).

Csh1p Is Involved in the Synthesis of MIPC-Csg1p and
Csg2p are known to be involved in MIPC synthesis (3,6), yet several issues remain unresolved. For instance, it is unclear whether one of them, or both, is a mannosyltransferase, or if any other factors are involved in this reaction. To gain insight into these problems, we investigated the involvement of Csh1p, which is highly homologous to Csg1p (71% similarity and 65% identity; Csg1p and Csh1p share especially high similarity at the conserved regions spanning amino acid residues 54 -148 and 63-155, respectively. On the other hand, there is little homology between their C termini. Csh1p, as well as Csg1p, is predicted to have three transmembrane segments (Fig. 2, underlining). Both contain potential N-glycosylation sites, five in Csg1p and three in Csh1p (Fig. 2, dots).
To examine whether Csh1p also functions in MIPC synthesis, we constructed ⌬csg1 and ⌬csh1 mutants, as well as a ⌬csg1 ⌬csh1 double mutant, and compared these with wildtype cells for their ability to synthesize sphingolipids. The cells were cultured in YPD medium containing [ 3 H]DHS for 1 h (Fig.  3, top panel) or 3 h (Fig. 3, bottom panel), and then lipids were extracted and separated by TLC. In wild-type cells labeled for 1 h, IPC, MIPC, and M(IP) 2 C, each with several hydroxylation states, were detected. Among these IPC-C was the most prominent, followed by IPC-A, M(IP) 2 C-A, M(IP) 2 C-B/BЈ, and M(IP) 2 C-C. In the 3-h labeling experiment, the amounts of M(IP) 2 C-C and MIPC-C had increased, whereas the amounts of IPC-A and MIPC-B/BЈ had decreased. In both labeling experiments, MIPC was found in much lower quantities than IPC or M(IP) 2 C. Thus, the MIPC-to-M(IP) 2 C conversion appears to be rapid.
In the ⌬csg1 mutant, MIPC-C was not detected at either time period. Moreover, the amount of M(IP) 2 C-C was greatly reduced compared with that in wild-type cells (10 and 30% for 1and 3-h labeling, respectively). Concomitantly, however, slightly higher amounts of the IPC species were observed, especially in the 3-h labeling. On the other hand, the lipid composition of the ⌬csh1 mutant was quite similar to that of the wild-type cells (Fig. 3, lane 3). In the ⌬csg1 ⌬csh1 double mutant, however, both MIPC and M(IP) 2 C were completely absent, and the levels of all IPC species were increased compared with those in the wild-type cells (Fig. 3, lane 4). MIPC and M(IP) 2 C were not detected in the ⌬csg1⌬csh1 double mutant, even after labeling for 24 h (data not shown). The effects of deleting the CSG1 and CSH1 genes on the sphingolipid synthesis pattern were not restricted to the yeast background used (KA31-1A). Similar data were obtained using another yeast background (BY4741) (data not shown). Taken together, these results indicate that both Csg1p and Csh1p have functions, albeit redundant, in MIPC synthesis.
Csg1p and Csh1p Exhibit Different Substrate Specificities-As shown in Fig. 3, the [ 3 H]DHS-labeled pattern of ⌬csg1 cells differed from that of ⌬csh1 cells. The most prominent differences were a reduction in M(IP) 2 C-C and a loss of MIPC-C in the ⌬csg1 cells. These results implied that Csg1p and Csh1p might exhibit different substrate specificities. To examine this possibility, we constructed ⌬csg1 and ⌬csh1 mutants in a ⌬ipt1, ⌬sur2, or ⌬scs7 background and then performed the [ 3 H]DHS labeling experiments.
We first investigated sphingolipid synthesis in the mutant of IPT1. Because Ipt1p is involved in converting MIPC to M(IP) 2 C, M(IP) 2 C is not synthesized in the ⌬ipt1 cells. Sphingolipid synthesis in the ⌬csh1 cells in the ⌬ipt1 background was quite similar to that in the control CSH1 ϩ cells (Fig. 4A,  lanes 1 and 3). On the other hand, the ⌬csg1 mutation caused a reduction in MIPC-C to 20% of that in the CSG1 ϩ cells, with a concomitant increase in IPC-C (Fig. 4A, lanes 1 and 2). MIPC-A and MIPC-B/BЈ levels observed in the ⌬csg1 mutant resembled those of the wild-type cells. The amounts IPC-D and MIPC-A also did not differ, although they were not separated by the original solvent system used (Fig. 4A, lanes 1-3). Another solvent system, in which the two separate, confirmed that the amounts of both lipids were unchanged by the ⌬csg1 mutation (Fig. 4A, lanes 4 and 5).
Because B and BЈ species exhibit very similar mobilities, we discriminated between these two using ⌬sur2 and ⌬scs7 mutations. We first investigated sphingolipid synthesis in the ⌬sur2 background, in which only the A and BЈ species are synthesized. The amounts of IPC, MIPC, and M(IP) 2 C in the ⌬csg1 ⌬sur2 double mutant were quite similar to those in the control ⌬sur2 mutant (Fig. 4B, lanes 1 and 2). The labeling pattern in the ⌬csh1⌬sur2 cells was nearly the same as that in the control  (36) program predicts both proteins to be integral membrane proteins with three transmembrane segments (underlined). The 93-amino acid segments of Csg1p and Csh1p that exhibit homology to the two yeast ␣-1, 6-mannosyltransferases Och1p and Hoc1p are designated by a dashed line. The dotted asparagine residues are potential N-glycosylation sites. CSH1 ϩ cells, although a slight decrease in the amount of MIPC-BЈ was observed in the ⌬csh1⌬sur2 cells.
We next investigated the substrate preferences of Csg1p and Csh1p using the ⌬scs7 background, in which only the A and B species are synthesized. Sphingolipid synthesis in the ⌬csh1 ⌬scs7 double mutant was quite similar to that in the CSH1 ϩ ⌬scs7 mutant (Fig. 4C, lanes 1 and 3). On the other hand, the ⌬csg1 mutation caused a reduction in M(IP) 2 C-B to 10% of that in the CSG1 ϩ cells (Fig. 4C, lanes 1 and 2). Taken together, these results lead to the conclusion that Csh1p exhibits a similar activity to Csg1p against IPC-A and IPC-BЈ but a weaker activity against IPC-B and IPC-C than Csg1p.
Csg2p Interacts with Both Csg1p and Csh1p-It has been reported that Csg2p is involved in MIPC synthesis (3,6). Therefore, we hypothesized that Csg2p may function together with Csg1p and Csh1p by forming complexes with them. To examine this hypothesis, we performed a pull-down assay using a Ni-NTA column. We constructed a pSU30 plasmid encoding Csg1p-His 6 -Myc, a pSU41 plasmid encoding Csh1p-His 6 -Myc, and a pSU8 plasmid encoding Csg2p-3ϫHA. Expression of these tagged proteins in transfected KHY195 cells was confirmed by immunoblotting using anti-Myc or anti-HA antibodies. Csg1p-His 6 -Myc was detected as a band of 46 kDa (the predicted molecular mass is 48 kDa) (Fig. 5A, lane 2). Csh1p-His 6 -Myc was detected as several bands of 75, 60, and 47 kDa, and bands with lower molecular masses (Fig. 5A, lane 4). Of these bands, the 47-kDa band was most likely to be the intact Csh1p-His 6 -Myc (predicted molecular mass is 48 kDa). It is possible that the 75-and 60-kDa forms are modified by glycosylation because Csh1p contains potential N-glycosylation sites (Fig. 2, dots). The bands with lower molecular masses may be degradation products. Csg2p-3ϫHA was detected as a weak band of 47 kDa (the predicted molecular mass is 49 kDa) (Fig.  5A, lane 6); several bands with faster gel mobilities, including a prominent band of 24 kDa, were also detected. These bands may be generated by degradation caused by nonphysiological overproduction.
Using these constructs, we next examined the possibility that Csg2p interacts with Csg1p and/or with Csh1p. Cell lysates prepared from KHY195 cells bearing pSU8 (CSG2-3ϫHA) together with either pRS423 (vector) or pSU30 (CSG1-HIS 6 -MYC) were solubilized with Triton X-100 and applied to a Ni-NTA column, which binds His 6 -tagged proteins. The bound proteins were then eluted with 250 mM imidazole. As expected, Csg1p-His 6 -Myc was collected in the eluted fraction (Fig. 5B,  lane 2). Although Csg2p-3ϫHA itself had no affinity to the Ni-NTA column, it was recovered in the eluted fraction when co-expressed with Csg1p-His 6 -Myc (Fig. 5B). These results indicate an association between Csg2p-3ϫHA and Csg1p-His 6 -Myc. Moreover, the putative degradated Csg2p-3ϫHA was also co-eluted with Csg1p-His 6 -Myc (Fig. 5B, asterisk), suggesting that the N-terminally truncated Csg2p retains the Csg1p-interacting site.
Determination of the Role of Csg2p in MIPC Synthesis-Next, we examined the effect of the ⌬csg2 mutation on the synthesis of MIPC. We constructed ⌬csg2, ⌬csg1 ⌬csg2, ⌬csh1 ⌬csg2, and ⌬csg1 ⌬csg2 ⌬csh1 cells. A [ 3 H]DHS labeling experiment using the ⌬csg2 cells demonstrated that deletion of the CSG2 gene results in a drastic decrease in the amounts of MIPC and M(IP) 2 C and an increase in the amount of IPC (Fig.  6A, lane 2). Only M(IP) 2 C-BЈ/B, and M(IP) 2 C-C were detected, barely (Fig. 6A, lane 2). This effect of the ⌬csg2 mutation was much greater than that observed in the ⌬csg1 or the ⌬csh1 cells (Fig. 3). Thus, Csg2p has important functions for both complexes (Csg1p-Csg2p and Csh1p-Csg2p) in eliciting their enzyme activities. Next, to examine whether Csg1p or Csh1p is responsible for the residual activity observed in the ⌬csg2 cells, we examined the labeling patterns of the ⌬csh1⌬csg2 and ⌬csg1⌬csg2 cells. The labeling pattern of the ⌬csh1 ⌬csg2 double mutant was identical to that of the ⌬csg2 mutant (Fig. 6A,  lane 3). In contrast, the small amounts of M(IP) 2 C-BЈ/B and MIPC-C observed in the ⌬csg2 cells were completely lacking after the introduction of the ⌬csg1 mutation (Fig. 6A, lane 4). These results indicate that, compared with Csg2p-associated Csg1p, the MIPC synthesis activity of free Csg1p is significantly lower, but free Csg1p does retain detectable MIPC-BЈ/B and MIPC-C synthesis activity. On the other hand, free Csh1p could not produce MIPC at all. Thus, Csg2p is essential for the MIPC synthesis activity of the Csh1p-Csg2p complex and is also quite important for that of the Csg1p-Csg2p complex.
Ca 2ϩ Sensitivity Correlates with MIPC Synthesis Activity-We next examined the effect of the ⌬csh1 mutation on Ca 2ϩ sensitivity. The cells were grown on a YPD plate containing 100 mM Ca 2ϩ at 30°C. As shown in Fig. 6B, the ⌬csg1 and ⌬csh1 mutants were resistant to exogenous Ca 2ϩ and exhibited normal growth. On the other hand, the ⌬csg2 mutant, as well as the ⌬csg1 ⌬csh1 double mutant, was highly sensitive to Ca 2ϩ and would not grow on the medium. Identical results were obtained when the cells were grown at 37°C (data not shown). Thus, the Ca 2ϩ sensitivities of these mutants correlate well with their MIPC synthesis activities. DISCUSSION Although previous studies have suggested the involvement of Csg1p and Csg2p in MIPC synthesis (3,6), their precise roles in that reaction have not been defined. We reveal here that Csg1p and Csg2p form a heterodimer that functions as a MIPC synthase. Additionally, we identify that Csh1p is also involved in the MIPC synthesis and can also function in a complex with Csg2p. Using yeast mutants carrying deletions for the CSG1 or CSH1 gene, in various gene-ablated backgrounds, we examined their substrate specificities. Csh1p exhibited similar activity against IPC-A and IPC-BЈ but lower activity against IPC-B and, especially, IPC-C as compared with Csg1p (Fig. 4). These results indicate that Csg1p and Csh1p are involved in the MIPC synthesis redundantly with different substrate specificities.
The exact catalytic subunit of MIPC synthesis (Csg1p and Csh1p, Csg2p, or other) has not been verified. Nevertheless, we propose a model that Csg1p and Csh1p are catalytic subunits with Csg2p as a regulatory subunit, based on several supportive results. First, Csg1p and Csh1p contain stretches of 93 amino acids with homology to those found in two yeast ␣-1,6mannosyltranferases, Och1p and Hoc1p (3), but Csg2p does not share this sequence. Second, previous studies have shown that although overexpression of Csg1p suppresses the Ca 2ϩ -sensitive phenotype of a csg2 mutant, in the inverse experiment, overexpression of Csg2p did not suppress the Ca 2ϩ sensitivity of a csg1 mutant (3). Thus, the role of Csg2p appears more regulatory rather than directly enzymatic. Finally, M(IP) 2 C was detected, albeit weakly, in the ⌬csh1⌬csg2 cells, suggesting that, in the absence of Csg2p, Csg1p could produce MIPC.
The growth of the ⌬csg1 mutant on a YPD plate containing Ca 2ϩ was as normal as that of the wild-type cells at both 30 and 37°C ( Fig. 6B and data not shown). In contrast, the ⌬csg2 cells and the ⌬csg1 ⌬csh1 double mutant cells, both of which exhibited more severe defects in MIPC synthesis, were highly sensitive to exogenous Ca 2ϩ . The Ca 2ϩ resistance of the ⌬csg1 cells in this study is not consistent with a previous report (3) that the ⌬csg1 mutant exhibits a Ca 2ϩ -sensitive phenotype at 37°C but not at 26°C (3). It is possible that the contribution of Csg1p to the total MIPC synthesis activity differs among different yeast backgrounds. Indeed, the ⌬csg1 cells used in the previous report, on a DBY685 background, exhibited a severe defect in MIPC synthesis (3).
Several lines of evidence suggest that the Ca 2ϩ sensitivity is caused by an accumulation of IPC-C, which contains a phytosphingosine-type long chain base and a hydroxylated very long chain fatty acid at the C2 position. The Ca 2ϩ -sensitive phenotype of the csg2 mutant was suppressed by several mutations that caused a reduction in the IPC-C concentrations (6). These mutations included genes involved in long chain base synthesis (lcb1, lcb2, tsc3, and tsc10) (6,9,15), palmitoyl-CoA synthesis (fas2) (9), elongation of very long chain fatty acids (tsc13) (9,16), conversion of DHS to phytosphingosine (sur2) (1), and hydroxylation of very long chain fatty acids (scs7) (1). Moreover, overproduction of Ccc2p, which may stimulate the conversion of IPC-C to IPC-D, also rescued the Ca 2ϩ -sensitive phenotype of the csg2 mutant (3). However, the reason the accumulation of IPC-C causes Ca 2ϩ sensitivity remains unclear. One possibility is that IPC-C itself acts as a signaling molecule. Recent studies have indicated that several sphingolipid metabolites and intermediates function as signaling molecules. In yeast, sphingoid long chain base is involved in endocytosis (17) and in cell cycle arrest at G 0 /G 1 (18), and sphingoid long chain base 1-phosphate has been implicated in heat stress resistance, diauxic shift, and Ca 2ϩ mobilization (19 -23). Moreover, accumulation of ceramide inhibits growth (24,25). Thus, it is possible that IPC-C is another potent signaling molecule involved in the Ca 2ϩ signaling pathway. Other suppressors of Ca 2ϩ sensitivity in the csg2 mutant were found to be genes involved in signal transduction. These included genes encoding the protein kinase TOR2 and the phosphatidylinositol-4-phosphate 5-kinase MSS4 (9). Tor2p functions in a nutrient-sensing signaling cascade and in cell cycle-dependent reorganization of actin cytoskeleton (26 -30). Mss4p is also known to be involved in the organization of actin cytoskeleton (31,32). Thus, IPC-C might connect the increase of Ca 2ϩ in the cells and the signaling pathways of Tor2p and Mss4p.
Another possibility is that IPC-C regulates Ca 2ϩ signaling by modulating the formation of microdomains. Recent theory suggests the existence of microscopic regions called microdomains (or lipid rafts) in the plasma membrane, which function as platforms for effective signal transduction (33). In yeast, the microdomain is composed of ergosterol and sphingolipids (34). It has been reported that several proteins such as Pma1p, Gas1p, and Nce2p associate with the microdomain (34). Consistent with this model, Ca 2ϩ changes the sphingolipid composition by stimulating the conversion of IPC to MIPC. 2 Because Csg2p contains an EF-Ca 2ϩ -binding domain (8,10), we propose that Csg1p or Csh1p activity may be regulated by Ca 2ϩ through Csg2p.