Cdc1 removes the ethanolamine phosphate of the first mannose of GPI anchors and thereby facilitates the integration of GPI proteins into the yeast cell wall.

Temperature-sensitive cdc1(ts) mutants are reported to stop the cell cycle upon a shift to 30°C in early G2, that is, as small budded cells having completed DNA replication but unable to duplicate the spindle pole body. A recent report showed that PGAP5, a human homologue of CDC1, acts as a phosphodiesterase removing an ethanolamine phosphate (EtN-P) from mannose 2 of the glycosylphosphatidylinositol (GPI) anchor, thus permitting efficient endoplasmic reticulum (ER)-to-Golgi transport of GPI proteins. We find that the essential CDC1 gene can be deleted in mcd4∆ cells, which do not attach EtN-P to mannose 1 of the GPI anchor, suggesting that Cdc1 removes the EtN-P added by Mcd4. Cdc1-314(ts) mutants do not accumulate GPI proteins in the ER but have a partial secretion block later in the secretory pathway. Growth tests and the genetic interaction profile of cdc1-314(ts) pinpoint a distinct cell wall defect. Osmotic support restores GPI protein secretion and actin polarization but not growth. Cell walls of cdc1-314(ts) mutants contain large amounts of GPI proteins that are easily released by β-glucanases and not attached to cell wall β1,6-glucans and that retain their original GPI anchor lipid. This suggests that the presumed transglycosidases Dfg5 and Dcw1 of cdc1-314(ts) transfer GPI proteins to cell wall β1,6-glucans inefficiently.

INTRODUCTION

Glycosylphosphatidylinositol (GPI) anchoring in yeast and mammals

In all eukaryotes GPI lipids are posttranslationally attached to the C-terminus of certain proteins in the lumen of the endoplasmic reticulum (ER). Genetic ablation of GPI anchoring leads to embryonic lethality in humans and lethality in yeast (Maeda and Kinoshita, 2011). While all GPI proteins in mammals are exposed at the plasma membrane, only about half of yeast GPI proteins stay in the plasma membrane; the other half loses the GPI lipid moiety and gets covalently attached to the cell wall β1,6-glucans (Caro ; Hamada ; de Groot ). Indeed, among the yeast cell wall proteins (CWPs) that are covalently attached to β-glucans, the majority are initially GPI anchored (GPI-CWPs), whereas a minority are linked via an alkali-sensitive linkage to β1,3-glucans (ASL-CWPs) (de Groot ).

The activity transferring GPI proteins from the plasma membrane onto β-glucans may reside with Dfg5 and Dcw1, two highly homologous GPI proteins (54% identity; Kitagaki ). During the transfer of GPI-CWPs to cell wall glucans, the glucosamine–inositol–phospholipid moiety of the anchor is lost, whereas four to five mannoses (Man), the phosphodiester bond, and the bridging ethanolamine (EtN) remain attached to the protein (Figure 1). Therefore Dfg5 and Dcw1, having homology with bacterial endomannosidases, are proposed to cleave the Manα1,4glucosamine linkage of the GPI anchor and to reattach the liberated Man residue with the attached GPI protein to β1,6-glucans of the cell wall (Kollar ; Fujii ; Kitagaki ). Simultaneous deletion of DFG5 and DCW1 is lethal, suggesting that the covalent attachment of GPI-CWPs to glucans is essential, and this remains true even if cells receive osmotic support (Kitagaki ).

FIGURE 1:

GPI anchor remodeling and subsequent attachment to the cell wall. Essential genes are indicated in blue. Mature GPI anchor lipids (left) contain EtN-Ps on Man1, Man2, and Man3; their addition is catalyzed by MCD4, GPI7, and GPI13, respectively. After the protein is attached to the GPI lipid by the transamidase complex, the lipid moiety is remodeled by Bst1, Per1, Gup1, and Cwh43 (middle). At some time during lipid remodeling, EtN-Ps may be removed from Man1 and Man2. At the plasma membrane, Man1 of the anchor of GPI-CWPs is transferred and covalently linked to β1,6-glucans. This reaction is presumably catalyzed by Dcw1 and Dfg5.

GPI lipid remodeling in yeast

GPI lipids in yeast and mammals are constructed in the ER by sequential addition of sugars to a phosphatidylinositol (PI) and then are attached to proteins, the entire process requiring more than 20 genes. The GPI lipid usually contains C16:0 and C18:1 fatty acids (Pittet and Conzelmann, 2007; Fujita and Jigami, 2008). After attachment of the GPI lipid to yeast GPI proteins, the lipid moiety is remodeled so as to contain a C26:0 fatty acid rather than C18:1 or a phytosphingosine-C26:0–type ceramide. Ceramide anchors are also found in the GPI anchors of other fungi, certain plants, Dictyostelium, and protozoan organisms (Fankhauser ; Pittet and Conzelmann, 2007; Fujita and Jigami, 2008). These remodeling reactions occur in the ER and are a prerequisite for the concentration of GPI proteins in the ER exit sites (Castillon , 2011).

By composition, yeast GPI anchors of detergent-extractable GPI proteins and GPI anchor remnants of GPI proteins linked to cell wall β-glucans did not contain EtN-phosphate (EtN-P) side chains on Man1 and Man2 (Figure 1; Fankhauser ; Fujii ). While such residues may have been lost during purification, these reports are certainly compatible with the idea that EtN-P side chains are removed during maturation of GPI proteins. Yeast CDC1 and TED1 show 21 and 23% identities to 73 and 39% of PGAP5 sequence, respectively, and TED1 shows 23% identity to 33% of CDC1 sequence. Moreover, ted1∆ mutants show a similar GPI protein transport defect as PGAP5 mutants (Haass ). Thus CDC1 and TED1 are candidates for enzymes removing EtN-P side chains.

Discovery of CDC1

CDC1 is an essential gene. Temperature sensitive (ts) cdc1 alleles were identified as cell cycle mutants accumulating upon a shift to nonpermissive temperature as cells with no or only a small bud, mostly duplicated DNA, a nonduplicated spindle pole body, and an undivided nucleus (Paidhungat and Garrett, 1998b). Subsequent work revealed that certain cdc1 alleles are rescued by supplementing media with Mn2+ or overexpression of plasma membrane Mn2+ transporters Smf1 or Smf2. Moreover, even wild-type (WT) cells, when deprived of Mn2+, stop cycling and exhibit small buds, duplicated DNA, and an undivided nucleus (Loukin and Kung, 1995; Supek ; Eguez ). These data led to the proposal that Cdc1 regulates the Mn2+ concentrations in cells, and this view was supported by the finding that deletion of PER1, now known to remove the sn2-linked fatty acid of the primary GPI anchor (Figure 1), allows for the deletion of CDC1 as long as Mn2+ is present in high concentrations in the media (Paidhungat and Garrett, 1998a).

A more recent study found strong evidence that Cdc1 is not regulating but is regulated by the intracellular Mn2+ concentration and that it is a Mn2+-dependent phosphodiesterase. Indeed, mutation of amino acids belonging to the Mn2+-binding motif caused a Cdc1-deficiency phenotype (Losev ). These studies also showed that many cdc1 cells at 30°C have an elevated Ca2+ content and that elevated cytosolic Ca2+ levels contribute to the growth phenotype, to actin depolarization, and, related to this, a Golgi inheritance defect, whereby these phenomena are suppressed upon deletion of plasma membrane calcium channel components Mid1 or Cch1 (Paidhungat and Garrett, 1997; Rossanese ; Losev ). However, abolition of the calcium channel did not allow growth of cdc1 cells at 37°C.

The above-mentioned GPI anchor modification function of the mammalian homologue PGAP5 drove us to investigate the effect of cdc1 mutants on GPI protein biosynthesis in yeast.

RESULTS

Does Cdc1 remove an EtN-P from either Man1 or Man2?

EtN-P is added to Man1, Man2, and Man3 of the GPI lipid precursor by Mcd4, Gpi7, and Gpi13, respectively (Figure 1). Among these three paralogues, only GPI7 is not essential. Previous data indicated that ted1∆ mutants retain the GPI protein Gas1 in the ER and that gpi7∆ and ted1∆ each induce a strong unfolded protein response (UPR), which does not get stronger in a gpi7∆/ted1∆ double mutant (Jonikas ). This mutual phenotypic suppression but absence of correlation of the phenotypic interaction profiles of ted1∆ and gpi7∆ raises the possibility that Ted1 removes the EtN-P from Man2, explaining why the UPRs of ted1∆ and gpi7∆ are not aggravating each other. This paradigm suggests that the lack of a EtN-P phosphodiesterase may be compensated by the lack of the EtN-P transferase adding the EtN-P that cannot be removed.

We did not find any negative genetic interaction of the temperature-sensitive cdc1-314 allele with gpi7∆ (see Supplemental Table S1) nor is such an interaction recorded in BioGRID (http://thebiogrid.org). Thus the genetic data do not indicate that Cdc1 would have the same function as Ted1.

To investigate whether Cdc1 may be involved in the removal of EtN-P from Man1, we produced a cdc1∆/mcd4∆ haploid strain harboring two different plasmids carrying yeast CDC1 and TbGPI10. MCD4 is an essential gene, because Gpi10, the mannosyltransferase adding Man3, does not work on GPI lipid intermediates lacking EtN-P on Man1, but MCD4 becomes nonessential if yeast harbors the GPI10 orthologue from Trypanosoma brucei, a species that does not add EtN-P onto Man1 (Zhu ). Mcd4∆ cells have dividing times three times longer than WT in liquid media and also grow badly on plates (Zhu ). As shown in Figure 2, cdc1∆ harboring TbGpi10 is unable to grow but mcd4∆/cdc1∆ harboring TbGpi10 grows as fast as mcd4∆ harboring TbGpi10. This indicates that deletion of CDC1 in a mcd4∆ background has no negative growth effect and our data (see Figure 6A, rows mcd4∆ and mcd4∆/cdc1, later in this article) show that it does not aggravate the cell wall defect of mcd4∆. Thus, whatever the problem caused by the deletion of the essential CDC1 gene may be, it is fully compensated by not adding EtN-P to Man1 during the biosynthesis of the GPI lipid precursor. This constellation strongly suggests that Cdc1 has specialized in removing EtN-P from Man1.

FIGURE 2:

The essential CDC1 gene can be deleted in the mcd4∆ mutant, which fails to add EtN-P onto Man1 of the GPI anchor. A diploid mcd4∆/MCD4 cdc1∆/CDC1 strain harboring vectors expressing GPI10 from Trypanosoma brucei (TbGPI10) (LEU2) and yeast CDC1 (URA3) was sporulated and dissected. A tetratype tetrad is shown at the top left; haploid offspring carrying mcd4∆ are producing very small colonies. The four colonies of this tetrad were grown, and 10-fold dilutions of cells were spotted on SC media with indicated supplements and grown at 30°C for 2–3 d.

FIGURE 6:

cdc1 cells have fragile cell walls. (A) Fourfold serial dilutions of the indicated strains were spotted onto YPD containing low concentrations of cell wall–perturbing agents and incubated at 30°C (mcd4∆ strains) or 27°C (all others). (B) Same as A but plates contained sorbitol (1 or 1.4 M) and were incubated at indicated temperatures. The same picture as in A was also found when the identical assays were repeated on SC medium (unpublished data). (C) WT and cdc1 were also tested for resistance to cell wall–perturbing agents in the presence of 1.4 M sorbitol, at 30°C for the plate containing CFW and at 32°C for all others. (D) CFW staining of WT and cdc1 cells grown in YPD at 30°C for 16 h. The pictures are directly comparable, as they were taken using the same exposure time and were processed in the same way. All plates were incubated 2 d except for the ones with mcd4∆ strains (4 d).

Genetic interactions of cdc1

We sought to better understand the function of Cdc1 by doing a systematic genetic array (SGA) study to find other mutations that would modulate the growth of a well-characterized temperature-sensitive cdc1-314 allele that was found to have a rather normal affinity for its essential cofactor Mn2+ but showed strong Ca2+ accumulation and an actin depolarization defect at 33.5°C (Losev ). In the rest of this paper, ordinary strains containing this allele will be referred to as cdc1 and query strains for SGA will be referred to as cdc1*.

As shown in Figure 3A, cdc1* was unable to grow at 30°C; its growth was restored by WT CDC1, but not cdc1D144A, which has a point mutation in the Mn2+-binding motif (Losev ). The CDC1* and cdc1* (cdc1-314) query strains were both crossed in quadruplicate to a library of 5850 mutants comprising deletions for each of the nonessential genes and 879 DAmP alleles of essential genes (Breslow ).

FIGURE 3:

Genetic interactions of cdc1. (A) Fourfold dilutions of cdc1, cdc1 harboring WT CDC1, or the functionally dead cdc1D144A allele were incubated at 24 or 30°C. (B) Small areas from plates used to obtain colony sizes and Z-scores for cdc1/geneX∆ double mutants are shown. The upper three and lower three areas each show the same mutants selected either at 24, 26, or 30°C. Two double mutants (in quadruplicate) showing significant negative interaction at 24 as well as 26°C, or only at 26°C, are boxed in yellow and red, respectively, one showing positive genetic interaction is boxed in blue. The same mutants at the temperatures at which they do not get significant Z-scores are in dotted-line boxes. (C) Z-scores observed at 24 and 26°C of ∼5500 deletion strains remaining after the first filtering (see Materials and Methods) are plotted. Significant hits (Z-scores > 2.7; p values < 0.01) found only at 24°C are shown in green, hits seen only at 26°C in red, and hits found in both screens in yellow. (D–F) Negative interactions found at 24 and 26°C (D and E) and positive ones found at 30°C (F) were manually clustered into functional categories. Functional classes enriched at both 24 and 26°C are in bold. Only interactions remaining after a second filtering (see Materials and Methods) are reported.

Screens were done in parallel at 24, 26, and 30°C to detect synthetic sick interactions at permissive (24°C) and semipermissive (26°C) temperatures and suppressors at nonpermissive (30°C) temperature. Typical examples of such interactions are seen in the colored boxes of Figure 3B. The scores of the interactions retained after a first filtering (see Materials and Methods) had a normal distribution, and more negative than positive significant interactions were found (Figure 3C, colored dots). A complete list of the strains used in the SGA screen and of significant interactions (p < 0.01) is given in Supplemental Table S1.

Significant hits were manually attributed to functional categories, as shown in Figure 3, D–F, and Table S2. Among negative interactors (Figure 3, D and E), significant enrichments were found in genes orchestrating cell wall biosynthesis, ER protein folding, N-glycosylation, antero- and retrograde transport between ER and Golgi, and vacuolar H+-ATPase assembly. The latter may score because cytosolic calcium levels are regulated in part by the vacuolar cation/H+-exchanger Vcx1. Indeed, deletions in the plasma membrane calcium channel components Mid1 or Cch1 were shown to rescue growth of cdc1-1 at 30°C on yeast–peptone–dextrose (YPD) medium (Paidhungat and Garrett, 1997; Losev ), but our screen did not pick up these suppressors, possibly because the Ca2+ concentration in synthetic complete (SC) medium used here is sixfold higher than in YPD. Interestingly, a strong negative genetic interaction of cdc1 with cnb1∆, deleting the obligatory regulatory subunit of the calcium-dependent protein phosphatase calcineurin, was seen in our screen and also noted in a previous report (Paidhungat and Garrett, 1997). It suggests that the elevation of cytosolic Ca2+ may also have a positive effect of unknown nature. This positive effect is not mediated by the calcineurin-dependent transcription factor Crz1, because crz1∆ had no negative effect on the growth of cdc1 (Table S1).

Also, the Mn2+ and metal-ion transporters Smf1 and Smf2 of the plasma membrane were identified as negative interactors, suggesting that cdc1-314 may still be somewhat sensitive to Mn2+ depletion (Losev ). At 26°C, genes affecting ribosomal protein production also were enriched (Figure 3F). It is conceivable that the response to loss of Cdc1 function involves active protein synthesis.

The 22 deletions rescuing growth at 30°C mostly fell into one of three classes: the ER-associated protein degradation (ERAD) pathway, ER-to-Golgi transport, and GPI remodeling. In the latter category fell gup1∆ and per1∆, which had previously been identified as suppressors of the growth defect of cdc1 mutants (Paidhungat and Garrett, 1998a). Some of the suppressor mutants were tested also by serial dilution assays, and all of them were found to enhance growth of cdc1 in serial dilution growth assays, as shown in Supplemental Figure S1.

In summary, the screen suggested several biological processes that either aggravate or mitigate the effects of the cdc1-314 mutation, the effects of which we explored further as described in the following sections.

Stability of Cdc1-314 and Cdc1 proteins

Because ERAD mutants rescued growth of cdc1 (Figure 3F), we hypothesized that ERAD mutants may stabilize Cdc1-314. To verify this, we expressed HA-Cdc1 or HA-Cdc1-314, that is, tagged versions carrying the hemagglutinin-derived HA epitope (HA) under the native promoter from a centromeric vector in WT cells and quantified the amounts of HA-tagged protein after shifting cells from 24° to 30° or 37°C, as shown in Figure 4A. In cells growing at 24°C, Cdc1-314 already was 3.8-fold less abundant than WT Cdc1. A shift to 30°C did not destabilize Cdc1-314, although cells depending only on mutant protein do not grow at this temperature (Figure 3A). When cells were shifted to 37°C, a third of Cdc1-314 was degraded within 15–30 min, but the protein remained stable thereafter. The same temperature shifts reduced the amounts of WT Cdc1, but more slowly, so that it took more than 4 h to reach equilibrium levels.

FIGURE 4:

Stability of Cdc1-314 and Cdc1 proteins. (A) WT cells harboring HA-tagged Cdc1 or Cdc1-314 proteins were incubated at 24°C overnight (ON) and then shifted to 30 and 37°C for the indicated times. Thereafter cell extracts were subjected to SDS–PAGE and Western blotting using anti-HA antibodies. Adh1 was detected as a loading control. Signals of two biological replicates were averaged and normalized using the Adh1 signals and reference samples (see Materials and Methods). (B) Stability of HA-tagged Cdc1 and Cdc1-314 proteins in cells having gene deletions that were identified as suppressors. Cells were incubated overnight at 30 and 37°C before being processed as in A. Quantifications of A and B are not directly comparable, as normalizations were done only within the experiments of each panel.

Deletion of the ERAD component HRD3 increased the amount of WT Cdc1 and Cdc1-314 ∼2.3-fold and fourfold, respectively, so Cdc1-314 in hrd3∆ is ∼1.8-fold more abundant than Cdc1 in HRD3 cells, at both 30 and 37°C (Figure 4B).

These results suggest that WT Cdc1 is constitutively turned over by ERAD and that the Cdc1-314 is turned over more rapidly than Cdc1 already at 24°C. Moreover, the results strongly suggest that ERAD mutants rescue cdc1 cells at 30°C through stabilization of Cdc1-314. A similar mechanism may also be invoked for the deletion of SVP26, which raises Cdc1-314 levels twofold at 30°C (Figure 4B). On the other hand, the suppressor effect of emp24∆ and gup1∆ cannot be explained in terms of Cdc1-314 stability.

Does cdc1 affect the ER exit of GPI proteins and induce a UPR?

Deletion of TED1 causes ER retention of Gas1 (Haass ) and therewith induces a UPR (Jonikas ). We therefore decided to verify whether cdc1 cells show similar ER retention of GPI proteins and display signs of ER stress. Gas1 is an abundant GPI-anchored β1,3-transglucosidase, which, after extensive N- and O-glycosylation in the ER, runs with an apparent molecular weight (MW) of 105 kDa on SDS–PAGE, but at 125 kDa after elongation of its glycans in the Golgi (Fankhauser and Conzelmann, 1991). Pulse-chase experiments shown in Figure 5A indicate that cdc1 cells export Gas1 out of the ER with normal kinetics, whereas its export is delayed in ted1∆ mutants. Similarly, as shown by Figure 5B, Western blotting of extracts of cdc1 cells having been at 24, 30, 33, or 37°C during 4 h show no significant accumulation of an immature ER form of Gas1, much in contrast to bst1∆ and per1∆ cells (Figure 1). Ccw12 is a GPI-CWP undergoing extensive elongation of its three N-glycans in the Golgi, a process that raises the apparent MW of its 50- and 58-kDa forms to >200 kDa (Ragni ). The disappearance of the 50/58-kDa forms of Ccw12 upon addition of cycloheximide was complete within 15 min and followed similar kinetics in cdc1 and WT cells, indicating that ER-to-Golgi transport and elongation of glycans are not compromised in the cdc1 mutant (Figure S2).

FIGURE 5:

Remodeling and export of GPI proteins and induction of UPR in cdc1. (A) WT and mutants were pulse labeled with [35S]Met/Cys and chased for indicated times. Gas1 was immunoprecipitated and detected by autoradiography. (B) Western blots using anti-Gas1 antibody of total protein extracts from strains grown for 4 h at 24, 30, 33, or 37°C. (C) WT and cdc1 carrying an UPRE-lacZ plasmid were preincubated for 5 or 3 h at 27 or 30°C, then further incubated in the presence or absence of DTT for 2 h. Thereafter UPR induction was assessed by measuring β-galactosidase activity, which was plotted as absolute activity (upper plots) or fold induction caused by DTT (lower plots). (D) WT or cdc1 cells were preincubated at 37°C for 60 min (upper panel) or 10 min (lower panel) and then labeled with [3H]myo-inositol for 2 h. Anchor peptides from the total of SDS-extractable GPI proteins were isolated, lipid moieties were released by HNO2 treatment, subjected or not to mild alkaline deacylation with NaOH, resolved by TLC, and detected by autoradiography. No lipids were seen when HNO2 treatment was omitted, indicating that only GPI anchor lipids had been isolated. An aliquot of free lipids (FL), not attached to GPI anchors, was run on the same TLC for comparison. (E) Lipid extracts from cells radiolabeled for the upper panel in D were analyzed by TLC and autoradiography before and after deacylation with NaOH. (F) Lipid extracts from [14C]serine-labeled cells preincubated and labeled at 30°C for 5 h were analyzed as in E.

The accumulation of Gas1 in per1∆ and bst1∆ cells is also observed in the cdc1 background, suggesting that the persistence of EtN-P on Man1 does not accelerate ER exit of GPI proteins (Figure S3A).

To measure a potential induction of the UPR, we introduced a multicopy plasmid with the UPR element (UPRE) upstream of lacZ into strains (Menzel ). As seen in Figure 5C, cdc1 showed less UPR activation than WT, at both nonpermissive (30°C) and semipermissive (27°C) temperatures. This, however, was not due to an inability to induce a UPR, since dithiothreitol (DTT) strongly induced lacZ in cdc1 (Figure 5C).

These results argue that the cdc1 mutation neither slows nor accelerates the transport of GPI proteins out of the ER.

The reason why emp24∆, per1∆, and gup1∆ restore the growth of cdc1 at 30°C may be that all these mutants reportedly accumulate GPI proteins in the ER (Fujita and Jigami, 2008; Castillon ; Figure S3B). This would prolong the time span during which newly synthesized GPI proteins can be acted on by the residual phosphodiesterase activity of Cdc1-314 in the ER, thereby increasing the percentage of correctly processed anchors. It may also be, however, that the strong UPR induced by deletions such as emp24∆, per1∆, and gup1∆ (Jonikas ) would help cdc1 cells to overcome its late secretion block (see Figure 7A later in this article). Indeed, the similar sec1 block is partially cured by a strong UPR (Chang ). On the other hand, subsistence of the acyl on the inositol moiety in bst1∆ may impede the interaction of the GPI anchor with Cdc1, leading to the strongly negative interaction of bst1∆ and cdc1 (Table S1).

FIGURE 7:

Sorbitol normalizes surface transport of GPI proteins and actin depolarization in cdc1 cells. WT and cdc1 cells expressing Gas1-GFP (A) or Crh2-GFP (B) under their own promoters and present on centromeric plasmids were grown overnight at the indicated temperatures without or with 1 M sorbitol (+ 1MS). Cells were viewed under a fluorescence microscope. Magnification is the same in all pictures; scale bars: 5 μm. (C) actin patches and cables were stained with rhodamine phalloidin in cells grown overnight under the indicated conditions. (D and E) Actin polarization in cells grown overnight at 33 (D) or 37°C (E) was quantified by measuring the phalloidin fluorescence density in buds and their mothers and then calculating the bud/mother density ratios. Each dot represents a budded cell; bud/mother density ratios are plotted as a function of the bud/mother size ratio and data were subjected to linear regression analysis. Note that these fluorescence density ratios are quite small, because they do not account for the volume but only the area occupied by buds and mothers.

Does cdc1 affect remodeling of GPI proteins?

In view of the several GPI anchor remodelases that showed either strongly negative (bst1∆) or strongly positive (per1∆, gup1∆) genetic interactions with cdc1 (Table S2), we wondered whether GPI remodeling was perturbed in cdc1. As shown in Figure 5D, detergent-extractable GPI proteins of cdc1 cells contain a normal amount of ceramide-based, inositol phosphorylceramide (IPC)/B-type anchors. In contrast, cdc1 cells made three- to fourfold more pG1-type anchors than WT. pG1 is a mild base–sensitive, diacylglycerol-type anchor lipid having C26 in sn2 (Figure 5D). The absolute and relative increase of pG1 raised the possibility that pG1-type GPI anchors are not efficiently transferred from the plasma membrane onto β1,6-glucans, a process, in which the GPI proteins lose the glucosamine–inositol–lipid moiety (Kollar ; Fujii ).

The profile of the free [3H]inositol-labeled membrane lipids of cdc1 was normal and showed a normal ratio of mild base–sensitive PI and base-resistant sphingolipids (Figure 5E). However, cdc1 displayed a consistent relative lack of IPC-C (Figure 5E).

We also labeled cdc1 cells with [14C]serine, a precursor that labels sphingolipids plus all mature glycerophospholipids, except for PI and mitochondrial lipids. As seen in Figure 5F, the profile of [14C]serine-labeled lipids of cdc1 cells was normal, whereby a relative reduction of IPCs was noted also by this approach. Apart from the reduction of IPCs, our data are in agreement with a previous report using metabolic labeling with 32PO4−, which concluded that the major sphingo- and phospholipids of the secretory apparatus are unchanged in cdc1 mutant; still, our method was too crude to detect up-regulation of a minor abnormal phospholipid observed in this previous report (Losev ).

Cdc1 cells have fragile cell walls

A previous report showed that growth at 30°C of several cdc1 strains was rescued by osmotic support, but concluded that CDC1 depletion does not affect cell wall integrity, because 1) cdc1 cells grown with osmotic support at 30°C were not lysed when put into hypotonic media, 2) osmotic support did not rescue growth of cdc1 at 36°C or of cdc1∆, and 3) there were no genetic interactions between cdc1 alleles and genes in the Pkc1-regulated mitogen-activated protein kinase (MAPK) cell integrity pathway (Paidhungat and Garrett, 1998b). The cdc1-314 allele studied here showed the same properties. When kept in YPD in 1 M or 1.4 M sorbitol at 37°C for 24 h, it remained viable (Figure S9) and did not lyse when resuspended thereafter in water (unpublished data); also, our genetic screen failed to show genetic interactions of cdc1 with the cell wall integrity (CWI) pathway. Yet, as mutants in genes required for cell wall biosynthesis were highly enriched among the negative interactors of cdc1 (Figure 3, D and E, and Table S2), we nevertheless investigated the susceptibility of cdc1 to various cell wall–perturbing agents.

As seen in Figure 6A, at 27°C, at which they still grow almost as well as WT, cdc1 cells were extremely sensitive to very low concentrations of SDS, the N-glycosylation inhibitor tunicamycin, calcofluor white (CFW), and the β1,3-glucan synthase inhibitor caspofungin. Sensitivity to these agents is characteristic for gene mutations affecting cell wall biosynthesis. The extreme tunicamycin hypersensitivity of cdc1 cannot be due to an inability to induce the UPR (Figure 5C) but may rather be due to the reduced export and functionality of GPI-CWPs and other surface mannoproteins under tunicamycin, an effect that may enhance the cell wall defect of cdc1. It should be noted that bst1∆ and per1∆ single mutants, known to also have cell wall defects, showed no hypersensitivity to the very low concentrations of cell wall–stressing reagents used in Figure 6A. However, as expected from SGA interactions, cdc1/bst1∆ were barely growing on YPD (but fully rescued by sorbitol), whereas cdc1/per1∆ grew much better than cdc1 on the cell wall–perturbing agents (Figure 6A), in agreement with the reported rescue of cdc1-1 cells at 30°C by per1∆ (= cos16∆) (Paidhungat and Garrett, 1998a) and the data in Figure S1.

As previously discussed, the hypersensitivities to cell wall–perturbing agents are phenotypically suppressed in the mcd4∆ background, with the possible exception of the CFW hypersensitivity (Figure 6A, rows mcd4∆ and mcd4∆/cdc1).

As shown in Figure 6B, growth of cdc1 cells was normalized by 1 M or 1.4 M sorbitol up to 30 and 35°C, respectively, but not at higher temperatures, and all attempts to generate cdc1∆/per1∆, cdc1∆/gup1∆, cdc1∆/hrd3∆, cdc1∆/svp26∆, or cdc1∆/emp24∆ mutants by dissecting tetrads from the corresponding diploid heterozygous double mutants on sorbitol plates failed (unpublished data). The results also suggest that the Cdc1-314 protein is at least partially functional at 35°C, but we cannot be sure that its activity is preserved at 37°C, at which temperature cdc1 cells, even when having suppressor mutations, stop growing, although the results shown in Figure 4B suggest that cells at 37°C may still have substantial amounts of Cdc1-314 protein.

As shown in Figure 6C, the cell wall defect of cdc1 cells was discernible even on 1.4 M sorbitol at 32°C when cells were exposed to SDS or tunicamycin. At the same time, they were not hypersensitive to CFW or caspofungin. Thus, although cdc1 cells maintain at high temperatures a cell wall structure that protects them from subsequent lysis in hypotonic conditions, they are hypersensitive to SDS and tunicamycin, indicating some sort of cell wall abnormality that does not seem to be related to chitin or β1,3-glucan biosynthesis.

Sensitivity of cdc1 to CFW can be explained by the fact that absolute CFW fluorescence in cdc1 was much stronger and appeared over the entire cell surface, whereas in WT cells, CFW stained mainly the bud neck and birth scars, as expected (Figure 6D). Occasionally CFW fluorescence was seen at the tip of small buds (Figure 6D, inset), as was also reported for in dcw1∆ mutant presumed to be deficient in the transglycosidase anchoring GPI-CWPs in the cell wall (Kitagaki ). Lateral chitin depositions are a typical compensation mechanism in mutants having defects in glucan biosynthesis (Popolo ; Ram ).

On the basis of the extreme hypersensitivity of cdc1 cells to cell wall–perturbing agents, we propose that cdc1 cells die at higher temperatures because of a severe cell wall deficiency. This also is supported by the fact that the CWI pathway is activated in cdc1 cells (Losev ). The structure of cell walls in cdc1 may, however, be quite different from the one in pkc1 or rho1 mutants, because only the latter two lyse in hypotonic media after having been on osmotic support (Paidhungat and Garrett, 1998b).

Sorbitol normalizes surface transport of GPI proteins and actin depolarization in cdc1 cells

The strong negative genetic interaction of cdc1 with numerous genes required for cell wall biosynthesis (Figure 3, D and E), as well as its hypersensitivity to cell wall–perturbing agents, raised the possibility that cdc1 cells may have a deficiency in the transport of CWPs later in the secretory apparatus, that is, in the Golgi and beyond. As seen in Figure 7, A and B, when cdc1 are incubated at 30 or 37°C, the GPI proteins Gas1 and Crh2 accumulate in internal punctate structures. This accumulation is already visible at 27°C (unpublished data). The same punctate structures were also observed for Cwp1 and Cwp2 (Figure S4). When sorbitol was added to media, the intracellular accumulation of GPI proteins was completely relieved (Figures 7A and S4). It appeared conceivable that the intracellular accumulation of GPI proteins in cdc1 mutants may be related to the loss of actin polarization and actin filaments (Losev ), because actin mutants or actin polymerization inhibitor cause a partial block in the late secretory pathway (Novick and Botstein, 1985; Karpova ). Similarly, defects in COF1, an actin-severing protein, lead to retention of secretory cargo (Bgl2) in the trans-Golgi network (Curwin ). We therefore tested whether sorbitol can correct actin defects of cdc1. As shown in Figure 7, C–E, 1 M sorbitol indeed fully corrected the actin polarization defect at 33°C and did so partially at 37°C.

These data indicate that 1 M sorbitol can rescue the secretion and actin polarization defect. They suggest that the secretion defect of cdc1 cells may be caused by the perturbation of the actin cytoskeleton (Novick and Botstein, 1985; Karpova ), but other data also raise the possibility that the secretion defect is upstream of actin depolarization (Aronov and Gerst, 2004).

Incidentally, the data also confirmed that most cdc1 cells arrest with small buds when shifted to 33°C, but no longer do so when shifted to 37°C. It has previously been reported that shifts to 30°C arrest cells in G2, but that part of the cells arrest in G1 after a shift to 37°C (Paidhungat and Garrett, 1998b).

Cdc1 establish fewer links between Cwp1 or Cwp2 and cell wall glucans than WT

The fact that sorbitol normalized surface transport of GPI proteins but did not allow growth of cells at 37°C made us believe that GPI proteins of cdc1 cells retaining an EtN-P on Man1 may not reach their proper destination or may not be properly transferred to the cell wall glucans. After mechanical disruption of yeast cells, the highly cross-linked cell wall glucans remain sedimentable by low-speed centrifugation and are dubbed “cell walls.” CWPs are defined as proteins that are not solubilized by boiling cell walls in SDS and 2-mercaptolethanol and remain sedimentable by low-speed centrifugation. Treatment of such SDS-extracted cell walls with linkage-specific glucanases releases proteins such that they no longer sediment with cell walls, enter polyacrylamide gels upon subsequent SDS–PAGE, and can be identified on Western blots (Kapteyn ). Classically, CWPs released by β1,6- or β1,3-glucanase are considered to be covalently attached to β1,6- or β1,3-glucans, respectively, whereby the linkage to β1,3-glucans can be either direct via a PIR domain or indirect via β1,6-glucans, which in turn are linked to β1,3-glucans (de Groot ). This approach was used to demonstrate that most CWPs are GPI-CWPs, that GPI-CWPs are linked to β1,3-glucans via β1,6-glucans (Kapteyn ), and that, in mutants harboring cell wall defects, GPI-CWPs may also be linked (via β1,6-glucans) to chitin (Kapteyn ). Moreover, for Cwp1 having a GPI anchor attachment signal and a PIR domain, it was shown that ∼40% is doubly linked 1) to β1,6-glucan via the GPI anchor and 2) to β1,3-glucan via a mild base–sensitive linkage involving the PIR domain, the latter linkage becoming more prominent in cells growing at low pH (Kapteyn ). This double linkage to both types of β-glucans explains why, in WT cells, part of Cwp1 is neither released by β1,6-glucanase nor base treatment, but only their combination (Kapteyn ). Doubly linked GPI-CWPs also may cross-link different β1,3-glucans to each other.

In our experiments done with cells growing in SC medium at low pH (4.2), Cwp1 of WT cells was also released only by combined treatment of β1,3- and β1,6-glucanases and chitinase (Figure 8A, top panel). In contrast, Cwp1 of cdc1 cells was already released by incubation at 60°C alone, and its release was increased by single digestions with either β1,6- or β1,3-glucanase or chitinase. That higher amounts of Cwp1 are released from cdc1 cell walls than from WT cell walls is due to the well-known induction of CWP1 upon cell wall stress (Ram ; García ; Figure S5). This is apparent also in SDS extracts of cells (Figure 8A, lanes 1–4, top panel).

FIGURE 8:

The transfer of GPI-CWPs onto β1,6-glucans is compromised in cdc1 cells. (A) SDS-treated cell walls (10 OD600 equivalents) from the indicated strains harboring a plasmid with HA-Cwp2 and having been grown in presence or absence of 1 M sorbitol were digested with the specified glycosidases or control incubated, boiled in sample buffer, and centrifuged. Supernatants were loaded and separated in a 4–20% gradient SDS–PAGE. Cwp1 (top) and HA-Cwp2 (bottom) were detected on Western blots. SDS extracts from the same cells, as well as proteins secreted into the media corresponding to 0.5 and 10 OD600 equivalents, respectively, are shown in lanes 1–4 and 33–36. The specificity of the polyclonal anti-Cwp1 antibody is shown in lanes 5, 15, 27, and 32. The antibodies used do not react with the glycosidases used (lanes 14, 20, and 26). (B) Cells of indicated genotypes were metabolically labeled with [3H]myo-inositol, mechanically disrupted, and divided into two equal parts. One part was subjected to extensive delipidation using organic solvents, proteins were extracted by repeated boiling in SDS, and nonsoluble material was removed by centrifugation (SDS extractable); in the other part, cell walls were prepared by repeated SDS extraction and were further delipidated with organic solvent (cell wall associated). In both parts, the remaining, GPI protein–associated radioactivity was detected by scintillation counting. (C) Same as in B but cells were grown overnight and labeled in media containing 1 M sorbitol. All strains in B and C, except for cwp1∆, had the Y8205 genetic background. (D) Same as B. Average and SD of three independent experiments and nine measurements are shown for B, and two independent experiments and six measurements are shown for C and D.

Cdc1 cells also secreted part of Cwp1 into the culture medium (Figure 8A, lanes 33–36). Secretion of Cwp1 or other CWPs is often observed in mutants having a cell wall defect, such as gas1∆, gup1∆, per1∆, gpi7∆, and dcw1∆ (Ram ; Kitagaki ; Richard ; Bosson ; Fujita ). As shown in Figure S6, the GPI protein Gas1 was also secreted by cdc1.

The classical interpretation of the data is to say that most Cwp1 of cdc1 cells is linked only to either β1,3- or β1,6-glucan or neither, but not both, as is seen in WT cells. Interestingly, the mobility of Cwp1 released from cdc1 cells by the combination of all glycosidases was slightly lower than the one released from WT cells (Figure 8A, lanes 28–31). This suggested to us that Cwp1 of cdc1 may have a structure different from the one released from WT.

We also investigated Cwp2, another potentially doubly linked GPI-CWP. Cdc1cells expressed less HA-Cwp2 than WT (Figure 8A, lanes 1–4, bottom, and Figure S5). When released from WT cell walls, HA-Cwp2 appeared as a high-MW smear after β1,3- and as several lower-MW bands (75–200 kDa) after β1,6-glucanase treatment (Figure 8A, lanes 10, 12, 16, and 18). The same lower-MW bands were also observed after combined treatment with all glycosidases (Figure 8A, lanes 28 and 30), although the SDS-soluble HA-Cwp2 runs as a doublet with much lower molecular mass (Figure 8A, lanes 1–4), as described previously (van der Vaart ). We assume that the HA-Cwp2 bands running with apparent MWs of 75–200 kDa (Figure 8A, lanes 16, 18, 28, and 30) represent HA-Cwp2 forms with incompletely digested, covalently linked β1,3-glucans, as often observed after β1,3-glucanase (Quantazyme) treatment (Kapteyn ). Similarly, the HA-Cwp2 released with β1,6-glucanase may indicate residual attachment to β1,3-glucan. Interestingly, similar to Cwp1, a substantial fraction of HA-Cwp2 was released as a low-MW form from cell walls of cdc1 cells that had already undergone single digestions with either β1,6-, β1,3-glucanase or chitinase, although combination of all three glycosidases enhanced the recovery (Figure 8A, lanes 11, 13, 17, 19, 23, and 25 vs. lanes 29 and 31). Similar to Cwp1, the smallest form of Cwp2 released from cdc1 had a slightly different mobility on SDS–PAGE than the one released from WT cell walls, suggesting that these two forms have structurally different C-termini. Part of the HA-Cwp2 released by glucanases from cdc1 was also running in the 75–200 kDa region (Figure 8A, lanes 11, 13, 17, 19, 29, and 31); this possibly represents HA-Cwp2 that was made before the temperature shift, when cells were growing at 24°C. We therefore used β1,3- and β1,6-glucanases to treat cell walls from cdc1 cells that were metabolically labeled with [35S]cysteine and [35S]methionine at 37°C. As can be seen in Figure S7, we also found in these experiments that proteins made under restrictive conditions were integrated into the cell wall (in spite of the partial secretion block) and that they could be released through single treatments with β1,6-glucanase.

Proteins may continue to be integrated into the cell wall even under restrictive conditions, either because Cdc1 is not required for this integration process or because the remaining activity of Cdc1-314 at 37°C is sufficient for mere cell wall integration.

Overall the classical interpretation of the data shown in Figure 8A would argue that, in cdc1 cells, GPI-CWPs continue to be attached via their GPI anchors to β1,6- and β1,3-glucans but that these linkages were established less frequently, as larger than normal proportions of Cwp1 and Cwp2 were released by treatment with only one enzyme and Cwp2 was recovered as a low-MW form devoid of remaining glucans.

The transfer of GPI-CWPs onto β1,6-glucans is compromised in cdc1 cells

It seemed possible that the anomalous subsistence of an EtN-P on Man1 in cdc1 (Figure 1) might interfere with the efficient transfer of GPI-CWPs onto β1,6-glucans. In support of this, we found a strongly negative genetic interaction between cdc1 and dfg5∆, Dfg5 being one of the supposed transglycosidases that operate this transfer (Kitagaki , 2004).

While the experiments in Figures 8A and S7 suggested that cdc1 cells continue to covalently add GPI-CWPs to β1,6-glucans, because such proteins can be released from the cell wall by β1,6-glucanase, we considered an alternative interpretation of these results. It seemed conceivable that the dense meshwork of glucans and chitin, possibly rendered thicker and more dense by the cell wall response, would trap GPI proteins still possessing their GPI anchor such that they cannot be released by boiling in SDS and 2-mercaptoethanol, that is, by the classical procedure of preparing cell walls (de Groot ). This hypothesis will be referred to as the “trapping model” in the rest of this paper.

We thus sought to test the possibility that proteins not linked covalently to cell walls would be released by β1,6-glucanase simply because this enzyme would render the meshwork of glucans less dense. We tested this by monitoring whether GPI-CWPs remaining in the cell wall fraction after SDS extraction still contain inositol. According to current understanding, this ought not be the case, because the inositol moiety normally is lost when GPI-CWPs are transferred to β1,6-glucans (Figure 1).

As shown in Figure 8B, SDS-extracted GPI proteins of all strains tested contained similar amounts of [3H]inositol, arguing that the incorporation of [3H]inositol into the fraction of free GPI proteins was similar among strains. The same was true for the incorporation of [3H]inositol into lipids (Figure S8). Surprisingly, cell walls of WT cells contained significant amounts of radioactivity (Figure 8B). Moreover, cdc1-314 mutants contained significantly more cell wall–associated radioactivity than the cell walls of WT, dcw1∆, or dfg5∆ mutants. Radioactivity was further increased in cell walls of cdc1/dfg5∆ but not of cdc1/dcw1∆ cells (Figure 8B).

While the presence of [3H]inositol associated with the SDS-extracted cell walls supports the trapping model, other explanations have to be considered. We can exclude that the cell wall–associated radioactivity is due to the presence of residual free lipids, because all cell wall preparations shown in Figure 8, B– D, after having been extracted four times by boiling in SDS plus 2-mercaptoethanol, were further extracted with chloroform–methanol–water (10:10:3), and these organic solvent extracts did not contain any radioactivity (unpublished data).

We also considered the possibility that radioactivity may be associated with the fraction of Cwp1 that can be released from the cell walls by alkali treatment alone (30% of total; Kapteyn ). This fraction of Cwp1 and possibly of three further doubly linkable GPI proteins (Klis ), linked only to β1,3-glucans, would retain its [3H]inositol moiety and could therefore account for the radioactivity in cell wall fractions of WT cells. The increase of cell wall–associated radioactivity in cdc1 cells could thus be the consequence of a cell wall response that induces expression of GPI-CWPs, including doubly linkable GPI-CWPs. However, the absolute amount of cell wall–associated [3H]inositol-containing GPI-CWPs of cdc1 and cdc1/dfg5∆ is comparable to or exceeds the total of SDS-extractable GPI proteins, whereas in WT, protein-associated [3H]inositol in the cell wall fraction is much lower than in the SDS extract. As it is estimated that about half of all GPI proteins remain associated with the plasma membrane (Caro ; Hamada ; de Groot ), the sheer relative amount of [3H]inositol-labeled CWPs in cdc1 and cdc1/dfg5∆ cells makes it rather unlikely that this material would consist merely of the singly linked fraction of doubly linkable GPI-CWPs.

To get further support for the trapping model, we also measured the cell wall–associated [3H]inositol in cell wall mutants such as gas1∆ and cwp1∆ cells. Gas1∆ is extremely CFW hypersensitive, and its cell wall contains 5.3-fold more chitin than WT (Kapteyn ), whereas cwp1∆ cells are only moderately CFW hypersensitive (van der Vaart ). In gas1∆ cells the cell walls contain drastically increased amounts of [3H]inositol-labeled GPI-CWPs (Figure 8D), whereas in cwp1∆ cells they contain normal amounts (Figure 8B). We also tested the [3H]inositol-labeled GPI-CWPs in cell walls of dcw1∆/dfg5∆ cells rescued by either dfg5-29 or DFG5 (Kitagaki ). Mutants rescued by dfg5-29 contained approximately twofold increased levels of [3H]inositol-labeled GPI-CWPs compared with those rescued by DFG5, and both strains had higher amounts of labeled proteins in their cell walls than WT (Figure 8D vs. 8B).

Astonishingly, addition of sorbitol to the media totally abolishes the association of [3H]inositol-labeled GPI proteins with the cell wall in mutant and also WT cells (Figure 8C), although the experiments in Figure 8A show that Cwp1 and Cwp2 are still associated with the cell walls of cdc1 when cells are grown in sorbitol (lanes 13, 19, 25, and 31) and that they can be released by glucanase treatment as low-MW forms, suggesting that they were not connected to residual β-glucans. To explain this paradox, our current working model comprises yet a further type of trapping that would not be relieved by sorbitol and concerns to-be GPI-CWPs having already lost their lipid moiety but being not yet connected to the cell wall glucans. Indeed, it seems conceivable that a first enzyme cleaves the glucosamineα1, 6inositol bond of the GPI anchor (Figure 1) and that the transglycosidation reaction cleaving the Manα1,4glucosamine linkage occurs in a separate, later step. In this scenario, cdc1 cells may be able to cleave the glucosamineα1,6inositol bond, but not the Manα1,4glucosamine bond, thus leading to a trapping/retention in the cell wall of GPI-CWPs making no covalent bonds to β-glucans but devoid of their lipid anchor.

DISCUSSION

No biochemical defect has so far been identified in cdc1 cell division–cycle mutants, and it was therefore difficult to make out cause and consequence among the numerous cell biological abnormalities that arise after a temperature upshift in cdc1 mutants. The possible biochemical defect was pinpointed by the groundbreaking study by the Taroh Kinoshita group on PGAP5 (Fujita ). Our data indicate that it is likely that the homologous Cdc1 has a function similar to that of PGAP5 and removes EtN-P from Man1 of the GPI anchor, whereas the literature suggests that Ted1 may remove EtN-P from Man2, as does PGAP5. Direct biochemical assays are required to test this hypothesis.

Three main explanations for the strong cell wall phenotype of cdc1 can be envisaged. Either cdc1 cells do not transfer GPI-CWPs to the cell wall glucans, because the persistence of EtN-P on Man1 renders the interaction of these proteins with the enzymes operating that transfer less efficient. Indeed, EtN-P is attached to the C2 atom of Man1 (Homans ) in the vicinity of the C1 engaged in the glycosidic bond that has to be cleaved in order to transfer Man1 onto β1,6-glucans (Kollar ; Fujii ). Second, the two presumed transferases Dfg5 and Dcw1 as well as Gas1, all being GPI proteins themselves, cannot reach the cell surface because of the secretion block that affects the surface transport of GPI proteins in cdc1 (Figure 7A). From the metabolic labeling studies with [32S]methionine, however, we know that this secretion block is only partial and that as much protein gets into the cell wall of cdc1 kept at 37°C as in WT (Figure S7). Moreover, cells were grown overnight at 24°C before being labeled at 37°C, and some Dcw1 and Dfg5 made at 24°C ought to still be present at the plasma membrane during the labeling. Thus this second hypothesis seems less likely. Third, the persistence of EtN-P on Man1 leads to mistargeting of GPI-CWPs to inappropriate locations of the cell wall or plasma membrane and thus causes lethality. Also in this scenario, the mislocalization of GPI-CWPs in cdc1 and the ensuing overproduction of chitin and β-glucans may trap/immobilize Dfg5p, Dcw1, and their substrates, that is, the to-be GPI-CWPs, in a way that they cannot interact. However, the very severe trapping of GPI proteins seen in the perfectly viable gas1∆ mutant would argue that at least some types of trapping are not lethal. Of course, it is possible that cell wall–associated [3H]inositol is elevated for different reasons in different mutants, for example, for hypothesis 1 in cdc1 and for hypothesis 3 in gas1∆.

Discrimination between hypothesis 1 and 3 will require a biochemical in vitro assay of Dcw1 and Dfg5 activities, for which preliminary assays so far have not shown positive results (Hiroshi Kitagaki, personal communication). Whatever the exact reason for the cell wall phenotype of cdc1, it currently is not clear whether the failure to remove EtN-P from Man1 concerns all GPI proteins, or only GPI-CWPs, or only a subgroup of them. The major increase in SDS-extractable GPI anchors having the pG1 lipid moiety (Figure 5D) may be an indication that Cdc1 primarily acts on pG1-type GPI anchors. It also could be envisaged that Cdc1 primes certain GPI proteins for transfer into the cell wall by removing the EtN-P from Man1. However, Cdc1 probably acts on more than one GPI protein, because none of the 64 predicted GPI proteins of Saccharomyces cerevisiae is essential, but we cannot exclude a dominant effect through which a single unprocessed and mislocalized GPI protein could kill cdc1 cells.

While Cdc1 may act on only a fraction of GPI proteins, it is clear that all GPI anchors of WT cells initially contain an EtN-P on Man1, because the Gpi10 mannosyltransferase 3 requires this element, as mentioned previously, and the only way to produce GPI anchors lacking EtN-P on Man1 is by way of Cdc1.

Interestingly, the lack of EtN-P on Man2 has been invoked as a potential cause for the mistargeting of Egt2, a daughter cell–specific presumed glucanase that is evenly distributed over the bud surface in gpi7∆ cells and digests their cell wall, whereas it is directed to the primary septum in WT cells (Fujita ). Thus some GPI proteins such as Egt2 may not be acted upon by Ted1, but the inhibitory ­effect of the gpi7∆ deletion on Cwh43-mediated GPI remodeling may also be responsible for this particular sorting defect (Benachour ).

Our proposal that cdc1 cells stop growing at 37°C because of a severe cell wall defect does not exclude the possibility that the growth arrest of cdc1 at 33°C is due to other reasons, for example, the Ca2+ influx, loss of actin filaments, actin polarization, Golgi inheritance (Losev ), and/or the partial secretion defect (Figures 7, A and B, and S3). Yet, on sorbitol-containing media, the actin polarization and secretion defects disappear, whereas growth does not resume. Thus severe CDC1 deficiency leads to a growth defect that cannot be attributed to the actin depolarization and ensuing phenotypes and may be due to the cell wall defect.

Dcw1∆ dfg5∆ double mutants rescued by a temperature-sensitive allele of either gene arrest after temperature upshift, having small buds, duplicated DNA, nonseparated spindle pole bodies, and depolarized actin cytoskeleton (Kitagaki ), a phenotype closely related to cdc1 mutants (Byers and Goetsch, 1974; Paidhungat and Garrett, 1998b; Losev ). Moreover, dfg5Δ/dcw1-3ts and dcw1Δ/dfg5-29ts cells could grow at 37°C in 1 M sorbitol, and CFW staining indicated atypical chitin deposition at the tips of small buds (Kitagaki ). This similarity argues that cdc1 mutants may be deficient in the same process that occurs in dfg5/dcw1 mutants, or one that is closely related.

Only dfg5∆, not dcw1∆, shows a negative genetic interaction with cdc1, and amounts of [3H]inositol-labeled GPI proteins in cell wall fractions phenocopies the genetic interactions (Figure 8B). One possible reason is that only DFG5 is induced under cell wall stress (Yoshimoto ; Hagen ), whereas a similar induction of DCW1 has not been reported. Alternatively, it may be that GPI proteins having an EtN-P on Man1 are processed only by Dfg5, not Dcw1. Also, the maximal transcription of DFG5 occurs in G1, while the maximal transcription of DCW1 occurs in S phase (Kitagaki ), and the precise cellular localization of both is currently unknown. It should be noted that some CWPs also tend to be transcribed at different moments in the cell cycle and to locate to different parts of the cell wall (Klis ).

The data in Figure 8 led us to propose the somewhat heretical trapping model, which may be especially applicable to mutants having mounted a cell wall response, whereby the trapping of inositol-containing GPI-CWPs is abolished when cells receive osmotic support. A mass spectrometric analysis of yeast CWPs released by mild base or hydrofluoric acid (to cleave the phosphodiester linkage of β1.6-glucan–linked GPI-CWPs) identified 19 CWPs, three of which had neither a GPI attachment signal nor a PIR consensus repeat (Q[IV]XDGQ[IVP]Q; Yin ). This supports the idea that, even in WT cells, some CWPs may simply be trapped or be anchored in another yet unknown manner.

MATERIALS AND METHODS

Yeast strains, media, and reagents

Materials were from sources described recently (Ramachandra and Conzelmann, 2013). [3H]myo-inositol and [14C]serine were from ARC Radiochemicals (St. Louis, MO); rhodamine phalloidin and Prolong Gold antifade reagent were from Invitrogen. Endo-β1,3-glucanase 81A and endochitinase 18A of Clostridium thermocellum were from NZYThec (Lisbon, Portugal); endo-β1,6-glucanase (ThermoActive Pustulanase-Cel 136) was from Prokazyme. Strains, plasmids, and primers used are listed in Supplemental Tables S3–S5. Cells were grown at 30°C on YPD supplemented with uracil and adenine or synthetic complete (SC) medium made from YNB (yeast nitrogen base; United States Biological), supplemented also with amino acids and containing 2% glucose.

Construction of strains and plasmids

Standard methods were used to construct strains for SGA and plasmids with HA- or green fluorescent protein (GFP)-tagged proteins. Detailed procedures are reported in the Supplemental Material.

Protein extraction and quantification

For experiments shown in Figures 4, 5B, S3, and S5, cells were put on ice after addition of 10 mM NaN3 and NaF. Proteins were extracted (Kushnirov, 2000), separated by SDS–PAGE, and processed for Western blotting, and the immunodetected protein bands were quantified using Image Studio Lite software from LI-COR (Lincoln, NE). For quantifications of Figure 4, the anti-HA signal of each lane was first normalized using Adh1 to correct small errors of protein loading during SDS–PAGE. All gels for the experiments in a given panel (A or B) contained a reference sample that was used to normalize and compare the values between different gels. The doubly normalized data were then used to get a mean and SD for each condition.

Measuring β-galactosidase activity

Assays were done as described by Guarente (1983) but with the modifications described in the Supplemental Material.

Pulse-chase metabolic labeling of proteins using [35S]methionine/cysteine

Pulse-chase experiments were done as previously described (Watanabe ) with the modifications described in the Supplemental Material.

Metabolic labeling of cells with [3H]myo-inositol or [14C]serine

For labeling with [3H]myo-inositol (Figures 5, D and E, and 8, B– D), cells were grown to exponential phase at 24°C in inositol-free SC medium containing or not 1 M sorbitol. Ten OD600 were collected, resuspended in 1 ml fresh medium of the same kind, and preincubated at 37°C for 10 or 60 min in a water bath with good aeration (shaking). At t = 0, 40 μCi of [3H]myo-inositol was added. Fresh medium (0.2 ml) from a 5× concentrated stock was added at t = 0, 40, and 80 min if cells had been preincubated for 60 min, and at t = 40 and 80 min if they had been preincubated for 10 min. Labeling was stopped after 2 h. GPI anchor lipids and free lipids (Figure 5, D and E) were isolated and analyzed as described by Guillas . Free lipids were migrated on TLC silica plates in chloroform–methanol–water (10:10:2.5), and anchor lipids were migrated in chloroform–methanol–0.25% KCl (55:45:10) solvent systems. GPI proteins (Figure 8, B– D) and lipids (Figure S8) to be used for scintillation counting were extracted from labeled cells as described for analysis of anchor lipids (Guillas ), that is, by breaking cells with glass beads in organic solvent, but delipidated protein pellets were incubated with ethanol–water–diethyl ether–pyridine–ammonia (15:15:5:1:0.018 vol/vol) at 60°C for 15 min, dried down, and further extracted with chloroform–methanol–water (10:10:2.5) to remove residual lipids. Final protein pellets containing labeled GPI proteins were solubilized by boiling them twice for 10 min in counting buffer (2% SDS, 5% 2-mercaptoethanol) and counted in a scintillation counter.

For labeling with [14C]serine (Figure 5F), cells were grown to exponential phase at 24°C in YPD. Three OD600 were collected, washed, and resuspended in 3 ml serine-free SC medium. Cells were then labeled with 3 μCi of [14C]serine in a shaking water bath at 30°C for 5 h. Thereafter equivalent numbers of OD600 units of cells were collected from all strains, and lipids were extracted as described by Guan but no standards were used. Lipids were separated on TLC silica plates in chloroform–methanol–4.2 N ammonia (9:7:2) solvent system.

Mild base treatment of lipids

Dried lipids were resuspended in 200 μl chloroform–methanol–water (10:10:3), 40 μl of 0.6 M NaOH in methanol was added, and samples were incubated at 37°C for 1 h. Hydrolysis was stopped with 40 μl of 0.8 M acetic acid, and the lipids were dried. Control samples followed the same procedure, but NaOH and acetic acid were added together at the end of the incubation. Dried lipids were desalted by butanol–water partitioning as described by Guillas .

Fluorescence microscopy

Fluorescence microscopy was done using an Olympus BX51 microscope. Detailed procedures are contained in the Supplemental Material.

Cell wall isolation and protein extraction

Cell walls (Figures 8 and S7) were isolated as described by de Groot but with the following modifications: cells were lysed with glass beads in 50 mM Tris-HCl (pH 7.5) containing 1 mM EDTA, 1 μM pepstatin, 1 mM phenylmethylsulfonyl fluoride, and Roche protease inhibitor cocktail. After lysis, SDS, NaCl, and β-mercaptoethanol were added to final concentrations of 4%, 100 mM, and 40 mM, respectively, and lysates were boiled for 10 min. Boiled lysates were centrifuged (16,000 × g, 5 min) and the pellets were extracted three more times by each time adding fresh extraction buffer (4% SDS, 100 mM NaCl and 40 mM β-mercaptoethanol), boiling for 2 min, and centrifuging. After having been extracted four times, the resulting cell walls were washed three times with water and used directly for glycosidase digestions. Cell walls corresponding to 50 OD600 units of cells were digested in 100 μl of 100 mM sodium phosphate (pH 6) with 0.2 U/μl of endo-β1,3-glucanase 81A, 0.05 U/μl of endo-β1,6-glucanase, and/or 0.0025 U/μl of endochitinase 18A. The buffer of the glycosidases was changed to 100 mM sodium phosphate (pH 6) using Vivaspin columns with a 10-kDa cutoff before being added to cell walls. All samples were first incubated for 5 h at 60°C, which is the optimal temperature for β1,3-glucanase and chitinase, then 5 h at 80°C, the optimal temperature for β1,6-glucanase. After the incubations, SDS sample buffer (60 mM Tris-HCl, pH 6.8, 3.3% SDS, 5% glycerol, 100 mM DTT, and 3 mM bromphenol blue) was added and samples were boiled for 10 min and centrifuged 5 min at 16,000 × g. Extracted proteins were resolved in 4–20% gradient SDS–PAGE and transferred to PVDF membranes. HA-Cwp2 was detected using anti-HA antibodies, Cwp1 and Gas1 were detected using specific antibodies against these proteins.

For detection of radiolabeled GPI proteins associated with the cell wall, SDS-treated cell walls (see preceding paragraph) from cells labeled with [3H]myo-inositol were washed three times with water and further delipidated two times with 500 μl chloroform–methanol–water (10:10:3), resuspended in counting buffer (2% SDS, 5% 2-mercaptoethanol), and boiled for 10 min before scintillation counting.

Precipitation of proteins in culture media

Culture media corresponding to 30 OD600 were concentrated down to 1.2 ml using Vivaspin columns with a cutoff of 10 kDa. Five micrograms of bovine serum albumin was added, and proteins were precipitated by adding 300 μl 100% trichloroacetic acid (20% final concentration) and incubated for 40 min at −20°C. Pellets were rinsed twice with ethanol/Tris-HCl,pH 8.8, dried, resuspended in SDS sample buffer, and boiled for 10 min.