The novel fission yeast (1,3)β-D-glucan synthase catalytic subunit Bgs4p is essential during both cytokinesis and polarized growth

The fission yeast Schizosaccharomyces pombe provides an ideal model to study cell morphogenesis (Brunner and Nurse, 2000). The cells are cylindrical and grow from their ends in a polarized fashion (Chang and Peter, 2003; Hayles and Nurse, 2001). Genetic studies have identified many genes important for the different steps of cytokinesis, a process similar to that in animal cells, such as the positioning and assembly of the actomyosin ring, the recruitment of actin patches to the medial ring and the activation of ring contraction by a network of regulatory proteins referred to as the septation initiation network (SIN) (Balasubramanian et al., 2004; Guertin et al., 2002; Krapp et al., 2004). Centripetal ring contraction and septum synthesis are perfectly coordinated, resulting in a three-layered septum structure between the membranes of both daughter cells that is composed of a middle disk structure called the primary septum, analogous to that of budding yeast (Schmidt et al., 2002), and flanked at both sides by the secondary septum. Once the septum has formed, cell division proceeds by dissolution of the primary septum (Martin-Cuadrado et al., 2003).

The S. pombe cell wall is a well-organized structure consisting of an outer layer enriched in glycoproteins and an inner layer of carbohydrates, including (1,3)β-D-, (1,6)β-D- and (1,3)α-D-glucans (for a review, see Duran and Perez, 2004; Perez and Ribas, 2004). (1,3)β-D-glucan is the major contributor to the cell wall framework. In situ localization studies (Humbel et al., 2001) have indicated that (1,6)β-D-branched (1,3)β-D-glucan is localized, forming filamentous structures all over the cell wall non-dense layer and throughout the septum; (1,6)β-D-glucan appears in the same cell wall non-dense layer and in the secondary septum, whereas linear (1,3)β-D-glucan is only present at the primary septum.

The enzyme complex involved in (1,3)β-D-glucan synthesis in yeast is (1,3)β-D-glucan synthase (GS). It is formed by at least two components: a regulatory and a catalytic subunit (Kang and Cabib, 1986; Ribas et al., 1991). The regulatory subunit is the GTPase Rho1p, which not only activates the GS (Arellano et al., 1996; Drgonova et al., 1996) but is also involved in other morphogenetic processes, such as organization of the actin cytoskeleton and genesis of the growing regions (Arellano et al., 1999a; Cabib et al., 2001; Drgonova et al., 1999).

S. pombe contains four putative GS catalytic subunits named Bgs1p-4p. bgs1⁺, bgs2⁺ and bgs3⁺ have been reported to be essential genes, at least for ascospore revival (Cortes et al., 2002; Liu et al., 2000; Liu et al., 1999; Martin et al., 2003; Martin et al., 2000), but nothing is known about bgs4⁺. The first gene to be identified was bgs1⁺ and this was initially named cps1⁺. It was cloned by complementation of several mutants unable to assemble the division septum, and it was suggested to be involved in cytokinesis, polarity and in a septation checkpoint that coordinates cytokinesis and the cell cycle (Ishiguro et al., 1997; Le Goff et al., 1999; Liu et al., 1999). Recently, Bgs1p has been localized to the contractile ring and septum in a manner dependent on the actomyosin ring and SIN (Cortes et al., 2002; Liu et al., 2002), as well as to the growing poles and to other regions of cell wall synthesis during the sexual phase of the life cycle (Cortes et al., 2002). However, neither GS activity nor cell wall (1,3)β-D-glucan synthesis has been shown to depend on Bgs1p function. Therefore, the involvement of Bgs1p in (1,3)β-D-glucan synthesis remains to be clarified. bgs2⁺ is not expressed in vegetative cells but is essential for proper spore wall maturation (Liu et al., 2000; Martin et al., 2000). bgs3⁺ is also essential for vegetative cells and Bgs3p has also been shown to be localized to sites of cell wall growth. Similarly, bgs3⁺ shut-off does not produce any apparent decrease in GS activity or cell wall (1,3)β-D-glucan, although the overall cell wall structure is altered (Martin et al., 2003). Saccharomyces cerevisiae contain two putative GS catalytic subunits, Fks1p and Fks2p, whose sequences share a high homology with Bgs proteins. FKS1 and FKS2 are redundant genes that are differentially expressed depending on the cycle or growth conditions. Single disruption of either FKS1 or FKS2 produces viable cells, but the double disruption is lethal (Mazur et al., 1995). A third FKS gene appears to be dispensable. Cell wall integrity is regulated by Rho GTPases and protein kinase C families (Arellano et al., 1999b). Alteration of these regulators dramatically affects GS activity, which in turn results in cell lysis. However, although the Bgs proteins code for putative GS catalytic subunits, neither Bgs1p nor Bgs3p elicits such a response, and indeed nothing is known about the GS catalytic subunit associated with this regulation that promotes such effects.

In this work, we characterized Bgs4p as the first Bgs subunit found to be part of the GS catalytic subunit. We show that the mutant cwg1-1, which displays reduced cell wall (1,3)β-D-glucan and GS catalytic subunit activity, is allelic to bgs4⁺. We also show that Bgs4p is crucial for maintaining cell integrity, as its absence promotes cell lysis at the poles and septum. The data concerning bgs4⁺ expression suggest that the GS catalytic subunit is a multiprotein complex. Bgs4p localizes to the growing poles, contractile ring and septum, and to each site of cell wall synthesis during the sexual phase of the life cycle. However, the requirements and timing of Bgs4p for proper positioning during septum formation as well as the localization pattern during spore maturation differ from those of Bgs1p. All the above data suggest that Bgs4p plays a role in each cell wall synthesis process of the S. pombe life cycle by synthesizing a (1,3)β-D-glucan that is essential for preserving cell integrity and reinforce its importance as an essential target for the design of specific antifungal drugs. In addition, the lytic phenotype observed during bgs4⁺ shut-off may be biotechnologically useful as an efficient and controlled cell-free system for heterologous protein expression and purification (Priority patent application ES200401042; April 30, 2004).

The S. pombe strains used in this work are listed in Table 1. bgs4Δ strain 561 containing an integrated GFP-bgs4⁺ copy was made by transforming strain 498 (bgs4Δ pJG33, his3⁺ selection) with StuI-digested pJG53 (leu1⁺ selection, see below), which directs its integration at the StuI site adjacent to bgs4Δ::ura4⁺, at position –1320 of the bgs4⁺ promoter sequence. The GFP-bgs4⁺ and bgs4::ura4⁺ sequences are separated by a pJK148 backbone. In order to eliminate pJG33, to analyze GFP-Bgs4p functionality and to confirm that integration was in the bgs4⁺ promoter sequence, this strain was crossed with strain 285 (Leu^–, Ura^–, His^–) and tetrad analysis was performed. Leu and Ura phenotypes always co-segregated 2+:2–. pJG33 was lost in most of the clones as result of the sporulation process. The GFP-bgs4⁺ bgs4Δ strain obtained displayed a wild-type phenotype under all conditions tested and expressed GFP-Bgs4p at physiological levels from a single integrated GFP-bgs4⁺ gene under the control of its own promoter.

Standard complete yeast growth (YES), selective (EMM) supplemented with the appropriate amino acids (Alfa et al., 1993) and sporulation media (SPA) (Egel, 1984) have been described elsewhere. EMM-N is EMM (0.5% NH₄Cl) from which NH₄Cl has been omitted. Stable sterile diploids for complementation analyses were obtained by mating mat2P-B102 and h^– strains (Egel, 1984). The general procedures for yeast culture and genetic manipulations were carried out as described (Moreno et al., 1991). Latrunculin A (Lat A) was from Sigma and was used at 100 μM final concentration (from a stock solution of 10 mM in DMSO). Cell growth was monitored by measuring the A₆₀₀ of cell cultures in a Coleman Junior II spectrophotometer (OD₆₀₀ 0.15=1×10⁷ cells/ml).

Escherichia coli strains DH5α and DH10B (Life Technologies) were used for routine propagation of plasmids (Sambrook and Russell, 2001) and CJ236 (Bio-Rad) was used for site-directed mutagenesis. Bacterial LB, 2×YT and TB media, supplemented with 100 μg/ml ampicillin or 50 μg/ml kanamycin, when appropriate and standard transformation methods (Sambrook and Russell, 2001) have been described elsewhere.

pJG1 (pAL-bgs4⁺) is the 8.84 kb PstI-NheI DNA fragment containing the bgs4⁺ gene sequence from cosmid clone SPCC1840c (35 kb insert and kanamycin resistance selection, The S. pombe Genome Sequencing Project, Sanger Institute, UK) cloned into the PstI-SpeI sites of S. pombe vector pAL-KS⁺ (S. pombe ars1⁺ and S. cerevisiae LEU2 selection).

pJG4 (pJK-bgs4⁺) is the PstI-NheI bgs4⁺ fragment from SPCC1840c cloned into the PstI-SpeI sites of the integrative plasmid pJK148 (S. pombe leu1⁺ selection). pJG9 is pJG1 with NotI from the multiple cloning site destroyed and with XhoI and PstI sites inserted by site-directed mutagenesis just before the start codon and just after the TAG stop codon of bgs4⁺, respectively. pJG17, pJG18 and pJG19 are pJR2-3XL, pJR2-41XL and pJR2-81XL (S. cerevisiae LEU2 selection and 3X, 41X and 81X versions of thiamine-repressible nmt1⁺ promoter, respectively) (Moreno et al., 2000) containing the XhoI-PstI bgs4⁺ ORF sequence from pJG9, respectively. pJG33 is pJR1-81XH (S. pombe his3⁺ selection) (Moreno et al., 2000) with the XhoI-ApaI bgs4⁺ ORF sequence from pJG18. The resulting S. pombe strains bgs4Δ pJG18, pJG19 and pJG33 displayed a lytic phenotype after repressing the nmt1⁺ promoter in the presence of thiamine.

pJR111 is pSK⁺-ura4⁺ (Moreno et al., 2000) with the 1.45 kb promoter and 1.0 kb terminator sequences of bgs4⁺, amplified by PCR and cloned into ApaI-ClaI (5′ to ura4⁺) and BamHI-XbaI (3′ to ura4⁺), respectively. This plasmid was used for PCR amplification of the bgs4⁺ deletion cassette, as described below.

pJG10 is pJG9 with a 9 bp insertion containing a NotI site just after the ATG initiation codon of the bgs4⁺ ORF. pJG27 (pAL-GFP-bgs4⁺, GFP inserted at amino acid 2) is pJG10 cut with NotI and with the 0.74 kb NotI-NotI GFP fragment from pKS⁺-GFP (GFP2-5, modified GFP sequence with changes S65T, V163A, I167T and S175G) (Fernandez-Abalos et al., 1998) inserted in-frame. GFP-Bgs4p from pJG27 proved to be non-functional as strain bgs4Δ pJG27 was not viable. Similarly, Bgs4p containing a C-terminal GFP fusion was not functional. In order to obtain a functional GFP-Bgs4p fusion, the four most hydrophilic and N-terminal regions, as deduced from hydropathy analysis, were selected. Then, the amino acids predicted to be at the boundaries between the secondary structures were chosen, and the 3xFLAG immunoreactive epitope was inserted in-frame at bases 85, 295, 694 and 970 (amino acids 29, 99, 232 and 324) of the bgs4⁺ sequence from pJG9. Only the plasmids with insertions at bases 85 and 970 expressed a functional 3xFLAG-bgs4⁺ fusion. pJG39 and pJG41 (pAL-GFP-bgs4⁺, GFP at base 85 and 970, respectively) are pJG9 with a 9 bp NotI insertion at bases 85 and 970 of the bg4⁺ ORF and with the GFP sequence inserted in-frame in the NotI site, respectively. Both pJG39 and pJG41 restored the wild-type phenotype to a bgs4Δ strain. pJG44 and pJG53 (pJK-GFP-bgs4⁺, GFP at base 85 and 970, respectively) are the integrative plasmid pJG4 with the PstI-BglII bgs4⁺ sequence replaced by the PstI-BglII GFP-bgs4⁺ sequence from pJG39 and pJG41, respectively. pJG53, but not pJG44, expressed a fully functional fusion protein and was therefore selected to create the GFP-bgs4⁺ bgs4Δ strains for Bgs4p localization studies at the physiological level.

The sequences of native tetrameric DsRed (pDsRed2-1) and of the dimeric HcRed (pHcRed1-1) proteins (Clontech) were used to create an alternative fusion with the bgs4⁺ sequence but none of these fusions resulted in a functional Bgs4p. Next, the sequence of the monomeric mutant form of DsRed protein (mRFP) (Campbell et al., 2002) was used to create a functional RFP-Bgs4p fusion. pRSETB-mRFP1 contains the 678 bp ORF sequence of mRFP1 cloned into BamHI-EcoRI of pRSETB (Invitrogen). This plasmid was modified to insert two 9 bp NotI sites, after the ATG initiation codon and before the TAA stop codon, producing pJG84. This mRFP NotI-NotI sequence can be inserted in the same NotI sites as those used for the GFP sequence. pJG52 is the integrative plasmid pJG4 with a 9 bp NotI site inserted at base 970 of bgs4⁺. pJG87 is pJG52 with NotI from the multiple cloning site destroyed (XbaI-SacII deletion) and with the mRFP sequence inserted in-frame in the NotI site of bgs4⁺ at base 970. pJG101 is pJG87 with the two StuI sites of the mRFP sequence destroyed, without changing the amino acid sequence. The integrated StuI-digested pJG101 plasmid expressed a functional fusion protein in a bgs4Δ strain and was therefore selected for localization studies.

S. pombe transformation was carried out using the lithium acetate procedure (Gietz et al., 1995). Other DNA manipulations were performed essentially as described (Moreno et al., 2000; Sambrook and Russell, 2001).

bgs4⁺ gene deletion was performed in the Leu^– Ura^– diploid strain 315. The 4.1 kb deletion cassette was generated by PCR amplification from pJR111. It eliminates the entire bgs4⁺ coding sequence and contains 1.4 kb of the bgs4⁺ promoter, the 1.76 kb ura4⁺ and 1.0 kb of bgs4⁺ terminator sequences. Correct deletion of the bgs4⁺ ORF was confirmed by PCR analysis of both ends of both bgs4⁺ and bgs4Δ::ura4⁺ sequences, using combinations of oligonucleotides external and internal to the deletion fragment (data not shown). Tetrad analysis of sporulated bgs4⁺/bgs4Δ diploids showed Bgs4p to be essential. Haploid bgs4Δ strains were obtained by random spore analysis of bgs4⁺/bgs4Δ strains transformed with the bgs4⁺-carrying plasmids pJG1, pJG17, pJG18 and pJG19, and selection of Leu⁺ (bgs4⁺ plasmids), Ura⁺ (bgs4Δ), Ade^– (haploidy marker) clones. Strain 498 bgs4Δ pJG33 (his3⁺ selection) was obtained by genetic cross of strains bgs4Δ pJG1 and 285 (Leu^– Ura^– His^–) transformed with pJG33. Additional bgs4Δ strains were obtained by genetic cross of wild-type strain 285 and bgs4Δ pJG33 transformed with the corresponding plasmid.

Cell extracts and GS assay were essentially as described (Cortes et al., 2002). Cell extracts were obtained from early log-phase cells grown in EMM as indicated for each case. Standard GS assay contained 15-25 μg protein of enzyme extract (3-5 mg protein/ml) in a total volume of 40 μl and the reaction was incubated at 30°C for 60-90 minutes. All reactions were carried out in duplicate and the values were calculated from three independent cell cultures.

[¹⁴C]glucose labeling and fractionation of cell wall polysaccharides were carried out essentially as described (Ishiguro et al., 1997). Exponentially growing cells were incubated as indicated in EMM supplemented with D-[U-¹⁴C]glucose, as specified for each case of bgs4⁺ shut-off (5 μCi/ml control cells and 9 μCi/ml shut-off cells) or overexpression induction (3 μCi/ml control cells and 4 μCi/ml overexpression cells), and they were collected at the indicated times. All determinations were carried out in duplicate, and the values for each strain were calculated from three independent experiments.

Cell colony and spore images were obtained directly from the growing plate with a Zeiss Axiophot microscope, a Plan-NEOFLUAR 40×/0.75 Ph2 objective, a RT Monochrome SPOT digital camera and SPOT V3.4.2 software (Diagnostic Instruments). For cell wall staining, early log-phase cells grown in YES or EMM liquid medium were concentrated and visualized directly by adding a solution of Calcofluor white (CW; 50 μg/ml final concentration), using the appropriate UV filter. For F-actin visualization, cells were fixed and stained with rhodamine-conjugated phalloidin, as described (Alfa et al., 1993). Nucleus and cell wall staining and direct GFP fluorescence visualization of early log-phase cells were essentially as described, suspending the cells in PBS containing 20 μg/ml CW and 10 μg/ml Hoechst 33258 (Sigma) (Cortes et al., 2002). Images were obtained with a Leica DM RXA fluorescence microscope as previously described (Cortes et al., 2002).

The gene coding for the fourth Bgs subunit, bgs4⁺, has been found by the S. pombe Genome Sequencing Project (SPCC1840.02c), and consists of an open reading frame of 5868 nucleotides; it encodes a predicted membrane protein of 1955 amino acids with a molecular weight of 225 kDa. Bgs4p shows sequence identity ranging from 53 to 59% with other Bgs proteins and from 49% to 57% with S. cerevisiae Fks proteins. The cosmid clone that contains bgs4⁺ is located close to and between the cosmids that contain the ppe1⁺ (=ppx1⁺/esp1⁺) and ade5⁺ genes.

Interestingly, a previous search for S. pombe mutants affected in cell wall (1,3)β-D-glucan biosynthesis led to the characterization of the cwg1-1 mutant (Ribas et al., 1991). This mutant showed a rounded thermosensitive lethal phenotype, accompanied by a considerable reduction in cell wall β-glucan (25% of wild-type amount) and GS activity (25% wild-type activity); this defect was associated with the insoluble catalytic fraction of the enzyme. Unfortunately, cloning of the corresponding cwg1⁺ gene using different gene libraries was unsuccessful. By genetic analysis, cwg1⁺ was mapped to the same region as bgs4⁺: on the right arm of chromosome III, linked to and mapping between ppe1⁺ and ade5⁺. Therefore, the close location of bgs4⁺ and cwg1⁺ and the hypothetical involvement of both gene products as part of the GS catalytic subunit suggest that bgs4⁺ and cwg1⁺ could be the same gene.

To test this notion, the bgs4⁺ gene was cloned from the cosmid clone SPCC1840c and the plasmid obtained, pAL-bgs4⁺ (pJG1), was used to transform the cwg1-1 mutant strain. It was found that bgs4⁺ was able to suppress all the cwg1-1 defects: (1) the lethal phenotype at 37°C (Fig. 1A); (2) the spherical morphology prior to lysis (data not shown); (3) the increased cell wall fluorescence in the presence of the fluorochrome Calcofluor white (CW), which preferentially binds cell wall material (data not shown); and (4) the low (25%) in vitro GS activity (Fig. 1B) and in vivo cell wall β-glucan content. In addition, the same suppression of all the phenotypes was obtained with another mutant allele: cwg1-2 (Perez and Ribas, 2004) (data not shown). Moreover, genetic cross between bgs4⁺:leu1⁺ leu1-32 and cwg1-1 leu1-32 strains showed no recombination (98 PD: 0 T: 0 NPD) between bgs4⁺ (Leu⁺ phenotype) and cwg1-1 (lytic phenotype) markers. This tight linkage (<0.51 cM) suggested that bgs4⁺ and cwg1⁺ are the same gene. This was finally corroborated by complementation analysis in diploid strains. The cwg1-1 mutant phenotype was recessive in heterozygosity, as described (Ribas et al., 1991), but the heterozygous cwg1-1/bgs4Δ strain displayed the same thermosensitive phenotype as the homozygous cwg1-1/cwg1-1 strain (Fig. 1C). This proves that cwg1⁺ and bgs4⁺ are the same gene.

Genetic analysis revealed a tight linkage between cwg1-1 and the thermosensitive mutant orb11-59, which has been described as rounded and showing a complete loss of cell polarity (Verde et al., 1995). Like cwg1-1, orb11-59 defects were not suppressed by multicopy expression of any of the three homologous genes bgs1⁺, bgs2⁺ or bgs3⁺ (data not shown). In contrast, both multicopy and integrated bgs4⁺ expression suppressed the lethal and morphological phenotypes of the orb11-59 strain (Fig. 1D), indicating that bgs4⁺, cwg1⁺ and orb11⁺ are the same gene.

In order to study the effect produced by the absence of Bgs4p in the cell, we constructed a heterozygous diploid strain, bgs4⁺/bgs4Δ, by replacing the entire bgs4⁺ open reading frame with the ura4⁺ gene (see Materials and Methods). Tetrad analysis of sporulated diploids showed that only the two Ura^- (bgs4⁺) clones of each tetrad were viable and sorbitol was unable to rescue the viability of the bgs4Δ spores (Fig. 2A). Therefore, bgs4⁺ is essential at least during spore germination, as reported for bgs1⁺ and bgs3⁺ (Cortes et al., 2002; Liu et al., 1999; Martin et al., 2003). Microscopic observation revealed that bgs4Δ spores were able to germinate but that they lysed before the first round of cell division (Fig. 2A). In contrast, bgs1Δ spores were able to germinate, producing enlarged and multiseptated cells that eventually died (data not shown) and bgs3Δ spores germinated and divided several times, producing small groups of rounded cells (Martin et al., 2003).

In order to analyze the effect of the absence of bgs4⁺, it was expressed under the control of the 41X (medium) and 81X (low) versions of the thiamine-regulated nmt1⁺ promoter (see Materials and Methods). Viable bgs4Δ cells were obtained only when the 41X or 81X promoter was induced (Fig. 2B). In both cases, the cells had a growth rate and morphology similar to wild-type cells. According to the different expression levels of these nmt1⁺ promoters (Moreno et al., 2000), S. pombe cells can support at least a 50-fold difference in bgs4⁺ expression. bgs4⁺ shut-off arrested cell growth and this occurred earlier in the p81X-bgs4⁺ strain (data not shown). After 14 hours of bgs4⁺ repression, the cells stopped growing (Fig. 2B), and unlike the cwg1-1, cwg1-2 and orb11-59 mutant alleles, osmotic stabilization only partially protected the cells, delaying the arrest of growth to 17 hours and maintaining a residual growth (Fig. 2B).

Morphological observations revealed that bgs4⁺ repression promoted cell lysis and the release of cytoplasmic material from the regions of cell wall synthesis, either the poles or the septum (Fig. 2C). Lysis at the septum appeared earlier and was more abundant than at the poles: 15% at 14-15 hours and increasing up to 70-80% at 21-23 hours after bgs4⁺ shut-off (data not shown). Initially, the lysis occurred at the border between the septum and cell wall of one of the daughter cells, giving rise to one dead cell whose cell wall remained attached on one side to the live cell; this probably retained most of the previously formed cell wall and hence had a better chance of survival. After longer times, lysis occurred at both sides of the septum (Fig. 2C), resulting in the death of both cells. Lysis at the poles usually appeared after 17 hours and increased up to 21-23 hours of shut-off, representing around 10% of the cells. Sorbitol delayed lysis at the poles but not at the septum (data not shown), and the surviving cells had a healthier aspect and were able to maintain a slow increase in cell density (Fig. 2B), although they eventually died. These results suggest that the function of Bgs4p at the cell wall-septum boundary is more critical than that at the poles. A small population was able to survive bgs4⁺ shut-off, giving rise to stable but sickly spherical revertants (data not shown), probably due to cell wall compensatory mechanisms (see Discussion). Actin staining during shut-off showed that after 8 hours, actin patches began to appear dispersed throughout the cytoplasm, disappearing from or present at lower amounts at both the septum and poles. However, the contractile actomyosin ring remained stable, even at the time of cell lysis (Fig. 2E).

Analysis of in vitro GS activity during bgs4⁺ shut-off and before cell lysis revealed a dramatic decrease in activity (Fig. 2D), similar to that reported for the bgs4⁺ mutant allele cwg1-1 (Ribas et al., 1991). Next, the cell wall composition was analyzed immediately before cell lysis had begun. Surprisingly, the amount of cell wall did not decrease; instead, it rose from 33 to 43% and this increase was only due to α-glucan, suggesting an in vivo compensatory mechanism (Table 2). The increase in α-glucan and total cell wall was attenuated in the presence of sorbitol (Table 2) and was enhanced when the cells were grown at 37°C (24% α-glucan and 50% cell wall, data not shown). Contrary to the GS activity, cell wall β-glucan did not decrease, probably owing to a compensatory mechanism (see Discussion). All these data indicate that Bgs4p is critical for maintaining cell integrity at any site where cell wall growth takes place.

bgs4Δ p3X-bgs4⁺ cells grow normally under repressed conditions (expression is fivefold higher than with the induced 81X promoter) (Moreno et al., 2000). However, bgs4⁺ overexpression promoted the arrest of cell growth 16-18 hours after induction and sorbitol merely delayed it to 20 hours and permitted a residual growth (Fig. 3A). Contrary to bgs4⁺ shut-off, bgs4⁺-overexpressing cells were viable for several days, with 80% and 70% of cells remaining viable after 15 and 30 hours of overexpression, respectively (data not shown). The cells maintained a slow growth pattern and increased their heterogeneous phenotype of rounded and elongated cells, some becoming refringent and containing aberrant depositions of CW-stained cell wall material (Fig. 3B). Actin staining showed that after 24 hours of overexpression the actin patches were delocalized throughout the cytoplasm in the rounded cells, disappearing mainly from the poles but still remaining localized to the septum to some extent (Fig. 3C).

Unexpectedly, GS activity decreased during bgs4⁺ overexpression. This activity was only 65% at 15 hours, when the cells still had no apparent phenotype, and decreased dramatically to 30% at 30 hours (Fig. 3D). This suggests that the GS catalytic subunit may be composed of a multiprotein complex rather than by Bgs4p alone (see Discussion). Cell wall analysis during bgs4⁺ overexpression revealed an increase in cell wall due to α-glucan, whereas β-glucan was not affected (Table 3), suggesting that, as for its shut-off, bgs4⁺ overexpression promotes the activation of cell wall compensatory mechanisms.

Neither a C-terminal nor an N-terminal GFP-Bgs4p fusion was able to maintain the function of Bgs4p. Similarly, other fusions with GFP inserted in the N-terminal region of Bgs4p were not functional either (see Materials and Methods). Only GFP-Bgs4p with GFP inserted at amino acid 324 completely restored the wild-type phenotype to a bgs4Δ strain and also showed localized GFP fluorescence. Therefore, this GFP-bgs4⁺ bgs4Δ strain (GFP at amino acid 324, integrated single copy, own promoter and absence of original bgs4⁺ gene) was chosen for Bgs4p localization studies.

The cells were visualized by GFP fluorescence to detect Bgs4p and by CW staining, which preferentially binds to cell wall material, probably (1,3)β-D-glucan, of the septum and growing zones. CW staining revealed fluorescence at the single growing pole during monopolar growth, at both poles during bipolar growth, and at the septum during cytokinesis. Bgs4p was found to localize to the growing ends and septum along the mitotic cycle, overlapping the CW staining (Fig. 4A).

Initially, some dividing cells were able to separate before resuming growth, Bgs4p appearing at the new pole formed after septum degradation. When cell growth began, Bgs4p accumulated at the old growing end. During bipolar growth, Bgs4p also localized to the opposite pole. Finally, after CW-stained septum material had become visible, the GFP disappeared from both poles and localized to the middle of the cell, concentrating as two faint dots at both sides of the emerging septum. Then, the GFP moved to the tip of the growing septum, concentrating in a narrow ring that moved centripetally with the edge of the growing septum. The GFP signal was stronger where the ring was intersected by the focal plane. An apparent signal remained along the invaginated membrane where new cell wall material had been synthesized (Fig. 4B). After the septum wall had been completed, the GFP appeared in two separated bands. During cell division, Bgs4p remained at both sides of the septum until the two daughter cells were ready to (or already had) separated, then changing its localization to the old end of each daughter cell, which then initiated monopolar growth (Fig. 4A,B).

Bgs1p has also been described to localize to septum and poles (Cortes et al., 2002) but its localization pattern during septation seemed to differ in some aspects from that of Bgs4p. Therefore, GFP-Bgs1p and GFP-Bgs4p were compared as regards their localization and displacement from the tips to the medial ring during mitosis (Fig. 4C). Bgs1p and Bgs4p localized to both cell ends until the end of anaphase, when nuclear division had been completed. When the nuclei were migrating to the cell poles Bgs1p, but not Bgs4p, disappeared from the poles, appearing as two GFP dots at the middle of the cell. Then, after the nuclei had reached the cell poles, two CW-stained septum dots appeared overlapping the bright GFP-Bgs1p signals. Simultaneously, GFP-Bgs4p began to disappear from the poles and to concentrate diffusely in the middle of the cell. Next, when the septum and GFP-Bgs1p dots were intense, GFP-bgs4p condensed at both sides of the septum. Finally, GFP-Bgs4p moved towards and appeared concentrated at the end of the growing septum, overlapping GFP-Bgs1p (Fig. 4C). Therefore, Bgs1p moves earlier than Bgs4p to the division site during mitosis, appearing before any septum material can be detected, whereas Bgs4p appears later, after the septum has become visible.

In order to study the requirements of Bgs4p localization during septation, several septation mutants were generated in a GFP-bgs4⁺ bgs4Δ background and GFP-Bgs4p localization was analyzed. The anillin-like Mid1p is located to the middle zone and is essential for correct positioning of the medial ring (Bahler et al., 1998; Sohrmann et al., 1996). The mid1 mutants show the actomyosin ring and septum positioned at random locations and angles (Chang et al., 1996), coincident with the localization of Bgs1p and with the CW-stained septum material (Cortes et al., 2002). Initially, Bgs4p positioning coincided with the altered ring and septa (Fig. 5), suggesting that Bgs4p localization is not dependent on Mid1p and that it can interact with the rest of the septation components and localizes to the abnormally positioned septa. However, cells with the aggravated mutant phenotype displayed aberrant CW-stained septum material, previously shown to be coincident with Bgs1p localization (Cortes et al., 2002), but that did not necessarily overlap Bgs4p localization (compare arrowhead and arrow positions in mid1-366 cells in Fig. 5). Considering that Bgs4p would remain functional as in wild-type bgs4Δ cells, this result suggests that Bgs4p is responsible for the synthesis of a (1,3)β-D-glucan not stained by CW and, reciprocally, that CW detects a cell wall component, probably glucan, synthesized by other proteins, such as Bgs1p.

The profilin Cdc3p is essential for the assembly of the medial ring actin filaments (Balasubramanian et al., 1994; Chang et al., 1996). cdc3-6 mutants show defects in septum formation but not in cell elongation and they accumulate CW-stained cell wall material at the site of septum formation (Nurse et al., 1976). Bgs4p was detected at the poles and/or overlapping with many of the cell wall structures synthesized at the middle of the cell (Fig. 5). The PCH family Cdc15p is essential for cell division and the mutants display impaired actin ring formation or form unstable ring structures, although they are not impaired in cell elongation (Chang et al., 1996; Fankhauser et al., 1995). In the cdc15-140 mutant, Bgs4p was detected either dispersed in the cytoplasm or at the cell poles but not at the middle of the cell. Also, the mutant sometimes showed thick or diffuse aberrant cell wall structures resembling the septa but in which Bgs4p was never present (Fig. 5). Cdc16p forms a two-component GAP for Spg1p GTPase that is required for negative regulation of septum formation (Cerutti and Simanis, 1999; Furge et al., 1998). The cdc16-116 mutants formed multiple septa because of their failure to turn off septation. Bgs4p was located at the poles and at many of the multiple septa (Fig. 5). In sum, these data indicate that the contractile ring is necessary for Bgs4p localization to the middle of the cell.

Next, we wondered whether the SIN proteins are necessary for Bgs4p displacement from the tips to the medial ring during mitosis or for its stable localization to the medial ring. To check this, CW and nucleus staining, and GFP were analyzed simultaneously in the same cells. The SIN proteins Cdc11p and Cdc14p are essential for septum formation (Guertin et al., 2000; Krapp et al., 2004; Krapp et al., 2001). The cdc11-119 and cdc14-118 mutants have cytokinesis blocked but assemble actomyosin rings during mitosis and accumulate actin patches at the cell ends during interphase. In these mutants, GFP-Bgs4p localized to the poles, coinciding with cell growth (Fig. 5 and data not shown). Some cdc14-118 cells exhibited CW-stained septum structures (shown by arrowheads in cdc14-118 cells in Fig. 5), formed before they expressed the thermosensitive phenotype. Analysis of cdc14-118 cells at different incubation times (3-6 hours) at the restrictive temperature, which results in cells containing two, four and eight nuclei, elicited no change in the number of cells with partial septum structures and absence of complete septa in all cases. This indicates that these septum structures are inactive and never progress to complete the corresponding septum. Analysis of both mutants revealed that Bgs4p never relocated to the medial ring in cells lacking a partial septum. However, the number of cells with no partial septum showing internalized Bgs4p was higher than that observed in wild-type cells (data not shown), suggesting that contrary to the requirements of Bgs1p (Cortes et al., 2002), SIN signaling is not necessary for post-anaphase Bgs4p displacement from the tips to the middle of the cell. Additionally, when a partial septum structure was present, Bgs4p could be found at the poles and absent from the medial ring, or located to the contractile ring (shown by arrows in cdc14-118 cells in Fig. 5) and absent from the poles, or at both sites simultaneously (Fig. 5). Therefore, SIN proteins are dispensable for Bgs4p localization to the medial ring when a partial septum structure is present.

Bgs4p was localized to the poles during both monopolar and bipolar growth. The tea1-1 and tea2-1 mutants do not recognize the old end and activate a new incorrectly positioned growing site owing to a failure to establish polarized growth along the long axis of the cell (Browning et al., 2003; Sawin and Snaith, 2004; Verde et al., 1995). In these mutants, Bgs4p appeared localized to the incorrect new growing tips or to the mislocalized septum. Similarly, the double mutant tea2-1 cdc11-119 activates several new growing ends in which Bgs4p was always present (Fig. 6A). Therefore, Bgs4p, like Bgs1p and Bgs3p (Cortes et al., 2002; Martin et al., 2003), remains polarized to the growing sites independently of its mislocalization owing to a failure of the polarity proteins to establish the correct growing ends.

F-actin localizes to either the growing tips or the dividing medial region, playing a role in cell wall synthesis and polarized growth (Kobori et al., 1989; Marks et al., 1986). Actin coincides with Bgs4p localization along the cell cycle and we were therefore prompted to analyze whether the actin cytoskeleton is involved in correct Bgs4p localization as well. The actin mutant cps8-188 shows rounded and enlarged cells in which the actin remains polymerized but randomly distributed and the middle ring is absent (Ishiguro and Kobayashi, 1996). Bgs4p localization extended from the septum and poles along the plasma membrane all around the cell (Fig. 6B). Intermediate phenotypes showed cells with aberrant contractile ring structures, in which Bgs4p localized simultaneously to the septum, poles, and regions of aberrant cell wall synthesis (data not shown). The profilin mutant cdc3-6 presents a terminal phenotype reflecting the absence of F-actin structures. The cdc3-6 cells contain aberrant septa and cell wall structures as a result of a previous intermediate phenotype in which Bgs4p was absent, appearing delocalized inside the cell as a low GFP background (data not shown). Thus, Bgs4p localization depends on the polymerized F-actin cytoskeleton and its specific localization to regions of active cell wall synthesis depends on the correct localization of polymerized F-actin, as reported for Bgs1p, Fks1p and the putative (1,3)α-D-glucan synthase Ags1p/Mok1p (Cortes et al., 2002; Katayama et al., 1999; Utsugi et al., 2002).

We next analyzed in detail the role of F-actin in Bgs4p localization. We used bgs4Δ RFP-bgs4⁺ crn1⁺-GFP cells to analyze, in the same cells, Bgs4p localization by RFP fluorescence and F-actin patch localization by GFP fluorescence of the coronin homologue Crn1p, a protein associated with and used as a marker for F-actin patches and cables (Pelham and Chang, 2002; Pelham and Chang, 2001). First, we wondered if the F-actin cytoskeleton is necessary for stable Bgs4p maintenance at the growing regions. The cells were treated for 15 minutes with latrunculin A (Lat A), which prevents actin polymerization. As a control, cells were incubated with DMSO, showing no effect on Bgs4p or actin patch localization (data not shown). Lat A-treated cells displayed the absence of polymerized actin patches, with the GFP fluorescence distributed uniformly in the cytoplasm, whereas Bgs4p remained localized, indistinguishable from control cells (Fig. 7A). A longer Lat A treatment of 30 minutes produced the same effect on Bgs4p and actin localization, but the cell recovery time required to relocalize both Bgs4p and F-actin increased up to 2 hours (data not shown). This indicates that F-actin is not needed for stable Bgs4p maintenance to the medial ring and poles. Then, Lat A was removed to analyze the recovery of actin patches. After 30 minutes of Lat A wash out, the actin patches reappeared and were randomly localized, whereas by contrast, Bgs4p had lost its polarization and appeared transiently delocalized inside the cell (Fig. 7A). At 1 hour, the actin patches were partially localized and Bgs4p had relocalized to the growing sites. At 3 hours, both the actin patches and Bgs4p localization were fully recovered (Fig. 7A). Next, we speculated whether the observed transitory Bgs4p delocalization and relocalization might be due to the transient random distribution of the repolymerized actin patches. Initially, the cells were exposed to stress conditions (2 hours at high temperature and absence of nitrogen). Under these conditions, the F-actin patches appeared randomly localized and partially depolymerized and Bgs4p was completely delocalized inside the cell (Fig. 7B). Then, the cells were transferred to fresh YES complete medium, in either the presence or absence of Lat A. In the presence of Lat A, actin appeared depolymerized in the cytoplasm and Bgs4p was unable to relocalize to the growing zones for at least 2 hours. After 1 hour in the absence of Lat A, the control cells already displayed a few actin patches localized to the growing regions and part of Bgs4p appeared relocalized. At 2 hours, F-actin and Bgs4p appeared completely relocalized to the growing sites (Fig. 7B). Therefore, the F-actin cytoskeleton is necessary for Bgs4p repolarization to the growing sites. In a third experiment, the cells were treated with Lat A as in Fig. 7A to depolymerize the F-actin, but maintaining Bgs4p polarized. Then, the cells were exposed to stress conditions in either the presence or absence of Lat A (Fig. 7C). In the presence of Lat A, actin remained depolymerized and Bgs4p remained polarized for at least 3 hours. In the control cells (absence of Lat A), it was observed that at 1 hour some actin had repolymerized as random actin patches and Bgs4p appeared delocalized inside the cell; this effect persisted for at least 3 hours (Fig. 7C). Hence, the F-actin cytoskeleton is also needed for Bgs4p delocalization from the growing zones.

Bgs4p plays a major role in cell wall synthesis during vegetative growth and is present at each site where growth takes place. Therefore, we wondered if Bgs4p might also be present at the cell wall growth sites during the sexual phase of the life cycle in a homothallic bgs4Δ GFP-bgs4⁺ h⁹⁰ strain.

Initially, Bgs4p appeared localized in only one of the mating cells. Then, the mating partner responded by forming a mating projection to which Bgs4p localized (Fig. 8). The cell walls of both cells fused and Bgs4p localized to the contact region that separates the cells. When the contact region had disappeared, Bgs4p remained as a wide band in the neck between both cells. Later, it disappeared from the neck membrane and began to fill in the whole cytoplasm, but not the fused nucleus, with a uniform fluorescence that persisted during the process of meiosis. Contrary to Bgs1p, Bgs4 never appeared inside or associated with the membranes of the oval-shaped prespores. Instead, it remained as a bright and uniform fluorescence in the cytoplasm and around the prespore. Although the prespores were perfectly distinguishable with phase-contrast microscopy (data not shown), CW staining only revealed the zygote shape. Following this, Bgs4p fluorescence decreased transiently and reappeared as bright patches close to or around the spore wall. CW began to stain the spores, probably due to the maturation of the prespore wall to become a spore. Finally, Bgs4p fluorescence gradually disappeared, remaining only localized to the spore periphery and as faint patches inside the mature spores (Fig. 8). In sum, these observations suggest Bgs4p may cooperate to form the mating projection and to protect the cell from an excess of cell wall degradation during cell fusion. In addition, Bgs4p may play a specific role in both prespore and spore wall synthesis.

We were interested in analyzing whether the initial asymmetry observed for the appearance of Bgs4p in the mating projections was related to a specific mating type. To do so, we analyzed the appearance of Bgs4p in the mating projections along the mating process of strains h⁺GFP-bgs4⁺ bgs4Δ × h^–RFP-bgs4⁺ bgs4Δ and h⁺RFP-bgs4⁺ bgs4Δ × h^–GFP-bgs4⁺ bgs4Δ (Fig. 9). Initially, Bgs4p was found delocalized inside the cells. Then, both GFP- and RFP-Bgs4p began to appear localized in some cells of both h⁺ strains (see arrowheads in Fig. 9), whereas no h^– cell showed localized Bgs4p. Following this, the amount of h⁺ cells with localized GFP- or RFP-Bgs4p increased, and all the h^– cells continued displaying delocalized Bgs4p. One hour after the appearance of polarized Bgs4p in h⁺ cells, some h^– cells showed the first localized GFP- or RFP-Bgs4p (see arrows in Fig. 9). Finally, the number of cells with localized Bgs4p increased to a similar extent in both mating types and the first zygotes appeared (Fig. 9). Therefore, during the mating progression Bgs4p appears localized first in P (h⁺) cells and later in M (h^–) cells, in agreement with the reported formation of mating projections in response to pheromones during the mating process (Nielsen and Davey, 1995).

Analysis of the localization of bgs4Δ GFP-bgs4⁺ h⁹⁰ spores showed that Bgs4p might also be involved in the germination process (Fig. 10). The spores were released by spontaneous lysis of the asci in the absence of any enzymatic treatment that might alter spore wall or Bgs4p localization. First, during isotropic spore growth Bgs4p fluorescence increased uniformly inside the spore, and more weakly all around the periphery. Then, the internal fluorescence disappeared and Bgs4p concentrated to the place where polarized growth began. CW stained the site of polarized growth, as in vegetative cells, and Bgs4p was localized to this growing pole. Next, Bgs4p reorganized to the middle of the cell, localizing to the medial ring and septum until the first cell was ready to separate. Finally, Bgs4p localized to the old end of the new cell but remained at the septation pole of the spore (Fig. 10), probably because growth will resume only at this pole, with a vegetative cell structure.

The fission yeast is an attractive morphogenetic model for understanding how fungi, and cells in general, regulate their polarized growth (Chang and Peter, 2003; Hayles and Nurse, 2001) and cytokinesis through the cell cycle (Balasubramanian et al., 2004; Guertin et al., 2002; Krapp et al., 2004). However, little is known about how fission yeast synthesizes and regulates its cell wall components during these processes; in particular (1,3)β-D-glucan as the main structural polysaccharide. S. pombe contains four putative GS catalytic subunits, although their involvement in (1,3)β-D-glucan synthesis has not yet been elucidated.

Here we cloned bgs4⁺ and we show that it is the only bgs gene found to be essential to maintain cell integrity during both cytokinesis and polarized growth. In addition, Bgs4p is the first subunit directly involved as part of the GS catalytic subunit. First, bgs4⁺ is the only gene of this family that has mutant alleles (cwg1-1 and cwg1-2) with a specific reduction in cell wall (1,3)β-D-glucan and GS activity, the latter due to a defect in the catalytic subunit (Perez and Ribas, 2004; Ribas et al., 1991). Second, bgs4⁺ shut-off promotes specific cell lysis at the growing regions, poles and septum; this is accompanied by a dramatic decrease in GS activity. A similar lysis occurred in germinating bgs4Δ spores, highlighting the importance of Bgs4p during spore germination. The specific localization of Bgs4p to the areas of active cell wall synthesis is consistent with these observations.

The predominant bgs4⁺ shut-off phenotype is that of lysis at the already synthesized septum at the beginning of cell division, suggesting that the function of Bgs4p is not only essential for cell wall growth but also for cell wall repair, compensating an excess of septum wall degradation during cytokinesis.

The three mutants described for Bgs1p show similar septation defects (Ishiguro et al., 1997; Le Goff et al., 1999; Liu et al., 1999) but none of them show a defect in cell wall (1,3)β-D-glucan or GS activity, although the mutant phenotype, homology with Fks proteins and Bgs1p localization led to the proposal that this subunit is responsible for the synthesis of a linear (1,3)β-D-glucan from the primary septum and growing poles (Cortes et al., 2002). Nonetheless, the true function of Bgs1p remains obscure. No bgs3 mutant has been described and bgs3⁺ shut-off does not show any relationship with (1,3)β-D-glucan synthesis, measured by cell wall composition or enzymatic activity (Martin et al., 2003). Also, the S. cerevisiae homologue FKS1 has been described from different mutant alleles, all of them either unrelated (Eng et al., 1994; Garrett-Engele et al., 1995; Ram et al., 1995) or indirectly related to GS activity owing to their resistance to the specific GS inhibitors papulacandin and echinocandin (Castro et al., 1995; Douglas et al., 1994). Fks1p has been implicated in GS activity because some of these mutants showed a reduction in GS or a general reduction in cell wall β-glucan and because partial purification of the GS membrane enzyme showed Fks1p to be present among the enriched proteins of the extract (Inoue et al., 1995). Conversely, different Fks1p mutations displayed a simultaneous reduction or increase in both cell wall (1,3)β and (1,6)β-D-glucan, and specific (1,6)β-D-glucan mutants exhibited a decrease in GS activity similar to that of the fks1Δ mutant (Dijkgraaf et al., 2002). In addition, although the fks1Δ fks2Δ double mutant is lethal, no cell lysis at the growing regions or release of cytoplasmic material because of a defect in cell wall integrity have been described. Additionally, the S. cerevisiae Fks3p homologue is not essential and has no known function. In sum, Bgs4p is the first subunit that shows a direct in vivo involvement in cell wall (1,3)β-D-glucan synthesis and GS activity, detected by its mutants and shut-off phenotypes, as well as in the control of cell structure and integrity.

In contrast to the cwg1-1 mutant allele, bgs4⁺ shut-off did not elicit a decrease in cell wall (1,3)β-D-glucan, suggesting a compensatory mechanism that activates other Bgs subunits. In fact, this is consistent with the appearance of stable bgs4Δ revertants and with the considerable increase in at least GFP-Bgs1p during bgs4⁺ shut-off (our unpublished results). The absence of Bgs4p would be necessary to activate this compensatory mechanism, as the presence of a partially non-functional protein in cwg1 mutants did not display it. A similar mechanism might also be induced during bgs3⁺ shut-off (Martin et al., 2003) but, unlike bgs4⁺ shut-off, there is no decrease in GS activity, suggesting that Bgs4p is the main agent responsible for the GS activity detected in vitro, as reported for S. cerevisiae Chs1p and chitin synthase activity (Roncero, 2002). Alternatively, the difference in cell wall β-glucan levels detected between the cwg1-1 mutant and bgs4⁺ shut-off could be due to a cumulative defect in the former that cannot be detected with a restricted but lethal defect produced by the absence of Bgs4p. In the case of Bgs1p, and contrary to the cwg1-1 mutant, the cps1-12 mutant did not show any decrease in GS or cell wall β-glucan. This is consistent with the idea of Bgs1p being responsible for the synthesis of a minor linear (1,3)β-D-glucan of the primary septum and nascent cell wall (Cortes et al., 2002).

We found that S. pombe also induces a stronger and more general compensatory mechanism in response to cell wall alterations and stress by increasing (1,3)α-glucan synthesis. This has been observed during bgs4 shut-off or overexpression, bgs3⁺ shut-off, expression of cwg1-1 or cps1-12 mutant phenotypes (Ishiguro et al., 1997; Martin et al., 2003; Ribas et al., 1991) and bgs1⁺ shut-off or overexpression (our unpublished results). bgs4⁺ overexpression did not increase cell wall β-glucan or GS activity. Instead, it resulted in a considerable decrease in GS activity. The same result has been described for bgs3⁺ (Martin et al., 2003). This effect is opposite to that observed when the single regulatory subunit Rho1p is overexpressed (Arellano et al., 1996) or with Ags1p/Mok1p overexpression and an increase in cell wall α-glucan (Katayama et al., 1999). It is unlikely this would be due to limiting concentrations of the Rho1p regulatory subunit, because in that case it would promote neither an increase nor a decrease in GS activity. This suggests that Bgs4p forms part of a multiprotein complex that constitutes the GS catalytic subunit. Thus, overproducing only Bgs4p would result in multiple partial complexes and a decrease in the enzymatic activity. Regarding its shut-off, cell wall (1,3)β-D-glucan would not decrease because of a compensatory mechanism from other Bgs proteins. Additional studies will be necessary to characterize the (1,3)β-D-glucan synthase catalytic components and the mechanism by which they and the Bgs proteins interact.

For Bgs4p localization studies, we attempted to mimic the Bgs4p physiological state: an integrated single copy gene, its own promoter and the wild-type phenotype in a bgs4Δ background. The conformation of Bgs4p seems to be different or less stable than that of Bgs1p and Bgs3p, as the N-terminal GFP fusion is functional with Bgs1p and Bgs3p but not with Bgs4p. As reported for Bgs1p and Fks1p (Cortes et al., 2002; Delley and Hall, 1999), the localization of Bgs4p is very unstable and extremely sensitive to cell wall stress, mainly at the poles, but less so at the septum, and more stable at the contractile ring. Bgs4p localization during septation was delayed with respect to that of Bgs1p, supporting the idea that Bgs1p would synthesize a (1,3)β-D-glucan of the primary septum (Cortes et al., 2002) and that Bgs4p would synthesize a major (1,3)β-D-glucan present in the secondary septum in a process that is delayed but overlapping with primary septum synthesis. In S. cerevisiae, the primary septum is made of chitin. S. pombe does not contain chitin and it is therefore reasonable to speculate that Bgs proteins and (1,3)β-D-glucans have specialized to assume multiple essential functions. In this sense, Bgs1p would mimic budding yeast Chs2p, responsible for the synthesis of the primary septum (Schmidt et al., 2002), whereas Bgs4p would mimic Fks1p and Fks2p, which synthesize the (1,3)β-D-glucan of the secondary septum and cell wall, or Chs1p, which protects the cell against an excess of cell wall degradation. However, the S. pombe model becomes more complicated because it contains two other proteins, Bgs3p and Ags1p/Mok1p, which are essential for cell wall construction. The specific role of each of these proteins is not yet known, although it is reasonable to assume that they have specific and separate functions during cell wall synthesis, as is the case for the three chitin synthases in S. cerevisiae (Shaw et al., 1991).

The requirements of Bgs4p localization during septation are only partially coincident with those of Bgs1p. Both depend on the contractile ring, but only Bgs1p fully depends on the SIN. Bgs4p does not depend on SIN proteins to internalize from the tips during mitosis. Furthermore, it depends only indirectly on the SIN to localize to the medial ring, and only when no septum has been initiated. When some septum material is present, as in the cdc14-118 mutant, Bgs4p is able to localize to the contractile ring in the absence of a functional SIN. Nonetheless, Bgs4p localization is not sufficient to continue septum synthesis and it eventually resumes its localization to the poles. This suggests that a pre-made septum material, probably (1,3)β-D-glucan from the primary septum, acts as a landmark for Bgs4p localization, and that Bgs1p, and hence the SIN, would be needed only for Bgs4p to continue septum synthesis.

In the ring positioning mutant mid1-366, it was observed that CW-stained septum material overlapped Bgs1p localization but not necessarily that of Bgs4p. This result, and the fact that in a normal septum CW staining appears before the arrival of Bgs4p, indicates that at least part of the CW-stained septum material is not made by Bgs4p. Furthermore, considering that GFP-Bgs4p is functional in the mid1 mutant, as it is in the wild-type GFP-bgs4⁺ bgs4Δ strain, the product made by Bgs4p would not be stained by CW (compare arrowhead and arrow positions in mid1 mutant in Fig. 5). This suggests that CW could discriminate two different (1,3)β-D-glucan products probably comprising two Bgs proteins, which is consistent with the idea that CW recognizes fibrillar structures such as linear (1,3)β-D-glucan, proposed to be synthesized by Bgs1p.

We analyzed the role of F-actin in Bgs4p localization and observed that F-actin function is necessary for Bgs4p delocalization and relocalization but not for its stable maintenance at the growing zones. Some of these requirements have also been reported for other Bgs/Fks homologues. F-actin is necessary for the localization, but not the maintenance, of Bgs1p at the cell division ring; it is not required for Bgs3p maintenance at the growing sites, and it is needed for Fks1p depolarization and repolarization (Delley and Hall, 1999; Liu et al., 2002; Martin et al., 2003). It is reasonable to assume that all Bgs/Fks family members share the same F-actin requirements needed for their transport to or their internalization from the plasma membrane but dispensable for their maintenance at the growing sites as part of the plasma membrane.

Bgs2p is essential for sporulation (Liu et al., 2000; Martin et al., 2000), but this does not imply that all the (1,3)β-D-glucan of the spore wall is due to Bgs2p. In fact, Bgs1p (Cortes et al., 2002) and Bgs4p display a specific but different localization pattern during spore formation, suggesting a specific role for Bgs1p and Bgs4p in spore wall synthesis. Although not confirmed, it is very likely that Bgs3p would also be present during spore maturation. In fact, gene expression analysis using DNA microarrays has shown that the four bgs genes are induced during the sporulation program (Mata et al., 2002). Although bgs2⁺ presents the highest induction of expression, bgs1⁺, bgs3⁺ and bgs4⁺ exhibit an induction similar to that of the essential sporulation-specific chitin synthase gene chs1⁺ (Arellano et al., 2000). Further study is necessary to ascertain the specific role of each bgs in spore wall formation. We found that during the mating progression Bgs4p appeared localized first in P (h⁺) cells and later in M (h^–) cells. This asymmetry suggests that Bgs4p localization forms part of the sequential pheromone response process between mating partners. At first, only M cells produce a pheromone (M factor), which induces P cells to form a mating projection and to produce P factor. Then, the P factor induces M cells to form the mating projections (Nielsen and Davey, 1995).

S. pombe is the only known model system containing four essential Bgs proteins that are presumed to synthesize similar (1,3)β-D-glucans. Bgs2p is specific for sporulation whereas Bgs1p and Bgs3p do not have a clear function. Bgs4p is the only subunit found to be necessary for cell integrity. Our Bgs4p results point to its involvement in the synthesis of an essential cell wall and septum (1,3)β-D-glucan, and also in a crucial repair mechanism against excess cell wall degradation during cell division. In summary, this is an appealing system but further experimental work will be needed to ascertain the functions of the other subunits, Bgs1p and Bgs3p, as possible GS catalytic subunits, as well as to determine their specific role in the fission yeast life cycle.

We thank P. Pérez and B. Santos for a critical reading of the manuscript, and N. Skinner for language revision. We thank P. Munz, R. Egel, M. Yanagida, P. Nurse and F. Chang for the generous gift of strains, R. Gwilliam for cosmid clones and R.Y. Tsien for kindly providing mRFP-containing plasmid. J.C.G.C. and E.C. acknowledge support from a fellowship granted by Consejo Superior de Investigaciones Científicas and Ministerio de Educación y Ciencia (Spain), respectively. This work was supported by grants BIO2000-1448 and BIO2003-01040 from the Comisión Interministerial de Ciencia y Tecnología, and CSI04/03 from the Junta de Castilla y León, Spain.