Role of the Synthase Domain of Ags1p in Cell Wall α-Glucan Biosynthesis in Fission Yeast*

The cell wall is important for maintenance of the structural integrity and morphology of fungal cells. Besides β-glucan and chitin, α-glucan is a major polysaccharide in the cell wall of many fungi. In the fission yeast Schizosaccharomyces pombe, cell wall α-glucan is an essential component, consisting mainly of (1,3)-α-glucan with ∼10% (1,4)-linked α-glucose residues. The multidomain protein Ags1p is required for α-glucan biosynthesis and is conserved among cell wall α-glucan-containing fungi. One of its domains shares amino acid sequence motifs with (1,4)-α-glucan synthases such as bacterial glycogen synthases and plant starch synthases. Whether Ags1p is involved in the synthesis of the (1,4)-α-glucan constituent of cell wall α-glucan had remained unclear. Here, we show that overexpression of Ags1p in S. pombe cells results in accumulation of (1,4)-α-glucan. To determine whether the synthase domain of Ags1p is responsible for this activity, we overexpressed Ags1p-E1526A, which carries a mutation in a putative catalytic residue of the synthase domain, but observed no accumulation of (1,4)-α-glucan. Compared with wild-type Ags1p, this mutant Ags1p showed a markedly reduced ability to complement the cell lysis phenotype of the temperature-sensitive ags1-1 mutant. Therefore, we conclude that, in S. pombe, the production of (1,4)-α-glucan by the synthase domain of Ags1p is important for the biosynthesis of cell wall α-glucan.

Distinct plasma membrane-localized synthases are responsible for the production of structural polysaccharides in the fungal cell wall, mostly (1,3)-␤-glucan, chitin, and ␣-glucan. (1,3)-␤-Glucan and chitin synthases were identified first in the budding yeast Saccharomyces cerevisiae (1,2) as integral membrane proteins with multiple transmembrane regions and a large cytoplasmic domain, which may be responsible for catalytic activity (3)(4)(5). Given that cell wall polysaccharides are absent in humans but crucial for maintaining the morphology and structural integrity of fungal cells, inhibitors of the synthases may function as antifungal drugs. For example, caspofungin is a (1,3)-␤-glucan synthase inhibitor proven to be effective against many fungi and also safe and well tolerated in humans (6). With regard to ␣-glucan synthase (Ags), 2 much less is known, and no inhibitors targeting this enzyme have been developed.
The ␣-glucan synthase Ags1p was identified in the fission yeast Schizosaccharomyces pombe using a temperature-sensitive mutant strain, ags1-1 ts , the cells of which lyse at the restrictive temperature due to a weakened cell wall unable to withstand internal osmotic pressure (7). These observations showed that cell wall ␣-glucan is an essential component of the S. pombe cell wall. Ags1p is a multidomain protein with two probable catalytic domains predicted to reside at opposite sides of the plasma membrane, as well as a multipass transmembrane domain. This overall domain structure of Ags1p is conserved among the four Ags1p homologs of S. pombe, Mok11p, Mok12p, Mok13p, and (partly) Mok14p, the genes of which are expressed during sporulation (8). Notably, this domain structure is also well conserved among Ags1p homologs in other cell wall ␣-glucan-containing fungi such as several human fungal pathogens in which cell wall ␣-glucan accounts for ϳ35% of the total cell wall polysaccharides. The genome of the filamentous fungus Aspergillus fumigatus contains three AGS genes, of which AGS1 and AGS3 appear to be directly involved in cell wall ␣-glucan biosynthesis (9,10). For the thermally dimorphic fungus Histoplasma capsulatum, the virulent yeast form contains substantial levels of cell wall ␣-glucan, and targeting of its sole AGS gene by RNA interference demonstrated directly that cell wall ␣-glucan is important for virulence of this pathogen (11). In the opportunistic yeast Cryptococcus neoformans, inhibition of expression of its sole AGS gene gives rise to acapsular cells, indicating that cell wall ␣-glucan plays a role in anchoring the capsule, a critical virulence factor for this pathogen (12).
Glycogen and starch synthases are glycosyltransferases that catalyze the formation of a (1,4)-␣-glucosidic bond and transfer the ␣-glucose moiety from UDP-glucose or ADP-glucose to the nonreducing end of a pre-existing (1,4)-␣-glucan primer. Based on amino acid sequence similarities, these synthases have been divided into two families, family 3 of glycosyltransferases (GT-3, with animal and fungal glycogen synthases) and family 5 of glycosyltransferases (GT-5, with archaeal and bacterial glycogen synthases and plant starch synthases) (13). Although these two families display only marginal sequence similarities, they appear to share certain structural and catalytic features (14). The recently reported crystal structures of the bacterial glycogen synthase GlgA from the bacterium Agrobacterium tumefaciens (15) and the Archaea Pyrococcus abyssi (16) provide a basis for our understanding of the catalytic mechanism of these synthases. Both the N-and C-terminal halves fold into a subdomain with a Rossmann-type fold, a classical structural motif characterized by a central ␤-sheet flanked by several ␣-helices. These subdomains are connected by a narrow hinge, creating a deep and wide central cleft between them. It has been proposed that, upon substrate binding, this cleft may close, bringing the nucleotide-glucose-binding motif (Lys/Arg)-X-Gly-Gly (17) on the N-terminal side of the cleft in close proximity to the catalytic motif Glu-X 7 -Glu (18) on the C-terminal side to create a functional active center (15,16).
In this study, we focused on the role of the putative intracellular domain of S. pombe Ags1p, denoted here as the synthase (SYN) domain. Our recent observation that cell wall ␣-glucan from S. pombe consists of ϳ10% (1,4)-linked ␣-glucose residues (19) prompted us to investigate whether the SYN domain is involved in the synthesis of this cell wall ␣-glucan constituent.
Here, we show that fungal Ags SYN domains share several sequence motifs with GT-3 and GT-5 enzymes, including the Glu-X 7 -Glu motif, the first Glu of which is a highly conserved catalytic residue. Furthermore, we show that (1,4)-␣-glucan accumulates in S. pombe cells overexpressing wild-type Ags1p, but not in S. pombe cells overexpressing the mutant Ags1p-E1526A, with the predicted catalytic residue mutated. Together, our data demonstrate that the S. pombe Ags1p SYN domain is involved in cell wall ␣-glucan biosynthesis by producing (1,4)-␣-glucan.

EXPERIMENTAL PROCEDURES
Strains and Culture Media-Escherichia coli strain DH5␣ (Invitrogen) was used for all plasmid isolations. The kanMX6-3nmt1 cassette (20) was integrated in front of the ags1 or mok11 open reading frame (ORF) at its chromosomal location in S. pombe wild-type strain 972. S. pombe plasmid transformations were performed using strains with genotype h Ϫ ura4-D18 or h Ϫ ags1-1 ts ura4-D18 using a lithium acetate method at pH 4.9 (21). Cells were grown in EMM2 medium (22) supplemented with 250 mg/liter adenine sulfate (EMMA medium). Following plasmid transformation, expression of cloned genes was repressed by growing the cells on EMMA plates supplemented with 100 M thiamine at 28°C (or at 21°C for ags1-1 ts strains) for 3-5 days. For overexpression experiments, cells were grown overnight at 28°C in EMMA medium containing 10 M thiamine, washed twice with EMMA medium (which lacks thiamine), and grown at 28°C in EMMA medium for 6 -24 h. The S. pombe strains used in this study are listed in Table 1.
Sequence Alignments, Phylogenetic Analyses, and Secondary Structure Prediction-The amino acid sequences used in this study are listed in supplemental Table S1. Sequencing projects used to obtain protein sequences were the Fungal Genome Initiative, Broad Institute of Harvard and MIT (www.broad.mit. edu); the United States Department of Energy Joint Genome Institute (www.jgi.doe.gov); the Institute for Genomic Research (www.tigr.org); and Genoscope (www.genoscope.cns.fr). The 79 selected sequences were aligned using the algorithms TCoffee (23) and ClustalW (24). The resulting multiple sequence alignments showed only minor differences between them, and the TCoffee alignment was chosen for phylogeny reconstruction based on the number of identities and conserved residues (supplemental Fig. S1). After removal of all positions containing gaps in Ͼ50% of the sequences using the program LIST-POS, 3 phylogenetic relationships were reconstructed using the neighbor-joining method (25), as well as FITCH (26), TREE-PUZZLE (27), and Bayesian (28) analyses. For the neighborjoining and FITCH methods, the programs NEIGHBOR and FITCH from the PHYLIP package (29) were used, respectively, and the distance matrices for these methods were calculated with the program PROTDIST using the Jones-Taylor-Thornton model. TREE-PUZZLE calculations were performed with the program TREE-PUZZLE, and Bayesian analysis was performed with the program MrBayes for 1,000,000 generations using a mixed model. The unrooted trees were rooted by the midpoint method using the program RETREE from the PHYLIP package, and the resulting rooted trees were converted to Scalable Vector Graphics format using SVGTREE (30). Neighbor-joining bootstrapping was performed by generating 1000 random bootstrap samples with the programs SEQBOOT, PROTDIST, and NEIGHBOR; trees generated from these samples were analyzed using the extended majority consensus method as implemented in the program CON-SENSE from the PHYLIP package. To calculate the percentage of identities in pairwise amino acid sequence alignments, the program GAP from the Wisconsin Package (Version 10.3; Accelrys, San Diego, CA) was used. Secondary structure predictions resulted from the program PSIPRED (Version 2.5) (31).

TABLE 1 S. pombe strains used in this study
All strains were constructed for this study, except strain 972 (P. Nurse, Rockefeller University, New York) and strain FH021 (7).

Strain
Genotype Construction of Plasmids-All PCR amplifications were performed using Phusion TM polymerase (Finnzymes). The primers used in this study are listed in supplemental Table S2. The ORF of ags1 (GenBank TM accession number AF061180) was cloned between the MluNI and BamHI restriction enzyme sites of pREP4 (32) using three fragments generated by PCR and two fragments subcloned from cosmid c17A7. Then, the SmaI site in the multiple cloning site of the resulting plasmid was removed, producing pAV011. The ORF of the A. tumefaciens glycogen synthase gene was amplified by PCR using primers AV098A and AV099A from plasmid pBG19 (kindly provided by R. A. Ugalde) (33), whereas the ORFs of the S. cerevisiae glycogen synthase gene GSY2 and its hyperactive version were amplified from genomic DNA of S. cerevisiae strain BY4741 using primers AV171 and AV172 and primers AV171 and AV173, respectively. These ORFs were cloned between the XhoI and BglII sites of the pREP4-based vector pAdH006 (34), generating pAV066, pAV111, and pAV112, respectively. To create pAgs1-⌬TGL, the fragment between the XhoI and SpeI sites in pAV011 was replaced with the Myc epitope tag-encoding hybridization product of oligonucleotides BS013 and BS014, producing pBS010. To create pAgs1-⌬SYN, a KpnI site was introduced in the ags1 ORF of pAV011 through introduction of the silent mutation C3351A following the QuikChange TM site-directed mutagenesis protocol (Stratagene) using primers AV219 and AV220, generating pAV127. Then, the fragment between the KpnI and SalI sites in pAV127 was replaced with the Myc epitope tag-encoding hybridization product of oligonucleotides AV221 and AV222, producing pAV128. To introduce defined mutations into the gene fragment encoding the SYN domain, the SmaI-SalI fragment was cloned into the pUC18 vector, and site-directed mutagenesis was performed using primers AV041 and AV042 (K1163Q), AV043 and AV044 (G1165A), AV045 and AV046 (E1526A), AV047 and AV048 (E1534A), or AV051 and AV052 (K1422Q). The SmaI-SalI fragment of pAV011 was replaced with the mutated SmaI-SalI fragments, generating pAV035, pAV036, pAV037, pAV038, or pAV040, respectively. To introduce the temperature-sensitive mutation (G696S) into the gene fragment encoding the TGL domain, the XhoI-EcoRI fragment was cloned into the pUC18-based vector pAdH100, and site-directed mutagenesis was performed using primers BS011 and BS012. The XhoI-SpeI fragment of pAV011 was replaced with the mutated XhoI-SpeI fragment, generating pAV137. All constructs used in this study were sequenced in both directions using a series of overlapping PCR amplification products (Big-Dye Terminator sequencing kit, Applied Biosystems). The plasmids used in this study are listed in Table 2.
Plate-based Assay for (1,4)-␣-Glucan Accumulation-Cells were grown in EMMA medium containing 10 M thiamine at 28°C to an optical density at 595 nm wavelength (OD 595 ) of 1.0 -1.5, washed, and taken up in Milli-Q water a final concentration of 2.5 ϫ 10 7 cells/ml. Four microliters of cell suspension (1 ϫ 10 5 cells) were spotted onto EMMA plates supplemented with or without 10 M thiamine. Plates were incubated at 28°C for 3 days and exposed to iodine vapor for 4 min.
(1,4)-␣-Glucan Determination-Cells (10 9 ) were taken up in 5 mM sodium azide and 20 mM Tris-HCl (pH 7.6), broken with acid-washed glass beads using FastPrep 120 (Bio 101, Inc.) at speed 6 for five intervals of 15 s, and centrifuged at 16,000 ϫ g for 10 min at 4°C. Cell pellets were resuspended in 2% (w/v) SDS, 100 mM EDTA, 20 mM dithiothreitol, and 50 mM Tris (pH 7.6); boiled for 10 min; and centrifuged again. The resulting pellets were washed twice with Milli-Q water and once in 10 mM sodium acetate (pH 5.6) and were divided equally over four fractions. Fractions were resuspended in 200 l of 10 mM sodium acetate (pH 5.6) and incubated at 37°C for 2.5 h in the presence of 1.8 units of ␣-amylase (catalog no. 102814, Roche Applied Science) and 6.0 milliunits of glucoamylase (catalog no. 1202332, Roche Applied Science), 0.1 unit of Zymolyase 100T (catalog no. 120493, Seikagaku Corp.), all three enzymes, or buffer only. In a control experiment, we confirmed that the amount of enzyme added was sufficient for complete digestion. After digestion, supernatants were collected, and the amount of liberated reducing ends was measured using a colorimetric assay (35). The reducing ends were converted to glucose equivalents; the background (buffer-only fraction) was subtracted; and the glucose equivalents/10 9 cells were corrected for the ␤-glucan amounts (Zymolyase 100T digestion). Digestion of (1,4)-␣-glucan was checked visually for completion by resuspending the remaining pellets in 100 l of Milli-Q water and adding 900 l of potassium triiodide reagent prepared from 0.01% (w/v) iodine and 0.1% (w/v) potassium iodide in 5 M calcium chloride.
Immunoblotting-Cells (2 ϫ 10 7 ) were taken up in 10% (w/v) trichloroacetic acid, broken with acid-washed glass beads using FastPrep 120 at speed 6 for three intervals of 15 s, incubated on ice for 10 min, and centrifuged at 16,000 ϫ g for 15 min at 4°C. The resulting pellets were washed with ice-cold acetone, resuspended in 200 l of SDS sample buffer containing 20 mM dithiothreitol, and incubated at 37°C for 10 min. Total cell lysates were separated on an SDS-6% polyacrylamide gel (or on an SDS-8% polyacrylamide gel for the ␣-tubulin control) under reducing conditions; blotted onto nitrocellulose membranes (0.45 m; Schleicher & Schüll); and probed with anti-␣-tubulin antibody (catalog no. T-5168, Sigma), mouse anti-Myc tag monoclonal antibody 9B11 (catalog no. 2276, Cell Signaling Technology), or anti-SYN domain antiserum SN269. This anti-SYN domain antiserum was generated in rabbits immunized

Plasmids used in this study
All plasmids were constructed for this study, except pREP4 (32).

Plasmid
Name Description with peptide H 2 N-CSQKYGRNSRSRSS-CONH 2 (based on residues 1618 -1631 of S. pombe Ags1p) coupled to keyhole limpet hemocyanin. Blots were incubated with horseradish peroxidase-conjugated goat anti-mouse IgG (Bio-Rad) or horseradish peroxidase-conjugated goat anti-rabbit IgG (Bio-Rad) and developed by chemiluminescence (ECL kit, Amersham Biosciences). Complementation Assay-Cells were grown in EMMA medium containing 100 M thiamine at 21°C to OD 595 ϳ 1.0, washed, and taken up in Milli-Q water to a final concentration of 2.5 ϫ 10 7 cells/ml. Four microliters of cell suspension were spotted in 10-fold serial dilutions onto EMMA plates containing 100 M thiamine. Plates were incubated at 28 or 36°C for 3-5 days.
␤-Glucanase Hypersensitivity Assay-Cells were grown in EMMA medium containing 100 M thiamine (repression conditions) at 21°C to OD 595 ϭ 1.0 -3.0, diluted with fresh EMMA medium containing 100 M thiamine to OD 595 ϭ 0.2, and incubated at 37°C for 4 h. Cells were washed once with 1 mM 2-mercaptoethanol and 10 mM Tris-HCl (pH 7.6) (digestion buffer) and taken up at OD 595 ϭ 0.2 in digestion buffer containing 0.4 units/ml of Zymolyase 20T (catalog no. 120491, Seikagaku Corp.). After a 1-h incubation at 28°C under continuous shaking, cell lysis was monitored by measurement of the optical density at 595 nm.

Phylogenetic Tree of the (1,4)-␣-Glucan Synthase Superfamily-
Close examination of the putative intracellular domain of S. pombe Ags1p (residues 1091-1986) revealed an N-terminal Lys-rich (19%) region (residues 1091-1144) and a C-terminal Ser-rich (15%) region (residues 1628 -1986), both of which flank a domain with amino acid sequence similarities to glycogen and starch synthases, referred to here as the SYN domain (Fig. 1A). The S. pombe Ags1p SYN domain displays ϳ26% amino acid sequence identity to the A. tumefaciens and P. abyssi GlgA proteins, the three-dimensional structures of which are known. Besides the genome of S. pombe, the genomes of 14 other ascomycetous fungi and three basidiomycetous fungi are known to encode Ags proteins. Their Ags SYN domains show ϳ70% amino acid sequence identity to one another, whereas they show ϳ27% identity to established GT-5 synthases, the archaeal and bacterial glycogen synthases and plant synthases (data not shown). To study the phylogenetic relationship of the fungal Ags SYN domains with glycogen and starch synthases, we compared their amino acid sequences using TCoffee (supplemental Fig. S1) and ClustalW (data not shown) and generated phylogenetic trees using the neighbor-joining method (Fig. 1B), as well as FITCH, PUZZLE, and Bayesian analyses (data not shown). Although minor differences were observed in the branching pattern within the major clusters, the mutual relationships among the major clusters were similar. The fungal Ags SYN domains clustered as a monophyletic group within the (1,4)-␣-glucan synthase superfamily, relating more closely to other GT-5 synthases than to GT-3 synthases (Fig. 1B). Furthermore, the previously described putative catalytic residues of E. coli GlgA and human muscle glycogen synthase are highly conserved in the Ags SYN domains (Fig. 2). These results sug- A, shown is a schematic representation of S. pombe Ags1p indicating its putative TGL and SYN domains, as well as the predicted N-terminal signal peptide (SP), putative transmembrane region (TM), and multipass transmembrane domain (PORE), which might form a pore. Relevant amino acid residues are indicated. B, a midpoint-rooted neighbor-joining tree was constructed for fungal Ags SYN domains, archaeal and bacterial glycogen synthases (GlgA), plant granule-bound (GBSS) and soluble (SS) starch synthases, and fungal and animal glycogen synthases (Gsy and Gys, respectively). The species and synthases included are listed in supplemental Table S1. Bootstrap support for this tree (%, upper number) was calculated from 1000 replicates. A similar topology was obtained using Bayesian analysis run for 1,000,000 generations, and the resulting partition probability values are indicated in the tree (%, lower number). The scale bar indicates the branch length corresponding to 0.2 amino acid substitutions/site. gest that the SYN domain may function as a (1,4)-␣-glucan synthase.
Ags1p Overexpression Causes (1,4)-␣-Glucan Accumulation-To investigate experimentally whether Ags1p synthesizes (1,4)-␣-glucan, we inserted the thiamine-repressible nmt1 promoter in front of the ags1 ϩ ORF in the genome of S. pombe wild-type strain 972. This well characterized promoter was chosen because it shows strong promoter activity when cells are grown in the absence of thiamine (induction conditions) (32). Furthermore, when grown in the presence of thiamine (repression conditions), cells maintain low expression levels of ags1 ϩ sufficient for normal vegetative growth as a result of residual promoter activity (36). To analyze these Pnmt1-ags1 ϩ cells, we developed a simple plate-based assay for the detection of (1,4)-␣-glucan accumulation. Approximately 100,000 cells were spotted onto plates with chemically defined culture medium (EMMA) with or without the addition of thiamine. Following incubation at 28°C for 3 days, the plates were developed by a 4-min exposure to iodine vapor, which stains (1,4)-␣-glucan, either branched (glycogen and starch) or unbranched (amylose). As a positive control, we used cells overexpressing the A. tumefaciens gene encoding GlgA under the control of the nmt1 promoter from the pREP4 plasmid for ectopic expression in S. pombe (32). Under repression conditions, these cells showed a yellowish staining (Fig. 3A, lower panel (LOW), lane 3) similar to wildtype cells of strain 972 or cells carrying an empty pREP4 plasmid (lanes 1 and 2). By contrast, under induction conditions, cells expressing A. tumefaciens GlgA showed an intense brown staining characteristic for accumulation of (1,4)-␣-glucan (Fig.  3A, upper panel (HIGH), lane 3). Remarkably, when we tested the Pnmt1-ags1 ϩ cells under induction conditions, we also observed a brown staining, indicating that also these cells accumulated (1,4)-␣-glucan (Fig. 3A, upper panel (HIGH), lane 5). This brown staining was due to ags1 overexpression because, when Pnmt1-ags1 ϩ cells were grown under repression conditions, the intensity of staining resembled that of control cells (Fig. 3A, lower panel (LOW), lane 5). Similarly, cells overexpressing the close ags1 ϩ homolog mok11 ϩ , which is expressed normally only during sporulation (8,37), also accumulated (1,4)-␣-glucan upon induction during vegetative growth (Fig.  3A, lane 6).
To demonstrate directly that the ags1 ϩ ORF was responsible for the (1,4)-␣-glucan accumulation, we cloned it behind the nmt1 promoter in the pREP4 plasmid and analyzed the ability of the resulting pAgs1 plasmid to induce (1,4)-␣-glucan accumulation in wild-type cells. In our plate-based assay, cells carrying plasmid pAgs1 showed a thiamine-repressible brown staining similar to that observed for Pnmt1-ags1 ϩ cells, indicating that both systems for ags1 ϩ overexpression resulted in comparable levels of (1,4)-␣-glucan accumulation (compare Fig. 3B, lane 2, with Fig. 3A, lane 5). Given that ags1 ϩ is an essential gene in S. pombe (7,8) and that, as a consequence, mutagenesis of the chromosomal copy could result in cell lysis, we chose the pREP4 expression system for the remainder of this study to overexpress mutant versions of Ags1p. We next analyzed whether overexpression of a functional version of Ags1p is required for (1,4)-␣-glucan accumulation. The ags1-1 ts mutant strain carries a single missense mutation in its ags1 gene, encoding Ags1p-G696S, the function of which is temperaturedependent (7). To analyze the ability of this Ags1p mutant to produce (1,4)-␣-glucan, we induced its overexpression in wildtype cells. Overexpression of Ags1p-G696S at 28°C, when this protein is functional, resulted in (1,4)-␣-glucan accumulation at levels observed for overexpression of wild-type Ags1p (Fig.  3C, compare lanes 2 and 3). By contrast, Ags1p-G696S overexpression at 36°C, when this protein is defective, did not result in (1,4)-␣-glucan accumulation (Fig. 3C, lane 5). We conclude that induction of functional Ags1p is required for accumulation of (1,4)-␣-glucan.
To confirm that iodine vapor was staining (1,4)-␣-glucan rather than other cell components such as lipids (38), we monitored the accumulation of (1,4)-␣-glucan by measuring the amount of glucose released from insoluble cell fractions after treatment with ␣-amylase and glucoamylase, endo-and exotype glucanases, respectively, specific for (1,4)-␣-glucan. Fol-  Table S1. Note that the first Glu is located in a loop (neither ␣-helix nor ␤-sheet), whereas the second Glu is predicted to be part of an ␣-helix (i.e. ␣13 in A. tumefaciens GlgA and P. abyssi GlgA).
lowing the shift of Pnmt1-ags1 ϩ cells from EMMA medium with thiamine to EMMA medium without thiamine, we analyzed cells at regular time intervals and observed a pronounced increase in the amount of cellular (1,4)-␣-glucan (Fig. 4A, upper  panel). In total lysates of matching cells, Ags1p overexpression levels were analyzed in immunoblot analyses using an antiserum directed against a peptide of the Ags1p SYN domain (residues 1618 -1631). The increasing levels of (1,4)-␣-glucan correlated closely with increasing levels of Ags1p, which resolved as a specific protein band at an apparent molecular mass of ϳ265 kDa (Fig. 4A, lower panel), consistent with the calculated molecular mass for Ags1p of 272.1 kDa. Under repression conditions, Pnmt1-ags1 ϩ cells maintained background levels of amylase-sensitive glucans (Fig. 4B, upper  panel), whereas Ags1p levels remained undetectable in immunoblot analyses using this antiserum (lower panel), as was the case for wild-type strain 972 (Fig. 4, A and B, lane 1). A similar correlation between Ags1p overexpression and (1,4)-␣-glucan accumulation was observed when Ags1p overexpression was driven from plasmid pAgs1 (Fig. 4C). Although induction was somewhat delayed, the maximum levels of cellular (1,4)-␣-glucan were similar to those observed for Pnmt1-ags1 ϩ cells (Fig.  4, compare A and C). Together, these results corroborate our assertion that overexpression of Ags1p results in accumulation of (1,4)-␣-glucan.

The Intracellular Domain of Ags1p Is Responsible for (1,4)-␣-Glucan Accumulation-In addition to the SYN domain, Ags1p
contains a putative extracellular domain predicted to exhibit transglycosylase activity (7), denoted here as the TGL domain. To explore the question of which domain was responsible for the observed (1,4)-␣-glucan accumulation, the SYN domain or the TGL domain, we deleted each domain separately and replaced them with a Myc epitope tag. First, we overexpressed Ags1p-⌬TGL (lacking residues 113-724) and observed induction of a protein with an apparent molecular mass of ϳ211 kDa, consistent with a calculated molecular mass of 203.6 kDa (Fig. 5, compare lanes 5 and 6). Clearly, Ags1p-⌬TGL was able to synthesize (1,4)-␣-glucan as determined in our plate-based assay (Fig. 3B, lane 4). Next, we overexpressed Ags1p-⌬SYN (lacking residues 1118 -1689, including the SYN epitope formed by residues 1618 -1631) and observed the appearance of an ϳ216-kDa protein, consistent with a calculated molecular mass of 209.5 kDa (Fig. 5, lanes 7 and 8). Notably, overexpression of Ags1p-⌬SYN did not result in accumulation of (1,4)-␣-glucan (Fig. 3B, lane 5). Together, these results demonstrate that the putative extracellular domain is not responsible for (1,4)-␣-glucan biosynthesis.
To investigate whether the SYN domain is involved in the synthesis of (1,4)-␣-glucan, we introduced mutations in several conserved residues of Ags1p. First, we examined the role of a putative catalytic residue (Glu 1526 ) that is conserved among all members of the (1,4)-␣-glucan synthase superfamily and that represents the first Glu in the Glu-X 7 -Glu motif (Fig. 2). Replacement of this Glu with Ala to create Ags1p-E1526A resulted in a complete reduction of brown straining in our plate-based assay (Fig. 3D, lane 2). Furthermore, no accumulation of (1,4)-␣-glucan was detected in our amylase assay, despite the overexpression of this mutant Ags1p (Fig. 6, lane 3). Similarly, substitution of the second Glu in the Glu-X 7 -Glu motif or the Lys or first Gly in the putative nucleotide-glucosebinding motif (Lys/Arg)-X-Gly-Gly reduced the brown staining in the plate-based assay (Fig. 3D, lanes 3-5), as well as (1,4)-␣glucan levels in the amylase assay (Fig. 6, lanes 4 -6), despite proper overexpression of the mutant Ags1 proteins. Interestingly, substitution of Gln for Ags1p Lys 1422 , whose equivalent (Lys 277 ) in E. coli GlgA was proposed to function as a catalytic residue (39), only modestly affected iodine staining in the platebased assay (Fig. 3D, lane 6), reducing (1,4)-␣-glucan accumulation by only ϳ50% during overexpression (Fig. 6, lane 7). In conclusion, these data support our hypothesis that the SYN domain of Ags1p is responsible for the synthesis of (1,4)-␣-glucan.
Given that many members of the (1,4)-␣-glucan synthase superfamily are dependent on a primer to initiate synthesis, we addressed whether such a primer is present in vegetatively grown S. pombe cells. We tested glycogen synthase II from S. cerevisiae (Gsy2p) as a reporter, which normally elongates (1,4)-␣glucan primers generated by glycogenin, but lacks the ability to initiate de novo (1,4)-␣-glucan synthesis (33,40). Overexpression of wildtype S. cerevisiae Gsy2p in S. pombe cells resulted in a brown staining, whereas low expression did not (Fig.  3E, lane 3). When we analyzed a hyperactive form of S. cerevisiae Gsy2p, Gsy2p-⌬643, which lacks the C-terminal phosphorylation sites for negative control of its synthase activity, brown staining was obtained even when expression levels were low (Fig. 3E, lane 4). These results demonstrate that S. cerevisiae Gsy2p is able to synthesize (1,4)-␣-glucan, indicating that S. pombe cells contain a primer for (1,4)-␣-glucan biosynthesis.
(1,4)-␣-Glucan Biosynthesis Is an Important Function of Ags1p-To determine whether (1,4)-␣-glucan biosynthesis is important for Ags1p function in vivo, we assessed the ability of mutant Ags1 proteins to complement the cell lysis and ␤-glucanase hypersensitivity phenotypes of the mutant strain ags1-1 ts (7). This temperature-sensitive strain grew well at a permissive temperature of 28°C, but unlike wild-type strain 972, it was unable to grow at a restrictive temperature of 36°C (Fig. 7A, compare rows 1 and 4). By contrast, cells of the ags1-1 ts strain carrying plasmid pAgs1 were able to grow at 36°C in the presence of thiamine, conditions that repress ags1 expression to low, wild type-like levels (Fig. 7A, row 6). Similarly, in a ␤-glucanase hypersensitivity assay in which the ags1-1 ts mutant was grown at 37°C for only 4 h to create a weakened cell wall structure at areas of new growth, plasmid-derived Ags1p was able to protect the ags1-1 ts mutant from lysis after exposure to purified ␤-glucanase (Table 3). Together, these data show that wildtype Ags1p can complement the cell lysis and ␤-glucanase hypersensitivity phenotypes of ags1-1 ts . We demonstrated that the temperature-sensitive version of Ags1p (Ags1p-G696S; used as a negative control) failed to complement the cell lysis phenotype of ags1-1 ts (Fig. 7B). Furthermore, nei- . Overexpression of Ags1p results in accumulation of (1,4)-␣-glucan. Cells were grown in EMMA medium without (high expression; ϪT) or with (low expression; ϩT) the addition of thiamine. At the indicated times following the start of induction, samples were taken. Upper panels, (1,4)-␣-glucan levels of the insoluble cell fraction were measured in an assay using ␣-amylase and glucoamylase. Lower panels, Ags1p levels were determined in total cell lysates by immunoblotting using anti-SYN domain antiserum. ␣-Tubulin (␣-tub) served as a loading control. The following cells were analyzed. A and B, cells of genotype Pnmt1-ags1 ϩ (strain ND025); C, cells containing pAgs1 (strain AV028); D, cells containing pREP4 (strain AV027). wt, wild type. Synthesizes (1,4) JUNE 29, 2007 • VOLUME 282 • NUMBER 26

JOURNAL OF BIOLOGICAL CHEMISTRY 18975
ther Ags1p-⌬TGL nor Ags1p-⌬SYN was able to complement the lysis phenotype (Fig. 7C), demonstrating that both domains are essential for Ags1p function in vivo.
Ags1p-E1534A, Ags1p-G1165A, Ags1p-K1422Q, and, to a lesser extent, Ags1p-K1163Q were found to be able to complement the cell lysis and ␤-glucanase hypersensitivity phenotypes of ags1-1 ts (Fig. 7D and Table 3). Notably, ags1-1 ts cells expressing Ags1p with a mutation in the putative catalytic Glu residue of the SYN domain (Ags1p-E1526A) were markedly reduced in their ability to grow at 36°C (Fig. 7D, row 3) and to protect the ags1-1 ts mutant from lysis after exposure to ␤-glucanase (Table  3). From these experiments, we conclude that Glu 1526 is important for Ags1p function in vivo.

DISCUSSION
In this study, we integrated the SYN domain of fungal ␣-glucan synthases into the superfamily of (1,4)-␣-glucan synthases, which already included glycogen and starch synthases. First, we demonstrated that Ags SYN domains display striking amino acid sequence similarities to archaeal and bacterial glycogen synthases and plant starch synthases, all belonging to the glycosyltransferase family GT-5 ( Fig. 2 and supplemental Fig. S1). Second, we showed that Ags SYN domains form a monophyletic group that clusters within the (1,4)-␣-glucan synthase superfamily (Fig. 1B), suggesting that a hypothetical ags1 ancestor gene existed before the divergence of the Ascomycota and Basidiomycota. Third, we demonstrated that overexpression of Ags1p in S. pombe cells resulted in accumulation of (1,4)-␣glucan, as detected by brown staining using iodine vapor (Fig. 3) and in a quantitative amylase assay (Fig. 4). Fourth, overexpression of Ags1p lacking the TGL domain but retaining the SYN domain induced accumulation of (1,4)-␣-glucan, whereas Ags1p without the SYN domain did not (Fig. 3). Fifth, replacement of a highly conserved and putative catalytic residue in the SYN domain (Ags1p-E1526A), the first Glu in the Glu-X 7 -Glu motif, abrogated (1,4)-␣-glucan synthesis. Similarly, mutations in the putative nucleotide-glucose-binding motif (K1163Q and G1165A) resulted in a major decrease in (1,4)-␣-glucan accumulation upon overexpression (Figs. 3 and 6). Finally, synthesis of (1,4)-␣-glucan was found to be an important function of Ags1p in vivo because cells dependent on Ags1p-E1526A grew slower than cells dependent on wild-type Ags1p, and their cell wall structure was weakened dramatically (Fig. 7 and Table 3). On the basis of these results, we conclude that, of the two probable catalytic domains of S. pombe Ags1p, the SYN domain is responsible for the production of the (1,4)-␣-glucan constituent of cell wall ␣-glucan.
The Glu-X 7 -Glu motif is conserved among archaeal and animal glycogen synthases, as well as fungal Ags SYN domains, whereas in bacterial glycogen synthases and plant starch synthases, only the Glu equivalent to the first Glu of  the Glu-X 7 -Glu motif is conserved (Fig. 2). Substitution of this Glu with Ala in E. coli GlgA (E377A) resulted in a 10,000-fold reduction of the catalytic activity without affecting its ability to bind the ADP-glucose or glycogen substrates, demonstrating that this residue is involved in catalysis rather than substrate binding (41). However, in the glycosyltransferase Gpi3p of S. cerevisiae and S. pombe, the second Glu of the Glu-X 7 -Glu motif is of greater importance than the first for in vivo function (42). To address the question of which of the two Glu residues of the Glu-X 7 -Glu motif is more important for the S. pombe Ags1p SYN domain, we mutagenized them individually. Replacement of the first Glu in Ags1p-E1526A abrogated (1,4)-␣-glucan production (Figs. 3 and 6) and dramatically reduced Ags1p function in vivo ( Fig. 7 and Table 3). By contrast, replacement of the second Glu in Ags1p-E1534A reduced (1,4)-␣-glucan accumulation upon overexpression, but did not result in a decrease in the function of Ags1p in vivo. Therefore, it appears that the first Glu of the Glu-X 7 -Glu motif is of greater importance than the second for the S. pombe Ags1p SYN domain. Attempts to definitively answer this question were thus far unsuccessful because we were unable to generate ags1⌬ strains expressing mutant versions of Ags1p from plasmid. For example, when we allowed ags1⌬/ags1 ϩ diploid cells carrying plasmid pAgs1 to sporulate, the ags1⌬ ascospores lysed during germination (apart from a few exceptions), presumably because Ags1p plays a critical role during germination and because the nmt1 promoter of pAgs1 failed to supply proper levels of Ags1p during this process.
Besides the catalytic Glu-X 7 -Glu motif, the nucleotideglucose-binding motif (Lys/Arg)-X-Gly-Gly is conserved among all members of the (1,4)-␣-glucan synthase superfamily, including the Ags SYN domains. Replacement of the equivalent residues (Lys 1163 and Gly 1165 ) showed that they play a role in (1,4)-␣-glucan production by S. pombe Ags1p (Figs. 3 and 6). However, we found no evidence for a catalytic role of Ags1p Lys 1422 . The equivalent Lys in E. coli GlgA (Lys 277 ) was predicted to be involved in catalysis (39), but our results would be consistent with more recent data on the crystal structure of A. tumefaciens GlgA, which suggested that this residue has a more distant effect on catalysis (15). Definitive identification of the catalytic residues of the Ags1p SYN domain will have to await the development of an in vitro assay for Ags1p activity. To our knowledge, no such assay has been described, and despite numerous attempts, we were unsuccessful in developing such an assay either with isolated membranes containing native Ags1p or with recombinant versions of the Ags1p SYN domain. Nonetheless, the results presented here suggest that the catalytic mechanism for the S. pombe Ags1p SYN domain appears to be similar to that of other members of the (1,4)-␣-glucan synthase superfamily.
Our previous finding that cell wall ␣-glucan is composed of two distinct constituents, (1,3)-␣-glucan and (1,4)-␣-glucan oligosaccharides (19), combined with our present finding that the SYN domain of Ags1p is able to synthesize the latter constituent, raises the question of which enzyme or enzyme domain produces the (1,3)-␣-glucan constituent. We cannot exclude the unlikely possibility that the SYN domain might be capable of catalyzing two distinct reactions, synthesizing not only (1,4)-␣-glucan but also (1,3)-␣-glucan. Also, the TGL domain of Ags1p seems to be an improbable candidate for (1,3)-␣-glucan synthesis because it has sequence similarities to bacterial and fungal amylases of the glycoside hydrolase family GH-13, enzymes known to hydrolyze or transglycosylate (1,4)-␣-glucan. Therefore, we hypothesize that an unidentified enzyme may be responsible for synthesis of the (1,3)-␣-glucan constituent. If so, this would require a further refinement of our speculative model for cell wall ␣-glucan biosynthesis (7,19). In this refined model, the (1,4)-␣-glucan oligosaccharides of cell wall ␣-glucan would be produced by the Ags1p SYN domain. The degree of polymerization of the (1,4)-␣-glucan produced by Ags1p during overexpression appears to be between 10 and 40, based on the red-brown color of the cells observed after exposure to iodine vapor (46). Interestingly, this size range includes the estimated length of the (1,4)-␣-glucan oligosaccharides observed in cell wall ␣-glucan, which are predicted to consist of ϳ12 (1,4)-linked ␣-glucose residues (19). The (1,4)-␣-glucan oligosaccharides might serve as a primer for the unknown (1,3)-␣-glucan synthase for elongation. The resulting building block would then be transported across the membrane via the putative pore domain of Ags1p and be coupled to another building block by the TGL domain, thereby forming mature cell wall ␣-glucan.
On the basis of the strict conservation of the SYN domain among fungal Ags proteins, we predict that, not only S. pombe, but indeed all fungi expressing an Ags1p homolog may contain (1,4)-linked ␣-glucose residues coupled to cell wall (1,3)-␣-glucan. Although the presence of (1,4)-␣-glucosidic linkages was noticed in cell wall ␣-glucan fractions from both ascomycetous and basidiomycetous fungi such as the filamentous fungus Aspergillus niger (47) and the medically important pathogen C. neoformans (48), additional work will be required to demon-strate that these linkages, rather than representing a contamination of glycogen, are linked covalently to (1,3)-␣-glucan. We hope that our present findings may contribute to a rational design for antifungal drugs directed against Ags proteins.