The Dual Activity Responsible for the Elongation and Branching of β-(1,3)-Glucan in the Fungal Cell Wall

ABSTRACT β-(1,3)-Glucan, the major fungal cell wall component, ramifies through β-(1,6)-glycosidic linkages, which facilitates its binding with other cell wall components contributing to proper cell wall assembly. Using Saccharomyces cerevisiae as a model, we developed a protocol to quantify β-(1,6)-branching on β-(1,3)-glucan. Permeabilized S. cerevisiae and radiolabeled substrate UDP-(14C)glucose allowed us to determine branching kinetics. A screening aimed at identifying deletion mutants with reduced branching among them revealed only two, the bgl2Δ and gas1Δ mutants, showing 15% and 70% reductions in the branching, respectively, compared to the wild-type strain. Interestingly, a recombinant Gas1p introduced β-(1,6)-branching on the β-(1,3)-oligomers following its β-(1,3)-elongase activity. Sequential elongation and branching activity of Gas1p occurred on linear β-(1,3)-oligomers as well as Bgl2p-catalyzed products [short β-(1,3)-oligomers linked by a linear β-(1,6)-linkage]. The double S. cerevisiae gas1Δ bgl2Δ mutant showed a drastically sick phenotype. An ScGas1p ortholog, Gel4p from Aspergillus fumigatus, also showed dual β-(1,3)-glucan elongating and branching activity. Both ScGas1p and A. fumigatus Gel4p sequences are endowed with a carbohydrate binding module (CBM), CBM43, which was required for the dual β-(1,3)-glucan elongating and branching activity. Our report unravels the β-(1,3)-glucan branching mechanism, a phenomenon occurring during construction of the cell wall which is essential for fungal life.

T he fungal cell wall plays an important role in maintaining cell shape and integrity and protects the fungal cells from the internal turgor pressure as well as from the external environment (1,2). This cell wall has a central core composed of a branched ␤-(1,3)-glucan to which other structural polysaccharides are bound (3)(4)(5)(6). Studies have suggested that interlinking between these different polysaccharides is essential for the strength and flexibility of the cell wall (6,7).

Solubilization of the S. cerevisiae cell wall alkali-insoluble (AI) fraction with recombinant endo-␤-(1,3)-glucanase releases both linear and nonlinear ␤-(1,3)oligosaccharides.
LamA [a recombinant endo-␤-(1,3)-glucanase (7,25)] was employed to solubilize the AI fraction from the wild-type S. cerevisiae strain (BY4741). Solubilized material resolved into five major peaks upon high-performance anion-exchange chromatography (HPAEC; Dionex) (Fig. 1A). The first peak and the second peak corresponded to the retention times of glucose and laminaribiose [L 2 , two glucose units joined by a ␤-(1,3)-glycosidic linkage], respectively, and the peak eluted at the end of the gradient run was ␤-(1,6)-glucan (7). The elution times of other two peaks (Br 1 and Br 2 ) did not correspond to linear ␤-(1,3)-oligomers of degrees of polymerization (DPs) of 2 (DP2) to DP6. Upon thin-layer chromatography (TLC), these two peak fractions migrated with R f values intermediary to those calculated for ␤-(1,3)-oligotriose-tetraose and ␤-(1,3)-oligotetraose-pentaose (Fig. 1B). Taken together, these results suggested that two additional peaks did not represent linear ␤-(1,3)-oligomers. To characterize them, the peak fractions were purified by gel permeation chromatography on a Biogel P2 column and subjected to nuclear magnetic resonance (NMR) analyses. Figure 1C presents one-dimensional (1D) 1 H-edited and two-dimensional 13 C-edited gradient heteronuclear single-quantum correlation spectroscopy (gHSQC) spectra; the anomeric region contained 12 signals corresponding to sugar residues that were arbitrarily labeled A, B, C, D, E, F, G, H, I, J, K, and L. Nearly all the 1 H and 13 C resonances could be assigned, although some signals overlapped heavily (see Tables S1 and S2 in the supplemental material). The chemical shifts and 1 H, 1 H coupling constant analyses confirmed that all residues corresponded to glucose. The 3 J H1,H2 and 1 J H1,C1 coupling constant values revealed that all residues except A and B were ␤-anomers. In the 13 C-edited gHSQC experiment, two distinct methylene groups can be observed within downfield-shifted H6,6=/C6 at 3.86 to 4.21/71.43 ppm assigned to G and H glucose residues, revealing that both glucose residues were 6-substituted (data in bold in Table S1 in the supplemental material). Furthermore, downfield-shifted C3 carbons characteristic of 3-substituted glucose residues (between 85 and 88 ppm) (Table S1, in bold) were observed for all glucose residues except for the E, F, and L residues identified as nonreducing-end residues from the absence of a chemical shift and for 6-substituted glucose residues (G and H) indicating the absence of disubstituted glucose residue.
Strong interactions between the anomeric proton (4.510 ppm) of the L-glucose residues and H6/H6= protons (3.855 to 4.206 ppm) of the G and H glucose residues were observed in the rotating-frame Overhauser effect spectroscopy (ROESY) experiment, corroborated by the presence in the gradient heteronuclear multiple-bond correlation spectroscopy (gHMBC) experiment of an H1/C6 correlation (4.510/71.43 ppm) between these residues, indicating the L (1 ¡ 6) G and L (1 ¡ 6) H sequence motifs that are ␤-Glcp-(1¡6)-␤-Glcp-(1¡ (data not shown). Dipolar interactions were also observed between the anomeric proton of the G residue (4.704 ppm) and the H3 proton of the I residue (3.734 ppm), suggesting the G (1¡3) I sequence motif that corresponds to ¡6)-␤-Glcp-(1¡3)-␤-Glc. This linkage was confirmed by the results of the gHMBC experiment with . Lane 1, glucose/laminarioligo standards L 2 to L 5 ; lanes 2 and 3, purified radiolabeled branched trimer and tetramer, respectively. All samples were run on the same TLC plate. Lane 1 was separated and revealed by orcinol-H 2 SO 4 treatment, whereas lanes 2 and 3 were subjected to autoradiography. After the samples were revealed, lane 1 was aligned with lanes 2 and 3 based on the sample application points on the TLC plate before migration. (F) Branching activity is localized in the cell wall. The cytosolic fraction did not incorporate radioactivity upon incubation with UDP-( 14 C)glucose followed by LamA treatment. (G) Membrane fractionation released glucose and laminaribiose, and the cell wall fraction profile showed peaks corresponding to branched oligomers (gradient run I, radiometric detection).
Neosynthesis of branched ␤-(1,3)-glucan by permeabilized S. cerevisiae. Permeabilized S. cerevisiae could incorporate radioactivity into the neosynthesized cell wall (7). The LamA-solubilized AI fraction of permeabilized cells incubated with UDP-( 14 C)glucose showed a chromatography profile (Fig. 1D) similar to that from the intact S. cerevisiae strain (Fig. 1A). Biogel P2 column-purified neosynthesized nonlinear peaks corresponded to branched ␤-(1,3)-oligosaccharides by TLC (revealed by autoradiography; Fig. 1E). This result showed that the permeabilized cells could also introduce ␤-(1,6)-linkages on the linear ␤-(1,3)-chains. Branched ␤-(1,3)-oligomers could be detected after 15 min of incubation of the permeabilized cells with UDP-( 14 C)glucose; the branching percentage increased until 1 h, after which it became stationary due to saturation of the enzyme activity. Optimal incorporation was obtained at pH 7.5 and between 25 and 30°C. Removal of ATP, GTP, and EDTA from the buffer reduced glucose incorporation into branching units (data not shown).
Further, S. cerevisiae cells were disrupted to separate the cytosolic, membrane, and cell wall fractions and incubated with UDP-( 14 C)glucose individually. There was no radioactivity incorporated with the cytosolic fraction, and the membrane fraction could synthesize only linear ␤-(1,3)-glucan (Fig. 1F). In contrast, in the cell wall fraction (which contained plasma membrane fragments), branched ␤-(1,3)-oligomers were detected (Fig. 1G), suggesting that branching occurs in the cell wall after the initial synthesis of linear ␤-(1,3)-glucan at the plasma membrane.
In terms of growth, the bgl2Δ mutant was similar to the wild-type strain and the gas1Δ mutant showed a 20% to 25% decrease, whereas the gas1Δ bgl2Δ mutant showed a 75% to 80% decrease after 48 h (Fig. 3A). Cell wall analyses of the single gas1Δ and the double gas1Δ bgl2Δ mutant strains showed altered compositions, while the bgl2Δ cell wall composition was comparable to that of the wild-type strain (Fig. 3B). The gas1Δ and gas1Δ bgl2Δ mutants showed a significant (2-fold to 8-fold) increase in the amount of cell wall chitin content (represented by GlcN) and a 2-fold decrease in the AI fraction glucose level. The amorphous (AS) fraction compositions were not altered by the single GAS1 and the double GAS1 BGL2 gene deletions. Calcofluor white (CFW) staining of the cells was more intense for the gas1Δ and gas1Δ bgl2Δ mutants than for the parental strain, supporting the data indicating increased cell wall chitin levels (Fig. 3C). CFW staining also showed dispersed distribution of bud scars on the gas1Δ and gas1Δ bgl2Δ mutant surfaces, in contrast to the presence of polarized bud scars on the wild-type strain and bgl2Δ mutant surfaces, suggesting a disorganized cell wall leading to nonpolarized budding in the gas1Δ and gas1Δ bgl2Δ mutants. In addition, the gas1Δ and gas1Δ bgl2Δ mutant cells were, respectively, 1.5-fold Ϯ 0.3-fold and 1.7-fold Ϯ 0.4-fold bigger than the wild-type strain and bgl2Δ mutant cells. Additionally, the gas1Δ bgl2Δ mutant and, to a lesser extent, the gas1Δ mutant were sensitive to cell wall-perturbing compounds such as Congo red and CFW (Fig. 3D).
Gas1p and Bgl2p are responsible for the branching of the cell wall ␤-(1,3)glucan in yeast. As the minimum length of ␤-(1,3)-oligomers required for Gas1p activity was 11 monomeric units (12), we tested ␤-(1,3)-oligomers with a degree of polymerization (DP) of 11 and above. Similar to previous reports (21,26), Bgl2p did introduce a ␤-(1,6)-linkage on the DP11 oligomer (Fig. 4A). Incubation of Gas1p with the DP11 oligomer initially resulted in the elongation of the ␤-(1,3)-chain (12), but during the later time course (Ͼ20 h of incubation) it introduced ␤-(1,6)-branching on the reaction products (Fig. 4B). Results of reactions performed with DP15 and DP24 were similar to those seen with DP11, but branching signals were detectable at earlier incubation times with increasing DPs of the oligomeric substrate: the first branching signal was observed when DP24 and DP15 were incubated with Gas1p for 12 h and 16 h, respectively (Fig. 4C, profile for DP24). This result suggested that an increase in the ␤-(1,3)-oligosaccharide length decreases the time required to introduce ␤-(1,6)-linkages on them by Gas1p. A medium pH range tested (pH 5.0 to 7.0) had no effect on the Gas1p branching activity. An increase in the branch signal upon coincubation of Gas1p and Bgl2p (Fig. 4D) compared to the results seen with their individual incubation with DP11 confirmed their cooperativity in ␤-(1,3)-glucan branching activity.

DISCUSSION
␤-Glucan is the major constituent of fungal cell wall, with its amount ranging between 30% and 80% of the cell wall dry mass depending on the fungal species (2).  (27). Branching ramifies ␤-(1,3)-glucan, facilitating its binding with other cell wall components, and hence it is considered to be essential for the cell wall architecture (3). Until now, ␤-(1,3)-glucan branching was a mystery, as membrane preparations synthesized only linear ␤-(1,3)-glucan in vitro (7,9). In the present study, we showed that mature ␤-(1,3)-glucan biosynthesis requires the presence of both cell wall and membrane fractions and that branching in the yeast S. cerevisiae occurs due to cooperative activity of two glycosyltransferases, Gas1p and Bgl2p, previously shown to display a unique genetic interaction (http://www.yeastgenome.org/locus/S000004924/interaction).
The gas1Δ mutant, in agreement with earlier reports (28,29), showed a mild growth defect and altered cell morphology with spherical cells and dispersed bud scars, unlike  the wild-type strain, which showed an ellipsoidal shape and bud scars concentrated at one pole. The double mutant also showed enlarged spherical cells with dispersed bud scars. In the gas1Δ and gas1Δ bgl2Δ mutants, there were 90% and 98% decreases in the cell wall ␤-(1,6)-glucan content, respectively, compared to the wild-type strain. Orlean (6) reported that the S. cerevisiae cell wall is organized in the order ␤-(1,3)-glucan¡␤-(1,6)-glucan¡mannoproteins. However, first, we did not find ␤-(1,6)-glucan in the gas1Δ bgl2Δ mutant cell wall, suggesting that the ␤-(1,3)-glucan ramification is essential for ␤-(1,6)-glucan attachment. Second, according to the described organization order, mannoproteins must be present in the fibrillar (AI) fraction of the cell wall and the absence of ␤-(1,6)-glucan must result in the loss of mannoproteins from the cell wall due to the lack of an anchoring structure. However, in our study, we extracted mannan mainly in the amorphous (AS) fraction of the wild-type strain and its amount in the mutants was similar to that in the wild-type strain, indicating that mannan is not covalently bound to the other cell wall components. Magnelli and coworkers were also able to extract mannan in the alkali-soluble (AS) fraction (30), whereas Ballou reported its extraction using citrate buffer (31), which supports our observation and the idea that, as a mannoprotein, mannan occurs as a fibrillar outer layer in the yeast cell wall (32). When the gas1Δ bgl2Δ mutant culture supernatant was analyzed, we did not find ␤-(1,6)-glucan, suggesting that ␤-(1,3)-glucan branching indeed affects ␤-(1,6)-glucan biosynthesis. There is still a debate about the site and mechanism of ␤-(1,6)-glucan biosynthesis (33)(34)(35)(36). But our study results suggest that the order of cell wall construction is branched ␤-(1,3)-glucan¡␤-(1,6)-glucan and reinforce the speculations published by earlier researchers that the maturation of ␤-(1,6)-glucan occurs in the cell wall (37). Moreover, the lack of ␤-(1,6)-glucan in the gas1Δ bgl2Δ mutant cell wall and the fact that the presence of proteins involved the ␤-(1,6)-glucan biosynthesis in the cell wall and/or plasma membrane-associated forms (38)(39)(40)(41)(42) led to the speculation that (i) the biosynthesis and maturation of ␤-(1,6)-glucan occur in the cell wall and (ii) Kre5p functions as a chaperon for the proteins involved in the ␤-(1,6)-glucan biosynthesis (36) rather than being involved in the synthesis of nascent ␤-1,6-glucan chains (42,43).  (46,47), respectively], suggesting that glucan-chitin linkage may not be an essential part of the cell wall fibrillar core.
Gas1p and Bgl2p are among the best-characterized glycosyltransferases (48). Our study data have allowed a better understanding of their role. The phenotypes of the single gas1Δ and double gas1Δ bgl2Δ mutants the we analyzed were in agreement with the observations by Plotnikova et al. (48) indicating that Gas1p and Bgl2p are functionally related. Bgl2p is one of the most abundant cell wall proteins and is able to introduce ␤-(1,6)-linkages on ␤-(1,3)-glucan (24). However, Bgl2p preferred shorter ␤-(1,3)-oligomers, as there was a decrease in its activity upon an increase in the length of the oligomeric substrate (see Fig. S5 in the supplemental material). Initially, recombinant Gas1p showed ␤-(1,3)-elongase activity followed by the introduction of ␤-(1,6)linkages on the ␤-(1,3)-glucan, suggesting that branching activity of Gas1p is dependent on the elongation of the ␤-(1,3)-glucan chain that generates an appropriate substrate for branching. In support of this hypothesis, with ␤-(1,3)-oligomers of greater chain length, there was a shorter incubation time before the appearance of branches. There was a significant increase in the branching when Gas1p and Bgl2p were incubated together with ␤-(1,3)-oligomers, suggesting their cooperative branching activity. Bgl2p preferring shorter ␤-(1,3)-oligomers and Gas1p elongating ␤-(1,3)-oligomers prior to its ␤-(1,6)-branching activity suggest the hypothesized mechanism of branching activity depicted in Fig. 8. Supporting our model, the branching signal seen in the LamA-digested AI fraction from the gas1Δ mutant could be destroyed completely upon  prior periodate oxidation-Smith degradation of the AI fraction, whereas the AI fractions from the wild-type and bgl2Δ mutant strains were resistant to such treatment, with the wild-type strain showing less than 10% destruction of the branched trimer (Fig. S6). This result confirms that, in the wild-type strain, Bgl2p introduces less than 15% of linear ␤-(1,6)-linkage on the short ␤-(1,3)-oligomers synthesized by the plasma membranebound glucan synthase complex, which could be destroyed by periodate oxidation-Smith degradation. In contrast, in the gas1Δ mutant, possibly more short ␤-(1,3)oligomers are available for Bgl2p to introduce ␤-(1,6)-linkages ( Fig. 2A; 30% instead of the 10% to 15% branching introduced by Bgl2p in the wild-type cell wall) due to the lack of Gas1p activity that initially elongates shorter ␤-(1,3)-oligomers synthesized by the glucan synthase complex; as these Bgl2p introduced ␤-(1,6)-linkages are linear, they could be completely destroyed by periodate oxidation-Smith degradation (Fig. S6).
The dual activity seen in our study with GH72 family fungal glycosyltransferases carrying a CBM is not an exception in biology. Adenylosuccinate lyase from Thermotoga maritima, which forms a homotetramer, catalyzes the addition of nitrogen at two different positions of AMP in a reaction involving the beta-elimination of fumarate (49). Its dual activity is attributed to a single 180°-bond rotation in the substrate between the first and the second enzymatic activities. AmiA, a chlamydial enzyme, acts both as a carboxypeptidase and an amidase, the former activity being associated with the presence of a penicillin-binding protein motif (50). A 175-kDa enzyme from Candida utilis showed trehalase-sucrase activity (51). In the present study, only those GH72 family members with a CBM showed dual activity, suggesting that proper positioning of the substrate by a CBM is essential. We did attempt to delete the CBM from GAS family members; however, such a deletion where a CBM is comprised of 90 to 100 amino acids (http://www.cazy.org/) resulted in the complete loss of both elongating and branching activity, possibly due to the loss of active enzyme structure.
Construction of gas1⌬ bgl2⌬ double deletion mutant strain. Primers used are listed in the Table  S4 in the supplemental material. GAS1 was deleted in the bgl2Δ strain by chromosomal integration of an 893-bp nourseothricin (NAT) PCR fragment. The integrated product was PCR amplified from the pFA6a-natNT2 plasmid DNA (containing the nourseothricin marker) using primers LB-GAS1DEL-FnatNT2 and LB-GAS1DEL-RnatNT2, including part of the GAS1 promoter and terminator regions, with the following program: 30 s at 98°C followed by 30 cycles of 10 s at 98°C, 30 s at 45°C, and 30 s at 72°C. The bgl2Δ strain was transformed with this construct according to the LiOAc method, and the yeast chromosomal DNA was extracted according to protocols described elsewhere (52). A 890-bp PCR fragment was amplified with primers GAS1ctrlF-PROM and natNT2REV (homologous to GAS1 promoter and NAT sequences), and a 1,239-bp PCR fragment was amplified using primers natNT2FOR and GAS1ctrlR-TERM (homologous to NAT and GAS1 terminator sequences) for five clones, confirming the deletion of GAS1 gene in the bgl2Δ strain. Each of the five clones was able to grow on YPD medium containing geneticin (300 g/ml) or nourseothrycin (100 g/ml) or both. Two of the five clones were used for the entire study. The Fungal Cell Wall ␤-(1,3)-Glucan Branching ® duplication of GAS1 in these two clones has also been ruled out using the primers GAS1geneF and Gas1geneR (primers with the sequences inside GAS1; Table S4). GAS1verif1 and GAS1verif2 (primer sequences outside the deletion cassette) were used to verify ectopic integration of the GAS1 deletion cassette; a 2,860-bp band was observed for the wild-type strain and a 2,000-bp band for the gas1Δ bgl2Δ mutant, confirming the presence of the nourseothricin deletion cassette at the right locus.
S. cerevisiae permeabilization, cell wall fractionation and solubilization, and periodate oxidation and characterization. Permeabilization, alkali-insoluble (AI) fraction extraction from the cell wall, its solubilization using endo-␤-(1,3)-glucanase, periodate oxidation-Smith degradation, high-performance anion-exchange chromatography (HPAEC/Dionex), thin-layer chromatography, and low-pressure liquid chromatography disruption of the cells to obtain cytosolic, membrane, and cell wall fractions were performed as described earlier (7). Dionex profiling was performed using PA1 and PA200 CarboPAC columns (Thermo-Fisher Scientific); the gradient run (flow rates, 1 ml/min for PA1 and 0.350 for PA200) was performed using solvent A (50 mM NaOH) and solvent B (500 mM sodium acetate-50 mM NaOH) as follows: for gradient run I, 0 to 2 min, isocratic ( B (gradients I and II were used for the PA1 column; gradient III was used for the PA200 column). Samples were detected on a pulsed electrochemical detector (PED; for nonradiolabeled samples) or using a radiometric detector (Packard Radiomatic Flo-One, equipped with a 500-l liquid-type cell) for 14 Cradiolabeled samples. 14 C-radiolabeled compounds were detected at 156 keV with a liquid scintillation flow rate of 2.0 ml/min. The representative Dionex profiles shown in the figures are reproducible, as each experiment was performed at least 3 to 5 times, sometimes over 10 times.
Nuclear magnetic resonance (NMR) spectroscopy. NMR spectra were acquired at 288 K on Varian Inova spectrometers operating at proton frequencies of 500 MHz and 600 MHz equipped with a triple-resonance 1 H( 13 C/ 15 N) Triax gradient probe and a cryogenically cooled triple-resonance 1 H( 13 C/ 15 N) pulsed-field gradient (PFG) probe, respectively. Sample lyophilized repeatedly in D 2 O was dissolved in 420 l D 2 O (99.97% 2 H atoms) (Euriso-top, CEA, Saclay, France) and transferred into a 5-mm-diameter Shigemi tube (Shigemi Inc., Alison Park, USA). 1 H chemical shifts were referenced to external DSS (4,4-dimethyl-4-silapentane-1-sulfonic acid; its methyl resonance was set to 0 ppm). 13 C chemical shifts were then calculated from the 1 H chemical shift and gamma ratio relative to DSS. A 13 C/ 1 H gamma ratio of 0.251449530 was used (53). The following nucleus assignment strategy was adopted. First, the proton resonances were assigned using a two-dimensional correlation spectroscopy (COSY) experiment (20). A relayed COSY experiment (RELAY) with one and two relays of 60 ms allowed us to follow connectivities from the anomeric proton up to the H4 proton of the glycosidic residues (54,55). The intraglycosidic residue spin systems were completed by means of a total correlation spectroscopy (TOCSY) experiment with a long mixing time (100 ms) (56). Second, a 1 H-13 C gradient heteronuclear single-quantum correlation spectroscopy (gHSQC) experiment and a 1 H-13 C gHSQC-TOCSY experiment with a mixing time of 80 ms allowed achieving assignment of 13 C chemical shifts from previously identified 1 H resonances (57). In addition, the CH2 groups were easily identified from the 13 C-edited gHSQC experiment. Then, analysis of 1 H, 1 H coupling constants ( 3 J H1, H2 ) from a 1D and/or COSY spectrum ( 1 H resolution of 0.1 Hz and 1.0 Hz, respectively) assessed the monosaccharide residue identity. Moreover, the anomeric configuration of monosaccharide residues was established from knowledge of 3 J H1, H2 values and confirmed via the measurement of the 1 J C1, H1 heteronuclear coupling constants in the 1 H dimension of the gradient heteronuclear multiple-bond correlation spectroscopy (gHMBC) spectrum ( 1 H resolution of 1.4 Hz) (57). Finally, glycosidic linkages were established via through-space dipolar interactions using a 1 H, 1 H rotating-frame Overhauser effect spectroscopy (ROESY) experiment (mixing time of 250 ms) and/or via three-bond interglycosidic 1 H, 13 C correlations using a 1 H, 13 C gHMBC experiment (long-range delay of 60 ms).
In situ branching activity assay. The assay mixture (in a total volume of 67 l) contained 50 mM Tris-HCl buffer (pH 7.5), 0.5 mM UDP-( 14 C)glucose (specific activity, 34 nmol/125 nCi in the final reaction mixture), 0.2 mM ATP, 20 M GTP␥S, EDTA (1 mM), and permeabilized cells (5 ϫ 10 8 cells) at room temperature. Neosynthesized polysaccharides were precipitated overnight by the use of two volumes of cold ethanol (Ϫ20°C). The precipitate thus obtained was washed three times with water (500 l each time) and treated twice with 500 l of 1 M NaOH containing 0.5 M NaBH 4 at 65°C for 1 h. The AI fraction was collected by centrifugation at 10,000 ϫ g for 10 min, washed with water, neutralized using acetic acid, and subjected further to LamA digestion. Incorporation of the radioactivity was measured at each step using a Wallac 1410 liquid scintillation counter (PerkinElmer Life Sciences).
Production of recombinant Gas1p and Bgl2p. Recombinant Gas1p was produced using a Pichia pastoris expression system (12). For Bgl2p, the amino acid sequence (CAA97313.1) was back-translated into a nucleotide sequence that was codon optimized for expression in Escherichia coli. The gene was synthesized with flanking NdeI and XhoI restriction sites (GeneArt; Life Technologies, Inc.). The fragment was ligated into pET28a(ϩ) expression vector (Novagen), creating a sequence with an N-terminal histidine tag. Cloning was done using E. coli DH5␣ and selection with 30 g/ml kanamycin; the final expression vector was transformed into SHuffle T7 competent E. coli cells (New England Biolabs). LB medium containing selection antibiotic was inoculated with the expression strain and shaken at 200 rpm and 30°C. At an optical density of 0.8, the culture was induced with a final concentration of 1 mM isopropyl-␤-D-1-thiogalactopyranoside (Sigma-Aldrich). Cells were collected after 4 h (centrifugation, 30 min at 4,000 ϫ g) and suspended in 50 mM Tris-Cl (pH 8.0) containing 200 g/ml lysozyme