Molecular Organization of the Alkali-insoluble Fraction of Aspergillus fumigatus Cell Wall*

Physical and biological properties of the fungal cell wall are determined by the composition and arrange-ment of the structural polysaccharides. Cell wall polymers of fungi are classically divided into two groups depending on their solubility in hot alkali. We have analyzed the alkali-insoluble fraction of the Aspergillus fumigatus cell wall, which is the fraction believed to be responsible for fungal cell wall rigidity. Using enzymatic digestions with recombinant endo- b -1,3-glu-canase and chitinase, fractionation by gel filtration, affinity chromatography with immobilized lectins, and high performance liquid chromatography, several fractions that contained specific interpolysaccharide covalent linkages were isolated. Unique features of the A. fumigatus cell wall are (i) the absence of b

The fungal cell wall is a physically rigid layer that protects the fungal cell from its environment, mediates cell-cell interaction, and is responsible for the shape of the cell. Despite its central role in growth and survival, the fungal cell wall remains poorly studied and its biosynthesis is insufficiently understood (1,2).
Cell wall polysaccharides are separated in two groups according to their solubility in hot alkali solution. The structural skeleton of the cell wall is alkali-insoluble. It has been known for a long time that ␤-1,3-glucan and chitin (linear polymer of ␤-1,4-N-acetylglucosamine) are the main components of the alkali-insoluble fraction. The alkali insolubility of glucan is due to its covalent linkage with chitin (3)(4)(5). The covalent bond between the two polysaccharides has been characterized in Saccharomyces cerevisiae by Kollar et al. (6), who showed that chitin is linked to the nonreducing end of a ␤-1,3-glucan chain. More recently, the same research group reported that the core of the yeast cell wall is a complex structure with a ␤-1,6and ␤-1,3-glucan to which chitin and some mannoproteins are attached (7). In yeast, cell wall-bound glycoproteins have been described to be covalently linked to ␤-1,6-glucan (8 -10). These proteins are originally GPI-anchored to the membrane (11,12) and then cleaved to be transferred onto ␤-1,6-glucan using the sugar moiety of GPI as a bridge (7,13). Ethanolamine and mannose residues, but not glucosamine and inositol, are in the GPI remnant involved in the protein-glucan linkage (14). Another family of cell wall proteins are directly bound to ␤-1,3glucan and released by mild alkali treatments (15). In contrast to yeast, the polymer organization of the cell wall of filamentous fungi has been poorly studied. Basically, it is only known that the alkali insolubility of their cell wall results, like in yeast, from the covalent association between glucan and chitin, with a concentration of chitin (around 10%) that is considerably higher than in yeast (2%) (16).
To better understand the organization of the cell wall components of a filamentous fungus and to gain further insight into the biosynthetic pathways involved in cell wall construction, we have focused our studies toward the chemical characterization of the interpolymer linkages occurring in the structural part of the cell wall, i.e. alkali-insoluble fraction of cell wall. The fungal model used is Aspergillus fumigatus. Using specific enzymatic digestion and various carbohydrate chemistry methods, we have shown that four polysaccharide components constituted this fraction: ␤-1,3-glucan was highly branched and was linked to chitin, galactomannan, and a linear ␤-1,3/1,4glucan never described before.

Preparation of Cell Walls
A. fumigatus CBS 144 -89 was grown in a 15-liter fermenter in a liquid medium containing a 2% glucose and 1% mycopeptone (Biokar Diagnostics) as described previously (17). After 24 h of culture (linear growth phase), the mycelia were collected by filtration, washed extensively with water and disrupted in a 50 mM Tris-HCl, pH 7.5 buffer containing 50 mM EDTA and 1 mM phenylmethylsulfonyl fluoride in a Dyno-mill (W. A. Bachofen AG, Basel, Switzerland) cell homogenizer in the presence of 1-mm-diameter glass beads at 4°C. The disrupted mycelial suspension was centrifuged (8000 ϫ g for 10 min), and the cell wall pellet was washed three times with the same buffer and stored at Ϫ20°C.
* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Fractionation of Cell Walls by Alkali and Enzymatic Treatments
Fractionation and digestion steps of the cell wall are summarized in Fig. 1. The disrupted cell wall pellet (50 g of wet weight, equivalent to 7.5 g of dried material) was incubated twice in 200 ml of 1 M NaOH at 65°C for 30 min. A third NaOH treatment did not release any extra material from the pellet. The alkali-insoluble pellet was washed five times with water and once with 50 mM Tris-HCl, pH 7.4 buffer. The pellet was resuspended in the same buffer supplemented with 5 mM sodium azide (60 ml) and incubated with 180 l of Quantazyme (50 units/l, recombinant endo-␤-1,3-glucanase; Quantum Industry, Quebec, Canada) at 37°C for 5 days. Endoglucanase digestion was repeated once. Pooled supernatants (QzSN) were kept frozen. After Quantazyme digestion, the insoluble pellet was treated twice with 1 M NaOH at 65°C for 30 min both to inactivate Quantazyme and to extract material that had become alkali-soluble after the glucanase treatment (accounting for 8% of the total alkali-insoluble starting material). After washing with water, the insoluble pellet residue was resuspended in 80 ml of 50 mM Tris-HCl, pH 8.0 containing 5 mM sodium azide and incubated at 37°C for 5 days with 4 ml of recombinant chitinase A (0.5 mg of protein/ml) from Serratia marcescens produced in Escherichia coli and purified as described previously (18). After centrifugation, the residual pellet was treated again with 1.2 ml of chitinase A in 40 ml of 50 mM Tris-HCl, pH 8.0, for 3 days. After centrifugation, supernatants (ChSN) were pooled. The Quantazyme-chitinase resistant pellet was treated with sodium hydroxide in the same conditions as described above releasing an alkali-soluble fraction, which accounted for 2% of the total alkali-insoluble starting material. The final pellet (FP) 1 was extensively washed with water before freeze drying. QzSN, ChSN, and FP were the three fractions analyzed.

Liquid Chromatography Procedures
Fractionation of Water-soluble Fractions (QzSN and ChSN) Released during Enzymatic Digestions-After concentration under vacuum, the QzSN and ChSN fractions were fractionated by gel filtration on a TSK HW40S column (90 ϫ 1.4 cm, ToyoPearl) eluted with 0.25% (v/v) acetic acid at 0.5 ml/min. The products were detected by refractometry. The excluded fractions, eluting at the void volume of the TSK HW40S column, were run on a Sephadex G100 column (90 ϫ 1.4 cm; Amersham Pharmacia Biotech) eluted with 50 mM sodium acetate, pH 6.0, at 9 ml/h. Polysaccharide sizes were estimated based on dextran standards (Amersham Pharmacia Biotech). All fractions were desalted by gel filtration on a Sephadex G15 column (35 ϫ 2.5 cm; Amersham Pharmacia Biotech) eluted with 20 mM acetic acid at 2 ml/min and freeze dried.

Chemical Degradation of Polysaccharides and Oligosaccharides
Release of galactofuranoside residues from galactomannan was obtained by mild acid hydrolysis with 15 mM HCl at 100°C for 20 h (17). Periodate oxidation was performed after incubation of 10 mg of material in 2 ml of sodium meta-periodate, 100 mM, during 7 days at 4°C in the dark (27). Excess of reagent was destroyed following addition of 200 l of ethylene glycol. After dialysis against water (membrane cut-off, 1000 Da) or gel filtration chromatography on Sephadex G15 column as described above, the oxidized product was reduced for 2 h in 100 mM ammonium hydroxide (2 ml) containing 20 mg of NaBH 4 . Excess of reagent was destroyed by addition of Dowex 50 ϫ 8 (H ϩ form) resin beads until a pH of 5-6 was reached. After co-distillations with methanol, Smith degradation was performed with 10% acetic acid at 100°C for 1 h. Degraded products were then separated by gel filtration chromatography on a TSK HW40S column as described above.
Acetolysis of soluble polymers was performed according to Ferguson (28). Peracetylated products (40 mg) were treated with 10 ml of an acetic acid/acetic anhydride/sulfuric acid solution (10:10:1 v/v/v) at 25 or 37°C for 3, 5, 7, and 24 h. The reaction was stopped by addition of 40 ml of pyridine and water (1:3 v/v). The peracetylated products were extracted with chloroform and washed with water. Deacetylation was performed in 300 mM NaOH and NaBH 4 (10 mg/ml) overnight at room temperature.
To remove GlcNAc from the terminal nonreducing end of an oligosaccharide, 5 mg of sample were incubated with 5 l of ␤-D-N-acetylglucosaminidase from Jack bean (10 units/160 l, Sigma) in 250 l of a 250 mM sodium acetate, pH 5.0 buffer at 25°C for 24 h. After addition of 50 l of 1 M NaOH, oligosaccharides were treated with NaBD 4 as described before.
Transgalactosylation of terminal nonreducing GlcNAc residues was performed using the following procedure: samples containing 5 mg of carbohydrate were incubated in 600 l of 50 mM Tris-HCl, pH 7.5, containing 1 mM MnCl 2, and 5 mM sodium azide, with 120 l of UDP-Gal (10 mg/200 l) and 30 l of galactosyl transferase (1 milliunit/ml; Roche Molecular Biochemicals) at 37°C during 3 days. A supplementary batch of 12 l of enzyme and 20 l of UDP-Gal was added to the sample, which was incubated for another 3 days at 37°C. The reaction was stopped by passing through two ion exchange columns (3 ml, Dowex 1 ϫ 2 acetate form, Dowex 50 ϫ 2, H ϩ form). The eluted products were desalted by gel filtration on a HW40S column as described above.
Transgalactosylated samples were applied to a column of Erythrina cristagalli lectin-agarose (4 ml, Vector) equilibrated in 50 mM HEPES, pH 7.5, containing 100 mM KCl, and 2 mM MgCl 2 . After washing with the same buffer, bound products were eluted with 200 mM lactose in the HEPES buffer (7). Fractions of 500-l volume were collected and desalted by gel filtration on a TSK HW40S column as described above.

Nuclear Magnetic Resonance Spectroscopy
Samples were deuterium exchanged by freeze drying solution in D 2 O and then dissolved in 99.95% D 2 O (Solvants Documentation Synthèse, Peypin, France). The polysaccharide concentrations were approximately 3, 40, and 15 mg/ml for QzSN II, QzSN I B , and ChSN II, respectively. All NMR spectra were collected between 31 and 40°C on a Varian Unity spectrometer operating at a proton frequency of 500 MHz and a 13 C frequency of 125 MHz and equipped with a z-gradient triple resonance ( 1 H, 15 N, 13 C) probe. The temperature was chosen for each sample to avoid superposition of the HOD signal with anomeric protons signals. Spectra were collected at 31, 40, and 35°C for QzSN II (fractions c, d, e, and f), QzSN I B , and ChSN II, respectively. Spectra were referenced to external trimethylsilyl-3 propionic acid-d 4 2,2,3,3sodium salt. The NMR signals were assigned by 1 H homonuclear experiments (DQF-COSY (30), RELAYH with single-and two-step relayed coherence transfer using delays of 60 ms (31) and TOCSY with mixing times of 80 -120 ms (32)) and 1 H-13 C heteronuclear NMR experiments (geHSQC leading to one-bond 1 H-13 C correlations and identification of methylenic carbons (33) and gHSQC-TOCSY with long mixing times of 80 and 120 ms (33,34)). This last experiment gave all the 13 C resonances of the same residue from the anomeric proton. Linkage assignments were made using two-dimensional nuclear Overhauser effect spectroscopy experiments with mixing time of 400 ms for QzSN II (35), off-resonance rotating frame two-dimensional nuclear overhauser effect spectroscopy with a mixing time of 200 ms (36), and gHMBC experiments with a delay of 60 ms for others samples (33). All twodimensional data, except for RELAYH and gHMBC, were collected in the phase sensitive mode using the States-Haberkorn method (37). Coupling constants were measured from one-dimensional spectrum recorded with a digital resolution of 0.30 Hz/point or from the DQF-COSY experiment with a digital resolution of 0.75 Hz/point after a zero filling.

Mass Spectrometry
Matrix-assisted desorption ionization/time of flight (MALDI-TOF) mass spectrometry. Mass spectra were measured on a reflectron-type Vision 2000 time of flight mass spectrometer (Finnigan MAT, Bremen, Germany). Samples were mounted on a x, y movable stage allowing irradiation of selected sample areas. A nitrogen laser with an emission wavelength of 337 nm and 3-ns pulse duration was used. The spectrum was recorded in the positive ion mode and accelerated to an energy of 10 keV before entering the flight tube. Ions were post-accelerated for detection to an energy of 10 keV. Samples were prepared by mixing directly on the target 1 l of oligosaccharide solution (about 25 pmol) and 1 l of 2.5-dihydroxybenzoic acid matrix solution (12 mg/ml dissolved in CH 3 OH/H 2 O (80:20 v/v)).

Gas Chromatography-Mass Spectrometry
GLC-MS analysis were recorded using a Automass II 30 quadripolar mass spectrometer interfaced with a Carlo Erba 8000 Top gas chromatograph (Finnigan, Argenteuil, France). Electron ionization spectra were recorded using an ionization energy of 70 eV. Positive chemical ionization spectra were obtained at 70 eV using ammonia as reagent gas. Gas chromatograph was equipped with a CP-Sil 5CB/MS capillary column (25 m ϫ 0.32 mm, Chrompack) gas vector, and helium was at the flow rate of 2 ml/min; the column temperature was 100 -240°C at 5°C/min.

Fractionation and General Composition of the Cell Wall
NaOH treatment of 7.5 g of cell wall dried material (equivalent to 50 g of wet weight) resulted in the production of 3 g of alkali-insoluble fraction (40% of the wall dry weight). Treatment of this fraction by sequential incubation with recombinant ␤-1,3-glucanase (Quantazyme) and chitinase, alternated with alkali treatment, made soluble 90% of the alkali-insoluble starting material. (Fig. 1 and Table I). Further purification of the soluble fractions by gel filtration chromatography on TSK-HW40S and Sephadex G100 columns resulted in the separation of nine fractions of different molecular mass (Figs. 2 and 3). The amounts of material of each fraction, expressed as percentages of original dry weight, and their sugar compositions are shown in Table I. Three out of the four fractions released by Quantazyme contained only glucose, whereas the high molecular weight fraction (QZSN I A ) contained glucose associated with galactose and mannose. In a similar way, low molecular weight fractions ChSN III and ChSN IV released by chitinase were exclusively composed of GlcNAc, whereas fractions ChSN I A , ChSN I B , and ChSN II contained GlcNAc associated with various amounts of glucose, galactose, and mannose. The chemical analysis of the different fractions resulted in the identifi-

Chemical Characterization of the Water-soluble Products
Released by Quantazyme Fraction QzSN III: Degradation Products Obtained with Endo-␤-1,3-glucanase-HPAEC, methylation, and MALDI-TOF mass spectrometry analysis showed that QzSN III corresponded to laminaripentaose, which is the product of hydrolysis of ␤-1,3-glucan by Quantazyme (data not shown) Fraction QzSN II: Mixture of Branched Laminarioligosaccharides-QzSN II (1.5-2.5 kDa) contained a mixture of laminarioligosaccharides that had been reduced with NaBH 4 before separation by HPAEC (Fig. 4). MALDI-TOF mass spectrometry analysis indicated that the degree of polymerization (dp) of oligosaccharide varied from 9 to 15 ( Fig. 4). Products with the same Mr gave two peaks on HPAEC analysis, suggesting they were chemically organized differently. Analysis of these oligosaccharides was performed by 1 H NMR spectroscopy on two couples of oligosaccharides with 10 and 11 glucose residues (QzSN IIc, QzSN IId, QzSNIIe, and QzSN IIf). The one-dimensional 1 H NMR spectra of QzSN IId and QzSN IIf oligosaccharides were similar to a linear ␤-1,3-glucan with a glucitol residue at the reducing end as described previously (Ref. 38 and data not shown). The one-dimensional spectrum of the QzSN IIe oligosaccharide contained five doublets and two close-lying doublets in the anomeric region between 4.5 and 4.8 ppm (Table II). These chemical shifts and the coupling constants 3 J 1,2 of these doublets were in good agreement with those published for a linear ␤-1,3-glucan containing a ␤-1,6 linkage (39,40). Each signal corresponded to one anomeric proton, except for the signals at 4.76 and 4.80 ppm, which accounted for two and four protons, respectively. The two close-lying doublets corresponding to one proton were due to different populations of conformers. Because glucitol did not give any signal in this part of the spectrum, the NMR results indicated the presence of 11 glucose units/QZSNIIe molecule, in agreement with the MALDI-TOF data. Because of severe overlap of  other proton resonances, even in the DQF-COSY experiment, RELAYH experiments were performed to identify H3 resonances for further sequential assignment and linkage determination. From the chemical shifts reported in Table II, it can be seen that four units (labeled F) are undistinguishable by their proton resonances, as expected for a linear ␤-1,3-glucan chain. Chemical shifts of H3 protons are comprised between 3.75 and 3.80 ppm, except for the two C units with H3 resonances at 3.53 ppm, typical value for ␤-Glc unsubstituted in C3 corresponding to nonreducing end units. These results, which showed the presence of two glucose residues at a nonreducing end of the QzSN IIe molecule, indicated that the oligosaccharide QzSN IIe was branched. This was confirmed by the analysis of the D residue, which was substituted in C6, inducing large resonance shifts at low field for protons H6 and H6Ј of 0.14 and 0.29 ppm, respectively, and smaller shifts up to the H4 as described previously (39). The position of H3 at 3.80 ppm in the D unit implied that it is substituted in C3; therefore, the D residue is a branching point in the oligosaccharide. The two-dimensional nuclear Overhauser effect spectroscopy experiment showed that all residues were linked via a ␤-1,3 linkage, except for residue A, which is connected to the D unit via a ␤-1,6 linkage. For each anomeric proton, intra-residue NOE was observed with H2 and also with H3 and H5 as expected for a ␤ anomer. An inter-residue NOE is observed with the proton of the neighboring unit where glycosidic linkage occurred. Complete sequential assignment was not possible because of the drastic overlap of H3 resonances for most units (␦ comprised between 3.75 and 3.80 ppm for A, B, D, E, and F). For the same reason, the ␤-1,6 linkage could not be localized in the sequence by NMR. The structure of QzSN IIe identified from the NMR data is as follows.
where x, y, and z are between 0 and 4, and x ϩ y ϩ z ϭ 4. The one-dimensional 1 H NMR spectrum of the QzSN IIc oligosaccharide is identical to that of QzSN IIe except for the signal intensities that correspond to a branched ␤-1,3-glucan decasaccharide (data not shown). In addition, GLC-MS of the methyl ethers obtained after methylation of all oligosaccharides recovered by HPLC (dp 9 -14) showed that half of the oligosaccharides with a given molecular weight were linear ␤-1,3-glucan, whereas the other half were branched ␤-1,3-glucan with a single 1,6-glucose linkage as indicated by a ratio 2:1:1 corresponding to glucose units at the nonreducing end (2,3,4,6-Glc), disubstituted glucose units (2,, and glucose units at the reducing end (1,2,4,5,6-Glc), respectively (data not shown).
Acetolysis degradation, which preferentially cleaves 1,6-glycosidic bonds, was performed on QzSN IIe. Degraded products were analyzed by HPAEC and MALDI-TOF mass spectrometry. A kinetic analysis showed that acetolysis released preferentially two laminarioligosaccharides: pentaose and hexaose from the QzSN IIe undecaoligosaccharide (Fig. 5). These results suggested that QzSN IIe contained one major oligosaccharide structure: a laminarihexaose branched with a laminaripentaose. However, like for the 1 H NMR and methylation data, the exact position of the branch point could not be identified with this technique. NMR, GLC-MS, and acetolysis data suggested that the structure of QzSN IIe as follows.
where x ϭ 2 or 3, y and z are between 0 and 2, and x ϩ y ϩ z ϭ 4. Fraction QzSN I B : 1-Laminarioligosaccharides of dp Ͼ 15 Are Also Branched-Fraction QzSN I B of 2-5 kDa isolated after gel filtration on a Sephadex G100 column was composed of glucose residues ( Fig. 3 and Table I). MALDI-TOF mass spectrometry and HPAEC showed that this fraction contained a mixture of oligosaccharides of dp 11-25, the major species having a dp of 16 -18 (data not shown). After reduction, methylation analysis produced five types of methyl ether of glucose units: 1,2,4,5,6-Glc, 2,3,4,6-Glc, 2,4,6-Glc, 2,3,6-Glc, and 2,4-Glc 2 with a molar ratio of 0.4:1.2:13.3:0.8:1. The presence of ␤-1,6 linkages in the ␤-1,3-glucan indicated by the methylation analysis was confirmed by 1 H and 13 C NMR analysis. The one-dimensional spectrum of the QzSN I B , obtained by 1 H NMR, contained seven doublets in the anomeric region between 4.5 and 4.8 ppm (Table III). These chemical shifts and coupling constant 3 J 1,2 were identical to the ones obtained with QzSN IIe (Tables II and III) except for (i) two signals at 4.67 and 5.24 ppm that corresponded to the ␣ and ␤ anomer of the glucose at the reducing end (whereas QzSN IIe was reduced before NMR analysis) and (ii) one signal at 4.76 ppm, which corresponded to 4-O substituted glucose (in agreement also with the methylation data). 13 C resonances were assigned from the geHSQC and gHSQC-TOCSY experiments (34) (Table III). ppm, which corresponded, respectively, to the ␤ and ␣ Glc moieties of the reducing end, respectively (39). The gHMBC experiment confirmed that all the units are ␤-1,3 linked except for the A residue, which was ␤-1,6 linked with the D residue. Further confirmation that QzSN I B contained a mixture of branched oligosaccharides was obtained by an acetolysis assay. In a way similar to QzSN IIe, acetolysis of QzSN I B released a large number of linear ␤-1,3 linked oligosaccharides of variable size (dp 2-20 and higher; Fig. 5).
In conclusion, data obtained by 1 H and 13 C NMR, methylation analysis, and HPAEC separation of QzSN I B submitted to acetolysis were all in agreement indicating that this fraction contained ␤-1,6 branched ␤-1,3-glucans, but it was of higher M r than QsSN II with oligosaccharides of dp higher than 15.
Fraction QzSN I B 2: Identification of a New ␤-1,3/1,4-Glucan-Methylation analysis of QzSN I B showed the presence of 2,3,6-Glc and 2,4,6-Glc in the molar ratio 1:16.6. The glucan containing 4-O-substituted glucose was also recovered from mycelium grown in a chemically defined medium in flasks without cotton plugs, indicating that its presence was not due to cellulose contamination but was an intrinsic component of the cell wall. The ␤-1,3/1,4-glucan was released from the alkaliinsoluble fraction of the cell wall by quantazyme treatment, suggesting that it was bound to the ␤-1,3/1,6-glucan. To separate the glucan containing 4-linked glucose from the branched ␤-1,3/1,6-glucan, QzSN I B was incubated with the 74-kDa endo-␤-1,3-glucanase from A. fumigatus (ENG1). ENG1 has the ability to cleave laminari-oligosaccharides of shorter dp than Quantazyme (29). As a consequence, ENG1 was able to cleave a branched ␤-1,3/1,6-glucan with short ␤-1,3-glucan side chains, so that Quantazyme would not be able to cleave. Indeed, ENG1 was able to cleave QzSN I B resulting from the action of Quantazyme. The ENG1-digested product was borohydride-reduced and submitted to gel filtration chromatogra-  phy on the TSK HW40S column (Fig. 6). Gel permeation profile showed laminari-oligosaccharides of low dp (mainly 2 and 3), branched ␤-1,3/1,6 laminari-oligosaccharides (dp 6 -9, fraction b), and a fraction a of high M r that was further analyzed. Methylation analysis showed that a polymer with an equimolar ratio of glucose residues substituted in position 3 and substituted in position 4 was recovered in the fraction a. 1 H and 13 C NMR data of this fraction are presented in Table IV and Fig. 7. Two main doublets were observed in the anomeric region at 4.53 and 4.77 with a 3 J 1,2 coupling constant value of 7.9 Hz typical of ␤-linked units. Integration of these two signals shows that A and B residues were in the ratio 1:1. H3 and H4 resonances determined with RELAYH and TOCSY experiments showed that residue A was substituted in position 3 and residue B in position 4. The gGHMBC experiment showed interglycosidic couplings between H1A and C4B and between H1B and C3A, indicating that A was linked to B in position 4 and B was linked to A in position 3 (Fig. 7). Degradation of this fraction by periodic oxidation and Smith degradation yielded a monosaccharide glycoside isolated from the TSK HW40S column as a dp 2 (Fig. 6). GLC analysis indicated that it was composed of glucose and erythritol. Erythritol residue was produced by cleavage of glucose residue substituted in position 4 by sodium periodate. GLC-MS analysis using the chemical ionization mode showed that the permethylated compound had a M r of 382, corresponding to an hexose plus erythritol. Analysis with electronic impact mode after methanolysis indicated that the glucose residue was bound to carbon 2 of erythritol. NMR and methylation data indicated that the linear ␤-1,3/1,4glucan has the following repeating unit: [3Glc␤1-4Glc␤1].
Fraction QzSN I A 1: Linkage between ␤-1,3-Glucan and Galactomannan-This fraction had a molecular mass of 40 kDa and contained galactose, mannose, and glucose residues in similar amounts (Table I). Methylation analysis of this fraction showed that methyl ethers of galactose and mannose residues corresponded to the galactomannan polymer, which is composed of a core chain of ␣-mannose residues with short side chain of ␤-1,5-galactofuranose residues, as described previously by Latgé 1.4 cm) was eluted with 0.25% (v/v) acetic acid solution at 0.5 ml/min. Products were applied to the column and monitored by refractometry. dp was established with malto-oligosaccharides as standard. The homonuclear proton coupling constant in Hz are written in the middle. The nonequivalent geminal proton resonating at lower field is denoted HЈ. The glucose residues were labeled A to G in order of increasing chemical shift of their anomeric protons. ments preferentially released a tetrasaccharide consistent with the presence of a repeating mannose unit [6Man␣1-2Man␣1-2Man␣1-2Man␣1] (data not shown). As indicated by methylation analysis, two of the ␣-1,2 linked mannose residues were substituted in position 3 or 6 and were branching point for the galactofuran side chain (Table V). Because galactomannan binds to Concanavalin A, the galactomannan containing molecules were purified by affinity chromatography on a ConA-Sepharose column. The fraction bound to the ConA-Sepharose and released with 0.2 M ␣-methylmannoside accounted for 80% of QzSN I A . It was composed of mannose, galactose, and glucose residues in a molar ratio of 2.5:2.7:1. This result indicated that the galactomannan polymer was covalently bound to the glucan moiety. The polysaccharide bound to ConA-Sepharose was reduced with NaBH 4 , hydrolyzed, and derivatized. GLC analysis showed the absence of mannitol and the presence of glucitol, indicating that a glucose residue was located at the re-ducing end (data not shown). This fraction was sequentially submitted to mild acid hydrolysis to remove galactofuran side chains and to periodate oxidation to degrade the mannan moiety The products were separated by gel filtration chromatography on TSK HW40S column (Fig. 8). Fraction b contained typical products of degradation by periodate oxidation. Fraction a was composed of a mixture of oligosaccharides with sizes varying between 2 and 11 residues. Only glucose and arabitol were detected by GLC analysis in the fraction a with a glucose/ arabitol ratio of 2.9. Arabitol was produced by periodate oxidation, which cleaves glucose residues at the reducing end between carbon 1 and 2, indicating that the ␤-1,3-glucan chain was at the reducing end. This result was in agreement with the composition analysis performed in the reduced undegraded QzSN I A . MALDI-TOF mass spectrometry analysis of the fraction a showed the presence of a series of oligosaccharides containing 1 arabitol residue plus an increasing number of glucose residues (Fig. 9).  where Man is ␣-mannose, Gal is ␤-galactofuranose, and Glc is ␤-1,3-glucose. Acetolysis degradation of the fraction bound to ConA-Sepharose resulted in the release of laminarioligosaccharides free from mannose residues. As acetolysis cleaves preferentially 1,6 glycosidic linkage, these data suggested, although indirectly, that the mannan was linked through a 1,6 linkage to the nonreducing end of the glucan chain.

Characterization of Water-soluble Products Released by Chitinase Fractions ChSN III and ChSN IV: Products of Degradation of Chitinase-Fractions ChSN III and ChSN IV contained only
N-acetylglucosamine residues (GlcNAc). MALDI-TOF mass spectrometry showed that ChSN IV corresponded to N-acetylchitobiose and ChSN III to a mixture of N-acetylchitotriose and N-acetylchitotetraose (data not shown).
Fraction ChSN II: Linkage between Chitin and Glucan-Fraction ChSN II (molecular mass, 1-2 kDa) was composed of glucose and GlcNAc residues in a molar ratio of 16:1 ( Table I). Because of the low amount of material recovered, this fraction was analyzed in toto without further purification. MALDI-TOF mass spectrometry showed that ChSN II contained a mixture of oligosaccharides of dp 7-13. After 9 h of chitinase digestion, half of the oligosaccharides contained one GlcNAc residue (Fig. 10). If chitinase incubation was prolonged for 24 -72 h, the amount of oligosaccharides containing the GlcNAc residue gradually decreased over time (data not shown), indicating that GlcNAc residues bound to glucan chain were hydrolyzed by the chitinase treatment. Methylation was performed after reduction with NaBD 4 of ChSN II treated with chitinase for 24 h. GLC analysis of methyl ethers showed the presence of 2,3,4,6-Glc, 2,4,6-Glc, 2,3,6-Glc, 2,4-Glc, and 3,4,6-O-methyl 2-Nmethyl 1,5-O-acetyl 2-N-acetyl glucosaminitol in a molar ratio of 0.7:10:1.6:0.4:0.6. A 24-h incubation of ChSN II with ␤-Dglucosaminidase removed 70% of the GlcNAc residues and 4-O-substituted glucose residues, indicating that GlcNAc was located at the nonreducing end of the oligosaccharide (data not shown). These methylation data suggested that the GlcNAc residue was linked to the nonreducing end of the ␤-1,3-glucan oligosaccharide via a ␤-1,4 glycosidic linkage.
The one-dimensional 1 H NMR spectrum of the ChSN II (after 24 h of chitinase digestion) fraction contained five doublets in the anomeric region between 4.5 and 4.8 ppm (Table VI). Chemical shifts of 4.52, 4.75, and 4.80 for residues A, C, and E and the associated coupling constants values 3 J 1,2 for these doublets were similar to those obtained with QzSN II and QzSN I B and indicated the presence of a linear ␤-1,3-glucan containing a ␤-1,6 branching point. H3 resonances, determined was equilibrated with 0.25% (v/v) acetic acid solution at 0.5 ml/min. Products were applied to the column and detected by refractometry. dp was established with malto-oligosaccharides as standard. with RELAYH experiments and 13 C resonances, assigned from the geHSQC and gHSQC-TOCSY experiments confirmed the presence of a ␤-1,3-glucan chain with ␤-1,6 linkage (Glc A, C, and E; Table VI) (34). The B residue was unambiguously identified to be a GlcNAc residue from the gHMBC experiment through the observed long range couplings between the carbon of the carbonyl group at 177.1 ppm and the proton chemical shift of the CH 3 group at 2.06 ppm on the one hand and the ring H2 proton at 3.74 ppm on the other hand. Proton chemical shift values were typical of a nonreducing end GlcNAc residue (41). Sequence analysis, obtained from the gHMBC experiment, showed that all the units are ␤-1,3 linked except for the nonreducing terminal GlcNAc residue, which is ␤-1,4-linked with the D unit and for the A residue, which is ␤-1,6 linked with the reducing-end residue. Specific anomeric 1 H and 13 C signals were not identified at the reducing end. Moreover, methylation and GLC-MS analysis showed the presence of an unknown methyl ether. Methylation and GLC-MS analysis of reducing ends in QzSN I B treated with hot NaOH showed that the modification of the reducing end was due to a peeling reaction, which was sequential and stopped at a 1,6 branching point. From these data, it can be deduced that GlcNAc was linked to ␤-1,3-glucan side chains through a 1,4 linkage according to the following structure. GlcNAc␤1-4Glc␤1-3Glc␤1-3Glc␤1-3Glc␤1-3Glc␤1
The only chemical difference between QzSN I B and ChSN I B was the presence of GlcNAc residues in ChSN I B . To identify the linkages and the position of GlcNAc residues to the sugar core, transgalactosylation of GlcNAc residues was performed. Acetolysis of ChSN I B released a mixture of oligosaccharides of variable size with a maximal dp of 10 (data not shown). The mixture of oligosaccharides was incubated with galactosyltransferase and UDP-galactose and then applied to a E. cristagalli lectin-agarose column. The fraction retarded on the affinity column was analyzed by GLC and MALDI-TOF mass spectrometry. The Glc/GlcNAc ratio was 6, and all oligosaccharides recovered contained the sequence Gal␤1-4GlcNAc, indicating that all GlcNAc residues were bound in a ␤ anomeric configuration to the branched ␤-1,3/1,6-glucan. The size of the side chains containing one GlcNAc residue varied from 2 to 8 residues (Fig. 11). These chemical data showed that all GlcNAc residues were bound to the nonreducing end of the branched ␤-1,3/1,6-glucan as in ChSN II, in agreement with the following structure.
where n is between 0 and 7.
Fraction ChSN I A : Chitin, ␤-1,3/1,4-Glucan and Galactomannan Are Linked to Different ␤-1,3-Glucan Branches-ChSN I A had a molecular mass of 40 kDa (Fig. 3) and contained galactose, mannose, and glucose in equivalent amounts with the presence of traces of GlcNAc residues in a Glc/GlcNAc ratio of 100 (Table I). Methylation analysis showed that the fraction released by the chitinase treatment was originally composed of galactomannan, ␤-1,3/1,4-glucan and chitin, as shown by the presence of GlcNAc in this fraction resulting from the action of chitinase (Table V). The main question addressed in the analysis of ChSN I A was the identification of the linkage between chitin, galactomannan, and glucan. To address this question, CHSN I A was incubated with the 74-kDa endo-␤-1,3-glucanase, ENG1. Degradation products were separated by gel filtration chromatography on a TSK HW40S column (Fig. 12). Monosaccharide composition showed that GlcNAc residues were only found in the tetraoligosaccharide peak (dp 4) and in fraction b containing oligosaccharides of dp 5-10, whereas galactose and mannose associated to glucose residues were in fraction a excluded at the void volume. MALDI-TOF analysis of fractions b and dp 4 showed that GlcNAc residues were covalently linked to the glucan fragment (Fig. 13). Methylation analysis of fraction b showed the presence of four methyl ethers: 1,2,4,5,6-Glc, 2,4,5,6-Glc, 2,4,6-Glc, and 2,4-Glc in a molar ratio of 0.7:2:2.9:1, respectively, indicating that this fraction consisted of a highly branched ␤-1,3/1,6-glucan with short chains of linear ␤-1,3glucans, which explained its resistance to the 74-kDa endo-␤-1,3-glucanase (see above). These data suggested that ChSN I A had a structure similar to that of QzSN I A with the presence of galactomannan and the ␤-1,3/1,4-glucan associated to ␤-1,3glucan and also similar to ChSN I B or ChSN II where GlcNAc residue was linked to branched ␤-1,3-glucan. This result indicated that GlcNAc, ␤-1,3/1,4-glucan and galactomannan were covalently associated to three different ␤-1,3-glucan side chains of the branched ␤-1,3/1,6 core glucan.
Characterization of the Quantazyme/Chitinase-resistant Pellet-The residual insoluble pellet (FP) was very rich in glucosamine and galactosamine, because hexosamine accounted for 80% of this material (Table I). The insolubility of this residual pellet made the analysis of this fraction difficult, and only partial information was obtained on the chemical organization of FP. The most striking features were the following: (i) the glucan moiety was a branched ␤-1,3/1,6 polymer similar in composition to the one analyzed in QzSN I A and ChSN I A ; (ii) treatment of FP with the 74-kDa endoglucanase (ENG1) solubilized 15% of FP. Gel permeation and MALDI-TOF analysis showed that the soluble oligosaccharides had a composition similar to those released by ENG1 from ChSN I A with a dp of 2-10 and the presence of GlcNAc linked to some of the laminarioligosaccharides (data not shown); (iii) D-N-acetylgalactosamine residues, which is the major monosaccharide of FP, was substituted in position 4 and should correspond to a poly-N-acetylgalactosamine polymer.

DISCUSSION
Cell wall biosynthesis is a key process in the formation, growth, and morphogenesis of fungal cells. Despite its essential function in fungal development, cell wall structure and biosynthesis, especially polymer cross-linking, remain poorly studied. In this study, a fractionation procedure using alkali extraction and enzymatic digestions of the cell wall, followed by a physicochemical analysis of the soluble polysaccharides, was followed to understand the general organization of the polysaccharides (␤-glucans, chitin, and galactomannan) in the alkali-insoluble fraction of the A. fumigatus cell wall. Care has been taken to use exclusively recombinant ␤-1,3-glucanase and chitinase, which are free of any contaminating enzymes. Most previous reported studies in cell wall chemistry, including the study of Kollar et al. (7) have used commercial enzymes such as laminarinase or zymolyase, which also contain protease and exo-␤glucanase activities, or chitinase, which is contaminated by chitobiase and ␤-glucanase activities. Such contaminating activities can mislead the identification of sugar linkages in the cell wall.
From the data presented in this paper, a tentative scheme of the organization of the ␤-1,3-glucan and its covalently associated polymers found in the cell wall of A. fumigatus can be presented (Fig. 14). ␤-1,3-Glucan is the main component of the alkali-insoluble fraction of A. fumigatus cell wall, and it is highly branched with ␤-1,6 linkages (4% of branch points), constituting a three-dimensional network with a large number of nonreducing ends. Chitin, galactomannan, and ␤-1,3/1,4glucan are covalently anchored onto these nonreducing ends, producing a large heteropolymer complex. Chitinase digestion (without previous Quantazyme treatment) degraded 80 -90% of the total amount of the chitin of the alkali-insoluble fraction. Chitinase treatment only released GlcNAc and did not release other water-soluble polymers such as ␤-glucan or galactomannan (data not shown), indicating that these components are not directly linked to chitin. Based on Fig. 14, our current hypothesis of the chronological events occurring in the biosynthesis of the A. fumigatus polysaccharide network is as follows: (i) biosynthesis of the individual polysaccharides (␤-1,3-glucan, ␤-1,3/1-4 glucan, chitin, and galactomannan); (ii) branching of ␤-1,3-glucan, increasing the number of acceptor sites; and (iii) covalent addition of chitin, galactomannan, and ␤-1,3/1,4-glucan to glucan branches.
The first class of biosynthetic events, synthesis of ␤-1,3glucan and chitin, is the only one known in A. fumigatus (42)(43)(44). Synthesis of ␤-1,3-glucan in A. fumigatus, like in yeast, is under the control of a transmembrane glucan-synthase complex, using UDP-glucose as substrate to extrude linear glucan chain in the periplasmic space (45). Two components of this complex, the regulatory GTPase Rho1 and the putative catalytic subunit FKS have been characterized at the molecular level in yeast (46 -50) and recently in A. fumigatus. 3 The enzymes responsible for the second and third classes of biosynthetic events (branching and cross-linking) are totally unknown. Two glucanosyltransferases have been recently characterized in A. fumigatus, but none of them could synthesize the ␤-1,3/1,6 branched glucan of the cell wall. The first glucanosyltransferase, homologous to Bgl2p in S. cerevisiae and Candida albicans, is able to construct a ␤-1,6 linkage inside a ␤-1,3glucan chain (39,51,52). However, this enzyme (i) requires a free reducing end to act on the ␤-1,3-glucan chain and (ii) does not make a branched glucan but makes a kinked linear glucan. Moreover, disruption of the gene encoding this protein in A. fumigatus and homologous yeast genes did not result in a phenotype distinct from the parental strain, suggesting that this glucanosyltransferase does not play a role in cell wall construction (40,53,54). In contrast, the second glucanosyltransferase identified in A. fumigatus (Gelp) plays a major role in fungal morphogenesis. This transferase is responsible for the elongation of ␤-1,3-glucan chains (38). Molecular characteriza-3 A. Beauvais and P. Mol, personal communication. tion of this enzyme in our laboratory showed that it was homologous to S. cerevisiae Gasp, C. albicans Phrp, and Candida maltosa Epdp. 4 Phenotypic analysis of the yeast and A. fumigatus mutants suggested that this enzyme activity was involved in hyphal growth, cell separation, and cell wall construction (55)(56)(57)(58)(59). 4 However, the exact biological function of these proteins remains unclear. It has been only shown that the transfer activity requires nonreducing ends of ␤-1,3-glucan chains, similar to the ones produced during branching of glucan. A putative role for this enzyme could be the elongation of the ␤-1,3glucan side chains of this ␤-1,3/1,6 branched glucan, reinforcing the primary role of the branching of this glucan in the construction of the cell wall.
The only other fungal cell wall that has been studied in detail is the one of S. cerevisiae. Comparison of polysaccharide structures showed that these two fungal species present striking similarities: (i) the ␤-glucan ore of S. cerevisiae is branched via ␤-1,6 linkages. The ␤-1,3/␤-1,6-glucan fraction represents 85% of the total cell wall glucans and contains 3% of branch points (60) and (ii) the alkali-insoluble fraction is a complex structure associating chitin to ␤-1,3-glucan via a ␤-1,4-linkage (6). This cross-linking is essential for the alkali insolubility of ␤-glucan in yeast as well as in filamentous fungi (3,4,5).
However, these two fungi present important differences in their cell wall composition: (i) ␤-1,6-glucans have a central role in S. cerevisiae, where they interconnect proteins, chitin, and ␤-1,3-glucans (7,13). In A. fumigatus, no ␤-1,6-glucan has been identified. (ii) Two new polysaccharides have been identified in A. fumigatus: A poly-N-acetylgalactosamine polymer and a linear ␤-1,3/1,4-glucan. A poly-N-acetylgalactosamine polymer has been already described in Aspergillus parasiticus (61,62). In A. fumigatus, the galactosamine polymer was also found in the culture filtrate (data not shown) where it became rapidly insoluble. 5 We have no evidence that this polymer is covalently associated to other cell wall polymers. Its presence in FP could be only due to its co-precipitation with the alkali-insoluble material. A linear ␤-1,3/1,4-glucan never described before in fungi was released with Quantazyme and/or chitinase from the alkali-insoluble cell wall fraction in A. fumigatus. This ␤-1,3/ 1,4-glucan from A. fumigatus has a structure close to vegetal and lichen glucan. Indeed, lichenin, isolated from Cetraria islandica, barley, and oat ␤-glucans are ␤-1,3/1,4-glucans constituted with ␤-1,4-glucooligosaccharides of dp 2 or 3 mainly, joined by a single ␤-1,3-linkage (63)(64)(65). Methylation data from the total soluble fraction released by Quantazyme suggested that it represented 10% of total ␤-glucan. The identification and characterization of this ␤-1,3/1-4-glucan could explained the presence of 4-O-substituted ␤ linked glucose residue reported 30 years ago in the cell wall of Ascomycete species (66). (iii) Galactomannan with galactofuran side chains has been only described in Penicillium and Aspergillus species. Among these filamentous fungi, this polymer has been previously isolated from a culture filtrate or from the alkali-soluble fraction of the mycelial wall (17,67,68). It is the first time that galactomannan was described to be covalently associated to ␤-1,3glucan in fungi.
In yeast, cell wall bound glycoproteins have been described to be covalently linked to ␤-1,3-glucan trough a ␤-1,6-glucan moiety (8,9,10). These proteins are originally GPI-anchored to the membrane (11,12) and then cleaved to be transferred onto ␤-1,6-glucan using the sugar moiety of GPI as a bridge (13,7). Another family of cell wall proteins are directly bound to ␤-1,3glucan through an unknown linkage (15). In our study, the use of hot NaOH to prepare the alkali -insoluble fraction of A. fumigatus cell wall destroyed all proteins putatively associated to the cell wall and did not allow to study the covalent incorporation of proteins in the A. fumigatus cell wall. The presence of such proteins is being presently investigated.
Despite the differences observed in the composition of the structural polysaccharides of S. cerevisiae and A. fumigatus, branching of ␤-1,3-glucan appears as a central event for the cell wall construction. Recent analysis of C. albicans cell wall also indicated the presence of branched ␤-1,3/1,6-glucan and the linkage of the ␤-1,6-glucan to the nonreducing end of the ␤-1,3glucan chain. 6 The enzymatic activity responsible for the branching of ␤-1,3-glucan has not yet been identified and is presently under study in our laboratory since the understanding of this biosynthetic step in the cell wall construction may lead to the discovery of new antifungal drugs for the treatment of life-threatening invasive aspergillosis. 4